[etherlab-users] Error events during long-term testing

[Jeroen Van den Keybus]

> I can’t increase the cycle-rate significantly due to application load (is now 80%).

That's fairly high. Sure you didn't starve the CPU at some point ?
Some parts of a network driver may require deferred cleanup work. If
that work starts getting postponed too much, you may run out of
buffers sooner or later.

You should be able to load your system with a significant amount of
additional work with lower priority (cpu 0% idle and lots of I/O).

> Network card is Freescale P2020 On-Chip MAC. Drivers is generic driver using the standard Linux gianfar driver.

I'm not familiar with that driver. I would suggest that you instrument
it to check the state of the buffers etc. when the problem occurs.


BTW please include the mailing list when responding.


J.