>
>In 2.0.34, I have a lot of debugging enabled in my network card
>(pcmcia Xircom CEM33, it has frequent lockups under heavy traffic).
>It seems to die when a transmit happens during a recieve. However,
>this raises a problem with printk() - printk() seems to have insufficient
>locking of a shared buffer, or something of that order. The line underlined
>by XXXXXXXXXXXX is where my network card stops receiving (and no more
>interrupts are recieved). EnterIntr is the start of the interupt routing,
>EnableIntr is done at the end.
printk() is not SMP and interrupt safe. This mean that if the kernel is
running printk() when an interrupt happens and the interrupt handler will
run printk() too, when the interrupt handler will return the first printk
will be corrupted (but I don' t know if this could cause major problems).
>This problem with printk() means I don't know if the EnableIntr is
>actually happening and getting lost or isn't happening.
I think you should try to comment out all printk() in interrupt handlers
and see if the machine will continue to lockup.
Andrea[s] Arcangeli
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu