linux/arch/x86/mm/srat_64.c
David Rientjes adc1938994 x86: Interleave emulated nodes over physical nodes
Add interleaved NUMA emulation support

This patch interleaves emulated nodes over the system's physical
nodes. This is required for interleave optimizations since
mempolicies, for example, operate by iterating over a nodemask and
act without knowledge of node distances.  It can also be used for
testing memory latencies and NUMA bugs in the kernel.

There're a couple of ways to do this:

 - divide the number of emulated nodes by the number of physical
   nodes and allocate the result on each physical node, or

 - allocate each successive emulated node on a different physical
   node until all memory is exhausted.

The disadvantage of the first option is, depending on the asymmetry
in node capacities of each physical node, emulated nodes may
substantially differ in size on a particular physical node compared
to another.

The disadvantage of the second option is, also depending on the
asymmetry in node capacities of each physical node, there may be
more emulated nodes allocated on a single physical node as another.

This patch implements the second option; we sacrifice the
possibility that we may have slightly more emulated nodes on a
particular physical node compared to another in lieu of node size
asymmetry.

 [ Note that "node capacity" of a physical node is not only a
   function of its addressable range, but also is affected by
   subtracting out the amount of reserved memory over that range.
   NUMA emulation only deals with available, non-reserved memory
   quantities. ]

We ensure there is at least a minimal amount of available memory
allocated to each node.  We also make sure that at least this
amount of available memory is available in ZONE_DMA32 for any node
that includes both ZONE_DMA32 and ZONE_NORMAL.

This patch also cleans the emulation code up by no longer passing
the statically allocated struct bootnode array among the various
functions. This init.data array is not allocated on the stack since
it may be very large and thus it may be accessed at file scope.

The WARN_ON() for nodes_cover_memory() when faking proximity
domains is removed since it relies on successive nodes always
having greater start addresses than previous nodes; with
interleaving this is no longer always true.

Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Ankita Garg <ankita@in.ibm.com>
Cc: Len Brown <len.brown@intel.com>
LKML-Reference: <alpine.DEB.1.00.0909251519150.14754@chino.kir.corp.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-12 22:56:46 +02:00

