=== foka_ is now known as foka | ||
vlowther | anyone here willing to help me debug a kernel-related suspend/resume problem on 2.6.24-17 on Hardy? It appears to be scheduler-related, but debugging it beyond what I can see in the log is a bit tricky. https://bugs.launchpad.net/ubuntu/+source/linux/+bug/212660 has details -- the short version is that suspend/resume last worked properly on 2.6.24-12 for me. | 01:43 |
---|---|---|
vlowther | from what I can tell, something broke in the way the scheduler freezes processes between 2.6.24-12 and 2.6.24-14 | 01:50 |
vlowther | because (and this is in the bug report if you read the suspend-failed.log attachments), the first suspend on kernels later than 2.6.24-12 work, but the second and subsequent attempts fail and spew loads of scheduler debugging information into the logs. | 01:53 |
vlowther | my read of them is that it fails to freeze pm-suspend | 01:53 |
vlowther | which is rather silly, because pm-suspend is the shell script that at that instant told the kernel to suspend by echo'ing "mem" to /sys/power/state | 01:56 |
pwnguin | the traditional solution to regressions is a bisection | 01:58 |
pwnguin | you nab the kernel git tree and use git-bisect to binary search for the commit that broke it | 01:58 |
pwnguin | this requires a number of rebuilds and general "able to build a kernel"-itude | 01:59 |
pwnguin | im also not sure that's the right team to assign a bug to | 02:00 |
vlowther | hm. | 02:01 |
vlowther | who would be a better team? | 02:01 |
pwnguin | I don't do much triaging myself =/ | 02:01 |
pwnguin | i think the wiki has a guide on that | 02:02 |
pwnguin | apparently the kernel team is too good to use the process the rest of ubuntu uses ;) | 02:02 |
vlowther | lurvley. | 02:02 |
vlowther | well, as far as triaging goes, my process went something like this: | 02:03 |
vlowther | boot to $latest-kernel | 02:03 |
vlowther | try to suspend/resume more than once. | 02:03 |
vlowther | observe failure | 02:03 |
vlowther | boot back to 2.6.24-12 | 02:03 |
vlowther | suspend/resume goes back to working reliably. | 02:04 |
vlowther | wait the next apt-get dist-upgrade introduces a new kernel | 02:05 |
vlowther | lather, rinse, repeat. | 02:05 |
pwnguin | git-bisect will find the change that "broke" it. possibly in the straw that broke the camel's back sense, though | 02:06 |
pwnguin | unfortunately, that nvidia module is bad news | 02:06 |
vlowther | that would be unfortunate | 02:08 |
vlowther | seeing as how suspend does not work at all without it, and the module version has not changed at all since 2.6.24-12. | 02:09 |
vlowther | (nv driver is not smart enough to do The Right Thing when reinitializing the video card on this system) | 02:09 |
pwnguin | a) i have yet to figure out how to rebuild linux-restricted-modules in conjunction with a kernel b) if it turns out to be nvidia, we can't even report it, let alone fix it | 02:11 |
pwnguin | on a related note, is iwl in linux-restricted drivers? | 02:11 |
vlowther | nah, it appears to be in linux-ubuntu modules | 02:13 |
vlowther | (or whatever the name of that package is) | 02:13 |
vlowther | based on a quick once-over of the /lib/modules/2.6.24-12 dir | 02:13 |
pwnguin | well, perhaps i just screwed up the config of upstream .25 then. no matter, as I accomplished what i needed to accomplish to report a bug to upstream | 02:15 |
vlowther | iirc, iwl is native on .25 | 02:19 |
vlowther | anyways, off to clone the ubuntu git repo | 02:23 |
vlowther | see if vgrep can find anything suspicious before trying to git-bisect. | 02:24 |
vlowther | if mdomsch is paying attention, mabye he could also help a brother out. | 02:26 |
vlowther | :) | 02:26 |
mdomsch | vlowther, how can I help? | 02:26 |
vlowther | woah, didn;t expect a response that fast. | 02:27 |
pwnguin | git bisect suspend / resume bug? | 02:27 |
vlowther | (or at least pointers on how to read sched-debug logs more accurately) | 02:28 |
