Description of problem: Getting a segmentation fault when trying to attach to a process on ppc64le. Only happens when attaching, running '$ strace cmd' works fine. Version-Release number of selected component: strace-4.8-5.fc21 Additional info: reporter: libreport-2.2.3 backtrace_rating: 4 cmdline: strace -p 3250 crash_function: init executable: /usr/bin/strace kernel: 3.14.0-1.fc21.ppc64le runlevel: N 5 type: CCpp uid: 1002 Truncated backtrace: Thread no. 1 (3 frames) #0 newoutf at strace.c:644 #1 startup_attach at strace.c:976 #2 init at strace.c:1836
Created attachment 926313 [details] File: backtrace
Created attachment 926314 [details] File: cgroup
Created attachment 926315 [details] File: core_backtrace
Created attachment 926316 [details] File: dso_list
Created attachment 926317 [details] File: environ
Created attachment 926318 [details] File: exploitable
Created attachment 926319 [details] File: limits
Created attachment 926320 [details] File: maps
Created attachment 926321 [details] File: open_fds
Created attachment 926322 [details] File: proc_pid_status
Created attachment 926323 [details] File: var_log_messages
Could you check whether upstream version from git://git.code.sf.net/p/strace/code passes "make check" on that platform, please?
The testsuite passes. The strace binary built within that tree does not have the problem - no segfault when attaching to system processes.
(In reply to Michal Toman from comment #13) > The testsuite passes. The strace binary built within that tree does not have > the problem - no segfault when attaching to system processes. Thanks for testing. So this bug will be fixed in strace-4.9 which I'm going to release soon.
note that this is very similar to an opensuse bug https://bugzilla.novell.com/show_bug.cgi?id=865151 that pointed to same error and was solved by upstream commit https://sourceforge.net/p/strace/code/ci/03ef0b2665363adf6100bc4af479a14b5f17bc27/
Created attachment 926378 [details] strace-ppc-ptrace-proto.patch The commit fixes the problem. Since the upstream patch does not apply to Fedora dist git, attaching a modified version that does.
(In reply to Dmitry V. Levin from comment #14) > (In reply to Michal Toman from comment #13) > > The testsuite passes. The strace binary built within that tree does not have > > the problem - no segfault when attaching to system processes. > > Thanks for testing. So this bug will be fixed in strace-4.9 which I'm going > to release soon. Dmitry, do you have any guess when it might happen?
(In reply to Dan Horák from comment #17) > (In reply to Dmitry V. Levin from comment #14) > > (In reply to Michal Toman from comment #13) > > > The testsuite passes. The strace binary built within that tree does not have > > > the problem - no segfault when attaching to system processes. > > > > Thanks for testing. So this bug will be fixed in strace-4.9 which I'm going > > to release soon. > > Dmitry, do you have any guess when it might happen? I hope it'll happen this week.