Re: [RFC 3/4] x86/signal/64: Re-add support for SS in the 64-bit signal context

[Andy Lutomirski]

On Sun, Oct 18, 2015 at 9:43 AM, Stas Sergeev <stsp@xxxxxxx> wrote:
> 18.10.2015 19:36, Andy Lutomirski ÐÐÑÐÑ:
>
>> On Sun, Oct 18, 2015 at 9:29 AM, Stas Sergeev <stsp@xxxxxxx> wrote:
>>>
>>> 18.10.2015 19:12, Andy Lutomirski ÐÐÑÐÑ:
>>>>
>>>> On Sun, Oct 18, 2015 at 6:36 AM, Stas Sergeev <stsp@xxxxxxx> wrote:
>>>>>
>>>>> 15.10.2015 00:41, Andy Lutomirski ÐÐÑÐÑ:
>>>>>>>
>>>>>>> If this my
>>>>>>> understanding is correct and the flag is just an indication rather
>>>>>>> than a requested action, perhaps the name should be different,
>>>>>>> e.g. UC_SIG_FROM_32BIT or the like?
>>>>>>> Anyway, this is minor. :)
>>>>>>> I'll try to test the patch within a few days, thanks for you time!
>>>>>>
>>>>>> No problem.  Thanks for being willing to test!
>>>>>
>>>>> Hello Andy, I am unlucky at testing this.
>>>>> dosemu doesn't even start for me on the git kernels.
>>>>> After a half day of debugging, it seems the kernel forgets
>>>>> to fill in the "err" field in the sigcontext struct when
>>>>> page fault occurs. That confuses the dosemu's instruction
>>>>> decoder.
>>>>> Does this ring any bells?
>>>>
>>>> No, but I can reproduce it on some kernels.  Let me see if I can fix it,
>>>> too.
>>>
>>> Thanks!
>>> You should really consider adding dosemu as your test-case.
>>> It feels very unhappy on all recent kernels. I was getting hard
>>> lock-ups under different circumstances (when starting windows,
>>> for example). But fedora-packaged kernels are quite good, as of
>>> yet. I fear the problems will soon populate to them too.
>>
>> I'll work on that.  It's a lot easier to have a packaged set of tests
>> that say yes/no, though, and it's handy when those tests are fully
>> open-source and easy to run (DOSEMU, in contrast, requires the
>> annoyingly impossible-to-compile freedos stuff iirc, and the
>> interesting bits need other test programs).  In any case, I'll add a
>> self-test for the
>> err thing once I figure out what's going on.
>
> I meant just a local test-case on your PC. :)
> It can't be a part of an automated test-suit of course.
> But after changing anything in vm86 or signal handling,
> running dosemu will never hurt, as it is probably the
> most rigorous test-case for them. I deduce that based
> on its breakage frequency with the kernel updates.

Point taken.

Anyway, the bug was an error in the patches I sent out and aren't in
anyone's tree yet, so there's no big worry here.  I'll send out new
patches after incorporating feedback and poking at it a bit.

--Andy
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/