On Thu, Oct 31, 2002 at 06:48:05AM -0800, kernel_learner wrote: > Rule of thumb: > > Nothing is going to stop u from doing anything in the > ISR. There are no 'programming' restrictions as > such..but to acheive the right goals..: Oh yes. There is something. There are things, that will inevitably lead to a crash sooner or later. That is especially everything, that can eventually call down to schedule(). Simply because the ISR is running with interrupts disabled and also because current is not set to anything sane. Bottom-halves have interrupts enabled (that's why they are used after all), so you can do longer operations there, but you still can't sleep as current is still not anything sane. > Don't sleep_on or use an API that can cause unbounded > delays (unpredictably).. You want to do as much as > possible in the ISR and also not block for too long.. That's about right. But the true reason is really schedule() and anything else that uses current. You also have to be careful with spinlocks (let alone semaphores - they call schedule()), because when the interrupt occurs while a lock is locked and wants to lock it, it will spin forever - deadlock. So the interrupt must not happen - that's what spin_lock_irqsave is for. ------------------------------------------------------------------------------- Jan 'Bulb' Hudec <bulb@ucw.cz> -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
