From f0e2d93c29dc39ffd24cac180a19d48f700c0706 Mon Sep 17 00:00:00 2001 From: Denys Vlasenko Date: Mon, 19 May 2008 16:31:57 +1000 Subject: [XFS] Remove unused arg from kmem_free() kmem_free() function takes (ptr, size) arguments but doesn't actually use second one. This patch removes size argument from all callsites. SGI-PV: 981498 SGI-Modid: xfs-linux-melb:xfs-kern:31050a Signed-off-by: Denys Vlasenko Signed-off-by: David Chinner Signed-off-by: Lachlan McIlroy --- fs/xfs/linux-2.6/xfs_buf.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'fs/xfs/linux-2.6/xfs_buf.c') diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c index 98e0e86093b4..ed03c6d3c9c1 100644 --- a/fs/xfs/linux-2.6/xfs_buf.c +++ b/fs/xfs/linux-2.6/xfs_buf.c @@ -310,8 +310,7 @@ _xfs_buf_free_pages( xfs_buf_t *bp) { if (bp->b_pages != bp->b_page_array) { - kmem_free(bp->b_pages, - bp->b_page_count * sizeof(struct page *)); + kmem_free(bp->b_pages); } } @@ -1398,7 +1397,7 @@ STATIC void xfs_free_bufhash( xfs_buftarg_t *btp) { - kmem_free(btp->bt_hash, (1<bt_hashshift) * sizeof(xfs_bufhash_t)); + kmem_free(btp->bt_hash); btp->bt_hash = NULL; } @@ -1444,7 +1443,7 @@ xfs_free_buftarg( xfs_unregister_buftarg(btp); kthread_stop(btp->bt_task); - kmem_free(btp, sizeof(*btp)); + kmem_free(btp); } STATIC int @@ -1575,7 +1574,7 @@ xfs_alloc_buftarg( return btp; error: - kmem_free(btp, sizeof(*btp)); + kmem_free(btp); return NULL; } -- cgit v1.2.3 From 19f354d4c3f4c48bf6b2a86227d8e3050e5f7d50 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 20 May 2008 11:31:13 +1000 Subject: [XFS] sort out opening and closing of the block devices Currently closing the rt/log block device is done in the wrong spot, and far too early. So revampt it: - xfs_blkdev_put moved out of xfs_free_buftarg into the caller so that it is done after tearing down the buftarg completely. - call to xfs_unmountfs_close moved from xfs_mountfs into caller so that it's done after tearing down the filesystem completely. - xfs_unmountfs_close is renamed to xfs_close_devices and made static in xfs_super.c - opening of the block devices is split into a helper xfs_open_devices that is symetric in use to xfs_close_devices - xfs_unmountfs can now lose struct cred - error handling around device opening sanitized in xfs_fs_fill_super SGI-PV: 981951 SGI-Modid: xfs-linux-melb:xfs-kern:31193a Signed-off-by: Christoph Hellwig Signed-off-by: David Chinner Signed-off-by: Lachlan McIlroy --- fs/xfs/linux-2.6/xfs_buf.c | 5 +- fs/xfs/linux-2.6/xfs_buf.h | 2 +- fs/xfs/linux-2.6/xfs_super.c | 192 +++++++++++++++++++++++++++---------------- fs/xfs/linux-2.6/xfs_super.h | 3 - fs/xfs/xfs_mount.c | 13 +-- fs/xfs/xfs_mount.h | 3 +- 6 files changed, 123 insertions(+), 95 deletions(-) (limited to 'fs/xfs/linux-2.6/xfs_buf.c') diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c index ed03c6d3c9c1..9cc8f0213095 100644 --- a/fs/xfs/linux-2.6/xfs_buf.c +++ b/fs/xfs/linux-2.6/xfs_buf.c @@ -1427,13 +1427,10 @@ xfs_unregister_buftarg( void xfs_free_buftarg( - xfs_buftarg_t *btp, - int external) + xfs_buftarg_t *btp) { xfs_flush_buftarg(btp, 1); xfs_blkdev_issue_flush(btp); - if (external) - xfs_blkdev_put(btp->bt_bdev); xfs_free_bufhash(btp); iput(btp->bt_mapping->host); diff --git a/fs/xfs/linux-2.6/xfs_buf.h b/fs/xfs/linux-2.6/xfs_buf.h index f948ec7ba9a4..29d1d4adc078 100644 --- a/fs/xfs/linux-2.6/xfs_buf.h +++ b/fs/xfs/linux-2.6/xfs_buf.h @@ -429,7 +429,7 @@ static inline void xfs_bdwrite(void *mp, xfs_buf_t *bp) * Handling of buftargs. */ extern xfs_buftarg_t *xfs_alloc_buftarg(struct block_device *, int); -extern void xfs_free_buftarg(xfs_buftarg_t *, int); +extern void xfs_free_buftarg(xfs_buftarg_t *); extern void xfs_wait_buftarg(xfs_buftarg_t *); extern int xfs_setsize_buftarg(xfs_buftarg_t *, unsigned int, unsigned int); extern int xfs_flush_buftarg(xfs_buftarg_t *, int); diff --git a/fs/xfs/linux-2.6/xfs_super.c b/fs/xfs/linux-2.6/xfs_super.c index 4b6ddf88d44e..055faa06ca22 100644 --- a/fs/xfs/linux-2.6/xfs_super.c +++ b/fs/xfs/linux-2.6/xfs_super.c @@ -766,6 +766,103 @@ xfs_blkdev_issue_flush( blkdev_issue_flush(buftarg->bt_bdev, NULL); } +STATIC void +xfs_close_devices( + struct xfs_mount *mp) +{ + if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { + xfs_free_buftarg(mp->m_logdev_targp); + xfs_blkdev_put(mp->m_logdev_targp->bt_bdev); + } + if (mp->m_rtdev_targp) { + xfs_free_buftarg(mp->m_rtdev_targp); + xfs_blkdev_put(mp->m_rtdev_targp->bt_bdev); + } + xfs_free_buftarg(mp->m_ddev_targp); +} + +/* + * 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( + struct xfs_mount *mp, + struct xfs_mount_args *args) +{ + 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. + */ + if (args->logname[0]) { + error = xfs_blkdev_get(mp, args->logname, &logdev); + if (error) + goto out; + } + + if (args->rtname[0]) { + error = xfs_blkdev_get(mp, args->rtname, &rtdev); + 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; + mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0); + if (!mp->m_ddev_targp) + goto out_close_rtdev; + + if (rtdev) { + mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1); + if (!mp->m_rtdev_targp) + goto out_free_ddev_targ; + } + + if (logdev && logdev != ddev) { + mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1); + 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) + xfs_free_buftarg(mp->m_rtdev_targp); + out_free_ddev_targ: + xfs_free_buftarg(mp->m_ddev_targp); + out_close_rtdev: + if (rtdev) + xfs_blkdev_put(rtdev); + out_close_logdev: + if (logdev && logdev != ddev) + xfs_blkdev_put(logdev); + out: + return error; +} + + + /* * XFS AIL push thread support */ @@ -1138,7 +1235,8 @@ xfs_fs_put_super( unmount_event_flags); } - xfs_unmountfs(mp, NULL); + xfs_unmountfs(mp); + xfs_close_devices(mp); xfs_qmops_put(mp); xfs_dmops_put(mp); kmem_free(mp); @@ -1585,16 +1683,6 @@ xfs_finish_flags( return 0; } -/* - * 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 vfsp->vfs_super->s_bdev. - */ STATIC int xfs_fs_fill_super( struct super_block *sb, @@ -1604,8 +1692,6 @@ xfs_fs_fill_super( struct