/srv/irclogs.ubuntu.com/2010/09/20/#ubuntu-kernel.txt

Kanohi, the .35 ubuntu kernel has got patches from .36 it seems so that fglrx does not compile00:21
melkorfglrx compiles with the .35 kernel?01:16
Kanowith mainline it does01:18
melkorI forgot, the reason I can't use the new fglrx is because my card isn't supported any more hd3650.01:29
Kanoit is supported01:29
Kanojust 10-9 is not for xserver 1.9, but you could try lucid with 2.6.35 mainline and it will work01:30
Kanojust not with the maverick kernel01:30
melkorLast I checked (months ago) it was supported with the 9.xx drivers, but not the 10.x.01:30
Kanothats incorrect01:30
Kanohd2+ is supported01:30
melkorBut I'd need to install the mainline kernel and not the backports maverick?01:31
Kanomainline works01:31
Kano100%01:31
melkorIs there a repo with mainline?01:32
Kanonot directly a repository01:33
Kanohttp://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.35.4-maverick/01:33
melkorKano: isn't that a maverick kernel?01:36
Kanomaybe it was compiled with maverick01:37
Kanobut the libc6 is compatible with lucid01:37
Kanoso headers work01:37
Kanoits even compatible with squeeze ;)01:38
bjf[afk]melkor: the -maverick indicates it was build with the maverick configuration settings01:40
bjf[afk]s/build/built/01:41
melkorSo after I install this I should be able to reboot to that kernel and install fglrx.  Sometimes I have trouble with fglrx trying to install on the wrong kernel.01:43
Kanomy script is able to install it01:45
Kanorm -f /etc/X11/xorg.conf*01:46
Kanohttp://kanotix.com/files/install-fglrx-debian.sh01:46
Kanouse my script with -z optoin and reboot then01:46
Kanowell use a supported kernel... dont forget to install both header debs01:46
melkorThat is one hell of a script.01:50
melkorHow do I check my x server version?01:51
KanoX -version01:51
melkor1.8.201:52
Kanono problem then01:52
Kano.35 kernel is supported, just not the u patched variant01:53
melkorwhat about using the x-swat fglrx package?01:53
Kanowill not be much different01:53
melkorI'm gonna give that a shot it seems easier to undo.01:54
Kanoi update the script in the minutes after a new fglrx when i know about it. it is even prepared for maverick01:54
Kanoas soon you can dl the file01:54
Kanobut not with maverick kernel ;)01:54
melkoralright I'm gonna give it a shot.01:55
smbRAOF, Are you still around?08:34
erez_helo08:59
erez_hello08:59
erez_I have a kernel oops. i want to report it, but I can't find where to attach the dmesg09:01
erez_who should i report it o ?09:01
smberez_, Report it against the linux package (ubuntu-bug linux). If you can write the dmesg to a file to add it to the generated bug report.09:03
erez_it doesn't let me add the dmesg log09:04
smberez_, Isn't there a little item to the bottom that says attach file?09:05
smbActually it says "Add attachment or patch" in the bug report09:05
erez_not that i saw. i'll try again...09:06
smbIf you open the bug report in the browser its very much to the bottom of the page. Above "What next"09:06
erez_no, it doesn't09:06
smbHm, are you logged into launchpad?09:07
erez_no, i used ubuntu-bug linux command09:07
smbThat should have opened a bug report or at least given you the bug number 09:07
erez_i didn't proceed when it said "send report" as the report wasn't finished09:09
erez_it also asks strange questions, i have to choose if this is a regression or not, but there is no "i don't know" option09:09
smberez_, Ah, ok. I cannot say from my memory how this exactly looks like. But it should not matter. Just finish it off and then the report can be completed 09:10
erez_ok, trying right now09:10
smberez_, Did it work before (with and older release or the same release)?09:10
erez_is an OOPS is a "Kernel Config" or "Other" or "I Don't know"09:10
erez_I know what a "regression is", i just didn't test this with a previous ubuntu 09:11
smbWas that still the question about regression or another one (the one with kernel config ....) Ok, in that case I would rather say no to regression09:12
smbIt also can be changed later, when more is known09:12
