o Rename kernload_ap to bp_kernelload. This to introduce a common prefix

    for variables that live in the boot page.
o   Add bp_trace (yes, it's in the boot page) that gets zeroed before we
    try to wake a core and to which the core being woken can write markers
    so that we know where the core was in case it doesn't wake up. The
    boot code does not yet write markers (too follow).
o   Disable the boot page translation to allow the last 4K page to be used
    for whatever we please. It would get mapped otherwise.
o   Fix kernstart in the case of SMP. The start argument is typically page
    aligned due to the alignment requirements that come with having a boot
    page. The point of using trunc_page is that we get the actual load
    address given that the entry point is immediately following the ELF
    headers. In the SMP case this ended up exactly 4K after the load
    address. Hence subtracting 1 from start.

3 files changed, 39 insertions, 19 deletions

diff --git a/sys/powerpc/booke/locore.S b/sys/powerpc/booke/locore.S
index 293eec3..0255677 100644
--- a/sys/powerpc/booke/locore.S
+++ b/sys/powerpc/booke/locore.S
@@ -242,14 +242,20 @@ done_mapping:
 __boot_page:
 	bl	1f
 
-	.globl	kernload_ap
-kernload_ap:
+	.globl	bp_trace
+bp_trace:
+	.long	0
+
+	.globl	bp_kernload
+bp_kernload:
 	.long	0
 
 /*
  * Initial configuration
  */
 1:
+	mflr    %r31		/* r31 hold the address of bp_trace */
+
 	/* Set HIDs */
 	lis	%r3, HID0_E500_DEFAULT_SET@h
 	ori	%r3, %r3, HID0_E500_DEFAULT_SET@l
@@ -318,15 +324,15 @@ kernload_ap:
 	mtspr	SPR_MAS2, %r3
 	isync
 
-	/* Retrieve kernel load [physical] address from kernload_ap */
+	/* Retrieve kernel load [physical] address from bp_kernload */
 	bl	4f
 4:	mflr	%r3
 	rlwinm	%r3, %r3, 0, 0, 19
-	lis	%r4, kernload_ap@h
-	ori	%r4, %r4, kernload_ap@l
+	lis	%r4, bp_kernload@h
+	ori	%r4, %r4, bp_kernload@l
 	lis	%r5, __boot_page@h
 	ori	%r5, %r5, __boot_page@l
-	sub	%r4, %r4, %r5	/* offset of kernload_ap within __boot_page */
+	sub	%r4, %r4, %r5	/* offset of bp_kernload within __boot_page */
 	lwzx	%r3, %r4, %r3
 
 	/* Set RPN and protection */
diff --git a/sys/powerpc/booke/platform_bare.c b/sys/powerpc/booke/platform_bare.c
index f04bf96..58fbb7f 100644
--- a/sys/powerpc/booke/platform_bare.c
+++ b/sys/powerpc/booke/platform_bare.c
@@ -56,7 +56,8 @@ __FBSDID("$FreeBSD$");
 #ifdef SMP
 extern void *ap_pcpu;
 extern uint8_t __boot_page[];		/* Boot page body */
-extern uint32_t kernload_ap;		/* Kernel physical load address */
+extern uint32_t bp_kernload;		/* Kernel physical load address */
+extern uint32_t bp_trace;		/* AP boot trace field */
 #endif
 
 extern uint32_t *bootinfo;
@@ -262,8 +263,8 @@ bare_smp_start_cpu(platform_t plat, struct pcpu *pc)
 
 	eebpcr = ccsr_read4(OCP85XX_EEBPCR);
 	if ((eebpcr & (1 << (pc->pc_cpuid + 24))) != 0) {
-		printf("%s: CPU=%d already out of hold-off state!\n",
-		    __func__, pc->pc_cpuid);
+		printf("SMP: CPU %d already out of hold-off state!\n",
+		    pc->pc_cpuid);
 		return (ENXIO);
 	}
 
@@ -273,12 +274,13 @@ bare_smp_start_cpu(platform_t plat, struct pcpu *pc)
 	/*
 	 * Set BPTR to the physical address of the boot page
 	 */
-	bptr = ((uint32_t)__boot_page - KERNBASE) + kernload_ap;
+	bptr = ((uint32_t)__boot_page - KERNBASE) + bp_kernload;
 	ccsr_write4(OCP85XX_BPTR, (bptr >> 12) | 0x80000000);
 
