Re: KA46/48 vsbus DMA fallthrough

[Felix Deichmann <m4j0rd0m0%gmail.com@localhost>]

2015-05-10 5:55 GMT+02:00 John Klos <john%ziaspace.com@localhost>:
>>> Do you think it would be a good idea to start searching/debugging in the
>>> match routine of if_le_vsbus.c now?
>>
>> Looking at the match in if_le_vsbus I would be tempted to increase i
>> for the DELAY loop from 100 to 1000, then print out the value after
>> the loop to see how long it takes (and if the longer wait helps).
>
> I noticed this problem on one of my 4000/60 machines, but because I only had
> the one datapoint, I didn't report it. I should have a machine ready for
> testing soon and can test this recommendation.

Did you test this already, John?
With a recent network loader now closing network devices (i. e.
setting STOP bit for le(4)) before jumping into the kernel, I cannot
reproduce the issue of a disappeared le(4) anymore (or easily) on my
4000/60.
Maybe the now longer period in which the Lance is stopped helped.

What is also interesting is a note in the KA46 System Board
Specification, section 9.9:
    "The network controller hardware requires up to five seconds after
power on to become stable. Self-test routines must delay at least this
time before attempting to use the controller for either internal or
external testing."

If re-initializing the Lance after a STOP condition is in a similar
league, we could be definitely too fast at the moment...

Regards,
Felix