[Bug 2088460] Re: valgrind falsely reports "Invalid read" on ppc64el

[Launchpad Bug Tracker]

This bug was fixed in the package valgrind-if-available -
3.18.1-1-1ubuntu1

---------------
valgrind-if-available (3.18.1-1-1ubuntu1) plucky; urgency=medium

  * Do not build for ppc64el as it produces false "Invalid read" reports if
    32KB buffers are located on the stack (LP: #2088460)

 -- Julian Andres Klode <juliank at ubuntu.com>  Mon, 18 Nov 2024 15:09:05
+0100

** Changed in: valgrind-if-available (Ubuntu)
       Status: Fix Committed => Fix Released

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to valgrind in Ubuntu.
https://bugs.launchpad.net/bugs/2088460

Title:
  valgrind falsely reports "Invalid read" on ppc64el

Status in valgrind package in Ubuntu:
  New
Status in valgrind-if-available package in Ubuntu:
  Fix Released
Status in valgrind package in Debian:
  New

Bug description:
  valgrind has a problem on ppc64el, similar to when stack clash
  protection is enabled on armhf in that a 32 KB buffer on the stack
  causes it to generate incorrect "Invalid read of size 8" (16 in
  Debian):

  442s -==46196== Invalid read of size 8
  442s -==46196==    at 0x4CDF9B0: ExtractTar::Go(pkgDirStream&) (in /usr/lib/powerpc64le-linux-gnu/libapt-pkg.so.6.0.0)
  442s -==46196==    by 0x4D0C823: debDebFile::MemControlExtract::Read(debDebFile&) (in /usr/lib/powerpc64le-linux-gnu/libapt-pkg.so.6.0.0)
  442s -==46196==    by 0x1938EB: ??? (in /usr/bin/apt-ftparchive)
  442s -==46196==    by 0x19B6FF: ??? (in /usr/bin/apt-ftparchive)
  442s -==46196==    by 0x1894D7: ??? (in /usr/bin/apt-ftparchive)
  442s -==46196==    by 0x19AA5B: ??? (in /usr/bin/apt-ftparchive)
  442s -==46196==    by 0x19B20F: ??? (in /usr/bin/apt-ftparchive)
  442s -==46196==    by 0x19549F: ??? (in /usr/bin/apt-ftparchive)
  442s -==46196==    by 0x4CCDFA7: CommandLine::DispatchArg(CommandLine::Dispatch const*, bool) (in /usr/lib/powerpc64le-linux-gnu/libapt-pkg.so.6.0.0)
  442s -==46196==    by 0x491FD4F: DispatchCommandLine(CommandLine&, std::vector<CommandLine::Dispatch, std::allocator<CommandLine::Dispatch> > const&) (in /usr/lib/powerpc64le-linux-gnu/libapt-private.so.0.0.0)
  442s -==46196==    by 0x18762F: ??? (in /usr/bin/apt-ftparchive)
  442s -==46196==    by 0x5356C43: (below main) (libc_start_call_main.h:58)
  442s -==46196==  Address 0x1fff00d820 is on thread 1's stack
  442s -==46196==  336 bytes below stack pointer

  
  This was fixed in APT in

  https://salsa.debian.org/apt-
  team/apt/-/commit/a5d029ea6474db4f7edf8e9b6d73afd2ae583250

  But we have since reached the conclusion in Debian to remove valgrind
  for armhf and ppc64el, hence the workaround has been reverted and the
  test suite changed to depend on valgrind-if-available.

  For Ubuntu, armhf is not affected, and we don't need to patch out the
  valgrind binary itself, but we can just drop ppc64el from the
  valgrind-if-available architecture list.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/valgrind/+bug/2088460/+subscriptions