Crash when loading module

Crash when loading module

[Zaki]

Hi,

I am new to ath10k.  Used to have ath9k-based module last time.  I am
using Freescale MPC8315ERDB board to host the QCA988X_2_0_DEVICE_ID
device.  I am forcing the module to use legacy interrupt handler
because my BSP doesn't support MSI or MSI-X just yet.
During the module loading, the system crash.  Looks like it failed at
some reset device or reset target function, but i am not too sure
which one causing it.  Below are the logs i collected.  I would
apopreciate any help.  Thanks a lot in advance.

# modprobe ath10k_pci.ko
[   42.465296] cfg80211: Calling CRDA to update world regulatory domain
[   42.744607] ath10k_pci 0001:02:00.0: BAR 0: assigned [mem 0xa0000000-0xa01ff]
[   42.752698] ath10k_pci 0001:02:00.0: enabling device (0000 -> 0002)
[   42.759300] ath10k: MSI-X didn't succeed (1), trying MSI
[   42.785492] ath10k: legacy interrupt handling
[   42.850072] Oops: Machine check, sig: 7 [#1]
[   42.854333] DEBUG_PAGEALLOC MPC831x RDB
[   42.858157] Modules linked in: ath10k_pci(+) ath10k_core ath mac80211 cfg802s
[   42.868341] CPU: 0 PID: 1328 Comm: modprobe Tainted: G        W    3.11.0-rc1
[   42.876313] task: c6ea5000 ti: c6cee000 task.ti: c6cee000
[   42.881688] NIP: 00000500 LR: c9a6d31c CTR: c201415c
[   42.886632] REGS: c6cefae0 TRAP: 0200   Tainted: G        W     (3.11.0-rc3-)
[   42.894249] MSR: 00041000 <ME>  CR: 22008842  XER: 20000000
[   42.899812]
GPR00: c9a6d314 c6cefb90 c6ea5000 00000000 00000001 00000005 c2219188 00000000
GPR08: 000006cc c25a708c 000006cc c25b6288 c201415c 100b1254 c6cefe44 c2580000
GPR16: c2420000 c6cefe2c c2580000 c2580000 00000000 c9a80000 c9a70000 c6d2ac00
GPR24: c72b4c80 00000001 c7858468 c9b00004 00000001 c9a80000 c72b4c80 00000063
[   42.929482] NIP [00000500] 0x500
[   42.932730] LR [c9a6d31c] ath10k_pci_target_is_awake+0x2c/0x64 [ath10k_pci]
[   42.939657] Call Trace:
[   42.942109] [c6cefb90] [c9a6d314] ath10k_pci_target_is_awake+0x24/0x64 [ath1)
[   42.950876] [c6cefbb0] [c9a6d51c] ath10k_pci_wait+0x40/0xa0 [ath10k_pci]
[   42.957563] [c6cefbc0] [c9a6e128] ath10k_pci_hif_power_up+0x12c/0x6d8 [ath10]
[   42.965130] [c6cefc10] [c9a48698] ath10k_core_register+0x24/0x304 [ath10k_co]
[   42.972425] [c6cefc40] [c9a6f9e4] ath10k_pci_probe+0x434/0x5bc [ath10k_pci]
[   42.979373] [c6cefc90] [c21f6a5c] local_pci_probe+0x34/0x58
[   42.984931] [c6cefca0] [c21f769c] pci_device_probe+0x84/0xa4
[   42.990577] [c6cefcd0] [c22256fc] driver_probe_device+0x17c/0x380
[   42.996653] [c6cefcf0] [c2225ccc] __driver_attach+0xa4/0xa8
[   43.002209] [c6cefd10] [c2223b10] bus_for_each_dev+0x80/0xbc
[   43.007851] [c6cefd40] [c2225384] driver_attach+0x24/0x34
[   43.013234] [c6cefd50] [c2224314] bus_add_driver+0x10c/0x27c
[   43.018877] [c6cefd70] [c22262e8] driver_register+0x6c/0x194
[   43.024519] [c6cefd90] [c21f7768] __pci_register_driver+0x3c/0x4c
[   43.030605] [c6cefda0] [c9a7602c] ath10k_pci_init+0x2c/0x80 [ath10k_pci]
[   43.037292] [c6cefdb0] [c2003bc0] do_one_initcall+0x118/0x1dc
[   43.043029] [c6cefe20] [c206af28] load_module+0xda8/0xec8
[   43.048414] [c6cefee0] [c206b324] SyS_init_module+0x118/0x158
[   43.054153] [c6ceff40] [c200f4d0] ret_from_syscall+0x0/0x38
[   43.059712] --- Exception: c01 at 0xff20a8c
[   43.059712]     LR = 0x10010a20
[   43.066991] Instruction dump:
[   43.085425] ---[ end trace 3eb4529c14633098 ]---
[   43.090019]
[   44.081557] Kernel panic - not syncing: Fatal exception
[   44.086762] Rebooting in 180 seconds..

