linux/fs/btrfs/file.c

1938 lines
49 KiB
C
Raw Normal View History

/*
* Copyright (C) 2007 Oracle. All rights reserved.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public
* License v2 as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* You should have received a copy of the GNU General Public
* License along with this program; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 021110-1307, USA.
*/
#include <linux/fs.h>
#include <linux/pagemap.h>
#include <linux/highmem.h>
#include <linux/time.h>
#include <linux/init.h>
#include <linux/string.h>
#include <linux/backing-dev.h>
#include <linux/mpage.h>
#include <linux/falloc.h>
#include <linux/swap.h>
#include <linux/writeback.h>
#include <linux/statfs.h>
#include <linux/compat.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>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "btrfs_inode.h"
#include "ioctl.h"
#include "print-tree.h"
#include "tree-log.h"
#include "locking.h"
Btrfs: Add workaround for AppArmor changing remove_suid() In openSUSE 10.3, AppArmor modifies remove_suid to take a struct path rather than just a dentry. This patch tests that the kernel is openSUSE 10.3 or newer and adjusts the call accordingly. Debian/Ubuntu with AppArmor applied will also need a similar patch. Maintainers of btrfs under those distributions should build on this patch or, alternatively, alter their package descriptions to add -DREMOVE_SUID_PATH to the compiler command line. Signed-off-by: Jeff Mahoney <jeffm@suse.com> - --- /dev/null 1970-01-01 00:00:00.000000000 +0000 +++ b/compat.h 2008-02-06 16:46:13.000000000 -0500 @@ -0,0 +1,15 @@ +#ifndef _COMPAT_H_ +#define _COMPAT_H_ + + +/* + * Even if AppArmor isn't enabled, it still has different prototypes. + * Add more distro/version pairs here to declare which has AppArmor applied. + */ +#if defined(CONFIG_SUSE_KERNEL) +# if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22) +# define REMOVE_SUID_PATH 1 +# endif +#endif + +#endif /* _COMPAT_H_ */ - --- a/file.c 2008-02-06 11:37:39.000000000 -0500 +++ b/file.c 2008-02-06 16:46:23.000000000 -0500 @@ -37,6 +37,7 @@ #include "ordered-data.h" #include "ioctl.h" #include "print-tree.h" +#include "compat.h" static int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes, @@ -790,7 +791,11 @@ static ssize_t btrfs_file_write(struct f goto out_nolock; if (count == 0) goto out_nolock; +#ifdef REMOVE_SUID_PATH + err = remove_suid(&file->f_path); +#else err = remove_suid(fdentry(file)); +#endif if (err) goto out_nolock; file_update_time(file); Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-05-02 21:03:58 +02:00
#include "compat.h"
/*
* when auto defrag is enabled we
* queue up these defrag structs to remember which
* inodes need defragging passes
*/
struct inode_defrag {
struct rb_node rb_node;
/* objectid */
u64 ino;
/*
* transid where the defrag was added, we search for
* extents newer than this
*/
u64 transid;
/* root objectid */
u64 root;
/* last offset we were able to defrag */
u64 last_offset;
/* if we've wrapped around back to zero once already */
int cycled;
};
static int __compare_inode_defrag(struct inode_defrag *defrag1,
struct inode_defrag *defrag2)
{
if (defrag1->root > defrag2->root)
return 1;
else if (defrag1->root < defrag2->root)
return -1;
else if (defrag1->ino > defrag2->ino)
return 1;
else if (defrag1->ino < defrag2->ino)
return -1;
else
return 0;
}
/* pop a record for an inode into the defrag tree. The lock
* must be held already
*
* If you're inserting a record for an older transid than an
* existing record, the transid already in the tree is lowered
*
* If an existing record is found the defrag item you
* pass in is freed
*/
static void __btrfs_add_inode_defrag(struct inode *inode,
struct inode_defrag *defrag)
{
struct btrfs_root *root = BTRFS_I(inode)->root;
struct inode_defrag *entry;
struct rb_node **p;
struct rb_node *parent = NULL;
int ret;
p = &root->fs_info->defrag_inodes.rb_node;
while (*p) {
parent = *p;
entry = rb_entry(parent, struct inode_defrag, rb_node);
ret = __compare_inode_defrag(defrag, entry);
if (ret < 0)
p = &parent->rb_left;
else if (ret > 0)
p = &parent->rb_right;
else {
/* if we're reinserting an entry for
* an old defrag run, make sure to
* lower the transid of our existing record
*/
if (defrag->transid < entry->transid)
entry->transid = defrag->transid;
if (defrag->last_offset > entry->last_offset)
entry->last_offset = defrag->last_offset;
goto exists;
}
}
set_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags);
rb_link_node(&defrag->rb_node, parent, p);
rb_insert_color(&defrag->rb_node, &root->fs_info->defrag_inodes);
return;
exists:
kfree(defrag);
return;
}
/*
* insert a defrag record for this inode if auto defrag is
* enabled
*/
int btrfs_add_inode_defrag(struct btrfs_trans_handle *trans,
struct inode *inode)
{
struct btrfs_root *root = BTRFS_I(inode)->root;
struct inode_defrag *defrag;
u64 transid;
if (!btrfs_test_opt(root, AUTO_DEFRAG))
return 0;
if (btrfs_fs_closing(root->fs_info))
return 0;
if (test_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags))
return 0;
if (trans)
transid = trans->transid;
else
transid = BTRFS_I(inode)->root->last_trans;
defrag = kzalloc(sizeof(*defrag), GFP_NOFS);
if (!defrag)
return -ENOMEM;
defrag->ino = btrfs_ino(inode);
defrag->transid = transid;
defrag->root = root->root_key.objectid;
spin_lock(&root->fs_info->defrag_inodes_lock);
if (!test_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags))
__btrfs_add_inode_defrag(inode, defrag);
else
kfree(defrag);
spin_unlock(&root->fs_info->defrag_inodes_lock);
return 0;
}
/*
* must be called with the defrag_inodes lock held
*/
struct inode_defrag *btrfs_find_defrag_inode(struct btrfs_fs_info *info,
u64 root, u64 ino,
struct rb_node **next)
{
struct inode_defrag *entry = NULL;
struct inode_defrag tmp;
struct rb_node *p;
struct rb_node *parent = NULL;
int ret;
tmp.ino = ino;
tmp.root = root;
p = info->defrag_inodes.rb_node;
while (p) {
parent = p;
entry = rb_entry(parent, struct inode_defrag, rb_node);
ret = __compare_inode_defrag(&tmp, entry);
if (ret < 0)
p = parent->rb_left;
else if (ret > 0)
p = parent->rb_right;
else
return entry;
}
if (next) {
while (parent && __compare_inode_defrag(&tmp, entry) > 0) {
parent = rb_next(parent);
entry = rb_entry(parent, struct inode_defrag, rb_node);
}
*next = parent;
}
return NULL;
}
/*
* run through the list of inodes in the FS that need
* defragging
*/
int btrfs_run_defrag_inodes(struct btrfs_fs_info *fs_info)
{
struct inode_defrag *defrag;
struct btrfs_root *inode_root;
struct inode *inode;
struct rb_node *n;
struct btrfs_key key;
struct btrfs_ioctl_defrag_range_args range;
u64 first_ino = 0;
u64 root_objectid = 0;
int num_defrag;
int defrag_batch = 1024;
memset(&range, 0, sizeof(range));
range.len = (u64)-1;
atomic_inc(&fs_info->defrag_running);
spin_lock(&fs_info->defrag_inodes_lock);
while(1) {
n = NULL;
/* find an inode to defrag */
defrag = btrfs_find_defrag_inode(fs_info, root_objectid,
first_ino, &n);
if (!defrag) {
if (n) {
defrag = rb_entry(n, struct inode_defrag,
rb_node);
} else if (root_objectid || first_ino) {
root_objectid = 0;
first_ino = 0;
continue;
} else {
break;
}
}
/* remove it from the rbtree */
first_ino = defrag->ino + 1;
root_objectid = defrag->root;
rb_erase(&defrag->rb_node, &fs_info->defrag_inodes);
if (btrfs_fs_closing(fs_info))
goto next_free;
spin_unlock(&fs_info->defrag_inodes_lock);
/* get the inode */
key.objectid = defrag->root;
btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
key.offset = (u64)-1;
inode_root = btrfs_read_fs_root_no_name(fs_info, &key);
if (IS_ERR(inode_root))
goto next;
key.objectid = defrag->ino;
btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
key.offset = 0;
inode = btrfs_iget(fs_info->sb, &key, inode_root, NULL);
if (IS_ERR(inode))
goto next;
/* do a chunk of defrag */
clear_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags);
range.start = defrag->last_offset;
num_defrag = btrfs_defrag_file(inode, NULL, &range, defrag->transid,
defrag_batch);
/*
* if we filled the whole defrag batch, there
* must be more work to do. Queue this defrag
* again
*/
if (num_defrag == defrag_batch) {
defrag->last_offset = range.start;
__btrfs_add_inode_defrag(inode, defrag);
/*
* we don't want to kfree defrag, we added it back to
* the rbtree
*/
defrag = NULL;
} else if (defrag->last_offset && !defrag->cycled) {
/*
* we didn't fill our defrag batch, but
* we didn't start at zero. Make sure we loop
* around to the start of the file.
*/
defrag->last_offset = 0;
defrag->cycled = 1;
__btrfs_add_inode_defrag(inode, defrag);
defrag = NULL;
}
iput(inode);
next:
spin_lock(&fs_info->defrag_inodes_lock);
next_free:
kfree(defrag);
}
spin_unlock(&fs_info->defrag_inodes_lock);
atomic_dec(&fs_info->defrag_running);
/*
* during unmount, we use the transaction_wait queue to
* wait for the defragger to stop
*/
wake_up(&fs_info->transaction_wait);
return 0;
}
/* simple helper to fault in pages and copy. This should go away
* and be replaced with calls into generic code.
*/
static noinline int btrfs_copy_from_user(loff_t pos, int num_pages,
size_t write_bytes,
struct page **prepared_pages,
struct iov_iter *i)
{
size_t copied = 0;
size_t total_copied = 0;
int pg = 0;
int offset = pos & (PAGE_CACHE_SIZE - 1);
while (write_bytes > 0) {
size_t count = min_t(size_t,
PAGE_CACHE_SIZE - offset, write_bytes);
struct page *page = prepared_pages[pg];
/*
* Copy data from userspace to the current page
*
* Disable pagefault to avoid recursive lock since
* the pages are already locked
*/
pagefault_disable();
copied = iov_iter_copy_from_user_atomic(page, i, offset, count);
pagefault_enable();
/* Flush processor's dcache for this page */
flush_dcache_page(page);
/*
* if we get a partial write, we can end up with
* partially up to date pages. These add
* a lot of complexity, so make sure they don't
* happen by forcing this copy to be retried.
*
* The rest of the btrfs_file_write code will fall
* back to page at a time copies after we return 0.
*/
if (!PageUptodate(page) && copied < count)
copied = 0;
iov_iter_advance(i, copied);
write_bytes -= copied;
total_copied += copied;
/* Return to btrfs_file_aio_write to fault page */
if (unlikely(copied == 0))
break;
if (unlikely(copied < PAGE_CACHE_SIZE - offset)) {
offset += copied;
} else {
pg++;
offset = 0;
}
}
return total_copied;
}
/*
* unlocks pages after btrfs_file_write is done with them
*/
void btrfs_drop_pages(struct page **pages, size_t num_pages)
{
size_t i;
for (i = 0; i < num_pages; i++) {
/* page checked is some magic around finding pages that
* have been modified without going through btrfs_set_page_dirty
* clear it here
*/
ClearPageChecked(pages[i]);
unlock_page(pages[i]);
mark_page_accessed(pages[i]);
page_cache_release(pages[i]);
}
}
/*
* after copy_from_user, pages need to be dirtied and we need to make
* sure holes are created between the current EOF and the start of
* any next extents (if required).
*
* this also makes the decision about creating an inline extent vs
* doing real data extents, marking pages dirty and delalloc as required.
*/
int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode,
struct page **pages, size_t num_pages,
loff_t pos, size_t write_bytes,
struct extent_state **cached)
{
int err = 0;
int i;
u64 num_bytes;
u64 start_pos;
u64 end_of_last_block;
u64 end_pos = pos + write_bytes;
loff_t isize = i_size_read(inode);
start_pos = pos & ~((u64)root->sectorsize - 1);
num_bytes = (write_bytes + pos - start_pos +
root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
end_of_last_block = start_pos + num_bytes - 1;
err = btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block,
cached);
if (err)
return err;
Btrfs: proper -ENOSPC handling At the start of a transaction we do a btrfs_reserve_metadata_space() and specify how many items we plan on modifying. Then once we've done our modifications and such, just call btrfs_unreserve_metadata_space() for the same number of items we reserved. For keeping track of metadata needed for data I've had to add an extent_io op for when we merge extents. This lets us track space properly when we are doing sequential writes, so we don't end up reserving way more metadata space than what we need. The only place where the metadata space accounting is not done is in the relocation code. This is because Yan is going to be reworking that code in the near future, so running btrfs-vol -b could still possibly result in a ENOSPC related panic. This patch also turns off the metadata_ratio stuff in order to allow users to more efficiently use their disk space. This patch makes it so we track how much metadata we need for an inode's delayed allocation extents by tracking how many extents are currently waiting for allocation. It introduces two new callbacks for the extent_io tree's, merge_extent_hook and split_extent_hook. These help us keep track of when we merge delalloc extents together and split them up. Reservations are handled prior to any actually dirty'ing occurs, and then we unreserve after we dirty. btrfs_unreserve_metadata_for_delalloc() will make the appropriate unreservations as needed based on the number of reservations we currently have and the number of extents we currently have. Doing the reservation outside of doing any of the actual dirty'ing lets us do things like filemap_flush() the inode to try and force delalloc to happen, or as a last resort actually start allocation on all delalloc inodes in the fs. This has survived dbench, fs_mark and an fsx torture test. Signed-off-by: Josef Bacik <jbacik@redhat.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:12:44 +02:00
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 19:49:59 +01:00
for (i = 0; i < num_pages; i++) {
struct page *p = pages[i];
SetPageUptodate(p);
ClearPageChecked(p);
set_page_dirty(p);
}
/*
* we've only changed i_size in ram, and we haven't updated
* the disk i_size. There is no need to log the inode
* at this time.
*/
if (end_pos > isize)
i_size_write(inode, end_pos);
return 0;
}
/*
* this drops all the extents in the cache that intersect the range
* [start, end]. Existing extents are split as required.
*/
int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
int skip_pinned)
{
struct extent_map *em;
struct extent_map *split = NULL;
struct extent_map *split2 = NULL;
struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
u64 len = end - start + 1;
int ret;
int testend = 1;
unsigned long flags;
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 19:49:59 +01:00
int compressed = 0;
WARN_ON(end < start);
if (end == (u64)-1) {
len = (u64)-1;
testend = 0;
}
while (1) {
if (!split)
split = alloc_extent_map();
if (!split2)
split2 = alloc_extent_map();
BUG_ON(!split || !split2); /* -ENOMEM */
write_lock(&em_tree->lock);
em = lookup_extent_mapping(em_tree, start, len);
if (!em) {
write_unlock(&em_tree->lock);
break;
}
flags = em->flags;
if (skip_pinned && test_bit(EXTENT_FLAG_PINNED, &em->flags)) {
if (testend && em->start + em->len >= start + len) {
free_extent_map(em);
write_unlock(&em_tree->lock);
break;
}
start = em->start + em->len;
if (testend)
len = start + len - (em->start + em->len);
free_extent_map(em);
write_unlock(&em_tree->lock);
continue;
}
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 19:49:59 +01:00
compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
clear_bit(EXTENT_FLAG_PINNED, &em->flags);
remove_extent_mapping(em_tree, em);
if (em->block_start < EXTENT_MAP_LAST_BYTE &&
em->start < start) {
split->start = em->start;
split->len = start - em->start;
split->orig_start = em->orig_start;
split->block_start = em->block_start;
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 19:49:59 +01:00
if (compressed)
split->block_len = em->block_len;
else
split->block_len = split->len;
split->bdev = em->bdev;
split->flags = flags;
split->compress_type = em->compress_type;
ret = add_extent_mapping(em_tree, split);
BUG_ON(ret); /* Logic error */
free_extent_map(split);
split = split2;
split2 = NULL;
}
if (em->block_start < EXTENT_MAP_LAST_BYTE &&
testend && em->start + em->len > start + len) {
u64 diff = start + len - em->start;
split->start = start + len;
split->len = em->start + em->len - (start + len);
split->bdev = em->bdev;
split->flags = flags;
split->compress_type = em->compress_type;
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 19:49:59 +01:00
if (compressed) {
split->block_len = em->block_len;
split->block_start = em->block_start;
split->orig_start = em->orig_start;
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 19:49:59 +01:00
} else {
split->block_len = split->len;
split->block_start = em->block_start + diff;
split->orig_start = split->start;
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 19:49:59 +01:00
}
ret = add_extent_mapping(em_tree, split);
BUG_ON(ret); /* Logic error */
free_extent_map(split);
split = NULL;
}
write_unlock(&em_tree->lock);
/* once for us */
free_extent_map(em);
/* once for the tree*/
free_extent_map(em);
}
if (split)
free_extent_map(split);
if (split2)
free_extent_map(split2);
return 0;
}
/*
* this is very complex, but the basic idea is to drop all extents
* in the range start - end. hint_block is filled in with a block number
* that would be a good hint to the block allocator for this file.