inode *root; struct xfs_mount *mp = NULL; struct xfs_mount_args *args = xfs_args_allocate(sb, silent); - struct block_device *ddev = sb->s_bdev; - struct block_device *logdev = NULL, *rtdev = NULL; int flags = 0, error; mp = xfs_mount_init(); @@ -1634,61 +1720,14 @@ xfs_fs_fill_super( goto fail_vfsop; error = xfs_qmops_get(mp, args); if (error) - goto fail_vfsop; + goto out_put_dmops; if (args->flags & XFSMNT_QUIET) flags |= XFS_MFSI_QUIET; - /* - * Open real time and log devices - order is important. - */ - if (args->logname[0]) { - error = xfs_blkdev_get(mp, args->logname, &logdev); - if (error) - goto fail_vfsop; - } - if (args->rtname[0]) { - error = xfs_blkdev_get(mp, args->rtname, &rtdev); - if (error) { - xfs_blkdev_put(logdev); - goto fail_vfsop; - } - - if (rtdev == ddev || rtdev == logdev) { - cmn_err(CE_WARN, - "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev."); - xfs_blkdev_put(logdev); - xfs_blkdev_put(rtdev); - error = EINVAL; - goto fail_vfsop; - } - } - - /* - * Setup xfs_mount buffer target pointers - */ - error = ENOMEM; - mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0); - if (!mp->m_ddev_targp) { - xfs_blkdev_put(logdev); - xfs_blkdev_put(rtdev); - goto fail_vfsop; - } - if (rtdev) { - mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1); - if (!mp->m_rtdev_targp) { - xfs_blkdev_put(logdev); - xfs_blkdev_put(rtdev); - goto error0; - } - } - mp->m_logdev_targp = (logdev && logdev != ddev) ? - xfs_alloc_buftarg(logdev, 1) : mp->m_ddev_targp; - if (!mp->m_logdev_targp) { - xfs_blkdev_put(logdev); - xfs_blkdev_put(rtdev); - goto error0; - } + error = xfs_open_devices(mp, args); + if (error) + goto out_put_qmops; /* * Setup flags based on mount(2) options and then the superblock @@ -1708,7 +1747,9 @@ xfs_fs_fill_super( */ error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize, mp->m_sb.sb_sectsize); - if (!error && logdev && logdev != ddev) { + if (error) + goto error2; + 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)) @@ -1716,13 +1757,16 @@ xfs_fs_fill_super( error = xfs_setsize_buftarg(mp->m_logdev_targp, mp->m_sb.sb_blocksize, log_sector_size); + if (error) + goto error2; } - if (!error && rtdev) + 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) - goto error2; + if (error) + goto error2; + } if (mp->m_flags & XFS_MOUNT_BARRIER) xfs_mountfs_check_barriers(mp); @@ -1778,13 +1822,14 @@ xfs_fs_fill_super( xfs_freesb(mp); error1: xfs_binval(mp->m_ddev_targp); - if (logdev && logdev != ddev) + if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) xfs_binval(mp->m_logdev_targp); - if (rtdev) + if (mp->m_rtdev_targp) xfs_binval(mp->m_rtdev_targp); - error0: - xfs_unmountfs_close(mp, NULL); + xfs_close_devices(mp); + out_put_qmops: xfs_qmops_put(mp); + out_put_dmops: xfs_dmops_put(mp); goto fail_vfsop; @@ -1810,7 +1855,8 @@ xfs_fs_fill_super( IRELE(mp->m_rootip); - xfs_unmountfs(mp, NULL); + xfs_unmountfs(mp); + xfs_close_devices(mp); xfs_qmops_put(mp); xfs_dmops_put(mp); kmem_free(mp); diff --git