• Locked Locked
  • Replies 2 replies
  • Subscribers 119 subscribers
  • Views 2793 views
  • Users 0 members are here
Options
Related
How was your experience today?
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

GIC Acknowledge Register read has no impact

Sebastian Meisheit
Note: This was originally posted on 30th July 2012 at http://forums.arm.com

Hi,

Setup:
I am using Pandaboard ES with Cortex-A9MP (2x Cortex-A9) and connect to processor_0 by JTAG and Lauterbach hardware debugger.
From GICD_TYPER [10] = 1 i know the GIC implements Security Extensions.
From the Peripheral ID2 Register [7:4] = 1 i know its GICv1.
I am not able to enter Secure World (SMC or CPSR mode change doesnt work on the PandaES), so every access would be non-secure.

Problem:
According to the ICCIAR definition, an interrupt pending on a CPU interface should change from pending to active only after the register ICCIAR is read. In my GIC the interrupt changes immediately to active, if the priority is sufficient. I use the ICCIAR to get the current interrupt ID, but it has no efffect on the CPU interface or the GIC registers.
Is the ICCIAR definition wrong here, or am i missing something?

Greets
Sebastian
  • Note: This was originally posted on 31st July 2012 at http://forums.arm.com

    Thanks for your thoughts!
    I found out it was human error once again. I watch memory around 0x48240100 (CPU Interrupt Interface registers sit here) while testing. And that means the debugger reads all the memory and thus memory mapped IO registers at that addresses.
    So by viewing the memory region, the IAR was read by the debugger :)
  • Arm Employee Badge
    Note: This was originally posted on 31st July 2012 at http://forums.arm.com

    You're correct that interrupts go from Pending to Active by reading the Ack register (GICC_AIR).

    NOTE... In GICv2 (i.e. A15/A7 orGIC-400) you can force an interrupt to the Active state by writing to the GICD_ISACTIVER, but that doesn't apply here. :-)

    Couple of thoughts... the mostly likely to my mind is that one of the other cores is acknowledging the interrupt.  You said that you're connected to CPU0, what is CPU1 doing?

    Other thought... the Secure world code could be acknowledging the interrupt.  This seems less likely as usually the Secure world code only handles secure interrupts - which you Normal world code would never get to see.