Re: [PATCH] compat: Fix endian issue in union sigval

[Chris Metcalf <cmetcalf@ezchip.com>]

On 2/14/2015 6:22 AM, Catalin Marinas wrote:
> 1. user populates sival_int compat_sigevent and invokes
>     compat_sys_mq_notify()
> 2. kernel get_compat_sigevent() copies compat_sigevent into the native
>     sigevent. compat and native sival_int are the same, no problem so
>     far. The other half of 64-bit sival_ptr is zeroed by a memset in this
>     function (this other half can be top or bottom, depending on
>     endianness)
> 3. signal is about to be delivered to user via arch code. The
>     compat_ptr(from->si_ptr) conversion always takes the least
>     significant part of the native si_ptr. On big endian 64-bit, this is
>     zero because get_compat_sigevent() populated the top part of si_ptr
>     with si_int.
>
> So delivering such signals to compat user always sets si_int to 0.
> Little endian is fine.

I looked at this again as I was getting ready to do a tile patch, and realized
why tile and arm64 are different here: tile does a field-by-field copy in
copy_siginfo_from_user32(), like parisc and s390.  As a result, we initialize
the 64-bit kernel si_ptr value by cast from the 32-bit user si_ptr value, rather
than blindly writing into the lower-addressed half of the 64-bit sigval.

As a result, I think I will leave the existing code alone, though unfortunately
that leaves it somewhat unique in manipulating the si_ptr field directly.
But I think the s390 and parisc copy_siginfo_from_user32 leave the high
bits of si_ptr uninitialized, which also strikes me as a bad idea in general.

-- 
Chris Metcalf, EZChip Semiconductor
http://www.ezchip.com