On multi-core, multi-threaded PPC systems, it is important that the threads

be brought up in the order they are enumerated in the device tree (in
particular, that thread 0 on each core be brought up first). The SLIST
through which we loop to start the CPUs has all of its entries added with
SLIST_INSERT_HEAD(), which means it is in reverse order of enumeration
and so AP startup would always fail in such situations (causing a machine
check or RTAS failure). Fix this by changing the SLIST into an STAILQ,
and inserting new CPUs at the end.

Reviewed by:	jhb

15 files changed, 25 insertions, 25 deletions

diff --git a/sys/i386/pci/pci_cfgreg.c b/sys/i386/pci/pci_cfgreg.c
index ae56990..ac641a8 100644
--- a/sys/i386/pci/pci_cfgreg.c
+++ b/sys/i386/pci/pci_cfgreg.c
@@ -553,7 +553,7 @@ pcie_cfgregopen(uint64_t base, uint8_t minbus, uint8_t maxbus)
 		    (uintmax_t)base);
 
 #ifdef SMP
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu)
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu)
 #endif
 	{
 
diff --git a/sys/ia64/ia64/machdep.c b/sys/ia64/ia64/machdep.c
index 41d2211..7252865 100644
--- a/sys/ia64/ia64/machdep.c
+++ b/sys/ia64/ia64/machdep.c
@@ -316,7 +316,7 @@ cpu_startup(void *dummy)
 	/*
 	 * Create sysctl tree for per-CPU information.
 	 */
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		snprintf(nodename, sizeof(nodename), "%u", pc->pc_cpuid);
 		sysctl_ctx_init(&pc->pc_md.sysctl_ctx);
 		pc->pc_md.sysctl_tree = SYSCTL_ADD_NODE(&pc->pc_md.sysctl_ctx,
diff --git a/sys/ia64/ia64/mp_machdep.c b/sys/ia64/ia64/mp_machdep.c
index 5804f8c..b6b0bef 100644
--- a/sys/ia64/ia64/mp_machdep.c
+++ b/sys/ia64/ia64/mp_machdep.c
@@ -357,7 +357,7 @@ cpu_mp_start()
 	/* Keep 'em spinning until we unleash them... */
 	ia64_ap_state.as_spin = 1;
 
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		pc->pc_md.current_pmap = kernel_pmap;
 		pc->pc_other_cpus = all_cpus & ~pc->pc_cpumask;
 		/* The BSP is obviously running already. */
@@ -424,7 +424,7 @@ cpu_mp_unleash(void *dummy)
 
 	cpus = 0;
 	smp_cpus = 0;
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		cpus++;
 		if (pc->pc_md.awake) {
 			kproc_create(ia64_store_mca_state, pc, NULL, 0, 0,
@@ -462,7 +462,7 @@ ipi_selected(cpumask_t cpus, int ipi)
 {
 	struct pcpu *pc;
 
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		if (cpus & pc->pc_cpumask)
 			ipi_send(pc, ipi);
 	}
@@ -486,7 +486,7 @@ ipi_all_but_self(int ipi)
 {
 	struct pcpu *pc;
 
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		if (pc != pcpup)
 			ipi_send(pc, ipi);
 	}
diff --git a/sys/ia64/ia64/pmap.c b/sys/ia64/ia64/pmap.c
index 5f10ad6..411d53a 100644
--- a/sys/ia64/ia64/pmap.c
+++ b/sys/ia64/ia64/pmap.c
@@ -535,7 +535,7 @@ pmap_invalidate_page(vm_offset_t va)
 	critical_enter();
 	vhpt_ofs = ia64_thash(va) - PCPU_GET(md.vhpt);
 	tag = ia64_ttag(va);
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		pte = (struct ia64_lpte *)(pc->pc_md.vhpt + vhpt_ofs);
 		atomic_cmpset_64(&pte->tag, tag, 1UL << 63);
 	}
diff --git a/sys/kern/kern_idle.c b/sys/kern/kern_idle.c
index af12d7d..f412d17 100644
--- a/sys/kern/kern_idle.c
+++ b/sys/kern/kern_idle.c
@@ -60,7 +60,7 @@ idle_setup(void *dummy)
 
