Paul Durrant wrote:
On 29/05/07, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
Paul Durrant wrote:
> On 28/05/07, Garrett D'Amore <[EMAIL PROTECTED]> wrote:
>> You should not attempt to acquire a lock which will be held by other
>> functions calling cv_wait, or its brethren. (There are ways to do
that
>> safely, but it requires a fair bit of effort to make sure you do it
>> safely.)
>>
>
> I was not aware of this restriction, nor can I find it documented.
> This suggests that one cannot call cv_signal() from within an
> interrupt handler. Is this true?
>
> Paul
>
Check the man page for cv_wait(9f) under CONTEXT.
Yes, I am aware of the restriction for cv_wait(). That makes sense.
As for using cv_signal
in an interrupt handler, no problem. Initially, at least, this was the
most common place for
cv_signal to be called.
But to call cv_signal() one must be holding the mutex that was
released by the call to cv_wait() in the other thread; thereby doing
something that Garrett suggests is unsafe.
Paul
I assume you mean the bit quoted from Garrett at the top of this email.
I think he was specifically talking about
timeout functions, and about other mutexes (not the one used within the
cv_wait call). You want to avoid this scenario:
Thread
A Thread B
(Interrupt context (or not))
mutex_enter(&l1);
mutex_enter(&l1);
/* problem is you don't get past here assuming */
/* Thread A got mutex l1 */
mutex_enter(&l2);
mutex_enter(&l2);
while (condition ! met) /*
condition has been met */
cv_wait(&cv, &l2); /* should release l1 first*/ cv_signal(&cv);
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
