Horton

Re: Detection of hadware feature => please do not use signal

Heikki Linnakangas <hlinnaka@iki.fi>

View thread View on postgr.es
From: Heikki Linnakangas <hlinnaka@iki.fi>
To: Bastien Roucariès <rouca@debian.org>, pgsql-bugs@lists.postgresql.org
Date: 2024-10-31T17:29:45Z
Lists: pgsql-bugs 
On 31/10/2024 17:41, Bastien Roucariès wrote:
> Hi,
> 
> On debian side we will like to avoid use of sigill to detect feature.
> 
> https://sources.debian.org/src/postgresql-17/17.0-1/src/port/pg_crc32c_armv8_choose.c/#L55
> 
> is really bad.

Why is it bad?

> Could you use the canonical way under linux
> 
> #include <sys/auxv.h>
> #include <asm/hwcap.h>
> 
>   #if defined(__aarch64__)
>     return !(getauxval(AT_HWCAP) &  HWCAP_CRC32);
>   #else
>     return !(getauxval(AT_HWCAP2) & HWCAP2_CRC32);
>   #endif

We used to do that, but it was changed in commit 1c72ec6f. Looking at 
the discussion that led to it, it's not clear to me why we switched [1]. 
Portablity to non-glibc systems maybe?

If there's a handy, portable way to check CPU capabilities, I agree it 
seem less hacky than catching SIGILL...

[1] 
https://www.postgresql.org/message-id/CAEepm%3D02Run-Pk3xyt%2BRV3p1N%2B7cKZxN95_MamaJw8Cnw%2BDwjQ%40mail.gmail.com

-- 
Heikki Linnakangas
Neon (https://neon.tech)

Commits

  1. Update configure probes for CFLAGS needed for ARM CRC instructions.

  2. Support runtime CRC feature probing on NetBSD/ARM using sysctl().

  3. Use auxv to check for CRC32 instructions on ARM.

  4. Improve our method for probing the availability of ARM CRC instructions.