*
* If an extent intersects the range but is not entirely inside the range
* it is either truncated or split. Anything entirely inside the range
* is deleted from the tree.
*/
int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct inode *inode,
u64 start, u64 end, u64 *hint_byte, int drop_cache)
{
struct btrfs_root *root = BTRFS_I(inode)->root;
struct extent_buffer *leaf;
struct btrfs_file_extent_item *fi;
struct btrfs_path *path;
struct btrfs_key key;
struct btrfs_key new_key;
u64 ino = btrfs_ino(inode);
u64 search_start = start;
u64 disk_bytenr = 0;
u64 num_bytes = 0;
u64 extent_offset = 0;
u64 extent_end = 0;
int del_nr = 0;
int del_slot = 0;
int extent_type;
int recow;
int ret;
int modify_tree = -1;
if (drop_cache)
btrfs_drop_extent_cache(inode, start, end - 1, 0);
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
if (start >= BTRFS_I(inode)->disk_i_size)
modify_tree = 0;
while (1) {
recow = 0;
ret = btrfs_lookup_file_extent(trans, root, path, ino,
search_start, modify_tree);
if (ret < 0)
break;
if (ret > 0 && path->slots[0] > 0 && search_start == start) {
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &key, path->slots[0] - 1);
if (key.objectid == ino &&
key.type == BTRFS_EXTENT_DATA_KEY)
path->slots[0]--;
}
ret = 0;
next_slot:
leaf = path->nodes[0];
if (path->slots[0] >= btrfs_header_nritems(leaf)) {
BUG_ON(del_nr > 0);
ret = btrfs_next_leaf(root, path);
if (ret < 0)
break;
if (ret > 0) {
ret = 0;
break;
}
leaf = path->nodes[0];
recow = 1;
}
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
if (key.objectid > ino ||
key.type > BTRFS_EXTENT_DATA_KEY || key.offset >= end)
break;
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
extent_type = btrfs_file_extent_type(leaf, fi);
if (extent_type == BTRFS_FILE_EXTENT_REG ||
extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
extent_offset = btrfs_file_extent_offset(leaf, fi);
extent_end = key.offset +
btrfs_file_extent_num_bytes(leaf, fi);
} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
extent_end = key.offset +
btrfs_file_extent_inline_len(leaf, fi);
} else {
WARN_ON(1);
extent_end = search_start;
}
if (extent_end <= search_start) {
path->slots[0]++;
goto next_slot;
}
search_start = max(key.offset, start);
if (recow || !modify_tree) {
modify_tree = -1;
btrfs_release_path(path);
continue;
}
/*
* | - range to drop - |
* | -------- extent -------- |
*/
if (start > key.offset && end < extent_end) {
BUG_ON(del_nr > 0);
BUG_ON(extent_type == BTRFS_FILE_EXTENT_INLINE);
memcpy(&new_key, &key, sizeof(new_key));
new_key.offset = start;
ret = btrfs_duplicate_item(trans, root, path,
&new_key);
if (ret == -EAGAIN) {
btrfs_release_path(path);
continue;
}
if (ret < 0)
break;
leaf = path->nodes[0];
fi = btrfs_item_ptr(leaf, path->slots[0] - 1,
struct btrfs_file_extent_item);
btrfs_set_file_extent_num_bytes(leaf, fi,
start - key.offset);
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
extent_offset += start - key.offset;
btrfs_set_file_extent_offset(leaf, fi, extent_offset);
btrfs_set_file_extent_num_bytes(leaf, fi,
extent_end - start);
btrfs_mark_buffer_dirty(leaf);
if (disk_bytenr > 0) {
ret = btrfs_inc_extent_ref(trans, root,
disk_bytenr, num_bytes, 0,
root->root_key.objectid,
new_key.objectid,
start - extent_offset, 0);
BUG_ON(ret); /* -ENOMEM */
*hint_byte = disk_bytenr;
}
key.offset = start;
}
/*
* | ---- range to drop ----- |
* | -------- extent -------- |
*/
if (start <= key.offset && end < extent_end) {
BUG_ON(extent_type == BTRFS_FILE_EXTENT_INLINE);
memcpy(&new_key, &key, sizeof(new_key));
new_key.offset = end;
btrfs_set_item_key_safe(trans, root, path, &new_key);
extent_offset += end - key.offset;
btrfs_set_file_extent_offset(leaf, fi, extent_offset);
btrfs_set_file_extent_num_bytes(leaf, fi,
extent_end - end);
btrfs_mark_buffer_dirty(leaf);
if (disk_bytenr > 0) {
inode_sub_bytes(inode, end - key.offset);
*hint_byte = disk_bytenr;
}
break;
}
search_start = extent_end;
/*
* | ---- range to drop ----- |
* | -------- extent -------- |
*/
if (start > key.offset && end >= extent_end) {
BUG_ON(del_nr > 0);
BUG_ON(extent_type == BTRFS_FILE_EXTENT_INLINE);
btrfs_set_file_extent_num_bytes(leaf, fi,
start - key.offset);
btrfs_mark_buffer_dirty(leaf);
if (disk_bytenr > 0) {
inode_sub_bytes(inode, extent_end - start);
*hint_byte = disk_bytenr;
}
if (end == extent_end)
break;
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 19:49:59 +01:00
path->slots[0]++;
goto next_slot;
}
/*
* | ---- range to drop ----- |
* | ------ extent ------ |
*/
if (start <= key.offset && end >= extent_end) {
if (del_nr == 0) {
del_slot = path->slots[0];
del_nr = 1;
} else {
BUG_ON(del_slot + del_nr != path->slots[0]);
del_nr++;
}
if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
inode_sub_bytes(inode,
extent_end - key.offset);
extent_end = ALIGN(extent_end,
root->sectorsize);
} else if (disk_bytenr > 0) {
ret = btrfs_free_extent(trans, root,
disk_bytenr, num_bytes, 0,
root->root_key.objectid,
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 16:45:14 +02:00
key.objectid, key.offset -
extent_offset, 0);
BUG_ON(ret); /* -ENOMEM */
inode_sub_bytes(inode,
extent_end - key.offset);
*hint_byte = disk_bytenr;
}
if (end == extent_end)
break;
if (path->slots[0] + 1 < btrfs_header_nritems(leaf)) {
path->slots[0]++;
goto next_slot;
}
ret = btrfs_del_items(trans, root, path, del_slot,
del_nr);
if (ret) {
btrfs_abort_transaction(trans, root, ret);
goto out;
}
del_nr = 0;
del_slot = 0;
btrfs_release_path(path);
continue;
}
BUG_ON(1);
}
if (!ret && del_nr > 0) {
ret = btrfs_del_items(trans, root, path, del_slot, del_nr);
if (ret)
btrfs_abort_transaction(trans, root, ret);
}
out:
btrfs_free_path(path);
return ret;
}
static int extent_mergeable(struct extent_buffer *leaf, int slot,
u64 objectid, u64 bytenr, u64 orig_offset,
u64 *start, u64 *end)
{
struct btrfs_file_extent_item *fi;
struct btrfs_key key;
u64 extent_end;
if (slot < 0 || slot >= btrfs_header_nritems(leaf))
return 0;
btrfs_item_key_to_cpu(leaf, &key, slot);
if (key.objectid != objectid || key.type != BTRFS_EXTENT_DATA_KEY)
return 0;
fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
if (btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_REG ||
btrfs_file_extent_disk_bytenr(leaf, fi) != bytenr ||
btrfs_file_extent_offset(leaf, fi) != key.offset - orig_offset ||
btrfs_file_extent_compression(leaf, fi) ||
btrfs_file_extent_encryption(leaf, fi) ||
btrfs_file_extent_other_encoding(leaf, fi))
return 0;
extent_end = key.offset + btrfs_file_extent_num_bytes(leaf, fi);
if ((*start && *start != key.offset) || (*end && *end != extent_end))
return 0;
*start = key.offset;
*end = extent_end;
return 1;
}
/*
* Mark extent in the range start - end as written.
*
* This changes extent type from 'pre-allocated' to 'regular'. If only
* part of extent is marked as written, the extent will be split into
* two or three.
