2006-10-11 10:20:50 +02:00
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* linux/fs/ext4/inode.c
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
|
|
|
* Copyright (C) 1992, 1993, 1994, 1995
|
|
|
|
* Remy Card (card@masi.ibp.fr)
|
|
|
|
* Laboratoire MASI - Institut Blaise Pascal
|
|
|
|
* Universite Pierre et Marie Curie (Paris VI)
|
|
|
|
*
|
|
|
|
* from
|
|
|
|
*
|
|
|
|
* linux/fs/minix/inode.c
|
|
|
|
*
|
|
|
|
* Copyright (C) 1991, 1992 Linus Torvalds
|
|
|
|
*
|
|
|
|
* Goal-directed block allocation by Stephen Tweedie
|
|
|
|
* (sct@redhat.com), 1993, 1998
|
|
|
|
* Big-endian to little-endian byte-swapping/bitmaps by
|
|
|
|
* David S. Miller (davem@caip.rutgers.edu), 1995
|
|
|
|
* 64-bit file support on 64-bit platforms by Jakub Jelinek
|
|
|
|
* (jj@sunsite.ms.mff.cuni.cz)
|
|
|
|
*
|
2006-10-11 10:20:53 +02:00
|
|
|
* Assorted race fixes, rewrite of ext4_get_block() by Al Viro, 2000
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/time.h>
|
2006-10-11 10:21:01 +02:00
|
|
|
#include <linux/jbd2.h>
|
2006-10-11 10:20:50 +02:00
|
|
|
#include <linux/highuid.h>
|
|
|
|
#include <linux/pagemap.h>
|
|
|
|
#include <linux/quotaops.h>
|
|
|
|
#include <linux/string.h>
|
|
|
|
#include <linux/buffer_head.h>
|
|
|
|
#include <linux/writeback.h>
|
2008-07-12 01:27:31 +02:00
|
|
|
#include <linux/pagevec.h>
|
2006-10-11 10:20:50 +02:00
|
|
|
#include <linux/mpage.h>
|
2008-12-19 21:47:15 +01:00
|
|
|
#include <linux/namei.h>
|
2006-10-11 10:20:50 +02:00
|
|
|
#include <linux/uio.h>
|
|
|
|
#include <linux/bio.h>
|
2009-09-28 21:48:41 +02:00
|
|
|
#include <linux/workqueue.h>
|
2010-03-04 22:14:02 +01:00
|
|
|
#include <linux/kernel.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 09:04:11 +01:00
|
|
|
#include <linux/slab.h>
|
2009-06-17 17:48:11 +02:00
|
|
|
|
2008-04-30 00:13:32 +02:00
|
|
|
#include "ext4_jbd2.h"
|
2006-10-11 10:20:50 +02:00
|
|
|
#include "xattr.h"
|
|
|
|
#include "acl.h"
|
2008-07-14 23:52:37 +02:00
|
|
|
#include "ext4_extents.h"
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2009-06-17 17:48:11 +02:00
|
|
|
#include <trace/events/ext4.h>
|
|
|
|
|
2008-08-20 03:55:02 +02:00
|
|
|
#define MPAGE_DA_EXTENT_TAIL 0x01
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
static inline int ext4_begin_ordered_truncate(struct inode *inode,
|
|
|
|
loff_t new_size)
|
|
|
|
{
|
2009-02-10 17:15:34 +01:00
|
|
|
return jbd2_journal_begin_ordered_truncate(
|
|
|
|
EXT4_SB(inode->i_sb)->s_journal,
|
|
|
|
&EXT4_I(inode)->jinode,
|
|
|
|
new_size);
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
static void ext4_invalidatepage(struct page *page, unsigned long offset);
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* Test whether an inode is a fast symlink.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static int ext4_inode_is_fast_symlink(struct inode *inode)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
int ea_blocks = EXT4_I(inode)->i_file_acl ?
|
2006-10-11 10:20:50 +02:00
|
|
|
(inode->i_sb->s_blocksize >> 9) : 0;
|
|
|
|
|
|
|
|
return (S_ISLNK(inode->i_mode) && inode->i_blocks - ea_blocks == 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Work out how many blocks we need to proceed with the next chunk of a
|
|
|
|
* truncate transaction.
|
|
|
|
*/
|
|
|
|
static unsigned long blocks_for_truncate(struct inode *inode)
|
|
|
|
{
|
2008-01-29 05:58:27 +01:00
|
|
|
ext4_lblk_t needed;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
needed = inode->i_blocks >> (inode->i_sb->s_blocksize_bits - 9);
|
|
|
|
|
|
|
|
/* Give ourselves just enough room to cope with inodes in which
|
|
|
|
* i_blocks is corrupt: we've seen disk corruptions in the past
|
|
|
|
* which resulted in random data in an inode which looked enough
|
2006-10-11 10:20:53 +02:00
|
|
|
* like a regular file for ext4 to try to delete it. Things
|
2006-10-11 10:20:50 +02:00
|
|
|
* will go a bit crazy if that happens, but at least we should
|
|
|
|
* try not to panic the whole kernel. */
|
|
|
|
if (needed < 2)
|
|
|
|
needed = 2;
|
|
|
|
|
|
|
|
/* But we need to bound the transaction so we don't overflow the
|
|
|
|
* journal. */
|
2006-10-11 10:20:53 +02:00
|
|
|
if (needed > EXT4_MAX_TRANS_DATA)
|
|
|
|
needed = EXT4_MAX_TRANS_DATA;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
return EXT4_DATA_TRANS_BLOCKS(inode->i_sb) + needed;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Truncate transactions can be complex and absolutely huge. So we need to
|
|
|
|
* be able to restart the transaction at a conventient checkpoint to make
|
|
|
|
* sure we don't overflow the journal.
|
|
|
|
*
|
|
|
|
* start_transaction gets us a new handle for a truncate transaction,
|
|
|
|
* and extend_transaction tries to extend the existing one a bit. If
|
|
|
|
* extend fails, we need to propagate the failure up and restart the
|
|
|
|
* transaction in the top-level truncate loop. --sct
|
|
|
|
*/
|
|
|
|
static handle_t *start_transaction(struct inode *inode)
|
|
|
|
{
|
|
|
|
handle_t *result;
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
result = ext4_journal_start(inode, blocks_for_truncate(inode));
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!IS_ERR(result))
|
|
|
|
return result;
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_std_error(inode->i_sb, PTR_ERR(result));
|
2006-10-11 10:20:50 +02:00
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Try to extend this transaction for the purposes of truncation.
|
|
|
|
*
|
|
|
|
* Returns 0 if we managed to create more room. If we can't create more
|
|
|
|
* room, and the transaction must be restarted we return 1.
|
|
|
|
*/
|
|
|
|
static int try_to_extend_transaction(handle_t *handle, struct inode *inode)
|
|
|
|
{
|
2009-01-07 06:06:22 +01:00
|
|
|
if (!ext4_handle_valid(handle))
|
|
|
|
return 0;
|
|
|
|
if (ext4_handle_has_enough_credits(handle, EXT4_RESERVE_TRANS_BLOCKS+1))
|
2006-10-11 10:20:50 +02:00
|
|
|
return 0;
|
2006-10-11 10:20:53 +02:00
|
|
|
if (!ext4_journal_extend(handle, blocks_for_truncate(inode)))
|
2006-10-11 10:20:50 +02:00
|
|
|
return 0;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Restart the transaction associated with *handle. This does a commit,
|
|
|
|
* so before we call here everything must be consistently dirtied against
|
|
|
|
* this transaction.
|
|
|
|
*/
|
2009-11-03 00:50:49 +01:00
|
|
|
int ext4_truncate_restart_trans(handle_t *handle, struct inode *inode,
|
2009-08-18 04:17:20 +02:00
|
|
|
int nblocks)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2009-08-18 04:17:20 +02:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
/*
|
2010-05-17 01:00:00 +02:00
|
|
|
* Drop i_data_sem to avoid deadlock with ext4_map_blocks. At this
|
2009-08-18 04:17:20 +02:00
|
|
|
* moment, get_block can be called only for blocks inside i_size since
|
|
|
|
* page cache has been already dropped and writes are blocked by
|
|
|
|
* i_mutex. So we can safely drop the i_data_sem here.
|
|
|
|
*/
|
2009-01-07 06:06:22 +01:00
|
|
|
BUG_ON(EXT4_JOURNAL(inode) == NULL);
|
2006-10-11 10:20:50 +02:00
|
|
|
jbd_debug(2, "restarting handle %p\n", handle);
|
2009-08-18 04:17:20 +02:00
|
|
|
up_write(&EXT4_I(inode)->i_data_sem);
|
|
|
|
ret = ext4_journal_restart(handle, blocks_for_truncate(inode));
|
|
|
|
down_write(&EXT4_I(inode)->i_data_sem);
|
2009-11-03 00:50:49 +01:00
|
|
|
ext4_discard_preallocations(inode);
|
2009-08-18 04:17:20 +02:00
|
|
|
|
|
|
|
return ret;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Called at the last iput() if i_nlink is zero.
|
|
|
|
*/
|
2008-09-09 04:25:24 +02:00
|
|
|
void ext4_delete_inode(struct inode *inode)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
handle_t *handle;
|
2008-08-03 03:10:38 +02:00
|
|
|
int err;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2010-03-03 15:05:06 +01:00
|
|
|
if (!is_bad_inode(inode))
|
2010-03-03 15:05:07 +01:00
|
|
|
dquot_initialize(inode);
|
2010-03-03 15:05:06 +01:00
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
if (ext4_should_order_data(inode))
|
|
|
|
ext4_begin_ordered_truncate(inode, 0);
|
2006-10-11 10:20:50 +02:00
|
|
|
truncate_inode_pages(&inode->i_data, 0);
|
|
|
|
|
|
|
|
if (is_bad_inode(inode))
|
|
|
|
goto no_delete;
|
|
|
|
|
2008-08-03 03:10:38 +02:00
|
|
|
handle = ext4_journal_start(inode, blocks_for_truncate(inode)+3);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (IS_ERR(handle)) {
|
2008-08-03 03:10:38 +02:00
|
|
|
ext4_std_error(inode->i_sb, PTR_ERR(handle));
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* If we're going to skip the normal cleanup, we still need to
|
|
|
|
* make sure that the in-core orphan linked list is properly
|
|
|
|
* cleaned up.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_orphan_del(NULL, inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
goto no_delete;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (IS_SYNC(inode))
|
2009-01-07 06:06:22 +01:00
|
|
|
ext4_handle_sync(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
inode->i_size = 0;
|
2008-08-03 03:10:38 +02:00
|
|
|
err = ext4_mark_inode_dirty(handle, inode);
|
|
|
|
if (err) {
|
2010-02-15 20:19:27 +01:00
|
|
|
ext4_warning(inode->i_sb,
|
2008-08-03 03:10:38 +02:00
|
|
|
"couldn't mark inode dirty (err %d)", err);
|
|
|
|
goto stop_handle;
|
|
|
|
}
|
2006-10-11 10:20:50 +02:00
|
|
|
if (inode->i_blocks)
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_truncate(inode);
|
2008-08-03 03:10:38 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* ext4_ext_truncate() doesn't reserve any slop when it
|
|
|
|
* restarts journal transactions; therefore there may not be
|
|
|
|
* enough credits left in the handle to remove the inode from
|
|
|
|
* the orphan list and set the dtime field.
|
|
|
|
*/
|
2009-01-07 06:06:22 +01:00
|
|
|
if (!ext4_handle_has_enough_credits(handle, 3)) {
|
2008-08-03 03:10:38 +02:00
|
|
|
err = ext4_journal_extend(handle, 3);
|
|
|
|
if (err > 0)
|
|
|
|
err = ext4_journal_restart(handle, 3);
|
|
|
|
if (err != 0) {
|
2010-02-15 20:19:27 +01:00
|
|
|
ext4_warning(inode->i_sb,
|
2008-08-03 03:10:38 +02:00
|
|
|
"couldn't extend journal (err %d)", err);
|
|
|
|
stop_handle:
|
|
|
|
ext4_journal_stop(handle);
|
|
|
|
goto no_delete;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* Kill off the orphan record which ext4_truncate created.
|
2006-10-11 10:20:50 +02:00
|
|
|
* AKPM: I think this can be inside the above `if'.
|
2006-10-11 10:20:53 +02:00
|
|
|
* Note that ext4_orphan_del() has to be able to cope with the
|
2006-10-11 10:20:50 +02:00
|
|
|
* deletion of a non-existent orphan - this is because we don't
|
2006-10-11 10:20:53 +02:00
|
|
|
* know if ext4_truncate() actually created an orphan record.
|
2006-10-11 10:20:50 +02:00
|
|
|
* (Well, we could do this if we need to, but heck - it works)
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_orphan_del(handle, inode);
|
|
|
|
EXT4_I(inode)->i_dtime = get_seconds();
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* One subtle ordering requirement: if anything has gone wrong
|
|
|
|
* (transaction abort, IO errors, whatever), then we can still
|
|
|
|
* do these next steps (the fs will already have been marked as
|
|
|
|
* having errors), but we can't free the inode if the mark_dirty
|
|
|
|
* fails.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
if (ext4_mark_inode_dirty(handle, inode))
|
2006-10-11 10:20:50 +02:00
|
|
|
/* If that failed, just do the required in-core inode clear. */
|
|
|
|
clear_inode(inode);
|
|
|
|
else
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_free_inode(handle, inode);
|
|
|
|
ext4_journal_stop(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
return;
|
|
|
|
no_delete:
|
|
|
|
clear_inode(inode); /* We must guarantee clearing of inode... */
|
|
|
|
}
|
|
|
|
|
|
|
|
typedef struct {
|
|
|
|
__le32 *p;
|
|
|
|
__le32 key;
|
|
|
|
struct buffer_head *bh;
|
|
|
|
} Indirect;
|
|
|
|
|
|
|
|
static inline void add_chain(Indirect *p, struct buffer_head *bh, __le32 *v)
|
|
|
|
{
|
|
|
|
p->key = *(p->p = v);
|
|
|
|
p->bh = bh;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_block_to_path - parse the block number into array of offsets
|
2006-10-11 10:20:50 +02:00
|
|
|
* @inode: inode in question (we are only interested in its superblock)
|
|
|
|
* @i_block: block number to be parsed
|
|
|
|
* @offsets: array to store the offsets in
|
2007-05-24 19:04:54 +02:00
|
|
|
* @boundary: set this non-zero if the referred-to block is likely to be
|
|
|
|
* followed (on disk) by an indirect block.
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
2006-10-11 10:20:53 +02:00
|
|
|
* To store the locations of file's data ext4 uses a data structure common
|
2006-10-11 10:20:50 +02:00
|
|
|
* for UNIX filesystems - tree of pointers anchored in the inode, with
|
|
|
|
* data blocks at leaves and indirect blocks in intermediate nodes.
|
|
|
|
* This function translates the block number into path in that tree -
|
|
|
|
* return value is the path length and @offsets[n] is the offset of
|
|
|
|
* pointer to (n+1)th node in the nth one. If @block is out of range
|
|
|
|
* (negative or too large) warning is printed and zero returned.
|
|
|
|
*
|
|
|
|
* Note: function doesn't find node addresses, so no IO is needed. All
|
|
|
|
* we need to know is the capacity of indirect blocks (taken from the
|
|
|
|
* inode->i_sb).
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Portability note: the last comparison (check that we fit into triple
|
|
|
|
* indirect block) is spelled differently, because otherwise on an
|
|
|
|
* architecture with 32-bit longs and 8Kb pages we might get into trouble
|
|
|
|
* if our filesystem had 8Kb blocks. We might use long long, but that would
|
|
|
|
* kill us on x86. Oh, well, at least the sign propagation does not matter -
|
|
|
|
* i_block would have to be negative in the very beginning, so we would not
|
|
|
|
* get there at all.
|
|
|
|
*/
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
static int ext4_block_to_path(struct inode *inode,
|
2009-06-14 23:45:34 +02:00
|
|
|
ext4_lblk_t i_block,
|
|
|
|
ext4_lblk_t offsets[4], int *boundary)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
int ptrs = EXT4_ADDR_PER_BLOCK(inode->i_sb);
|
|
|
|
int ptrs_bits = EXT4_ADDR_PER_BLOCK_BITS(inode->i_sb);
|
|
|
|
const long direct_blocks = EXT4_NDIR_BLOCKS,
|
2006-10-11 10:20:50 +02:00
|
|
|
indirect_blocks = ptrs,
|
|
|
|
double_blocks = (1 << (ptrs_bits * 2));
|
|
|
|
int n = 0;
|
|
|
|
int final = 0;
|
|
|
|
|
2009-08-11 04:47:22 +02:00
|
|
|
if (i_block < direct_blocks) {
|
2006-10-11 10:20:50 +02:00
|
|
|
offsets[n++] = i_block;
|
|
|
|
final = direct_blocks;
|
2008-09-09 04:25:24 +02:00
|
|
|
} else if ((i_block -= direct_blocks) < indirect_blocks) {
|
2006-10-11 10:20:53 +02:00
|
|
|
offsets[n++] = EXT4_IND_BLOCK;
|
2006-10-11 10:20:50 +02:00
|
|
|
offsets[n++] = i_block;
|
|
|
|
final = ptrs;
|
|
|
|
} else if ((i_block -= indirect_blocks) < double_blocks) {
|
2006-10-11 10:20:53 +02:00
|
|
|
offsets[n++] = EXT4_DIND_BLOCK;
|
2006-10-11 10:20:50 +02:00
|
|
|
offsets[n++] = i_block >> ptrs_bits;
|
|
|
|
offsets[n++] = i_block & (ptrs - 1);
|
|
|
|
final = ptrs;
|
|
|
|
} else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) {
|
2006-10-11 10:20:53 +02:00
|
|
|
offsets[n++] = EXT4_TIND_BLOCK;
|
2006-10-11 10:20:50 +02:00
|
|
|
offsets[n++] = i_block >> (ptrs_bits * 2);
|
|
|
|
offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1);
|
|
|
|
offsets[n++] = i_block & (ptrs - 1);
|
|
|
|
final = ptrs;
|
|
|
|
} else {
|
2010-02-15 20:19:27 +01:00
|
|
|
ext4_warning(inode->i_sb, "block %lu > max in inode %lu",
|
2009-06-14 23:45:34 +02:00
|
|
|
i_block + direct_blocks +
|
|
|
|
indirect_blocks + double_blocks, inode->i_ino);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
if (boundary)
|
|
|
|
*boundary = final - 1 - (i_block & (ptrs - 1));
|
|
|
|
return n;
|
|
|
|
}
|
|
|
|
|
2009-03-31 14:36:10 +02:00
|
|
|
static int __ext4_check_blockref(const char *function, struct inode *inode,
|
2009-05-17 21:38:01 +02:00
|
|
|
__le32 *p, unsigned int max)
|
|
|
|
{
|
2009-04-08 00:46:47 +02:00
|
|
|
__le32 *bref = p;
|
2009-05-17 21:38:01 +02:00
|
|
|
unsigned int blk;
|
|
|
|
|
2009-03-31 14:36:10 +02:00
|
|
|
while (bref < p+max) {
|
2009-05-17 21:38:01 +02:00
|
|
|
blk = le32_to_cpu(*bref++);
|
2009-06-14 23:45:34 +02:00
|
|
|
if (blk &&
|
|
|
|
unlikely(!ext4_data_block_valid(EXT4_SB(inode->i_sb),
|
2009-05-17 21:38:01 +02:00
|
|
|
blk, 1))) {
|
2010-05-17 03:00:00 +02:00
|
|
|
ext4_error_inode(function, inode,
|
|
|
|
"invalid block reference %u", blk);
|
2009-06-14 23:45:34 +02:00
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return 0;
|
2009-03-31 14:36:10 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
#define ext4_check_indirect_blockref(inode, bh) \
|
2009-06-14 23:45:34 +02:00
|
|
|
__ext4_check_blockref(__func__, inode, (__le32 *)(bh)->b_data, \
|
2009-03-31 14:36:10 +02:00
|
|
|
EXT4_ADDR_PER_BLOCK((inode)->i_sb))
|
|
|
|
|
|
|
|
#define ext4_check_inode_blockref(inode) \
|
2009-06-14 23:45:34 +02:00
|
|
|
__ext4_check_blockref(__func__, inode, EXT4_I(inode)->i_data, \
|
2009-03-31 14:36:10 +02:00
|
|
|
EXT4_NDIR_BLOCKS)
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/**
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_get_branch - read the chain of indirect blocks leading to data
|
2006-10-11 10:20:50 +02:00
|
|
|
* @inode: inode in question
|
|
|
|
* @depth: depth of the chain (1 - direct pointer, etc.)
|
|
|
|
* @offsets: offsets of pointers in inode/indirect blocks
|
|
|
|
* @chain: place to store the result
|
|
|
|
* @err: here we store the error value
|
|
|
|
*
|
|
|
|
* Function fills the array of triples <key, p, bh> and returns %NULL
|
|
|
|
* if everything went OK or the pointer to the last filled triple
|
|
|
|
* (incomplete one) otherwise. Upon the return chain[i].key contains
|
|
|
|
* the number of (i+1)-th block in the chain (as it is stored in memory,
|
|
|
|
* i.e. little-endian 32-bit), chain[i].p contains the address of that
|
|
|
|
* number (it points into struct inode for i==0 and into the bh->b_data
|
|
|
|
* for i>0) and chain[i].bh points to the buffer_head of i-th indirect
|
|
|
|
* block for i>0 and NULL for i==0. In other words, it holds the block
|
|
|
|
* numbers of the chain, addresses they were taken from (and where we can
|
|
|
|
* verify that chain did not change) and buffer_heads hosting these
|
|
|
|
* numbers.
|
|
|
|
*
|
|
|
|
* Function stops when it stumbles upon zero pointer (absent block)
|
|
|
|
* (pointer to last triple returned, *@err == 0)
|
|
|
|
* or when it gets an IO error reading an indirect block
|
|
|
|
* (ditto, *@err == -EIO)
|
|
|
|
* or when it reads all @depth-1 indirect blocks successfully and finds
|
|
|
|
* the whole chain, all way to the data (returns %NULL, *err == 0).
|
2008-01-29 05:58:27 +01:00
|
|
|
*
|
|
|
|
* Need to be called with
|
2008-01-29 05:58:26 +01:00
|
|
|
* down_read(&EXT4_I(inode)->i_data_sem)
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
2008-01-29 05:58:27 +01:00
|
|
|
static Indirect *ext4_get_branch(struct inode *inode, int depth,
|
|
|
|
ext4_lblk_t *offsets,
|
2006-10-11 10:20:50 +02:00
|
|
|
Indirect chain[4], int *err)
|
|
|
|
{
|
|
|
|
struct super_block *sb = inode->i_sb;
|
|
|
|
Indirect *p = chain;
|
|
|
|
struct buffer_head *bh;
|
|
|
|
|
|
|
|
*err = 0;
|
|
|
|
/* i_data is not going away, no lock needed */
|
2008-09-09 04:25:24 +02:00
|
|
|
add_chain(chain, NULL, EXT4_I(inode)->i_data + *offsets);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!p->key)
|
|
|
|
goto no_block;
|
|
|
|
while (--depth) {
|
2009-03-31 14:36:10 +02:00
|
|
|
bh = sb_getblk(sb, le32_to_cpu(p->key));
|
|
|
|
if (unlikely(!bh))
|
2006-10-11 10:20:50 +02:00
|
|
|
goto failure;
|
2009-06-14 23:45:34 +02:00
|
|
|
|
2009-03-31 14:36:10 +02:00
|
|
|
if (!bh_uptodate_or_lock(bh)) {
|
|
|
|
if (bh_submit_read(bh) < 0) {
|
|
|
|
put_bh(bh);
|
|
|
|
goto failure;
|
|
|
|
}
|
|
|
|
/* validate block references */
|
|
|
|
if (ext4_check_indirect_blockref(inode, bh)) {
|
|
|
|
put_bh(bh);
|
|
|
|
goto failure;
|
|
|
|
}
|
|
|
|
}
|
2009-06-14 23:45:34 +02:00
|
|
|
|
2008-09-09 04:25:24 +02:00
|
|
|
add_chain(++p, bh, (__le32 *)bh->b_data + *++offsets);
|
2006-10-11 10:20:50 +02:00
|
|
|
/* Reader: end */
|
|
|
|
if (!p->key)
|
|
|
|
goto no_block;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
failure:
|
|
|
|
*err = -EIO;
|
|
|
|
no_block:
|
|
|
|
return p;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_find_near - find a place for allocation with sufficient locality
|
2006-10-11 10:20:50 +02:00
|
|
|
* @inode: owner
|
|
|
|
* @ind: descriptor of indirect block.
|
|
|
|
*
|
2008-04-22 00:45:55 +02:00
|
|
|
* This function returns the preferred place for block allocation.
|
2006-10-11 10:20:50 +02:00
|
|
|
* It is used when heuristic for sequential allocation fails.
|
|
|
|
* Rules are:
|
|
|
|
* + if there is a block to the left of our position - allocate near it.
|
|
|
|
* + if pointer will live in indirect block - allocate near that block.
|
|
|
|
* + if pointer will live in inode - allocate in the same
|
|
|
|
* cylinder group.
|
|
|
|
*
|
|
|
|
* In the latter case we colour the starting block by the callers PID to
|
|
|
|
* prevent it from clashing with concurrent allocations for a different inode
|
|
|
|
* in the same block group. The PID is used here so that functionally related
|
|
|
|
* files will be close-by on-disk.
|
|
|
|
*
|
|
|
|
* Caller must make sure that @ind is valid and will stay that way.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static ext4_fsblk_t ext4_find_near(struct inode *inode, Indirect *ind)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
struct ext4_inode_info *ei = EXT4_I(inode);
|
2008-09-09 04:25:24 +02:00
|
|
|
__le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data;
|
2006-10-11 10:20:50 +02:00
|
|
|
__le32 *p;
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_fsblk_t bg_start;
|
2008-02-15 19:43:07 +01:00
|
|
|
ext4_fsblk_t last_block;
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_grpblk_t colour;
|
2009-03-12 17:18:34 +01:00
|
|
|
ext4_group_t block_group;
|
|
|
|
int flex_size = ext4_flex_bg_size(EXT4_SB(inode->i_sb));
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/* Try to find previous block */
|
|
|
|
for (p = ind->p - 1; p >= start; p--) {
|
|
|
|
if (*p)
|
|
|
|
return le32_to_cpu(*p);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* No such thing, so let's try location of indirect block */
|
|
|
|
if (ind->bh)
|
|
|
|
return ind->bh->b_blocknr;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* It is going to be referred to from the inode itself? OK, just put it
|
|
|
|
* into the same cylinder group then.
|
|
|
|
*/
|
2009-03-12 17:18:34 +01:00
|
|
|
block_group = ei->i_block_group;
|
|
|
|
if (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) {
|
|
|
|
block_group &= ~(flex_size-1);
|
|
|
|
if (S_ISREG(inode->i_mode))
|
|
|
|
block_group++;
|
|
|
|
}
|
|
|
|
bg_start = ext4_group_first_block_no(inode->i_sb, block_group);
|
2008-02-15 19:43:07 +01:00
|
|
|
last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1;
|
|
|
|
|
2009-03-12 17:18:34 +01:00
|
|
|
/*
|
|
|
|
* If we are doing delayed allocation, we don't need take
|
|
|
|
* colour into account.
|
|
|
|
*/
|
|
|
|
if (test_opt(inode->i_sb, DELALLOC))
|
|
|
|
return bg_start;
|
|
|
|
|
2008-02-15 19:43:07 +01:00
|
|
|
if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block)
|
|
|
|
colour = (current->pid % 16) *
|
2006-10-11 10:20:53 +02:00
|
|
|
(EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
|
2008-02-15 19:43:07 +01:00
|
|
|
else
|
|
|
|
colour = (current->pid % 16) * ((last_block - bg_start) / 16);
|
2006-10-11 10:20:50 +02:00
|
|
|
return bg_start + colour;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2008-04-22 00:45:55 +02:00
|
|
|
* ext4_find_goal - find a preferred place for allocation.
|
2006-10-11 10:20:50 +02:00
|
|
|
* @inode: owner
|
|
|
|
* @block: block we want
|
|
|
|
* @partial: pointer to the last triple within a chain
|
|
|
|
*
|
2008-04-22 00:45:55 +02:00
|
|
|
* Normally this function find the preferred place for block allocation,
|
2008-02-06 10:40:16 +01:00
|
|
|
* returns it.
|
ext4: limit block allocations for indirect-block files to < 2^32
Today, the ext4 allocator will happily allocate blocks past
2^32 for indirect-block files, which results in the block
numbers getting truncated, and corruption ensues.
This patch limits such allocations to < 2^32, and adds
BUG_ONs if we do get blocks larger than that.
This should address RH Bug 519471, ext4 bitmap allocator
must limit blocks to < 2^32
* ext4_find_goal() is modified to choose a goal < UINT_MAX,
so that our starting point is in an acceptable range.
* ext4_xattr_block_set() is modified such that the goal block
is < UINT_MAX, as above.
* ext4_mb_regular_allocator() is modified so that the group
search does not continue into groups which are too high
* ext4_mb_use_preallocated() has a check that we don't use
preallocated space which is too far out
* ext4_alloc_blocks() and ext4_xattr_block_set() add some BUG_ONs
No attempt has been made to limit inode locations to < 2^32,
so we may wind up with blocks far from their inodes. Doing
this much already will lead to some odd ENOSPC issues when the
"lower 32" gets full, and further restricting inodes could
make that even weirder.
For high inodes, choosing a goal of the original, % UINT_MAX,
may be a bit odd, but then we're in an odd situation anyway,
and I don't know of a better heuristic.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2009-09-16 20:45:10 +02:00
|
|
|
* Because this is only used for non-extent files, we limit the block nr
|
|
|
|
* to 32 bits.
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
2008-01-29 05:58:27 +01:00
|
|
|
static ext4_fsblk_t ext4_find_goal(struct inode *inode, ext4_lblk_t block,
|
2009-06-14 23:45:34 +02:00
|
|
|
Indirect *partial)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
ext4: limit block allocations for indirect-block files to < 2^32
Today, the ext4 allocator will happily allocate blocks past
2^32 for indirect-block files, which results in the block
numbers getting truncated, and corruption ensues.
This patch limits such allocations to < 2^32, and adds
BUG_ONs if we do get blocks larger than that.
This should address RH Bug 519471, ext4 bitmap allocator
must limit blocks to < 2^32
* ext4_find_goal() is modified to choose a goal < UINT_MAX,
so that our starting point is in an acceptable range.
* ext4_xattr_block_set() is modified such that the goal block
is < UINT_MAX, as above.
* ext4_mb_regular_allocator() is modified so that the group
search does not continue into groups which are too high
* ext4_mb_use_preallocated() has a check that we don't use
preallocated space which is too far out
* ext4_alloc_blocks() and ext4_xattr_block_set() add some BUG_ONs
No attempt has been made to limit inode locations to < 2^32,
so we may wind up with blocks far from their inodes. Doing
this much already will lead to some odd ENOSPC issues when the
"lower 32" gets full, and further restricting inodes could
make that even weirder.
For high inodes, choosing a goal of the original, % UINT_MAX,
may be a bit odd, but then we're in an odd situation anyway,
and I don't know of a better heuristic.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2009-09-16 20:45:10 +02:00
|
|
|
ext4_fsblk_t goal;
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
2008-10-10 15:40:52 +02:00
|
|
|
* XXX need to get goal block from mballoc's data structures
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
|
|
|
|
ext4: limit block allocations for indirect-block files to < 2^32
Today, the ext4 allocator will happily allocate blocks past
2^32 for indirect-block files, which results in the block
numbers getting truncated, and corruption ensues.
This patch limits such allocations to < 2^32, and adds
BUG_ONs if we do get blocks larger than that.
This should address RH Bug 519471, ext4 bitmap allocator
must limit blocks to < 2^32
* ext4_find_goal() is modified to choose a goal < UINT_MAX,
so that our starting point is in an acceptable range.
* ext4_xattr_block_set() is modified such that the goal block
is < UINT_MAX, as above.
* ext4_mb_regular_allocator() is modified so that the group
search does not continue into groups which are too high
* ext4_mb_use_preallocated() has a check that we don't use
preallocated space which is too far out
* ext4_alloc_blocks() and ext4_xattr_block_set() add some BUG_ONs
No attempt has been made to limit inode locations to < 2^32,
so we may wind up with blocks far from their inodes. Doing
this much already will lead to some odd ENOSPC issues when the
"lower 32" gets full, and further restricting inodes could
make that even weirder.
For high inodes, choosing a goal of the original, % UINT_MAX,
may be a bit odd, but then we're in an odd situation anyway,
and I don't know of a better heuristic.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2009-09-16 20:45:10 +02:00
|
|
|
goal = ext4_find_near(inode, partial);
|
|
|
|
goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
|
|
|
|
return goal;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_blks_to_allocate: Look up the block map and count the number
|
2006-10-11 10:20:50 +02:00
|
|
|
* of direct blocks need to be allocated for the given branch.
|
|
|
|
*
|
|
|
|
* @branch: chain of indirect blocks
|
|
|
|
* @k: number of blocks need for indirect blocks
|
|
|
|
* @blks: number of data blocks to be mapped.
|
|
|
|
* @blocks_to_boundary: the offset in the indirect block
|
|
|
|
*
|
|
|
|
* return the total number of blocks to be allocate, including the
|
|
|
|
* direct and indirect blocks.
|
|
|
|
*/
|
2008-11-05 06:14:04 +01:00
|
|
|
static int ext4_blks_to_allocate(Indirect *branch, int k, unsigned int blks,
|
2009-06-14 23:45:34 +02:00
|
|
|
int blocks_to_boundary)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2008-11-05 06:14:04 +01:00
|
|
|
unsigned int count = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Simple case, [t,d]Indirect block(s) has not allocated yet
|
|
|
|
* then it's clear blocks on that path have not allocated
|
|
|
|
*/
|
|
|
|
if (k > 0) {
|
|
|
|
/* right now we don't handle cross boundary allocation */
|
|
|
|
if (blks < blocks_to_boundary + 1)
|
|
|
|
count += blks;
|
|
|
|
else
|
|
|
|
count += blocks_to_boundary + 1;
|
|
|
|
return count;
|
|
|
|
}
|
|
|
|
|
|
|
|
count++;
|
|
|
|
while (count < blks && count <= blocks_to_boundary &&
|
|
|
|
le32_to_cpu(*(branch[0].p + count)) == 0) {
|
|
|
|
count++;
|
|
|
|
}
|
|
|
|
return count;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_alloc_blocks: multiple allocate blocks needed for a branch
|
2006-10-11 10:20:50 +02:00
|
|
|
* @indirect_blks: the number of blocks need to allocate for indirect
|
|
|
|
* blocks
|
|
|
|
*
|
|
|
|
* @new_blocks: on return it will store the new block numbers for
|
|
|
|
* the indirect blocks(if needed) and the first direct block,
|
|
|
|
* @blks: on return it will store the total number of allocated
|
|
|
|
* direct blocks
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static int ext4_alloc_blocks(handle_t *handle, struct inode *inode,
|
2009-06-14 23:45:34 +02:00
|
|
|
ext4_lblk_t iblock, ext4_fsblk_t goal,
|
|
|
|
int indirect_blks, int blks,
|
|
|
|
ext4_fsblk_t new_blocks[4], int *err)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2009-01-02 05:59:43 +01:00
|
|
|
struct ext4_allocation_request ar;
|
2006-10-11 10:20:50 +02:00
|
|
|
int target, i;
|
2008-07-12 01:27:31 +02:00
|
|
|
unsigned long count = 0, blk_allocated = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
int index = 0;
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_fsblk_t current_block = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Here we try to allocate the requested multiple blocks at once,
|
|
|
|
* on a best-effort basis.
|
|
|
|
* To build a branch, we should allocate blocks for
|
|
|
|
* the indirect blocks(if not allocated yet), and at least
|
|
|
|
* the first direct block of this branch. That's the
|
|
|
|
* minimum number of blocks need to allocate(required)
|
|
|
|
*/
|
2008-07-12 01:27:31 +02:00
|
|
|
/* first we try to allocate the indirect blocks */
|
|
|
|
target = indirect_blks;
|
|
|
|
while (target > 0) {
|
2006-10-11 10:20:50 +02:00
|
|
|
count = target;
|
|
|
|
/* allocating blocks for indirect blocks and direct blocks */
|
2008-07-12 01:27:31 +02:00
|
|
|
current_block = ext4_new_meta_blocks(handle, inode,
|
|
|
|
goal, &count, err);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (*err)
|
|
|
|
goto failed_out;
|
|
|
|
|
2010-03-02 17:46:09 +01:00
|
|
|
if (unlikely(current_block + count > EXT4_MAX_BLOCK_FILE_PHYS)) {
|
|
|
|
EXT4_ERROR_INODE(inode,
|
|
|
|
"current_block %llu + count %lu > %d!",
|
|
|
|
current_block, count,
|
|
|
|
EXT4_MAX_BLOCK_FILE_PHYS);
|
|
|
|
*err = -EIO;
|
|
|
|
goto failed_out;
|
|
|
|
}
|
ext4: limit block allocations for indirect-block files to < 2^32
Today, the ext4 allocator will happily allocate blocks past
2^32 for indirect-block files, which results in the block
numbers getting truncated, and corruption ensues.
This patch limits such allocations to < 2^32, and adds
BUG_ONs if we do get blocks larger than that.
This should address RH Bug 519471, ext4 bitmap allocator
must limit blocks to < 2^32
* ext4_find_goal() is modified to choose a goal < UINT_MAX,
so that our starting point is in an acceptable range.
* ext4_xattr_block_set() is modified such that the goal block
is < UINT_MAX, as above.
* ext4_mb_regular_allocator() is modified so that the group
search does not continue into groups which are too high
* ext4_mb_use_preallocated() has a check that we don't use
preallocated space which is too far out
* ext4_alloc_blocks() and ext4_xattr_block_set() add some BUG_ONs
No attempt has been made to limit inode locations to < 2^32,
so we may wind up with blocks far from their inodes. Doing
this much already will lead to some odd ENOSPC issues when the
"lower 32" gets full, and further restricting inodes could
make that even weirder.
For high inodes, choosing a goal of the original, % UINT_MAX,
may be a bit odd, but then we're in an odd situation anyway,
and I don't know of a better heuristic.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2009-09-16 20:45:10 +02:00
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
target -= count;
|
|
|
|
/* allocate blocks for indirect blocks */
|
|
|
|
while (index < indirect_blks && count) {
|
|
|
|
new_blocks[index++] = current_block++;
|
|
|
|
count--;
|
|
|
|
}
|
2008-07-12 01:27:31 +02:00
|
|
|
if (count > 0) {
|
|
|
|
/*
|
|
|
|
* save the new block number
|
|
|
|
* for the first direct block
|
|
|
|
*/
|
|
|
|
new_blocks[index] = current_block;
|
|
|
|
printk(KERN_INFO "%s returned more blocks than "
|
|
|
|
"requested\n", __func__);
|
|
|
|
WARN_ON(1);
|
2006-10-11 10:20:50 +02:00
|
|
|
break;
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
target = blks - count ;
|
|
|
|
blk_allocated = count;
|
|
|
|
if (!target)
|
|
|
|
goto allocated;
|
|
|
|
/* Now allocate data blocks */
|
2009-01-02 05:59:43 +01:00
|
|
|
memset(&ar, 0, sizeof(ar));
|
|
|
|
ar.inode = inode;
|
|
|
|
ar.goal = goal;
|
|
|
|
ar.len = target;
|
|
|
|
ar.logical = iblock;
|
|
|
|
if (S_ISREG(inode->i_mode))
|
|
|
|
/* enable in-core preallocation only for regular files */
|
|
|
|
ar.flags = EXT4_MB_HINT_DATA;
|
|
|
|
|
|
|
|
current_block = ext4_mb_new_blocks(handle, &ar, err);
|
2010-03-02 17:46:09 +01:00
|
|
|
if (unlikely(current_block + ar.len > EXT4_MAX_BLOCK_FILE_PHYS)) {
|
|
|
|
EXT4_ERROR_INODE(inode,
|
|
|
|
"current_block %llu + ar.len %d > %d!",
|
|
|
|
current_block, ar.len,
|
|
|
|
EXT4_MAX_BLOCK_FILE_PHYS);
|
|
|
|
*err = -EIO;
|
|
|
|
goto failed_out;
|
|
|
|
}
|
2009-01-02 05:59:43 +01:00
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
if (*err && (target == blks)) {
|
|
|
|
/*
|
|
|
|
* if the allocation failed and we didn't allocate
|
|
|
|
* any blocks before
|
|
|
|
*/
|
|
|
|
goto failed_out;
|
|
|
|
}
|
|
|
|
if (!*err) {
|
|
|
|
if (target == blks) {
|
2009-06-14 23:45:34 +02:00
|
|
|
/*
|
|
|
|
* save the new block number
|
|
|
|
* for the first direct block
|
|
|
|
*/
|
2008-07-12 01:27:31 +02:00
|
|
|
new_blocks[index] = current_block;
|
|
|
|
}
|
2009-01-02 05:59:43 +01:00
|
|
|
blk_allocated += ar.len;
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
|
|
|
allocated:
|
2006-10-11 10:20:50 +02:00
|
|
|
/* total number of blocks allocated for direct blocks */
|
2008-07-12 01:27:31 +02:00
|
|
|
ret = blk_allocated;
|
2006-10-11 10:20:50 +02:00
|
|
|
*err = 0;
|
|
|
|
return ret;
|
|
|
|
failed_out:
|
2008-09-09 04:25:24 +02:00
|
|
|
for (i = 0; i < index; i++)
|
2009-11-23 13:17:05 +01:00
|
|
|
ext4_free_blocks(handle, inode, 0, new_blocks[i], 1, 0);
|
2006-10-11 10:20:50 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_alloc_branch - allocate and set up a chain of blocks.
|
2006-10-11 10:20:50 +02:00
|
|
|
* @inode: owner
|
|
|
|
* @indirect_blks: number of allocated indirect blocks
|
|
|
|
* @blks: number of allocated direct blocks
|
|
|
|
* @offsets: offsets (in the blocks) to store the pointers to next.
|
|
|
|
* @branch: place to store the chain in.
|
|
|
|
*
|
|
|
|
* This function allocates blocks, zeroes out all but the last one,
|
|
|
|
* links them into chain and (if we are synchronous) writes them to disk.
|
|
|
|
* In other words, it prepares a branch that can be spliced onto the
|
|
|
|
* inode. It stores the information about that chain in the branch[], in
|
2006-10-11 10:20:53 +02:00
|
|
|
* the same format as ext4_get_branch() would do. We are calling it after
|
2006-10-11 10:20:50 +02:00
|
|
|
* we had read the existing part of chain and partial points to the last
|
|
|
|
* triple of that (one with zero ->key). Upon the exit we have the same
|
2006-10-11 10:20:53 +02:00
|
|
|
* picture as after the successful ext4_get_block(), except that in one
|
2006-10-11 10:20:50 +02:00
|
|
|
* place chain is disconnected - *branch->p is still zero (we did not
|
|
|
|
* set the last link), but branch->key contains the number that should
|
|
|
|
* be placed into *branch->p to fill that gap.
|
|
|
|
*
|
|
|
|
* If allocation fails we free all blocks we've allocated (and forget
|
|
|
|
* their buffer_heads) and return the error value the from failed
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_alloc_block() (normally -ENOSPC). Otherwise we set the chain
|
2006-10-11 10:20:50 +02:00
|
|
|
* as described above and return 0.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static int ext4_alloc_branch(handle_t *handle, struct inode *inode,
|
2009-06-14 23:45:34 +02:00
|
|
|
ext4_lblk_t iblock, int indirect_blks,
|
|
|
|
int *blks, ext4_fsblk_t goal,
|
|
|
|
ext4_lblk_t *offsets, Indirect *branch)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
int blocksize = inode->i_sb->s_blocksize;
|
|
|
|
int i, n = 0;
|
|
|
|
int err = 0;
|
|
|
|
struct buffer_head *bh;
|
|
|
|
int num;
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_fsblk_t new_blocks[4];
|
|
|
|
ext4_fsblk_t current_block;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
num = ext4_alloc_blocks(handle, inode, iblock, goal, indirect_blks,
|
2006-10-11 10:20:50 +02:00
|
|
|
*blks, new_blocks, &err);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
branch[0].key = cpu_to_le32(new_blocks[0]);
|
|
|
|
/*
|
|
|
|
* metadata blocks and data blocks are allocated.
|
|
|
|
*/
|
|
|
|
for (n = 1; n <= indirect_blks; n++) {
|
|
|
|
/*
|
|
|
|
* Get buffer_head for parent block, zero it out
|
|
|
|
* and set the pointer to new one, then send
|
|
|
|
* parent to disk.
|
|
|
|
*/
|
|
|
|
bh = sb_getblk(inode->i_sb, new_blocks[n-1]);
|
|
|
|
branch[n].bh = bh;
|
|
|
|
lock_buffer(bh);
|
|
|
|
BUFFER_TRACE(bh, "call get_create_access");
|
2006-10-11 10:20:53 +02:00
|
|
|
err = ext4_journal_get_create_access(handle, bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (err) {
|
2009-07-17 16:54:08 +02:00
|
|
|
/* Don't brelse(bh) here; it's done in
|
|
|
|
* ext4_journal_forget() below */
|
2006-10-11 10:20:50 +02:00
|
|
|
unlock_buffer(bh);
|
|
|
|
goto failed;
|
|
|
|
}
|
|
|
|
|
|
|
|
memset(bh->b_data, 0, blocksize);
|
|
|
|
branch[n].p = (__le32 *) bh->b_data + offsets[n];
|
|
|
|
branch[n].key = cpu_to_le32(new_blocks[n]);
|
|
|
|
*branch[n].p = branch[n].key;
|
2008-09-09 04:25:24 +02:00
|
|
|
if (n == indirect_blks) {
|
2006-10-11 10:20:50 +02:00
|
|
|
current_block = new_blocks[n];
|
|
|
|
/*
|
|
|
|
* End of chain, update the last new metablock of
|
|
|
|
* the chain to point to the new allocated
|
|
|
|
* data blocks numbers
|
|
|
|
*/
|
2009-06-14 23:45:34 +02:00
|
|
|
for (i = 1; i < num; i++)
|
2006-10-11 10:20:50 +02:00
|
|
|
*(branch[n].p + i) = cpu_to_le32(++current_block);
|
|
|
|
}
|
|
|
|
BUFFER_TRACE(bh, "marking uptodate");
|
|
|
|
set_buffer_uptodate(bh);
|
|
|
|
unlock_buffer(bh);
|
|
|
|
|
2009-01-07 06:06:22 +01:00
|
|
|
BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
|
|
|
|
err = ext4_handle_dirty_metadata(handle, inode, bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (err)
|
|
|
|
goto failed;
|
|
|
|
}
|
|
|
|
*blks = num;
|
|
|
|
return err;
|
|
|
|
failed:
|
|
|
|
/* Allocation failed, free what we already allocated */
|
2009-11-23 13:17:05 +01:00
|
|
|
ext4_free_blocks(handle, inode, 0, new_blocks[0], 1, 0);
|
2006-10-11 10:20:50 +02:00
|
|
|
for (i = 1; i <= n ; i++) {
|
2010-05-17 13:00:00 +02:00
|
|
|
/*
|
2009-11-23 13:17:05 +01:00
|
|
|
* branch[i].bh is newly allocated, so there is no
|
|
|
|
* need to revoke the block, which is why we don't
|
|
|
|
* need to set EXT4_FREE_BLOCKS_METADATA.
|
2009-11-23 03:00:13 +01:00
|
|
|
*/
|
2009-11-23 13:17:05 +01:00
|
|
|
ext4_free_blocks(handle, inode, 0, new_blocks[i], 1,
|
|
|
|
EXT4_FREE_BLOCKS_FORGET);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2009-11-23 13:17:05 +01:00
|
|
|
for (i = n+1; i < indirect_blks; i++)
|
|
|
|
ext4_free_blocks(handle, inode, 0, new_blocks[i], 1, 0);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2009-11-23 13:17:05 +01:00
|
|
|
ext4_free_blocks(handle, inode, 0, new_blocks[i], num, 0);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_splice_branch - splice the allocated branch onto inode.
|
2006-10-11 10:20:50 +02:00
|
|
|
* @inode: owner
|
|
|
|
* @block: (logical) number of block we are adding
|
|
|
|
* @chain: chain of indirect blocks (with a missing link - see
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_alloc_branch)
|
2006-10-11 10:20:50 +02:00
|
|
|
* @where: location of missing link
|
|
|
|
* @num: number of indirect blocks we are adding
|
|
|
|
* @blks: number of direct blocks we are adding
|
|
|
|
*
|
|
|
|
* This function fills the missing link and does all housekeeping needed in
|
|
|
|
* inode (->i_blocks, etc.). In case of success we end up with the full
|
|
|
|
* chain to new block and return 0.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static int ext4_splice_branch(handle_t *handle, struct inode *inode,
|
2009-06-14 23:45:34 +02:00
|
|
|
ext4_lblk_t block, Indirect *where, int num,
|
|
|
|
int blks)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
int err = 0;
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_fsblk_t current_block;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If we're splicing into a [td]indirect block (as opposed to the
|
|
|
|
* inode) then we need to get write access to the [td]indirect block
|
|
|
|
* before the splice.
|
|
|
|
*/
|
|
|
|
if (where->bh) {
|
|
|
|
BUFFER_TRACE(where->bh, "get_write_access");
|
2006-10-11 10:20:53 +02:00
|
|
|
err = ext4_journal_get_write_access(handle, where->bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (err)
|
|
|
|
goto err_out;
|
|
|
|
}
|
|
|
|
/* That's it */
|
|
|
|
|
|
|
|
*where->p = where->key;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update the host buffer_head or inode to point to more just allocated
|
|
|
|
* direct blocks blocks
|
|
|
|
*/
|
|
|
|
if (num == 0 && blks > 1) {
|
|
|
|
current_block = le32_to_cpu(where->key) + 1;
|
|
|
|
for (i = 1; i < blks; i++)
|
2008-09-09 04:25:24 +02:00
|
|
|
*(where->p + i) = cpu_to_le32(current_block++);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* We are done with atomic stuff, now do the rest of housekeeping */
|
|
|
|
/* had we spliced it onto indirect block? */
|
|
|
|
if (where->bh) {
|
|
|
|
/*
|
|
|
|
* If we spliced it onto an indirect block, we haven't
|
|
|
|
* altered the inode. Note however that if it is being spliced
|
|
|
|
* onto an indirect block at the very end of the file (the
|
|
|
|
* file is growing) then we *will* alter the inode to reflect
|
|
|
|
* the new i_size. But that is not done here - it is done in
|
2006-10-11 10:20:53 +02:00
|
|
|
* generic_commit_write->__mark_inode_dirty->ext4_dirty_inode.