a/fs/xfs/linux-2.6/xfs_super.h b/fs/xfs/linux-2.6/xfs_super.h index 3efb7c6d3303..212bdc7a7897 100644 --- a/fs/xfs/linux-2.6/xfs_super.h +++ b/fs/xfs/linux-2.6/xfs_super.h @@ -107,9 +107,6 @@ extern void xfs_initialize_vnode(struct xfs_mount *mp, bhv_vnode_t *vp, extern void xfs_flush_inode(struct xfs_inode *); extern void xfs_flush_device(struct xfs_inode *); -extern int xfs_blkdev_get(struct xfs_mount *, const char *, - struct block_device **); -extern void xfs_blkdev_put(struct block_device *); extern void xfs_blkdev_issue_flush(struct xfs_buftarg *); extern const struct export_operations xfs_export_operations; diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c index ee5df5fae829..c67f8a9ae418 100644 --- a/fs/xfs/xfs_mount.c +++ b/fs/xfs/xfs_mount.c @@ -1278,7 +1278,7 @@ xfs_mountfs( * log and makes sure that incore structures are freed. */ int -xfs_unmountfs(xfs_mount_t *mp, struct cred *cr) +xfs_unmountfs(xfs_mount_t *mp) { __uint64_t resblks; int error = 0; @@ -1345,7 +1345,6 @@ xfs_unmountfs(xfs_mount_t *mp, struct cred *cr) */ ASSERT(mp->m_inodes == NULL); - xfs_unmountfs_close(mp, cr); if ((mp->m_flags & XFS_MOUNT_NOUUID) == 0) uuid_table_remove(&mp->m_sb.sb_uuid); @@ -1356,16 +1355,6 @@ xfs_unmountfs(xfs_mount_t *mp, struct cred *cr) return 0; } -void -xfs_unmountfs_close(xfs_mount_t *mp, struct cred *cr) -{ - if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) - xfs_free_buftarg(mp->m_logdev_targp, 1); - if (mp->m_rtdev_targp) - xfs_free_buftarg(mp->m_rtdev_targp, 1); - xfs_free_buftarg(mp->m_ddev_targp, 0); -} - STATIC void xfs_unmountfs_wait(xfs_mount_t *mp) { diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h index 4aff0c125ad3..5acde7fd9c0a 100644 --- a/fs/xfs/xfs_mount.h +++ b/fs/xfs/xfs_mount.h @@ -518,8 +518,7 @@ extern void xfs_mount_free(xfs_mount_t *mp); extern int xfs_mountfs(xfs_mount_t *mp, int); extern void xfs_mountfs_check_barriers(xfs_mount_t *mp); -extern int xfs_unmountfs(xfs_mount_t *, struct cred *); -extern void xfs_unmountfs_close(xfs_mount_t *, struct cred *); +extern int xfs_unmountfs(xfs_mount_t *); extern int xfs_unmountfs_writesb(xfs_mount_t *); extern int xfs_unmount_flush(xfs_mount_t *, int); extern int xfs_mod_incore_sb(xfs_mount_t *, xfs_sb_field_t, int64_t, int); -- cgit v1.2.3 From 3790689fa3c771bba6bafb7dee3e8389dd0b55bc Mon Sep 17 00:00:00 2001 From: Lachlan McIlroy Date: Wed, 13 Aug 2008 15:42:10 +1000 Subject: [XFS] Do not access buffers after dropping reference count We should not access a buffer after dropping it's reference count otherwise we could race with another thread that releases the final reference count and frees the buffer causing us to access potentially unmapped memory. The bug this change fixes only occured on DEBUG XFS since the offending code was in an ASSERT. SGI-PV: 984429 SGI-Modid: xfs-linux-melb:xfs-kern:31715a Signed-off-by: Lachlan McIlroy Signed-off-by: David Chinner --- fs/xfs/linux-2.6/xfs_buf.