ath11k: QCN9074: Incompatible board-2.bin for 6GHz

ath11k: QCN9074: Incompatible board-2.bin for 6GHz

[Sven Eckelmann]

Hi,

I wanted to use an 8devices Pineapple6 device on x86. But the board_id of this 
card came always in as 255 (and not as a valid 6GHz capable board_id). I've 
contacted the support and they told me to use the QCA BDF with the board_id 
162 for this card. So I've downloaded 
2.4.0.1.r1/WLAN.HK.2.4.0.1.r1-00019-QCAHKSWPL_SILICONZ-1/board-2.bin [1], 
extracted it and moved it to /lib/firmware/ath11k/QCN9074/hw1.0/board.bin. 
I've also added the (not yet upstreamed) 6GHz regulatory support to 
ath11k [2]. But to my surprise, the WMI_REG_CHAN_LIST_CC_EXT_EVENTID never 
came in. Instead, a WMI_REG_CHAN_LIST_CC_EVENTID was received. This resulted 
in a output like:

    ath11k_pci 0000:01:00.0: No reg rules available
    ath11k_pci 0000:01:00.0: failed to extract regulatory info from received event
    ------------[ cut here ]------------
    WARNING: at drivers/net/wireless/ath/ath11k/wmi.c:7255

The warning from ath11k (which breaks the device btw) is from

    if (!(reg_info->num_2g_reg_rules + reg_info->num_5g_reg_rules)) {
    	ath11k_warn(ab, "No regulatory rules available in the event info\n");
    	kfree(tb);
    	return -EINVAL;
    }

I found out that this is actually caused by the BDF. If I've extracted the 
board.bin from the TIP board-2.bin [3] (which seems to come from the QCA 
ATH.11.4 CS release). With this version, it seems to work fine. I haven't yet 
found out what the relevant change is - but if someone has hints then this 
would could be interesting for other devices which should be added to things 
like OpenWrt.

Maybe Kalle Valo could upload a new board-2.bin for QCN9074 hw1.0 which has
the problem solved.


And it would also be nice when the authors of the 
WMI_REG_CHAN_LIST_CC_EXT_EVENTID support would send this stuff upstream. I 
just find it interesting that the assoc patches are already part of
upstream [4,5] - even when the reg support is still missing. There are a lot 
more patches in there which are missing upstream.

Kind regards,
	Sven

[1] https://github.com/kvalo/ath11k-firmware/commit/ae4f4f64ed14097dc24a8d686fa60e932cba4868/QCN9074/hw1.0/testing/2.4.0.1.r1/WLAN.HK.2.4.0.1.r1-00019-QCAHKSWPL_SILICONZ-1/board-2.bin
[2] https://source.codeaurora.org/quic/qsdk/oss/system/feeds/wlan-open/tree/mac80211/patches/302-ath11k-Add-support-to-parse-new-wmi-event-for-6G-reg.patch?h=NHSS.QSDK.11.4.0.5
[3] https://github.com/Telecominfraproject/wlan-ap/blob/main/feeds/wifi-ax/ath11k-firmware/files/QCN9000/board-2.bin
[4] https://source.codeaurora.org/quic/qsdk/oss/system/feeds/wlan-open/tree/mac80211/patches/206-ath11k-add-6ghz-params-in-peer-assoc-cmd.patch?h=NHSS.QSDK.11.4.0.5
[5] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=c3a7d7eb4c9853bb457b792cef42ddd4a029a914
[6] https://source.codeaurora.org/quic/qsdk/oss/system/feeds/wlan-open/tree/mac80211/patches/302-ath11k-add-177-channel-to-5G-channel-list.patch?h=NHSS.QSDK.11.4.0.5



-- 
ath11k mailing list
ath11k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath11k

Re: ath11k: QCN9074: Incompatible board-2.bin for 6GHz

[John Crispin]

On 26.11.21 11:52, Sven Eckelmann wrote:
> I found out that this is actually caused by the BDF. If I've extracted the
> board.bin from the TIP board-2.bin [3] (which seems to come from the QCA
> ATH.11.4 CS release). With this version, it seems to work fine. I haven't yet
> found out what the relevant change is - but if someone has hints then this
> would could be interesting for other devices which should be added to things
> like OpenWrt.

