Fix false warning in inode_to_wb

PATCH In review Andreas Gruenbacher 3 participants 8 messages Apr 12, 2025 – Apr 14, 2025

Subsystems (4)

Memory management VFS cgroup GFS2 file system

lore

Share & tools

lore

Andreas Gruenbacher

author maintainer

2025-04-12 16:39 UTC #0 lore raw

v3 cover Fix false warning in inode_to_wb

Changes in v3:

- In gfs2_fill_super(), initialize sdp->sd_inode after initializing ->s_blocksize_bits so that ->i_blkbits will be set correctly.

- inode_to_wb(const struct inode *inode) no longer calls inode_cgwb_enabled(struct inode *inode), so its argument can remain const.

Changes in v2:

- In inode_to_wb(), only check if the SB_I_CGROUPWB flag is set so that the consistency checks remain active even when cgroup writeback isn't actively used.

- When new_inode() fails in gfs2_fill_super(), clean up correctly.

- Improve the description of "gfs2: replace sd_aspace with sd_inode".

Andreas Gruenbacher (1):
gfs2: replace sd_aspace with sd_inode

Jan Kara (1):
writeback: Fix false warning in inode_to_wb()

 fs/gfs2/glock.c             |  3 +--
 fs/gfs2/glops.c             |  4 ++--
 fs/gfs2/incore.h            |  9 ++++++++-
 fs/gfs2/meta_io.c           |  2 +-
 fs/gfs2/meta_io.h           |  4 +---
 fs/gfs2/ops_fstype.c        | 31 ++++++++++++++++++-------------
 fs/gfs2/super.c             |  2 +-
 include/linux/backing-dev.h |  1 +
 8 files changed, 33 insertions(+), 23 deletions(-)

Signature · 1 line

2.48.1

Andreas Gruenbacher

author maintainer

2025-04-12 16:39 UTC #1 lore raw

PATCH v3 1/2 gfs2: replace sd_aspace with sd_inode

Currently, sdp->sd_aspace and the per-inode metadata address spaces use sb->s_bdev->bd_mapping->host as their ->host; folios in those address spaces will thus appear to be on bdev rather than on gfs2 filesystems. This is a problem because gfs2 doesn't support cgroup writeback (SB_I_CGROUPWB), but bdev does.

Fix that by using a "dummy" gfs2 inode as ->host in those address spaces. When coming from a folio, folio->mapping->host->i_sb will then be a gfs2 super block and the SB_I_CGROUPWB flag will not be set in sb->s_iflags.

Based on a previous version from Bob Peterson from several years ago. Thanks to Tetsuo Handa, Jan Kara, and Rafael Aquini for helping figure this out.

Signed-off-by Andreas Gruenbacher <[email protected]>

---
 fs/gfs2/glock.c      |  3 +--
 fs/gfs2/glops.c      |  4 ++--
 fs/gfs2/incore.h     |  9 ++++++++-
 fs/gfs2/meta_io.c    |  2 +-
 fs/gfs2/meta_io.h    |  4 +---
 fs/gfs2/ops_fstype.c | 31 ++++++++++++++++++-------------
 fs/gfs2/super.c      |  2 +-
 7 files changed, 32 insertions(+), 23 deletions(-)

fs/gfs2/glock.c +1 -2

index d7220a6fe8f5..ba25b884169e 100644
@@ -1166,7 +1166,6 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 		   const struct gfs2_glock_operations *glops, int create,
 		   struct gfs2_glock **glp)
 {
-	struct super_block *s = sdp->sd_vfs;
 	struct lm_lockname name = { .ln_number = number,
 				    .ln_type = glops->go_type,
 				    .ln_sbd = sdp };