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'fs/xfs/linux-2.6/xfs_buf.c') diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c index 9cc8f0213095..9f45c74f1a84 100644 --- a/fs/xfs/linux-2.6/xfs_buf.c +++ b/fs/xfs/linux-2.6/xfs_buf.c @@ -838,6 +838,7 @@ xfs_buf_rele( return; } + ASSERT(atomic_read(&bp->b_hold) > 0); if (atomic_dec_and_lock(&bp->b_hold, &hash->bh_lock)) { if (bp->b_relse) { atomic_inc(&bp->b_hold); @@ -851,11 +852,6 @@ xfs_buf_rele( spin_unlock(&hash->bh_lock); xfs_buf_free(bp); } - } else { - /* - * Catch reference count leaks - */ - ASSERT(atomic_read(&bp->b_hold) >= 0); } } -- cgit v1.2.3 From d63f154a36c4ec7c2826874371887cecae375ebb Mon Sep 17 00:00:00 2001 From: Lachlan McIlroy Date: Wed, 13 Aug 2008 16:28:40 +1000 Subject: [XFS] Fix compile failure in xfs_buf_trace() SGI-PV: 957103 SGI-Modid: xfs-linux-melb:xfs-kern:31804a Signed-off-by: Lachlan McIlroy --- fs/xfs/linux-2.6/xfs_buf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/xfs/linux-2.6/xfs_buf.c') diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c index 9f45c74f1a84..a01daa7b5922 100644 --- a/fs/xfs/linux-2.6/xfs_buf.c +++ b/fs/xfs/linux-2.6/xfs_buf.c @@ -58,7 +58,7 @@ xfs_buf_trace( bp, id, (void *)(unsigned long)bp->b_flags, (void *)(unsigned long)bp->b_hold.counter, - (void *)(unsigned long)bp->b_sema.count.counter, + (void *)(unsigned long)bp->b_sema.count, (void *)current, data, ra, (void *)(unsigned long)((bp->b_file_offset>>32) & 0xffffffff), -- cgit v1.2.3 From b4dd330b9e0c9c78ebff754e72563b148f05e9e0 Mon Sep 17 00:00:00 2001 From: David Chinner Date: Wed, 13 Aug 2008 16:36:11 +1000 Subject: [XFS] replace the XFS buf iodone semaphore with a completion The xfs_buf_t b_iodonesema is really just a semaphore that wants to be a completion. Change it to a completion and remove the last user of the sema_t from XFS. SGI-PV: 981498 SGI-Modid: xfs-linux-melb:xfs-kern:31815a Signed-off-by: David Chinner Signed-off-by: Lachlan McIlroy --- fs/xfs/linux-2.6/xfs_buf.c | 6 +++--- fs/xfs/linux-2.6/xfs_buf.h | 4 ++-- fs/xfs/xfs_buf_item.c | 2 +- fs/xfs/xfs_rw.c | 2 +- 4 files changed, 7 insertions(+), 7 deletions(-) (limited to 'fs/xfs/linux-2.6/xfs_buf.c') diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c index a01daa7b5922..a0072d9315a9 100644 --- a/fs/xfs/linux-2.6/xfs_buf.c +++ b/fs/xfs/linux-2.6/xfs_buf.c @@ -253,7 +253,7 @@ _xfs_buf_initialize( memset(bp, 0, sizeof(xfs_buf_t)); atomic_set(&bp->b_hold, 1); - init_MUTEX_LOCKED(&bp->b_iodonesema); + init_completion(&bp->b_iowait); INIT_LIST_HEAD(&bp->b_list); INIT_LIST_HEAD(&bp->b_hash_list); init_MUTEX_LOCKED(&bp->b_sema); /* held, no waiters */ @@ -1033,7 +1033,7 @@ xfs_buf_ioend( xfs_buf_iodone_work(&bp->b_iodone_work); } } else { - up(&bp->b_iodonesema); + complete(&bp->b_iowait); } } @@ -1271,7 +1271,7 @@ xfs_buf_iowait( XB_TRACE(bp, "iowait", 0); if (atomic_read(&bp->b_io_remaining)) blk_run_address_space(bp->b_target->bt_mapping); - down(&bp->b_iodonesema); + wait_for_completion(&bp->b_iowait); XB_TRACE(bp, "iowaited", (long)bp->b_error); return bp->b_error; } diff --git a/fs/xfs/linux-2.6/xfs_buf.h