Re: [PATCH] Fix compat regression in process_vm_rw()

[Christoph Hellwig]

On Tue, Oct 27, 2020 at 12:09:20AM +0000, Al Viro wrote:
> On Mon, Oct 26, 2020 at 06:03:18PM -0600, Jens Axboe wrote:
> > The removal of compat_process_vm_{readv,writev} didn't change
> > process_vm_rw(), which always assumes it's not doing a compat syscall.
> > Instead of passing in 'false' unconditionally for 'compat', make it
> > conditional on in_compat_syscall().
> > 
> > Fixes: c3973b401ef2 ("mm: remove compat_process_vm_{readv,writev}")
> > Reported-by: Kyle Huey <me@xxxxxxxxxxxx>
> > Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
> 
> ACK with some reservations - I suspect that we want an explicit flag
> for process_vm_{read,write}v() that would force the 64bit layout for
> the vector refering to the foreign process.  It's not relevant for
> regression fix; however, as it is these syscalls are not usable for
> 32bit process trying to access memory of 64bit one - there's no way
> to specify the addresses past 4G.

Independent of this fix I think we just need to explicitly prohibit
cross-access.