Re: [PATCH] Fix sync. in blkdev_write_iter() acessing i_flags

[Al Viro]

On Fri, Dec 07, 2018 at 05:10:15PM +0100, Alexander Lochmann wrote:
> 
> inode.i_flags might be altered without proper
> synchronisation when the inode belongs to devtmpfs.
> blkdev_write_iter() starts writing via __generic_file_write_iter()
> which sets S_NOSEC bit without any synchronisation.
> The following stacktrace shows how to get there:
> 13: entry_SYSENTER_32:460
> 12: do_fast_syscall_32:410
> 11: _static_cpu_has:146
> 10: do_syscall_32_irqs_on:322
> 09: SyS_pwrite64:636
> 08: SYSC_pwrite64:650
> 07: fdput:38
> 06: vfs_write:560
> 05: __vfs_write:512
> 04: new_sync_write:500
> 03: blkdev_write_iter:1977
> 02: __generic_file_write_iter:2897
> 01: file_remove_privs:1818
> 00: inode_has_no_xattr:3163
> If S_NOSEC is *not* set, i_rwsem is acquired around
> __generic_file_write_iter().

> +	 * Ensure excl. access to i_flags in __generic_file_write_iter().
> +	 * Otherwise, it would race with chmod adding SUID bit.
> +	 */

_What_ SUID bit?  We are talking about a write to block device, for fsck sake...