- Do not handle the per-CPU containers in mbuf code as though the cpuids

  were indices in a dense array. The cpuids are a sparse set and treat
  them as such, setting up containers only for CPUs activated during
  mb_init().

- Fix netstat(1) and systat(1) to treat the per-CPU stats area as a sparse
  map, in accordance with the above.

This allows us to properly boot with certain CPUs disactivated. However, if
we later decide to re-activate said CPUs, we will barf until we decide to
implement CPU spinon/spinoff callback hooks to allow for said CPUs' per-CPU
containers to get configured on their activation.

Reported by: mjacob
Partially (sys/ diffs) Submitted by: mjacob

11 files changed, 68 insertions, 29 deletions

diff --git a/sys/alpha/alpha/mp_machdep.c b/sys/alpha/alpha/mp_machdep.c
index 74500b8..e042183 100644
--- a/sys/alpha/alpha/mp_machdep.c
+++ b/sys/alpha/alpha/mp_machdep.c
@@ -301,6 +301,11 @@ cpu_mp_probe(void)
 
 	/* XXX: Need to check for valid platforms here. */
 
+	boot_cpu_id = PCPU_GET(cpuid);
+	KASSERT(boot_cpu_id == hwrpb->rpb_primary_cpu_id,
+	    ("cpu_mp_probe() called on non-primary CPU"));
+	all_cpus = 1 << boot_cpu_id;
+
 	mp_ncpus = 1;
 
 	/* Make sure we have at least one secondary CPU. */
@@ -324,16 +329,12 @@ cpu_mp_probe(void)
 }
 
 void
-cpu_mp_start()
+cpu_mp_start(void)
 {
 	int i;
 
 	mtx_init(&ap_boot_mtx, "ap boot", MTX_SPIN);
 
-	boot_cpu_id = PCPU_GET(cpuid);
-	KASSERT(boot_cpu_id == hwrpb->rpb_primary_cpu_id,
-	    ("mp_start() called on non-primary CPU"));
-	all_cpus = 1 << boot_cpu_id;
 	for (i = 0; i < hwrpb->rpb_pcs_cnt; i++) {
 		struct pcs *pcsp;
 
diff --git a/sys/amd64/amd64/mp_machdep.c b/sys/amd64/amd64/mp_machdep.c
index d5af7b3..eaf508c 100644
--- a/sys/amd64/amd64/mp_machdep.c
+++ b/sys/amd64/amd64/mp_machdep.c
@@ -406,6 +406,12 @@ found:
 int
 cpu_mp_probe(void)
 {
+	/*
+	 * Record BSP in CPU map
+	 * This is done here so that MBUF init code works correctly.
+	 */
+	all_cpus = 1;
+
 	return (mp_capable);
 }
 
@@ -1929,9 +1935,6 @@ start_all_aps(u_int boot_addr)
 	mpbiosreason = inb(CMOS_DATA);
 #endif
 
-	/* record BSP in CPU map */
-	all_cpus = 1;
-
 	/* set up temporary P==V mapping for AP boot */
 	/* XXX this is a hack, we should boot the AP on its own stack/PTD */
 	kptbase = (uintptr_t)(void *)KPTphys;
diff --git a/sys/amd64/amd64/mptable.c b/sys/amd64/amd64/mptable.c
index d5af7b3..eaf508c 100644
--- a/sys/amd64/amd64/mptable.c
+++ b/sys/amd64/amd64/mptable.c
@@ -406,6 +406,12 @@ found:
 int
 cpu_mp_probe(void)
 {
+	/*
+	 * Record BSP in CPU map
+	 * This is done here so that MBUF init code works correctly.
+	 */
+	all_cpus = 1;
+
 	return (mp_capable);
 }
 
@@ -1929,9 +1935,6 @@ start_all_aps(u_int boot_addr)
 	mpbiosreason = inb(CMOS_DATA);
 #endif
 
-	/* record BSP in CPU map */
-	all_cpus = 1;
-
 	/* set up temporary P==V mapping for AP boot */
 	/* XXX this is a hack, we should boot the AP on its own stack/PTD */
 	kptbase = (uintptr_t)(void *)KPTphys;
diff --git a/sys/amd64/include/mptable.h b/sys/amd64/include/mptable.h
index d5af7b3..eaf508c 100644
--- a/sys/amd64/include/mptable.h
+++ b/sys/amd64/include/mptable.h
@@ -406,6 +406,12 @@ found:
 int
 cpu_mp_probe(void)
 {
+	/*
+	 * Record BSP in CPU map
+	 * This is done here so that MBUF init code works correctly.
+	 */
+	all_cpus = 1;
+
 	return (mp_capable);
 }
 
