From f59cf5c29919d17b61913c3360a7bd29b72975c1 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 4 Dec 2017 17:32:55 -0800 Subject: xfs: remove "no-allocation" reservations for file creations If we create a new file we will need an inode, and usually some metadata in the parent direction. Aiming for everything to go well despite the lack of a reservation leads to dirty transactions cancelled under a heavy create/delete load. This patch removes those nospace transactions, which will lead to slightly earlier ENOSPC on some workloads, but instead prevent file system shutdowns due to cancelling dirty transactions for others. A customer could observe assertations failures and shutdowns due to cancelation of dirty transactions during heavy NFS workloads as shown below: 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728125] XFS: Assertion failed: error != -ENOSPC, file: fs/xfs/xfs_inode.c, line: 1262 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728222] Call Trace: 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728246] [] dump_stack+0x63/0x81 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728262] [] warn_slowpath_common+0x8a/0xc0 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728264] [] warn_slowpath_null+0x1a/0x20 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728285] [] asswarn+0x33/0x40 [xfs] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728308] [] xfs_create+0x7be/0x7d0 [xfs] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728329] [] xfs_generic_create+0x1fb/0x2e0 [xfs] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728348] [] xfs_vn_mknod+0x14/0x20 [xfs] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728366] [] xfs_vn_create+0x13/0x20 [xfs] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728380] [] vfs_create+0xd5/0x140 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728390] [] do_nfsd_create+0x499/0x610 [nfsd] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728396] [] nfsd3_proc_create+0x135/0x210 [nfsd] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728401] [] nfsd_dispatch+0xc3/0x210 [nfsd] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728416] [] svc_process_common+0x453/0x6f0 [sunrpc] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728423] [] svc_process+0x113/0x1f0 [sunrpc] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728427] [] nfsd+0x10f/0x180 [nfsd] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728432] [] ? nfsd_destroy+0x80/0x80 [nfsd] 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728438] [] kthread+0xd8/0xf0 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728441] [] ? kthread_create_on_node+0x1b0/0x1b0 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728451] [] ret_from_fork+0x42/0x70 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728453] [] ? kthread_create_on_node+0x1b0/0x1b0 2017-05-30 21:17:06 kernel: WARNING: [ 2670.728454] ---[ end trace f9822c842fec81d4 ]--- 2017-05-30 21:17:06 kernel: ALERT: [ 2670.728477] XFS (sdb): Internal error xfs_trans_cancel at line 983 of file fs/xfs/xfs_trans.c. Caller xfs_create+0x4ee/0x7d0 [xfs] 2017-05-30 21:17:06 kernel: ALERT: [ 2670.728684] XFS (sdb): Corruption of in-memory data detected. Shutting down filesystem 2017-05-30 21:17:06 kernel: ALERT: [ 2670.728685] XFS (sdb): Please umount the filesystem and rectify the problem(s) Signed-off-by: Christoph Hellwig Reviewed-by: Darrick J. Wong Signed-off-by: Darrick J. Wong --- fs/xfs/xfs_inode.