- 29 Jul, 2020 23 commits
-
-
Darrick J. Wong authored
Hoist the code that adjusts the incore quota reservation count adjustments into a separate function, both to reduce the level of indentation and also to reduce the amount of open-coded logic. Signed-off-by:
Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by:
Allison Collins <allison.henderson@oracle.com> Reviewed-by:
Chandan Babu R <chandanrlinux@gmail.com> Reviewed-by:
Christoph Hellwig <hch@lst.de>
-
Darrick J. Wong authored
Now that we've refactored the resource usage and limits into per-resource structures, we can refactor some of the open-coded reservation limit checking in xfs_trans_dqresv. Signed-off-by:
-
Darrick J. Wong authored
Now that we can pass around quota resource and limit structures, clean up the open-coded field setting in xfs_qm_scall_setqlim. Signed-off-by:
-
Darrick J. Wong authored
Refactor the open-coded test for whether or not we're over quota. Signed-off-by:
-
Darrick J. Wong authored
struct xfs_dquot already has a pointer to the xfs mount, so remove the redundant parameter from xfs_qm_adjust_dq*. Signed-off-by:
-
Darrick J. Wong authored
Now that we've split up the dquot resource fields into separate structs, do the same for the default limits to enable further refactoring. Signed-off-by:
-
Darrick J. Wong authored
Now that we've stopped using qcore entirely, drop it from the incore dquot. Signed-off-by:
-
Darrick J. Wong authored
Add timers fields to the incore dquot, and use that instead of the ones in qcore. This eliminates a bunch of endian conversions and will eventually allow us to remove qcore entirely. Signed-off-by:
-
Darrick J. Wong authored
Add warning counter fields to the incore dquot, and use that instead of the ones in qcore. This eliminates a bunch of endian conversions and will eventually allow us to remove qcore entirely. Signed-off-by:
-
Darrick J. Wong authored
Add counter fields to the incore dquot, and use that instead of the ones in qcore. This eliminates a bunch of endian conversions and will eventually allow us to remove qcore entirely. Signed-off-by:
-
Darrick J. Wong authored
Add limits fields in the incore dquot, and use that instead of the ones in qcore. This eliminates a bunch of endian conversions and will eventually allow us to remove qcore entirely. Signed-off-by:
-
Darrick J. Wong authored
Introduce a new struct xfs_dquot_res that we'll use to track all the incore data for a particular resource type (block, inode, rt block). This will help us (once we've eliminated q_core) to declutter quota functions that currently open-code field access or pass around fields around explicitly. Signed-off-by:
-
Darrick J. Wong authored
Add a dquot id field to the incore dquot, and use that instead of the one in qcore. This eliminates a bunch of endian conversions and will eventually allow us to remove qcore entirely. We also rearrange the start of xfs_dquot to remove padding holes, saving 8 bytes. Signed-off-by:
-
Darrick J. Wong authored
Use the incore dq_flags to figure out the dquot type. This is the first step towards removing xfs_disk_dquot from the incore dquot. Signed-off-by:
-
Darrick J. Wong authored
Move the dquot cluster size #define to xfs_format.h. It is an important part of the ondisk format because the ondisk dquot record size is not an even power of two, which means that the buffer size we use is significant here because the kernel leaves slack space at the end of the buffer to avoid having to deal with a dquot record crossing a block boundary. This is also an excuse to fix one of the longstanding discrepancies between kernel and userspace libxfs headers. Signed-off-by:
-
Darrick J. Wong authored
Rename the existing incore dquot "dq_flags" field to "q_flags" to match everything else in the structure, then move the two actual dquot state flags to the XFS_DQFLAG_ namespace from XFS_DQ_. Signed-off-by:
-
Darrick J. Wong authored
We only use the XFS_QMOPT flags in quotacheck to signal the quota type, so rip out all the flags handling and just pass the type all the way through. Signed-off-by:
-
Darrick J. Wong authored
Since xfs_qm_scall_trunc_qfiles can take a bitset of quota types that we want to truncate, change the flags argument to take XFS_QMOPT_[UGP}QUOTA so that the next patch can start to deprecate XFS_DQ_*. Signed-off-by:
-
Darrick J. Wong authored
