#162 Sometimes onnode fails with error 95

open
nobody
Filesystem (49)
5
2008-06-11
2008-06-11
John Hughes
No

$ onnode 8 pwd
can't execute pwd, errno=95.

In the syslog on node 8 I find:

rmtfb_getcli_id: non-exportable filesystem (devpts)

Weird.

When a node is in this state it stays like that until it's rebooted.

Currently being seen on my 2.6.12 based system. Don't know whether same problem exists on 2.6.11

Discussion

  • John Hughes
    John Hughes
    2008-06-12

    Logged In: YES
    user_id=166336
    Originator: YES

    So the code in rmtfb_getcli_id is:

    [...]
    file = reop_make_file(path, flags, svr);
    [...]
    if (file->private_data) {
    printk(KERN_ERR
    "rmtfb_getcli_id: non-exportable filesystem (%s)\n",
    file->f_vfsmnt->mnt_sb->s_type->name);

    "path" is presumably "/dev/pts/xxx".

     
  • Roger Tsang
    Roger Tsang
    2008-06-15

    Logged In: YES
    user_id=1246761
    Originator: NO

    Haven't seen this on 2.6.11-ssi but could be a symptom of vproc_release() bug 1842982.

     
  • Logged In: NO

    Some more things happened to this cluster:

    On a non-init node:

    rmtfb_getcli_id: non-exportable filesystem (cfs)
    rmtfb_getcli_id: non-exportable filesystem (cfs)
    rmtfb_badops: stale client structure: f74079e0
    rmtfb_badops: stale client structure: f74079e0
    rmtfb_badops: stale client structure: f77aaba0

    On the init node:

    rmtfb_getcli_id: non-exportable filesystem (devpts)
    Node 8 has gone down!!!
    rmtfb_badops: stale client structure: d5ff3980
    rmtfb_badops: stale client structure: d25a4520
    nm_add_node: Node 8 added
    Badness in kref_get at lib/kref.c:32
    [<c0107a3e>] dump_stack+0x1e/0x20
    [<c02b73d5>] kref_get+0x45/0x50
    [<c02b6949>] kobject_get+0x19/0x20
    [<c017c326>] cdev_get+0x36/0x70
    [<c017c611>] exact_lock+0x11/0x20
    [<c03362e6>] kobj_lookup+0x106/0x1b0
    [<c017c4b8>] chrdev_open+0x108/0x1a0
    [<c017080a>] dentry_open+0x17a/0x260
    [<c017068a>] filp_open+0x5a/0x60
    [<c0170ade>] sys_open+0x5e/0xa0
    [<c0106a41>] syscall_call+0x7/0xb
    Badness in kref_get at lib/kref.c:32

    Yow!