Skip to contentSkip to conversation

xfs: journal geometry is not properly bounds checked

PATCH In review Dave Chinner 2 participants 3 messages Jun 19, 2023 XFS filesystem

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 xfs: journal geometry is not properly bounds checked

From: Dave Chinner <[email protected]>

If the journal geometry results in a sector or log stripe unit validation problem, it indicates that we cannot set the log up to safely write to the the journal. In these cases, we must abort the mount because the corruption needs external intervention to resolve. Similarly, a journal that is too large cannot be written to safely, either, so we shouldn't allow those geometries to mount, either.

If the log is too small, we risk having transaction reservations overruning the available log space and the system hanging waiting for space it can never provide. This is purely a runtime hang issue, not a corruption issue as per the first cases listed above. We abort mounts of the log is too small for V5 filesystems, but we must allow v4 filesystems to mount because, historically, there was no log size validity checking and so some systems may still be out there with undersized logs.

The problem is that on V4 filesystems, when we discover a log geometry problem, we skip all the remaining checks and then allow the log to continue mounting. This mean that if one of the log size checks fails, we skip the log stripe unit check. i.e. we allow the mount because a "non-fatal" geometry is violated, and then fail to check the hard fail geometries that should fail the mount.

Move all these fatal checks to the superblock verifier, and add a new check for the two log sector size geometry variables having the same values. This will prevent any attempt to mount a log that has invalid or inconsistent geometries long before we attempt to mount the log.

However, for the minimum log size checks, we can only do that once we've setup up the log and calculated all the iclog sizes and roundoffs. Hence this needs to remain in the log mount code after the log has been initialised. It is also the only case where we should allow a v4 filesystem to continue running, so leave that handling in place, too.

---
 fs/xfs/libxfs/xfs_sb.c | 59 +++++++++++++++++++++++++++++++++++++++++-
 fs/xfs/xfs_log.c       | 45 ++++++++++----------------------
 2 files changed, 72 insertions(+), 32 deletions(-)
fs/xfs/libxfs/xfs_sb.c +58 -1
index ba0f17bc1dc0..0daf6bb37741 100644
@@ -412,7 +412,6 @@ xfs_validate_sb_common(
 	    sbp->sb_inodelog < XFS_DINODE_MIN_LOG			||
 	    sbp->sb_inodelog > XFS_DINODE_MAX_LOG			||
 	    sbp->sb_inodesize != (1 << sbp->sb_inodelog)		||
-	    sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE			||
 	    sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
 	    XFS_FSB_TO_B(mp, sbp->sb_agblocks) < XFS_MIN_AG_BYTES	||
 	    XFS_FSB_TO_B(mp, sbp->sb_agblocks) > XFS_MAX_AG_BYTES	||
@@ -430,6 +429,64 @@ xfs_validate_sb_common(
 		return -EFSCORRUPTED;
 	}
 
