Linux EXT4 FS development
 help / color / mirror / Atom feed
* [PATCH] ext4: reject a non-regular boot loader inode in EXT4_IOC_SWAP_BOOT
@ 2026-11-17  6:18 William Chuang
  2026-11-17  6:43 ` sashiko-bot
  2026-11-17  6:56 ` [PATCH v2] " William Chuang
  0 siblings, 2 replies; 4+ messages in thread
From: William Chuang @ 2026-11-17  6:18 UTC (permalink / raw)
  To: Theodore Ts'o
  Cc: Andreas Dilger, Jan Kara, Ojaswin Mujoo, Ritesh Harjani,
	linux-ext4, linux-kernel

swap_inode_boot_loader() gets the boot loader inode (inode 5) with
EXT4_IGET_BAD.  On a new filesystem that inode is unused, its mode is
0, and ext4_iget() returns it as a bad inode, which make_bad_inode()
sets to S_IFREG.  On a corrupted or crafted filesystem, ext4_iget()
can instead return inode 5 as a directory or another non-regular type,
fully set up with that type's inode and file operations.

Such an inode is first passed to lock_two_nondirectories(), which
warns if it is a directory:

  WARNING: fs/inode.c:1271 at lock_two_nondirectories
   swap_inode_boot_loader
   __ext4_ioctl

The "never used as a boot loader" branch then sets inode_bl->i_mode
to S_IFREG, changing the type of a live in-core inode while it keeps
the operations of its old type.

Check the type right after ext4_iget(): report the corruption and
return -EFSCORRUPTED unless the inode is a regular file.  The unused
case is unaffected, since make_bad_inode() has already made it
S_IFREG.  This is the check Ted Ts'o suggested when the warning was
first reported.

Tested on UML at 666539eec: inode 5 set to a directory or a character
device with debugfs, then EXT4_IOC_SWAP_BOOT issued twice.  Before
this patch, the directory case warns, and both cases succeed and
leave inode 5 as a regular file on disk.  After it, both return
EFSCORRUPTED, the error is logged, and inode 5 is unchanged.  On a
new filesystem the ioctl still succeeds, twice.  Not tested: KASAN
builds and syzbot's 2023 reproducer.

Fixes: 991ed014de08 ("ext4: fix bug_on in __es_tree_search caused by bad boot loader inode")
Suggested-by: Theodore Ts'o <tytso@mit.edu>
Link: https://lore.kernel.org/all/tencent_DABB2333139E8D1BCF4B5D1B2725FABA9108@qq.com/T/
Link: https://sashiko.dev/#/bug/linux-8cfc8e22-8259-4c89-8a8b-64c16e4481a5
Signed-off-by: William Chuang <founder@logarcheon.com>
---
 fs/ext4/ioctl.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/fs/ext4/ioctl.c b/fs/ext4/ioctl.c
index e8e579498..cf005fe2d 100644
--- a/fs/ext4/ioctl.c
+++ b/fs/ext4/ioctl.c
@@ -390,6 +390,20 @@ static long swap_inode_boot_loader(struct super_block *sb,
 		return PTR_ERR(inode_bl);
 	ei_bl = EXT4_I(inode_bl);
 
+	/*
+	 * An unused boot loader inode comes back from ext4_iget() as a bad
+	 * inode, which make_bad_inode() leaves as S_IFREG.  Any other type
+	 * means the on-disk inode is corrupted; its type cannot be changed
+	 * in place below.
+	 */
+	if (!S_ISREG(inode_bl->i_mode)) {
+		EXT4_ERROR_INODE(inode_bl,
+				 "boot loader inode is not a regular file (mode 0%o)",
+				 inode_bl->i_mode);
+		iput(inode_bl);
+		return -EFSCORRUPTED;
+	}
+
 	/* Protect orig inodes against a truncate and make sure,
 	 * that only 1 swap_inode_boot_loader is running. */
 	lock_two_nondirectories(inode, inode_bl);
-- 
2.43.0



^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH] ext4: reject a non-regular boot loader inode in EXT4_IOC_SWAP_BOOT
  2026-11-17  6:18 [PATCH] ext4: reject a non-regular boot loader inode in EXT4_IOC_SWAP_BOOT William Chuang
@ 2026-11-17  6:43 ` sashiko-bot
  2026-11-17  6:56 ` [PATCH v2] " William Chuang
  1 sibling, 0 replies; 4+ messages in thread
From: sashiko-bot @ 2026-11-17  6:43 UTC (permalink / raw)
  To: William Chuang; +Cc: linux-ext4, tytso

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Medium] ext4: missing make_bad_inode() before iput() in
  swap_inode_boot_loader()
--

commit f760f0ea1cf04164f03ebca7f4659e4130a0b1ed
Author: William Chuang <founder@logarcheon.com>

ext4: reject a non-regular boot loader inode in EXT4_IOC_SWAP_BOOT

This commit rejects a boot loader inode (inode 5) in
swap_inode_boot_loader() with -EFSCORRUPTED if its mode after ext4_iget()
is not S_IFREG.

