Re: glibc redefines SIGSTKSZ and MINSTKSZ (Re: pagure pushed to stress-ng (rawhide). "Fix build with glibc 2.34 (..more)")

[Florian Weimer]

* Tomasz Kłoczko:

> Is it not kind of a mistake in case of glibc to introduce such
> changes? (just asking to only have confirmation that it was not actual
> mistake and not to start another flame)

We don't have a choice.  The old interfaces are simply incompatible with
modern hardware.  CPUs have massive register files nowadays.
Applications using minimal signal stacks need to adapt in some way.

This is fallout from a hardware change.  I expect that related bugs will
be quite visible as Tiger Lake devices become more common.

Thanks,
Florian
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org
Do not reply to spam on the list, report it: 
https://pagure.io/fedora-infrastructure

glibc redefines SIGSTKSZ and MINSTKSZ (Re: pagure pushed to stress-ng (rawhide). "Fix build with glibc 2.34 (..more)")

[Tomasz Kłoczko]

On Thu, 25 Feb 2021 at 10:52,  wrote:

> Notification time stamped 2021-02-25 10:17:40 UTC
>
> From a9bc8b5947d10f13996e6e5d7280e5860473bbd4 Mon Sep 17 00:00:00 2001
> From: Yaakov Selkowitz 
> Date: Feb 25 2021 03:05:37 +
> Subject: Fix build with glibc 2.34
>
>
> https://github.com/ColinIanKing/stress-ng/issues/107
>
> ---
>
> diff --git a/stress-ng-0.12.03-sigstksz.patch
> b/stress-ng-0.12.03-sigstksz.patch
> new file mode 100644
> index 000..b05c935
> --- /dev/null
> +++ b/stress-ng-0.12.03-sigstksz.patch
> @@ -0,0 +1,661 @@
> +From 7c4f74761089177127c2cfe6685b7886aa231885 Mon Sep 17 00:00:00 2001
> +From: Colin Ian King 
> +Date: Thu, 25 Feb 2021 00:33:17 +
> +Subject: [PATCH] stack handling: use _SC_SIGSTKSZ and _SC_MINSIGSTKSZ
> +
> +New versions of glibc will define SIGSTKSZ and MINSTKSZ
> +on the sysconf values for _SC_SIGSTKSZ and _SC_MINSIGSTKSZ
> +respectively.  Define two helper functions to determine the
> +stack sizes by trying to use cached sysconf values, fetching
> +and caching the sysconf values or falling back to the
> +traditional SIGSTKSZ or MINSTKSZ defined values, or hard
> +coded 8K limits if all else fails.
> +
> +Define STRESS_SIGSTKSZ and STRESS_MINSTKSZ that call the
> +helper functions and hide the details.  Since these sizes
> +are dynamic, replace all statically allocated and stack
> +allocated alternative stacks with mmap'd versions and add
> +in allocation failure error handling.
> +
> +Finally remove the MLOCKED_DATA macros now that the mlocked
> +alt stacks are no longer used.
> +
> +Signed-off-by: Colin Ian King 
> +---
> + core-helper.c | 85 +--
> + stress-bad-altstack.c | 39 ++--
> + stress-context.c  | 19 +++---
> + stress-ng.c   |  3 --
> + stress-ng.h   | 10 -
> + stress-rlimit.c   | 15 +++-
> + stress-stack.c| 18 -
> + stress-stackmmap.c| 70 ---
> + stress-vforkmany.c| 14 +--
> + 9 files changed, 180 insertions(+), 93 deletions(-)
>

Hmm .. looks like there will probably be mny packages affected by this
issue because of the recent glibc changes.
The same is with ocaml https://github.com/ocaml/ocaml/issues/10250

Is it not kind of a mistake in case of glibc to introduce such changes?
(just asking to only have confirmation that it was not actual mistake and
not to start another flame)

kloczek
-- 
Tomasz Kłoczko | LinkedIn: http://lnkd.in/FXPWxH
___
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org
Do not reply to spam on the list, report it: 
https://pagure.io/fedora-infrastructure