Re: valgrind memcheck + drd error on alpine 3.6 container

[Rich Felker <dalias@libc.org>]

On Thu, Jun 15, 2017 at 08:28:41AM +0300, Timo Teras wrote:
> On Wed, 14 Jun 2017 20:16:26 -0400
> Rich Felker <dalias@libc.org> wrote:
> 
> > On Wed, Jun 14, 2017 at 07:51:30PM +0200, Eugenio Pérez wrote:
> > > I'm having an issue trying to use valgrind drd in the simplest
> > > musl-linked program (just return 0).
> > > 
> > > drd refuses to even run main, and give me this error:
> > > 
> > > ==24== drd, a thread error detector
> > > ==24== Copyright (C) 2006-2015, and GNU GPL'd, by Bart Van Assche.
> > > ==24== Using Valgrind-3.12.0 and LibVEX; rerun with -h for
> > > copyright info ==24== Command: ./f2k
> > > ==24==
> > > 
> > > drd: drd_malloc_wrappers.c:115 (handle_free): Assertion 'success'
> > > failed.
> > > 
> > > host stacktrace:
> > > ==24==    at 0x3805EF1D: show_sched_status_wrk (m_libcassert.c:343)
> > > ==24==    by 0x3805F208: report_and_quit (m_libcassert.c:419)
> > > ==24==    by 0x3805F3E9: vgPlain_assert_fail (m_libcassert.c:485)
> > > ==24==    by 0x38057635: handle_free (drd_malloc_wrappers.c:115)
> > > ==24==    by 0x380A51B9: do_client_request (scheduler.c:1861)
> > > ==24==    by 0x380A51B9: vgPlain_scheduler (scheduler.c:1425)
> > > ==24==    by 0x380B25DA: thread_wrapper (syswrap-linux.c:103)
> > > ==24==    by 0x380B25DA: run_a_thread_NORETURN (syswrap-linux.c:156)
> > > 
> > > sched status:
> > >   running_tid=1
> > > 
> > > Thread 1: status = VgTs_Runnable (lwpid 24)
> > > ==24==    at 0x4C96951: free (vg_replace_malloc.c:530)
> > > ==24==    by 0x4057A19: ??? (in /lib/ld-musl-x86_64.so.1)
> > > 
> > > I would like to compare it with glibc, but I'm unable to use it in
> > > alpine linux properly. If it helps, memcheck gives me  this error:
> > > 
> > > ==163== Invalid free() / delete / delete[] / realloc()
> > > ==163==    at 0x4C939EA: free (vg_replace_malloc.c:530)
> > > ==163==    by 0x4057A19: ??? (in /lib/ld-musl-x86_64.so.1)
> > > ==163==  Address 0x4e9b180 is in a rw- mapped file
> > > /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so segment
> > > ==163==
> > > ==163==
> > > ==163== HEAP SUMMARY:
> > > ==163==     in use at exit: 404 bytes in 1 blocks
> > > ==163==   total heap usage: 1 allocs, 1 frees, 404 bytes allocated
> > > ==163==
> > > ==163== 404 bytes in 1 blocks are still reachable in loss record 1
> > > of 1 ==163==    at 0x4C9461F: calloc (vg_replace_malloc.c:711)
> > > ==163==    by 0x4058B45: ??? (in /lib/ld-musl-x86_64.so.1)
> > > ==163==    by 0x4059774: __dls3 (in /lib/ld-musl-x86_64.so.1)
> > > ==163==    by 0xFFF000CB7: ???
> > > ==163==    by 0xFFF000D27: ???
> > > ==163==
> > > ==163== LEAK SUMMARY:
> > > ==163==    definitely lost: 0 bytes in 0 blocks
> > > ==163==    indirectly lost: 0 bytes in 0 blocks
> > > ==163==      possibly lost: 0 bytes in 0 blocks
> > > ==163==    still reachable: 404 bytes in 1 blocks
> > > ==163==         suppressed: 0 bytes in 0 blocks
> > > 
> > > And that is before even reach main() function! However, memchecks
> > > continue after error; drd does not. ¿What could I do to keep
> > > diagnosing this error?
> > 
> > It's not an error, just lack of a suppressions file. You can ignore it
> > or figure out how to write the suppressions file, or get one from
> > somewhere (maybe a distro) that's already done it for musl.
> 
> Could you please re-consider applying the patch that moves the "donate
> memory" part to malloc.c, and stops misusing free() like this. It
> really helps readability, maintainability and modularity to not
> hardcode malloc internals in dynlink.c. I think I had sent a patch for
> this in IRC, but could not found it immediately.

I think that's a good idea, but I don't have it in front of me. It
does need to be checked very carefully for correctness though. I
remember there were subtle bugs in one version which is why I was
hesitant to go forward without time to spend checking it in detail.

Rich