Skip to contentSkip to conversation

zoned growfs fix v2

PATCH In review Christoph Hellwig 4 participants 8 messages Jun 10

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hi all,

this series fixes growfs on zoned file systems, which so far has been completely broken. A test to verify it properly will be sent out as well.

Changes since v1:
- rename xfs_add_free_zone to xfs_zone_mark_free
- use xfs_rtgs_to_rfsbs to avoid overflows
- improve a commit message
- add a comment explaining the zoned growfs space accounting logic

Diffstat:

 xfs_rtalloc.c    |   16 ++++++++++++----
 xfs_zone_alloc.c |   11 +++++++++--
 xfs_zone_alloc.h |    1 +
 xfs_zone_gc.c    |    4 +---
 4 files changed, 23 insertions(+), 9 deletions(-)
#1 lore raw

PATCH 1/3 xfs: factor out a xfs_zone_mark_free helper

Add a helper for adding a zone to the free pool in preparation of adding another caller.

---
 fs/xfs/xfs_zone_alloc.c | 11 +++++++++--
 fs/xfs/xfs_zone_alloc.h |  1 +
 fs/xfs/xfs_zone_gc.c    |  4 +---
 3 files changed, 11 insertions(+), 5 deletions(-)
fs/xfs/xfs_zone_alloc.c +9 -2
index 5e297b75a85f..08d8b34f467e 100644
@@ -944,6 +944,14 @@ xfs_zone_rgbno_is_valid(
 			rtg_rgno(rtg), XFS_RTG_FREE);
 }
 
+void
+xfs_zone_mark_free(
+	struct xfs_rtgroup	*rtg)
+{
+	xfs_group_set_mark(rtg_group(rtg), XFS_RTG_FREE);
+	atomic_inc(&rtg_mount(rtg)->m_zone_info->zi_nr_free_zones);
+}
+
 static void
 xfs_free_open_zones(
 	struct xfs_zone_info	*zi)
@@ -1082,8 +1090,7 @@ xfs_init_zone(
 
 	if (write_pointer == 0) {
 		/* zone is empty */
-		atomic_inc(&zi->zi_nr_free_zones);
-		xfs_group_set_mark(rtg_group(rtg), XFS_RTG_FREE);
+		xfs_zone_mark_free(rtg);
 		iz->available += rtg_blocks(rtg);
 	} else if (write_pointer < rtg_blocks(rtg)) {
 		/* zone is open */
fs/xfs/xfs_zone_alloc.h +1 -0
index 8b2ef98c81ef..abf8358bb266 100644
@@ -42,6 +42,7 @@ void xfs_zoned_wake_all(struct xfs_mount *mp);
 bool xfs_zone_rgbno_is_valid(struct xfs_rtgroup *rtg, xfs_rgnumber_t rgbno);
 void xfs_mark_rtg_boundary(struct iomap_ioend *ioend);
 
+void xfs_zone_mark_free(struct xfs_rtgroup *rtg);
 uint64_t xfs_zoned_default_resblks(struct xfs_mount *mp,
 		enum xfs_free_counter ctr);
 void xfs_zoned_show_stats(struct seq_file *m, struct xfs_mount *mp);
fs/xfs/xfs_zone_gc.c +1 -3
index f03211e4354a..f76a09130852 100644
@@ -924,9 +924,7 @@ xfs_zone_gc_finish_reset(
 		goto out;
 	}
 
-	xfs_group_set_mark(rtg_group(rtg), XFS_RTG_FREE);
-	atomic_inc(&zi->zi_nr_free_zones);
-
+	xfs_zone_mark_free(rtg);
 	xfs_zoned_add_available(mp, rtg_blocks(rtg));
 
 	wake_up_all(&zi->zi_zone_wait);
Signature · 1 line
2.53.0
#2 lore raw

PATCH 2/3 xfs: add newly added RTGs to the free pool in growfs

In reply to Christoph Hellwig

When growing a zoned RT section, the newly added RTGs also need to be tagged as free in the radix tree and add to the nr_free_zones counters. Call xfs_add_free_zone to do that, otherwise using up the newly added space will wait for free zones forever.

