[PATCH v3 0/3] Updates to pxm2node mapping and nodeID sizing

[PATCH v3 0/3] Updates to pxm2node mapping and nodeID sizing

[Boris Ostrovsky]

A set of patches for better px2<->node mapping and consistent use of
nodeID datatype.


Boris Ostrovsky (3):
  x86/numa: Allow arbitrary value of PXM in PXM<->node mapping
  x86/numa: Adjust datatypes for node and pxm
  mm: MEMF_node should handle changes in nodeid_t size

 xen/arch/x86/irq.c                  |    4 +-
 xen/arch/x86/numa.c                 |   15 ++--
 xen/arch/x86/setup.c                |    2 +-
 xen/arch/x86/smpboot.c              |    6 +-
 xen/arch/x86/srat.c                 |  121 ++++++++++++++++++++++++-----------
 xen/arch/x86/x86_64/mm.c            |    5 +-
 xen/common/page_alloc.c             |    8 ++-
 xen/drivers/passthrough/vtd/iommu.c |    5 +-
 xen/include/asm-arm/numa.h          |    4 +-
 xen/include/asm-x86/irq.h           |    3 +-
 xen/include/asm-x86/numa.h          |   24 ++++---
 xen/include/xen/mm.h                |    5 +-
 12 files changed, 133 insertions(+), 69 deletions(-)

[PATCH v3 1/3] x86/numa: Allow arbitrary value of PXM in PXM<->node mapping

[Boris Ostrovsky]

ACPI defines proximity domain identifier as a 32-bit integer. While
in most cases the values will be zero-based this is not guaranteed,
making current pxm2node[256] mapping structure not appropriate.

We will instead use MAX_NUMNODES-sized array of struct pxm2node to
store PXM-to-node mapping. To accommodate common case of zero-based
and contiguios PXMs we will, whenever possible, try to use PXM as
index into this array array for fast lookups.

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
---

Changes in v3:
* Renamed p2n to pxm2node (and struct pxm2node to struct pxm_to_node)
* Changed node type in pxm_to_node to u8
* pxm_to_node(), setup_node() and node_to_pxm()) use unsigned as node
  argument
* Reinitialized pxm2node in abd_srat();
* Added BUILD_BUG_ON(MAX_NUMNODES >= NUMA_NO_NODE);
* Made changes to K&R style (which is what this file uses)

 xen/arch/x86/srat.c        |   96 ++++++++++++++++++++++++++++++++------------
 xen/include/asm-x86/numa.h |    4 +-
 2 files changed, 72 insertions(+), 28 deletions(-)

diff --git a/xen/arch/x86/srat.c b/xen/arch/x86/srat.c
index 29fc724..1664393 100644
--- a/xen/arch/x86/srat.c
+++ b/xen/arch/x86/srat.c
@@ -27,35 +27,78 @@ static nodemask_t memory_nodes_parsed __initdata;
 static nodemask_t processor_nodes_parsed __initdata;
 static nodemask_t nodes_found __initdata;
 static struct node nodes[MAX_NUMNODES] __initdata;
-static u8 __read_mostly pxm2node[256] = { [0 ... 255] = NUMA_NO_NODE };
 
+struct pxm_to_node {
+	unsigned pxm;
+	u8 node;
+};
+static struct pxm_to_node __read_mostly pxm2node[MAX_NUMNODES] =
+	{ [0 ... MAX_NUMNODES - 1] = {.node = NUMA_NO_NODE} };
+
+static int node_to_pxm(unsigned n);
 
 static int num_node_memblks;
 static struct node node_memblk_range[NR_NODE_MEMBLKS];
 static int memblk_nodeid[NR_NODE_MEMBLKS];
 
+static inline bool_t node_found(unsigned idx, unsigned pxm)
+{
+	return ((pxm2node[idx].pxm == pxm) &&
+		(pxm2node[idx].node != NUMA_NO_NODE));
+}
 
