xfs: use setattr_copy to set vfs inode attributes
authorDarrick J. Wong <djwong@kernel.org>
Tue, 8 Mar 2022 18:51:16 +0000 (10:51 -0800)
committerDarrick J. Wong <djwong@kernel.org>
Mon, 14 Mar 2022 17:23:16 +0000 (10:23 -0700)
Filipe Manana pointed out that XFS' behavior w.r.t. setuid/setgid
revocation isn't consistent with btrfs[1] or ext4.  Those two
filesystems use the VFS function setattr_copy to convey certain
attributes from struct iattr into the VFS inode structure.

Andrey Zhadchenko reported[2] that XFS uses the wrong user namespace to
decide if it should clear setgid and setuid on a file attribute update.
This is a second symptom of the problem that Filipe noticed.

XFS, on the other hand, open-codes setattr_copy in xfs_setattr_mode,
xfs_setattr_nonsize, and xfs_setattr_time.  Regrettably, setattr_copy is
/not/ a simple copy function; it contains additional logic to clear the
setgid bit when setting the mode, and XFS' version no longer matches.

The VFS implements its own setuid/setgid stripping logic, which
establishes consistent behavior.  It's a tad unfortunate that it's
scattered across notify_change, should_remove_suid, and setattr_copy but
XFS should really follow the Linux VFS.  Adapt XFS to use the VFS
functions and get rid of the old functions.

[1] https://lore.kernel.org/fstests/CAL3q7H47iNQ=Wmk83WcGB-KBJVOEtR9+qGczzCeXJ9Y2KCV25Q@mail.gmail.com/
[2] https://lore.kernel.org/linux-xfs/20220221182218.748084-1-andrey.zhadchenko@virtuozzo.com/

Fixes: 7fa294c8991c ("userns: Allow chown and setgid preservation")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Christian Brauner <brauner@kernel.org>
fs/xfs/xfs_iops.c
fs/xfs/xfs_pnfs.c

index a65217f..db97cfe 100644 (file)
@@ -613,37 +613,6 @@ xfs_vn_getattr(
        return 0;
 }
 
-static void
-xfs_setattr_mode(
-       struct xfs_inode        *ip,
-       struct iattr            *iattr)
-{
-       struct inode            *inode = VFS_I(ip);
-       umode_t                 mode = iattr->ia_mode;
-
-       ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
-
-       inode->i_mode &= S_IFMT;
-       inode->i_mode |= mode & ~S_IFMT;
-}
-
-void
-xfs_setattr_time(
-       struct xfs_inode        *ip,
-       struct iattr            *iattr)
-{
-       struct inode            *inode = VFS_I(ip);
-
-       ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
-
-       if (iattr->ia_valid & ATTR_ATIME)
-               inode->i_atime = iattr->ia_atime;
-       if (iattr->ia_valid & ATTR_CTIME)
-               inode->i_ctime = iattr->ia_ctime;
-       if (iattr->ia_valid & ATTR_MTIME)
-               inode->i_mtime = iattr->ia_mtime;
-}
-
 static int
 xfs_vn_change_ok(
        struct user_namespace   *mnt_userns,
@@ -742,16 +711,6 @@ xfs_setattr_nonsize(
                gid = (mask & ATTR_GID) ? iattr->ia_gid : igid;
                uid = (mask & ATTR_UID) ? iattr->ia_uid : iuid;
 
-               /*
-                * CAP_FSETID overrides the following restrictions:
-                *
-                * The set-user-ID and set-group-ID bits of a file will be
-                * cleared upon successful return from chown()
-                */
-               if ((inode->i_mode & (S_ISUID|S_ISGID)) &&
-                   !capable(CAP_FSETID))
-                       inode->i_mode &= ~(S_ISUID|S_ISGID);
-
                /*
                 * Change the ownerships and register quota modifications
                 * in the transaction.
@@ -763,7 +722,6 @@ xfs_setattr_nonsize(
                                olddquot1 = xfs_qm_vop_chown(tp, ip,
                                                        &ip->i_udquot, udqp);
                        }
-                       inode->i_uid = uid;
                }
                if (!gid_eq(igid, gid)) {
                        if (XFS_IS_GQUOTA_ON(mp)) {
@@ -774,15 +732,10 @@ xfs_setattr_nonsize(
                                olddquot2 = xfs_qm_vop_chown(tp, ip,
                                                        &ip->i_gdquot, gdqp);
                        }
-                       inode->i_gid = gid;
                }
        }
 
-       if (mask & ATTR_MODE)
-               xfs_setattr_mode(ip, iattr);
-       if (mask & (ATTR_ATIME|ATTR_CTIME|ATTR_MTIME))
-               xfs_setattr_time(ip, iattr);
-
+       setattr_copy(mnt_userns, inode, iattr);
        xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 
        XFS_STATS_INC(mp, xs_ig_attrchg);
@@ -1006,11 +959,8 @@ xfs_setattr_size(
                xfs_inode_clear_eofblocks_tag(ip);
        }
 
-       if (iattr->ia_valid & ATTR_MODE)
-               xfs_setattr_mode(ip, iattr);
-       if (iattr->ia_valid & (ATTR_ATIME|ATTR_CTIME|ATTR_MTIME))
-               xfs_setattr_time(ip, iattr);
-
+       ASSERT(!(iattr->ia_valid & (ATTR_UID | ATTR_GID)));
+       setattr_copy(mnt_userns, inode, iattr);
        xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 
        XFS_STATS_INC(mp, xs_ig_attrchg);
index 4abe173..37a24f0 100644 (file)
@@ -319,7 +319,8 @@ xfs_fs_commit_blocks(
        xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
        xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 
-       xfs_setattr_time(ip, iattr);
+       ASSERT(!(iattr->ia_valid & (ATTR_UID | ATTR_GID)));
+       setattr_copy(&init_user_ns, inode, iattr);
        if (update_isize) {
                i_size_write(inode, iattr->ia_size);
                ip->i_disk_size = iattr->ia_size;