Re: bad vme interrupt 0 or uninitialized vme interrupt

I don't know exactly how long it takes becuase I have a basecounter but
it can't be call from the ISR and I think that I don't have my BSP
compiled to use the others hardware counters.
My ISR worked very fast but I had interrupt problems (explained in last
emails). Now I have included an EIEIO and a CACHE_FLUSH after each bus
access and it seems to work fine, because I don't have any interrupt
error message. I don't know If it is working well because it is the
right solution or because these macro calls include a delay big enough
to make my system work. I say that because my prior temporal solution
was to include a delay (aprox 20 microsecs) at the ISR and everything
worked.


tony wrote:
Do you know how long to finish your code between intLock and
intUnlock? Make it short as soon as possible!
apazosm@xxxxxxxxx wrote:
Could be that the new macro functions that I am including (EIEIO, CACHE
FLUSH), introduce a delay at the ISR that makes my system work?


apazosm@xxxxxxxxx wrote:
I tried the solution that you propose me (CACHE_PIPE_FLUSH) together
with an EIEIO. It had some effect because the program was working
well... but after one night running it crashed and the LCU is blocked
so I have to reboot. I saw the same effect when I included a delay at
the ISR.
I know that we are on the right direction but it is not still
completely solved.

Thank you for your help.
Andrés M. Pazos

cxz@xxxxxxxxxxxxx wrote:
yes, it may also be a macro like,
CACHE_PIPE_FLUSH();

wish it will be helpful

apazosm@xxxxxxxxx wrote:
Thank you for your answer, but how can I reflash my cache? is something
related whith isync?

Thank you again,
Greetings
Andrés M. Pazos

cxz@xxxxxxxxxxxxx wrote:
It may alse be the problem of the faster CPU speed than your bus, which
i meet last week.
You can just reflash your cache and put the cached data into RAM.

Wish it will be helpful.
Good Luck!

apazosm@xxxxxxxxx wrote:
I am trying to solve a problem with the handle of vme interrupts at
high speed, 1khz or more, when I am latching the position of IK320 vme
cards.

My system:
I have several VME encoder cards (heidenhain-IK320) that are latched at
1 kHz by a CPU Motorola MVME5100-2263 MPC-7410 under VxWorks 5.5.1
kernel 2.6.
They are configured at the IRQ 3 and with the interrupt vectors: C1, C2
and C3. (I also tried different interrupt configurations and happen the
same)

The problem:
When I start latching the position I receive an unintialized vme
interrupt at the group address of the cards, in this case at the C0
vector, and the cards do not answer any more.
If I include in my code the commands "eieio" and "sync" I also receive
a bad vme interrupt 0.

Detailed description:
The cards have a flag "TRANSFER MARKER" that must be set to 0 to
indicate the card that is ready to write again.
I also need to set the INTERRUPT STATUS to 0 to reset the interrupts
and be ready for the next one.
I have located that the problem is at the ISR.
1) If I reset the interrupt status at the end of the ISR the system
fails after 45 minutes aprox.
2) If I reset the interrupt status at the begining of the ISR the
system fails immediately
3) I have a partial solution that is to include a delay in the ISR so
the system seems to work well. Depending of the delay it crashes after
hours.

I also tried to include in my code the orders EIEIO, SYNC and read the
interrupt status but I didn't solve the problem

This is the code that I am executing:

void IK320Driver::interruptService_()
{
unsigned char lowByte;
unsigned char highByte;

//Lock the interrupts
short key = intLock();

highByte = (unsigned char) ((ik320_->interruptStatus & 0xFF00) >> 8);
lowByte = (unsigned char) ( ik320_->interruptStatus & 0x00FF );

//Clear interrupt Status
ik320_->interruptStatus = 0; // Clear interrupt status

// Check interrupt Status
switch (highByte)
{
//read the data
//reset the transfer marker
}

/Unlock the interrupts
intUnlock(key);

}

I really appreciate any comment.

Andrés M. Pazos
apazosm@xxxxxxxxx

.

Relevant Pages

  • Re: bad vme interrupt 0 or uninitialized vme interrupt
    ... introduce a delay at the ISR that makes my system work? ... I have several VME encoder cards that are latched at ... They are configured at the IRQ 3 and with the interrupt vectors: ...
    (comp.os.vxworks)
  • Re: Interrupts are not coming sometime in NIC miniport
    ... An ISR should do only what is absolutely necessary, ... This is usually done by reading some interrupt register on the ... The DPC then does all the real work. ... Fill up blob with data from the NICs data buffers. ...
    (microsoft.public.development.device.drivers)
  • Re: Installable ISR - general questions
    ... The ISR (Interrupt Service Routine) is called by the interrupt ... handler installed by the OS. ... SYSINTR_RESCHEDULE tells the kernel that it should check if a thread ...
    (microsoft.public.windowsce.platbuilder)
  • Re: bad vme interrupt 0 or uninitialized vme interrupt
    ... introduce a delay at the ISR that makes my system work? ... I have several VME encoder cards that are latched at ... They are configured at the IRQ 3 and with the interrupt vectors: ...
    (comp.os.vxworks)
  • Re: softirqd
    ... > 2) ksoftirqd is not a real time thread ... 'Real-time' related to an 'interrupt' has no meaning. ... your ISR code gets control, the time is bounded by some ...
    (Linux-Kernel)