http://bugme.osdl.org/show_bug.cgi?id=1057
Summary: oops performing AIO write with O_DIRECT to block device
Kernel Version: 2.6.0-test2-mm2
Status: NEW
Severity: blocking
Owner: akpm@digeo.com
Submitter: slpratt@us.ibm.com
Distribution:SLES8
Hardware Environment:
8way PIII 700Mhz, 4xServRAID with 80 9GB disks (20 RAID0 arrays)
Software Environment:
Problem Description:
Doing multithreaded AIO to multiple drives produces the following oops
08/04/03-18:26:32 processing command: run seqaiowrite-O_DIR rawio -m 20 -p 160
-d 6 -n 102400 -x -z -t 13 -s 1024
new log requested
08/04/03-18:26:32 Running command rawio -m 20 -p 160 -d 6 -n 102400 -x -z -t 13
-s 1024
Oops: 0000 [#1]
SMP
CPU: 2
EIP: 0060:[<c0307a9e>] Not tainted VLI
EFLAGS: 00010086
EIP is at as_move_to_dispatch+0xe/0x130
eax: 00000000 ebx: f7a8b0a0 ecx: 00000000 edx: fffffff9
esi: 00000000 edi: 00000000 ebp: 00000001 esp: f5bf1c78
ds: 007b es: 007b ss: 0068
Process rawread (pid: 24164, threadinfo=f5bf0000 task=f5e90690)
Stack: 00000001 f59e2e40 f7a4ec00 00000000 f7a8b0a0 00000000 00000000 c0307cf5
f7a8b0a0 00000000 c036727d f7a8b0a0 00000000 f79e1a00 f7a4ec00 c0307e68
f7a8b0a0 f657d200 f79e1a00 c02ff626 f79e1a00 f657d200 f79e0800 f657d200
Call Trace:
[<c0307cf5>] as_dispatch_request+0x135/0x270
[<c036727d>] ips_queue+0xdd/0x170
[<c0307e68>] as_next_request+0x38/0x50
[<c02ff626>] elv_next_request+0x16/0x100
[<c03432e8>] scsi_request_fn+0x38/0x2c0
[<c0301467>] generic_unplug_device+0x67/0x70
[<c03015c6>] blk_run_queues+0x86/0xa0
[<c017c73d>] direct_io_worker+0x28d/0x3f0
[<c033e9b5>] scsi_finish_command+0x75/0xb0
[<c017c9f2>] blockdev_direct_IO+0x152/0x162
[<c0160470>] blkdev_get_blocks+0x0/0x80
[<c0160556>] blkdev_direct_IO+0x66/0x70
[<c0160470>] blkdev_get_blocks+0x0/0x80
[<c013d06d>] generic_file_direct_IO+0x9d/0xc0
[<c013c8d6>] generic_file_aio_write_nolock+0x886/0xb30
[<c010ba3e>] apic_timer_interrupt+0x1a/0x20
[<c011d3ec>] schedule+0x2dc/0x5f0
[<c0161606>] blkdev_file_aio_write+0x36/0x40
[<c017e370>] aio_pwrite+0x40/0xc0
[<c017e330>] aio_pwrite+0x0/0xc0
[<c017d84a>] aio_run_iocb+0x8a/0x180
[<c017e79d>] io_submit_one+0x1bd/0x220
[<c017e8bc>] sys_io_submit+0xbc/0x140
[<c010b04f>] syscall_call+0x7/0xb
Code: 01 00 00 00 a1 00 8c 48 c0 83 ea 28 39 42 30 0f 49 fe eb b0 90 90 8d b4 26
00 00 00 00 57 56 53 83 ec 10 8b 74 24 24 8b 5c 24 20 <8b> 7e 34 83 7e 04 02 0f
84 04 01 00 00 89 1c 24 e8 bd f5 ff ff
Steps to reproduce:
Benchmark can be obtained from:
http://www-124.ibm.com/developerworks/opensource/linuxperf/rawread/rawread.html
rawread -m 20 -p 160 -d 6 -n 102400 -x -z -t 13 -s 1024
This invocation runs on 20 logical disks starting at /dev/sdf. This test does
not fail when run against only 1 disk. Have not tried binary search to see how
many disks it takes to cause the failure.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Aug 07 2003 - 22:00:40 EST