Re: [PATCH] Chaining sg lists for big IO commands v5

[Jens Axboe <jens.axboe@oracle.com>]

On Thu, May 24 2007, FUJITA Tomonori wrote:
> From: Jens Axboe <jens.axboe@oracle.com>
> Subject: Re: [PATCH] Chaining sg lists for big IO commands v5
> Date: Thu, 24 May 2007 11:34:41 +0200
> 
> > On Tue, May 22 2007, Badari Pulavarty wrote:
> > > > If iommu=off works, can you try a normal boot but with this applied on
> > > > top of the sglist patches? That should fix gart mapping.
> > > > 
> > > > diff --git a/arch/x86_64/kernel/pci-gart.c b/arch/x86_64/kernel/pci-gart.c
> > > > index 2e22a3a..b16384f 100644
> > > > --- a/arch/x86_64/kernel/pci-gart.c
> > > > +++ b/arch/x86_64/kernel/pci-gart.c
> > > 
> > > Yes. With this patch + kzalloc() for ide probe took care of all my
> > > problems on x86-64.
> > > 
> > > But my ppc64 box is still sick :(
> > 
> > :-(
> > 
> > > Oops: Kernel access of bad area, sig: 11 [#1]
> > > SMP NR_CPUS=32 NUMA pSeries
> > > Modules linked in: qla2xxx scsi_transport_fc
> > > NIP: c0000000000414a0 LR: c00000000004162c CTR: 0000000000000001
> > > REGS: c0000000047bb130 TRAP: 0300   Not tainted  (2.6.22-rc1)
> > > MSR: 8000000000001032 <ME,IR,DR>  CR: 28000022  XER: 00000008
> > > DAR: 0000000000000000, DSISR: 0000000040000000
> > > TASK = c0000000047a6aa0[0] 'swapper' THREAD: c0000000047b8000 CPU: 7
> > > GPR00: 0000000000000080 c0000000047bb3b0 c000000000692358 c0000000047a6aa0
> > > GPR04: 0000000000000000 0000000000000070 0000000000000000 c0000000005ac0b8
> > > GPR08: 000000000000e4b4 0000000000000001 0000000000000000 0000000000000280
> > > GPR12: 0000000000000020 c0000000005a3e80 0000000000000000 0000000007a8dd70
> > > GPR16: 0000000000000000 0000000000000000 0000000000000000 c0000000047b8000
> > > GPR20: 000000003b9aca00 c0000000047a6c50 0000000000000001 0000000000000000
> > > GPR24: d000000000909048 0001dee6d30c0d30 c000000003b7dd80 c0000000047a6aa0
> > > GPR28: 00000001079027ca c0000000047a6aa0 c0000000005b7cb0 c000000000472c9c
> > > NIP [c0000000000414a0] .dequeue_task+0x0/0x9c
> > > LR [c00000000004162c] .deactivate_task+0x40/0x60
> > > Call Trace:
> > > [c0000000047bb3b0] [c00000000004bccc] .printk+0x38/0x48 (unreliable)
> > > [c0000000047bb440] [c000000000471704] .schedule+0x1fc/0x8dc
> > > [c0000000047bb540] [c000000000472c9c] .schedule_timeout+0xa8/0xe8
> > > [c0000000047bb610] [c000000000057260] .msleep+0x20/0x38
> > > [c0000000047bb690] [c00000000003f5ec] .eeh_dn_check_failure+0x114/0x268
> > > [c0000000047bb740] [c00000000003fc64] .eeh_check_failure+0xec/0x114
> > > [c0000000047bb7c0] [d00000000086190c] .qla2300_fw_dump+0x1130/0x1c00 [qla2xxx]
> > > [c0000000047bb8a0] [d000000000858d50] .qla2300_intr_handler+0x1e8/0x60c [qla2xxx]
> > > [c0000000047bb950] [c000000000078368] .handle_IRQ_event+0x70/0xe4
> > > [c0000000047bb9f0] [c00000000007a7e0] .handle_fasteoi_irq+0x11c/0x1d0
> > > [c0000000047bba90] [c00000000000c178] .do_IRQ+0x90/0xec
> > > [c0000000047bbb10] [c000000000004790] hardware_interrupt_entry+0x18/0x1c
> > 
> > Not good. The qla changes are non-trivial (that hardware has a really
> > funky sg setup), so I may have botched a part of it. I'll review the
> > qla changes and get back to you.
> 
> I guess that qla2xxx needs more for_each_sg macros. My qla2xxx cleanup
> patch might help:
> 
> http://git.kernel.org/?p=linux/kernel/git/tomo/linux-2.6-bidi.git;a=commitdiff;h=fb2747a9f2465ac4a1ea9e3533b92510fbcaae0c

Neat, will take a look! I should be able to just reuse that.

-- 
Jens Axboe