On Wed, 2003-07-02 at 18:28, Marcelo Tosatti wrote:
> On Wed, 2 Jul 2003, Marcelo Tosatti wrote:
> 
> >
> > Hello people,
> >
> > What is the status of the IO scheduler fixes for increased fairness for
> > 2.4 ?
> >
> > I haven't had time to read and think about everything you guys discussed,
> > so a brief summary would be very helpful for me.

Ok, based on mail from Andrea, it seems like reusing the existing
elvtune ioctl is the way to go for now.  The patch below uses
max_bomb_segments for two things.

1) the number of sectors allowed on a given request queue before new io
triggers an unplug (q->max_queue_sectors)

2) to enable or disable q->full checks.  If max_bomb_segments is odd, it
means the q->full checks are on, even means they are off.

The ioctl code on the kernel side is setup to allow this:

elvtune -b 1 /dev/xxx 
elvtune -b 0 /dev/xxx

For just switching q->full on and off without changing
q->max_queue_sectors.

elvtune -b 8192 /dev/xxx will get you the current default behavior (4MB
in flight, q->full checks off)

-chris