pwnguin | mermaidman: thats nice, but from wikipedia: The file system is currently marked as developmental and is titled "ext4dev".[5] It is considered unstable and so is not recommended for use in production environments, since data corruption is possible in this early stage of development." | 00:09 |
---|---|---|
pwnguin | im kinda thinking anyone interested in ext4 outta learn how to build kernels anyways | 00:09 |
mermaidman | pwnguin: Umm you have to add some special test_fs code to ext3 in order to mount ext4 which the ubuntu devs dont understand | 00:10 |
pwnguin | mermaidman: when you say "dont understand" | 00:12 |
pwnguin | is there specific evidence? | 00:12 |
mermaidman | they never respond to all my suggestions | 00:12 |
pwnguin | im not sure those are equivilent | 00:12 |
pwnguin | where can i read these suggestions? | 00:13 |
mermaidman | forums | 00:13 |
pwnguin | oh well duh | 00:13 |
pwnguin | you'd get more attention from IRC or the mailing list | 00:14 |
mermaidman | http://ubuntuforums.org/showthread.php?t=837589 | 00:14 |
mermaidman | went there to still no answer | 00:14 |
pwnguin | on -kernel? | 00:14 |
mermaidman | yup | 00:15 |
pwnguin | gotta grab dinner | 00:15 |
pwnguin | but i'll check the archive | 00:15 |
pwnguin | john nelson? | 00:17 |
mermaidman | yeah | 00:18 |
pwnguin | We need to talk about effective writing | 00:19 |
mermaidman | i was in a rush that day | 00:21 |
pwnguin | https://lists.ubuntu.com/archives/kernel-team/2008-July/002793.html | 00:21 |
pwnguin | well, then the little time you did spend was wasted =/ | 00:21 |
mermaidman | Since i have all the time of the night I will write another one right now | 00:22 |
pwnguin | if you communicate intelligently, you may find people are more willing to talk with you | 00:22 |
pwnguin | but I believe there's also some structural bias | 00:24 |
pwnguin | from what ive seen many in the kernel team are paid canonical employees who work 9-5 | 00:24 |
mermaidman | just sent it its in the moderation queue now | 00:24 |
pwnguin | being the weekend, there's probably a bit less attention to go around, if you know what i mean | 00:25 |
mermaidman | Yeah | 00:25 |
pwnguin | well pizza time | 00:29 |
mermaidman | cool | 00:29 |
pwnguin | you know what, if ted doesn't think it's ready, i don't think we should bother enabling it | 00:49 |
mermaidman | ted Tso? | 00:50 |
pwnguin | yes | 00:50 |
mermaidman | finally finshed the annoying fsck on ext3 | 00:53 |
=== emma_ is now known as emma | ||
=== alex_jon1 is now known as alex_joni | ||
benje | hi there a bug in the se401 driver i get this [ 3335.080589] se401: probe of 3-1:1.0 failed with error -5 | 17:16 |
benje | with a kensington webcam Bus 003 Device 004: ID 047d:5001 Kensington | 17:16 |
benje | under hardy | 17:16 |
sn9 | benje: you're sure it's not bug 88746? | 17:17 |
sn9 | !bug 88746 | 17:18 |
sn9 | where's ubottu? | 17:18 |
benje | i go to see i just searching there | 17:18 |
benje | shure sn9 it's about ehci but it using usb 3-1: new full speed USB device using uhci_hcd and address 4 | 17:19 |
sn9 | any hub involved? | 17:20 |
benje | since a long time this webcam not working | 17:20 |
benje | no directly on face pc (dell 1800) | 17:20 |
sn9 | ok... | 17:20 |
benje | it was working a day | 17:20 |
benje | do you want full dmesg ? | 17:21 |
benje | sn9, http://pastebin.com/m3d7c3f1e | 17:22 |
sn9 | 404 | 17:24 |
benje | sn9, :/ | 17:25 |
benje | this is working for me | 17:25 |
benje | sn9, do you have an other paste | 17:25 |
benje | prefered | 17:25 |
sn9 | i probably won't learn anything from it, anyway | 17:26 |