While loading dquot records off disk, make sure that the quota type flags are the same between the incore dquot and the ondisk dquot. Signed-off-by:
-
Darrick J. Wong authored
xfs_trans_dqresv is the function that we use to make reservations against resource quotas. Each resource contains two counters: the q_core counter, which tracks resources allocated on disk; and the dquot reservation counter, which tracks how much of that resource has either been allocated or reserved by threads that are working on metadata updates. For disk blocks, we compare the proposed reservation counter against the hard and soft limits to decide if we're going to fail the operation. However, for inodes we inexplicably compare against the q_core counter, not the incore reservation count. Since the q_core counter is always lower than the reservation count and we unlock the dquot between reservation and transaction commit, this means that multiple threads can reserve the last inode count before we hit the hard limit, and when they commit, we'll be well over the hard limit. Fix this by checking against the incore inode reservation counter, since we would appear to maintain that correctly (and that's what we report in GETQUOTA). Signed-off-by:
-
Darrick J. Wong authored
In commit 8d3d7e2b, we changed xfs_qm_dqpurge to bail out if we can't lock the dquot buf to flush the dquot. This prevents the AIL from blocking on the dquot, but it also forgets to clear the FREEING flag on its way out. A subsequent purge attempt will see the FREEING flag is set and bail out, which leads to dqpurge_all failing to purge all the dquots. (copy-pasting from Dave Chinner's identical patch) This was found by inspection after having xfs/305 hang 1 in ~50 iterations in a quotaoff operation: [ 8872.301115] xfs_quota D13888 92262 91813 0x00004002 [ 8872.302538] Call Trace: [ 8872.303193] __schedule+0x2d2/0x780 [ 8872.304108] ? do_raw_spin_unlock+0x57/0xd0 [ 8872.305198] schedule+0x6e/0xe0 [ 8872.306021] schedule_timeout+0x14d/0x300 [ 8872.307060] ? __next_timer_interrupt+0xe0/0xe0 [ 8872.308231] ? xfs_qm_dqusage_adjust+0x200/0x200 [ 8872.309422] schedule_timeout_uninterruptible+0x2a/0x30 [ 8872.310759] xfs_qm_dquot_walk.isra.0+0x15a/0x1b0 [ 8872.311971] xfs_qm_dqpurge_all+0x7f/0x90 [ 8872.313022] xfs_qm_scall_quotaoff+0x18d/0x2b0 [ 8872.314163] xfs_quota_disable+0x3a/0x60 [ 8872.315179] kernel_quotactl+0x7e2/0x8d0 [ 8872.316196] ? __do_sys_newstat+0x51/0x80 [ 8872.317238] __x64_sys_quotactl+0x1e/0x30 [ 8872.318266] do_syscall_64+0x46/0x90 [ 8872.319193] entry_SYSCALL_64_after_hwframe+0x44/0xa9 [ 8872.320490] RIP: 0033:0x7f46b5490f2a [ 8872.321414] Code: Bad RIP value. Returning -EAGAIN from xfs_qm_dqpurge() without clearing the XFS_DQ_FREEING flag means the xfs_qm_dqpurge_all() code can never free the dquot, and we loop forever waiting for the XFS_DQ_FREEING flag to go away on the dquot that leaked it via -EAGAIN. Fixes: 8d3d7e2b ("xfs: trylock underlying buffer on dquot flush") Signed-off-by:
Brian Foster <bfoster@redhat.com> Signed-off-by:
Dave Chinner <dchinner@redhat.com> Reviewed-by:
-
Brian Foster authored
The block reservation calculation for inode allocation is supposed to consist of the blocks required for the inode chunk plus (maxlevels-1) of the inode btree multiplied by the number of inode btrees in the fs (2 when finobt is enabled, 1 otherwise). Instead, the macro returns (ialloc_blocks + 2) due to a precedence error in the calculation logic. This leads to block reservation overruns via generic/531 on small block filesystems with finobt enabled. Add braces to fix the calculation and reserve the appropriate number of blocks. Fixes: 9d43b180 ("xfs: update inode allocation/free transaction reservations for finobt") Signed-off-by:
-
Brian Foster authored
