From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4) Gecko/20030709 Description of problem: When trying to mount an exported filesystem from a s390x the rpc.mountd runs amok. It doesn't mount the filesystem and rpc.mountd runs at 100% cpu time and loops in svc_run.c:my_svc_run() line 71. Strace on the process continously shows this: select(1024, [3 4], NULL, NULL, NULL) = 1 (in [4]) I suspect is has something to do with either the ACL or the cache patches for nfs-utils. nfs-utils-1.0.3-6 works fine btw. I already looked at the diff between the 2 versions, but the changes are quite big. I'll be looking into the problem today and hopefully come up witha fixed version later today. Will attach my results here. Read ya, Phil Version-Release number of selected component (if applicable): nfs-utils-1.0.5-2 How reproducible: Always Steps to Reproduce: 1. Export a filesystem 2. Start nfs service 3. Try to mount or showmount the filesystem 4. Watch rpc.mountd eat up a whole CPU. :-) Actual Results: rpc.mountd runs amok and filesystem doesn't get mounted. Expected Results: rpc.mountd shouldn't run amok and i should be able to mount the fs. :-) Additional info:
Does ethereal or tcpdump show anything going over the wire?
Not sure, didn't check that. I tried mounting from the localhost, so i'd need to tcpdump lo0 i guess. Might do that if i'm stuck. I already narrowed it down to the new my_svc_run() function. If you use the default svc_run() in mountd.c it works again, but i guess that then the new auth methods aren't being used for ACL etc. Read ya, Phil
OK, probably found the root of the problem. It's actually not a problem with nfs-utils itself but with a function in glibc. I've fixed the same bug about 1 year ago and submitted the fix upstream, but obviously it got lost somewhere on the way. Will fix glibc tonight and send a patch to Jakub and Ulrich for review. Read ya, Phil
OK, fixed glibc fixes the problem. Patch will be included in one of the next glibc builds, it has been included upstream now. Read ya, Phil