/srv/irclogs.ubuntu.com/2010/12/21/#ubuntu-kernel.txt

=== jjohansen is now known as jj-afk
dstoHello, would you mind to help me moving this bug to the correct package? meta-package "wireless backports generic" depends on a kernel level matching package which is currently missing in the repo. Bug 445931, thanks, dsto07:37
ubot2Launchpad bug 445931 in linux (Ubuntu) "linux-backports-modules-karmic not installable, dependency problem (affects: 3) (heat: 10)" [Undecided,New] https://launchpad.net/bugs/44593107:37
apwdsto, moved09:14
dstoapw: thx09:14
apwdsto this bug is a karmic bug, yet your symtoms are maverick09:17
apwdsto, are you sure this bug still exists?  the archive is showing a consistent set versions wise09:18
apwdsto, though the versions are -24 not -23 in the archive right now09:18
dstoapw: I've seen it this morning, will try on another machine. sec09:19
apw     linux | 2.6.35-24.42 | maverick-updates | source09:19
apwlinux-backports-modules-2.6.35 | 2.6.35-24.15 | maverick-updates | source09:19
apwboth are 35-24, so if you have -23 your kernel would be out of date09:20
apwi am supprised that there wouldn't be a -23 of both currently though as they should be held for --security09:20
apwlinux-backports-modules-2.6.35 | 2.6.35-23.13 | maverick-security | source09:22
apwso that should be in the pool and available.  hrm09:22
apwdsto, it is possible the packages were promoted t09:26
apwthis morning to -updates and the publisher published half of the updates, if the rare cases that occurs it normally only persists for an hour09:26
apwso it may be resolved now09:27
dstoapw: I confirm, it's gone at 2.6.35-24 level. 09:31
apwdsto, great, thanks09:32
apwi'll clean up that bug09:32
dstofine, thank you 09:32
apwdsto, yeah its an issue with the way the publisher works, if a package is promoted at exactly the wrong moment an inconsistent package set can be published for an hour, which is confusing09:33
dstothe good part: I did not open a new bug and we were able to clean an old one ;-)09:34
apwdsto, indeed :)09:34
dstopleasure having worked with you, take care, have a nice day. 09:35
apwyou too09:35
apwcjwatson, ok confirmed ... adding that memset turns grub into a hard reboot on both warm and _cold_ reboots09:46
apwcjwatson, so i assume that that buffer is the issue09:46
* apw goes restest with the memset in but the buffer at 0x09...09:47
cjwatsonapw: so I think that may well just mean that we got past one bug only to hit another10:22
cjwatsonapw: and probably means that my valgrind test wasn't extensive enough10:22
cjwatsonthough odd for it to break cold reboots10:22
apwcjwatson, i suspect that for it to break cold it is not something valgrind can see, i suspect there is something we care about in the area10:23
apwcjwatson, normally we only damage a small section of it, the memset is splatting a far larger chunk and the symtoms are catestrophic, breaking grub itself10:24
cjwatsonrebooting at run-time usually means a unhandled trap - something that would ordinarily a segfault would be sufficient10:24
cjwatsonanyway, let me beat on it for a bit relative to the upstream code and I'll see what I can come up with10:24
cjwatsonif it's breaking cold reboots now then that should be accessible with kvm10:24
apwcjwatson, sounds good, i have other issues to poke sadly10:25
=== lifeless_ is now known as lifeless
KeybukI'm noticing a pattern here10:43
* apw is all ears10:43
Keybukbrcm80211 works on days with a 'u' in them10:43
Keybukwl works on the other days10:44
apwnice, a solution for every day :)10:44
