From: Bastien =?ISO-8859-1?Q?Roucari=E8s?= <rouca@debian.org>
To: pgsql-bugs@lists.postgresql.org, Heikki Linnakangas <hlinnaka@iki.fi>,
 pgsql-bugs@lists.postgresql.org, Heikki Linnakangas <hlinnaka@iki.fi>
Subject: Re: Detection of hadware feature => please do not use signal
Date: Thu, 31 Oct 2024 17:42:09 +0000
Message-ID: <193820057.q8YS0ZdlI3@portable-bastien>
In-Reply-To: <af84e934-4e9c-4380-95ee-8921dab1254b@iki.fi>
References: 
 <4496616.iHFcN1HehY@portable-bastien>
 <af84e934-4e9c-4380-95ee-8921dab1254b@iki.fi>
MIME-Version: 1.0
Content-Type: multipart/signed; boundary="nextPart11275106.XtdUW7qJID";
 micalg="pgp-sha512"; protocol="application/pgp-signature"
Archived-At: 
 <https://www.postgresql.org/message-id/193820057.q8YS0ZdlI3%40portable-bastien>
Precedence: bulk

--nextPart11275106.XtdUW7qJID
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="UTF-8"; protected-headers="v1"
From: Bastien =?ISO-8859-1?Q?Roucari=E8s?= <rouca@debian.org>
Subject: Re: Detection of hadware feature => please do not use signal
Date: Thu, 31 Oct 2024 17:41:53 +0000
Message-ID: <193820057.q8YS0ZdlI3@portable-bastien>
In-Reply-To: <af84e934-4e9c-4380-95ee-8921dab1254b@iki.fi>
MIME-Version: 1.0

Le jeudi 31 octobre 2024, 17:29:45 UTC Heikki Linnakangas a =C3=A9crit :
> On 31/10/2024 17:41, Bastien Roucari=C3=A8s wrote:
> > Hi,
> >=20
> > On debian side we will like to avoid use of sigill to detect feature.
> >=20
> > https://sources.debian.org/src/postgresql-17/17.0-1/src/port/pg_crc32c_=
armv8_choose.c/#L55
> >=20
> > is really bad.
>=20
> Why is it bad?

for a binary it is maybe safe for a library it is unsafe in Multi thread co=
ntext to capture signal...
And we are burned by a recent CVE in ssh due to signal handling.
>=20
> > Could you use the canonical way under linux
> >=20
> > #include <sys/auxv.h>
> > #include <asm/hwcap.h>
> >=20
> >   #if defined(__aarch64__)
> >     return !(getauxval(AT_HWCAP) &  HWCAP_CRC32);
> >   #else
> >     return !(getauxval(AT_HWCAP2) & HWCAP2_CRC32);
> >   #endif
>=20
> We used to do that, but it was changed in commit 1c72ec6f. Looking at=20
> the discussion that led to it, it's not clear to me why we switched [1].=
=20
> Portablity to non-glibc systems maybe?

Yes I suppose but a
#ifdef __GLIBC__

#else

#endif

Will be safe
>=20
> If there's a handy, portable way to check CPU capabilities, I agree it=20
> seem less hacky than catching SIGILL...

unfortunatly no
>=20
> [1]=20
> https://www.postgresql.org/message-id/CAEepm%3D02Run-Pk3xyt%2BRV3p1N%2B7c=
KZxN95_MamaJw8Cnw%2BDwjQ%40mail.gmail.com
>=20
>=20


--nextPart11275106.XtdUW7qJID
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: This is a digitally signed message part.
Content-Transfer-Encoding: 7Bit

-----BEGIN PGP SIGNATURE-----
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=iSnE
-----END PGP SIGNATURE-----

--nextPart11275106.XtdUW7qJID--