| praveen_ | looking for the defination of dlsym | 02:05 |
|---|---|---|
| praveen_ | anh pointers ? | 02:05 |
| praveen_ | probably its libc code | 02:07 |
| praveen_ | I think I can find it | 02:07 |
| aaron__ | hi i'm trying to follow this doc: https://help.ubuntu.com/community/Kernel/Compile | 02:27 |
| aaron__ | but I can't find the 'debian/config' directory it keeps referringto | 02:27 |
| aaron__ | oh, sorry this looks like a dev channel | 02:27 |
| jjohansen | aaron__: did you check out the ubuntu git tree? debian/config doesn't exist in mainline trees | 02:30 |
| aaron__ | jjohansen: no i'm just following that doc. does it mean that's a directory within the source package? | 02:31 |
| aaron__ | i'm d/ling that at the moment | 02:32 |
| aaron__ | (apg-get source linux-image-$(uname -r)) | 02:32 |
| jjohansen | aaron__: yes it is a directory in the kernel directory | 02:32 |
| aaron__ | ok, then i should be able to find my way around... where does that land, /usr/share somewhere? | 02:33 |
| jjohansen | aaron__: /usr/src/linux-2.6.XXXX | 02:35 |
| aaron__ | cool, thanks | 02:35 |
| jjohansen | aaron__: or that would be my guess, standard kernel location. I have never actually used the deb source package | 02:36 |
| aaron__ | hmm | 02:39 |
| aaron__ | huh, interesting, it puts it right in / | 02:45 |
| jjohansen | aaron__: interesting | 02:57 |
| === nightwis- is now known as nightwish | ||
| jjohansen | aaron__: just a warning about the scripts, they stick the .debs etc in the dir above the kernel dir | 02:57 |
| aaron__ | ok | 03:01 |
| aaron__ | if i want to customize a standard ubuntu flavor i think i have to take the standard config, copy to .config, run make xconfig, and then copy it back and then gothrough the standard ubuntu build | 03:01 |
| aaron__ | fakeroot debian/rules etc. | 03:02 |
| aaron__ | or maybe i can point xconfig at a certain config file | 03:02 |
| aaron__ | hmm doesn't look like make xconfig loaded the standard config optiosn | 03:09 |
| spO | do any of you know how to compile fglrx/ati drivers for a custom ubuntu kernel (after you uinstall karmic package)? | 03:33 |
| aaron__ | omg i forgot how many freakin options there were | 03:45 |
| Yingying | anyone know the kernel version of Ubuntu 8.04.3 LTS? | 03:51 |
| maswan | ikepanhc: give me a prod if you want me to take a new netboot kernel for a spin for testing things out | 08:36 |
| hyperair | aren't all kernels nebootable? | 09:10 |
| hyperair | net* | 09:10 |
| maswan | hyperair: Oh, context that was only clearly available in my head: Bug 360966, bnx2x missing or broken | 09:13 |
| ubot3 | Malone bug 360966 in linux "bnx2x missing in initrd for install media" [Medium,In progress] https://launchpad.net/bugs/360966 | 09:13 |
| hyperair | ah | 09:13 |
| maswan | sorry for the confusion | 09:13 |
| hyperair | np | 09:13 |
| apw | maswan, didn't see the context, but i though that was resolved at least for karmic? | 09:48 |
| maswan | apw: yes, except for bnx2x not loading due to missing crc32c dependancy | 09:50 |
| apw | i thought there was a dep on something crc related added | 09:50 |
| apw | +libcrc32c | 09:52 |
| apw | yeah there was a deliberate attempt to get that resolved in the patch | 09:52 |
| apw | i see ikepanhc is looking at it | 09:54 |
| apw | maswan, are you able to boot these blades from a usb stick? | 10:09 |
| maswan | apw: Not currently, no. That's something else I'm trying to debug, the blade virtual media applet just gives me a grey square, and no virtual media. | 10:26 |
| maswan | But I doubt that's the kernel at fault. :) | 10:27 |
| apw | how annoying | 10:27 |
| maswan | yup, and the bladecenter is 1200 km away so I have to arrange with someone else that then has to go onsite, etc, etc, if I want to try and plug something in. But is is a bit annoying to have all this new and shiny hardware without being able to use it right now. :) | 10:34 |
| apw | does any netboot work for any release? | 10:35 |
| apw | hardy for instance? | 10:35 |
| maswan | hardy doesn't have bnx2x support at all, jaunty almost has it, and karmic is closest so far with module existing but failing to load. | 10:36 |
| apw | i assume you are gettitng to a busy box? | 10:37 |
