Sample 18-ntptime freezes after some seconds if a keyboard and mouse are attached about circle HOT 5 CLOSED

A short delay in the main loop in CKernel::Run() also mitigates the problem but does not solve it.

from circle.

I think I have found the problem. After doing some traces using the CTracer class I found a very strange system behaviour occurring. There were nested USB interrupts which should not happen and cannot be handled. The problem could not reproduced on the Raspberry Pi 2.

After checking the relevant paths in the source code (not a lot) I was guessing a side effect and changed the data cache clean and invalidate operation before/after an USB slave DMA operation from touching the entire data cache to the DMA data buffer range only which was already the case on the Raspberry Pi 2.

Now I cannot reproduce the problem in a ping test running one hour. It occurred after some seconds before the change. So I suppose the problem is fixed now. But such problems are difficult and further testing may be needed.

from circle.

Hi,

This afternoon I've tested the patch sending near 10.000 packets from two
sessions on same IP and a third session from another IP. Around two
uninterrupted hours, without packets lost and a keyboard and a mouse
working correctly.

Great work again!. :)

Thanks by your efforts.
José Luis

2015-09-27 15:10 GMT+02:00 rsta2 [email protected]:

I think I have found the problem. After doing some traces using the
CTracer class I found a very strange system behaviour occurring. There were
nested USB interrupts which should not happen and cannot be handled. The
problem could not reproduced on the Raspberry Pi 2.

After checking the relevant paths in the source code (not a lot) I was
guessing a side effect and changed the data cache clean and invalidate
operation before/after an USB slave DMA operation from touching the
entire data cache to the DMA data buffer range only which was already the
case on the Raspberry Pi 2.

Now I cannot reproduce the problem in a ping test running one hour. It
occurred after some seconds before the change. So I suppose the problem is
fixed now. But such problems are difficult and further testing may be
needed.

—
Reply to this email directly or view it on GitHub
#10 (comment).

from circle.

That sounds very good. Thanks for this valuable testing work!

from circle.

This problem should have been fixed.

from circle.