@@ -1229,7 +1228,7 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 	mapping = gfs2_glock2aspace(gl);
 	if (mapping) {
                 mapping->a_ops = &gfs2_meta_aops;
-		mapping->host = s->s_bdev->bd_mapping->host;
+		mapping->host = sdp->sd_inode;
 		mapping->flags = 0;
 		mapping_set_gfp_mask(mapping, GFP_NOFS);
 		mapping->i_private_data = NULL;

fs/gfs2/glops.c +2 -2

index eb4714f299ef..116efe335c32 100644
@@ -168,7 +168,7 @@ void gfs2_ail_flush(struct gfs2_glock *gl, bool fsync)
 static int gfs2_rgrp_metasync(struct gfs2_glock *gl)
 {
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
-	struct address_space *metamapping = &sdp->sd_aspace;
+	struct address_space *metamapping = gfs2_aspace(sdp);
 	struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl);
 	const unsigned bsize = sdp->sd_sb.sb_bsize;
 	loff_t start = (rgd->rd_addr * bsize) & PAGE_MASK;

@@ -225,7 +225,7 @@ static int rgrp_go_sync(struct gfs2_glock *gl)
 static void rgrp_go_inval(struct gfs2_glock *gl, int flags)
 {
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
-	struct address_space *mapping = &sdp->sd_aspace;
+	struct address_space *mapping = gfs2_aspace(sdp);
 	struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl);
 	const unsigned bsize = sdp->sd_sb.sb_bsize;
 	loff_t start, end;

fs/gfs2/incore.h +8 -1

index 74abbd4970f8..0a41c4e76b32 100644
@@ -795,7 +795,7 @@ struct gfs2_sbd {
 
 	/* Log stuff */
 
-	struct address_space sd_aspace;
+	struct inode *sd_inode;
 
 	spinlock_t sd_log_lock;

@@ -851,6 +851,13 @@ struct gfs2_sbd {
 	unsigned long sd_glock_dqs_held;
 };
 
+#define GFS2_BAD_INO 1
+
+static inline struct address_space *gfs2_aspace(struct gfs2_sbd *sdp)
+{
+	return sdp->sd_inode->i_mapping;
+}
+
 static inline void gfs2_glstats_inc(struct gfs2_glock *gl, int which)
 {
 	gl->gl_stats.stats[which]++;

fs/gfs2/meta_io.c +1 -1

index 198cc7056637..9dc8885c95d0 100644
@@ -132,7 +132,7 @@ struct buffer_head *gfs2_getbuf(struct gfs2_glock *gl, u64 blkno, int create)
 	unsigned int bufnum;
 
 	if (mapping == NULL)
-		mapping = &sdp->sd_aspace;
+		mapping = gfs2_aspace(sdp);
 
 	shift = PAGE_SHIFT - sdp->sd_sb.sb_bsize_shift;
 	index = blkno >> shift;             /* convert block to page */

fs/gfs2/meta_io.h +1 -3

index 831d988c2ceb..b7c8a6684d02 100644
@@ -44,9 +44,7 @@ static inline struct gfs2_sbd *gfs2_mapping2sbd(struct address_space *mapping)
 		struct gfs2_glock_aspace *gla =
 			container_of(mapping, struct gfs2_glock_aspace, mapping);
 		return gla->glock.gl_name.ln_sbd;
-	} else if (mapping->a_ops == &gfs2_rgrp_aops)
-		return container_of(mapping, struct gfs2_sbd, sd_aspace);
-	else
+	} else
 		return inode->i_sb->s_fs_info;
 }

fs/gfs2/ops_fstype.c +18 ~~-13~~

index e83d293c3614..6ce475e1c6d6 100644
@@ -72,7 +72,6 @@ void free_sbd(struct gfs2_sbd *sdp)
 static struct gfs2_sbd *init_sbd(struct super_block *sb)
 {
 	struct gfs2_sbd *sdp;
-	struct address_space *mapping;
 
 	sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
 	if (!sdp)

@@ -109,16 +108,6 @@ static struct gfs2_sbd *init_sbd(struct super_block *sb)
 
 	INIT_LIST_HEAD(&sdp->sd_sc_inodes_list);
 
-	mapping = &sdp->sd_aspace;
-
-	address_space_init_once(mapping);
-	mapping->a_ops = &gfs2_rgrp_aops;
-	mapping->host = sb->s_bdev->bd_mapping->host;
-	mapping->flags = 0;
-	mapping_set_gfp_mask(mapping, GFP_NOFS);
-	mapping->i_private_data = NULL;
-	mapping->writeback_index = 0;
-
 	spin_lock_init(&sdp->sd_log_lock);
 	atomic_set(&sdp->sd_log_pinned, 0);
 	INIT_LIST_HEAD(&sdp->sd_log_revokes);

@@ -1135,6 +1124,7 @@ static int gfs2_fill_super(struct super_block *sb, struct fs_context *fc)
 	int silent = fc->sb_flags & SB_SILENT;
 	struct gfs2_sbd *sdp;
 	struct gfs2_holder mount_gh;
+	struct address_space *mapping;
 	int error;
 
 	sdp = init_sbd(sb);

@@ -1156,6 +1146,7 @@ static int gfs2_fill_super(struct super_block *sb, struct fs_context *fc)
 	sb->s_flags |= SB_NOSEC;
 	sb->s_magic = GFS2_MAGIC;
 	sb->s_op = &gfs2_super_ops;
+
 	sb->s_d_op = &gfs2_dops;
 	sb->s_export_op = &gfs2_export_ops;
 	sb->s_qcop = &gfs2_quotactl_ops;

@@ -1181,9 +1172,21 @@ static int gfs2_fill_super(struct super_block *sb, struct fs_context *fc)
 		sdp->sd_tune.gt_statfs_quantum = 30;
 	}
 
+	/* Set up an address space for metadata writes */
+	sdp->sd_inode = new_inode(sb);
+	error = -ENOMEM;
+	if (!sdp->sd_inode)
+		goto fail_free;
+	sdp->sd_inode->i_ino = GFS2_BAD_INO;
+	sdp->sd_inode->i_size = OFFSET_MAX;
+
+	mapping = gfs2_aspace(sdp);
+	mapping->a_ops = &gfs2_rgrp_aops;
+	mapping_set_gfp_mask(mapping, GFP_NOFS);
+
 	error = init_names(sdp, silent);
 	if (error)
-		goto fail_free;
+		goto fail_iput;
 
 	snprintf(sdp->sd_fsname, sizeof(sdp->sd_fsname), "%s", sdp->sd_table_name);

@@ -1192,7 +1195,7 @@ static int gfs2_fill_super(struct super_block *sb, struct fs_context *fc)
 			WQ_MEM_RECLAIM | WQ_HIGHPRI | WQ_FREEZABLE, 0,
 			sdp->sd_fsname);
 	if (!sdp->sd_glock_wq)
-		goto fail_free;
+		goto fail_iput;
 
 	sdp->sd_delete_wq = alloc_workqueue("gfs2-delete/%s",
 			WQ_MEM_RECLAIM | WQ_FREEZABLE, 0, sdp->sd_fsname);

@@ -1309,6 +1312,8 @@ static int gfs2_fill_super(struct super_block *sb, struct fs_context *fc)
 fail_glock_wq:
 	if (sdp->sd_glock_wq)
 		destroy_workqueue(sdp->sd_glock_wq);
+fail_iput:
+	iput(sdp->sd_inode);
 fail_free:
 	free_sbd(sdp);
 	sb->s_fs_info = NULL;

fs/gfs2/super.c +1 -1

index 44e5658b896c..4529b7dda8ca 100644
@@ -648,7 +648,7 @@ static void gfs2_put_super(struct super_block *sb)
 	gfs2_jindex_free(sdp);
 	/*  Take apart glock structures and buffer lists  */
 	gfs2_gl_hash_clear(sdp);
-	truncate_inode_pages_final(&sdp->sd_aspace);
+	iput(sdp->sd_inode);
 	gfs2_delete_debugfs_file(sdp);
 
 	gfs2_sys_fs_del(sdp);

Signature · 1 line

2.48.1

Andreas Gruenbacher

author maintainer

2025-04-12 16:39 UTC #2 lore raw

PATCH v3 2/2 writeback: Fix false warning in inode_to_wb()

In reply to Andreas Gruenbacher

From: Jan Kara <[email protected]>

inode_to_wb() is used also for filesystems that don't support cgroup writeback. For these filesystems inode->i_wb is stable during the lifetime of the inode (it points to bdi->wb) and there's no need to hold locks protecting the inode->i_wb dereference. Improve the warning in inode_to_wb() to not trigger for these filesystems.