Rgds,
Zack.

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Crash when loading module

[Kalle Valo]

Zaki <zaki.bm@gmail.com> writes:

> Hi,
>
> I am new to ath10k.  Used to have ath9k-based module last time.  I am
> using Freescale MPC8315ERDB board to host the QCA988X_2_0_DEVICE_ID
> device.

Are you sure it's QCA9880 v2? For some reason v1 has same PCI id as v2.

> I am forcing the module to use legacy interrupt handler because my BSP
> doesn't support MSI or MSI-X just yet.

How did you force the legacy interrupt handler?

What version of ath10k are you using? Are you using ath.git master
branch or something else?

> During the module loading, the system crash. Looks like it failed at
> some reset device or reset target function, but i am not too sure
> which one causing it. Below are the logs i collected. I would
> apopreciate any help. Thanks a lot in advance.
>
> # modprobe ath10k_pci.ko
> [   42.465296] cfg80211: Calling CRDA to update world regulatory domain
> [   42.744607] ath10k_pci 0001:02:00.0: BAR 0: assigned [mem 0xa0000000-0xa01ff]
> [   42.752698] ath10k_pci 0001:02:00.0: enabling device (0000 -> 0002)
> [   42.759300] ath10k: MSI-X didn't succeed (1), trying MSI
> [   42.785492] ath10k: legacy interrupt handling
> [   42.850072] Oops: Machine check, sig: 7 [#1]
> [   42.854333] DEBUG_PAGEALLOC MPC831x RDB

According to the link below this error might indicate that the PCI
device is not responding. I wonder if you are using v1 board?

https://lkml.org/lkml/2007/8/23/41

-- 
Kalle Valo

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Crash when loading module

[Kalle Valo]

Hi Zaki,

please CC the mailing list when replying. I do not have time to give
private support.

Zaki <zaki.bm@gmail.com> writes:

> The PCIe module is taken from Archer C7 V1 product's board.  I am not
> so sure if it is V2 hardware, though. I just depends on the debugging
> msg that says it is V2 device.
>
> The way i force it to use legacy interrupt is by modifying the code to
> skip MSI-X and MSI.  The ath10k code is taken from ath.git master
> branch.
>
> Thanks a lot for the link.  If someone can verify that the module on
> Archer C7 V1 is actually a V1 hardware, then it will be very helpful
> also.

To my knowledge Archer C7 has QCA9880 v1. Because the PCI ids are the
same I'm not currently aware of any way to detect from ath10k if QCA9880
v1 or v2 :(

-- 
Kalle Valo

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Crash when loading module

[Zaki]

Apologize. I should have pressed the Reply All button just now.

On Wed, Aug 28, 2013 at 2:32 PM, Kalle Valo <kvalo@qca.qualcomm.com> wrote:
> Hi Zaki,
>
> please CC the mailing list when replying. I do not have time to give
> private support.
>
> Zaki <zaki.bm@gmail.com> writes:
>
>> The PCIe module is taken from Archer C7 V1 product's board.  I am not
>> so sure if it is V2 hardware, though. I just depends on the debugging
>> msg that says it is V2 device.
>>
>> The way i force it to use legacy interrupt is by modifying the code to
>> skip MSI-X and MSI.  The ath10k code is taken from ath.git master
>> branch.
>>
>> Thanks a lot for the link.  If someone can verify that the module on
>> Archer C7 V1 is actually a V1 hardware, then it will be very helpful
>> also.
>
> To my knowledge Archer C7 has QCA9880 v1. Because the PCI ids are the
> same I'm not currently aware of any way to detect from ath10k if QCA9880
> v1 or v2 :(
>
> --
> Kalle Valo

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Crash when loading module

[Bartosz Markowski]

Hi,

Here's a topic worth looking at:
https://forum.openwrt.org/viewtopic.php?pid=210840#p210840
I'm not sure about the source and if the information is confirmed, but
please check your Archer 7 revision and compare.

