Hi Lucian,

IMHO better to have something like this

CANSendMessage() {
    // Write data to FIFO or similar
    if ( NOT CANTX IRQ_Enabled  ) {
           // Write data into at  least one of the CAN buffers
           // Enable CANTX IRQ
    }
}


Irq_routine {
     If ( data left to write ) {
           fill CAN buffer(s)
     }
     else {
         disable CANTXIRQ
     }
}

If you use a FIFO the IRQ routine will read from the read pointer and you
will write to the write pointer. No conflict between the two.

Cheers
/Ake


On 3/31/06, Lucian Copat <lucian.copat@audiotel.ro> wrote:
>
> Maybe I'm missing something obsious here...
>
> Let's say the filling function is called CANPut(), and it fills a CAN
> message, activates the corresponding interrupt and return True if there
> was space for the message. I have a function CANSendMessage which does:
>
> void CANSendMessage() {
>
>     //fill the 100-byte buffer
>
>     while (CANPut());
> }
>
> In the ISR I have:
>
> if (can TX interrupt) {
>     while (CANPut());
> }
>
> Until here am I correct ?
>
> The compiler will give: CANPut() is called from different.... -
> reentrant code.
>
> Lucian
>
> YAP wrote:
>
> >On 3/31/06, Lucian Copat <lucian.copat@audiotel.ro> wrote:
> >
> >
> >>I know that a message has at maximum 8 bytes in CAN, that's why I need
> >>to develop my own higher layer protocol to fragment the data. But when
> >>the interrupt is enabled, it doesn't activate if the buffer is empty,
> >>but when the buffer becomes empty (after having sent a CAN message).
> >>
> >>In this way, I can only avoid reentrance by writing the same code twice
> >>or by setting a flag in the interrupt.
> >>
> >>Lucian
> >>
> >>
> >
> >
> >I can't understand the problem here either if you use the method Gerhard
> >suggest.
> >
> >1.) When you have data to send fill all free buffers. Enable interrupt.
> >2.) When you get a frame sent interrupt fill data again if any to all
> free
> >buffers.
> >
> >No need to rewrite code use a function call to fill the buffers. It will
> >only be called by one process anyway and don't need to be reentrant.
> >
> >/Ake
> >
> >
> >
> >Gerhard Fiedler wrote:
> >
> >
> >>>Lucian Copat wrote:
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>>My question regards the E-CAN module in PIC18F4680. I have a routine
> >>>>which loads a message (100 bytes max) into a temporary sending buffer.
> >>>>
> >>>>
> >>>>
> >>>>
> >>>You know that a single CAN message can only have 8 bytes payload,
> right?
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>>I would like that after loading this buffer, when I activate the TX
> >>>>interrupts, the ISR to be called if there are some free buffers. E-CAN
> >>>>module does generate an interrupt only when a buffer _becomes_ empty,
> >>>>which means I should load it first. If I load it first and also in
> ISR,
> >>>>I would have reentrant code. Is there any method to pass this problem
> ?
> >>>>
> >>>>
> >>>>
> >>>>
> >>>You would typically disable the TX interrupt in the ISR when it sees
> that
> >>>there are no more bytes. Then, after initiating the next tx sequence,
> you
> >>>reenable it. This way, there's no problem with reentrance.
> >>>
> >>>Gerhard
> >>>
> >>>
> >>>
> >>>
> >>>
> >>--
> >>*Lucian Copat
> >>Software Team Leader
> >>Audiotel Comlabs
> >>E-mail: lucian.copat@audiotel.ro
> >>Mobile: (+4)0722.729.658
> >>Web: http://www.audiotel.ro*
> >>--
> >>http://www.piclist.com PIC/SX FAQ & list archive
> >>View/change your membership options at
> >>http://mailman.mit.edu/mailman/listinfo/piclist
> >>
> >>
> >>
>
> --
> *Lucian Copat
> Software Team Leader
> Audiotel Comlabs
> E-mail: lucian.copat@audiotel.ro
> Mobile: (+4)0722.729.658
> Web: http://www.audiotel.ro*
> --
> http://www.piclist.com PIC/SX FAQ & list archive
> View/change your membership options at
> http://mailman.mit.edu/mailman/listinfo/piclist
>
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist