**** BEGIN LOGGING AT Sat Oct 30 02:59:57 2010 Oct 30 06:15:51 hi Oct 30 06:15:57 who know upstart? Oct 30 06:16:07 com.ubuntu.Upstart.c Oct 30 06:16:17 job_process.c Oct 30 06:23:35 exit Oct 30 11:50:07 ho Oct 30 21:22:38 rcn-ee, oooooo Oct 30 21:23:21 rcn-ee, I got a nice crashing bug with ks8851 Oct 30 21:23:42 hey cwillu, catching up on the irc log, so some issues with ^ ;) just got back from texas.. so how nasty is the bug.. Oct 30 21:23:54 rcn-ee, completely locks up the board Oct 30 21:24:16 intermittent, had no idea what was causing it until I tripped over a solid repro case by mistake Oct 30 21:25:21 rcn-ee, remember my muttering about how the driver's 'cable connected' state didn't seem to make it to network-manager, even though mii-tool reported it fine? Oct 30 21:26:10 does it occur just idle, or in a tx state or rx state.. i talked to micrel before, and we had some work arounds for a tx silicon bug.. Oct 30 21:26:30 rcn-ee, tx for sure, not certain about the others Oct 30 21:26:54 I've been seeing hangs for 'no reason', without any network traffic to speak of Oct 30 21:27:36 but if I pull half a meg from the beagle over the network, while mii-tool is in its polling mode, and restart network manager during the transfer, the whole system will gradually lock up Oct 30 21:28:16 sysrq-w (?) ends up showing a couple dozen tasks blocked in uninterruptable sleep (many of which have no network activity involved in them) Oct 30 21:29:04 so, disabling my fake-mii job (which works by restarting network-manager whenever mii-tool detects a change) occasionally (but often several times in a row) locks up the system Oct 30 21:29:30 I know there's no cable state changes involved though, which confuses me: the crash doesn't occur unless network manager is restarted Oct 30 21:29:38 humm, that sounds like the same old TX bug.. I'll set up my rig again, something must have changed to bring it back.. which kernel version? Oct 30 21:29:44 .35 Oct 30 21:30:05 I've been running on .35rc5 for a while, but it's reproduceable on the latest .35 as well Oct 30 21:30:33 incidently, do you know if there's any fixes for mii state not getting reported properly? Oct 30 21:30:48 2.6.35.7-l6 is the one I was running Oct 30 21:34:53 checking out micrel's ftp site, the last i had from them was April 2010.. there might be something newer, otherise i'm not sure about the mii state.. Oct 30 21:35:40 now that I know what it is, I've set things up to restart the network when a particular connection dies instead of using mii-tool, but yeah, it'd be nice if cable-plug events worked :) Oct 30 21:37:31 rcn-ee, this definitely wasn't related to 'heavy traffic' (looking at "fix ks8851 snl transmit problem", if that's the patch you're referring to) Oct 30 21:38:56 rcn-ee, http://patchwork.ozlabs.org/patch/40247/ Oct 30 21:39:18 hang on, I can get you the oops that came along with the crashes Oct 30 21:41:42 yeap, that is the tx fix for the silicon bug, i have that in my patchset. (from micrel's patch dump) Oct 30 21:43:05 nothing newer on their ftp site.. Oct 30 21:47:04 rcn-ee, http://pastebin.com/srkcrPAM Oct 30 21:49:18 interesting, very weird the omap spi side crashed.. (so 50/50 ks8851/omap spi)... on the dumps i saw before it was the ks8851 side that crashed and it would lock up the whole board.. Oct 30 21:50:23 ? Oct 30 21:50:39 it does lock up the whole board :p Oct 30 21:50:42 kinda sorta Oct 30 21:51:18 sorry, I need to run, but I'll be around later, and I can do testing on this now that I know how to reproduce it Oct 30 21:51:33 this is what Tristan's patch was fixing : http://pastebin.com/rWbCsnhN Oct 31 00:23:19 aloha **** ENDING LOGGING AT Sun Oct 31 02:59:58 2010