-static int node_to_pxm(int n);
-
-int pxm_to_node(int pxm)
+int pxm_to_node(unsigned pxm)
 {
-	if ((unsigned)pxm >= 256)
-		return -1;
+	unsigned i;
+
+	if ((pxm < MAX_NUMNODES) && node_found(pxm, pxm))
+		return pxm2node[pxm].node;
+
+	for (i = 0; i < MAX_NUMNODES; i++)
+		if (node_found(i, pxm))
+			return pxm2node[i].node;
+
 	/* Extend 0xff to (int)-1 */
-	return (signed char)pxm2node[pxm];
+	return (signed char)NUMA_NO_NODE;
 }
 
-__devinit int setup_node(int pxm)
+__devinit int setup_node(unsigned pxm)
 {
-	unsigned node = pxm2node[pxm];
-	if (node == 0xff) {
-		if (nodes_weight(nodes_found) >= MAX_NUMNODES)
-			return -1;
-		node = first_unset_node(nodes_found); 
-		node_set(node, nodes_found);
-		pxm2node[pxm] = node;
+	int node;
+	unsigned idx;
+	static bool_t warned;
+
+	BUILD_BUG_ON(MAX_NUMNODES >= NUMA_NO_NODE);
+
+	if (pxm < MAX_NUMNODES) {
+		if (node_found(pxm, pxm))
+			return pxm2node[pxm].node;
+
+		/* Try to maintain indexing of pxm2node by pxm */
+		if (pxm2node[pxm].node == NUMA_NO_NODE) {
+			idx = pxm;
+			goto finish;
+		}
+	}
+
+	for (idx = 0; idx < MAX_NUMNODES; idx++)
+		if (pxm2node[idx].node == NUMA_NO_NODE)
+			goto finish;
+
+	if (!warned) {
+		printk(XENLOG_WARNING "More PXMs than available nodes\n");
+		warned = 1;
 	}
-	return pxm2node[pxm];
+
+	return (signed char)NUMA_NO_NODE;
+
+ finish:
+	node = first_unset_node(nodes_found);
+	node_set(node, nodes_found);
+	pxm2node[idx].pxm = pxm;
+	pxm2node[idx].node = node;
+
+	return node;
 }
 
 int valid_numa_range(u64 start, u64 end, int node)
@@ -111,8 +154,8 @@ static __init void bad_srat(void)
 	acpi_numa = -1;
 	for (i = 0; i < MAX_LOCAL_APIC; i++)
 		apicid_to_node[i] = NUMA_NO_NODE;
-	for (i = 0; i < ARRAY_SIZE(pxm2node); i++)
-		pxm2node[i] = NUMA_NO_NODE;
+	for (i = 0; i < MAX_NUMNODES; i++)
+		pxm2node[i].node = NUMA_NO_NODE;
 	mem_hotplug = 0;
 }
 
@@ -438,15 +481,16 @@ int __init acpi_scan_nodes(u64 start, u64 end)
 	return 0;
 }
 
-static int node_to_pxm(int n)
+static int node_to_pxm(unsigned n)
 {
-       int i;
-       if (pxm2node[n] == n)
-               return n;
-       for (i = 0; i < 256; i++)
-               if (pxm2node[i] == n)
-                       return i;
-       return 0;
+	unsigned i;
+
+	if ((n <  MAX_NUMNODES) && (pxm2node[n].node == n))
+		return pxm2node[n].pxm;
+	for (i = 0; i < MAX_NUMNODES; i++)
+		if (pxm2node[i].node == n)
+			return pxm2node[i].pxm;
+	return 0;
 }
 
 int __node_distance(int a, int b)
diff --git a/xen/include/asm-x86/numa.h b/xen/include/asm-x86/numa.h
index 5959860..e491a9e 100644
--- a/xen/include/asm-x86/numa.h
+++ b/xen/include/asm-x86/numa.h
@@ -21,7 +21,7 @@ struct node {
 
 extern int compute_hash_shift(struct node *nodes, int numnodes,
 			      int *nodeids);
-extern int pxm_to_node(int nid);
+extern int pxm_to_node(unsigned pxm);
 
 #define ZONE_ALIGN (1UL << (MAX_ORDER+PAGE_SHIFT))
 #define VIRTUAL_BUG_ON(x) 
@@ -33,7 +33,7 @@ extern int numa_off;
 
 extern int srat_disabled(void);
 extern void numa_set_node(int cpu, int node);
-extern int setup_node(int pxm);
+extern int setup_node(unsigned pxm);
 extern void srat_detect_node(int cpu);
 
 extern void setup_node_bootmem(int nodeid, u64 start, u64 end);
-- 
1.7.1

[PATCH v3 2/3] x86/numa: Adjust datatypes for node and pxm

[Boris Ostrovsky]

Use u8-sized node IDs and unsigned PXMs consistently throughout
code (and introduce nodeid_t type).

