freeze for the upcoming glibc 2.38 release

freeze for the upcoming glibc 2.38 release

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 1583 bytes --]

Hello everyone, 

as you may have already heard I'm volunteering as release manager for the
upcoming glibc 2.38.

We follow approximately the same schedule as last time.


- July 6-11, "slushy ABI freeze"
  * No new ABI additions.
  * ABI corrections should be reviewed, RM approval not required.
  * Avoid any changes that require action by the ports (including all
    that require ulps regeneration)
  * Bug fixes.
  * Monday patch review switches to reviewing critical bug list and
    ABI issues.

  We still have two "desirables" on the wiki page, 
  * "Add clone3 support for multiple architectures" - can we wrap this up
    safely next week?
    https://patchwork.sourceware.org/project/glibc/list/?series=16730
  * "Flip defaults for some security-relevant build configuration flags"
    https://patchwork.sourceware.org/project/glibc/list/?series=21908
    The series consists of 4 patches. 2 (stack-protector-strong), 3 
    (bind-now) and 4 (disable-crypt) should be uncontroversial.
    ## The list thread on 1 (autoconf-2.71) ended without a clear
       conclusion. What's the status here? ##

  Anything else?


- July 12-31, "hard ABI freeze"
  * Noncritical bug fixes.
  * No ABI changes without RM discussion.
  * High risk bug fixes should be discussed on the list.


- August 1
  * glibc 2.38 release

  
If you have any comments or objections, please reply to the list.
  
Cheers,
Andreas


-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: freeze for the upcoming glibc 2.38 release

[Xi Ruoyao]

On Thu, 2023-07-06 at 20:28 +0200, Andreas K. Huettel via Libc-alpha
wrote:

> - July 6-11, "slushy ABI freeze"
>   * No new ABI additions.
>   * ABI corrections should be reviewed, RM approval not required.
>   * Avoid any changes that require action by the ports (including all
>     that require ulps regeneration)
>   * Bug fixes.
>   * Monday patch review switches to reviewing critical bug list and
>     ABI issues.
> 
>   We still have two "desirables" on the wiki page, 
>   * "Add clone3 support for multiple architectures" - can we wrap this
> up
>     safely next week?
>     https://patchwork.sourceware.org/project/glibc/list/?series=16730
>   * "Flip defaults for some security-relevant build configuration
> flags"
>     https://patchwork.sourceware.org/project/glibc/list/?series=21908
>     The series consists of 4 patches. 2 (stack-protector-strong), 3 
>     (bind-now) and 4 (disable-crypt) should be uncontroversial.
>     ## The list thread on 1 (autoconf-2.71) ended without a clear
>        conclusion. What's the status here? ##
> 
>   Anything else?

May I add
https://patchwork.sourceware.org/project/glibc/list/?series=22006 here?
The reason:

- It's trivial.
- Applying it here will stop GCC analyzer developers from implementing
these as special cases.

-- 
Xi Ruoyao <xry111@xry111.site>
School of Aerospace Science and Technology, Xidian University

Re: freeze for the upcoming glibc 2.38 release

[Paul Zimmermann]

       Dear Andreas,

>   Anything else?

I would include "elf: Fix slow tls access after dlopen [BZ #19924]"
(patch from January 6).

Best regards,
Paul

Re: freeze for the upcoming glibc 2.38 release

[Siddhesh Poyarekar]

On 2023-07-06 14:28, Andreas K. Huettel wrote:
> Hello everyone,
> 
> as you may have already heard I'm volunteering as release manager for the
> upcoming glibc 2.38.
> 
> We follow approximately the same schedule as last time.
> 
> 
> - July 6-11, "slushy ABI freeze"
>    * No new ABI additions.
>    * ABI corrections should be reviewed, RM approval not required.
>    * Avoid any changes that require action by the ports (including all
>      that require ulps regeneration)
>    * Bug fixes.
>    * Monday patch review switches to reviewing critical bug list and
>      ABI issues.
> 
>    We still have two "desirables" on the wiki page,
>    * "Add clone3 support for multiple architectures" - can we wrap this up
>      safely next week?
>      https://patchwork.sourceware.org/project/glibc/list/?series=16730
>    * "Flip defaults for some security-relevant build configuration flags"
>      https://patchwork.sourceware.org/project/glibc/list/?series=21908
>      The series consists of 4 patches. 2 (stack-protector-strong), 3
>      (bind-now) and 4 (disable-crypt) should be uncontroversial.
>      ## The list thread on 1 (autoconf-2.71) ended without a clear
>         conclusion. What's the status here? ##

I addressed Joseph's concern about sysdep configure and preconfigure 
snippets in v2; I clarified in the git commit log that the sysdep 
scripts did change but they were all trivial transformations that do not 
affect functionality.

