On Thu, Aug 5, 2010 at 4:47 PM, Nirmal Guhan <vavat...@gmail.com> wrote: > On Tue, Aug 3, 2010 at 1:22 AM, Toby Corkindale > <toby.corkind...@strategicdata.com.au> wrote: >> On 03/08/10 17:44, Nirmal Guhan wrote: >>> On Mon, Aug 2, 2010 at 6:07 PM, Toby Corkindale >>> <toby.corkind...@strategicdata.com.au> wrote: >>>> On 03/08/10 09:04, Nirmal Guhan wrote: >>>>> 4) Hot swap does not work within the container. After usb device is >>>>> reinserted, container cannot recognize it but host can. >>>>> 5) "mount" within the container always displays just one single line >>>>> while I have few more in fstab including the above /media stuff. >>>>> none on /proc/sys/fs/binfmt_misc type binfmt_misc (rw) >>>> >>>> Again, that's because of the way LXC works with the filesystem. >>>> >>>> Perhaps you could just bind-mount the whole /media directory into the >>>> guest containers, to their /media directory? That might work better for >>>> you, although still not quite what you want. >>> >>> Thanks Toby. I doubt if this will address #1 and #4 above. Basically, >>> how to make hot swap work? Or what are the workaround to get >>> notifications if I have to manually mount/umount. >> >> >> I think you would need to adjust the devices permissions for your >> container, for the usb nodes.. but I'm not sure.. then the container >> could talk to the USB devices over USB, and handle the hotswapping. >> However I'd expect conflicts to occur with the host trying to talk to >> the same device. >> Try adding this to your config file? >> >> lxc.cgroup.devices.allow = c 189:* rwm >> >> I'm no expert and I haven't tried this myself, mind, so you might prefer >> to wait for someone more qualified to answer. >> >> -Toby >> > Can you explain what this config means and 189 in particular? I tried > this but did not help in hot swap. I believe that is because udev is > not supported within the container and yes, you are correct about > host/guest accessing the same device.
the 189:* corresponds to some device with a major number 189, and any minor number. i'm guessing 189 must correspond to USB drivers. that config option is basically saying "allow the container create character device nodes matching 189:*" check out "man mknod" for more info. > Any plans to support udev and usb pass-through in lxc ? IMO "lxc" itself is a very specific technology: to add namespaces to processes/users/network/etc... it really doesn't have anything to do with full blown virtualization, it's just a use case. to support udev, a daemon (or is it kernel?) on the host needs to send uevents to the container namespace to "fake" shared and private devices intended for the container. this probably needs a lot of config... in the *nix spirit, that should be handled by a separate specialized app... i think the last paragraph is correct :-) anyways, you could try to use shared/slave mounts along with the bind mount to accomplish what you want. from mount(8): "Since Linux 2.6.15 it is possible to mark a mount and its submounts as shared, private, slave or unbindable. A shared mount provides ability to create mirrors of that mount such that mounts and umounts within any of the mirrors propagate to the other mirror. A slave mount receives propagation from its master, but any not vice-versa. A private mount carries no propagation abilities. A unbindable mount is a private mount which cannot cloned through a bind operation. Detailed semantics is documented in Documentation/sharedsubtree.txt file in the kernel source tree." it might be as easy as: mount --make-shared /media providing /media itself is a mount... else you might have to: mount --make-shared / let us know if/how it goes. C Anthony ------------------------------------------------------------------------------ This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev _______________________________________________ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users