Link: https://lore.kernel.org/all/tencent_DABB2333139E8D1BCF4B5D1B2725FABA9108@qq.com/T/
Link: https://sashiko.dev/#/bug/linux-8cfc8e22-8259-4c89-8a8b-64c16e4481a5

> diff --git a/fs/ext4/ioctl.c b/fs/ext4/ioctl.c
> index e8e579498c990..cf005fe2da34f 100644
> --- a/fs/ext4/ioctl.c
> +++ b/fs/ext4/ioctl.c
> @@ -390,6 +390,20 @@ static long swap_inode_boot_loader(struct super_block *sb,
>  		return PTR_ERR(inode_bl);
>  	ei_bl = EXT4_I(inode_bl);
>  
> +	/*
> +	 * An unused boot loader inode comes back from ext4_iget() as a bad
> +	 * inode, which make_bad_inode() leaves as S_IFREG.  Any other type
> +	 * means the on-disk inode is corrupted; its type cannot be changed
> +	 * in place below.
> +	 */
> +	if (!S_ISREG(inode_bl->i_mode)) {
> +		EXT4_ERROR_INODE(inode_bl,
> +				 "boot loader inode is not a regular file (mode 0%o)",
> +				 inode_bl->i_mode);
> +		iput(inode_bl);
> +		return -EFSCORRUPTED;
> +	}

[Severity: Medium]
Should this error path call make_bad_inode(inode_bl) before iput(inode_bl),
similar to ext4_get_journal_inode()?

__ext4_iget() exempts EXT4_BOOT_LOADER_INO from the i_nlink == 0 check and
only marks it bad when i_mode is not a recognized file type:

fs/ext4/inode.c:__ext4_iget() {
    ...
	if (inode->i_nlink == 0) {
		if ((inode->i_mode == 0 || flags & EXT4_IGET_SPECIAL ||
		     !(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_ORPHAN_FS)) &&
		    ino != EXT4_BOOT_LOADER_INO) {
    ...
}

When a corrupted inode 5 has a valid non-regular i_mode (such as S_IFDIR or
S_IFCHR) and i_nlink == 0, __ext4_iget() returns a live, non-bad inode.
Under errors=continue, iput(inode_bl) then enters ext4_evict_inode(), which
only skips truncation and on-disk deletion when is_bad_inode(inode) is
true, causing it to truncate inode 5, dirty it on disk, and call
ext4_free_inode() on the reserved inode.  If i_nlink > 0, the corrupted
non-regular inode stays hashed in the inode cache instead of being evicted.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261117061822.3826-1-founder@logarcheon.com?part=1

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [PATCH v2] ext4: reject a non-regular boot loader inode in EXT4_IOC_SWAP_BOOT
  2026-11-17  6:18 [PATCH] ext4: reject a non-regular boot loader inode in EXT4_IOC_SWAP_BOOT William Chuang
  2026-11-17  6:43 ` sashiko-bot
@ 2026-11-17  6:56 ` William Chuang
  2026-11-17  7:15   ` sashiko-bot
  1 sibling, 1 reply; 4+ messages in thread
From: William Chuang @ 2026-11-17  6:56 UTC (permalink / raw)
  To: Theodore Ts'o
  Cc: Andreas Dilger, Jan Kara, Ojaswin Mujoo, Ritesh Harjani,
	linux-ext4, linux-kernel

swap_inode_boot_loader() gets the boot loader inode (inode 5) with
EXT4_IGET_BAD.  On a new filesystem that inode is unused, its mode is
0, and ext4_iget() returns it as a bad inode, which make_bad_inode()
sets to S_IFREG.  On a corrupted or crafted filesystem, ext4_iget()
can instead return inode 5 as a directory or another non-regular type,
fully set up with that type's inode and file operations.

Such an inode is first passed to lock_two_nondirectories(), which
warns if it is a directory:

  WARNING: fs/inode.c:1271 at lock_two_nondirectories
   swap_inode_boot_loader
   __ext4_ioctl

The "never used as a boot loader" branch then sets inode_bl->i_mode
to S_IFREG, changing the type of a live in-core inode while it keeps
the operations of its old type.

