John,
If possible, I would like you to do one more test on that troublesome netbook.
Please build a kernel with CONFIG_MMIOTRACE=y and build the Broadcom wl driver
for it. Starting with wl unloaded, issue the following commands (as root):
echo 10000 > /sys/kernel/debug/tracing/buffer_size_kb
echo mmiotrace > /sys/kernel/debug/tracing/current_tracer
cat /sys/kernel/debug/tracing/trace_pipe > wl_dump.txt &
modprobe -v wl
After the wireless network device gets created, etc., then
echo nop > /sys/kernel/debug/tracing/current_tracer
sleep 1
killproc cat
and send me the file wl_dump.txt. With it, I will be able to see if the Broadcom
driver is actually finding an SPROM that b43 is missing.
Thanks,
Larry
On 03/30/2010 03:28 PM, John W. Linville wrote:
> On Tue, Mar 30, 2010 at 10:23:51AM -0500, Larry Finger wrote:
>
>> If possible, I would like you to do one more test on that troublesome netbook.
>> Please build a kernel with CONFIG_MMIOTRACE=y and build the Broadcom wl driver
>> for it. Starting with wl unloaded, issue the following commands (as root):
>>
>> echo 10000 > /sys/kernel/debug/tracing/buffer_size_kb
>> echo mmiotrace > /sys/kernel/debug/tracing/current_tracer
>> cat /sys/kernel/debug/tracing/trace_pipe > wl_dump.txt &
>> modprobe -v wl
>>
>> After the wireless network device gets created, etc., then
>>
>> echo nop > /sys/kernel/debug/tracing/current_tracer
>> sleep 1
>> killproc cat
>>
>> and send me the file wl_dump.txt. With it, I will be able to see if the Broadcom
>> driver is actually finding an SPROM that b43 is missing.
>
> Done:
>
> http://www.kernel.org/pub/linux/kernel/people/linville/ssb-debug/wl_dump.txt.gz
Thanks. After using the mmio reformatting tool on your data, I found the following:
read16 #1 block at 0xfeafc800 of 220 (0xdc) elements:
0xfeafc800: 3801 0000 1507 103c 0078 6dbe 0000 2bc4 2a64 2964 2c64 3ce7
0xfeafc818: 46ff 477f 0c00 7840 00a0 ffff 0000 ffff ffff ffff ffff ffff
0xfeafc830: ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc848: ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc860: 4315 8000 0002 0000 1800 1800 0000 0000 ffff ffff ffff ffff
0xfeafc878: ffff ffff ffff ffff 5372 1306 0a01 0006 0000 0000 0025 5681
0xfeafc890: 783e 5553 0000 03ff ffff ffff 0003 ff02 ffff ff11 000f ffff
0xfeafc8a8: 006c ffff fffa ffff ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc8c0: 3e44 17ec fa34 fe73 ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc8d8: ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc8f0: ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc908: ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc920: ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc938: ffff ffff ffff ffff ffff 0008 ffff ffff ffff ffff ffff ffff
0xfeafc950: ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc968: ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc980: ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc998: ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff ffff
0xfeafc9b0: ffff ffff ffff ed08
That is the reading of a valid, rev. 8 SPROM. Your device does indeed have one.
Why we avoided the lockup by skipping the readout remains a question. Now to
build wl here and compare the differences.
Thanks,
Larry
On 03/30/2010 03:28 PM, John W. Linville wrote:
> On Tue, Mar 30, 2010 at 10:23:51AM -0500, Larry Finger wrote:
>
>> If possible, I would like you to do one more test on that troublesome netbook.
>> Please build a kernel with CONFIG_MMIOTRACE=y and build the Broadcom wl driver
>> for it. Starting with wl unloaded, issue the following commands (as root):
>>
>> echo 10000 > /sys/kernel/debug/tracing/buffer_size_kb
>> echo mmiotrace > /sys/kernel/debug/tracing/current_tracer
>> cat /sys/kernel/debug/tracing/trace_pipe > wl_dump.txt &
>> modprobe -v wl
>>
>> After the wireless network device gets created, etc., then
>>
>> echo nop > /sys/kernel/debug/tracing/current_tracer
>> sleep 1
>> killproc cat
>>
>> and send me the file wl_dump.txt. With it, I will be able to see if the Broadcom
>> driver is actually finding an SPROM that b43 is missing.
>
> Done:
>
> http://www.kernel.org/pub/linux/kernel/people/linville/ssb-debug/wl_dump.txt.gz
>
> Hth!
It has already. The clue was in the fragment I sent you earlier. Every other
BCM43XX card that we have seen had its SPROM starting at offset 0x1000 relative
to the start of the MMIO mapping. Yours is at 0x0800! I would have hoped for an
"incorrect CRC" message, but we got the hang instead. I now have to find the
hidden flag that lets us know about the different location. I wonder what else
has moved.
As a warning to users of wl. I tried to run it from the 2.6.34-rc3 on Linus's
tree. Broadcom (I really like the other variation of that name!) has not (yet)
discovered the change in the DMA API in 2.6.34. As a result, I got a few "tries
to free DMA memory it has not allocated" warnings followed by a lot of other
stuff concluding in a kernel panic - the first I've had for a long time. Note: I
will not be debugging this problem. I have enough to do already.
Larry
On Tue, Mar 30, 2010 at 10:23:51AM -0500, Larry Finger wrote:
> If possible, I would like you to do one more test on that troublesome netbook.
> Please build a kernel with CONFIG_MMIOTRACE=y and build the Broadcom wl driver
> for it. Starting with wl unloaded, issue the following commands (as root):
>
> echo 10000 > /sys/kernel/debug/tracing/buffer_size_kb
> echo mmiotrace > /sys/kernel/debug/tracing/current_tracer
> cat /sys/kernel/debug/tracing/trace_pipe > wl_dump.txt &
> modprobe -v wl
>
> After the wireless network device gets created, etc., then
>
> echo nop > /sys/kernel/debug/tracing/current_tracer
> sleep 1
> killproc cat
>
> and send me the file wl_dump.txt. With it, I will be able to see if the Broadcom
> driver is actually finding an SPROM that b43 is missing.
Done:
http://www.kernel.org/pub/linux/kernel/people/linville/ssb-debug/wl_dump.txt.gz
Hth!
John
--
John W. Linville Someday the world will need a hero, and you
[email protected] might be all we have. Be ready.