510 lines
13 KiB
C

/*
* ACPI 3.0 based NUMA setup
* Copyright 2004 Andi Kleen, SuSE Labs.
*
* Reads the ACPI SRAT table to figure out what memory belongs to which CPUs.
*
* Called from acpi_numa_init while reading the SRAT and SLIT tables.
* Assumes all memory regions belonging to a single proximity domain
* are in one chunk. Holes between them will be included in the node.
*/
#include <linux/kernel.h>
#include <linux/acpi.h>
#include <linux/mmzone.h>
#include <linux/bitmap.h>
#include <linux/module.h>
#include <linux/topology.h>
#include <linux/bootmem.h>
#include <linux/mm.h>
#include <asm/proto.h>
#include <asm/numa.h>
#include <asm/e820.h>
#include <asm/apic.h>
#include <asm/uv/uv.h>
int acpi_numa __initdata;
static struct acpi_table_slit *acpi_slit;
static nodemask_t nodes_parsed __initdata;
static nodemask_t cpu_nodes_parsed __initdata;
static struct bootnode nodes[MAX_NUMNODES] __initdata;
static struct bootnode nodes_add[MAX_NUMNODES];
static int num_node_memblks __initdata;
static struct bootnode node_memblk_range[NR_NODE_MEMBLKS] __initdata;
static int memblk_nodeid[NR_NODE_MEMBLKS] __initdata;
static __init int setup_node(int pxm)
{
return acpi_map_pxm_to_node(pxm);
}
static __init int conflicting_memblks(unsigned long start, unsigned long end)
{
int i;
for (i = 0; i < num_node_memblks; i++) {
struct bootnode *nd = &node_memblk_range[i];
if (nd->start == nd->end)
continue;
if (nd->end > start && nd->start < end)
return memblk_nodeid[i];
if (nd->end == end && nd->start == start)
return memblk_nodeid[i];
}
return -1;
}
static __init void cutoff_node(int i, unsigned long start, unsigned long end)
{
struct bootnode *nd = &nodes[i];
if (nd->start < start) {
nd->start = start;
if (nd->end < nd->start)
nd->start = nd->end;
}
if (nd->end > end) {
nd->end = end;
if (nd->start > nd->end)
nd->start = nd->end;
}
}
static __init void bad_srat(void)
{
int i;
printk(KERN_ERR "SRAT: SRAT not used.\n");
acpi_numa = -1;
for (i = 0; i < MAX_LOCAL_APIC; i++)
apicid_to_node[i] = NUMA_NO_NODE;
for (i = 0; i < MAX_NUMNODES; i++) {
nodes[i].start = nodes[i].end = 0;
nodes_add[i].start = nodes_add[i].end = 0;
}
remove_all_active_ranges();
}
static __init inline int srat_disabled(void)
{
return numa_off || acpi_numa < 0;
}
/* Callback for SLIT parsing */
void __init acpi_numa_slit_init(struct acpi_table_slit *slit)
{
unsigned length;
unsigned long phys;
length = slit->header.length;
phys = find_e820_area(0, max_pfn_mapped<<PAGE_SHIFT, length,
PAGE_SIZE);
if (phys == -1L)
panic(" Can not save slit!\n");
acpi_slit = __va(phys);
memcpy(acpi_slit, slit, length);
reserve_early(phys, phys + length, "ACPI SLIT");
}
/* Callback for Proximity Domain -> x2APIC mapping */
void __init
acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa)
{
int pxm, node;
int apic_id;
if (srat_disabled())
return;
if (pa->header.length < sizeof(struct acpi_srat_x2apic_cpu_affinity)) {
bad_srat();
return;
}
if ((pa->flags & ACPI_SRAT_CPU_ENABLED) == 0)
return;
pxm = pa->proximity_domain;
node = setup_node(pxm);
if (node < 0) {
printk(KERN_ERR "SRAT: Too many proximity domains %x\n", pxm);
bad_srat();
return;
}
apic_id = pa->apic_id;
apicid_to_node[apic_id] = node;
node_set(node, cpu_nodes_parsed);
acpi_numa = 1;
printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> Node %u\n",
pxm, apic_id, node);
}
/* Callback for Proximity Domain -> LAPIC mapping */
void __init
acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa)
{
int pxm, node;
int apic_id;
if (srat_disabled())
return;
if (pa->header.length != sizeof(struct acpi_srat_cpu_affinity)) {
bad_srat();
return;
}
