I have found an interesting workaround that seems quite effective and requires no code changes for a Fedora user with my hardware... would be interested to see if anyone else can reproduce this or explain why it fixes the problem.
Recall I was trying to audit the Fedora kernel ASPM related patches - I tried but really don't have enough knowledge of PCIE and related stuff to completely make sense of it all. Also I have to treat all the patches and code comments with healthy skepticism because they seem to have an air of black art about them. As if they are guesses at a solution rather than something guranteed to conform to some specification. Anyway reading the code and understanding a little about the changes that had worked, I saw that there was another option for pcie_aspm. Tony had tried pcie_aspm=off previously but it seemed to me that pcie_aspm=force might be worth a try. I tried it with the 2.6.38.8-35.fc15.i686 kernel straight from Fedora (ie. without any patches or hacking from me) and it worked. So for me at least, no more hacking bleeding edge wireless packages or building specially patched kernels. Just add: pcie_aspm=force to the kernel command line. Comments? -Cam On Thu, Jul 14, 2011 at 8:02 AM, Camilo Mesias <[email protected]> wrote: > OK, I took my current Fedora Kernel and applied a patch to remove the > commit mentioned above, basically removing aspm_clear_state ... I > found that it DIDN'T cure the problem. I am wondering if any of the > other Fedora patches (some are ASPM related) might not help. I did > have to go beyond the changes in the commit as there was one further > mention of aspm_clear_state. > > The other patches are: > linux-2.6-defaults-aspm.patch > pci-enable-aspm-state-clearing-regardless-of-policy.patch > pci-pcie-links-may-not-get-configured-for-aspm-under-powersave-mode.patch > > The kernel I used was 2.6.38.8-35.fc15.i686 > > If I can tonight, I will read those patches and see which might be > involved in this mess, but for now I have to say I can't confirm any > fix based on simply backing out that commit. > > -Cam > > On Tue, Jul 12, 2011 at 2:46 PM, Tony Houghton <[email protected]> wrote: >> On Mon, 27 Jun 2011 18:37:00 +0800 >> Adrian Chadd <[email protected]> wrote: >> >>> Camilo, Tony, etc: >>> >>> Please read this article and try what they're suggesting. It has to do >>> with an APSM commit which has negatively impacted some users (eg by >>> causing hangs.) Sound familiar? :) >>> >>> http://www.phoronix.com/scan.php?page=article&item=linux_2638_aspm&num=2 >>> >>> (Obviously with my ath9k patch backed out.) >> >> Sorry about the delay, I don't regularly use the offending netbook any >> more. But I've reverted that APSM commit and removed your patch, and it >> does seem to cure the problem! I'm quite surprised, because the >> timescales/releases of the bug and that commit don't seem to match. So >> as before, I hope Camilo and/or others can confirm this. >> >> It also seems to fix a similar issue I had experienced with rt2800pci. >> Is there some sort of bug report or discussion for that APSM commit? I >> should probably report my findings there. >> >> > _______________________________________________ ath9k-devel mailing list [email protected] https://lists.ath9k.org/mailman/listinfo/ath9k-devel
