This should be fixed in the development cvs tree.
Steve
> Dual-processor Dell with AMI MegaRAID's running SGI XFS 1.0 Red Hat
> 7.1 (2.4.2-SGI_XFS_1.0smp). One of the MegaRAID fs's is xfs and is
> exported to a couple of SGI O200's. Around the same time I did a chown
> and chmod on the XFS fs via NFS, the Linux box reported:
>
> May 30 09:47:12 daacsnfs kernel: xfs_iget_core: ambiguous vns: vp/0xf38e7ab0,
> invp/0xf7067570
> May 30 09:47:12 daacsnfs kernel: Unable to handle kernel NULL pointer derefer
> ence at virtual address 00000008
> May 30 09:47:12 daacsnfs kernel: printing eip:
> May 30 09:47:12 daacsnfs kernel: c01e7892
> May 30 09:47:12 daacsnfs kernel: pgd entry f697b000: 0000000000000000
> May 30 09:47:12 daacsnfs kernel: pmd entry f697b000: 0000000000000000
> May 30 09:47:12 daacsnfs kernel: ... pmd not present!
> May 30 09:47:12 daacsnfs kernel: Oops: 0000
> May 30 09:47:12 daacsnfs kernel: CPU: 1
> May 30 09:47:12 daacsnfs kernel: EIP: 0010:[vn_revalidate+34/232]
> May 30 09:47:12 daacsnfs kernel: EIP: 0010:[<c01e7892>]
> May 30 09:47:12 daacsnfs kernel: EFLAGS: 00010286
> May 30 09:47:12 daacsnfs kernel: eax: 00000084 ebx: f7067570 ecx: f7ba640
> 0 edx: 00000000
> May 30 09:47:12 daacsnfs kernel: esi: f7067570 edi: 00000084 ebp: f706757
> 0 esp: f6b7dc20
> May 30 09:47:12 daacsnfs kernel: ds: 0018 es: 0018 ss: 0018
> May 30 09:47:12 daacsnfs kernel: Process nfsd (pid: 719, stackpage=f6b7d000)
> May 30 09:47:12 daacsnfs kernel: Stack: f7067570 f7067570 f89048a8 00000001 1
> 4003fff f89048c8 ffffffff ffffffff
> May 30 09:47:12 daacsnfs kernel: 00000019 f713fe60 00000002 00000000 0
> 0000001 00000002 00000000 00000000
> May 30 09:47:12 daacsnfs kernel: 00000001 c22ea000 f6b7dc94 00000082 f
> 89048ac 00000246 f89048a8 00000046
> May 30 09:47:12 daacsnfs kernel: Call Trace: [eepro100:__insmod_eepro100_O/li
> b/modules/2.4.2-SGI_XFS_1.0smp/kernel+-329560/96] [eepro100:__insmod_eepro100
> _O/lib/modules/2.4.2-SGI_XFS_1.0smp/kernel+-329528/96] [eepro100:__insmod_eep
> ro100_O/lib/modules/2.4.2-SGI_XFS_1.0smp/kernel+-329556/96] [eepro100:__insmo
> d_eepro100_O/lib/modules/2.4.2-SGI_XFS_1.0smp/kernel+-329560/96] [eepro100:__
> insmod_eepro100_O/lib/modules/2.4.2-SGI_XFS_1.0smp/kernel+-329524/96] [xfs_ig
> et_core+1916/1956] [ip_rcv+926/992]
> May 30 09:47:12 daacsnfs kernel: Call Trace: [<f89048a8>] [<f89048c8>] [<f890
> 48ac>] [<f89048a8>] [<f89048cc>] [<c01bd320>] [<c0258c5e>]
> May 30 09:47:12 daacsnfs kernel: [eepro100:__insmod_eepro100_O/lib/mod
> ules/2.4.2-SGI_XFS_1.0smp/kernel+-329556/96] [xfs_vn_iget+52/60] [vn_initiali
> ze+213/344] [linvfs_read_inode+30/80] [get_new_inode+227/376] [iget4+221/232]
> [<f8981d1d>] [<f89822fd>]
> May 30 09:47:12 daacsnfs kernel: [<f89048ac>] [<c01bd3b8>] [<c01e7649>
> ] [<c01e6b86>] [<c01505f7>] [<c0150945>] [<f8981d1d>] [<f89822fd>]
> May 30 09:47:12 daacsnfs kernel: [<f89827a4>] [<f89838bd>] [<f8983ec0>
> ] [do_ccupdate_local+0/64] [smp_call_function_interrupt+50/88] [csum_partial+
> 185/232] [call_call_function_interrupt+5/12] [csum_partial+185/232]
> May 30 09:47:12 daacsnfs kernel: [<f89827a4>] [<f89838bd>] [<f8983ec0>
> ] [<c012f074>] [<c01134da>] [<c0291a69>] [<c029f12d>] [<c0291a69>]
> May 30 09:47:12 daacsnfs kernel: [<f8989173>] [<f8991480>] [<f89805c3>
> ] [<f8991480>] [<f89656f8>] [<f8991340>] [<f8990e18>] [<f8980369>]
> May 30 09:47:12 daacsnfs kernel: [kernel_thread+35/48]
> May 30 09:47:12 daacsnfs kernel: [<c01075e3>]
> May 30 09:47:12 daacsnfs kernel:
> May 30 09:47:12 daacsnfs kernel: Code: 8b 4a 08 6a 00 25 80 00 00 00 50 8d 44
> 24 18 50 52 8b 41 14
>
> I was able to unmount all the filesystems on the SGI's, but the nfsd's
> on the Linux box wouldn't die. Had to power cycle to get it back
> up. (Why isn't there a way to kill processes in I/O wait?)
>
> Is this a known bug that would be fixed by installing the current CVS'd
> kernel?
>
> -Dave
|