> I tried using setitimer as a scheduling source for lowlatency testing
> with generally good results. I left my test running over the weekend
> and found some large latencies had occurred. It appears that since
> the signal generation occurs in timer_bh, it is possible for a user
> task to run between when the timer_interrupt occurs and when timer_bh
> runs (~2msec in my case).
that should not be possible (except on SMP where there might be a window).
You are not running SMP, right? The TIMER_BH is run right when the IRQ
handler returns to non-irq context (which might be user-space or kernel
space), so it's not possible to 'lose' a TIMER_BH event due to some task
running. BHs might get delayed due to interrupts though, so you might want
to profile how long various interrupt sources execute.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to email@example.com
Please read the FAQ at http://www.tux.org/lkml/