Skip to contentSkip to conversation

xfs: don't assume perags are initialised when trimming AGs

PATCH Applied Dave Chinner 4 participants 7 messages Apr 30, 2025 – May 9, 2025 XFS filesystem

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 xfs: don't assume perags are initialised when trimming AGs

From: Dave Chinner <[email protected]>

When running fstrim immediately after mounting a V4 filesystem, the fstrim fails to trim all the free space in the filesystem. It only trims the first extent in the by-size free space tree in each AG and then returns. If a second fstrim is then run, it runs correctly and the entire free space in the filesystem is iterated and discarded correctly.

The problem lies in the setup of the trim cursor - it assumes that pag->pagf_longest is valid without either reading the AGF first or checking if xfs_perag_initialised_agf(pag) is true or not.

As a result, when a filesystem is mounted without reading the AGF (e.g. a clean mount on a v4 filesystem) and the first operation is a fstrim call, pag->pagf_longest is zero and so the free extent search starts at the wrong end of the by-size btree and exits after discarding the first record in the tree.

Fix this by deferring the initialisation of tcur->count to after we have locked the AGF and guaranteed that the perag is properly initialised. We trigger this on tcur->count == 0 after locking the AGF, as this will only occur on the first call to xfs_trim_gather_extents() for each AG. If we need to iterate, tcur->count will be set to the length of the record we need to restart at, so we can use this to ensure we only sample a valid pag->pagf_longest value for the iteration.

---
 fs/xfs/xfs_discard.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)
fs/xfs/xfs_discard.c +16 -1
index c1a306268ae4..94d0873bcd62 100644
@@ -167,6 +167,14 @@ xfs_discard_extents(
 	return error;
 }
 
+/*
+ * Care must be taken setting up the trim cursor as the perags may not have been
+ * initialised when the cursor is initialised. e.g. a clean mount which hasn't
+ * read in AGFs and the first operation run on the mounted fs is a trim. This
+ * can result in perag fields that aren't initialised until
+ * xfs_trim_gather_extents() calls xfs_alloc_read_agf() to lock down the AG for
+ * the free space search.
+ */
 struct xfs_trim_cur {
 	xfs_agblock_t	start;
 	xfs_extlen_t	count;
@@ -204,6 +212,14 @@ xfs_trim_gather_extents(
 	if (error)
 		goto out_trans_cancel;
 
+	/*
+	 * First time through tcur->count will not have been initialised as
+	 * pag->pagf_longest is not guaranteed to be valid before we read
+	 * the AGF buffer above.
+	 */
+	if (!tcur->count)
+		tcur->count = pag->pagf_longest;
+
 	if (tcur->by_bno) {
 		/* sub-AG discard request always starts at tcur->start */
 		cur = xfs_bnobt_init_cursor(mp, tp, agbp, pag);
@@ -350,7 +366,6 @@ xfs_trim_perag_extents(
 {
 	struct xfs_trim_cur	tcur = {
 		.start		= start,
-		.count		= pag->pagf_longest,
 		.end		= end,
 		.minlen		= minlen,
 	};
Signature · 1 line
2.45.2
Reviewed-by on v1 1/1
#1 lore raw
Quoted 28 lines · Dave Chinner

From: Dave Chinner <[email protected]>

When running fstrim immediately after mounting a V4 filesystem, the fstrim fails to trim all the free space in the filesystem. It only trims the first extent in the by-size free space tree in each AG and then returns. If a second fstrim is then run, it runs correctly and the entire free space in the filesystem is iterated and discarded correctly.

The problem lies in the setup of the trim cursor - it assumes that pag->pagf_longest is valid without either reading the AGF first or checking if xfs_perag_initialised_agf(pag) is true or not.

As a result, when a filesystem is mounted without reading the AGF (e.g. a clean mount on a v4 filesystem) and the first operation is a fstrim call, pag->pagf_longest is zero and so the free extent search starts at the wrong end of the by-size btree and exits after discarding the first record in the tree.