+	/*
+	 * Logs that are too large are not supported at all. Reject them
+	 * outright. Logs that are too small are tolerated on v4 filesystems,
+	 * but we can only check that when mounting the log. Hence we skip
+	 * those checks here.
+	 */
+	if (sbp->sb_logblocks > XFS_MAX_LOG_BLOCKS) {
+		xfs_notice(mp,
+		"Log size 0x%x blocks too large, maximum size is 0x%llx blocks",
+			 sbp->sb_logblocks, XFS_MAX_LOG_BLOCKS);
+		return -EFSCORRUPTED;
+	}
+
+	if (XFS_FSB_TO_B(mp, sbp->sb_logblocks) > XFS_MAX_LOG_BYTES) {
+		xfs_warn(mp,
+		"log size 0x%llx bytes too large, maximum size is 0x%llx bytes",
+			 XFS_FSB_TO_B(mp, sbp->sb_logblocks),
+			 XFS_MAX_LOG_BYTES);
+		return -EFSCORRUPTED;
+	}
+
+	/*
+	 * Do not allow filesystems with corrupted log sector or stripe units to
+	 * be mounted. We cannot safely size the iclogs or write to the log if
+	 * the log stripe unit is not valid.
+	 */
+	if (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT) {
+		if (sbp->sb_logsectsize != (1U << sbp->sb_logsectlog)) {
+			xfs_notice(mp,
+			"log sector size in bytes/log2 (0x%x/0x%x) must match",
+				sbp->sb_logsectsize, 1U << sbp->sb_logsectlog);
+			return -EFSCORRUPTED;
+		}
+	} else if (sbp->sb_logsectsize || sbp->sb_logsectlog) {
+		xfs_notice(mp,
+		"log sector size in bytes/log2 (0x%x/0x%x) are not zero",
+			sbp->sb_logsectsize, sbp->sb_logsectlog);
+		return -EFSCORRUPTED;
+	}
+
+	if (sbp->sb_logsunit > 1) {
+		if (sbp->sb_logsunit % sbp->sb_blocksize) {
+			xfs_notice(mp,
+		"log stripe unit %u bytes must be a multiple of block size",
+				sbp->sb_logsunit);
+			return -EFSCORRUPTED;
+		}
+		if (sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE) {
+			xfs_notice(mp,
+		"log stripe unit %u bytes must be a multiple of block size",
+				sbp->sb_logsunit);
+			return -EFSCORRUPTED;
+		}
+	}
+
+
+
+
 	/* Validate the realtime geometry; stolen from xfs_repair */
 	if (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE ||
 	    sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) {
fs/xfs/xfs_log.c +14 -31

72 lines not on this page · raw message

Signature · 1 line
2.40.1
#1 lore raw
Quoted 101 lines · Dave Chinner

From: Dave Chinner <[email protected]>

If the journal geometry results in a sector or log stripe unit validation problem, it indicates that we cannot set the log up to safely write to the the journal. In these cases, we must abort the mount because the corruption needs external intervention to resolve. Similarly, a journal that is too large cannot be written to safely, either, so we shouldn't allow those geometries to mount, either.

If the log is too small, we risk having transaction reservations overruning the available log space and the system hanging waiting for space it can never provide. This is purely a runtime hang issue, not a corruption issue as per the first cases listed above. We abort mounts of the log is too small for V5 filesystems, but we must allow v4 filesystems to mount because, historically, there was no log size validity checking and so some systems may still be out there with undersized logs.

The problem is that on V4 filesystems, when we discover a log geometry problem, we skip all the remaining checks and then allow the log to continue mounting. This mean that if one of the log size checks fails, we skip the log stripe unit check. i.e. we allow the mount because a "non-fatal" geometry is violated, and then fail to check the hard fail geometries that should fail the mount.

Move all these fatal checks to the superblock verifier, and add a new check for the two log sector size geometry variables having the same values. This will prevent any attempt to mount a log that has invalid or inconsistent geometries long before we attempt to mount the log.

However, for the minimum log size checks, we can only do that once we've setup up the log and calculated all the iclog sizes and roundoffs. Hence this needs to remain in the log mount code after the log has been initialised. It is also the only case where we should allow a v4 filesystem to continue running, so leave that handling in place, too.

---
 fs/xfs/libxfs/xfs_sb.c | 59 +++++++++++++++++++++++++++++++++++++++++-
 fs/xfs/xfs_log.c       | 45 ++++++++++----------------------
 2 files changed, 72 insertions(+), 32 deletions(-)
fs/xfs/libxfs/xfs_sb.c +41 -1
index ba0f17bc1dc0..0daf6bb37741 100644
@@ -412,7 +412,6 @@ xfs_validate_sb_common(
 	    sbp->sb_inodelog < XFS_DINODE_MIN_LOG			||
 	    sbp->sb_inodelog > XFS_DINODE_MAX_LOG			||
 	    sbp->sb_inodesize != (1 << sbp->sb_inodelog)		||
-	    sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE			||
 	    sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
 	    XFS_FSB_TO_B(mp, sbp->sb_agblocks) < XFS_MIN_AG_BYTES	||
 	    XFS_FSB_TO_B(mp, sbp->sb_agblocks) > XFS_MAX_AG_BYTES	||
@@ -430,6 +429,64 @@ xfs_validate_sb_common(
 		return -EFSCORRUPTED;
 	}
 