*/
int btrfs_mark_extent_written(struct btrfs_trans_handle *trans,
struct inode *inode, u64 start, u64 end)
{
struct btrfs_root *root = BTRFS_I(inode)->root;
struct extent_buffer *leaf;
struct btrfs_path *path;
struct btrfs_file_extent_item *fi;
struct btrfs_key key;
struct btrfs_key new_key;
u64 bytenr;
u64 num_bytes;
u64 extent_end;
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 16:45:14 +02:00
u64 orig_offset;
u64 other_start;
u64 other_end;
u64 split;
int del_nr = 0;
int del_slot = 0;
int recow;
int ret;
u64 ino = btrfs_ino(inode);
btrfs_drop_extent_cache(inode, start, end - 1, 0);
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
again:
recow = 0;
split = start;
key.objectid = ino;
key.type = BTRFS_EXTENT_DATA_KEY;
key.offset = split;
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
if (ret < 0)
goto out;
if (ret > 0 && path->slots[0] > 0)
path->slots[0]--;
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
BUG_ON(key.objectid != ino || key.type != BTRFS_EXTENT_DATA_KEY);
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
BUG_ON(btrfs_file_extent_type(leaf, fi) !=
BTRFS_FILE_EXTENT_PREALLOC);
extent_end = key.offset + btrfs_file_extent_num_bytes(leaf, fi);
BUG_ON(key.offset > start || extent_end < end);
bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 16:45:14 +02:00
orig_offset = key.offset - btrfs_file_extent_offset(leaf, fi);
memcpy(&new_key, &key, sizeof(new_key));
if (start == key.offset && end < extent_end) {
other_start = 0;
other_end = start;
if (extent_mergeable(leaf, path->slots[0] - 1,
ino, bytenr, orig_offset,
&other_start, &other_end)) {
new_key.offset = end;
btrfs_set_item_key_safe(trans, root, path, &new_key);
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
btrfs_set_file_extent_num_bytes(leaf, fi,
extent_end - end);
btrfs_set_file_extent_offset(leaf, fi,
end - orig_offset);
fi = btrfs_item_ptr(leaf, path->slots[0] - 1,
struct btrfs_file_extent_item);
btrfs_set_file_extent_num_bytes(leaf, fi,
end - other_start);
btrfs_mark_buffer_dirty(leaf);
goto out;
}
}
if (start > key.offset && end == extent_end) {
other_start = end;
other_end = 0;
if (extent_mergeable(leaf, path->slots[0] + 1,
ino, bytenr, orig_offset,
&other_start, &other_end)) {
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
btrfs_set_file_extent_num_bytes(leaf, fi,
start - key.offset);
path->slots[0]++;
new_key.offset = start;
btrfs_set_item_key_safe(trans, root, path, &new_key);
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
btrfs_set_file_extent_num_bytes(leaf, fi,
other_end - start);
btrfs_set_file_extent_offset(leaf, fi,
start - orig_offset);
btrfs_mark_buffer_dirty(leaf);
goto out;
}
}
while (start > key.offset || end < extent_end) {
if (key.offset == start)
split = end;
new_key.offset = split;
ret = btrfs_duplicate_item(trans, root, path, &new_key);
if (ret == -EAGAIN) {
btrfs_release_path(path);
goto again;
}
if (ret < 0) {
btrfs_abort_transaction(trans, root, ret);
goto out;
}
leaf = path->nodes[0];
fi = btrfs_item_ptr(leaf, path->slots[0] - 1,
struct btrfs_file_extent_item);
btrfs_set_file_extent_num_bytes(leaf, fi,
split - key.offset);
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
btrfs_set_file_extent_offset(leaf, fi, split - orig_offset);
btrfs_set_file_extent_num_bytes(leaf, fi,
extent_end - split);
btrfs_mark_buffer_dirty(leaf);
ret = btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, 0,
root->root_key.objectid,
ino, orig_offset, 0);
BUG_ON(ret); /* -ENOMEM */
if (split == start) {
key.offset = start;
} else {
BUG_ON(start != key.offset);
path->slots[0]--;
extent_end = end;
}
recow = 1;
}
other_start = end;
other_end = 0;
if (extent_mergeable(leaf, path->slots[0] + 1,
ino, bytenr, orig_offset,
&other_start, &other_end)) {
if (recow) {
btrfs_release_path(path);
goto again;
}
extent_end = other_end;
del_slot = path->slots[0] + 1;
del_nr++;
ret = btrfs_free_extent(trans, root, bytenr, num_bytes,
0, root->root_key.objectid,
ino, orig_offset, 0);
BUG_ON(ret); /* -ENOMEM */
}
other_start = 0;
other_end = start;
if (extent_mergeable(leaf, path->slots[0] - 1,
ino, bytenr, orig_offset,
&other_start, &other_end)) {
if (recow) {
btrfs_release_path(path);
goto again;
}
key.offset = other_start;
del_slot = path->slots[0];
del_nr++;
ret = btrfs_free_extent(trans, root, bytenr, num_bytes,
0, root->root_key.objectid,
ino, orig_offset, 0);
BUG_ON(ret); /* -ENOMEM */
}
if (del_nr == 0) {
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
btrfs_set_file_extent_type(leaf, fi,
BTRFS_FILE_EXTENT_REG);
btrfs_mark_buffer_dirty(leaf);
} else {
fi = btrfs_item_ptr(leaf, del_slot - 1,
struct btrfs_file_extent_item);
btrfs_set_file_extent_type(leaf, fi,
BTRFS_FILE_EXTENT_REG);
btrfs_set_file_extent_num_bytes(leaf, fi,
extent_end - key.offset);
btrfs_mark_buffer_dirty(leaf);
ret = btrfs_del_items(trans, root, path, del_slot, del_nr);
if (ret < 0) {
btrfs_abort_transaction(trans, root, ret);
goto out;
}
}
out:
btrfs_free_path(path);
return 0;
}
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
/*
* on error we return an unlocked page and the error value
* on success we return a locked page and 0
*/
static int prepare_uptodate_page(struct page *page, u64 pos,
bool force_uptodate)
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
{
int ret = 0;
if (((pos & (PAGE_CACHE_SIZE - 1)) || force_uptodate) &&
!PageUptodate(page)) {
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
ret = btrfs_readpage(NULL, page);
if (ret)
return ret;
lock_page(page);
if (!PageUptodate(page)) {
unlock_page(page);
return -EIO;
}
}
return 0;
}
/*
* this gets pages into the page cache and locks them down, it also properly
* waits for data=ordered extents to finish before allowing the pages to be
* modified.
*/
static noinline int prepare_pages(struct btrfs_root *root, struct file *file,
struct page **pages, size_t num_pages,
loff_t pos, unsigned long first_index,
size_t write_bytes, bool force_uptodate)
{
struct extent_state *cached_state = NULL;
int i;
unsigned long index = pos >> PAGE_CACHE_SHIFT;
struct inode *inode = fdentry(file)->d_inode;
gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
int err = 0;
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
int faili = 0;
u64 start_pos;
u64 last_pos;
start_pos = pos & ~((u64)root->sectorsize - 1);
last_pos = ((u64)index + num_pages) << PAGE_CACHE_SHIFT;
again:
for (i = 0; i < num_pages; i++) {
pages[i] = find_or_create_page(inode->i_mapping, index + i,
mask | __GFP_WRITE);
if (!pages[i]) {
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
faili = i - 1;
err = -ENOMEM;
goto fail;
}
if (i == 0)
err = prepare_uptodate_page(pages[i], pos,
force_uptodate);
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
if (i == num_pages - 1)
err = prepare_uptodate_page(pages[i],
pos + write_bytes, false);
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
if (err) {
page_cache_release(pages[i]);
faili = i - 1;
goto fail;
}
wait_on_page_writeback(pages[i]);
}
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
err = 0;
if (start_pos < inode->i_size) {
struct btrfs_ordered_extent *ordered;
lock_extent_bits(&BTRFS_I(inode)->io_tree,
start_pos, last_pos - 1, 0, &cached_state);
ordered = btrfs_lookup_first_ordered_extent(inode,
last_pos - 1);
if (ordered &&
ordered->file_offset + ordered->len > start_pos &&
ordered->file_offset < last_pos) {
btrfs_put_ordered_extent(ordered);
unlock_extent_cached(&BTRFS_I(inode)->io_tree,
start_pos, last_pos - 1,
&cached_state, GFP_NOFS);
for (i = 0; i < num_pages; i++) {
unlock_page(pages[i]);
page_cache_release(pages[i]);
}
btrfs_wait_ordered_range(inode, start_pos,
last_pos - start_pos);
goto again;
}
if (ordered)
btrfs_put_ordered_extent(ordered);
clear_extent_bit(&BTRFS_I(inode)->io_tree, start_pos,
last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
GFP_NOFS);
unlock_extent_cached(&BTRFS_I(inode)->io_tree,
start_pos, last_pos - 1, &cached_state,
GFP_NOFS);
}
for (i = 0; i < num_pages; i++) {
if (clear_page_dirty_for_io(pages[i]))
account_page_redirty(pages[i]);
set_page_extent_mapped(pages[i]);
WARN_ON(!PageLocked(pages[i]));
}
return 0;
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
fail:
while (faili >= 0) {
unlock_page(pages[faili]);
page_cache_release(pages[faili]);
faili--;
}
return err;
}
static noinline ssize_t __btrfs_buffered_write(struct file *file,
struct iov_iter *i,
loff_t pos)
{
struct inode *inode = fdentry(file)->d_inode;
struct btrfs_root *root = BTRFS_I(inode)->root;
struct page **pages = NULL;
unsigned long first_index;
size_t num_written = 0;
int nrptrs;
int ret = 0;
bool force_page_uptodate = false;
nrptrs = min((iov_iter_count(i) + PAGE_CACHE_SIZE - 1) /
PAGE_CACHE_SIZE, PAGE_CACHE_SIZE /
(sizeof(struct page *)));
nrptrs = min(nrptrs, current->nr_dirtied_pause - current->nr_dirtied);
nrptrs = max(nrptrs, 8);
pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL);
if (!pages)
return -ENOMEM;
first_index = pos >> PAGE_CACHE_SHIFT;
while (iov_iter_count(i) > 0) {
size_t offset = pos & (PAGE_CACHE_SIZE - 1);
size_t write_bytes = min(iov_iter_count(i),
nrptrs * (size_t)PAGE_CACHE_SIZE -
offset);
size_t num_pages = (write_bytes + offset +
PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
size_t dirty_pages;
size_t copied;
WARN_ON(num_pages > nrptrs);
/*
* Fault pages before locking them in prepare_pages
* to avoid recursive lock
*/
if (unlikely(iov_iter_fault_in_readable(i, write_bytes))) {
ret = -EFAULT;
break;
}
ret = btrfs_delalloc_reserve_space(inode,
num_pages << PAGE_CACHE_SHIFT);
if (ret)
break;
/*
* This is going to setup the pages array with the number of
* pages we want, so we don't really need to worry about the
* contents of pages from loop to loop
*/
ret = prepare_pages(root, file, pages, num_pages,
pos, first_index, write_bytes,
force_page_uptodate);
if (ret) {
btrfs_delalloc_release_space(inode,
num_pages << PAGE_CACHE_SHIFT);
break;
}
copied = btrfs_copy_from_user(pos, num_pages,
write_bytes, pages, i);
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
/*
* if we have trouble faulting in the pages, fall
* back to one page at a time
*/
if (copied < write_bytes)
nrptrs = 1;
if (copied == 0) {
force_page_uptodate = true;
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
dirty_pages = 0;
} else {
force_page_uptodate = false;
Btrfs: fix regressions in copy_from_user handling Commit 914ee295af418e936ec20a08c1663eaabe4cd07a fixed deadlocks in btrfs_file_write where we would catch page faults on pages we had locked. But, there were a few problems: 1) The x86-32 iov_iter_copy_from_user_atomic code always fails to copy data when the amount to copy is more than 4K and the offset to start copying from is not page aligned. The result was btrfs_file_write looping forever retrying the iov_iter_copy_from_user_atomic We deal with this by changing btrfs_file_write to drop down to single page copies when iov_iter_copy_from_user_atomic starts returning failure. 2) The btrfs_file_write code was leaking delalloc reservations when iov_iter_copy_from_user_atomic returned zero. The looping above would result in the entire filesystem running out of delalloc reservations and constantly trying to flush things to disk. 3) btrfs_file_write will lock down page cache pages, make sure any writeback is finished, do the copy_from_user and then release them. Before the loop runs we check the first and last pages in the write to see if they are only being partially modified. If the start or end of the write isn't aligned, we make sure the corresponding pages are up to date so that we don't introduce garbage into the file. With the copy_from_user changes, we're allowing the VM to reclaim the pages after a partial update from copy_from_user, but we're not making sure the page cache page is up to date when we loop around to resume the write. We deal with this by pushing the up to date checks down into the page prep code. This fits better with how the rest of file_write works. Signed-off-by: Chris Mason <chris.mason@oracle.com> Reported-by: Mitch Harder <mitch.harder@sabayonlinux.org> cc: stable@kernel.org
2011-02-28 15:52:08 +01:00
dirty_pages = (copied + offset +
PAGE_CACHE_SIZE - 1) >>
PAGE_CACHE_SHIFT;
}
/*
* If we had a short copy we need to release the excess delaloc
* bytes we reserved. We need to increment outstanding_extents
* because btrfs_delalloc_release_space will decrement it, but
* we still have an outstanding extent for the chunk we actually
* managed to copy.
*/
if (num_pages > dirty_pages) {
if (copied > 0) {
spin_lock(&BTRFS_I(inode)->lock);
BTRFS_I(inode)->outstanding_extents++;
spin_unlock(&BTRFS_I(inode)->lock);
}
btrfs_delalloc_release_space(inode,
(num_pages - dirty_pages) <<
PAGE_CACHE_SHIFT);
}
if (copied > 0) {
ret = btrfs_dirty_pages(root, inode, pages,
dirty_pages, pos, copied,
NULL);
if (ret) {
btrfs_delalloc_release_space(inode,
dirty_pages << PAGE_CACHE_SHIFT);
btrfs_drop_pages(pages, num_pages);
break;
}
}
btrfs_drop_pages(pages, num_pages);
cond_resched();
balance_dirty_pages_ratelimited_nr(inode->i_mapping,
dirty_pages);
if (dirty_pages < (root->leafsize >> PAGE_CACHE_SHIFT) + 1)
btrfs_btree_balance_dirty(root, 1);
pos += copied;
num_written += copied;
}
kfree(pages);
return num_written ? num_written : ret;
}
static ssize_t __btrfs_direct_write(struct kiocb *iocb,
const struct iovec *iov,
unsigned long nr_segs, loff_t pos,
loff_t *ppos, size_t count, size_t ocount)
{
struct file *file = iocb->ki_filp;
struct inode *inode = fdentry(file)->d_inode;
struct iov_iter i;
ssize_t written;
ssize_t written_buffered;
loff_t endbyte;
int err;
written = generic_file_direct_write(iocb, iov, &nr_segs, pos, ppos,
count, ocount);
/*
* the generic O_DIRECT will update in-memory i_size after the
* DIOs are done. But our endio handlers that update the on
* disk i_size never update past the in memory i_size. So we
* need one more update here to catch any additions to the
* file
*/
if (inode->i_size != BTRFS_I(inode)->disk_i_size) {
btrfs_ordered_update_i_size(inode, inode->i_size, NULL);
mark_inode_dirty(inode);
}
if (written < 0 || written == count)
return written;
pos += written;
count -= written;
iov_iter_init(&i, iov, nr_segs, count, written);
written_buffered = __btrfs_buffered_write(file, &i, pos);
if (written_buffered < 0) {
err = written_buffered;
goto out;
}
endbyte = pos + written_buffered - 1;
err = filemap_write_and_wait_range(file->f_mapping, pos, endbyte);
if (err)
goto out;
written += written_buffered;
*ppos = pos + written_buffered;
invalidate_mapping_pages(file->f_mapping, pos >> PAGE_CACHE_SHIFT,
endbyte >> PAGE_CACHE_SHIFT);
out:
return written ? written : err;
}
static ssize_t btrfs_file_aio_write(struct kiocb *iocb,
const struct iovec *iov,
unsigned long nr_segs, loff_t pos)
{
struct file *file = iocb->ki_filp;
struct inode *inode = fdentry(file)->d_inode;
struct btrfs_root *root = BTRFS_I(inode)->root;
loff_t *ppos = &iocb->ki_pos;
u64 start_pos;
ssize_t num_written = 0;
ssize_t err = 0;
size_t count, ocount;
vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
mutex_lock(&inode->i_mutex);
err = generic_segment_checks(iov, &nr_segs, &ocount, VERIFY_READ);
if (err) {
mutex_unlock(&inode->i_mutex);
goto out;
}
count = ocount;
current->backing_dev_info = inode->i_mapping->backing_dev_info;
err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
if (err) {
mutex_unlock(&inode->i_mutex);
goto out;
}
if (count == 0) {
mutex_unlock(&inode->i_mutex);
goto out;
}
err = file_remove_suid(file);
if (err) {
mutex_unlock(&inode->i_mutex);
goto out;
}
/*
* If BTRFS flips readonly due to some impossible error
* (fs_info->fs_state now has BTRFS_SUPER_FLAG_ERROR),
* although we have opened a file as writable, we have
* to stop this write operation to ensure FS consistency.
