[PATCH] sigaction: don't sign-extend sa_flags

[Carlos O'Donell]

On 6/2/25 8:58 AM, Adhemerval Zanella Netto wrote:
> 
> 
> On 02/06/25 06:24, Florian Weimer wrote:
>> * наб:
>>
>>> Before:
>>>    rt_sigaction(SIGBUS, {sa_handler=0x55abb9960139, sa_mask=[], sa_flags=SA_RESTORER|SA_RESETHAND|SA_SIGINFO|0xffffffff00000000, sa_restorer=0x7fb1b2a82050}, NULL, 8) = 0
>>>
>>> After:
>>>    rt_sigaction(SIGBUS, {sa_handler=0x7f6a70dce139, sa_mask=[], sa_flags=SA_RESTORER|SA_RESETHAND|SA_SIGINFO, sa_restorer=0x7f6a70c28f60}, NULL, 8) = 0
>>>
>>> Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
>>
>> I wonder if this is something that should be fixed in strace?
>>
>> It's not that the kernel can start using the upper bits with some
>> compatibility check (e.g., bits 32 and 33 have to be distinct).
> 
> POSIX defines sa_flags is an 'int' [1], while the rt signal interface
> uses unsigned (alpha seems to be an outlier, which uses 'int' as well).
> 
> So I think it should be fixed on strace as well; the high bits usage would
> be possible only on 64-bit ABIs.
> 
> [1] https://pubs.opengroup.org/onlinepubs/9799919799/functions/sigaction.html

While I agree that we can't use the high bits, the benefit here is that we'd
fix older strace output that doesn't yet have the fix Florian is recommending?

I'd be willing to accept this change (with the formatting fixed up) just to
make strace output better for developers.

-- 
Cheers,
Carlos.