It is possible to render a system with utrace support unusable in ~10 seconds
(but not mainline kernel where Ctrl+C will kill process).
Created attachment 149446 [details]
fix for utrace/ptrace leak and crash bugs with MT exec
This fixes both OOM and BUG_ON failure modes and another crash failure mode
from the same problem. I can't reproduce any problems in the test case after
this patch, where I got all three failure modes before.
Created attachment 149481 [details]
modified test program
Here is the version of the test I have been using.
It prints out some more info, but not so much that it perturbed the behavior
Fedora bug 227952 is the same problems as this.
Created attachment 149546 [details]
fix for utrace/ptrace leak and crash bugs with MT exec + utrace_attach spin failure mode
This includes an additional fix that I think resolves a failure mode where
utrace_attach spins when you try to kill the test with SIGKILL.
A patch for this issue has been included in zstream build 2.6.18-8.1.2.el5.
fix verified on an x86_64 test system, I had to run 3 concurrent copies of
ptrace_thrash to reproduce the problem initially.
An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on the solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.