On 14 August 2011 02:09, Rajkumar Manoharan <rmanohar@xxxxxxxxxxxxxxxx> wrote: > On Sat, Aug 13, 2011 at 06:53:12PM +0800, Adrian Chadd wrote: >> .. again, I'll just point out that because I was seeing this when the >> PCU honestly had hit the end of the RX descriptor list (lots of radar >> phyerr frames), I needed to kick the PCU to start again. >> > so without PCU restart you are still getting rxeol. am i right? Yup. That's because I've actually hit the end of the RX descriptor list and since the 11n NIC code (so ath9k, and FreeBSD's ath when using an 11n NIC) doesn't enable the self-link descriptor trick, the PCU is now "stopped". I hit this when doing radar phyerr frame reception and (artificially) generated lots of radar phyerr frames whilst keeping the CPU busy. The RX thread would be CPU starved and it wouldn't be able to service the RX descriptor list in time. I've never hit it in normal operation, because (a) I wasn't doing RX phyerr reception, and (b) I wasn't _trying_ to break RX under load. :-) The next problem (which I hinted at in another email) is that I'm seeing an AR9280 NIC (Ubiquiti SR71-E) get itself into a situation where its unable to do any kind of DMA, and constantly fires off RXEOL interrupts, even if I then reset the PCU RX on RXEOL. The only fix is a cold reset - a warm reset doesn't trigger it. So, I wonder what is causing your underlying issues on Merlin, as it doesn't sound like you've hit the end of the RX descriptor list (or you'd have to restart the PCU w/ a new RX descriptor list to fix it.) Your problem may be tangentally related to what I'm seeing :-) Adrian -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
