[pxc] Problems with unexpected IRQ with three frame grabbers

[Luis Lloret]

Hi all.

    We are having some trouble when trying to handle three frame grabbers
simultaneously. When we execute three instances of pxc_live, one with
/dev/pxc0Hpgm, other with /dev/pxc1Hpgm and other with /dev/pxc2Hpgm, we get
this kind of errors (through dmesg):

pxc200: Unexpected IRQ 0xXXXXXXX, IP 0xXXXXXX X times
pxc200:     irq bit 02: hsync
pxc200:     irq bit 03: overflow
pxc200:     irq bit 12: fifo overrun, drop data

and

pxc200: lockup detected
pxc200: serious error (X times), restart
pxc200:     irq bit 03: overflow
pxc200:     irq bit 11: RISC irq
pxc200:     irq bit 18: opcode error


When we are using only two of the instances of pxc_live, no errors appear,
but as soon as we start the other... they do. We also see that some of the
lines of the image appear in black, as though they hadn´t been captured.
The same happens with our applications, when we use three or more grabbers.

If we execute in field mode instead of frame mode, then no errors appear.

I suspected of IRQ conflicts, but each PXC200AL has its own interrupt (5, 7,
10). Maybe we are beyond the PCI bandwidth or something like that?

We are using Linux 2.4.18 + RTAI 24.1.9, on a Pentium III with 512MB of RAM.
No RTAI modules loaded while testing this.

Any help or idea will be greatly appreciated (the sooner, the better).

Thanks a lot.


--------------------------------------------------------
Luis Lloret Portillo
VISAR CONTROL INDUSTRIAL, S.L.
Jefe Área Informática
Tfn: +34 918 041 013
Fax: +34 918 043 195
e-mail: luislloret@visarcontrol.es
web: http://www.visarcontrol.es
--------------------------------------------------------