*/
if (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
mutex_unlock(&inode->i_mutex);
err = -EROFS;
goto out;
}
err = btrfs_update_time(file);
if (err) {
mutex_unlock(&inode->i_mutex);
goto out;
}
start_pos = round_down(pos, root->sectorsize);
if (start_pos > i_size_read(inode)) {
err = btrfs_cont_expand(inode, i_size_read(inode), start_pos);
if (err) {
mutex_unlock(&inode->i_mutex);
goto out;
}
}
if (unlikely(file->f_flags & O_DIRECT)) {
num_written = __btrfs_direct_write(iocb, iov, nr_segs,
pos, ppos, count, ocount);
} else {
struct iov_iter i;
iov_iter_init(&i, iov, nr_segs, count, num_written);
num_written = __btrfs_buffered_write(file, &i, pos);
if (num_written > 0)
*ppos = pos + num_written;
}
mutex_unlock(&inode->i_mutex);
Btrfs: add extra flushing for renames and truncates Renames and truncates are both common ways to replace old data with new data. The filesystem can make an effort to make sure the new data is on disk before actually replacing the old data. This is especially important for rename, which many application use as though it were atomic for both the data and the metadata involved. The current btrfs code will happily replace a file that is fully on disk with one that was just created and still has pending IO. If we crash after transaction commit but before the IO is done, we'll end up replacing a good file with a zero length file. The solution used here is to create a list of inodes that need special ordering and force them to disk before the commit is done. This is similar to the ext3 style data=ordering, except it is only done on selected files. Btrfs is able to get away with this because it does not wait on commits very often, even for fsync (which use a sub-commit). For renames, we order the file when it wasn't already on disk and when it is replacing an existing file. Larger files are sent to filemap_flush right away (before the transaction handle is opened). For truncates, we order if the file goes from non-zero size down to zero size. This is a little different, because at the time of the truncate the file has no dirty bytes to order. But, we flag the inode so that it is added to the ordered list on close (via release method). We also immediately add it to the ordered list of the current transaction so that we can try to flush down any writes the application sneaks in before commit. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-31 19:27:11 +02:00
/*
* we want to make sure fsync finds this change
* but we haven't joined a transaction running right now.
*
* Later on, someone is sure to update the inode and get the
* real transid recorded.
*
* We set last_trans now to the fs_info generation + 1,
* this will either be one more than the running transaction
* or the generation used for the next transaction if there isn't
* one running right now.
*/
BTRFS_I(inode)->last_trans = root->fs_info->generation + 1;
if (num_written > 0 || num_written == -EIOCBQUEUED) {
err = generic_write_sync(file, pos, num_written);
if (err < 0 && num_written > 0)
num_written = err;
}
out:
current->backing_dev_info = NULL;
return num_written ? num_written : err;
}
int btrfs_release_file(struct inode *inode, struct file *filp)
{
Btrfs: add extra flushing for renames and truncates Renames and truncates are both common ways to replace old data with new data. The filesystem can make an effort to make sure the new data is on disk before actually replacing the old data. This is especially important for rename, which many application use as though it were atomic for both the data and the metadata involved. The current btrfs code will happily replace a file that is fully on disk with one that was just created and still has pending IO. If we crash after transaction commit but before the IO is done, we'll end up replacing a good file with a zero length file. The solution used here is to create a list of inodes that need special ordering and force them to disk before the commit is done. This is similar to the ext3 style data=ordering, except it is only done on selected files. Btrfs is able to get away with this because it does not wait on commits very often, even for fsync (which use a sub-commit). For renames, we order the file when it wasn't already on disk and when it is replacing an existing file. Larger files are sent to filemap_flush right away (before the transaction handle is opened). For truncates, we order if the file goes from non-zero size down to zero size. This is a little different, because at the time of the truncate the file has no dirty bytes to order. But, we flag the inode so that it is added to the ordered list on close (via release method). We also immediately add it to the ordered list of the current transaction so that we can try to flush down any writes the application sneaks in before commit. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-31 19:27:11 +02:00
/*
* ordered_data_close is set by settattr when we are about to truncate
* a file from a non-zero size to a zero size. This tries to
* flush down new bytes that may have been written if the
* application were using truncate to replace a file in place.
*/
if (test_and_clear_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
&BTRFS_I(inode)->runtime_flags)) {
Btrfs: add extra flushing for renames and truncates Renames and truncates are both common ways to replace old data with new data. The filesystem can make an effort to make sure the new data is on disk before actually replacing the old data. This is especially important for rename, which many application use as though it were atomic for both the data and the metadata involved. The current btrfs code will happily replace a file that is fully on disk with one that was just created and still has pending IO. If we crash after transaction commit but before the IO is done, we'll end up replacing a good file with a zero length file. The solution used here is to create a list of inodes that need special ordering and force them to disk before the commit is done. This is similar to the ext3 style data=ordering, except it is only done on selected files. Btrfs is able to get away with this because it does not wait on commits very often, even for fsync (which use a sub-commit). For renames, we order the file when it wasn't already on disk and when it is replacing an existing file. Larger files are sent to filemap_flush right away (before the transaction handle is opened). For truncates, we order if the file goes from non-zero size down to zero size. This is a little different, because at the time of the truncate the file has no dirty bytes to order. But, we flag the inode so that it is added to the ordered list on close (via release method). We also immediately add it to the ordered list of the current transaction so that we can try to flush down any writes the application sneaks in before commit. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-31 19:27:11 +02:00
btrfs_add_ordered_operation(NULL, BTRFS_I(inode)->root, inode);
if (inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT)
filemap_flush(inode->i_mapping);
}
if (filp->private_data)
btrfs_ioctl_trans_end(filp);
return 0;
}
/*
* fsync call for both files and directories. This logs the inode into
* the tree log instead of forcing full commits whenever possible.
*
* It needs to call filemap_fdatawait so that all ordered extent updates are
* in the metadata btree are up to date for copying to the log.
*
* It drops the inode mutex before doing the tree log commit. This is an
* important optimization for directories because holding the mutex prevents
* new operations on the dir while we write to disk.
*/
int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
{
struct dentry *dentry = file->f_path.dentry;
struct inode *inode = dentry->d_inode;
struct btrfs_root *root = BTRFS_I(inode)->root;
int ret = 0;
struct btrfs_trans_handle *trans;
Btrfs: add initial tracepoint support for btrfs Tracepoints can provide insight into why btrfs hits bugs and be greatly helpful for debugging, e.g dd-7822 [000] 2121.641088: btrfs_inode_request: root = 5(FS_TREE), gen = 4, ino = 256, blocks = 8, disk_i_size = 0, last_trans = 8, logged_trans = 0 dd-7822 [000] 2121.641100: btrfs_inode_new: root = 5(FS_TREE), gen = 8, ino = 257, blocks = 0, disk_i_size = 0, last_trans = 0, logged_trans = 0 btrfs-transacti-7804 [001] 2146.935420: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29368320 (orig_level = 0), cow_buf = 29388800 (cow_level = 0) btrfs-transacti-7804 [001] 2146.935473: btrfs_cow_block: root = 1(ROOT_TREE), refs = 2, orig_buf = 29364224 (orig_level = 0), cow_buf = 29392896 (cow_level = 0) btrfs-transacti-7804 [001] 2146.972221: btrfs_transaction_commit: root = 1(ROOT_TREE), gen = 8 flush-btrfs-2-7821 [001] 2155.824210: btrfs_chunk_alloc: root = 3(CHUNK_TREE), offset = 1103101952, size = 1073741824, num_stripes = 1, sub_stripes = 0, type = DATA flush-btrfs-2-7821 [001] 2155.824241: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29388800 (orig_level = 0), cow_buf = 29396992 (cow_level = 0) flush-btrfs-2-7821 [001] 2155.824255: btrfs_cow_block: root = 4(DEV_TREE), refs = 2, orig_buf = 29372416 (orig_level = 0), cow_buf = 29401088 (cow_level = 0) flush-btrfs-2-7821 [000] 2155.824329: btrfs_cow_block: root = 3(CHUNK_TREE), refs = 2, orig_buf = 20971520 (orig_level = 0), cow_buf = 20975616 (cow_level = 0) btrfs-endio-wri-7800 [001] 2155.898019: btrfs_cow_block: root = 5(FS_TREE), refs = 2, orig_buf = 29384704 (orig_level = 0), cow_buf = 29405184 (cow_level = 0) btrfs-endio-wri-7800 [001] 2155.898043: btrfs_cow_block: root = 7(CSUM_TREE), refs = 2, orig_buf = 29376512 (orig_level = 0), cow_buf = 29409280 (cow_level = 0) Here is what I have added: 1) ordere_extent: btrfs_ordered_extent_add btrfs_ordered_extent_remove btrfs_ordered_extent_start btrfs_ordered_extent_put These provide critical information to understand how ordered_extents are updated. 2) extent_map: btrfs_get_extent extent_map is used in both read and write cases, and it is useful for tracking how btrfs specific IO is running. 3) writepage: __extent_writepage btrfs_writepage_end_io_hook Pages are cirtical resourses and produce a lot of corner cases during writeback, so it is valuable to know how page is written to disk. 4) inode: btrfs_inode_new btrfs_inode_request btrfs_inode_evict These can show where and when a inode is created, when a inode is evicted. 5) sync: btrfs_sync_file btrfs_sync_fs These show sync arguments. 6) transaction: btrfs_transaction_commit In transaction based filesystem, it will be useful to know the generation and who does commit. 7) back reference and cow: btrfs_delayed_tree_ref btrfs_delayed_data_ref btrfs_delayed_ref_head btrfs_cow_block Btrfs natively supports back references, these tracepoints are helpful on understanding btrfs's COW mechanism. 8) chunk: btrfs_chunk_alloc btrfs_chunk_free Chunk is a link between physical offset and logical offset, and stands for space infomation in btrfs, and these are helpful on tracing space things. 9) reserved_extent: btrfs_reserved_extent_alloc btrfs_reserved_extent_free These can show how btrfs uses its space. Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-03-24 12:18:59 +01:00
trace_btrfs_sync_file(file, datasync);
mutex_lock(&inode->i_mutex);
/*
* we wait first, since the writeback may change the inode, also wait
* ordered range does a filemape_write_and_wait_range which is why we
* don't do it above like other file systems.