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/xfs/xfs_inode.h') diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h index cc13c3763721..b2136af9289f 100644 --- a/fs/xfs/xfs_inode.h +++ b/fs/xfs/xfs_inode.h @@ -428,7 +428,7 @@ xfs_extlen_t xfs_get_extsz_hint(struct xfs_inode *ip); xfs_extlen_t xfs_get_cowextsz_hint(struct xfs_inode *ip); int xfs_dir_ialloc(struct xfs_trans **, struct xfs_inode *, umode_t, - xfs_nlink_t, dev_t, prid_t, int, + xfs_nlink_t, dev_t, prid_t, struct xfs_inode **, int *); /* from xfs_file.c */ -- cgit v1.2.3 From 91aae6be4139b9e3902656d819e6af66e051bd7a Mon Sep 17 00:00:00 2001 From: "Darrick J. Wong" Date: Thu, 14 Dec 2017 15:42:22 -0800 Subject: xfs: track cowblocks separately in i_flags The EOFBLOCKS/COWBLOCKS tags are totally separate things, so track them with separate i_flags. Right now we're abusing IEOFBLOCKS for both, which is totally bogus because we won't tag the inode with COWBLOCKS if IEOFBLOCKS was set by a previous tagging of the inode with EOFBLOCKS. Found by wiring up clonerange to fsstress in xfs/017. Signed-off-by: Darrick J. Wong Reviewed-by: Dave Chinner Reviewed-by: Christoph Hellwig --- fs/xfs/xfs_icache.c | 33 ++++++++++++++++++++++++--------- fs/xfs/xfs_inode.h | 1 + 2 files changed, 25 insertions(+), 9 deletions(-) (limited to 'fs/xfs/xfs_inode.h') diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 43005fbe8b1e..58d2d4253c8e 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -1536,8 +1536,23 @@ xfs_inode_free_quota_eofblocks( return __xfs_inode_free_quota_eofblocks(ip, xfs_icache_free_eofblocks); } +static inline unsigned long +xfs_iflag_for_tag( + int tag) +{ + switch (tag) { + case XFS_ICI_EOFBLOCKS_TAG: + return XFS_IEOFBLOCKS; + case XFS_ICI_COWBLOCKS_TAG: + return XFS_ICOWBLOCKS; + default: + ASSERT(0); + return 0; + } +} + static void -__xfs_inode_set_eofblocks_tag( +__xfs_inode_set_blocks_tag( xfs_inode_t *ip, void (*execute)(struct xfs_mount *mp), void (*set_tp)(struct xfs_mount *mp, xfs_agnumber_t agno, @@ -1552,10 +1567,10 @@ __xfs_inode_set_eofblocks_tag( * Don't bother locking the AG and looking up in the radix trees * if we already know that we have the tag set. */ - if (ip->i_flags & XFS_IEOFBLOCKS) + if (ip->i_flags & xfs_iflag_for_tag(tag)) return; spin_lock(&ip->i_flags_lock); - ip->i_flags |= XFS_IEOFBLOCKS; + ip->i_flags |= xfs_iflag_for_tag(tag); spin_unlock(&ip->i_flags_lock); pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino)); @@ -1587,13 +1602,13 @@ xfs_inode_set_eofblocks_tag( xfs_inode_t *ip) { trace_xfs_inode_set_eofblocks_tag(ip); - return __xfs_inode_set_eofblocks_tag(ip, xfs_queue_eofblocks, + return __xfs_inode_set_blocks_tag(ip, xfs_queue_eofblocks, trace_xfs_perag_set_eofblocks, XFS_ICI_EOFBLOCKS_TAG); } static void -__xfs_inode_clear_eofblocks_tag( +__xfs_inode_clear_blocks_tag( xfs_inode_t *ip, void (*clear_tp)(struct xfs_mount *mp, xfs_agnumber_t agno, int error, unsigned long caller_ip), @@ -1603,7 +1618,7 @@ __xfs_inode_clear_eofblocks_tag( struct xfs_perag *pag; spin_lock(&ip->i_flags_lock); - ip->i_flags &= ~XFS_IEOFBLOCKS; + ip->i_flags &= ~xfs_iflag_for_tag(tag); spin_unlock(&ip->i_flags_lock); pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino)); @@ -1630,7 +1645,7 @@ xfs_inode_clear_eofblocks_tag( xfs_inode_t *ip) { trace_xfs_inode_clear_eofblocks_tag(ip); - return __xfs_inode_clear_eofblocks_tag(ip, + return __xfs_inode_clear_blocks_tag(ip, trace_xfs_perag_clear_eofblocks, XFS_ICI_EOFBLOCKS_TAG); } @@ -1724,7 +1739,7 @@ xfs_inode_set_cowblocks_tag( xfs_inode_t *ip) { trace_xfs_inode_set_cowblocks_tag(ip); - return __xfs_inode_set_eofblocks_tag(ip, xfs_queue_cowblocks, + return __xfs_inode_set_blocks_tag(ip, xfs_queue_cowblocks, trace_xfs_perag_set_cowblocks, XFS_ICI_COWBLOCKS_TAG); } @@ -1734,6 +1749,6 @@ xfs_inode_clear_cowblocks_tag( xfs_inode_t *ip) { trace_xfs_inode_clear_cowblocks_tag(ip); - return __xfs_inode_clear_eofblocks_tag(ip, + return __xfs_inode_clear_blocks_tag(ip, trace_xfs_perag_clear_cowblocks, XFS_ICI_COWBLOCKS_TAG); } diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h index b2136af9289f..d383e392ec9d 100644 --- a/fs/xfs/xfs_inode.h +++ b/fs/xfs/xfs_inode.h @@ -232,6 +232,7 @@ static inline bool xfs_is_reflink_inode(struct xfs_inode *ip) * log recovery to replay a bmap operation on the inode. */ #define XFS_IRECOVERY (1 << 11) +#define XFS_ICOWBLOCKS (1 << 12)/* has the cowblocks tag set */ /* * Per-lifetime flags need to be reset when re-using a reclaimable inode during -- cgit v1.2.3