Re: Oops in 3.0.30/fsl_usb2_device_register

[Guenter Roeck <guenter.roeck@ericsson.com>]

On Tue, 2012-05-01 at 13:12 -0400, Greg Kroah-Hartman wrote:
> On Tue, May 01, 2012 at 09:22:53AM -0700, Guenter Roeck wrote:
> > Hi all,
> > 
> > I see the following oops in 3.0.30. This is with a Freescale 8536.
> > 
> > Unable to handle kernel paging request for data at address 0x00000000
> > Faulting instruction address: 0xc03f7e0c
> > Oops: Kernel access of bad area, sig: 11 [#1]
> > PREEMPT SE8000_lc8536
> > Modules linked in:
> > NIP: c03f7e0c LR: c03f7dd4 CTR: 00000000
> > REGS: ef837d50 TRAP: 0300   Not tainted  (3.0.30-767-gfd1550a)
> > MSR: 00029000 <EE,ME,CE>  CR: 82002024  XER: 20000000
> > DEAR: 00000000, ESR: 00800000
> > TASK = ef838000[1] 'swapper' THREAD: ef836000
> > GPR00: ef912008 ef837e00 ef838000 ef8cbc00 ef837e28 00000050 00000000 00000000
> > GPR08: 00000000 00000000 ffffffff ef9120f0 24002024 1017f868 deabfc4c 72947cde
> > GPR16: 31e01609 00000800 00000000 04100000 1f07fde8 1f180000 c09c8000 fffff000
> > GPR24: ef837e28 c0ae7d14 c09c75b4 ef906ae0 00000002 ef837e28 ef8cbc00 ef912000
> > NIP [c03f7e0c] fsl_usb2_device_register+0x74/0xf8
> > LR [c03f7dd4] fsl_usb2_device_register+0x3c/0xf8
> > Call Trace:
> > [ef837e00] [c03f7dd4] fsl_usb2_device_register+0x3c/0xf8 (unreliable)
> > [ef837e20] [c03f819c] fsl_usb2_mph_dr_of_probe+0x238/0x2bc
> > [ef837ea0] [c0285a00] platform_drv_probe+0x20/0x30
> > [ef837eb0] [c0283e6c] driver_probe_device+0xc8/0x1f8
> > [ef837ed0] [c0284070] __driver_attach+0xd4/0xd8
> > [ef837ef0] [c02833fc] bus_for_each_dev+0x7c/0xc4
> > [ef837f20] [c0283c34] driver_attach+0x24/0x34
> > [ef837f30] [c0282aa0] bus_add_driver+0xdc/0x2dc
> > [ef837f60] [c0284900] driver_register+0x94/0x198
> > [ef837f80] [c028609c] platform_driver_register+0x6c/0x7c
> > [ef837f90] [c04dd288] fsl_usb2_mph_dr_init+0x18/0x28
> > [ef837fa0] [c0001ddc] do_one_initcall+0x50/0x200
> > [ef837fd0] [c04bf1bc] kernel_init+0x9c/0x14c
> > [ef837ff0] [c000dd28] kernel_thread+0x4c/0x68
> > Instruction dump:
> > 813f0080 381f0008 815f0084 7fa4eb78 901e0008 38a00050 913e0080 915e0084
> > 817f0078 811e0078 812b0000 814b0004
> >  91480004 4be8e679 2c030000
> >  ---[ end trace 3a8eb627b8e9e0fb ]---
> > 
> > Problem was introduced with commit d39514c14bd941232976b68e2750dc725b90e724.
> > The commit claims that dma_mask is already set in arch_setup_pdev_archdata().
> > However, that only seems to be true for the current kernel, not for 3.0.
> > As a result, dma_mask is NULL in fsl_usb2_device_register, causing the above oops.
> > 
> > Question now is if d39514c14bd941232976b68e2750dc725b90e724 should be reverted from 3.0,
> > or if there is a better solution. Any thoughts/ideas ?
> 
> This patch is already queued up to be reverted from the next 3.0-stable
> release, so yes, it should be reverted :)
> 
> The next 3.0-stable tree should have this fixed for you, sorry about
> that.

Sounds good.

Thanks,
Guenter