Comment 38 for bug 1730924

I've spent some time looking at the source code and trying to debug what is happening. I'm using the kvalo wireless-drivers-next pending branch, and I've applied recent patches sent to lkml that seem valuable; but the problems persist.

It seems as though in the 'normal' case we get an unusual number of 'max count reached while accessing sleep cookie' messages. The usleep calls in that case make me suspicious, and being slow to wake up would seem to be a good explanation of the performance we see. (That is, sending a packet is fast, but responding is slow). It also seems like we're in a rather startling loop of awake/sleep events; doing multiple spins within a millisecond, as far as I can tell.

In the disconnect case, I see an 'invalid cmd resp' error followed by 'There is no command but got a command response'.

I have now started booting with debug_mask=0xffffffff; I'll replace my logs with those, as they appear to be more complete.

For reference, here are related areas of the code:

Sleep message:
https://kernel.googlesource.com/pub/scm/linux/kernel/git/kvalo/wireless-drivers-next/+/pending/drivers/net/wireless/marvell/mwifiex/pcie.c#377

There is no command:
https://kernel.googlesource.com/pub/scm/linux/kernel/git/kvalo/wireless-drivers-next/+/pending/drivers/net/wireless/marvell/mwifiex/pcie.c#1637
(although intriguingly to my naive mind, that code block is related to the sleep check as well).