+	/*
+	 * Logs that are too large are not supported at all. Reject them
+	 * outright. Logs that are too small are tolerated on v4 filesystems,
+	 * but we can only check that when mounting the log. Hence we skip
+	 * those checks here.
+	 */
+	if (sbp->sb_logblocks > XFS_MAX_LOG_BLOCKS) {
+		xfs_notice(mp,
+		"Log size 0x%x blocks too large, maximum size is 0x%llx blocks",
+			 sbp->sb_logblocks, XFS_MAX_LOG_BLOCKS);
+		return -EFSCORRUPTED;
+	}
+
+	if (XFS_FSB_TO_B(mp, sbp->sb_logblocks) > XFS_MAX_LOG_BYTES) {
+		xfs_warn(mp,
+		"log size 0x%llx bytes too large, maximum size is 0x%llx bytes",
+			 XFS_FSB_TO_B(mp, sbp->sb_logblocks),
+			 XFS_MAX_LOG_BYTES);
+		return -EFSCORRUPTED;
+	}
+
+	/*
+	 * Do not allow filesystems with corrupted log sector or stripe units to
+	 * be mounted. We cannot safely size the iclogs or write to the log if
+	 * the log stripe unit is not valid.
+	 */
+	if (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT) {
+		if (sbp->sb_logsectsize != (1U << sbp->sb_logsectlog)) {
+			xfs_notice(mp,
+			"log sector size in bytes/log2 (0x%x/0x%x) must match",
+				sbp->sb_logsectsize, 1U << sbp->sb_logsectlog);
+			return -EFSCORRUPTED;
+		}
+	} else if (sbp->sb_logsectsize || sbp->sb_logsectlog) {
+		xfs_notice(mp,
+		"log sector size in bytes/log2 (0x%x/0x%x) are not zero",
+			sbp->sb_logsectsize, sbp->sb_logsectlog);
+		return -EFSCORRUPTED;
+	}
+
+	if (sbp->sb_logsunit > 1) {

fs/xfs/libxfs/xfs_sb.c xfs_validate_sb_common( · line 482

Quoted 10 lines
fs/xfs/libxfs/xfs_sb.c +10 -0
+		if (sbp->sb_logsunit % sbp->sb_blocksize) {
+			xfs_notice(mp,
+		"log stripe unit %u bytes must be a multiple of block size",
+				sbp->sb_logsunit);
+			return -EFSCORRUPTED;
+		}
+		if (sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE) {
+			xfs_notice(mp,
+		"log stripe unit %u bytes must be a multiple of block size",
+				sbp->sb_logsunit);

I guess this xfs_notice message needs to be corrected.

fs/xfs/libxfs/xfs_sb.c xfs_validate_sb_common( · line 487

Quoted 7 lines
fs/xfs/libxfs/xfs_sb.c +7 -0
+			return -EFSCORRUPTED;
+		}
+	}
+
+
+
+

too many new lines here ^^^

Quoted 41 lines
fs/xfs/libxfs/xfs_sb.c +0 -0
 	/* Validate the realtime geometry; stolen from xfs_repair */
 	if (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE ||
 	    sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) {
fs/xfs/xfs_log.c +12 -9
index fc61cc024023..397374d07f73 100644
@@ -639,7 +639,6 @@ xfs_log_mount(
 	int		num_bblks)
 {
 	struct xlog	*log;
-	bool		fatal = xfs_has_crc(mp);
 	int		error = 0;
 	int		min_logfsbs;
 