@@ -1929,9 +1935,6 @@ start_all_aps(u_int boot_addr)
 	mpbiosreason = inb(CMOS_DATA);
 #endif
 
-	/* record BSP in CPU map */
-	all_cpus = 1;
-
 	/* set up temporary P==V mapping for AP boot */
 	/* XXX this is a hack, we should boot the AP on its own stack/PTD */
 	kptbase = (uintptr_t)(void *)KPTphys;
diff --git a/sys/i386/i386/mp_machdep.c b/sys/i386/i386/mp_machdep.c
index d5af7b3..eaf508c 100644
--- a/sys/i386/i386/mp_machdep.c
+++ b/sys/i386/i386/mp_machdep.c
@@ -406,6 +406,12 @@ found:
 int
 cpu_mp_probe(void)
 {
+	/*
+	 * Record BSP in CPU map
+	 * This is done here so that MBUF init code works correctly.
+	 */
+	all_cpus = 1;
+
 	return (mp_capable);
 }
 
@@ -1929,9 +1935,6 @@ start_all_aps(u_int boot_addr)
 	mpbiosreason = inb(CMOS_DATA);
 #endif
 
-	/* record BSP in CPU map */
-	all_cpus = 1;
-
 	/* set up temporary P==V mapping for AP boot */
 	/* XXX this is a hack, we should boot the AP on its own stack/PTD */
 	kptbase = (uintptr_t)(void *)KPTphys;
diff --git a/sys/i386/i386/mptable.c b/sys/i386/i386/mptable.c
index d5af7b3..eaf508c 100644
--- a/sys/i386/i386/mptable.c
+++ b/sys/i386/i386/mptable.c
@@ -406,6 +406,12 @@ found:
 int
 cpu_mp_probe(void)
 {
+	/*
+	 * Record BSP in CPU map
+	 * This is done here so that MBUF init code works correctly.
+	 */
+	all_cpus = 1;
+
 	return (mp_capable);
 }
 
@@ -1929,9 +1935,6 @@ start_all_aps(u_int boot_addr)
 	mpbiosreason = inb(CMOS_DATA);
 #endif
 
-	/* record BSP in CPU map */
-	all_cpus = 1;
-
 	/* set up temporary P==V mapping for AP boot */
 	/* XXX this is a hack, we should boot the AP on its own stack/PTD */
 	kptbase = (uintptr_t)(void *)KPTphys;
diff --git a/sys/i386/include/mptable.h b/sys/i386/include/mptable.h
index d5af7b3..eaf508c 100644
--- a/sys/i386/include/mptable.h
+++ b/sys/i386/include/mptable.h
@@ -406,6 +406,12 @@ found:
 int
 cpu_mp_probe(void)
 {
+	/*
+	 * Record BSP in CPU map
+	 * This is done here so that MBUF init code works correctly.
+	 */
+	all_cpus = 1;
+
 	return (mp_capable);
 }
 
@@ -1929,9 +1935,6 @@ start_all_aps(u_int boot_addr)
 	mpbiosreason = inb(CMOS_DATA);
 #endif
 
-	/* record BSP in CPU map */
-	all_cpus = 1;
-
 	/* set up temporary P==V mapping for AP boot */
 	/* XXX this is a hack, we should boot the AP on its own stack/PTD */
 	kptbase = (uintptr_t)(void *)KPTphys;
diff --git a/sys/ia64/ia64/mp_machdep.c b/sys/ia64/ia64/mp_machdep.c
index d97c551..b46819d 100644
--- a/sys/ia64/ia64/mp_machdep.c
+++ b/sys/ia64/ia64/mp_machdep.c
@@ -71,6 +71,7 @@ SYSCTL_INT(_machdep, OID_AUTO, forward_irq_enabled, CTLFLAG_RW,
 int
 cpu_mp_probe()
 {
+	all_cpus = 1;	/* Needed for MB init code */
 	return (0);
 }
 