Check the type right after ext4_iget(): report the corruption and
return -EFSCORRUPTED unless the inode is a regular file.  Mark the
rejected inode bad before iput(), as ext4_get_journal_inode() does, so
that a corrupted inode 5 with i_nlink == 0 is not passed to the
deletion path of ext4_evict_inode().  The unused
case is unaffected, since make_bad_inode() has already made it
S_IFREG.  This is the check Ted Ts'o suggested when the warning was
first reported.

Tested on UML at 666539eec: inode 5 set with debugfs to a directory
(with i_links_count 1, and with 0) or a character device, then
EXT4_IOC_SWAP_BOOT issued twice.  Before this patch, the directory
cases warn, and all cases succeed and leave inode 5 as a regular file
on disk.  After it, all return EFSCORRUPTED, the error is logged,
inode 5 is unchanged, and no attempt is made to free it.  On a
new filesystem the ioctl still succeeds, twice.  Not tested: KASAN
builds and syzbot's 2023 reproducer.

Fixes: 991ed014de08 ("ext4: fix bug_on in __es_tree_search caused by bad boot loader inode")
Suggested-by: Theodore Ts'o <tytso@mit.edu>
Link: https://lore.kernel.org/all/tencent_DABB2333139E8D1BCF4B5D1B2725FABA9108@qq.com/T/
Link: https://sashiko.dev/#/bug/linux-8cfc8e22-8259-4c89-8a8b-64c16e4481a5
Signed-off-by: William Chuang <founder@logarcheon.com>
---
v2: mark the rejected inode bad before iput(), as ext4_get_journal_inode()
    does (Sashiko review).  Without it, a corrupted inode 5 with
    i_links_count 0 went to the deletion path on iput(), which logged
    "ext4_free_inode: reserved or nonexistent inode 5".  Added that case
    to the tests.
v1: https://lore.kernel.org/all/20261117061822.3826-1-founder@logarcheon.com/

 fs/ext4/ioctl.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/fs/ext4/ioctl.c b/fs/ext4/ioctl.c
index e8e579498..8990122f5 100644
--- a/fs/ext4/ioctl.c
+++ b/fs/ext4/ioctl.c
@@ -390,6 +390,26 @@ static long swap_inode_boot_loader(struct super_block *sb,
 		return PTR_ERR(inode_bl);
 	ei_bl = EXT4_I(inode_bl);
 
+	/*
+	 * An unused boot loader inode comes back from ext4_iget() as a bad
+	 * inode, which make_bad_inode() leaves as S_IFREG.  Any other type
+	 * means the on-disk inode is corrupted; its type cannot be changed
+	 * in place below.
+	 */
+	if (!S_ISREG(inode_bl->i_mode)) {
+		EXT4_ERROR_INODE(inode_bl,
+				 "boot loader inode is not a regular file (mode 0%o)",
+				 inode_bl->i_mode);
+		/*
+		 * Drop it as a bad inode, as ext4_get_journal_inode() does:
+		 * with i_nlink == 0, a plain iput() would try to delete the
+		 * reserved inode, and otherwise it would stay cached.
+		 */
+		make_bad_inode(inode_bl);
+		iput(inode_bl);
+		return -EFSCORRUPTED;
+	}
+
 	/* Protect orig inodes against a truncate and make sure,
 	 * that only 1 swap_inode_boot_loader is running. */
 	lock_two_nondirectories(inode, inode_bl);
-- 
2.43.0



^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH v2] ext4: reject a non-regular boot loader inode in EXT4_IOC_SWAP_BOOT
  2026-11-17  6:56 ` [PATCH v2] " William Chuang
@ 2026-11-17  7:15   ` sashiko-bot
  0 siblings, 0 replies; 4+ messages in thread
From: sashiko-bot @ 2026-11-17  7:15 UTC (permalink / raw)
  To: William Chuang; +Cc: linux-ext4, tytso

Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [Medium] ext4: unlocked make_bad_inode() in swap_inode_boot_loader()
  races with iget
- [Low] ext4: redundant !S_ISREG() check left in
  swap_inode_boot_loader()
--

commit 92e1647560c592675448976b5391a26f9a8b9d29
Author: William Chuang <founder@logarcheon.com>

ext4: reject a non-regular boot loader inode in EXT4_IOC_SWAP_BOOT

This commit rejects EXT4_IOC_SWAP_BOOT when the boot loader inode is not a
regular file, marking the inode bad before dropping the reference.

