aptitude crashes qemu-m68k with handle_cpu_signal received signal outside vCPU context

When building a package with sbuild on Debian, sbuild can use aptitude to resolve dependencies.

Recently, some changes introduced to aptitude or related packages cause qemu to crash:

(sid-m68k-sbuild)root@nofan:/# aptitude -y --without-recommends -o Dpkg::Options::=--force-confold -o Aptitude::CmdLine::Ignore-Trust-Violations=false -o Aptitude::ProblemResolver::StepScore=100 -o Aptitude::ProblemResolver::SolutionCost="safety, priority, non-default-versions" -o Aptitude::ProblemResolver::Hints::KeepDummy="reject sbuild-build-depends-core-dummy :UNINST" -o Aptitude::ProblemResolver::Keep-All-Level=55000 -o Aptitude::ProblemResolver::Remove-Essential-Level=maximum install vim
Warning: Invalid locale (please review locale settings, this might lead to problems later):
  locale::facet::_S_create_c_locale name not valid
The following NEW packages will be installed:
  libgpm2{a} vim vim-common{a} vim-runtime{a} xxd{a} 
0 packages upgraded, 5 newly installed, 0 to remove and 1 not upgraded.
Need to get 7225 kB/7260 kB of archives. After unpacking 33.5 MB will be used.
qemu:handle_cpu_signal received signal outside vCPU context @ pc=0x6019d1bf
qemu:handle_cpu_signal received signal outside vCPU context @ pc=0x601b64ab
Segmentation fault
(sid-m68k-sbuild)root@nofan:/#

The crash does not reproduce on real hardware running Debian unstable.

It seems it crashes during futex syscall:

...
[pid     4] getpid()                    = 4
[pid     4] tgkill(4, 24, SIGRT_1)      = 0
[pid    24] <... futex resumed> )       = ? ERESTARTSYS (To be restarted if SA_RESTART is set)
[pid    24] --- SIGRT_1 {si_signo=SIGRT_1, si_code=SI_TKILL, si_pid=4, si_uid=0} ---
[pid     4] futex(0x7f77abb4f610, FUTEX_WAIT_PRIVATE, 16777216, NULL <unfinished ...>
[pid    24] getpid()                    = 4
[pid    24] --- SIGSEGV {si_signo=SIGSEGV, si_code=SEGV_MAPERR, si_addr=0x10} ---
...

On 2/15/19 1:47 PM, Laurent Vivier wrote:
> It seems it crashes during futex syscall:
> 
> ...
> [pid     4] getpid()                    = 4
> [pid     4] tgkill(4, 24, SIGRT_1)      = 0
> [pid    24] <... futex resumed> )       = ? ERESTARTSYS (To be restarted if SA_RESTART is set)
> [pid    24] --- SIGRT_1 {si_signo=SIGRT_1, si_code=SI_TKILL, si_pid=4, si_uid=0} ---
> [pid     4] futex(0x7f77abb4f610, FUTEX_WAIT_PRIVATE, 16777216, NULL <unfinished ...>
> [pid    24] getpid()                    = 4
> [pid    24] --- SIGSEGV {si_signo=SIGSEGV, si_code=SEGV_MAPERR, si_addr=0x10} ---
> ...

The crash also reproduces with qemu-sh4, so it's not specific to m68k.


I think this is probably a duplicate of bug #1594394, in which case it seems to be fixed in 5.0.0+.

John, can you still reproduce it with the latest version of QEMU?

Just verified it with a very recently compiled version of QEMU from git master and, indeed, the bug seems to be fixed as I can no longer reproduce the crash. The command executes correctly.

I guess it's safe to mark this as fixed.