ARM Cortex-A9 Intall Interrupt Handler bug?

Kernel
1

johnwilkes wrote on Monday, January 28, 2013:

The vPortInstallInterruptHandler() function in the ARM_Cortex-A9 port.c file has the following:

unsigned long ulBank16 = 4 * ( ulVector / 16 );

/* Is it Edge Triggered?. */
if ( 0 != ucEdgeTriggered )
{
portGIC_SET( (portGIC_ICDICR_BASE(puxGICDistributorAddress + ulBank16) ),
     ( portGIC_READ(puxGICDistributorAddress + ulBank16) | ( 0x02 << ( ulOffset16 * 2 ) ) ) );
}

I think it wants to set the edge-triggered bit in the appropriate Interrupt Interrupt Configuration Register, but it looks like it is OR’ing the edge-triggered bit with the contents of the Interrupt Controller Type Register at GCDistributorAddress+4 (e.g. for “vector” 29).

This makes no sense to me. Am I missing something? Why does it read (puxGICDistributorAddress + ulBank16) and OR the edge-triggered bit into that value?

Seems like it should be something like this:

    portGIC_SET( (portGIC_ICDICR_BASE(puxGICDistributorAddress + ulBank16) ), ( ( 0x02 << ( ulOffset16 * 2 ) ) ) );

-jw-

2

rtel wrote on Monday, January 28, 2013:

There is not (yet) an official Real Time Engineers ltd maintained Cortex-A9 port.  Did you get your code from the FreeRTOS Interactive site?  If so, please let me know which post the code came from and I may be able to ping the authour for a reply.

Regards.

3

johnwilkes wrote on Monday, January 28, 2013:

I found the Cortex-A9 posting and download on the FreeRTOS Interactive site:

http://interactive.freertos.org/entries/21035433

Thanks!

4

rtel wrote on Monday, January 28, 2013:

Did you check it against this version?
http://interactive.freertos.org/entries/21558347-arm-cortex-a9-versatileexpress-quad-core-port

Regards.

5

johnwilkes wrote on Monday, January 28, 2013:

I hadn’t seen that version, but it has the same edge-triggered code in vPortInstallInterruptHandler().

6

johnwilkes wrote on Tuesday, January 29, 2013:

I have convinced myself that this is a bug in the code, but I don’t think it matters for the Cortex-A9 MPCore, as ICDICR0 and ICDICR1 are read-only registers. Different implementations, e.g. single Cortex-A9 with a GIC, might have a problem.

-jw-

7

rtel wrote on Wednesday, January 30, 2013:

Thanks for taking the time to mention this.  I will let the authors know.

Regards.