Compex WLE900N5-23

Michal Kazior michal.kazior at tieto.com
Thu Apr 30 00:05:19 PDT 2015


On 30 April 2015 at 08:46, Yeoh Chun-Yeow <yeohchunyeow at gmail.com> wrote:
> Hi, Michal
>
> Attached is the partial log. Hopefully it is useful for you.

Thanks.

It looks like the device goes through the cold reset. I would expect
the device to work after that. Maybe WLE900N5-23 board is wired
differently or 10.2 doesn't reset a few HW registers anymore.

Anyway, the device seems to crash after it's started and upped. It
gets to transmit 2 beacons and then it crashes. I didn't see any Rx
(WMI nor HTT) events which leads me to think that the HW MAC Rx may
contain some invalid/stale DMA pointers which subsequently lead to CE
failure and device CPU crash. If you could run the device on a x86
with IOMMU enabled it'd be probably possible to narrow down the search
by looking at DMAR error addresses. Hmm.. I recall there are a few HW
MAC Rx rings. Maybe 10.2 uses more than 10.1 did but doesn't reset
them properly?

I assume you need to reboot the machine before it starts to work
again, no? Or does driver reload alone help? Did you try to `echo 1 >
/sys/bus/pci/devices/xxxx:xx:xx.x/remove` and `echo 1 >
/sys/bus/pci/rescan`? Or did you try to swap fw from 10.2 to 10.1
(needs driver reload) after the "bad bad crashes" start to appear?


Michał



More information about the ath10k mailing list