Thanks,
Sid

Re: freeze for the upcoming glibc 2.38 release

[Adhemerval Zanella Netto]

On 06/07/23 15:28, Andreas K. Huettel wrote:
> Hello everyone, 
> 
> as you may have already heard I'm volunteering as release manager for the
> upcoming glibc 2.38.
> 
> We follow approximately the same schedule as last time.
> 
> 
> - July 6-11, "slushy ABI freeze"
>   * No new ABI additions.
>   * ABI corrections should be reviewed, RM approval not required.
>   * Avoid any changes that require action by the ports (including all
>     that require ulps regeneration)
>   * Bug fixes.
>   * Monday patch review switches to reviewing critical bug list and
>     ABI issues.
> 
>   We still have two "desirables" on the wiki page, 
>   * "Add clone3 support for multiple architectures" - can we wrap this up
>     safely next week?

The missing patches are for s390x, arm, and mips. I think it should be ok
to push them, I have test all on them on hardware (s390x) and qemu system
(s390x and mips).

It would be good to have the pidfd and cgroupv2 extension as well [1], since
systemd is asking for such interface for some time.

[1] https://patchwork.sourceware.org/project/glibc/list/?series=16730

Re: freeze for the upcoming glibc 2.38 release

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 804 bytes --]

> > 
> >   We still have two "desirables" on the wiki page, 
> >   * "Add clone3 support for multiple architectures" - can we wrap this up
> >     safely next week?
> 
> The missing patches are for s390x, arm, and mips. I think it should be ok
> to push them, I have test all on them on hardware (s390x) and qemu system
> (s390x and mips).
> 
> It would be good to have the pidfd and cgroupv2 extension as well [1], since
> systemd is asking for such interface for some time.
> 
> [1] https://patchwork.sourceware.org/project/glibc/list/?series=16730

That's clone3, I guess you mean this one?

https://patchwork.sourceware.org/project/glibc/list/?series=22140


-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: freeze for the upcoming glibc 2.38 release

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 525 bytes --]

> >   * "Add clone3 support for multiple architectures" - can we wrap this up
> >     safely next week?
> 
> The missing patches are for s390x, arm, and mips. I think it should be ok
> to push them, I have test all on them on hardware (s390x) and qemu system
> (s390x and mips).

"Checked on mips64el-linux-gnueabihf and mipsel-linux-gnu."

How about N32 ABI (where things like to explode)?

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: freeze for the upcoming glibc 2.38 release

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 619 bytes --]

> May I add
> https://patchwork.sourceware.org/project/glibc/list/?series=22006 here?
> The reason:
> 
> - It's trivial.
> - Applying it here will stop GCC analyzer developers from implementing
> these as special cases.

OK from me

1) please get it technically reviewed from someone as usual (should be a
   formality only in this case)

2) question, should we add v3 (safer but less functional) now and update
   it to v4 (stricter but more functional) in master after release?

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: freeze for the upcoming glibc 2.38 release

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 423 bytes --]

> I would include "elf: Fix slow tls access after dlopen [BZ #19924]"
> (patch from January 6).
> 
> Best regards,
> Paul

That's a performance fix and also rather complex code.

I'd prefer to defer this to after the release; it looks to me like it could
be backported after more testing.

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: freeze for the upcoming glibc 2.38 release

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 850 bytes --]


> >    * "Flip defaults for some security-relevant build configuration flags"
> >      https://patchwork.sourceware.org/project/glibc/list/?series=21908
> >      The series consists of 4 patches. 2 (stack-protector-strong), 3
> >      (bind-now) and 4 (disable-crypt) should be uncontroversial.
> >      ## The list thread on 1 (autoconf-2.71) ended without a clear
> >         conclusion. What's the status here? ##
> 
> I addressed Joseph's concern about sysdep configure and preconfigure 
> snippets in v2; I clarified in the git commit log that the sysdep 
> scripts did change but they were all trivial transformations that do not 
> affect functionality.

Great. OK from me, please get a review as usual.

