Re: BUG when doing parallel NFS mounts (WAS: Re: Merge window closed: v2.6.19-rc1)

[Peter Osterlund <petero2@telia.com>]

Peter Osterlund <petero2@telia.com> writes:

> Linus Torvalds <torvalds@osdl.org> writes:
> 
> > so please give it a good testing, and let's see if there are any 
> > regressions.
> 
> If I run this one-liner script:
> 
> while true ; do echo -n . ; mount p4:/home /p4home -ohard,intr & mount p4:/extra /p4extra -ohard,intr ; umount /p4home ; umount /p4extra ; done
> 
> I get a BUG after a few iterations, see below. If I replace the '&' with
> a ';' so the mounts happen one at a time, it works as expected.
> 
> The NFS server runs kernel 2.6.17.11.
> 
> ------------[ cut here ]------------
> kernel BUG at fs/nfs/client.c:352!
> invalid opcode: 0000 [#1]
> PREEMPT
> Modules linked in: nfs radeon nfsd exportfs lockd autofs4 sunrpc michael_mic arc4 ecb blkcipher cryptomgr crypto_algapi ieee80211_crypt_tkip binfmt_misc dm_mirror dm_mod lp parport_pc parport pcmcia ehci_hcd joydev ohci1394 ieee1394 ohci_hcd bcm43xx yenta_socket snd_atiixp_modem rsrc_nonstatic ide_cd ieee80211softmac ieee80211 serio_raw 8250_pci pcmcia_core 8250 ieee80211_crypt pcspkr 8139too serial_core cdrom rtc usbcore
> CPU:    0
> EIP:    0060:[<e0c836b5>]    Not tainted VLI
> EFLAGS: 00010202   (2.6.19-rc1 #53)
> EIP is at nfs_create_server+0x9f1/0xa6e [nfs]
> eax: 00000000   ebx: d5e11b40   ecx: ca01a000   edx: 00000000
> esi: 00000001   edi: ca01bdac   ebp: ca01bd08   esp: ca01bbf8
> ds: 007b   es: 007b   ss: 0068
> Process mount (pid: 4052, ti=ca01a000 task=ca078030 task.ti=ca01a000)
> Stack: ca01bc58 c1405c68 ca01bc2c c1405c40 c1405c40 00000296 c05cc5d0 ca078030
>        ca07857c d6b03380 00000003 ca97704c ca97705c ca078030 00000000 ca01bc8c
>        c0135211 c01eea52 ca078570 00000246 ca01bc6c 00000246 00000002 ca845cb0
> Call Trace:
>  [<e0c899b2>] nfs_get_sb+0xbb/0x531 [nfs]
>  [<c0166416>] vfs_kern_mount+0x42/0x8c
>  [<c01664c4>] do_kern_mount+0x3b/0x4e
>  [<c017a8ca>] do_mount+0x25a/0x6a5
>  [<c017adaa>] sys_mount+0x95/0xd4
>  [<c01030c1>] sysenter_past_esp+0x56/0x8d
>  [<b7fa9410>] 0xb7fa9410
>  =======================
> Code: 85 28 ff ff ff 89 44 24 04 c7 04 24 00 61 c9 e0 e8 16 8e 49 df 8b 9d 28 ff ff ff 31 d2 81 fb 00 f0 ff ff 0f 97 c2 e9 62 f9 ff ff <0f> 0b 60 01 40 50 c9 e0 e9 34 f9 ff ff 81 fa 60 ea 00 00 0f 86
> EIP: [<e0c836b5>] nfs_create_server+0x9f1/0xa6e [nfs] SS:ESP 0068:ca01bbf8

According to "git bisect", this is the bad commit:

commit 54ceac4515986030c2502960be620198dd8fe25b
Author: David Howells <dhowells@redhat.com>
Date:   Tue Aug 22 20:06:13 2006 -0400

    NFS: Share NFS superblocks per-protocol per-server per-FSID

    The attached patch makes NFS share superblocks between mounts from the same
    server and FSID over the same protocol.

-- 
Peter Osterlund - petero2@telia.com
http://web.telia.com/~u89404340