 	p = NULL; /* start with no idle process */
 #ifdef SMP
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 #endif
 #ifdef SMP
 		error = kproc_kthread_add(sched_idletd, NULL, &p, &td,
diff --git a/sys/kern/sched_4bsd.c b/sys/kern/sched_4bsd.c
index fef9e25..519cae5 100644
--- a/sys/kern/sched_4bsd.c
+++ b/sys/kern/sched_4bsd.c
@@ -1081,7 +1081,7 @@ forward_wakeup(int cpunum)
 	dontuse = me | stopped_cpus | hlt_cpus_mask;
 	map2 = 0;
 	if (forward_wakeup_use_loop) {
-		SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+		STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 			id = pc->pc_cpumask;
 			if ((id & dontuse) == 0 &&
 			    pc->pc_curthread == pc->pc_idlethread) {
@@ -1112,7 +1112,7 @@ forward_wakeup(int cpunum)
 	}
 	if (map) {
 		forward_wakeups_delivered++;
-		SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+		STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 			id = pc->pc_cpumask;
 			if ((map & id) == 0)
 				continue;
diff --git a/sys/kern/subr_kdb.c b/sys/kern/subr_kdb.c
index 342c5ca..5d68ae2 100644
--- a/sys/kern/subr_kdb.c
+++ b/sys/kern/subr_kdb.c
@@ -412,7 +412,7 @@ kdb_thr_ctx(struct thread *thr)
 		return (&kdb_pcb);
 
 #if defined(SMP) && defined(KDB_STOPPEDPCB)
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu)  {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu)  {
 		if (pc->pc_curthread == thr && (stopped_cpus & pc->pc_cpumask))
 			return (KDB_STOPPEDPCB(pc));
 	}
diff --git a/sys/kern/subr_pcpu.c b/sys/kern/subr_pcpu.c
index de5cafc..5cb4f26 100644
--- a/sys/kern/subr_pcpu.c
+++ b/sys/kern/subr_pcpu.c
@@ -74,7 +74,7 @@ static TAILQ_HEAD(, dpcpu_free) dpcpu_head = TAILQ_HEAD_INITIALIZER(dpcpu_head);
 static struct sx dpcpu_lock;
 uintptr_t dpcpu_off[MAXCPU];
 struct pcpu *cpuid_to_pcpu[MAXCPU];
-struct cpuhead cpuhead = SLIST_HEAD_INITIALIZER(cpuhead);
+struct cpuhead cpuhead = STAILQ_HEAD_INITIALIZER(cpuhead);
 
 /*
  * Initialize the MI portions of a struct pcpu.
@@ -89,7 +89,7 @@ pcpu_init(struct pcpu *pcpu, int cpuid, size_t size)
 	pcpu->pc_cpuid = cpuid;
 	pcpu->pc_cpumask = 1 << cpuid;
 	cpuid_to_pcpu[cpuid] = pcpu;
-	SLIST_INSERT_HEAD(&cpuhead, pcpu, pc_allcpu);
+	STAILQ_INSERT_TAIL(&cpuhead, pcpu, pc_allcpu);
 	cpu_pcpu_init(pcpu, cpuid, size);
 	pcpu->pc_rm_queue.rmq_next = &pcpu->pc_rm_queue;
 	pcpu->pc_rm_queue.rmq_prev = &pcpu->pc_rm_queue;
@@ -245,7 +245,7 @@ void
 pcpu_destroy(struct pcpu *pcpu)
 {
 
-	SLIST_REMOVE(&cpuhead, pcpu, pcpu, pc_allcpu);
+	STAILQ_REMOVE(&cpuhead, pcpu, pcpu, pc_allcpu);
 	cpuid_to_pcpu[pcpu->pc_cpuid] = NULL;
 	dpcpu_off[pcpu->pc_cpuid] = 0;
 }
diff --git a/sys/mips/mips/mp_machdep.c b/sys/mips/mips/mp_machdep.c
index e945736..7191b37 100644
--- a/sys/mips/mips/mp_machdep.c
+++ b/sys/mips/mips/mp_machdep.c
@@ -86,7 +86,7 @@ ipi_selected(cpumask_t cpus, int ipi)
 
 	CTR3(KTR_SMP, "%s: cpus: %x, ipi: %x\n", __func__, cpus, ipi);
 
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		if ((cpus & pc->pc_cpumask) != 0)
 			ipi_send(pc, ipi);
 	}
diff --git a/sys/net/netisr.c b/sys/net/netisr.c
index 67ec160..127cf67 100644
--- a/sys/net/netisr.c
+++ b/sys/net/netisr.c
@@ -1221,7 +1221,7 @@ netisr_start(void *arg)
 {
 	struct pcpu *pc;
 
