From: David Laight > Sent: 01 November 2022 13:30 > > From: Mark Rutland > > Sent: 01 November 2022 13:13 > > > > On Tue, Nov 01, 2022 at 12:19:51PM +0000, David Laight wrote: > > > From: 'Szabolcs Nagy' <szabolcs.nagy@xxxxxxx> > > > > Sent: 01 November 2022 11:45 > > > > > > > > The 11/01/2022 10:02, David Laight wrote: > > > > > From: Szabolcs Nagy > > > > > > kernel code: > > > > ------------ > > > > SYSCALL_DEFINE3(fcntl, unsigned int, fd, unsigned int, cmd, unsigned long, arg) > > > > { > > > > > > That is just a wrapper and calls do_fcntl(). > > > which needs changing to be add: > > > arg &= ~0U; > > > before the switch(cmd) { > > > > Just to check, do you mean the switch in do_fcntl(), or the switch within memfd_fcntl() ? > > > > The former handles other APIs which do expect arg to be a long (e.g. > > F_SET_RW_HINT and F_GET_RW_HINT expect it to hold a full 64-bit pointer), so > > that'd break things. > > The assignment to argv is earlier. Clearly I meant argp :-) Alternatively all the helper functions should be changed to have an 'unsigned int' argument instead of 'unsigned long'. That might show up any that try to read user buffers. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)