xfsaild is racy with respect to transaction abort and shutdown in that the task can idle or exit with an empty AIL but buffers still on the delwri queue. This was partly addressed by cancelling the delwri queue before the task exits to prevent memory leaks, but it's also possible for xfsaild to empty and idle with buffers on the delwri queue. For example, a transaction that pins a buffer that also happens to sit on the AIL delwri queue will explicitly remove the associated log item from the AIL if the transaction aborts. The side effect of this is an unmount hang in xfs_wait_buftarg() as the associated buffers remain held by the delwri queue indefinitely. This is reproduced on repeated runs of generic/531 with an fs format (-mrmapbt=1 -bsize=1k) that happens to also reproduce transaction aborts. Update xfsaild to not idle until both the AIL and associated delwri queue are empty and update the push code to continue delwri queue submission attempts even when the AIL is empty. This allows the AIL to eventually release aborted buffers stranded on the delwri queue when they are unlocked by the associated transaction. This should have no significant effect on normal runtime behavior because the xfsaild currently idles only when the AIL is empty and in practice the AIL is rarely empty with a populated delwri queue. The items must be AIL resident to land in the queue in the first place and generally aren't removed until writeback completes. Note that the pre-existing delwri queue cancel logic in the exit path is retained because task stop is external, could technically come at any point, and xfsaild is still responsible to release its buffer references before it exits. Signed-off-by:
-
- 17 Jul, 2020 1 commit
-
-
Eric Sandeen authored
The MS_I_VERSION mount flag is exposed via the VFS, as documented in the mount manpages etc; see the iversion and noiversion mount options in mount(8). As a result, mount -o remount looks for this option in /proc/mounts and will only send the I_VERSION flag back in during remount it it is present. Since it's not there, a remount will /remove/ the I_VERSION flag at the vfs level, and iversion functionality is lost. xfs v5 superblocks intend to always have i_version enabled; it is set as a default at mount time, but is lost during remount for the reasons above. The generic fix would be to expose this documented option in /proc/mounts, but since that was rejected, fix it up again in the xfs remount path instead, so that at least xfs won't suffer from this misbehavior. Signed-off-by:
Eric Sandeen <sandeen@redhat.com> Reviewed-by:
-
- 14 Jul, 2020 3 commits
-
-
YueHaibing authored
Remove duplicated include. Signed-off-by:
YueHaibing <yuehaibing@huawei.com> Reviewed-by:
Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
-
Christoph Hellwig authored
These two definitions are unused now. Signed-off-by:
-
Gao Xiang authored
In the course of some operations, we look up the perag from the mount multiple times to get or change perag information. These are often very short pieces of code, so while the lookup cost is generally low, the cost of the lookup is far higher than the cost of the operation we are doing on the perag. Since we changed buffers to hold references to the perag they are cached in, many modification contexts already hold active references to the perag that are held across these operations. This is especially true for any operation that is serialised by an allocation group header buffer. In these cases, we can just use the buffer's reference to the perag to avoid needing to do lookups to access the perag. This means that many operations don't need to do perag lookups at all to access the perag because they've already looked up objects that own persistent references and hence can use that reference instead. Cc: Dave Chinner <dchinner@redhat.com> Cc: "Darrick J. Wong" <darrick.wong@oracle.com> Signed-off-by:
Gao Xiang <hsiangkao@redhat.com> Reviewed-by:
-
- 09 Jul, 2020 1 commit
-
-
Waiman Long authored
Depending on the workloads, the following circular locking dependency warning between sb_internal (a percpu rwsem) and fs_reclaim (a pseudo lock) may show up: ====================================================== WARNING: possible circular locking dependency detected 5.0.0-rc1+ #60 Tainted: G W ------------------------------------------------------ fsfreeze/4346 is trying to acquire lock: 0000000026f1d784 (fs_reclaim){+.+.