[00:06] <Mead> hello
[00:22] <Mead> I just got a cheap netbook with a ARM processor and am interested in replacing windows CE with linux.  Anyone have some advice or resources that could be helpful?
[01:58] <rcn-ee> cwillu_at_work, weird.. is this with 2.6.35-rc5-dl6?  there is a couple patches micrel has tried pushing but unsucessful..
[01:59] <cwillu> rcn-ee, this is 2.6.33 up to 2.6.34;  2.6.35 may or may not have the same issue, can't tell due to a null pointer dereference I get on any significant network activity
[01:59] <cwillu> I've tried the most recent -35rc, still unable to use it
[02:00] <cwillu> however, knowing where to look now, I can determine if a kernel has the problem in a couple minutes, rather than the two day test I needed before
[02:00] <cwillu> specifically, increasing size-2048 allocations
[02:01] <cwillu> rcn-ee, how easy would it be to roll a 2.6.35 and/or a 2.6.34 minus the ks8851 patches?
[02:02] <cwillu> rcn-ee, also, did you see my comments in #beagle?  I put more information there
[02:02] <cwillu> I have to run again right away, but I'll be back in about 2 hours
[02:03] <cwillu> rcn-ee, also, I can provide you remote access to the board if that would be useful
[02:04] <rcn-ee> weird...  2.6.35-rc6-dl5 will be without the ks8851 patches..
[02:05] <rcn-ee> my zippy2 is at work, so i'll test it monday...  do you have a test pattern i can confirm and then send to my contacts at micrel?
[02:06] <cwillu> rcn-ee, "slabtop -s s", watch the size-2048 entry
[02:06] <cwillu> network interface should be up and connected to a network, haven't tested if it depends on actual activity yet though
[02:06] <cwillu> (i'll check that tonight)
[02:07] <rcn-ee> that's still bad if leaks with no connection..
[02:07] <cwillu> I know, but it might narrow things down if it requires activity
[02:07] <cwillu> you should see another allocation every second or so
[02:08] <cwillu> the problems become really noticeable when there's ~20,000 allocations in that slab
[02:08] <cwillu> i.e., oom's even though there's lots of memory available, etc
[02:08] <cwillu> (because the memory gets fragmented, hence my asking about the defrag patches before :p)
[02:08] <cwillu> okay, gotta go, I'll ping you when I get back
[02:09] <rcn-ee> okay later.. i'll keep an eye out..
=== kmargar is now known as markos_
[08:29] <cwillu_at_work> rcn-ee, back
[08:31] <JaMa> cwillu_at_work, rcn-ee: can you please point me to patches used to fix "unknown rellocations" in 2.6.35-rc*?
[08:32] <cwillu_at_work> JaMa, you might have missed him by about 5\ hours
[08:32] <JaMa> I can find out, if you point me to repo where it's fixed..
[08:33] <cwillu_at_work> later 2.6.35 builds from http://rcn-ee.net/deb/lucid/
[08:33] <cwillu_at_work> it'll either be in the patch file (among other things), or it's already in mainline git
[08:34] <JaMa> thanks
[08:40] <JaMa> ok found it
=== lag is now known as Guest77371
[12:06] <cwillu_at_work> rcn-ee, building from your 2.6.35-devel with the micrel patches disabled still results in the null pointer
[12:06] <cwillu_at_work> going to try again against 2.6.34
[13:28] <cwillu_at_work> rcn-ee, looks like that null pointer thing is purely a 2.6.35 thing, of which you may or may not be aware
[13:28] <cwillu_at_work> checking for the memory leak under 2.6.34 right now, with micrel patches disabled
[13:29] <cwillu_at_work> rcn-ee, ... and the verdict is nack
[13:29] <cwillu_at_work> I'll run it over night (need to go to bed anyway :p)
[13:29] <cwillu_at_work> slabtop -s s is showing a consistent increase in allocations to the size-2048 pool, and the usual rate
[13:32] <cwillu_at_work> rcn-ee, send me your id_rsa.pub, I'll set you up so you can log into the beagle images I'm making if you want
[13:32] <cwillu_at_work> cwillu@cwillu.com
[13:33] <cwillu_at_work> hmm, actually, there's a flicker of hope
[13:33] <cwillu_at_work> I guess I should give the box a couple minutes to calm down
[13:33] <cwillu_at_work> it _seems_ to be holding at around 230-250
[13:34] <cwillu_at_work> so I'll let you know in a few hours (i.e., when I wake up and check :p) if it's increased from that
[13:34] <cwillu_at_work> I was seeing values in the 20,000's,
[13:35] <cwillu_at_work> so if it stays under a thousand for a day or so, I'll be satisfied that disabling the micrel patchset eliminates the behaviour
[13:35] <cwillu_at_work> might even be able to convince me to bisect it :p
[13:40] <cwillu_at_work> ...
[13:41] <cwillu_at_work> (I'll still check tomorrow, but it appears to still be climbing, maybe a little less consistently than before, or I might not have noticed the bumping around before)
[13:41] <cwillu_at_work> it's up at 450 now
=== JaMa is now known as JaMa|Gym
=== JaMa|Gym is now known as JaMa
[22:12] <cwillu> rcn-ee, koen on #beagle mentioned that there was some patches on the linux-omap mailing list for a micrel memory leak
=== JaMa is now known as JaMa|Zz