Re: CONFIG_DEVTMPFS
On Thursday 17 December 2009 10:49:33 Jacek Konieczny wrote: > The problem is udev providing the nodes won't help if tmpfs was already > mounted at /dev when udev is installed just in case, you weren't thinking so far, or just didn't know, you can still install the udev nodes if /dev is mounted. mount --move /dev /mnt rpm -Uhv udev.rpm mount --move /mnt /dev or if you don't want/can't to move: mount --bind / /mnt cp -a /somewhere/dev /mnt/dev # or mknod directly umount /mnt but imho, we should put /dev contents of udev also into /lib/udev/devices, so that in case /dev is empty due tmpfs mount when it was installed, you still get at least the initial console (our /sbin/start_udev supports /lib/udev/devices), or _at_least_ /dev/console, as if /dev/console is missing you can't login even with ssh as it seems to stall somewhere at startup. -- glen ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: CONFIG_DEVTMPFS
On Thu, Dec 17, 2009 at 09:40:32AM +0100, Patryk Zawadzki wrote: > > Unfortunately didn't get around to integrating it to distros > > (hmm... maybe today is the day? ;-), only installed by hand > > on a few systems. > > We have udev package providing the minimum required to launch shells > (/dev/zero, /dev/console etc.). The problem is udev providing the nodes won't help if tmpfs was already mounted at /dev when udev is installed (often a case during chroot installs or upgrades). And the minimum is not enough for e.g. proper initrd building (which may fail during chroot install/upgrade when udev tmps is not mounted). We can say these problems are human mistakes, not package defects, but CONFIG_DEVTMPFS could help a lot in such cases. Of course, when it is implemented correctly. Let's wait. Greets, Jacek ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: CONFIG_DEVTMPFS
On Thu, Dec 17, 2009 at 09:40:32AM +0100, Patryk Zawadzki wrote: > We have udev package providing the minimum required to launch > shells (/dev/zero, /dev/console etc.). Ah, cute. I've done it when udev wasn't pretty requisite. > In case of DEVTMPFS Thanks, already skimmed over 2.6.32 relnotes back then. I'm considering it (with vm_deadlock/swap-over-nfs patches) for custom LTSP kernels, so am naturally interested in ditching udev if no hardware changes are to occur (e.g. when we have 16M RAM or less and can't pull in all the machinery to handle usbflash -- even full-blown udev might be too much). -- while at it, http://www.altlinux.org/LTSP [ru]: LTSP5 with the better pieces of LTSP4 back in ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: CONFIG_DEVTMPFS
On Wed, Dec 16, 2009 at 7:26 PM, Michael Shigorin wrote: > JFYI, I've done dev-minimal package in ALT Linux > hoping to use it for this very reason: some bare /dev > in case modern stuff fails again. > > Unfortunately didn't get around to integrating it to distros > (hmm... maybe today is the day? ;-), only installed by hand > on a few systems. We have udev package providing the minimum required to launch shells (/dev/zero, /dev/console etc.). It's all nice but still requires initramfs to create its own device nodes. In case of DEVTMPFS you don't have to do anything at boot time, just chmod the nodes after switch_root (or let udev rules do it) :) -- Patryk Zawadzki ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en