| maswan | apw: yeah, the installer starts even and I can chose language etc, then it doesn't get very far without any network connectivity [sorry for the slow reply, collegues pulled me away from lunch] | 11:19 |
| apw | maswan, so that implies you have access to the emergency holographic shell on VT-2 or wherever it is | 11:29 |
| apw | so you should be able to open that and try the modprobe yourself to see what errors it reports | 11:29 |
| apw | and to find out if the modules are in there | 11:29 |
| apw | and indeed if the firmware is actually there | 11:29 |
| maswan | the module fails to load due to missing symbols that look like they should come from the crc32-thingie somewhere | 11:30 |
| apw | and is that modules on there? | 11:30 |
| apw | libcrc32c that is | 11:30 |
| apw | maswan, also if its there does loading it directly work, and if so does modprobe bnx2x work? | 11:36 |
| maswan | Working on it, I did try most of this last night, but getting a console on another server now to try it "properly". | 11:38 |
| maswan | apw: unknown sumbol in module | 11:53 |
| maswan | might libcrc32c possibly depend on crc32c, which isn't there.. | 11:55 |
| maswan | I should probably add that to the bug. | 11:58 |
| apw | crc32 should be there already in another one | 11:59 |
| apw | does it not tell you the dep you need when you try and load it? | 11:59 |
| maswan | find . -name '*crc* in /lib/modules only finds libcrc32c.ko, crc-itu-t.ko and crc-ccitt.ko | 11:59 |
| maswan | Nope, just "Unknown symbol in module, or unknown parameter (see dmesg)", and it seems like dmesg is having issues, I haven't gotten any updates since timestamp 125.9. | 12:01 |
| maswan | no, that's not dmesg issue, just that the libcrc32c modprobe doesn't put anything there | 12:02 |
| apw | debian/d-i/modules/crypto-modules:crc32c | 12:10 |
| apw | debian/d-i/modules/nic-shared-modules:libcrc32c | 12:10 |
| apw | we are meant to be supplying crc32c at least in theory in crypto modules | 12:10 |
| maswan | onlything in kernel/crypto is arc4.ko | 12:11 |
| apw | bibble ... so sounds like installer failure ... | 12:11 |
| maswan | of course, this is before I have a chance to load any udebs | 12:11 |
| apw | could you report all that in the bug. that the crc32c seems to be missing too, and not in crypto | 12:12 |
| maswan | ack | 12:12 |
| apw | rtg those packaging changes seem to be holding up so far. the build i have in progress seems to at least have produced sane packages for lpia | 14:02 |
| rtg | apw, thanks for the note. now I have to write up how complex I'm making the stable maintenance job. Perhaps you could review the rebase scripts in each of the topic branches. I forgot to look at them. | 14:03 |
| maswan | And with a rebuilt initrd with copies of the modules in crypto-modules-2.6.31-5-generic-di_2.6.31-5.24_amd64.udeb the network interfaces are indeed discovered (after a manual modprobe crc32c) | 14:04 |
| apw | maswan, you had to manually probe the crc32c? without the probe of the network would not work? | 14:05 |
| rtg | maswan, are you doing a network boot? | 14:05 |
| maswan | apw: I'll retry it once more, it might just have needed to wait very long | 14:07 |
| maswan | rtg: yes, with bnx2x nic | 14:07 |
| apw | i would not expect a pause, but if you can confirm the load does not occur unless you manually load the crc32c then it implies there is a missing dep from libcrc32c to it, and that would likely explain why its missing | 14:08 |
| rtg | maswan, indeed, you are probably correct. | 14:08 |
| rtg | maswan, can you start a bug report on that? Please make sure to list the module dependencies in the report. | 14:09 |
| * apw notes that empathy sucks... | 14:10 | |
| maswan | rtg: it's currently a followup in bug 360966, I'll add some more info there (or open a new bug if you'd prefer) once I verify that it won't load unless I manually load crc32c | 14:11 |
| ubot3 | Malone bug 360966 in linux "bnx2x missing in initrd for install media" [Medium,In progress] https://launchpad.net/bugs/360966 | 14:11 |
| maswan | Anyway, need to spend some "quality time" with hardware in a local machine room, so I'll let it try load bnx2x without manual interference for a while | 14:12 |