diff --git a/sys/kern/subr_mbuf.c b/sys/kern/subr_mbuf.c
index 97ea586..8be31a0 100644
--- a/sys/kern/subr_mbuf.c
+++ b/sys/kern/subr_mbuf.c
@@ -48,7 +48,12 @@
 /*
  * Maximum number of PCPU containers. If you know what you're doing you could
  * explicitly define MBALLOC_NCPU to be exactly the number of CPUs on your
- * system during compilation, and thus prevent kernel structure bloats.
+ * system during compilation, and thus prevent kernel structure bloat.
+ *
+ * SMP and non-SMP kernels clearly have a different number of possible cpus,
+ * but because we cannot assume a dense array of CPUs, we always allocate
+ * and traverse PCPU containers up to NCPU amount and merely check for
+ * CPU availability.
  */
 #ifdef	MBALLOC_NCPU
 #define	NCPU	MBALLOC_NCPU
@@ -57,12 +62,18 @@
 #endif
 
 /*
- * SMP and non-SMP kernels clearly have a different number of possible cpus.
+ * Macros allowing us to determine whether or not a given CPU's container
+ * should be configured during mb_init().
+ * XXX: Eventually we may want to provide hooks for CPU spinon/spinoff that
+ *      will allow us to configure the containers on spinon/spinoff. As it
+ *      stands, booting with CPU x disactivated and activating CPU x only
+ *      after bootup will lead to disaster and CPU x's container will be
+ *      uninitialized.
  */
 #ifdef	SMP
-#define	NCPU_PRESENT	mp_ncpus
+#define	CPU_ABSENT(x)	((all_cpus & (1 << x)) == 0)
 #else
-#define	NCPU_PRESENT	1
+#define	CPU_ABSENT(x)	0
 #endif
 
 /*
@@ -388,7 +399,10 @@ mb_init(void *dummy)
 	/*
 	 * Allocate and initialize PCPU containers.
 	 */
-	for (i = 0; i < NCPU_PRESENT; i++) {
+	for (i = 0; i < NCPU; i++) {
+		if (CPU_ABSENT(i))
+			continue;
+
 		mb_list_mbuf.ml_cntlst[i] = malloc(sizeof(struct mb_pcpu_list),
 		    M_MBUF, M_NOWAIT);
 		mb_list_clust.ml_cntlst[i] = malloc(sizeof(struct mb_pcpu_list),
@@ -401,6 +415,7 @@ mb_init(void *dummy)
 		mb_list_mbuf.ml_cntlst[i]->mb_cont.mc_lock =
 		    mb_list_clust.ml_cntlst[i]->mb_cont.mc_lock = &mbuf_pcpu[i];
 
+		mb_statpcpu[i].mb_active = 1;
 		mb_list_mbuf.ml_cntlst[i]->mb_cont.mc_numowner =
 		    mb_list_clust.ml_cntlst[i]->mb_cont.mc_numowner = i;
 		mb_list_mbuf.ml_cntlst[i]->mb_cont.mc_starved =
@@ -626,7 +641,9 @@ mb_alloc_wait(struct mb_lstmngr *mb_list)
 	 * Cycle all the PCPU containers. Increment starved counts if found
 	 * empty.
 	 */
-	for (i = 0; i < NCPU_PRESENT; i++) {
+	for (i = 0; i < NCPU; i++) {
+		if (CPU_ABSENT(i))
+			continue;
 		cnt_lst = MB_GET_PCPU_LIST_NUM(mb_list, i);
 		MB_LOCK_CONT(cnt_lst);
 
diff --git a/sys/powerpc/powerpc/mp_machdep.c b/sys/powerpc/powerpc/mp_machdep.c
index a8e5819..5e4c339 100644
--- a/sys/powerpc/powerpc/mp_machdep.c
+++ b/sys/powerpc/powerpc/mp_machdep.c
@@ -51,6 +51,7 @@ int			boot_cpu_id;
 int
 cpu_mp_probe(void)
 {
+	all_cpus = 1;	/* needed for MB init code */
 	return 0;
 }
 
diff --git a/sys/sys/mbuf.h b/sys/sys/mbuf.h
index ec496d4..d0eed39 100644
--- a/sys/sys/mbuf.h
+++ b/sys/sys/mbuf.h
@@ -216,6 +216,7 @@ struct mbpstat {
 	u_long	mb_mbpgs;
 	u_long	mb_clfree;
 	u_long	mb_clpgs;
+	short	mb_active;
 };
 
 /*