- 15 Mar, 2012 1 commit
-
-
Dave Chinner authored
When we get concurrent lookups of the same inode that is not in the per-AG inode cache, there is a race condition that triggers warnings in unlock_new_inode() indicating that we are initialising an inode that isn't in a the correct state for a new inode. When we do an inode lookup via a file handle or a bulkstat, we don't serialise lookups at a higher level through the dentry cache (i.e. pathless lookup), and so we can get concurrent lookups of the same inode. The race condition is between the insertion of the inode into the cache in the case of a cache miss and a concurrently lookup: Thread 1 Thread 2 xfs_iget() xfs_iget_cache_miss() xfs_iread() lock radix tree radix_tree_insert() rcu_read_lock radix_tree_lookup lock inode flags XFS_INEW not set igrab() unlock inode flags rcu_read_unlock use uninitialised inode ..... lock inode flags set XFS_INEW unlock inode flags unlock radix tree xfs_setup_inode() inode flags = I_NEW unlock_new_inode() WARNING as inode flags != I_NEW This can lead to inode corruption, inode list corruption, etc, and is generally a bad thing to occur. Fix this by setting XFS_INEW before inserting the inode into the radix tree. This will ensure any concurrent lookup will find the new inode with XFS_INEW set and that forces the lookup to wait until the XFS_INEW flag is removed before allowing the lookup to succeed. cc: <stable@vger.kernel.org> # for 3.0.x, 3.2.x Signed-off-by:Dave Chinner <dchinner@redhat.com> Reviewed-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
Ben Myers <bpm@sgi.com>
-
- 14 Mar, 2012 6 commits
-
-
Dave Chinner authored
Signed-off-by:
-
Christoph Hellwig authored
If we initialize the slab caches for the quota code when XFS is loaded there is no need for a global and reference counted quota manager structure. Drop all this overhead and also fix the error handling during quota initialization. Reviewed-by:
-
Christoph Hellwig authored
Instead of keeping a separate per-filesystem list of dquots we can walk the radix tree for the two places where we need to iterate all quota structures. Reviewed-by:
-
Christoph Hellwig authored
Replace the global hash tables for looking up in-memory dquot structures with per-filesystem radix trees to allow scaling to a large number of in-memory dquot structures. Reviewed-by:
-
Christoph Hellwig authored
Replace the global dquot lru lists with a per-filesystem one. Note that the shrinker isn't wire up to the per-superblock VFS shrinker infrastructure as would have problems summing up and splitting the counts for inodes and dquots. I don't think this is a major problem as the quota cache isn't as interwinded with the inode cache as the dentry cache is, because an inode that is dropped from the cache will generally release a dquot reference, but most of the time it won't be the last one. Signed-off-by:
-
Christoph Hellwig authored
Switch the quota code over to use the generic XFS statistics infrastructure. While the legacy /proc/fs/xfs/xqm and /proc/fs/xfs/xqmstats interfaces are preserved for now the statistics that still have a meaning with the current code are now also available from /proc/fs/xfs/stats. Signed-off-by:
-
- 13 Mar, 2012 5 commits
-
-
Christoph Hellwig authored
Add an in-memory only flag to say we logged timestamps only, and use it to check if fdatasync can optimize away the log force. Reviewed-by:
Mark Tinguely <tinguely@sgi.com> Signed-off-by:
-
Christoph Hellwig authored
Add a new ili_fields member to the inode log item to isolate the in-memory flags from the ones that actually go to the log. This will allow tracking timestamp-only updates for fdatasync and O_DSYNC in the next patch and prepares for divorcing the on-disk log format from the in-memory log item a little further down the road. Reviewed-by:
-
Christoph Hellwig authored
Move all code messing with the inode log item flags into xfs_inode_item_format to make sure xfs_inode_item_size really only calculates the the number of vectors, but doesn't modify any state of the inode item. Reviewed-by:
-
Christoph Hellwig authored
Timestamps on regular files are the last metadata that XFS does not update transactionally. Now that we use the delaylog mode exclusively and made the log scode scale extremly well there is no need to bypass that code for timestamp updates. Logging all updates allows to drop a lot of code, and will allow for further performance improvements later on. Note that this patch drops optimized handling of fdatasync - it will be added back in a separate commit. Reviewed-by:
-
Christoph Hellwig authored