Keybukexcept for the weekend, brcm80211 works all weekend10:44
Keybuk(I actually expect it's something to do with the WiFi router)10:44
Keybukwl's failure mode on some is never to lock to a channel, despite being associated, and bounce between 2.4Ghz g, n and 5Ghz a & n10:45
Keybukbrcm80211 works on those, but it's failure mode on the routers that wl seems to work fine on is ... to kernel panic10:45
apwKeybuk, the joy of having two drivers10:46
apwat least they are concentrating resource on the open one10:46
Keybukyeah10:47
Keybukthe best plan at the moment seems to be to have brcm80211 blacklisted and use wl10:47
Keybukbut then manually switch if wl isn't working10:47
Keybukbecause at least then you switch when the wireless is poo10:47
Keybukrather than reboot after a kernel panic and switch in a hurry10:48
apwKeybuk, did you need to do anything special to get brcm80211 working in general, firmware etc ?10:52
KeybukI just grabbed the source from git10:52
Keybukfiddled with the Makefile to make it build against the maverick kernel10:52
Keybukand grabbed the firmware out of git10:52
apwKeybuk, ahh on maverick10:52
apwok as the driver is built by the looks of it in natty10:52
Keybukyeah, I'm staying on maverick this cycle10:52
Keybukthe firmware for it was in the "linux-firmware" git repo10:53
apwcool10:53
Keybukactually checking my history10:53
KeybukI was git clone'ing that10:53
Keybukbut since I was on wl at the time, in the end I cheated10:53
Keybukand installed the brcm80211-firmware package from Debian10:53
apwheh ... got it to blammo in about 10s ... sweet11:00
apwKeybuk, heh enabling the brcm driver seems to stop my lappy booting11:09
Keybukheh11:10
Keybuklike I said, the failure mode of that driver is a kernel panic11:11
Keybuk(or hang)11:11
apwcirtainly not ready for prime time in natty 11:31
=== diwic is now known as diwic_afk
=== diwic_afk is now known as diwic
=== yofel_ is now known as yofel
czrhughhalf, thanks for the ftdi-regression kick btw12:46
=== jussi01 is now known as jussi
Keybukhmm, here's an interesting gmail test13:02
Keybukif I tag all of LKML with one label13:02
Keybukand I tag certain people with another13:02
macothey get two tags13:02
Keybukif those people post to LKML, does the entire thread they post in end up under the other label13:02
Keybukor just their posts?13:02
macoin imap or in gmail interface?13:03
Keybukgmail interface13:03
macoentire thread13:03
Keybukwin13:03
apwyeah its all thread based as far as i know, 'conversations' sorry13:19
=== jepler- is now known as jepler
=== doko__ is now known as doko
=== BenC__ is now known as BenC
cjwatsonapw: you're right, grub_memset doesn't act the way I thought15:20
cjwatsonapw: http://paste.ubuntu.com/546286/ is no longer in a reboot loop for me - can you see if it fixes your problem?15:20
cjwatson(more or less the same as before, just with 'for (i = 0; i < n; i++) chosen[i] = -1;'15:20
cjwatson)15:20
tgardnercjwatson, apw is out for a bit. he should be back in an hour or so.15:22
cjwatsonno rush15:22
apwcjwatson, will do16:13
cjwatsonta16:13
apwcjwatson, how funny we are back to the one combination i didn't test16:14
cjwatsonheh, and the one I asked you not to test earlier16:15
cjwatsonoh well16:15
apwcjwatson, shit happens every day, can't worry about such things that is for sure16:15
apwcjwatson, reality is if my machine compiled this thing faster i would have tested it, so i blame h/w16:18
=== jj-afk is now known as jjohansen
apwcjwatson, preliminary result is positive ... need to 100% confirm i have the right kernel bits but looking _good_16:45