Fixes 01b71e64bb87 ("xfs: support growfs on zoned file systems")
Cc <[email protected]> # v6.15
---
 fs/xfs/xfs_rtalloc.c | 8 ++++++++
 1 file changed, 8 insertions(+)
fs/xfs/xfs_rtalloc.c +8 -0
index 153f3c378f9f..debbcefdf07f 100644
@@ -933,6 +933,14 @@ xfs_growfs_rt_zoned(
 	mp->m_features |= XFS_FEAT_REALTIME;
 	xfs_rtrmapbt_compute_maxlevels(mp);
 	xfs_rtrefcountbt_compute_maxlevels(mp);
+
+	/*
+	 * Finally add the newly added zone to the freelist and add the space
+	 * to the available counter.  The order is important here: only add
+	 * the available space after the zones, as available space guarantees
+	 * that zones to back it are available.
+	 */
+	xfs_zone_mark_free(rtg);
 	xfs_zoned_add_available(mp, freed_rtx);
 out_free:
 	kfree(nmp);
Signature · 1 line
2.53.0
#3 lore raw

PATCH 3/3 xfs: only log freed extents for the current RTG in zoned growfs

In reply to Christoph Hellwig

Otherwise a power fail or crash during growfs could lead to an elevated sb_rblocks counter.

Note that the step function is much simpler compared to the classic RT allocator as zoned RT sections must be aligned to real time group boundaries.

Fixes 01b71e64bb87 ("xfs: support growfs on zoned file systems")
---
 fs/xfs/xfs_rtalloc.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
fs/xfs/xfs_rtalloc.c +4 -4
index debbcefdf07f..7a3f97686989 100644
@@ -890,8 +890,7 @@ xfs_growfs_rt_sb_fields(
 
 static int
 xfs_growfs_rt_zoned(
-	struct xfs_rtgroup	*rtg,
-	xfs_rfsblock_t		nrblocks)
+	struct xfs_rtgroup	*rtg)
 {
 	struct xfs_mount	*mp = rtg_mount(rtg);
 	struct xfs_mount	*nmp;
@@ -903,7 +902,8 @@ xfs_growfs_rt_zoned(
 	 * Calculate new sb and mount fields for this round.  Also ensure the
 	 * rtg_extents value is uptodate as the rtbitmap code relies on it.
 	 */
-	nmp = xfs_growfs_rt_alloc_fake_mount(mp, nrblocks,
+	nmp = xfs_growfs_rt_alloc_fake_mount(mp,
+			xfs_rtgs_to_rfsbs(mp, rtg_rgno(rtg) + 1),
 			mp->m_sb.sb_rextsize);
 	if (!nmp)
 		return -ENOMEM;
@@ -1226,7 +1226,7 @@ xfs_growfs_rtg(
 	}
 
 	if (xfs_has_zoned(mp)) {
-		error = xfs_growfs_rt_zoned(rtg, nrblocks);
+		error = xfs_growfs_rt_zoned(rtg);
 		goto out_rele;
 	}
 
Signature · 1 line
2.53.0
Reviewed-by on v1 3/3
#4 lore raw

Re: [PATCH 3/3] xfs: only log freed extents for the current RTG in zoned growfs

Quoted 28 lines · Christoph Hellwig

Otherwise a power fail or crash during growfs could lead to an elevated sb_rblocks counter.

Note that the step function is much simpler compared to the classic RT allocator as zoned RT sections must be aligned to real time group boundaries.

