[FFe] 2.36-0ubuntu4 update (various bugfixes + ARC enablement)

[Simon Chopin]

Hi,

We'd like to upload a new glibc to Kinetic.
It ships a fix for https://bugs.launchpad.net/ubuntu/+source/glibc/+bug/1991652 as well as multiple patches from the upstream stable release branch, for the following bugs:

https://sourceware.org/bugzilla/show_bug.cgi?id=28937
Dynamic loader issue

https://sourceware.org/bugzilla/show_bug.cgi?id=29583
iconv issue on 32-bit architectures with large files. Might impact armhf.

https://sourceware.org/bugzilla/show_bug.cgi?id=29611
Potentially incorrect use of BMI instructions on AVX2 optimised routines

https://sourceware.org/bugzilla/show_bug.cgi?id=29607
https://sourceware.org/bugzilla/show_bug.cgi?id=29605
Couple of nscd regressions

https://sourceware.org/bugzilla/show_bug.cgi?id=29638
arc4random issue on older kernels (including the Trusty ones)

I believe the BMI issue is particularly important to fix. While most hardware has both BMI and AVX2, there are still some without BMI support, or even some that only gained BMI support via microcode updates. While we could assume that users have up-to-date microcode on a running system, this issue could manifest when trying to install Ubuntu on a fresh computer *before* the user had a chance to update the microcode.

[Simon Chopin (schopin)]

Test build: https://launchpad.net/~mwhudson/+archive/ubuntu/devirt/+packages?field.name_filter=glibc&field.status_filter=published&field.series_filter=kinetic

Debdiff is attached.

[Utkarsh Gupta (utkarsh)]

Hi Simon,

Thank for you filing the FFe bug. The desctiption, links, and the debdiff looks reasonable enough. Looks good for FFe to be granted. Please wait for an official Release team member's ACK before uploading, however. Thanks!

[Graham Inggs (ginggs)]

Thanks Utkarsh, FFe granted. Please remember to close this bug in the changelog.

[Simon Chopin (schopin)]

Thanks, uploaded.

[Steve Langasek (vorlon)]

> While we could assume that users have up-to-date microcode on a running
> system, this issue could manifest when trying to install Ubuntu on a fresh
> computer *before* the user had a chance to update the microcode.

This doesn't block acceptance of the upload, but microcode updates are supposed to be applied to the CPU from early boot by the kernel via the initramfs, including in install images; so why is this a concern in practice?

[Michael Hudson-Doyle (mwhudson)]

On Sat, 8 Oct 2022 at 07:00, Steve Langasek <email address hidden>
wrote:

> > While we could assume that users have up-to-date microcode on a running
> > system, this issue could manifest when trying to install Ubuntu on a
> fresh
> > computer *before* the user had a chance to update the microcode.
>
> This doesn't block acceptance of the upload, but microcode updates are
> supposed to be applied to the CPU from early boot by the kernel via the
> initramfs, including in install images; so why is this a concern in
> practice?
>

I think the more realistic scenario is a 22.10 container on an OS that does
not mandate microcode updates. Also I don't know if all processors that are
affected by this will get the relevant instructions enabled by an
up-to-date microcode (I haven't even looked into this at all though).

[Simon Chopin (schopin)]

Quoting Steve Langasek (2022-10-07 19:51:28)
> > While we could assume that users have up-to-date microcode on a running
> > system, this issue could manifest when trying to install Ubuntu on a fresh
> > computer *before* the user had a chance to update the microcode.
>
> This doesn't block acceptance of the upload, but microcode updates are
> supposed to be applied to the CPU from early boot by the kernel via the
> initramfs, including in install images; so why is this a concern in
> practice?

My bad, I now realize I was thinking of Debian's installer, which
doesn't ship (or didn't use to?) microcode in the installer, you had to
enable non-free later on. Pardon my brain fart on that one.

[Launchpad Janitor (janitor)]

This bug was fixed in the package glibc - 2.36-0ubuntu4

---------------
glibc (2.36-0ubuntu4) kinetic; urgency=medium

  [ Michael Hudson-Doyle ]
  * Enable building for the 'arc' architecture. (LP: #1991652)
  * Cherry-pick patches from upstream maintenance branch (LP: #1991898):
    - 0020-01-scripts-dso-ordering-test.py-Generate-program-run-ti.patch
    - 0020-02-elf-Rename-_dl_sort_maps-parameter-from-skip-to-forc.patch
    - 0021-elf-Implement-force_first-handling-in-_dl_sort_maps_.patch
    - 0022-gconv-Use-64-bit-interfaces-in-gconv_parseconfdir-bu.patch
    - 0023-01-x86-include-BMI1-and-BMI2-in-x86-64-v3-level.patch
    - 0023-02-x86-64-Require-BMI2-for-AVX2-str-n-casecmp-implement.patch
    - 0023-03-x86-64-Require-BMI2-for-AVX2-strcmp-implementation.patch
    - 0023-04-x86-64-Require-BMI2-for-AVX2-strncmp-implementation.patch
    - 0023-05-x86-64-Require-BMI2-for-AVX2-wcs-n-cmp-implementatio.patch
    - 0023-06-x86-64-Require-BMI2-for-AVX2-raw-w-memchr-implementa.patch
    - 0023-07-x86-64-Require-BMI2-and-LZCNT-for-AVX2-memrchr-imple.patch
    - 0023-08-x86-64-Require-BMI1-BMI2-for-AVX2-strrchr-and-wcsrch.patch
    - 0024-nscd-Drop-local-address-tuple-variable-BZ-29607.patch
    - 0025-get_nscd_addresses-Fix-subscript-typos-BZ-29605.patch
    - 0026-stdlib-Fix-__getrandom_nocancel-type-and-arc4random-.patch

 -- Simon Chopin <email address hidden> Fri, 07 Oct 2022 10:13:35 +0200