Re: wireless: Problems with ath5k and 2.6.35

From: Jiri Slaby <jirislaby@gmail.com>
To: Daniel Mack <daniel@caiaq.de>
Cc: linux-kernel@vger.kernel.org, linux-wireless@vger.kernel.org,
	Sven Neumann <s.neumann@raumfeld.com>,
	Michael Hirsch <m.hirsch@raumfeld.com>,
	ath5k-devel@lists.ath5k.org
Subject: Re: wireless: Problems with ath5k and 2.6.35
Date: Sun, 08 Aug 2010 10:20:17 +0200	[thread overview]
Message-ID: <4C5E68C1.9070403@gmail.com> (raw)
In-Reply-To: <20100807172108.GA15974@buzzloop.caiaq.de>

On 08/07/2010 07:21 PM, Daniel Mack wrote:
> Hi,
> 
> we're having problems using an ath5k-based mini-PCI card on a mainboard
> with a Geode CS5535 chipset. Below is the output generated by 2.6.35
> vanilla, the first block is reported at boot time, the second one some
> seconds later.
> 
> Don't know if this has been reported before, but at least I couldn't
> find anything in recent LKML postings.
> 
> 
> Thanks,
> Daniel
> 
> 
> [    2.142771] ath5k 0000:00:0c.0: guessed PCI INT A -> IRQ 9
> [    2.148319] ath5k 0000:00:0c.0: sharing IRQ 9 with 0000:00:01.2
> [    2.154548] ath5k 0000:00:0c.0: registered as 'phy0'
> [    3.126593] irq 9: nobody cared (try booting with the "irqpoll" option)
> [    3.130015] Pid: 1, comm: swapper Not tainted 2.6.35+ #1
> [    3.130015] Call Trace:
> [    3.130015]  [<c013e984>] __report_bad_irq+0x2e/0x6f
> [    3.130015]  [<c013eab5>] note_interrupt+0xf0/0x147
> [    3.130015]  [<c013f076>] handle_level_irq+0x6d/0x99
> [    3.130015]  [<c0103bf6>] handle_irq+0x1a/0x20
> [    3.130015]  [<c0103810>] do_IRQ+0x33/0x88
> [    3.130015]  [<c0102a69>] common_interrupt+0x29/0x30
> [    3.130015]  [<c0120e3c>] ? __do_softirq+0x34/0xca
> [    3.130015]  [<c0120ef7>] do_softirq+0x25/0x2a
> [    3.130015]  [<c0120fb7>] irq_exit+0x28/0x57
> [    3.130015]  [<c0103850>] do_IRQ+0x73/0x88
> [    3.130015]  [<c0102a69>] common_interrupt+0x29/0x30
> [    3.130015]  [<c013007b>] ? __async_schedule+0x38/0xe9
> [    3.130015]  [<c013e420>] ? __setup_irq+0x1d8/0x259
> [    3.130015]  [<c013e544>] request_threaded_irq+0xa3/0xd0
> [    3.130015]  [<c02486e5>] ? ath5k_intr+0x0/0x28b
> [    3.130015]  [<c032e15c>] ath5k_pci_probe+0x2f4/0x13c5

The interrupt came right after enabled in probe function with
ATH_STAT_INVALID still set. Is this really a 2.6.34 regression? The only
difference I see that we now (in .35) use MIB interrrupts.

> [    3.130015] handlers:
> [    3.130015] [<c02486e5>] (ath5k_intr+0x0/0x28b)
> [    3.130015] Disabling IRQ #9
> [    4.014683] ath5k phy0: Atheros AR5414 chip found (MAC: 0xa5, PHY: 0x61)
> 
> [...]
> 
> [   13.510198] BUG: unable to handle kernel NULL pointer dereference at 0000009b
> [   13.517396] IP: [<c1138a54>] ath5k_hw_is_intr_pending+0x1/0x19
> [   13.520166] *pde = 00000000 
> [   13.520166] Oops: 0000 [#1] 
> [   13.520166] last sysfs file: 
> [   13.520166] Modules linked in:
> [   13.520166] 
> [   13.520166] Pid: 0, comm: swapper Not tainted 2.6.34 #1 /
> [   13.520166] EIP: 0060:[<c1138a54>] EFLAGS: 00010046 CPU: 0
> [   13.520166] EIP is at ath5k_hw_is_intr_pending+0x1/0x19
> [   13.520166] EAX: ffffffff EBX: cf9289a0 ECX: c12f5f24 EDX: cf9289a0
> [   13.520166] ESI: c1303920 EDI: ffffffff EBP: c12f5ee4 ESP: c12f5e98
> [   13.520166]  DS: 007b ES: 007b FS: 0000 GS: 0000 SS: 0068
> [   13.520166] Process swapper (pid: 0, ti=c12f4000 task=c12fa300 task.ti=c12f4000)
> [   13.520166] Stack:
> [   13.520166]  c12f5ee4 c1143674 00000082 c12f5eb4 c103df6b 00000000 00000030 c12f5ec0
> [   13.520166] <0> c12f5ec0 c10201ff c12f5ed8 c100393c c12f5f88 cfbdba00 00000001 c141e080
> [   13.520166] <0> cfbe1de0 c1303920 00000000 c12f5efc c103d773 00000009 00000009 c141e080
> [   13.520166] Call Trace:
> [   13.520166]  [<c1143674>] ? ath5k_intr+0x25/0x1f5

This is from .34. However I don't understand how this can happen. It
means that sc->ah is -25. But this cannot be true as sc->ah = kzalloc.
"Code:" line would be helpful too. Does the "nobody cared" appear in .34
too?

thanks,
-- 
js