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		if (nws_count >= netisr_maxthreads)
 			break;
 		/* XXXRW: Is skipping absent CPUs still required here? */
diff --git a/sys/powerpc/booke/pmap.c b/sys/powerpc/booke/pmap.c
index aff7901..cabe58f 100644
--- a/sys/powerpc/booke/pmap.c
+++ b/sys/powerpc/booke/pmap.c
@@ -390,7 +390,7 @@ tlb_miss_lock(void)
 	if (!smp_started)
 		return;
 
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		if (pc != pcpup) {
 
 			CTR3(KTR_PMAP, "%s: tlb miss LOCK of CPU=%d, "
@@ -416,7 +416,7 @@ tlb_miss_unlock(void)
 	if (!smp_started)
 		return;
 
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		if (pc != pcpup) {
 			CTR2(KTR_PMAP, "%s: tlb miss UNLOCK of CPU=%d",
 			    __func__, pc->pc_cpuid);
diff --git a/sys/powerpc/powerpc/mp_machdep.c b/sys/powerpc/powerpc/mp_machdep.c
index 02920da..577d4dc 100644
--- a/sys/powerpc/powerpc/mp_machdep.c
+++ b/sys/powerpc/powerpc/mp_machdep.c
@@ -212,7 +212,7 @@ cpu_mp_unleash(void *dummy)
 
 	cpus = 0;
 	smp_cpus = 0;
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		cpus++;
 		pc->pc_other_cpus = all_cpus & ~pc->pc_cpumask;
 		if (!pc->pc_bsp) {
@@ -347,7 +347,7 @@ ipi_selected(cpumask_t cpus, int ipi)
 {
 	struct pcpu *pc;
 
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		if (cpus & pc->pc_cpumask)
 			ipi_send(pc, ipi);
 	}
@@ -367,7 +367,7 @@ ipi_all_but_self(int ipi)
 {
 	struct pcpu *pc;
 
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		if (pc != pcpup)
 			ipi_send(pc, ipi);
 	}
diff --git a/sys/sparc64/sparc64/mp_machdep.c b/sys/sparc64/sparc64/mp_machdep.c
index 8700f89..4d9151e 100644
--- a/sys/sparc64/sparc64/mp_machdep.c
+++ b/sys/sparc64/sparc64/mp_machdep.c
@@ -383,7 +383,7 @@ cpu_mp_unleash(void *v)
 	ctx_inc = (TLB_CTX_USER_MAX - 1) / mp_ncpus;
 	csa = &cpu_start_args;
 	csa->csa_count = mp_ncpus;
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
 		pc->pc_tlb_ctx = ctx_min;
 		pc->pc_tlb_ctx_min = ctx_min;
 		pc->pc_tlb_ctx_max = ctx_min + ctx_inc;
diff --git a/sys/sparc64/sparc64/pmap.c b/sys/sparc64/sparc64/pmap.c
index 09482b9..c34fc45 100644
--- a/sys/sparc64/sparc64/pmap.c
+++ b/sys/sparc64/sparc64/pmap.c
@@ -1278,7 +1278,7 @@ pmap_release(pmap_t pm)
 	 *   to a kernel thread, leaving the pmap pointer unchanged.
 	 */
 	mtx_lock_spin(&sched_lock);
-	SLIST_FOREACH(pc, &cpuhead, pc_allcpu)
+	STAILQ_FOREACH(pc, &cpuhead, pc_allcpu)
 		if (pc->pc_pmap == pm)
 			pc->pc_pmap = NULL;
 	mtx_unlock_spin(&sched_lock);
diff --git a/sys/sys/pcpu.h b/sys/sys/pcpu.h
index ad1cf33..0bb2cbd 100644
--- a/sys/sys/pcpu.h
+++ b/sys/sys/pcpu.h
@@ -164,7 +164,7 @@ struct pcpu {
 	u_int		pc_cpuid;		/* This cpu number */
 	cpumask_t	pc_cpumask;		/* This cpu mask */
 	cpumask_t	pc_other_cpus;		/* Mask of all other cpus */
-	SLIST_ENTRY(pcpu) pc_allcpu;
+	STAILQ_ENTRY(pcpu) pc_allcpu;
 	struct lock_list_entry *pc_spinlocks;
 #ifdef KTR
 	char		pc_name[PCPU_NAME_LEN];	/* String name for KTR */
@@ -201,7 +201,7 @@ struct pcpu {
 
 #ifdef _KERNEL
 
-SLIST_HEAD(cpuhead, pcpu);
+STAILQ_HEAD(cpuhead, pcpu);
 
 extern struct cpuhead cpuhead;
 extern struct pcpu *cpuid_to_pcpu[MAXCPU];