Do not use unlogged metadata updates and the VFS dirty bit for updating the file size after writeback. In addition to causing various problems with updates getting delayed for far too long this also drags in the unscalable VFS dirty tracking, and is one of the few remaining unlogged metadata updates. Reviewed-by:
-
- 05 Mar, 2012 3 commits
-
-
Christoph Hellwig authored
If we convert and unwritten extent past the current i_size log the size update as part of the extent manipulation transactions instead of doing an unlogged metadata update later. Reviewed-by:
-
Christoph Hellwig authored
Replace xfs_ioend_new_eof with a new inline xfs_new_eof helper that doesn't require and ioend, and is available also outside of xfs_aops.c. Also make the code a bit more clear by using a normal if statement instead of a slightly misleading MIN(). Reviewed-by:
-
Christoph Hellwig authored
The new concurrency managed workqueues are cheap enough that we can create per-filesystem instead of global workqueues. This allows us to remove the trylock or defer scheme on the ilock, which is not helpful once we have outstanding log reservations until finishing a size update. Also allow the default concurrency on this workqueues so that I/O completions blocking on the ilock for one inode do not block process for another inode. Reviewed-by:
-
- 29 Feb, 2012 3 commits
-
-
Christoph Hellwig authored
Now that XFS takes quota reservations into account there is no need to flush anything before reporting quotas - in addition to beeing fully transactional all quota information is also 100% coherent with the rest of the filesystem now. Reviewed-by:
-
Christoph Hellwig authored
Report all quota usage including the currently pending reservations. This avoids the need to flush delalloc space before gathering quota information, and matches quota enforcement, which already takes the reservations into account. This fixes xfstests 270. Reviewed-by:
-
Christoph Hellwig authored
The is no good reason to have these two separate, and for the next change we would need the full struct xfs_dquot in xfs_qm_export_dquot, so better just fold the code now instead of changing it spuriously. Reviewed-by:
-
- 25 Feb, 2012 1 commit
-
-
Alex Elder authored
At the end of xfs_reclaim_inode(), the inode is locked in order to we wait for a possible concurrent lookup to complete before the inode is freed. This synchronization step was taking both the ILOCK and the IOLOCK, but the latter was causing lockdep to produce reports of the possibility of deadlock. It turns out that there's no need to acquire the IOLOCK at this point anyway. It may have been required in some earlier version of the code, but there should be no need to take the IOLOCK in xfs_iget(), so there's no (longer) any need to get it here for synchronization. Add an assertion in xfs_iget() as a reminder of this assumption. Dave Chinner diagnosed this on IRC, and Christoph Hellwig suggested no longer including the IOLOCK. I just put together the patch. Signed-off-by:
Alex Elder <elder@dreamhost.com> Reviewed-by:
-
- 23 Feb, 2012 13 commits
-
-
Christoph Hellwig authored
Split the log regrant case out of xfs_log_reserve into a separate function, and merge xlog_grant_log_space and xlog_regrant_write_log_space into their respective callers. Also replace the XFS_LOG_PERM_RESERV flag, which easily got misused before the previous cleanups with a simple boolean parameter. Signed-off-by:
-
Christoph Hellwig authored
Reviewed-by:
-
Christoph Hellwig authored
Reviewed-by:
-
Christoph Hellwig authored
Reviewed-by:
-
Christoph Hellwig authored
Reviewed-by:
-
Christoph Hellwig authored
Reviewed-by:
-
Christoph Hellwig authored
Add a new data structure to allow sharing code between the log grant and regrant code. Reviewed-by:
-
Christoph Hellwig authored
The tic->t_wait waitqueues can never have more than a single waiter on them, so we can easily replace them with a task_struct pointer and wake_up_process. Reviewed-by:
-
Christoph Hellwig authored
Remove the now unused opportunistic parameter, and use the the xlog_writeq_wake and xlog_reserveq_wake helpers now that we don't have to care about the opportunistic wakeups. Reviewed-by:
-
Christoph Hellwig authored
There is no reason to wake up log space waiters when unlocking inodes or dquots, and the commit log has no explanation for this function either. Given that we now have exact log space wakeups everywhere we can assume the reason for this function was to paper over log space races in earlier XFS versions. Reviewed-by:
-
Christoph Hellwig authored
The only reason that xfs_log_space_wake had to do opportunistic wakeups was that the old xfs_log_move_tail calling convention didn't allow for exact wakeups when not updating the log tail LSN. Since this issue has been fixed we can do exact wakeups now. Reviewed-by:
-
Christoph Hellwig authored
