Skip to contentSkip to conversation

mkfs: validate start and end of aligned logs

RFC PATCH In review Darrick J. Wong 2 participants 3 messages May 3, 2019

lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 mkfs: validate start and end of aligned logs

From: Darrick J. Wong <[email protected]>

Validate that the start and end of the log stay within a single AG if we adjust either end to align to stripe units.

---
 mkfs/xfs_mkfs.c |   11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)
mkfs/xfs_mkfs.c +10 -1
index 3ca8c9dc..0862621a 100644
@@ -3070,11 +3070,20 @@ align_internal_log(
 	if ((cfg->logstart % sunit) != 0)
 		cfg->logstart = ((cfg->logstart + (sunit - 1)) / sunit) * sunit;
 
+	/* if our log start rounds into the next AG we're done */
+	if (!xfs_verify_fsbno(mp, cfg->logstart)) {
+			fprintf(stderr,
+_("Due to stripe alignment, the internal log start (%lld) cannot be aligned\n"
+  "within an allocation group.\n"),
+			(long long) cfg->logstart);
+		usage();
+	}
+
 	/* round up/down the log size now */
 	align_log_size(cfg, sunit);
 
 	/* check the aligned log still fits in an AG. */
-	if (cfg->logblocks > cfg->agsize - XFS_FSB_TO_AGBNO(mp, cfg->logstart)) {
+	if (!xfs_verify_fsbno(mp, cfg->logstart + cfg->logblocks - 1)) {
 		fprintf(stderr,
 _("Due to stripe alignment, the internal log size (%lld) is too large.\n"
   "Must fit within an allocation group.\n"),
#1 lore raw
Quoted 23 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

Validate that the start and end of the log stay within a single AG if we adjust either end to align to stripe units.

---
 mkfs/xfs_mkfs.c |   11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)
mkfs/xfs_mkfs.c +5 -0
index 3ca8c9dc..0862621a 100644
@@ -3070,11 +3070,20 @@ align_internal_log(
 	if ((cfg->logstart % sunit) != 0)
 		cfg->logstart = ((cfg->logstart + (sunit - 1)) / sunit) * sunit;
 
+	/* if our log start rounds into the next AG we're done */
+	if (!xfs_verify_fsbno(mp, cfg->logstart)) {
+			fprintf(stderr,
+_("Due to stripe alignment, the internal log start (%lld) cannot be aligned\n"
+  "within an allocation group.\n"),

mkfs/xfs_mkfs.c align_internal_log( · line 3086

Quoted 10 lines
mkfs/xfs_mkfs.c +5 -1
+			(long long) cfg->logstart);
+		usage();
+	}
+
 	/* round up/down the log size now */
 	align_log_size(cfg, sunit);
 
 	/* check the aligned log still fits in an AG. */
-	if (cfg->logblocks > cfg->agsize - XFS_FSB_TO_AGBNO(mp, cfg->logstart)) {
+	if (!xfs_verify_fsbno(mp, cfg->logstart + cfg->logblocks - 1)) {

This used to see if the aligned log size was actually smaller than the AG.

Your new check just makes sure that the end block doesn't land on metadata, right?

i.e. we could end up with:

[ AG 0 ][ AG 1 ]
[ log ]

and pass your new test, because the end of the log doesn't stomp on ag metadata, even though it goes past the end of the start AG... right?

-Eric

Quoted 4 lines
mkfs/xfs_mkfs.c +0 -0
 		fprintf(stderr,
 _("Due to stripe alignment, the internal log size (%lld) is too large.\n"
   "Must fit within an allocation group.\n"),
#2 lore raw
Quoted 42 lines · Eric Sandeen
Quoted 23 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

Validate that the start and end of the log stay within a single AG if we adjust either end to align to stripe units.

---
 mkfs/xfs_mkfs.c |   11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)
mkfs/xfs_mkfs.c +5 -0
index 3ca8c9dc..0862621a 100644
@@ -3070,11 +3070,20 @@ align_internal_log(
 	if ((cfg->logstart % sunit) != 0)
 		cfg->logstart = ((cfg->logstart + (sunit - 1)) / sunit) * sunit;
 
+	/* if our log start rounds into the next AG we're done */
+	if (!xfs_verify_fsbno(mp, cfg->logstart)) {
+			fprintf(stderr,
+_("Due to stripe alignment, the internal log start (%lld) cannot be aligned\n"
+  "within an allocation group.\n"),

mkfs/xfs_mkfs.c align_internal_log( · line 3086

Quoted 10 lines
mkfs/xfs_mkfs.c +5 -1
+			(long long) cfg->logstart);
+		usage();
+	}
+
 	/* round up/down the log size now */
 	align_log_size(cfg, sunit);
 
 	/* check the aligned log still fits in an AG. */
-	if (cfg->logblocks > cfg->agsize - XFS_FSB_TO_AGBNO(mp, cfg->logstart)) {
+	if (!xfs_verify_fsbno(mp, cfg->logstart + cfg->logblocks - 1)) {

This used to see if the aligned log size was actually smaller than the AG.

Your new check just makes sure that the end block doesn't land on metadata, right?

Quoted 7 lines

i.e. we could end up with:

[ AG 0 ][ AG 1 ]
[ log ]

and pass your new test, because the end of the log doesn't stomp on ag metadata, even though it goes past the end of the start AG... right?

DOH. Yes. Somewhere in there I coded up a FSB_TO_AGNO(logstart) == FSB_TO_AGNO(logstart + logblocks - 1) but clearly it fell out.

Derp derp try again tomorrow. :(

--D

Quoted 6 lines

-Eric

Quoted 4 lines
mkfs/xfs_mkfs.c +0 -0
 		fprintf(stderr,
 _("Due to stripe alignment, the internal log size (%lld) is too large.\n"
   "Must fit within an allocation group.\n"),