b/fs/xfs/linux-2.6/xfs_buf.h index 29d1d4adc078..fe0109956656 100644 --- a/fs/xfs/linux-2.6/xfs_buf.h +++ b/fs/xfs/linux-2.6/xfs_buf.h @@ -157,7 +157,7 @@ typedef struct xfs_buf { xfs_buf_iodone_t b_iodone; /* I/O completion function */ xfs_buf_relse_t b_relse; /* releasing function */ xfs_buf_bdstrat_t b_strat; /* pre-write function */ - struct semaphore b_iodonesema; /* Semaphore for I/O waiters */ + struct completion b_iowait; /* queue for I/O waiters */ void *b_fspriv; void *b_fspriv2; void *b_fspriv3; @@ -352,7 +352,7 @@ extern void xfs_buf_trace(xfs_buf_t *, char *, void *, void *); #define XFS_BUF_CPSEMA(bp) (xfs_buf_cond_lock(bp) == 0) #define XFS_BUF_VSEMA(bp) xfs_buf_unlock(bp) #define XFS_BUF_PSEMA(bp,x) xfs_buf_lock(bp) -#define XFS_BUF_V_IODONESEMA(bp) up(&bp->b_iodonesema); +#define XFS_BUF_FINISH_IOWAIT(bp) complete(&bp->b_iowait); #define XFS_BUF_SET_TARGET(bp, target) ((bp)->b_target = (target)) #define XFS_BUF_TARGET(bp) ((bp)->b_target) diff --git a/fs/xfs/xfs_buf_item.c b/fs/xfs/xfs_buf_item.c index d86ca2c03a70..215c36dee324 100644 --- a/fs/xfs/xfs_buf_item.c +++ b/fs/xfs/xfs_buf_item.c @@ -1056,7 +1056,7 @@ xfs_buf_iodone_callbacks( anyway. */ XFS_BUF_SET_BRELSE_FUNC(bp,xfs_buf_error_relse); XFS_BUF_DONE(bp); - XFS_BUF_V_IODONESEMA(bp); + XFS_BUF_FINISH_IOWAIT(bp); } return; } diff --git a/fs/xfs/xfs_rw.c b/fs/xfs/xfs_rw.c index b0f31c09a76d..3a82576dde9a 100644 --- a/fs/xfs/xfs_rw.c +++ b/fs/xfs/xfs_rw.c @@ -314,7 +314,7 @@ xfs_bioerror_relse( * ASYNC buffers. */ XFS_BUF_ERROR(bp, EIO); - XFS_BUF_V_IODONESEMA(bp); + XFS_BUF_FINISH_IOWAIT(bp); } else { xfs_buf_relse(bp); } -- cgit v1.2.3 From 5695ef46ef02ba1c6658daa46e6879a2d4f52f5f Mon Sep 17 00:00:00 2001 From: Lachlan McIlroy Date: Wed, 13 Aug 2008 16:51:57 +1000 Subject: [XFS] Use KM_NOFS for debug trace buffers Use KM_NOFS to prevent recursion back into the filesystem which can cause deadlocks. In the case of xfs_iread() we hold the lock on the inode cluster buffer while allocating memory for the trace buffers. If we recurse back into XFS to flush data that may require a transaction to allocate extents which needs log space. This can deadlock with the xfsaild thread which can't push the tail of the log because it is trying to get the inode cluster buffer lock. SGI-PV: 981498 SGI-Modid: xfs-linux-melb:xfs-kern:31838a Signed-off-by: Lachlan McIlroy Signed-off-by: David Chinner --- fs/xfs/linux-2.6/xfs_buf.c | 2 +- fs/xfs/quota/xfs_dquot.c | 2 +- fs/xfs/xfs_buf_item.c | 2 +- fs/xfs/xfs_filestream.c | 2 +- fs/xfs/xfs_inode.c | 12 ++++++------ fs/xfs/xfs_log.c | 2 +- 6 files changed, 11 insertions(+), 11 deletions(-) (limited to 'fs/xfs/linux-2.6/xfs_buf.c') diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c index a0072d9315a9..986061ae1b9b 100644 --- a/fs/xfs/linux-2.6/xfs_buf.c +++ b/fs/xfs/linux-2.6/xfs_buf.c @@ -1795,7 +1795,7 @@ int __init xfs_buf_init(void) { #ifdef XFS_BUF_TRACE - xfs_buf_trace_buf = ktrace_alloc(XFS_BUF_TRACE_SIZE, KM_SLEEP); + xfs_buf_trace_buf = ktrace_alloc(XFS_BUF_TRACE_SIZE, KM_NOFS); #endif