From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040115 Description of problem: Running mkisofs, with option '-h' results in a machine hang. Input and can be either from NFS or local disk. Input has to be at least abt. 50 MB large. Typically, when 95 to 99% is done (as says the -v option), the program stalls, and consumes 100% CPU. No extra memory seems to be consumed. A few seconds later, the machine grinds to a halt completely. No swapping noise, no disk io. It does not happen always, but definitely more than 90% of the time. Without option '-h', all works perfectly. An strace reported a 'munmap' as the last syscall. Version-Release number of selected component (if applicable): kernel-smp-2.6.5-1.358 and mkisofs-2.01-0.a27.3 How reproducible: Sometimes Steps to Reproduce: 1. Choose a directory tree with abt. 50 MB or more. 2. mkisofs -v -h -o rubbish.iso <this_directory_tree> 3. Wait for 95% finished. 4. Hang... Actual Results: The machine uses 100% CPU (Which I could see running it on an SMP). This lasts for a short time (seconds). Then there's a complete freeze. Expected Results: Mkisofs should have exited normally. Or if need be crash. But should never trigger a machine hang. Additional info:
Update: with kernel-2.6.6-1.406, I can no longer trigger it; but in SMP with kernel-smp-2.6.6-1.406 it is still present.
Update: kernel-2.6.6-1.422 seems not to have it. kernel-smp-2.6.6-1.422 in SMP still has it, though slightly differently: the mkisofs-process goes into 100% system, is un-interrruptible; console of machine is dead, but remote access to the machine is okay. Hard reboot is still necessary though.
We're at 2.6.6-1.427smp now, with mkisofs-2.01-0.a27.4. Still getting a 100% CPU un-interruptible proces. Not getting any wiser from the /proc info for that process neither.
A collegue of mine figured out that the 4G/4G patch is the culprit: without it, no problems.
we fixed a rare cornercase in the 4g/4g patch in 435 kernel (out as erratum now) that may explain this....
Yap, problem fixed in 2.6.6-1.435smp, thanks!