On Wed, 2 Aug 2000, Manuel Bouyer wrote:

> On Wed, Aug 02, 2000 at 08:05:47AM -0700, Eduardo Horvath wrote:
> > On Wed, 2 Aug 2000, Manuel Bouyer wrote:
> > 
> > > This timeout should already be enouth (it's something like 10s).
> > > Also, it looks strange that the disk spins up just after spinning down.
> > > Looks like it went off in the middle of a command.
> > 
> > I don't really think that 10s is a sufficient timeout for a disk.  Maybe
> > for IDE, but some SCSI disks can take 30s or more to spin up.  ISTR that
> > when I was dealing with fibre-channel disks they took the better part of a
> > minute to spin up.  10s is cutting it pretty close for spinups even on
> > some of the faster disks.
> 
> Hum more than 10s would make power management on laptop almost unusable :)

Then maybe the disk driver should keep track of how long the disk has been
idle and always send the start/stop unit command if it's been idle more
than a specific amount of time.

Eduardo Horvath