Skip to contentSkip to conversation

ovl: fix wrong WARN_ON() in ovl_cache_update_ino()

PATCH In review Amir Goldstein 2 participants 5 messages Dec 23, 2019 – Jan 24, 2020 Overlay filesystem

lore
Share & tools
lore

Amir Goldstein

author maintainer
#0 lore raw

PATCH 1/1 ovl: fix wrong WARN_ON() in ovl_cache_update_ino()

The WARN_ON() that child entry is always on overlay st_dev became wrong when we allowed this function to update d_ino in non-samefs setup with xino enabled.

It is not true in case of xino bits overflow on a non-dir inode. Leave the WARN_ON() only for directories, where assertion is still true.

Fixes adbf4f7ea834 ("ovl: consistent d_ino for non-samefs with xino")
Cc <[email protected]> # v4.17+

---

Hi Miklos,

Another fall out from testing nested xino setup.

Thanks,
Amir.

 fs/overlayfs/readdir.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
fs/overlayfs/readdir.c +7 -1
index 47a91c9733a5..7255e6a5838f 100644
@@ -504,7 +504,13 @@ static int ovl_cache_update_ino(struct path *path, struct ovl_cache_entry *p)
 		if (err)
 			goto fail;
 
-		WARN_ON_ONCE(dir->d_sb->s_dev != stat.dev);
+		/*
+		 * Directory inode is always on overlay st_dev.
+		 * Non-dir with ovl_same_dev() could be on pseudo st_dev in case
+		 * of xino bits overflow.
+		 */
+		WARN_ON_ONCE(S_ISDIR(stat.mode) &&
+			     dir->d_sb->s_dev != stat.dev);
 		ino = stat.ino;
 	} else if (xinobits && !OVL_TYPE_UPPER(type)) {
 		ino = ovl_remap_lower_ino(ino, xinobits,
Signature · 1 line
2.17.1
author maintainer
#1 lore raw
Quoted 5 lines · Amir Goldstein

The WARN_ON() that child entry is always on overlay st_dev became wrong when we allowed this function to update d_ino in non-samefs setup with xino enabled.

Quoted 8 lines

It is not true in case of xino bits overflow on a non-dir inode. Leave the WARN_ON() only for directories, where assertion is still true.

Fixes adbf4f7ea834 ("ovl: consistent d_ino for non-samefs with xino")
Cc <[email protected]> # v4.17+

---

Miklos,

If you have time, please send this one to Linus for v5.5.
It is a simple fix and the only one causing failure in the new xfstests [1] that I posted.

Thanks,
Amir.

[1] https://lore.kernel.org/fstests/[email protected]/

Quoted 26 lines
 fs/overlayfs/readdir.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
fs/overlayfs/readdir.c +7 -2
index 47a91c9733a5..7255e6a5838f 100644
@@ -504,7 +504,13 @@ static int ovl_cache_update_ino(struct path *path, struct ovl_cache_entry *p)
                if (err)
                        goto fail;
 
-               WARN_ON_ONCE(dir->d_sb->s_dev != stat.dev);
+               /*
+                * Directory inode is always on overlay st_dev.
+                * Non-dir with ovl_same_dev() could be on pseudo st_dev in case
+                * of xino bits overflow.
+                */
+               WARN_ON_ONCE(S_ISDIR(stat.mode) &&
+                            dir->d_sb->s_dev != stat.dev);
                ino = stat.ino;
        } else if (xinobits && !OVL_TYPE_UPPER(type)) {
                ino = ovl_remap_lower_ino(ino, xinobits,
--

2.17.1

author maintainer
#2 lore raw
Quoted 16 lines · Amir Goldstein
Quoted 5 lines · Amir Goldstein

The WARN_ON() that child entry is always on overlay st_dev became wrong when we allowed this function to update d_ino in non-samefs setup with xino enabled.

Quoted 8 lines

It is not true in case of xino bits overflow on a non-dir inode. Leave the WARN_ON() only for directories, where assertion is still true.

Fixes adbf4f7ea834 ("ovl: consistent d_ino for non-samefs with xino")
Cc <[email protected]> # v4.17+

---

Quoted 6 lines

Miklos,

If you have time, please send this one to Linus for v5.5.
It is a simple fix and the only one causing failure in the new xfstests [1] that I posted.

Gentle nudge..

Thanks,
Amir.

maintainer
#3 lore raw
Quoted 18 lines · Amir Goldstein
Quoted 16 lines · Amir Goldstein
Quoted 13 lines · Amir Goldstein

The WARN_ON() that child entry is always on overlay st_dev became wrong when we allowed this function to update d_ino in non-samefs setup with xino enabled.

It is not true in case of xino bits overflow on a non-dir inode. Leave the WARN_ON() only for directories, where assertion is still true.

Fixes adbf4f7ea834 ("ovl: consistent d_ino for non-samefs with xino")
Cc <[email protected]> # v4.17+

---

Quoted 8 lines
Quoted 6 lines

Miklos,

If you have time, please send this one to Linus for v5.5.
It is a simple fix and the only one causing failure in the new xfstests [1] that I posted.

Gentle nudge..

I'm working this into the next batch bound for 5.6, unless something more urgent comes up before that.

Thanks,
Miklos

author maintainer
#4 lore raw
Quoted 3 lines

I'm working this into the next batch bound for 5.6, unless something more urgent comes up before that.

overlayfs-next tested.
ovl-ino rebased.

You would probably want to squash the "merge conflict" below to overlayfs-next.

Thanks,
Amir.

fs/overlayfs/super.c +0 -0
@@ -1348,7 +1348,7 @@ static int ovl_get_layers(struct super_block

*sb, struct ovl_fs *ofs,

         */
        err = get_anon_bdev(&ofs->fs[0].pseudo_dev);
        if (err) {
-               pr_err("overlayfs: failed to get anonymous bdev for
upper fs\n");
+               pr_err("failed to get anonymous bdev for upper fs\n");
                goto out;
        }

@@ -1409,7 +1409,7 @@ static int ovl_get_layers(struct super_block *sb, struct ovl_fs *ofs,

         */
        if (ofs->numfs - !ofs->upper_mnt == 1) {
                if (ofs->config.xino == OVL_XINO_ON)
-                       pr_info("overlayfs: \"xino=on\" is useless
with all layers on same fs, ignore.\n");
+                       pr_info("\"xino=on\" is useless with all
layers on same fs, ignore.\n");
                ofs->xino_mode = 0;
        } else if (ofs->config.xino == OVL_XINO_ON && ofs->xino_mode < 0) {