 	/*
 	 * Release AP from hold-off state
 	 */
+	bp_trace = 0;
 	eebpcr |= (1 << (pc->pc_cpuid + 24));
 	ccsr_write4(OCP85XX_EEBPCR, eebpcr);
 	__asm __volatile("isync; msync");
@@ -287,6 +289,16 @@ bare_smp_start_cpu(platform_t plat, struct pcpu *pc)
 	while (!pc->pc_awake && timeout--)
 		DELAY(1000);	/* wait 1ms */
 
+	/*
+	 * Disable boot page translation so that the 4K page at the default
+	 * address (= 0xfffff000) isn't permanently remapped and thus not
+	 * usable otherwise.
+	 */
+	ccsr_write4(OCP85XX_BPTR, 0);
+
+	if (!pc->pc_awake)
+		printf("SMP: CPU %d didn't wake up (trace code %#x).\n",
+		    pc->pc_awake, bp_trace);
 	return ((pc->pc_awake) ? 0 : EBUSY);
 #else
 	/* No SMP support */
diff --git a/sys/powerpc/booke/pmap.c b/sys/powerpc/booke/pmap.c
index 712967f..ca14f46 100644
--- a/sys/powerpc/booke/pmap.c
+++ b/sys/powerpc/booke/pmap.c
@@ -115,7 +115,7 @@ extern unsigned char _end[];
 extern uint32_t *bootinfo;
 
 #ifdef SMP
-extern uint32_t kernload_ap;
+extern uint32_t bp_kernload;
 #endif
 
 vm_paddr_t kernload;
@@ -967,10 +967,9 @@ mmu_booke_bootstrap(mmu_t mmu, vm_offset_t start, vm_offset_t kernelend)
 	debugf("mmu_booke_bootstrap: entered\n");
 
 #ifdef SMP
-	kernload_ap = kernload;
+	bp_kernload = kernload;
 #endif
 
-
 	/* Initialize invalidation mutex */
 	mtx_init(&tlbivax_mutex, "tlbivax", NULL, MTX_SPIN);
 
@@ -981,8 +980,13 @@ mmu_booke_bootstrap(mmu_t mmu, vm_offset_t start, vm_offset_t kernelend)
 	 * Align kernel start and end address (kernel image).
 	 * Note that kernel end does not necessarily relate to kernsize.
 	 * kernsize is the size of the kernel that is actually mapped.
+	 * Also note that "start - 1" is deliberate. With SMP, the
+	 * entry point is exactly a page from the actual load address.
+	 * As such, trunc_page() has no effect and we're off by a page.
+	 * Since we always have the ELF header between the load address
+	 * and the entry point, we can safely subtract 1 to compensate.
 	 */
-	kernstart = trunc_page(start);
+	kernstart = trunc_page(start - 1);
 	data_start = round_page(kernelend);
 	data_end = data_start;
 
@@ -1233,9 +1237,9 @@ mmu_booke_bootstrap(mmu_t mmu, vm_offset_t start, vm_offset_t kernelend)
 	 * entries, but for pte_vatopa() to work correctly with kernel area
 	 * addresses.
 	 */
-	for (va = KERNBASE; va < data_end; va += PAGE_SIZE) {
+	for (va = kernstart; va < data_end; va += PAGE_SIZE) {
 		pte = &(kernel_pmap->pm_pdir[PDIR_IDX(va)][PTBL_IDX(va)]);
-		pte->rpn = kernload + (va - KERNBASE);
+		pte->rpn = kernload + (va - kernstart);
 		pte->flags = PTE_M | PTE_SR | PTE_SW | PTE_SX | PTE_WIRED |
 		    PTE_VALID;
 	}
@@ -1397,9 +1401,7 @@ mmu_booke_kenter(mmu_t mmu, vm_offset_t va, vm_offset_t pa)
 	KASSERT(((va >= VM_MIN_KERNEL_ADDRESS) &&
 	    (va <= VM_MAX_KERNEL_ADDRESS)), ("mmu_booke_kenter: invalid va"));
 
-	flags = 0;
-	flags |= (PTE_SR | PTE_SW | PTE_SX | PTE_WIRED | PTE_VALID);
-	flags |= PTE_M;
+	flags = PTE_M | PTE_SR | PTE_SW | PTE_SX | PTE_WIRED | PTE_VALID;
 
 	pte = &(kernel_pmap->pm_pdir[pdir_idx][ptbl_idx]);