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
|
|
|
jbd_debug(5, "splicing indirect only\n");
|
2009-01-07 06:06:22 +01:00
|
|
|
BUFFER_TRACE(where->bh, "call ext4_handle_dirty_metadata");
|
|
|
|
err = ext4_handle_dirty_metadata(handle, inode, where->bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (err)
|
|
|
|
goto err_out;
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* OK, we spliced it into the inode itself on a direct block.
|
|
|
|
*/
|
2009-06-15 09:41:23 +02:00
|
|
|
ext4_mark_inode_dirty(handle, inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
jbd_debug(5, "splicing direct\n");
|
|
|
|
}
|
|
|
|
return err;
|
|
|
|
|
|
|
|
err_out:
|
|
|
|
for (i = 1; i <= num; i++) {
|
2010-05-17 13:00:00 +02:00
|
|
|
/*
|
2009-11-23 13:17:05 +01:00
|
|
|
* branch[i].bh is newly allocated, so there is no
|
|
|
|
* need to revoke the block, which is why we don't
|
|
|
|
* need to set EXT4_FREE_BLOCKS_METADATA.
|
2009-11-23 03:00:13 +01:00
|
|
|
*/
|
2009-11-23 13:17:05 +01:00
|
|
|
ext4_free_blocks(handle, inode, where[i].bh, 0, 1,
|
|
|
|
EXT4_FREE_BLOCKS_FORGET);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2009-11-23 13:17:05 +01:00
|
|
|
ext4_free_blocks(handle, inode, 0, le32_to_cpu(where[num].key),
|
|
|
|
blks, 0);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2010-05-17 01:00:00 +02:00
|
|
|
* The ext4_ind_map_blocks() function handles non-extents inodes
|
2009-05-14 06:54:29 +02:00
|
|
|
* (i.e., using the traditional indirect/double-indirect i_blocks
|
2010-05-17 01:00:00 +02:00
|
|
|
* scheme) for ext4_map_blocks().
|
2009-05-14 06:54:29 +02:00
|
|
|
*
|
2006-10-11 10:20:50 +02:00
|
|
|
* Allocation strategy is simple: if we have to allocate something, we will
|
|
|
|
* have to go the whole way to leaf. So let's do it before attaching anything
|
|
|
|
* to tree, set linkage between the newborn blocks, write them if sync is
|
|
|
|
* required, recheck the path, free and repeat if check fails, otherwise
|
|
|
|
* set the last missing link (that will protect us from any truncate-generated
|
|
|
|
* removals - all blocks on the path are immune now) and possibly force the
|
|
|
|
* write on the parent block.
|
|
|
|
* That has a nice additional property: no special recovery from the failed
|
|
|
|
* allocations is needed - we simply release blocks and do not touch anything
|
|
|
|
* reachable from inode.
|
|
|
|
*
|
|
|
|
* `handle' can be NULL if create == 0.
|
|
|
|
*
|
|
|
|
* return > 0, # of blocks mapped or allocated.
|
|
|
|
* return = 0, if plain lookup failed.
|
|
|
|
* return < 0, error case.
|
2008-01-29 05:58:27 +01:00
|
|
|
*
|
2009-05-14 06:54:29 +02:00
|
|
|
* The ext4_ind_get_blocks() function should be called with
|
|
|
|
* down_write(&EXT4_I(inode)->i_data_sem) if allocating filesystem
|
|
|
|
* blocks (i.e., flags has EXT4_GET_BLOCKS_CREATE set) or
|
|
|
|
* down_read(&EXT4_I(inode)->i_data_sem) if not allocating file system
|
|
|
|
* blocks.
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
2010-05-17 01:00:00 +02:00
|
|
|
static int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
|
|
|
|
struct ext4_map_blocks *map,
|
2009-06-14 23:45:34 +02:00
|
|
|
int flags)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
int err = -EIO;
|
2008-01-29 05:58:27 +01:00
|
|
|
ext4_lblk_t offsets[4];
|
2006-10-11 10:20:50 +02:00
|
|
|
Indirect chain[4];
|
|
|
|
Indirect *partial;
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_fsblk_t goal;
|
2006-10-11 10:20:50 +02:00
|
|
|
int indirect_blks;
|
|
|
|
int blocks_to_boundary = 0;
|
|
|
|
int depth;
|
|
|
|
int count = 0;
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_fsblk_t first_block = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2010-05-17 04:00:00 +02:00
|
|
|
J_ASSERT(!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)));
|
2009-05-14 06:58:52 +02:00
|
|
|
J_ASSERT(handle != NULL || (flags & EXT4_GET_BLOCKS_CREATE) == 0);
|
2010-05-17 01:00:00 +02:00
|
|
|
depth = ext4_block_to_path(inode, map->m_lblk, offsets,
|
2009-06-14 23:45:34 +02:00
|
|
|
&blocks_to_boundary);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
if (depth == 0)
|
|
|
|
goto out;
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
partial = ext4_get_branch(inode, depth, offsets, chain, &err);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/* Simplest case - block found, no allocation needed */
|
|
|
|
if (!partial) {
|
|
|
|
first_block = le32_to_cpu(chain[depth - 1].key);
|
|
|
|
count++;
|
|
|
|
/*map more blocks*/
|
2010-05-17 01:00:00 +02:00
|
|
|
while (count < map->m_len && count <= blocks_to_boundary) {
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_fsblk_t blk;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
blk = le32_to_cpu(*(chain[depth-1].p + count));
|
|
|
|
|
|
|
|
if (blk == first_block + count)
|
|
|
|
count++;
|
|
|
|
else
|
|
|
|
break;
|
|
|
|
}
|
2008-01-29 05:58:27 +01:00
|
|
|
goto got_it;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Next simple case - plain lookup or failed read of indirect block */
|
2009-05-14 06:58:52 +02:00
|
|
|
if ((flags & EXT4_GET_BLOCKS_CREATE) == 0 || err == -EIO)
|
2006-10-11 10:20:50 +02:00
|
|
|
goto cleanup;
|
|
|
|
|
|
|
|
/*
|
2008-10-10 15:40:52 +02:00
|
|
|
* Okay, we need to do block allocation.
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
2010-05-17 01:00:00 +02:00
|
|
|
goal = ext4_find_goal(inode, map->m_lblk, partial);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/* the number of blocks need to allocate for [d,t]indirect blocks */
|
|
|
|
indirect_blks = (chain + depth) - partial - 1;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Next look up the indirect map to count the totoal number of
|
|
|
|
* direct blocks to allocate for this branch.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
count = ext4_blks_to_allocate(partial, indirect_blks,
|
2010-05-17 01:00:00 +02:00
|
|
|
map->m_len, blocks_to_boundary);
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* Block out ext4_truncate while we alter the tree
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
2010-05-17 01:00:00 +02:00
|
|
|
err = ext4_alloc_branch(handle, inode, map->m_lblk, indirect_blks,
|
2009-06-14 23:45:34 +02:00
|
|
|
&count, goal,
|
|
|
|
offsets + (partial - chain), partial);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* The ext4_splice_branch call will free and forget any buffers
|
2006-10-11 10:20:50 +02:00
|
|
|
* on the new chain if there is a failure, but that risks using
|
|
|
|
* up transaction credits, especially for bitmaps where the
|
|
|
|
* credits cannot be returned. Can we handle this somehow? We
|
|
|
|
* may need to return -EAGAIN upwards in the worst case. --sct
|
|
|
|
*/
|
|
|
|
if (!err)
|
2010-05-17 01:00:00 +02:00
|
|
|
err = ext4_splice_branch(handle, inode, map->m_lblk,
|
2009-06-14 23:45:34 +02:00
|
|
|
partial, indirect_blks, count);
|
2009-11-23 13:24:48 +01:00
|
|
|
if (err)
|
2006-10-11 10:20:50 +02:00
|
|
|
goto cleanup;
|
|
|
|
|
2010-05-17 01:00:00 +02:00
|
|
|
map->m_flags |= EXT4_MAP_NEW;
|
2009-12-09 05:51:10 +01:00
|
|
|
|
|
|
|
ext4_update_inode_fsync_trans(handle, inode, 1);
|
2006-10-11 10:20:50 +02:00
|
|
|
got_it:
|
2010-05-17 01:00:00 +02:00
|
|
|
map->m_flags |= EXT4_MAP_MAPPED;
|
|
|
|
map->m_pblk = le32_to_cpu(chain[depth-1].key);
|
|
|
|
map->m_len = count;
|
2006-10-11 10:20:50 +02:00
|
|
|
if (count > blocks_to_boundary)
|
2010-05-17 01:00:00 +02:00
|
|
|
map->m_flags |= EXT4_MAP_BOUNDARY;
|
2006-10-11 10:20:50 +02:00
|
|
|
err = count;
|
|
|
|
/* Clean up and exit */
|
|
|
|
partial = chain + depth - 1; /* the whole chain */
|
|
|
|
cleanup:
|
|
|
|
while (partial > chain) {
|
|
|
|
BUFFER_TRACE(partial->bh, "call brelse");
|
|
|
|
brelse(partial->bh);
|
|
|
|
partial--;
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2009-12-14 13:21:14 +01:00
|
|
|
#ifdef CONFIG_QUOTA
|
|
|
|
qsize_t *ext4_get_reserved_space(struct inode *inode)
|
2009-01-22 18:13:05 +01:00
|
|
|
{
|
2009-12-14 13:21:14 +01:00
|
|
|
return &EXT4_I(inode)->i_reserved_quota;
|
2009-01-22 18:13:05 +01:00
|
|
|
}
|
2009-12-14 13:21:14 +01:00
|
|
|
#endif
|
2010-01-01 08:41:30 +01:00
|
|
|
|
2008-07-17 22:12:08 +02:00
|
|
|
/*
|
|
|
|
* Calculate the number of metadata blocks need to reserve
|
2010-01-01 08:41:30 +01:00
|
|
|
* to allocate a new block at @lblocks for non extent file based file
|
2008-07-17 22:12:08 +02:00
|
|
|
*/
|
2010-01-01 08:41:30 +01:00
|
|
|
static int ext4_indirect_calc_metadata_amount(struct inode *inode,
|
|
|
|
sector_t lblock)
|
2008-07-17 22:12:08 +02:00
|
|
|
{
|
2010-01-01 08:41:30 +01:00
|
|
|
struct ext4_inode_info *ei = EXT4_I(inode);
|
2010-03-14 23:17:54 +01:00
|
|
|
sector_t dind_mask = ~((sector_t)EXT4_ADDR_PER_BLOCK(inode->i_sb) - 1);
|
2010-01-01 08:41:30 +01:00
|
|
|
int blk_bits;
|
2008-07-17 22:12:08 +02:00
|
|
|
|
2010-01-01 08:41:30 +01:00
|
|
|
if (lblock < EXT4_NDIR_BLOCKS)
|
|
|
|
return 0;
|
2008-07-17 22:12:08 +02:00
|
|
|
|
2010-01-01 08:41:30 +01:00
|
|
|
lblock -= EXT4_NDIR_BLOCKS;
|
2008-07-17 22:12:08 +02:00
|
|
|
|
2010-01-01 08:41:30 +01:00
|
|
|
if (ei->i_da_metadata_calc_len &&
|
|
|
|
(lblock & dind_mask) == ei->i_da_metadata_calc_last_lblock) {
|
|
|
|
ei->i_da_metadata_calc_len++;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
ei->i_da_metadata_calc_last_lblock = lblock & dind_mask;
|
|
|
|
ei->i_da_metadata_calc_len = 1;
|
2010-03-14 23:17:54 +01:00
|
|
|
blk_bits = order_base_2(lblock);
|
2010-01-01 08:41:30 +01:00
|
|
|
return (blk_bits / EXT4_ADDR_PER_BLOCK_BITS(inode->i_sb)) + 1;
|
2008-07-17 22:12:08 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Calculate the number of metadata blocks need to reserve
|
2010-01-01 08:41:30 +01:00
|
|
|
* to allocate a block located at @lblock
|
2008-07-17 22:12:08 +02:00
|
|
|
*/
|
2010-01-01 08:41:30 +01:00
|
|
|
static int ext4_calc_metadata_amount(struct inode *inode, sector_t lblock)
|
2008-07-17 22:12:08 +02:00
|
|
|
{
|
2010-05-17 04:00:00 +02:00
|
|
|
if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
|
2010-01-01 08:41:30 +01:00
|
|
|
return ext4_ext_calc_metadata_amount(inode, lblock);
|
2008-07-17 22:12:08 +02:00
|
|
|
|
2010-01-01 08:41:30 +01:00
|
|
|
return ext4_indirect_calc_metadata_amount(inode, lblock);
|
2008-07-17 22:12:08 +02:00
|
|
|
}
|
|
|
|
|
2009-12-30 20:20:45 +01:00
|
|
|
/*
|
|
|
|
* Called with i_data_sem down, which is important since we can call
|
|
|
|
* ext4_discard_preallocations() from here.
|
|
|
|
*/
|
2010-01-25 10:00:31 +01:00
|
|
|
void ext4_da_update_reserve_space(struct inode *inode,
|
|
|
|
int used, int quota_claim)
|
2008-07-17 22:12:08 +02:00
|
|
|
{
|
|
|
|
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
|
2009-12-30 20:20:45 +01:00
|
|
|
struct ext4_inode_info *ei = EXT4_I(inode);
|
|
|
|
|
|
|
|
spin_lock(&ei->i_block_reservation_lock);
|
2010-01-01 07:00:21 +01:00
|
|
|
trace_ext4_da_update_reserve_space(inode, used);
|
2009-12-30 20:20:45 +01:00
|
|
|
if (unlikely(used > ei->i_reserved_data_blocks)) {
|
|
|
|
ext4_msg(inode->i_sb, KERN_NOTICE, "%s: ino %lu, used %d "
|
|
|
|
"with only %d reserved data blocks\n",
|
|
|
|
__func__, inode->i_ino, used,
|
|
|
|
ei->i_reserved_data_blocks);
|
|
|
|
WARN_ON(1);
|
|
|
|
used = ei->i_reserved_data_blocks;
|
|
|
|
}
|
2008-07-17 22:12:08 +02:00
|
|
|
|
2009-12-30 20:20:45 +01:00
|
|
|
/* Update per-inode reservations */
|
|
|
|
ei->i_reserved_data_blocks -= used;
|
|
|
|
ei->i_reserved_meta_blocks -= ei->i_allocated_meta_blocks;
|
2010-05-16 17:00:00 +02:00
|
|
|
percpu_counter_sub(&sbi->s_dirtyblocks_counter,
|
|
|
|
used + ei->i_allocated_meta_blocks);
|
2009-12-30 20:20:45 +01:00
|
|
|
ei->i_allocated_meta_blocks = 0;
|
2008-10-10 15:39:00 +02:00
|
|
|
|
2009-12-30 20:20:45 +01:00
|
|
|
if (ei->i_reserved_data_blocks == 0) {
|
|
|
|
/*
|
|
|
|
* We can release all of the reserved metadata blocks
|
|
|
|
* only when we have written all of the delayed
|
|
|
|
* allocation blocks.
|
|
|
|
*/
|
2010-05-16 17:00:00 +02:00
|
|
|
percpu_counter_sub(&sbi->s_dirtyblocks_counter,
|
|
|
|
ei->i_reserved_meta_blocks);
|
2010-01-01 08:36:15 +01:00
|
|
|
ei->i_reserved_meta_blocks = 0;
|
2010-01-01 08:41:30 +01:00
|
|
|
ei->i_da_metadata_calc_len = 0;
|
2008-10-10 15:39:00 +02:00
|
|
|
}
|
2008-07-17 22:12:08 +02:00
|
|
|
spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
|
2009-01-22 18:13:05 +01:00
|
|
|
|
2010-05-16 17:00:00 +02:00
|
|
|
/* Update quota subsystem for data blocks */
|
|
|
|
if (quota_claim)
|
2010-03-03 15:05:00 +01:00
|
|
|
dquot_claim_block(inode, used);
|
2010-05-16 17:00:00 +02:00
|
|
|
else {
|
2010-01-25 10:00:31 +01:00
|
|
|
/*
|
|
|
|
* We did fallocate with an offset that is already delayed
|
|
|
|
* allocated. So on delayed allocated writeback we should
|
2010-05-16 17:00:00 +02:00
|
|
|
* not re-claim the quota for fallocated blocks.
|
2010-01-25 10:00:31 +01:00
|
|
|
*/
|
2010-05-16 17:00:00 +02:00
|
|
|
dquot_release_reservation_block(inode, used);
|
2010-01-25 10:00:31 +01:00
|
|
|
}
|
2009-03-28 03:36:43 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If we have done all the pending block allocations and if
|
|
|
|
* there aren't any writers on the inode, we can discard the
|
|
|
|
* inode's preallocations.
|
|
|
|
*/
|
2009-12-30 20:20:45 +01:00
|
|
|
if ((ei->i_reserved_data_blocks == 0) &&
|
|
|
|
(atomic_read(&inode->i_writecount) == 0))
|
2009-03-28 03:36:43 +01:00
|
|
|
ext4_discard_preallocations(inode);
|
2008-07-17 22:12:08 +02:00
|
|
|
}
|
|
|
|
|
2010-05-17 03:00:00 +02:00
|
|
|
static int check_block_validity(struct inode *inode, const char *func,
|
|
|
|
struct ext4_map_blocks *map)
|
2009-05-17 21:38:01 +02:00
|
|
|
{
|
2010-05-17 03:00:00 +02:00
|
|
|
if (!ext4_data_block_valid(EXT4_SB(inode->i_sb), map->m_pblk,
|
|
|
|
map->m_len)) {
|
|
|
|
ext4_error_inode(func, inode,
|
|
|
|
"lblock %lu mapped to illegal pblock %llu "
|
|
|
|
"(length %d)", (unsigned long) map->m_lblk,
|
|
|
|
map->m_pblk, map->m_len);
|
2009-05-17 21:38:01 +02:00
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-09-29 19:31:31 +02:00
|
|
|
/*
|
2009-10-01 04:57:41 +02:00
|
|
|
* Return the number of contiguous dirty pages in a given inode
|
|
|
|
* starting at page frame idx.
|
2009-09-29 19:31:31 +02:00
|
|
|
*/
|
|
|
|
static pgoff_t ext4_num_dirty_pages(struct inode *inode, pgoff_t idx,
|
|
|
|
unsigned int max_pages)
|
|
|
|
{
|
|
|
|
struct address_space *mapping = inode->i_mapping;
|
|
|
|
pgoff_t index;
|
|
|
|
struct pagevec pvec;
|
|
|
|
pgoff_t num = 0;
|
|
|
|
int i, nr_pages, done = 0;
|
|
|
|
|
|
|
|
if (max_pages == 0)
|
|
|
|
return 0;
|
|
|
|
pagevec_init(&pvec, 0);
|
|
|
|
while (!done) {
|
|
|
|
index = idx;
|
|
|
|
nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
|
|
|
|
PAGECACHE_TAG_DIRTY,
|
|
|
|
(pgoff_t)PAGEVEC_SIZE);
|
|
|
|
if (nr_pages == 0)
|
|
|
|
break;
|
|
|
|
for (i = 0; i < nr_pages; i++) {
|
|
|
|
struct page *page = pvec.pages[i];
|
|
|
|
struct buffer_head *bh, *head;
|
|
|
|
|
|
|
|
lock_page(page);
|
|
|
|
if (unlikely(page->mapping != mapping) ||
|
|
|
|
!PageDirty(page) ||
|
|
|
|
PageWriteback(page) ||
|
|
|
|
page->index != idx) {
|
|
|
|
done = 1;
|
|
|
|
unlock_page(page);
|
|
|
|
break;
|
|
|
|
}
|
2009-10-01 04:57:41 +02:00
|
|
|
if (page_has_buffers(page)) {
|
|
|
|
bh = head = page_buffers(page);
|
|
|
|
do {
|
|
|
|
if (!buffer_delay(bh) &&
|
|
|
|
!buffer_unwritten(bh))
|
|
|
|
done = 1;
|
|
|
|
bh = bh->b_this_page;
|
|
|
|
} while (!done && (bh != head));
|
|
|
|
}
|
2009-09-29 19:31:31 +02:00
|
|
|
unlock_page(page);
|
|
|
|
if (done)
|
|
|
|
break;
|
|
|
|
idx++;
|
|
|
|
num++;
|
|
|
|
if (num >= max_pages)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
pagevec_release(&pvec);
|
|
|
|
}
|
|
|
|
return num;
|
|
|
|
}
|
|
|
|
|
2008-02-25 21:29:55 +01:00
|
|
|
/*
|
2010-05-17 01:00:00 +02:00
|
|
|
* The ext4_map_blocks() function tries to look up the requested blocks,
|
2008-07-26 22:15:44 +02:00
|
|
|
* and returns if the blocks are already mapped.
|
2008-02-25 21:29:55 +01:00
|
|
|
*
|
|
|
|
* Otherwise it takes the write lock of the i_data_sem and allocate blocks
|
|
|
|
* and store the allocated blocks in the result buffer head and mark it
|
|
|
|
* mapped.
|
|
|
|
*
|
2010-05-17 01:00:00 +02:00
|
|
|
* If file type is extents based, it will call ext4_ext_map_blocks(),
|
|
|
|
* Otherwise, call with ext4_ind_map_blocks() to handle indirect mapping
|
2008-02-25 21:29:55 +01:00
|
|
|
* based files
|
|
|
|
*
|
|
|
|
* On success, it returns the number of blocks being mapped or allocate.
|
|
|
|
* if create==0 and the blocks are pre-allocated and uninitialized block,
|
|
|
|
* the result buffer head is unmapped. If the create ==1, it will make sure
|
|
|
|
* the buffer head is mapped.
|
|
|
|
*
|
|
|
|
* It returns 0 if plain look up failed (blocks have not been allocated), in
|
|
|
|
* that casem, buffer head is unmapped
|
|
|
|
*
|
|
|
|
* It returns the error in case of allocation failure.
|
|
|
|
*/
|
2010-05-17 01:00:00 +02:00
|
|
|
int ext4_map_blocks(handle_t *handle, struct inode *inode,
|
|
|
|
struct ext4_map_blocks *map, int flags)
|
2008-01-29 05:58:26 +01:00
|
|
|
{
|
|
|
|
int retval;
|
2008-02-25 21:29:55 +01:00
|
|
|
|
2010-05-17 01:00:00 +02:00
|
|
|
map->m_flags = 0;
|
|
|
|
ext_debug("ext4_map_blocks(): inode %lu, flag %d, max_blocks %u,"
|
|
|
|
"logical block %lu\n", inode->i_ino, flags, map->m_len,
|
|
|
|
(unsigned long) map->m_lblk);
|
2008-01-29 05:58:29 +01:00
|
|
|
/*
|
2009-05-14 06:54:29 +02:00
|
|
|
* Try to see if we can get the block without requesting a new
|
|
|
|
* file system block.
|
2008-01-29 05:58:29 +01:00
|
|
|
*/
|
|
|
|
down_read((&EXT4_I(inode)->i_data_sem));
|
2010-05-17 04:00:00 +02:00
|
|
|
if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
|
2010-05-17 01:00:00 +02:00
|
|
|
retval = ext4_ext_map_blocks(handle, inode, map, 0);
|
2008-01-29 05:58:26 +01:00
|
|
|
} else {
|
2010-05-17 01:00:00 +02:00
|
|
|
retval = ext4_ind_map_blocks(handle, inode, map, 0);
|
2008-01-29 05:58:26 +01:00
|
|
|
}
|
2008-01-29 05:58:29 +01:00
|
|
|
up_read((&EXT4_I(inode)->i_data_sem));
|
2008-02-25 21:29:55 +01:00
|
|
|
|
2010-05-17 01:00:00 +02:00
|
|
|
if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
|
2010-05-17 03:00:00 +02:00
|
|
|
int ret = check_block_validity(inode, __func__, map);
|
2009-05-17 21:38:01 +02:00
|
|
|
if (ret != 0)
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-02-25 21:29:55 +01:00
|
|
|
/* If it is only a block(s) look up */
|
2009-05-14 06:58:52 +02:00
|
|
|
if ((flags & EXT4_GET_BLOCKS_CREATE) == 0)
|
2008-02-25 21:29:55 +01:00
|
|
|
return retval;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Returns if the blocks have already allocated
|
|
|
|
*
|
|
|
|
* Note that if blocks have been preallocated
|
|
|
|
* ext4_ext_get_block() returns th create = 0
|
|
|
|
* with buffer head unmapped.
|
|
|
|
*/
|
2010-05-17 01:00:00 +02:00
|
|
|
if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED)
|
2008-01-29 05:58:29 +01:00
|
|
|
return retval;
|
|
|
|
|
2009-05-14 23:05:39 +02:00
|
|
|
/*
|
|
|
|
* When we call get_blocks without the create flag, the
|
|
|
|
* BH_Unwritten flag could have gotten set if the blocks
|
|
|
|
* requested were part of a uninitialized extent. We need to
|
|
|
|
* clear this flag now that we are committed to convert all or
|
|
|
|
* part of the uninitialized extent to be an initialized
|
|
|
|
* extent. This is because we need to avoid the combination
|
|
|
|
* of BH_Unwritten and BH_Mapped flags being simultaneously
|
|
|
|
* set on the buffer_head.
|
|
|
|
*/
|
2010-05-17 01:00:00 +02:00
|
|
|
map->m_flags &= ~EXT4_MAP_UNWRITTEN;
|
2009-05-14 23:05:39 +02:00
|
|
|
|
2008-01-29 05:58:29 +01:00
|
|
|
/*
|
2008-02-25 21:29:55 +01:00
|
|
|
* New blocks allocate and/or writing to uninitialized extent
|
|
|
|
* will possibly result in updating i_data, so we take
|
|
|
|
* the write lock of i_data_sem, and call get_blocks()
|
|
|
|
* with create == 1 flag.
|
2008-01-29 05:58:29 +01:00
|
|
|
*/
|
|
|
|
down_write((&EXT4_I(inode)->i_data_sem));
|
2008-07-14 23:52:37 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* if the caller is from delayed allocation writeout path
|
|
|
|
* we have already reserved fs blocks for allocation
|
|
|
|
* let the underlying get_block() function know to
|
|
|
|
* avoid double accounting
|
|
|
|
*/
|
2009-05-14 06:58:52 +02:00
|
|
|
if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
|
2008-07-14 23:52:37 +02:00
|
|
|
EXT4_I(inode)->i_delalloc_reserved_flag = 1;
|
2008-01-29 05:58:29 +01:00
|
|
|
/*
|
|
|
|
* We need to check for EXT4 here because migrate
|
|
|
|
* could have changed the inode type in between
|
|
|
|
*/
|
2010-05-17 04:00:00 +02:00
|
|
|
if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
|
2010-05-17 01:00:00 +02:00
|
|
|
retval = ext4_ext_map_blocks(handle, inode, map, flags);
|
2008-01-29 05:58:26 +01:00
|
|
|
} else {
|
2010-05-17 01:00:00 +02:00
|
|
|
retval = ext4_ind_map_blocks(handle, inode, map, flags);
|
2008-04-29 14:11:12 +02:00
|
|
|
|
2010-05-17 01:00:00 +02:00
|
|
|
if (retval > 0 && map->m_flags & EXT4_MAP_NEW) {
|
2008-04-29 14:11:12 +02:00
|
|
|
/*
|
|
|
|
* We allocated new blocks which will result in
|
|
|
|
* i_data's format changing. Force the migrate
|
|
|
|
* to fail by clearing migrate flags
|
|
|
|
*/
|
2010-01-24 20:34:07 +01:00
|
|
|
ext4_clear_inode_state(inode, EXT4_STATE_EXT_MIGRATE);
|
2008-04-29 14:11:12 +02:00
|
|
|
}
|
2008-07-14 23:52:37 +02:00
|
|
|
|
2010-01-25 10:00:31 +01:00
|
|
|
/*
|
|
|
|
* Update reserved blocks/metadata blocks after successful
|
|
|
|
* block allocation which had been deferred till now. We don't
|
|
|
|
* support fallocate for non extent files. So we can update
|
|
|
|
* reserve space here.
|
|
|
|
*/
|
|
|
|
if ((retval > 0) &&
|
2010-01-15 07:27:59 +01:00
|
|
|
(flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE))
|
2010-01-25 10:00:31 +01:00
|
|
|
ext4_da_update_reserve_space(inode, retval, 1);
|
|
|
|
}
|
2009-05-14 19:57:08 +02:00
|
|
|
if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
|
2008-07-14 23:52:37 +02:00
|
|
|
EXT4_I(inode)->i_delalloc_reserved_flag = 0;
|
2009-05-14 19:57:08 +02:00
|
|
|
|
2008-01-29 05:58:29 +01:00
|
|
|
up_write((&EXT4_I(inode)->i_data_sem));
|
2010-05-17 01:00:00 +02:00
|
|
|
if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
|
2010-05-17 03:00:00 +02:00
|
|
|
int ret = check_block_validity(inode,
|
|
|
|
"ext4_map_blocks_after_alloc",
|
|
|
|
map);
|
2009-05-17 21:38:01 +02:00
|
|
|
if (ret != 0)
|
|
|
|
return ret;
|
|
|
|
}
|
2008-01-29 05:58:26 +01:00
|
|
|
return retval;
|
|
|
|
}
|
|
|
|
|
2008-08-20 04:16:03 +02:00
|
|
|
/* Maximum number of blocks we map for direct IO at once. */
|
|
|
|
#define DIO_MAX_BLOCKS 4096
|
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
static int _ext4_get_block(struct inode *inode, sector_t iblock,
|
|
|
|
struct buffer_head *bh, int flags)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2007-02-10 10:46:35 +01:00
|
|
|
handle_t *handle = ext4_journal_current_handle();
|
2010-05-17 02:00:00 +02:00
|
|
|
struct ext4_map_blocks map;
|
2008-02-10 07:08:38 +01:00
|
|
|
int ret = 0, started = 0;
|
2008-08-20 04:16:03 +02:00
|
|
|
int dio_credits;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
map.m_lblk = iblock;
|
|
|
|
map.m_len = bh->b_size >> inode->i_blkbits;
|
|
|
|
|
|
|
|
if (flags && !handle) {
|
2008-02-10 07:08:38 +01:00
|
|
|
/* Direct IO write... */
|
2010-05-17 02:00:00 +02:00
|
|
|
if (map.m_len > DIO_MAX_BLOCKS)
|
|
|
|
map.m_len = DIO_MAX_BLOCKS;
|
|
|
|
dio_credits = ext4_chunk_trans_blocks(inode, map.m_len);
|
2008-08-20 04:16:03 +02:00
|
|
|
handle = ext4_journal_start(inode, dio_credits);
|
2008-02-10 07:08:38 +01:00
|
|
|
if (IS_ERR(handle)) {
|
2006-10-11 10:20:50 +02:00
|
|
|
ret = PTR_ERR(handle);
|
2010-05-17 02:00:00 +02:00
|
|
|
return ret;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2008-02-10 07:08:38 +01:00
|
|
|
started = 1;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
ret = ext4_map_blocks(handle, inode, &map, flags);
|
2008-02-10 07:08:38 +01:00
|
|
|
if (ret > 0) {
|
2010-05-17 02:00:00 +02:00
|
|
|
map_bh(bh, inode->i_sb, map.m_pblk);
|
|
|
|
bh->b_state = (bh->b_state & ~EXT4_MAP_FLAGS) | map.m_flags;
|
|
|
|
bh->b_size = inode->i_sb->s_blocksize * map.m_len;
|
2008-02-10 07:08:38 +01:00
|
|
|
ret = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2008-02-10 07:08:38 +01:00
|
|
|
if (started)
|
|
|
|
ext4_journal_stop(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
int ext4_get_block(struct inode *inode, sector_t iblock,
|
|
|
|
struct buffer_head *bh, int create)
|
|
|
|
{
|
|
|
|
return _ext4_get_block(inode, iblock, bh,
|
|
|
|
create ? EXT4_GET_BLOCKS_CREATE : 0);
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* `handle' can be NULL if create is zero
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
struct buffer_head *ext4_getblk(handle_t *handle, struct inode *inode,
|
2008-01-29 05:58:27 +01:00
|
|
|
ext4_lblk_t block, int create, int *errp)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2010-05-17 02:00:00 +02:00
|
|
|
struct ext4_map_blocks map;
|
|
|
|
struct buffer_head *bh;
|
2006-10-11 10:20:50 +02:00
|
|
|
int fatal = 0, err;
|
|
|
|
|
|
|
|
J_ASSERT(handle != NULL || create == 0);
|
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
map.m_lblk = block;
|
|
|
|
map.m_len = 1;
|
|
|
|
err = ext4_map_blocks(handle, inode, &map,
|
|
|
|
create ? EXT4_GET_BLOCKS_CREATE : 0);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
if (err < 0)
|
|
|
|
*errp = err;
|
|
|
|
if (err <= 0)
|
|
|
|
return NULL;
|
|
|
|
*errp = 0;
|
|
|
|
|
|
|
|
bh = sb_getblk(inode->i_sb, map.m_pblk);
|
|
|
|
if (!bh) {
|
|
|
|
*errp = -EIO;
|
|
|
|
return NULL;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2010-05-17 02:00:00 +02:00
|
|
|
if (map.m_flags & EXT4_MAP_NEW) {
|
|
|
|
J_ASSERT(create != 0);
|
|
|
|
J_ASSERT(handle != NULL);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
/*
|
|
|
|
* Now that we do not always journal data, we should
|
|
|
|
* keep in mind whether this should always journal the
|
|
|
|
* new buffer as metadata. For now, regular file
|
|
|
|
* writes use ext4_get_block instead, so it's not a
|
|
|
|
* problem.
|
|
|
|
*/
|
|
|
|
lock_buffer(bh);
|
|
|
|
BUFFER_TRACE(bh, "call get_create_access");
|
|
|
|
fatal = ext4_journal_get_create_access(handle, bh);
|
|
|
|
if (!fatal && !buffer_uptodate(bh)) {
|
|
|
|
memset(bh->b_data, 0, inode->i_sb->s_blocksize);
|
|
|
|
set_buffer_uptodate(bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2010-05-17 02:00:00 +02:00
|
|
|
unlock_buffer(bh);
|
|
|
|
BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
|
|
|
|
err = ext4_handle_dirty_metadata(handle, inode, bh);
|
|
|
|
if (!fatal)
|
|
|
|
fatal = err;
|
|
|
|
} else {
|
|
|
|
BUFFER_TRACE(bh, "not a new buffer");
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2010-05-17 02:00:00 +02:00
|
|
|
if (fatal) {
|
|
|
|
*errp = fatal;
|
|
|
|
brelse(bh);
|
|
|
|
bh = NULL;
|
|
|
|
}
|
|
|
|
return bh;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
struct buffer_head *ext4_bread(handle_t *handle, struct inode *inode,
|
2008-01-29 05:58:27 +01:00
|
|
|
ext4_lblk_t block, int create, int *err)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2008-09-09 04:25:24 +02:00
|
|
|
struct buffer_head *bh;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
bh = ext4_getblk(handle, inode, block, create, err);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!bh)
|
|
|
|
return bh;
|
|
|
|
if (buffer_uptodate(bh))
|
|
|
|
return bh;
|
|
|
|
ll_rw_block(READ_META, 1, &bh);
|
|
|
|
wait_on_buffer(bh);
|
|
|
|
if (buffer_uptodate(bh))
|
|
|
|
return bh;
|
|
|
|
put_bh(bh);
|
|
|
|
*err = -EIO;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2008-09-09 04:25:24 +02:00
|
|
|
static int walk_page_buffers(handle_t *handle,
|
|
|
|
struct buffer_head *head,
|
|
|
|
unsigned from,
|
|
|
|
unsigned to,
|
|
|
|
int *partial,
|
|
|
|
int (*fn)(handle_t *handle,
|
|
|
|
struct buffer_head *bh))
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
struct buffer_head *bh;
|
|
|
|
unsigned block_start, block_end;
|
|
|
|
unsigned blocksize = head->b_size;
|
|
|
|
int err, ret = 0;
|
|
|
|
struct buffer_head *next;
|
|
|
|
|
2008-09-09 04:25:24 +02:00
|
|
|
for (bh = head, block_start = 0;
|
|
|
|
ret == 0 && (bh != head || !block_start);
|
2009-06-14 23:45:34 +02:00
|
|
|
block_start = block_end, bh = next) {
|
2006-10-11 10:20:50 +02:00
|
|
|
next = bh->b_this_page;
|
|
|
|
block_end = block_start + blocksize;
|
|
|
|
if (block_end <= from || block_start >= to) {
|
|
|
|
if (partial && !buffer_uptodate(bh))
|
|
|
|
*partial = 1;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
err = (*fn)(handle, bh);
|
|
|
|
if (!ret)
|
|
|
|
ret = err;
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* To preserve ordering, it is essential that the hole instantiation and
|
|
|
|
* the data write be encapsulated in a single transaction. We cannot
|
2006-10-11 10:20:53 +02:00
|
|
|
* close off a transaction and start a new one between the ext4_get_block()
|
2006-10-11 10:21:01 +02:00
|
|
|
* and the commit_write(). So doing the jbd2_journal_start at the start of
|
2006-10-11 10:20:50 +02:00
|
|
|
* prepare_write() is the right place.
|
|
|
|
*
|
2006-10-11 10:20:53 +02:00
|
|
|
* Also, this function can nest inside ext4_writepage() ->
|
|
|
|
* block_write_full_page(). In that case, we *know* that ext4_writepage()
|
2006-10-11 10:20:50 +02:00
|
|
|
* has generated enough buffer credits to do the whole page. So we won't
|
|
|
|
* block on the journal in that case, which is good, because the caller may
|
|
|
|
* be PF_MEMALLOC.
|
|
|
|
*
|
2006-10-11 10:20:53 +02:00
|
|
|
* By accident, ext4 can be reentered when a transaction is open via
|
2006-10-11 10:20:50 +02:00
|
|
|
* quota file writes. If we were to commit the transaction while thus
|
|
|
|
* reentered, there can be a deadlock - we would be holding a quota
|
|
|
|
* lock, and the commit would never complete if another thread had a
|
|
|
|
* transaction open and was blocking on the quota lock - a ranking
|
|
|
|
* violation.
|
|
|
|
*
|
2006-10-11 10:21:01 +02:00
|
|
|
* So what we do is to rely on the fact that jbd2_journal_stop/journal_start
|
2006-10-11 10:20:50 +02:00
|
|
|
* will _not_ run commit under these circumstances because handle->h_ref
|
|
|
|
* is elevated. We'll still have enough credits for the tiny quotafile
|
|
|
|
* write.
|
|
|
|
*/
|
|
|
|
static int do_journal_get_write_access(handle_t *handle,
|
2009-06-14 23:45:34 +02:00
|
|
|
struct buffer_head *bh)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
if (!buffer_mapped(bh) || buffer_freed(bh))
|
|
|
|
return 0;
|
2006-10-11 10:20:53 +02:00
|
|
|
return ext4_journal_get_write_access(handle, bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2009-12-09 03:24:33 +01:00
|
|
|
/*
|
|
|
|
* Truncate blocks that were not used by write. We have to truncate the
|
|
|
|
* pagecache as well so that corresponding buffers get properly unmapped.
|
|
|
|
*/
|
|
|
|
static void ext4_truncate_failed_write(struct inode *inode)
|
|
|
|
{
|
|
|
|
truncate_inode_pages(inode->i_mapping, inode->i_size);
|
|
|
|
ext4_truncate(inode);
|
|
|
|
}
|
|
|
|
|
2010-03-04 22:14:02 +01:00
|
|
|
static int ext4_get_block_write(struct inode *inode, sector_t iblock,
|
|
|
|
struct buffer_head *bh_result, int create);
|
2007-10-16 10:25:05 +02:00
|
|
|
static int ext4_write_begin(struct file *file, struct address_space *mapping,
|
2009-06-14 23:45:34 +02:00
|
|
|
loff_t pos, unsigned len, unsigned flags,
|
|
|
|
struct page **pagep, void **fsdata)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2008-09-09 04:25:24 +02:00
|
|
|
struct inode *inode = mapping->host;
|
2009-06-05 07:00:26 +02:00
|
|
|
int ret, needed_blocks;
|
2006-10-11 10:20:50 +02:00
|
|
|
handle_t *handle;
|
|
|
|
int retries = 0;
|
2008-09-09 04:25:24 +02:00
|
|
|
struct page *page;
|
2009-06-14 23:45:34 +02:00
|
|
|
pgoff_t index;
|
2008-09-09 04:25:24 +02:00
|
|
|
unsigned from, to;
|
2007-10-16 10:25:05 +02:00
|
|
|
|
2009-06-17 17:48:11 +02:00
|
|
|
trace_ext4_write_begin(inode, pos, len, flags);
|
2009-06-05 07:00:26 +02:00
|
|
|
/*
|
|
|
|
* Reserve one block more for addition to orphan list in case
|
|
|
|
* we allocate blocks but write fails for some reason
|
|
|
|
*/
|
|
|
|
needed_blocks = ext4_writepage_trans_blocks(inode) + 1;
|
2009-06-14 23:45:34 +02:00
|
|
|
index = pos >> PAGE_CACHE_SHIFT;
|
2008-09-09 04:25:24 +02:00
|
|
|
from = pos & (PAGE_CACHE_SIZE - 1);
|
|
|
|
to = from + len;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
retry:
|
2008-09-09 04:25:24 +02:00
|
|
|
handle = ext4_journal_start(inode, needed_blocks);
|
|
|
|
if (IS_ERR(handle)) {
|
|
|
|
ret = PTR_ERR(handle);
|
|
|
|
goto out;
|
2007-04-02 08:49:44 +02:00
|
|
|
}
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2009-02-23 03:09:59 +01:00
|
|
|
/* We cannot recurse into the filesystem as the transaction is already
|
|
|
|
* started */
|
|
|
|
flags |= AOP_FLAG_NOFS;
|
|
|
|
|
fs: symlink write_begin allocation context fix
With the write_begin/write_end aops, page_symlink was broken because it
could no longer pass a GFP_NOFS type mask into the point where the
allocations happened. They are done in write_begin, which would always
assume that the filesystem can be entered from reclaim. This bug could
cause filesystem deadlocks.
The funny thing with having a gfp_t mask there is that it doesn't really
allow the caller to arbitrarily tinker with the context in which it can be
called. It couldn't ever be GFP_ATOMIC, for example, because it needs to
take the page lock. The only thing any callers care about is __GFP_FS
anyway, so turn that into a single flag.
Add a new flag for write_begin, AOP_FLAG_NOFS. Filesystems can now act on
this flag in their write_begin function. Change __grab_cache_page to
accept a nofs argument as well, to honour that flag (while we're there,
change the name to grab_cache_page_write_begin which is more instructive
and does away with random leading underscores).
This is really a more flexible way to go in the end anyway -- if a
filesystem happens to want any extra allocations aside from the pagecache
ones in ints write_begin function, it may now use GFP_KERNEL (rather than
GFP_NOFS) for common case allocations (eg. ocfs2_alloc_write_ctxt, for a
random example).
