Bug#763695: console-setup: even worse than assumed

[Eduard Bloch <edi@gmx.de>]

Hallo,
* Samuel Thibault [Tue, Aug 25 2015, 09:56:39PM]:
> Eduard Bloch, le Tue 25 Aug 2015 21:35:36 +0200, a écrit :
> > * Samuel Thibault [Tue, Aug 25 2015, 08:54:35PM]:
> > > Just so that people reading the log don't take it too exagerated:
> > 
> > > Eduard Bloch, le Tue 25 Aug 2015 20:24:22 +0200, a écrit :
> > > > I checked again... about 20% of the time is spent in two calls of pidof
> > > 
> > > We're here talking about ~15ms each.
> > 
> > I don't. It's ~270ms each (I added another log with timestamps).
> 
> How long is the readlink /proc/*/exe | grep?

~70ms (strace -f -tt -o logfile sh -c 'readlink /proc/*/exe | fgrep slash')

> > > > And about 70% is spent by loadkeys executing the mentioned storm of ioctls.
> > > 
> > > And about 70ms.
> > 
> > Or ~1.8s on the mentioned device.
> 
> Ouch. The optimization I mentioned could perhaps reduce that, in a not
> too complex setup (shift, alt, control, altgr, capslock modifiers) by a
> 4x factor. To go further a kernel patch is really needed, there is no
> way around it.

Ok, thanks for investigation.

Regards,
Eduard.