xfs_buf_zone = kmem_zone_init_flags(sizeof(xfs_buf_t), "xfs_buf", diff --git a/fs/xfs/quota/xfs_dquot.c b/fs/xfs/quota/xfs_dquot.c index 32290578c661..f2705f2fd43c 100644 --- a/fs/xfs/quota/xfs_dquot.c +++ b/fs/xfs/quota/xfs_dquot.c @@ -112,7 +112,7 @@ xfs_qm_dqinit( complete(&dqp->q_flush); #ifdef XFS_DQUOT_TRACE - dqp->q_trace = ktrace_alloc(DQUOT_TRACE_SIZE, KM_SLEEP); + dqp->q_trace = ktrace_alloc(DQUOT_TRACE_SIZE, KM_NOFS); xfs_dqtrace_entry(dqp, "DQINIT"); #endif } else { diff --git a/fs/xfs/xfs_buf_item.c b/fs/xfs/xfs_buf_item.c index 215c36dee324..608c30c3f76b 100644 --- a/fs/xfs/xfs_buf_item.c +++ b/fs/xfs/xfs_buf_item.c @@ -737,7 +737,7 @@ xfs_buf_item_init( bip->bli_format.blf_len = (ushort)BTOBB(XFS_BUF_COUNT(bp)); bip->bli_format.blf_map_size = map_size; #ifdef XFS_BLI_TRACE - bip->bli_trace = ktrace_alloc(XFS_BLI_TRACE_SIZE, KM_SLEEP); + bip->bli_trace = ktrace_alloc(XFS_BLI_TRACE_SIZE, KM_NOFS); #endif #ifdef XFS_TRANS_DEBUG diff --git a/fs/xfs/xfs_filestream.c b/fs/xfs/xfs_filestream.c index c38fd14fca29..f3bb75da384e 100644 --- a/fs/xfs/xfs_filestream.c +++ b/fs/xfs/xfs_filestream.c @@ -400,7 +400,7 @@ xfs_filestream_init(void) if (!item_zone) return -ENOMEM; #ifdef XFS_FILESTREAMS_TRACE - xfs_filestreams_trace_buf = ktrace_alloc(XFS_FSTRM_KTRACE_SIZE, KM_SLEEP); + xfs_filestreams_trace_buf = ktrace_alloc(XFS_FSTRM_KTRACE_SIZE, KM_NOFS); #endif return 0; } diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index 41371c56f360..358511b85ced 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -835,22 +835,22 @@ xfs_iread( * Do this before xfs_iformat in case it adds entries. */ #ifdef XFS_INODE_TRACE - ip->i_trace = ktrace_alloc(INODE_TRACE_SIZE, KM_SLEEP); + ip->i_trace = ktrace_alloc(INODE_TRACE_SIZE, KM_NOFS); #endif #ifdef XFS_BMAP_TRACE - ip->i_xtrace = ktrace_alloc(XFS_BMAP_KTRACE_SIZE, KM_SLEEP); + ip->i_xtrace = ktrace_alloc(XFS_BMAP_KTRACE_SIZE, KM_NOFS); #endif #ifdef XFS_BMBT_TRACE - ip->i_btrace = ktrace_alloc(XFS_BMBT_KTRACE_SIZE, KM_SLEEP); + ip->i_btrace = ktrace_alloc(XFS_BMBT_KTRACE_SIZE, KM_NOFS); #endif #ifdef XFS_RW_TRACE - ip->i_rwtrace = ktrace_alloc(XFS_RW_KTRACE_SIZE, KM_SLEEP); + ip->i_rwtrace = ktrace_alloc(XFS_RW_KTRACE_SIZE, KM_NOFS); #endif #ifdef XFS_ILOCK_TRACE - ip->i_lock_trace = ktrace_alloc(XFS_ILOCK_KTRACE_SIZE, KM_SLEEP); + ip->i_lock_trace = ktrace_alloc(XFS_ILOCK_KTRACE_SIZE, KM_NOFS); #endif #ifdef XFS_DIR2_TRACE - ip->i_dir_trace = ktrace_alloc(XFS_DIR2_KTRACE_SIZE, KM_SLEEP); + ip->i_dir_trace = ktrace_alloc(XFS_DIR2_KTRACE_SIZE, KM_NOFS); #endif /* diff --git a/fs/xfs/xfs_log.c b/fs/xfs/xfs_log.c index 9c23864d4fd0..1f6f780dbd39 100644 --- a/fs/xfs/xfs_log.c +++ b/fs/xfs/xfs_log.c @@ -160,7 +160,7 @@ void xlog_trace_iclog(xlog_in_core_t *iclog, uint state) { if (!iclog->ic_trace) - iclog->ic_trace = ktrace_alloc(256, KM_SLEEP); + iclog->ic_trace = ktrace_alloc(256, KM_NOFS); ktrace_enter(iclog->ic_trace, (void *)((unsigned long)state), (void *)((unsigned long)current_pid()), -- cgit v1.2.3