Kalle's repo is outdated by several releases. This is the new upstream 
repo.

the error you are seeing is an ABI breakage between pre/post sfp11.4 
fw/bdf. using a fw >= 2.5.x requires a BDF generated with / derviced 
from >= sfp11.4

the BDFs inside the TIP repo come from here
--> https://github.com/quic/upstream-wifi-fw/

	John

-- 
ath11k mailing list
ath11k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath11k

Re: ath11k: QCN9074: Incompatible board-2.bin for 6GHz

[Mark Herbert]

Will there be a new FW for QCA6390?  It seems to be the most outdated in 
Kalle's repo :(

On 26.11.2021 17:54, John Crispin wrote:
>
>
> On 26.11.21 11:52, Sven Eckelmann wrote:
>> I found out that this is actually caused by the BDF. If I've 
>> extracted the
>> board.bin from the TIP board-2.bin [3] (which seems to come from the QCA
>> ATH.11.4 CS release). With this version, it seems to work fine. I 
>> haven't yet
>> found out what the relevant change is - but if someone has hints then 
>> this
>> would could be interesting for other devices which should be added to 
>> things
>> like OpenWrt.
>
> Kalle's repo is outdated by several releases. This is the new upstream 
> repo.
>
> the error you are seeing is an ABI breakage between pre/post sfp11.4 
> fw/bdf. using a fw >= 2.5.x requires a BDF generated with / derviced 
> from >= sfp11.4
>
> the BDFs inside the TIP repo come from here
> --> https://github.com/quic/upstream-wifi-fw/
>
>     John
>

-- 
ath11k mailing list
ath11k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath11k

Re: ath11k: QCN9074: Incompatible board-2.bin for 6GHz

[Sven Eckelmann]

On Friday, 26 November 2021 11:52:42 CET Sven Eckelmann wrote:
> I found out that this is actually caused by the BDF. If I've extracted the 
> board.bin from the TIP board-2.bin [3] (which seems to come from the QCA 
> ATH.11.4 CS release). With this version, it seems to work fine. I haven't yet 
> found out what the relevant change is - but if someone has hints then this 
> would could be interesting for other devices which should be added to things 
> like OpenWrt.

I had are look at that for another board. And I found out that they changed 
the size of various sections in the BDF and moved stuff accordingly. And the 
one in Kalle's ath11k-firmware repository is even using different IDs. This 
caused the ABI breakage which John described in the other reply.

No idea how the firmware is supposed to handle this. At the moment it is only 
clear that it didn't handle it :(

Kind regards,
	Sven




-- 
ath11k mailing list
ath11k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath11k

Re: ath11k: QCN9074: Incompatible board-2.bin for 6GHz

[Kalle Valo]

Sven Eckelmann <sven@narfation.org> writes:

> On Friday, 26 November 2021 11:52:42 CET Sven Eckelmann wrote:
>> I found out that this is actually caused by the BDF. If I've extracted the 
>> board.bin from the TIP board-2.bin [3] (which seems to come from the QCA 
>> ATH.11.4 CS release). With this version, it seems to work fine. I haven't yet 
>> found out what the relevant change is - but if someone has hints then this 
>> would could be interesting for other devices which should be added to things 
>> like OpenWrt.
>
> I had are look at that for another board. And I found out that they changed 
> the size of various sections in the BDF and moved stuff accordingly. And the 
> one in Kalle's ath11k-firmware repository is even using different IDs. This 
> caused the ABI breakage which John described in the other reply.
>
> No idea how the firmware is supposed to handle this. At the moment it is only 
> clear that it didn't handle it :(

I have told the firmware team numerious times that the board file should
be backwards compatible as it creates problems for upstream. And I do
try to keep everything in ath11k-firmware (and in linux-firmware)
backwards compatible, but if they are not please do let me know. For
other repos I cannot do much, if any.

-- 
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches

-- 
ath11k mailing list
ath11k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath11k

Re: ath11k: QCN9074: Incompatible board-2.bin for 6GHz

[Sven Eckelmann]

[-- Attachment #1.1: Type: text/plain, Size: 1984 bytes --]

On Tuesday, 21 December 2021 15:52:03 CET Kalle Valo wrote:
> I have told the firmware team numerious times that the board file should
> be backwards compatible as it creates problems for upstream. And I do
> try to keep everything in ath11k-firmware (and in linux-firmware)
> backwards compatible, but if they are not please do let me know.

Please check the first message to find such a problem. Ok, it was not about 
board files being really backward compatible - it actually is about the 
firmware not being backward compatible and not working with the board-2.bin/
BDFs in the same repository.

So if I get
https://github.com/kvalo/ath11k-firmware/commit/ae4f4f64ed14097dc24a8d686fa60e932cba4868/QCN9074/hw1.0/testing/2.4.0.1.r1/WLAN.HK.2.4.0.1.r1-00019-QCAHKSWPL_SILICONZ-1/board-2.bin
(extract board_id 162 as board.bin - because nobody seems to initialize the 
board_id in the OTP)

and use it with

https://github.com/kvalo/ath11k-firmware/tree/master/QCN9074/hw1.0/2.5.0.1/WLAN.HK.2.5.0.1-01100-QCAHKSWPL_SILICONZ-1

then it causes the initialization error described in the first mail. And the 
firmware used here is the only non-testing firmware available for QCN9074. And 
the board-2.bin is the only available one from your repository for QCN9074.

