After upgrading to udev-079-r1, the device nodes are no longer created at boot time. This, of course, leads to a failed boot. Downgrading to udev-079-r1 fixes this issue.
Created attachment 77839 [details] Console log from booting
After booting into single mode, the following device nodes are present: # ls -l /dev/ total 0 crw------- 1 root tty 5, 1 Jan 22 21:00 console lrwxrwxrwx 1 root root 13 Jan 22 20:59 fd -> /proc/self/fd prw------- 1 root root 0 Jan 22 20:59 initctl crw-rw-rw- 1 root root 1, 3 Jan 22 18:05 null drwxr-xr-x 2 root root 0 Jan 22 20:59 pts drwxrwxrwt 2 root root 40 Jan 22 20:59 shm lrwxrwxrwx 1 root root 4 Jan 22 20:59 stderr -> fd/2 lrwxrwxrwx 1 root root 4 Jan 22 20:59 stdin -> fd/0 lrwxrwxrwx 1 root root 4 Jan 22 20:59 stdout -> fd/1 crw-rw-rw- 1 root root 1, 5 Jan 22 18:05 zero
(In reply to comment #2) > After booting into single mode, the following device nodes are present: After booting into single user mode and running `udevstart`, the correct device nodes appears in /dev/
Created attachment 77844 [details] Output from `emerge --info` This is a GeodeGX1 based Soekris net4801 system.
Created attachment 77845 [details] linux-2.6.15.1 .config Kernel configuration used.
Bug #119894 seems to cover a similar issue.
udevstart should not be run for 2.6.15+ kernels. Also, what version is not working? In your original description, you say 079-r1 failed, so you went back to 079-r1 and it worked. I'm a bit confused here :)
Heh, you have disabled CONFIG_HOTPLUG in your kernel .config. Care to enable that and then this will probably work again...
(In reply to comment #7) > Also, what version is not working? In your original description, you say > 079-r1 failed, so you went back to 079-r1 and it worked. Sorry, meant to say that I went back to 070-r1, which worked. (In reply to comment #8) > Heh, you have disabled CONFIG_HOTPLUG in your kernel .config. Care to enable > that and then this will probably work again... Testing that right now, thanks.
(In reply to comment #8) > Heh, you have disabled CONFIG_HOTPLUG in your kernel .config. Care to enable > that and then this will probably work again... Enabling CONFIG_HOTPLUG made it work again, thanks.
Would it make sense to add a check (using linux-info.eclass) to see if CONFIG_HOTPLUG (and perhaps CONFIG_KOBJECT_UEVENT?) are enabled, and warn the user if they are not? I'll be happy to prepare a small patch for the ebuild.
No, because those config options are pretty much just going away, and will be very hard for people to disable them (2.6.16-rc1 already has this change.)
Ok :)