anonymous wrote on Tuesday, November 07, 2006:
(I thought I posted this earlier, but it didn’t appear so I am trying again)
Richard, et al
I have recently been addressing spurious interrupt
issues on an LPC2000 processor. There are a variety
of issues - related to the pipeline, the VIC, and some
peripherals.
The issue I bring to this forum is one in the ARM core
itself, due to the pipeline architecture. It involves
a scenario in which an IRQ can happen with IRQ/FIQ
interrupts disabled. This is discussed fairly well in
the Philips User Manuals - I used UM10114_1.pdf.
My concern is with the point that if such an interrupt
is processed without taking any special action, then FIQs
remain disabled during the whole IRQ processing. There
are several possible fixes, but for various reasons I
like the one in which interrupts are disabled by first
disabling IRQs, then disabling FIQs.
I have made this change in my version of FreeRTOS, but
I suggest it should be in the distributed code. This would
involve changing everywhere:
MRS R0, CPSR
ORR R0, R0, #0xC0
MSR CPSR, R0
to
MRS R0, CPSR
ORR R0, R0, #0x80
MSR CPSR, R0
ORR R0, R0, #0x40
MSR CPSR, R0
Glen