[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <166811139478.327006.13879198441587445544.stgit@magnolia>
Date: Thu, 10 Nov 2022 12:16:34 -0800
From: "Darrick J. Wong" <djwong@...nel.org>
To: djwong@...nel.org, tytso@....edu
Cc: catherine.hoang@...cle.com, linux-ext4@...r.kernel.org
Subject: [PATCH 2/2] ext4: don't fail GETFSUUID when the caller provides a
long buffer
From: Darrick J. Wong <djwong@...nel.org>
If userspace provides a longer UUID buffer than is required, we
shouldn't fail the call with EINVAL -- rather, we can fill the caller's
buffer with the bytes we /can/ fill, and update the length field to
reflect what we copied. This doesn't break the UAPI since we're
enabling a case that currently fails, and so far Ted hasn't released a
version of e2fsprogs that uses the new ext4 ioctl.
Signed-off-by: Darrick J. Wong <djwong@...nel.org>
---
fs/ext4/ioctl.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/fs/ext4/ioctl.c b/fs/ext4/ioctl.c
index 5f91f3ad3e50..31e643795016 100644
--- a/fs/ext4/ioctl.c
+++ b/fs/ext4/ioctl.c
@@ -1159,14 +1159,16 @@ static int ext4_ioctl_getuuid(struct ext4_sb_info *sbi,
return 0;
}
- if (fsuuid.fsu_len != UUID_SIZE || fsuuid.fsu_flags != 0)
+ if (fsuuid.fsu_len < UUID_SIZE || fsuuid.fsu_flags != 0)
return -EINVAL;
lock_buffer(sbi->s_sbh);
memcpy(uuid, sbi->s_es->s_uuid, UUID_SIZE);
unlock_buffer(sbi->s_sbh);
- if (copy_to_user(&ufsuuid->fsu_uuid[0], uuid, UUID_SIZE))
+ fsuuid.fsu_len = UUID_SIZE;
+ if (copy_to_user(ufsuuid, &fsuuid, sizeof(fsuuid)) ||
+ copy_to_user(&ufsuuid->fsu_uuid[0], uuid, UUID_SIZE))
return -EFAULT;
return 0;
}
Powered by blists - more mailing lists