Fix this by deferring the initialisation of tcur->count to after we have locked the AGF and guaranteed that the perag is properly initialised. We trigger this on tcur->count == 0 after locking the AGF, as this will only occur on the first call to xfs_trim_gather_extents() for each AG. If we need to iterate, tcur->count will be set to the length of the record we need to restart at, so we can use this to ensure we only sample a valid pag->pagf_longest value for the iteration.

Quoted 1 line

Makes sense to me. Please add the following trailers on merge:

Cc <[email protected]> # v6.10
Fixes b0ffe661fab4b9 ("xfs: fix performance problems when fstrimming a subset of a fragmented AG")

--D

Quoted 50 lines
---
 fs/xfs/xfs_discard.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)
fs/xfs/xfs_discard.c +16 -1
index c1a306268ae4..94d0873bcd62 100644
@@ -167,6 +167,14 @@ xfs_discard_extents(
 	return error;
 }
 
+/*
+ * Care must be taken setting up the trim cursor as the perags may not have been
+ * initialised when the cursor is initialised. e.g. a clean mount which hasn't
+ * read in AGFs and the first operation run on the mounted fs is a trim. This
+ * can result in perag fields that aren't initialised until
+ * xfs_trim_gather_extents() calls xfs_alloc_read_agf() to lock down the AG for
+ * the free space search.
+ */
 struct xfs_trim_cur {
 	xfs_agblock_t	start;
 	xfs_extlen_t	count;
@@ -204,6 +212,14 @@ xfs_trim_gather_extents(
 	if (error)
 		goto out_trans_cancel;
 
+	/*
+	 * First time through tcur->count will not have been initialised as
+	 * pag->pagf_longest is not guaranteed to be valid before we read
+	 * the AGF buffer above.
+	 */
+	if (!tcur->count)
+		tcur->count = pag->pagf_longest;
+
 	if (tcur->by_bno) {
 		/* sub-AG discard request always starts at tcur->start */
 		cur = xfs_bnobt_init_cursor(mp, tp, agbp, pag);
@@ -350,7 +366,6 @@ xfs_trim_perag_extents(
 {
 	struct xfs_trim_cur	tcur = {
 		.start		= start,
-		.count		= pag->pagf_longest,
 		.end		= end,
 		.minlen		= minlen,
 	};
Signature · 1 line
2.45.2
#2 lore raw
Quoted 29 lines · Darrick J. Wong
Quoted 28 lines · Dave Chinner

From: Dave Chinner <[email protected]>

When running fstrim immediately after mounting a V4 filesystem, the fstrim fails to trim all the free space in the filesystem. It only trims the first extent in the by-size free space tree in each AG and then returns. If a second fstrim is then run, it runs correctly and the entire free space in the filesystem is iterated and discarded correctly.

The problem lies in the setup of the trim cursor - it assumes that pag->pagf_longest is valid without either reading the AGF first or checking if xfs_perag_initialised_agf(pag) is true or not.

As a result, when a filesystem is mounted without reading the AGF (e.g. a clean mount on a v4 filesystem) and the first operation is a fstrim call, pag->pagf_longest is zero and so the free extent search starts at the wrong end of the by-size btree and exits after discarding the first record in the tree.

Fix this by deferring the initialisation of tcur->count to after we have locked the AGF and guaranteed that the perag is properly initialised. We trigger this on tcur->count == 0 after locking the AGF, as this will only occur on the first call to xfs_trim_gather_extents() for each AG. If we need to iterate, tcur->count will be set to the length of the record we need to restart at, so we can use this to ensure we only sample a valid pag->pagf_longest value for the iteration.

Quoted 6 lines
Quoted 1 line

Makes sense to me. Please add the following trailers on merge:

Cc <[email protected]> # v6.10
Fixes b0ffe661fab4b9 ("xfs: fix performance problems when fstrimming a subset of a fragmented AG")

Those tags are incorrect. The regression was introduced by commit 89cfa899608f ("xfs: reduce AGF hold times during fstrim operations") a few releases before that change....