On 28 August 2013 08:32, Kalle Valo <kvalo@qca.qualcomm.com> wrote:
> Hi Zaki,
>
> please CC the mailing list when replying. I do not have time to give
> private support.
>
> Zaki <zaki.bm@gmail.com> writes:
>
>> The PCIe module is taken from Archer C7 V1 product's board.  I am not
>> so sure if it is V2 hardware, though. I just depends on the debugging
>> msg that says it is V2 device.
>>
>> The way i force it to use legacy interrupt is by modifying the code to
>> skip MSI-X and MSI.  The ath10k code is taken from ath.git master
>> branch.
>>
>> Thanks a lot for the link.  If someone can verify that the module on
>> Archer C7 V1 is actually a V1 hardware, then it will be very helpful
>> also.
>
> To my knowledge Archer C7 has QCA9880 v1. Because the PCI ids are the
> same I'm not currently aware of any way to detect from ath10k if QCA9880
> v1 or v2 :(
>
> --
> Kalle Valo
>
> _______________________________________________
> ath10k mailing list
> ath10k@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/ath10k



-- 
Bartosz

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Crash when loading module

[Zaki]

Bartosz,  thanks for your input.  I will check it and do the compare.

On Wed, Aug 28, 2013 at 2:58 PM, Bartosz Markowski
<bartosz.markowski@tieto.com> wrote:
> Hi,
>
> Here's a topic worth looking at:
> https://forum.openwrt.org/viewtopic.php?pid=210840#p210840
> I'm not sure about the source and if the information is confirmed, but
> please check your Archer 7 revision and compare.
>
> On 28 August 2013 08:32, Kalle Valo <kvalo@qca.qualcomm.com> wrote:
>> Hi Zaki,
>>
>> please CC the mailing list when replying. I do not have time to give
>> private support.
>>
>> Zaki <zaki.bm@gmail.com> writes:
>>
>>> The PCIe module is taken from Archer C7 V1 product's board.  I am not
>>> so sure if it is V2 hardware, though. I just depends on the debugging
>>> msg that says it is V2 device.
>>>
>>> The way i force it to use legacy interrupt is by modifying the code to
>>> skip MSI-X and MSI.  The ath10k code is taken from ath.git master
>>> branch.
>>>
>>> Thanks a lot for the link.  If someone can verify that the module on
>>> Archer C7 V1 is actually a V1 hardware, then it will be very helpful
>>> also.
>>
>> To my knowledge Archer C7 has QCA9880 v1. Because the PCI ids are the
>> same I'm not currently aware of any way to detect from ath10k if QCA9880
>> v1 or v2 :(
>>
>> --
>> Kalle Valo
>>
>> _______________________________________________
>> ath10k mailing list
>> ath10k@lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/ath10k
>
>
>
> --
> Bartosz

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Crash when loading module

[Zaki]

OK. Just confirmed that my module is "QCA9880-AR1A NFH090.001C 1251".
According to the link posted by Bartosz, it is a V1 hardware.  Too bad
for me, then :( .

