Bug #22253
closed"rbd info" crashed: stack smashing detected
0%
Description
Environment: quite small vstart cluster.
This is the stack trace:
#3 0x00007fffed44711c in __GI___fortify_fail (msg=<optimized out>, msg@entry=0x7fffed4bd441 "stack smashing detected") at fortify_fail.c:37 #4 0x00007fffed4470c0 in __stack_chk_fail () at stack_chk_fail.c:28 #5 0x00007ffff78f0beb in librbd::ImageCtx::perf_start (this=this@entry=0x555555b7bf70, name="librbd-8c39e2ae8944a-rbd-huge2") at /home/sebastian/Repos/ceph/src/librbd/ImageCtx.cc:397 #6 0x00007ffff78f3cb4 in librbd::ImageCtx::init (this=0x555555b7bf70) at /home/sebastian/Repos/ceph/src/librbd/ImageCtx.cc:275 #7 0x00007ffff799dacd in librbd::image::OpenRequest<librbd::ImageCtx>::send_register_watch (this=this@entry=0x555555b7fe60) at /home/sebastian/Repos/ceph/src/librbd/image/OpenRequest.cc:477 #8 0x00007ffff79a3102 in librbd::image::OpenRequest<librbd::ImageCtx>::handle_v2_apply_metadata (this=this@entry=0x555555b7fe60, result=result@entry=0x7fffb77fa374) at /home/sebastian/Repos/ceph/src/librbd/image/OpenRequest.cc:471 #9 0x00007ffff79a351f in librbd::util::detail::rados_state_callback<librbd::image::OpenRequest<librbd::ImageCtx>, &librbd::image::OpenRequest<librbd::ImageCtx>::handle_v2_apply_metadata, true> (c=<optimized out>, arg=0x555555b7fe60) at /home/sebastian/Repos/ceph/src/librbd/Utils.h:39 #10 0x00007ffff75d678d in librados::C_AioComplete::finish (this=0x7fffd0001b60, r=<optimized out>) at /home/sebastian/Repos/ceph/src/librados/AioCompletionImpl.h:169 #11 0x0000555555613949 in Context::complete (this=0x7fffd0001b60, r=<optimized out>) at /home/sebastian/Repos/ceph/src/include/Context.h:70 #12 0x00007fffeeab6010 in Finisher::finisher_thread_entry (this=0x555555acb3e8) at /home/sebastian/Repos/ceph/src/common/Finisher.cc:72 #13 0x00007fffee3a86ba in start_thread (arg=0x7fffb77fe700) at pthread_create.c:333 #14 0x00007fffed4353dd in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:109
Files
Updated by Sebastian Wagner over 6 years ago
my correct version number is:
git describe $ v12.2.0-1124-g5e519ae
Updated by Jason Dillaman over 6 years ago
- Status changed from New to Need More Info
@Sebastian I.: please retest on the latest available version. Your line numbers do not align with v12.2.0.
Updated by Sebastian Wagner over 6 years ago
I don't think it is easy to reproduce it, because
- That RBD is untouched, thus no data was ever written to this RBD, no snapshots were created, it wasn't resized, etc.
- I had restarted the cluster many times since the creation of this RBD.
Updated by Jason Dillaman over 6 years ago
@Sebastian I.: without line numbers that actually align to the code at the mentioned version, there really isn't much I can do to assist. Perhaps attempt to run it through "valgrind --tool=memcheck rbd XYZ".
Updated by Sebastian Wagner over 6 years ago
- File valgrind_out.log valgrind_out.log added
Added valgrind output.
@Jason Borden, should I recompile and retest on the latest luminous branch or on v12.2.1?
Updated by Jason Dillaman over 6 years ago
@Sebastian I.: that Valgrind output doesn't help since it failed on an "unknown instruction" error. Can you reproduce on distro or Ceph-provided packages instead of your home-grown build?
Updated by Sebastian Wagner over 6 years ago
Jason Dillaman wrote:
Can you reproduce on distro or Ceph-provided packages instead of your home-grown build?
If there is a documentation of how to run a cluster created by vstart.sh with distro packages, I'd give it a try.
Updated by Jason Dillaman over 6 years ago
@Sebastian I.: vstart is for development. Just install the Ceph client packages on a VM, copy the vstart-generated ceph.conf to that host, and retest.
Updated by Sebastian Wagner over 6 years ago
Jason Dillaman wrote:
@Sebastian I.: vstart is for development. Just install the Ceph client packages on a VM, copy the vstart-generated ceph.conf to that host, and retest.
Ok, but I also have to make sure, daemons are using the existing data, e.g OSDs.
Updated by Jason Dillaman over 6 years ago
@Sebastian I.: yup, that's why you would copy the "ceph.conf" so that the VM can connect to your vstart-created cluster.
Updated by Sebastian Wagner over 6 years ago
So, I recompiled 12.2.1 and can no longer reproduce this one. seems to be gone now.
Updated by Jason Dillaman over 6 years ago
- Status changed from Need More Info to Can't reproduce