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: freeze for the upcoming glibc 2.38 release

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 749 bytes --]

Am Freitag, 7. Juli 2023, 23:40:52 CEST schrieb Andreas K. Huettel via Libc-alpha:
> > >   * "Add clone3 support for multiple architectures" - can we wrap this up
> > >     safely next week?
> > 
> > The missing patches are for s390x, arm, and mips. I think it should be ok
> > to push them, I have test all on them on hardware (s390x) and qemu system
> > (s390x and mips).
> 
> "Checked on mips64el-linux-gnueabihf and mipsel-linux-gnu."
> 
> How about N32 ABI (where things like to explode)?

Apart from this ^, looks OK to me.

Would be good to get an Ack from someone familiar with the ports still though.

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: freeze for the upcoming glibc 2.38 release

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 714 bytes --]

> > It would be good to have the pidfd and cgroupv2 extension as well [1], since
> > systemd is asking for such interface for some time.
> > 
> > [1] https://patchwork.sourceware.org/project/glibc/list/?series=16730
> 
> That's clone3, I guess you mean this one?
> 
> https://patchwork.sourceware.org/project/glibc/list/?series=22140

If you're talking about this one ^, "Add pidfd and cgroupv2 support for process
creation (resend)", this is in my opinion too much code too late.

I realize this is desirable, so let's add it to master as soon after the release
as possible.

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: freeze for the upcoming glibc 2.38 release

[Paul Zimmermann]

       Dear Andreas,

I feared that answer. It was a very similar reason why it was not in 2.37.
It is really a pity that nobody takes the time to look at this patch.

Best regards,
Paul Zimmermann

> From: "Andreas K. Huettel" <dilfridge@gentoo.org>
> Cc: adhemerval.zanella@linaro.org, siddhesh@sourceware.org,
>  libc-alpha@sourceware.org, carlos@redhat.com,
>  Paul Zimmermann <Paul.Zimmermann@inria.fr>
> Date: Sat, 08 Jul 2023 19:44:58 +0200
> Organization: Gentoo Linux
> 
> 
> [1:text/plain Hide]
> 
> > I would include "elf: Fix slow tls access after dlopen [BZ #19924]"
> > (patch from January 6).
> > 
> > Best regards,
> > Paul
> 
> That's a performance fix and also rather complex code.
> 
> I'd prefer to defer this to after the release; it looks to me like it could
> be backported after more testing.
> 
> -- 
> Andreas K. Hüttel
> dilfridge@gentoo.org
> Gentoo Linux developer
> (council, toolchain, base-system, perl, libreoffice)
> 
> [2:application/pgp-signature Show Save:signature.asc (981B)]
> 

Re: freeze for the upcoming glibc 2.38 release

[Xi Ruoyao]

On Sat, 2023-07-08 at 19:42 +0200, Andreas K. Huettel wrote:
> > May I add
> > https://patchwork.sourceware.org/project/glibc/list/?series=22006 here?
> > The reason:
> > 
> > - It's trivial.
> > - Applying it here will stop GCC analyzer developers from implementing
> > these as special cases.
> 
> OK from me
> 
> 1) please get it technically reviewed from someone as usual (should be a
>    formality only in this case)

I hope it can be done in the next Monday review...  It seems v2 (same as
v3 but lacking a S-o-b) has already been reviewed as OK.

> 2) question, should we add v3 (safer but less functional) now and update
>    it to v4 (stricter but more functional) in master after release?

Maybe.

-- 
Xi Ruoyao <xry111@xry111.site>
School of Aerospace Science and Technology, Xidian University

Re: freeze for the upcoming glibc 2.38 release

[Adhemerval Zanella Netto]

On 08/07/23 14:48, Andreas K. Huettel wrote:
> Am Freitag, 7. Juli 2023, 23:40:52 CEST schrieb Andreas K. Huettel via Libc-alpha:
>>>>   * "Add clone3 support for multiple architectures" - can we wrap this up
>>>>     safely next week?
>>>
>>> The missing patches are for s390x, arm, and mips. I think it should be ok
>>> to push them, I have test all on them on hardware (s390x) and qemu system
>>> (s390x and mips).
>>
>> "Checked on mips64el-linux-gnueabihf and mipsel-linux-gnu."
>>
>> How about N32 ABI (where things like to explode)?
> 
> Apart from this ^, looks OK to me.
> 
> Would be good to get an Ack from someone familiar with the ports still though.
> 

And testing on mips64el-n32 it triggered a failure not really related to n32
(it was a wrong name for the SETUP_GP64_STACK macro).  I will send an update
version for mips.

