Return-path: Received: from senator.holtmann.net ([87.106.208.187]:59566 "EHLO mail.holtmann.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752218AbYKATLr (ORCPT ); Sat, 1 Nov 2008 15:11:47 -0400 Cc: linux-wireless Message-Id: (sfid-20081101_201154_385904_67344998) From: Marcel Holtmann To: Tomas Winkler In-Reply-To: <1ba2fa240810311635t69165473hef0ef64bf4426d3a@mail.gmail.com> Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes Mime-Version: 1.0 (Apple Message framework v929.2) Subject: Re: Still 4965 issues with wireless-testing Date: Sat, 1 Nov 2008 20:11:42 +0100 References: <5A6314C4-958B-43E7-B94D-FE5BABED89EA@linux.intel.com> <1ba2fa240810310652t351b0b77k819a7e03659503eb@mail.gmail.com> <1ba2fa240810311409i5203bcccu2b944dcc505deb30@mail.gmail.com> <1ba2fa240810311635t69165473hef0ef64bf4426d3a@mail.gmail.com> Sender: linux-wireless-owner@vger.kernel.org List-ID: Hi Tomas, >>>>>>> so besides the point that the Ubuntu 8.10 kernel has a broken >>>>>>> driver >>>>>>> for >>>>>>> the Intel 4965 cards which will crash my X61 after less than one >>>>>>> minute >>>>>>> uptime when it connects to my access point, I have my dmesg >>>>>>> full with >>>>>>> warnings from the latest wireless-testing kernel. >>>>>>> >>>>>>> I know that we have discussed this before and Johannes' patch >>>>>>> at least >>>>>>> turns it into warnings instead of killing my system >>>>>>> completely, but I >>>>>>> am >>>>>>> still surprised why this hasn't been fixed. >>>>>> >>>>>> so this becomes more and more annoying. I get this WARN_ON now >>>>>> multiple >>>>>> times pre minute. Only the first few minutes after loading the >>>>>> firmware >>>>>> everything is okay and then it keeps filling my dmesg. And on >>>>>> top of >>>>>> that, I >>>>>> now see these: >>>>>> >>>>>> __ratelimit: 1189 callbacks suppressed >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> __ratelimit: 57695 callbacks suppressed >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> __ratelimit: 3345 callbacks suppressed >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> iwlagn: Can not allocate SKB buffers >>>>>> >>>>>> What is going on here? This is a normal setup and nothing >>>>>> fancy. Access >>>>>> point is N-Capable from D-Link btw. >>>>> >>>>> >>>>> I have D-Link too I don't see this. What kernel version is this? >>>>> Is >>>>> this your ubuntu version or wireless-testing? >>>>> Any more data you can provide (.config. slabinfo) >>>> >>>> this is from wireless-testing. The Ubuntu 2.6.27 kernel is totally >>>> broken. >>>> It will crash the kernel since we are missing the patch to turn >>>> this into >>>> a >>>> warning :) >>>> >>>> Attached is my config. The slabinfo will follow once the SKB >>>> errors come >>>> up >>>> again. It normally takes a few hours before they show. Just >>>> resumed that >>>> machine. >>> >>> I've downloaded whole moblin last night over 2.6.27.4, it's a bit >>> slow >>> because the missing calibrations but it didn't crashed. >>> >>> I will test your config, your machine is T61 am I right? >> >> I have a X61 and running the 64-bit version of kernel and userspace. > > I found in sky2 driver comment > > * The buffer returned from netdev_alloc_skb is > * aligned except if slab debugging is enabled > > Since you have slub debug enabled, can you run this debug patch on > your system (might be withe space broken, I've just pasted in) I applied your patch and ran the latest wireless-testing for over 24 hours now. The iwl_tx_cmd_complete issues shows up on a regular basis and the allocate SKB buffers, too. However it didn't trigger the extra WARN_ON you added. Regards Marcel