Kernel hang: unable to handle kernel paging request at abcddcba when trying to modprobe zfs.ko

[robinsmidsrod]

Using Ubuntu 10.04 LTS on 32-bit hardware. 3.5GB RAM. ZFS installed from PPA. I was trying to figure out if this was a duplicate to some of the other kernel hang bug reports, but I couldn't see any similar traces. Machine has been running stable for quite some time, so hardware error is somewhat unlikely. This bug hangs the bootup process, and only way to fix the problem is to boot with init=/bin/bash and remove all the dkms-installed kernel modules and rebuild the initrd.

[ 2497.470188] BUG: unable to handle kernel paging request at abcddcba
[ 2497.470791] IP: [<abcddcba>] 0xabcddcba
[ 2497.474111] *pdpt = 000000003690e001 *pde = 0000000000000000
[ 2497.474111] Oops: 0000 [#1] SMP
[ 2497.474111] last sysfs file: /sys/devices/system/cpu/cpu3/topology/core_siblings
[ 2497.474111] Modules linked in: spl(+) zlib_deflate iscsi_trgt crc32c nfs lockd nfs_acl auth_rpcgss sunrpc autofs4 ib_iser rdma_cm ib_cm iw_cm ib_sa ib_mad ib_core ib_addr iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi fbcon scb2_flash mtd tileblit ppdev chipreg font map_funcs bitblit parport_pc i2c_piix4 softcursor psmouse vga16fb vgastate serio_raw lp parport raid10 raid456 async_raid6_recov async_pq raid6_pq async_xor xor async_memcpy async_tx raid1 raid0 multipath linear vesafb floppy pata_serverworks tg3 cciss sworks_agp agpgart ramzswap xvmalloc lzo_decompress lzo_compress
[ 2497.474111]
[ 2497.474111] Pid: 3790, comm: modprobe Not tainted (2.6.32-41-generic-pae #94-Ubuntu) ProLiant ML350 G3
[ 2497.474111] EIP: 0060:[<abcddcba>] EFLAGS: 00010282 CPU: 3
[ 2497.474111] EIP is at 0xabcddcba
[ 2497.474111] EAX: f8704ade EBX: ee0452f0 ECX: 00000001 EDX: 00000286
[ 2497.474111] ESI: f8709320 EDI: 00000000 EBP: f692def8 ESP: f692def0
[ 2497.474111]  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
[ 2497.474111] Process modprobe (pid: 3790, ti=f692c000 task=edeebfc0 task.ti=f692c000)
[ 2497.474111] Stack:
[ 2497.474111]  f86f6806 ee0452f0 f692df5c f873a6ac 00000000 00000000 00000244 00000000
[ 2497.474111] <0> 00000000 f8709320 f692df28 00000001 00000000 c07b0d20 00000000 f8704d39
[ 2497.474111] <0> f8704d40 f8705ea0 00000000 f8704d2e f8704d36 f8705ef4 00000000 00000004
[ 2497.474111] Call Trace:
[ 2497.474111]  [<f86f6806>] ? spl_kmem_init_kallsyms_lookup+0x16/0x150 [spl]
[ 2497.474111]  [<f873a6ac>] ? spl_init+0x6ac/0x8e9 [spl]
[ 2497.474111]  [<c0103041>] ? do_one_initcall+0x31/0x190
[ 2497.474111]  [<f873a000>] ? spl_init+0x0/0x8e9 [spl]
[ 2497.474111]  [<c018baf0>] ? sys_init_module+0xb0/0x210
[ 2497.474111]  [<c01097cc>] ? syscall_call+0x7/0xb
[ 2497.474111]  [<c05b0000>] ? early_init_intel+0x5f/0x15f
[ 2497.474111] Code:  Bad EIP value.
[ 2497.474111] EIP: [<abcddcba>] 0xabcddcba SS:ESP 0068:f692def0
[ 2497.474111] CR2: 00000000abcddcba
[ 2498.000739] ---[ end trace c88a1acb48a77dc0 ]---

[ryao]

ZFSOnLinux will build on i686 hardware, but unfortunately, it does not work very well on them. The only ZFS implementation known to work well on i686 is in FreeBSD.

[robinsmidsrod]

I actually have Debian kFreeBSD running on another similar box (32-bit as well), and to be honest I don't think that implementation is production level quality either. I've seen it kernel panic several times because of out-of-memory when it should be able to organize its own use of memory when it is the only thing running on the machine (it's a file server).

I guess I should try it on a 64-bit capable machine to see how it behaves there.

[behlendorf]

Indeed. Please stick to the 64-bit implementation, I should really add a larger warning to the zfs website and perhaps a warning message at build time as well.

[behlendorf]

I was able to reproduce the issue under Lucid, it's fixed by commit openzfs/spl@034f1b3 which was just merged in to master.

[behlendorf]

Oh and it was NOT a 32-bit issue after all.

[robinsmidsrod]

Is this fix included in the current stable PPA release for Lucid? If not, when it is expected to be included?

[dajhorn]

Is this fix included in the current stable PPA release for Lucid?

No.

If not, when it is expected to be included?

The daily PPA will probably get it Friday evening. The stable PPA will get it after the next release candidate is tagged.