Signed-off-by Jan Kara <[email protected]>

Reviewed-by Andreas Gruenbacher <[email protected]>

---
 include/linux/backing-dev.h | 1 +
 1 file changed, 1 insertion(+)

include/linux/backing-dev.h +1 -0

index 8e7af9a03b41..e721148c95d0 100644
@@ -249,6 +249,7 @@ static inline struct bdi_writeback *inode_to_wb(const struct inode *inode)
 {
 #ifdef CONFIG_LOCKDEP
 	WARN_ON_ONCE(debug_locks &&
+		     (inode->i_sb->s_iflags & SB_I_CGROUPWB) &&
 		     (!lockdep_is_held(&inode->i_lock) &&
 		      !lockdep_is_held(&inode->i_mapping->i_pages.xa_lock) &&
 		      !lockdep_is_held(&inode->i_wb->list_lock)));

Signature · 1 line

2.48.1

Jan Kara

reviewer

2025-04-14 13:55 UTC #3 lore raw

Re: [PATCH v3 2/2] writeback: Fix false warning in inode_to_wb()

Hi Andrew!

Can you please take this patch through MM tree? Thanks! Andreas is already taking the first patch in the series through GFS2 tree.

Honza

Quoted 29 lines · Andreas Gruenbacher

From: Jan Kara <[email protected]>
inode_to_wb() is used also for filesystems that don't support cgroup writeback. For these filesystems inode->i_wb is stable during the lifetime of the inode (it points to bdi->wb) and there's no need to hold locks protecting the inode->i_wb dereference. Improve the warning in inode_to_wb() to not trigger for these filesystems.
Signed-off-by Jan Kara <[email protected]>
Reviewed-by Andreas Gruenbacher <[email protected]>
---
 include/linux/backing-dev.h | 1 +
 1 file changed, 1 insertion(+)
include/linux/backing-dev.h +1 -0
index 8e7af9a03b41..e721148c95d0 100644
@@ -249,6 +249,7 @@ static inline struct bdi_writeback *inode_to_wb(const struct inode *inode)
 {
 #ifdef CONFIG_LOCKDEP
 	WARN_ON_ONCE(debug_locks &&
+		     (inode->i_sb->s_iflags & SB_I_CGROUPWB) &&
 		     (!lockdep_is_held(&inode->i_lock) &&
 		      !lockdep_is_held(&inode->i_mapping->i_pages.xa_lock) &&
 		      !lockdep_is_held(&inode->i_wb->list_lock)));
Signature · 1 line
2.48.1

Signature · 2 lines

Jan Kara <[email protected]>
SUSE Labs, CR

Andrew Morton

maintainer

2025-04-14 21:47 UTC #4 lore raw

Re: [PATCH v3 2/2] writeback: Fix false warning in inode_to_wb()

In reply to Andreas Gruenbacher

Quoted 13 lines · On Sat, 12 Apr 2025 18:39:12 +0200 Andreas Gruenbacher <[email protected]>…

From: Jan Kara <[email protected]>
inode_to_wb() is used also for filesystems that don't support cgroup writeback. For these filesystems inode->i_wb is stable during the lifetime of the inode (it points to bdi->wb) and there's no need to hold locks protecting the inode->i_wb dereference. Improve the warning in inode_to_wb() to not trigger for these filesystems.
...

include/linux/backing-dev.h static inline struct bdi_writeback *inode_to_wb(const struct inode *inode) · line 255

Quoted 7 lines

include/linux/backing-dev.h +1 -0

@@ -249,6 +249,7 @@ static inline struct bdi_writeback *inode_to_wb(const struct inode *inode)
 {
 #ifdef CONFIG_LOCKDEP
 	WARN_ON_ONCE(debug_locks &&
+		     (inode->i_sb->s_iflags & SB_I_CGROUPWB) &&
 		     (!lockdep_is_held(&inode->i_lock) &&
 		      !lockdep_is_held(&inode->i_mapping->i_pages.xa_lock) &&
 		      !lockdep_is_held(&inode->i_wb->list_lock)));

Is this a does-nothing now GFS2 has been altered?