smbOtherwise an oops is rather unlikely a kernel config. So I'd say either other or don't know 09:13
smberez_, One note, if you are using the system which had the oops for the generation of the bug report, then ubuntu-bug adds the dmesg output automatically09:15
erez_ok, thanks09:17
erez_after i finished the bug report, it opened lunchpad on my browser. i can login, however - it doesn't seem to associate that to the ubuntu-bug i just used 09:20
smbHm, it has been a little while when I used it. I thought it would redirect you to the bug report after login. Was there any printing of the generated bug number somewhere?09:22
erez_no09:23
erez_i will just have to fill it manually09:23
erez_ok, thanks smb, bug commited09:30
erez_bye09:30
tseliotapw, smb: there's no gpl code in fglrx hence I can't use that workaround that you suggested. Furthermore it seems that things broke in Lucid too10:56
tseliotas shown in make.log here: https://bugs.launchpad.net/ubuntu/+source/fglrx-installer/+bug/573748/comments/2810:57
ubot2Ubuntu bug 573748 in fglrx-installer (Ubuntu) "[MASTER] fglrx does not build on 2.6.33 kernel and higher (affects: 103) (dups: 33) (heat: 618)" [Medium,Triaged]10:57
tseliotapw, smb-afk: the problem is that even if I patch the driver to call arch_compat_alloc_user_space instead of compat_alloc_user_space when dealing with 2.6.32 and 2.6.35 I can't check the ABI and the patch may or may not break the module compilation11:00
* tseliot repeats things for smb-afk11:00
tseliotsmb: there's no gpl code in fglrx hence I can't use that workaround that you suggested. Furthermore it seems that things broke in Lucid too, as shown in make.log here: https://bugs.launchpad.net/ubuntu/+source/fglrx-installer/+bug/573748/comments/2811:01
ubot2Ubuntu bug 573748 in fglrx-installer (Ubuntu) "[MASTER] fglrx does not build on 2.6.33 kernel and higher (affects: 103) (dups: 33) (heat: 618)" [Medium,Triaged]11:01
=== smb-afk is now known as smb
=== smb is now known as smb-afk
=== ivoks-afk is now known as ivoks
smbtseliot, That is bad. In hindsight it might have been better to modify the security patches to export normally11:14
tseliotsmb: yep. What do you suggest that we do now?11:15
tseliotpatch the kernel and undo the breakage?11:16
smbtseliot, I'll have to talk to the security team. Problem atm likely is that no binary gfx driver works (or is this restricted to fglrx only11:18
smb?)11:18
smbThe question would be whether to revert it as a single upload or have it packaged into the next security release11:18
tseliotsmb: nvidia doesn't seem to use that function so hopefully it's just fglrx11:19
smbtseliot, Ok, at least something but I agree that it should be reverted. I just hope that changing the type of export does not cause an abi bump. But on the other hand it did not when the function was made gpl only.11:21
tseliotsmb: ok. Are you referring only to Lucid's kernel or also to Maverick's?11:22
smbThat would be only Lucid and before, though. Long term (and that would inlcude Maverick) the fglrx driver likely needs somehow to cope11:22
smbThis change is upstream that way and I don't think it will change there11:23
tseliotsmb: I suspected that. In Maverick I'll have no way to check the availability of that function at compilation time so my patch won't work on kernels with older ABIs11:24
tseliot:/11:24
smbtseliot, But to as back about the workaround: what was the problem that may break older things?11:25
tseliotsmb: I would call arch_compat_alloc_user_space instead of compat_alloc_user_space11:25
smbtseliot, When you posted that I was trying to get an irc proxy working, so I might have missed details11:26
tseliot^11:26
smbok, there was no place for the wrapper? calling arch would not be very good. 11:26
smbNeither in old code which does not have it nor in new where it is not doing any checking11:27
tseliotsmb: the wrapper that you suggested would work well only if there were GPL files in the fglrx driver11:27
tseliotAmd told me that there isn't any11:27
smbtseliot, Would it not be possible to add a file and make that gpl?11:27
