=== rsalveti-afk is now known as rsalveti === rsalveti is now known as rsalveti-afk === rsalveti-afk is now known as rsalveti === rsalveti is now known as rsalveti-afk === rsalveti-afk is now known as rsalveti === rsalveti is now known as rsalveti-afk === rsalveti-afk is now known as rsalveti === rsalveti is now known as rsalveti-afk === rsalveti-afk is now known as rsalveti === rsalveti is now known as rsalveti-afk === rsalveti-afk is now known as rsalveti === rsalveti is now known as rsalveti-afk [05:30] wow [05:30] er [06:19] Hi all...Can anyone let know what is the maximum memory that can be allocated for a kernel module in x86_64???? [06:45] What is the maximum memory that can be allocated to a kernel module in x86_64 bit system?? [06:45] -1 I presume [06:46] I didnot get -1?? [06:47] Is there any reason to expect that a kernel module should be limited in the amount of memory it can allocate on an x86_64 system? [06:48] I did not aware of any limitation for single module, but I know there are limits for kmalloc [06:49] IIRC for x86 it is 128k, if you need memory more then 128k [06:49] you need to use get_free_page [06:50] I need to allocate static memory... need to create queues of total size 2 GB...but memory allocation failing... [06:50] what kernel API you use? [06:52] I m using rvmalloc(). [06:53] for 2G buffer, you need to use get_free_pages() and manage them. [06:53] its almost impossible to ask kernel to arrange 2G not-fragment-dram for you [06:54] ok i will try that... [06:55] I was able to do for 1 GB... after that it was failing.. [06:56] the idea of 2GB in-kernel terrifies me. [06:56] Just saying. [06:57] yah its true... actually trying to write a kernel module for high speed packet processing.. [06:57] lots of processing required in kernel and need to buffer the packets... [07:00] That's a seriously huge buffer! Is processing so slow? [07:02] yah but traffic rate is also high... 5gbps.. [07:02] that doesn't mean you have to be inkernel [07:02] you just have to pass out sensible regions to userspace - lots of work on efficient userspace drivers/processing has been done [07:04] yah I am trying to do that also... but getting drop in the card level... userspace copying also is a bottleneck [07:05] I am getting drop in nw driver level.. [07:05] IIRC the bottom neck for 5gbps data flow is not usually on content switch, its usually on processor frequency [07:06] rocky_IITM: hand out entire pages, no copying involved surely. [07:08] yah true... what we are trying is to create a multi-threaded driver and process it in different cores... [07:08] you will make things a lot more robust if you're only privileged when you *need to be*. [07:24] Is there any maximum limit to vmalloc_32()?? [07:28] lifeless: Is there any maximum limit to vmalloc_32()?? [07:30] rocky_IITM: I don't know, but haven't you already been pointed at get_free_page? [07:30] And isn't asking for 2GB of contiguous memory asking for your code to fail in all but the most controlled of situations? [07:33] hmm ic... but my system is having 64 GB of RAM... so why its not possible to do contiguous memory allocation?? [07:35] fragmentation [07:35] only takes 33 widely separated pages and you can't do any single 2G allocation [07:40] see ours is not a single allocation... we are trying to create queue... each node is of 16KB and the total size of the queue is 2GB [07:42] then why are you asking if there is a limit to the allocation facility of a single call? [07:43] lifeless: see when we are crossing 1 GB the allocation is failing...and we are not quite sure where is the problem [07:45] So, after you've allocated 134217728 separate 16KB memory blocks it then starts to fail? [07:47] when I have allocated 16KB * 65536 nodes then further allocation fails [07:49] RAOF: when I allocate 65536 seperate 16KB blocks then it starts to fail [08:10] smb: sorry, but I could not test the gobi loader patch last week - the hardware was already sent back [08:11] TeTeT, Well, as long as there is no testing the changes will be simply on hold. [08:11] smb: understood [08:45] * apw yawns [08:46] * smb lazily waves [08:47] apw and smb morning [08:49] hiya cooloney [08:53] * smb moans. Ok, children in the kindergarten sing songs, but why does that need amplifying? [08:54] smb: you're running a kindergarten at home? [08:54] amitk, No, there is just one close by. And today they seemed to have "escaped" [08:55] open windows, nice weather and all that, then [08:55] yup [09:04] moin [09:04] moin [10:09] hello [10:11] I compiled a kernel 2.6.29 patched rtai for lucid but when I reboot the mouse and keybourd not work [10:17] hello [10:17] I compiled a kernel 2.6.29 patched rtai for lucid but when I reboot the mouse and keybourd not work === amitk is now known as amitk-afk [10:45] salvarane, what is rtai === smb is now known as smb-afk === amitk-afk is now known as amitk [11:50] realtime for linux + emc2 cnc control [11:52] salvarane, where did you get the kernel configuration you used to build it? [11:59] /boot/config-2.6.32-23-generic-pae ; make oldconfig [12:02] salvarane, you might try the config from a version of ubuntu before 2.6.29, jaunty for example ... as the config may have changed between versions of the upstream kernel [12:02] salvarane, you might also want to try a 2.6.29 kernel from the mainline builds, and see if the keyboard works there [12:21] thanks , how can I download in synaptics a kernel 2.6.29 ? [12:21] for ubuntu [12:21] in this manner I copy config file for my kernel rtai [12:22] if all function's very well === smb-afk is now known as smb [12:52] anyone knows : /usr/bin/ld: skipping incompatible /lib/libc.so.6 when searching for /lib/libc.so.6 , how to fix? === rsalveti-afk is now known as rsalveti [12:58] * abogani waves [12:59] apw, May i disturb you for a minute? :-) [13:25] Question: [13:25] I am trying to populate the platform_driver struct [13:25] i.e. [13:25] .probe [13:25] .release [13:26] I am doing so from arch code [13:26] How do I 'see' the module's symbol? [13:26] I've tried to EXPORT_SYMBOL it and in the arch prototype I have declared it as an extern [13:27] Ideas on a postcard === rsalveti-afk is now known as rsalveti [14:38] apw: does Maverick's kernel have a known I/O eating CPU issue? [14:38] shoudl we add one if there isnt one ? :) [14:39] heh [14:39] my icore 7 feels like an old 386 when extracting packages, etc. [14:44] that's a dpkg issue [14:44] aiui [14:44] it's syncing the disk all the time [14:46] i'm more fed up with the gnome-help taking 45 seconds to start.. [14:46] and that's not a kernel problem, so I'll leave it there ;) [14:46] or, I don't think it is.. [14:47] tjaalton: ah, I know who I can bug about dpkg :-D thanks [14:48] it was a deliberate decision iirc.. [14:50] i think it only happens with ext4 [14:50] yeah [14:50] and is an issue with fsync calls [14:51] do you know what's the relevant bug report? [14:51] must be in the lucid release notes somewhere [14:51] probably [14:52] I've never had this problem in lucid with ext4 [14:52] https://wiki.ubuntu.com/LucidLynx/ReleaseNotes#Performance regressions with ext4 under certain workloads [14:53] bug 570805 it seems [14:53] Launchpad bug 570805 in dpkg (Ubuntu Lucid) (and 3 other projects) "[regression] dpkg's fsync causes massive regression in Ubuntu Server and Alternate installation times (affects: 16) (dups: 1) (heat: 125)" [High,Fix committed] https://launchpad.net/bugs/570805 [14:53] https://bugzilla.kernel.org/show_bug.cgi?id=15910 [14:54] bugzilla.kernel.org bug 15910 in ext4 "zero-length files and performance degradation" [Normal,New] [14:54] whee, seems like a fix is coming [14:54] maybe I/O is worse in maverick and this is why I'm noticing the problem now [15:25] hi [15:34] hi [15:48] apw, hi [15:49] Please guide me about my post on https://lists.ubuntu.com/archives/ubuntu-server/2010-July/004402.html [15:49] hi [15:50] kaushal, did you try the preseed change recommended in the email you are replying to there? [15:50] apw, I am using ks.cfg kickstart file [15:51] kaushal, and did you try the > d-i base-installer/kernel/override-image string linux-server [15:52] apw, as i said i dont use the preseed method at all [15:52] I use kickstart method [15:53] hrm, you have already left my area of expertise :/ ... you probabally need to ask if that file can change the kernel, i would ask on #ubuntu-devel, the installer folk mostly hang out there [15:53] ah ok [15:53] Thanks apw [15:53] np [15:55] apw, is there a mailing list for ubuntu-devel ? [15:56] i think if you are emailing them there is an ubuntu-installer list [15:56] i would ask on #ubuntu-devel first [15:57] kaushal, you give very few hints as to what your problem is from your irc request, which makes it hard for the right person to know what you need [16:05] kaushal, looks like you got your answer on #ubuntu-devel ... [16:05] apw, sure === bjf is now known as bjf[afk] === rsalveti is now known as rsalveti-lunch [17:10] it turns out that my partition was busted. No I/O issues with the kernel === amitk is now known as amitk-afk [17:46] tseliot, thats good news [17:47] * tseliot nods === rsalveti-lunch is now known as rsalveti === rsalveti is now known as rsalveti-afk