On Fri, May 29, 2015 at 11:08:43AM +0800, Chen-Yu Tsai wrote:
+.globl psci_fiq_enter +psci_fiq_enter:
push {r0-r12}@ Switch to securemrc p15, 0, r7, c1, c1, 0bic r8, r7, #1mcr p15, 0, r8, c1, c1, 0isb@ Validate reason based on IAR and acknowledgemovw r8, #(GICC_BASE & 0xffff)movt r8, #(GICC_BASE >> 16)ldr r9, [r8, #GICC_IAR]movw r10, #0x3ffmovt r10, #0cmp r9, r10 @ skip spurious interrupt 1023beq outmovw r10, #0x3fe @ ...and 1022Maybe we could add some defines for these spurious interrupts values ?
From the GIC spec:
The read returns a spurious interrupt ID of 1023 if any of the following apply:
- forwarding of interrupts by the Distributor to the CPU interface is disabled
- signaling of interrupts by the CPU interface to the connected processor is disabled
- no pending interrupt on the CPU interface has sufficient priority for the interface to signal it to the processor.
Likewise, 1022 means group 0 secure interrupts aren't forwarded to the processor.
So my understanding is that 1022/1023 just means there's nothing for the processor to process. Maybe a note referring to the spec would be enough?
Sorry, that's not what I meant.
What I meant was to do
#define GIC_SPURIOUS_INTERRUPT 1023
movw r10, #GIC_SPURIOUS_INTERRUPT.
+.globl psci_arch_init +psci_arch_init:
mov r6, lrmovw r4, #(GICD_BASE & 0xffff)movt r4, #(GICD_BASE >> 16)ldr r5, [r4, #GICD_IGROUPRn]bic r5, r5, #(1 << 15) @ SGI15 as Group-0str r5, [r4, #GICD_IGROUPRn]mov r5, #0 @ Set SGI15 priority to 0strb r5, [r4, #(GICD_IPRIORITYRn + 15)]add r4, r4, #0x1000 @ GICC addressmov r5, #0xffstr r5, [r4, #GICC_PMR] @ Be cool with non-secureldr r5, [r4, #GICC_CTLR]orr r5, r5, #(1 << 3) @ Switch FIQEn onstr r5, [r4, #GICC_CTLR]mrc p15, 0, r5, c1, c1, 0 @ Read SCRorr r5, r5, #4 @ Enable FIQ in monitor modebic r5, r5, #1 @ Secure modemcr p15, 0, r5, c1, c1, 0 @ Write SCRisbbl psci_get_cpu_id @ CPU ID => r0bl psci_get_cpu_stack_top @ stack top => r0mov sp, r0bx r6.globl psci_text_endIsn't it exactly the same function than the A20's? Maybe that can be shared?
Good idea. This might even be shared by some other platforms. "psci_text_end" should always be linked in last though.
Maybe that can be generated directly by the linker script at the end of the PSCI text section?
Maxime