}, at: fs_reclaim_acquire.part.19+0x5/0x30 but task is already holding lock: 0000000072bfc54b (sb_internal){++++}, at: percpu_down_write+0xb4/0x650 which lock already depends on the new lock. : Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(sb_internal); lock(fs_reclaim); lock(sb_internal); lock(fs_reclaim); *** DEADLOCK *** 4 locks held by fsfreeze/4346: #0: 00000000b478ef56 (sb_writers#8){++++}, at: percpu_down_write+0xb4/0x650 #1: 000000001ec487a9 (&type->s_umount_key#28){++++}, at: freeze_super+0xda/0x290 #2: 000000003edbd5a0 (sb_pagefaults){++++}, at: percpu_down_write+0xb4/0x650 #3: 0000000072bfc54b (sb_internal){++++}, at: percpu_down_write+0xb4/0x650 stack backtrace: Call Trace: dump_stack+0xe0/0x19a print_circular_bug.isra.10.cold.34+0x2f4/0x435 check_prev_add.constprop.19+0xca1/0x15f0 validate_chain.isra.14+0x11af/0x3b50 __lock_acquire+0x728/0x1200 lock_acquire+0x269/0x5a0 fs_reclaim_acquire.part.19+0x29/0x30 fs_reclaim_acquire+0x19/0x20 kmem_cache_alloc+0x3e/0x3f0 kmem_zone_alloc+0x79/0x150 xfs_trans_alloc+0xfa/0x9d0 xfs_sync_sb+0x86/0x170 xfs_log_sbcount+0x10f/0x140 xfs_quiesce_attr+0x134/0x270 xfs_fs_freeze+0x4a/0x70 freeze_super+0x1af/0x290 do_vfs_ioctl+0xedc/0x16c0 ksys_ioctl+0x41/0x80 __x64_sys_ioctl+0x73/0xa9 do_syscall_64+0x18f/0xd23 entry_SYSCALL_64_after_hwframe+0x49/0xbe This is a false positive as all the dirty pages are flushed out before the filesystem can be frozen. One way to avoid this splat is to add GFP_NOFS to the affected allocation calls by using the memalloc_nofs_save()/memalloc_nofs_restore() pair. This shouldn't matter unless the system is really running out of memory. In that particular case, the filesystem freeze operation may fail while it was succeeding previously. Without this patch, the command sequence below will show that the lock dependency chain sb_internal -> fs_reclaim exists. # fsfreeze -f /home # fsfreeze --unfreeze /home # grep -i fs_reclaim -C 3 /proc/lockdep_chains | grep -C 5 sb_internal After applying the patch, such sb_internal -> fs_reclaim lock dependency chain can no longer be found. Because of that, the locking dependency warning will not be shown. Suggested-by:Dave Chinner <david@fromorbit.com> Signed-off-by:
Waiman Long <longman@redhat.com> Reviewed-by:
-
- 07 Jul, 2020 12 commits
-
-
Darrick J. Wong authored
Make sure the rtbitmap is large enough to store the entire bitmap. Signed-off-by:
-
Darrick J. Wong authored
Ensure that the realtime bitmap file is backed entirely by written extents. No holes, no unwritten blocks, etc. Signed-off-by:
-
Dave Chinner authored
This debug code is called on every xfs_iflush() call, which then checks every inode in the buffer for non-zero unlinked list field. Hence it checks every inode in the cluster buffer every time a single inode on that cluster it flushed. This is resulting in: - 38.91% 5.33% [kernel] [k] xfs_iflush - 17.70% xfs_iflush - 9.93% xfs_inobp_check 4.36% xfs_buf_offset 10% of the CPU time spent flushing inodes is repeatedly checking unlinked fields in the buffer. We don't need to do this. The other place we call xfs_inobp_check() is xfs_iunlink_update_dinode(), and this is after we've done this assert for the agino we are about to write into that inode: ASSERT(xfs_verify_agino_or_null(mp, agno, next_agino)); which means we've already checked that the agino we are about to write is not 0 on debug kernels. The inode buffer verifiers do everything else we need, so let's just remove this debug code. Signed-off-by: -
Dave Chinner authored
xfs_iflush_done() does 3 distinct operations to the inodes attached to the buffer. Separate these operations out into functions so that it is easier to modify these operations independently in future. Signed-off-by:
-
Dave Chinner authored
Now that we have all the dirty inodes attached to the cluster buffer, we don't actually have to do radix tree lookups to find them. Sure, the radix tree is efficient, but walking a linked list of just the dirty inodes attached to the buffer is much better. We are also no longer dependent on having a locked inode passed into the function to determine where to start the lookup. This means we can drop it from the function call and treat all inodes the same. We also make xfs_iflush_cluster skip inodes marked with XFS_IRECLAIM. This we avoid races with inodes that reclaim is actively referencing or are being re-initialised by inode lookup. If they are actually dirty, they'll get written by a future cluster flush.... We also add a shutdown check after obtaining the flush lock so that we catch inodes that are dirty in memory and may have inconsistent state due to the shutdown in progress. We abort these inodes directly and so they remove themselves directly from the buffer list and the AIL rather than having to wait for the buffer to be failed and callbacks run to be processed correctly. Signed-off-by:
