Re: [lwip-users] mutex API in system layer is necessary ?

[Simon Goldschmidt]

Mason <address@hidden> wrote:

> AFAIU, mutexes do not make a system immune to priority inversion.
> 
> Consider 3 processes of increasing priority L, M, H.
> 
> Suppose L locks a mutex; then H waits for the mutex; then M starts
> running and preempts L : H will never run, as long as M runs.

That's exactly the case we want to be able to fix by using mutexes instead of 
binary semaphores.

> Unless you were thinking of specific implementations of mutexes
> (priority mutexes) which temporarily boost the priority of tasks
> that lock that specific mutex?

Of course, if your mutex implementation doesn't support this, you can use 
binary semaphores as well. However, in all embedded ("realtime") OSes I've 
worked with, L gets the priority of H while L owns the mutex that H waits for.

Simon