On 2012-04-20 19:27, Vikram Seth wrote: > Hi Jens, > > Recently we saw an issue where fio created kernel panic because of 0 > size requests. > Pasting the stack trace below. > > <d>Pid: 13846, comm: fio Tainted: P M 2.6.38.4- > <d>RIP: 0010:[<ffffffff8111d90d>] [<ffffffff8111d90d>] > __blockdev_direct_IO+0x6a7/0x9ce > <d>RSP: 0018:ffff8823fb7e9c18 EFLAGS: 00010246 > <d>RAX: 0000000000000000 RBX: ffffea007d55d460 RCX: 000000000000000a > <d>RDX: 0000000000000c00 RSI: 0000000000000000 RDI: ffff8812342b80e0 > <d>RBP: ffff8823fb7e9cd8 R08: 0000000000000001 R09: 0000000000000000 > <d>R10: 0000000000000000 R11: ffffea0000000000 R12: ffff881233d64400 > <d>R13: 0000000000000003 R14: ffff881233d64498 R15: ffff8812342b80e0 > <d>FS: 00007f048cddb6e0(0000) GS:ffff88007e200000(0000) knlGS:0000000000000000 > <d>CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 > <d>CR2: 000000000040d040 CR3: 000000238ea69000 CR4: 00000000000406f0 > <d>DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 > <d>DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 > Process fio (pid: 13846, threadinfo ffff8823fb7e8000, task ffff882432ead120) > <0>Stack: > <c> ffff8823fb7e9c68<c> ffffffff814478ce<c> ffff882300000001<c> > ffff8823cf614000<c> > <c> ffff882300000004<c> ffffffff814478ce<c> fffffffffffffc00<c> > 00000000814478ce<c> > <c> ffff8823fb7e8000<c> 0000000000000400<c> 0000000a00000001<c> > ffff8823fb7e8000<c> > > Checking that function it's > > 0xffffffff8111d90d is in __blockdev_direct_IO (fs/direct-io.c:955). > > and that points to following line in direct-io.c: > > 955 BUG_ON(this_chunk_bytes == 0); > > We also tested with following patch from kernel.org > > f9b5570 fs: simplify handling of zero sized reads in __blockdev_direct_IO > > That resolved the panic problem. > > Can you please add check to fio that it should error out OR skip 0 > size accesses ? It seems you forgot to include the job file that triggered this :-) Fio should not be making zero sized reads, but pretty embarassing that the kernel oopsed on getting one. It could be a file system issue, a bug there triggering a zero sized read. So how did you trigger this? And on what fs? -- Jens Axboe -- To unsubscribe from this list: send the line "unsubscribe fio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