[kosaki.motohiro@jp.fujitsu.com: fix ubifs]
[kosaki.motohiro@jp.fujitsu.com: fix fuse]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: <stable@kernel.org> [2.6.28.x]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ Cleaned up the calling convention: just pass in the AOP flags
untouched to the grab_cache_page_write_begin() function. That
just simplifies everybody, and may even allow future expansion of the
logic. - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-04 21:00:53 +01:00
|
|
|
page = grab_cache_page_write_begin(mapping, index, flags);
|
2008-07-12 01:27:31 +02:00
|
|
|
if (!page) {
|
|
|
|
ext4_journal_stop(handle);
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
*pagep = page;
|
|
|
|
|
2010-03-04 22:14:02 +01:00
|
|
|
if (ext4_should_dioread_nolock(inode))
|
|
|
|
ret = block_write_begin(file, mapping, pos, len, flags, pagep,
|
|
|
|
fsdata, ext4_get_block_write);
|
|
|
|
else
|
|
|
|
ret = block_write_begin(file, mapping, pos, len, flags, pagep,
|
|
|
|
fsdata, ext4_get_block);
|
2007-10-16 10:25:05 +02:00
|
|
|
|
|
|
|
if (!ret && ext4_should_journal_data(inode)) {
|
2006-10-11 10:20:50 +02:00
|
|
|
ret = walk_page_buffers(handle, page_buffers(page),
|
|
|
|
from, to, NULL, do_journal_get_write_access);
|
|
|
|
}
|
2007-10-16 10:25:05 +02:00
|
|
|
|
|
|
|
if (ret) {
|
2008-09-09 04:25:24 +02:00
|
|
|
unlock_page(page);
|
|
|
|
page_cache_release(page);
|
2008-09-13 19:10:25 +02:00
|
|
|
/*
|
|
|
|
* block_write_begin may have instantiated a few blocks
|
|
|
|
* outside i_size. Trim these off again. Don't need
|
|
|
|
* i_size_read because we hold i_mutex.
|
2009-06-05 07:00:26 +02:00
|
|
|
*
|
|
|
|
* Add inode to orphan list in case we crash before
|
|
|
|
* truncate finishes
|
2008-09-13 19:10:25 +02:00
|
|
|
*/
|
2009-07-13 22:22:22 +02:00
|
|
|
if (pos + len > inode->i_size && ext4_can_truncate(inode))
|
2009-06-05 07:00:26 +02:00
|
|
|
ext4_orphan_add(handle, inode);
|
|
|
|
|
|
|
|
ext4_journal_stop(handle);
|
|
|
|
if (pos + len > inode->i_size) {
|
2009-12-09 03:24:33 +01:00
|
|
|
ext4_truncate_failed_write(inode);
|
2009-06-14 23:45:34 +02:00
|
|
|
/*
|
2009-07-13 22:22:22 +02:00
|
|
|
* If truncate failed early the inode might
|
2009-06-05 07:00:26 +02:00
|
|
|
* still be on the orphan list; we need to
|
|
|
|
* make sure the inode is removed from the
|
|
|
|
* orphan list in that case.
|
|
|
|
*/
|
|
|
|
if (inode->i_nlink)
|
|
|
|
ext4_orphan_del(NULL, inode);
|
|
|
|
}
|
2007-10-16 10:25:05 +02:00
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
|
2006-10-11 10:20:50 +02:00
|
|
|
goto retry;
|
2007-04-02 08:49:44 +02:00
|
|
|
out:
|
2006-10-11 10:20:50 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2007-10-16 10:25:05 +02:00
|
|
|
/* For write_end() in data=journal mode */
|
|
|
|
static int write_end_fn(handle_t *handle, struct buffer_head *bh)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
if (!buffer_mapped(bh) || buffer_freed(bh))
|
|
|
|
return 0;
|
|
|
|
set_buffer_uptodate(bh);
|
2009-01-07 06:06:22 +01:00
|
|
|
return ext4_handle_dirty_metadata(handle, NULL, bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2009-06-05 06:56:49 +02:00
|
|
|
static int ext4_generic_write_end(struct file *file,
|
2009-06-14 23:45:34 +02:00
|
|
|
struct address_space *mapping,
|
|
|
|
loff_t pos, unsigned len, unsigned copied,
|
|
|
|
struct page *page, void *fsdata)
|
2009-06-05 06:56:49 +02:00
|
|
|
{
|
|
|
|
int i_size_changed = 0;
|
|
|
|
struct inode *inode = mapping->host;
|
|
|
|
handle_t *handle = ext4_journal_current_handle();
|
|
|
|
|
|
|
|
copied = block_write_end(file, mapping, pos, len, copied, page, fsdata);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* No need to use i_size_read() here, the i_size
|
|
|
|
* cannot change under us because we hold i_mutex.
|
|
|
|
*
|
|
|
|
* But it's important to update i_size while still holding page lock:
|
|
|
|
* page writeout could otherwise come in and zero beyond i_size.
|
|
|
|
*/
|
|
|
|
if (pos + copied > inode->i_size) {
|
|
|
|
i_size_write(inode, pos + copied);
|
|
|
|
i_size_changed = 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (pos + copied > EXT4_I(inode)->i_disksize) {
|
|
|
|
/* We need to mark inode dirty even if
|
|
|
|
* new_i_size is less that inode->i_size
|
|
|
|
* bu greater than i_disksize.(hint delalloc)
|
|
|
|
*/
|
|
|
|
ext4_update_i_disksize(inode, (pos + copied));
|
|
|
|
i_size_changed = 1;
|
|
|
|
}
|
|
|
|
unlock_page(page);
|
|
|
|
page_cache_release(page);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Don't mark the inode dirty under page lock. First, it unnecessarily
|
|
|
|
* makes the holding time of page lock longer. Second, it forces lock
|
|
|
|
* ordering of page lock and transaction start for journaling
|
|
|
|
* filesystems.
|
|
|
|
*/
|
|
|
|
if (i_size_changed)
|
|
|
|
ext4_mark_inode_dirty(handle, inode);
|
|
|
|
|
|
|
|
return copied;
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* We need to pick up the new inode size which generic_commit_write gave us
|
|
|
|
* `file' can be NULL - eg, when called from page_symlink().
|
|
|
|
*
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4 never places buffers on inode->i_mapping->private_list. metadata
|
2006-10-11 10:20:50 +02:00
|
|
|
* buffers are managed internally.
|
|
|
|
*/
|
2007-10-16 10:25:05 +02:00
|
|
|
static int ext4_ordered_write_end(struct file *file,
|
2009-06-14 23:45:34 +02:00
|
|
|
struct address_space *mapping,
|
|
|
|
loff_t pos, unsigned len, unsigned copied,
|
|
|
|
struct page *page, void *fsdata)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
handle_t *handle = ext4_journal_current_handle();
|
2008-07-12 01:27:31 +02:00
|
|
|
struct inode *inode = mapping->host;
|
2006-10-11 10:20:50 +02:00
|
|
|
int ret = 0, ret2;
|
|
|
|
|
2009-06-17 17:48:11 +02:00
|
|
|
trace_ext4_ordered_write_end(inode, pos, len, copied);
|
2008-07-12 01:27:31 +02:00
|
|
|
ret = ext4_jbd2_file_inode(handle, inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
if (ret == 0) {
|
2009-06-05 06:56:49 +02:00
|
|
|
ret2 = ext4_generic_write_end(file, mapping, pos, len, copied,
|
2007-10-16 10:25:05 +02:00
|
|
|
page, fsdata);
|
2008-04-30 04:01:18 +02:00
|
|
|
copied = ret2;
|
2009-07-13 22:22:22 +02:00
|
|
|
if (pos + len > inode->i_size && ext4_can_truncate(inode))
|
2009-06-05 06:56:49 +02:00
|
|
|
/* if we have allocated more blocks and copied
|
|
|
|
* less. We will have blocks allocated outside
|
|
|
|
* inode->i_size. So truncate them
|
|
|
|
*/
|
|
|
|
ext4_orphan_add(handle, inode);
|
2008-04-30 04:01:18 +02:00
|
|
|
if (ret2 < 0)
|
|
|
|
ret = ret2;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2006-10-11 10:20:53 +02:00
|
|
|
ret2 = ext4_journal_stop(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!ret)
|
|
|
|
ret = ret2;
|
2007-10-16 10:25:05 +02:00
|
|
|
|
2009-06-05 06:56:49 +02:00
|
|
|
if (pos + len > inode->i_size) {
|
2009-12-09 03:24:33 +01:00
|
|
|
ext4_truncate_failed_write(inode);
|
2009-06-14 23:45:34 +02:00
|
|
|
/*
|
2009-07-13 22:22:22 +02:00
|
|
|
* If truncate failed early the inode might still be
|
2009-06-05 06:56:49 +02:00
|
|
|
* on the orphan list; we need to make sure the inode
|
|
|
|
* is removed from the orphan list in that case.
|
|
|
|
*/
|
|
|
|
if (inode->i_nlink)
|
|
|
|
ext4_orphan_del(NULL, inode);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2007-10-16 10:25:05 +02:00
|
|
|
return ret ? ret : copied;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2007-10-16 10:25:05 +02:00
|
|
|
static int ext4_writeback_write_end(struct file *file,
|
2009-06-14 23:45:34 +02:00
|
|
|
struct address_space *mapping,
|
|
|
|
loff_t pos, unsigned len, unsigned copied,
|
|
|
|
struct page *page, void *fsdata)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
handle_t *handle = ext4_journal_current_handle();
|
2008-07-12 01:27:31 +02:00
|
|
|
struct inode *inode = mapping->host;
|
2006-10-11 10:20:50 +02:00
|
|
|
int ret = 0, ret2;
|
|
|
|
|
2009-06-17 17:48:11 +02:00
|
|
|
trace_ext4_writeback_write_end(inode, pos, len, copied);
|
2009-06-05 06:56:49 +02:00
|
|
|
ret2 = ext4_generic_write_end(file, mapping, pos, len, copied,
|
2007-10-16 10:25:05 +02:00
|
|
|
page, fsdata);
|
2008-04-30 04:01:18 +02:00
|
|
|
copied = ret2;
|
2009-07-13 22:22:22 +02:00
|
|
|
if (pos + len > inode->i_size && ext4_can_truncate(inode))
|
2009-06-05 06:56:49 +02:00
|
|
|
/* if we have allocated more blocks and copied
|
|
|
|
* less. We will have blocks allocated outside
|
|
|
|
* inode->i_size. So truncate them
|
|
|
|
*/
|
|
|
|
ext4_orphan_add(handle, inode);
|
|
|
|
|
2008-04-30 04:01:18 +02:00
|
|
|
if (ret2 < 0)
|
|
|
|
ret = ret2;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
ret2 = ext4_journal_stop(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!ret)
|
|
|
|
ret = ret2;
|
2007-10-16 10:25:05 +02:00
|
|
|
|
2009-06-05 06:56:49 +02:00
|
|
|
if (pos + len > inode->i_size) {
|
2009-12-09 03:24:33 +01:00
|
|
|
ext4_truncate_failed_write(inode);
|
2009-06-14 23:45:34 +02:00
|
|
|
/*
|
2009-07-13 22:22:22 +02:00
|
|
|
* If truncate failed early the inode might still be
|
2009-06-05 06:56:49 +02:00
|
|
|
* on the orphan list; we need to make sure the inode
|
|
|
|
* is removed from the orphan list in that case.
|
|
|
|
*/
|
|
|
|
if (inode->i_nlink)
|
|
|
|
ext4_orphan_del(NULL, inode);
|
|
|
|
}
|
|
|
|
|
2007-10-16 10:25:05 +02:00
|
|
|
return ret ? ret : copied;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2007-10-16 10:25:05 +02:00
|
|
|
static int ext4_journalled_write_end(struct file *file,
|
2009-06-14 23:45:34 +02:00
|
|
|
struct address_space *mapping,
|
|
|
|
loff_t pos, unsigned len, unsigned copied,
|
|
|
|
struct page *page, void *fsdata)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
handle_t *handle = ext4_journal_current_handle();
|
2007-10-16 10:25:05 +02:00
|
|
|
struct inode *inode = mapping->host;
|
2006-10-11 10:20:50 +02:00
|
|
|
int ret = 0, ret2;
|
|
|
|
int partial = 0;
|
2007-10-16 10:25:05 +02:00
|
|
|
unsigned from, to;
|
2008-09-13 19:06:18 +02:00
|
|
|
loff_t new_i_size;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2009-06-17 17:48:11 +02:00
|
|
|
trace_ext4_journalled_write_end(inode, pos, len, copied);
|
2007-10-16 10:25:05 +02:00
|
|
|
from = pos & (PAGE_CACHE_SIZE - 1);
|
|
|
|
to = from + len;
|
|
|
|
|
|
|
|
if (copied < len) {
|
|
|
|
if (!PageUptodate(page))
|
|
|
|
copied = 0;
|
|
|
|
page_zero_new_buffers(page, from+copied, to);
|
|
|
|
}
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
ret = walk_page_buffers(handle, page_buffers(page), from,
|
2007-10-16 10:25:05 +02:00
|
|
|
to, &partial, write_end_fn);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!partial)
|
|
|
|
SetPageUptodate(page);
|
2008-09-13 19:06:18 +02:00
|
|
|
new_i_size = pos + copied;
|
|
|
|
if (new_i_size > inode->i_size)
|
2007-10-16 10:25:05 +02:00
|
|
|
i_size_write(inode, pos+copied);
|
2010-01-24 20:34:07 +01:00
|
|
|
ext4_set_inode_state(inode, EXT4_STATE_JDATA);
|
2008-09-13 19:06:18 +02:00
|
|
|
if (new_i_size > EXT4_I(inode)->i_disksize) {
|
|
|
|
ext4_update_i_disksize(inode, new_i_size);
|
2006-10-11 10:20:53 +02:00
|
|
|
ret2 = ext4_mark_inode_dirty(handle, inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!ret)
|
|
|
|
ret = ret2;
|
|
|
|
}
|
2007-10-16 10:25:05 +02:00
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
unlock_page(page);
|
2009-06-05 06:56:49 +02:00
|
|
|
page_cache_release(page);
|
2009-07-13 22:22:22 +02:00
|
|
|
if (pos + len > inode->i_size && ext4_can_truncate(inode))
|
2009-06-05 06:56:49 +02:00
|
|
|
/* if we have allocated more blocks and copied
|
|
|
|
* less. We will have blocks allocated outside
|
|
|
|
* inode->i_size. So truncate them
|
|
|
|
*/
|
|
|
|
ext4_orphan_add(handle, inode);
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
ret2 = ext4_journal_stop(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!ret)
|
|
|
|
ret = ret2;
|
2009-06-05 06:56:49 +02:00
|
|
|
if (pos + len > inode->i_size) {
|
2009-12-09 03:24:33 +01:00
|
|
|
ext4_truncate_failed_write(inode);
|
2009-06-14 23:45:34 +02:00
|
|
|
/*
|
2009-07-13 22:22:22 +02:00
|
|
|
* If truncate failed early the inode might still be
|
2009-06-05 06:56:49 +02:00
|
|
|
* on the orphan list; we need to make sure the inode
|
|
|
|
* is removed from the orphan list in that case.
|
|
|
|
*/
|
|
|
|
if (inode->i_nlink)
|
|
|
|
ext4_orphan_del(NULL, inode);
|
|
|
|
}
|
2007-10-16 10:25:05 +02:00
|
|
|
|
|
|
|
return ret ? ret : copied;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2008-07-14 23:52:37 +02:00
|
|
|
|
2010-01-01 08:41:30 +01:00
|
|
|
/*
|
|
|
|
* Reserve a single block located at lblock
|
|
|
|
*/
|
|
|
|
static int ext4_da_reserve_space(struct inode *inode, sector_t lblock)
|
2008-07-14 23:52:37 +02:00
|
|
|
{
|
2008-09-09 05:14:50 +02:00
|
|
|
int retries = 0;
|
2009-01-22 18:13:05 +01:00
|
|
|
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
|
2009-12-30 20:20:45 +01:00
|
|
|
struct ext4_inode_info *ei = EXT4_I(inode);
|
2010-05-16 17:00:00 +02:00
|
|
|
unsigned long md_needed;
|
2010-03-03 15:05:00 +01:00
|
|
|
int ret;
|
2008-07-14 23:52:37 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* recalculate the amount of metadata blocks to reserve
|
|
|
|
* in order to allocate nrblocks
|
|
|
|
* worse case is one extent per block
|
|
|
|
*/
|
2008-09-09 05:14:50 +02:00
|
|
|
repeat:
|
2009-12-30 20:20:45 +01:00
|
|
|
spin_lock(&ei->i_block_reservation_lock);
|
2010-01-01 08:41:30 +01:00
|
|
|
md_needed = ext4_calc_metadata_amount(inode, lblock);
|
2010-01-01 07:00:21 +01:00
|
|
|
trace_ext4_da_reserve_space(inode, md_needed);
|
2009-12-30 20:20:45 +01:00
|
|
|
spin_unlock(&ei->i_block_reservation_lock);
|
2008-07-14 23:52:37 +02:00
|
|
|
|
2009-01-22 18:13:05 +01:00
|
|
|
/*
|
2010-05-16 17:00:00 +02:00
|
|
|
* We will charge metadata quota at writeout time; this saves
|
|
|
|
* us from metadata over-estimation, though we may go over by
|
|
|
|
* a small amount in the end. Here we just reserve for data.
|
2009-01-22 18:13:05 +01:00
|
|
|
*/
|
2010-05-16 17:00:00 +02:00
|
|
|
ret = dquot_reserve_block(inode, 1);
|
2010-03-03 15:05:00 +01:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
2010-05-16 17:00:00 +02:00
|
|
|
/*
|
|
|
|
* We do still charge estimated metadata to the sb though;
|
|
|
|
* we cannot afford to run out of free blocks.
|
|
|
|
*/
|
2010-01-01 08:41:30 +01:00
|
|
|
if (ext4_claim_free_blocks(sbi, md_needed + 1)) {
|
2010-05-16 17:00:00 +02:00
|
|
|
dquot_release_reservation_block(inode, 1);
|
2008-09-09 05:14:50 +02:00
|
|
|
if (ext4_should_retry_alloc(inode->i_sb, &retries)) {
|
|
|
|
yield();
|
|
|
|
goto repeat;
|
|
|
|
}
|
2008-07-14 23:52:37 +02:00
|
|
|
return -ENOSPC;
|
|
|
|
}
|
2009-12-30 20:20:45 +01:00
|
|
|
spin_lock(&ei->i_block_reservation_lock);
|
2010-01-01 08:41:30 +01:00
|
|
|
ei->i_reserved_data_blocks++;
|
2009-12-30 20:20:45 +01:00
|
|
|
ei->i_reserved_meta_blocks += md_needed;
|
|
|
|
spin_unlock(&ei->i_block_reservation_lock);
|
2009-12-10 17:36:27 +01:00
|
|
|
|
2008-07-14 23:52:37 +02:00
|
|
|
return 0; /* success */
|
|
|
|
}
|
|
|
|
|
2008-07-17 22:12:08 +02:00
|
|
|
static void ext4_da_release_space(struct inode *inode, int to_free)
|
2008-07-14 23:52:37 +02:00
|
|
|
{
|
|
|
|
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
|
2009-12-30 20:20:45 +01:00
|
|
|
struct ext4_inode_info *ei = EXT4_I(inode);
|
2008-07-14 23:52:37 +02:00
|
|
|
|
2008-08-20 04:16:59 +02:00
|
|
|
if (!to_free)
|
|
|
|
return; /* Nothing to release, exit */
|
|
|
|
|
2008-07-14 23:52:37 +02:00
|
|
|
spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
|
2008-08-20 04:16:59 +02:00
|
|
|
|
2010-05-17 08:00:00 +02:00
|
|
|
trace_ext4_da_release_space(inode, to_free);
|
2009-12-30 20:20:45 +01:00
|
|
|
if (unlikely(to_free > ei->i_reserved_data_blocks)) {
|
2008-08-20 04:16:59 +02:00
|
|
|
/*
|
2009-12-30 20:20:45 +01:00
|
|
|
* if there aren't enough reserved blocks, then the
|
|
|
|
* counter is messed up somewhere. Since this
|
|
|
|
* function is called from invalidate page, it's
|
|
|
|
* harmless to return without any action.
|
2008-08-20 04:16:59 +02:00
|
|
|
*/
|
2009-12-30 20:20:45 +01:00
|
|
|
ext4_msg(inode->i_sb, KERN_NOTICE, "ext4_da_release_space: "
|
|
|
|
"ino %lu, to_free %d with only %d reserved "
|
|
|
|
"data blocks\n", inode->i_ino, to_free,
|
|
|
|
ei->i_reserved_data_blocks);
|
|
|
|
WARN_ON(1);
|
|
|
|
to_free = ei->i_reserved_data_blocks;
|
2008-08-20 04:16:59 +02:00
|
|
|
}
|
2009-12-30 20:20:45 +01:00
|
|
|
ei->i_reserved_data_blocks -= to_free;
|
2008-08-20 04:16:59 +02:00
|
|
|
|
2009-12-30 20:20:45 +01:00
|
|
|
if (ei->i_reserved_data_blocks == 0) {
|
|
|
|
/*
|
|
|
|
* We can release all of the reserved metadata blocks
|
|
|
|
* only when we have written all of the delayed
|
|
|
|
* allocation blocks.
|
|
|
|
*/
|
2010-05-16 17:00:00 +02:00
|
|
|
percpu_counter_sub(&sbi->s_dirtyblocks_counter,
|
|
|
|
ei->i_reserved_meta_blocks);
|
2010-01-01 08:36:15 +01:00
|
|
|
ei->i_reserved_meta_blocks = 0;
|
2010-01-01 08:41:30 +01:00
|
|
|
ei->i_da_metadata_calc_len = 0;
|
2009-12-30 20:20:45 +01:00
|
|
|
}
|
2008-07-14 23:52:37 +02:00
|
|
|
|
2010-05-16 17:00:00 +02:00
|
|
|
/* update fs dirty data blocks counter */
|
2009-12-30 20:20:45 +01:00
|
|
|
percpu_counter_sub(&sbi->s_dirtyblocks_counter, to_free);
|
2008-07-14 23:52:37 +02:00
|
|
|
|
|
|
|
spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
|
2009-01-22 18:13:05 +01:00
|
|
|
|
2010-03-03 15:05:00 +01:00
|
|
|
dquot_release_reservation_block(inode, to_free);
|
2008-07-14 23:52:37 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void ext4_da_page_release_reservation(struct page *page,
|
2009-06-14 23:45:34 +02:00
|
|
|
unsigned long offset)
|
2008-07-14 23:52:37 +02:00
|
|
|
{
|
|
|
|
int to_release = 0;
|
|
|
|
struct buffer_head *head, *bh;
|
|
|
|
unsigned int curr_off = 0;
|
|
|
|
|
|
|
|
head = page_buffers(page);
|
|
|
|
bh = head;
|
|
|
|
do {
|
|
|
|
unsigned int next_off = curr_off + bh->b_size;
|
|
|
|
|
|
|
|
if ((offset <= curr_off) && (buffer_delay(bh))) {
|
|
|
|
to_release++;
|
|
|
|
clear_buffer_delay(bh);
|
|
|
|
}
|
|
|
|
curr_off = next_off;
|
|
|
|
} while ((bh = bh->b_this_page) != head);
|
2008-07-17 22:12:08 +02:00
|
|
|
ext4_da_release_space(page->mapping->host, to_release);
|
2008-07-14 23:52:37 +02:00
|
|
|
}
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* Delayed allocation stuff
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* mpage_da_submit_io - walks through extent of pages and try to write
|
2008-08-20 03:55:02 +02:00
|
|
|
* them with writepage() call back
|
2008-07-12 01:27:31 +02:00
|
|
|
*
|
|
|
|
* @mpd->inode: inode
|
|
|
|
* @mpd->first_page: first page of the extent
|
|
|
|
* @mpd->next_page: page after the last page of the extent
|
|
|
|
*
|
|
|
|
* By the time mpage_da_submit_io() is called we expect all blocks
|
|
|
|
* to be allocated. this may be wrong if allocation failed.
|
|
|
|
*
|
|
|
|
* As pages are already locked by write_cache_pages(), we can't use it
|
|
|
|
*/
|
|
|
|
static int mpage_da_submit_io(struct mpage_da_data *mpd)
|
|
|
|
{
|
2008-10-16 16:10:36 +02:00
|
|
|
long pages_skipped;
|
2009-01-06 03:50:43 +01:00
|
|
|
struct pagevec pvec;
|
|
|
|
unsigned long index, end;
|
|
|
|
int ret = 0, err, nr_pages, i;
|
|
|
|
struct inode *inode = mpd->inode;
|
|
|
|
struct address_space *mapping = inode->i_mapping;
|
2008-07-12 01:27:31 +02:00
|
|
|
|
|
|
|
BUG_ON(mpd->next_page <= mpd->first_page);
|
2009-01-06 03:50:43 +01:00
|
|
|
/*
|
|
|
|
* We need to start from the first_page to the next_page - 1
|
|
|
|
* to make sure we also write the mapped dirty buffer_heads.
|
2009-02-23 12:46:01 +01:00
|
|
|
* If we look at mpd->b_blocknr we would only be looking
|
2009-01-06 03:50:43 +01:00
|
|
|
* at the currently mapped buffer_heads.
|
|
|
|
*/
|
2008-07-12 01:27:31 +02:00
|
|
|
index = mpd->first_page;
|
|
|
|
end = mpd->next_page - 1;
|
|
|
|
|
2009-01-06 03:50:43 +01:00
|
|
|
pagevec_init(&pvec, 0);
|
2008-07-12 01:27:31 +02:00
|
|
|
while (index <= end) {
|
2009-01-06 03:50:43 +01:00
|
|
|
nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE);
|
2008-07-12 01:27:31 +02:00
|
|
|
if (nr_pages == 0)
|
|
|
|
break;
|
|
|
|
for (i = 0; i < nr_pages; i++) {
|
|
|
|
struct page *page = pvec.pages[i];
|
|
|
|
|
2009-01-06 03:50:43 +01:00
|
|
|
index = page->index;
|
|
|
|
if (index > end)
|
|
|
|
break;
|
|
|
|
index++;
|
|
|
|
|
|
|
|
BUG_ON(!PageLocked(page));
|
|
|
|
BUG_ON(PageWriteback(page));
|
|
|
|
|
2008-10-16 16:10:36 +02:00
|
|
|
pages_skipped = mpd->wbc->pages_skipped;
|
2008-08-20 03:55:02 +02:00
|
|
|
err = mapping->a_ops->writepage(page, mpd->wbc);
|
2008-10-16 16:10:36 +02:00
|
|
|
if (!err && (pages_skipped == mpd->wbc->pages_skipped))
|
|
|
|
/*
|
|
|
|
* have successfully written the page
|
|
|
|
* without skipping the same
|
|
|
|
*/
|
2008-08-20 03:55:02 +02:00
|
|
|
mpd->pages_written++;
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* In error case, we have to continue because
|
|
|
|
* remaining pages are still locked
|
|
|
|
* XXX: unlock and re-dirty them?
|
|
|
|
*/
|
|
|
|
if (ret == 0)
|
|
|
|
ret = err;
|
|
|
|
}
|
|
|
|
pagevec_release(&pvec);
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* mpage_put_bnr_to_bhs - walk blocks and assign them actual numbers
|
|
|
|
*
|
|
|
|
* the function goes through all passed space and put actual disk
|
2009-05-12 22:30:27 +02:00
|
|
|
* block numbers into buffer heads, dropping BH_Delay and BH_Unwritten
|
2008-07-12 01:27:31 +02:00
|
|
|
*/
|
2010-05-17 02:00:00 +02:00
|
|
|
static void mpage_put_bnr_to_bhs(struct mpage_da_data *mpd,
|
|
|
|
struct ext4_map_blocks *map)
|
2008-07-12 01:27:31 +02:00
|
|
|
{
|
|
|
|
struct inode *inode = mpd->inode;
|
|
|
|
struct address_space *mapping = inode->i_mapping;
|
2010-05-17 02:00:00 +02:00
|
|
|
int blocks = map->m_len;
|
|
|
|
sector_t pblock = map->m_pblk, cur_logical;
|
2008-07-12 01:27:31 +02:00
|
|
|
struct buffer_head *head, *bh;
|
2008-08-20 03:55:02 +02:00
|
|
|
pgoff_t index, end;
|
2008-07-12 01:27:31 +02:00
|
|
|
struct pagevec pvec;
|
|
|
|
int nr_pages, i;
|
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
index = map->m_lblk >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
|
|
|
|
end = (map->m_lblk + blocks - 1) >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
|
2008-07-12 01:27:31 +02:00
|
|
|
cur_logical = index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
|
|
|
|
|
|
|
|
pagevec_init(&pvec, 0);
|
|
|
|
|
|
|
|
while (index <= end) {
|
|
|
|
/* XXX: optimize tail */
|
|
|
|
nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE);
|
|
|
|
if (nr_pages == 0)
|
|
|
|
break;
|
|
|
|
for (i = 0; i < nr_pages; i++) {
|
|
|
|
struct page *page = pvec.pages[i];
|
|
|
|
|
|
|
|
index = page->index;
|
|
|
|
if (index > end)
|
|
|
|
break;
|
|
|
|
index++;
|
|
|
|
|
|
|
|
BUG_ON(!PageLocked(page));
|
|
|
|
BUG_ON(PageWriteback(page));
|
|
|
|
BUG_ON(!page_has_buffers(page));
|
|
|
|
|
|
|
|
bh = page_buffers(page);
|
|
|
|
head = bh;
|
|
|
|
|
|
|
|
/* skip blocks out of the range */
|
|
|
|
do {
|
2010-05-17 02:00:00 +02:00
|
|
|
if (cur_logical >= map->m_lblk)
|
2008-07-12 01:27:31 +02:00
|
|
|
break;
|
|
|
|
cur_logical++;
|
|
|
|
} while ((bh = bh->b_this_page) != head);
|
|
|
|
|
|
|
|
do {
|
2010-05-17 02:00:00 +02:00
|
|
|
if (cur_logical >= map->m_lblk + blocks)
|
2008-07-12 01:27:31 +02:00
|
|
|
break;
|
2009-05-12 22:30:27 +02:00
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
if (buffer_delay(bh) || buffer_unwritten(bh)) {
|
2009-05-12 22:30:27 +02:00
|
|
|
|
|
|
|
BUG_ON(bh->b_bdev != inode->i_sb->s_bdev);
|
|
|
|
|
|
|
|
if (buffer_delay(bh)) {
|
|
|
|
clear_buffer_delay(bh);
|
|
|
|
bh->b_blocknr = pblock;
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* unwritten already should have
|
|
|
|
* blocknr assigned. Verify that
|
|
|
|
*/
|
|
|
|
clear_buffer_unwritten(bh);
|
|
|
|
BUG_ON(bh->b_blocknr != pblock);
|
|
|
|
}
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
} else if (buffer_mapped(bh))
|
2008-07-12 01:27:31 +02:00
|
|
|
BUG_ON(bh->b_blocknr != pblock);
|
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
if (map->m_flags & EXT4_MAP_UNINIT)
|
2010-03-04 22:14:02 +01:00
|
|
|
set_buffer_uninit(bh);
|
2008-07-12 01:27:31 +02:00
|
|
|
cur_logical++;
|
|
|
|
pblock++;
|
|
|
|
} while ((bh = bh->b_this_page) != head);
|
|
|
|
}
|
|
|
|
pagevec_release(&pvec);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2008-08-20 03:08:18 +02:00
|
|
|
static void ext4_da_block_invalidatepages(struct mpage_da_data *mpd,
|
|
|
|
sector_t logical, long blk_cnt)
|
|
|
|
{
|
|
|
|
int nr_pages, i;
|
|
|
|
pgoff_t index, end;
|
|
|
|
struct pagevec pvec;
|
|
|
|
struct inode *inode = mpd->inode;
|
|
|
|
struct address_space *mapping = inode->i_mapping;
|
|
|
|
|
|
|
|
index = logical >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
|
|
|
|
end = (logical + blk_cnt - 1) >>
|
|
|
|
(PAGE_CACHE_SHIFT - inode->i_blkbits);
|
|
|
|
while (index <= end) {
|
|
|
|
nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE);
|
|
|
|
if (nr_pages == 0)
|
|
|
|
break;
|
|
|
|
for (i = 0; i < nr_pages; i++) {
|
|
|
|
struct page *page = pvec.pages[i];
|
2010-03-03 22:19:32 +01:00
|
|
|
if (page->index > end)
|
2008-08-20 03:08:18 +02:00
|
|
|
break;
|
|
|
|
BUG_ON(!PageLocked(page));
|
|
|
|
BUG_ON(PageWriteback(page));
|
|
|
|
block_invalidatepage(page, 0);
|
|
|
|
ClearPageUptodate(page);
|
|
|
|
unlock_page(page);
|
|
|
|
}
|
2010-03-03 22:19:32 +01:00
|
|
|
index = pvec.pages[nr_pages - 1]->index + 1;
|
|
|
|
pagevec_release(&pvec);
|
2008-08-20 03:08:18 +02:00
|
|
|
}
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2008-09-09 05:05:34 +02:00
|
|
|
static void ext4_print_free_blocks(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
|
2009-09-26 23:43:59 +02:00
|
|
|
printk(KERN_CRIT "Total free blocks count %lld\n",
|
|
|
|
ext4_count_free_blocks(inode->i_sb));
|
|
|
|
printk(KERN_CRIT "Free/Dirty block details\n");
|
|
|
|
printk(KERN_CRIT "free_blocks=%lld\n",
|
|
|
|
(long long) percpu_counter_sum(&sbi->s_freeblocks_counter));
|
|
|
|
printk(KERN_CRIT "dirty_blocks=%lld\n",
|
|
|
|
(long long) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
|
|
|
|
printk(KERN_CRIT "Block reservation details\n");
|
|
|
|
printk(KERN_CRIT "i_reserved_data_blocks=%u\n",
|
|
|
|
EXT4_I(inode)->i_reserved_data_blocks);
|
|
|
|
printk(KERN_CRIT "i_reserved_meta_blocks=%u\n",
|
|
|
|
EXT4_I(inode)->i_reserved_meta_blocks);
|
2008-09-09 05:05:34 +02:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* mpage_da_map_blocks - go through given space
|
|
|
|
*
|
2009-02-23 12:46:01 +01:00
|
|
|
* @mpd - bh describing space
|
2008-07-12 01:27:31 +02:00
|
|
|
*
|
|
|
|
* The function skips space we know is already mapped to disk blocks.
|
|
|
|
*
|
|
|
|
*/
|
2009-02-23 16:48:07 +01:00
|
|
|
static int mpage_da_map_blocks(struct mpage_da_data *mpd)
|
2008-07-12 01:27:31 +02:00
|
|
|
{
|
2009-05-14 19:57:08 +02:00
|
|
|
int err, blks, get_blocks_flags;
|
2010-05-17 02:00:00 +02:00
|
|
|
struct ext4_map_blocks map;
|
2009-05-14 15:29:45 +02:00
|
|
|
sector_t next = mpd->b_blocknr;
|
|
|
|
unsigned max_blocks = mpd->b_size >> mpd->inode->i_blkbits;
|
|
|
|
loff_t disksize = EXT4_I(mpd->inode)->i_disksize;
|
|
|
|
handle_t *handle = NULL;
|
2008-07-12 01:27:31 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* We consider only non-mapped and non-allocated blocks
|
|
|
|
*/
|
2009-02-23 12:46:01 +01:00
|
|
|
if ((mpd->b_state & (1 << BH_Mapped)) &&
|
2009-05-12 22:30:27 +02:00
|
|
|
!(mpd->b_state & (1 << BH_Delay)) &&
|
|
|
|
!(mpd->b_state & (1 << BH_Unwritten)))
|
2008-08-20 03:08:18 +02:00
|
|
|
return 0;
|
2009-05-14 15:29:45 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If we didn't accumulate anything to write simply return
|
|
|
|
*/
|
|
|
|
if (!mpd->b_size)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
handle = ext4_journal_current_handle();
|
|
|
|
BUG_ON(!handle);
|
|
|
|
|
2009-05-13 21:13:42 +02:00
|
|
|
/*
|
2009-05-14 19:57:08 +02:00
|
|
|
* Call ext4_get_blocks() to allocate any delayed allocation
|
|
|
|
* blocks, or to convert an uninitialized extent to be
|
|
|
|
* initialized (in the case where we have written into
|
|
|
|
* one or more preallocated blocks).
|
|
|
|
*
|
|
|
|
* We pass in the magic EXT4_GET_BLOCKS_DELALLOC_RESERVE to
|
|
|
|
* indicate that we are on the delayed allocation path. This
|
|
|
|
* affects functions in many different parts of the allocation
|
|
|
|
* call path. This flag exists primarily because we don't
|
|
|
|
* want to change *many* call functions, so ext4_get_blocks()
|
|
|
|
* will set the magic i_delalloc_reserved_flag once the
|
|
|
|
* inode's allocation semaphore is taken.
|
|
|
|
*
|
|
|
|
* If the blocks in questions were delalloc blocks, set
|
|
|
|
* EXT4_GET_BLOCKS_DELALLOC_RESERVE so the delalloc accounting
|
|
|
|
* variables are updated after the blocks have been allocated.
|
2009-05-13 21:13:42 +02:00
|
|
|
*/
|
2010-05-17 02:00:00 +02:00
|
|
|
map.m_lblk = next;
|
|
|
|
map.m_len = max_blocks;
|
2010-01-15 07:27:59 +01:00
|
|
|
get_blocks_flags = EXT4_GET_BLOCKS_CREATE;
|
2010-03-04 22:14:02 +01:00
|
|
|
if (ext4_should_dioread_nolock(mpd->inode))
|
|
|
|
get_blocks_flags |= EXT4_GET_BLOCKS_IO_CREATE_EXT;
|
2009-05-14 19:57:08 +02:00
|
|
|
if (mpd->b_state & (1 << BH_Delay))
|
2010-01-15 07:27:59 +01:00
|
|
|
get_blocks_flags |= EXT4_GET_BLOCKS_DELALLOC_RESERVE;
|
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
blks = ext4_map_blocks(handle, mpd->inode, &map, get_blocks_flags);
|
2009-05-14 15:29:45 +02:00
|
|
|
if (blks < 0) {
|
|
|
|
err = blks;
|
2009-02-23 16:48:07 +01:00
|
|
|
/*
|
|
|
|
* If get block returns with error we simply
|
|
|
|
* return. Later writepage will redirty the page and
|
|
|
|
* writepages will find the dirty page again
|
2008-08-20 03:08:18 +02:00
|
|
|
*/
|
|
|
|
if (err == -EAGAIN)
|
|
|
|
return 0;
|
2008-09-09 05:05:34 +02:00
|
|
|
|
|
|
|
if (err == -ENOSPC &&
|
2009-02-23 16:48:07 +01:00
|
|
|
ext4_count_free_blocks(mpd->inode->i_sb)) {
|
2008-09-09 05:05:34 +02:00
|
|
|
mpd->retval = err;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-08-20 03:08:18 +02:00
|
|
|
/*
|
2009-02-23 16:48:07 +01:00
|
|
|
* get block failure will cause us to loop in
|
|
|
|
* writepages, because a_ops->writepage won't be able
|
|
|
|
* to make progress. The page will be redirtied by
|
|
|
|
* writepage and writepages will again try to write
|
|
|
|
* the same.
|
2008-08-20 03:08:18 +02:00
|
|
|
*/
|
2009-09-26 23:43:59 +02:00
|
|
|
ext4_msg(mpd->inode->i_sb, KERN_CRIT,
|
|
|
|
"delayed block allocation failed for inode %lu at "
|
|
|
|
"logical offset %llu with max blocks %zd with "
|
2010-05-16 19:00:00 +02:00
|
|
|
"error %d", mpd->inode->i_ino,
|
2009-09-26 23:43:59 +02:00
|
|
|
(unsigned long long) next,
|
|
|
|
mpd->b_size >> mpd->inode->i_blkbits, err);
|
|
|
|
printk(KERN_CRIT "This should not happen!! "
|
|
|
|
"Data will be lost\n");
|
2008-09-09 05:14:50 +02:00
|
|
|
if (err == -ENOSPC) {
|
2008-09-09 05:05:34 +02:00
|
|
|
ext4_print_free_blocks(mpd->inode);
|
2008-09-09 05:14:50 +02:00
|
|
|
}
|
2009-05-14 15:29:45 +02:00
|
|
|
/* invalidate all the pages */
|
2008-08-20 03:08:18 +02:00
|
|
|
ext4_da_block_invalidatepages(mpd, next,
|
2009-02-23 12:46:01 +01:00
|
|
|
mpd->b_size >> mpd->inode->i_blkbits);
|
2008-08-20 03:08:18 +02:00
|
|
|
return err;
|
|
|
|
}
|
2009-05-14 15:29:45 +02:00
|
|
|
BUG_ON(blks == 0);
|
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
if (map.m_flags & EXT4_MAP_NEW) {
|
|
|
|
struct block_device *bdev = mpd->inode->i_sb->s_bdev;
|
|
|
|
int i;
|
2008-07-12 01:27:31 +02:00
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
for (i = 0; i < map.m_len; i++)
|
|
|
|
unmap_underlying_metadata(bdev, map.m_pblk + i);
|
|
|
|
}
|
2008-07-12 01:27:31 +02:00
|
|
|
|
2008-08-20 03:55:02 +02:00
|
|
|
/*
|
|
|
|
* If blocks are delayed marked, we need to
|
|
|
|
* put actual blocknr and drop delayed bit
|
|
|
|
*/
|
2009-02-23 12:46:01 +01:00
|
|
|
if ((mpd->b_state & (1 << BH_Delay)) ||
|
|
|
|
(mpd->b_state & (1 << BH_Unwritten)))
|
2010-05-17 02:00:00 +02:00
|
|
|
mpage_put_bnr_to_bhs(mpd, &map);
|
2008-07-12 01:27:31 +02:00
|
|
|
|
2009-05-14 15:29:45 +02:00
|
|
|
if (ext4_should_order_data(mpd->inode)) {
|
|
|
|
err = ext4_jbd2_file_inode(handle, mpd->inode);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2009-06-09 06:17:05 +02:00
|
|
|
* Update on-disk size along with block allocation.
|
2009-05-14 15:29:45 +02:00
|
|
|
*/
|
|
|
|
disksize = ((loff_t) next + blks) << mpd->inode->i_blkbits;
|
|
|
|
if (disksize > i_size_read(mpd->inode))
|
|
|
|
disksize = i_size_read(mpd->inode);
|
|
|
|
if (disksize > EXT4_I(mpd->inode)->i_disksize) {
|
|
|
|
ext4_update_i_disksize(mpd->inode, disksize);
|
|
|
|
return ext4_mark_inode_dirty(handle, mpd->inode);
|
|
|
|
}
|
|
|
|
|
2008-08-20 03:08:18 +02:00
|
|
|
return 0;
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
|
|
|
|
2008-08-20 04:16:43 +02:00
|
|
|
#define BH_FLAGS ((1 << BH_Uptodate) | (1 << BH_Mapped) | \
|
|
|
|
(1 << BH_Delay) | (1 << BH_Unwritten))
|
2008-07-12 01:27:31 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* mpage_add_bh_to_extent - try to add one more block to extent of blocks
|
|
|
|
*
|
|
|
|
* @mpd->lbh - extent of blocks
|
|
|
|
* @logical - logical number of the block in the file
|
|
|
|
* @bh - bh of the block (used to access block's state)
|
|
|
|
*
|
|
|
|
* the function is used to collect contig. blocks in same state
|
|
|
|
*/
|
|
|
|
static void mpage_add_bh_to_extent(struct mpage_da_data *mpd,
|
2009-02-23 12:46:01 +01:00
|
|
|
sector_t logical, size_t b_size,
|
|
|
|
unsigned long b_state)
|
2008-07-12 01:27:31 +02:00
|
|
|
{
|
|
|
|
sector_t next;
|
2009-02-23 12:46:01 +01:00
|
|
|
int nrblocks = mpd->b_size >> mpd->inode->i_blkbits;
|
2008-07-12 01:27:31 +02:00
|
|
|
|
ext4: don't scan/accumulate more pages than mballoc will allocate
There was a bug reported on RHEL5 that a 10G dd on a 12G box
had a very, very slow sync after that.
At issue was the loop in write_cache_pages scanning all the way
to the end of the 10G file, even though the subsequent call
to mpage_da_submit_io would only actually write a smallish amt; then
we went back to the write_cache_pages loop ... wasting tons of time
in calling __mpage_da_writepage for thousands of pages we would
just revisit (many times) later.
Upstream it's not such a big issue for sys_sync because we get
to the loop with a much smaller nr_to_write, which limits the loop.
However, talking with Aneesh he realized that fsync upstream still
gets here with a very large nr_to_write and we face the same problem.
This patch makes mpage_add_bh_to_extent stop the loop after we've
accumulated 2048 pages, by setting mpd->io_done = 1; which ultimately
causes the write_cache_pages loop to break.
Repeating the test with a dirty_ratio of 80 (to leave something for
fsync to do), I don't see huge IO performance gains, but the reduction
in cpu usage is striking: 80% usage with stock, and 2% with the
below patch. Instrumenting the loop in write_cache_pages clearly
shows that we are wasting time here.