Currently xfs_log_move_tail has a tail_lsn argument that is horribly overloaded: it may contain either an actual lsn to assign to the log tail, 0 as a special case to use the last sync LSN, or 1 to indicate that no tail LSN assignment should be performed, and we should opportunisticly wake up at one task waiting for log space even if we did not move the LSN. Remove the tail lsn assigned from xfs_log_move_tail and make the two callers use xlog_assign_tail_lsn instead of the current variant of partially using the code in xfs_log_move_tail and partially opencoding it. Note that means we grow an addition lock roundtrip on the AIL lock for each bulk update or delete, which is still far less than what we had before introducing the bulk operations. If this proves to be a problem we can still add a variant of xlog_assign_tail_lsn that expects the lock to be held already. Also rename the remainder of xfs_log_move_tail to xfs_log_space_wake as that name describes its functionality much better. Reviewed-by:
-
Mitsuo Hayasaka authored
This patch is a cleanup of quota check on disk blocks and inodes reservations, and changes it as follows. (1) add a total_count variable to store the total number of current usages and new reservations for disk blocks and inodes, respectively. (2) make it more readable to check if the local variables softlimit and hardlimit are positive. It has been changed as follows. if (softlimit > 0ULL) -> if (softlimit) if (hardlimit > 0ULL) -> if (hardlimit) This is because they are defined as xfs_qcnt_t which is unsigned. Signed-off-by:Mitsuo Hayasaka <mitsuo.hayasaka.hu@hitachi.com> Cc: Ben Myers <bpm@sgi.com> Cc: Alex Elder <elder@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Reviewed-by:
-
- 21 Feb, 2012 2 commits
-
-
Mitsuo Hayasaka authored
The xfs checks quota when reserving disk blocks and inodes. In the block reservation, it checks if the total number of blocks including current usage and new reservation exceed quota. In the inode reservation, it checks using the total number of inodes including only current usage without new reservation. However, this inode quota check works well since the caller of xfs_trans_dquot() always sets the argument of the number of new inode reservation to 1 or 0 and inode is reserved one by one in current xfs. To make it more general, this patch changes it to the same way as the block quota check. Signed-off-by:
-
Mitsuo Hayasaka authored
In general, quota allows us to use disk blocks and inodes up to each limit, that is, they are available if they don't exceed their limitations. Current xfs sets their available ranges to lower than them except disk inode quota check. So, this patch changes the ranges to not beyond them. Signed-off-by:
-
- 13 Feb, 2012 1 commit
-
-
Jesper Juhl authored
It looks to me like the two ASSERT()s in xfs_trans_add_item() really want to do a compare (==) rather than assignment (=). This patch changes it from the latter to the former. Signed-off-by:
Jesper Juhl <jj@chaosbits.net> Signed-off-by:
-
- 10 Feb, 2012 1 commit
-
-
Christoph Hellwig authored
Stop reusing dquots from the freelist when allocating new ones directly, and implement a shrinker that actually follows the specifications for the interface. The shrinker implementation is still highly suboptimal at this point, but we can gradually work on it. This also fixes an bug in the previous lock ordering, where we would take the hash and dqlist locks inside of the freelist lock against the normal lock ordering. This is only solvable by introducing the dispose list, and thus not when using direct reclaim of unused dquots for new allocations. As a side-effect the quota upper bound and used to free ratio values in /proc/fs/xfs/xqm are set to 0 as these values don't make any sense in the new world order. Signed-off-by:
-
- 03 Feb, 2012 4 commits
-
-
Chandra Seetharaman authored
Define new macro XFS_ALL_QUOTA_ACTIVE and simply some usage of quota macros. Signed-off-by:
Chandra Seetharaman <sekharan@us.ibm.com> Reviewed-by:
-
Chandra Seetharaman authored
Change xfs_sb_from_disk() interface to take a mount pointer instead of a superblock pointer. This is to print mount point specific error messages in future fixes. Signed-off-by:
-
Chandra Seetharaman authored
Define a new function xfs_inode_dquot() that takes a inode pointer and a disk quota type and returns the quota pointer for the specified quota type. This simplifies the xfs_qm_dqget() error path significantly. Signed-off-by:
-
Chandra Seetharaman authored
Create a new function xfs_this_quota_on() that takes a xfs_mount data structure and a disk quota type and returns true if the specified type of quota is ON in the xfs_mount data structure. Signed-off-by:
-