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
---

Changes n v3:
* Moved node declaration in cpu_add() into its scope.

 xen/arch/x86/irq.c                  |    4 +-
 xen/arch/x86/numa.c                 |   15 ++++++------
 xen/arch/x86/setup.c                |    2 +-
 xen/arch/x86/smpboot.c              |    6 +++-
 xen/arch/x86/srat.c                 |   41 ++++++++++++++++++----------------
 xen/arch/x86/x86_64/mm.c            |    5 ++-
 xen/common/page_alloc.c             |    5 ++-
 xen/drivers/passthrough/vtd/iommu.c |    5 ++-
 xen/include/asm-arm/numa.h          |    4 ++-
 xen/include/asm-x86/irq.h           |    3 +-
 xen/include/asm-x86/numa.h          |   24 +++++++++++---------
 11 files changed, 64 insertions(+), 50 deletions(-)

diff --git a/xen/arch/x86/irq.c b/xen/arch/x86/irq.c
index f214072..9be8840 100644
--- a/xen/arch/x86/irq.c
+++ b/xen/arch/x86/irq.c
@@ -153,7 +153,7 @@ int __init bind_irq_vector(int irq, int vector, const cpumask_t *cpu_mask)
 /*
  * Dynamic irq allocate and deallocation for MSI
  */
-int create_irq(int node)
+int create_irq(nodeid_t node)
 {
     int irq, ret;
     struct irq_desc *desc;
@@ -173,7 +173,7 @@ int create_irq(int node)
     {
         cpumask_t *mask = NULL;
 
-        if (node != NUMA_NO_NODE && node >= 0)
+        if ( node != NUMA_NO_NODE )
         {
             mask = &node_to_cpumask(node);
             if (cpumask_empty(mask))
diff --git a/xen/arch/x86/numa.c b/xen/arch/x86/numa.c
index 628a40a..7a01923 100644
--- a/xen/arch/x86/numa.c
+++ b/xen/arch/x86/numa.c
@@ -35,13 +35,13 @@ static typeof(*memnodemap) _memnodemap[64];
 unsigned long memnodemapsize;
 u8 *memnodemap;
 
-unsigned char cpu_to_node[NR_CPUS] __read_mostly = {
+nodeid_t cpu_to_node[NR_CPUS] __read_mostly = {
     [0 ... NR_CPUS-1] = NUMA_NO_NODE
 };
 /*
  * Keep BIOS's CPU2node information, should not be used for memory allocaion
  */
-unsigned char apicid_to_node[MAX_LOCAL_APIC] __cpuinitdata = {
+nodeid_t apicid_to_node[MAX_LOCAL_APIC] __cpuinitdata = {
     [0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
 };
 cpumask_t node_to_cpumask[MAX_NUMNODES] __read_mostly;
@@ -65,7 +65,7 @@ int srat_disabled(void)
  * -1 if node overlap or lost ram (shift too big)
  */
 static int __init populate_memnodemap(const struct node *nodes,
-                                      int numnodes, int shift, int *nodeids)
+                                      int numnodes, int shift, nodeid_t *nodeids)
 {
     unsigned long spdx, epdx;
     int i, res = -1;
@@ -150,7 +150,7 @@ static int __init extract_lsb_from_nodes(const struct node *nodes,
 }
 
 int __init compute_hash_shift(struct node *nodes, int numnodes,
-                              int *nodeids)
+                              nodeid_t *nodeids)
 {
     int shift;
 
@@ -172,7 +172,7 @@ int __init compute_hash_shift(struct node *nodes, int numnodes,
     return shift;
 }
 /* initialize NODE_DATA given nodeid and start/end */
-void __init setup_node_bootmem(int nodeid, u64 start, u64 end)
+void __init setup_node_bootmem(nodeid_t nodeid, u64 start, u64 end)
 { 
     unsigned long start_pfn, end_pfn;
 