Eventually we need to change mpage_da_map_pages() also submit its I/O
to the block layer, subsuming mpage_da_submit_io(), and then change it
call ext4_get_blocks() multiple times.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2010-05-16 10:00:00 +02:00
|
|
|
/*
|
|
|
|
* XXX Don't go larger than mballoc is willing to allocate
|
|
|
|
* This is a stopgap solution. We eventually need to fold
|
|
|
|
* mpage_da_submit_io() into this function and then call
|
|
|
|
* ext4_get_blocks() multiple times in a loop
|
|
|
|
*/
|
|
|
|
if (nrblocks >= 8*1024*1024/mpd->inode->i_sb->s_blocksize)
|
|
|
|
goto flush_it;
|
|
|
|
|
2008-08-20 04:15:58 +02:00
|
|
|
/* check if thereserved journal credits might overflow */
|
2010-05-17 04:00:00 +02:00
|
|
|
if (!(ext4_test_inode_flag(mpd->inode, EXT4_INODE_EXTENTS))) {
|
2008-08-20 04:15:58 +02:00
|
|
|
if (nrblocks >= EXT4_MAX_TRANS_DATA) {
|
|
|
|
/*
|
|
|
|
* With non-extent format we are limited by the journal
|
|
|
|
* credit available. Total credit needed to insert
|
|
|
|
* nrblocks contiguous blocks is dependent on the
|
|
|
|
* nrblocks. So limit nrblocks.
|
|
|
|
*/
|
|
|
|
goto flush_it;
|
|
|
|
} else if ((nrblocks + (b_size >> mpd->inode->i_blkbits)) >
|
|
|
|
EXT4_MAX_TRANS_DATA) {
|
|
|
|
/*
|
|
|
|
* Adding the new buffer_head would make it cross the
|
|
|
|
* allowed limit for which we have journal credit
|
|
|
|
* reserved. So limit the new bh->b_size
|
|
|
|
*/
|
|
|
|
b_size = (EXT4_MAX_TRANS_DATA - nrblocks) <<
|
|
|
|
mpd->inode->i_blkbits;
|
|
|
|
/* we will do mpage_da_submit_io in the next loop */
|
|
|
|
}
|
|
|
|
}
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* First block in the extent
|
|
|
|
*/
|
2009-02-23 12:46:01 +01:00
|
|
|
if (mpd->b_size == 0) {
|
|
|
|
mpd->b_blocknr = logical;
|
|
|
|
mpd->b_size = b_size;
|
|
|
|
mpd->b_state = b_state & BH_FLAGS;
|
2008-07-12 01:27:31 +02:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2009-02-23 12:46:01 +01:00
|
|
|
next = mpd->b_blocknr + nrblocks;
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* Can we merge the block to our big extent?
|
|
|
|
*/
|
2009-02-23 12:46:01 +01:00
|
|
|
if (logical == next && (b_state & BH_FLAGS) == mpd->b_state) {
|
|
|
|
mpd->b_size += b_size;
|
2008-07-12 01:27:31 +02:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2008-08-20 04:15:58 +02:00
|
|
|
flush_it:
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* We couldn't merge the block to our extent, so we
|
|
|
|
* need to flush current extent and start new one
|
|
|
|
*/
|
2008-08-20 03:08:18 +02:00
|
|
|
if (mpage_da_map_blocks(mpd) == 0)
|
|
|
|
mpage_da_submit_io(mpd);
|
2008-08-20 03:55:02 +02:00
|
|
|
mpd->io_done = 1;
|
|
|
|
return;
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
|
|
|
|
2009-06-14 23:57:10 +02:00
|
|
|
static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh)
|
2009-05-12 22:30:27 +02:00
|
|
|
{
|
2009-06-14 23:57:10 +02:00
|
|
|
return (buffer_delay(bh) || buffer_unwritten(bh)) && buffer_dirty(bh);
|
2009-05-12 22:30:27 +02:00
|
|
|
}
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* __mpage_da_writepage - finds extent of pages and blocks
|
|
|
|
*
|
|
|
|
* @page: page to consider
|
|
|
|
* @wbc: not used, we just follow rules
|
|
|
|
* @data: context
|
|
|
|
*
|
|
|
|
* The function finds extents of pages and scan them for all blocks.
|
|
|
|
*/
|
|
|
|
static int __mpage_da_writepage(struct page *page,
|
|
|
|
struct writeback_control *wbc, void *data)
|
|
|
|
{
|
|
|
|
struct mpage_da_data *mpd = data;
|
|
|
|
struct inode *inode = mpd->inode;
|
2009-02-23 12:46:01 +01:00
|
|
|
struct buffer_head *bh, *head;
|
2008-07-12 01:27:31 +02:00
|
|
|
sector_t logical;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Can we merge this page to current extent?
|
|
|
|
*/
|
|
|
|
if (mpd->next_page != page->index) {
|
|
|
|
/*
|
|
|
|
* Nope, we can't. So, we map non-allocated blocks
|
2008-08-20 03:55:02 +02:00
|
|
|
* and start IO on them using writepage()
|
2008-07-12 01:27:31 +02:00
|
|
|
*/
|
|
|
|
if (mpd->next_page != mpd->first_page) {
|
2008-08-20 03:08:18 +02:00
|
|
|
if (mpage_da_map_blocks(mpd) == 0)
|
|
|
|
mpage_da_submit_io(mpd);
|
2008-08-20 03:55:02 +02:00
|
|
|
/*
|
|
|
|
* skip rest of the page in the page_vec
|
|
|
|
*/
|
|
|
|
mpd->io_done = 1;
|
|
|
|
redirty_page_for_writepage(wbc, page);
|
|
|
|
unlock_page(page);
|
|
|
|
return MPAGE_DA_EXTENT_TAIL;
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Start next extent of pages ...
|
|
|
|
*/
|
|
|
|
mpd->first_page = page->index;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* ... and blocks
|
|
|
|
*/
|
2009-02-23 12:46:01 +01:00
|
|
|
mpd->b_size = 0;
|
|
|
|
mpd->b_state = 0;
|
|
|
|
mpd->b_blocknr = 0;
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
mpd->next_page = page->index + 1;
|
|
|
|
logical = (sector_t) page->index <<
|
|
|
|
(PAGE_CACHE_SHIFT - inode->i_blkbits);
|
|
|
|
|
|
|
|
if (!page_has_buffers(page)) {
|
2009-02-23 12:46:01 +01:00
|
|
|
mpage_add_bh_to_extent(mpd, logical, PAGE_CACHE_SIZE,
|
|
|
|
(1 << BH_Dirty) | (1 << BH_Uptodate));
|
2008-08-20 03:55:02 +02:00
|
|
|
if (mpd->io_done)
|
|
|
|
return MPAGE_DA_EXTENT_TAIL;
|
2008-07-12 01:27:31 +02:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Page with regular buffer heads, just add all dirty ones
|
|
|
|
*/
|
|
|
|
head = page_buffers(page);
|
|
|
|
bh = head;
|
|
|
|
do {
|
|
|
|
BUG_ON(buffer_locked(bh));
|
2009-01-06 03:50:43 +01:00
|
|
|
/*
|
|
|
|
* We need to try to allocate
|
|
|
|
* unmapped blocks in the same page.
|
|
|
|
* Otherwise we won't make progress
|
2009-06-14 23:58:45 +02:00
|
|
|
* with the page in ext4_writepage
|
2009-01-06 03:50:43 +01:00
|
|
|
*/
|
2009-06-14 23:57:10 +02:00
|
|
|
if (ext4_bh_delay_or_unwritten(NULL, bh)) {
|
2009-02-23 12:46:01 +01:00
|
|
|
mpage_add_bh_to_extent(mpd, logical,
|
|
|
|
bh->b_size,
|
|
|
|
bh->b_state);
|
2008-08-20 03:55:02 +02:00
|
|
|
if (mpd->io_done)
|
|
|
|
return MPAGE_DA_EXTENT_TAIL;
|
2009-01-06 03:50:43 +01:00
|
|
|
} else if (buffer_dirty(bh) && (buffer_mapped(bh))) {
|
|
|
|
/*
|
|
|
|
* mapped dirty buffer. We need to update
|
|
|
|
* the b_state because we look at
|
|
|
|
* b_state in mpage_da_map_blocks. We don't
|
|
|
|
* update b_size because if we find an
|
|
|
|
* unmapped buffer_head later we need to
|
|
|
|
* use the b_state flag of that buffer_head.
|
|
|
|
*/
|
2009-02-23 12:46:01 +01:00
|
|
|
if (mpd->b_size == 0)
|
|
|
|
mpd->b_state = bh->b_state & BH_FLAGS;
|
2008-08-20 03:55:02 +02:00
|
|
|
}
|
2008-07-12 01:27:31 +02:00
|
|
|
logical++;
|
|
|
|
} while ((bh = bh->b_this_page) != head);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2009-05-14 06:54:29 +02:00
|
|
|
* This is a special get_blocks_t callback which is used by
|
|
|
|
* ext4_da_write_begin(). It will either return mapped block or
|
|
|
|
* reserve space for a single block.
|
2009-05-12 22:30:27 +02:00
|
|
|
*
|
|
|
|
* For delayed buffer_head we have BH_Mapped, BH_New, BH_Delay set.
|
|
|
|
* We also have b_blocknr = -1 and b_bdev initialized properly
|
|
|
|
*
|
|
|
|
* For unwritten buffer_head we have BH_Mapped, BH_New, BH_Unwritten set.
|
|
|
|
* We also have b_blocknr = physicalblock mapping unwritten extent and b_bdev
|
|
|
|
* initialized properly.
|
2008-07-12 01:27:31 +02:00
|
|
|
*/
|
|
|
|
static int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
|
2010-05-17 02:00:00 +02:00
|
|
|
struct buffer_head *bh, int create)
|
2008-07-12 01:27:31 +02:00
|
|
|
{
|
2010-05-17 02:00:00 +02:00
|
|
|
struct ext4_map_blocks map;
|
2008-07-12 01:27:31 +02:00
|
|
|
int ret = 0;
|
2009-05-12 20:40:37 +02:00
|
|
|
sector_t invalid_block = ~((sector_t) 0xffff);
|
|
|
|
|
|
|
|
if (invalid_block < ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es))
|
|
|
|
invalid_block = ~0;
|
2008-07-12 01:27:31 +02:00
|
|
|
|
|
|
|
BUG_ON(create == 0);
|
2010-05-17 02:00:00 +02:00
|
|
|
BUG_ON(bh->b_size != inode->i_sb->s_blocksize);
|
|
|
|
|
|
|
|
map.m_lblk = iblock;
|
|
|
|
map.m_len = 1;
|
2008-07-12 01:27:31 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* first, we need to know whether the block is allocated already
|
|
|
|
* preallocated blocks are unmapped but should treated
|
|
|
|
* the same as allocated blocks.
|
|
|
|
*/
|
2010-05-17 02:00:00 +02:00
|
|
|
ret = ext4_map_blocks(NULL, inode, &map, 0);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
if (ret == 0) {
|
|
|
|
if (buffer_delay(bh))
|
|
|
|
return 0; /* Not sure this could or should happen */
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* XXX: __block_prepare_write() unmaps passed block,
|
|
|
|
* is it OK?
|
|
|
|
*/
|
2010-01-01 08:41:30 +01:00
|
|
|
ret = ext4_da_reserve_space(inode, iblock);
|
2008-07-14 23:52:37 +02:00
|
|
|
if (ret)
|
|
|
|
/* not enough space to reserve */
|
|
|
|
return ret;
|
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
map_bh(bh, inode->i_sb, invalid_block);
|
|
|
|
set_buffer_new(bh);
|
|
|
|
set_buffer_delay(bh);
|
|
|
|
return 0;
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
map_bh(bh, inode->i_sb, map.m_pblk);
|
|
|
|
bh->b_state = (bh->b_state & ~EXT4_MAP_FLAGS) | map.m_flags;
|
|
|
|
|
|
|
|
if (buffer_unwritten(bh)) {
|
|
|
|
/* A delayed write to unwritten bh should be marked
|
|
|
|
* new and mapped. Mapped ensures that we don't do
|
|
|
|
* get_block multiple times when we write to the same
|
|
|
|
* offset and new ensures that we do proper zero out
|
|
|
|
* for partial write.
|
|
|
|
*/
|
|
|
|
set_buffer_new(bh);
|
|
|
|
set_buffer_mapped(bh);
|
|
|
|
}
|
|
|
|
return 0;
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
2008-07-12 01:27:31 +02:00
|
|
|
|
2009-05-14 06:54:29 +02:00
|
|
|
/*
|
|
|
|
* This function is used as a standard get_block_t calback function
|
|
|
|
* when there is no desire to allocate any blocks. It is used as a
|
|
|
|
* callback function for block_prepare_write(), nobh_writepage(), and
|
|
|
|
* block_write_full_page(). These functions should only try to map a
|
|
|
|
* single block at a time.
|
|
|
|
*
|
|
|
|
* Since this function doesn't do block allocations even if the caller
|
|
|
|
* requests it by passing in create=1, it is critically important that
|
|
|
|
* any caller checks to make sure that any buffer heads are returned
|
|
|
|
* by this function are either all already mapped or marked for
|
|
|
|
* delayed allocation before calling nobh_writepage() or
|
|
|
|
* block_write_full_page(). Otherwise, b_blocknr could be left
|
|
|
|
* unitialized, and the page write functions will be taken by
|
|
|
|
* surprise.
|
|
|
|
*/
|
|
|
|
static int noalloc_get_block_write(struct inode *inode, sector_t iblock,
|
2008-07-12 01:27:31 +02:00
|
|
|
struct buffer_head *bh_result, int create)
|
|
|
|
{
|
2009-05-12 19:51:29 +02:00
|
|
|
BUG_ON(bh_result->b_size != inode->i_sb->s_blocksize);
|
2010-05-17 02:00:00 +02:00
|
|
|
return _ext4_get_block(inode, iblock, bh_result, 0);
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
|
|
|
|
2009-06-14 23:59:34 +02:00
|
|
|
static int bget_one(handle_t *handle, struct buffer_head *bh)
|
|
|
|
{
|
|
|
|
get_bh(bh);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int bput_one(handle_t *handle, struct buffer_head *bh)
|
|
|
|
{
|
|
|
|
put_bh(bh);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __ext4_journalled_writepage(struct page *page,
|
|
|
|
unsigned int len)
|
|
|
|
{
|
|
|
|
struct address_space *mapping = page->mapping;
|
|
|
|
struct inode *inode = mapping->host;
|
|
|
|
struct buffer_head *page_bufs;
|
|
|
|
handle_t *handle = NULL;
|
|
|
|
int ret = 0;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
page_bufs = page_buffers(page);
|
|
|
|
BUG_ON(!page_bufs);
|
|
|
|
walk_page_buffers(handle, page_bufs, 0, len, NULL, bget_one);
|
|
|
|
/* As soon as we unlock the page, it can go away, but we have
|
|
|
|
* references to buffers so we are safe */
|
|
|
|
unlock_page(page);
|
|
|
|
|
|
|
|
handle = ext4_journal_start(inode, ext4_writepage_trans_blocks(inode));
|
|
|
|
if (IS_ERR(handle)) {
|
|
|
|
ret = PTR_ERR(handle);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = walk_page_buffers(handle, page_bufs, 0, len, NULL,
|
|
|
|
do_journal_get_write_access);
|
|
|
|
|
|
|
|
err = walk_page_buffers(handle, page_bufs, 0, len, NULL,
|
|
|
|
write_end_fn);
|
|
|
|
if (ret == 0)
|
|
|
|
ret = err;
|
|
|
|
err = ext4_journal_stop(handle);
|
|
|
|
if (!ret)
|
|
|
|
ret = err;
|
|
|
|
|
|
|
|
walk_page_buffers(handle, page_bufs, 0, len, NULL, bput_one);
|
2010-01-24 20:34:07 +01:00
|
|
|
ext4_set_inode_state(inode, EXT4_STATE_JDATA);
|
2009-06-14 23:59:34 +02:00
|
|
|
out:
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2010-03-04 22:14:02 +01:00
|
|
|
static int ext4_set_bh_endio(struct buffer_head *bh, struct inode *inode);
|
|
|
|
static void ext4_end_io_buffer_write(struct buffer_head *bh, int uptodate);
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
2009-06-14 23:58:45 +02:00
|
|
|
* Note that we don't need to start a transaction unless we're journaling data
|
|
|
|
* because we should have holes filled from ext4_page_mkwrite(). We even don't
|
|
|
|
* need to file the inode to the transaction's list in ordered mode because if
|
|
|
|
* we are writing back data added by write(), the inode is already there and if
|
|
|
|
* we are writing back data modified via mmap(), noone guarantees in which
|
|
|
|
* transaction the data will hit the disk. In case we are journaling data, we
|
|
|
|
* cannot start transaction directly because transaction start ranks above page
|
|
|
|
* lock so we have to do some magic.
|
|
|
|
*
|
2009-05-14 06:54:29 +02:00
|
|
|
* This function can get called via...
|
|
|
|
* - ext4_da_writepages after taking page lock (have journal handle)
|
|
|
|
* - journal_submit_inode_data_buffers (no journal handle)
|
|
|
|
* - shrink_page_list via pdflush (no journal handle)
|
|
|
|
* - grab_page_cache when doing write_begin (have journal handle)
|
2009-06-14 23:58:45 +02:00
|
|
|
*
|
|
|
|
* We don't do any block allocation in this function. If we have page with
|
|
|
|
* multiple blocks we need to write those buffer_heads that are mapped. This
|
|
|
|
* is important for mmaped based write. So if we do with blocksize 1K
|
|
|
|
* truncate(f, 1024);
|
|
|
|
* a = mmap(f, 0, 4096);
|
|
|
|
* a[0] = 'a';
|
|
|
|
* truncate(f, 4096);
|
|
|
|
* we have in the page first buffer_head mapped via page_mkwrite call back
|
|
|
|
* but other bufer_heads would be unmapped but dirty(dirty done via the
|
|
|
|
* do_wp_page). So writepage should write the first block. If we modify
|
|
|
|
* the mmap area beyond 1024 we will again get a page_fault and the
|
|
|
|
* page_mkwrite callback will do the block allocation and mark the
|
|
|
|
* buffer_heads mapped.
|
|
|
|
*
|
|
|
|
* We redirty the page if we have any buffer_heads that is either delay or
|
|
|
|
* unwritten in the page.
|
|
|
|
*
|
|
|
|
* We can get recursively called as show below.
|
|
|
|
*
|
|
|
|
* ext4_writepage() -> kmalloc() -> __alloc_pages() -> page_launder() ->
|
|
|
|
* ext4_writepage()
|
|
|
|
*
|
|
|
|
* But since we don't do any block allocation we should not deadlock.
|
|
|
|
* Page also have the dirty flag cleared so we don't get recurive page_lock.
|
2008-07-12 01:27:31 +02:00
|
|
|
*/
|
2009-06-14 23:58:45 +02:00
|
|
|
static int ext4_writepage(struct page *page,
|
2009-06-14 23:59:34 +02:00
|
|
|
struct writeback_control *wbc)
|
2008-07-12 01:27:31 +02:00
|
|
|
{
|
|
|
|
int ret = 0;
|
2008-07-12 01:27:31 +02:00
|
|
|
loff_t size;
|
2008-11-05 06:14:04 +01:00
|
|
|
unsigned int len;
|
2010-03-04 22:14:02 +01:00
|
|
|
struct buffer_head *page_bufs = NULL;
|
2008-07-12 01:27:31 +02:00
|
|
|
struct inode *inode = page->mapping->host;
|
|
|
|
|
2009-06-14 23:58:45 +02:00
|
|
|
trace_ext4_writepage(inode, page);
|
2008-07-12 01:27:31 +02:00
|
|
|
size = i_size_read(inode);
|
|
|
|
if (page->index == size >> PAGE_CACHE_SHIFT)
|
|
|
|
len = size & ~PAGE_CACHE_MASK;
|
|
|
|
else
|
|
|
|
len = PAGE_CACHE_SIZE;
|
2008-07-12 01:27:31 +02:00
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
if (page_has_buffers(page)) {
|
2008-07-12 01:27:31 +02:00
|
|
|
page_bufs = page_buffers(page);
|
2008-07-12 01:27:31 +02:00
|
|
|
if (walk_page_buffers(NULL, page_bufs, 0, len, NULL,
|
2009-06-14 23:57:10 +02:00
|
|
|
ext4_bh_delay_or_unwritten)) {
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
2008-07-12 01:27:31 +02:00
|
|
|
* We don't want to do block allocation
|
|
|
|
* So redirty the page and return
|
2008-07-12 01:27:31 +02:00
|
|
|
* We may reach here when we do a journal commit
|
|
|
|
* via journal_submit_inode_data_buffers.
|
|
|
|
* If we don't have mapping block we just ignore
|
2008-07-12 01:27:31 +02:00
|
|
|
* them. We can also reach here via shrink_page_list
|
|
|
|
*/
|
|
|
|
redirty_page_for_writepage(wbc, page);
|
|
|
|
unlock_page(page);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* The test for page_has_buffers() is subtle:
|
|
|
|
* We know the page is dirty but it lost buffers. That means
|
|
|
|
* that at some moment in time after write_begin()/write_end()
|
|
|
|
* has been called all buffers have been clean and thus they
|
|
|
|
* must have been written at least once. So they are all
|
|
|
|
* mapped and we can happily proceed with mapping them
|
|
|
|
* and writing the page.
|
|
|
|
*
|
|
|
|
* Try to initialize the buffer_heads and check whether
|
|
|
|
* all are mapped and non delay. We don't want to
|
|
|
|
* do block allocation here.
|
|
|
|
*/
|
2009-06-04 14:06:06 +02:00
|
|
|
ret = block_prepare_write(page, 0, len,
|
2009-05-14 06:54:29 +02:00
|
|
|
noalloc_get_block_write);
|
2008-07-12 01:27:31 +02:00
|
|
|
if (!ret) {
|
|
|
|
page_bufs = page_buffers(page);
|
|
|
|
/* check whether all are mapped and non delay */
|
|
|
|
if (walk_page_buffers(NULL, page_bufs, 0, len, NULL,
|
2009-06-14 23:57:10 +02:00
|
|
|
ext4_bh_delay_or_unwritten)) {
|
2008-07-12 01:27:31 +02:00
|
|
|
redirty_page_for_writepage(wbc, page);
|
|
|
|
unlock_page(page);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* We can't do block allocation here
|
|
|
|
* so just redity the page and unlock
|
|
|
|
* and return
|
2008-07-12 01:27:31 +02:00
|
|
|
*/
|
|
|
|
redirty_page_for_writepage(wbc, page);
|
|
|
|
unlock_page(page);
|
|
|
|
return 0;
|
|
|
|
}
|
2008-11-07 15:06:45 +01:00
|
|
|
/* now mark the buffer_heads as dirty and uptodate */
|
2009-06-04 14:06:06 +02:00
|
|
|
block_commit_write(page, 0, len);
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
|
|
|
|
2009-06-14 23:58:45 +02:00
|
|
|
if (PageChecked(page) && ext4_should_journal_data(inode)) {
|
|
|
|
/*
|
|
|
|
* It's mmapped pagecache. Add buffers and journal it. There
|
|
|
|
* doesn't seem much point in redirtying the page here.
|
|
|
|
*/
|
|
|
|
ClearPageChecked(page);
|
2009-11-24 17:15:44 +01:00
|
|
|
return __ext4_journalled_writepage(page, len);
|
2009-06-14 23:58:45 +02:00
|
|
|
}
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
if (test_opt(inode->i_sb, NOBH) && ext4_should_writeback_data(inode))
|
2009-05-14 06:54:29 +02:00
|
|
|
ret = nobh_writepage(page, noalloc_get_block_write, wbc);
|
2010-03-04 22:14:02 +01:00
|
|
|
else if (page_bufs && buffer_uninit(page_bufs)) {
|
|
|
|
ext4_set_bh_endio(page_bufs, inode);
|
|
|
|
ret = block_write_full_page_endio(page, noalloc_get_block_write,
|
|
|
|
wbc, ext4_end_io_buffer_write);
|
|
|
|
} else
|
2009-05-14 06:54:29 +02:00
|
|
|
ret = block_write_full_page(page, noalloc_get_block_write,
|
|
|
|
wbc);
|
2008-07-12 01:27:31 +02:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
2008-08-20 04:15:58 +02:00
|
|
|
* This is called via ext4_da_writepages() to
|
|
|
|
* calulate the total number of credits to reserve to fit
|
|
|
|
* a single extent allocation into a single transaction,
|
|
|
|
* ext4_da_writpeages() will loop calling this before
|
|
|
|
* the block allocation.
|
2008-07-12 01:27:31 +02:00
|
|
|
*/
|
2008-08-20 04:15:58 +02:00
|
|
|
|
|
|
|
static int ext4_da_writepages_trans_blocks(struct inode *inode)
|
|
|
|
{
|
|
|
|
int max_blocks = EXT4_I(inode)->i_reserved_data_blocks;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* With non-extent format the journal credit needed to
|
|
|
|
* insert nrblocks contiguous block is dependent on
|
|
|
|
* number of contiguous block. So we will limit
|
|
|
|
* number of contiguous block to a sane value
|
|
|
|
*/
|
2010-05-17 04:00:00 +02:00
|
|
|
if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) &&
|
2008-08-20 04:15:58 +02:00
|
|
|
(max_blocks > EXT4_MAX_TRANS_DATA))
|
|
|
|
max_blocks = EXT4_MAX_TRANS_DATA;
|
|
|
|
|
|
|
|
return ext4_chunk_trans_blocks(inode, max_blocks);
|
|
|
|
}
|
2008-07-12 01:27:31 +02:00
|
|
|
|
2010-05-17 00:00:00 +02:00
|
|
|
/*
|
|
|
|
* write_cache_pages_da - walk the list of dirty pages of the given
|
|
|
|
* address space and call the callback function (which usually writes
|
|
|
|
* the pages).
|
|
|
|
*
|
|
|
|
* This is a forked version of write_cache_pages(). Differences:
|
|
|
|
* Range cyclic is ignored.
|
|
|
|
* no_nrwrite_index_update is always presumed true
|
|
|
|
*/
|
|
|
|
static int write_cache_pages_da(struct address_space *mapping,
|
|
|
|
struct writeback_control *wbc,
|
|
|
|
struct mpage_da_data *mpd)
|
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
int done = 0;
|
|
|
|
struct pagevec pvec;
|
|
|
|
int nr_pages;
|
|
|
|
pgoff_t index;
|
|
|
|
pgoff_t end; /* Inclusive */
|
|
|
|
long nr_to_write = wbc->nr_to_write;
|
|
|
|
|
|
|
|
pagevec_init(&pvec, 0);
|
|
|
|
index = wbc->range_start >> PAGE_CACHE_SHIFT;
|
|
|
|
end = wbc->range_end >> PAGE_CACHE_SHIFT;
|
|
|
|
|
|
|
|
while (!done && (index <= end)) {
|
|
|
|
int i;
|
|
|
|
|
|
|
|
nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
|
|
|
|
PAGECACHE_TAG_DIRTY,
|
|
|
|
min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1);
|
|
|
|
if (nr_pages == 0)
|
|
|
|
break;
|
|
|
|
|
|
|
|
for (i = 0; i < nr_pages; i++) {
|
|
|
|
struct page *page = pvec.pages[i];
|
|
|
|
|
|
|
|
/*
|
|
|
|
* At this point, the page may be truncated or
|
|
|
|
* invalidated (changing page->mapping to NULL), or
|
|
|
|
* even swizzled back from swapper_space to tmpfs file
|
|
|
|
* mapping. However, page->index will not change
|
|
|
|
* because we have a reference on the page.
|
|
|
|
*/
|
|
|
|
if (page->index > end) {
|
|
|
|
done = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
lock_page(page);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Page truncated or invalidated. We can freely skip it
|
|
|
|
* then, even for data integrity operations: the page
|
|
|
|
* has disappeared concurrently, so there could be no
|
|
|
|
* real expectation of this data interity operation
|
|
|
|
* even if there is now a new, dirty page at the same
|
|
|
|
* pagecache address.
|
|
|
|
*/
|
|
|
|
if (unlikely(page->mapping != mapping)) {
|
|
|
|
continue_unlock:
|
|
|
|
unlock_page(page);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!PageDirty(page)) {
|
|
|
|
/* someone wrote it for us */
|
|
|
|
goto continue_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (PageWriteback(page)) {
|
|
|
|
if (wbc->sync_mode != WB_SYNC_NONE)
|
|
|
|
wait_on_page_writeback(page);
|
|
|
|
else
|
|
|
|
goto continue_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
BUG_ON(PageWriteback(page));
|
|
|
|
if (!clear_page_dirty_for_io(page))
|
|
|
|
goto continue_unlock;
|
|
|
|
|
|
|
|
ret = __mpage_da_writepage(page, wbc, mpd);
|
|
|
|
if (unlikely(ret)) {
|
|
|
|
if (ret == AOP_WRITEPAGE_ACTIVATE) {
|
|
|
|
unlock_page(page);
|
|
|
|
ret = 0;
|
|
|
|
} else {
|
|
|
|
done = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (nr_to_write > 0) {
|
|
|
|
nr_to_write--;
|
|
|
|
if (nr_to_write == 0 &&
|
|
|
|
wbc->sync_mode == WB_SYNC_NONE) {
|
|
|
|
/*
|
|
|
|
* We stop writing back only if we are
|
|
|
|
* not doing integrity sync. In case of
|
|
|
|
* integrity sync we have to keep going
|
|
|
|
* because someone may be concurrently
|
|
|
|
* dirtying pages, and we might have
|
|
|
|
* synced a lot of newly appeared dirty
|
|
|
|
* pages, but have not synced all of the
|
|
|
|
* old dirty pages.
|
|
|
|
*/
|
|
|
|
done = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
pagevec_release(&pvec);
|
|
|
|
cond_resched();
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
static int ext4_da_writepages(struct address_space *mapping,
|
2008-08-20 03:55:02 +02:00
|
|
|
struct writeback_control *wbc)
|
2008-07-12 01:27:31 +02:00
|
|
|
{
|
2008-10-16 16:10:36 +02:00
|
|
|
pgoff_t index;
|
|
|
|
int range_whole = 0;
|
2008-07-12 01:27:31 +02:00
|
|
|
handle_t *handle = NULL;
|
2008-09-09 05:05:34 +02:00
|
|
|
struct mpage_da_data mpd;
|
2008-08-19 00:00:57 +02:00
|
|
|
struct inode *inode = mapping->host;
|
2008-11-05 06:14:04 +01:00
|
|
|
int pages_written = 0;
|
|
|
|
long pages_skipped;
|
2009-09-29 19:31:31 +02:00
|
|
|
unsigned int max_pages;
|
2009-02-14 16:42:58 +01:00
|
|
|
int range_cyclic, cycled = 1, io_done = 0;
|
2009-09-29 19:31:31 +02:00
|
|
|
int needed_blocks, ret = 0;
|
|
|
|
long desired_nr_to_write, nr_to_writebump = 0;
|
2009-08-31 23:00:59 +02:00
|
|
|
loff_t range_start = wbc->range_start;
|
2008-08-19 00:00:57 +02:00
|
|
|
struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
|
2008-07-12 01:27:31 +02:00
|
|
|
|
2009-06-17 17:48:11 +02:00
|
|
|
trace_ext4_da_writepages(inode, wbc);
|
2009-01-04 02:03:21 +01:00
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* No pages to write? This is mainly a kludge to avoid starting
|
|
|
|
* a transaction for special inodes like journal inode on last iput()
|
|
|
|
* because that could violate lock ordering on umount
|
|
|
|
*/
|
2008-08-20 03:55:02 +02:00
|
|
|
if (!mapping->nrpages || !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
|
2008-07-12 01:27:31 +02:00
|
|
|
return 0;
|
2008-11-05 15:22:24 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If the filesystem has aborted, it is read-only, so return
|
|
|
|
* right away instead of dumping stack traces later on that
|
|
|
|
* will obscure the real source of the problem. We test
|
2009-06-13 16:09:36 +02:00
|
|
|
* EXT4_MF_FS_ABORTED instead of sb->s_flag's MS_RDONLY because
|
2008-11-05 15:22:24 +01:00
|
|
|
* the latter could be true if the filesystem is mounted
|
|
|
|
* read-only, and in that case, ext4_da_writepages should
|
|
|
|
* *never* be called, so if that ever happens, we would want
|
|
|
|
* the stack trace.
|
|
|
|
*/
|
2009-06-13 16:09:36 +02:00
|
|
|
if (unlikely(sbi->s_mount_flags & EXT4_MF_FS_ABORTED))
|
2008-11-05 15:22:24 +01:00
|
|
|
return -EROFS;
|
|
|
|
|
2008-10-16 16:10:36 +02:00
|
|
|
if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
|
|
|
|
range_whole = 1;
|
2008-07-12 01:27:31 +02:00
|
|
|
|
2009-02-14 16:42:58 +01:00
|
|
|
range_cyclic = wbc->range_cyclic;
|
|
|
|
if (wbc->range_cyclic) {
|
2008-10-16 16:10:36 +02:00
|
|
|
index = mapping->writeback_index;
|
2009-02-14 16:42:58 +01:00
|
|
|
if (index)
|
|
|
|
cycled = 0;
|
|
|
|
wbc->range_start = index << PAGE_CACHE_SHIFT;
|
|
|
|
wbc->range_end = LLONG_MAX;
|
|
|
|
wbc->range_cyclic = 0;
|
|
|
|
} else
|
2008-10-16 16:10:36 +02:00
|
|
|
index = wbc->range_start >> PAGE_CACHE_SHIFT;
|
2008-08-20 03:55:02 +02:00
|
|
|
|
2009-09-29 19:31:31 +02:00
|
|
|
/*
|
|
|
|
* This works around two forms of stupidity. The first is in
|
|
|
|
* the writeback code, which caps the maximum number of pages
|
|
|
|
* written to be 1024 pages. This is wrong on multiple
|
|
|
|
* levels; different architectues have a different page size,
|
|
|
|
* which changes the maximum amount of data which gets
|
|
|
|
* written. Secondly, 4 megabytes is way too small. XFS
|
|
|
|
* forces this value to be 16 megabytes by multiplying
|
|
|
|
* nr_to_write parameter by four, and then relies on its
|
|
|
|
* allocator to allocate larger extents to make them
|
|
|
|
* contiguous. Unfortunately this brings us to the second
|
|
|
|
* stupidity, which is that ext4's mballoc code only allocates
|
|
|
|
* at most 2048 blocks. So we force contiguous writes up to
|
|
|
|
* the number of dirty blocks in the inode, or
|
|
|
|
* sbi->max_writeback_mb_bump whichever is smaller.
|
|
|
|
*/
|
|
|
|
max_pages = sbi->s_max_writeback_mb_bump << (20 - PAGE_CACHE_SHIFT);
|
|
|
|
if (!range_cyclic && range_whole)
|
|
|
|
desired_nr_to_write = wbc->nr_to_write * 8;
|
|
|
|
else
|
|
|
|
desired_nr_to_write = ext4_num_dirty_pages(inode, index,
|
|
|
|
max_pages);
|
|
|
|
if (desired_nr_to_write > max_pages)
|
|
|
|
desired_nr_to_write = max_pages;
|
|
|
|
|
|
|
|
if (wbc->nr_to_write < desired_nr_to_write) {
|
|
|
|
nr_to_writebump = desired_nr_to_write - wbc->nr_to_write;
|
|
|
|
wbc->nr_to_write = desired_nr_to_write;
|
|
|
|
}
|
|
|
|
|
2008-09-09 05:05:34 +02:00
|
|
|
mpd.wbc = wbc;
|
|
|
|
mpd.inode = mapping->host;
|
|
|
|
|
2008-10-16 16:10:36 +02:00
|
|
|
pages_skipped = wbc->pages_skipped;
|
|
|
|
|
2009-02-14 16:42:58 +01:00
|
|
|
retry:
|
2008-10-16 16:10:36 +02:00
|
|
|
while (!ret && wbc->nr_to_write > 0) {
|
2008-08-20 03:55:02 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* we insert one extent at a time. So we need
|
|
|
|
* credit needed for single extent allocation.
|
|
|
|
* journalled mode is currently not supported
|
|
|
|
* by delalloc
|
|
|
|
*/
|
|
|
|
BUG_ON(ext4_should_journal_data(inode));
|
2008-08-20 04:15:58 +02:00
|
|
|
needed_blocks = ext4_da_writepages_trans_blocks(inode);
|
2008-08-20 03:55:02 +02:00
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
/* start a new transaction*/
|
|
|
|
handle = ext4_journal_start(inode, needed_blocks);
|
|
|
|
if (IS_ERR(handle)) {
|
|
|
|
ret = PTR_ERR(handle);
|
2009-09-26 23:43:59 +02:00
|
|
|
ext4_msg(inode->i_sb, KERN_CRIT, "%s: jbd2_start: "
|
2010-05-16 19:00:00 +02:00
|
|
|
"%ld pages, ino %lu; err %d", __func__,
|
2008-08-20 03:55:02 +02:00
|
|
|
wbc->nr_to_write, inode->i_ino, ret);
|
2008-07-12 01:27:31 +02:00
|
|
|
goto out_writepages;
|
|
|
|
}
|
2009-02-23 22:42:39 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Now call __mpage_da_writepage to find the next
|
|
|
|
* contiguous region of logical blocks that need
|
|
|
|
* blocks to be allocated by ext4. We don't actually
|
|
|
|
* submit the blocks for I/O here, even though
|
|
|
|
* write_cache_pages thinks it will, and will set the
|
|
|
|
* pages as clean for write before calling
|
|
|
|
* __mpage_da_writepage().
|
|
|
|
*/
|
|
|
|
mpd.b_size = 0;
|
|
|
|
mpd.b_state = 0;
|
|
|
|
mpd.b_blocknr = 0;
|
|
|
|
mpd.first_page = 0;
|
|
|
|
mpd.next_page = 0;
|
|
|
|
mpd.io_done = 0;
|
|
|
|
mpd.pages_written = 0;
|
|
|
|
mpd.retval = 0;
|
2010-05-17 00:00:00 +02:00
|
|
|
ret = write_cache_pages_da(mapping, wbc, &mpd);
|
2009-02-23 22:42:39 +01:00
|
|
|
/*
|
tree-wide: fix assorted typos all over the place
That is "success", "unknown", "through", "performance", "[re|un]mapping"
, "access", "default", "reasonable", "[con]currently", "temperature"
, "channel", "[un]used", "application", "example","hierarchy", "therefore"
, "[over|under]flow", "contiguous", "threshold", "enough" and others.
Signed-off-by: André Goddard Rosa <andre.goddard@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2009-11-14 16:09:05 +01:00
|
|
|
* If we have a contiguous extent of pages and we
|
2009-02-23 22:42:39 +01:00
|
|
|
* haven't done the I/O yet, map the blocks and submit
|
|
|
|
* them for I/O.
|
|
|
|
*/
|
|
|
|
if (!mpd.io_done && mpd.next_page != mpd.first_page) {
|
|
|
|
if (mpage_da_map_blocks(&mpd) == 0)
|
|
|
|
mpage_da_submit_io(&mpd);
|
|
|
|
mpd.io_done = 1;
|
|
|
|
ret = MPAGE_DA_EXTENT_TAIL;
|
|
|
|
}
|
2009-09-01 05:13:11 +02:00
|
|
|
trace_ext4_da_write_pages(inode, &mpd);
|
2009-02-23 22:42:39 +01:00
|
|
|
wbc->nr_to_write -= mpd.pages_written;
|
2008-09-09 05:05:34 +02:00
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
ext4_journal_stop(handle);
|
2008-09-09 05:05:34 +02:00
|
|
|
|
2009-02-26 06:57:35 +01:00
|
|
|
if ((mpd.retval == -ENOSPC) && sbi->s_journal) {
|
2008-10-16 16:10:36 +02:00
|
|
|
/* commit the transaction which would
|
|
|
|
* free blocks released in the transaction
|
|
|
|
* and try again
|
|
|
|
*/
|
2008-09-09 05:05:34 +02:00
|
|
|
jbd2_journal_force_commit_nested(sbi->s_journal);
|
2008-10-16 16:10:36 +02:00
|
|
|
wbc->pages_skipped = pages_skipped;
|
|
|
|
ret = 0;
|
|
|
|
} else if (ret == MPAGE_DA_EXTENT_TAIL) {
|
2008-08-20 03:55:02 +02:00
|
|
|
/*
|
|
|
|
* got one extent now try with
|
|
|
|
* rest of the pages
|
|
|
|
*/
|
2008-10-16 16:10:36 +02:00
|
|
|
pages_written += mpd.pages_written;
|
|
|
|
wbc->pages_skipped = pages_skipped;
|
2008-08-20 03:55:02 +02:00
|
|
|
ret = 0;
|
2009-02-14 16:42:58 +01:00
|
|
|
io_done = 1;
|
2008-10-16 16:10:36 +02:00
|
|
|
} else if (wbc->nr_to_write)
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* There is no more writeout needed
|
|
|
|
* or we requested for a noblocking writeout
|
|
|
|
* and we found the device congested
|
|
|
|
*/
|
|
|
|
break;
|
2008-08-20 03:55:02 +02:00
|
|
|
}
|
2009-02-14 16:42:58 +01:00
|
|
|
if (!io_done && !cycled) {
|
|
|
|
cycled = 1;
|
|
|
|
index = 0;
|
|
|
|
wbc->range_start = index << PAGE_CACHE_SHIFT;
|
|
|
|
wbc->range_end = mapping->writeback_index - 1;
|
|
|
|
goto retry;
|
|
|
|
}
|
2008-10-16 16:10:36 +02:00
|
|
|
if (pages_skipped != wbc->pages_skipped)
|
2009-09-26 23:43:59 +02:00
|
|
|
ext4_msg(inode->i_sb, KERN_CRIT,
|
|
|
|
"This should not happen leaving %s "
|
2010-05-16 19:00:00 +02:00
|
|
|
"with nr_to_write = %ld ret = %d",
|
2009-09-26 23:43:59 +02:00
|
|
|
__func__, wbc->nr_to_write, ret);
|
2008-10-16 16:10:36 +02:00
|
|
|
|
|
|
|
/* Update index */
|
|
|
|
index += pages_written;
|
2009-02-14 16:42:58 +01:00
|
|
|
wbc->range_cyclic = range_cyclic;
|
2008-10-16 16:10:36 +02:00
|
|
|
if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
|
|
|
|
/*
|
|
|
|
* set the writeback_index so that range_cyclic
|
|
|
|
* mode will write it back later
|
|
|
|
*/
|
|
|
|
mapping->writeback_index = index;
|
2008-08-20 03:55:02 +02:00
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
out_writepages:
|
2009-12-25 21:46:07 +01:00
|
|
|
wbc->nr_to_write -= nr_to_writebump;
|
2009-08-31 23:00:59 +02:00
|
|
|
wbc->range_start = range_start;
|
2009-06-17 17:48:11 +02:00
|
|
|
trace_ext4_da_writepages_result(inode, wbc, ret, pages_written);
|
2008-07-12 01:27:31 +02:00
|
|
|
return ret;
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
|
|
|
|
2008-10-09 05:13:30 +02:00
|
|
|
#define FALL_BACK_TO_NONDELALLOC 1
|
|
|
|
static int ext4_nonda_switch(struct super_block *sb)
|
|
|
|
{
|
|
|
|
s64 free_blocks, dirty_blocks;
|
|
|
|
struct ext4_sb_info *sbi = EXT4_SB(sb);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* switch to non delalloc mode if we are running low
|
|
|
|
* on free block. The free block accounting via percpu
|
2009-01-06 23:41:04 +01:00
|
|
|
* counters can get slightly wrong with percpu_counter_batch getting
|
2008-10-09 05:13:30 +02:00
|
|
|
* accumulated on each CPU without updating global counters
|
|
|
|
* Delalloc need an accurate free block accounting. So switch
|
|
|
|
* to non delalloc when we are near to error range.
|
|
|
|
*/
|
|
|
|
free_blocks = percpu_counter_read_positive(&sbi->s_freeblocks_counter);
|
|
|
|
dirty_blocks = percpu_counter_read_positive(&sbi->s_dirtyblocks_counter);
|
|
|
|
if (2 * free_blocks < 3 * dirty_blocks ||
|
|
|
|
free_blocks < (dirty_blocks + EXT4_FREEBLOCKS_WATERMARK)) {
|
|
|
|
/*
|
2009-12-23 13:58:12 +01:00
|
|
|
* free block count is less than 150% of dirty blocks
|
|
|
|
* or free blocks is less than watermark
|
2008-10-09 05:13:30 +02:00
|
|
|
*/
|
|
|
|
return 1;
|
|
|
|
}
|
2009-12-23 13:58:12 +01:00
|
|
|
/*
|
|
|
|
* Even if we don't switch but are nearing capacity,
|
|
|
|
* start pushing delalloc when 1/2 of free blocks are dirty.
|
|
|
|
*/
|
|
|
|
if (free_blocks < 2 * dirty_blocks)
|
|
|
|
writeback_inodes_sb_if_idle(sb);
|
|
|
|
|
2008-10-09 05:13:30 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
static int ext4_da_write_begin(struct file *file, struct address_space *mapping,
|
2009-06-14 23:45:34 +02:00
|
|
|
loff_t pos, unsigned len, unsigned flags,
|
|
|
|
struct page **pagep, void **fsdata)
|
2008-07-12 01:27:31 +02:00
|
|
|
{
|
2010-05-16 17:00:00 +02:00
|
|
|
int ret, retries = 0;
|
2008-07-12 01:27:31 +02:00
|
|
|
struct page *page;
|
|
|
|
pgoff_t index;
|
|
|
|
unsigned from, to;
|
|
|
|
struct inode *inode = mapping->host;
|
|
|
|
handle_t *handle;
|
|
|
|
|
|
|
|
index = pos >> PAGE_CACHE_SHIFT;
|
|
|
|
from = pos & (PAGE_CACHE_SIZE - 1);
|
|
|
|
to = from + len;
|
2008-10-09 05:13:30 +02:00
|
|
|
|
|
|
|
if (ext4_nonda_switch(inode->i_sb)) {
|
|
|
|
*fsdata = (void *)FALL_BACK_TO_NONDELALLOC;
|
|
|
|
return ext4_write_begin(file, mapping, pos,
|
|
|
|
len, flags, pagep, fsdata);
|
|
|
|
}
|
|
|
|
*fsdata = (void *)0;
|
2009-06-17 17:48:11 +02:00
|
|
|
trace_ext4_da_write_begin(inode, pos, len, flags);
|
2008-07-14 23:52:37 +02:00
|
|
|
retry:
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* With delayed allocation, we don't log the i_disksize update
|
|
|
|
* if there is delayed block allocation. But we still need
|
|
|
|
* to journalling the i_disksize update if writes to the end
|
|
|
|
* of file which has an already mapped buffer.
|
|
|
|
*/
|
|
|
|
handle = ext4_journal_start(inode, 1);
|
|
|
|
if (IS_ERR(handle)) {
|
|
|
|
ret = PTR_ERR(handle);
|
|
|
|
goto out;
|
|
|
|
}
|
2009-02-23 03:09:59 +01:00
|
|
|
/* We cannot recurse into the filesystem as the transaction is already
|
|
|
|
* started */
|
|
|
|
flags |= AOP_FLAG_NOFS;
|
2008-07-12 01:27:31 +02:00
|
|
|
|
fs: symlink write_begin allocation context fix
With the write_begin/write_end aops, page_symlink was broken because it
could no longer pass a GFP_NOFS type mask into the point where the
allocations happened. They are done in write_begin, which would always
assume that the filesystem can be entered from reclaim. This bug could
cause filesystem deadlocks.
The funny thing with having a gfp_t mask there is that it doesn't really
allow the caller to arbitrarily tinker with the context in which it can be
called. It couldn't ever be GFP_ATOMIC, for example, because it needs to
take the page lock. The only thing any callers care about is __GFP_FS
anyway, so turn that into a single flag.
Add a new flag for write_begin, AOP_FLAG_NOFS. Filesystems can now act on
this flag in their write_begin function. Change __grab_cache_page to
accept a nofs argument as well, to honour that flag (while we're there,
change the name to grab_cache_page_write_begin which is more instructive
and does away with random leading underscores).
This is really a more flexible way to go in the end anyway -- if a
filesystem happens to want any extra allocations aside from the pagecache
ones in ints write_begin function, it may now use GFP_KERNEL (rather than
GFP_NOFS) for common case allocations (eg. ocfs2_alloc_write_ctxt, for a
random example).
[kosaki.motohiro@jp.fujitsu.com: fix ubifs]
[kosaki.motohiro@jp.fujitsu.com: fix fuse]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: <stable@kernel.org> [2.6.28.x]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ Cleaned up the calling convention: just pass in the AOP flags
untouched to the grab_cache_page_write_begin() function. That
just simplifies everybody, and may even allow future expansion of the
logic. - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-04 21:00:53 +01:00
|
|
|
page = grab_cache_page_write_begin(mapping, index, flags);
|
2008-08-03 00:51:06 +02:00
|
|
|
if (!page) {
|
|
|
|
ext4_journal_stop(handle);
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
2008-07-12 01:27:31 +02:00
|
|
|
*pagep = page;
|
|
|
|
|
|
|
|
ret = block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
|
2009-05-14 06:54:29 +02:00
|
|
|
ext4_da_get_block_prep);
|
2008-07-12 01:27:31 +02:00
|
|
|
if (ret < 0) {
|
|
|
|
unlock_page(page);
|
|
|
|
ext4_journal_stop(handle);
|
|
|
|
page_cache_release(page);
|
2008-09-13 19:10:25 +02:00
|
|
|
/*
|
|
|
|
* block_write_begin may have instantiated a few blocks
|
|
|
|
* outside i_size. Trim these off again. Don't need
|
|
|
|
* i_size_read because we hold i_mutex.
|
|
|
|
*/
|
|
|
|
if (pos + len > inode->i_size)
|
2009-12-09 03:24:33 +01:00
|
|
|
ext4_truncate_failed_write(inode);
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
|
|
|
|
2008-07-14 23:52:37 +02:00
|
|
|
if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
|
|
|
|
goto retry;
|
2008-07-12 01:27:31 +02:00
|
|
|
out:
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
ext4: fix delalloc i_disksize early update issue
Ext4_da_write_end() used walk_page_buffers() with a callback function of
ext4_bh_unmapped_or_delay() to check if it extended the file size
without allocating any blocks (since in this case i_disksize needs to be
updated). However, this is didn't work proprely because the buffer head
has not been marked dirty yet --- this is done later in
block_commit_write() --- which caused ext4_bh_unmapped_or_delay() to
always return false.
In addition, walk_page_buffers() checks all of the buffer heads covering
the page, and the only buffer_head that should be checked is the one
covering the end of the write. Otherwise, given a 1k blocksize
filesystem and a 4k page size, the buffer head covering the first 1k
stripe of the file could be unmapped (because it was a sparse file), and
the second or third buffer_head covering that page could be mapped, and
using walk_page_buffers() would fail in this case since it would stop at
the first unmapped buffer_head and return true.
The core problem is that walk_page_buffers() was intended to do work in
a callback function, and a non-zero return value indicated a failure,
which termined the walk of the buffer heads covering the page. It was
not intended to be used with a boolean function, such as
ext4_bh_unmapped_or_delay().
Add addtional fix from Aneesh to protect i_disksize update rave with truncate.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* Check if we should update i_disksize
|
|
|
|
* when write to the end of file but not require block allocation
|
|
|
|
*/
|
|
|
|
static int ext4_da_should_update_i_disksize(struct page *page,
|
2009-06-14 23:45:34 +02:00
|
|
|
unsigned long offset)
|
ext4: fix delalloc i_disksize early update issue
Ext4_da_write_end() used walk_page_buffers() with a callback function of
ext4_bh_unmapped_or_delay() to check if it extended the file size
without allocating any blocks (since in this case i_disksize needs to be
updated). However, this is didn't work proprely because the buffer head
has not been marked dirty yet --- this is done later in
block_commit_write() --- which caused ext4_bh_unmapped_or_delay() to
always return false.
In addition, walk_page_buffers() checks all of the buffer heads covering
the page, and the only buffer_head that should be checked is the one
covering the end of the write. Otherwise, given a 1k blocksize
filesystem and a 4k page size, the buffer head covering the first 1k
stripe of the file could be unmapped (because it was a sparse file), and
the second or third buffer_head covering that page could be mapped, and
using walk_page_buffers() would fail in this case since it would stop at
the first unmapped buffer_head and return true.
The core problem is that walk_page_buffers() was intended to do work in
a callback function, and a non-zero return value indicated a failure,
which termined the walk of the buffer heads covering the page. It was
not intended to be used with a boolean function, such as
ext4_bh_unmapped_or_delay().
Add addtional fix from Aneesh to protect i_disksize update rave with truncate.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-07-12 01:27:31 +02:00
|
|
|
{
|
|
|
|
struct buffer_head *bh;
|
|
|
|
struct inode *inode = page->mapping->host;
|
|
|
|
unsigned int idx;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
bh = page_buffers(page);
|
|
|
|
idx = offset >> inode->i_blkbits;
|
|
|
|
|
2008-09-09 04:25:24 +02:00
|
|
|
for (i = 0; i < idx; i++)
|
ext4: fix delalloc i_disksize early update issue
Ext4_da_write_end() used walk_page_buffers() with a callback function of
ext4_bh_unmapped_or_delay() to check if it extended the file size
without allocating any blocks (since in this case i_disksize needs to be
updated). However, this is didn't work proprely because the buffer head
has not been marked dirty yet --- this is done later in
block_commit_write() --- which caused ext4_bh_unmapped_or_delay() to
always return false.