Quoted 1 line

Thanks.

-Dave.

Signature · 2 lines
Dave Chinner
[email protected]
maintainer
#3 lore raw
Quoted 32 lines · Dave Chinner
Quoted 31 lines · Darrick J. Wong
Quoted 29 lines · Dave Chinner

From: Dave Chinner <[email protected]>

When running fstrim immediately after mounting a V4 filesystem, the fstrim fails to trim all the free space in the filesystem. It only trims the first extent in the by-size free space tree in each AG and then returns. If a second fstrim is then run, it runs correctly and the entire free space in the filesystem is iterated and discarded correctly.

The problem lies in the setup of the trim cursor - it assumes that pag->pagf_longest is valid without either reading the AGF first or checking if xfs_perag_initialised_agf(pag) is true or not.

As a result, when a filesystem is mounted without reading the AGF (e.g. a clean mount on a v4 filesystem) and the first operation is a fstrim call, pag->pagf_longest is zero and so the free extent search starts at the wrong end of the by-size btree and exits after discarding the first record in the tree.

Fix this by deferring the initialisation of tcur->count to after we have locked the AGF and guaranteed that the perag is properly initialised. We trigger this on tcur->count == 0 after locking the AGF, as this will only occur on the first call to xfs_trim_gather_extents() for each AG. If we need to iterate, tcur->count will be set to the length of the record we need to restart at, so we can use this to ensure we only sample a valid pag->pagf_longest value for the iteration.

Quoted 8 lines
Quoted 4 lines

Makes sense to me. Please add the following trailers on merge:

Cc <[email protected]> # v6.10
Fixes b0ffe661fab4b9 ("xfs: fix performance problems when fstrimming a subset of a fragmented AG")

Those tags are incorrect. The regression was introduced by commit 89cfa899608f ("xfs: reduce AGF hold times during fstrim operations") a few releases before that change....

This sounds right, introduced in v6.6.

Darrick, I'll add a stable #v6.6 tag then.

Quoted 10 lines
Quoted 2 lines

Thanks.

-Dave.
--
Dave Chinner
[email protected]

#4 lore raw
Quoted 38 lines · Carlos Maiolino
Quoted 37 lines · Dave Chinner
Quoted 35 lines · Darrick J. Wong
Quoted 28 lines · Dave Chinner

From: Dave Chinner <[email protected]>

When running fstrim immediately after mounting a V4 filesystem, the fstrim fails to trim all the free space in the filesystem. It only trims the first extent in the by-size free space tree in each AG and then returns. If a second fstrim is then run, it runs correctly and the entire free space in the filesystem is iterated and discarded correctly.

The problem lies in the setup of the trim cursor - it assumes that pag->pagf_longest is valid without either reading the AGF first or checking if xfs_perag_initialised_agf(pag) is true or not.

As a result, when a filesystem is mounted without reading the AGF (e.g. a clean mount on a v4 filesystem) and the first operation is a fstrim call, pag->pagf_longest is zero and so the free extent search starts at the wrong end of the by-size btree and exits after discarding the first record in the tree.

Fix this by deferring the initialisation of tcur->count to after we have locked the AGF and guaranteed that the perag is properly initialised. We trigger this on tcur->count == 0 after locking the AGF, as this will only occur on the first call to xfs_trim_gather_extents() for each AG. If we need to iterate, tcur->count will be set to the length of the record we need to restart at, so we can use this to ensure we only sample a valid pag->pagf_longest value for the iteration.

Quoted 1 line

Makes sense to me. Please add the following trailers on merge:

Cc <[email protected]> # v6.10
Fixes b0ffe661fab4b9 ("xfs: fix performance problems when fstrimming a subset of a fragmented AG")
Quoted 7 lines
Quoted 3 lines

Those tags are incorrect. The regression was introduced by commit 89cfa899608f ("xfs: reduce AGF hold times during fstrim operations") a few releases before that change....

This sounds right, introduced in v6.6.