@@ -294,7 +294,7 @@ __cpuinit void numa_add_cpu(int cpu)
     cpumask_set_cpu(cpu, &node_to_cpumask[cpu_to_node(cpu)]);
 } 
 
-void __cpuinit numa_set_node(int cpu, int node)
+void __cpuinit numa_set_node(int cpu, nodeid_t node)
 {
     cpu_to_node[cpu] = node;
 }
@@ -340,7 +340,8 @@ static __init int numa_setup(char *opt)
  */
 void __init init_cpu_to_node(void)
 {
-    int i, node;
+    unsigned i;
+    nodeid_t node;
 
     for ( i = 0; i < nr_cpu_ids; i++ )
     {
diff --git a/xen/arch/x86/setup.c b/xen/arch/x86/setup.c
index c27c49c..84da722 100644
--- a/xen/arch/x86/setup.c
+++ b/xen/arch/x86/setup.c
@@ -188,7 +188,7 @@ static void __init init_idle_domain(void)
 
 void __devinit srat_detect_node(int cpu)
 {
-    unsigned node;
+    nodeid_t node;
     u32 apicid = x86_cpu_to_apicid[cpu];
 
     node = apicid_to_node[apicid];
diff --git a/xen/arch/x86/smpboot.c b/xen/arch/x86/smpboot.c
index c54be7e..ea84fe0 100644
--- a/xen/arch/x86/smpboot.c
+++ b/xen/arch/x86/smpboot.c
@@ -843,7 +843,7 @@ void __cpu_die(unsigned int cpu)
 
 int cpu_add(uint32_t apic_id, uint32_t acpi_id, uint32_t pxm)
 {
-    int node, cpu = -1;
+    int cpu = -1;
 
     dprintk(XENLOG_DEBUG, "cpu_add apic_id %x acpi_id %x pxm %x\n",
             apic_id, acpi_id, pxm);
@@ -877,7 +877,9 @@ int cpu_add(uint32_t apic_id, uint32_t acpi_id, uint32_t pxm)
 
     if ( !srat_disabled() )
     {
-        if ( (node = setup_node(pxm)) < 0 )
+        nodeid_t node = setup_node(pxm);
+
+        if ( node == NUMA_NO_NODE )
         {
             dprintk(XENLOG_WARNING,
                     "Setup node failed for pxm %x\n", pxm);
diff --git a/xen/arch/x86/srat.c b/xen/arch/x86/srat.c
index 1664393..5a237a3 100644
--- a/xen/arch/x86/srat.c
+++ b/xen/arch/x86/srat.c
@@ -30,16 +30,16 @@ static struct node nodes[MAX_NUMNODES] __initdata;
 
 struct pxm_to_node {
 	unsigned pxm;
-	u8 node;
+	nodeid_t node;
 };
 static struct pxm_to_node __read_mostly pxm2node[MAX_NUMNODES] =
 	{ [0 ... MAX_NUMNODES - 1] = {.node = NUMA_NO_NODE} };
 
-static int node_to_pxm(unsigned n);
+static unsigned node_to_pxm(nodeid_t n);
 
 static int num_node_memblks;
 static struct node node_memblk_range[NR_NODE_MEMBLKS];
-static int memblk_nodeid[NR_NODE_MEMBLKS];
+static nodeid_t memblk_nodeid[NR_NODE_MEMBLKS];
 
 static inline bool_t node_found(unsigned idx, unsigned pxm)
 {
@@ -47,7 +47,7 @@ static inline bool_t node_found(unsigned idx, unsigned pxm)
 		(pxm2node[idx].node != NUMA_NO_NODE));
 }
 
-int pxm_to_node(unsigned pxm)
+nodeid_t pxm_to_node(unsigned pxm)
 {
 	unsigned i;
 
@@ -58,13 +58,12 @@ int pxm_to_node(unsigned pxm)
 		if (node_found(i, pxm))
 			return pxm2node[i].node;
 
-	/* Extend 0xff to (int)-1 */
-	return (signed char)NUMA_NO_NODE;
+	return NUMA_NO_NODE;
 }
 