Re: freeze for the upcoming glibc 2.38 release

[Noah Goldstein]

On Thu, Jul 6, 2023 at 1:29 PM Andreas K. Huettel via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
> Hello everyone,
>
> as you may have already heard I'm volunteering as release manager for the
> upcoming glibc 2.38.
>
> We follow approximately the same schedule as last time.
>
>
> - July 6-11, "slushy ABI freeze"
>   * No new ABI additions.
>   * ABI corrections should be reviewed, RM approval not required.
>   * Avoid any changes that require action by the ports (including all
>     that require ulps regeneration)
>   * Bug fixes.
>   * Monday patch review switches to reviewing critical bug list and
>     ABI issues.
>
>   We still have two "desirables" on the wiki page,
>   * "Add clone3 support for multiple architectures" - can we wrap this up
>     safely next week?
>     https://patchwork.sourceware.org/project/glibc/list/?series=16730
>   * "Flip defaults for some security-relevant build configuration flags"
>     https://patchwork.sourceware.org/project/glibc/list/?series=21908
>     The series consists of 4 patches. 2 (stack-protector-strong), 3
>     (bind-now) and 4 (disable-crypt) should be uncontroversial.
>     ## The list thread on 1 (autoconf-2.71) ended without a clear
>        conclusion. What's the status here? ##
>
>   Anything else?

I would like to make sure any necessary follow ups for tuning the
`non_temporal_threshold` on X86 machines gets into 2.38.
>
>
> - July 12-31, "hard ABI freeze"
>   * Noncritical bug fixes.
>   * No ABI changes without RM discussion.
>   * High risk bug fixes should be discussed on the list.
>
>
> - August 1
>   * glibc 2.38 release
>
>
> If you have any comments or objections, please reply to the list.
>
> Cheers,
> Andreas
>
>
> --
> Andreas K. Hüttel
> dilfridge@gentoo.org
> Gentoo Linux developer
> (council, toolchain, base-system, perl, libreoffice)

Re: freeze for the upcoming glibc 2.38 release

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 1090 bytes --]

> >   We still have two "desirables" on the wiki page,
> >   * "Add clone3 support for multiple architectures" - can we wrap this up
> >     safely next week?
> >     https://patchwork.sourceware.org/project/glibc/list/?series=16730
> >   * "Flip defaults for some security-relevant build configuration flags"
> >     https://patchwork.sourceware.org/project/glibc/list/?series=21908
> >     The series consists of 4 patches. 2 (stack-protector-strong), 3
> >     (bind-now) and 4 (disable-crypt) should be uncontroversial.
> >     ## The list thread on 1 (autoconf-2.71) ended without a clear
> >        conclusion. What's the status here? ##
> >
> >   Anything else?
> 
> I would like to make sure any necessary follow ups for tuning the
> `non_temporal_threshold` on X86 machines gets into 2.38.

Let's be careful and reluctant with pure performance tuning.

Bug fixes are of course fine (and yes I know the boundary is somewhat
ill-defined).

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: freeze for the upcoming glibc 2.38 release

[Noah Goldstein]

On Thu, Jul 13, 2023 at 4:35 PM Andreas K. Huettel <dilfridge@gentoo.org> wrote:
>
> > >   We still have two "desirables" on the wiki page,
> > >   * "Add clone3 support for multiple architectures" - can we wrap this up
> > >     safely next week?
> > >     https://patchwork.sourceware.org/project/glibc/list/?series=16730
> > >   * "Flip defaults for some security-relevant build configuration flags"
> > >     https://patchwork.sourceware.org/project/glibc/list/?series=21908
> > >     The series consists of 4 patches. 2 (stack-protector-strong), 3
> > >     (bind-now) and 4 (disable-crypt) should be uncontroversial.
> > >     ## The list thread on 1 (autoconf-2.71) ended without a clear
> > >        conclusion. What's the status here? ##
> > >
> > >   Anything else?
> >
> > I would like to make sure any necessary follow ups for tuning the
> > `non_temporal_threshold` on X86 machines gets into 2.38.
>
> Let's be careful and reluctant with pure performance tuning.
>
> Bug fixes are of course fine (and yes I know the boundary is somewhat
> ill-defined).

I think this one is more in the former category. At the very least want
to add a bound that ensure new number is never less than older value
(to avoid any blatant regressions). But hoping to hear from Sajan before
then to see if we can get a proper fix in.

>
> --
> Andreas K. Hüttel
> dilfridge@gentoo.org
> Gentoo Linux developer
> (council, toolchain, base-system, perl, libreoffice)