Monday Patch Queue Review update (2024-09-09)

[Noah Goldstein]

On Mon, Sep 9, 2024 at 11:03 AM Carlos O'Donell <carlos@redhat.com> wrote:
>
> Most recent meeting status is always here:
> https://sourceware.org/glibc/wiki/PatchworkReviewMeetings#Update
>
> Meeting: 2024-09-09 @ 0900h EST5EDT
>
> Video/Audio: https://bbb.linuxfoundation.org/room/adm-alk-1uu-7fu
>
> IRC: #glibc on OFTC.
>
> Review new patches and restart review at the top.
>
> * State NEW and delegate NOBODY at 457 patches.
>  * Carlos's SLI at 356 days average patch age in queue and 167306 accumulated patch days.
>  * Starting at 97340.
>  * [v2] math: Improve layout of expf data (Wilco)
>   * Reviewer Pool: Adhemerval
>   * Carlos sent DJ a note about the failed to apply not having text for it.
>  * [v3] debug: Fix read error handling in pcprofiledump (Florian)
>   * Needs a v4.
>  * tst: Extend cross-test-ssh.sh to support passing glibc tunables (Yury)
>   * Reviewer Pool: Arjun
>  * [v2] nss: reject invalid port passed to getaddrinfo [BZ #16208] (Antonio)
>   * Reviewer Pool: Carlos
>  * aarch64: Avoid redundant MOVs in AdvSIMD F32 logs (Joe)
>   * Already pushed.
>   * Note that Wilco's response didn't get included in the thread. Need to review why.
>  * [v2] libio: Fix crash in fputws [BZ #20632] (Peter)
>   * Assigned to Florian for review.
>  * Optimization and benchmarking of bsearch() (Kuwan-Wei)
>   * Patch 2 reviewed by Noah Goldstein already.
>   * Need review of patch 1 for the benchmark.
>   * Reviewer Pool: DJ Delorie.

Reviewed benchmarks as well. But a second set of eyes never hurts.
>  * elf: Initialize tunables_env_alias with loop to avoid memset (Jesse)
>   * Patch dropped.
>  * [v3] libio: Set _vtable_offset before calling _IO_link_in [BZ #32148] (HJ)
>   * Reviewer Pool: Florian
>  * Stopping at 97282.
>  * Anything to raise for review?
>  * Adhemerval: Yes, [v4] linux: Add support for getrandom vDSO (96923)
>   * Florian: Surprised that this isn't faster than 2x?
>   * Adhemerval: The larger stressng buffer makes this slower. For a small buffer the speedup is up to 10x speedup.
>   * Discussed adding cancellation to the implementation and that you could internally call with GRND_NONBLOCK first and then fall back to a cancellable syscall.
>   * Does the small size buffer fast path still remain cheap?
>   * Note about vDSO unwinding: https://bugzilla.redhat.com/show_bug.cgi?id=2044483#c3
>  * Florian: The asprintf patch: [v4] libio: asprintf should write NULL upon failure
>   * Carlos: It is in my review queue, but getenv hardening is ahead of this.
>  * Adhemerval: What about the RELRO patch series? "RELRO linkmaps"?
>   * Florian: There is an attack vector around DT_DEBUG and the ld.so link map. If you have memory protection keys it is really fast. You need a memory protection key with signal semantics that work on lazy binding.
>  * Adhemerval: I don't want to block the sched_setattr, sched_getattr addition, just note the tradeoff.
>
> --
> Cheers,
> Carlos.
>