David Brownell <[email protected]> writes: > On Monday 11 May 2009, David Brownell wrote: >> On Monday 11 May 2009, Kevin Hilman wrote: >> > /* temporarily mask (level sensitive) parent IRQ */ >> > + desc->chip->mask(irq); >> > desc->chip->ack(irq); >> > while (1) { >> >> That looks like a clear bugfix to me ... even makes the code >> match the comment! Do any of the other parts of this patch >> really matter? I'd send that part upstream, by itself. > > Testing results with *ONLY* that change: nuttcp ran for > many hours, 100 million IRQs, no dm9000 timeouts. (*) > > Same test without adding that line: timeout within the > first hundred packets.
Nice! thanks for the extra testing. > I wouldn't change the irq_chip methods though; as I mentioned > elsewhere, that seems incorrect. OK. Note that I changed the irq_chip methods and changed from a 'simple' to an 'edge' handler. I was going on the assumption that there were some nasty nesting problems happening and that maybe the genirq handler would help with those. Adding that mask alone will prevent any of these nesting problems as well so I agree that the other changes are unncesary. I'll push a fix with just the mask. Kevin _______________________________________________ Davinci-linux-open-source mailing list [email protected] http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
