BUG: brcmfmac: add firmware feature detection for gscan feature breaks 43430a1

BUG: brcmfmac: add firmware feature detection for gscan feature breaks 43430a1

[Ian Molton]

Resent due to subject error. Apologies, Hante :)

Hi Folks,

9fe929aaace6  brcmfmac: add firmware feature detection for gscan feature

Completely breaks 43430a1 support.

This was tested on top of my last patchset - but the same occurs on top
of vanilla 4.13-rc1

[   15.703014]  [1 ] core 0x800:49 base 0x18000000 wrap 0x18100000
[   15.703025]  [2 ] core 0x812:39 base 0x18001000 wrap 0x18101000
[   15.703036]  [3 ] core 0x829:21 base 0x18002000 wrap 0x18102000
[   15.703044]  [4 ] core 0x82a:9  base 0x18003000 wrap 0x18103000
[   15.703052]  [5 ] core 0x80e:22 base 0x18004000 wrap 0x18104000
[   15.703059]  [6 ] core 0x135:0  base 0x00000000 wrap 0x18105000
[   15.703067]  [7 ] core 0x240:0  base 0x00000000 wrap 0x18106000
[   15.703070] Enter
[   15.735280] RAM: base=0x0 size=524288 (0x80000) sr=65536 (0x10000)
[   15.735293] Found BCM43430, AXI backplane, rev=1
[   15.735461] ccrev=49, pmurev=24, pmucaps=0x39d25f18
[   15.904051] Enter
[   15.974045] Enter
[   15.984378] brcmfmac: brcmf_c_preinit_dcmds: Firmware version = wl0:
Aug 29 2016 20:48:16 version 7.45.41.26 (r640327) FWID 01-4527cfab
[   16.009363] brcmfmac: brcmf_sdio_hostmail: Unknown mailbox data
content: 0x40012
...
[   18.566406] brcmfmac: brcmf_sdio_bus_rxctl: resumed on timeout
...
[   21.126419] brcmfmac: brcmf_sdio_bus_rxctl: resumed on timeout
[   23.686371] brcmfmac: brcmf_sdio_bus_rxctl: resumed on timeout
...
[   26.246357] brcmfmac: brcmf_sdio_bus_rxctl: resumed on timeout
[   28.806347] brcmfmac: brcmf_sdio_bus_rxctl: resumed on timeout
[   31.366355] brcmfmac: brcmf_sdio_bus_rxctl: resumed on timeout
[   33.926319] brcmfmac: brcmf_proto_bcdc_query_dcmd:
brcmf_proto_bcdc_msg failed w/status -110
[   36.486448] brcmfmac: brcmf_proto_bcdc_query_dcmd:
brcmf_proto_bcdc_msg failed w/status -110
[   36.497332] brcmfmac: brcmf_cfg80211_attach: Failed to get D11
version (-110)
[   36.504059] brcmfmac: brcmf_bus_started: failed: -12
[   36.517686] brcmfmac: brcmf_sdio_firmware_callback: dongle is not
responding
[   36.586256] Enter

-Ian

Re: BUG: brcmfmac: add firmware feature detection for gscan feature breaks 43430a1

[Ian Molton]

On 17/07/17 18:50, Ian Molton wrote:
> Resent due to subject error. Apologies, Hante :)
> 
> Hi Folks,
> 
> 9fe929aaace6  brcmfmac: add firmware feature detection for gscan feature
> 
> Completely breaks 43430a1 support.

Digging a bit deeper, this would *appear* to be one of a couple of
possibilities;

1) brcmf_sdio_hostmail() is missing code to handle a response of type
   0x40012
2) The firmware is failing when asked to handle the new requiests, and
   is going to la-la land.

Decoding the reply gives us HMB_DATA_DEVREADY or-ed with 0x40010

Since the upper half-word is masked, that leaves us with 0x10 which does
not appear to be a defined state.

The driver does not handle this well, in any event.

My guess is that you are working on top of newer firmware than I have.
Can you let me know what the latest public firmware is, and where I can
get it?

I have version wl0: Aug 29 2016 20:48:16 version 7.45.41.26 (r640327)
FWID 01-4527cfab

Thanks,

-Ian

Re: BUG: brcmfmac: add firmware feature detection for gscan feature breaks 43430a1

[Arend van Spriel]

On 17-07-17 20:31, Ian Molton wrote:
> On 17/07/17 18:50, Ian Molton wrote:
>> Resent due to subject error. Apologies, Hante :)
>>
>> Hi Folks,
>>
>> 9fe929aaace6  brcmfmac: add firmware feature detection for gscan feature
>>
>> Completely breaks 43430a1 support.
> 
> Digging a bit deeper, this would *appear* to be one of a couple of
> possibilities;
> 
> 1) brcmf_sdio_hostmail() is missing code to handle a response of type
>    0x40012
> 2) The firmware is failing when asked to handle the new requiests, and
>    is going to la-la land.

taken a wrong turn straight into la-la land.

> Decoding the reply gives us HMB_DATA_DEVREADY or-ed with 0x40010
> 
> Since the upper half-word is masked, that leaves us with 0x10 which does
> not appear to be a defined state.
> 
> The driver does not handle this well, in any event.
> 
> My guess is that you are working on top of newer firmware than I have.
> Can you let me know what the latest public firmware is, and where I can
> get it?
> 
> I have version wl0: Aug 29 2016 20:48:16 version 7.45.41.26 (r640327)
> FWID 01-4527cfab

Not so much newer firmware, but I used a different chip when
implementing the multi-scheduled scan patches. I tried some other
devices, but not the bcm4343{0,8}. Now as far as I can see over here the
brcmfmac43430-sdio.bin is not built with gscan support so it should just
return firmware errorcode, ie. -23. So it would be interesting to see
what firmware has to say. Can you load the driver with
ignore_probe_fail=1 and dump the forensics file in debugfs.

Regards,
Arend

Re: BUG: brcmfmac: add firmware feature detection for gscan feature breaks 43430a1

[Ian Molton]

On 17/07/17 20:53, Arend van Spriel wrote:
>> 2) The firmware is failing when asked to handle the new requiests, and
>>    is going to la-la land.
> taken a wrong turn straight into la-la land.

Doh!

>> I have version wl0: Aug 29 2016 20:48:16 version 7.45.41.26 (r640327)
>> FWID 01-4527cfab
> Not so much newer firmware, but I used a different chip when
> implementing the multi-scheduled scan patches. I tried some other
> devices, but not the bcm4343{0,8}. Now as far as I can see over here the
> brcmfmac43430-sdio.bin is not built with gscan support so it should just
> return firmware errorcode, ie. -23. So it would be interesting to see
> what firmware has to say. Can you load the driver with
> ignore_probe_fail=1 and dump the forensics file in debugfs.

I don't appear to get anything useful in debugfs. Where should I be looking?

-Ian

Re: BUG: brcmfmac: add firmware feature detection for gscan feature breaks 43430a1

[Arend van Spriel]

On 17-07-17 22:13, Ian Molton wrote:
> On 17/07/17 20:53, Arend van Spriel wrote:
>>> 2) The firmware is failing when asked to handle the new requiests, and
>>>    is going to la-la land.
>> taken a wrong turn straight into la-la land.
> 
> Doh!
> 
>>> I have version wl0: Aug 29 2016 20:48:16 version 7.45.41.26 (r640327)
>>> FWID 01-4527cfab
>> Not so much newer firmware, but I used a different chip when
>> implementing the multi-scheduled scan patches. I tried some other
>> devices, but not the bcm4343{0,8}. Now as far as I can see over here the
>> brcmfmac43430-sdio.bin is not built with gscan support so it should just
>> return firmware errorcode, ie. -23. So it would be interesting to see
>> what firmware has to say. Can you load the driver with
>> ignore_probe_fail=1 and dump the forensics file in debugfs.
> 
> I don't appear to get anything useful in debugfs. Where should I be looking?

In <debugfs_mountpoint>/brcmfmac/*/forensics, but you have to load the
driver with modules parameter 'ignore_probe_fail=1' to avoid the driver
remove tearing down all debugfs stuff.

Re: BUG: brcmfmac: add firmware feature detection for gscan feature breaks 43430a1

[Ian Molton]

On 17/07/17 21:36, Arend van Spriel wrote:

> In <debugfs_mountpoint>/brcmfmac/*/forensics, but you have to load the
> driver with modules parameter 'ignore_probe_fail=1' to avoid the driver
> remove tearing down all debugfs stuff.

modinfo shows:

parm:           txglomsz:Maximum tx packet chain size [SDIO] (int)
parm:           debug:Level of debug output (int)
parm:           p2pon:Enable legacy p2p management functionality (int)
parm:           feature_disable:Disable features (int)
parm:           alternative_fw_path:Alternative firmware path (string)
parm:           fcmode:Mode of firmware signalled flow control (int)
parm:           roamoff:Do not use internal roaming engine (int)

Re: BUG: brcmfmac: add firmware feature detection for gscan feature breaks 43430a1

[Arend van Spriel]