-__devinit int setup_node(unsigned pxm)
+__devinit nodeid_t setup_node(unsigned pxm)
 {
-	int node;
+	nodeid_t node;
 	unsigned idx;
 	static bool_t warned;
 
@@ -90,7 +89,7 @@ __devinit int setup_node(unsigned pxm)
 		warned = 1;
 	}
 
-	return (signed char)NUMA_NO_NODE;
+	return NUMA_NO_NODE;
 
  finish:
 	node = first_unset_node(nodes_found);
@@ -101,7 +100,7 @@ __devinit int setup_node(unsigned pxm)
 	return node;
 }
 
-int valid_numa_range(u64 start, u64 end, int node)
+int valid_numa_range(u64 start, u64 end, nodeid_t node)
 {
 	int i;
 
@@ -205,8 +204,9 @@ void __init acpi_numa_slit_init(struct acpi_table_slit *slit)
 void __init
 acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa)
 {
-	int pxm, node;
-	int apic_id;
+	unsigned pxm;
+	nodeid_t node;
+	u32 apic_id;
 
 	if (srat_disabled())
 		return;
@@ -218,7 +218,7 @@ acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa)
 		return;
 	pxm = pa->proximity_domain;
 	node = setup_node(pxm);
-	if (node < 0) {
+	if (node == NUMA_NO_NODE) {
 		printk(KERN_ERR "SRAT: Too many proximity domains %x\n", pxm);
 		bad_srat();
 		return;
@@ -235,7 +235,9 @@ acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa)
 void __init
 acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa)
 {
-	int pxm, node;
+	unsigned pxm;
+	nodeid_t node;
+
 	if (srat_disabled())
 		return;
 	if (pa->header.length != sizeof(struct acpi_srat_cpu_affinity)) {
@@ -251,7 +253,7 @@ acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa)
 		pxm |= pa->proximity_domain_hi[2] << 24;
 	}
 	node = setup_node(pxm);
-	if (node < 0) {
+	if (node == NUMA_NO_NODE) {
 		printk(KERN_ERR "SRAT: Too many proximity domains %x\n", pxm);
 		bad_srat();
 		return;
@@ -269,7 +271,8 @@ acpi_numa_memory_affinity_init(struct acpi_srat_mem_affinity *ma)
 {
 	struct node *nd;
 	u64 start, end;
-	int node, pxm;
+	unsigned pxm;
+	nodeid_t node;
 	int i;
 
 	if (srat_disabled())
@@ -295,7 +298,7 @@ acpi_numa_memory_affinity_init(struct acpi_srat_mem_affinity *ma)
 	if (srat_rev < 2)
 		pxm &= 0xff;
 	node = setup_node(pxm);
-	if (node < 0) {
+	if (node == NUMA_NO_NODE) {
 		printk(KERN_ERR "SRAT: Too many proximity domains.\n");
 		bad_srat();
 		return;
@@ -481,7 +484,7 @@ int __init acpi_scan_nodes(u64 start, u64 end)
 	return 0;
 }
 
-static int node_to_pxm(unsigned n)
+static unsigned node_to_pxm(nodeid_t n)
 {
 	unsigned i;
 
@@ -493,7 +496,7 @@ static int node_to_pxm(unsigned n)
 	return 0;
 }
 
-int __node_distance(int a, int b)
+int __node_distance(nodeid_t a, nodeid_t b)
 {
 	int index;
 
diff --git a/xen/arch/x86/x86_64/mm.c b/xen/arch/x86/x86_64/mm.c
index d631aee..6875c92 100644
--- a/xen/arch/x86/x86_64/mm.c
+++ b/xen/arch/x86/x86_64/mm.c
@@ -1343,7 +1343,8 @@ int mem_hotadd_check(unsigned long spfn, unsigned long epfn)
 int memory_add(unsigned long spfn, unsigned long epfn, unsigned int pxm)
 {
     struct mem_hotadd_info info;
-    int ret, node;
+    int ret;
+    nodeid_t node;
     unsigned long old_max = max_page, old_total = total_pages;
     unsigned long old_node_start, old_node_span, orig_online;
     unsigned long i;
@@ -1353,7 +1354,7 @@ int memory_add(unsigned long spfn, unsigned long epfn, unsigned int pxm)
     if ( !mem_hotadd_check(spfn, epfn) )
         return -EINVAL;
 
