On 4/20/07, Christian Boon <c.boon@xxxxxxxxxxxxxxxxxxxx> wrote:
pradeep singh wrote: > On 4/20/07, Christian Boon <c.boon@xxxxxxxxxxxxxxxxxxxx> wrote: >> Hello, >> >> i've got a question about local_irq_save() >> >> When i use this function to disable interrupts, does this mean that when >> an interrupt occurs >> when interrupts are disabled, for example from the ethernet chip, this >> interrupt is lost or does it mean >> that the interrupt isnt serviced at the moment but serviced when >> interrupts are enabled again? > the interrupt is lost, if it maskable and not unmaskable interrupts > IMHO.Because local_irq_save() macro inturn makes use of cli assembly > instruction. > Somebody please CMIIW. But isnt it bad that i loose for example an ethernet or timer interrupt?
But if you do allow that you may end up into a race condition, because this means you will have to use a lock to protect a critical section between interrupt handlers.Or may be you'll beed a recursive lock, which is bad. Moreover i don't think you usually miss that much interrupts as it seems.
>> >> i've got 2 interrupts in my driver which both need to write data in the >> same buffer so when i keep >> interrupts enabled, then i get bad behaviour. > Is the buffer too large.Try writing from different ends on the same > buffer for each interrupts.Does this helps anyway? I need to write on the same location, meaning the next buffer position.
Sad locking is not an option here. RCU mechanism is of any help to you.Please check, i dont have much idea about it. HTH ~psr
>> >> Is there a way to lock a buffer from two interrupt service routines >> without disabling interrupts? >> >> thanks in advance, >> >> Chris.
-- play the game -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