@@ -663,51 +662,35 @@ xfs_log_mount(
 	mp->m_log = log;
 
 	/*
-	 * Validate the given log space and drop a critical message via syslog
-	 * if the log size is too small that would lead to some unexpected
-	 * situations in transaction log space reservation stage.
+	 * Now that we have set up the log and it's internal geometry
+	 * parameters, we can validate the given log space and drop a critical
+	 * message via syslog if the log size is too small. A log that is too
+	 * small can lead to unexpected situations in transaction log space
+	 * reservation stage. The superblock verifier has already validated all
+	 * the other log geometry constraints, so we don't have to check those
+	 * here.
 	 *
-	 * Note: we can't just reject the mount if the validation fails.  This
-	 * would mean that people would have to downgrade their kernel just to
-	 * remedy the situation as there is no way to grow the log (short of
-	 * black magic surgery with xfs_db).
+	 * Note: For v4 filesystems, we can't just reject the mount if the
+	 * validation fails.  This would mean that people would have to
+	 * downgrade their kernel just to remedy the situation as there is no
+	 * way to grow the log (short of black magic surgery with xfs_db).
 	 *
-	 * We can, however, reject mounts for CRC format filesystems, as the
+	 * We can, however, reject mounts for V5 format filesystems, as the

fs/xfs/xfs_log.c xfs_log_mount( · line 687

Quoted 10 lines
fs/xfs/xfs_log.c +0 -1
 	 * mkfs binary being used to make the filesystem should never create a
 	 * filesystem with a log that is too small.
 	 */
 	min_logfsbs = xfs_log_calc_minimum_size(mp);
-
 	if (mp->m_sb.sb_logblocks < min_logfsbs) {
 		xfs_warn(mp,
 		"Log size %d blocks too small, minimum size is %d blocks",
 			 mp->m_sb.sb_logblocks, min_logfsbs);
 		error = -EINVAL;

Are we using this error now?

Quoted 20 lines
fs/xfs/xfs_log.c +0 -20
-	} else if (mp->m_sb.sb_logblocks > XFS_MAX_LOG_BLOCKS) {
-		xfs_warn(mp,
-		"Log size %d blocks too large, maximum size is %lld blocks",
-			 mp->m_sb.sb_logblocks, XFS_MAX_LOG_BLOCKS);
-		error = -EINVAL;
-	} else if (XFS_FSB_TO_B(mp, mp->m_sb.sb_logblocks) > XFS_MAX_LOG_BYTES) {
-		xfs_warn(mp,
-		"log size %lld bytes too large, maximum size is %lld bytes",
-			 XFS_FSB_TO_B(mp, mp->m_sb.sb_logblocks),
-			 XFS_MAX_LOG_BYTES);
-		error = -EINVAL;
-	} else if (mp->m_sb.sb_logsunit > 1 &&
-		   mp->m_sb.sb_logsunit % mp->m_sb.sb_blocksize) {
-		xfs_warn(mp,
-		"log stripe unit %u bytes must be a multiple of block size",
-			 mp->m_sb.sb_logsunit);
-		error = -EINVAL;
-		fatal = true;
-	}
-	if (error) {

fs/xfs/xfs_log.c xfs_log_mount( · line 696

Quoted 10 lines
fs/xfs/xfs_log.c +2 -1
+
 		/*
 		 * Log check errors are always fatal on v5; or whenever bad
 		 * metadata leads to a crash.
 		 */
-		if (fatal) {
+		if (xfs_has_crc(mp)) {
 			xfs_crit(mp, "AAIEEE! Log failed size checks. Abort!");
 			ASSERT(0);
 			goto out_free_log;

yes, only here in goto out_free_log we will return "error". Then why not shift error = -EINVAL in this if block?

