Re: [patch] softirq-2.4.5-A1

[Andrea Arcangeli]

On Sat, May 26, 2001 at 09:33:59PM +0200, Ingo Molnar wrote:
>    interrupts disabled, otherwise we can end up getting another softirq
>    right after the test. (and this causes a missed softirq.)

an irq that could mark the softirq active under entry.S will also run
do_softirq itself before iret to entry.S. If the softirq remains active
after an irq it it because it was marked active again under do_softirq
and ksoftirq is the way to go for fixing that case I think.

Andrea