| rtg | maswan, its odd. debian/d-i/modules/nic-shared-modules:libcrc32c which implies that it should be getting loaded, though the network boot stuff may be different wrt initramfs. | 14:12 |
| maswan | rtg: AIUI, libcrc32c is a small wrapper aroudn crc32c as of 2.6.29 | 14:12 |
| maswan | rtg: and the missing part being crc32c which makes libcrc32c/bnx2x not load | 14:13 |
| maswan | But I'm no expert in these matters | 14:13 |
| apw | rtg libcrc32c is being loaded, but crc32c is not... even though it is listed in our .udeb lists | 14:13 |
| rtg | maswan, right. I think, as you've discovered, that the crc32 stuff needs to be in the initramfs when loaded by PXE. Right? | 14:14 |
| maswan | rtg: Yes, I've gotten that far. Now there is a second potential issue, and that is bnx2x not pulling crc32c in automatically but needing a manual modprobe as well. | 14:14 |
| maswan | I have a custom initrd with the modules from crypto-modules that I'm working off right now. Of course, that might lack some dependancy info, I guess. | 14:15 |
| maswan | anyway, I'll let it stand at "detecting network hardware" for a while and see how it goes if I'm patient (aka doing other stuff) | 14:16 |
| rtg | maswan, can you add the output of 'lsmod|grep bnx' to the report? That should fully describe the dependencies. | 14:17 |
| ikepanhc | maswan: hi, you mean we still need crc32c.ko for bnx2x? | 14:24 |
| maswan | ikepanhc: Yes, that's what it looks like from here. After I added that module to the initrd bnx2x works. | 14:40 |
| maswan | init takes a really long time, but it did eventually complete | 14:41 |
| ikepanhc | maswan: you mean after you add the crc32c.ko into initrd, it takes a long time, but it works? | 14:42 |
| maswan | ikepanhc: yes | 14:42 |
| ikepanhc | I remember I have tested, the libcrc32c.ko works fine without crc32c.ko. Its seems I need to double check it | 14:44 |
| maswan | ikepanhc: while googling around, I found this: http://lkml.indiana.edu/hypermail/linux/kernel/0811.0/02433.html | 14:45 |
| maswan | and it seems like that went in as of 2.6.29.something | 14:45 |
| maswan | That's funny, lsmod doesn't claim crc32c to be used by any other module | 14:51 |
| ikepanhc | maswan: I saw that too, so I trust what modules.dep tells me | 14:53 |
| ikepanhc | maswan: but looking into the code, looks like libcrc32c.ko shall depends on crc32c.ko | 14:54 |
| maswan | but before crc32c is loaded, bnx2x timesout on init of libcrc32c with some other message about missing symbol crc32c (sorry, it has scrolled off now) | 14:54 |
| rtg | ikepanhc, the driver could be doing a dynamic module load instead of a direct symbol reference. | 14:57 |
| ikepanhc | rtg: you mean he just register and leave? | 14:58 |
| rtg | ikepanhc, no, probably something like load_module() | 14:58 |
| * ikepanhc digs in | 14:59 | |
| rtg | ikepanhc, its in lib/libcrc32c.c. it loads the crc32c module dynamically. | 15:04 |
| rtg | it eventually boils down to crypto_alloc_tfm() | 15:05 |
| ikepanhc | rtg: ya, looks like my fault, I shall make one initrd.img for testing, not only test with insmod | 15:08 |
| * ikepanhc changes the bug status | 15:09 | |
| === jMyles_ is now known as jMyles | ||
| maswan | ikepanhc: please let me know if there is anything you want me to test | 15:25 |
| ikepanhc | maswan: thanks a lot. | 15:26 |
| maswan | ikepanhc: I'm trying it again now (fiddled around with the external networking so I can get further), and it isn't that bad a slowness, but still, 2-4 minutes of waiting for "detecting network hardware" might be a bit on the long side | 16:40 |
| === zul_ is now known as zul | ||
| maswan | then it's 50 second timeouts to see if the nics have link, x8 for all the interfaces | 16:42 |
| maswan | hm. which gets repeated several times over all interfaces, which is why it takes so long | 16:46 |
| maswan | but that might be an installer issue or driver issue, rather than kernel/initrd inclusion issue | 16:46 |
| spO | is this about custom kernels? | 20:36 |
| hyperair | spO: is what about custom kernels? | 20:50 |
| === JanC_ is now known as JanC | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!