In addition, walk_page_buffers() checks all of the buffer heads covering
the page, and the only buffer_head that should be checked is the one
covering the end of the write. Otherwise, given a 1k blocksize
filesystem and a 4k page size, the buffer head covering the first 1k
stripe of the file could be unmapped (because it was a sparse file), and
the second or third buffer_head covering that page could be mapped, and
using walk_page_buffers() would fail in this case since it would stop at
the first unmapped buffer_head and return true.
The core problem is that walk_page_buffers() was intended to do work in
a callback function, and a non-zero return value indicated a failure,
which termined the walk of the buffer heads covering the page. It was
not intended to be used with a boolean function, such as
ext4_bh_unmapped_or_delay().
Add addtional fix from Aneesh to protect i_disksize update rave with truncate.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-07-12 01:27:31 +02:00
|
|
|
bh = bh->b_this_page;
|
|
|
|
|
2009-05-12 22:30:27 +02:00
|
|
|
if (!buffer_mapped(bh) || (buffer_delay(bh)) || buffer_unwritten(bh))
|
ext4: fix delalloc i_disksize early update issue
Ext4_da_write_end() used walk_page_buffers() with a callback function of
ext4_bh_unmapped_or_delay() to check if it extended the file size
without allocating any blocks (since in this case i_disksize needs to be
updated). However, this is didn't work proprely because the buffer head
has not been marked dirty yet --- this is done later in
block_commit_write() --- which caused ext4_bh_unmapped_or_delay() to
always return false.
In addition, walk_page_buffers() checks all of the buffer heads covering
the page, and the only buffer_head that should be checked is the one
covering the end of the write. Otherwise, given a 1k blocksize
filesystem and a 4k page size, the buffer head covering the first 1k
stripe of the file could be unmapped (because it was a sparse file), and
the second or third buffer_head covering that page could be mapped, and
using walk_page_buffers() would fail in this case since it would stop at
the first unmapped buffer_head and return true.
The core problem is that walk_page_buffers() was intended to do work in
a callback function, and a non-zero return value indicated a failure,
which termined the walk of the buffer heads covering the page. It was
not intended to be used with a boolean function, such as
ext4_bh_unmapped_or_delay().
Add addtional fix from Aneesh to protect i_disksize update rave with truncate.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-07-12 01:27:31 +02:00
|
|
|
return 0;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
static int ext4_da_write_end(struct file *file,
|
2009-06-14 23:45:34 +02:00
|
|
|
struct address_space *mapping,
|
|
|
|
loff_t pos, unsigned len, unsigned copied,
|
|
|
|
struct page *page, void *fsdata)
|
2008-07-12 01:27:31 +02:00
|
|
|
{
|
|
|
|
struct inode *inode = mapping->host;
|
|
|
|
int ret = 0, ret2;
|
|
|
|
handle_t *handle = ext4_journal_current_handle();
|
|
|
|
loff_t new_i_size;
|
ext4: fix delalloc i_disksize early update issue
Ext4_da_write_end() used walk_page_buffers() with a callback function of
ext4_bh_unmapped_or_delay() to check if it extended the file size
without allocating any blocks (since in this case i_disksize needs to be
updated). However, this is didn't work proprely because the buffer head
has not been marked dirty yet --- this is done later in
block_commit_write() --- which caused ext4_bh_unmapped_or_delay() to
always return false.
In addition, walk_page_buffers() checks all of the buffer heads covering
the page, and the only buffer_head that should be checked is the one
covering the end of the write. Otherwise, given a 1k blocksize
filesystem and a 4k page size, the buffer head covering the first 1k
stripe of the file could be unmapped (because it was a sparse file), and
the second or third buffer_head covering that page could be mapped, and
using walk_page_buffers() would fail in this case since it would stop at
the first unmapped buffer_head and return true.
The core problem is that walk_page_buffers() was intended to do work in
a callback function, and a non-zero return value indicated a failure,
which termined the walk of the buffer heads covering the page. It was
not intended to be used with a boolean function, such as
ext4_bh_unmapped_or_delay().
Add addtional fix from Aneesh to protect i_disksize update rave with truncate.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-07-12 01:27:31 +02:00
|
|
|
unsigned long start, end;
|
2008-10-09 05:13:30 +02:00
|
|
|
int write_mode = (int)(unsigned long)fsdata;
|
|
|
|
|
|
|
|
if (write_mode == FALL_BACK_TO_NONDELALLOC) {
|
|
|
|
if (ext4_should_order_data(inode)) {
|
|
|
|
return ext4_ordered_write_end(file, mapping, pos,
|
|
|
|
len, copied, page, fsdata);
|
|
|
|
} else if (ext4_should_writeback_data(inode)) {
|
|
|
|
return ext4_writeback_write_end(file, mapping, pos,
|
|
|
|
len, copied, page, fsdata);
|
|
|
|
} else {
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
}
|
ext4: fix delalloc i_disksize early update issue
Ext4_da_write_end() used walk_page_buffers() with a callback function of
ext4_bh_unmapped_or_delay() to check if it extended the file size
without allocating any blocks (since in this case i_disksize needs to be
updated). However, this is didn't work proprely because the buffer head
has not been marked dirty yet --- this is done later in
block_commit_write() --- which caused ext4_bh_unmapped_or_delay() to
always return false.
In addition, walk_page_buffers() checks all of the buffer heads covering
the page, and the only buffer_head that should be checked is the one
covering the end of the write. Otherwise, given a 1k blocksize
filesystem and a 4k page size, the buffer head covering the first 1k
stripe of the file could be unmapped (because it was a sparse file), and
the second or third buffer_head covering that page could be mapped, and
using walk_page_buffers() would fail in this case since it would stop at
the first unmapped buffer_head and return true.
The core problem is that walk_page_buffers() was intended to do work in
a callback function, and a non-zero return value indicated a failure,
which termined the walk of the buffer heads covering the page. It was
not intended to be used with a boolean function, such as
ext4_bh_unmapped_or_delay().
Add addtional fix from Aneesh to protect i_disksize update rave with truncate.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-07-12 01:27:31 +02:00
|
|
|
|
2009-06-17 17:48:11 +02:00
|
|
|
trace_ext4_da_write_end(inode, pos, len, copied);
|
ext4: fix delalloc i_disksize early update issue
Ext4_da_write_end() used walk_page_buffers() with a callback function of
ext4_bh_unmapped_or_delay() to check if it extended the file size
without allocating any blocks (since in this case i_disksize needs to be
updated). However, this is didn't work proprely because the buffer head
has not been marked dirty yet --- this is done later in
block_commit_write() --- which caused ext4_bh_unmapped_or_delay() to
always return false.
In addition, walk_page_buffers() checks all of the buffer heads covering
the page, and the only buffer_head that should be checked is the one
covering the end of the write. Otherwise, given a 1k blocksize
filesystem and a 4k page size, the buffer head covering the first 1k
stripe of the file could be unmapped (because it was a sparse file), and
the second or third buffer_head covering that page could be mapped, and
using walk_page_buffers() would fail in this case since it would stop at
the first unmapped buffer_head and return true.
The core problem is that walk_page_buffers() was intended to do work in
a callback function, and a non-zero return value indicated a failure,
which termined the walk of the buffer heads covering the page. It was
not intended to be used with a boolean function, such as
ext4_bh_unmapped_or_delay().
Add addtional fix from Aneesh to protect i_disksize update rave with truncate.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-07-12 01:27:31 +02:00
|
|
|
start = pos & (PAGE_CACHE_SIZE - 1);
|
2008-09-09 04:25:24 +02:00
|
|
|
end = start + copied - 1;
|
2008-07-12 01:27:31 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* generic_write_end() will run mark_inode_dirty() if i_size
|
|
|
|
* changes. So let's piggyback the i_disksize mark_inode_dirty
|
|
|
|
* into that.
|
|
|
|
*/
|
|
|
|
|
|
|
|
new_i_size = pos + copied;
|
ext4: fix delalloc i_disksize early update issue
Ext4_da_write_end() used walk_page_buffers() with a callback function of
ext4_bh_unmapped_or_delay() to check if it extended the file size
without allocating any blocks (since in this case i_disksize needs to be
updated). However, this is didn't work proprely because the buffer head
has not been marked dirty yet --- this is done later in
block_commit_write() --- which caused ext4_bh_unmapped_or_delay() to
always return false.
In addition, walk_page_buffers() checks all of the buffer heads covering
the page, and the only buffer_head that should be checked is the one
covering the end of the write. Otherwise, given a 1k blocksize
filesystem and a 4k page size, the buffer head covering the first 1k
stripe of the file could be unmapped (because it was a sparse file), and
the second or third buffer_head covering that page could be mapped, and
using walk_page_buffers() would fail in this case since it would stop at
the first unmapped buffer_head and return true.
The core problem is that walk_page_buffers() was intended to do work in
a callback function, and a non-zero return value indicated a failure,
which termined the walk of the buffer heads covering the page. It was
not intended to be used with a boolean function, such as
ext4_bh_unmapped_or_delay().
Add addtional fix from Aneesh to protect i_disksize update rave with truncate.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-07-12 01:27:31 +02:00
|
|
|
if (new_i_size > EXT4_I(inode)->i_disksize) {
|
|
|
|
if (ext4_da_should_update_i_disksize(page, end)) {
|
|
|
|
down_write(&EXT4_I(inode)->i_data_sem);
|
|
|
|
if (new_i_size > EXT4_I(inode)->i_disksize) {
|
|
|
|
/*
|
|
|
|
* Updating i_disksize when extending file
|
|
|
|
* without needing block allocation
|
|
|
|
*/
|
|
|
|
if (ext4_should_order_data(inode))
|
|
|
|
ret = ext4_jbd2_file_inode(handle,
|
|
|
|
inode);
|
2008-07-12 01:27:31 +02:00
|
|
|
|
ext4: fix delalloc i_disksize early update issue
Ext4_da_write_end() used walk_page_buffers() with a callback function of
ext4_bh_unmapped_or_delay() to check if it extended the file size
without allocating any blocks (since in this case i_disksize needs to be
updated). However, this is didn't work proprely because the buffer head
has not been marked dirty yet --- this is done later in
block_commit_write() --- which caused ext4_bh_unmapped_or_delay() to
always return false.
In addition, walk_page_buffers() checks all of the buffer heads covering
the page, and the only buffer_head that should be checked is the one
covering the end of the write. Otherwise, given a 1k blocksize
filesystem and a 4k page size, the buffer head covering the first 1k
stripe of the file could be unmapped (because it was a sparse file), and
the second or third buffer_head covering that page could be mapped, and
using walk_page_buffers() would fail in this case since it would stop at
the first unmapped buffer_head and return true.
The core problem is that walk_page_buffers() was intended to do work in
a callback function, and a non-zero return value indicated a failure,
which termined the walk of the buffer heads covering the page. It was
not intended to be used with a boolean function, such as
ext4_bh_unmapped_or_delay().
Add addtional fix from Aneesh to protect i_disksize update rave with truncate.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-07-12 01:27:31 +02:00
|
|
|
EXT4_I(inode)->i_disksize = new_i_size;
|
|
|
|
}
|
|
|
|
up_write(&EXT4_I(inode)->i_data_sem);
|
2008-09-13 19:06:18 +02:00
|
|
|
/* We need to mark inode dirty even if
|
|
|
|
* new_i_size is less that inode->i_size
|
|
|
|
* bu greater than i_disksize.(hint delalloc)
|
|
|
|
*/
|
|
|
|
ext4_mark_inode_dirty(handle, inode);
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
ext4: fix delalloc i_disksize early update issue
Ext4_da_write_end() used walk_page_buffers() with a callback function of
ext4_bh_unmapped_or_delay() to check if it extended the file size
without allocating any blocks (since in this case i_disksize needs to be
updated). However, this is didn't work proprely because the buffer head
has not been marked dirty yet --- this is done later in
block_commit_write() --- which caused ext4_bh_unmapped_or_delay() to
always return false.
In addition, walk_page_buffers() checks all of the buffer heads covering
the page, and the only buffer_head that should be checked is the one
covering the end of the write. Otherwise, given a 1k blocksize
filesystem and a 4k page size, the buffer head covering the first 1k
stripe of the file could be unmapped (because it was a sparse file), and
the second or third buffer_head covering that page could be mapped, and
using walk_page_buffers() would fail in this case since it would stop at
the first unmapped buffer_head and return true.
The core problem is that walk_page_buffers() was intended to do work in
a callback function, and a non-zero return value indicated a failure,
which termined the walk of the buffer heads covering the page. It was
not intended to be used with a boolean function, such as
ext4_bh_unmapped_or_delay().
Add addtional fix from Aneesh to protect i_disksize update rave with truncate.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-07-12 01:27:31 +02:00
|
|
|
}
|
2008-07-12 01:27:31 +02:00
|
|
|
ret2 = generic_write_end(file, mapping, pos, len, copied,
|
|
|
|
page, fsdata);
|
|
|
|
copied = ret2;
|
|
|
|
if (ret2 < 0)
|
|
|
|
ret = ret2;
|
|
|
|
ret2 = ext4_journal_stop(handle);
|
|
|
|
if (!ret)
|
|
|
|
ret = ret2;
|
|
|
|
|
|
|
|
return ret ? ret : copied;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ext4_da_invalidatepage(struct page *page, unsigned long offset)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Drop reserved blocks
|
|
|
|
*/
|
|
|
|
BUG_ON(!PageLocked(page));
|
|
|
|
if (!page_has_buffers(page))
|
|
|
|
goto out;
|
|
|
|
|
2008-07-14 23:52:37 +02:00
|
|
|
ext4_da_page_release_reservation(page, offset);
|
2008-07-12 01:27:31 +02:00
|
|
|
|
|
|
|
out:
|
|
|
|
ext4_invalidatepage(page, offset);
|
|
|
|
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2009-02-26 07:04:07 +01:00
|
|
|
/*
|
|
|
|
* Force all delayed allocation blocks to be allocated for a given inode.
|
|
|
|
*/
|
|
|
|
int ext4_alloc_da_blocks(struct inode *inode)
|
|
|
|
{
|
2009-09-17 01:30:40 +02:00
|
|
|
trace_ext4_alloc_da_blocks(inode);
|
|
|
|
|
2009-02-26 07:04:07 +01:00
|
|
|
if (!EXT4_I(inode)->i_reserved_data_blocks &&
|
|
|
|
!EXT4_I(inode)->i_reserved_meta_blocks)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We do something simple for now. The filemap_flush() will
|
|
|
|
* also start triggering a write of the data blocks, which is
|
|
|
|
* not strictly speaking necessary (and for users of
|
|
|
|
* laptop_mode, not even desirable). However, to do otherwise
|
|
|
|
* would require replicating code paths in:
|
2009-06-14 23:45:34 +02:00
|
|
|
*
|
2009-02-26 07:04:07 +01:00
|
|
|
* ext4_da_writepages() ->
|
|
|
|
* write_cache_pages() ---> (via passed in callback function)
|
|
|
|
* __mpage_da_writepage() -->
|
|
|
|
* mpage_add_bh_to_extent()
|
|
|
|
* mpage_da_map_blocks()
|
|
|
|
*
|
|
|
|
* The problem is that write_cache_pages(), located in
|
|
|
|
* mm/page-writeback.c, marks pages clean in preparation for
|
|
|
|
* doing I/O, which is not desirable if we're not planning on
|
|
|
|
* doing I/O at all.
|
|
|
|
*
|
|
|
|
* We could call write_cache_pages(), and then redirty all of
|
|
|
|
* the pages by calling redirty_page_for_writeback() but that
|
|
|
|
* would be ugly in the extreme. So instead we would need to
|
|
|
|
* replicate parts of the code in the above functions,
|
|
|
|
* simplifying them becuase we wouldn't actually intend to
|
|
|
|
* write out the pages, but rather only collect contiguous
|
|
|
|
* logical block extents, call the multi-block allocator, and
|
|
|
|
* then update the buffer heads with the block allocations.
|
2009-06-14 23:45:34 +02:00
|
|
|
*
|
2009-02-26 07:04:07 +01:00
|
|
|
* For now, though, we'll cheat by calling filemap_flush(),
|
|
|
|
* which will map the blocks, and start the I/O, but not
|
|
|
|
* actually wait for the I/O to complete.
|
|
|
|
*/
|
|
|
|
return filemap_flush(inode->i_mapping);
|
|
|
|
}
|
2008-07-12 01:27:31 +02:00
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* bmap() is special. It gets used by applications such as lilo and by
|
|
|
|
* the swapper to find the on-disk block of a specific piece of data.
|
|
|
|
*
|
|
|
|
* Naturally, this is dangerous if the block concerned is still in the
|
2006-10-11 10:20:53 +02:00
|
|
|
* journal. If somebody makes a swapfile on an ext4 data-journaling
|
2006-10-11 10:20:50 +02:00
|
|
|
* filesystem and enables swap, then they may get a nasty shock when the
|
|
|
|
* data getting swapped to that swapfile suddenly gets overwritten by
|
|
|
|
* the original zero's written out previously to the journal and
|
|
|
|
* awaiting writeback in the kernel's buffer cache.
|
|
|
|
*
|
|
|
|
* So, if we see any bmap calls here on a modified, data-journaled file,
|
|
|
|
* take extra steps to flush any blocks which might be in the cache.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static sector_t ext4_bmap(struct address_space *mapping, sector_t block)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
struct inode *inode = mapping->host;
|
|
|
|
journal_t *journal;
|
|
|
|
int err;
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
|
|
|
|
test_opt(inode->i_sb, DELALLOC)) {
|
|
|
|
/*
|
|
|
|
* With delalloc we want to sync the file
|
|
|
|
* so that we can make sure we allocate
|
|
|
|
* blocks for file
|
|
|
|
*/
|
|
|
|
filemap_write_and_wait(mapping);
|
|
|
|
}
|
|
|
|
|
2010-01-24 20:34:07 +01:00
|
|
|
if (EXT4_JOURNAL(inode) &&
|
|
|
|
ext4_test_inode_state(inode, EXT4_STATE_JDATA)) {
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* This is a REALLY heavyweight approach, but the use of
|
|
|
|
* bmap on dirty files is expected to be extremely rare:
|
|
|
|
* only if we run lilo or swapon on a freshly made file
|
|
|
|
* do we expect this to happen.
|
|
|
|
*
|
|
|
|
* (bmap requires CAP_SYS_RAWIO so this does not
|
|
|
|
* represent an unprivileged user DOS attack --- we'd be
|
|
|
|
* in trouble if mortal users could trigger this path at
|
|
|
|
* will.)
|
|
|
|
*
|
2006-10-11 10:20:53 +02:00
|
|
|
* NB. EXT4_STATE_JDATA is not set on files other than
|
2006-10-11 10:20:50 +02:00
|
|
|
* regular files. If somebody wants to bmap a directory
|
|
|
|
* or symlink and gets confused because the buffer
|
|
|
|
* hasn't yet been flushed to disk, they deserve
|
|
|
|
* everything they get.
|
|
|
|
*/
|
|
|
|
|
2010-01-24 20:34:07 +01:00
|
|
|
ext4_clear_inode_state(inode, EXT4_STATE_JDATA);
|
2006-10-11 10:20:53 +02:00
|
|
|
journal = EXT4_JOURNAL(inode);
|
2006-10-11 10:21:01 +02:00
|
|
|
jbd2_journal_lock_updates(journal);
|
|
|
|
err = jbd2_journal_flush(journal);
|
|
|
|
jbd2_journal_unlock_updates(journal);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
if (err)
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-09-09 04:25:24 +02:00
|
|
|
return generic_block_bmap(mapping, block, ext4_get_block);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
static int ext4_readpage(struct file *file, struct page *page)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
return mpage_readpage(page, ext4_get_block);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_readpages(struct file *file, struct address_space *mapping,
|
2006-10-11 10:20:50 +02:00
|
|
|
struct list_head *pages, unsigned nr_pages)
|
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
return mpage_readpages(mapping, pages, nr_pages, ext4_get_block);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2010-03-04 22:14:02 +01:00
|
|
|
static void ext4_free_io_end(ext4_io_end_t *io)
|
|
|
|
{
|
|
|
|
BUG_ON(!io);
|
|
|
|
if (io->page)
|
|
|
|
put_page(io->page);
|
|
|
|
iput(io->inode);
|
|
|
|
kfree(io);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ext4_invalidatepage_free_endio(struct page *page, unsigned long offset)
|
|
|
|
{
|
|
|
|
struct buffer_head *head, *bh;
|
|
|
|
unsigned int curr_off = 0;
|
|
|
|
|
|
|
|
if (!page_has_buffers(page))
|
|
|
|
return;
|
|
|
|
head = bh = page_buffers(page);
|
|
|
|
do {
|
|
|
|
if (offset <= curr_off && test_clear_buffer_uninit(bh)
|
|
|
|
&& bh->b_private) {
|
|
|
|
ext4_free_io_end(bh->b_private);
|
|
|
|
bh->b_private = NULL;
|
|
|
|
bh->b_end_io = NULL;
|
|
|
|
}
|
|
|
|
curr_off = curr_off + bh->b_size;
|
|
|
|
bh = bh->b_this_page;
|
|
|
|
} while (bh != head);
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
static void ext4_invalidatepage(struct page *page, unsigned long offset)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
journal_t *journal = EXT4_JOURNAL(page->mapping->host);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2010-03-04 22:14:02 +01:00
|
|
|
/*
|
|
|
|
* free any io_end structure allocated for buffers to be discarded
|
|
|
|
*/
|
|
|
|
if (ext4_should_dioread_nolock(page->mapping->host))
|
|
|
|
ext4_invalidatepage_free_endio(page, offset);
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* If it's a full truncate we just forget about the pending dirtying
|
|
|
|
*/
|
|
|
|
if (offset == 0)
|
|
|
|
ClearPageChecked(page);
|
|
|
|
|
2009-01-07 06:06:22 +01:00
|
|
|
if (journal)
|
|
|
|
jbd2_journal_invalidatepage(journal, page, offset);
|
|
|
|
else
|
|
|
|
block_invalidatepage(page, offset);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
static int ext4_releasepage(struct page *page, gfp_t wait)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
journal_t *journal = EXT4_JOURNAL(page->mapping->host);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
WARN_ON(PageChecked(page));
|
|
|
|
if (!page_has_buffers(page))
|
|
|
|
return 0;
|
2009-01-07 06:06:22 +01:00
|
|
|
if (journal)
|
|
|
|
return jbd2_journal_try_to_free_buffers(journal, page, wait);
|
|
|
|
else
|
|
|
|
return try_to_free_buffers(page);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2009-09-28 21:48:41 +02:00
|
|
|
* O_DIRECT for ext3 (or indirect map) based files
|
|
|
|
*
|
2006-10-11 10:20:50 +02:00
|
|
|
* If the O_DIRECT write will extend the file then add this inode to the
|
|
|
|
* orphan list. So recovery will truncate it back to the original size
|
|
|
|
* if the machine crashes during the write.
|
|
|
|
*
|
|
|
|
* If the O_DIRECT write is intantiating holes inside i_size and the machine
|
2008-02-10 07:08:38 +01:00
|
|
|
* crashes then stale disk data _may_ be exposed inside the file. But current
|
|
|
|
* VFS code falls back into buffered path in that case so we are safe.
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
2009-09-28 21:48:41 +02:00
|
|
|
static ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
|
2009-06-14 23:45:34 +02:00
|
|
|
const struct iovec *iov, loff_t offset,
|
|
|
|
unsigned long nr_segs)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
struct file *file = iocb->ki_filp;
|
|
|
|
struct inode *inode = file->f_mapping->host;
|
2006-10-11 10:20:53 +02:00
|
|
|
struct ext4_inode_info *ei = EXT4_I(inode);
|
2008-02-10 07:08:38 +01:00
|
|
|
handle_t *handle;
|
2006-10-11 10:20:50 +02:00
|
|
|
ssize_t ret;
|
|
|
|
int orphan = 0;
|
|
|
|
size_t count = iov_length(iov, nr_segs);
|
2009-10-03 03:20:55 +02:00
|
|
|
int retries = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
if (rw == WRITE) {
|
|
|
|
loff_t final_size = offset + count;
|
|
|
|
|
|
|
|
if (final_size > inode->i_size) {
|
2008-02-10 07:08:38 +01:00
|
|
|
/* Credits for sb + inode write */
|
|
|
|
handle = ext4_journal_start(inode, 2);
|
|
|
|
if (IS_ERR(handle)) {
|
|
|
|
ret = PTR_ERR(handle);
|
|
|
|
goto out;
|
|
|
|
}
|
2006-10-11 10:20:53 +02:00
|
|
|
ret = ext4_orphan_add(handle, inode);
|
2008-02-10 07:08:38 +01:00
|
|
|
if (ret) {
|
|
|
|
ext4_journal_stop(handle);
|
|
|
|
goto out;
|
|
|
|
}
|
2006-10-11 10:20:50 +02:00
|
|
|
orphan = 1;
|
|
|
|
ei->i_disksize = inode->i_size;
|
2008-02-10 07:08:38 +01:00
|
|
|
ext4_journal_stop(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-10-03 03:20:55 +02:00
|
|
|
retry:
|
2010-03-02 19:26:36 +01:00
|
|
|
if (rw == READ && ext4_should_dioread_nolock(inode))
|
|
|
|
ret = blockdev_direct_IO_no_locking(rw, iocb, inode,
|
|
|
|
inode->i_sb->s_bdev, iov,
|
|
|
|
offset, nr_segs,
|
|
|
|
ext4_get_block, NULL);
|
|
|
|
else
|
|
|
|
ret = blockdev_direct_IO(rw, iocb, inode,
|
|
|
|
inode->i_sb->s_bdev, iov,
|
2006-10-11 10:20:50 +02:00
|
|
|
offset, nr_segs,
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_get_block, NULL);
|
2009-10-03 03:20:55 +02:00
|
|
|
if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
|
|
|
|
goto retry;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2008-02-10 07:08:38 +01:00
|
|
|
if (orphan) {
|
2006-10-11 10:20:50 +02:00
|
|
|
int err;
|
|
|
|
|
2008-02-10 07:08:38 +01:00
|
|
|
/* Credits for sb + inode write */
|
|
|
|
handle = ext4_journal_start(inode, 2);
|
|
|
|
if (IS_ERR(handle)) {
|
|
|
|
/* This is really bad luck. We've written the data
|
|
|
|
* but cannot extend i_size. Bail out and pretend
|
|
|
|
* the write failed... */
|
|
|
|
ret = PTR_ERR(handle);
|
2010-03-02 05:15:02 +01:00
|
|
|
if (inode->i_nlink)
|
|
|
|
ext4_orphan_del(NULL, inode);
|
|
|
|
|
2008-02-10 07:08:38 +01:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
if (inode->i_nlink)
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_orphan_del(handle, inode);
|
2008-02-10 07:08:38 +01:00
|
|
|
if (ret > 0) {
|
2006-10-11 10:20:50 +02:00
|
|
|
loff_t end = offset + ret;
|
|
|
|
if (end > inode->i_size) {
|
|
|
|
ei->i_disksize = end;
|
|
|
|
i_size_write(inode, end);
|
|
|
|
/*
|
|
|
|
* We're going to return a positive `ret'
|
|
|
|
* here due to non-zero-length I/O, so there's
|
|
|
|
* no way of reporting error returns from
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_mark_inode_dirty() to userspace. So
|
2006-10-11 10:20:50 +02:00
|
|
|
* ignore it.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_mark_inode_dirty(handle, inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
}
|
2006-10-11 10:20:53 +02:00
|
|
|
err = ext4_journal_stop(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (ret == 0)
|
|
|
|
ret = err;
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2010-05-17 02:00:00 +02:00
|
|
|
/*
|
|
|
|
* ext4_get_block used when preparing for a DIO write or buffer write.
|
|
|
|
* We allocate an uinitialized extent if blocks haven't been allocated.
|
|
|
|
* The extent will be converted to initialized after the IO is complete.
|
|
|
|
*/
|
2010-03-02 19:28:44 +01:00
|
|
|
static int ext4_get_block_write(struct inode *inode, sector_t iblock,
|
2009-09-28 21:48:41 +02:00
|
|
|
struct buffer_head *bh_result, int create)
|
|
|
|
{
|
2010-03-02 19:28:44 +01:00
|
|
|
ext4_debug("ext4_get_block_write: inode %lu, create flag %d\n",
|
2009-09-28 21:48:29 +02:00
|
|
|
inode->i_ino, create);
|
2010-05-17 02:00:00 +02:00
|
|
|
return _ext4_get_block(inode, iblock, bh_result,
|
|
|
|
EXT4_GET_BLOCKS_IO_CREATE_EXT);
|
2009-09-28 21:48:41 +02:00
|
|
|
}
|
|
|
|
|
2010-03-02 19:28:44 +01:00
|
|
|
static void dump_completed_IO(struct inode * inode)
|
2009-09-28 21:48:29 +02:00
|
|
|
{
|
|
|
|
#ifdef EXT4_DEBUG
|
|
|
|
struct list_head *cur, *before, *after;
|
|
|
|
ext4_io_end_t *io, *io0, *io1;
|
2010-03-04 22:14:02 +01:00
|
|
|
unsigned long flags;
|
2009-09-28 21:48:29 +02:00
|
|
|
|
2010-03-02 19:28:44 +01:00
|
|
|
if (list_empty(&EXT4_I(inode)->i_completed_io_list)){
|
|
|
|
ext4_debug("inode %lu completed_io list is empty\n", inode->i_ino);
|
2009-09-28 21:48:29 +02:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2010-03-02 19:28:44 +01:00
|
|
|
ext4_debug("Dump inode %lu completed_io list \n", inode->i_ino);
|
2010-03-04 22:14:02 +01:00
|
|
|
spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
|
2010-03-02 19:28:44 +01:00
|
|
|
list_for_each_entry(io, &EXT4_I(inode)->i_completed_io_list, list){
|
2009-09-28 21:48:29 +02:00
|
|
|
cur = &io->list;
|
|
|
|
before = cur->prev;
|
|
|
|
io0 = container_of(before, ext4_io_end_t, list);
|
|
|
|
after = cur->next;
|
|
|
|
io1 = container_of(after, ext4_io_end_t, list);
|
|
|
|
|
|
|
|
ext4_debug("io 0x%p from inode %lu,prev 0x%p,next 0x%p\n",
|
|
|
|
io, inode->i_ino, io0, io1);
|
|
|
|
}
|
2010-03-04 22:14:02 +01:00
|
|
|
spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
|
2009-09-28 21:48:29 +02:00
|
|
|
#endif
|
|
|
|
}
|
2009-09-28 21:48:41 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* check a range of space and convert unwritten extents to written.
|
|
|
|
*/
|
2010-03-02 19:28:44 +01:00
|
|
|
static int ext4_end_io_nolock(ext4_io_end_t *io)
|
2009-09-28 21:48:41 +02:00
|
|
|
{
|
|
|
|
struct inode *inode = io->inode;
|
|
|
|
loff_t offset = io->offset;
|
2010-02-05 05:58:38 +01:00
|
|
|
ssize_t size = io->size;
|
2009-09-28 21:48:41 +02:00
|
|
|
int ret = 0;
|
|
|
|
|
2010-03-02 19:28:44 +01:00
|
|
|
ext4_debug("ext4_end_io_nolock: io 0x%p from inode %lu,list->next 0x%p,"
|
2009-09-28 21:48:29 +02:00
|
|
|
"list->prev 0x%p\n",
|
|
|
|
io, inode->i_ino, io->list.next, io->list.prev);
|
|
|
|
|
|
|
|
if (list_empty(&io->list))
|
|
|
|
return ret;
|
|
|
|
|
2010-03-02 19:28:44 +01:00
|
|
|
if (io->flag != EXT4_IO_UNWRITTEN)
|
2009-09-28 21:48:29 +02:00
|
|
|
return ret;
|
|
|
|
|
2010-03-04 22:14:02 +01:00
|
|
|
ret = ext4_convert_unwritten_extents(inode, offset, size);
|
2009-09-28 21:48:29 +02:00
|
|
|
if (ret < 0) {
|
2009-09-28 21:48:41 +02:00
|
|
|
printk(KERN_EMERG "%s: failed to convert unwritten"
|
2009-09-28 21:48:29 +02:00
|
|
|
"extents to written extents, error is %d"
|
|
|
|
" io is still on inode %lu aio dio list\n",
|
|
|
|
__func__, ret, inode->i_ino);
|
|
|
|
return ret;
|
|
|
|
}
|
2009-09-28 21:48:41 +02:00
|
|
|
|
2009-09-28 21:48:29 +02:00
|
|
|
/* clear the DIO AIO unwritten flag */
|
|
|
|
io->flag = 0;
|
|
|
|
return ret;
|
2009-09-28 21:48:41 +02:00
|
|
|
}
|
2010-03-02 19:28:44 +01:00
|
|
|
|
2009-09-28 21:48:29 +02:00
|
|
|
/*
|
|
|
|
* work on completed aio dio IO, to convert unwritten extents to extents
|
|
|
|
*/
|
2010-03-02 19:28:44 +01:00
|
|
|
static void ext4_end_io_work(struct work_struct *work)
|
2009-09-28 21:48:29 +02:00
|
|
|
{
|
2010-03-04 22:14:02 +01:00
|
|
|
ext4_io_end_t *io = container_of(work, ext4_io_end_t, work);
|
|
|
|
struct inode *inode = io->inode;
|
|
|
|
struct ext4_inode_info *ei = EXT4_I(inode);
|
|
|
|
unsigned long flags;
|
|
|
|
int ret;
|
2009-09-28 21:48:41 +02:00
|
|
|
|
2009-09-28 21:48:29 +02:00
|
|
|
mutex_lock(&inode->i_mutex);
|
2010-03-02 19:28:44 +01:00
|
|
|
ret = ext4_end_io_nolock(io);
|
2010-03-04 22:14:02 +01:00
|
|
|
if (ret < 0) {
|
|
|
|
mutex_unlock(&inode->i_mutex);
|
|
|
|
return;
|
2009-09-28 21:48:29 +02:00
|
|
|
}
|
2010-03-04 22:14:02 +01:00
|
|
|
|
|
|
|
spin_lock_irqsave(&ei->i_completed_io_lock, flags);
|
|
|
|
if (!list_empty(&io->list))
|
|
|
|
list_del_init(&io->list);
|
|
|
|
spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
|
2009-09-28 21:48:29 +02:00
|
|
|
mutex_unlock(&inode->i_mutex);
|
2010-03-04 22:14:02 +01:00
|
|
|
ext4_free_io_end(io);
|
2009-09-28 21:48:29 +02:00
|
|
|
}
|
2010-03-02 19:28:44 +01:00
|
|
|
|
2009-09-28 21:48:29 +02:00
|
|
|
/*
|
|
|
|
* This function is called from ext4_sync_file().
|
|
|
|
*
|
2010-03-02 19:28:44 +01:00
|
|
|
* When IO is completed, the work to convert unwritten extents to
|
|
|
|
* written is queued on workqueue but may not get immediately
|
2009-09-28 21:48:29 +02:00
|
|
|
* scheduled. When fsync is called, we need to ensure the
|
|
|
|
* conversion is complete before fsync returns.
|
2010-03-02 19:28:44 +01:00
|
|
|
* The inode keeps track of a list of pending/completed IO that
|
|
|
|
* might needs to do the conversion. This function walks through
|
|
|
|
* the list and convert the related unwritten extents for completed IO
|
|
|
|
* to written.
|
|
|
|
* The function return the number of pending IOs on success.
|
2009-09-28 21:48:29 +02:00
|
|
|
*/
|
2010-03-02 19:28:44 +01:00
|
|
|
int flush_completed_IO(struct inode *inode)
|
2009-09-28 21:48:29 +02:00
|
|
|
{
|
|
|
|
ext4_io_end_t *io;
|
2010-03-04 22:14:02 +01:00
|
|
|
struct ext4_inode_info *ei = EXT4_I(inode);
|
|
|
|
unsigned long flags;
|
2009-09-28 21:48:29 +02:00
|
|
|
int ret = 0;
|
|
|
|
int ret2 = 0;
|
|
|
|
|
2010-03-04 22:14:02 +01:00
|
|
|
if (list_empty(&ei->i_completed_io_list))
|
2009-09-28 21:48:29 +02:00
|
|
|
return ret;
|
|
|
|
|
2010-03-02 19:28:44 +01:00
|
|
|
dump_completed_IO(inode);
|
2010-03-04 22:14:02 +01:00
|
|
|
spin_lock_irqsave(&ei->i_completed_io_lock, flags);
|
|
|
|
while (!list_empty(&ei->i_completed_io_list)){
|
|
|
|
io = list_entry(ei->i_completed_io_list.next,
|
2009-09-28 21:48:29 +02:00
|
|
|
ext4_io_end_t, list);
|
|
|
|
/*
|
2010-03-02 19:28:44 +01:00
|
|
|
* Calling ext4_end_io_nolock() to convert completed
|
2009-09-28 21:48:29 +02:00
|
|
|
* IO to written.
|
|
|
|
*
|
|
|
|
* When ext4_sync_file() is called, run_queue() may already
|
|
|
|
* about to flush the work corresponding to this io structure.
|
|
|
|
* It will be upset if it founds the io structure related
|
|
|
|
* to the work-to-be schedule is freed.
|
|
|
|
*
|
|
|
|
* Thus we need to keep the io structure still valid here after
|
|
|
|
* convertion finished. The io structure has a flag to
|
|
|
|
* avoid double converting from both fsync and background work
|
|
|
|
* queue work.
|
|
|
|
*/
|
2010-03-04 22:14:02 +01:00
|
|
|
spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
|
2010-03-02 19:28:44 +01:00
|
|
|
ret = ext4_end_io_nolock(io);
|
2010-03-04 22:14:02 +01:00
|
|
|
spin_lock_irqsave(&ei->i_completed_io_lock, flags);
|
2009-09-28 21:48:29 +02:00
|
|
|
if (ret < 0)
|
|
|
|
ret2 = ret;
|
|
|
|
else
|
|
|
|
list_del_init(&io->list);
|
|
|
|
}
|
2010-03-04 22:14:02 +01:00
|
|
|
spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
|
2009-09-28 21:48:29 +02:00
|
|
|
return (ret2 < 0) ? ret2 : 0;
|
|
|
|
}
|
|
|
|
|
2010-03-04 22:14:02 +01:00
|
|
|
static ext4_io_end_t *ext4_init_io_end (struct inode *inode, gfp_t flags)
|
2009-09-28 21:48:41 +02:00
|
|
|
{
|
|
|
|
ext4_io_end_t *io = NULL;
|
|
|
|
|
2010-03-04 22:14:02 +01:00
|
|
|
io = kmalloc(sizeof(*io), flags);
|
2009-09-28 21:48:41 +02:00
|
|
|
|
|
|
|
if (io) {
|
2009-09-28 21:48:29 +02:00
|
|
|
igrab(inode);
|
2009-09-28 21:48:41 +02:00
|
|
|
io->inode = inode;
|
2009-09-28 21:48:29 +02:00
|
|
|
io->flag = 0;
|
2009-09-28 21:48:41 +02:00
|
|
|
io->offset = 0;
|
|
|
|
io->size = 0;
|
2010-03-04 22:14:02 +01:00
|
|
|
io->page = NULL;
|
2010-03-02 19:28:44 +01:00
|
|
|
INIT_WORK(&io->work, ext4_end_io_work);
|
2009-09-28 21:48:29 +02:00
|
|
|
INIT_LIST_HEAD(&io->list);
|
2009-09-28 21:48:41 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
return io;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset,
|
|
|
|
ssize_t size, void *private)
|
|
|
|
{
|
|
|
|
ext4_io_end_t *io_end = iocb->private;
|
|
|
|
struct workqueue_struct *wq;
|
2010-03-04 22:14:02 +01:00
|
|
|
unsigned long flags;
|
|
|
|
struct ext4_inode_info *ei;
|
2009-09-28 21:48:41 +02:00
|
|
|
|
2009-11-03 20:44:54 +01:00
|
|
|
/* if not async direct IO or dio with 0 bytes write, just return */
|
|
|
|
if (!io_end || !size)
|
|
|
|
return;
|
|
|
|
|
2009-09-28 21:48:29 +02:00
|
|
|
ext_debug("ext4_end_io_dio(): io_end 0x%p"
|
|
|
|
"for inode %lu, iocb 0x%p, offset %llu, size %llu\n",
|
|
|
|
iocb->private, io_end->inode->i_ino, iocb, offset,
|
|
|
|
size);
|
|
|
|
|
|
|
|
/* if not aio dio with unwritten extents, just free io and return */
|
2010-03-02 19:28:44 +01:00
|
|
|
if (io_end->flag != EXT4_IO_UNWRITTEN){
|
2009-09-28 21:48:29 +02:00
|
|
|
ext4_free_io_end(io_end);
|
|
|
|
iocb->private = NULL;
|
2009-09-28 21:48:41 +02:00
|
|
|
return;
|
2009-09-28 21:48:29 +02:00
|
|
|
}
|
|
|
|
|
2009-09-28 21:48:41 +02:00
|
|
|
io_end->offset = offset;
|
|
|
|
io_end->size = size;
|
2010-03-04 22:14:02 +01:00
|
|
|
io_end->flag = EXT4_IO_UNWRITTEN;
|
2009-09-28 21:48:41 +02:00
|
|
|
wq = EXT4_SB(io_end->inode->i_sb)->dio_unwritten_wq;
|
|
|
|
|
2009-09-28 21:48:29 +02:00
|
|
|
/* queue the work to convert unwritten extents to written */
|
2009-09-28 21:48:41 +02:00
|
|
|
queue_work(wq, &io_end->work);
|
|
|
|
|
2009-09-28 21:48:29 +02:00
|
|
|
/* Add the io_end to per-inode completed aio dio list*/
|
2010-03-04 22:14:02 +01:00
|
|
|
ei = EXT4_I(io_end->inode);
|
|
|
|
spin_lock_irqsave(&ei->i_completed_io_lock, flags);
|
|
|
|
list_add_tail(&io_end->list, &ei->i_completed_io_list);
|
|
|
|
spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
|
2009-09-28 21:48:41 +02:00
|
|
|
iocb->private = NULL;
|
|
|
|
}
|
2010-03-02 19:28:44 +01:00
|
|
|
|
2010-03-04 22:14:02 +01:00
|
|
|
static void ext4_end_io_buffer_write(struct buffer_head *bh, int uptodate)
|
|
|
|
{
|
|
|
|
ext4_io_end_t *io_end = bh->b_private;
|
|
|
|
struct workqueue_struct *wq;
|
|
|
|
struct inode *inode;
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
if (!test_clear_buffer_uninit(bh) || !io_end)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (!(io_end->inode->i_sb->s_flags & MS_ACTIVE)) {
|
|
|
|
printk("sb umounted, discard end_io request for inode %lu\n",
|
|
|
|
io_end->inode->i_ino);
|
|
|
|
ext4_free_io_end(io_end);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
io_end->flag = EXT4_IO_UNWRITTEN;
|
|
|
|
inode = io_end->inode;
|
|
|
|
|
|
|
|
/* Add the io_end to per-inode completed io list*/
|
|
|
|
spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
|
|
|
|
list_add_tail(&io_end->list, &EXT4_I(inode)->i_completed_io_list);
|
|
|
|
spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
|
|
|
|
|
|
|
|
wq = EXT4_SB(inode->i_sb)->dio_unwritten_wq;
|
|
|
|
/* queue the work to convert unwritten extents to written */
|
|
|
|
queue_work(wq, &io_end->work);
|
|
|
|
out:
|
|
|
|
bh->b_private = NULL;
|
|
|
|
bh->b_end_io = NULL;
|
|
|
|
clear_buffer_uninit(bh);
|
|
|
|
end_buffer_async_write(bh, uptodate);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int ext4_set_bh_endio(struct buffer_head *bh, struct inode *inode)
|
|
|
|
{
|
|
|
|
ext4_io_end_t *io_end;
|
|
|
|
struct page *page = bh->b_page;
|
|
|
|
loff_t offset = (sector_t)page->index << PAGE_CACHE_SHIFT;
|
|
|
|
size_t size = bh->b_size;
|
|
|
|
|
|
|
|
retry:
|
|
|
|
io_end = ext4_init_io_end(inode, GFP_ATOMIC);
|
|
|
|
if (!io_end) {
|
|
|
|
if (printk_ratelimit())
|
|
|
|
printk(KERN_WARNING "%s: allocation fail\n", __func__);
|
|
|
|
schedule();
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
io_end->offset = offset;
|
|
|
|
io_end->size = size;
|
|
|
|
/*
|
|
|
|
* We need to hold a reference to the page to make sure it
|
|
|
|
* doesn't get evicted before ext4_end_io_work() has a chance
|
|
|
|
* to convert the extent from written to unwritten.
|
|
|
|
*/
|
|
|
|
io_end->page = page;
|
|
|
|
get_page(io_end->page);
|
|
|
|
|
|
|
|
bh->b_private = io_end;
|
|
|
|
bh->b_end_io = ext4_end_io_buffer_write;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-09-28 21:48:41 +02:00
|
|
|
/*
|
|
|
|
* For ext4 extent files, ext4 will do direct-io write to holes,
|
|
|
|
* preallocated extents, and those write extend the file, no need to
|
|
|
|
* fall back to buffered IO.
|
|
|
|
*
|
|
|
|
* For holes, we fallocate those blocks, mark them as unintialized
|
|
|
|
* If those blocks were preallocated, we mark sure they are splited, but
|
|
|
|
* still keep the range to write as unintialized.
|
|
|
|
*
|
2009-09-28 21:48:29 +02:00
|
|
|
* The unwrritten extents will be converted to written when DIO is completed.
|
|
|
|
* For async direct IO, since the IO may still pending when return, we
|
|
|
|
* set up an end_io call back function, which will do the convertion
|
|
|
|
* when async direct IO completed.
|
2009-09-28 21:48:41 +02:00
|
|
|
*
|
|
|
|
* If the O_DIRECT write will extend the file then add this inode to the
|
|
|
|
* orphan list. So recovery will truncate it back to the original size
|
|
|
|
* if the machine crashes during the write.
