Return-path: Received: from ey-out-2122.google.com ([74.125.78.27]:29812 "EHLO ey-out-2122.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751875AbYISCwS (ORCPT ); Thu, 18 Sep 2008 22:52:18 -0400 Received: by ey-out-2122.google.com with SMTP id 6so71388eyi.37 for ; Thu, 18 Sep 2008 19:52:16 -0700 (PDT) Message-ID: (sfid-20080919_045231_207195_CEBC0A8A) Date: Thu, 18 Sep 2008 19:52:16 -0700 From: "Steven Noonan" To: "Luis R. Rodriguez" Subject: Re: [ath9k-devel] ath9k: massive unexplained latency in 2.6.27 (rc5, rc6, probably others) Cc: "Ingo Molnar" , "ath9k-devel@lists.ath9k.org" , linux-wireless , LKML , "Luis Rodriguez" In-Reply-To: <43e72e890809181610h3a7729d8s4c8484d97b21932e@mail.gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 References: <43e72e890809181134ybbec8fdxcb2a466aa17fe390@mail.gmail.com> <43e72e890809181142n7738cd99g522e6688e68d11ce@mail.gmail.com> <43e72e890809181344q416b5944w3332ee5a33db048c@mail.gmail.com> <20080918220102.GE7408@tesla> <43e72e890809181508w5232a14ewbf2bf18fe90a92d5@mail.gmail.com> <43e72e890809181610h3a7729d8s4c8484d97b21932e@mail.gmail.com> Sender: linux-wireless-owner@vger.kernel.org List-ID: This IRQ debugging is lots of fun. My /var/log/messages jumped in size to 241MB. :) My test case was a file transfer that basically copies 50MB from a remote machine's /dev/urandom and writes it to /dev/null locally. That should be a pretty nice large RX. And jeez, I'm seeing tremendous download speeds now and I have no idea what is different about my configuration. I'm getting up to 24 megabit per second transfer rates (actual), which is incredibly unusual for ath9k. I wasn't able to reproduce the IRQ problems on -tip or Linus' current tree (both with the group rekeying fix applied, which may or may not be a good idea for reproducing the issue). Even tried doing rmmod/modprobe in fast succession multiple times. Couldn't get it to throw up. I wonder... My rmmod/modprobe originally was used after the improper group rekeying before in order to get it back to normal. Perhaps the encryption has to go wonky before this IRQ issue surfaces? Another possibility is the earlier mentioned "ACPI: EC: GPE storm detected, disabling EC GPE". I don't know if it's related or not, but I haven't waited for it to show up before testing. Oh well, if the IRQ problem resurfaces (or if I get a ridiculous latency tomorrow morning with irqpoll off), I'll try to get some useful data to help debug it.