-
Dave Chinner authored
with xfs_iflush() gone, we can rename xfs_iflush_int() back to xfs_iflush(). Also move it up above xfs_iflush_cluster() so we don't need the forward definition any more. Signed-off-by:
Amir Goldstein <amir73il@gmail.com> Reviewed-by:
-
Dave Chinner authored
Now we have a cached buffer on inode log items, we don't need to do buffer lookups when flushing inodes anymore - all we need to do is lock the buffer and we are ready to go. This largely gets rid of the need for xfs_iflush(), which is essentially just a mechanism to look up the buffer and flush the inode to it. Instead, we can just call xfs_iflush_cluster() with a few modifications to ensure it also flushes the inode we already hold locked. This allows the AIL inode item pushing to be almost entirely non-blocking in XFS - we won't block unless memory allocation for the cluster inode lookup blocks or the block device queues are full. Writeback during inode reclaim becomes a little more complex because we now have to lock the buffer ourselves, but otherwise this change is largely a functional no-op that removes a whole lot of code. Signed-off-by:
-
Dave Chinner authored
Rather than attach inodes to the cluster buffer just when we are doing IO, attach the inodes to the cluster buffer when they are dirtied. The means the buffer always carries a list of dirty inodes that reference it, and we can use that list to make more fundamental changes to inode writeback that aren't otherwise possible. Signed-off-by:
-
Dave Chinner authored
Once we have inodes pinning the cluster buffer and attached whenever they are dirty, we no longer have a guarantee that the items are flush locked when we lock the cluster buffer. Hence we cannot just walk the buffer log item list and modify the attached inodes. If the inode is not flush locked, we have to ILOCK it first and then flush lock it to do all the prerequisite checks needed to avoid races with other code. This is already handled by xfs_ifree_get_one_inode(), so rework the inode iteration loop and function to update all inodes in cache whether they are attached to the buffer or not. Note: we also remove the copying of the log item lsn to the ili_flush_lsn as xfs_iflush_done() now uses the XFS_ISTALE flag to trigger aborts and so flush lsn matching is not needed in IO completion for processing freed inodes. Signed-off-by:
-
Dave Chinner authored
Inode reclaim is quite different now to the way described in various comments, so update all the comments explaining what it does and how it works. Signed-off-by:
-
Dave Chinner authored
Clean up xfs_reclaim_inodes() callers. Most callers want blocking behaviour, so just make the existing SYNC_WAIT behaviour the default. For the xfs_reclaim_worker(), just call xfs_reclaim_inodes_ag() directly because we just want optimistic clean inode reclaim to be done in the background. For xfs_quiesce_attr() we can just remove the inode reclaim calls as they are a historic relic that was required to flush dirty inodes that contained unlogged changes. We now log all changes to the inodes, so the sync AIL push from xfs_log_quiesce() called by xfs_quiesce_attr() will do all the required inode writeback for freeze. Seeing as we now want to loop until all reclaimable inodes have been reclaimed, make xfs_reclaim_inodes() loop on the XFS_ICI_RECLAIM_TAG tag rather than having xfs_reclaim_inodes_ag() tell it that inodes were skipped. This is much more reliable and will always loop until all reclaimable inodes are reclaimed. Signed-off-by:
-
Dave Chinner authored
All background reclaim is SYNC_TRYLOCK already, and even blocking reclaim (SYNC_WAIT) can use trylock mechanisms as xfs_reclaim_inodes_ag() will keep cycling until there are no more reclaimable inodes. Hence we can kill SYNC_TRYLOCK from inode reclaim and make everything unconditionally non-blocking. We remove all the optimistic "avoid blocking on locks" checks done in xfs_reclaim_inode_grab() as nothing blocks on locks anymore. Further, checking XFS_IFLOCK optimistically can result in detecting inodes in the process of being cleaned (i.e. between being removed from the AIL and having the flush lock dropped), so for xfs_reclaim_inodes() to reliably reclaim all inodes we need to drop these checks anyway. Signed-off-by:
-