smbSomehow I thought they had some issues with other functions that became gpl exported only before11:29
tseliotsmb: I'm not sure. I guess that would be a problem too11:30
tseliotsmb: the problem is that they moved compat_alloc_user_space from asm/compat.h to kernel/compat.c and did EXPORT_SYMBOL_GPL(compat_alloc_user_space)11:32
smbtseliot, Hm, I guess I need to think about that. I need to leave for lunch, too. Let me think and I'll get back to you.11:32
tseliotsmb: ok, thanks11:32
smbtseliot, remind me of the source package name for fglrx11:32
tseliotsmb: fglrx-installer11:33
smbok, ta11:33
=== amitk is now known as amitk-afk
=== amitk-afk is now known as amitk
tseliotsmb: this is the correct bug report: https://bugs.launchpad.net/ubuntu/+source/fglrx-installer/+bug/64251814:04
ubot2Ubuntu bug 642518 in fglrx-installer (Ubuntu Lucid) (and 1 other project) "[MASTER] package fglrx 2:8.723.1-0ubuntu4 failed to install/upgrade: Kernel fix for CVE-2010-3081 breaks fglrx (affects: 176) (dups: 165) (heat: 1408)" [High,Triaged]14:04
smbtseliot, Ok, I just sent out an email. I believe this could work.14:05
tseliotsmb: did you CC me?14:06
smbtseliot, no, I To'ed you14:06
tseliotah, even better14:06
smbIn essence you may be able to decide whether to use arch_compat_... or compat_... based on whether compat_alloc_user_space is an exported symbol in kallsyms14:07
smbtseliot, I have not tested against old kernels yet, but I believe the principle should work14:08
* tseliot reads the email14:10
smbDarn, most of my systems have ATI cards that are no longer supported by fglrx14:15
tseliotsmb: now, that's clever :)14:15
smbtseliot, Clever of ATI :-P They used to be previously but them removed support apparently14:16
tseliotsmb: yes, I was referring to your patch ;) I'm more than fine with the open driver even with cards that are supported by fglrx14:17
tseliot~$ grep "T compat_alloc_user_space" /proc/kallsyms14:17
tseliotffffffff810a5630 T compat_alloc_user_space14:17
tseliotthis is with 2.6.35-22-generic14:18
smbAh I thought you referred to the cards no longer being supported. :) I just checked on an older Lucid kernel, there is no such entry in kallsyms14:19
smbSeems I found a system that has a more recent card14:19
tseliotgood14:22
smb*argh* that is a 32bit installation...14:22
tseliothey at least we can use the workaround in Maverick14:23
tseliotI can try a Lucid livecd14:23
smbWe could potentially use the workaround on Lucid, too. Somehow I suspect we get the fglrx package updated faster than the kernel14:24
smb(not to forget that I know how much pain it will be to do from the amount of work)14:24
smbtseliot, Oh, so actually the breakage is limited to 64bit as well...14:26
smbgiven that this 32bit install is happy with the most recent kernel. OK, should have known by the fact that this is in compat_...14:27
tseliotsmb: yes, sorry, I should have pointed that out before14:27
smbtseliot, As I said I should have known from the code. :)14:28
tseliothehe ok14:30
smbtseliot, How much effort would it be for you to update fglrx for the older releases as well? I am winching about making a kernel update as that not only means new kernel packages for all the releases but also merging things around and creating new proposed uploads for Karmic and Lucid... again...14:36
tseliotsmb: if your workaround works as expected on Lucid, I think I can safely do that. Oh does it also affect Karmic?14:37
=== ivoks is now known as ivoks-brb
smbtseliot, Unfortunately that change went in back to Dapper, so it would affect all that have som fglrx support. (I think Hardy its envy)14:38
tseliotsmb: ouch14:39
smbYeah. :( 14:39
tseliotok, it looks like I'll have some additional work to do14:40
smbtseliot, Hm I am not even sure just changing EXPORT_SYMBOL_GPL to EXPORT_SYMBOL would be enough as the function declaration is in linux/compat.h and I think you mentioned that the fglrx driver only includes asm/compat.h. *urgh*14:41
tseliotsmb: yes, it can't include linux/compat.h14:42
* smb offers to by tseliot beers in Orlando (though I was told you might not enjoy that)14:43
