We have a rather large autofs setup at our site with four different maps/mountpoints and one them being quite large with >1000 entries. However, on any one box rarely will over 20 mounts be used at once. We recently installed a FC2 box and everything works fine until somebody starts up a KDE application. Soon after that we will notice automounting stops working and several NFS mounts are unresponsive. Doing a 'ps' shows several 'mount' and 'umount' process hung in the 'D' state. Looking at the kernel log I find: Badness in interruptible_sleep_on at kernel/sched.c:1927 Call Trace: [<0229fe72>] interruptible_sleep_on+0x5a/0xc6 [<0211b419>] default_wake_function+0x0/0xc [<021783d1>] proc_pid_lookup+0x199/0x1e1 [<f889b4f9>] autofs4_wait+0x1fe/0x25e [autofs4] [<f889a576>] try_to_fill_dentry+0x39/0x100 [autofs4] [<02159950>] do_lookup+0x54/0x72 [<02159f70>] link_path_walk+0x602/0x7d0 [<021616a7>] dput+0x18/0x16c [<0215c557>] vfs_follow_link+0x114/0x177 [<02159d42>] link_path_walk+0x3d4/0x7d0 [<021616a7>] dput+0x18/0x16c [<0215c557>] vfs_follow_link+0x114/0x177 [<02159d42>] link_path_walk+0x3d4/0x7d0 [<0215a42b>] path_lookup+0x13f/0x16f [<0215a567>] __user_walk+0x21/0x51 [<02156177>] vfs_stat+0x14/0x3a [<021566e9>] sys_stat64+0xf/0x23 This message if followed by some nfs server timeouts and then some other identical "Badness" messages depending on how long i wait to reboot. The thing with KDE that I suspect as the most likely culprit is that KDE starts the 'fam' process which does filesystem monitoring. I don't think that 'fam' is actually a critically needed process so I am removing it to do a test. Of course the bug is not necessarily with 'fam', but could be in the kernel or autofs. Just 'fam' is what I suspect triggers it. -- --------------------------------------------------------------- Paul Raines email: raines@xxxxxxxxxxxxxxxxxxx MGH/MIT/HMS Athinoula A. Martinos Center for Biomedical Imaging 149 (2301) 13th Street Charlestown, MA 02129 USA