-    if ( (node = setup_node(pxm)) == -1 )
+    if ( (node = setup_node(pxm)) == NUMA_NO_NODE )
         return -EINVAL;
 
     if ( !valid_numa_range(spfn << PAGE_SHIFT, epfn << PAGE_SHIFT, node) )
diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
index 7b4092d..124fa88 100644
--- a/xen/common/page_alloc.c
+++ b/xen/common/page_alloc.c
@@ -581,7 +581,7 @@ static struct page_info *alloc_heap_pages(
     struct domain *d)
 {
     unsigned int first_node, i, j, zone = 0, nodemask_retry = 0;
-    unsigned int node = (uint8_t)((memflags >> _MEMF_node) - 1);
+    nodeid_t node = (nodeid_t)((memflags >> _MEMF_node) - 1);
     unsigned long request = 1UL << order;
     struct page_info *pg;
     nodemask_t nodemask = (d != NULL ) ? d->node_affinity : node_online_map;
@@ -1278,7 +1278,8 @@ static void __init smp_scrub_heap_pages(void *data)
     unsigned long mfn, start, end;
     struct page_info *pg;
     struct scrub_region *r;
-    unsigned int temp_cpu, node, cpu_idx = 0;
+    unsigned int temp_cpu, cpu_idx = 0;
+    nodeid_t node;
     unsigned int cpu = smp_processor_id();
 
     if ( data )
diff --git a/xen/drivers/passthrough/vtd/iommu.c b/xen/drivers/passthrough/vtd/iommu.c
index 19d8165..2e0af6f 100644
--- a/xen/drivers/passthrough/vtd/iommu.c
+++ b/xen/drivers/passthrough/vtd/iommu.c
@@ -190,14 +190,15 @@ u64 alloc_pgtable_maddr(struct acpi_drhd_unit *drhd, unsigned long npages)
     struct acpi_rhsa_unit *rhsa;
     struct page_info *pg, *cur_pg;
     u64 *vaddr;
-    int node = -1, i;
+    nodeid_t node = NUMA_NO_NODE;
+    unsigned i;
 
     rhsa = drhd_to_rhsa(drhd);
     if ( rhsa )
         node =  pxm_to_node(rhsa->proximity_domain);
 
     pg = alloc_domheap_pages(NULL, get_order_from_pages(npages),
-                             (node == -1 ) ? 0 : MEMF_node(node));
+                             (node == NUMA_NO_NODE) ? 0 : MEMF_node(node));
     if ( !pg )
         return 0;
 
diff --git a/xen/include/asm-arm/numa.h b/xen/include/asm-arm/numa.h
index 06a9d5a..a00cb7c 100644
--- a/xen/include/asm-arm/numa.h
+++ b/xen/include/asm-arm/numa.h
@@ -1,11 +1,13 @@
 #ifndef __ARCH_ARM_NUMA_H
 #define __ARCH_ARM_NUMA_H
 
+typedef u8 nodeid_t;
+
 /* Fake one node for now. See also node_online_map. */
 #define cpu_to_node(cpu) 0
 #define node_to_cpumask(node)   (cpu_online_map)
 
-static inline __attribute__((pure)) int phys_to_nid(paddr_t addr)
+static inline __attribute__((pure)) nodeid_t phys_to_nid(paddr_t addr)
 {
     return 0;
 }
diff --git a/xen/include/asm-x86/irq.h b/xen/include/asm-x86/irq.h
index d3c55f3..a44305e 100644
--- a/xen/include/asm-x86/irq.h
+++ b/xen/include/asm-x86/irq.h
@@ -5,6 +5,7 @@
 
 #include <xen/config.h>
 #include <asm/atomic.h>
+#include <asm/numa.h>
 #include <xen/cpumask.h>
 #include <xen/smp.h>
 #include <xen/hvm/irq.h>
@@ -155,7 +156,7 @@ int  init_irq_data(void);
 void clear_irq_vector(int irq);
 
 int irq_to_vector(int irq);
-int create_irq(int node);
+int create_irq(nodeid_t node);
 void destroy_irq(unsigned int irq);
 int assign_irq_vector(int irq, const cpumask_t *);
 
diff --git a/xen/include/asm-x86/numa.h b/xen/include/asm-x86/numa.h
index e491a9e..419f770 100644
--- a/xen/include/asm-x86/numa.h
+++ b/xen/include/asm-x86/numa.h
@@ -5,6 +5,8 @@
 