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
static ssize_t ext4_ext_direct_IO(int rw, struct kiocb *iocb,
|
|
|
|
const struct iovec *iov, loff_t offset,
|
|
|
|
unsigned long nr_segs)
|
|
|
|
{
|
|
|
|
struct file *file = iocb->ki_filp;
|
|
|
|
struct inode *inode = file->f_mapping->host;
|
|
|
|
ssize_t ret;
|
|
|
|
size_t count = iov_length(iov, nr_segs);
|
|
|
|
|
|
|
|
loff_t final_size = offset + count;
|
|
|
|
if (rw == WRITE && final_size <= inode->i_size) {
|
|
|
|
/*
|
2009-09-28 21:48:29 +02:00
|
|
|
* We could direct write to holes and fallocate.
|
|
|
|
*
|
|
|
|
* Allocated blocks to fill the hole are marked as uninitialized
|
2009-09-28 21:48:41 +02:00
|
|
|
* to prevent paralel buffered read to expose the stale data
|
|
|
|
* before DIO complete the data IO.
|
2009-09-28 21:48:29 +02:00
|
|
|
*
|
|
|
|
* As to previously fallocated extents, ext4 get_block
|
2009-09-28 21:48:41 +02:00
|
|
|
* will just simply mark the buffer mapped but still
|
|
|
|
* keep the extents uninitialized.
|
|
|
|
*
|
2009-09-28 21:48:29 +02:00
|
|
|
* for non AIO case, we will convert those unwritten extents
|
|
|
|
* to written after return back from blockdev_direct_IO.
|
|
|
|
*
|
|
|
|
* for async DIO, the conversion needs to be defered when
|
|
|
|
* the IO is completed. The ext4 end_io callback function
|
|
|
|
* will be called to take care of the conversion work.
|
|
|
|
* Here for async case, we allocate an io_end structure to
|
|
|
|
* hook to the iocb.
|
2009-09-28 21:48:41 +02:00
|
|
|
*/
|
2009-09-28 21:48:29 +02:00
|
|
|
iocb->private = NULL;
|
|
|
|
EXT4_I(inode)->cur_aio_dio = NULL;
|
|
|
|
if (!is_sync_kiocb(iocb)) {
|
2010-03-04 22:14:02 +01:00
|
|
|
iocb->private = ext4_init_io_end(inode, GFP_NOFS);
|
2009-09-28 21:48:29 +02:00
|
|
|
if (!iocb->private)
|
|
|
|
return -ENOMEM;
|
|
|
|
/*
|
|
|
|
* we save the io structure for current async
|
|
|
|
* direct IO, so that later ext4_get_blocks()
|
|
|
|
* could flag the io structure whether there
|
|
|
|
* is a unwritten extents needs to be converted
|
|
|
|
* when IO is completed.
|
|
|
|
*/
|
|
|
|
EXT4_I(inode)->cur_aio_dio = iocb->private;
|
|
|
|
}
|
|
|
|
|
2009-09-28 21:48:41 +02:00
|
|
|
ret = blockdev_direct_IO(rw, iocb, inode,
|
|
|
|
inode->i_sb->s_bdev, iov,
|
|
|
|
offset, nr_segs,
|
2010-03-02 19:28:44 +01:00
|
|
|
ext4_get_block_write,
|
2009-09-28 21:48:41 +02:00
|
|
|
ext4_end_io_dio);
|
2009-09-28 21:48:29 +02:00
|
|
|
if (iocb->private)
|
|
|
|
EXT4_I(inode)->cur_aio_dio = NULL;
|
|
|
|
/*
|
|
|
|
* The io_end structure takes a reference to the inode,
|
|
|
|
* that structure needs to be destroyed and the
|
|
|
|
* reference to the inode need to be dropped, when IO is
|
|
|
|
* complete, even with 0 byte write, or failed.
|
|
|
|
*
|
|
|
|
* In the successful AIO DIO case, the io_end structure will be
|
|
|
|
* desctroyed and the reference to the inode will be dropped
|
|
|
|
* after the end_io call back function is called.
|
|
|
|
*
|
|
|
|
* In the case there is 0 byte write, or error case, since
|
|
|
|
* VFS direct IO won't invoke the end_io call back function,
|
|
|
|
* we need to free the end_io structure here.
|
|
|
|
*/
|
|
|
|
if (ret != -EIOCBQUEUED && ret <= 0 && iocb->private) {
|
|
|
|
ext4_free_io_end(iocb->private);
|
|
|
|
iocb->private = NULL;
|
2010-01-24 20:34:07 +01:00
|
|
|
} else if (ret > 0 && ext4_test_inode_state(inode,
|
|
|
|
EXT4_STATE_DIO_UNWRITTEN)) {
|
2009-11-10 16:48:08 +01:00
|
|
|
int err;
|
2009-09-28 21:48:29 +02:00
|
|
|
/*
|
|
|
|
* for non AIO case, since the IO is already
|
|
|
|
* completed, we could do the convertion right here
|
|
|
|
*/
|
2009-11-10 16:48:08 +01:00
|
|
|
err = ext4_convert_unwritten_extents(inode,
|
|
|
|
offset, ret);
|
|
|
|
if (err < 0)
|
|
|
|
ret = err;
|
2010-01-24 20:34:07 +01:00
|
|
|
ext4_clear_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
|
2009-11-10 16:48:08 +01:00
|
|
|
}
|
2009-09-28 21:48:41 +02:00
|
|
|
return ret;
|
|
|
|
}
|
2009-09-28 21:48:29 +02:00
|
|
|
|
|
|
|
/* for write the the end of file case, we fall back to old way */
|
2009-09-28 21:48:41 +02:00
|
|
|
return ext4_ind_direct_IO(rw, iocb, iov, offset, nr_segs);
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t ext4_direct_IO(int rw, struct kiocb *iocb,
|
|
|
|
const struct iovec *iov, loff_t offset,
|
|
|
|
unsigned long nr_segs)
|
|
|
|
{
|
|
|
|
struct file *file = iocb->ki_filp;
|
|
|
|
struct inode *inode = file->f_mapping->host;
|
|
|
|
|
2010-05-17 04:00:00 +02:00
|
|
|
if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
|
2009-09-28 21:48:41 +02:00
|
|
|
return ext4_ext_direct_IO(rw, iocb, iov, offset, nr_segs);
|
|
|
|
|
|
|
|
return ext4_ind_direct_IO(rw, iocb, iov, offset, nr_segs);
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* Pages can be marked dirty completely asynchronously from ext4's journalling
|
2006-10-11 10:20:50 +02:00
|
|
|
* activity. By filemap_sync_pte(), try_to_unmap_one(), etc. We cannot do
|
|
|
|
* much here because ->set_page_dirty is called under VFS locks. The page is
|
|
|
|
* not necessarily locked.
|
|
|
|
*
|
|
|
|
* We cannot just dirty the page and leave attached buffers clean, because the
|
|
|
|
* buffers' dirty state is "definitive". We cannot just set the buffers dirty
|
|
|
|
* or jbddirty because all the journalling code will explode.
|
|
|
|
*
|
|
|
|
* So what we do is to mark the page "pending dirty" and next time writepage
|
|
|
|
* is called, propagate that into the buffers appropriately.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static int ext4_journalled_set_page_dirty(struct page *page)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
SetPageChecked(page);
|
|
|
|
return __set_page_dirty_nobuffers(page);
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
static const struct address_space_operations ext4_ordered_aops = {
|
vfs: pagecache usage optimization for pagesize!=blocksize
When we read some part of a file through pagecache, if there is a
pagecache of corresponding index but this page is not uptodate, read IO
is issued and this page will be uptodate.
I think this is good for pagesize == blocksize environment but there is
room for improvement on pagesize != blocksize environment. Because in
this case a page can have multiple buffers and even if a page is not
uptodate, some buffers can be uptodate.
So I suggest that when all buffers which correspond to a part of a file
that we want to read are uptodate, use this pagecache and copy data from
this pagecache to user buffer even if a page is not uptodate. This can
reduce read IO and improve system throughput.
I wrote a benchmark program and got result number with this program.
This benchmark do:
1: mount and open a test file.
2: create a 512MB file.
3: close a file and umount.
4: mount and again open a test file.
5: pwrite randomly 300000 times on a test file. offset is aligned
by IO size(1024bytes).
6: measure time of preading randomly 100000 times on a test file.
The result was:
2.6.26
330 sec
2.6.26-patched
226 sec
Arch:i386
Filesystem:ext3
Blocksize:1024 bytes
Memory: 1GB
On ext3/4, a file is written through buffer/block. So random read/write
mixed workloads or random read after random write workloads are optimized
with this patch under pagesize != blocksize environment. This test result
showed this.
The benchmark program is as follows:
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <time.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mount.h>
#define LEN 1024
#define LOOP 1024*512 /* 512MB */
main(void)
{
unsigned long i, offset, filesize;
int fd;
char buf[LEN];
time_t t1, t2;
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
memset(buf, 0, LEN);
fd = open("/root/test1/testfile", O_CREAT|O_RDWR|O_TRUNC);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
for (i = 0; i < LOOP; i++)
write(fd, buf, LEN);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
fd = open("/root/test1/testfile", O_RDWR);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
filesize = LEN * LOOP;
for (i = 0; i < 300000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pwrite(fd, buf, LEN, offset);
}
printf("start test\n");
time(&t1);
for (i = 0; i < 100000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pread(fd, buf, LEN, offset);
}
time(&t2);
printf("%ld sec\n", t2-t1);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
}
Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Jan Kara <jack@ucw.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:36 +02:00
|
|
|
.readpage = ext4_readpage,
|
|
|
|
.readpages = ext4_readpages,
|
2009-06-14 23:58:45 +02:00
|
|
|
.writepage = ext4_writepage,
|
vfs: pagecache usage optimization for pagesize!=blocksize
When we read some part of a file through pagecache, if there is a
pagecache of corresponding index but this page is not uptodate, read IO
is issued and this page will be uptodate.
I think this is good for pagesize == blocksize environment but there is
room for improvement on pagesize != blocksize environment. Because in
this case a page can have multiple buffers and even if a page is not
uptodate, some buffers can be uptodate.
So I suggest that when all buffers which correspond to a part of a file
that we want to read are uptodate, use this pagecache and copy data from
this pagecache to user buffer even if a page is not uptodate. This can
reduce read IO and improve system throughput.
I wrote a benchmark program and got result number with this program.
This benchmark do:
1: mount and open a test file.
2: create a 512MB file.
3: close a file and umount.
4: mount and again open a test file.
5: pwrite randomly 300000 times on a test file. offset is aligned
by IO size(1024bytes).
6: measure time of preading randomly 100000 times on a test file.
The result was:
2.6.26
330 sec
2.6.26-patched
226 sec
Arch:i386
Filesystem:ext3
Blocksize:1024 bytes
Memory: 1GB
On ext3/4, a file is written through buffer/block. So random read/write
mixed workloads or random read after random write workloads are optimized
with this patch under pagesize != blocksize environment. This test result
showed this.
The benchmark program is as follows:
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <time.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mount.h>
#define LEN 1024
#define LOOP 1024*512 /* 512MB */
main(void)
{
unsigned long i, offset, filesize;
int fd;
char buf[LEN];
time_t t1, t2;
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
memset(buf, 0, LEN);
fd = open("/root/test1/testfile", O_CREAT|O_RDWR|O_TRUNC);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
for (i = 0; i < LOOP; i++)
write(fd, buf, LEN);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
fd = open("/root/test1/testfile", O_RDWR);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
filesize = LEN * LOOP;
for (i = 0; i < 300000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pwrite(fd, buf, LEN, offset);
}
printf("start test\n");
time(&t1);
for (i = 0; i < 100000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pread(fd, buf, LEN, offset);
}
time(&t2);
printf("%ld sec\n", t2-t1);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
}
Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Jan Kara <jack@ucw.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:36 +02:00
|
|
|
.sync_page = block_sync_page,
|
|
|
|
.write_begin = ext4_write_begin,
|
|
|
|
.write_end = ext4_ordered_write_end,
|
|
|
|
.bmap = ext4_bmap,
|
|
|
|
.invalidatepage = ext4_invalidatepage,
|
|
|
|
.releasepage = ext4_releasepage,
|
|
|
|
.direct_IO = ext4_direct_IO,
|
|
|
|
.migratepage = buffer_migrate_page,
|
|
|
|
.is_partially_uptodate = block_is_partially_uptodate,
|
2009-09-16 11:50:16 +02:00
|
|
|
.error_remove_page = generic_error_remove_page,
|
2006-10-11 10:20:50 +02:00
|
|
|
};
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
static const struct address_space_operations ext4_writeback_aops = {
|
vfs: pagecache usage optimization for pagesize!=blocksize
When we read some part of a file through pagecache, if there is a
pagecache of corresponding index but this page is not uptodate, read IO
is issued and this page will be uptodate.
I think this is good for pagesize == blocksize environment but there is
room for improvement on pagesize != blocksize environment. Because in
this case a page can have multiple buffers and even if a page is not
uptodate, some buffers can be uptodate.
So I suggest that when all buffers which correspond to a part of a file
that we want to read are uptodate, use this pagecache and copy data from
this pagecache to user buffer even if a page is not uptodate. This can
reduce read IO and improve system throughput.
I wrote a benchmark program and got result number with this program.
This benchmark do:
1: mount and open a test file.
2: create a 512MB file.
3: close a file and umount.
4: mount and again open a test file.
5: pwrite randomly 300000 times on a test file. offset is aligned
by IO size(1024bytes).
6: measure time of preading randomly 100000 times on a test file.
The result was:
2.6.26
330 sec
2.6.26-patched
226 sec
Arch:i386
Filesystem:ext3
Blocksize:1024 bytes
Memory: 1GB
On ext3/4, a file is written through buffer/block. So random read/write
mixed workloads or random read after random write workloads are optimized
with this patch under pagesize != blocksize environment. This test result
showed this.
The benchmark program is as follows:
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <time.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mount.h>
#define LEN 1024
#define LOOP 1024*512 /* 512MB */
main(void)
{
unsigned long i, offset, filesize;
int fd;
char buf[LEN];
time_t t1, t2;
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
memset(buf, 0, LEN);
fd = open("/root/test1/testfile", O_CREAT|O_RDWR|O_TRUNC);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
for (i = 0; i < LOOP; i++)
write(fd, buf, LEN);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
fd = open("/root/test1/testfile", O_RDWR);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
filesize = LEN * LOOP;
for (i = 0; i < 300000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pwrite(fd, buf, LEN, offset);
}
printf("start test\n");
time(&t1);
for (i = 0; i < 100000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pread(fd, buf, LEN, offset);
}
time(&t2);
printf("%ld sec\n", t2-t1);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
}
Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Jan Kara <jack@ucw.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:36 +02:00
|
|
|
.readpage = ext4_readpage,
|
|
|
|
.readpages = ext4_readpages,
|
2009-06-14 23:58:45 +02:00
|
|
|
.writepage = ext4_writepage,
|
vfs: pagecache usage optimization for pagesize!=blocksize
When we read some part of a file through pagecache, if there is a
pagecache of corresponding index but this page is not uptodate, read IO
is issued and this page will be uptodate.
I think this is good for pagesize == blocksize environment but there is
room for improvement on pagesize != blocksize environment. Because in
this case a page can have multiple buffers and even if a page is not
uptodate, some buffers can be uptodate.
So I suggest that when all buffers which correspond to a part of a file
that we want to read are uptodate, use this pagecache and copy data from
this pagecache to user buffer even if a page is not uptodate. This can
reduce read IO and improve system throughput.
I wrote a benchmark program and got result number with this program.
This benchmark do:
1: mount and open a test file.
2: create a 512MB file.
3: close a file and umount.
4: mount and again open a test file.
5: pwrite randomly 300000 times on a test file. offset is aligned
by IO size(1024bytes).
6: measure time of preading randomly 100000 times on a test file.
The result was:
2.6.26
330 sec
2.6.26-patched
226 sec
Arch:i386
Filesystem:ext3
Blocksize:1024 bytes
Memory: 1GB
On ext3/4, a file is written through buffer/block. So random read/write
mixed workloads or random read after random write workloads are optimized
with this patch under pagesize != blocksize environment. This test result
showed this.
The benchmark program is as follows:
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <time.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mount.h>
#define LEN 1024
#define LOOP 1024*512 /* 512MB */
main(void)
{
unsigned long i, offset, filesize;
int fd;
char buf[LEN];
time_t t1, t2;
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
memset(buf, 0, LEN);
fd = open("/root/test1/testfile", O_CREAT|O_RDWR|O_TRUNC);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
for (i = 0; i < LOOP; i++)
write(fd, buf, LEN);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
fd = open("/root/test1/testfile", O_RDWR);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
filesize = LEN * LOOP;
for (i = 0; i < 300000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pwrite(fd, buf, LEN, offset);
}
printf("start test\n");
time(&t1);
for (i = 0; i < 100000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pread(fd, buf, LEN, offset);
}
time(&t2);
printf("%ld sec\n", t2-t1);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
}
Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Jan Kara <jack@ucw.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:36 +02:00
|
|
|
.sync_page = block_sync_page,
|
|
|
|
.write_begin = ext4_write_begin,
|
|
|
|
.write_end = ext4_writeback_write_end,
|
|
|
|
.bmap = ext4_bmap,
|
|
|
|
.invalidatepage = ext4_invalidatepage,
|
|
|
|
.releasepage = ext4_releasepage,
|
|
|
|
.direct_IO = ext4_direct_IO,
|
|
|
|
.migratepage = buffer_migrate_page,
|
|
|
|
.is_partially_uptodate = block_is_partially_uptodate,
|
2009-09-16 11:50:16 +02:00
|
|
|
.error_remove_page = generic_error_remove_page,
|
2006-10-11 10:20:50 +02:00
|
|
|
};
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
static const struct address_space_operations ext4_journalled_aops = {
|
vfs: pagecache usage optimization for pagesize!=blocksize
When we read some part of a file through pagecache, if there is a
pagecache of corresponding index but this page is not uptodate, read IO
is issued and this page will be uptodate.
I think this is good for pagesize == blocksize environment but there is
room for improvement on pagesize != blocksize environment. Because in
this case a page can have multiple buffers and even if a page is not
uptodate, some buffers can be uptodate.
So I suggest that when all buffers which correspond to a part of a file
that we want to read are uptodate, use this pagecache and copy data from
this pagecache to user buffer even if a page is not uptodate. This can
reduce read IO and improve system throughput.
I wrote a benchmark program and got result number with this program.
This benchmark do:
1: mount and open a test file.
2: create a 512MB file.
3: close a file and umount.
4: mount and again open a test file.
5: pwrite randomly 300000 times on a test file. offset is aligned
by IO size(1024bytes).
6: measure time of preading randomly 100000 times on a test file.
The result was:
2.6.26
330 sec
2.6.26-patched
226 sec
Arch:i386
Filesystem:ext3
Blocksize:1024 bytes
Memory: 1GB
On ext3/4, a file is written through buffer/block. So random read/write
mixed workloads or random read after random write workloads are optimized
with this patch under pagesize != blocksize environment. This test result
showed this.
The benchmark program is as follows:
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <time.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mount.h>
#define LEN 1024
#define LOOP 1024*512 /* 512MB */
main(void)
{
unsigned long i, offset, filesize;
int fd;
char buf[LEN];
time_t t1, t2;
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
memset(buf, 0, LEN);
fd = open("/root/test1/testfile", O_CREAT|O_RDWR|O_TRUNC);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
for (i = 0; i < LOOP; i++)
write(fd, buf, LEN);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
fd = open("/root/test1/testfile", O_RDWR);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
filesize = LEN * LOOP;
for (i = 0; i < 300000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pwrite(fd, buf, LEN, offset);
}
printf("start test\n");
time(&t1);
for (i = 0; i < 100000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pread(fd, buf, LEN, offset);
}
time(&t2);
printf("%ld sec\n", t2-t1);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
}
Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Jan Kara <jack@ucw.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:36 +02:00
|
|
|
.readpage = ext4_readpage,
|
|
|
|
.readpages = ext4_readpages,
|
2009-06-14 23:58:45 +02:00
|
|
|
.writepage = ext4_writepage,
|
vfs: pagecache usage optimization for pagesize!=blocksize
When we read some part of a file through pagecache, if there is a
pagecache of corresponding index but this page is not uptodate, read IO
is issued and this page will be uptodate.
I think this is good for pagesize == blocksize environment but there is
room for improvement on pagesize != blocksize environment. Because in
this case a page can have multiple buffers and even if a page is not
uptodate, some buffers can be uptodate.
So I suggest that when all buffers which correspond to a part of a file
that we want to read are uptodate, use this pagecache and copy data from
this pagecache to user buffer even if a page is not uptodate. This can
reduce read IO and improve system throughput.
I wrote a benchmark program and got result number with this program.
This benchmark do:
1: mount and open a test file.
2: create a 512MB file.
3: close a file and umount.
4: mount and again open a test file.
5: pwrite randomly 300000 times on a test file. offset is aligned
by IO size(1024bytes).
6: measure time of preading randomly 100000 times on a test file.
The result was:
2.6.26
330 sec
2.6.26-patched
226 sec
Arch:i386
Filesystem:ext3
Blocksize:1024 bytes
Memory: 1GB
On ext3/4, a file is written through buffer/block. So random read/write
mixed workloads or random read after random write workloads are optimized
with this patch under pagesize != blocksize environment. This test result
showed this.
The benchmark program is as follows:
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <time.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mount.h>
#define LEN 1024
#define LOOP 1024*512 /* 512MB */
main(void)
{
unsigned long i, offset, filesize;
int fd;
char buf[LEN];
time_t t1, t2;
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
memset(buf, 0, LEN);
fd = open("/root/test1/testfile", O_CREAT|O_RDWR|O_TRUNC);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
for (i = 0; i < LOOP; i++)
write(fd, buf, LEN);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
fd = open("/root/test1/testfile", O_RDWR);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
filesize = LEN * LOOP;
for (i = 0; i < 300000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pwrite(fd, buf, LEN, offset);
}
printf("start test\n");
time(&t1);
for (i = 0; i < 100000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pread(fd, buf, LEN, offset);
}
time(&t2);
printf("%ld sec\n", t2-t1);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
}
Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Jan Kara <jack@ucw.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:36 +02:00
|
|
|
.sync_page = block_sync_page,
|
|
|
|
.write_begin = ext4_write_begin,
|
|
|
|
.write_end = ext4_journalled_write_end,
|
|
|
|
.set_page_dirty = ext4_journalled_set_page_dirty,
|
|
|
|
.bmap = ext4_bmap,
|
|
|
|
.invalidatepage = ext4_invalidatepage,
|
|
|
|
.releasepage = ext4_releasepage,
|
|
|
|
.is_partially_uptodate = block_is_partially_uptodate,
|
2009-09-16 11:50:16 +02:00
|
|
|
.error_remove_page = generic_error_remove_page,
|
2006-10-11 10:20:50 +02:00
|
|
|
};
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
static const struct address_space_operations ext4_da_aops = {
|
vfs: pagecache usage optimization for pagesize!=blocksize
When we read some part of a file through pagecache, if there is a
pagecache of corresponding index but this page is not uptodate, read IO
is issued and this page will be uptodate.
I think this is good for pagesize == blocksize environment but there is
room for improvement on pagesize != blocksize environment. Because in
this case a page can have multiple buffers and even if a page is not
uptodate, some buffers can be uptodate.
So I suggest that when all buffers which correspond to a part of a file
that we want to read are uptodate, use this pagecache and copy data from
this pagecache to user buffer even if a page is not uptodate. This can
reduce read IO and improve system throughput.
I wrote a benchmark program and got result number with this program.
This benchmark do:
1: mount and open a test file.
2: create a 512MB file.
3: close a file and umount.
4: mount and again open a test file.
5: pwrite randomly 300000 times on a test file. offset is aligned
by IO size(1024bytes).
6: measure time of preading randomly 100000 times on a test file.
The result was:
2.6.26
330 sec
2.6.26-patched
226 sec
Arch:i386
Filesystem:ext3
Blocksize:1024 bytes
Memory: 1GB
On ext3/4, a file is written through buffer/block. So random read/write
mixed workloads or random read after random write workloads are optimized
with this patch under pagesize != blocksize environment. This test result
showed this.
The benchmark program is as follows:
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <time.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mount.h>
#define LEN 1024
#define LOOP 1024*512 /* 512MB */
main(void)
{
unsigned long i, offset, filesize;
int fd;
char buf[LEN];
time_t t1, t2;
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
memset(buf, 0, LEN);
fd = open("/root/test1/testfile", O_CREAT|O_RDWR|O_TRUNC);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
for (i = 0; i < LOOP; i++)
write(fd, buf, LEN);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
fd = open("/root/test1/testfile", O_RDWR);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
filesize = LEN * LOOP;
for (i = 0; i < 300000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pwrite(fd, buf, LEN, offset);
}
printf("start test\n");
time(&t1);
for (i = 0; i < 100000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pread(fd, buf, LEN, offset);
}
time(&t2);
printf("%ld sec\n", t2-t1);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
}
Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Jan Kara <jack@ucw.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:36 +02:00
|
|
|
.readpage = ext4_readpage,
|
|
|
|
.readpages = ext4_readpages,
|
2009-06-14 23:58:45 +02:00
|
|
|
.writepage = ext4_writepage,
|
vfs: pagecache usage optimization for pagesize!=blocksize
When we read some part of a file through pagecache, if there is a
pagecache of corresponding index but this page is not uptodate, read IO
is issued and this page will be uptodate.
I think this is good for pagesize == blocksize environment but there is
room for improvement on pagesize != blocksize environment. Because in
this case a page can have multiple buffers and even if a page is not
uptodate, some buffers can be uptodate.
So I suggest that when all buffers which correspond to a part of a file
that we want to read are uptodate, use this pagecache and copy data from
this pagecache to user buffer even if a page is not uptodate. This can
reduce read IO and improve system throughput.
I wrote a benchmark program and got result number with this program.
This benchmark do:
1: mount and open a test file.
2: create a 512MB file.
3: close a file and umount.
4: mount and again open a test file.
5: pwrite randomly 300000 times on a test file. offset is aligned
by IO size(1024bytes).
6: measure time of preading randomly 100000 times on a test file.
The result was:
2.6.26
330 sec
2.6.26-patched
226 sec
Arch:i386
Filesystem:ext3
Blocksize:1024 bytes
Memory: 1GB
On ext3/4, a file is written through buffer/block. So random read/write
mixed workloads or random read after random write workloads are optimized
with this patch under pagesize != blocksize environment. This test result
showed this.
The benchmark program is as follows:
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <time.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mount.h>
#define LEN 1024
#define LOOP 1024*512 /* 512MB */
main(void)
{
unsigned long i, offset, filesize;
int fd;
char buf[LEN];
time_t t1, t2;
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
memset(buf, 0, LEN);
fd = open("/root/test1/testfile", O_CREAT|O_RDWR|O_TRUNC);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
for (i = 0; i < LOOP; i++)
write(fd, buf, LEN);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) {
perror("cannot mount\n");
exit(1);
}
fd = open("/root/test1/testfile", O_RDWR);
if (fd < 0) {
perror("cannot open file\n");
exit(1);
}
filesize = LEN * LOOP;
for (i = 0; i < 300000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pwrite(fd, buf, LEN, offset);
}
printf("start test\n");
time(&t1);
for (i = 0; i < 100000; i++){
offset = (random() % filesize) & (~(LEN - 1));
pread(fd, buf, LEN, offset);
}
time(&t2);
printf("%ld sec\n", t2-t1);
close(fd);
if (umount("/root/test1/") < 0) {
perror("cannot umount\n");
exit(1);
}
}
Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Jan Kara <jack@ucw.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-29 00:46:36 +02:00
|
|
|
.writepages = ext4_da_writepages,
|
|
|
|
.sync_page = block_sync_page,
|
|
|
|
.write_begin = ext4_da_write_begin,
|
|
|
|
.write_end = ext4_da_write_end,
|
|
|
|
.bmap = ext4_bmap,
|
|
|
|
.invalidatepage = ext4_da_invalidatepage,
|
|
|
|
.releasepage = ext4_releasepage,
|
|
|
|
.direct_IO = ext4_direct_IO,
|
|
|
|
.migratepage = buffer_migrate_page,
|
|
|
|
.is_partially_uptodate = block_is_partially_uptodate,
|
2009-09-16 11:50:16 +02:00
|
|
|
.error_remove_page = generic_error_remove_page,
|
2008-07-12 01:27:31 +02:00
|
|
|
};
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
void ext4_set_aops(struct inode *inode)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2008-07-12 01:27:31 +02:00
|
|
|
if (ext4_should_order_data(inode) &&
|
|
|
|
test_opt(inode->i_sb, DELALLOC))
|
|
|
|
inode->i_mapping->a_ops = &ext4_da_aops;
|
|
|
|
else if (ext4_should_order_data(inode))
|
2006-10-11 10:20:53 +02:00
|
|
|
inode->i_mapping->a_ops = &ext4_ordered_aops;
|
2008-07-12 01:27:31 +02:00
|
|
|
else if (ext4_should_writeback_data(inode) &&
|
|
|
|
test_opt(inode->i_sb, DELALLOC))
|
|
|
|
inode->i_mapping->a_ops = &ext4_da_aops;
|
2006-10-11 10:20:53 +02:00
|
|
|
else if (ext4_should_writeback_data(inode))
|
|
|
|
inode->i_mapping->a_ops = &ext4_writeback_aops;
|
2006-10-11 10:20:50 +02:00
|
|
|
else
|
2006-10-11 10:20:53 +02:00
|
|
|
inode->i_mapping->a_ops = &ext4_journalled_aops;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_block_truncate_page() zeroes out a mapping from file offset `from'
|
2006-10-11 10:20:50 +02:00
|
|
|
* up to the end of the block which corresponds to `from'.
|
|
|
|
* This required during truncate. We need to physically zero the tail end
|
|
|
|
* of that block so it doesn't yield old data if the file is later grown.
|
|
|
|
*/
|
2008-07-12 01:27:31 +02:00
|
|
|
int ext4_block_truncate_page(handle_t *handle,
|
2006-10-11 10:20:50 +02:00
|
|
|
struct address_space *mapping, loff_t from)
|
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_fsblk_t index = from >> PAGE_CACHE_SHIFT;
|
2006-10-11 10:20:50 +02:00
|
|
|
unsigned offset = from & (PAGE_CACHE_SIZE-1);
|
2008-01-29 05:58:27 +01:00
|
|
|
unsigned blocksize, length, pos;
|
|
|
|
ext4_lblk_t iblock;
|
2006-10-11 10:20:50 +02:00
|
|
|
struct inode *inode = mapping->host;
|
|
|
|
struct buffer_head *bh;
|
2008-07-12 01:27:31 +02:00
|
|
|
struct page *page;
|
2006-10-11 10:20:50 +02:00
|
|
|
int err = 0;
|
|
|
|
|
2009-07-06 04:08:16 +02:00
|
|
|
page = find_or_create_page(mapping, from >> PAGE_CACHE_SHIFT,
|
|
|
|
mapping_gfp_mask(mapping) & ~__GFP_FS);
|
2008-07-12 01:27:31 +02:00
|
|
|
if (!page)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
blocksize = inode->i_sb->s_blocksize;
|
|
|
|
length = blocksize - (offset & (blocksize - 1));
|
|
|
|
iblock = index << (PAGE_CACHE_SHIFT - inode->i_sb->s_blocksize_bits);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For "nobh" option, we can only work if we don't need to
|
|
|
|
* read-in the page - otherwise we create buffers to do the IO.
|
|
|
|
*/
|
|
|
|
if (!page_has_buffers(page) && test_opt(inode->i_sb, NOBH) &&
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_should_writeback_data(inode) && PageUptodate(page)) {
|
2008-02-05 07:28:29 +01:00
|
|
|
zero_user(page, offset, length);
|
2006-10-11 10:20:50 +02:00
|
|
|
set_page_dirty(page);
|
|
|
|
goto unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!page_has_buffers(page))
|
|
|
|
create_empty_buffers(page, blocksize, 0);
|
|
|
|
|
|
|
|
/* Find the buffer that contains "offset" */
|
|
|
|
bh = page_buffers(page);
|
|
|
|
pos = blocksize;
|
|
|
|
while (offset >= pos) {
|
|
|
|
bh = bh->b_this_page;
|
|
|
|
iblock++;
|
|
|
|
pos += blocksize;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = 0;
|
|
|
|
if (buffer_freed(bh)) {
|
|
|
|
BUFFER_TRACE(bh, "freed: skip");
|
|
|
|
goto unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!buffer_mapped(bh)) {
|
|
|
|
BUFFER_TRACE(bh, "unmapped");
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_get_block(inode, iblock, bh, 0);
|
2006-10-11 10:20:50 +02:00
|
|
|
/* unmapped? It's a hole - nothing to do */
|
|
|
|
if (!buffer_mapped(bh)) {
|
|
|
|
BUFFER_TRACE(bh, "still unmapped");
|
|
|
|
goto unlock;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Ok, it's mapped. Make sure it's up-to-date */
|
|
|
|
if (PageUptodate(page))
|
|
|
|
set_buffer_uptodate(bh);
|
|
|
|
|
|
|
|
if (!buffer_uptodate(bh)) {
|
|
|
|
err = -EIO;
|
|
|
|
ll_rw_block(READ, 1, &bh);
|
|
|
|
wait_on_buffer(bh);
|
|
|
|
/* Uhhuh. Read error. Complain and punt. */
|
|
|
|
if (!buffer_uptodate(bh))
|
|
|
|
goto unlock;
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
if (ext4_should_journal_data(inode)) {
|
2006-10-11 10:20:50 +02:00
|
|
|
BUFFER_TRACE(bh, "get write access");
|
2006-10-11 10:20:53 +02:00
|
|
|
err = ext4_journal_get_write_access(handle, bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (err)
|
|
|
|
goto unlock;
|
|
|
|
}
|
|
|
|
|
2008-02-05 07:28:29 +01:00
|
|
|
zero_user(page, offset, length);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
BUFFER_TRACE(bh, "zeroed end of block");
|
|
|
|
|
|
|
|
err = 0;
|
2006-10-11 10:20:53 +02:00
|
|
|
if (ext4_should_journal_data(inode)) {
|
2009-01-07 06:06:22 +01:00
|
|
|
err = ext4_handle_dirty_metadata(handle, inode, bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
} else {
|
2006-10-11 10:20:53 +02:00
|
|
|
if (ext4_should_order_data(inode))
|
2008-07-12 01:27:31 +02:00
|
|
|
err = ext4_jbd2_file_inode(handle, inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
mark_buffer_dirty(bh);
|
|
|
|
}
|
|
|
|
|
|
|
|
unlock:
|
|
|
|
unlock_page(page);
|
|
|
|
page_cache_release(page);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Probably it should be a library function... search for first non-zero word
|
|
|
|
* or memcmp with zero_page, whatever is better for particular architecture.
|
|
|
|
* Linus?
|
|
|
|
*/
|
|
|
|
static inline int all_zeroes(__le32 *p, __le32 *q)
|
|
|
|
{
|
|
|
|
while (p < q)
|
|
|
|
if (*p++)
|
|
|
|
return 0;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_find_shared - find the indirect blocks for partial truncation.
|
2006-10-11 10:20:50 +02:00
|
|
|
* @inode: inode in question
|
|
|
|
* @depth: depth of the affected branch
|
2006-10-11 10:20:53 +02:00
|
|
|
* @offsets: offsets of pointers in that branch (see ext4_block_to_path)
|
2006-10-11 10:20:50 +02:00
|
|
|
* @chain: place to store the pointers to partial indirect blocks
|
|
|
|
* @top: place to the (detached) top of branch
|
|
|
|
*
|
2006-10-11 10:20:53 +02:00
|
|
|
* This is a helper function used by ext4_truncate().
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
|
|
|
* When we do truncate() we may have to clean the ends of several
|
|
|
|
* indirect blocks but leave the blocks themselves alive. Block is
|
|
|
|
* partially truncated if some data below the new i_size is refered
|
|
|
|
* from it (and it is on the path to the first completely truncated
|
|
|
|
* data block, indeed). We have to free the top of that path along
|
|
|
|
* with everything to the right of the path. Since no allocation
|
2006-10-11 10:20:53 +02:00
|
|
|
* past the truncation point is possible until ext4_truncate()
|
2006-10-11 10:20:50 +02:00
|
|
|
* finishes, we may safely do the latter, but top of branch may
|
|
|
|
* require special attention - pageout below the truncation point
|
|
|
|
* might try to populate it.
|
|
|
|
*
|
|
|
|
* We atomically detach the top of branch from the tree, store the
|
|
|
|
* block number of its root in *@top, pointers to buffer_heads of
|
|
|
|
* partially truncated blocks - in @chain[].bh and pointers to
|
|
|
|
* their last elements that should not be removed - in
|
|
|
|
* @chain[].p. Return value is the pointer to last filled element
|
|
|
|
* of @chain.
|
|
|
|
*
|
|
|
|
* The work left to caller to do the actual freeing of subtrees:
|
|
|
|
* a) free the subtree starting from *@top
|
|
|
|
* b) free the subtrees whose roots are stored in
|
|
|
|
* (@chain[i].p+1 .. end of @chain[i].bh->b_data)
|
|
|
|
* c) free the subtrees growing from the inode past the @chain[0].
|
|
|
|
* (no partially truncated stuff there). */
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
static Indirect *ext4_find_shared(struct inode *inode, int depth,
|
2009-06-14 23:45:34 +02:00
|
|
|
ext4_lblk_t offsets[4], Indirect chain[4],
|
|
|
|
__le32 *top)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
Indirect *partial, *p;
|
|
|
|
int k, err;
|
|
|
|
|
|
|
|
*top = 0;
|
2009-10-28 20:11:03 +01:00
|
|
|
/* Make k index the deepest non-null offset + 1 */
|
2006-10-11 10:20:50 +02:00
|
|
|
for (k = depth; k > 1 && !offsets[k-1]; k--)
|
|
|
|
;
|
2006-10-11 10:20:53 +02:00
|
|
|
partial = ext4_get_branch(inode, k, offsets, chain, &err);
|
2006-10-11 10:20:50 +02:00
|
|
|
/* Writer: pointers */
|
|
|
|
if (!partial)
|
|
|
|
partial = chain + k-1;
|
|
|
|
/*
|
|
|
|
* If the branch acquired continuation since we've looked at it -
|
|
|
|
* fine, it should all survive and (new) top doesn't belong to us.
|
|
|
|
*/
|
|
|
|
if (!partial->key && *partial->p)
|
|
|
|
/* Writer: end */
|
|
|
|
goto no_top;
|
2008-09-09 04:25:24 +02:00
|
|
|
for (p = partial; (p > chain) && all_zeroes((__le32 *) p->bh->b_data, p->p); p--)
|
2006-10-11 10:20:50 +02:00
|
|
|
;
|
|
|
|
/*
|
|
|
|
* OK, we've found the last block that must survive. The rest of our
|
|
|
|
* branch should be detached before unlocking. However, if that rest
|
|
|
|
* of branch is all ours and does not grow immediately from the inode
|
|
|
|
* it's easier to cheat and just decrement partial->p.
|
|
|
|
*/
|
|
|
|
if (p == chain + k - 1 && p > chain) {
|
|
|
|
p->p--;
|
|
|
|
} else {
|
|
|
|
*top = *p->p;
|
2006-10-11 10:20:53 +02:00
|
|
|
/* Nope, don't do this in ext4. Must leave the tree intact */
|
2006-10-11 10:20:50 +02:00
|
|
|
#if 0
|
|
|
|
*p->p = 0;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
/* Writer: end */
|
|
|
|
|
2008-09-09 04:25:24 +02:00
|
|
|
while (partial > p) {
|
2006-10-11 10:20:50 +02:00
|
|
|
brelse(partial->bh);
|
|
|
|
partial--;
|
|
|
|
}
|
|
|
|
no_top:
|
|
|
|
return partial;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Zero a number of block pointers in either an inode or an indirect block.
|
|
|
|
* If we restart the transaction we must again get write access to the
|
|
|
|
* indirect block for further modification.
|
|
|
|
*
|
|
|
|
* We release `count' blocks on disk, but (last - first) may be greater
|
|
|
|
* than `count' because there can be holes in there.
|
|
|
|
*/
|
ext4: Add block validity check when truncating indirect block mapped inodes
Add checks to ext4_free_branches() to make sure a block number found
in an indirect block are valid before trying to free it. If a bad
block number is found, stop freeing the indirect block immediately,
since the file system is corrupt and we will need to run fsck anyway.
This also avoids spamming the logs, and specifically avoids
driver-level "attempt to access beyond end of device" errors obscure
what is really going on.
If you get *really*, *really*, *really* unlucky, without this patch, a
supposed indirect block containing garbage might contain a reference
to a primary block group descriptor, in which case
ext4_free_branches() could end up zero'ing out a block group
descriptor block, and if then one of the block bitmaps for a block
group described by that bg descriptor block is not in memory, and is
read in by ext4_read_block_bitmap(). This function calls
ext4_valid_block_bitmap(), which assumes that bg_inode_table() was
validated at mount time and hasn't been modified since. Since this
assumption is no longer valid, it's possible for the value
(ext4_inode_table(sb, desc) - group_first_block) to go negative, which
will cause ext4_find_next_zero_bit() to trigger a kernel GPF.
Addresses-Google-Bug: #2220436
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2010-01-22 23:40:42 +01:00
|
|
|
static int ext4_clear_blocks(handle_t *handle, struct inode *inode,
|
|
|
|
struct buffer_head *bh,
|
|
|
|
ext4_fsblk_t block_to_free,
|
|
|
|
unsigned long count, __le32 *first,
|
|
|
|
__le32 *last)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
__le32 *p;
|
ext4: Add block validity check when truncating indirect block mapped inodes
Add checks to ext4_free_branches() to make sure a block number found
in an indirect block are valid before trying to free it. If a bad
block number is found, stop freeing the indirect block immediately,
since the file system is corrupt and we will need to run fsck anyway.
This also avoids spamming the logs, and specifically avoids
driver-level "attempt to access beyond end of device" errors obscure
what is really going on.
If you get *really*, *really*, *really* unlucky, without this patch, a
supposed indirect block containing garbage might contain a reference
to a primary block group descriptor, in which case
ext4_free_branches() could end up zero'ing out a block group
descriptor block, and if then one of the block bitmaps for a block
group described by that bg descriptor block is not in memory, and is
read in by ext4_read_block_bitmap(). This function calls
ext4_valid_block_bitmap(), which assumes that bg_inode_table() was
validated at mount time and hasn't been modified since. Since this
assumption is no longer valid, it's possible for the value
(ext4_inode_table(sb, desc) - group_first_block) to go negative, which
will cause ext4_find_next_zero_bit() to trigger a kernel GPF.
Addresses-Google-Bug: #2220436
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2010-01-22 23:40:42 +01:00
|
|
|
int flags = EXT4_FREE_BLOCKS_FORGET | EXT4_FREE_BLOCKS_VALIDATED;
|
2009-11-23 13:17:05 +01:00
|
|
|
|
|
|
|
if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
|
|
|
|
flags |= EXT4_FREE_BLOCKS_METADATA;
|
2009-11-23 13:17:34 +01:00
|
|
|
|
ext4: Add block validity check when truncating indirect block mapped inodes
Add checks to ext4_free_branches() to make sure a block number found
in an indirect block are valid before trying to free it. If a bad
block number is found, stop freeing the indirect block immediately,
since the file system is corrupt and we will need to run fsck anyway.
This also avoids spamming the logs, and specifically avoids
driver-level "attempt to access beyond end of device" errors obscure
what is really going on.
If you get *really*, *really*, *really* unlucky, without this patch, a
supposed indirect block containing garbage might contain a reference
to a primary block group descriptor, in which case
ext4_free_branches() could end up zero'ing out a block group
descriptor block, and if then one of the block bitmaps for a block
group described by that bg descriptor block is not in memory, and is
read in by ext4_read_block_bitmap(). This function calls
ext4_valid_block_bitmap(), which assumes that bg_inode_table() was
validated at mount time and hasn't been modified since. Since this
assumption is no longer valid, it's possible for the value
(ext4_inode_table(sb, desc) - group_first_block) to go negative, which
will cause ext4_find_next_zero_bit() to trigger a kernel GPF.
Addresses-Google-Bug: #2220436
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2010-01-22 23:40:42 +01:00
|
|
|
if (!ext4_data_block_valid(EXT4_SB(inode->i_sb), block_to_free,
|
|
|
|
count)) {
|
2010-05-17 03:00:00 +02:00
|
|
|
EXT4_ERROR_INODE(inode, "attempt to clear invalid "
|
|
|
|
"blocks %llu len %lu",
|
|
|
|
(unsigned long long) block_to_free, count);
|
ext4: Add block validity check when truncating indirect block mapped inodes
Add checks to ext4_free_branches() to make sure a block number found
in an indirect block are valid before trying to free it. If a bad
block number is found, stop freeing the indirect block immediately,
since the file system is corrupt and we will need to run fsck anyway.
This also avoids spamming the logs, and specifically avoids
driver-level "attempt to access beyond end of device" errors obscure
what is really going on.
If you get *really*, *really*, *really* unlucky, without this patch, a
supposed indirect block containing garbage might contain a reference
to a primary block group descriptor, in which case
ext4_free_branches() could end up zero'ing out a block group
descriptor block, and if then one of the block bitmaps for a block
group described by that bg descriptor block is not in memory, and is
read in by ext4_read_block_bitmap(). This function calls
ext4_valid_block_bitmap(), which assumes that bg_inode_table() was
validated at mount time and hasn't been modified since. Since this
assumption is no longer valid, it's possible for the value
(ext4_inode_table(sb, desc) - group_first_block) to go negative, which
will cause ext4_find_next_zero_bit() to trigger a kernel GPF.
Addresses-Google-Bug: #2220436
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2010-01-22 23:40:42 +01:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
if (try_to_extend_transaction(handle, inode)) {
|
|
|
|
if (bh) {
|
2009-01-07 06:06:22 +01:00
|
|
|
BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
|
|
|
|
ext4_handle_dirty_metadata(handle, inode, bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_mark_inode_dirty(handle, inode);
|
2009-08-18 04:17:20 +02:00
|
|
|
ext4_truncate_restart_trans(handle, inode,
|
|
|
|
blocks_for_truncate(inode));
|
2006-10-11 10:20:50 +02:00
|
|
|
if (bh) {
|
|
|
|
BUFFER_TRACE(bh, "retaking write access");
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_journal_get_write_access(handle, bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-11-23 13:17:05 +01:00
|
|
|
for (p = first; p < last; p++)
|
|
|
|
*p = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2009-11-23 13:17:05 +01:00
|
|
|
ext4_free_blocks(handle, inode, 0, block_to_free, count, flags);
|
ext4: Add block validity check when truncating indirect block mapped inodes
Add checks to ext4_free_branches() to make sure a block number found
in an indirect block are valid before trying to free it. If a bad
block number is found, stop freeing the indirect block immediately,
since the file system is corrupt and we will need to run fsck anyway.
This also avoids spamming the logs, and specifically avoids
driver-level "attempt to access beyond end of device" errors obscure
what is really going on.
If you get *really*, *really*, *really* unlucky, without this patch, a
supposed indirect block containing garbage might contain a reference
to a primary block group descriptor, in which case
ext4_free_branches() could end up zero'ing out a block group
descriptor block, and if then one of the block bitmaps for a block
group described by that bg descriptor block is not in memory, and is
read in by ext4_read_block_bitmap(). This function calls
ext4_valid_block_bitmap(), which assumes that bg_inode_table() was
validated at mount time and hasn't been modified since. Since this
assumption is no longer valid, it's possible for the value
(ext4_inode_table(sb, desc) - group_first_block) to go negative, which
will cause ext4_find_next_zero_bit() to trigger a kernel GPF.
Addresses-Google-Bug: #2220436
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2010-01-22 23:40:42 +01:00
|
|
|
return 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_free_data - free a list of data blocks
|
2006-10-11 10:20:50 +02:00
|
|
|
* @handle: handle for this transaction
|
|
|
|
* @inode: inode we are dealing with
|
|
|
|
* @this_bh: indirect buffer_head which contains *@first and *@last
|
|
|
|
* @first: array of block numbers
|
|
|
|
* @last: points immediately past the end of array
|
|
|
|
*
|
|
|
|
* We are freeing all blocks refered from that array (numbers are stored as
|
|
|
|
* little-endian 32-bit) and updating @inode->i_blocks appropriately.
|
|
|
|
*
|
|
|
|
* We accumulate contiguous runs of blocks to free. Conveniently, if these
|
|
|
|
* blocks are contiguous then releasing them at one time will only affect one
|
|
|
|
* or two bitmap blocks (+ group descriptor(s) and superblock) and we won't
|
|
|
|
* actually use a lot of journal space.
|
|
|
|
*
|
|
|
|
* @this_bh will be %NULL if @first and @last point into the inode's direct
|
|
|
|
* block pointers.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static void ext4_free_data(handle_t *handle, struct inode *inode,
|
2006-10-11 10:20:50 +02:00
|
|
|
struct buffer_head *this_bh,
|
|
|
|
__le32 *first, __le32 *last)
|
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_fsblk_t block_to_free = 0; /* Starting block # of a run */
|
2006-10-11 10:20:50 +02:00
|
|
|
unsigned long count = 0; /* Number of blocks in the run */
|
|
|
|
__le32 *block_to_free_p = NULL; /* Pointer into inode/ind
|
|
|
|
corresponding to
|
|
|
|
block_to_free */
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_fsblk_t nr; /* Current block # */
|
2006-10-11 10:20:50 +02:00
|
|
|
__le32 *p; /* Pointer into inode/ind
|
|
|
|
for current block */
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if (this_bh) { /* For indirect block */
|
|
|
|
BUFFER_TRACE(this_bh, "get_write_access");
|
2006-10-11 10:20:53 +02:00
|
|
|
err = ext4_journal_get_write_access(handle, this_bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
/* Important: if we can't update the indirect pointers
|
|
|
|
* to the blocks, we can't free them. */
|
|
|
|
if (err)
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (p = first; p < last; p++) {
|
|
|
|
nr = le32_to_cpu(*p);
|
|
|
|
if (nr) {
|
|
|
|
/* accumulate blocks to free if they're contiguous */
|
|
|
|
if (count == 0) {
|
|
|
|
block_to_free = nr;
|
|
|
|
block_to_free_p = p;
|
|
|
|
count = 1;
|
|
|
|
} else if (nr == block_to_free + count) {
|
|
|
|
count++;
|
|
|
|
} else {
|
ext4: Add block validity check when truncating indirect block mapped inodes
Add checks to ext4_free_branches() to make sure a block number found
in an indirect block are valid before trying to free it. If a bad
block number is found, stop freeing the indirect block immediately,
since the file system is corrupt and we will need to run fsck anyway.
This also avoids spamming the logs, and specifically avoids
driver-level "attempt to access beyond end of device" errors obscure
what is really going on.
If you get *really*, *really*, *really* unlucky, without this patch, a
supposed indirect block containing garbage might contain a reference
to a primary block group descriptor, in which case
ext4_free_branches() could end up zero'ing out a block group
descriptor block, and if then one of the block bitmaps for a block
group described by that bg descriptor block is not in memory, and is
read in by ext4_read_block_bitmap(). This function calls
ext4_valid_block_bitmap(), which assumes that bg_inode_table() was
validated at mount time and hasn't been modified since. Since this
assumption is no longer valid, it's possible for the value
(ext4_inode_table(sb, desc) - group_first_block) to go negative, which
will cause ext4_find_next_zero_bit() to trigger a kernel GPF.
Addresses-Google-Bug: #2220436
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2010-01-22 23:40:42 +01:00
|
|
|
if (ext4_clear_blocks(handle, inode, this_bh,
|
|
|
|
block_to_free, count,
|
|
|
|
block_to_free_p, p))
|
|
|
|
break;
|
2006-10-11 10:20:50 +02:00
|
|
|
block_to_free = nr;
|
|
|
|
block_to_free_p = p;
|
|
|
|
count = 1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (count > 0)
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_clear_blocks(handle, inode, this_bh, block_to_free,
|
2006-10-11 10:20:50 +02:00
|
|
|
count, block_to_free_p, p);
|
|
|
|
|
|
|
|
if (this_bh) {
|
2009-01-07 06:06:22 +01:00
|
|
|
BUFFER_TRACE(this_bh, "call ext4_handle_dirty_metadata");
|
2008-07-12 01:27:31 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The buffer head should have an attached journal head at this
|
|
|
|
* point. However, if the data is corrupted and an indirect
|
|
|
|
* block pointed to itself, it would have been detached when
|
|
|
|
* the block was cleared. Check for this instead of OOPSing.
|
|
|
|
*/
|
2009-01-20 15:50:19 +01:00
|
|
|
if ((EXT4_JOURNAL(inode) == NULL) || bh2jh(this_bh))
|
2009-01-07 06:06:22 +01:00
|
|
|
ext4_handle_dirty_metadata(handle, inode, this_bh);
|
2008-07-12 01:27:31 +02:00
|
|
|
else
|
2010-05-17 03:00:00 +02:00
|
|
|
EXT4_ERROR_INODE(inode,
|
|
|
|
"circular indirect block detected at "
|
|
|
|
"block %llu",
|
|
|
|
(unsigned long long) this_bh->b_blocknr);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_free_branches - free an array of branches
|
2006-10-11 10:20:50 +02:00
|
|
|
* @handle: JBD handle for this transaction
|
|
|
|
* @inode: inode we are dealing with
|
|
|
|
* @parent_bh: the buffer_head which contains *@first and *@last
|
|
|
|
* @first: array of block numbers
|
|
|
|
* @last: pointer immediately past the end of array
|
|
|
|
* @depth: depth of the branches to free
|
|
|
|
*
|
|
|
|
* We are freeing all blocks refered from these branches (numbers are
|
|
|
|
* stored as little-endian 32-bit) and updating @inode->i_blocks
|
|
|
|
* appropriately.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static void ext4_free_branches(handle_t *handle, struct inode *inode,
|
2006-10-11 10:20:50 +02:00
|
|
|
struct buffer_head *parent_bh,
|
|
|
|
__le32 *first, __le32 *last, int depth)
|
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_fsblk_t nr;
|
2006-10-11 10:20:50 +02:00
|
|
|
__le32 *p;
|
|
|
|
|
2009-01-07 06:06:22 +01:00
|
|
|
if (ext4_handle_is_aborted(handle))
|
2006-10-11 10:20:50 +02:00
|
|
|
return;
|
|
|
|
|
|
|
|
if (depth--) {
|
|
|
|
struct buffer_head *bh;
|
2006-10-11 10:20:53 +02:00
|
|
|
int addr_per_block = EXT4_ADDR_PER_BLOCK(inode->i_sb);
|
2006-10-11 10:20:50 +02:00
|
|
|
p = last;
|
|
|
|
while (--p >= first) {
|
|
|
|
nr = le32_to_cpu(*p);
|
|
|
|
if (!nr)
|
|
|
|
continue; /* A hole */
|
|
|
|
|
ext4: Add block validity check when truncating indirect block mapped inodes
Add checks to ext4_free_branches() to make sure a block number found
in an indirect block are valid before trying to free it. If a bad
block number is found, stop freeing the indirect block immediately,
since the file system is corrupt and we will need to run fsck anyway.
This also avoids spamming the logs, and specifically avoids
driver-level "attempt to access beyond end of device" errors obscure
what is really going on.
If you get *really*, *really*, *really* unlucky, without this patch, a
supposed indirect block containing garbage might contain a reference
to a primary block group descriptor, in which case
ext4_free_branches() could end up zero'ing out a block group
descriptor block, and if then one of the block bitmaps for a block
group described by that bg descriptor block is not in memory, and is
read in by ext4_read_block_bitmap(). This function calls
ext4_valid_block_bitmap(), which assumes that bg_inode_table() was
validated at mount time and hasn't been modified since. Since this
assumption is no longer valid, it's possible for the value
(ext4_inode_table(sb, desc) - group_first_block) to go negative, which
will cause ext4_find_next_zero_bit() to trigger a kernel GPF.
Addresses-Google-Bug: #2220436
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2010-01-22 23:40:42 +01:00
|
|
|
if (!ext4_data_block_valid(EXT4_SB(inode->i_sb),
|
|
|
|
nr, 1)) {
|
2010-05-17 03:00:00 +02:00
|
|
|
EXT4_ERROR_INODE(inode,
|
|
|
|
"invalid indirect mapped "
|
|
|
|
"block %lu (level %d)",
|
|
|
|
(unsigned long) nr, depth);
|
ext4: Add block validity check when truncating indirect block mapped inodes
Add checks to ext4_free_branches() to make sure a block number found
in an indirect block are valid before trying to free it. If a bad
block number is found, stop freeing the indirect block immediately,
since the file system is corrupt and we will need to run fsck anyway.
This also avoids spamming the logs, and specifically avoids
driver-level "attempt to access beyond end of device" errors obscure
what is really going on.
If you get *really*, *really*, *really* unlucky, without this patch, a
supposed indirect block containing garbage might contain a reference
to a primary block group descriptor, in which case
ext4_free_branches() could end up zero'ing out a block group
descriptor block, and if then one of the block bitmaps for a block
group described by that bg descriptor block is not in memory, and is
read in by ext4_read_block_bitmap(). This function calls
ext4_valid_block_bitmap(), which assumes that bg_inode_table() was
validated at mount time and hasn't been modified since. Since this
assumption is no longer valid, it's possible for the value
(ext4_inode_table(sb, desc) - group_first_block) to go negative, which
will cause ext4_find_next_zero_bit() to trigger a kernel GPF.
Addresses-Google-Bug: #2220436
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2010-01-22 23:40:42 +01:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/* Go read the buffer for the next level down */
|
|
|
|
bh = sb_bread(inode->i_sb, nr);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* A read failure? Report error and clear slot
|
|
|
|
* (should be rare).
|
|
|
|
*/
|
|
|
|
if (!bh) {
|
2010-05-17 03:00:00 +02:00
|
|
|
EXT4_ERROR_INODE(inode,
|
|
|
|
"Read failure block=%llu",
|
|
|
|
(unsigned long long) nr);
|
2006-10-11 10:20:50 +02:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* This zaps the entire block. Bottom up. */
|
|
|
|
BUFFER_TRACE(bh, "free child branches");
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_free_branches(handle, inode, bh,
|
2008-09-09 04:25:24 +02:00
|
|
|
(__le32 *) bh->b_data,
|
|
|
|
(__le32 *) bh->b_data + addr_per_block,
|
|
|
|
depth);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* We've probably journalled the indirect block several
|
|
|
|
* times during the truncate. But it's no longer
|
|
|
|
* needed and we now drop it from the transaction via
|
2006-10-11 10:21:01 +02:00
|
|
|
* jbd2_journal_revoke().
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
|
|
|
* That's easy if it's exclusively part of this
|
|
|
|
* transaction. But if it's part of the committing
|
2006-10-11 10:21:01 +02:00
|
|
|
* transaction then jbd2_journal_forget() will simply
|
2006-10-11 10:20:50 +02:00
|
|
|
* brelse() it. That means that if the underlying
|
2006-10-11 10:20:53 +02:00
|
|
|
* block is reallocated in ext4_get_block(),
|
2006-10-11 10:20:50 +02:00
|
|
|
* unmap_underlying_metadata() will find this block
|
|
|
|
* and will try to get rid of it. damn, damn.
|
|
|
|
*
|
|
|
|
* If this block has already been committed to the
|
|
|
|
* journal, a revoke record will be written. And
|
|
|
|
* revoke records must be emitted *before* clearing
|
|
|
|
* this block's bit in the bitmaps.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_forget(handle, 1, inode, bh, bh->b_blocknr);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Everything below this this pointer has been
|
|
|
|
* released. Now let this top-of-subtree go.
|
|
|
|
*
|
|
|
|
* We want the freeing of this indirect block to be
|
|
|
|
* atomic in the journal with the updating of the
|
|
|
|
* bitmap block which owns it. So make some room in
|
|
|
|
* the journal.
|
|
|
|
*
|
|
|
|
* We zero the parent pointer *after* freeing its
|
|
|
|
* pointee in the bitmaps, so if extend_transaction()
|
|
|
|
* for some reason fails to put the bitmap changes and
|
|
|
|
* the release into the same transaction, recovery
|
|
|
|
* will merely complain about releasing a free block,
|
|
|
|
* rather than leaking blocks.
|
|
|
|
*/
|
2009-01-07 06:06:22 +01:00
|
|
|
if (ext4_handle_is_aborted(handle))
|
2006-10-11 10:20:50 +02:00
|
|
|
return;
|
|
|
|
if (try_to_extend_transaction(handle, inode)) {
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_mark_inode_dirty(handle, inode);
|
2009-08-18 04:17:20 +02:00
|
|
|
ext4_truncate_restart_trans(handle, inode,
|
|
|
|
blocks_for_truncate(inode));
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2009-11-23 13:17:05 +01:00
|
|
|
ext4_free_blocks(handle, inode, 0, nr, 1,
|
|
|
|
EXT4_FREE_BLOCKS_METADATA);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
if (parent_bh) {
|
|
|
|
/*
|
|
|
|
* The block which we have just freed is
|
|
|
|
* pointed to by an indirect block: journal it
|
|
|
|
*/
|
|
|
|
BUFFER_TRACE(parent_bh, "get_write_access");
|
2006-10-11 10:20:53 +02:00
|
|
|
if (!ext4_journal_get_write_access(handle,
|
2006-10-11 10:20:50 +02:00
|
|
|
parent_bh)){
|
|
|
|
*p = 0;
|
|
|
|
BUFFER_TRACE(parent_bh,
|
2009-01-07 06:06:22 +01:00
|
|
|
"call ext4_handle_dirty_metadata");
|
|
|
|
ext4_handle_dirty_metadata(handle,
|
|
|
|
inode,
|
|
|
|
parent_bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
/* We have reached the bottom of the tree. */
|
|
|
|
BUFFER_TRACE(parent_bh, "free data blocks");
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_free_data(handle, inode, parent_bh, first, last);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
int ext4_can_truncate(struct inode *inode)
|
|
|
|
{
|
|
|
|
if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
|
|
|
|
return 0;
|
|
|
|
if (S_ISREG(inode->i_mode))
|
|
|
|
return 1;
|
|
|
|
if (S_ISDIR(inode->i_mode))
|
|
|
|
return 1;
|
|
|
|
if (S_ISLNK(inode->i_mode))
|
|
|
|
return !ext4_inode_is_fast_symlink(inode);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_truncate()
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
2006-10-11 10:20:53 +02:00
|
|
|
* We block out ext4_get_block() block instantiations across the entire
|
|
|
|
* transaction, and VFS/VM ensures that ext4_truncate() cannot run
|
2006-10-11 10:20:50 +02:00
|
|
|
* simultaneously on behalf of the same inode.
|
|
|
|
*
|
|
|
|
* As we work through the truncate and commmit bits of it to the journal there
|
|
|
|
* is one core, guiding principle: the file's tree must always be consistent on
|
|
|
|
* disk. We must be able to restart the truncate after a crash.
|
|
|
|
*
|
|
|
|
* The file's tree may be transiently inconsistent in memory (although it
|
|
|
|
* probably isn't), but whenever we close off and commit a journal transaction,
|
|
|
|
* the contents of (the filesystem + the journal) must be consistent and
|
|
|
|
* restartable. It's pretty simple, really: bottom up, right to left (although
|
|
|
|
* left-to-right works OK too).
|
|
|
|
*
|
|
|
|
* Note that at recovery time, journal replay occurs *before* the restart of
|
|
|
|
* truncate against the orphan inode list.
|
|
|
|
*
|
|
|
|
* The committed inode has the new, desired i_size (which is the same as
|
2006-10-11 10:20:53 +02:00
|
|
|
* i_disksize in this case). After a crash, ext4_orphan_cleanup() will see
|
2006-10-11 10:20:50 +02:00
|
|
|
* that this inode's truncate did not complete and it will again call
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_truncate() to have another go. So there will be instantiated blocks
|
|
|
|
* to the right of the truncation point in a crashed ext4 filesystem. But
|
2006-10-11 10:20:50 +02:00
|
|
|
* that's fine - as long as they are linked from the inode, the post-crash
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_truncate() run will find them and release them.
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
void ext4_truncate(struct inode *inode)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
handle_t *handle;
|
2006-10-11 10:20:53 +02:00
|
|
|
struct ext4_inode_info *ei = EXT4_I(inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
__le32 *i_data = ei->i_data;
|
2006-10-11 10:20:53 +02:00
|
|
|
int addr_per_block = EXT4_ADDR_PER_BLOCK(inode->i_sb);
|
2006-10-11 10:20:50 +02:00
|
|
|
struct address_space *mapping = inode->i_mapping;
|
2008-01-29 05:58:27 +01:00
|
|
|
ext4_lblk_t offsets[4];
|
2006-10-11 10:20:50 +02:00
|
|
|
Indirect chain[4];
|
|
|
|
Indirect *partial;
|
|
|
|
__le32 nr = 0;
|
|
|
|
int n;
|
2008-01-29 05:58:27 +01:00
|
|
|
ext4_lblk_t last_block;
|
2006-10-11 10:20:50 +02:00
|
|
|
unsigned blocksize = inode->i_sb->s_blocksize;
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
if (!ext4_can_truncate(inode))
|
2006-10-11 10:20:50 +02:00
|
|
|
return;
|
|
|
|
|
2010-05-17 04:00:00 +02:00
|
|
|
ext4_clear_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
|
2010-02-24 15:52:53 +01:00
|
|
|
|
2009-09-17 15:34:16 +02:00
|
|
|
if (inode->i_size == 0 && !test_opt(inode->i_sb, NO_AUTO_DA_ALLOC))
|
2010-01-24 20:34:07 +01:00
|
|
|
ext4_set_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE);
|
2009-02-24 14:21:14 +01:00
|
|
|
|
2010-05-17 04:00:00 +02:00
|
|
|
if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
|
2008-07-12 01:27:31 +02:00
|
|
|
ext4_ext_truncate(inode);
|
2008-01-29 05:58:27 +01:00
|
|
|
return;
|
|
|
|
}
|
2006-10-11 10:21:03 +02:00
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
handle = start_transaction(inode);
|
2008-07-12 01:27:31 +02:00
|
|
|
if (IS_ERR(handle))
|
2006-10-11 10:20:50 +02:00
|
|
|
return; /* AKPM: return what? */
|
|
|
|
|
|
|
|
last_block = (inode->i_size + blocksize-1)
|
2006-10-11 10:20:53 +02:00
|
|
|
>> EXT4_BLOCK_SIZE_BITS(inode->i_sb);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
if (inode->i_size & (blocksize - 1))
|
|
|
|
if (ext4_block_truncate_page(handle, mapping, inode->i_size))
|
|
|
|
goto out_stop;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
n = ext4_block_to_path(inode, last_block, offsets, NULL);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (n == 0)
|
|
|
|
goto out_stop; /* error */
|
|
|
|
|
|
|
|
/*
|
|
|
|
* OK. This truncate is going to happen. We add the inode to the
|
|
|
|
* orphan list, so that if this truncate spans multiple transactions,
|
|
|
|
* and we crash, we will resume the truncate when the filesystem
|
|
|
|
* recovers. It also marks the inode dirty, to catch the new size.
|
|
|
|
*
|
|
|
|
* Implication: the file must always be in a sane, consistent
|
|
|
|
* truncatable state while each transaction commits.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
if (ext4_orphan_add(handle, inode))
|
2006-10-11 10:20:50 +02:00
|
|
|
goto out_stop;
|
|
|
|
|
ext4: fix delalloc i_disksize early update issue
Ext4_da_write_end() used walk_page_buffers() with a callback function of
ext4_bh_unmapped_or_delay() to check if it extended the file size
without allocating any blocks (since in this case i_disksize needs to be
updated). However, this is didn't work proprely because the buffer head
has not been marked dirty yet --- this is done later in
block_commit_write() --- which caused ext4_bh_unmapped_or_delay() to
always return false.
In addition, walk_page_buffers() checks all of the buffer heads covering
the page, and the only buffer_head that should be checked is the one
covering the end of the write. Otherwise, given a 1k blocksize
filesystem and a 4k page size, the buffer head covering the first 1k
stripe of the file could be unmapped (because it was a sparse file), and
the second or third buffer_head covering that page could be mapped, and
using walk_page_buffers() would fail in this case since it would stop at
the first unmapped buffer_head and return true.
The core problem is that walk_page_buffers() was intended to do work in
a callback function, and a non-zero return value indicated a failure,
which termined the walk of the buffer heads covering the page. It was
not intended to be used with a boolean function, such as
ext4_bh_unmapped_or_delay().
Add addtional fix from Aneesh to protect i_disksize update rave with truncate.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* From here we block out all ext4_get_block() callers who want to
|
|
|
|
* modify the block allocation tree.
|
|
|
|
*/
|
|
|
|
down_write(&ei->i_data_sem);
|
2008-08-14 03:44:34 +02:00
|
|
|
|
2008-10-10 15:40:52 +02:00
|
|
|
ext4_discard_preallocations(inode);
|
2008-08-14 03:44:34 +02:00
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* The orphan list entry will now protect us from any crash which
|
|
|
|
* occurs before the truncate completes, so it is now safe to propagate
|
|
|
|
* the new, shorter inode size (held for now in i_size) into the
|
|
|
|
* on-disk inode. We do this via i_disksize, which is the value which
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4 *really* writes onto the disk inode.
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
|
|
|
ei->i_disksize = inode->i_size;
|
|
|
|
|
|
|
|
if (n == 1) { /* direct blocks */
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_free_data(handle, inode, NULL, i_data+offsets[0],
|
|
|
|
i_data + EXT4_NDIR_BLOCKS);
|
2006-10-11 10:20:50 +02:00
|
|
|
goto do_indirects;
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
partial = ext4_find_shared(inode, n, offsets, chain, &nr);
|
2006-10-11 10:20:50 +02:00
|
|
|
/* Kill the top of shared branch (not detached) */
|
|
|
|
if (nr) {
|
|
|
|
if (partial == chain) {
|
|
|
|
/* Shared branch grows from the inode */
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_free_branches(handle, inode, NULL,
|
2006-10-11 10:20:50 +02:00
|
|
|
&nr, &nr+1, (chain+n-1) - partial);
|
|
|
|
*partial->p = 0;
|
|
|
|
/*
|
|
|
|
* We mark the inode dirty prior to restart,
|
|
|
|
* and prior to stop. No need for it here.
|
|
|
|
*/
|
|
|
|
} else {
|
|
|
|
/* Shared branch grows from an indirect block */
|
|
|
|
BUFFER_TRACE(partial->bh, "get_write_access");
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_free_branches(handle, inode, partial->bh,
|
2006-10-11 10:20:50 +02:00
|
|
|
partial->p,
|
|
|
|
partial->p+1, (chain+n-1) - partial);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
/* Clear the ends of indirect blocks on the shared branch */
|
|
|
|
while (partial > chain) {
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_free_branches(handle, inode, partial->bh, partial->p + 1,
|
2006-10-11 10:20:50 +02:00
|
|
|
(__le32*)partial->bh->b_data+addr_per_block,
|
|
|
|
(chain+n-1) - partial);
|
|
|
|
BUFFER_TRACE(partial->bh, "call brelse");
|
2009-06-14 23:45:34 +02:00
|
|
|
brelse(partial->bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
partial--;
|
|
|
|
}
|
|
|
|
do_indirects:
|
|
|
|
/* Kill the remaining (whole) subtrees */
|
|
|
|
switch (offsets[0]) {
|
|
|
|
default:
|
2006-10-11 10:20:53 +02:00
|
|
|
nr = i_data[EXT4_IND_BLOCK];
|
2006-10-11 10:20:50 +02:00
|
|
|
if (nr) {
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_free_branches(handle, inode, NULL, &nr, &nr+1, 1);
|
|
|
|
i_data[EXT4_IND_BLOCK] = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2006-10-11 10:20:53 +02:00
|
|
|
case EXT4_IND_BLOCK:
|
|
|
|
nr = i_data[EXT4_DIND_BLOCK];
|
2006-10-11 10:20:50 +02:00
|
|
|
if (nr) {
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_free_branches(handle, inode, NULL, &nr, &nr+1, 2);
|
|
|
|
i_data[EXT4_DIND_BLOCK] = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2006-10-11 10:20:53 +02:00
|
|
|
case EXT4_DIND_BLOCK:
|
|
|
|
nr = i_data[EXT4_TIND_BLOCK];
|
2006-10-11 10:20:50 +02:00
|
|
|
if (nr) {
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_free_branches(handle, inode, NULL, &nr, &nr+1, 3);
|
|
|
|
i_data[EXT4_TIND_BLOCK] = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2006-10-11 10:20:53 +02:00
|
|
|
case EXT4_TIND_BLOCK:
|
2006-10-11 10:20:50 +02:00
|
|
|
;
|
|
|
|
}
|
|
|
|
|
2008-01-29 05:58:26 +01:00
|
|
|
up_write(&ei->i_data_sem);
|
2007-07-18 15:15:20 +02:00
|
|
|
inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_mark_inode_dirty(handle, inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* In a multi-transaction truncate, we only make the final transaction
|
|
|
|
* synchronous
|
|
|
|
*/
|
|
|
|
if (IS_SYNC(inode))
|
2009-01-07 06:06:22 +01:00
|
|
|
ext4_handle_sync(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
out_stop:
|
|
|
|
/*
|
|
|
|
* If this was a simple ftruncate(), and the file will remain alive
|
|
|
|
* then we need to clear up the orphan record which we created above.
|
|
|
|
* However, if this was a real unlink then we were called by
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_delete_inode(), and we allow that function to clean up the
|
2006-10-11 10:20:50 +02:00
|
|
|
* orphan info for us.
|
|
|
|
*/
|
|
|
|
if (inode->i_nlink)
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_orphan_del(handle, inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_journal_stop(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_get_inode_loc returns with an extra refcount against the inode's
|
2006-10-11 10:20:50 +02:00
|
|
|
* underlying buffer_head on success. If 'in_mem' is true, we have all
|
|
|
|
* data in memory that is needed to recreate the on-disk version of this
|
|
|
|
* inode.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static int __ext4_get_inode_loc(struct inode *inode,
|
|
|
|
struct ext4_iloc *iloc, int in_mem)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2008-10-10 05:53:47 +02:00
|
|
|
struct ext4_group_desc *gdp;
|
|
|
|
struct buffer_head *bh;
|
|
|
|
struct super_block *sb = inode->i_sb;
|
|
|
|
ext4_fsblk_t block;
|
|
|
|
int inodes_per_block, inode_offset;
|
|
|
|
|
2008-11-22 21:04:59 +01:00
|
|
|
iloc->bh = NULL;
|
2008-10-10 05:53:47 +02:00
|
|
|
if (!ext4_valid_inum(sb, inode->i_ino))
|
|
|
|
return -EIO;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2008-10-10 05:53:47 +02:00
|
|
|
iloc->block_group = (inode->i_ino - 1) / EXT4_INODES_PER_GROUP(sb);
|
|
|
|
gdp = ext4_get_group_desc(sb, iloc->block_group, NULL);
|
|
|
|
if (!gdp)
|
2006-10-11 10:20:50 +02:00
|
|
|
return -EIO;
|
|
|
|
|
2008-10-10 05:53:47 +02:00
|
|
|
/*
|
|
|
|
* Figure out the offset within the block group inode table
|
|
|
|
*/
|
|
|
|
inodes_per_block = (EXT4_BLOCK_SIZE(sb) / EXT4_INODE_SIZE(sb));
|
|
|
|
inode_offset = ((inode->i_ino - 1) %
|
|
|
|
EXT4_INODES_PER_GROUP(sb));
|
|
|
|
block = ext4_inode_table(sb, gdp) + (inode_offset / inodes_per_block);
|
|
|
|
iloc->offset = (inode_offset % inodes_per_block) * EXT4_INODE_SIZE(sb);
|
|
|
|
|
|
|
|
bh = sb_getblk(sb, block);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!bh) {
|
2010-05-17 03:00:00 +02:00
|
|
|
EXT4_ERROR_INODE(inode, "unable to read inode block - "
|
|
|
|
"block %llu", block);
|
2006-10-11 10:20:50 +02:00
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
if (!buffer_uptodate(bh)) {
|
|
|
|
lock_buffer(bh);
|
2008-07-26 22:39:26 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If the buffer has the write error flag, we have failed
|
|
|
|
* to write out another inode in the same block. In this
|
|
|
|
* case, we don't have to read the block because we may
|
|
|
|
* read the old inode data successfully.
|
|
|
|
*/
|
|
|
|
if (buffer_write_io_error(bh) && !buffer_uptodate(bh))
|
|
|
|
set_buffer_uptodate(bh);
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
if (buffer_uptodate(bh)) {
|
|
|
|
/* someone brought it uptodate while we waited */
|
|
|
|
unlock_buffer(bh);
|
|
|
|
goto has_buffer;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we have all information of the inode in memory and this
|
|
|
|
* is the only valid inode in the block, we need not read the
|
|
|
|
* block.
|
|
|
|
*/
|
|
|
|
if (in_mem) {
|
|
|
|
struct buffer_head *bitmap_bh;
|
2008-10-10 05:53:47 +02:00
|
|
|
int i, start;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2008-10-10 05:53:47 +02:00
|
|
|
start = inode_offset & ~(inodes_per_block - 1);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2008-10-10 05:53:47 +02:00
|
|
|
/* Is the inode bitmap in cache? */
|
|
|
|
bitmap_bh = sb_getblk(sb, ext4_inode_bitmap(sb, gdp));
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!bitmap_bh)
|
|
|
|
goto make_io;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the inode bitmap isn't in cache then the
|
|
|
|
* optimisation may end up performing two reads instead
|
|
|
|
* of one, so skip it.
|
|
|
|
*/
|
|
|
|
if (!buffer_uptodate(bitmap_bh)) {
|
|
|
|
brelse(bitmap_bh);
|
|
|
|
goto make_io;
|
|
|
|
}
|
2008-10-10 05:53:47 +02:00
|
|
|
for (i = start; i < start + inodes_per_block; i++) {
|
2006-10-11 10:20:50 +02:00
|
|
|
if (i == inode_offset)
|
|
|
|
continue;
|
2006-10-11 10:20:53 +02:00
|
|
|
if (ext4_test_bit(i, bitmap_bh->b_data))
|
2006-10-11 10:20:50 +02:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
brelse(bitmap_bh);
|
2008-10-10 05:53:47 +02:00
|
|
|
if (i == start + inodes_per_block) {
|
2006-10-11 10:20:50 +02:00
|
|
|
/* all other inodes are free, so skip I/O */
|
|
|
|
memset(bh->b_data, 0, bh->b_size);
|
|
|
|
set_buffer_uptodate(bh);
|
|
|
|
unlock_buffer(bh);
|
|
|
|
goto has_buffer;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
make_io:
|
2008-10-10 05:53:47 +02:00
|
|
|
/*
|
|
|
|
* If we need to do any I/O, try to pre-readahead extra
|
|
|
|
* blocks from the inode table.
|
|
|
|
*/
|
|
|
|
if (EXT4_SB(sb)->s_inode_readahead_blks) {
|
|
|
|
ext4_fsblk_t b, end, table;
|
|
|
|
unsigned num;
|
|
|
|
|
|
|
|
table = ext4_inode_table(sb, gdp);
|
2009-03-31 15:11:14 +02:00
|
|
|
/* s_inode_readahead_blks is always a power of 2 */
|
2008-10-10 05:53:47 +02:00
|
|
|
b = block & ~(EXT4_SB(sb)->s_inode_readahead_blks-1);
|
|
|
|
if (table > b)
|
|
|
|
b = table;
|
|
|
|
end = b + EXT4_SB(sb)->s_inode_readahead_blks;
|
|
|
|
num = EXT4_INODES_PER_GROUP(sb);
|
|
|
|
if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
|
|
|
|
EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
|
2009-01-06 04:20:24 +01:00
|
|
|
num -= ext4_itable_unused_count(sb, gdp);
|
2008-10-10 05:53:47 +02:00
|
|
|
table += num / inodes_per_block;
|
|
|
|
if (end > table)
|
|
|
|
end = table;
|
|
|
|
while (b <= end)
|
|
|
|
sb_breadahead(sb, b++);
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* There are other valid inodes in the buffer, this inode
|
|
|
|
* has in-inode xattrs, or we don't have this inode in memory.
|
|
|
|
* Read the block from disk.
|
|
|
|
*/
|
|
|
|
get_bh(bh);
|
|
|
|
bh->b_end_io = end_buffer_read_sync;
|
|
|
|
submit_bh(READ_META, bh);
|
|
|
|
wait_on_buffer(bh);
|
|
|
|
if (!buffer_uptodate(bh)) {
|
2010-05-17 03:00:00 +02:00
|
|
|
EXT4_ERROR_INODE(inode, "unable to read inode "
|
|
|
|
"block %llu", block);
|
2006-10-11 10:20:50 +02:00
|
|
|
brelse(bh);
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
has_buffer:
|
|
|
|
iloc->bh = bh;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
int ext4_get_inode_loc(struct inode *inode, struct ext4_iloc *iloc)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
/* We have all inode data except xattrs in memory here. */
|
2006-10-11 10:20:53 +02:00
|
|
|
return __ext4_get_inode_loc(inode, iloc,
|
2010-01-24 20:34:07 +01:00
|
|
|
!ext4_test_inode_state(inode, EXT4_STATE_XATTR));
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
void ext4_set_inode_flags(struct inode *inode)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
unsigned int flags = EXT4_I(inode)->i_flags;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
|
2006-10-11 10:20:53 +02:00
|
|
|
if (flags & EXT4_SYNC_FL)
|
2006-10-11 10:20:50 +02:00
|
|
|
inode->i_flags |= S_SYNC;
|
2006-10-11 10:20:53 +02:00
|
|
|
if (flags & EXT4_APPEND_FL)
|
2006-10-11 10:20:50 +02:00
|
|
|
inode->i_flags |= S_APPEND;
|
2006-10-11 10:20:53 +02:00
|
|
|
if (flags & EXT4_IMMUTABLE_FL)
|
2006-10-11 10:20:50 +02:00
|
|
|
inode->i_flags |= S_IMMUTABLE;
|
2006-10-11 10:20:53 +02:00
|
|
|
if (flags & EXT4_NOATIME_FL)
|
2006-10-11 10:20:50 +02:00
|
|
|
inode->i_flags |= S_NOATIME;
|
2006-10-11 10:20:53 +02:00
|
|
|
if (flags & EXT4_DIRSYNC_FL)
|
2006-10-11 10:20:50 +02:00
|
|
|
inode->i_flags |= S_DIRSYNC;
|
|
|
|
}
|
|
|
|
|
2007-07-18 15:24:20 +02:00
|
|
|
/* Propagate flags from i_flags to EXT4_I(inode)->i_flags */
|
|
|
|
void ext4_get_inode_flags(struct ext4_inode_info *ei)
|
|
|
|
{
|
2010-06-05 17:51:27 +02:00
|
|
|
unsigned int vfs_fl;
|
|
|
|
unsigned long old_fl, new_fl;
|
|
|
|
|
|
|
|
do {
|
|
|
|
vfs_fl = ei->vfs_inode.i_flags;
|
|
|
|
old_fl = ei->i_flags;
|
|
|
|
new_fl = old_fl & ~(EXT4_SYNC_FL|EXT4_APPEND_FL|
|
|
|
|
EXT4_IMMUTABLE_FL|EXT4_NOATIME_FL|
|
|
|
|
EXT4_DIRSYNC_FL);
|
|
|
|
if (vfs_fl & S_SYNC)
|
|
|
|
new_fl |= EXT4_SYNC_FL;
|
|
|
|
if (vfs_fl & S_APPEND)
|
|
|
|
new_fl |= EXT4_APPEND_FL;
|
|
|
|
if (vfs_fl & S_IMMUTABLE)
|
|
|
|
new_fl |= EXT4_IMMUTABLE_FL;
|
|
|
|
if (vfs_fl & S_NOATIME)
|
|
|
|
new_fl |= EXT4_NOATIME_FL;
|
|
|
|
if (vfs_fl & S_DIRSYNC)
|
|
|
|
new_fl |= EXT4_DIRSYNC_FL;
|
|
|
|
} while (cmpxchg(&ei->i_flags, old_fl, new_fl) != old_fl);
|
2007-07-18 15:24:20 +02:00
|
|
|
}
|
2009-06-14 23:45:34 +02:00
|
|
|
|
2008-01-29 05:58:26 +01:00
|
|
|
static blkcnt_t ext4_inode_blocks(struct ext4_inode *raw_inode,
|
2009-06-14 23:45:34 +02:00
|
|
|
struct ext4_inode_info *ei)
|
2008-01-29 05:58:26 +01:00
|
|
|
{
|
|
|
|
blkcnt_t i_blocks ;
|
2008-01-29 05:58:27 +01:00
|
|
|
struct inode *inode = &(ei->vfs_inode);
|
|
|
|
struct super_block *sb = inode->i_sb;
|
2008-01-29 05:58:26 +01:00
|
|
|
|
|
|
|
if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
|
|
|
|
EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
|
|
|
|
/* we are using combined 48 bit field */
|
|
|
|
i_blocks = ((u64)le16_to_cpu(raw_inode->i_blocks_high)) << 32 |
|
|
|
|
le32_to_cpu(raw_inode->i_blocks_lo);
|
2008-01-29 05:58:27 +01:00
|
|
|
if (ei->i_flags & EXT4_HUGE_FILE_FL) {
|
|
|
|
/* i_blocks represent file system block size */
|
|
|
|
return i_blocks << (inode->i_blkbits - 9);
|
|
|
|
} else {
|
|
|
|
return i_blocks;
|
|
|
|
}
|
2008-01-29 05:58:26 +01:00
|
|
|
} else {
|
|
|
|
return le32_to_cpu(raw_inode->i_blocks_lo);
|
|
|
|
}
|
|
|
|
}
|
2007-07-18 15:24:20 +02:00
|
|
|
|
2008-02-07 09:15:37 +01:00
|
|
|
struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
struct ext4_iloc iloc;
|
|
|
|
struct ext4_inode *raw_inode;
|
2008-02-07 09:15:37 +01:00
|
|
|
struct ext4_inode_info *ei;
|
|
|
|
struct inode *inode;
|
2009-12-09 05:51:10 +01:00
|
|
|
journal_t *journal = EXT4_SB(sb)->s_journal;
|
2008-02-07 09:15:37 +01:00
|
|
|
long ret;
|
2006-10-11 10:20:50 +02:00
|
|
|
int block;
|
|
|
|
|
2008-02-07 09:15:37 +01:00
|
|
|
inode = iget_locked(sb, ino);
|
|
|
|
if (!inode)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
if (!(inode->i_state & I_NEW))
|
|
|
|
return inode;
|
|
|
|
|
|
|
|
ei = EXT4_I(inode);
|
2009-11-14 14:19:05 +01:00
|
|
|
iloc.bh = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2008-02-07 09:15:37 +01:00
|
|
|
ret = __ext4_get_inode_loc(inode, &iloc, 0);
|
|
|
|
if (ret < 0)
|
2006-10-11 10:20:50 +02:00
|
|
|
goto bad_inode;
|
2006-10-11 10:20:53 +02:00
|
|
|
raw_inode = ext4_raw_inode(&iloc);
|
2006-10-11 10:20:50 +02:00
|
|
|
inode->i_mode = le16_to_cpu(raw_inode->i_mode);
|
|
|
|
inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
|
|
|
|
inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
|
2008-09-09 04:25:24 +02:00
|
|
|
if (!(test_opt(inode->i_sb, NO_UID32))) {
|
2006-10-11 10:20:50 +02:00
|
|
|
inode->i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
|
|
|
|
inode->i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
|
|
|
|
}
|
|
|
|
inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);
|
|
|
|
|
2010-01-24 20:34:07 +01:00
|
|
|
ei->i_state_flags = 0;
|
2006-10-11 10:20:50 +02:00
|
|
|
ei->i_dir_start_lookup = 0;
|
|
|
|
ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
|
|
|
|
/* We now have enough fields to check if the inode was active or not.
|
|
|
|
* This is needed because nfsd might try to access dead inodes
|
|
|
|
* the test is that same one that e2fsck uses
|
|
|
|
* NeilBrown 1999oct15
|
|
|
|
*/
|
|
|
|
if (inode->i_nlink == 0) {
|
|
|
|
if (inode->i_mode == 0 ||
|
2006-10-11 10:20:53 +02:00
|
|
|
!(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_ORPHAN_FS)) {
|
2006-10-11 10:20:50 +02:00
|
|
|
/* this inode is deleted */
|
2008-02-07 09:15:37 +01:00
|
|
|
ret = -ESTALE;
|
2006-10-11 10:20:50 +02:00
|
|
|
goto bad_inode;
|
|
|
|
}
|
|
|
|
/* The only unlinked inodes we let through here have
|
|
|
|
* valid i_mode and are being read by the orphan
|
|
|
|
* recovery code: that's fine, we're about to complete
|
|
|
|
* the process of deleting those. */
|
|
|
|
}
|
|
|
|
ei->i_flags = le32_to_cpu(raw_inode->i_flags);
|
2008-01-29 05:58:26 +01:00
|
|
|
inode->i_blocks = ext4_inode_blocks(raw_inode, ei);
|
2008-01-29 05:58:27 +01:00
|
|
|
ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl_lo);
|
2009-04-24 22:11:18 +02:00
|
|
|
if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT))
|
2006-10-11 10:21:09 +02:00
|
|
|
ei->i_file_acl |=
|
|
|
|
((__u64)le16_to_cpu(raw_inode->i_file_acl_high)) << 32;
|
2008-01-29 05:58:27 +01:00
|
|
|
inode->i_size = ext4_isize(raw_inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
ei->i_disksize = inode->i_size;
|
2009-12-14 13:21:14 +01:00
|
|
|
#ifdef CONFIG_QUOTA
|
|
|
|
ei->i_reserved_quota = 0;
|
|
|
|
#endif
|
2006-10-11 10:20:50 +02:00
|
|
|
inode->i_generation = le32_to_cpu(raw_inode->i_generation);
|
|
|
|
ei->i_block_group = iloc.block_group;
|
2009-03-12 17:18:34 +01:00
|
|
|
ei->i_last_alloc_group = ~0;
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* NOTE! The in-memory inode i_data array is in little-endian order
|
|
|
|
* even on big-endian machines: we do NOT byteswap the block numbers!
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
for (block = 0; block < EXT4_N_BLOCKS; block++)
|
2006-10-11 10:20:50 +02:00
|
|
|
ei->i_data[block] = raw_inode->i_block[block];
|
|
|
|
INIT_LIST_HEAD(&ei->i_orphan);
|
|
|
|
|
2009-12-09 05:51:10 +01:00
|
|
|
/*
|
|
|
|
* Set transaction id's of transactions that have to be committed
|
|
|
|
* to finish f[data]sync. We set them to currently running transaction
|
|
|
|
* as we cannot be sure that the inode or some of its metadata isn't
|
|
|
|
* part of the transaction - the inode could have been reclaimed and
|
|
|
|
* now it is reread from disk.
|
|
|
|
*/
|
|
|
|
if (journal) {
|
|
|
|
transaction_t *transaction;
|
|
|
|
tid_t tid;
|
|
|
|
|
|
|
|
spin_lock(&journal->j_state_lock);
|
|
|
|
if (journal->j_running_transaction)
|
|
|
|
transaction = journal->j_running_transaction;
|
|
|
|
else
|
|
|
|
transaction = journal->j_committing_transaction;
|
|
|
|
if (transaction)
|
|
|
|
tid = transaction->t_tid;
|
|
|
|
else
|
|
|
|
tid = journal->j_commit_sequence;
|
|
|
|
spin_unlock(&journal->j_state_lock);
|
|
|
|
ei->i_sync_tid = tid;
|
|
|
|
ei->i_datasync_tid = tid;
|
|
|
|
}
|
|
|
|
|
2008-02-06 04:36:43 +01:00
|
|
|
if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
|
2006-10-11 10:20:50 +02:00
|
|
|
ei->i_extra_isize = le16_to_cpu(raw_inode->i_extra_isize);
|
2006-10-11 10:20:53 +02:00
|
|
|
if (EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize >
|
2007-06-24 02:16:51 +02:00
|
|
|
EXT4_INODE_SIZE(inode->i_sb)) {
|
2008-02-07 09:15:37 +01:00
|
|
|
ret = -EIO;
|
2006-10-11 10:20:50 +02:00
|
|
|
goto bad_inode;
|
2007-06-24 02:16:51 +02:00
|
|
|
}
|
2006-10-11 10:20:50 +02:00
|
|
|
if (ei->i_extra_isize == 0) {
|
|
|
|
/* The extra space is currently unused. Use it. */
|
2006-10-11 10:20:53 +02:00
|
|
|
ei->i_extra_isize = sizeof(struct ext4_inode) -
|
|
|
|
EXT4_GOOD_OLD_INODE_SIZE;
|
2006-10-11 10:20:50 +02:00
|
|
|
} else {
|
|
|
|
__le32 *magic = (void *)raw_inode +
|
2006-10-11 10:20:53 +02:00
|
|
|
EXT4_GOOD_OLD_INODE_SIZE +
|
2006-10-11 10:20:50 +02:00
|
|
|
ei->i_extra_isize;
|
2006-10-11 10:20:53 +02:00
|
|
|
if (*magic == cpu_to_le32(EXT4_XATTR_MAGIC))
|
2010-01-24 20:34:07 +01:00
|
|
|
ext4_set_inode_state(inode, EXT4_STATE_XATTR);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
} else
|
|
|
|
ei->i_extra_isize = 0;
|
|
|
|
|
2007-07-18 15:15:20 +02:00
|
|
|
EXT4_INODE_GET_XTIME(i_ctime, inode, raw_inode);
|
|
|
|
EXT4_INODE_GET_XTIME(i_mtime, inode, raw_inode);
|
|
|
|
EXT4_INODE_GET_XTIME(i_atime, inode, raw_inode);
|
|
|
|
EXT4_EINODE_GET_XTIME(i_crtime, ei, raw_inode);
|
|
|
|
|
2008-01-29 05:58:27 +01:00
|
|
|
inode->i_version = le32_to_cpu(raw_inode->i_disk_version);
|
|
|
|
if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
|
|
|
|
if (EXT4_FITS_IN_INODE(raw_inode, ei, i_version_hi))
|
|
|
|
inode->i_version |=
|
|
|
|
(__u64)(le32_to_cpu(raw_inode->i_version_hi)) << 32;
|
|
|
|
}
|
|
|
|
|
ext4: Do not try to validate extents on special files
The EXTENTS_FL flag should never be set on special files, but if it
is, don't bother trying to validate that the extents tree is valid,
since only files, directories, and non-fast symlinks will ever have an
extent data structure. We perhaps should flag the filesystem as being
corrupted if we see a special file (named pipes, device nodes, Unix
domain sockets, etc.) with the EXTENTS_FL flag, but e2fsck doesn't
currently check this case, so we'll just ignore this for now, since
it's harmless.
Without this fix, a special device with the extents flag is flagged as
an error by the kernel, so it is impossible to access or delete the
inode, but e2fsck doesn't see it as a problem, leading to
confused/frustrated users.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2009-04-25 00:45:35 +02:00
|
|
|
ret = 0;
|
2009-04-24 19:43:20 +02:00
|
|
|
if (ei->i_file_acl &&
|
2009-11-15 21:29:56 +01:00
|
|
|
!ext4_data_block_valid(EXT4_SB(sb), ei->i_file_acl, 1)) {
|
2010-05-17 03:00:00 +02:00
|
|
|
EXT4_ERROR_INODE(inode, "bad extended attribute block %llu",
|
|
|
|
ei->i_file_acl);
|
2009-04-24 19:43:20 +02:00
|
|
|
ret = -EIO;
|
|
|
|
goto bad_inode;
|
|
|
|
} else if (ei->i_flags & EXT4_EXTENTS_FL) {
|
ext4: Do not try to validate extents on special files
The EXTENTS_FL flag should never be set on special files, but if it
is, don't bother trying to validate that the extents tree is valid,
since only files, directories, and non-fast symlinks will ever have an
extent data structure. We perhaps should flag the filesystem as being
corrupted if we see a special file (named pipes, device nodes, Unix
domain sockets, etc.) with the EXTENTS_FL flag, but e2fsck doesn't
currently check this case, so we'll just ignore this for now, since
it's harmless.