> For other repos I cannot do much, if any.

The described problem was NOT that files from other repositories were mixed. 
Actually, the solution was to mix the firmware with the BDF from a different 
repository.

And I didn't only see this with BDFs from your repository but also from non-
upstreamed BDFs from manufacturers which embedded a pine "card" on the PCB of 
their AP.

I was able to fix the problem for the non-upstream BDFs for this AP by 
rearranging the BDF sections to match the new layout. I didn't try this with 
the BDFs in your repository because also the IDs seemed to have changed (but I 
didn't analyze it further). Simply using the BDFs from a different repository 
was easier.

Kind regards,
	Sven

[-- Attachment #1.2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[-- Attachment #2: Type: text/plain, Size: 102 bytes --]

-- 
ath11k mailing list
ath11k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath11k

Re: ath11k: QCN9074: Incompatible board-2.bin for 6GHz

[Kalle Valo]

Sven Eckelmann <sven@narfation.org> writes:

> On Tuesday, 21 December 2021 15:52:03 CET Kalle Valo wrote:
>> I have told the firmware team numerious times that the board file should
>> be backwards compatible as it creates problems for upstream. And I do
>> try to keep everything in ath11k-firmware (and in linux-firmware)
>> backwards compatible, but if they are not please do let me know.
>
> Please check the first message to find such a problem. Ok, it was not about 
> board files being really backward compatible - it actually is about the 
> firmware not being backward compatible and not working with the board-2.bin/
> BDFs in the same repository.
>
> So if I get
> https://github.com/kvalo/ath11k-firmware/commit/ae4f4f64ed14097dc24a8d686fa60e932cba4868/QCN9074/hw1.0/testing/2.4.0.1.r1/WLAN.HK.2.4.0.1.r1-00019-QCAHKSWPL_SILICONZ-1/board-2.bin
> (extract board_id 162 as board.bin - because nobody seems to initialize the 
> board_id in the OTP)
>
> and use it with
>
> https://github.com/kvalo/ath11k-firmware/tree/master/QCN9074/hw1.0/2.5.0.1/WLAN.HK.2.5.0.1-01100-QCAHKSWPL_SILICONZ-1
>
> then it causes the initialization error described in the first mail. And the 
> firmware used here is the only non-testing firmware available for QCN9074. And 
> the board-2.bin is the only available one from your repository for QCN9074.

Ah, I forgot to mention that everything under "testing" directory is
work-in-progress, and might break backwards compatibility. To keep
things simple I now removed all backwards incompatible files from my
ath11k-firmware. Also now there's a proper board-2.bin for QCN9074,
hopefully that works better. Do let me know if you have any issues.

-- 
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches

-- 
ath11k mailing list
ath11k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath11k

Re: ath11k: QCN9074: Incompatible board-2.bin for 6GHz

[Sven Eckelmann]

[-- Attachment #1.1: Type: text/plain, Size: 579 bytes --]

On Wednesday, 13 April 2022 09:06:27 CEST Kalle Valo wrote:
[...]
> Ah, I forgot to mention that everything under "testing" directory is
> work-in-progress, and might break backwards compatibility. To keep
> things simple I now removed all backwards incompatible files from my
> ath11k-firmware. Also now there's a proper board-2.bin for QCN9074,
> hopefully that works better. Do let me know if you have any issues.

I have now used WLAN.HK.2.5.0.1-01208-QCAHKSWPL_SILICONZ-1 (+new board-2.bin) 
with the Pineapple6. And at least on first glance it seems to work.

Thanks,
	Sven

[-- Attachment #1.2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[-- Attachment #2: Type: text/plain, Size: 102 bytes --]

-- 
ath11k mailing list
ath11k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath11k

Re: ath11k: QCN9074: Incompatible board-2.bin for 6GHz

[Kalle Valo]

Sven Eckelmann <sven@narfation.org> writes:

> On Wednesday, 13 April 2022 09:06:27 CEST Kalle Valo wrote:
> [...]
>> Ah, I forgot to mention that everything under "testing" directory is
>> work-in-progress, and might break backwards compatibility. To keep
>> things simple I now removed all backwards incompatible files from my
>> ath11k-firmware. Also now there's a proper board-2.bin for QCN9074,
>> hopefully that works better. Do let me know if you have any issues.
>
> I have now used WLAN.HK.2.5.0.1-01208-QCAHKSWPL_SILICONZ-1 (+new board-2.bin) 
> with the Pineapple6. And at least on first glance it seems to work.

Great, thanks for testing.

-- 
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches

-- 
ath11k mailing list
ath11k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath11k