*/
root->log_batch++;
btrfs_wait_ordered_range(inode, start, end);
root->log_batch++;
/*
* check the transaction that last modified this inode
* and see if its already been committed
*/
if (!BTRFS_I(inode)->last_trans) {
mutex_unlock(&inode->i_mutex);
goto out;
}
/*
* if the last transaction that changed this file was before
* the current transaction, we can bail out now without any
* syncing
*/
Btrfs: kill trans_mutex We use trans_mutex for lots of things, here's a basic list 1) To serialize trans_handles joining the currently running transaction 2) To make sure that no new trans handles are started while we are committing 3) To protect the dead_roots list and the transaction lists Really the serializing trans_handles joining is not too hard, and can really get bogged down in acquiring a reference to the transaction. So replace the trans_mutex with a trans_lock spinlock and use it to do the following 1) Protect fs_info->running_transaction. All trans handles have to do is check this, and then take a reference of the transaction and keep on going. 2) Protect the fs_info->trans_list. This doesn't get used too much, basically it just holds the current transactions, which will usually just be the currently committing transaction and the currently running transaction at most. 3) Protect the dead roots list. This is only ever processed by splicing the list so this is relatively simple. 4) Protect the fs_info->reloc_ctl stuff. This is very lightweight and was using the trans_mutex before, so this is a pretty straightforward change. 5) Protect fs_info->no_trans_join. Because we don't hold the trans_lock over the entirety of the commit we need to have a way to block new people from creating a new transaction while we're doing our work. So we set no_trans_join and in join_transaction we test to see if that is set, and if it is we do a wait_on_commit. 6) Make the transaction use count atomic so we don't need to take locks to modify it when we're dropping references. 7) Add a commit_lock to the transaction to make sure multiple people trying to commit the same transaction don't race and commit at the same time. 8) Make open_ioctl_trans an atomic so we don't have to take any locks for ioctl trans. I have tested this with xfstests, but obviously it is a pretty hairy change so lots of testing is greatly appreciated. Thanks, Signed-off-by: Josef Bacik <josef@redhat.com>
2011-04-11 23:25:13 +02:00
smp_mb();
if (BTRFS_I(inode)->last_trans <=
root->fs_info->last_trans_committed) {
BTRFS_I(inode)->last_trans = 0;
mutex_unlock(&inode->i_mutex);
goto out;
}
/*
* ok we haven't committed the transaction yet, lets do a commit
*/
if (file->private_data)
btrfs_ioctl_trans_end(file);
trans = btrfs_start_transaction(root, 0);
if (IS_ERR(trans)) {
ret = PTR_ERR(trans);
mutex_unlock(&inode->i_mutex);
goto out;
}
ret = btrfs_log_dentry_safe(trans, root, dentry);
if (ret < 0) {
mutex_unlock(&inode->i_mutex);
goto out;
}
/* we've logged all the items and now have a consistent
* version of the file in the log. It is possible that
* someone will come in and modify the file, but that's
* fine because the log is consistent on disk, and we
* have references to all of the file's extents
*
* It is possible that someone will come in and log the
* file again, but that will end up using the synchronization
* inside btrfs_sync_log to keep things safe.
*/
mutex_unlock(&inode->i_mutex);
if (ret != BTRFS_NO_LOG_SYNC) {
if (ret > 0) {
Btrfs: tree logging unlink/rename fixes The tree logging code allows individual files or directories to be logged without including operations on other files and directories in the FS. It tries to commit the minimal set of changes to disk in order to fsync the single file or directory that was sent to fsync or O_SYNC. The tree logging code was allowing files and directories to be unlinked if they were part of a rename operation where only one directory in the rename was in the fsync log. This patch adds a few new rules to the tree logging. 1) on rename or unlink, if the inode being unlinked isn't in the fsync log, we must force a full commit before doing an fsync of the directory where the unlink was done. The commit isn't done during the unlink, but it is forced the next time we try to log the parent directory. Solution: record transid of last unlink/rename per directory when the directory wasn't already logged. For renames this is only done when renaming to a different directory. mkdir foo/some_dir normal commit rename foo/some_dir foo2/some_dir mkdir foo/some_dir fsync foo/some_dir/some_file The fsync above will unlink the original some_dir without recording it in its new location (foo2). After a crash, some_dir will be gone unless the fsync of some_file forces a full commit 2) we must log any new names for any file or dir that is in the fsync log. This way we make sure not to lose files that are unlinked during the same transaction. 2a) we must log any new names for any file or dir during rename when the directory they are being removed from was logged. 2a is actually the more important variant. Without the extra logging a crash might unlink the old name without recreating the new one 3) after a crash, we must go through any directories with a link count of zero and redo the rm -rf mkdir f1/foo normal commit rm -rf f1/foo fsync(f1) The directory f1 was fully removed from the FS, but fsync was never called on f1, only its parent dir. After a crash the rm -rf must be replayed. This must be able to recurse down the entire directory tree. The inode link count fixup code takes care of the ugly details. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-24 15:24:20 +01:00
ret = btrfs_commit_transaction(trans, root);
} else {
ret = btrfs_sync_log(trans, root);
if (ret == 0)
ret = btrfs_end_transaction(trans, root);
else
ret = btrfs_commit_transaction(trans, root);
}
} else {
ret = btrfs_end_transaction(trans, root);
}
out:
return ret > 0 ? -EIO : ret;
}
static const struct vm_operations_struct btrfs_file_vm_ops = {
.fault = filemap_fault,
.page_mkwrite = btrfs_page_mkwrite,
};
static int btrfs_file_mmap(struct file *filp, struct vm_area_struct *vma)
{
struct address_space *mapping = filp->f_mapping;
if (!mapping->a_ops->readpage)
return -ENOEXEC;
file_accessed(filp);
vma->vm_ops = &btrfs_file_vm_ops;
vma->vm_flags |= VM_CAN_NONLINEAR;
return 0;
}
static long btrfs_fallocate(struct file *file, int mode,
loff_t offset, loff_t len)
{
struct inode *inode = file->f_path.dentry->d_inode;
struct extent_state *cached_state = NULL;
u64 cur_offset;
u64 last_byte;
u64 alloc_start;
u64 alloc_end;
u64 alloc_hint = 0;
u64 locked_end;
u64 mask = BTRFS_I(inode)->root->sectorsize - 1;
struct extent_map *em;
int ret;
alloc_start = offset & ~mask;
alloc_end = (offset + len + mask) & ~mask;
/* We only support the FALLOC_FL_KEEP_SIZE mode */
if (mode & ~FALLOC_FL_KEEP_SIZE)
return -EOPNOTSUPP;
/*
* Make sure we have enough space before we do the
* allocation.
*/
ret = btrfs_check_data_free_space(inode, len);
if (ret)
return ret;
/*
* wait for ordered IO before we have any locks. We'll loop again
* below with the locks held.
