Re: [ECOS] ecos questions:kill safty

[Nick Garnett <nickg at calivar dot demon dot co dot uk>]

Please always CC messages to the ecos-discuss list. This way others
may benefit from the discussion, and the messages will also be
archived for future readers.

zhlg_shuhan <zhlg_shuhan@sina.com> writes:

> Thank you ,Sir! I have get it according your hint!
> i feel you are  the only engineer  i can communicate with,for not many people will think so more about RTOS!
>   another question:
>    Is it possible that sched_lock<0?
>    for example,a thread runnig with no lock(sched_lock=0 ),at that
> time,a interrupt occurs,so interrupt get cpu and runing till
> interrupt_end() in which unlock() was called,so shed_lock<0. right?

No. The scheduler lock is incremented by the HAL interrupt handling
code. The unlock in interrupt_end() just undoes this.

>     running  unlock_inner() and HAL_THREAD_SWITCH_CONTEXT( &current->stack_ptr,&next->stack_ptr )
>  Is it possible that interrupt occurs? because before entering HAL_THREAD_SWITCH_CONTEXT(),it seem that interrupt maybe keep enable.(except came from ISR)
>     unlock()
>        |
>    unlock_inner(){//sched_lock=1
>        
>      HAL_THREAD_SWITCH_CONTEXT<--interrupted by ISR(******)
>                                                  |
>                                        isr_end(){
>                                                      
>                                        unlock()<-attention!!
>                                            |
>                                   unlock_iner()//sched_lock=1
>  
>                                          loop once
>     }
>  }
>  i "assume" that context (note as *******) can be saved by cpu automatically,but maybe dangerous!
> 
>  right?
>                                                       zhlg 

Again no. You have missed the lock increment in the HAL.

-- 
Nick Garnett - eCos Kernel Architect


-- 
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss