On Thu, 18 Dec 1997 09:01-EST  "Robert.V.Baron" wrote:
> 
> This isn't rocket science.  Modify the irq allocator to not allow use of
> certain irq's.  (I keep suggesting this over and over.)  If you are
> opposed to this on general principles, then at least pass a #define
> constant down to the pcmcia saying that you are building a boot kernel
> and it should use a restrictive set of irq's.  NOTE: there is alreay
> such a mask in the pcmcia scheme -- so assuming it works, we are only
> quibling what values it should have.  You need at most 3 good irq's
> one for the pcmcia interrupt and one for each of two cards.  I believe
> that 9, 10, 11 are pretty safe.  (Note: according to the pcmcia notes
> 10 is used by the NEC VERSA dock for scsi -- who cares.  So you cann't
> boot a NEC with a dock.  [I've never seen anyone else reserve 10.]

Perhaps it is (rocket science).

9 isn't as safe as you might think; on a SAGER NP7500A 9 is unusable
(see pr #4464; it doesn't even have to be connected to a dock).

Further more irq 5 isn't safe necessarily either (see pr #4717).

The problem is, that any notebook manufacturer can use any interupt
they want to reserve for special hardware (ie docking stations for
example).  As a result, no matter what mask is choosen, there will
be hardware for which it is unsuitable.

Usually irqs in use are discovered by 1) hardware that autoconfs,
but is unusable due to the conflict; 2) the machine simply locks
up.

Perhaps 1) could be handled with timeout timers.  However 2) is
down right ugly to deal with.

This has been hashed out before by Jason & Co on this list; I see
no robust solution at this time; there I conclude is it "rocket
science" from my POV.

If you have a solution; I'm sure many people would love to see it
go into the tree...