if ((pa->flags & ACPI_SRAT_CPU_ENABLED) == 0)
return;
pxm = pa->proximity_domain_lo;
node = setup_node(pxm);
if (node < 0) {
printk(KERN_ERR "SRAT: Too many proximity domains %x\n", pxm);
bad_srat();
return;
}
if (get_uv_system_type() >= UV_X2APIC)
apic_id = (pa->apic_id << 8) | pa->local_sapic_eid;
else
apic_id = pa->apic_id;
apicid_to_node[apic_id] = node;
node_set(node, cpu_nodes_parsed);
acpi_numa = 1;
printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> Node %u\n",
pxm, apic_id, node);
}
#ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
static inline int save_add_info(void) {return 1;}
#else
static inline int save_add_info(void) {return 0;}
#endif
/*
* Update nodes_add[]
* This code supports one contiguous hot add area per node
*/
static void __init
update_nodes_add(int node, unsigned long start, unsigned long end)
{
unsigned long s_pfn = start >> PAGE_SHIFT;
unsigned long e_pfn = end >> PAGE_SHIFT;
int changed = 0;
struct bootnode *nd = &nodes_add[node];
/* I had some trouble with strange memory hotadd regions breaking
the boot. Be very strict here and reject anything unexpected.
If you want working memory hotadd write correct SRATs.
The node size check is a basic sanity check to guard against
mistakes */
if ((signed long)(end - start) < NODE_MIN_SIZE) {
printk(KERN_ERR "SRAT: Hotplug area too small\n");
return;
}
/* This check might be a bit too strict, but I'm keeping it for now. */
if (absent_pages_in_range(s_pfn, e_pfn) != e_pfn - s_pfn) {
printk(KERN_ERR
"SRAT: Hotplug area %lu -> %lu has existing memory\n",
s_pfn, e_pfn);
return;
}
/* Looks good */
if (nd->start == nd->end) {
nd->start = start;
nd->end = end;
changed = 1;
} else {
if (nd->start == end) {
nd->start = start;
changed = 1;
}
if (nd->end == start) {
nd->end = end;
changed = 1;
}
if (!changed)
printk(KERN_ERR "SRAT: Hotplug zone not continuous. Partly ignored\n");
}
if (changed)
printk(KERN_INFO "SRAT: hot plug zone found %Lx - %Lx\n",
nd->start, nd->end);
}
/* Callback for parsing of the Proximity Domain <-> Memory Area mappings */
void __init
acpi_numa_memory_affinity_init(struct acpi_srat_mem_affinity *ma)
{
struct bootnode *nd, oldnode;
unsigned long start, end;
int node, pxm;
int i;
if (srat_disabled())
return;
if (ma->header.length != sizeof(struct acpi_srat_mem_affinity)) {
bad_srat();
return;
}
if ((ma->flags & ACPI_SRAT_MEM_ENABLED) == 0)
return;
if ((ma->flags & ACPI_SRAT_MEM_HOT_PLUGGABLE) && !save_add_info())
return;
start = ma->base_address;
end = start + ma->length;
pxm = ma->proximity_domain;
node = setup_node(pxm);
if (node < 0) {
printk(KERN_ERR "SRAT: Too many proximity domains.\n");
bad_srat();
return;
}
i = conflicting_memblks(start, end);
if (i == node) {
printk(KERN_WARNING
"SRAT: Warning: PXM %d (%lx-%lx) overlaps with itself (%Lx-%Lx)\n",
pxm, start, end, nodes[i].start, nodes[i].end);
} else if (i >= 0) {
printk(KERN_ERR
"SRAT: PXM %d (%lx-%lx) overlaps with PXM %d (%Lx-%Lx)\n",
pxm, start, end, node_to_pxm(i),
nodes[i].start, nodes[i].end);
bad_srat();
return;
}
nd = &nodes[node];
oldnode = *nd;
if (!node_test_and_set(node, nodes_parsed)) {
nd->start = start;
nd->end = end;
} else {
if (start < nd->start)
nd->start = start;
if (nd->end < end)
nd->end = end;
}
printk(KERN_INFO "SRAT: Node %u PXM %u %lx-%lx\n", node, pxm,
start, end);
if (ma->flags & ACPI_SRAT_MEM_HOT_PLUGGABLE) {
update_nodes_add(node, start, end);
/* restore nodes[node] */
*nd = oldnode;
if ((nd->start | nd->end) == 0)
node_clear(node, nodes_parsed);
}
node_memblk_range[num_node_memblks].start = start;
node_memblk_range[num_node_memblks].end = end;
memblk_nodeid[num_node_memblks] = node;
