[XFS] Prevent ENOSPC from aborting transactions that need to succeed

During delayed allocation extent conversion or unwritten extent
conversion, we need to reserve some blocks for transactions reservations.
We need to reserve these blocks in case a btree split occurs and we need
to allocate some blocks.

Unfortunately, we've only ever reserved the number of data blocks we are
allocating, so in both the unwritten and delalloc case we can get ENOSPC
to the transaction reservation. This is bad because in both cases we
cannot report the failure to the writing application.

The fix is two-fold:

1 - leverage the reserved block infrastructure XFS already
has to reserve a small pool of blocks by default to allow
specially marked transactions to dip into when we are at
ENOSPC.
Default setting is min(5%, 1024 blocks).

2 - convert critical transaction reservations to be allowed
to dip into this pool. Spots changed are delalloc
conversion, unwritten extent conversion and growing a
filesystem at ENOSPC.
This also allows growing the filesytsem to succeed at ENOSPC.

SGI-PV: 964468
SGI-Modid: xfs-linux-melb:xfs-kern:28865a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>

3 files changed, 50 insertions, 19 deletions

diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index ddd45e5..2251a49 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -177,6 +177,7 @@ xfs_growfs_data_private(
 		up_write(&mp->m_peraglock);
 	}
 	tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
+	tp->t_flags |= XFS_TRANS_RESERVE;
 	if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
 			XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
 		xfs_trans_cancel(tp, 0);
@@ -499,8 +500,9 @@ xfs_reserve_blocks(
 	unsigned long		s;
 
 	/* If inval is null, report current values and return */
-
 	if (inval == (__uint64_t *)NULL) {
+		if (!outval)
+			return EINVAL;
 		outval->resblks = mp->m_resblks;
 		outval->resblks_avail = mp->m_resblks_avail;
 		return 0;
@@ -563,8 +565,10 @@ retry:
 		}
 	}
 out:
-	outval->resblks = mp->m_resblks;
-	outval->resblks_avail = mp->m_resblks_avail;
+	if (outval) {
+		outval->resblks = mp->m_resblks;
+		outval->resblks_avail = mp->m_resblks_avail;
+	}
 	XFS_SB_UNLOCK(mp, s);
 
 	if (fdblks_delta) {
diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c
index 3f2b9f2..ab50621 100644
--- a/fs/xfs/xfs_iomap.c
+++ b/fs/xfs/xfs_iomap.c
@@ -489,13 +489,13 @@ xfs_iomap_write_direct(
 	if (unlikely(rt)) {
 		resrtextents = qblocks = resaligned;
 		resrtextents /= mp->m_sb.sb_rextsize;
-  		resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0);
-  		quota_flag = XFS_QMOPT_RES_RTBLKS;
-  	} else {
-  		resrtextents = 0;
+		resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0);
+		quota_flag = XFS_QMOPT_RES_RTBLKS;
+	} else {
+		resrtextents = 0;
 		resblks = qblocks = XFS_DIOSTRAT_SPACE_RES(mp, resaligned);
-  		quota_flag = XFS_QMOPT_RES_REGBLKS;
-  	}
+		quota_flag = XFS_QMOPT_RES_REGBLKS;
+	}
 
 	/*
 	 * Allocate and setup the transaction
@@ -788,18 +788,12 @@ xfs_iomap_write_allocate(
 		nimaps = 0;
 		while (nimaps == 0) {
 			tp = xfs_trans_alloc(mp, XFS_TRANS_STRAT_WRITE);
+			tp->t_flags |= XFS_TRANS_RESERVE;
 			nres = XFS_EXTENTADD_SPACE_RES(mp, XFS_DATA_FORK);
 			error = xfs_trans_reserve(tp, nres,
 					XFS_WRITE_LOG_RES(mp),
 					0, XFS_TRANS_PERM_LOG_RES,
 					XFS_WRITE_LOG_COUNT);
-			if (error == ENOSPC) {
-				error = xfs_trans_reserve(tp, 0,
-						XFS_WRITE_LOG_RES(mp),
-						0,
-						XFS_TRANS_PERM_LOG_RES,
-						XFS_WRITE_LOG_COUNT);
-			}
 			if (error) {
 				xfs_trans_cancel(tp, 0);
 				return XFS_ERROR(error);
@@ -917,8 +911,8 @@ xfs_iomap_write_unwritten(
 		 * from unwritten to real. Do allocations in a loop until
 		 * we have covered the range passed in.
 		 */
-
 		tp = xfs_trans_alloc(mp, XFS_TRANS_STRAT_WRITE);
+		tp->t_flags |= XFS_TRANS_RESERVE;
 		error = xfs_trans_reserve(tp, resblks,
 				XFS_WRITE_LOG_RES(mp), 0,
 				XFS_TRANS_PERM_LOG_RES,
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index 39cf6f32..31453ca 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -725,7 +725,7 @@ xfs_mountfs(
 	bhv_vnode_t	*rvp = NULL;
 	int		readio_log, writeio_log;
 	xfs_daddr_t	d;
-	__uint64_t	ret64;
+	__uint64_t	resblks;
 	__int64_t	update_flags;
 	uint		quotamount, quotaflags;
 	int		agno;
@@ -842,6 +842,7 @@ xfs_mountfs(
 	 */
 	if ((mfsi_flags & XFS_MFSI_SECOND) == 0 &&
 	    (mp->m_flags & XFS_MOUNT_NOUUID) == 0) {
+		__uint64_t	ret64;
 		if (xfs_uuid_mount(mp)) {
 			error = XFS_ERROR(EINVAL);
 			goto error1;
@@ -1135,13 +1136,27 @@ xfs_mountfs(
 		goto error4;
 	}
 
-
 	/*
 	 * Complete the quota initialisation, post-log-replay component.
 	 */
 	if ((error = XFS_QM_MOUNT(mp, quotamount, quotaflags, mfsi_flags)))
 		goto error4;
 
+	/*
+	 * Now we are mounted, reserve a small amount of unused space for
+	 * privileged transactions. This is needed so that transaction
+	 * space required for critical operations can dip into this pool
+	 * when at ENOSPC. This is needed for operations like create with
+	 * attr, unwritten extent conversion at ENOSPC, etc. Data allocations
+	 * are not allowed to use this reserved space.
+	 *
+	 * We default to 5% or 1024 fsbs of space reserved, whichever is smaller.
+	 * This may drive us straight to ENOSPC on mount, but that implies
+	 * we were already there on the last unmount.
+	 */
+	resblks = min_t(__uint64_t, mp->m_sb.sb_dblocks / 20, 1024);
+	xfs_reserve_blocks(mp, &resblks, NULL);
+
 	return 0;
 
  error4:
@@ -1181,6 +1196,7 @@ xfs_unmountfs(xfs_mount_t *mp, struct cred *cr)
 #if defined(DEBUG) || defined(INDUCE_IO_ERROR)
 	int64_t		fsid;
 #endif
+	__uint64_t	resblks;
 
 	/*
 	 * We can potentially deadlock here if we have an inode cluster
@@ -1209,6 +1225,23 @@ xfs_unmountfs(xfs_mount_t *mp, struct cred *cr)
 		xfs_binval(mp->m_rtdev_targp);
 	}
 
+	/*
+	 * Unreserve any blocks we have so that when we unmount we don't account
+	 * the reserved free space as used. This is really only necessary for
+	 * lazy superblock counting because it trusts the incore superblock
+	 * counters to be aboslutely correct on clean unmount.
+	 *
+	 * We don't bother correcting this elsewhere for lazy superblock
+	 * counting because on mount of an unclean filesystem we reconstruct the
+	 * correct counter value and this is irrelevant.
+	 *
+	 * For non-lazy counter filesystems, this doesn't matter at all because
+	 * we only every apply deltas to the superblock and hence the incore
+	 * value does not matter....
+	 */
+	resblks = 0;
+	xfs_reserve_blocks(mp, &resblks, NULL);
+
 	xfs_log_sbcount(mp, 1);
 	xfs_unmountfs_writesb(mp);
 	xfs_unmountfs_wait(mp); 		/* wait for async bufs */