Otherwise, a bogus WARN is something we'll want to eliminate from -stable kernels also. Are we able to identify a Fixes: for this?

Thanks.

Andreas Gruenbacher

author maintainer

2025-04-14 23:18 UTC #5 lore raw

Re: [PATCH v3 2/2] writeback: Fix false warning in inode_to_wb()

Quoted 24 lines · Andrew Morton

Quoted 13 lines · On Sat, 12 Apr 2025 18:39:12 +0200 Andreas Gruenbacher <[email protected]>…
From: Jan Kara <[email protected]>
inode_to_wb() is used also for filesystems that don't support cgroup writeback. For these filesystems inode->i_wb is stable during the lifetime of the inode (it points to bdi->wb) and there's no need to hold locks protecting the inode->i_wb dereference. Improve the warning in inode_to_wb() to not trigger for these filesystems.
...
include/linux/backing-dev.h static inline struct bdi_writeback *inode_to_wb(const struct inode *inode) · line 255
Quoted 7 lines
include/linux/backing-dev.h +1 -0
@@ -249,6 +249,7 @@ static inline struct bdi_writeback *inode_to_wb(const struct inode *inode)
 {
 #ifdef CONFIG_LOCKDEP
      WARN_ON_ONCE(debug_locks &&
+                  (inode->i_sb->s_iflags & SB_I_CGROUPWB) &&
                   (!lockdep_is_held(&inode->i_lock) &&
                    !lockdep_is_held(&inode->i_mapping->i_pages.xa_lock) &&
                    !lockdep_is_held(&inode->i_wb->list_lock)));
Is this a does-nothing now GFS2 has been altered?

Quoted 2 lines

Otherwise, a bogus WARN is something we'll want to eliminate from -stable kernels also. Are we able to identify a Fixes: for this?

The excess warnings started with commit:

Fixes aaa2cacf8184 ("writeback: add lockdep annotation to inode_to_wb()")

Getting rid of them requires this change, together with "gfs2: replace sd_aspace with sd_inode" from gfs2 for-next:

https://web.git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git/commit/?h=

    a5fb828aba730d08aa6dec6bce3839f25e1f7a9d

Thanks,
Andreas

Andrew Morton

maintainer

2025-04-14 21:51 UTC #6 lore raw

Re: [PATCH v3 2/2] writeback: Fix false warning in inode_to_wb()

In reply to Andreas Gruenbacher

Quoted 8 lines · On Sat, 12 Apr 2025 18:39:12 +0200 Andreas Gruenbacher <[email protected]>…

inode_to_wb() is used also for filesystems that don't support cgroup writeback. For these filesystems inode->i_wb is stable during the lifetime of the inode (it points to bdi->wb) and there's no need to hold locks protecting the inode->i_wb dereference. Improve the warning in inode_to_wb() to not trigger for these filesystems.
Signed-off-by Jan Kara <[email protected]>
Reviewed-by Andreas Gruenbacher <[email protected]>

Yoo were on the patch delivery path so there should be a signed-off-by:Andreas somewhere. I made that change to the mm.git copy of this patch.

Andreas Gruenbacher

author maintainer

2025-04-14 23:22 UTC #7 lore raw

Re: [PATCH v3 2/2] writeback: Fix false warning in inode_to_wb()

Quoted 8 lines · Andrew Morton

Quoted 6 lines · On Sat, 12 Apr 2025 18:39:12 +0200 Andreas Gruenbacher <[email protected]>…
inode_to_wb() is used also for filesystems that don't support cgroup writeback. For these filesystems inode->i_wb is stable during the lifetime of the inode (it points to bdi->wb) and there's no need to hold locks protecting the inode->i_wb dereference. Improve the warning in inode_to_wb() to not trigger for these filesystems.

Quoted 6 lines

Quoted 2 lines
Signed-off-by Jan Kara <[email protected]>
Reviewed-by Andreas Gruenbacher <[email protected]>
Yoo were on the patch delivery path so there should be a signed-off-by:Andreas somewhere. I made that change to the mm.git copy of this patch.

I guess that's fine as long as Jan is credited as the author.

Thanks,
Andreas