Description of problem: POSIX specifies that Version-Release number of selected component (if applicable): 2.6.9-rc2 How reproducible: 100% Steps to Reproduce: 1. Compile attached program with -lpthread. 2. Run it. See it writes 3 bytes. 3. Realize it should have died with SIGXFSZ. Actual results: no SIGXFSZ Expected results: killed by SIGXFSZ Additional info: Fix on the way.
Created attachment 104249 [details] test program for multithreaded behavior of setrlimit gcc -g -o mt-rlimit mt-rlimit.c -lpthread Run with some args, demonstrates the single-threaded case and dies with SIGXFSZ. Run with no args, demonstrates the multi-threaded case and only dies with SIGXFSZ if the kernel is fixed.
Created attachment 104251 [details] patch vs 2.6.9-rc2-bk9 to fix the behavior I've submitted this patch upstream, waiting for feedback.
This patch has gone in upstream, after 2.6.9; hopefully in 2.6.10 release.
2.6.10 has this fixed. Only the RLIMIT_CPU behavior is not quite correct, everything else should be right now.