AT91 UDC softirq problem
Russell King - ARM Linux
linux at arm.linux.org.uk
Wed Sep 1 03:55:00 EDT 2010
On Fri, Aug 27, 2010 at 08:02:44AM +0000, Mars Steeve wrote:
> Then I compiled a 2.6.36-rc2 kernel (with the at91_udc softirq patch) and now my
> link is stable but I always have a warning message when I plug my device on the
> host:
>
> [ 2.170000] ------------[ cut here ]------------
> [ 2.170000] WARNING: at kernel/softirq.c:143 local_bh_enable+0x80/0xac()
> [ 2.170000] Modules linked in:
> [ 2.170000] [<c0025cb0>] (unwind_backtrace+0x0/0xf4) from [<c0032574>]
> (warn_slowpath_common+0x4c/0x64)
> [ 2.170000] [<c0032574>] (warn_slowpath_common+0x4c/0x64) from [<c00325a8>]
> (warn_slowpath_null+0x1c/0x24)
> [ 2.170000] [<c00325a8>] (warn_slowpath_null+0x1c/0x24) from [<c00389f4>]
> (local_bh_enable+0x80/0xac)
> [ 2.170000] [<c00389f4>] (local_bh_enable+0x80/0xac) from [<c01e012c>]
> (dev_txq_stats_fold+0xa0/0xec)
> [ 2.170000] [<c01e012c>] (dev_txq_stats_fold+0xa0/0xec) from [<c01e0268>]
> (dev_get_stats+0xf0/0xf8)
> [ 2.170000] [<c01e0268>] (dev_get_stats+0xf0/0xf8) from [<c01c45c8>]
> (gen_ndis_query_resp+0x58/0x4a0)
> [ 2.170000] [<c01c45c8>] (gen_ndis_query_resp+0x58/0x4a0) from [<c01c4c7c>]
> (rndis_msg_parser+0x26c/0x458)
> [ 2.170000] [<c01c4c7c>] (rndis_msg_parser+0x26c/0x458) from [<c01c4e8c>]
> (rndis_command_complete+0x24/0x58)
> [ 2.170000] [<c01c4e8c>] (rndis_command_complete+0x24/0x58) from [<c01c01c4>]
> (done+0x84/0xcc)
> [ 2.170000] [<c01c01c4>] (done+0x84/0xcc) from [<c01c11e4>]
> (read_fifo+0xf8/0x10c)
> [ 2.170000] [<c01c11e4>] (read_fifo+0xf8/0x10c) from [<c01c1830>]
> (at91_udc_irq+0x3a8/0x97c)
> [ 2.170000] [<c01c1830>] (at91_udc_irq+0x3a8/0x97c) from [<c0063a18>]
> (handle_IRQ_event+0x40/0x118)
> [ 2.170000] [<c0063a18>] (handle_IRQ_event+0x40/0x118) from [<c0065c58>]
> (handle_level_irq+0xa8/0x17c)
> [ 2.170000] [<c0065c58>] (handle_level_irq+0xa8/0x17c) from [<c0020044>]
> (asm_do_IRQ+0x44/0xa4)
> [ 2.170000] [<c0020044>] (asm_do_IRQ+0x44/0xa4) from [<c0020b28>]
> (__irq_svc+0x48/0x8c)
> [ 2.170000] Exception stack(0xc0319f80 to 0xc0319fc8)
> [ 2.170000] 9f80: c0318000 60000013 00000001 60000013 c0318000 c03397a8
> c031bc90 c031bc88
> [ 2.170000] 9fa0: 2001c838 41069265 2001c804 00000000 00000000 c0319fc8
> c0022180 c0022120
> [ 2.170000] 9fc0: 60000013 ffffffff
> [ 2.170000] [<c0020b28>] (__irq_svc+0x48/0x8c) from [<c0022120>]
> (cpu_idle+0x28/0xc0)
> [ 2.170000] [<c0022120>] (cpu_idle+0x28/0xc0) from [<c0008924>]
> (start_kernel+0x230/0x274)
> [ 2.170000] [<c0008924>] (start_kernel+0x230/0x274) from [<20008034>]
> (0x20008034)
> [ 2.170000] ---[ end trace d1eb86bf710a14c5 ]---
>
> Is this message important or can I still use this version without problem
> in the future ?
Yes, it means someone is calling code from IRQ context which must not
be called from IRQ context. In other words, the AT91 UDC driver is
broken.
My guess is that dev_get_stats() must not be called from IRQ context,
which raises the question why gen_ndis_query_resp() is doing so - maybe
it's a ndis bug? In any case, I think you need to report this to USB
people.
More information about the linux-arm-kernel
mailing list