 #define NODES_SHIFT 6
 
+typedef u8 nodeid_t;
+
 extern int srat_rev;
 
 extern unsigned char cpu_to_node[];
@@ -20,8 +22,8 @@ struct node {
 };
 
 extern int compute_hash_shift(struct node *nodes, int numnodes,
-			      int *nodeids);
-extern int pxm_to_node(unsigned pxm);
+			      nodeid_t *nodeids);
+extern nodeid_t pxm_to_node(unsigned pxm);
 
 #define ZONE_ALIGN (1UL << (MAX_ORDER+PAGE_SHIFT))
 #define VIRTUAL_BUG_ON(x) 
@@ -32,12 +34,12 @@ extern int numa_off;
 
 
 extern int srat_disabled(void);
-extern void numa_set_node(int cpu, int node);
-extern int setup_node(unsigned pxm);
+extern void numa_set_node(int cpu, nodeid_t node);
+extern nodeid_t setup_node(unsigned pxm);
 extern void srat_detect_node(int cpu);
 
-extern void setup_node_bootmem(int nodeid, u64 start, u64 end);
-extern unsigned char apicid_to_node[];
+extern void setup_node_bootmem(nodeid_t nodeid, u64 start, u64 end);
+extern nodeid_t apicid_to_node[];
 #ifdef CONFIG_NUMA
 extern void init_cpu_to_node(void);
 
@@ -54,14 +56,14 @@ extern u8 *memnodemap;
 struct node_data {
     unsigned long node_start_pfn;
     unsigned long node_spanned_pages;
-    unsigned int  node_id;
+    nodeid_t      node_id;
 };
 
 extern struct node_data node_data[];
 
-static inline __attribute__((pure)) int phys_to_nid(paddr_t addr) 
+static inline __attribute__((pure)) nodeid_t phys_to_nid(paddr_t addr) 
 { 
-	unsigned nid;
+	nodeid_t nid;
 	VIRTUAL_BUG_ON((paddr_to_pdx(addr) >> memnode_shift) >= memnodemapsize);
 	nid = memnodemap[paddr_to_pdx(addr) >> memnode_shift]; 
 	VIRTUAL_BUG_ON(nid >= MAX_NUMNODES || !node_data[nid]); 
@@ -75,7 +77,7 @@ static inline __attribute__((pure)) int phys_to_nid(paddr_t addr)
 #define node_end_pfn(nid)       (NODE_DATA(nid)->node_start_pfn + \
 				 NODE_DATA(nid)->node_spanned_pages)
 
-extern int valid_numa_range(u64 start, u64 end, int node);
+extern int valid_numa_range(u64 start, u64 end, nodeid_t node);
 #else
 #define init_cpu_to_node() do {} while (0)
 #define clear_node_cpumask(cpu) do {} while (0)
@@ -83,6 +85,6 @@ extern int valid_numa_range(u64 start, u64 end, int node);
 #endif
 
 void srat_parse_regions(u64 addr);
-extern int __node_distance(int a, int b);
+extern int __node_distance(nodeid_t a, nodeid_t b);
 
 #endif
-- 
1.7.1

[PATCH v3 3/3] mm: MEMF_node should handle changes in nodeid_t size

[Boris Ostrovsky]

Instead of using a hardcoded constant to extract nodeID from
memflags use a macro whose value is based on nodeid_t size.

Also provide a macro for extracting nodeID from memflags so that
users don't need to remember to decrement the value.

