2005-04-17 00:20:36 +02:00
|
|
|
/*
|
2006-06-19 00:40:27 +02:00
|
|
|
* Copyright (c) 2000-2006 Silicon Graphics, Inc.
|
2005-11-02 04:58:39 +01:00
|
|
|
* All Rights Reserved.
|
2005-04-17 00:20:36 +02:00
|
|
|
*
|
2005-11-02 04:58:39 +01:00
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public License as
|
2005-04-17 00:20:36 +02:00
|
|
|
* published by the Free Software Foundation.
|
|
|
|
*
|
2005-11-02 04:58:39 +01:00
|
|
|
* This program is distributed in the hope that it would 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.
|
2005-04-17 00:20:36 +02:00
|
|
|
*
|
2005-11-02 04:58:39 +01:00
|
|
|
* You should have received a copy of the GNU General Public License
|
|
|
|
* along with this program; if not, write the Free Software Foundation,
|
|
|
|
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2009-12-15 00:14:59 +01:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_bit.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_log.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_inum.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_trans.h"
|
|
|
|
#include "xfs_sb.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_ag.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_dir2.h"
|
|
|
|
#include "xfs_alloc.h"
|
|
|
|
#include "xfs_quota.h"
|
|
|
|
#include "xfs_mount.h"
|
|
|
|
#include "xfs_bmap_btree.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_alloc_btree.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_ialloc_btree.h"
|
|
|
|
#include "xfs_dinode.h"
|
|
|
|
#include "xfs_inode.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_btree.h"
|
2008-10-30 06:55:13 +01:00
|
|
|
#include "xfs_btree_trace.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_ialloc.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_bmap.h"
|
|
|
|
#include "xfs_rtalloc.h"
|
|
|
|
#include "xfs_error.h"
|
|
|
|
#include "xfs_itable.h"
|
2007-10-11 10:11:14 +02:00
|
|
|
#include "xfs_fsops.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_attr.h"
|
|
|
|
#include "xfs_buf_item.h"
|
|
|
|
#include "xfs_utils.h"
|
2007-08-29 02:58:01 +02:00
|
|
|
#include "xfs_vnodeops.h"
|
2007-11-23 06:29:32 +01:00
|
|
|
#include "xfs_log_priv.h"
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
#include "xfs_trans_priv.h"
|
2008-05-20 03:30:39 +02:00
|
|
|
#include "xfs_filestream.h"
|
2008-07-18 09:11:46 +02:00
|
|
|
#include "xfs_da_btree.h"
|
|
|
|
#include "xfs_extfree_item.h"
|
|
|
|
#include "xfs_mru_cache.h"
|
|
|
|
#include "xfs_inode_item.h"
|
2008-10-30 07:06:08 +01:00
|
|
|
#include "xfs_sync.h"
|
2009-12-15 00:14:59 +01:00
|
|
|
#include "xfs_trace.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
#include <linux/namei.h>
|
|
|
|
#include <linux/init.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>
|
2005-04-17 00:20:36 +02:00
|
|
|
#include <linux/mount.h>
|
2005-09-02 08:58:49 +02:00
|
|
|
#include <linux/mempool.h>
|
2005-04-17 00:20:36 +02:00
|
|
|
#include <linux/writeback.h>
|
2005-09-05 00:34:18 +02:00
|
|
|
#include <linux/kthread.h>
|
2006-12-07 05:34:23 +01:00
|
|
|
#include <linux/freezer.h>
|
2008-07-18 09:12:36 +02:00
|
|
|
#include <linux/parser.h>
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2009-09-22 02:01:09 +02:00
|
|
|
static const struct super_operations xfs_super_operations;
|
2007-02-10 08:34:56 +01:00
|
|
|
static kmem_zone_t *xfs_ioend_zone;
|
2005-09-02 08:58:49 +02:00
|
|
|
mempool_t *xfs_ioend_pool;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-11-23 06:29:32 +01:00
|
|
|
#define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
|
|
|
|
#define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
|
|
|
|
#define MNTOPT_LOGDEV "logdev" /* log device */
|
|
|
|
#define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
|
|
|
|
#define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
|
|
|
|
#define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
|
|
|
|
#define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
|
|
|
|
#define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
|
|
|
|
#define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
|
|
|
|
#define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
|
|
|
|
#define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
|
|
|
|
#define MNTOPT_MTPT "mtpt" /* filesystem mount point */
|
|
|
|
#define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
|
|
|
|
#define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
|
|
|
|
#define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
|
|
|
|
#define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
|
|
|
|
#define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
|
|
|
|
#define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
|
|
|
|
#define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
|
|
|
|
* unwritten extent conversion */
|
|
|
|
#define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
|
|
|
|
#define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
|
|
|
|
#define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
|
|
|
|
#define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
|
|
|
|
#define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
|
|
|
|
#define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
|
|
|
|
* in stat(). */
|
|
|
|
#define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
|
|
|
|
#define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
|
|
|
|
#define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
|
|
|
|
#define MNTOPT_QUOTA "quota" /* disk quotas (user) */
|
|
|
|
#define MNTOPT_NOQUOTA "noquota" /* no quotas */
|
|
|
|
#define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
|
|
|
|
#define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
|
|
|
|
#define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
|
|
|
|
#define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
|
|
|
|
#define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
|
|
|
|
#define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
|
|
|
|
#define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
|
|
|
|
#define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
|
|
|
|
#define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
|
|
|
|
#define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
|
xfs: Introduce delayed logging core code
The delayed logging code only changes in-memory structures and as
such can be enabled and disabled with a mount option. Add the mount
option and emit a warning that this is an experimental feature that
should not be used in production yet.
We also need infrastructure to track committed items that have not
yet been written to the log. This is what the Committed Item List
(CIL) is for.
The log item also needs to be extended to track the current log
vector, the associated memory buffer and it's location in the Commit
Item List. Extend the log item and log vector structures to enable
this tracking.
To maintain the current log format for transactions with delayed
logging, we need to introduce a checkpoint transaction and a context
for tracking each checkpoint from initiation to transaction
completion. This includes adding a log ticket for tracking space
log required/used by the context checkpoint.
To track all the changes we need an io vector array per log item,
rather than a single array for the entire transaction. Using the new
log vector structure for this requires two passes - the first to
allocate the log vector structures and chain them together, and the
second to fill them out. This log vector chain can then be passed
to the CIL for formatting, pinning and insertion into the CIL.
Formatting of the log vector chain is relatively simple - it's just
a loop over the iovecs on each log vector, but it is made slightly
more complex because we re-write the iovec after the copy to point
back at the memory buffer we just copied into.
This code also needs to pin log items. If the log item is not
already tracked in this checkpoint context, then it needs to be
pinned. Otherwise it is already pinned and we don't need to pin it
again.
The only other complexity is calculating the amount of new log space
the formatting has consumed. This needs to be accounted to the
transaction in progress, and the accounting is made more complex
becase we need also to steal space from it for log metadata in the
checkpoint transaction. Calculate all this at insert time and update
all the tickets, counters, etc correctly.
Once we've formatted all the log items in the transaction, attach
the busy extents to the checkpoint context so the busy extents live
until checkpoint completion and can be processed at that point in
time. Transactions can then be freed at this point in time.
Now we need to issue checkpoints - we are tracking the amount of log space
used by the items in the CIL, so we can trigger background checkpoints when the
space usage gets to a certain threshold. Otherwise, checkpoints need ot be
triggered when a log synchronisation point is reached - a log force event.
Because the log write code already handles chained log vectors, writing the
transaction is trivial, too. Construct a transaction header, add it
to the head of the chain and write it into the log, then issue a
commit record write. Then we can release the checkpoint log ticket
and attach the context to the log buffer so it can be called during
Io completion to complete the checkpoint.
We also need to allow for synchronising multiple in-flight
checkpoints. This is needed for two things - the first is to ensure
that checkpoint commit records appear in the log in the correct
sequence order (so they are replayed in the correct order). The
second is so that xfs_log_force_lsn() operates correctly and only
flushes and/or waits for the specific sequence it was provided with.
To do this we need a wait variable and a list tracking the
checkpoint commits in progress. We can walk this list and wait for
the checkpoints to change state or complete easily, an this provides
the necessary synchronisation for correct operation in both cases.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-05-21 06:37:18 +02:00
|
|
|
#define MNTOPT_DELAYLOG "delaylog" /* Delayed loging enabled */
|
|
|
|
#define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed loging disabled */
|
2007-11-23 06:29:32 +01:00
|
|
|
|
2008-07-18 09:12:36 +02:00
|
|
|
/*
|
|
|
|
* Table driven mount option parser.
|
|
|
|
*
|
|
|
|
* Currently only used for remount, but it will be used for mount
|
|
|
|
* in the future, too.
|
|
|
|
*/
|
|
|
|
enum {
|
|
|
|
Opt_barrier, Opt_nobarrier, Opt_err
|
|
|
|
};
|
|
|
|
|
2008-10-13 11:46:57 +02:00
|
|
|
static const match_table_t tokens = {
|
2008-07-18 09:12:36 +02:00
|
|
|
{Opt_barrier, "barrier"},
|
|
|
|
{Opt_nobarrier, "nobarrier"},
|
|
|
|
{Opt_err, NULL}
|
|
|
|
};
|
|
|
|
|
|
|
|
|
2007-11-23 06:29:32 +01:00
|
|
|
STATIC unsigned long
|
|
|
|
suffix_strtoul(char *s, char **endp, unsigned int base)
|
|
|
|
{
|
|
|
|
int last, shift_left_factor = 0;
|
|
|
|
char *value = s;
|
|
|
|
|
|
|
|
last = strlen(value) - 1;
|
|
|
|
if (value[last] == 'K' || value[last] == 'k') {
|
|
|
|
shift_left_factor = 10;
|
|
|
|
value[last] = '\0';
|
|
|
|
}
|
|
|
|
if (value[last] == 'M' || value[last] == 'm') {
|
|
|
|
shift_left_factor = 20;
|
|
|
|
value[last] = '\0';
|
|
|
|
}
|
|
|
|
if (value[last] == 'G' || value[last] == 'g') {
|
|
|
|
shift_left_factor = 30;
|
|
|
|
value[last] = '\0';
|
|
|
|
}
|
|
|
|
|
|
|
|
return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
|
|
|
|
}
|
|
|
|
|
2008-10-30 07:53:24 +01:00
|
|
|
/*
|
|
|
|
* This function fills in xfs_mount_t fields based on mount args.
|
|
|
|
* Note: the superblock has _not_ yet been read in.
|
|
|
|
*
|
|
|
|
* Note that this function leaks the various device name allocations on
|
|
|
|
* failure. The caller takes care of them.
|
|
|
|
*/
|
2007-11-23 06:29:32 +01:00
|
|
|
STATIC int
|
|
|
|
xfs_parseargs(
|
|
|
|
struct xfs_mount *mp,
|
2010-06-23 10:11:15 +02:00
|
|
|
char *options)
|
2007-11-23 06:29:32 +01:00
|
|
|
{
|
2008-10-30 07:53:24 +01:00
|
|
|
struct super_block *sb = mp->m_super;
|
2007-11-23 06:29:32 +01:00
|
|
|
char *this_char, *value, *eov;
|
2008-10-30 07:53:24 +01:00
|
|
|
int dsunit = 0;
|
|
|
|
int dswidth = 0;
|
|
|
|
int iosize = 0;
|
2009-02-09 08:37:39 +01:00
|
|
|
__uint8_t iosizelog = 0;
|
2008-10-30 07:53:24 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Copy binary VFS mount flags we are interested in.
|
|
|
|
*/
|
|
|
|
if (sb->s_flags & MS_RDONLY)
|
|
|
|
mp->m_flags |= XFS_MOUNT_RDONLY;
|
|
|
|
if (sb->s_flags & MS_DIRSYNC)
|
|
|
|
mp->m_flags |= XFS_MOUNT_DIRSYNC;
|
|
|
|
if (sb->s_flags & MS_SYNCHRONOUS)
|
|
|
|
mp->m_flags |= XFS_MOUNT_WSYNC;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set some default flags that could be cleared by the mount option
|
|
|
|
* parsing.
|
|
|
|
*/
|
|
|
|
mp->m_flags |= XFS_MOUNT_BARRIER;
|
|
|
|
mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
|
|
|
|
mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
|
2007-11-23 06:29:32 +01:00
|
|
|
|
2008-10-30 07:53:24 +01:00
|
|
|
/*
|
|
|
|
* These can be overridden by the mount option parsing.
|
|
|
|
*/
|
|
|
|
mp->m_logbufs = -1;
|
|
|
|
mp->m_logbsize = -1;
|
2007-11-23 06:29:32 +01:00
|
|
|
|
|
|
|
if (!options)
|
|
|
|
goto done;
|
|
|
|
|
|
|
|
while ((this_char = strsep(&options, ",")) != NULL) {
|
|
|
|
if (!*this_char)
|
|
|
|
continue;
|
|
|
|
if ((value = strchr(this_char, '=')) != NULL)
|
|
|
|
*value++ = 0;
|
|
|
|
|
|
|
|
if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
|
|
|
|
if (!value || !*value) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: %s option requires an argument",
|
|
|
|
this_char);
|
|
|
|
return EINVAL;
|
|
|
|
}
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_logbufs = simple_strtoul(value, &eov, 10);
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
|
|
|
|
if (!value || !*value) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: %s option requires an argument",
|
|
|
|
this_char);
|
|
|
|
return EINVAL;
|
|
|
|
}
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_logbsize = suffix_strtoul(value, &eov, 10);
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
|
|
|
|
if (!value || !*value) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: %s option requires an argument",
|
|
|
|
this_char);
|
|
|
|
return EINVAL;
|
|
|
|
}
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
|
|
|
|
if (!mp->m_logname)
|
|
|
|
return ENOMEM;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_MTPT)) {
|
2010-06-23 10:11:15 +02:00
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: %s option not allowed on this system",
|
|
|
|
this_char);
|
|
|
|
return EINVAL;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_RTDEV)) {
|
|
|
|
if (!value || !*value) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: %s option requires an argument",
|
|
|
|
this_char);
|
|
|
|
return EINVAL;
|
|
|
|
}
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
|
|
|
|
if (!mp->m_rtname)
|
|
|
|
return ENOMEM;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
|
|
|
|
if (!value || !*value) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: %s option requires an argument",
|
|
|
|
this_char);
|
|
|
|
return EINVAL;
|
|
|
|
}
|
|
|
|
iosize = simple_strtoul(value, &eov, 10);
|
2008-10-30 07:55:08 +01:00
|
|
|
iosizelog = ffs(iosize) - 1;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
|
|
|
|
if (!value || !*value) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: %s option requires an argument",
|
|
|
|
this_char);
|
|
|
|
return EINVAL;
|
|
|
|
}
|
|
|
|
iosize = suffix_strtoul(value, &eov, 10);
|
2008-10-30 07:53:24 +01:00
|
|
|
iosizelog = ffs(iosize) - 1;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_GRPID) ||
|
|
|
|
!strcmp(this_char, MNTOPT_BSDGROUPS)) {
|
|
|
|
mp->m_flags |= XFS_MOUNT_GRPID;
|
|
|
|
} else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
|
|
|
|
!strcmp(this_char, MNTOPT_SYSVGROUPS)) {
|
|
|
|
mp->m_flags &= ~XFS_MOUNT_GRPID;
|
|
|
|
} else if (!strcmp(this_char, MNTOPT_WSYNC)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags |= XFS_MOUNT_WSYNC;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags |= XFS_MOUNT_NORECOVERY;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags |= XFS_MOUNT_NOALIGN;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags |= XFS_MOUNT_SWALLOC;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_SUNIT)) {
|
|
|
|
if (!value || !*value) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: %s option requires an argument",
|
|
|
|
this_char);
|
|
|
|
return EINVAL;
|
|
|
|
}
|
|
|
|
dsunit = simple_strtoul(value, &eov, 10);
|
|
|
|
} else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
|
|
|
|
if (!value || !*value) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: %s option requires an argument",
|
|
|
|
this_char);
|
|
|
|
return EINVAL;
|
|
|
|
}
|
|
|
|
dswidth = simple_strtoul(value, &eov, 10);
|
|
|
|
} else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
|
2007-11-23 06:29:32 +01:00
|
|
|
#if !XFS_BIG_INUMS
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: %s option not allowed on this system",
|
|
|
|
this_char);
|
|
|
|
return EINVAL;
|
|
|
|
#endif
|
|
|
|
} else if (!strcmp(this_char, MNTOPT_NOUUID)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags |= XFS_MOUNT_NOUUID;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_BARRIER)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags |= XFS_MOUNT_BARRIER;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags &= ~XFS_MOUNT_BARRIER;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_IKEEP)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags |= XFS_MOUNT_IKEEP;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags &= ~XFS_MOUNT_IKEEP;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_ATTR2)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags |= XFS_MOUNT_ATTR2;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags &= ~XFS_MOUNT_ATTR2;
|
|
|
|
mp->m_flags |= XFS_MOUNT_NOATTR2;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_flags |= XFS_MOUNT_FILESTREAMS;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_qflags &= ~(XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
|
|
|
|
XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
|
2008-10-30 07:54:57 +01:00
|
|
|
XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
|
2008-10-30 07:53:24 +01:00
|
|
|
XFS_UQUOTA_ENFD | XFS_OQUOTA_ENFD);
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_QUOTA) ||
|
|
|
|
!strcmp(this_char, MNTOPT_UQUOTA) ||
|
|
|
|
!strcmp(this_char, MNTOPT_USRQUOTA)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
|
|
|
|
XFS_UQUOTA_ENFD);
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
|
|
|
|
!strcmp(this_char, MNTOPT_UQUOTANOENF)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
|
|
|
|
mp->m_qflags &= ~XFS_UQUOTA_ENFD;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
|
|
|
|
!strcmp(this_char, MNTOPT_PRJQUOTA)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
|
|
|
|
XFS_OQUOTA_ENFD);
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
|
|
|
|
mp->m_qflags &= ~XFS_OQUOTA_ENFD;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
|
|
|
|
!strcmp(this_char, MNTOPT_GRPQUOTA)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
|
|
|
|
XFS_OQUOTA_ENFD);
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
|
|
|
|
mp->m_qflags &= ~XFS_OQUOTA_ENFD;
|
xfs: Introduce delayed logging core code
The delayed logging code only changes in-memory structures and as
such can be enabled and disabled with a mount option. Add the mount
option and emit a warning that this is an experimental feature that
should not be used in production yet.
We also need infrastructure to track committed items that have not
yet been written to the log. This is what the Committed Item List
(CIL) is for.
The log item also needs to be extended to track the current log
vector, the associated memory buffer and it's location in the Commit
Item List. Extend the log item and log vector structures to enable
this tracking.
To maintain the current log format for transactions with delayed
logging, we need to introduce a checkpoint transaction and a context
for tracking each checkpoint from initiation to transaction
completion. This includes adding a log ticket for tracking space
log required/used by the context checkpoint.
To track all the changes we need an io vector array per log item,
rather than a single array for the entire transaction. Using the new
log vector structure for this requires two passes - the first to
allocate the log vector structures and chain them together, and the
second to fill them out. This log vector chain can then be passed
to the CIL for formatting, pinning and insertion into the CIL.
Formatting of the log vector chain is relatively simple - it's just
a loop over the iovecs on each log vector, but it is made slightly
more complex because we re-write the iovec after the copy to point
back at the memory buffer we just copied into.
This code also needs to pin log items. If the log item is not
already tracked in this checkpoint context, then it needs to be
pinned. Otherwise it is already pinned and we don't need to pin it
again.
The only other complexity is calculating the amount of new log space
the formatting has consumed. This needs to be accounted to the
transaction in progress, and the accounting is made more complex
becase we need also to steal space from it for log metadata in the
checkpoint transaction. Calculate all this at insert time and update
all the tickets, counters, etc correctly.
Once we've formatted all the log items in the transaction, attach
the busy extents to the checkpoint context so the busy extents live
until checkpoint completion and can be processed at that point in
time. Transactions can then be freed at this point in time.
Now we need to issue checkpoints - we are tracking the amount of log space
used by the items in the CIL, so we can trigger background checkpoints when the
space usage gets to a certain threshold. Otherwise, checkpoints need ot be
triggered when a log synchronisation point is reached - a log force event.
Because the log write code already handles chained log vectors, writing the
transaction is trivial, too. Construct a transaction header, add it
to the head of the chain and write it into the log, then issue a
commit record write. Then we can release the checkpoint log ticket
and attach the context to the log buffer so it can be called during
Io completion to complete the checkpoint.
We also need to allow for synchronising multiple in-flight
checkpoints. This is needed for two things - the first is to ensure
that checkpoint commit records appear in the log in the correct
sequence order (so they are replayed in the correct order). The
second is so that xfs_log_force_lsn() operates correctly and only
flushes and/or waits for the specific sequence it was provided with.
To do this we need a wait variable and a list tracking the
checkpoint commits in progress. We can walk this list and wait for
the checkpoints to change state or complete easily, an this provides
the necessary synchronisation for correct operation in both cases.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-05-21 06:37:18 +02:00
|
|
|
} else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
|
|
|
|
mp->m_flags |= XFS_MOUNT_DELAYLOG;
|
|
|
|
} else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
|
|
|
|
mp->m_flags &= ~XFS_MOUNT_DELAYLOG;
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, "ihashsize")) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: ihashsize no longer used, option is deprecated.");
|
|
|
|
} else if (!strcmp(this_char, "osyncisdsync")) {
|
|
|
|
cmn_err(CE_WARN,
|
2010-07-20 09:50:52 +02:00
|
|
|
"XFS: osyncisdsync has no effect, option is deprecated.");
|
|
|
|
} else if (!strcmp(this_char, "osyncisosync")) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: osyncisosync has no effect, option is deprecated.");
|
2007-11-23 06:29:32 +01:00
|
|
|
} else if (!strcmp(this_char, "irixsgid")) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
|
|
|
|
} else {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: unknown mount option [%s].", this_char);
|
|
|
|
return EINVAL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-10-30 07:53:24 +01:00
|
|
|
/*
|
|
|
|
* no recovery flag requires a read-only mount
|
|
|
|
*/
|
|
|
|
if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
|
|
|
|
!(mp->m_flags & XFS_MOUNT_RDONLY)) {
|
|
|
|
cmn_err(CE_WARN, "XFS: no-recovery mounts must be read-only.");
|
|
|
|
return EINVAL;
|
2007-11-23 06:29:32 +01:00
|
|
|
}
|
|
|
|
|
2008-10-30 07:53:24 +01:00
|
|
|
if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
|
2007-11-23 06:29:32 +01:00
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: sunit and swidth options incompatible with the noalign option");
|
|
|
|
return EINVAL;
|
|
|
|
}
|
|
|
|
|
2009-06-08 15:33:32 +02:00
|
|
|
#ifndef CONFIG_XFS_QUOTA
|
|
|
|
if (XFS_IS_QUOTA_RUNNING(mp)) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: quota support not available in this kernel.");
|
|
|
|
return EINVAL;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2008-10-30 07:53:24 +01:00
|
|
|
if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
|
|
|
|
(mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) {
|
2007-11-23 06:29:32 +01:00
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: cannot mount with both project and group quota");
|
|
|
|
return EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: sunit and swidth must be specified together");
|
|
|
|
return EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (dsunit && (dswidth % dsunit != 0)) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
|
|
|
|
dswidth, dsunit);
|
|
|
|
return EINVAL;
|
|
|
|
}
|
|
|
|
|
2008-10-30 07:53:24 +01:00
|
|
|
done:
|
|
|
|
if (!(mp->m_flags & XFS_MOUNT_NOALIGN)) {
|
|
|
|
/*
|
|
|
|
* At this point the superblock has not been read
|
|
|
|
* in, therefore we do not know the block size.
|
|
|
|
* Before the mount call ends we will convert
|
|
|
|
* these to FSBs.
|
|
|
|
*/
|
2007-11-23 06:29:32 +01:00
|
|
|
if (dsunit) {
|
2008-10-30 07:53:24 +01:00
|
|
|
mp->m_dalign = dsunit;
|
|
|
|
mp->m_flags |= XFS_MOUNT_RETERR;
|
2007-11-23 06:29:32 +01:00
|
|
|
}
|
2008-10-30 07:53:24 +01:00
|
|
|
|
|
|
|
if (dswidth)
|
|
|
|
mp->m_swidth = dswidth;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (mp->m_logbufs != -1 &&
|
|
|
|
mp->m_logbufs != 0 &&
|
|
|
|
(mp->m_logbufs < XLOG_MIN_ICLOGS ||
|
|
|
|
mp->m_logbufs > XLOG_MAX_ICLOGS)) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: invalid logbufs value: %d [not %d-%d]",
|
|
|
|
mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
|
|
|
|
return XFS_ERROR(EINVAL);
|
|
|
|
}
|
|
|
|
if (mp->m_logbsize != -1 &&
|
|
|
|
mp->m_logbsize != 0 &&
|
|
|
|
(mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
|
|
|
|
mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
|
|
|
|
!is_power_of_2(mp->m_logbsize))) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
|
|
|
|
mp->m_logbsize);
|
|
|
|
return XFS_ERROR(EINVAL);
|
|
|
|
}
|
|
|
|
|
|
|
|
mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
|
|
|
|
if (!mp->m_fsname)
|
|
|
|
return ENOMEM;
|
|
|
|
mp->m_fsname_len = strlen(mp->m_fsname) + 1;
|
|
|
|
|
|
|
|
if (iosizelog) {
|
|
|
|
if (iosizelog > XFS_MAX_IO_LOG ||
|
|
|
|
iosizelog < XFS_MIN_IO_LOG) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: invalid log iosize: %d [not %d-%d]",
|
|
|
|
iosizelog, XFS_MIN_IO_LOG,
|
|
|
|
XFS_MAX_IO_LOG);
|
|
|
|
return XFS_ERROR(EINVAL);
|
|
|
|
}
|
|
|
|
|
|
|
|
mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
|
|
|
|
mp->m_readio_log = iosizelog;
|
|
|
|
mp->m_writeio_log = iosizelog;
|
2007-11-23 06:29:32 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct proc_xfs_info {
|
|
|
|
int flag;
|
|
|
|
char *str;
|
|
|
|
};
|
|
|
|
|
|
|
|
STATIC int
|
|
|
|
xfs_showargs(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
struct seq_file *m)
|
|
|
|
{
|
|
|
|
static struct proc_xfs_info xfs_info_set[] = {
|
|
|
|
/* the few simple ones we can get from the mount struct */
|
2008-02-29 03:58:40 +01:00
|
|
|
{ XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
|
2007-11-23 06:29:32 +01:00
|
|
|
{ XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
|
|
|
|
{ XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
|
|
|
|
{ XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
|
|
|
|
{ XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
|
|
|
|
{ XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
|
|
|
|
{ XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
|
|
|
|
{ XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
|
|
|
|
{ XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
|
xfs: Introduce delayed logging core code
The delayed logging code only changes in-memory structures and as
such can be enabled and disabled with a mount option. Add the mount
option and emit a warning that this is an experimental feature that
should not be used in production yet.
We also need infrastructure to track committed items that have not
yet been written to the log. This is what the Committed Item List
(CIL) is for.
The log item also needs to be extended to track the current log
vector, the associated memory buffer and it's location in the Commit
Item List. Extend the log item and log vector structures to enable
this tracking.
To maintain the current log format for transactions with delayed
logging, we need to introduce a checkpoint transaction and a context
for tracking each checkpoint from initiation to transaction
completion. This includes adding a log ticket for tracking space
log required/used by the context checkpoint.
To track all the changes we need an io vector array per log item,
rather than a single array for the entire transaction. Using the new
log vector structure for this requires two passes - the first to
allocate the log vector structures and chain them together, and the
second to fill them out. This log vector chain can then be passed
to the CIL for formatting, pinning and insertion into the CIL.
Formatting of the log vector chain is relatively simple - it's just
a loop over the iovecs on each log vector, but it is made slightly
more complex because we re-write the iovec after the copy to point
back at the memory buffer we just copied into.
This code also needs to pin log items. If the log item is not
already tracked in this checkpoint context, then it needs to be
pinned. Otherwise it is already pinned and we don't need to pin it
again.
The only other complexity is calculating the amount of new log space
the formatting has consumed. This needs to be accounted to the
transaction in progress, and the accounting is made more complex
becase we need also to steal space from it for log metadata in the
checkpoint transaction. Calculate all this at insert time and update
all the tickets, counters, etc correctly.
Once we've formatted all the log items in the transaction, attach
the busy extents to the checkpoint context so the busy extents live
until checkpoint completion and can be processed at that point in
time. Transactions can then be freed at this point in time.
Now we need to issue checkpoints - we are tracking the amount of log space
used by the items in the CIL, so we can trigger background checkpoints when the
space usage gets to a certain threshold. Otherwise, checkpoints need ot be
triggered when a log synchronisation point is reached - a log force event.
Because the log write code already handles chained log vectors, writing the
transaction is trivial, too. Construct a transaction header, add it
to the head of the chain and write it into the log, then issue a
commit record write. Then we can release the checkpoint log ticket
and attach the context to the log buffer so it can be called during
Io completion to complete the checkpoint.
We also need to allow for synchronising multiple in-flight
checkpoints. This is needed for two things - the first is to ensure
that checkpoint commit records appear in the log in the correct
sequence order (so they are replayed in the correct order). The
second is so that xfs_log_force_lsn() operates correctly and only
flushes and/or waits for the specific sequence it was provided with.
To do this we need a wait variable and a list tracking the
checkpoint commits in progress. We can walk this list and wait for
the checkpoints to change state or complete easily, an this provides
the necessary synchronisation for correct operation in both cases.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-05-21 06:37:18 +02:00
|
|
|
{ XFS_MOUNT_DELAYLOG, "," MNTOPT_DELAYLOG },
|
2007-11-23 06:29:32 +01:00
|
|
|
{ 0, NULL }
|
|
|
|
};
|
|
|
|
static struct proc_xfs_info xfs_info_unset[] = {
|
|
|
|
/* the few simple ones we can get from the mount struct */
|
|
|
|
{ XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
|
|
|
|
{ XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
|
|
|
|
{ XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
|
|
|
|
{ 0, NULL }
|
|
|
|
};
|
|
|
|
struct proc_xfs_info *xfs_infop;
|
|
|
|
|
|
|
|
for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
|
|
|
|
if (mp->m_flags & xfs_infop->flag)
|
|
|
|
seq_puts(m, xfs_infop->str);
|
|
|
|
}
|
|
|
|
for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
|
|
|
|
if (!(mp->m_flags & xfs_infop->flag))
|
|
|
|
seq_puts(m, xfs_infop->str);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
|
|
|
|
seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
|
|
|
|
(int)(1 << mp->m_writeio_log) >> 10);
|
|
|
|
|
|
|
|
if (mp->m_logbufs > 0)
|
|
|
|
seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
|
|
|
|
if (mp->m_logbsize > 0)
|
|
|
|
seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
|
|
|
|
|
|
|
|
if (mp->m_logname)
|
|
|
|
seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
|
|
|
|
if (mp->m_rtname)
|
|
|
|
seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
|
|
|
|
|
|
|
|
if (mp->m_dalign > 0)
|
|
|
|
seq_printf(m, "," MNTOPT_SUNIT "=%d",
|
|
|
|
(int)XFS_FSB_TO_BB(mp, mp->m_dalign));
|
|
|
|
if (mp->m_swidth > 0)
|
|
|
|
seq_printf(m, "," MNTOPT_SWIDTH "=%d",
|
|
|
|
(int)XFS_FSB_TO_BB(mp, mp->m_swidth));
|
|
|
|
|
|
|
|
if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
|
|
|
|
seq_puts(m, "," MNTOPT_USRQUOTA);
|
|
|
|
else if (mp->m_qflags & XFS_UQUOTA_ACCT)
|
|
|
|
seq_puts(m, "," MNTOPT_UQUOTANOENF);
|
|
|
|
|
xfs: xfs_showargs() reports group *and* project quotas enabled
If you enable group or project quotas on an XFS file system, then the
mount table presented through /proc/self/mounts erroneously shows
that both options are in effect for the file system. The root of
the problem is some bad logic in the xfs_showargs() function, which
is used to format the file system type-specific options in effect
for a file system.
The problem originated in this GIT commit:
Move platform specific mount option parse out of core XFS code
Date: 11/22/07
Author: Dave Chinner
SHA1 ID: a67d7c5f5d25d0b13a4dfb182697135b014fa478
For XFS quotas, project and group quota management are mutually
exclusive--only one can be in effect at a time. There are two
parts to managing quotas: aggregating usage information; and
enforcing limits. It is possible to have a quota in effect
(aggregating usage) but not enforced.
These features are recorded on an XFS mount point using these flags:
XFS_PQUOTA_ACCT - Project quotas are aggregated
XFS_GQUOTA_ACCT - Group quotas are aggregated
XFS_OQUOTA_ENFD - Project/group quotas are enforced
The code in error is in fs/xfs/linux-2.6/xfs_super.c:
if (mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
seq_puts(m, "," MNTOPT_PRJQUOTA);
else if (mp->m_qflags & XFS_PQUOTA_ACCT)
seq_puts(m, "," MNTOPT_PQUOTANOENF);
if (mp->m_qflags & (XFS_GQUOTA_ACCT|XFS_OQUOTA_ENFD))
seq_puts(m, "," MNTOPT_GRPQUOTA);
else if (mp->m_qflags & XFS_GQUOTA_ACCT)
seq_puts(m, "," MNTOPT_GQUOTANOENF);
The problem is that XFS_OQUOTA_ENFD will be set in mp->m_qflags
if either group or project quotas are enforced, and as a result
both MNTOPT_PRJQUOTA and MNTOPT_GRPQUOTA will be shown as mount
options.
Signed-off-by: Alex Elder <aelder@sgi.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Felix Blyakher <felixb@sgi.com>
2009-09-03 00:02:24 +02:00
|
|
|
/* Either project or group quotas can be active, not both */
|
|
|
|
|
|
|
|
if (mp->m_qflags & XFS_PQUOTA_ACCT) {
|
|
|
|
if (mp->m_qflags & XFS_OQUOTA_ENFD)
|
|
|
|
seq_puts(m, "," MNTOPT_PRJQUOTA);
|
|
|
|
else
|
|
|
|
seq_puts(m, "," MNTOPT_PQUOTANOENF);
|
|
|
|
} else if (mp->m_qflags & XFS_GQUOTA_ACCT) {
|
|
|
|
if (mp->m_qflags & XFS_OQUOTA_ENFD)
|
|
|
|
seq_puts(m, "," MNTOPT_GRPQUOTA);
|
|
|
|
else
|
|
|
|
seq_puts(m, "," MNTOPT_GQUOTANOENF);
|
|
|
|
}
|
2007-11-23 06:29:32 +01:00
|
|
|
|
|
|
|
if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
|
|
|
|
seq_puts(m, "," MNTOPT_NOQUOTA);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
__uint64_t
|
|
|
|
xfs_max_file_offset(
|
|
|
|
unsigned int blockshift)
|
|
|
|
{
|
|
|
|
unsigned int pagefactor = 1;
|
|
|
|
unsigned int bitshift = BITS_PER_LONG - 1;
|
|
|
|
|
|
|
|
/* Figure out maximum filesize, on Linux this can depend on
|
|
|
|
* the filesystem blocksize (on 32 bit platforms).
|
2010-10-06 10:47:23 +02:00
|
|
|
* __block_write_begin does this in an [unsigned] long...
|
2005-04-17 00:20:36 +02:00
|
|
|
* page->index << (PAGE_CACHE_SHIFT - bbits)
|
|
|
|
* So, for page sized blocks (4K on 32 bit platforms),
|
|
|
|
* this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
|
|
|
|
* (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
|
|
|
|
* but for smaller blocksizes it is less (bbits = log2 bsize).
|
|
|
|
* Note1: get_block_t takes a long (implicit cast from above)
|
|
|
|
* Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
|
|
|
|
* can optionally convert the [unsigned] long from above into
|
|
|
|
* an [unsigned] long long.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#if BITS_PER_LONG == 32
|
2009-06-19 08:08:50 +02:00
|
|
|
# if defined(CONFIG_LBDAF)
|
2005-04-17 00:20:36 +02:00
|
|
|
ASSERT(sizeof(sector_t) == 8);
|
|
|
|
pagefactor = PAGE_CACHE_SIZE;
|
|
|
|
bitshift = BITS_PER_LONG;
|
|
|
|
# else
|
|
|
|
pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
|
|
|
|
# endif
|
|
|
|
#endif
|
|
|
|
|
|
|
|
return (((__uint64_t)pagefactor) << bitshift) - 1;
|
|
|
|
}
|
|
|
|
|
2009-03-04 19:34:10 +01:00
|
|
|
STATIC int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_blkdev_get(
|
|
|
|
xfs_mount_t *mp,
|
|
|
|
const char *name,
|
|
|
|
struct block_device **bdevp)
|
|
|
|
{
|
|
|
|
int error = 0;
|
|
|
|
|
2008-02-23 01:50:45 +01:00
|
|
|
*bdevp = open_bdev_exclusive(name, FMODE_READ|FMODE_WRITE, mp);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (IS_ERR(*bdevp)) {
|
|
|
|
error = PTR_ERR(*bdevp);
|
|
|
|
printk("XFS: Invalid device [%s], error=%d\n", name, error);
|
|
|
|
}
|
|
|
|
|
|
|
|
return -error;
|
|
|
|
}
|
|
|
|
|
2009-03-04 19:34:10 +01:00
|
|
|
STATIC void
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_blkdev_put(
|
|
|
|
struct block_device *bdev)
|
|
|
|
{
|
|
|
|
if (bdev)
|
2008-02-23 01:50:45 +01:00
|
|
|
close_bdev_exclusive(bdev, FMODE_READ|FMODE_WRITE);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2005-11-02 00:26:59 +01:00
|
|
|
/*
|
|
|
|
* Try to write out the superblock using barriers.
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_barrier_test(
|
|
|
|
xfs_mount_t *mp)
|
|
|
|
{
|
|
|
|
xfs_buf_t *sbp = xfs_getsb(mp, 0);
|
|
|
|
int error;
|
|
|
|
|
|
|
|
XFS_BUF_UNDONE(sbp);
|
|
|
|
XFS_BUF_UNREAD(sbp);
|
|
|
|
XFS_BUF_UNDELAYWRITE(sbp);
|
|
|
|
XFS_BUF_WRITE(sbp);
|
|
|
|
XFS_BUF_UNASYNC(sbp);
|
|
|
|
XFS_BUF_ORDERED(sbp);
|
|
|
|
|
|
|
|
xfsbdstrat(mp, sbp);
|
2010-10-06 20:41:18 +02:00
|
|
|
error = xfs_buf_iowait(sbp);
|
2005-11-02 00:26:59 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Clear all the flags we set and possible error state in the
|
|
|
|
* buffer. We only did the write to try out whether barriers
|
|
|
|
* worked and shouldn't leave any traces in the superblock
|
|
|
|
* buffer.
|
|
|
|
*/
|
|
|
|
XFS_BUF_DONE(sbp);
|
|
|
|
XFS_BUF_ERROR(sbp, 0);
|
|
|
|
XFS_BUF_UNORDERED(sbp);
|
|
|
|
|
|
|
|
xfs_buf_relse(sbp);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2009-07-02 07:09:33 +02:00
|
|
|
STATIC void
|
2005-11-02 00:26:59 +01:00
|
|
|
xfs_mountfs_check_barriers(xfs_mount_t *mp)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (mp->m_logdev_targp != mp->m_ddev_targp) {
|
|
|
|
xfs_fs_cmn_err(CE_NOTE, mp,
|
|
|
|
"Disabling barriers, not supported with external log device");
|
|
|
|
mp->m_flags &= ~XFS_MOUNT_BARRIER;
|
2006-01-11 05:27:18 +01:00
|
|
|
return;
|
2005-11-02 00:26:59 +01:00
|
|
|
}
|
|
|
|
|
2006-07-28 09:05:13 +02:00
|
|
|
if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
|
|
|
|
xfs_fs_cmn_err(CE_NOTE, mp,
|
|
|
|
"Disabling barriers, underlying device is readonly");
|
|
|
|
mp->m_flags &= ~XFS_MOUNT_BARRIER;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2005-11-02 00:26:59 +01:00
|
|
|
error = xfs_barrier_test(mp);
|
|
|
|
if (error) {
|
|
|
|
xfs_fs_cmn_err(CE_NOTE, mp,
|
|
|
|
"Disabling barriers, trial barrier write failed");
|
|
|
|
mp->m_flags &= ~XFS_MOUNT_BARRIER;
|
2006-01-11 05:27:18 +01:00
|
|
|
return;
|
2005-11-02 00:26:59 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
xfs_blkdev_issue_flush(
|
|
|
|
xfs_buftarg_t *buftarg)
|
|
|
|
{
|
2010-09-16 20:51:46 +02:00
|
|
|
blkdev_issue_flush(buftarg->bt_bdev, GFP_KERNEL, NULL);
|
2005-11-02 00:26:59 +01:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-05-20 03:31:13 +02:00
|
|
|
STATIC void
|
|
|
|
xfs_close_devices(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
|
2008-07-18 09:13:12 +02:00
|
|
|
struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
|
2009-03-03 20:48:37 +01:00
|
|
|
xfs_free_buftarg(mp, mp->m_logdev_targp);
|
2008-07-18 09:13:12 +02:00
|
|
|
xfs_blkdev_put(logdev);
|
2008-05-20 03:31:13 +02:00
|
|
|
}
|
|
|
|
if (mp->m_rtdev_targp) {
|
2008-07-18 09:13:12 +02:00
|
|
|
struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
|
2009-03-03 20:48:37 +01:00
|
|
|
xfs_free_buftarg(mp, mp->m_rtdev_targp);
|
2008-07-18 09:13:12 +02:00
|
|
|
xfs_blkdev_put(rtdev);
|
2008-05-20 03:31:13 +02:00
|
|
|
}
|
2009-03-03 20:48:37 +01:00
|
|
|
xfs_free_buftarg(mp, mp->m_ddev_targp);
|
2008-05-20 03:31:13 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The file system configurations are:
|
|
|
|
* (1) device (partition) with data and internal log
|
|
|
|
* (2) logical volume with data and log subvolumes.
|
|
|
|
* (3) logical volume with data, log, and realtime subvolumes.
|
|
|
|
*
|
|
|
|
* We only have to handle opening the log and realtime volumes here if
|
|
|
|
* they are present. The data subvolume has already been opened by
|
|
|
|
* get_sb_bdev() and is stored in sb->s_bdev.
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_open_devices(
|
2008-10-30 07:53:24 +01:00
|
|
|
struct xfs_mount *mp)
|
2008-05-20 03:31:13 +02:00
|
|
|
{
|
|
|
|
struct block_device *ddev = mp->m_super->s_bdev;
|
|
|
|
struct block_device *logdev = NULL, *rtdev = NULL;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Open real time and log devices - order is important.
|
|
|
|
*/
|
2008-10-30 07:53:24 +01:00
|
|
|
if (mp->m_logname) {
|
|
|
|
error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
|
2008-05-20 03:31:13 +02:00
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2008-10-30 07:53:24 +01:00
|
|
|
if (mp->m_rtname) {
|
|
|
|
error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
|
2008-05-20 03:31:13 +02:00
|
|
|
if (error)
|
|
|
|
goto out_close_logdev;
|
|
|
|
|
|
|
|
if (rtdev == ddev || rtdev == logdev) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
|
|
|
|
error = EINVAL;
|
|
|
|
goto out_close_rtdev;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Setup xfs_mount buffer target pointers
|
|
|
|
*/
|
|
|
|
error = ENOMEM;
|
2010-09-22 02:47:20 +02:00
|
|
|
mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, 0, mp->m_fsname);
|
2008-05-20 03:31:13 +02:00
|
|
|
if (!mp->m_ddev_targp)
|
|
|
|
goto out_close_rtdev;
|
|
|
|
|
|
|
|
if (rtdev) {
|
2010-09-22 02:47:20 +02:00
|
|
|
mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, 1,
|
|
|
|
mp->m_fsname);
|
2008-05-20 03:31:13 +02:00
|
|
|
if (!mp->m_rtdev_targp)
|
|
|
|
goto out_free_ddev_targ;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (logdev && logdev != ddev) {
|
2010-09-22 02:47:20 +02:00
|
|
|
mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, 1,
|
|
|
|
mp->m_fsname);
|
2008-05-20 03:31:13 +02:00
|
|
|
if (!mp->m_logdev_targp)
|
|
|
|
goto out_free_rtdev_targ;
|
|
|
|
} else {
|
|
|
|
mp->m_logdev_targp = mp->m_ddev_targp;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_free_rtdev_targ:
|
|
|
|
if (mp->m_rtdev_targp)
|
2009-03-03 20:48:37 +01:00
|
|
|
xfs_free_buftarg(mp, mp->m_rtdev_targp);
|
2008-05-20 03:31:13 +02:00
|
|
|
out_free_ddev_targ:
|
2009-03-03 20:48:37 +01:00
|
|
|
xfs_free_buftarg(mp, mp->m_ddev_targp);
|
2008-05-20 03:31:13 +02:00
|
|
|
out_close_rtdev:
|
|
|
|
if (rtdev)
|
|
|
|
xfs_blkdev_put(rtdev);
|
|
|
|
out_close_logdev:
|
|
|
|
if (logdev && logdev != ddev)
|
|
|
|
xfs_blkdev_put(logdev);
|
|
|
|
out:
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2008-05-20 07:10:36 +02:00
|
|
|
/*
|
|
|
|
* Setup xfs_mount buffer target pointers based on superblock
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_setup_devices(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
int error;
|
2008-05-20 03:31:13 +02:00
|
|
|
|
2008-05-20 07:10:36 +02:00
|
|
|
error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
|
|
|
|
mp->m_sb.sb_sectsize);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
|
|
|
|
unsigned int log_sector_size = BBSIZE;
|
|
|
|
|
|
|
|
if (xfs_sb_version_hassector(&mp->m_sb))
|
|
|
|
log_sector_size = mp->m_sb.sb_logsectsize;
|
|
|
|
error = xfs_setsize_buftarg(mp->m_logdev_targp,
|
|
|
|
mp->m_sb.sb_blocksize,
|
|
|
|
log_sector_size);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
if (mp->m_rtdev_targp) {
|
|
|
|
error = xfs_setsize_buftarg(mp->m_rtdev_targp,
|
|
|
|
mp->m_sb.sb_blocksize,
|
|
|
|
mp->m_sb.sb_sectsize);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2008-05-20 03:31:13 +02:00
|
|
|
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
/*
|
|
|
|
* XFS AIL push thread support
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
xfsaild_wakeup(
|
2008-10-30 07:38:26 +01:00
|
|
|
struct xfs_ail *ailp,
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
xfs_lsn_t threshold_lsn)
|
|
|
|
{
|
2010-12-17 10:08:04 +01:00
|
|
|
/* only ever move the target forwards */
|
|
|
|
if (XFS_LSN_CMP(threshold_lsn, ailp->xa_target) > 0) {
|
|
|
|
ailp->xa_target = threshold_lsn;
|
|
|
|
wake_up_process(ailp->xa_task);
|
|
|
|
}
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
}
|
|
|
|
|
2009-03-04 19:34:10 +01:00
|
|
|
STATIC int
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
xfsaild(
|
|
|
|
void *data)
|
|
|
|
{
|
2008-10-30 07:38:26 +01:00
|
|
|
struct xfs_ail *ailp = data;
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
xfs_lsn_t last_pushed_lsn = 0;
|
2010-01-11 12:49:58 +01:00
|
|
|
long tout = 0; /* milliseconds */
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
|
|
|
|
while (!kthread_should_stop()) {
|
2010-12-17 10:08:04 +01:00
|
|
|
/*
|
|
|
|
* for short sleeps indicating congestion, don't allow us to
|
|
|
|
* get woken early. Otherwise all we do is bang on the AIL lock
|
|
|
|
* without making progress.
|
|
|
|
*/
|
|
|
|
if (tout && tout <= 20)
|
|
|
|
__set_current_state(TASK_KILLABLE);
|
|
|
|
else
|
|
|
|
__set_current_state(TASK_INTERRUPTIBLE);
|
|
|
|
schedule_timeout(tout ?
|
|
|
|
msecs_to_jiffies(tout) : MAX_SCHEDULE_TIMEOUT);
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
|
|
|
|
/* swsusp */
|
|
|
|
try_to_freeze();
|
|
|
|
|
2008-10-30 07:38:26 +01:00
|
|
|
ASSERT(ailp->xa_mount->m_log);
|
|
|
|
if (XFS_FORCED_SHUTDOWN(ailp->xa_mount))
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
continue;
|
|
|
|
|
2008-10-30 07:38:26 +01:00
|
|
|
tout = xfsaild_push(ailp, &last_pushed_lsn);
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
} /* xfsaild */
|
|
|
|
|
|
|
|
int
|
|
|
|
xfsaild_start(
|
2008-10-30 07:38:26 +01:00
|
|
|
struct xfs_ail *ailp)
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
{
|
2008-10-30 07:38:26 +01:00
|
|
|
ailp->xa_target = 0;
|
2010-03-22 23:52:55 +01:00
|
|
|
ailp->xa_task = kthread_run(xfsaild, ailp, "xfsaild/%s",
|
|
|
|
ailp->xa_mount->m_fsname);
|
2008-10-30 07:38:26 +01:00
|
|
|
if (IS_ERR(ailp->xa_task))
|
|
|
|
return -PTR_ERR(ailp->xa_task);
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
xfsaild_stop(
|
2008-10-30 07:38:26 +01:00
|
|
|
struct xfs_ail *ailp)
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
{
|
2008-10-30 07:38:26 +01:00
|
|
|
kthread_stop(ailp->xa_task);
|
[XFS] Move AIL pushing into it's own thread
When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
2008-02-05 02:13:32 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2008-10-30 07:36:14 +01:00
|
|
|
/* Catch misguided souls that try to use this interface on XFS */
|
2005-04-17 00:20:36 +02:00
|
|
|
STATIC struct inode *
|
2006-03-14 04:06:18 +01:00
|
|
|
xfs_fs_alloc_inode(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct super_block *sb)
|
|
|
|
{
|
2008-10-30 07:36:14 +01:00
|
|
|
BUG();
|
2008-10-30 07:36:52 +01:00
|
|
|
return NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2008-10-30 07:36:14 +01:00
|
|
|
/*
|
2008-10-30 07:36:40 +01:00
|
|
|
* Now that the generic code is guaranteed not to be accessing
|
|
|
|
* the linux inode, we can reclaim the inode.
|
2008-10-30 07:36:14 +01:00
|
|
|
*/
|
2005-04-17 00:20:36 +02:00
|
|
|
STATIC void
|
2006-03-14 04:06:18 +01:00
|
|
|
xfs_fs_destroy_inode(
|
2009-09-29 15:48:56 +02:00
|
|
|
struct inode *inode)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2009-09-29 15:48:56 +02:00
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
|
|
|
|
2010-06-24 03:57:09 +02:00
|
|
|
trace_xfs_destroy_inode(ip);
|
2008-10-30 07:36:40 +01:00
|
|
|
|
|
|
|
XFS_STATS_INC(vn_reclaim);
|
2009-09-29 15:48:56 +02:00
|
|
|
|
|
|
|
/* bad inode, get out here ASAP */
|
|
|
|
if (is_bad_inode(inode))
|
|
|
|
goto out_reclaim;
|
|
|
|
|
|
|
|
xfs_ioend_wait(ip);
|
|
|
|
|
|
|
|
ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We should never get here with one of the reclaim flags already set.
|
|
|
|
*/
|
|
|
|
ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
|
|
|
|
ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
|
|
|
|
|
|
|
|
/*
|
2010-01-11 00:51:47 +01:00
|
|
|
* We always use background reclaim here because even if the
|
|
|
|
* inode is clean, it still may be under IO and hence we have
|
|
|
|
* to take the flush lock. The background reclaim path handles
|
|
|
|
* this more efficiently than we can here, so simply let background
|
|
|
|
* reclaim tear down all inodes.
|
2009-09-29 15:48:56 +02:00
|
|
|
*/
|
|
|
|
out_reclaim:
|
2010-01-11 00:51:47 +01:00
|
|
|
xfs_inode_set_reclaim_tag(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2008-10-30 06:11:59 +01:00
|
|
|
/*
|
|
|
|
* Slab object creation initialisation for the XFS inode.
|
|
|
|
* This covers only the idempotent fields in the XFS inode;
|
|
|
|
* all other fields need to be initialised on allocation
|
|
|
|
* from the slab. This avoids the need to repeatedly intialise
|
|
|
|
* fields in the xfs inode that left in the initialise state
|
|
|
|
* when freeing the inode.
|
|
|
|
*/
|
2008-10-30 07:36:14 +01:00
|
|
|
STATIC void
|
|
|
|
xfs_fs_inode_init_once(
|
2008-10-30 06:11:59 +01:00
|
|
|
void *inode)
|
|
|
|
{
|
|
|
|
struct xfs_inode *ip = inode;
|
|
|
|
|
|
|
|
memset(ip, 0, sizeof(struct xfs_inode));
|
2008-10-30 07:36:14 +01:00
|
|
|
|
|
|
|
/* vfs inode */
|
|
|
|
inode_init_once(VFS_I(ip));
|
|
|
|
|
|
|
|
/* xfs inode */
|
2008-10-30 06:11:59 +01:00
|
|
|
atomic_set(&ip->i_iocount, 0);
|
|
|
|
atomic_set(&ip->i_pincount, 0);
|
|
|
|
spin_lock_init(&ip->i_flags_lock);
|
|
|
|
init_waitqueue_head(&ip->i_ipin_wait);
|
|
|
|
/*
|
|
|
|
* Because we want to use a counting completion, complete
|
|
|
|
* the flush completion once to allow a single access to
|
|
|
|
* the flush completion without blocking.
|
|
|
|
*/
|
|
|
|
init_completion(&ip->i_flush);
|
|
|
|
complete(&ip->i_flush);
|
|
|
|
|
|
|
|
mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
|
|
|
|
"xfsino", ip->i_ino);
|
|
|
|
}
|
|
|
|
|
2009-10-06 22:29:26 +02:00
|
|
|
/*
|
|
|
|
* Dirty the XFS inode when mark_inode_dirty_sync() is called so that
|
2010-09-28 04:27:25 +02:00
|
|
|
* we catch unlogged VFS level updates to the inode.
|
2009-10-06 22:29:26 +02:00
|
|
|
*
|
|
|
|
* We need the barrier() to maintain correct ordering between unlogged
|
|
|
|
* updates and the transaction commit code that clears the i_update_core
|
|
|
|
* field. This requires all updates to be completed before marking the
|
|
|
|
* inode dirty.
|
|
|
|
*/
|
|
|
|
STATIC void
|
|
|
|
xfs_fs_dirty_inode(
|
|
|
|
struct inode *inode)
|
|
|
|
{
|
|
|
|
barrier();
|
|
|
|
XFS_I(inode)->i_update_core = 1;
|
|
|
|
}
|
|
|
|
|
2010-02-09 01:43:49 +01:00
|
|
|
STATIC int
|
|
|
|
xfs_log_inode(
|
|
|
|
struct xfs_inode *ip)
|
|
|
|
{
|
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
|
|
|
struct xfs_trans *tp;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_SHARED);
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_FSYNC_TS);
|
|
|
|
error = xfs_trans_reserve(tp, 0, XFS_FSYNC_TS_LOG_RES(mp), 0, 0, 0);
|
|
|
|
|
|
|
|
if (error) {
|
|
|
|
xfs_trans_cancel(tp, 0);
|
|
|
|
/* we need to return with the lock hold shared */
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_SHARED);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Note - it's possible that we might have pushed ourselves out of the
|
|
|
|
* way during trans_reserve which would flush the inode. But there's
|
|
|
|
* no guarantee that the inode buffer has actually gone out yet (it's
|
|
|
|
* delwri). Plus the buffer could be pinned anyway if it's part of
|
|
|
|
* an inode in another recent transaction. So we play it safe and
|
|
|
|
* fire off the transaction anyway.
|
|
|
|
*/
|
2010-06-24 03:36:58 +02:00
|
|
|
xfs_trans_ijoin(tp, ip);
|
2010-02-09 01:43:49 +01:00
|
|
|
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
|
|
|
|
error = xfs_trans_commit(tp, 0);
|
|
|
|
xfs_ilock_demote(ip, XFS_ILOCK_EXCL);
|
|
|
|
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
STATIC int
|
2006-03-14 04:06:18 +01:00
|
|
|
xfs_fs_write_inode(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct inode *inode,
|
2010-03-05 09:21:37 +01:00
|
|
|
struct writeback_control *wbc)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2008-11-28 04:23:33 +01:00
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
2009-02-04 09:36:19 +01:00
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
2010-02-09 01:43:49 +01:00
|
|
|
int error = EAGAIN;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2010-06-24 03:57:09 +02:00
|
|
|
trace_xfs_write_inode(ip);
|
2009-02-04 09:36:19 +01:00
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
|
|
|
return XFS_ERROR(EIO);
|
|
|
|
|
2010-03-05 09:21:37 +01:00
|
|
|
if (wbc->sync_mode == WB_SYNC_ALL) {
|
2010-02-09 01:43:49 +01:00
|
|
|
/*
|
2010-06-24 03:39:25 +02:00
|
|
|
* Make sure the inode has made it it into the log. Instead
|
|
|
|
* of forcing it all the way to stable storage using a
|
|
|
|
* synchronous transaction we let the log force inside the
|
|
|
|
* ->sync_fs call do that for thus, which reduces the number
|
|
|
|
* of synchronous log foces dramatically.
|
2010-02-09 01:43:49 +01:00
|
|
|
*/
|
2010-04-20 09:00:59 +02:00
|
|
|
xfs_ioend_wait(ip);
|
2009-02-04 09:36:19 +01:00
|
|
|
xfs_ilock(ip, XFS_ILOCK_SHARED);
|
2010-02-09 01:43:49 +01:00
|
|
|
if (ip->i_update_core) {
|
|
|
|
error = xfs_log_inode(ip);
|
|
|
|
if (error)
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
2009-02-04 09:36:19 +01:00
|
|
|
} else {
|
2010-02-09 01:43:49 +01:00
|
|
|
/*
|
|
|
|
* We make this non-blocking if the inode is contended, return
|
|
|
|
* EAGAIN to indicate to the caller that they did not succeed.
|
|
|
|
* This prevents the flush path from blocking on inodes inside
|
2010-06-24 03:39:25 +02:00
|
|
|
* another operation right now, they get caught later by
|
|
|
|
* xfs_sync.
|
2010-02-09 01:43:49 +01:00
|
|
|
*/
|
2009-02-04 09:36:19 +01:00
|
|
|
if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
|
|
|
|
goto out;
|
2010-02-09 01:43:49 +01:00
|
|
|
|
2010-06-24 03:39:25 +02:00
|
|
|
if (xfs_ipincount(ip) || !xfs_iflock_nowait(ip))
|
|
|
|
goto out_unlock;
|
2009-02-04 09:36:19 +01:00
|
|
|
|
2010-06-24 03:39:25 +02:00
|
|
|
/*
|
|
|
|
* Now we have the flush lock and the inode is not pinned, we
|
|
|
|
* can check if the inode is really clean as we know that
|
|
|
|
* there are no pending transaction completions, it is not
|
|
|
|
* waiting on the delayed write queue and there is no IO in
|
|
|
|
* progress.
|
|
|
|
*/
|
|
|
|
if (xfs_inode_clean(ip)) {
|
|
|
|
xfs_ifunlock(ip);
|
|
|
|
error = 0;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
error = xfs_iflush(ip, 0);
|
2009-02-04 09:36:19 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_SHARED);
|
|
|
|
out:
|
2007-10-12 03:13:35 +02:00
|
|
|
/*
|
|
|
|
* if we failed to write out the inode then mark
|
|
|
|
* it dirty again so we'll try again later.
|
|
|
|
*/
|
|
|
|
if (error)
|
2008-11-28 04:23:33 +01:00
|
|
|
xfs_mark_inode_dirty_sync(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
return -error;
|
|
|
|
}
|
|
|
|
|
|
|
|
STATIC void
|
2010-06-07 20:34:48 +02:00
|
|
|
xfs_fs_evict_inode(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct inode *inode)
|
|
|
|
{
|
2007-08-29 03:46:47 +02:00
|
|
|
xfs_inode_t *ip = XFS_I(inode);
|
2005-09-05 00:23:54 +02:00
|
|
|
|
2010-06-07 20:34:48 +02:00
|
|
|
trace_xfs_evict_inode(ip);
|
2010-06-24 03:57:09 +02:00
|
|
|
|
2010-06-07 20:34:48 +02:00
|
|
|
truncate_inode_pages(&inode->i_data, 0);
|
|
|
|
end_writeback(inode);
|
2008-10-30 07:36:40 +01:00
|
|
|
XFS_STATS_INC(vn_rele);
|
|
|
|
XFS_STATS_INC(vn_remove);
|
|
|
|
XFS_STATS_DEC(vn_active);
|
|
|
|
|
2009-10-19 06:05:26 +02:00
|
|
|
/*
|
|
|
|
* The iolock is used by the file system to coordinate reads,
|
|
|
|
* writes, and block truncates. Up to this point the lock
|
|
|
|
* protected concurrent accesses by users of the inode. But
|
|
|
|
* from here forward we're doing some final processing of the
|
|
|
|
* inode because we're done with it, and although we reuse the
|
|
|
|
* iolock for protection it is really a distinct lock class
|
|
|
|
* (in the lockdep sense) from before. To keep lockdep happy
|
|
|
|
* (and basically indicate what we are doing), we explicitly
|
|
|
|
* re-init the iolock here.
|
|
|
|
*/
|
|
|
|
ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
|
|
|
|
mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
|
2010-12-23 01:57:13 +01:00
|
|
|
lockdep_set_class_and_name(&ip->i_iolock.mr_lock,
|
|
|
|
&xfs_iolock_reclaimable, "xfs_iolock_reclaimable");
|
2009-10-19 06:05:26 +02:00
|
|
|
|
2008-10-30 07:36:40 +01:00
|
|
|
xfs_inactive(ip);
|
2005-09-05 00:23:54 +02:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-08-13 08:04:05 +02:00
|
|
|
STATIC void
|
|
|
|
xfs_free_fsname(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
kfree(mp->m_fsname);
|
|
|
|
kfree(mp->m_rtname);
|
|
|
|
kfree(mp->m_logname);
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
STATIC void
|
2006-03-14 04:06:18 +01:00
|
|
|
xfs_fs_put_super(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct super_block *sb)
|
|
|
|
{
|
2007-08-30 09:20:39 +02:00
|
|
|
struct xfs_mount *mp = XFS_M(sb);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2010-07-12 08:40:58 +02:00
|
|
|
/*
|
|
|
|
* Unregister the memory shrinker before we tear down the mount
|
|
|
|
* structure so we don't have memory reclaim racing with us here.
|
|
|
|
*/
|
|
|
|
xfs_inode_shrinker_unregister(mp);
|
2008-10-30 07:06:18 +01:00
|
|
|
xfs_syncd_stop(mp);
|
2009-06-08 15:35:48 +02:00
|
|
|
|
2008-05-20 03:30:52 +02:00
|
|
|
/*
|
|
|
|
* Blow away any referenced inode in the filestreams cache.
|
|
|
|
* This can and will cause log traffic as inodes go inactive
|
|
|
|
* here.
|
|
|
|
*/
|
|
|
|
xfs_filestream_unmount(mp);
|
|
|
|
|
|
|
|
XFS_bflush(mp->m_ddev_targp);
|
|
|
|
|
2008-05-20 03:31:13 +02:00
|
|
|
xfs_unmountfs(mp);
|
2008-08-13 08:50:21 +02:00
|
|
|
xfs_freesb(mp);
|
2008-05-20 07:10:52 +02:00
|
|
|
xfs_icsb_destroy_counters(mp);
|
2008-05-20 03:31:13 +02:00
|
|
|
xfs_close_devices(mp);
|
2008-08-13 08:04:05 +02:00
|
|
|
xfs_free_fsname(mp);
|
2008-05-20 07:10:52 +02:00
|
|
|
kfree(mp);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
STATIC int
|
2009-10-06 22:29:28 +02:00
|
|
|
xfs_fs_sync_fs(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct super_block *sb,
|
|
|
|
int wait)
|
|
|
|
{
|
2007-08-30 09:20:39 +02:00
|
|
|
struct xfs_mount *mp = XFS_M(sb);
|
2006-06-09 08:48:30 +02:00
|
|
|
int error;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-10-12 03:13:35 +02:00
|
|
|
/*
|
2009-10-06 22:29:28 +02:00
|
|
|
* Not much we can do for the first async pass. Writing out the
|
|
|
|
* superblock would be counter-productive as we are going to redirty
|
|
|
|
* when writing out other data and metadata (and writing out a single
|
|
|
|
* block is quite fast anyway).