Darrick, I'll add a stable #v6.6 tag then.

Ok thanks.

--D

Quoted 11 lines
Quoted 10 lines
Quoted 2 lines

Thanks.

-Dave.
--
Dave Chinner
[email protected]

Reviewed-by on v1 1/1
#5 lore raw
In reply to Dave Chinner
Quoted 28 lines · Dave Chinner

From: Dave Chinner <[email protected]>

When running fstrim immediately after mounting a V4 filesystem, the fstrim fails to trim all the free space in the filesystem. It only trims the first extent in the by-size free space tree in each AG and then returns. If a second fstrim is then run, it runs correctly and the entire free space in the filesystem is iterated and discarded correctly.

The problem lies in the setup of the trim cursor - it assumes that pag->pagf_longest is valid without either reading the AGF first or checking if xfs_perag_initialised_agf(pag) is true or not.

As a result, when a filesystem is mounted without reading the AGF (e.g. a clean mount on a v4 filesystem) and the first operation is a fstrim call, pag->pagf_longest is zero and so the free extent search starts at the wrong end of the by-size btree and exits after discarding the first record in the tree.

Fix this by deferring the initialisation of tcur->count to after we have locked the AGF and guaranteed that the perag is properly initialised. We trigger this on tcur->count == 0 after locking the AGF, as this will only occur on the first call to xfs_trim_gather_extents() for each AG. If we need to iterate, tcur->count will be set to the length of the record we need to restart at, so we can use this to ensure we only sample a valid pag->pagf_longest value for the iteration.

Quoted 2 lines

---

Needs a "fixes" note in the description. Otherwise, lgtm.

Quoted 49 lines
 fs/xfs/xfs_discard.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)
fs/xfs/xfs_discard.c +16 -1
index c1a306268ae4..94d0873bcd62 100644
@@ -167,6 +167,14 @@ xfs_discard_extents(
 	return error;
 }
 
+/*
+ * Care must be taken setting up the trim cursor as the perags may not have been
+ * initialised when the cursor is initialised. e.g. a clean mount which hasn't
+ * read in AGFs and the first operation run on the mounted fs is a trim. This
+ * can result in perag fields that aren't initialised until
+ * xfs_trim_gather_extents() calls xfs_alloc_read_agf() to lock down the AG for
+ * the free space search.
+ */
 struct xfs_trim_cur {
 	xfs_agblock_t	start;
 	xfs_extlen_t	count;
@@ -204,6 +212,14 @@ xfs_trim_gather_extents(
 	if (error)
 		goto out_trans_cancel;
 
+	/*
+	 * First time through tcur->count will not have been initialised as
+	 * pag->pagf_longest is not guaranteed to be valid before we read
+	 * the AGF buffer above.
+	 */
+	if (!tcur->count)
+		tcur->count = pag->pagf_longest;
+
 	if (tcur->by_bno) {
 		/* sub-AG discard request always starts at tcur->start */
 		cur = xfs_bnobt_init_cursor(mp, tp, agbp, pag);
@@ -350,7 +366,6 @@ xfs_trim_perag_extents(
 {
 	struct xfs_trim_cur	tcur = {
 		.start		= start,
-		.count		= pag->pagf_longest,
 		.end		= end,
 		.minlen		= minlen,
 	};
Signature · 1 line
2.45.2
maintainer Applied
#6 lore raw
In reply to Dave Chinner
Quoted 8 lines · Dave Chinner

When running fstrim immediately after mounting a V4 filesystem, the fstrim fails to trim all the free space in the filesystem. It only trims the first extent in the by-size free space tree in each AG and then returns. If a second fstrim is then run, it runs correctly and the entire free space in the filesystem is iterated and discarded correctly.

[...]

Applied to for-next, thanks!

[1/1] xfs: don't assume perags are initialised when trimming AGs

      commit: 23be716b1c4f3f3a6c00ee38d51a57ef7db9ef7d

Best regards,

Signature · 1 line
Carlos Maiolino <[email protected]>