Greetings,
I've recently upgraded udev from 063 to 082 and then 084.
With 063, startup of udev was near-instant; with both 082 and 084, it
takes a significant ammount of time (~15s) to create the base devices
using udevstart or udevsynthesize (this one is taken from Debian, which
apparently in turn taken it from SuSE; the rest of codebase is vanilla).
This issue appears on kernel 2.6.15.1 with SquashFS 2.2r2, SWSUP2 2.2
and VesaFB-TNG 1.0-rc1-r3 patches.
The init script used simply mounts 10MiB tmpfs onto /dev, creates
/dev/.udev/{db,queue} directories, then runs udevd --daemon and then
udevsynthesize or udevstart (tried both, same result).
I'm quite out of ideas, I don't think downgrading udev is the best
solution, so I wonder: what takes such a long time in udevstart? What
can I alter at my end, or is this a known bug (or feature)?
Thanks in advance for reply
-- iSteve
On Mon, Jan 30, 2006 at 11:04:14PM +0100, iSteve wrote:
> Greetings,
> I've recently upgraded udev from 063 to 082 and then 084.
>
> With 063, startup of udev was near-instant; with both 082 and 084, it
> takes a significant ammount of time (~15s) to create the base devices
> using udevstart or udevsynthesize (this one is taken from Debian, which
> apparently in turn taken it from SuSE; the rest of codebase is vanilla).
With the 2.6.15 kernel, udevstart is no longer needed. Please use the
recommended shell script instead (as posted to the linux-hotplug-devel
mailing list.)
> This issue appears on kernel 2.6.15.1 with SquashFS 2.2r2, SWSUP2 2.2
> and VesaFB-TNG 1.0-rc1-r3 patches.
All of which are not patches included in the mainline kernel. Can you
try it without these?
> The init script used simply mounts 10MiB tmpfs onto /dev, creates
> /dev/.udev/{db,queue} directories, then runs udevd --daemon and then
> udevsynthesize or udevstart (tried both, same result).
Again, don't do that :)
Also, udev specific questions like these are best asked on the
linux-hotplug-devel mailing list, as the udev documentation states.
thanks,
greg k-h