On Fri, Jan 10, 2025 at 05:37:26PM +0100, Günther Noack wrote:

On Fri, Jan 10, 2025 at 04:39:13PM +0100, Mickaël Salaün wrote:

quoted

A corrupted filesystem (e.g. bcachefs) might return weird files.
Instead of throwing a warning and allowing access to such file, treat
them as regular files.

Cc: Dave Chinner <david@fromorbit.com>
Cc: Günther Noack <gnoack@google.com>
Cc: Kent Overstreet <kent.overstreet@linux.dev>
Cc: Paul Moore <paul@paul-moore.com>
Reported-by: syzbot+34b68f850391452207df@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/r/000000000000a65b35061cffca61@google.com (local)
Reported-by: syzbot+360866a59e3c80510a62@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/r/67379b3f.050a0220.85a0.0001.GAE@google.com (local)
Reported-by: Ubisectech Sirius <redacted>
Closes: https://lore.kernel.org/r/c426821d-8380-46c4-a494-7008bbd7dd13.bugreport@ubisectech.com (local)
Fixes: cb2c7d1a1776 ("landlock: Support filesystem access-control")
Signed-off-by: Mickaël Salaün <mic@digikod.net>
---
 security/landlock/fs.c | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/security/landlock/fs.c b/security/landlock/fs.c
index e31b97a9f175..7adb25150488 100644
--- a/security/landlock/fs.c
+++ b/security/landlock/fs.c

@@ -937,10 +937,6 @@ static access_mask_t get_mode_access(const umode_t mode)
 	switch (mode & S_IFMT) {
 	case S_IFLNK:
 		return LANDLOCK_ACCESS_FS_MAKE_SYM;
-	case 0:
-		/* A zero mode translates to S_IFREG. */
-	case S_IFREG:
-		return LANDLOCK_ACCESS_FS_MAKE_REG;
 	case S_IFDIR:
 		return LANDLOCK_ACCESS_FS_MAKE_DIR;
 	case S_IFCHR:

@@ -951,9 +947,12 @@ static access_mask_t get_mode_access(const umode_t mode)
 		return LANDLOCK_ACCESS_FS_MAKE_FIFO;
 	case S_IFSOCK:
 		return LANDLOCK_ACCESS_FS_MAKE_SOCK;
+	case S_IFREG:
+	case 0:
+		/* A zero mode translates to S_IFREG. */
 	default:
-		WARN_ON_ONCE(1);
-		return 0;
+		/* Treats weird files as regular files. */
+		return LANDLOCK_ACCESS_FS_MAKE_REG;
 	}
 }
 

-- 
2.47.1

Reviewed-by: Günther Noack <redacted>

Makes sense to me, since this is enforcing a stronger check than before
and can only happen in the case of corruption.

I do not have a good intuition about what happens afterwards when the
file system is in such a state.  I imagine that this will usually give
an error shortly afterwards, as the opening of the file continues?  Is
that right?

I guess it depends on the filesystem implementation.  For instance, XFS
returns an error if a weird file is detected [1], whereas bcachefs
ignores it (which is considered a bug, but not fixed yet) [2].

[1] https://lore.kernel.org/all/Zpc46HEacI%2Fwd7Rg@dread.disaster.area/ (local)
[2] https://lore.kernel.org/all/4hohnthh54adx35lnxzedop3oxpntpmtygxso4iraiexfdlt4d@6m7ssepvjyar/ (local)

–Günther