Re: BUG: NULL pointer dereference at 00000000 -- IP: [<f8e783d5>] :b43:b43_dma_mapping_error+0x16/0x155

[Larry Finger <Larry.Finger@lwfinger.net>]

Miles Lane wrote:
> On Tue, Jun 10, 2008 at 7:15 PM, Larry Finger <Larry.Finger@lwfinger.net> wrote:
>> Miles Lane wrote:
>>> I got a slightly different BUG with this patch applied:
>>>
>>> BUG: unable to handle kernel NULL pointer dereference at 00000000
>>> IP: [<f89e3d41>] :b43:b43_dma_mapping_error+0x16/0x97
>>> *pde = 00000000
>>> Oops: 0000 [#1] PREEMPT DEBUG_PAGEALLOC
>>> last sysfs file:
>>>
>>> /sys/devices/pci0000:00/0000:00:08.0/0000:01:06.0/ssb0:0/firmware/ssb0:0/loading
>>> Modules linked in: sbs sbshc wmi battery iptable_filter ip_tables
>>> x_tables ac sbp2 rtc arc4 ecb crypto_blkcipher cryptomgr crypto_algapi
>>> b43 mac80211 cfg80211 led_class snd_intel8x0 snd_ac97_codec ac97_bus
>>> snd_pcm_oss snd_mixer_oss snd_pcm snd_seq_dummy snd_seq_oss
>>> snd_seq_midi_event snd_seq snd_timer snd_seq_device snd soundcore
>>> snd_page_alloc i2c_nforce2 i2c_core button ohci1394 3c59x ieee1394
>>> forcedeth ehci_hcd ohci_hcd usbcore thermal processor fan thermal_sys
>> Miles,
>>
>> If you have time, please cd to the kernel root directory and run the command
>> 'objdump -d -l drivers/net/wireless/b43/dma.o > dmadump'. You may have to
>> install the objdump package. In the file dmadump, find the line that
>> contains b43_dma_mapping_error. On my system it is 0000000000000085
>> <b43_dma_mapping_error>: Note, I use x86_64 and yours will likely be
>> different.
>>
>> Note the address (0x85 in my case), add the offset (0x16) and find the line
>> with the error. You can use KCalc in hexadecimal mode if you are not
>> comfortable with doing addition in base 16. Please report the source line
>> and instruction that failed.
> 
> For me, the line is:
> "000000ab <b43_dma_mapping_error>:"
> so, 0xab + 0x16 = 0xc1
> and here's the dump giving some context around that line:
> b43_dma_mapping_error():
>       ab:	55                   	push   %ebp
>       ac:	89 e5                	mov    %esp,%ebp
>       ae:	57                   	push   %edi
>       af:	89 d7                	mov    %edx,%edi
>       b1:	56                   	push   %esi
>       b2:	89 ce                	mov    %ecx,%esi
>       b4:	53                   	push   %ebx
>       b5:	83 ec 08             	sub    $0x8,%esp
>       b8:	8a 55 08             	mov    0x8(%ebp),%dl
>       bb:	88 55 ef             	mov    %dl,-0x11(%ebp)
>       be:	8b 50 58             	mov    0x58(%eax),%edx
>       c1:	8b 12                	mov    (%edx),%edx
>       c3:	8b 52 08             	mov    0x8(%edx),%edx
>       c6:	89 55 f0             	mov    %edx,-0x10(%ebp)
>       c9:	8b 40 34             	mov    0x34(%eax),%eax
>       cc:	83 f8 1e             	cmp    $0x1e,%eax
>       cf:	74 07                	je     d8 <b43_dma_mapping_error+0x2d>
>       d1:	83 f8 20             	cmp    $0x20,%eax
>       d4:	75 5b                	jne    131 <b43_dma_mapping_error+0x86>
>       d6:	eb 15                	jmp    ed <b43_dma_mapping_error+0x42>

The line that fails is

if (unlikely(dma_mapping_error(ring->dev->dev->dma_dev, addr)))

The value at 0x58(%eax) is zero, which corresponds to the "struct 
b43_wldev *dev" line in struct b43_dmaring. In other words, ring->dev 
is NULL, which causes the error. I'll leave it to FUJITA Tomonori to 
figure out why.

Larry