tseliotsmb: hehe, right, I don't drink. Let's put it this way, you can offer me a coke and I'll buy you a beer for your patch ;)14:44
smbtseliot, Ah ok. NP with that. (though I was actually referring to the beer quality) :)14:45
tseliotsmb: I wouldn't have noticed anyway ;)14:45
smbI see that know. :)14:45
tseliothehe14:45
smbtseliot, Should I update the bug report with those facts we have discussed here?14:47
tseliotsmb: yes, sure that would be welcome14:47
smbok, will do14:47
tseliotsmb: your patch works well in Maverick. I'll test it in Lucid soon15:58
smbtseliot, OK, cool. :)15:59
=== ivoks-brb is now known as ivoks
aquariusafter upgrading to maverick, when I tell my machine (dell m1330) to restart, it jumps to the "ubuntu" shutdown screen with the 5 dots, and the 5 dots continue to tick over forever, but the machine never shuts down/restarts. It worked under lucid, so this is a regression somewhere. Which information do I need to provide to help you chaps fix it?17:17
aquariusis "ubuntu-bug -p linux" sufficient?17:19
jjohansenaquarius: its a start, along with a description of what is going on17:19
jjohansenaquarius: if you still have the Lucid kernel installed it is worth booting with it too17:20
aquariusjjohansen, ok, I'll do that17:20
jjohansenthat would help isolate whether its actually the kernel or maybe upstart/plymouth17:20
aquariuser. Warning: The options -p/-P are deprecated, please do not use them. :)17:20
aquariuswill try rebooting with the lucid  kernel next time I reboot before fiiling a bug17:22
ogasawarasmb: your misc blueprint work item to pull back the Lucid version of fsl-imx51 to karmic-fsl-imx51... didn't you mention we won't do that?  ie I can mark that one off the list?17:27
smbogasawara, Oh, I think I should make that announcement I wanted to but forgot with oall the other stuff going on17:28
ogasawarasmb: also, I thought that you, sconklin, and bjf[afk] have been documenting the stable team processes.  Anything left to document there?  or can I mark that work item as done also.17:29
=== bjf[afk] is now known as bjf
bjfogasawara: please mark it done though with all kernel wiki documentation its really never ending17:30
ogasawarabjf: ack17:30
smbogasawara, I have the feeling that there might still be something. But as bjf says17:30
=== amitk is now known as amitk-afk
sconklinogasawara: I think that we have enough documentation for Brad and I to consult, but not of a quality to be generally useful. We've done a lot. Maybe it's appropriate to close it and open a new task for next cycle.17:34
smbogra, Is there a specific mailing list for arm or is this ubuntu-mobile?17:39
=== ivoks is now known as away
=== away is now known as ivoks-away
=== ivoks-away is now known as ivoks
ftacould someone please have look at bug 638024? looks like a kernel panic (with a patch in the upstream bug)19:23
ubot2Launchpad bug 638024 in chromium-browser (Ubuntu) "Crashes ubuntu on application exit (affects: 3) (heat: 18)" [Undecided,New] https://launchpad.net/bugs/63802419:23
jjohansenfta: I'll take a look in a bit19:33
ogasawarafta: comment #32 - http://code.google.com/p/chromium/issues/detail?id=54617#c32 notes this doesn't seem to be an issue with the 2.6.35 kernel19:36
ogasawarafta: so I assume this shouldn't be an issue for Maverick19:37
ftathe OP seems to be using lucid19:39
ogasawarafta: indeed, so it seems they should confirm if the issue still exists in Maverick before any further work is done trying to backport anything to Lucid.19:42
=== ivoks is now known as ivoks-dinner
=== jsalisbury is now known as jsalisbury_brb
jjohansen-> lunch19:54
* ogasawara lunch20:20
=== jsalisbury_brb is now known as jsalisbury
=== amitk-afk is now known as amitk
=== ivoks-dinner is now known as ivoks
=== bjf is now known as bjf[afk]
Keybukmjg59: here's a random one for you, when you're up23:44
Keybukthe current MBPs have both an Intel GMA HD and an nVidia card in them, right?23:44
KeybukLinux only sees the NV23:44
=== ivoks is now known as ivoks-afk

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!