Priority inversion and deadlock

1、 priority reverse

When resources are locked, low priority tasks preempt high priority tasks .
give an example ： Mission A、B、C, The priority from low to high is 1、2、3, All tasks are completely preemptive , Mission A and C Shared resources R.
1. Mission A function , And for resources R Lock ;
2. Mission C Be activated to preempt tasks A, And try to resource R Lock , therefore C Get into WAITING state ,A Recover from the occupied place ;
3. Mission B Be activated and preempt the task A, Mission B Successful execution ;
4. Mission A Resume execution and release resources R, Mission A Execution completed ;
5. Mission C Recover and restore resources R Lock and complete the execution .
We observe the task B Priority is lower than task C But actually occupied the task C, This is priority reversal . In systems with high real-time requirements, it is unacceptable for such high priority tasks to be unexpectedly delayed . More serious is due to priority reversal , Mission C The delay time of is uncertain , Because anything better than the task A Tasks with high priority can preempt tasks A.

2、 Deadlock

Locking resources causes a conflict between two tasks , Each task locks the resources needed by another task , Make every task impossible .
give an example ： Mission A、B, Share two resources R1、R2, Other situations are the same as those in priority inversion .
1. Mission A Yes R1 Lock ;
2. Mission B preemption A, Also on R2 Lock , Then try to R1 Lock , because R1 It's locked , So the task B Get into WAITING state ;
3. Mission A Resume execution , Try to be on R2 Lock , because R2 It's locked , So the task A Get into WAITING state ;
4. Other tasks enter the execution state , And the task A and B Can never continue .
Deadlock is likely not to be found in normal tests , After delivery to the customer, it is more difficult to be checked because the problem is difficult to recur .

There are many ways to solve priority inversion and deadlock , stay OSEK In the use of OS Priority ceiling protocol （priority ceiling protocol） It's the most effective way .

RT-thread  Mutually exclusive Priority inheritance mechanism in quantity

2018-11-01 13:23:58

The three threads are H Threads 、M Threads 、L Threads   
1. The priority order of the three threads is ：H Threads > M Threads > L Threads ;
2. During normal operation H Threads can break M Threads and L Threads , M Threads can break L Threads ;
3. Suppose a resource in the system is protected , At this time, the resource is L Threads are using , At some point H The thread needs to use this resource , but L The thread is not used up ,H The thread cannot apply and enters the blocking state , At this time, priority reversal has occurred ;

When there is no priority inheritance mechanism ：

If L When a thread is executing ,M The thread just woke up , because M Threads have a high priority , Will interrupt L Threads , preemption CPU Right to use , until M Thread execution complete ;
M Thread will CPU The right to use L Threads ,L The thread continues to execute ,L Release the resource after execution ;
H The thread gets the resource , Remove from the blocking state ;
This process , High priority H Waiting for the M Threads +L Thread time , If there are more threads , Then the system crashed , I can't afford to wait ！

When there is a priority inheritance mechanism ：
stay H When Xianchen applied for resources, he entered the blocking state because he could not apply for resources , The system will put the currently used resources L The priority of the thread is temporarily increased to H Threads have the same priority ;
here M The thread will not be interrupted when it is awakened L Threads ;
L Thread execution completed , Release resources ,H The thread obtains resources and continues to execute ,H The waiting time of the thread is only L Thread execution time .

original text ：
OSEK study notes – Priority inversion and deadlock