On 17-07-17 22:41, Ian Molton wrote:
> On 17/07/17 21:36, Arend van Spriel wrote:
> 
>> In <debugfs_mountpoint>/brcmfmac/*/forensics, but you have to load the
>> driver with modules parameter 'ignore_probe_fail=1' to avoid the driver
>> remove tearing down all debugfs stuff.
> 
> modinfo shows:
> 
> parm:           txglomsz:Maximum tx packet chain size [SDIO] (int)
> parm:           debug:Level of debug output (int)
> parm:           p2pon:Enable legacy p2p management functionality (int)
> parm:           feature_disable:Disable features (int)
> parm:           alternative_fw_path:Alternative firmware path (string)
> parm:           fcmode:Mode of firmware signalled flow control (int)
> parm:           roamoff:Do not use internal roaming engine (int)
> 

Seems you did not build with CONFIG_BRCMDBG?

common.c:

#ifdef DEBUG
/* always succeed brcmf_bus_started() */
static int brcmf_ignore_probe_fail;
module_param_named(ignore_probe_fail, brcmf_ignore_probe_fail, int, 0);
MODULE_PARM_DESC(ignore_probe_fail, "always succeed probe for debugging");
#endif

Regards,
Arend

Re: BUG: brcmfmac: add firmware feature detection for gscan feature breaks 43430a1

[Ian Molton]

On 17/07/17 22:31, Arend van Spriel wrote:
> 
> 
> On 17-07-17 22:41, Ian Molton wrote:
>> On 17/07/17 21:36, Arend van Spriel wrote:
>>
>>> In <debugfs_mountpoint>/brcmfmac/*/forensics, but you have to load the
>>> driver with modules parameter 'ignore_probe_fail=1' to avoid the driver
>>> remove tearing down all debugfs stuff.
>>
>> modinfo shows:
>>
>> parm:           txglomsz:Maximum tx packet chain size [SDIO] (int)
>> parm:           debug:Level of debug output (int)
>> parm:           p2pon:Enable legacy p2p management functionality (int)
>> parm:           feature_disable:Disable features (int)
>> parm:           alternative_fw_path:Alternative firmware path (string)
>> parm:           fcmode:Mode of firmware signalled flow control (int)
>> parm:           roamoff:Do not use internal roaming engine (int)
>>
> 
> Seems you did not build with CONFIG_BRCMDBG?

No - it failed the build last time I tried, and I havent had a chance to
look into why yet.

Possibly its fixed since - last I tried was around 4.12-rc4 or something.

-Ian

Re: BUG: brcmfmac: add firmware feature detection for gscan feature breaks 43430a1

[Ian Molton]

Hi Arend,

This is all there is in the forensics file, hope its of some use.

# cat /sys/kernel/debug/brcmfmac/mmc1\:0001\:1/forensics
Enter
hndarm_armr addr: 0x18003000, cr4_idx: 0
000000.000
RTE (SDIO-CDC) 7.45.41.26 (r640327) on BCM43430 r1 @ 37.4/81.0/81.0MHz
000000.001 sdpcmdcdc0: Broadcom SDPCMD CDC driver
000000.005 reclaim section 0: Returned 47716 bytes to the heap
000000.007 wlc_bmac_info_init: host_enab 1
000000.025 wl0: Broadcom BCM43430 802.11 Wireless Controller 7.45.41.26
(r640327)
000000.026 TCAM: 256 used: 179 exceed:0
000000.027 reclaim section 1: Returned 81268 bytes to the heap

-Ian

Re: BUG: brcmfmac: add firmware feature detection for gscan feature breaks 43430a1

[Arend van Spriel]

On 18-07-17 00:10, Ian Molton wrote:
> Hi Arend,
> 
> This is all there is in the forensics file, hope its of some use.
> 
> # cat /sys/kernel/debug/brcmfmac/mmc1\:0001\:1/forensics
> Enter
> hndarm_armr addr: 0x18003000, cr4_idx: 0
> 000000.000
> RTE (SDIO-CDC) 7.45.41.26 (r640327) on BCM43430 r1 @ 37.4/81.0/81.0MHz
> 000000.001 sdpcmdcdc0: Broadcom SDPCMD CDC driver
> 000000.005 reclaim section 0: Returned 47716 bytes to the heap
> 000000.007 wlc_bmac_info_init: host_enab 1
> 000000.025 wl0: Broadcom BCM43430 802.11 Wireless Controller 7.45.41.26
> (r640327)
> 000000.026 TCAM: 256 used: 179 exceed:0
> 000000.027 reclaim section 1: Returned 81268 bytes to the heap

You probably guessed it, but this does not show anything useful. The
behavior suggested a firmware crash, but that does not seem to be
happening. Maybe it is the SDIO core driver on the firmware side which
is in la-la land.

Regards,
Arend