On Wed, Aug 28, 2013 at 3:13 PM, Zaki <zaki.bm@gmail.com> wrote:
> Bartosz,  thanks for your input.  I will check it and do the compare.
>
> On Wed, Aug 28, 2013 at 2:58 PM, Bartosz Markowski
> <bartosz.markowski@tieto.com> wrote:
>> Hi,
>>
>> Here's a topic worth looking at:
>> https://forum.openwrt.org/viewtopic.php?pid=210840#p210840
>> I'm not sure about the source and if the information is confirmed, but
>> please check your Archer 7 revision and compare.
>>
>> On 28 August 2013 08:32, Kalle Valo <kvalo@qca.qualcomm.com> wrote:
>>> Hi Zaki,
>>>
>>> please CC the mailing list when replying. I do not have time to give
>>> private support.
>>>
>>> Zaki <zaki.bm@gmail.com> writes:
>>>
>>>> The PCIe module is taken from Archer C7 V1 product's board.  I am not
>>>> so sure if it is V2 hardware, though. I just depends on the debugging
>>>> msg that says it is V2 device.
>>>>
>>>> The way i force it to use legacy interrupt is by modifying the code to
>>>> skip MSI-X and MSI.  The ath10k code is taken from ath.git master
>>>> branch.
>>>>
>>>> Thanks a lot for the link.  If someone can verify that the module on
>>>> Archer C7 V1 is actually a V1 hardware, then it will be very helpful
>>>> also.
>>>
>>> To my knowledge Archer C7 has QCA9880 v1. Because the PCI ids are the
>>> same I'm not currently aware of any way to detect from ath10k if QCA9880
>>> v1 or v2 :(
>>>
>>> --
>>> Kalle Valo
>>>
>>> _______________________________________________
>>> ath10k mailing list
>>> ath10k@lists.infradead.org
>>> http://lists.infradead.org/mailman/listinfo/ath10k
>>
>>
>>
>> --
>> Bartosz

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Crash when loading module

[Kalle Valo]

Zaki <zaki.bm@gmail.com> writes:

> OK. Just confirmed that my module is "QCA9880-AR1A NFH090.001C 1251".
> According to the link posted by Bartosz, it is a V1 hardware.  Too bad
> for me, then :( .

There is this, but I can't make any promises:

http://lists.infradead.org/pipermail/ath10k/2013-July/000021.html

-- 
Kalle Valo

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Crash when loading module

[Kalle Valo]

Kalle Valo <kvalo@qca.qualcomm.com> writes:

> Zaki <zaki.bm@gmail.com> writes:
>
>> The PCIe module is taken from Archer C7 V1 product's board.  I am not
>> so sure if it is V2 hardware, though. I just depends on the debugging
>> msg that says it is V2 device.
>>
>> The way i force it to use legacy interrupt is by modifying the code to
>> skip MSI-X and MSI.  The ath10k code is taken from ath.git master
>> branch.
>>
>> Thanks a lot for the link.  If someone can verify that the module on
>> Archer C7 V1 is actually a V1 hardware, then it will be very helpful
>> also.
>
> To my knowledge Archer C7 has QCA9880 v1. Because the PCI ids are the
> same I'm not currently aware of any way to detect from ath10k if QCA9880
> v1 or v2 :(

Found a way now:

http://lists.infradead.org/pipermail/ath10k/2013-August/000279.html

Zaki, I would appreciate if you could test that patch and see if ath10k
still crashes (it shouldn't). And if you want to provide your full name
I can add your name to the Reported-by tag.

-- 
Kalle Valo

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Crash when loading module

[Kalle Valo]

Christian Lamparter <chunkeey@googlemail.com> writes:

>> Found a way now:
>>
>> http://lists.infradead.org/pipermail/ath10k/2013-August/000279.html
>>
>> Zaki, I would appreciate if you could test that patch and see if ath10k
>> still crashes (it shouldn't). And if you want to provide your full name
>> I can add your name to the Reported-by tag.
>
> I get this before it crashes now. (This is a QCA-9880-AR1A from TP-Link Archer C7)
>
> [  271.140000] ath10k: Warning: chip id 0x043200ff is not known, expect problems
>
> Could it be, that this is just a typo, or are there several v1.0 chip_ids?

I found out (after I sent the patch) that the chip id actually contains
more than just the hardware revision. So when testing for hard revision
(hw1.0 vs hw2.0) we need to do:

(chip_id & 0xf00) >> 8

0x2 will be hw2.0 and 0x0 will be hw1.0. This seems to match with the
chip ids my boards have and also you what you have. I'll send a v2,
thanks for the feedback.

-- 
Kalle Valo

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k