Description of problem: "strace -f" fails to trace child processes. Version-Release number of selected component (if applicable): kernel-2.6.26.3-29.fc9.i686 strace-4.5.16-5.fc9.i386 How reproducible: Every time Steps to Reproduce: 1. strace -f -o strace.out bash -c '(sleep 1)' (The parentheses are needed to get bash to start a subprocess. I used a sleep in order to rule out a race where the child exits before strace can attempt to trace it.) Actual results: strace does not trace the child "sleep" process, and it outputs "PANIC: attached pid # exited" (where # represents the PID of the child) when the one-second sleep ends. Expected results: strace traces the "sleep" process. Additional info: The problem does not occur with kernel-2.6.25.11-92.fc9.i686. That is why I put this bug in the kernel component.
The problem does not occur with vanilla kernel 82a28c794f27aac17d7a3ebd7f14d731a11a5532 .
Possibly a utrace bug.
I can duplicate this problem. This kernel is now in stable.
Upstream kernel is not affected. More precisely: 2.6.27-rc4 is ok (this is the upstream I tested against) 2.6.25.14-108 is ok 2.6.27-0.322.rc6 is bad 2.6.27-0.324.rc6.git1 is bad I am testing some more kernels at the moment, will add info as it becomes available.
Tested x86_64 kernel RPMs from koji, results: 2.6.27-0.144.rc0.git2 is ok 2.6.27-0.166.rc0.git8 is ok intervening kernels had build problems in koji, next successfully built one: 2.6.27-0.173.rc0.git11 is bad: prints "PANIC: attached pid NNNN exited" and strace exits 2.6.27-0.180.rc0.git11 is bad, same symptoms 2.6.27-0.186.rc0.git15 is bad, same symptoms 2.6.27-0.191.rc0.git16 is bad, same symptoms and sometime later symptoms change: 2.6.27-0.322.rc6 is bad: prints "PANIC: attached pid NNNN exited" and strace hangs, I KILLed it 2.6.27-0.324.rc6.git1 is bad, same symptoms as 2.6.27-0.322.rc6 Also tested newer upstream (yesterday's Linus tree) 2.6.27-rc6 and it is ok.
Created attachment 316855 [details] Testcase The bug is that clone(... CLONE_PTRACE ...) doesn't stop with SIGSTOP anymore. This affects strace because it patches clone() syscalls with this flag in order to get a trap just after a clone() in the child, not sometime after it.
Verified clone-ptrace works on vanilla upstream du jour and fails on my utrace development branch. I'm looking into it.
Fixed in utrace branch, in rawhide building kernel-2_6_27-0_372_rc8_fc10
Thanks Roland, just tested this kernel and it fixes bug 461941 too! Bug 461456 might be fixed also (not tested yet). Just FYI: some unrelated change in that kernel interferes with X, at least on my laptop. Created bug 464775.
Verified as still broken on: kernel-2.6.26.6-71.fc9.x86_64 Could you please backport it for F9? (kernel-2.6.25.10-86.fc9.x86_64 has disfunctional arrow keys in VNC/KVM, kernel-2.6.27-3.fc10.x86_64 has X locking up for ~5 seconds on any drawing, there is now no kernel to choose from)
(In reply to comment #10) > Verified as still broken on: kernel-2.6.26.6-71.fc9.x86_64 > Could you please backport it for F9? > Please test kernel-2.6.26.6-79.fc9
I tested kernel-2.6.26.6-79.fc9 and the problem is fixed.
*** Bug 467214 has been marked as a duplicate of this bug. ***
Will this be ported to Fedora 8?
(In reply to comment #14) > Will this be ported to Fedora 8? It already is, in 2.6.26.6-49.fc8