The edited patch set went on with no problems and compiled cleanly; the module works normally.
I rebuilt the daemon with these patches: autofs-5.0.2-dns-name-lookup.patch autofs-5.0.3-dont-fail-on-empty-master-fix-2.patch autofs-5.0.3-mount-thread-create-cond-handling.patch autofs-5.0.3-mount-thread-create-cond-handling-fix.patch autofs-5.0.3-dont-abuse-ap-ghost-field.patch autofs-5.0.3-submount-shutdown-recovery-7.patch At first I thought it was still failing as before, but on looking more closely at the output I see a new behavior which is much better than failing. After 35 minutes and about 1200 mount/unmount cycles, it took over 25 seconds to mount (and readdir) 12 filesystems distributed over 6 machines. 2 minutes later these all had been mounted, and 4 filesystems on 2 other machines were stuck. 7 minutes later these were finished and 5 filesystems on one machine were stuck. This continued (12 groups total) until I stopped the test 30 mins later. Syslog (debug priority) had not a single line from autofs or NFS for the entire period. I think I need to adjust my test program to accomodate this new behavior. I'll report back when that's done. If you want to see 0.5 Mbyte of backtraces from the above test, let me know. James F. Carter Voice 310 825 2897 FAX 310 206 6673 UCLA-Mathnet; 6115 MSA; 520 Portola Plaza; Los Angeles, CA, USA 90095-1555 Email: [EMAIL PROTECTED] http://www.math.ucla.edu/~jimc (q.v. for PGP key) _______________________________________________ autofs mailing list [email protected] http://linux.kernel.org/mailman/listinfo/autofs