cjwatsonwhee16:45
apwcjwatson, what _does_ grub_memset actually do as it doesn't set memory?16:45
cjwatsonI got an ack from upstream for the previous version as well, so I don't think I need to get a separate ack for this one16:45
cjwatsonit *does* set memory, it's a standalone memset16:45
cjwatsonobviously there's some subtle typecasting mistake I missed, but I'm not really interested :)16:46
apwso really the question is why does it blow its brains out with memset and not with the loop16:46
apwfair enough indeed16:46
cjwatsonthis code is clearer anyway, as you observed earlier16:46
apwyeah it would need to be well large to make any significant difference we might care about16:47
cjwatsonapw: shall I reassign bug 686705 to grub2 + me, then?16:49
ubot2Launchpad bug 686705 in linux (Ubuntu Natty) (and 1 other project) "System hangs at GRUB loading screen every warm boot since 2.6.37-3.11 seemingly due to nx-emu patch (affects: 1) (heat: 10)" [High,Confirmed] https://launchpad.net/bugs/68670516:49
apwcjwatson, ok stock kernel reboots just fine with that fixed grub16:50
apwso 'Tested-by: Andy Whitcroft <apw@canonical.com>'16:50
apwSarvatt, about ?16:50
Sarvattapw: yeah16:51
Sarvattapw: http://paste.ubuntu.com/546286/ ?16:51
cjwatsoncommitted upstream16:51
apwSarvatt, i have some binaries for that if you could test16:51
Sarvattsure thing! save me a year long compile :)16:52
apwcjwatson, yes all yours on the bug16:52
cjwatsonSarvatt: wuss16:52
* Sarvatt digs out that machine16:52
apwhttp://people.canonical.com/~apw/misc/grub/ ... -pc and -common for i386 in there ... work on my box :)16:52
apwSarvatt, yeah she isn't a quick build, about 15 mins here16:53
Sarvattoh nice, it apparently hangs after a few days powered off too :)16:53
cjwatsonyou can always edit debian/control before building and comment out all the binary stanzas except grub-common and grub-pc; that speeds it up a fair bit16:53
apwSarvatt, hrm!16:53
cjwatsonrandom memory contents I suppose16:53
cjwatsonnot a huge surprised16:54
cjwatson*surprise16:54
apwcjwatson, normally the bios clears it doesn't it?16:54
apwyou'd hope anyhow16:54
cjwatsonI wouldn't like to rely on that16:54
apwbut yeah it is odd it works reliably the first time if anything ... and perhaps it doesn't :)16:54
cjwatsonwouldn't it take ages anyway?16:54
apwcjwatson, well it does on large machines yes16:54
cjwatsonin bios terms16:54
apwthough they have to as its all ECCd so has to be cleared to be used16:55
apwso perhaps not any more with small machines16:55
* apw reboots that machine for the 4th time16:55
apwsucessfully!16:56
rsalvetiapw: seems we're still missing a meta update for omap 4, the new kernel is there already but it's not being pulled by default17:00
Sarvattapw: dingding we have a winner! :)17:02
Sarvatt3 reboots fine so far17:02
tgardnerrsalveti, which release?17:02
apwSarvatt, thank $deity for that17:02
apwcjwatson, ^^ i think we have it nailed17:03
rsalvetitgardner: latest is linux-image-2.6.35-1101-omap4, and meta is 2.6.35.1100.117:03
tgardnerrsalveti, then you mean natty17:03
apwtgardner, natty17:03
rsalvetitgardner: oh, sure17:03
cjwatsonapw: hooray17:03
cjwatsonmerging down the line through Debian experimental now17:03
apwcjwatson, we were lucky to find that one17:03
tgardnerrsalveti, ok, gimme a bit17:04
apwtgardner, thanks17:04
cjwatsonyeah17:04
apwcjwatson, i think i mostly noticied it cause they had also mucked up the size bit17:04
* apw happily closes that one on his todo list17:05
rsalvetitgardner: sure, thanks17:05
Sarvattapw, cjwatson: thanks a ton \o/17:05