Link: https://lore.kernel.org/all/tencent_DABB2333139E8D1BCF4B5D1B2725FABA9108@qq.com/T/
Link: https://sashiko.dev/#/bug/linux-8cfc8e22-8259-4c89-8a8b-64c16e4481a5

> diff --git a/fs/ext4/ioctl.c b/fs/ext4/ioctl.c
> index e8e579498c990..8990122f585df 100644
> --- a/fs/ext4/ioctl.c
> +++ b/fs/ext4/ioctl.c
> @@ -390,6 +390,26 @@ static long swap_inode_boot_loader(struct super_block *sb,
>  		return PTR_ERR(inode_bl);
>  	ei_bl = EXT4_I(inode_bl);
>  
> +	/*
> +	 * An unused boot loader inode comes back from ext4_iget() as a bad
> +	 * inode, which make_bad_inode() leaves as S_IFREG.  Any other type
> +	 * means the on-disk inode is corrupted; its type cannot be changed
> +	 * in place below.
> +	 */
> +	if (!S_ISREG(inode_bl->i_mode)) {
> +		EXT4_ERROR_INODE(inode_bl,
> +				 "boot loader inode is not a regular file (mode 0%o)",
> +				 inode_bl->i_mode);
> +		/*
> +		 * Drop it as a bad inode, as ext4_get_journal_inode() does:
> +		 * with i_nlink == 0, a plain iput() would try to delete the
> +		 * reserved inode, and otherwise it would stay cached.
> +		 */
> +		make_bad_inode(inode_bl);
> +		iput(inode_bl);
> +		return -EFSCORRUPTED;
> +	}
> +
>  	/* Protect orig inodes against a truncate and make sure,
>  	 * that only 1 swap_inode_boot_loader is running. */
>  	lock_two_nondirectories(inode, inode_bl);

[Severity: Medium]
Can calling make_bad_inode(inode_bl) on a live, unlocked inode race with a
concurrent swap_inode_boot_loader() caller and trick it into treating the
corrupted inode as an unused boot loader inode?

When the on-disk boot loader inode has a valid non-regular mode,
__ext4_iget() unlocks and returns it while it is hashed in the inode cache.
make_bad_inode() overwrites inode_bl->i_mode with S_IFREG and sets
inode_bl->i_op to &bad_inode_ops, which is the same state
swap_inode_boot_loader() uses to identify an unused boot loader inode:

CPU0                                  CPU1
swap_inode_boot_loader()              swap_inode_boot_loader()
  inode_bl = ext4_iget(...)
  !S_ISREG(inode_bl->i_mode)
  EXT4_ERROR_INODE(...)
                                        inode_bl = ext4_iget(...)
                                          /* gets same cached inode_bl */
  make_bad_inode(inode_bl)
    inode_bl->i_mode = S_IFREG
    inode_bl->i_op = &bad_inode_ops
  iput(inode_bl)
                                        !S_ISREG(inode_bl->i_mode) /* false */
                                        ...
                                        if (is_bad_inode(inode_bl) || ...)
                                          /* reinitializes & swaps inode_bl */

CPU1 then bypasses the !S_ISREG(inode_bl->i_mode) check, enters the
is_bad_inode(inode_bl) branch, and swaps the corrupted inode (preserving
its stale i_blocks, i_bytes, and union fields) with the caller's file.

[Severity: Low]
This isn't a bug, but now that non-regular boot loader inodes are rejected
upfront here, is the !S_ISREG(inode_bl->i_mode) check added by
991ed014de08 ("ext4: fix bug_on in __es_tree_search caused by bad boot
loader inode") further down in swap_inode_boot_loader() now redundant?

fs/ext4/ioctl.c:swap_inode_boot_loader() {
    ...
	if (is_bad_inode(inode_bl) || !S_ISREG(inode_bl->i_mode)) {
		/* this inode has never been used as a BOOT_LOADER */
    ...
}

Dropping !S_ISREG(inode_bl->i_mode) from that condition would keep the two
checks consistent.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261117065645.4409-1-founder@logarcheon.com?part=1

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2026-11-17  7:15 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-17  6:18 [PATCH] ext4: reject a non-regular boot loader inode in EXT4_IOC_SWAP_BOOT William Chuang
2026-11-17  6:43 ` sashiko-bot
2026-11-17  6:56 ` [PATCH v2] " William Chuang
2026-11-17  7:15   ` sashiko-bot

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox