On 19. 6. 2024 17:06, Jens Axboe wrote: > On 6/19/24 9:04 AM, Jakub Kicinski wrote: >> On Wed, 19 Jun 2024 07:40:40 -0600 Jens Axboe wrote: >>> On 6/18/24 6:49 PM, Jakub Kicinski wrote: >>>> On Fri, 14 Jun 2024 12:30:44 -0400 Gabriel Krisman Bertazi wrote: >>>>> io_uring holds a reference to the file and maintains a >>>>> sockaddr_storage address. Similarly to what was done to >>>>> __sys_connect_file, split an internal helper for __sys_bind in >>>>> preparation to supporting an io_uring bind command. >>>>> >>>>> Reviewed-by: Jens Axboe <axboe@xxxxxxxxx> >>>>> Signed-off-by: Gabriel Krisman Bertazi <krisman@xxxxxxx> >>>> >>>> Acked-by: Jakub Kicinski <kuba@xxxxxxxxxx> >>> >>> Are you fine with me queueing up 1-2 via the io_uring branch? >>> I'm guessing the risk of conflict should be very low, so doesn't >>> warrant a shared branch. >> >> Yup, exactly, these can go via io_uring without branch juggling. > > Great thanks! > Please fix io_bind and io_listen to not pass NULL ptr to related helpers __sys_bind_socket and __sys_listen_socket. The first helper's argument shouldn't be NULL, as related security hooks expect a valid socket object. See the syzkaller's bug report: https://lore.kernel.org/linux-security-module/0000000000007b7ce6061d1caec0@xxxxxxxxxx/ Thanks, mY