pwnguin | vlowther: perhaps you should repeat the question for him | 02:28 |
vlowther | lp bug 212660 for backstory | 02:28 |
vlowther | short version is that every hardy kernel since 2.6.24-12 suspens and resumes properly once, second and subsequent attempts spew scheduler info instead fo freezing processes | 02:29 |
vlowther | de sched-debug info appears to say that pm-suspend refused to freeze, which is rather odd | 02:30 |
vlowther | as that is just a bash script that kicked off the kernel suspend process with 'echo "mem" >/sys/power/state' | 02:31 |
mdomsch | laptop suspend/resume is a mystery even to me | 02:31 |
vlowther | no kidding | 02:31 |
mdomsch | if it's a Dell, poke rez kabir | 02:31 |
vlowther | it is. | 02:31 |
vlowther | he on irc/ | 02:31 |
vlowther | er, ? | 02:31 |
mdomsch | uh, no | 02:31 |
vlowther | ah | 02:31 |
mdomsch | he'll be in PS2 in the morning though | 02:32 |
vlowther | ok | 02:33 |
pwnguin | hmm | 02:33 |
vlowther | heck, if the git-bisect gods are smiling I might even have a patch by then. | 02:33 |
pwnguin | "the morning" is of course, a relative term ;) | 02:34 |
pwnguin | is ps2 an acpi term for awake? | 02:34 |
vlowther | (or is that the git-clone gods? Hard to tell with how fast this is going) | 02:34 |
pwnguin | if not, i have no idea what that meant =( | 02:34 |
vlowther | not in this instance. | 02:34 |
vlowther | seekrit code. | 02:34 |
vlowther | ;) | 02:34 |
pwnguin | ah, you're both in austin | 02:35 |
mdomsch | pwnguin, "Parmer South 2" is one of the Dell buildings | 02:36 |
mdomsch | was mine until 6 months ago | 02:36 |
pwnguin | "mine" | 02:36 |
mdomsch | 4 years 50 hours/week, yeah, "mine" | 02:37 |
mdomsch | it saw more of me than my wife | 02:37 |
mdomsch | vlowther, I added the Dell team to the bug in LP | 02:39 |
vlowther | ok | 02:39 |
mdomsch | rez should be familiar with the D820 | 02:39 |
mdomsch | he did a boatload of work on it about a year ago | 02:40 |
vlowther | kewl. | 02:40 |
vlowther | it has always worked nicely for me | 02:40 |
* vlowther makes a note for tomorrow. | 02:44 | |
vlowther | compile compile compile | 03:34 |
vlowther | damn -- 45 mins to compile the ubuntu kernel. | 04:17 |
vlowther | whiskey tango foxtrot! | 04:17 |
=== asac_ is now known as asac | ||
=== gnomefre1k is now known as gnomefreak | ||
=== BenC__ is now known as BenC | ||
BenC | Good morning everyone | 13:59 |
abogani | BenC: Good morning, Sir. | 14:06 |
rtg_ | BenC: are you working on the Gutsy security update for kees ? | 14:14 |
BenC | rtg: Yeah | 14:22 |
rtg_ | BenC: I was just noticing the state of the git tree is a little weird. the changelog was updated awhile ago by smb, but it never got uploaded. | 14:24 |
rtg_ | I also pushed an SRU late yesterday that I wanted to make sure you picked up. | 14:24 |
psufan | https://bugs.launchpad.net/ubuntu/+source/linux-source-2.6.15/+bug/58170 | 14:43 |
BenC | rtg: Ok | 14:50 |
=== gnomefre2k is now known as gnomefreak | ||
=== asac_ is now known as asac | ||
psufan | ] <psufan> ugh | 19:31 |
psufan | [14:31] <psufan> I think git did something to the kernel source | 19:31 |
psufan | [14:31] <psufan> it never took 4+ hrs to make a kerne; | 19:31 |
psufan | it's almost like it's looping a build | 19:33 |
psufan | < ivoks> psufan: cd /tmp ; wget http://launchpadlibrarian.net/8736376/linux-source-2.6.15_nfsv4client.patch ; sudo apt-get install git-core ; git clone git://kernel.ubuntu.com/ubuntu/ubuntu-dapper.git ubuntu-dapper ; cd ubuntu-dapper ; patch -p1 -i ../linux-source-2.6.15_nfsv4client.patch ; sudo apt-get build-dep linux-image-`uname -r` ; sudo apt-get install fakeroot build-essential ; fakeroot dpkg-buildpackage # at this point, you | 19:43 |
Mactaylor | why is reiser4progs provided but reiser4 is not in the kernel? | 20:51 |
=== asac_ is now known as asac | ||
=== asac_ is now known as asac |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!