Fixes 01b71e64bb87 ("xfs: support growfs on zoned file systems")
---
 fs/xfs/xfs_rtalloc.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
fs/xfs/xfs_rtalloc.c +1 -2
index debbcefdf07f..7a3f97686989 100644
@@ -890,8 +890,7 @@ xfs_growfs_rt_sb_fields(
 
 static int
 xfs_growfs_rt_zoned(
-	struct xfs_rtgroup	*rtg,
-	xfs_rfsblock_t		nrblocks)
+	struct xfs_rtgroup	*rtg)
 {
 	struct xfs_mount	*mp = rtg_mount(rtg);
 	struct xfs_mount	*nmp;

fs/xfs/xfs_rtalloc.c xfs_growfs_rt_zoned( · line 906

Quoted 6 lines
fs/xfs/xfs_rtalloc.c +2 -1
@@ -903,7 +902,8 @@ xfs_growfs_rt_zoned(
 	 * Calculate new sb and mount fields for this round.  Also ensure the
 	 * rtg_extents value is uptodate as the rtbitmap code relies on it.
 	 */
-	nmp = xfs_growfs_rt_alloc_fake_mount(mp, nrblocks,
+	nmp = xfs_growfs_rt_alloc_fake_mount(mp,
+			xfs_rtgs_to_rfsbs(mp, rtg_rgno(rtg) + 1),

Heh, I forgot we had a helper for this!

--D

Quoted 16 lines
fs/xfs/xfs_rtalloc.c +1 -1
 			mp->m_sb.sb_rextsize);
 	if (!nmp)
 		return -ENOMEM;
@@ -1226,7 +1226,7 @@ xfs_growfs_rtg(
 	}
 
 	if (xfs_has_zoned(mp)) {
-		error = xfs_growfs_rt_zoned(rtg, nrblocks);
+		error = xfs_growfs_rt_zoned(rtg);
 		goto out_rele;
 	}
 
Signature · 1 line
2.53.0
#5 lore raw

Re: [PATCH 3/3] xfs: only log freed extents for the current RTG in zoned growfs

Quoted 5 lines · Darrick J. Wong

fs/xfs/xfs_rtalloc.c xfs_growfs_rt_zoned( · line 906

Quoted 3 lines
fs/xfs/xfs_rtalloc.c +2 -1
-	nmp = xfs_growfs_rt_alloc_fake_mount(mp, nrblocks,
+	nmp = xfs_growfs_rt_alloc_fake_mount(mp,
+			xfs_rtgs_to_rfsbs(mp, rtg_rgno(rtg) + 1),

Heh, I forgot we had a helper for this!

Yeah, if only the person who added it would remember to consistently use it..

Reviewed-by on v1 3/3
#6 lore raw

Re: [PATCH 3/3] xfs: only log freed extents for the current RTG in zoned growfs

In reply to Christoph Hellwig
Quoted 9 lines · Christoph Hellwig

Otherwise a power fail or crash during growfs could lead to an elevated sb_rblocks counter.

Note that the step function is much simpler compared to the classic RT allocator as zoned RT sections must be aligned to real time group boundaries.

Fixes 01b71e64bb87 ("xfs: support growfs on zoned file systems")
Signature · 2 lines
Damien Le Moal
Western Digital Research
maintainer Applied
#7 lore raw
In reply to Christoph Hellwig
Quoted 11 lines · Christoph Hellwig

this series fixes growfs on zoned file systems, which so far has been completely broken. A test to verify it properly will be sent out as well.

Changes since v1:
- rename xfs_add_free_zone to xfs_zone_mark_free
- use xfs_rtgs_to_rfsbs to avoid overflows
- improve a commit message
- add a comment explaining the zoned growfs space accounting logic

[...]

Applied to for-next, thanks!

[1/3] xfs: factor out a xfs_zone_mark_free helper

      commit: 16fc9d358d12e6e9cc553162b4064ad58e7fa869
[2/3] xfs: add newly added RTGs to the free pool in growfs
      commit: ae3692c7f440c1ff577aae1a51202415ec4a794b
[3/3] xfs: only log freed extents for the current RTG in zoned growfs
      commit: 44cccefe65749821d9a13523c8b763bf1262ef73

Best regards,

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