[OpenWrt-Devel] ath10k memory leak on 19.07 branch and mikrotik RB952Ui-5ac2nD?

Hauke Mehrtens hauke at hauke-m.de
Sat Dec 7 08:19:17 EST 2019


On 12/6/19 7:02 PM, Ben Greear wrote:
> On 12/6/19 9:44 AM, Joe Ayers wrote:
>>>
>>>
>>>> Possibly the same symptoms don't exist on 128MB RAM devices.
>>>
>>> Like there is some if condition, which is doing some nasty things on 64M
>>> devices? I admit, that I don't have ath10k-ct source code under my
>>> pillow, but
>>> it doesn't make much sense to me.
>>>
>>>> Comparable results between above and my 64MB device.   However, if the
>>>> sleep time is extended the consumption is more
>>>
>>> Ok, I'll let it run overnight with 120s sleep in between.
>>>
>>>> I suspect this is not the intended behavior.
>>>
>>> No its not and it's even strange, that I'm not seeing it here if it
>>> should
>>> happen in the "default setup". Maybe its because:
>>>
>>> 1. You're using custom image (I'm using official prebuilt images)
>>> 2. You're not providing all the steps needed to reproduce the issue
>>> 3. I've way different hardware
>>>
>>> Every detail could make huge difference.
>>>
>>> -- ynezz
>>
>> On the device I am testing, it is both (2GHz) ath9k and (5GHz) ath10k.
>>    These look to be related patches to this issue:
>>
>> 960-0010-ath10k-limit-htt-rx-ring-size.patch
>> 960-0011-ath10k-limit-pci-buffer-size.patch
>>
>> In the v19.07.0-rc2 build for the rb-nor-flash-16M-ac ar71xx image,
>> these patches are applied to backports-4.19.85-1, but don't seem to be
>> applied to ath10k-ct-2019-09-09-5e8cd86f.    Should ath10k-ct have
>> these and other patches?    The device's installed packages do include
>> ath10k-ct (from downloads.openwrt.org installed image).
> 
> I think that if you need the patches for upstream ath10k, then you
> should also apply
> the patches to ath10k-ct.
> 
> Platforms with more memory probably do not need or benefit from those
> patches.
> 
> Thanks,
> Ben
> 
Hi,

I am fine with adding these patches to ath10k-ct if they solve problems.

It would be nice if a description could be added to the patches so I
know what kind of buffers are reduced there.

Will someone send a patch?

Hauke

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://lists.infradead.org/pipermail/openwrt-devel/attachments/20191207/7b18ea38/attachment.sig>
-------------- next part --------------
_______________________________________________
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel


More information about the openwrt-devel mailing list