Re: [PATCH] x86_64: handle iret faults better

[Andi Kleen <ak@suse.de>]

On Sat, Apr 23, 2005 at 05:50:37PM -0700, Roland McGrath wrote:
> This is the x86_64 variant of the i386 fix I just submitted.  I think
> iret can only produce these faults when returning to user mode in a
> 32-bit process.  The failure mode is even more mysterious on x86_64,
> because it exits with -9999&0x7f instead of 11 (SIGSEGV), so it says
> "Unknown signal 113 (core dumped)" when it exits without actually
> trying to dump a core file.

I agree that handling this better is a good idea.

But I really hate your is_iret hack in traps.c. Cant you just force the signal
after fixing up the stack? I dont want such a ugly complicated special case 
there that only handles this extremly exotic case.
If you cant do it in a cleaner way it would be better not to fix it.
But I suppose forcing a signal directly is doable.

-Andi