Philip Blundell (email@example.com)
Mon, 31 Aug 1998 16:26:23 +0100
>The irq now is owned by the parport highlevel driver. This allow us to
>detect irq out of cad and avoiding the many request_irq() free_irq() done
>during the parport scheduling.
I still don't think this is the right way to solve the problem.
>what does it mean cad ? ;-)
I think it stands for `currently active device'.
>in parport_share.c will produce 0 asm code with parport_pc, if
>PARPORT_OTHERS is not #defined.
Having the lowlevel driver handle the interrupt would have no overhead
regardless of the setting of PARPORT_OTHERS. At the moment you have
parport_share handle the interrupt and make a call both down to parport_pc and
up to lp (or whatever). If parport_pc handled the interrupts directly it
would only need to call up to lp and parport_share would not be involved.
>I have not said this code has to go in 2.2 but at least I just implemented
For 2.2 I guess it's slightly better than what we have now (fewer request/free
irq calls) and maybe it's a less invasive fix than doing it properly. Even if
we do put this patch into 2.2 I'd like to change to the other scheme for
-- with the single word "unsubscribe" in the body of the message. --
This archive was generated by hypermail 2.0b3 on Wed 30 Dec 1998 - 10:18:11 EST