If T1 acquires te spin lock, its IRQL is of no interest in te discussion.
T2, independent of its IRQL, will have to wait for T1 to release the lock.
Similarly, if T2 acquires the lock, then T1 must wait, independent of its
IRQL. T1 cannot “preempt” T2 while T2 holds the spin lock, because while
it is holding the spin lock, T2 is at DISPATCH_LEVEL, and therefore cannot
be scheduled (or descheduled). Now consider a muticore machine, where T1
is running one one core and T2 on another. T1, at DISPATCH_LEVEL, doesn’t
care about T2, and therefore the concept of “preemption” is irrelevant.
T1 can acquire the spin lock, which will prevent T2 from acquiring it, but
that does not constitute “preemption”. Similarly, if T2 acquires te spin
lock, T1 has to wait for it, and from the scheduler viewpoint, this cannot
constitute “preemption”. Note that as long as T2 holds the lock, it
cannot be touched by the scheduler, because it’s at DISPATCH_LEVEL There
is an issue where a *logically* lower-priority thread blocks a *logically*
higher-priority thread. If the lower-priority thread can be descheduled,
then the higher-priority thread is blocked, and in a pure
highest-priority-thread-wins model, you now have something known as
“priority inversion”. But this is not possible in Windows in the scenario
you present, because the “lower-priority” thread is made non-preemptible
by being raised to DISPATCH_LEVEL during the scope of the lock.
If you discover that the order of acquisition means your code doesn’t
work, it means your code is wrong.
joe
Potential problem in volume filter driver.I have two different threads
running at different IRQl’s.
But the different IRQLs are competing for the same resource.
For Ex: Thread on T1 is running at Dispatch level.
Thread two T2 is running at Passive level.
Both the thread are accessing the same resource upto some time but
suddenly
T1 is preempting T2.also for very less time T2 is Preempting T1.
Due to that the sequence of operation getting failed and control is coming
out of the loop abruptly.???
Is this a Deadlock kind of situation what can i do in this case???
How do i manage the IRQL levels for accessing the same resource.
Thanks.
NTDEV is sponsored by OSR
Visit the list at: http://www.osronline.com/showlists.cfm?list=ntdev
OSR is HIRING!! See http://www.osr.com/careers
For our schedule of WDF, WDM, debugging and other seminars visit:
http://www.osr.com/seminars
To unsubscribe, visit the List Server section of OSR Online at
http://www.osronline.com/page.cfm?name=ListServer