Without this fix, a special device with the extents flag is flagged as
an error by the kernel, so it is impossible to access or delete the
inode, but e2fsck doesn't see it as a problem, leading to
confused/frustrated users.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2009-04-25 00:45:35 +02:00
|
|
|
if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
|
|
|
|
(S_ISLNK(inode->i_mode) &&
|
|
|
|
!ext4_inode_is_fast_symlink(inode)))
|
|
|
|
/* Validate extent which is part of inode */
|
|
|
|
ret = ext4_ext_check_inode(inode);
|
2009-06-14 23:45:34 +02:00
|
|
|
} else if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
|
2009-03-31 14:36:10 +02:00
|
|
|
(S_ISLNK(inode->i_mode) &&
|
|
|
|
!ext4_inode_is_fast_symlink(inode))) {
|
2009-06-14 23:45:34 +02:00
|
|
|
/* Validate block references which are part of inode */
|
2009-03-31 14:36:10 +02:00
|
|
|
ret = ext4_check_inode_blockref(inode);
|
|
|
|
}
|
2009-11-14 14:19:05 +01:00
|
|
|
if (ret)
|
2009-06-14 23:45:34 +02:00
|
|
|
goto bad_inode;
|
2009-03-27 21:39:58 +01:00
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
if (S_ISREG(inode->i_mode)) {
|
2006-10-11 10:20:53 +02:00
|
|
|
inode->i_op = &ext4_file_inode_operations;
|
|
|
|
inode->i_fop = &ext4_file_operations;
|
|
|
|
ext4_set_aops(inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
} else if (S_ISDIR(inode->i_mode)) {
|
2006-10-11 10:20:53 +02:00
|
|
|
inode->i_op = &ext4_dir_inode_operations;
|
|
|
|
inode->i_fop = &ext4_dir_operations;
|
2006-10-11 10:20:50 +02:00
|
|
|
} else if (S_ISLNK(inode->i_mode)) {
|
2008-12-19 21:47:15 +01:00
|
|
|
if (ext4_inode_is_fast_symlink(inode)) {
|
2006-10-11 10:20:53 +02:00
|
|
|
inode->i_op = &ext4_fast_symlink_inode_operations;
|
2008-12-19 21:47:15 +01:00
|
|
|
nd_terminate_link(ei->i_data, inode->i_size,
|
|
|
|
sizeof(ei->i_data) - 1);
|
|
|
|
} else {
|
2006-10-11 10:20:53 +02:00
|
|
|
inode->i_op = &ext4_symlink_inode_operations;
|
|
|
|
ext4_set_aops(inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2009-03-26 05:06:19 +01:00
|
|
|
} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
|
|
|
|
S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
|
2006-10-11 10:20:53 +02:00
|
|
|
inode->i_op = &ext4_special_inode_operations;
|
2006-10-11 10:20:50 +02:00
|
|
|
if (raw_inode->i_block[0])
|
|
|
|
init_special_inode(inode, inode->i_mode,
|
|
|
|
old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
|
|
|
|
else
|
|
|
|
init_special_inode(inode, inode->i_mode,
|
|
|
|
new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
|
2009-03-26 05:06:19 +01:00
|
|
|
} else {
|
|
|
|
ret = -EIO;
|
2010-05-17 03:00:00 +02:00
|
|
|
EXT4_ERROR_INODE(inode, "bogus i_mode (%o)", inode->i_mode);
|
2009-03-26 05:06:19 +01:00
|
|
|
goto bad_inode;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
2008-09-09 04:25:24 +02:00
|
|
|
brelse(iloc.bh);
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_set_inode_flags(inode);
|
2008-02-07 09:15:37 +01:00
|
|
|
unlock_new_inode(inode);
|
|
|
|
return inode;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
bad_inode:
|
2009-11-14 14:19:05 +01:00
|
|
|
brelse(iloc.bh);
|
2008-02-07 09:15:37 +01:00
|
|
|
iget_failed(inode);
|
|
|
|
return ERR_PTR(ret);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2008-01-29 05:58:26 +01:00
|
|
|
static int ext4_inode_blocks_set(handle_t *handle,
|
|
|
|
struct ext4_inode *raw_inode,
|
|
|
|
struct ext4_inode_info *ei)
|
|
|
|
{
|
|
|
|
struct inode *inode = &(ei->vfs_inode);
|
|
|
|
u64 i_blocks = inode->i_blocks;
|
|
|
|
struct super_block *sb = inode->i_sb;
|
|
|
|
|
|
|
|
if (i_blocks <= ~0U) {
|
|
|
|
/*
|
|
|
|
* i_blocks can be represnted in a 32 bit variable
|
|
|
|
* as multiple of 512 bytes
|
|
|
|
*/
|
2008-01-29 05:58:27 +01:00
|
|
|
raw_inode->i_blocks_lo = cpu_to_le32(i_blocks);
|
2008-01-29 05:58:26 +01:00
|
|
|
raw_inode->i_blocks_high = 0;
|
2010-06-05 17:51:27 +02:00
|
|
|
ext4_clear_inode_flag(inode, EXT4_INODE_HUGE_FILE);
|
2008-10-17 04:50:48 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE))
|
|
|
|
return -EFBIG;
|
|
|
|
|
|
|
|
if (i_blocks <= 0xffffffffffffULL) {
|
2008-01-29 05:58:26 +01:00
|
|
|
/*
|
|
|
|
* i_blocks can be represented in a 48 bit variable
|
|
|
|
* as multiple of 512 bytes
|
|
|
|
*/
|
2008-01-29 05:58:27 +01:00
|
|
|
raw_inode->i_blocks_lo = cpu_to_le32(i_blocks);
|
2008-01-29 05:58:26 +01:00
|
|
|
raw_inode->i_blocks_high = cpu_to_le16(i_blocks >> 32);
|
2010-06-05 17:51:27 +02:00
|
|
|
ext4_clear_inode_flag(inode, EXT4_INODE_HUGE_FILE);
|
2008-01-29 05:58:26 +01:00
|
|
|
} else {
|
2010-06-05 17:51:27 +02:00
|
|
|
ext4_set_inode_flag(inode, EXT4_INODE_HUGE_FILE);
|
2008-01-29 05:58:27 +01:00
|
|
|
/* i_block is stored in file system block size */
|
|
|
|
i_blocks = i_blocks >> (inode->i_blkbits - 9);
|
|
|
|
raw_inode->i_blocks_lo = cpu_to_le32(i_blocks);
|
|
|
|
raw_inode->i_blocks_high = cpu_to_le16(i_blocks >> 32);
|
2008-01-29 05:58:26 +01:00
|
|
|
}
|
2008-10-17 04:50:48 +02:00
|
|
|
return 0;
|
2008-01-29 05:58:26 +01:00
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* Post the struct inode info into an on-disk inode location in the
|
|
|
|
* buffer-cache. This gobbles the caller's reference to the
|
|
|
|
* buffer_head in the inode location struct.
|
|
|
|
*
|
|
|
|
* The caller must have write access to iloc->bh.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static int ext4_do_update_inode(handle_t *handle,
|
2006-10-11 10:20:50 +02:00
|
|
|
struct inode *inode,
|
2009-09-29 16:07:47 +02:00
|
|
|
struct ext4_iloc *iloc)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
struct ext4_inode *raw_inode = ext4_raw_inode(iloc);
|
|
|
|
struct ext4_inode_info *ei = EXT4_I(inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
struct buffer_head *bh = iloc->bh;
|
|
|
|
int err = 0, rc, block;
|
|
|
|
|
|
|
|
/* For fields not not tracking in the in-memory inode,
|
|
|
|
* initialise them to zero for new inodes. */
|
2010-01-24 20:34:07 +01:00
|
|
|
if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
|
2006-10-11 10:20:53 +02:00
|
|
|
memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2007-07-18 15:24:20 +02:00
|
|
|
ext4_get_inode_flags(ei);
|
2006-10-11 10:20:50 +02:00
|
|
|
raw_inode->i_mode = cpu_to_le16(inode->i_mode);
|
2008-09-09 04:25:24 +02:00
|
|
|
if (!(test_opt(inode->i_sb, NO_UID32))) {
|
2006-10-11 10:20:50 +02:00
|
|
|
raw_inode->i_uid_low = cpu_to_le16(low_16_bits(inode->i_uid));
|
|
|
|
raw_inode->i_gid_low = cpu_to_le16(low_16_bits(inode->i_gid));
|
|
|
|
/*
|
|
|
|
* Fix up interoperability with old kernels. Otherwise, old inodes get
|
|
|
|
* re-used with the upper 16 bits of the uid/gid intact
|
|
|
|
*/
|
2008-09-09 04:25:24 +02:00
|
|
|
if (!ei->i_dtime) {
|
2006-10-11 10:20:50 +02:00
|
|
|
raw_inode->i_uid_high =
|
|
|
|
cpu_to_le16(high_16_bits(inode->i_uid));
|
|
|
|
raw_inode->i_gid_high =
|
|
|
|
cpu_to_le16(high_16_bits(inode->i_gid));
|
|
|
|
} else {
|
|
|
|
raw_inode->i_uid_high = 0;
|
|
|
|
raw_inode->i_gid_high = 0;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
raw_inode->i_uid_low =
|
|
|
|
cpu_to_le16(fs_high2lowuid(inode->i_uid));
|
|
|
|
raw_inode->i_gid_low =
|
|
|
|
cpu_to_le16(fs_high2lowgid(inode->i_gid));
|
|
|
|
raw_inode->i_uid_high = 0;
|
|
|
|
raw_inode->i_gid_high = 0;
|
|
|
|
}
|
|
|
|
raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
|
2007-07-18 15:15:20 +02:00
|
|
|
|
|
|
|
EXT4_INODE_SET_XTIME(i_ctime, inode, raw_inode);
|
|
|
|
EXT4_INODE_SET_XTIME(i_mtime, inode, raw_inode);
|
|
|
|
EXT4_INODE_SET_XTIME(i_atime, inode, raw_inode);
|
|
|
|
EXT4_EINODE_SET_XTIME(i_crtime, ei, raw_inode);
|
|
|
|
|
2008-01-29 05:58:26 +01:00
|
|
|
if (ext4_inode_blocks_set(handle, raw_inode, ei))
|
|
|
|
goto out_brelse;
|
2006-10-11 10:20:50 +02:00
|
|
|
raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
|
2009-09-17 14:32:22 +02:00
|
|
|
raw_inode->i_flags = cpu_to_le32(ei->i_flags);
|
2006-10-11 10:21:13 +02:00
|
|
|
if (EXT4_SB(inode->i_sb)->s_es->s_creator_os !=
|
|
|
|
cpu_to_le32(EXT4_OS_HURD))
|
2006-10-11 10:21:09 +02:00
|
|
|
raw_inode->i_file_acl_high =
|
|
|
|
cpu_to_le16(ei->i_file_acl >> 32);
|
2008-01-29 05:58:27 +01:00
|
|
|
raw_inode->i_file_acl_lo = cpu_to_le32(ei->i_file_acl);
|
2008-01-29 05:58:27 +01:00
|
|
|
ext4_isize_set(raw_inode, ei->i_disksize);
|
|
|
|
if (ei->i_disksize > 0x7fffffffULL) {
|
|
|
|
struct super_block *sb = inode->i_sb;
|
|
|
|
if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
|
|
|
|
EXT4_FEATURE_RO_COMPAT_LARGE_FILE) ||
|
|
|
|
EXT4_SB(sb)->s_es->s_rev_level ==
|
|
|
|
cpu_to_le32(EXT4_GOOD_OLD_REV)) {
|
|
|
|
/* If this is the first large file
|
|
|
|
* created, add a flag to the superblock.
|
|
|
|
*/
|
|
|
|
err = ext4_journal_get_write_access(handle,
|
|
|
|
EXT4_SB(sb)->s_sbh);
|
|
|
|
if (err)
|
|
|
|
goto out_brelse;
|
|
|
|
ext4_update_dynamic_rev(sb);
|
|
|
|
EXT4_SET_RO_COMPAT_FEATURE(sb,
|
2006-10-11 10:20:53 +02:00
|
|
|
EXT4_FEATURE_RO_COMPAT_LARGE_FILE);
|
2008-01-29 05:58:27 +01:00
|
|
|
sb->s_dirt = 1;
|
2009-01-07 06:06:22 +01:00
|
|
|
ext4_handle_sync(handle);
|
2010-02-16 21:06:29 +01:00
|
|
|
err = ext4_handle_dirty_metadata(handle, NULL,
|
2008-01-29 05:58:27 +01:00
|
|
|
EXT4_SB(sb)->s_sbh);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
raw_inode->i_generation = cpu_to_le32(inode->i_generation);
|
|
|
|
if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
|
|
|
|
if (old_valid_dev(inode->i_rdev)) {
|
|
|
|
raw_inode->i_block[0] =
|
|
|
|
cpu_to_le32(old_encode_dev(inode->i_rdev));
|
|
|
|
raw_inode->i_block[1] = 0;
|
|
|
|
} else {
|
|
|
|
raw_inode->i_block[0] = 0;
|
|
|
|
raw_inode->i_block[1] =
|
|
|
|
cpu_to_le32(new_encode_dev(inode->i_rdev));
|
|
|
|
raw_inode->i_block[2] = 0;
|
|
|
|
}
|
2009-06-14 23:45:34 +02:00
|
|
|
} else
|
|
|
|
for (block = 0; block < EXT4_N_BLOCKS; block++)
|
|
|
|
raw_inode->i_block[block] = ei->i_data[block];
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2008-01-29 05:58:27 +01:00
|
|
|
raw_inode->i_disk_version = cpu_to_le32(inode->i_version);
|
|
|
|
if (ei->i_extra_isize) {
|
|
|
|
if (EXT4_FITS_IN_INODE(raw_inode, ei, i_version_hi))
|
|
|
|
raw_inode->i_version_hi =
|
|
|
|
cpu_to_le32(inode->i_version >> 32);
|
2006-10-11 10:20:50 +02:00
|
|
|
raw_inode->i_extra_isize = cpu_to_le16(ei->i_extra_isize);
|
2008-01-29 05:58:27 +01:00
|
|
|
}
|
|
|
|
|
2009-09-29 16:07:47 +02:00
|
|
|
BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
|
2010-02-16 21:06:29 +01:00
|
|
|
rc = ext4_handle_dirty_metadata(handle, NULL, bh);
|
2009-09-29 16:07:47 +02:00
|
|
|
if (!err)
|
|
|
|
err = rc;
|
2010-01-24 20:34:07 +01:00
|
|
|
ext4_clear_inode_state(inode, EXT4_STATE_NEW);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2009-12-09 05:51:10 +01:00
|
|
|
ext4_update_inode_fsync_trans(handle, inode, 0);
|
2006-10-11 10:20:50 +02:00
|
|
|
out_brelse:
|
2008-09-09 04:25:24 +02:00
|
|
|
brelse(bh);
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_std_error(inode->i_sb, err);
|
2006-10-11 10:20:50 +02:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_write_inode()
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
|
|
|
* We are called from a few places:
|
|
|
|
*
|
|
|
|
* - Within generic_file_write() for O_SYNC files.
|
|
|
|
* Here, there will be no transaction running. We wait for any running
|
|
|
|
* trasnaction to commit.
|
|
|
|
*
|
|
|
|
* - Within sys_sync(), kupdate and such.
|
|
|
|
* We wait on commit, if tol to.
|
|
|
|
*
|
|
|
|
* - Within prune_icache() (PF_MEMALLOC == true)
|
|
|
|
* Here we simply return. We can't afford to block kswapd on the
|
|
|
|
* journal commit.
|
|
|
|
*
|
|
|
|
* In all cases it is actually safe for us to return without doing anything,
|
|
|
|
* because the inode has been copied into a raw inode buffer in
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_mark_inode_dirty(). This is a correctness thing for O_SYNC and for
|
2006-10-11 10:20:50 +02:00
|
|
|
* knfsd.
|
|
|
|
*
|
|
|
|
* Note that we are absolutely dependent upon all inode dirtiers doing the
|
|
|
|
* right thing: they *must* call mark_inode_dirty() after dirtying info in
|
|
|
|
* which we are interested.
|
|
|
|
*
|
|
|
|
* It would be a bug for them to not do this. The code:
|
|
|
|
*
|
|
|
|
* mark_inode_dirty(inode)
|
|
|
|
* stuff();
|
|
|
|
* inode->i_size = expr;
|
|
|
|
*
|
|
|
|
* is in error because a kswapd-driven write_inode() could occur while
|
|
|
|
* `stuff()' is running, and the new i_size will be lost. Plus the inode
|
|
|
|
* will no longer be on the superblock's dirty inode list.
|
|
|
|
*/
|
2010-03-05 09:21:37 +01:00
|
|
|
int ext4_write_inode(struct inode *inode, struct writeback_control *wbc)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2009-09-10 04:33:47 +02:00
|
|
|
int err;
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
if (current->flags & PF_MEMALLOC)
|
|
|
|
return 0;
|
|
|
|
|
2009-09-10 04:33:47 +02:00
|
|
|
if (EXT4_SB(inode->i_sb)->s_journal) {
|
|
|
|
if (ext4_journal_current_handle()) {
|
|
|
|
jbd_debug(1, "called recursively, non-PF_MEMALLOC!\n");
|
|
|
|
dump_stack();
|
|
|
|
return -EIO;
|
|
|
|
}
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2010-03-05 09:21:37 +01:00
|
|
|
if (wbc->sync_mode != WB_SYNC_ALL)
|
2009-09-10 04:33:47 +02:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
err = ext4_force_commit(inode->i_sb);
|
|
|
|
} else {
|
|
|
|
struct ext4_iloc iloc;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2010-04-03 22:45:06 +02:00
|
|
|
err = __ext4_get_inode_loc(inode, &iloc, 0);
|
2009-09-10 04:33:47 +02:00
|
|
|
if (err)
|
|
|
|
return err;
|
2010-03-05 09:21:37 +01:00
|
|
|
if (wbc->sync_mode == WB_SYNC_ALL)
|
2009-09-29 16:07:47 +02:00
|
|
|
sync_dirty_buffer(iloc.bh);
|
|
|
|
if (buffer_req(iloc.bh) && !buffer_uptodate(iloc.bh)) {
|
2010-05-17 03:00:00 +02:00
|
|
|
EXT4_ERROR_INODE(inode,
|
|
|
|
"IO error syncing inode (block=%llu)",
|
|
|
|
(unsigned long long) iloc.bh->b_blocknr);
|
2009-09-29 16:07:47 +02:00
|
|
|
err = -EIO;
|
|
|
|
}
|
2010-04-03 23:44:16 +02:00
|
|
|
brelse(iloc.bh);
|
2009-09-10 04:33:47 +02:00
|
|
|
}
|
|
|
|
return err;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_setattr()
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
|
|
|
* Called from notify_change.
|
|
|
|
*
|
|
|
|
* We want to trap VFS attempts to truncate the file as soon as
|
|
|
|
* possible. In particular, we want to make sure that when the VFS
|
|
|
|
* shrinks i_size, we put the inode on the orphan list and modify
|
|
|
|
* i_disksize immediately, so that during the subsequent flushing of
|
|
|
|
* dirty pages and freeing of disk blocks, we can guarantee that any
|
|
|
|
* commit will leave the blocks being flushed in an unused state on
|
|
|
|
* disk. (On recovery, the inode will get truncated and the blocks will
|
|
|
|
* be freed, so we have a strong guarantee that no future commit will
|
|
|
|
* leave these blocks visible to the user.)
|
|
|
|
*
|
2008-07-12 01:27:31 +02:00
|
|
|
* Another thing we have to assure is that if we are in ordered mode
|
|
|
|
* and inode is still attached to the committing transaction, we must
|
|
|
|
* we start writeout of all the dirty pages which are being truncated.
|
|
|
|
* This way we are sure that all the data written in the previous
|
|
|
|
* transaction are already on disk (truncate waits for pages under
|
|
|
|
* writeback).
|
|
|
|
*
|
|
|
|
* Called with inode->i_mutex down.
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
int ext4_setattr(struct dentry *dentry, struct iattr *attr)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
struct inode *inode = dentry->d_inode;
|
|
|
|
int error, rc = 0;
|
|
|
|
const unsigned int ia_valid = attr->ia_valid;
|
|
|
|
|
|
|
|
error = inode_change_ok(inode, attr);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
2010-04-08 20:04:20 +02:00
|
|
|
if (is_quota_modification(inode, attr))
|
2010-03-03 15:05:07 +01:00
|
|
|
dquot_initialize(inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
|
|
|
|
(ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid)) {
|
|
|
|
handle_t *handle;
|
|
|
|
|
|
|
|
/* (user+group)*(old+new) structure, inode write (sb,
|
|
|
|
* inode block, ? - but truncate inode update has it) */
|
2009-12-09 04:42:15 +01:00
|
|
|
handle = ext4_journal_start(inode, (EXT4_MAXQUOTAS_INIT_BLOCKS(inode->i_sb)+
|
2009-12-09 04:42:28 +01:00
|
|
|
EXT4_MAXQUOTAS_DEL_BLOCKS(inode->i_sb))+3);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (IS_ERR(handle)) {
|
|
|
|
error = PTR_ERR(handle);
|
|
|
|
goto err_out;
|
|
|
|
}
|
2010-03-03 15:05:03 +01:00
|
|
|
error = dquot_transfer(inode, attr);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (error) {
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_journal_stop(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
/* Update corresponding info in inode so that everything is in
|
|
|
|
* one transaction */
|
|
|
|
if (attr->ia_valid & ATTR_UID)
|
|
|
|
inode->i_uid = attr->ia_uid;
|
|
|
|
if (attr->ia_valid & ATTR_GID)
|
|
|
|
inode->i_gid = attr->ia_gid;
|
2006-10-11 10:20:53 +02:00
|
|
|
error = ext4_mark_inode_dirty(handle, inode);
|
|
|
|
ext4_journal_stop(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
2008-01-29 05:58:27 +01:00
|
|
|
if (attr->ia_valid & ATTR_SIZE) {
|
2010-05-17 04:00:00 +02:00
|
|
|
if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) {
|
2008-01-29 05:58:27 +01:00
|
|
|
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
|
|
|
|
|
|
|
|
if (attr->ia_size > sbi->s_bitmap_maxbytes) {
|
|
|
|
error = -EFBIG;
|
|
|
|
goto err_out;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
if (S_ISREG(inode->i_mode) &&
|
2010-02-24 15:52:53 +01:00
|
|
|
attr->ia_valid & ATTR_SIZE &&
|
|
|
|
(attr->ia_size < inode->i_size ||
|
2010-05-17 04:00:00 +02:00
|
|
|
(ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS)))) {
|
2006-10-11 10:20:50 +02:00
|
|
|
handle_t *handle;
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
handle = ext4_journal_start(inode, 3);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (IS_ERR(handle)) {
|
|
|
|
error = PTR_ERR(handle);
|
|
|
|
goto err_out;
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
error = ext4_orphan_add(handle, inode);
|
|
|
|
EXT4_I(inode)->i_disksize = attr->ia_size;
|
|
|
|
rc = ext4_mark_inode_dirty(handle, inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!error)
|
|
|
|
error = rc;
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_journal_stop(handle);
|
2008-07-12 01:27:31 +02:00
|
|
|
|
|
|
|
if (ext4_should_order_data(inode)) {
|
|
|
|
error = ext4_begin_ordered_truncate(inode,
|
|
|
|
attr->ia_size);
|
|
|
|
if (error) {
|
|
|
|
/* Do as much error cleanup as possible */
|
|
|
|
handle = ext4_journal_start(inode, 3);
|
|
|
|
if (IS_ERR(handle)) {
|
|
|
|
ext4_orphan_del(NULL, inode);
|
|
|
|
goto err_out;
|
|
|
|
}
|
|
|
|
ext4_orphan_del(handle, inode);
|
|
|
|
ext4_journal_stop(handle);
|
|
|
|
goto err_out;
|
|
|
|
}
|
|
|
|
}
|
2010-02-24 15:52:53 +01:00
|
|
|
/* ext4_truncate will clear the flag */
|
2010-05-17 04:00:00 +02:00
|
|
|
if ((ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS)))
|
2010-02-24 15:52:53 +01:00
|
|
|
ext4_truncate(inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
rc = inode_setattr(inode, attr);
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
/* If inode_setattr's call to ext4_truncate failed to get a
|
2006-10-11 10:20:50 +02:00
|
|
|
* transaction handle at all, we need to clean up the in-core
|
|
|
|
* orphan list manually. */
|
|
|
|
if (inode->i_nlink)
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_orphan_del(NULL, inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
if (!rc && (ia_valid & ATTR_MODE))
|
2006-10-11 10:20:53 +02:00
|
|
|
rc = ext4_acl_chmod(inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
err_out:
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_std_error(inode->i_sb, error);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!error)
|
|
|
|
error = rc;
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2008-07-12 01:27:31 +02:00
|
|
|
int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
|
|
|
|
struct kstat *stat)
|
|
|
|
{
|
|
|
|
struct inode *inode;
|
|
|
|
unsigned long delalloc_blocks;
|
|
|
|
|
|
|
|
inode = dentry->d_inode;
|
|
|
|
generic_fillattr(inode, stat);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We can't update i_blocks if the block allocation is delayed
|
|
|
|
* otherwise in the case of system crash before the real block
|
|
|
|
* allocation is done, we will have i_blocks inconsistent with
|
|
|
|
* on-disk file blocks.
|
|
|
|
* We always keep i_blocks updated together with real
|
|
|
|
* allocation. But to not confuse with user, stat
|
|
|
|
* will return the blocks that include the delayed allocation
|
|
|
|
* blocks for this file.
|
|
|
|
*/
|
|
|
|
spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
|
|
|
|
delalloc_blocks = EXT4_I(inode)->i_reserved_data_blocks;
|
|
|
|
spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
|
|
|
|
|
|
|
|
stat->blocks += (delalloc_blocks << inode->i_sb->s_blocksize_bits)>>9;
|
|
|
|
return 0;
|
|
|
|
}
|
2006-10-11 10:20:50 +02:00
|
|
|
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
static int ext4_indirect_trans_blocks(struct inode *inode, int nrblocks,
|
|
|
|
int chunk)
|
|
|
|
{
|
|
|
|
int indirects;
|
|
|
|
|
|
|
|
/* if nrblocks are contiguous */
|
|
|
|
if (chunk) {
|
|
|
|
/*
|
|
|
|
* With N contiguous data blocks, it need at most
|
|
|
|
* N/EXT4_ADDR_PER_BLOCK(inode->i_sb) indirect blocks
|
|
|
|
* 2 dindirect blocks
|
|
|
|
* 1 tindirect block
|
|
|
|
*/
|
|
|
|
indirects = nrblocks / EXT4_ADDR_PER_BLOCK(inode->i_sb);
|
|
|
|
return indirects + 3;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* if nrblocks are not contiguous, worse case, each block touch
|
|
|
|
* a indirect block, and each indirect block touch a double indirect
|
|
|
|
* block, plus a triple indirect block
|
|
|
|
*/
|
|
|
|
indirects = nrblocks * 2 + 1;
|
|
|
|
return indirects;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int ext4_index_trans_blocks(struct inode *inode, int nrblocks, int chunk)
|
|
|
|
{
|
2010-05-17 04:00:00 +02:00
|
|
|
if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
|
2008-11-06 22:49:36 +01:00
|
|
|
return ext4_indirect_trans_blocks(inode, nrblocks, chunk);
|
|
|
|
return ext4_ext_index_trans_blocks(inode, nrblocks, chunk);
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
}
|
2008-11-06 22:49:36 +01:00
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
* Account for index blocks, block groups bitmaps and block group
|
|
|
|
* descriptor blocks if modify datablocks and index blocks
|
|
|
|
* worse case, the indexs blocks spread over different block groups
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
* If datablocks are discontiguous, they are possible to spread over
|
tree-wide: fix assorted typos all over the place
That is "success", "unknown", "through", "performance", "[re|un]mapping"
, "access", "default", "reasonable", "[con]currently", "temperature"
, "channel", "[un]used", "application", "example","hierarchy", "therefore"
, "[over|under]flow", "contiguous", "threshold", "enough" and others.
Signed-off-by: André Goddard Rosa <andre.goddard@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2009-11-14 16:09:05 +01:00
|
|
|
* different block groups too. If they are contiuguous, with flexbg,
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
* they could still across block group boundary.
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
* Also account for superblock, inode, quota and xattr blocks
|
|
|
|
*/
|
|
|
|
int ext4_meta_trans_blocks(struct inode *inode, int nrblocks, int chunk)
|
|
|
|
{
|
2009-05-01 14:50:38 +02:00
|
|
|
ext4_group_t groups, ngroups = ext4_get_groups_count(inode->i_sb);
|
|
|
|
int gdpblocks;
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
int idxblocks;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* How many index blocks need to touch to modify nrblocks?
|
|
|
|
* The "Chunk" flag indicating whether the nrblocks is
|
|
|
|
* physically contiguous on disk
|
|
|
|
*
|
|
|
|
* For Direct IO and fallocate, they calls get_block to allocate
|
|
|
|
* one single extent at a time, so they could set the "Chunk" flag
|
|
|
|
*/
|
|
|
|
idxblocks = ext4_index_trans_blocks(inode, nrblocks, chunk);
|
|
|
|
|
|
|
|
ret = idxblocks;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Now let's see how many group bitmaps and group descriptors need
|
|
|
|
* to account
|
|
|
|
*/
|
|
|
|
groups = idxblocks;
|
|
|
|
if (chunk)
|
|
|
|
groups += 1;
|
|
|
|
else
|
|
|
|
groups += nrblocks;
|
|
|
|
|
|
|
|
gdpblocks = groups;
|
2009-05-01 14:50:38 +02:00
|
|
|
if (groups > ngroups)
|
|
|
|
groups = ngroups;
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
if (groups > EXT4_SB(inode->i_sb)->s_gdb_count)
|
|
|
|
gdpblocks = EXT4_SB(inode->i_sb)->s_gdb_count;
|
|
|
|
|
|
|
|
/* bitmaps and block group descriptor blocks */
|
|
|
|
ret += groups + gdpblocks;
|
|
|
|
|
|
|
|
/* Blocks for super block, inode, quota and xattr blocks */
|
|
|
|
ret += EXT4_META_TRANS_BLOCKS(inode->i_sb);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Calulate the total number of credits to reserve to fit
|
2008-08-20 04:16:03 +02:00
|
|
|
* the modification of a single pages into a single transaction,
|
|
|
|
* which may include multiple chunks of block allocations.
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
2008-08-20 04:15:58 +02:00
|
|
|
* This could be called via ext4_write_begin()
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
2008-08-20 04:15:58 +02:00
|
|
|
* We need to consider the worse case, when
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
* one new block per extent.
|
2006-10-11 10:20:50 +02:00
|
|
|
*/
|
2006-10-11 10:21:03 +02:00
|
|
|
int ext4_writepage_trans_blocks(struct inode *inode)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
int bpp = ext4_journal_blocks_per_page(inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
int ret;
|
|
|
|
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
ret = ext4_meta_trans_blocks(inode, bpp, 0);
|
2006-10-11 10:21:03 +02:00
|
|
|
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
/* Account for data blocks for journalled mode */
|
2006-10-11 10:20:53 +02:00
|
|
|
if (ext4_should_journal_data(inode))
|
ext4: journal credits calulation cleanup and fix for non-extent writepage
When considering how many journal credits are needed for modifying a
chunk of data, we need to account for the super block, inode block,
quota blocks and xattr block, indirect/index blocks, also, group bitmap
and group descriptor blocks for new allocation (including data and
indirect/index blocks). There are many places in ext4 do the calculation
on their own and often missed one or two meta blocks, and often they
assume single block allocation, and did not considering the multile
chunk of allocation case.
This patch is trying to cleanup current journal credit code, provides
some common helper funtion to calculate the journal credits, to be used
for writepage, writepages, DIO, fallocate, migration, defrag, and for
both nonextent and extent files.
This patch modified the writepage/write_begin credit caculation for
nonextent files, to use the new helper function. It also fixed the
problem that writepage on nonextent files did not consider the case
blocksize <pagesize, thus could possibelly need multiple block
allocation in a single transaction.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-08-20 04:16:07 +02:00
|
|
|
ret += bpp;
|
2006-10-11 10:20:50 +02:00
|
|
|
return ret;
|
|
|
|
}
|
2008-08-20 04:16:03 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Calculate the journal credits for a chunk of data modification.
|
|
|
|
*
|
|
|
|
* This is called from DIO, fallocate or whoever calling
|
tree-wide: fix assorted typos all over the place
That is "success", "unknown", "through", "performance", "[re|un]mapping"
, "access", "default", "reasonable", "[con]currently", "temperature"
, "channel", "[un]used", "application", "example","hierarchy", "therefore"
, "[over|under]flow", "contiguous", "threshold", "enough" and others.
Signed-off-by: André Goddard Rosa <andre.goddard@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2009-11-14 16:09:05 +01:00
|
|
|
* ext4_get_blocks() to map/allocate a chunk of contiguous disk blocks.
|
2008-08-20 04:16:03 +02:00
|
|
|
*
|
|
|
|
* journal buffers for data blocks are not included here, as DIO
|
|
|
|
* and fallocate do no need to journal data buffers.
|
|
|
|
*/
|
|
|
|
int ext4_chunk_trans_blocks(struct inode *inode, int nrblocks)
|
|
|
|
{
|
|
|
|
return ext4_meta_trans_blocks(inode, nrblocks, 1);
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* The caller must have previously called ext4_reserve_inode_write().
|
2006-10-11 10:20:50 +02:00
|
|
|
* Give this, we know that the caller already has write access to iloc->bh.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
int ext4_mark_iloc_dirty(handle_t *handle,
|
2009-06-14 23:45:34 +02:00
|
|
|
struct inode *inode, struct ext4_iloc *iloc)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
int err = 0;
|
|
|
|
|
2008-01-29 05:58:27 +01:00
|
|
|
if (test_opt(inode->i_sb, I_VERSION))
|
|
|
|
inode_inc_iversion(inode);
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/* the do_update_inode consumes one bh->b_count */
|
|
|
|
get_bh(iloc->bh);
|
|
|
|
|
2006-10-11 10:21:01 +02:00
|
|
|
/* ext4_do_update_inode() does jbd2_journal_dirty_metadata */
|
2009-09-29 16:07:47 +02:00
|
|
|
err = ext4_do_update_inode(handle, inode, iloc);
|
2006-10-11 10:20:50 +02:00
|
|
|
put_bh(iloc->bh);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* On success, We end up with an outstanding reference count against
|
|
|
|
* iloc->bh. This _must_ be cleaned up later.
|
|
|
|
*/
|
|
|
|
|
|
|
|
int
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
|
|
|
|
struct ext4_iloc *iloc)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2009-01-07 06:06:22 +01:00
|
|
|
int err;
|
|
|
|
|
|
|
|
err = ext4_get_inode_loc(inode, iloc);
|
|
|
|
if (!err) {
|
|
|
|
BUFFER_TRACE(iloc->bh, "get_write_access");
|
|
|
|
err = ext4_journal_get_write_access(handle, iloc->bh);
|
|
|
|
if (err) {
|
|
|
|
brelse(iloc->bh);
|
|
|
|
iloc->bh = NULL;
|
2006-10-11 10:20:50 +02:00
|
|
|
}
|
|
|
|
}
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_std_error(inode->i_sb, err);
|
2006-10-11 10:20:50 +02:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2007-07-18 15:19:57 +02:00
|
|
|
/*
|
|
|
|
* Expand an inode by new_extra_isize bytes.
|
|
|
|
* Returns 0 on success or negative error number on failure.
|
|
|
|
*/
|
2008-01-29 05:58:27 +01:00
|
|
|
static int ext4_expand_extra_isize(struct inode *inode,
|
|
|
|
unsigned int new_extra_isize,
|
|
|
|
struct ext4_iloc iloc,
|
|
|
|
handle_t *handle)
|
2007-07-18 15:19:57 +02:00
|
|
|
{
|
|
|
|
struct ext4_inode *raw_inode;
|
|
|
|
struct ext4_xattr_ibody_header *header;
|
|
|
|
struct ext4_xattr_entry *entry;
|
|
|
|
|
|
|
|
if (EXT4_I(inode)->i_extra_isize >= new_extra_isize)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
raw_inode = ext4_raw_inode(&iloc);
|
|
|
|
|
|
|
|
header = IHDR(inode, raw_inode);
|
|
|
|
entry = IFIRST(header);
|
|
|
|
|
|
|
|
/* No extended attributes present */
|
2010-01-24 20:34:07 +01:00
|
|
|
if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR) ||
|
|
|
|
header->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC)) {
|
2007-07-18 15:19:57 +02:00
|
|
|
memset((void *)raw_inode + EXT4_GOOD_OLD_INODE_SIZE, 0,
|
|
|
|
new_extra_isize);
|
|
|
|
EXT4_I(inode)->i_extra_isize = new_extra_isize;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* try to expand with EAs present */
|
|
|
|
return ext4_expand_extra_isize_ea(inode, new_extra_isize,
|
|
|
|
raw_inode, handle);
|
|
|
|
}
|
|
|
|
|
2006-10-11 10:20:50 +02:00
|
|
|
/*
|
|
|
|
* What we do here is to mark the in-core inode as clean with respect to inode
|
|
|
|
* dirtiness (it may still be data-dirty).
|
|
|
|
* This means that the in-core inode may be reaped by prune_icache
|
|
|
|
* without having to perform any I/O. This is a very good thing,
|
|
|
|
* because *any* task may call prune_icache - even ones which
|
|
|
|
* have a transaction open against a different journal.
|
|
|
|
*
|
|
|
|
* Is this cheating? Not really. Sure, we haven't written the
|
|
|
|
* inode out, but prune_icache isn't a user-visible syncing function.
|
|
|
|
* Whenever the user wants stuff synced (sys_sync, sys_msync, sys_fsync)
|
|
|
|
* we start and wait on commits.
|
|
|
|
*
|
|
|
|
* Is this efficient/effective? Well, we're being nice to the system
|
|
|
|
* by cleaning up our inodes proactively so they can be reaped
|
|
|
|
* without I/O. But we are potentially leaving up to five seconds'
|
|
|
|
* worth of inodes floating about which prune_icache wants us to
|
|
|
|
* write out. One way to fix that would be to get prune_icache()
|
|
|
|
* to do a write_super() to free up some memory. It has the desired
|
|
|
|
* effect.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
struct ext4_iloc iloc;
|
2007-07-18 15:19:57 +02:00
|
|
|
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
|
|
|
|
static unsigned int mnt_count;
|
|
|
|
int err, ret;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
might_sleep();
|
2006-10-11 10:20:53 +02:00
|
|
|
err = ext4_reserve_inode_write(handle, inode, &iloc);
|
2009-01-07 06:06:22 +01:00
|
|
|
if (ext4_handle_valid(handle) &&
|
|
|
|
EXT4_I(inode)->i_extra_isize < sbi->s_want_extra_isize &&
|
2010-01-24 20:34:07 +01:00
|
|
|
!ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND)) {
|
2007-07-18 15:19:57 +02:00
|
|
|
/*
|
|
|
|
* We need extra buffer credits since we may write into EA block
|
|
|
|
* with this same handle. If journal_extend fails, then it will
|
|
|
|
* only result in a minor loss of functionality for that inode.
|
|
|
|
* If this is felt to be critical, then e2fsck should be run to
|
|
|
|
* force a large enough s_min_extra_isize.
|
|
|
|
*/
|
|
|
|
if ((jbd2_journal_extend(handle,
|
|
|
|
EXT4_DATA_TRANS_BLOCKS(inode->i_sb))) == 0) {
|
|
|
|
ret = ext4_expand_extra_isize(inode,
|
|
|
|
sbi->s_want_extra_isize,
|
|
|
|
iloc, handle);
|
|
|
|
if (ret) {
|
2010-01-24 20:34:07 +01:00
|
|
|
ext4_set_inode_state(inode,
|
|
|
|
EXT4_STATE_NO_EXPAND);
|
2007-10-17 00:38:25 +02:00
|
|
|
if (mnt_count !=
|
|
|
|
le16_to_cpu(sbi->s_es->s_mnt_count)) {
|
2010-02-15 20:19:27 +01:00
|
|
|
ext4_warning(inode->i_sb,
|
2007-07-18 15:19:57 +02:00
|
|
|
"Unable to expand inode %lu. Delete"
|
|
|
|
" some EAs or run e2fsck.",
|
|
|
|
inode->i_ino);
|
2007-10-17 00:38:25 +02:00
|
|
|
mnt_count =
|
|
|
|
le16_to_cpu(sbi->s_es->s_mnt_count);
|
2007-07-18 15:19:57 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!err)
|
2006-10-11 10:20:53 +02:00
|
|
|
err = ext4_mark_iloc_dirty(handle, inode, &iloc);
|
2006-10-11 10:20:50 +02:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_dirty_inode() is called from __mark_inode_dirty()
|
2006-10-11 10:20:50 +02:00
|
|
|
*
|
|
|
|
* We're really interested in the case where a file is being extended.
|
|
|
|
* i_size has been changed by generic_commit_write() and we thus need
|
|
|
|
* to include the updated inode in the current transaction.
|
|
|
|
*
|
2010-03-03 15:05:00 +01:00
|
|
|
* Also, dquot_alloc_block() will always dirty the inode when blocks
|
2006-10-11 10:20:50 +02:00
|
|
|
* are allocated to the file.
|
|
|
|
*
|
|
|
|
* If the inode is marked synchronous, we don't honour that here - doing
|
|
|
|
* so would cause a commit on atime updates, which we don't bother doing.
|
|
|
|
* We handle synchronous inodes at the highest possible level.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
void ext4_dirty_inode(struct inode *inode)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
handle_t *handle;
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
handle = ext4_journal_start(inode, 2);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (IS_ERR(handle))
|
|
|
|
goto out;
|
2009-09-29 22:06:01 +02:00
|
|
|
|
|
|
|
ext4_mark_inode_dirty(handle, inode);
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_journal_stop(handle);
|
2006-10-11 10:20:50 +02:00
|
|
|
out:
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
#if 0
|
|
|
|
/*
|
|
|
|
* Bind an inode's backing buffer_head into this transaction, to prevent
|
|
|
|
* it from being flushed to disk early. Unlike
|
2006-10-11 10:20:53 +02:00
|
|
|
* ext4_reserve_inode_write, this leaves behind no bh reference and
|
2006-10-11 10:20:50 +02:00
|
|
|
* returns no iloc structure, so the caller needs to repeat the iloc
|
|
|
|
* lookup to mark the inode dirty later.
|
|
|
|
*/
|
2006-10-11 10:20:53 +02:00
|
|
|
static int ext4_pin_inode(handle_t *handle, struct inode *inode)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
2006-10-11 10:20:53 +02:00
|
|
|
struct ext4_iloc iloc;
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
int err = 0;
|
|
|
|
if (handle) {
|
2006-10-11 10:20:53 +02:00
|
|
|
err = ext4_get_inode_loc(inode, &iloc);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!err) {
|
|
|
|
BUFFER_TRACE(iloc.bh, "get_write_access");
|
2006-10-11 10:21:01 +02:00
|
|
|
err = jbd2_journal_get_write_access(handle, iloc.bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (!err)
|
2009-01-07 06:06:22 +01:00
|
|
|
err = ext4_handle_dirty_metadata(handle,
|
2010-02-16 21:06:29 +01:00
|
|
|
NULL,
|
2009-01-07 06:06:22 +01:00
|
|
|
iloc.bh);
|
2006-10-11 10:20:50 +02:00
|
|
|
brelse(iloc.bh);
|
|
|
|
}
|
|
|
|
}
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_std_error(inode->i_sb, err);
|
2006-10-11 10:20:50 +02:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
int ext4_change_inode_journal_flag(struct inode *inode, int val)
|
2006-10-11 10:20:50 +02:00
|
|
|
{
|
|
|
|
journal_t *journal;
|
|
|
|
handle_t *handle;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We have to be very careful here: changing a data block's
|
|
|
|
* journaling status dynamically is dangerous. If we write a
|
|
|
|
* data block to the journal, change the status and then delete
|
|
|
|
* that block, we risk forgetting to revoke the old log record
|
|
|
|
* from the journal and so a subsequent replay can corrupt data.
|
|
|
|
* So, first we make sure that the journal is empty and that
|
|
|
|
* nobody is changing anything.
|
|
|
|
*/
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
journal = EXT4_JOURNAL(inode);
|
2009-01-07 06:06:22 +01:00
|
|
|
if (!journal)
|
|
|
|
return 0;
|
2007-07-18 14:33:51 +02:00
|
|
|
if (is_journal_aborted(journal))
|
2006-10-11 10:20:50 +02:00
|
|
|
return -EROFS;
|
|
|
|
|
2006-10-11 10:21:01 +02:00
|
|
|
jbd2_journal_lock_updates(journal);
|
|
|
|
jbd2_journal_flush(journal);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* OK, there are no updates running now, and all cached data is
|
|
|
|
* synced to disk. We are now in a completely consistent state
|
|
|
|
* which doesn't have anything in the journal, and we know that
|
|
|
|
* no filesystem updates are running, so it is safe to modify
|
|
|
|
* the inode's in-core data-journaling state flag now.
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (val)
|
2010-05-17 04:00:00 +02:00
|
|
|
ext4_set_inode_flag(inode, EXT4_INODE_JOURNAL_DATA);
|
2006-10-11 10:20:50 +02:00
|
|
|
else
|
2010-05-17 04:00:00 +02:00
|
|
|
ext4_clear_inode_flag(inode, EXT4_INODE_JOURNAL_DATA);
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_set_aops(inode);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
2006-10-11 10:21:01 +02:00
|
|
|
jbd2_journal_unlock_updates(journal);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
/* Finally we can mark the inode as dirty. */
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
handle = ext4_journal_start(inode, 1);
|
2006-10-11 10:20:50 +02:00
|
|
|
if (IS_ERR(handle))
|
|
|
|
return PTR_ERR(handle);
|
|
|
|
|
2006-10-11 10:20:53 +02:00
|
|
|
err = ext4_mark_inode_dirty(handle, inode);
|
2009-01-07 06:06:22 +01:00
|
|
|
ext4_handle_sync(handle);
|
2006-10-11 10:20:53 +02:00
|
|
|
ext4_journal_stop(handle);
|
|
|
|
ext4_std_error(inode->i_sb, err);
|
2006-10-11 10:20:50 +02:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
2008-07-12 01:27:31 +02:00
|
|
|
|
|
|
|
static int ext4_bh_unmapped(handle_t *handle, struct buffer_head *bh)
|
|
|
|
{
|
|
|
|
return !buffer_mapped(bh);
|
|
|
|
}
|
|
|
|
|
2009-04-01 00:23:21 +02:00
|
|
|
int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
|
2008-07-12 01:27:31 +02:00
|
|
|
{
|
2009-04-01 00:23:21 +02:00
|
|
|
struct page *page = vmf->page;
|
2008-07-12 01:27:31 +02:00
|
|
|
loff_t size;
|
|
|
|
unsigned long len;
|
|
|
|
int ret = -EINVAL;
|
2008-10-09 05:13:30 +02:00
|
|
|
void *fsdata;
|
2008-07-12 01:27:31 +02:00
|
|
|
struct file *file = vma->vm_file;
|
|
|
|
struct inode *inode = file->f_path.dentry->d_inode;
|
|
|
|
struct address_space *mapping = inode->i_mapping;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get i_alloc_sem to stop truncates messing with the inode. We cannot
|
|
|
|
* get i_mutex because we are already holding mmap_sem.
|
|
|
|
*/
|
|
|
|
down_read(&inode->i_alloc_sem);
|
|
|
|
size = i_size_read(inode);
|
|
|
|
if (page->mapping != mapping || size <= page_offset(page)
|
|
|
|
|| !PageUptodate(page)) {
|
|
|
|
/* page got truncated from under us? */
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
ret = 0;
|
|
|
|
if (PageMappedToDisk(page))
|
|
|
|
goto out_unlock;
|
|
|
|
|
|
|
|
if (page->index == size >> PAGE_CACHE_SHIFT)
|
|
|
|
len = size & ~PAGE_CACHE_MASK;
|
|
|
|
else
|
|
|
|
len = PAGE_CACHE_SIZE;
|
|
|
|
|
2009-09-10 04:36:03 +02:00
|
|
|
lock_page(page);
|
|
|
|
/*
|
|
|
|
* return if we have all the buffers mapped. This avoid
|
|
|
|
* the need to call write_begin/write_end which does a
|
|
|
|
* journal_start/journal_stop which can block and take
|
|
|
|
* long time
|
|
|
|
*/
|
2008-07-12 01:27:31 +02:00
|
|
|
if (page_has_buffers(page)) {
|
|
|
|
if (!walk_page_buffers(NULL, page_buffers(page), 0, len, NULL,
|
2009-09-10 04:36:03 +02:00
|
|
|
ext4_bh_unmapped)) {
|
|
|
|
unlock_page(page);
|
2008-07-12 01:27:31 +02:00
|
|
|
goto out_unlock;
|
2009-09-10 04:36:03 +02:00
|
|
|
}
|
2008-07-12 01:27:31 +02:00
|
|
|
}
|
2009-09-10 04:36:03 +02:00
|
|
|
unlock_page(page);
|
2008-07-12 01:27:31 +02:00
|
|
|
/*
|
|
|
|
* OK, we need to fill the hole... Do write_begin write_end
|
|
|
|
* to do block allocation/reservation.We are not holding
|
|
|
|
* inode.i__mutex here. That allow * parallel write_begin,
|
|
|
|
* write_end call. lock_page prevent this from happening
|
|
|
|
* on the same page though
|
|
|
|
*/
|
|
|
|
ret = mapping->a_ops->write_begin(file, mapping, page_offset(page),
|
2008-10-09 05:13:30 +02:00
|
|
|
len, AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata);
|
2008-07-12 01:27:31 +02:00
|
|
|
if (ret < 0)
|
|
|
|
goto out_unlock;
|
|
|
|
ret = mapping->a_ops->write_end(file, mapping, page_offset(page),
|
2008-10-09 05:13:30 +02:00
|
|
|
len, len, page, fsdata);
|
2008-07-12 01:27:31 +02:00
|
|
|
if (ret < 0)
|
|
|
|
goto out_unlock;
|
|
|
|
ret = 0;
|
|
|
|
out_unlock:
|
2009-04-01 00:23:21 +02:00
|
|
|
if (ret)
|
|
|
|
ret = VM_FAULT_SIGBUS;
|
2008-07-12 01:27:31 +02:00
|
|
|
up_read(&inode->i_alloc_sem);
|
|
|
|
return ret;
|
|
|
|
}
|