|
|
|
|
*
|
|
|
|
* Try to asynchronously kick off quota syncing at least.
|
2007-10-12 03:13:35 +02:00
|
|
|
*/
|
2009-10-06 22:29:28 +02:00
|
|
|
if (!wait) {
|
|
|
|
xfs_qm_sync(mp, SYNC_TRYLOCK);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
error = xfs_quiesce_data(mp);
|
|
|
|
if (error)
|
|
|
|
return -error;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2009-10-06 22:29:28 +02:00
|
|
|
if (laptop_mode) {
|
2007-08-30 09:21:22 +02:00
|
|
|
int prev_sync_seq = mp->m_sync_seq;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The disk must be active because we're syncing.
|
|
|
|
* We schedule xfssyncd now (now that the disk is
|
|
|
|
* active) instead of later (when it might not be).
|
|
|
|
*/
|
2007-08-30 09:21:22 +02:00
|
|
|
wake_up_process(mp->m_sync_task);
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* We have to wait for the sync iteration to complete.
|
|
|
|
* If we don't, the disk activity caused by the sync
|
|
|
|
* will come after the sync is completed, and that
|
|
|
|
* triggers another sync from laptop mode.
|
|
|
|
*/
|
2007-08-30 09:21:22 +02:00
|
|
|
wait_event(mp->m_wait_single_sync_task,
|
|
|
|
mp->m_sync_seq != prev_sync_seq);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2009-10-06 22:29:28 +02:00
|
|
|
return 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
STATIC int
|
2006-03-14 04:06:18 +01:00
|
|
|
xfs_fs_statfs(
|
2006-06-23 11:02:58 +02:00
|
|
|
struct dentry *dentry,
|
2005-04-17 00:20:36 +02:00
|
|
|
struct kstatfs *statp)
|
|
|
|
{
|
2007-10-11 10:09:40 +02:00
|
|
|
struct xfs_mount *mp = XFS_M(dentry->d_sb);
|
|
|
|
xfs_sb_t *sbp = &mp->m_sb;
|
2009-06-08 15:33:32 +02:00
|
|
|
struct xfs_inode *ip = XFS_I(dentry->d_inode);
|
2007-10-11 10:09:40 +02:00
|
|
|
__uint64_t fakeinos, id;
|
|
|
|
xfs_extlen_t lsize;
|
2010-08-24 03:46:05 +02:00
|
|
|
__int64_t ffree;
|
2007-10-11 10:09:40 +02:00
|
|
|
|
|
|
|
statp->f_type = XFS_SB_MAGIC;
|
|
|
|
statp->f_namelen = MAXNAMELEN - 1;
|
|
|
|
|
|
|
|
id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
|
|
|
|
statp->f_fsid.val[0] = (u32)id;
|
|
|
|
statp->f_fsid.val[1] = (u32)(id >> 32);
|
|
|
|
|
2008-04-22 09:34:37 +02:00
|
|
|
xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
|
2007-10-11 10:09:40 +02:00
|
|
|
|
|
|
|
spin_lock(&mp->m_sb_lock);
|
|
|
|
statp->f_bsize = sbp->sb_blocksize;
|
|
|
|
lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
|
|
|
|
statp->f_blocks = sbp->sb_dblocks - lsize;
|
|
|
|
statp->f_bfree = statp->f_bavail =
|
|
|
|
sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
|
|
|
|
fakeinos = statp->f_bfree << sbp->sb_inopblog;
|
|
|
|
statp->f_files =
|
|
|
|
MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
|
|
|
|
if (mp->m_maxicount)
|
2009-03-29 09:51:08 +02:00
|
|
|
statp->f_files = min_t(typeof(statp->f_files),
|
|
|
|
statp->f_files,
|
|
|
|
mp->m_maxicount);
|
2010-08-24 03:46:05 +02:00
|
|
|
|
|
|
|
/* make sure statp->f_ffree does not underflow */
|
|
|
|
ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
|
|
|
|
statp->f_ffree = max_t(__int64_t, ffree, 0);
|
|
|
|
|
2007-10-11 10:09:40 +02:00
|
|
|
spin_unlock(&mp->m_sb_lock);
|
|
|
|
|
2009-06-08 15:33:32 +02:00
|
|
|
if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) ||
|
|
|
|
((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))) ==
|
|
|
|
(XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
|
|
|
|
xfs_qm_statvfs(ip, statp);
|
2007-10-11 10:09:40 +02:00
|
|
|
return 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2010-02-05 23:59:53 +01:00
|
|
|
STATIC void
|
|
|
|
xfs_save_resvblks(struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
__uint64_t resblks = 0;
|
|
|
|
|
|
|
|
mp->m_resblks_save = mp->m_resblks;
|
|
|
|
xfs_reserve_blocks(mp, &resblks, NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
STATIC void
|
|
|
|
xfs_restore_resvblks(struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
__uint64_t resblks;
|
|
|
|
|
|
|
|
if (mp->m_resblks_save) {
|
|
|
|
resblks = mp->m_resblks_save;
|
|
|
|
mp->m_resblks_save = 0;
|
|
|
|
} else
|
|
|
|
resblks = xfs_default_resblks(mp);
|
|
|
|
|
|
|
|
xfs_reserve_blocks(mp, &resblks, NULL);
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
STATIC int
|
2006-03-14 04:06:18 +01:00
|
|
|
xfs_fs_remount(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct super_block *sb,
|
|
|
|
int *flags,
|
|
|
|
char *options)
|
|
|
|
{
|
2007-08-30 09:20:39 +02:00
|
|
|
struct xfs_mount *mp = XFS_M(sb);
|
2008-07-18 09:12:36 +02:00
|
|
|
substring_t args[MAX_OPT_ARGS];
|
|
|
|
char *p;
|
2009-01-19 02:04:07 +01:00
|
|
|
int error;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-07-18 09:12:36 +02:00
|
|
|
while ((p = strsep(&options, ",")) != NULL) {
|
|
|
|
int token;
|
2008-05-20 07:10:44 +02:00
|
|
|
|
2008-07-18 09:12:36 +02:00
|
|
|
if (!*p)
|
|
|
|
continue;
|
2008-05-20 03:30:39 +02:00
|
|
|
|
2008-07-18 09:12:36 +02:00
|
|
|
token = match_token(p, tokens, args);
|
|
|
|
switch (token) {
|
|
|
|
case Opt_barrier:
|
2008-05-20 03:30:39 +02:00
|
|
|
mp->m_flags |= XFS_MOUNT_BARRIER;
|
2008-07-18 09:12:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Test if barriers are actually working if we can,
|
|
|
|
* else delay this check until the filesystem is
|
|
|
|
* marked writeable.
|
|
|
|
*/
|
|
|
|
if (!(mp->m_flags & XFS_MOUNT_RDONLY))
|
|
|
|
xfs_mountfs_check_barriers(mp);
|
|
|
|
break;
|
|
|
|
case Opt_nobarrier:
|
2008-05-20 03:30:39 +02:00
|
|
|
mp->m_flags &= ~XFS_MOUNT_BARRIER;
|
2008-07-18 09:12:36 +02:00
|
|
|
break;
|
|
|
|
default:
|
2008-09-17 08:49:33 +02:00
|
|
|
/*
|
|
|
|
* Logically we would return an error here to prevent
|
|
|
|
* users from believing they might have changed
|
|
|
|
* mount options using remount which can't be changed.
|
|
|
|
*
|
|
|
|
* But unfortunately mount(8) adds all options from
|
|
|
|
* mtab and fstab to the mount arguments in some cases
|
|
|
|
* so we can't blindly reject options, but have to
|
|
|
|
* check for each specified option if it actually
|
|
|
|
* differs from the currently set option and only
|
|
|
|
* reject it if that's the case.
|
|
|
|
*
|
|
|
|
* Until that is implemented we return success for
|
|
|
|
* every remount request, and silently ignore all
|
|
|
|
* options that we can't actually change.
|
|
|
|
*/
|
|
|
|
#if 0
|
2008-07-18 09:12:36 +02:00
|
|
|
printk(KERN_INFO
|
|
|
|
"XFS: mount option \"%s\" not supported for remount\n", p);
|
|
|
|
return -EINVAL;
|
2008-09-17 08:49:33 +02:00
|
|
|
#else
|
2008-10-12 14:30:44 +02:00
|
|
|
break;
|
2008-09-17 08:49:33 +02:00
|
|
|
#endif
|
2008-05-20 03:30:39 +02:00
|
|
|
}
|
2008-07-18 09:12:36 +02:00
|
|
|
}
|
|
|
|
|
2009-01-19 02:04:07 +01:00
|
|
|
/* ro -> rw */
|
2008-07-18 09:12:36 +02:00
|
|
|
if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
|
|
|
|
mp->m_flags &= ~XFS_MOUNT_RDONLY;
|
|
|
|
if (mp->m_flags & XFS_MOUNT_BARRIER)
|
|
|
|
xfs_mountfs_check_barriers(mp);
|
2009-01-19 02:04:07 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If this is the first remount to writeable state we
|
|
|
|
* might have some superblock changes to update.
|
|
|
|
*/
|
|
|
|
if (mp->m_update_flags) {
|
|
|
|
error = xfs_mount_log_sb(mp, mp->m_update_flags);
|
|
|
|
if (error) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: failed to write sb changes");
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
mp->m_update_flags = 0;
|
|
|
|
}
|
xfs: don't hold onto reserved blocks on remount,ro
If we hold onto reserved blocks when doing a remount,ro we end
up writing the blocks used count to disk that includes the reserved
blocks. Reserved blocks are not actually used, so this results in
the values in the superblock being incorrect.
Hence if we run xfs_check or xfs_repair -n while the filesystem is
mounted remount,ro we end up with an inconsistent filesystem being
reported. Also, running xfs_copy on the remount,ro filesystem will
result in an inconsistent image being generated.
To fix this, unreserve the blocks when doing the remount,ro, and
reserved them again on remount,rw. This way a remount,ro filesystem
will appear consistent on disk to all utilities.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2010-01-26 05:08:49 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Fill out the reserve pool if it is empty. Use the stashed
|
|
|
|
* value if it is non-zero, otherwise go with the default.
|
|
|
|
*/
|
2010-02-05 23:59:53 +01:00
|
|
|
xfs_restore_resvblks(mp);
|
2008-07-18 09:12:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* rw -> ro */
|
|
|
|
if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
|
xfs: don't hold onto reserved blocks on remount,ro
If we hold onto reserved blocks when doing a remount,ro we end
up writing the blocks used count to disk that includes the reserved
blocks. Reserved blocks are not actually used, so this results in
the values in the superblock being incorrect.
Hence if we run xfs_check or xfs_repair -n while the filesystem is
mounted remount,ro we end up with an inconsistent filesystem being
reported. Also, running xfs_copy on the remount,ro filesystem will
result in an inconsistent image being generated.
To fix this, unreserve the blocks when doing the remount,ro, and
reserved them again on remount,rw. This way a remount,ro filesystem
will appear consistent on disk to all utilities.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2010-01-26 05:08:49 +01:00
|
|
|
/*
|
|
|
|
* After we have synced the data but before we sync the
|
|
|
|
* metadata, we need to free up the reserve block pool so that
|
|
|
|
* the used block count in the superblock on disk is correct at
|
|
|
|
* the end of the remount. Stash the current reserve pool size
|
|
|
|
* so that if we get remounted rw, we can return it to the same
|
|
|
|
* size.
|
|
|
|
*/
|
|
|
|
|
2008-10-30 07:15:50 +01:00
|
|
|
xfs_quiesce_data(mp);
|
2010-02-05 23:59:53 +01:00
|
|
|
xfs_save_resvblks(mp);
|
2008-10-30 07:16:21 +01:00
|
|
|
xfs_quiesce_attr(mp);
|
2008-05-20 03:30:39 +02:00
|
|
|
mp->m_flags |= XFS_MOUNT_RDONLY;
|
|
|
|
}
|
|
|
|
|
2008-07-18 09:12:36 +02:00
|
|
|
return 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2007-10-11 10:11:14 +02:00
|
|
|
/*
|
|
|
|
* Second stage of a freeze. The data is already frozen so we only
|
2008-10-30 07:16:21 +01:00
|
|
|
* need to take care of the metadata. Once that's done write a dummy
|
2007-10-11 10:11:14 +02:00
|
|
|
* record to dirty the log in case of a crash while frozen.
|
|
|
|
*/
|
2009-01-10 01:40:58 +01:00
|
|
|
STATIC int
|
|
|
|
xfs_fs_freeze(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct super_block *sb)
|
|
|
|
{
|
2007-10-11 10:11:14 +02:00
|
|
|
struct xfs_mount *mp = XFS_M(sb);
|
|
|
|
|
2010-02-05 23:59:53 +01:00
|
|
|
xfs_save_resvblks(mp);
|
2008-10-30 07:16:21 +01:00
|
|
|
xfs_quiesce_attr(mp);
|
2010-08-24 03:46:31 +02:00
|
|
|
return -xfs_fs_log_dummy(mp, SYNC_WAIT);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2010-02-05 23:59:53 +01:00
|
|
|
STATIC int
|
|
|
|
xfs_fs_unfreeze(
|
|
|
|
struct super_block *sb)
|
|
|
|
{
|
|
|
|
struct xfs_mount *mp = XFS_M(sb);
|
|
|
|
|
|
|
|
xfs_restore_resvblks(mp);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
STATIC int
|
2006-03-14 04:06:18 +01:00
|
|
|
xfs_fs_show_options(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct seq_file *m,
|
|
|
|
struct vfsmount *mnt)
|
|
|
|
{
|
2007-08-30 09:20:39 +02:00
|
|
|
return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2008-05-20 03:30:59 +02:00
|
|
|
/*
|
|
|
|
* This function fills in xfs_mount_t fields based on mount args.
|
|
|
|
* Note: the superblock _has_ now been read in.
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_finish_flags(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
|
|
|
|
|
2008-10-16 19:02:37 +02:00
|
|
|
/* Fail a mount where the logbuf is smaller than the log stripe */
|
2008-05-20 03:30:59 +02:00
|
|
|
if (xfs_sb_version_haslogv2(&mp->m_sb)) {
|
2008-10-30 07:53:24 +01:00
|
|
|
if (mp->m_logbsize <= 0 &&
|
|
|
|
mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
|
2008-05-20 03:30:59 +02:00
|
|
|
mp->m_logbsize = mp->m_sb.sb_logsunit;
|
2008-10-30 07:53:24 +01:00
|
|
|
} else if (mp->m_logbsize > 0 &&
|
|
|
|
mp->m_logbsize < mp->m_sb.sb_logsunit) {
|
2008-05-20 03:30:59 +02:00
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: logbuf size must be greater than or equal to log stripe size");
|
|
|
|
return XFS_ERROR(EINVAL);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
/* Fail a mount if the logbuf is larger than 32K */
|
2008-10-30 07:53:24 +01:00
|
|
|
if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
|
2008-05-20 03:30:59 +02:00
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: logbuf size for version 1 logs must be 16K or 32K");
|
|
|
|
return XFS_ERROR(EINVAL);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* mkfs'ed attr2 will turn on attr2 mount unless explicitly
|
|
|
|
* told by noattr2 to turn it off
|
|
|
|
*/
|
|
|
|
if (xfs_sb_version_hasattr2(&mp->m_sb) &&
|
2008-10-30 07:53:24 +01:00
|
|
|
!(mp->m_flags & XFS_MOUNT_NOATTR2))
|
2008-05-20 03:30:59 +02:00
|
|
|
mp->m_flags |= XFS_MOUNT_ATTR2;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* prohibit r/w mounts of read-only filesystems
|
|
|
|
*/
|
|
|
|
if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
|
|
|
|
cmn_err(CE_WARN,
|
|
|
|
"XFS: cannot mount a read-only filesystem as read-write");
|
|
|
|
return XFS_ERROR(EROFS);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
STATIC int
|
2006-03-14 04:06:18 +01:00
|
|
|
xfs_fs_fill_super(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct super_block *sb,
|
|
|
|
void *data,
|
|
|
|
int silent)
|
|
|
|
{
|
2008-03-27 08:00:54 +01:00
|
|
|
struct inode *root;
|
2007-08-30 09:20:39 +02:00
|
|
|
struct xfs_mount *mp = NULL;
|
2008-05-20 07:10:52 +02:00
|
|
|
int flags = 0, error = ENOMEM;
|
2008-05-20 07:10:44 +02:00
|
|
|
|
2008-05-20 07:10:52 +02:00
|
|
|
mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
|
|
|
|
if (!mp)
|
2008-10-30 07:53:24 +01:00
|
|
|
goto out;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-05-20 07:10:52 +02:00
|
|
|
spin_lock_init(&mp->m_sb_lock);
|
|
|
|
mutex_init(&mp->m_growlock);
|
|
|
|
atomic_set(&mp->m_active_trans, 0);
|
2007-08-30 09:21:22 +02:00
|
|
|
INIT_LIST_HEAD(&mp->m_sync_list);
|
|
|
|
spin_lock_init(&mp->m_sync_lock);
|
|
|
|
init_waitqueue_head(&mp->m_wait_single_sync_task);
|
|
|
|
|
2007-08-30 09:21:30 +02:00
|
|
|
mp->m_super = sb;
|
|
|
|
sb->s_fs_info = mp;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2010-06-23 10:11:15 +02:00
|
|
|
error = xfs_parseargs(mp, (char *)data);
|
2007-08-30 09:20:39 +02:00
|
|
|
if (error)
|
2008-10-30 07:53:24 +01:00
|
|
|
goto out_free_fsname;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
sb_min_blocksize(sb, BBSIZE);
|
2008-06-23 05:23:01 +02:00
|
|
|
sb->s_xattr = xfs_xattr_handlers;
|
2006-03-14 04:06:18 +01:00
|
|
|
sb->s_export_op = &xfs_export_operations;
|
2009-02-09 08:47:34 +01:00
|
|
|
#ifdef CONFIG_XFS_QUOTA
|
2006-03-14 04:06:18 +01:00
|
|
|
sb->s_qcop = &xfs_quotactl_operations;
|
2009-02-09 08:47:34 +01:00
|
|
|
#endif
|
2006-03-14 04:06:18 +01:00
|
|
|
sb->s_op = &xfs_super_operations;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-10-30 07:53:24 +01:00
|
|
|
if (silent)
|
2008-05-20 03:30:59 +02:00
|
|
|
flags |= XFS_MFSI_QUIET;
|
|
|
|
|
2008-10-30 07:53:24 +01:00
|
|
|
error = xfs_open_devices(mp);
|
2008-05-20 03:31:13 +02:00
|
|
|
if (error)
|
2010-06-23 10:11:15 +02:00
|
|
|
goto out_free_fsname;
|
2008-05-20 03:30:59 +02:00
|
|
|
|
2010-09-30 04:25:54 +02:00
|
|
|
error = xfs_icsb_init_counters(mp);
|
|
|
|
if (error)
|
|
|
|
goto out_close_devices;
|
2008-05-20 07:10:52 +02:00
|
|
|
|
2008-05-20 03:30:59 +02:00
|
|
|
error = xfs_readsb(mp, flags);
|
|
|
|
if (error)
|
2008-10-30 07:53:24 +01:00
|
|
|
goto out_destroy_counters;
|
|
|
|
|
|
|
|
error = xfs_finish_flags(mp);
|
2008-05-20 03:30:59 +02:00
|
|
|
if (error)
|
2008-05-20 07:11:05 +02:00
|
|
|
goto out_free_sb;
|
2008-05-20 03:30:59 +02:00
|
|
|
|
2008-05-20 07:10:36 +02:00
|
|
|
error = xfs_setup_devices(mp);
|
2008-05-20 03:31:13 +02:00
|
|
|
if (error)
|
2008-05-20 07:11:05 +02:00
|
|
|
goto out_free_sb;
|
2008-05-20 03:30:59 +02:00
|
|
|
|
|
|
|
if (mp->m_flags & XFS_MOUNT_BARRIER)
|
|
|
|
xfs_mountfs_check_barriers(mp);
|
|
|
|
|
|
|
|
error = xfs_filestream_mount(mp);
|
|
|
|
if (error)
|
2008-05-20 07:11:05 +02:00
|
|
|
goto out_free_sb;
|
2008-05-20 03:30:59 +02:00
|
|
|
|
2008-08-13 08:49:32 +02:00
|
|
|
error = xfs_mountfs(mp);
|
2008-05-20 03:30:59 +02:00
|
|
|
if (error)
|
2008-05-20 07:11:11 +02:00
|
|
|
goto out_filestream_unmount;
|
2008-05-20 03:30:59 +02:00
|
|
|
|
2007-10-11 10:09:40 +02:00
|
|
|
sb->s_magic = XFS_SB_MAGIC;
|
|
|
|
sb->s_blocksize = mp->m_sb.sb_blocksize;
|
|
|
|
sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
|
|
|
|
sb->s_time_gran = 1;
|
|
|
|
set_posix_acl_flag(sb);
|
|
|
|
|
2008-08-13 07:45:15 +02:00
|
|
|
root = igrab(VFS_I(mp->m_rootip));
|
2008-03-27 08:00:54 +01:00
|
|
|
if (!root) {
|
2008-02-05 02:14:01 +01:00
|
|
|
error = ENOENT;
|
2005-04-17 00:20:36 +02:00
|
|
|
goto fail_unmount;
|
2008-02-05 02:14:01 +01:00
|
|
|
}
|
2008-03-27 08:00:54 +01:00
|
|
|
if (is_bad_inode(root)) {
|
|
|
|
error = EINVAL;
|
2005-04-17 00:20:36 +02:00
|
|
|
goto fail_vnrele;
|
|
|
|
}
|
2008-03-27 08:00:54 +01:00
|
|
|
sb->s_root = d_alloc_root(root);
|
|
|
|
if (!sb->s_root) {
|
|
|
|
error = ENOMEM;
|
2005-04-17 00:20:36 +02:00
|
|
|
goto fail_vnrele;
|
|
|
|
}
|
2007-08-30 09:21:22 +02:00
|
|
|
|
2008-10-30 07:06:18 +01:00
|
|
|
error = xfs_syncd_init(mp);
|
|
|
|
if (error)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto fail_vnrele;
|
2007-08-30 09:21:22 +02:00
|
|
|
|
2010-04-29 01:55:50 +02:00
|
|
|
xfs_inode_shrinker_register(mp);
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
return 0;
|
|
|
|
|
2008-05-20 07:11:11 +02:00
|
|
|
out_filestream_unmount:
|
|
|
|
xfs_filestream_unmount(mp);
|
2008-05-20 07:11:05 +02:00
|
|
|
out_free_sb:
|
|
|
|
xfs_freesb(mp);
|
2008-10-30 07:53:24 +01:00
|
|
|
out_destroy_counters:
|
2008-05-20 07:10:52 +02:00
|
|
|
xfs_icsb_destroy_counters(mp);
|
2010-09-30 04:25:54 +02:00
|
|
|
out_close_devices:
|
2008-05-20 03:31:13 +02:00
|
|
|
xfs_close_devices(mp);
|
2008-10-30 07:53:24 +01:00
|
|
|
out_free_fsname:
|
|
|
|
xfs_free_fsname(mp);
|
2008-05-20 07:10:52 +02:00
|
|
|
kfree(mp);
|
2008-10-30 07:53:24 +01:00
|
|
|
out:
|
2008-05-20 07:10:52 +02:00
|
|
|
return -error;
|
2008-05-20 03:30:59 +02:00
|
|
|
|
|
|
|
fail_vnrele:
|
2005-04-17 00:20:36 +02:00
|
|
|
if (sb->s_root) {
|
|
|
|
dput(sb->s_root);
|
|
|
|
sb->s_root = NULL;
|
|
|
|
} else {
|
2008-03-27 08:00:54 +01:00
|
|
|
iput(root);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2008-05-20 03:30:59 +02:00
|
|
|
fail_unmount:
|
2008-05-20 03:30:52 +02:00
|
|
|
/*
|
|
|
|
* Blow away any referenced inode in the filestreams cache.
|
|
|
|
* This can and will cause log traffic as inodes go inactive
|
|
|
|
* here.