-ritesh

#2 lore raw
Quoted 15 lines · Ritesh Harjani

fs/xfs/libxfs/xfs_sb.c xfs_validate_sb_common( · line 482

Quoted 11 lines · Dave Chinner
fs/xfs/libxfs/xfs_sb.c +11 -0
+	if (sbp->sb_logsunit > 1) {
+		if (sbp->sb_logsunit % sbp->sb_blocksize) {
+			xfs_notice(mp,
+		"log stripe unit %u bytes must be a multiple of block size",
+				sbp->sb_logsunit);
+			return -EFSCORRUPTED;
+		}
+		if (sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE) {
+			xfs_notice(mp,
+		"log stripe unit %u bytes must be a multiple of block size",
+				sbp->sb_logsunit);

I guess this xfs_notice message needs to be corrected.

Quoted 9 lines

fs/xfs/libxfs/xfs_sb.c xfs_validate_sb_common( · line 487

Quoted 7 lines
fs/xfs/libxfs/xfs_sb.c +7 -0
+			return -EFSCORRUPTED;
+		}
+	}
+
+
+
+

too many new lines here ^^^

Fixed.

Quoted 15 lines

fs/xfs/xfs_log.c xfs_log_mount( · line 687

Quoted 13 lines
fs/xfs/xfs_log.c +1 -2
-	 * We can, however, reject mounts for CRC format filesystems, as the
+	 * We can, however, reject mounts for V5 format filesystems, as the
 	 * mkfs binary being used to make the filesystem should never create a
 	 * filesystem with a log that is too small.
 	 */
 	min_logfsbs = xfs_log_calc_minimum_size(mp);
-
 	if (mp->m_sb.sb_logblocks < min_logfsbs) {
 		xfs_warn(mp,
 		"Log size %d blocks too small, minimum size is %d blocks",
 			 mp->m_sb.sb_logblocks, min_logfsbs);
 		error = -EINVAL;

Are we using this error now?

I'm not changing the existing error for this case.

Quoted 26 lines
Quoted 26 lines
fs/xfs/xfs_log.c +1 -20
-	} else if (mp->m_sb.sb_logblocks > XFS_MAX_LOG_BLOCKS) {
-		xfs_warn(mp,
-		"Log size %d blocks too large, maximum size is %lld blocks",
-			 mp->m_sb.sb_logblocks, XFS_MAX_LOG_BLOCKS);
-		error = -EINVAL;
-	} else if (XFS_FSB_TO_B(mp, mp->m_sb.sb_logblocks) > XFS_MAX_LOG_BYTES) {
-		xfs_warn(mp,
-		"log size %lld bytes too large, maximum size is %lld bytes",
-			 XFS_FSB_TO_B(mp, mp->m_sb.sb_logblocks),
-			 XFS_MAX_LOG_BYTES);
-		error = -EINVAL;
-	} else if (mp->m_sb.sb_logsunit > 1 &&
-		   mp->m_sb.sb_logsunit % mp->m_sb.sb_blocksize) {
-		xfs_warn(mp,
-		"log stripe unit %u bytes must be a multiple of block size",
-			 mp->m_sb.sb_logsunit);
-		error = -EINVAL;
-		fatal = true;
-	}
-	if (error) {
+
 		/*
 		 * Log check errors are always fatal on v5; or whenever bad
 		 * metadata leads to a crash.
 		 */
Quoted 8 lines

fs/xfs/xfs_log.c xfs_log_mount( · line 696

Quoted 5 lines
fs/xfs/xfs_log.c +1 -1
-		if (fatal) {
+		if (xfs_has_crc(mp)) {
 			xfs_crit(mp, "AAIEEE! Log failed size checks. Abort!");
 			ASSERT(0);
 			goto out_free_log;

yes, only here in goto out_free_log we will return "error". Then why not shift error = -EINVAL in this if block?

Because I tried to change as little as needed to fix the issue. I didn't think it was necessary to move it from where it was....

Fixed.

-Dave.

Signature · 2 lines
Dave Chinner
[email protected]