benje | yes nothing particular | 17:26 |
benje | detect w"ebcam firware format | 17:26 |
benje | but [ 3335.080563] /build/buildd/linux-2.6.24/drivers/media/video/se401.c: Bayer format not supported! | 17:26 |
benje | the light at webcam stay lighting | 17:28 |
benje | sn9, what is it need for post bug about this ? i have even post bug | 17:29 |
sn9 | have you looked at upstream info about the driver? | 17:29 |
benje | but what will be usefull | 17:29 |
benje | there a patch for version http://members.chello.nl/~j.vreeken/se401/ | 17:30 |
benje | but only for epcam | 17:31 |
benje | it's now in kernel tree | 17:31 |
benje | no sn9 | 17:31 |
sn9 | do you know whether this works in later kernels? | 17:31 |
benje | sn9, here http://cateee.net/lkddb/web-lkddb/USB_SE401.html . | 17:31 |
benje | no i don't know | 17:32 |
benje | this was working in dapper and edgy at start | 17:32 |
benje | at bigining | 17:32 |
benje | begining | 17:32 |
benje | but she is in all list as working :/ | 17:33 |
benje | http://lxr.linux.no/linux/Documentation/video4linux/se401.txt | 17:33 |
sn9 | so, it's a regression? | 17:35 |
benje | yes | 17:35 |
benje | it 's avery old webcam doesn't work in xp | 17:36 |
benje | ;) | 17:36 |
benje | no driver but it always working under linux | 17:37 |
benje | sn9, where can i find upstream of ubuntu ? | 17:37 |
sn9 | if the driver is in the mainline kernel tree, the upstream is often the mainline kernel itself, unless patches are applied | 17:39 |
sn9 | ubuntu maintains local kernel trees for each version on kernel.org, too | 17:39 |
benje | ok | 17:40 |
benje | i haven't try with an other distrib | 17:40 |
benje | i try under debian etch | 17:40 |
sn9 | under debian, you may need to use module-assistant | 17:41 |
benje | same error | 17:41 |
benje | :/ | 17:42 |
benje | not needed ;) | 17:43 |
benje | thanks sn9 | 17:43 |
benje | sn9, do you know what is the problem ? | 17:43 |
benje | or seems | 17:44 |
sn9 | it sounds like the upstream driver might be broken | 17:44 |
benje | ok thanks sn9 i will post bug or do you think it s not necessary | 17:46 |
sn9 | search for existing bug reports first | 17:46 |
benje | i found nothing about se401 and kensington webcam | 17:47 |
sn9 | no se401 bugs at all? file it, then | 17:48 |
benje | no, oki | 17:49 |
benje | no now i discution with friend but i will ;) | 17:49 |
benje | thanks and bye | 18:27 |
DHR | I reported this problem a few days ago. IOCTL interface to MTRRs appears broken in 8.04.1. Thoughts? https://bugs.launchpad.net/ubuntu/+source/linux-meta/+bug/253204 | 21:07 |
pwnguin | who moderates mail to ubuntu-kernel? | 21:54 |
sn9 | it should say at the bottom of the listinfo page for that list | 21:56 |
BenC | pwnguin: rtg does | 21:57 |
BenC | DHR: looking at it now | 21:57 |
pwnguin | ok thanks; a guy came by yesterday asking about ext4 and he said it was waiting moderation approval | 21:58 |
pwnguin | probably it'll be taken care of on monday | 21:58 |
pwnguin | a bit unfortunate for our weekend hackers | 21:58 |
DHR | BenC: thanks! | 21:59 |
BenC | DHR: it seems to be specific to the system, so the fact that it works on fedora9 is misleading | 22:00 |
DHR | BenC: in what sense is it specific? | 22:01 |
BenC | DHR: on my laptop, it prints all of them, but on my 8-way Core2 box, it only prints the first two | 22:01 |
BenC | and they are running the exact same kernel | 22:02 |
DHR | OK. | 22:02 |
BenC | my laptop has 4 registers enabled, and it prints all 4...the workstation has 3 registers and it only prints the first 2 | 22:03 |
DHR | There seem to be some problems with MTRRs on Intel systems with >4G. | 22:03 |
BenC | Nope | 22:03 |
BenC | neither of these has > 4G | 22:03 |
DHR | Sorry: not *this* problem, but it is the reason I care about MTRRs. | 22:03 |
BenC | Ah | 22:03 |
DHR | Some BIOSes set MTRRs with overlapping ranges. X drivers cannot then change their frame buffer to Write Combining. | 22:04 |
DHR | So: I wanted to write a C program to eliminate overlapping ranges. | 22:04 |
BenC | isn't that something you can do in your BIOS? | 22:06 |
DHR | There is a kernel patch to do this. I'm trying to understand the code (right now). I'm not convinced it is great code. | 22:06 |
DHR | No, I cannot write a new BIOS easily :-) | 22:06 |
BenC | No, I mean you can set it in your BIOS config :P | 22:06 |
DHR | no. Not normally. The usual work-around is to remove memory! | 22:07 |
DHR | You can also hack away with echos to /proc/mtrr. I thought ioctl interface would be cleaner. But it is broken, I think. | 22:08 |
DHR | kernel patch as discussed on LKML: http://kerneltrap.org/mailarchive/linux-kernel/2008/5/1/1686464 | 22:10 |
DHR | example of the problems caused by overlapping MTRRs: https://bugs.launchpad.net/ubuntu/+source/linux-restricted-modules-2.6.24/+bug/224404 | 22:11 |
DHR | BenC: since you have observed to problem with the ioctl interface too, could you add to https://bugs.launchpad.net/ubuntu/+source/linux-meta/+bug/253204 | 22:13 |
BenC | DHR: ... | 22:17 |
BenC | /* Hide entries that go above 4GB */ | 22:17 |
BenC | if (gentry.base + size - 1 >= (1UL << (8 * sizeof(gentry.size) - PAGE_SHIFT)) | 22:17 |
BenC | || size >= (1UL << (8 * sizeof(gentry.size) - PAGE_SHIFT))) | 22:17 |
BenC | gentry.base = gentry.size = gentry.type = 0; | 22:17 |
BenC | that's from the MTRRIOC_GET_ENTRY code in the kernel | 22:17 |
DHR | interesting! | 22:18 |
BenC | So it has nothing to do with > 4G of memory, just mapped to > 4G | 22:18 |
DHR | what file is that in? | 22:18 |
BenC | and sure enough, but 3rd entry on the workstation is mapped at 4G | 22:18 |
BenC | arch/x86/kernel/cpu/mtrr/if.c | 22:18 |
BenC | s/but/my/ | 22:19 |
DHR | Why would that be the correct thing to do? Especialy on amd64 (which I'm on)? | 22:19 |
mjg59 | What's the size of the field in the structure that the address gets returned in? | 22:20 |
BenC | mjg59: ah, it's unsigned int | 22:21 |
DHR | "unsigned int size" | 22:21 |
BenC | that's pretty stupid | 22:21 |
mjg59 | Yeah, so ABI concerns | 22:22 |
BenC | well, that still shouldn't be a problem...the base is unsigned long, so it should still be able to print up to 4G size | 22:22 |
mjg59 | Probably best to add a new ioctl... | 22:23 |
mjg59 | (nngh ioctls) | 22:23 |
BenC | yeah, best to just shove it in /sys somewhere and add something to X to use that | 22:23 |
DHR | if you look at /usr/include/asm/mtrr.h, i386 differs from amd64. So no excuse. | 22:23 |
BenC | ioctl's are just ugly anyway, I don't blame them for not fixing it by now | 22:24 |
BenC | plus they are working on something better for X to use, so focus is probably there | 22:25 |
DHR | /proc is ugly too: race conditions and all. | 22:25 |
mjg59 | You probably want to be using PAT on newer systems anyway | 22:27 |
BenC | right, PAT is what I was thinking of | 22:28 |
DHR | if the IOCTL cannot return the correct value, it ought to return an error. | 22:29 |
DHR | EOVERFLOW Value too large to be stored in data type (POSIX.1) ?? | 22:31 |
DHR | ERANGE Result too large (POSIX.1, C99) ?? | 22:32 |
DHR | is a size >= 4G legal on i386? It could well be, given PAE. | 22:34 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!