| Subject: |
|
Re: New port questions |
| Name: |
|
Miro Samek |
| Date Posted: |
|
Apr 17, 08 - 5:17 PM |
| Email: |
|
miro@quantum-leaps.com |
| Website: |
|
http://www.quantum-leaps.com |
| Message: |
|
Mike,
1) In QK-nano v4, you don't need to change the QK priority upon the interrupt entry or exit. Instead, you inform the kernel about entering and exiting ISR by incrementing and decrementing the QK_intNest_ variable, which is less expensive in stack space.
2) If an ISR does not call any QP-nano services, then the ISR does not need to inform QK-nano about entry or exit. For such ISRs you don't need to call QK_ISR_ENTRY() and QK_ISR_EXIT(). (but be careful and place a big comment in such ISR, for the next person who will add a call to QF_postISR())
3) I'm not sure that your port is correct. In fact, it probably isn't because I'm concerned that upon the interrupt entry the interrupts can nest even before you disable interrupts and increment QK_intNest_. In that case, the preempting interrupt won't account for one level of nesting, and might call the scheduler. I will need to study the ColdFire documentation to understand exactly what happens upon the ISR entry to avoid such a race condition.
An official Quantum Leaps port to ColdFire is scheduled to be released relatively soon. I hope before the end of April. The first port will be for QP/C.
Best regars,
Miro |
|
Replies:
|
|
|
|
|
Reply
Forum
bravenet.com