/srv/irclogs.ubuntu.com/2016/04/05/#ubuntu-kernel.txt

=== shuduo_ is now known as shuduo
pixel6692Hello is there way to get/make/build linux-image-extra for 4.5 (15.10)? I need aufs module and I also need to use 4.5 because of my ethernet driver08:42
xnoxpixel6692, usually one can find ethernet driver as a dkms module on e.g. github and use that with 15.10 kernel.09:06
xnoxpixel6692, note that e.g. even 16.04 has 4.4 kernel, but with some backports09:06
pixel6692I am currently running 4.5 from here http://kernel.ubuntu.com/~kernel-ppa/mainline/v4.5-wily/ I just need aufs module09:10
pixel6692that is typicaly in package linux-image-extra (which is not in this tree), I am trying to build it with this module now09:10
pixel6692my ethernet driver was repaired in 4.509:11
=== xnox_ is now known as xnox
=== psivaa_ is now known as psivaa
=== inaddy_ is now known as inaddy
=== TJ_Remix is now known as TJ-
cryptohwlinuxhelwhat function can I use from the kernel to copy a __user pointer to another __user pointer? can I just do memcpy?14:19
=== shuduo is now known as shuduo-afk
apwcryptohwlinuxhel, it seems unintuitive to want to copy userspace to userspace (within the same process) in a system call, i beleieve you cannot use mempcy as the kernel handles faults for userspace addresses differently if you are in a call it expects to touch userspace14:36
cryptohwlinuxhelyeah there is a flag I have to set to bypass what my driver does14:37
cryptohwlinuxhelso the bypass is to simply do a memory copy from one _user pointer to another14:38
cryptohwlinuxhelit cannot be done from outside the kernel since it is a driver14:38
cryptohwlinuxhelalso the kernel gives this flag a shared common memory that all processes can see14:38
cryptohwlinuxhelso what can I use if I cannot use memcpy?14:39
cryptohwlinuxhelI can always allocate local memory copyfromuser to my local buffer and then copytouser from my buffer but that is not efficient and I would like to avoid it14:40
apwi am not sure there is one, but i have not looked, you will find copyfromuser etc are marked specially as copiers, so you could look at what other calls are the same14:43
cryptohwlinuxhelwhat do you mean they are marked as copiers?15:00
cryptohwlinuxhelhttp://lxr.free-electrons.com/source/arch/arm/include/asm/uaccess.h#L55215:01
cryptohwlinuxhelI see no markings15:01
tych0i'm trying to build a kernel from xenial's current master, and i'm getting, http://paste.ubuntu.com/15633017/16:25
tych0any ideas?16:25
genkgojsalisbury: nice achievements again on the backup issue16:47
jsalisburygenkgo, getting closer.  I may have to perform a bisect of the entire tree, which will take a few more days of testing16:48
pesarihey, have you seen this with xenial-proposed 4.4.0-17 ? https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/156495117:00
ubot5`Launchpad bug 1564951 in systemd (Ubuntu) "systemd-timesyncd: Failed to call clock_adjtime(): Invalid argument" [Undecided,Confirmed]17:00
genkgojsalisbury: i am really really wondering what it is17:02
genkgojsalisbury: probably nobody ever thought of17:02
genkgo*something17:02
genkgoand might fixes other things too17:02
jsalisburygenkgo, Yeah, it isn't any of the hv commits or virtio commits between these two releases.  That's why sometimes a bisect is needed to identify the real offending commit.17:08
genkgojsalisbury: the virtio commits are also not the problem?17:09
jsalisburygenkgo, no.  I built a test kernel with them all reverted and I still hit the bug in about an hour.17:09
genkgook17:09
kamaltych0, your problem there is (I think) that you're trying to build with just 'make' ...  to build an Ubuntu kernel, follow the "Building the kernel" instructions here:17:12
kamalhttps://wiki.ubuntu.com/Kernel/BuildYourOwnKernel17:12
kamalspecifically:17:12
kamalfakeroot debian/rules clean17:12
kamalfakeroot debian/rules binary-generic17:12
tych0kamal: ah, interesting, ok17:13
kamaltych0, that ^^ will build security/apparmor/ just fine.    (whereas, with bare 'make', the thing you reported is indeed the first failure I get too).17:14
tych0interesting; it used to work with make bindeb-pkg before that patch :)17:14
genkgojsalisbury: thanks, looking forward to the end result! you must tired of all these backups by now ;)17:14
tych0although perhaps i was doing something horribly wrong17:14
jsalisburygenkgo, heh, not tired yet.  It too much of a fun puzzle.  The wait while testing is the pain.17:15
genkgojsalisbury: hehe, alright. glad to hear you still got the spirit!17:16
kamaltych0, hmmm.  well if it used to work (when exactly?) then it probably should still -- maybe we actually do have a new glitch to sort out.17:16
tych0kamal: yeah, it used to work as of a few weeks ago with the xenial kernel's master17:16
tych0let me look for the exact ref17:16
jsalisburygenkgo, We'll get to the bottom of this. 17:17
tych0kamal: f35781cd96795075436fde15abfc1b94bdffb99c17:17
kamaltych0, yup, you've got a point there!17:18
tyhickstych0: looks like this would do it: https://paste.ubuntu.com/15634420/17:19
* tych0 builds17:20
tych0tyhicks: but yeah, that looks suspicious :)17:21
tych0tyhicks: yeah, that fixes it17:22
tyhickstych0: thanks - I'll send a patch to the kteam17:22
tych0tyhicks: cool, thanks17:22
kamaltych0, tyhicks: nice job ty-team!  ;-)  thanks17:23
tyhicksheh :)17:23
tych0i have to be useful every once in a while if i'm going to troll tyhicks' tab completion everywhere17:24
kamal:-)17:24
tyhicksjjohansen: ^ FYI so you know the reason for the patch that I'm about to send out17:24
jjohansenI'm not sure why our builds didn't catch that17:27
tyhicksme neither17:27
tyhicksjjohansen: are you wanting to understand it more or should I go ahead and send out the patch?17:29
jjohansentyhicks: just send the patch17:29

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