Automount daemon hangs when multiple processes attempt to simultaneously access the same nonexistant, automounted filesystem. perl -e 'fork; foreach(0..100){print "$$ $_\n";stat("/foo/bar");}' will hang, where /foo has an associated map of the form * -rw,rsize=8192,wsize=8192,noquota,bg,hard,intr,retrans=6 foo:/foo/& and foo:/foo/bar is not exported.
Please upgrade to the latest version of autofs (3.1.4-3) from Raw Hide (ftp://ftp.redhat.com/pub/rawhide/i386/RedHat/RPMS/), which may already have fixed the problem (6.0 shipped with autofs 3.1.3). However, it may be a kernel problem. If you can use 'ps' to find the PID of the hung automount process, try running "strace -p <PID>" to see if it's hung waiting on the kernel or something else.
Autofs 3.1.4 seems to have the the same problem. PS shows two automount daemon processes for the mountpoint. One is the parent process, and seems to be ok. The second process is a child, and is hung in a call to mkdir.
Then I suspect a race condition in the kernel. Looks like it's in fs/autofs/root.c somewhere. Reassigning to the kernel package. You may want to re-post that last part to the autofs mailing list at autofs.org, because this looks like a non-distribution- specific problem to me.
A race in the kernel was indeed fixed - I suspect the 6.2 or 6.2 errata kernel will be solid
Closing due to old age.