*/
btrfs_wait_ordered_range(inode, alloc_start, alloc_end - alloc_start);
mutex_lock(&inode->i_mutex);
ret = inode_newsize_ok(inode, alloc_end);
if (ret)
goto out;
if (alloc_start > inode->i_size) {
ret = btrfs_cont_expand(inode, i_size_read(inode),
alloc_start);
if (ret)
goto out;
}
locked_end = alloc_end - 1;
while (1) {
struct btrfs_ordered_extent *ordered;
/* the extent lock is ordered inside the running
* transaction
*/
lock_extent_bits(&BTRFS_I(inode)->io_tree, alloc_start,
locked_end, 0, &cached_state);
ordered = btrfs_lookup_first_ordered_extent(inode,
alloc_end - 1);
if (ordered &&
ordered->file_offset + ordered->len > alloc_start &&
ordered->file_offset < alloc_end) {
btrfs_put_ordered_extent(ordered);
unlock_extent_cached(&BTRFS_I(inode)->io_tree,
alloc_start, locked_end,
&cached_state, GFP_NOFS);
/*
* we can't wait on the range with the transaction
* running or with the extent lock held
*/
btrfs_wait_ordered_range(inode, alloc_start,
alloc_end - alloc_start);
} else {
if (ordered)
btrfs_put_ordered_extent(ordered);
break;
}
}
cur_offset = alloc_start;
while (1) {
u64 actual_end;
em = btrfs_get_extent(inode, NULL, 0, cur_offset,
alloc_end - cur_offset, 0);
if (IS_ERR_OR_NULL(em)) {
if (!em)
ret = -ENOMEM;
else
ret = PTR_ERR(em);
break;
}
last_byte = min(extent_map_end(em), alloc_end);
actual_end = min_t(u64, extent_map_end(em), offset + len);
last_byte = (last_byte + mask) & ~mask;
if (em->block_start == EXTENT_MAP_HOLE ||
(cur_offset >= inode->i_size &&
!test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) {
ret = btrfs_prealloc_file_range(inode, mode, cur_offset,
last_byte - cur_offset,
1 << inode->i_blkbits,
offset + len,
&alloc_hint);
if (ret < 0) {
free_extent_map(em);
break;
}
} else if (actual_end > inode->i_size &&
!(mode & FALLOC_FL_KEEP_SIZE)) {
/*
* We didn't need to allocate any more space, but we
* still extended the size of the file so we need to
* update i_size.
*/
inode->i_ctime = CURRENT_TIME;
i_size_write(inode, actual_end);
btrfs_ordered_update_i_size(inode, actual_end, NULL);
}
free_extent_map(em);
cur_offset = last_byte;
if (cur_offset >= alloc_end) {
ret = 0;
break;
}
}
unlock_extent_cached(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
&cached_state, GFP_NOFS);
out:
mutex_unlock(&inode->i_mutex);
/* Let go of our reservation. */
btrfs_free_reserved_data_space(inode, len);
return ret;
}
static int find_desired_extent(struct inode *inode, loff_t *offset, int origin)
{
struct btrfs_root *root = BTRFS_I(inode)->root;
struct extent_map *em;
struct extent_state *cached_state = NULL;
u64 lockstart = *offset;
u64 lockend = i_size_read(inode);
u64 start = *offset;
u64 orig_start = *offset;
u64 len = i_size_read(inode);
u64 last_end = 0;
int ret = 0;
lockend = max_t(u64, root->sectorsize, lockend);
if (lockend <= lockstart)
lockend = lockstart + root->sectorsize;
len = lockend - lockstart + 1;
len = max_t(u64, len, root->sectorsize);
if (inode->i_size == 0)
return -ENXIO;
lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend, 0,
&cached_state);
/*
* Delalloc is such a pain. If we have a hole and we have pending
* delalloc for a portion of the hole we will get back a hole that
* exists for the entire range since it hasn't been actually written
* yet. So to take care of this case we need to look for an extent just
* before the position we want in case there is outstanding delalloc
* going on here.
*/
if (origin == SEEK_HOLE && start != 0) {
if (start <= root->sectorsize)
em = btrfs_get_extent_fiemap(inode, NULL, 0, 0,
root->sectorsize, 0);
else
em = btrfs_get_extent_fiemap(inode, NULL, 0,
start - root->sectorsize,
root->sectorsize, 0);
if (IS_ERR(em)) {
ret = PTR_ERR(em);
goto out;
}
last_end = em->start + em->len;
if (em->block_start == EXTENT_MAP_DELALLOC)
last_end = min_t(u64, last_end, inode->i_size);
free_extent_map(em);
}
while (1) {
em = btrfs_get_extent_fiemap(inode, NULL, 0, start, len, 0);
if (IS_ERR(em)) {
ret = PTR_ERR(em);
break;
}
if (em->block_start == EXTENT_MAP_HOLE) {
if (test_bit(EXTENT_FLAG_VACANCY, &em->flags)) {
if (last_end <= orig_start) {
free_extent_map(em);
ret = -ENXIO;
break;
}
}
if (origin == SEEK_HOLE) {
*offset = start;
free_extent_map(em);
break;
}
} else {
if (origin == SEEK_DATA) {
if (em->block_start == EXTENT_MAP_DELALLOC) {
if (start >= inode->i_size) {
free_extent_map(em);
ret = -ENXIO;
break;
}
}
*offset = start;
free_extent_map(em);
break;
}
}
start = em->start + em->len;
last_end = em->start + em->len;
if (em->block_start == EXTENT_MAP_DELALLOC)
last_end = min_t(u64, last_end, inode->i_size);
if (test_bit(EXTENT_FLAG_VACANCY, &em->flags)) {
free_extent_map(em);
ret = -ENXIO;
break;
}
free_extent_map(em);
cond_resched();
}
if (!ret)
*offset = min(*offset, inode->i_size);
out:
unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
&cached_state, GFP_NOFS);
return ret;
}
static loff_t btrfs_file_llseek(struct file *file, loff_t offset, int origin)
{
struct inode *inode = file->f_mapping->host;
int ret;
mutex_lock(&inode->i_mutex);
switch (origin) {
case SEEK_END:
case SEEK_CUR:
vfs: do (nearly) lockless generic_file_llseek The i_mutex lock use of generic _file_llseek hurts. Independent processes accessing the same file synchronize over a single lock, even though they have no need for synchronization at all. Under high utilization this can cause llseek to scale very poorly on larger systems. This patch does some rethinking of the llseek locking model: First the 64bit f_pos is not necessarily atomic without locks on 32bit systems. This can already cause races with read() today. This was discussed on linux-kernel in the past and deemed acceptable. The patch does not change that. Let's look at the different seek variants: SEEK_SET: Doesn't really need any locking. If there's a race one writer wins, the other loses. For 32bit the non atomic update races against read() stay the same. Without a lock they can also happen against write() now. The read() race was deemed acceptable in past discussions, and I think if it's ok for read it's ok for write too. => Don't need a lock. SEEK_END: This behaves like SEEK_SET plus it reads the maximum size too. Reading the maximum size would have the 32bit atomic problem. But luckily we already have a way to read the maximum size without locking (i_size_read), so we can just use that instead. Without i_mutex there is no synchronization with write() anymore, however since the write() update is atomic on 64bit it just behaves like another racy SEEK_SET. On non atomic 32bit it's the same as SEEK_SET. => Don't need a lock, but need to use i_size_read() SEEK_CUR: This has a read-modify-write race window on the same file. One could argue that any application doing unsynchronized seeks on the same file is already broken. But for the sake of not adding a regression here I'm using the file->f_lock to synchronize this. Using this lock is much better than the inode mutex because it doesn't synchronize between processes. => So still need a lock, but can use a f_lock. This patch implements this new scheme in generic_file_llseek. I dropped generic_file_llseek_unlocked and changed all callers. Signed-off-by: Andi Kleen <ak@linux.intel.com> Signed-off-by: Christoph Hellwig <hch@lst.de>
2011-09-16 01:06:48 +02:00
offset = generic_file_llseek(file, offset, origin);
goto out;
case SEEK_DATA:
case SEEK_HOLE:
if (offset >= i_size_read(inode)) {
mutex_unlock(&inode->i_mutex);
return -ENXIO;
}
ret = find_desired_extent(inode, &offset, origin);
if (ret) {
mutex_unlock(&inode->i_mutex);
return ret;
}
}
if (offset < 0 && !(file->f_mode & FMODE_UNSIGNED_OFFSET)) {
offset = -EINVAL;
goto out;
}
if (offset > inode->i_sb->s_maxbytes) {
offset = -EINVAL;
goto out;
}
/* Special lock needed here? */
if (offset != file->f_pos) {
file->f_pos = offset;
file->f_version = 0;
}
out:
mutex_unlock(&inode->i_mutex);
return offset;
}
const struct file_operations btrfs_file_operations = {
.llseek = btrfs_file_llseek,
.read = do_sync_read,
.write = do_sync_write,
.aio_read = generic_file_aio_read,
.splice_read = generic_file_splice_read,
.aio_write = btrfs_file_aio_write,
.mmap = btrfs_file_mmap,
.open = generic_file_open,
.release = btrfs_release_file,
.fsync = btrfs_sync_file,
.fallocate = btrfs_fallocate,
.unlocked_ioctl = btrfs_ioctl,
#ifdef CONFIG_COMPAT
.compat_ioctl = btrfs_ioctl,
#endif
};