|
|
|
|
*/
|
|
|
|
xfs_filestream_unmount(mp);
|
|
|
|
|
|
|
|
XFS_bflush(mp->m_ddev_targp);
|
|
|
|
|
2008-05-20 03:31:13 +02:00
|
|
|
xfs_unmountfs(mp);
|
2008-08-13 08:50:21 +02:00
|
|
|
goto out_free_sb;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2010-07-24 22:46:55 +02:00
|
|
|
STATIC struct dentry *
|
|
|
|
xfs_fs_mount(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct file_system_type *fs_type,
|
|
|
|
int flags,
|
|
|
|
const char *dev_name,
|
2010-07-24 22:46:55 +02:00
|
|
|
void *data)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2010-07-24 22:46:55 +02:00
|
|
|
return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
|
2006-03-14 04:06:18 +01:00
|
|
|
}
|
|
|
|
|
2009-09-22 02:01:09 +02:00
|
|
|
static const struct super_operations xfs_super_operations = {
|
2006-03-14 04:06:18 +01:00
|
|
|
.alloc_inode = xfs_fs_alloc_inode,
|
|
|
|
.destroy_inode = xfs_fs_destroy_inode,
|
2009-10-06 22:29:26 +02:00
|
|
|
.dirty_inode = xfs_fs_dirty_inode,
|
2006-03-14 04:06:18 +01:00
|
|
|
.write_inode = xfs_fs_write_inode,
|
2010-06-07 20:34:48 +02:00
|
|
|
.evict_inode = xfs_fs_evict_inode,
|
2006-03-14 04:06:18 +01:00
|
|
|
.put_super = xfs_fs_put_super,
|
2009-10-06 22:29:28 +02:00
|
|
|
.sync_fs = xfs_fs_sync_fs,
|
2009-01-10 01:40:58 +01:00
|
|
|
.freeze_fs = xfs_fs_freeze,
|
2010-02-05 23:59:53 +01:00
|
|
|
.unfreeze_fs = xfs_fs_unfreeze,
|
2006-03-14 04:06:18 +01:00
|
|
|
.statfs = xfs_fs_statfs,
|
|
|
|
.remount_fs = xfs_fs_remount,
|
|
|
|
.show_options = xfs_fs_show_options,
|
2005-04-17 00:20:36 +02:00
|
|
|
};
|
|
|
|
|
2007-02-20 22:57:47 +01:00
|
|
|
static struct file_system_type xfs_fs_type = {
|
2005-04-17 00:20:36 +02:00
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.name = "xfs",
|
2010-07-24 22:46:55 +02:00
|
|
|
.mount = xfs_fs_mount,
|
2005-04-17 00:20:36 +02:00
|
|
|
.kill_sb = kill_block_super,
|
|
|
|
.fs_flags = FS_REQUIRES_DEV,
|
|
|
|
};
|
|
|
|
|
2008-07-18 09:11:46 +02:00
|
|
|
STATIC int __init
|
|
|
|
xfs_init_zones(void)
|
|
|
|
{
|
|
|
|
|
|
|
|
xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
|
|
|
|
if (!xfs_ioend_zone)
|
2008-10-30 07:36:14 +01:00
|
|
|
goto out;
|
2008-07-18 09:11:46 +02:00
|
|
|
|
|
|
|
xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
|
|
|
|
xfs_ioend_zone);
|
|
|
|
if (!xfs_ioend_pool)
|
|
|
|
goto out_destroy_ioend_zone;
|
|
|
|
|
|
|
|
xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
|
|
|
|
"xfs_log_ticket");
|
|
|
|
if (!xfs_log_ticket_zone)
|
|
|
|
goto out_destroy_ioend_pool;
|
|
|
|
|
|
|
|
xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
|
|
|
|
"xfs_bmap_free_item");
|
|
|
|
if (!xfs_bmap_free_item_zone)
|
|
|
|
goto out_destroy_log_ticket_zone;
|
2008-10-30 07:36:14 +01:00
|
|
|
|
2008-07-18 09:11:46 +02:00
|
|
|
xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
|
|
|
|
"xfs_btree_cur");
|
|
|
|
if (!xfs_btree_cur_zone)
|
|
|
|
goto out_destroy_bmap_free_item_zone;
|
|
|
|
|
|
|
|
xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
|
|
|
|
"xfs_da_state");
|
|
|
|
if (!xfs_da_state_zone)
|
|
|
|
goto out_destroy_btree_cur_zone;
|
|
|
|
|
|
|
|
xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
|
|
|
|
if (!xfs_dabuf_zone)
|
|
|
|
goto out_destroy_da_state_zone;
|
|
|
|
|
|
|
|
xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
|
|
|
|
if (!xfs_ifork_zone)
|
|
|
|
goto out_destroy_dabuf_zone;
|
|
|
|
|
|
|
|
xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
|
|
|
|
if (!xfs_trans_zone)
|
|
|
|
goto out_destroy_ifork_zone;
|
|
|
|
|
2010-06-23 10:11:15 +02:00
|
|
|
xfs_log_item_desc_zone =
|
|
|
|
kmem_zone_init(sizeof(struct xfs_log_item_desc),
|
|
|
|
"xfs_log_item_desc");
|
|
|
|
if (!xfs_log_item_desc_zone)
|
|
|
|
goto out_destroy_trans_zone;
|
|
|
|
|
2008-07-18 09:11:46 +02:00
|
|
|
/*
|
|
|
|
* The size of the zone allocated buf log item is the maximum
|
|
|
|
* size possible under XFS. This wastes a little bit of memory,
|
|
|
|
* but it is much faster.
|
|
|
|
*/
|
|
|
|
xfs_buf_item_zone = kmem_zone_init((sizeof(xfs_buf_log_item_t) +
|
2010-05-07 03:05:19 +02:00
|
|
|
(((XFS_MAX_BLOCKSIZE / XFS_BLF_CHUNK) /
|
2008-07-18 09:11:46 +02:00
|
|
|
NBWORD) * sizeof(int))), "xfs_buf_item");
|
|
|
|
if (!xfs_buf_item_zone)
|
2010-06-23 10:11:15 +02:00
|
|
|
goto out_destroy_log_item_desc_zone;
|
2008-07-18 09:11:46 +02:00
|
|
|
|
|
|
|
xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
|
|
|
|
((XFS_EFD_MAX_FAST_EXTENTS - 1) *
|
|
|
|
sizeof(xfs_extent_t))), "xfs_efd_item");
|
|
|
|
if (!xfs_efd_zone)
|
|
|
|
goto out_destroy_buf_item_zone;
|
|
|
|
|
|
|
|
xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
|
|
|
|
((XFS_EFI_MAX_FAST_EXTENTS - 1) *
|
|
|
|
sizeof(xfs_extent_t))), "xfs_efi_item");
|
|
|
|
if (!xfs_efi_zone)
|
|
|
|
goto out_destroy_efd_zone;
|
|
|
|
|
|
|
|
xfs_inode_zone =
|
|
|
|
kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
|
2008-10-30 07:36:14 +01:00
|
|
|
KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
|
|
|
|
xfs_fs_inode_init_once);
|
2008-07-18 09:11:46 +02:00
|
|
|
if (!xfs_inode_zone)
|
|
|
|
goto out_destroy_efi_zone;
|
|
|
|
|
|
|
|
xfs_ili_zone =
|
|
|
|
kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
|
|
|
|
KM_ZONE_SPREAD, NULL);
|
|
|
|
if (!xfs_ili_zone)
|
|
|
|
goto out_destroy_inode_zone;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_destroy_inode_zone:
|
|
|
|
kmem_zone_destroy(xfs_inode_zone);
|
|
|
|
out_destroy_efi_zone:
|
|
|
|
kmem_zone_destroy(xfs_efi_zone);
|
|
|
|
out_destroy_efd_zone:
|
|
|
|
kmem_zone_destroy(xfs_efd_zone);
|
|
|
|
out_destroy_buf_item_zone:
|
|
|
|
kmem_zone_destroy(xfs_buf_item_zone);
|
2010-06-23 10:11:15 +02:00
|
|
|
out_destroy_log_item_desc_zone:
|
|
|
|
kmem_zone_destroy(xfs_log_item_desc_zone);
|
2008-07-18 09:11:46 +02:00
|
|
|
out_destroy_trans_zone:
|
|
|
|
kmem_zone_destroy(xfs_trans_zone);
|
|
|
|
out_destroy_ifork_zone:
|
|
|
|
kmem_zone_destroy(xfs_ifork_zone);
|
|
|
|
out_destroy_dabuf_zone:
|
|
|
|
kmem_zone_destroy(xfs_dabuf_zone);
|
|
|
|
out_destroy_da_state_zone:
|
|
|
|
kmem_zone_destroy(xfs_da_state_zone);
|
|
|
|
out_destroy_btree_cur_zone:
|
|
|
|
kmem_zone_destroy(xfs_btree_cur_zone);
|
|
|
|
out_destroy_bmap_free_item_zone:
|
|
|
|
kmem_zone_destroy(xfs_bmap_free_item_zone);
|
|
|
|
out_destroy_log_ticket_zone:
|
|
|
|
kmem_zone_destroy(xfs_log_ticket_zone);
|
|
|
|
out_destroy_ioend_pool:
|
|
|
|
mempool_destroy(xfs_ioend_pool);
|
|
|
|
out_destroy_ioend_zone:
|
|
|
|
kmem_zone_destroy(xfs_ioend_zone);
|
|
|
|
out:
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
STATIC void
|
|
|
|
xfs_destroy_zones(void)
|
|
|
|
{
|
|
|
|
kmem_zone_destroy(xfs_ili_zone);
|
|
|
|
kmem_zone_destroy(xfs_inode_zone);
|
|
|
|
kmem_zone_destroy(xfs_efi_zone);
|
|
|
|
kmem_zone_destroy(xfs_efd_zone);
|
|
|
|
kmem_zone_destroy(xfs_buf_item_zone);
|
2010-06-23 10:11:15 +02:00
|
|
|
kmem_zone_destroy(xfs_log_item_desc_zone);
|
2008-07-18 09:11:46 +02:00
|
|
|
kmem_zone_destroy(xfs_trans_zone);
|
|
|
|
kmem_zone_destroy(xfs_ifork_zone);
|
|
|
|
kmem_zone_destroy(xfs_dabuf_zone);
|
|
|
|
kmem_zone_destroy(xfs_da_state_zone);
|
|
|
|
kmem_zone_destroy(xfs_btree_cur_zone);
|
|
|
|
kmem_zone_destroy(xfs_bmap_free_item_zone);
|
|
|
|
kmem_zone_destroy(xfs_log_ticket_zone);
|
|
|
|
mempool_destroy(xfs_ioend_pool);
|
|
|
|
kmem_zone_destroy(xfs_ioend_zone);
|
|
|
|
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
STATIC int __init
|
2008-07-18 09:11:46 +02:00
|
|
|
init_xfs_fs(void)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
2008-11-28 04:23:33 +01:00
|
|
|
printk(KERN_INFO XFS_VERSION_STRING " with "
|
|
|
|
XFS_BUILD_OPTIONS " enabled\n");
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-12-03 12:20:39 +01:00
|
|
|
xfs_ioend_init();
|
2008-07-18 09:11:46 +02:00
|
|
|
xfs_dir_startup();
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-03-14 03:18:19 +01:00
|
|
|
error = xfs_init_zones();
|
2008-07-18 09:11:46 +02:00
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
error = xfs_mru_cache_init();
|
|
|
|
if (error)
|
2009-12-15 00:14:59 +01:00
|
|
|
goto out_destroy_zones;
|
2008-07-18 09:11:46 +02:00
|
|
|
|
|
|
|
error = xfs_filestream_init();
|
|
|
|
if (error)
|
|
|
|
goto out_mru_cache_uninit;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-01-11 05:39:08 +01:00
|
|
|
error = xfs_buf_init();
|
2008-07-18 09:11:46 +02:00
|
|
|
if (error)
|
|
|
|
goto out_filestream_uninit;
|
|
|
|
|
|
|
|
error = xfs_init_procfs();
|
|
|
|
if (error)
|
|
|
|
goto out_buf_terminate;
|
|
|
|
|
|
|
|
error = xfs_sysctl_register();
|
|
|
|
if (error)
|
|
|
|
goto out_cleanup_procfs;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
vfs_initquota();
|
|
|
|
|
|
|
|
error = register_filesystem(&xfs_fs_type);
|
|
|
|
if (error)
|
2008-07-18 09:11:46 +02:00
|
|
|
goto out_sysctl_unregister;
|
2005-04-17 00:20:36 +02:00
|
|
|
return 0;
|
|
|
|
|
2008-07-18 09:11:46 +02:00
|
|
|
out_sysctl_unregister:
|
|
|
|
xfs_sysctl_unregister();
|
|
|
|
out_cleanup_procfs:
|
|
|
|
xfs_cleanup_procfs();
|
|
|
|
out_buf_terminate:
|
2006-01-11 05:39:08 +01:00
|
|
|
xfs_buf_terminate();
|
2008-07-18 09:11:46 +02:00
|
|
|
out_filestream_uninit:
|
|
|
|
xfs_filestream_uninit();
|
|
|
|
out_mru_cache_uninit:
|
|
|
|
xfs_mru_cache_uninit();
|
|
|
|
out_destroy_zones:
|
2006-03-14 03:18:19 +01:00
|
|
|
xfs_destroy_zones();
|
2008-07-18 09:11:46 +02:00
|
|
|
out:
|
2005-04-17 00:20:36 +02:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
STATIC void __exit
|
2008-07-18 09:11:46 +02:00
|
|
|
exit_xfs_fs(void)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
vfs_exitquota();
|
|
|
|
unregister_filesystem(&xfs_fs_type);
|
2008-07-18 09:11:46 +02:00
|
|
|
xfs_sysctl_unregister();
|
|
|
|
xfs_cleanup_procfs();
|
2006-01-11 05:39:08 +01:00
|
|
|
xfs_buf_terminate();
|
2008-07-18 09:11:46 +02:00
|
|
|
xfs_filestream_uninit();
|
|
|
|
xfs_mru_cache_uninit();
|
2006-03-14 03:18:19 +01:00
|
|
|
xfs_destroy_zones();
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
module_init(init_xfs_fs);
|
|
|
|
module_exit(exit_xfs_fs);
|
|
|
|
|
|
|
|
MODULE_AUTHOR("Silicon Graphics, Inc.");
|
|
|
|
MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
|
|
|
|
MODULE_LICENSE("GPL");
|