[LEDE-DEV] QCA Dakota support

[Alexis Green]

Hi Christian,

Thank you for the usb wireless device hint! I had an ath9k-htc dongle
kicking about and I'm seeing memory wastage with that device as well
(albeit slower and only happening while I'm pumping a lot of traffic
through it), so ath10k is likely off the hook for this one. In fact, I
went even farther, checking to see if sending traffic via ethernet
port causes problems and, again, there's memory wastage. Maybe it's a
kernel bug after all. One potential thing to try is to roll kernel to
an older version.

Best regards,

Alexis

On Mon, Nov 28, 2016 at 9:37 AM, Christian Lamparter
<chunkeey at googlemail.com> wrote:
> Hello Alexis,
>
> On Sunday, November 27, 2016 12:18:00 PM CET Alexis Green wrote:
>> Could you post .config for your build? I cloned your repo and
>> successfully built and installed an image, but I'm seeing some rather
>> weird behavior.
> I added it, but I don't have any magic values in the .config. I played
> around with adding LPAE for KVM but as it turns out the DMA driver doesn't
> like that.
>
> And there are definitely bugs. If you are looking for support, your best
> option is to ask: Matthew McClintock. Since developed most of the platform
> code while he was working for QCA, so he more familiar with the hardware
> and drivers as I just got this and put a port together for just my hardware...
> It's too bad that he didn't get to keep his board.
>
>> I get kernel oops (null derefrence) in bridge code when I connect a
>> client to WPA2 AP that is bridged. The oops is gone after I removed
>> the following patches (I'm sure it's just one of them, but I have not
>> had a chance  to figure out which one exactly it is).
>> target/linux/generic/patches-4.8/120-bridge_allow_receiption_on_disabled_port.patch
>> target/linux/generic/patches-4.8/640-bridge_no_eap_forward.patch
>> target/linux/generic/patches-4.8/641-bridge_always_accept_eap.patch
> I have to add Alvaro for this (I'm using his experimental 4.8-rc series
> which he is using for his raspberrypi 3 code). As for which one causes
> the issue: I think 640-bridge_no_eap_forward.patch is the one. It looks
> like the following patch in 4.8. changed the way the skb and skb2 works:
>
> commit b35c5f632b630183396a2ea2e2247ff8bbf2c94f
> Author: Nikolay Aleksandrov <nikolay at cumulusnetworks.com>
> Date:   Thu Jul 14 06:10:01 2016 +0300
>
>     net: bridge: drop skb2/skb0 variables and use a local_rcv boolean [0]
>
> Looking at this commit and the original patch [1]:
> I think we can drop the skb = NULL there (and maybe put the comment
> in front of the local_rcv = true;)
>
>> I'm also seeing rather fast memory leak/wastage when wireless devices
>> are up - takes 10-15 min to run out of memory. I tried using kmemleak,
>> but it doesn't report any suspected leaks. I guess I'll try some
>> tracing next.
> I've added a few new patches to mac80211-package from the current
> wireless-testing.git. If the issue persists, can you report this to
> the ath10k-devel mailing list?
>
> Note: You should check, if it's really the ath10k driver. This can
> be done by plugging in a usb-wifi device and setting it up. If it
> also fails then there's probably another issue with the patchset.
>
> Regards,
> Christian
>
> [0] <https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/net/bridge/br_input.c?id=b35c5f632b630183396a2ea2e2247ff8bbf2c94f>
>
> [1] <https://git.lede-project.org/?p=source.git;a=blob;f=target/linux/generic/patches-4.4/640-bridge_no_eap_forward.patch>