Forwarded: [PATCH] nilfs2: reject CLEAN_SEGMENTS ioctl with nsegs exceeding ns_nsegments

[syzbot]

For archival purposes, forwarding an incoming command email to
linux-kernel@xxxxxxxxxxxxxxx, syzkaller-bugs@xxxxxxxxxxxxxxxx.

***

Subject: [PATCH] nilfs2: reject CLEAN_SEGMENTS ioctl with nsegs exceeding ns_nsegments
Author: kartikey406@xxxxxxxxx

#syz test: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master


Syzbot reported a hung task in nilfs_transaction_begin() where multiple
tasks performing chmod() on a nilfs2 mount blocked for over 143 seconds
waiting to acquire ns_segctor_sem for read:

  INFO: task syz.0.17:5918 blocked for more than 143 seconds.
        Not tainted syzkaller #0
  Call Trace:
   schedule+0x164/0x360
   rwsem_down_read_slowpath+0x6d9/0x940
   down_read+0x99/0x2e0
   nilfs_transaction_begin+0x364/0x710 fs/nilfs2/segment.c:221
   nilfs_setattr+0x124/0x2c0 fs/nilfs2/inode.c:921
   notify_change+0xc1a/0xf40
   chmod_common+0x273/0x4a0
   do_fchmodat+0x12d/0x230

The writer holding ns_segctor_sem was a concurrent NILFS_IOCTL_CLEAN_SEGMENTS
caller, stuck inside printk while emitting per-element warnings from
nilfs_sufile_updatev():

   __nilfs_msg+0x373/0x450 fs/nilfs2/super.c:78
   nilfs_sufile_updatev+0x21c/0x6d0 fs/nilfs2/sufile.c:186
   nilfs_sufile_freev fs/nilfs2/sufile.h:93 [inline]
   nilfs_free_segments fs/nilfs2/segment.c:1140 [inline]
   nilfs_segctor_collect_blocks fs/nilfs2/segment.c:1261 [inline]
   nilfs_segctor_do_construct+0x1f55/0x76c0
   nilfs_clean_segments+0x3bd/0xa50
   nilfs_ioctl_clean_segments fs/nilfs2/ioctl.c:922 [inline]
   nilfs_ioctl+0x261f/0x2780

The root cause is that nilfs_ioctl_clean_segments() does not bound
argv[4].v_nmembs (nsegs) against the actual number of segments on the
filesystem.  Userspace can therefore pass an arbitrarily large array of
segment numbers, which is copied in via memdup_array_user() and then
walked under both the sufile mi_sem and ns_segctor_sem (held for write
by nilfs_clean_segments()).  Each invalid entry produces a nilfs_warn()
inside that critical section, and on a slow console the cumulative
printk latency keeps ns_segctor_sem held long enough to trip the
hung_task watchdog.  Any concurrent operation needing
ns_segctor_sem for read -- e.g. chmod() through nilfs_setattr() -- then
stalls for the duration.

While argv[0..3] are bounded by nsegs * ns_blocks_per_segment, nsegs
itself is the root of that dependency chain and has no upper bound.
There is no legitimate reason for a CLEAN_SEGMENTS request to reference
more segments than exist on the filesystem.

Reject such requests at the ioctl entry point so that the malformed
input is refused before any FS-wide lock is acquired.

Reported-by: syzbot+62f0f99d2f2bb8e3bbd7@xxxxxxxxxxxxxxxxxxxxxxxxx
Closes: https://syzkaller.appspot.com/bug?extid=62f0f99d2f2bb8e3bbd7
Signed-off-by: Deepanshu Kartikey <kartikey406@xxxxxxxxx>
---
 fs/nilfs2/ioctl.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/nilfs2/ioctl.c b/fs/nilfs2/ioctl.c
index e0a606643e87..a1688e940f7a 100644
--- a/fs/nilfs2/ioctl.c
+++ b/fs/nilfs2/ioctl.c
@@ -863,6 +863,9 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
 	if (argv[4].v_size != argsz[4])
 		goto out;
 
+	nilfs = inode->i_sb->s_fs_info;
+	if (nsegs > nilfs->ns_nsegments)
+		goto out;
 	/*
 	 * argv[4] points to segment numbers this ioctl cleans.  We
 	 * use kmalloc() for its buffer because the memory used for the
@@ -874,7 +877,6 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
 		ret = PTR_ERR(kbufs[4]);
 		goto out;
 	}
-	nilfs = inode->i_sb->s_fs_info;
 
 	for (n = 0; n < 4; n++) {
 		ret = -EINVAL;
-- 
2.43.0