From: Carlos O'Donell <carlos@redhat.com>
To: "H.J. Lu" <hjl.tools@gmail.com>, libc-alpha@sourceware.org
Subject: Re: [PATCH] x86: Install <bits/platform/x86.h> [BZ #27958]
Date: Fri, 9 Jul 2021 16:01:44 -0400 [thread overview]
Message-ID: <83077104-ea01-0afc-5636-87e1039d463a@redhat.com> (raw)
In-Reply-To: <20210605135947.469959-1-hjl.tools@gmail.com>
On 6/5/21 9:59 AM, H.J. Lu via Libc-alpha wrote:
> Install <bits/platform/x86.h> for <sys/platform/x86.h> which includes
> <bits/platform/x86.h>.
>
> Fixes BZ #27958.
The constants in bits/platform/x86.h are largely ABI given the behaviour
of the cpuid instruction. Likewise we do a consistent mapping between
the cpuid_array <-> usable_array without exposing internal details.
The API in sys/platform/x86.h has already been reviewed, discussed, and
exposes HAS_CPU_FEATURE(name) and CPU_FEATURE_USABLE(name).
Given that we get one more chance at review let me ask a few final questions.
(1) API prefixes in macros help developers remember names.
Consistent prefix for APIs help developers remember.
We use HAS_* but also CPU_* which requires the programmer remember
two distinct naming strategies.
Suggestion: CPU_FEATURE_PRESENT(), CPU_FEATURE_USABLE()?
Note: We do this in the underlying name e.g. x86_cpu_*
has_feature (could be is_present) vs. is_usable.
(2) ABI testing?
- How are we making sure we don't accidentally break ABI?
- Do we need any further testing?
- Do we have a decoupled test to ensure a refactor doesn't break
things?
- We have tst-cpu-features-cpuinfo.c, which should cover
comparison to the decoupled cpuinfo.
Notes:
- We will not be able to avoid in-place-update failures, in that rpm
will do an atomic rename that unlinks the old libc.so.6 with the
new libc.so.6 and if ld.so is not yet updated or updated first then
a process that starts will crash. This makes it error prone to update
the ABI in downstream minor updates.
> ---
> sysdeps/x86/Makefile | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/sysdeps/x86/Makefile b/sysdeps/x86/Makefile
> index 346ec491b3..567ea54243 100644
> --- a/sysdeps/x86/Makefile
> +++ b/sysdeps/x86/Makefile
> @@ -5,7 +5,7 @@ endif
> ifeq ($(subdir),elf)
> sysdep_routines += get-cpuid-feature-leaf
> sysdep-dl-routines += dl-get-cpu-features
> -sysdep_headers += sys/platform/x86.h
> +sysdep_headers += sys/platform/x86.h bits/platform/x86.h
>
> CFLAGS-get-cpuid-feature-leaf.o += $(no-stack-protector)
--
Cheers,
Carlos.
next prev parent reply other threads:[~2021-07-09 20:01 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-06-05 13:59 H.J. Lu
2021-07-09 20:01 ` Carlos O'Donell [this message]
2021-07-10 17:21 ` H.J. Lu
2021-07-10 17:21 H.J. Lu
2021-07-19 2:13 ` Carlos O'Donell
2021-07-19 15:20 ` H.J. Lu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=83077104-ea01-0afc-5636-87e1039d463a@redhat.com \
--to=carlos@redhat.com \
--cc=hjl.tools@gmail.com \
--cc=libc-alpha@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).