cjwatsonreassigned to me then17:06
apwkees, ^^-- seems that the 'NX issue' is indeed a grub bug and now fixed17:06
=== jjohansen is now known as jj-afk
apwcjwatson, well i learned a lot about grub2 (that i'd rather not have known) and had practice with quilt patches in debian, so overall worthwhile17:06
tgardnerrsalveti, uploaded linux-meta-ti-omap4_2.6.35.1101.217:11
rsalvetitgardner: great, thanks a lot!17:12
* apw invokes marjo17:15
cjwatsonsigh, "attempt to move .org backwards" only happens with gcc-4.4 not gcc-4.5, that explains why I didn't see it when working upstream17:32
cjwatsonevidently gcc-4.5 generates slightly smaller code17:33
keesapw, cjwatson: \o/ so glad you guys found that. *whew* nice work. :)17:34
apwkees, i did think that i was going mad indeed17:35
apwcjwatson, that explains things17:35
apw-#define GRUB_KERNEL_I386_PC_RAW_SIZE0xc9017:35
apw+//#define GRUB_KERNEL_I386_PC_RAW_SIZE0xc9017:35
apw+#define GRUB_KERNEL_I386_PC_RAW_SIZE0xca817:35
apw 17:35
apw-#define GRUB_KERNEL_I386_PC_NO_REED_SOLOMON_PART 0x6f817:35
apw+//#define GRUB_KERNEL_I386_PC_NO_REED_SOLOMON_PART 0x6f817:35
apw+#define GRUB_KERNEL_I386_PC_NO_REED_SOLOMON_PART 0x71017:35
apwcjwatson, ^^ i used that on mine17:36
keesapw: I bet! I spent a few hours comparing the NX patches between maverick and natty trying to see any differences early on. :P17:36
cjwatsonthe growth is 20 bytes so I think I'll make it 0x70c17:36
cjwatsonjust to parallel that exactly17:36
apwcjwatson, sensible17:36
apwyou do need to move both i think17:36
cjwatsonyes, you're right17:37
cjwatsonhm, wonder if it needs to be eight-byte-aligned17:37
cjwatsondoesn't look like it, I think four-byte-aligned will do17:38
apwcjwatson, i'd be supprised if its not 4 or cache aligned ... ie 8 sounds unexpected17:38
cjwatsonthe code uses .p2align 2 in various places which is 4 bytes17:38
cjwatsonok, fixed that upstream too17:42
apwcjwatson, round and round :)18:49
* tgardner --> lunch19:17
tgardnerapw, I see 2.6.37-rc7 is out.20:36
apwtgardner, excellent will suck that up now, as i was about to do an upload20:37
tgardnerapw, looks like it wasn't more then an hour orso ago20:37
apwtgardner, can't have been indeed as i updated about 2 hours ago to see myself20:37
apwtgardner, perfect timing, this means i can get the cciss bug uploading20:38
apwand unblock cert20:38
tgardnerapw, I don't think its mirrored yet. I'm not getting the -rc7 tag20:39
apwtgardner, i get mine from master.k.o and it is there :)20:39
tgardnercool20:39
apw   b0c3844..90a8a73  master     -> linus/master20:39
apw * [new tag]         v2.6.37-rc7 -> v2.6.37-rc720:39
tgardnerapw, do you know anything about the ALSA c-o-t-d builds?21:18
apwtgardner, hmmm, no not really i think they are bjf's fun arn't they21:27
tgardnerapw, well, he's got the right perms on his zinc repo, so I was able to push an ABI bumper commit.21:28
tgardnerhopefully the automated scripts will pick it up21:28
apwtgardner, ahh i see ... they run on tang i think21:29
tgardnerapw, do you suppose he pulls from his zinc repo?21:30
apwtgardner, may i suggest you add the subject to the rally agenda, so he can document/transfer the knowledge21:30
tgardnerapw, what a good idea :)21:31
apwtgardner, ok not tang then as he only makes the isos there21:31
tgardnerapw, dang, linux-meta_2.6.35.24.29_source.changes rejected again. wtf?21:32
apwbubble21:32
hughhalfczr (belatedly - you're welcome)21:33
tgardnerapw, I think I've lost interest for the day. see you in my AM21:33
=== jk-- is now known as jk-

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