srp | howdy. Is there any way to compare kernel config params (i.e. the CONFIG_ bits in config-<kernel-version>) between two different versions? | 00:59 |
---|---|---|
srp | other than writing a program to do so or comparing by hand, I meant | 01:00 |
gerald_ | q | 06:19 |
=== Mikee_C_afk is now known as Mikee_C | ||
apw | srp (N,BFTL), not really, mostly by writing comparitors | 07:58 |
leighman | hey, trying to load a backported drm module on 14.04 using dkms but it says it is tainted | 10:02 |
leighman | is this to do with signing? can I enable signing somehow? | 10:02 |
leighman | taint code is 4098 | 10:02 |
leighman | drm and i915 module | 10:02 |
apw | leighman, i think it tells you why when it taints, in dmesg | 10:17 |
apw | though it is very likley them not being signed is what it says | 10:17 |
apw | you cannot sign external modules as the key used to sign internal modules has been destroyed deliberatly | 10:18 |
apw | why do you care that it is tainted | 10:18 |
leighman | apw: it disables trace events | 10:18 |
leighman | due to http://lwn.net/Articles/588799/ I think? | 10:19 |
leighman | /proc/.../tainted gives 4098 | 10:20 |
leighman | which seems to be TAINT_OOT_MODULE and TAINT_FORCED_MODULE | 10:20 |
leighman | TAINT_FORCED_MODULE causing trace events to be disabled as in the lwn article | 10:20 |
leighman | so I think I might be out of luck? | 10:20 |
leighman | but eg. vboxdrv does not appear in modinfo to be signed but is not listed as tainted | 10:22 |
apw | no idea how vbox would avoid that, unless it specifically messes with those flags somehow | 10:23 |
apw | leighman, it sounds like you have a valid usecase for the patch mentioned, no idea how ugly it would be to port | 10:23 |
apw | you might want to start a bug on this against "linux" so we can consider fixing that | 10:24 |
apw | as we have a lot of OOT modules which tracing sounds useful in | 10:24 |
apw | actually the patch looks pretty self contained, so it might be applicable | 10:24 |
apw | leighman, ^ | 10:24 |
leighman | does look like a nice simple patch actually *he says* | 10:26 |
leighman | okay, will file a bug against linux and subscribe you? | 10:26 |
leighman | apw: ^ | 10:28 |
apw | let me know the bug # here as i won't find it otherwise | 10:30 |
soren | smb: Thanks for looking into https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1358162 | 10:31 |
ubot5 | Ubuntu bug 1358162 in linux (Ubuntu Trusty) "skb_get_rxhash prematurely renamed to skb_get_hash" [Undecided,Fix committed] | 10:31 |
soren | smb: I have a small request, though. | 10:31 |
soren | smb: Here's that patch I applied to handle it: https://github.com/JioCloud/contrail-vrouter/commit/c31ab9b1d05acb6814a3ee8ddda16882cdcc0672 | 10:32 |
soren | smb: After your patch gets applied, skb_get_rxhash will be #defined twice. | 10:33 |
apw | then i suggest you make yours say | 10:33 |
apw | #ifndef foo | 10:33 |
soren | smb: What would you suggest? A more specific KERNEL_VERSION (does that even change between kernel revisions in Ubuntu)? | 10:33 |
apw | #define foo | 10:33 |
soren | apw: That's what I was meaning to do, but I wasn't sure if that'd be sufficient (since it depends on import order). | 10:34 |
apw | in the version check? | 10:34 |
apw | soren, though this new version will be a new ABI i am sure | 10:34 |
apw | (our new version) | 10:34 |
soren | apw: How does that help me? | 10:35 |
apw | you know you only need the compat for that one abi | 10:35 |
soren | apw: Anyway, what I was wondering was if you could wrap your #define in #ifdef as well... Or am I completely off here? | 10:35 |
soren | apw: ...so that it won't depend on a specific order of #include's. | 10:36 |
leighman | apw: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1359670 | 10:36 |
ubot5 | Ubuntu bug 1359670 in linux (Ubuntu) "Unsigned oot modules are wrongly tainted and trace events disabled" [Undecided,New] | 10:36 |
* apw suspects if anything we should #undef it | 10:36 | |
soren | apw: I dunno. It might not be a problem at all. I just saw the change and wasn't sure if #ifndef'ing on my side would be sufficient. | 10:36 |
apw | soren, might not, i think you are lucky in that it has not been uploaded yet. I suspect for us we should #undef before defining it to be sure we are getting what we expect. | 10:43 |
apw | smb, ^ | 10:43 |
soren | apw: I was exactly hoping to get this addressed before the upload. | 10:44 |
soren | apw: Thanks for considering it. | 10:44 |
apw | leighman, i assume this is 14.04 which is giving you trouble | 10:46 |
leighman | apw: yup | 10:46 |
apw | leighman, ok i've pushed some test kernels to your bug, if you could do the do on those. | 11:07 |
leighman | apw: is there a linux-tools-3.13.0-35 - I get unsatisfiable dependency on linux-tools-3.13.0-35-generic | 11:35 |
leighman | apw: anyway taint bitmask looks correct and no messages in dmesg about disabling tracepoints | 11:36 |
apw | leighman, that is good enough for me, i'll push that out for review | 11:36 |
leighman | apw: awesome, thanks | 11:38 |
leighman | apw: when would it land if so? | 11:38 |
apw | if accepted, in the next SRU round, of the order of 3 weeks | 11:40 |
leighman | apw: cool,thanks | 11:41 |
=== JanC_ is now known as JanC | ||
rtg | stgraber, https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1344049/comments/3 | 17:14 |
ubot5 | Ubuntu bug 1344049 in linux (Ubuntu Trusty) "Please cherry-pick tc rule fix for userns" [Undecided,Fix committed] | 17:14 |
=== Mikee_C is now known as Mikee_C_afk |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!