tinydns: allocation failures with glibc 2.38/s390x

[Simon Chopin]

When using glibc 2.38, in some conditions tinydns fails with the following error:

/usr/sbin/tinydns: error while loading shared libraries: libc.so.6: cannot map zero-fill pages

This error shows up at least on the s390x autopkgtests for djbdns.

This is blocking the glibc transition.

[Danilo Egea Gondolfo (danilogondolfo)]

The problem seems to be related to the "run" scripts.

Increasing the data segment limit makes it work:

# envuidgid root envdir ./env softlimit -d 400000 tinydns
tinydns: error while loading shared libraries: libc.so.6: cannot map zero-fill pages

# envuidgid root envdir ./env softlimit -d 4000000 tinydns
starting tinydns

[Danilo Egea Gondolfo (danilogondolfo)]

debian/patches/0011-datalimit.patch probably needs to be updated...

[Simon Chopin (schopin)]

Bumping the limit sounds like a good idea as a first step, but we might
want to dig a bit deeper to see if it's just a straw that broke the
camel's back or if there was instead a dramatic increase to memory
consumption on s390x with this release.

If you prepare it I shall sponsor it ;)

[Danilo Egea Gondolfo (danilogondolfo)]

Just created an MP. I tested it in different archs and autopkgtests are passing. It should unblock glibc.

[Simon Chopin (schopin)]

Tagging this as rls-mm-incoming as we might want to investigate the fairly big memory bump there. Re-opening the glibc task as well, but with a lower priority since we can still get things moving.

[Launchpad Janitor (janitor)]

This bug was fixed in the package djbdns - 1:1.05-15ubuntu1

---------------
djbdns (1:1.05-15ubuntu1) mantic; urgency=medium

  * d/p/0011-datalimit.patch: double the data segment limit. (LP: #2031909)
    The tinydns binary was failing with "libc.so.6: cannot map zero-fill
    pages". The new value was determined empirically. Slightly increasing it
    will cause a different issue ("cannot allocate TLS data structures for
    initial thread").

 -- Danilo Egea Gondolfo <email address hidden> Tue, 22 Aug 2023 10:01:34 +0100

[Danilo Egea Gondolfo (danilogondolfo)]

One more piece of information: I tried increasing it to 600k and got "cannot allocate TLS data structures for initial thread". Then I tried 700k and it worked, but I decided to change it to 800k for safety.

[Florian Weimer (fweimer)]

I think Adam Jackson found the root cause in glibc:

[PATCH] libio: Fix oversized __io_vtables
<https://<email address hidden>/T/#u>

[Launchpad Janitor (janitor)]

This bug was fixed in the package glibc - 2.38-1ubuntu5

---------------
glibc (2.38-1ubuntu5) mantic; urgency=medium

  * Update from upstream release branche:
    - CVE-2023-4527: Stack read overflow with large TCP responses in
      no-aaaa mode
    - CVE-2023-4806: use after free in getcanonname
    - LP: #2031909: Fix oversized __io_vtables
  * d/p/u/0001-Fix-leak-in-getaddrinfo-introduced-by-the-fix-for-CV:
    Cherry-picked to fix a regression in one of the previous CVE fixes
    (LP: #2037516, CVE-2023-5156)
  * d/p/lp2032624.patch: add an escape hatch in arm64 math-vector.h.
    This should help fixing multiple FTBFS (LP: #2032624)

 -- Simon Chopin <email address hidden> Wed, 27 Sep 2023 16:38:18 +0200