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
---

Changes in v3:
* Fixed shift for MEMF_node_mask and moved it a few lines down
* Added BUILD_BUG_ON((_MEMF_bits - _MEMF_node) >= (8 * sizeof(nodeid_t))) to
  alloc_heap_pages() (Not sure it's the best place).

 xen/common/page_alloc.c |    5 ++++-
 xen/include/xen/mm.h    |    5 ++++-
 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
index 124fa88..08d8ed6 100644
--- a/xen/common/page_alloc.c
+++ b/xen/common/page_alloc.c
@@ -581,13 +581,16 @@ static struct page_info *alloc_heap_pages(
     struct domain *d)
 {
     unsigned int first_node, i, j, zone = 0, nodemask_retry = 0;
-    nodeid_t node = (nodeid_t)((memflags >> _MEMF_node) - 1);
+    nodeid_t node = MEMF2NODE(memflags);
     unsigned long request = 1UL << order;
     struct page_info *pg;
     nodemask_t nodemask = (d != NULL ) ? d->node_affinity : node_online_map;
     bool_t need_tlbflush = 0;
     uint32_t tlbflush_timestamp = 0;
 
+    /* Make sure there are enough bits in memflags for nodeID */
+    BUILD_BUG_ON((_MEMF_bits - _MEMF_node) < (8 * sizeof(nodeid_t)));
+
     if ( node == NUMA_NO_NODE )
     {
         memflags &= ~MEMF_exact_node;
diff --git a/xen/include/xen/mm.h b/xen/include/xen/mm.h
index 74a65a6..788fbdd 100644
--- a/xen/include/xen/mm.h
+++ b/xen/include/xen/mm.h
@@ -121,10 +121,13 @@ struct npfec {
 #define _MEMF_exact_node  4
 #define  MEMF_exact_node  (1U<<_MEMF_exact_node)
 #define _MEMF_node        8
-#define  MEMF_node(n)     ((((n)+1)&0xff)<<_MEMF_node)
+#define  MEMF_node_mask   ((1U<<(8*sizeof(nodeid_t)))-1)
+#define  MEMF_node(n)     ((((n)+1)&MEMF_node_mask)<<_MEMF_node)
 #define _MEMF_bits        24
 #define  MEMF_bits(n)     ((n)<<_MEMF_bits)
 
+#define MEMF2NODE(memflags) (MASK_EXTR(memflags,MEMF_node_mask)-1)
+
 #ifdef CONFIG_PAGEALLOC_MAX_ORDER
 #define MAX_ORDER CONFIG_PAGEALLOC_MAX_ORDER
 #else
-- 
1.7.1

Re: [PATCH v3 1/3] x86/numa: Allow arbitrary value of PXM in PXM<->node mapping

[Jan Beulich]

>>> On 26.02.15 at 00:41, <boris.ostrovsky@oracle.com> wrote:
> * Made changes to K&R style (which is what this file uses)

I don't find anything K&R style here, and I think the compiler would
have prevented you from doing any such (by way for issuing
warnings and us using -Werror). I'm sure you mean Linux instead.

> @@ -111,8 +154,8 @@ static __init void bad_srat(void)
>  	acpi_numa = -1;
>  	for (i = 0; i < MAX_LOCAL_APIC; i++)
>  		apicid_to_node[i] = NUMA_NO_NODE;
> -	for (i = 0; i < ARRAY_SIZE(pxm2node); i++)
> -		pxm2node[i] = NUMA_NO_NODE;
> +	for (i = 0; i < MAX_NUMNODES; i++)
> +		pxm2node[i].node = NUMA_NO_NODE;

It escapes me why the ARRAY_SIZE() use needed replacing. I'm
going to undo that before committing, and in fact will adjust other
similar bounds checks to use it too.

Jan

Re: [PATCH v3 1/3] x86/numa: Allow arbitrary value of PXM in PXM<->node mapping

[Dario Faggioli]

[-- Attachment #1.1: Type: text/plain, Size: 586 bytes --]

On Wed, 2015-02-25 at 18:41 -0500, Boris Ostrovsky wrote:
> ACPI defines proximity domain identifier as a 32-bit integer. While
> in most cases the values will be zero-based this is not guaranteed,
> making current pxm2node[256] mapping structure not appropriate.
> 
> We will instead use MAX_NUMNODES-sized array of struct pxm2node to
> store PXM-to-node mapping. To accommodate common case of zero-based
> and contiguios PXMs we will, whenever possible, try to use PXM as
> index into this array array for fast lookups.
> 
                ^array for fast lookups.

Dario

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel