From: Tony B. <tb...@gm...> - 2006-03-09 19:28:47
|
Found a better patch but it is for the kernel. The uevent patch. It will b= e included in the 2.6.16 kernel so it might be best to just wait till 2.6.16goes stable. These 3 lines will then replace the whole udevstart/coldplug issue and much better than udevsynthesize. for i in /sys/block/*/*/uevent; do echo 1 > $i; done for i in /sys/class/*/*/uevent; do echo 1 > $i; done for i in /sys/bus/*/devices/*/uevent; do echo 1 > $i; done Let me know what you folks think. I can patch the stock kernel we use if w= e want to play with it before 2.6.16 comes out. Regards, Tony On 3/9/06, Tony Brijeski <tb...@gm...> wrote: > > I already have the package and the patch. I will make it and then we can > disable hotplug altogether and see how it works. > > > > > On 3/9/06, uel archuletta <ue...@gm...> wrote: > > > > Tony, > > > > please go ahead and make the package. I know udev is up to 0.87 you can > > get it here http://www.us.kernel.org/pub/linux/utils/kernel/hotplug/ude= v-087.tar.bz2 > > > > once the package is done we can make sure all the udev rules, scripts, > > and permissions files are complete, > > and begin testing it > > > > > > Thanks > > Uelsk8s > > > > > > On 3/9/06, Tony Brijeski <tb...@gm...> wrote: > > > > > > I did some testing last night and it made no difference. I also foun= d > > > that hotplug is running but it is sysvinit. I made a coldplug script= but it > > > didn't work. > > > > > > The other option we have is the udevsynthesize patch for udev. It > > > removes the need for udevstart and coldplug and can actually go throu= gh and > > > setup all the /dev/ entries and load modules in about 1 second on a > > > 1.7Gig processor. Hotplug on my AthlonXP 2400+(which is overclocked > > > like a 2900+ or better) takes 7 seconds in addition to the udevstart = time. > > > > > > Also while looking into the whole udev thing there is a lot of > > > confusion. The VL package on SOHO says 0.71 version but in the > > > package contents it is udev-0.64. What is this about? It looks like > > > the packlage was created manually so an uninstall of that package wou= ld fail > > > as the slack tools strip out the 1st 2 characters which should be ./ = so > > > whomever packaged that up really needs to package it up using this ty= pe of > > > command from the root of the filesystem: > > > > > > tar cvzpf udev-0.71.....tgz ./install ./usr...... etc - there must > > > always be a ./ before the folder name so that the slack install doesn= 't > > > strip it out. > > > > > > If you folks want, I can patch udev with udevsynthesize and create a > > > proper new package. This would completely eliminate the need for hot= plug > > > but we would have to make sure all of our udev rules, scripts, and > > > permissions files were complete. > > > > > > Regards, > > > Tony > > > > > > > > > > > > On 3/8/06, Tony Brijeski <tb...@gm...> wrote: > > > > > > > > I have done some digging and coldplug is actually just the old > > > > hotplug *.rc scripts. We have hotplug installed and there but it n= eeds to > > > > be called after udev is started. > > > > > > > > rc.hotplug is what coldplug really is. If it is run after udev, > > > > then the devices will be setup. I will do some testing and if all = is well, > > > > this should be a wash to just add to rc.S after starting udev. > > > > > > > > I'm testing it now and will let you all know. > > > > > > > > Regards, > > > > Tony > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > On 3/7/06, uel archuletta < ue...@gm...> wrote: > > > > > > > > > > I also have the udev and rules from zenwalk and it replaces > > > > > hotplug as well > > > > > so besides X setup like you said se have other options. > > > > > > > > > > > > > > > Uel > > > > > > > > > > > > > > > > > > > > On 3/7/06, Tony Brijeski <tb...@gm...> wrote: > > > > > > > > > > > > How much of vlsetup is still being used to configure the > > > > > > hardware and how much of udev/hotplug are we now using? > > > > > > > > > > > > I think we should concentrate on setting up udev to handle > > > > > > pretty much everything vlsetup used to but it will be on the fl= y during > > > > > > boot. > > > > > > > > > > > > Mandriva has setup some great scripts and rules for udev to tak= e > > > > > > care of most of the hardware so a hardware probe is not necessa= ry anymore > > > > > > other than for X setup. > > > > > > > > > > > > Let me know your thoughts on this. If we still want to keep > > > > > > vlsetup, then I will either have to try to find the source for = it in one of > > > > > > my very old backups or I will have to take a look at the curren= t hwsetup > > > > > > from knoppix and start from scratch again. > > > > > > > > > > > > I am thinking that the current state of udev makes most of > > > > > > vlsetup useless. If we have the right scripts and rules, there= should be no > > > > > > problems with letting udev take care of all of it. > > > > > > > > > > > > Regards, > > > > > > Tony > > > > > > > > > > > > > > > > > > > > > > > > > > > |