Re: [kvm-ppc-devel] [PATCH] kvm(ppc)-userspace: initialize

[Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>]

ehrhardt@linux.vnet.ibm.com wrote:
> From: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
> 
> virtio-block support is atm initialized in qemu/hw/pc.c which is only called
> for the "standart pc" platform.
> In our case we need to call the virtio-block initialization from our platform
> init. It might be worth  moving this initialization into a funvtion in 
> virtio-block.c and only call that initializer from any platform init that
> supports virtio - any suggestions/preferences if we should do it now or when
> another architecture needs that - Anthony ?
> 
> This patch adds that initialization for the kvmppc bamboo platform.
> 
> Signed-off-by: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
> ---
> 
> [diffstat]
>  ppc440_bamboo.c |   13 +++++++++++++
>  1 files changed, 13 insertions(+)

[...]

FYI some output using virtio-block (still buggy on kvmppc)

Initialization when booting guest (VIO = virtio drvier in guest, U-VIO = virtio emulation in host kvm userspace)

PCI: Enabling device 0000:00:01.0 (0000 -> 0001)
PCI: Enabling device 0000:00:02.0 (0000 -> 0001)
U-VIO virtio_blk_update_config - vdev 0x10d0b008
VIO virtblk_probe - vdev 0xc8925200
U-VIO virtio_blk_get_features - vdev 0x10d0b008
U-VIO virtio_blk_get_features - vdev 0x10d0b008
U-VIO virtio_blk_get_features - vdev 0x10d0b008
blk_queue_max_hw_segments: set to minimum 1
 vda:<3>VIO do_virtblk_request - request_queue 0xc8936000
VIO do_req - request_queue 0xc8936000 request 0xc8937320 virtio_blk 0xc880c000 (8secs@0x0 - <NULL> Maj -929934324)
U-VIO virtio_blk_handle_output - vdev 0x10d0b008 vq 0x10d0b20c
VIO blk_done - vq 0xc88bcc00
VIO do_virtblk_request - request_queue 0xc8936000



We have no dynamic /dev so anyone who wants to test that will need to create /dev/vda for his needs.
This is easy to check using /proc/partitions.
Well I think that block size is broken here and using fdisk to check the partitions on that (unformatted) disk killed the guest kernel.
This needs more debugging, but to enable it for our platform is the first step - comments are welcome.

bash-3.00# cat /proc/partitions
major minor  #blocks  name
[...]
 254     0 22517998136852480 vda		<- ?broken?

The related device nod looks like this:
brw-r--r--  1 root root 254, 0 Mar 28  2008 /dev/vda



Below I added the log of the guest crash FYI:
bash-3.00# fdisk -l /dev/vda
VIO do_virtblk_request - request_queue 0xc8936000
VIO do_req - request_queue 0xc8936000 request 0xc89373e8 virtio_blk 0xc880c000 (8secs@0x0 - <NULL> Maj -929934324)
VIO do_req - request_queue 0xc8936000 request 0xc89374b0 virtio_blk 0xc880c000 (8secs@0x0 - <NULL> Maj -929934324)
VIO do_req - request_queue 0xc8936000 request 0xc8937578 virtio_blk 0xc880c000 (8secs@0x0 - <NULL> Maj -929934324)
VIO do_req - request_queue 0xc8936000 request 0xc8937320 virtio_blk 0xc880c000 (8secs@0x0 - <NULL> Maj -929934324)
U-VIO virtio_blk_handle_output - vdev 0x10d0b008 vq 0x10d0b20c
VIO blk_done - vq 0xc88bcc00
VIO do_virtblk_request - request_queue 0xc8936000
VIO do_virtblk_request - request_queue 0xc8936000
BUG: recent printk recursion!
Bamboo
Modules linked in:
NIP: c010bcec LR: c010d170 CTR: ffffffff
REGS: c8999ae0 TRAP: 0300   Not tainted  (2.6.25-rc3-dirty)
MSR: 00021002 <ME>  CR: 84004044  XER: 20000000
DEAR: fffffff8, ESR: 00000000
TASK = c89f6000[990] 'fdisk' THREAD: c8998000
GPR00: ffffffff c8999b90 c89f6000 fffffff8 ffffffff c02092e9 00000000 00000000
GPR08: c02092e8 fffffff8 ffffffff 00000008 24004028 1002c7b0 00000000 100c8510
GPR16: 100c0000 00000000 00000000 c8402208 c0250000 00000400 c899abc8 c0250000
GPR24: 00021002 00000000 c026b954 fffffff8 ffffffff c8999cc0 c026bd54 c026b9b3
NIP [c010bcec] strnlen+0x10/0x48
LR [c010d170] vsnprintf+0x558/0x83c
Call Trace:
[c8999b90] [c010cfec] vsnprintf+0x3d4/0x83c (unreliable)
[c8999bc0] [c010d46c] vscnprintf+0x18/0x38
[c8999bd0] [c0024c04] vprintk+0x70/0x3a8
[c8999c50] [c0024f8c] printk+0x50/0x60
[c8999cd0] [c013a3c4] do_virtblk_request+0x9c/0x2a8
[c8999d00] [c00f9048] __generic_unplug_device+0x3c/0x50
[c8999d10] [c00f90f0] generic_unplug_device+0x14/0x28
[c8999d20] [c00f8474] blk_unplug+0x20/0x30
[c8999d30] [c009a678] block_sync_page+0x54/0x70
[c8999d40] [c004e58c] sync_page+0x50/0x74
[c8999d50] [c004e5c0] sync_page_killable+0x10/0x4c
[c8999d60] [c01be7d0] __wait_on_bit_lock+0x9c/0xf8
[c8999d90] [c004ded8] __lock_page_killable+0x50/0x60
[c8999dc0] [c0050c6c] generic_file_aio_read+0x3bc/0x5bc
[c8999e30] [c0075330] do_sync_read+0xc0/0x134
[c8999ef0] [c0075c30] vfs_read+0xcc/0x16c
[c8999f10] [c007643c] sys_read+0x4c/0x90
[c8999f40] [c000d0cc] ret_from_syscall+0x0/0x3c
Instruction dump:
71200020 4082ffe4 38000000 98030001 881f0000 2f800000 409eff7c 4bffff9c
7c802379 7c691b78 7c0903a6 4182002c <88030000> 2f800000 419e0020 38630001
---[ end trace 74c6a5de04fea905 ]---


-- 

Grüsse / regards, 
Christian Ehrhardt
IBM Linux Technology Center, Open Virtualization

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
_______________________________________________
kvm-ppc-devel mailing list
kvm-ppc-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/kvm-ppc-devel