num_node_memblks++;
}
/* Sanity check to catch more bad SRATs (they are amazingly common).
Make sure the PXMs cover all memory. */
static int __init nodes_cover_memory(const struct bootnode *nodes)
{
int i;
unsigned long pxmram, e820ram;
pxmram = 0;
for_each_node_mask(i, nodes_parsed) {
unsigned long s = nodes[i].start >> PAGE_SHIFT;
unsigned long e = nodes[i].end >> PAGE_SHIFT;
pxmram += e - s;
pxmram -= absent_pages_in_range(s, e);
if ((long)pxmram < 0)
pxmram = 0;
}
e820ram = max_pfn - (e820_hole_size(0, max_pfn<<PAGE_SHIFT)>>PAGE_SHIFT);
/* We seem to lose 3 pages somewhere. Allow 1M of slack. */
if ((long)(e820ram - pxmram) >= (1<<(20 - PAGE_SHIFT))) {
printk(KERN_ERR
"SRAT: PXMs only cover %luMB of your %luMB e820 RAM. Not used.\n",
(pxmram << PAGE_SHIFT) >> 20,
(e820ram << PAGE_SHIFT) >> 20);
return 0;
}
return 1;
}
void __init acpi_numa_arch_fixup(void) {}
int __init acpi_get_nodes(struct bootnode *physnodes)
{
int i;
int ret = 0;
for_each_node_mask(i, nodes_parsed) {
physnodes[ret].start = nodes[i].start;
physnodes[ret].end = nodes[i].end;
ret++;
}
return ret;
}
/* Use the information discovered above to actually set up the nodes. */
int __init acpi_scan_nodes(unsigned long start, unsigned long end)
{
int i;
if (acpi_numa <= 0)
return -1;
/* First clean up the node list */
for (i = 0; i < MAX_NUMNODES; i++)
cutoff_node(i, start, end);
memnode_shift = compute_hash_shift(node_memblk_range, num_node_memblks,
memblk_nodeid);
if (memnode_shift < 0) {
printk(KERN_ERR
"SRAT: No NUMA node hash function found. Contact maintainer\n");
bad_srat();
return -1;
}
for_each_node_mask(i, nodes_parsed)
e820_register_active_regions(i, nodes[i].start >> PAGE_SHIFT,
nodes[i].end >> PAGE_SHIFT);
if (!nodes_cover_memory(nodes)) {
bad_srat();
return -1;
}
/* Account for nodes with cpus and no memory */
nodes_or(node_possible_map, nodes_parsed, cpu_nodes_parsed);
/* Finally register nodes */
for_each_node_mask(i, node_possible_map)
setup_node_bootmem(i, nodes[i].start, nodes[i].end);
/* Try again in case setup_node_bootmem missed one due
to missing bootmem */
for_each_node_mask(i, node_possible_map)
if (!node_online(i))
setup_node_bootmem(i, nodes[i].start, nodes[i].end);
for (i = 0; i < nr_cpu_ids; i++) {
int node = early_cpu_to_node(i);
if (node == NUMA_NO_NODE)
continue;
if (!node_online(node))
numa_clear_node(i);
}
numa_init_array();
return 0;
}
#ifdef CONFIG_NUMA_EMU
static int fake_node_to_pxm_map[MAX_NUMNODES] __initdata = {
[0 ... MAX_NUMNODES-1] = PXM_INVAL
};
static s16 fake_apicid_to_node[MAX_LOCAL_APIC] __initdata = {
[0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
};
static int __init find_node_by_addr(unsigned long addr)
{
int ret = NUMA_NO_NODE;
int i;
for_each_node_mask(i, nodes_parsed) {
/*
* Find the real node that this emulated node appears on. For
* the sake of simplicity, we only use a real node's starting
* address to determine which emulated node it appears on.
*/
if (addr >= nodes[i].start && addr < nodes[i].end) {
ret = i;
break;
}
}
return ret;
}
/*
* In NUMA emulation, we need to setup proximity domain (_PXM) to node ID
* mappings that respect the real ACPI topology but reflect our emulated
* environment. For each emulated node, we find which real node it appears on
* and create PXM to NID mappings for those fake nodes which mirror that
* locality. SLIT will now represent the correct distances between emulated
* nodes as a result of the real topology.
*/
void __init acpi_fake_nodes(const struct bootnode *fake_nodes, int num_nodes)
{
int i, j;
printk(KERN_INFO "Faking PXM affinity for fake nodes on real "
"topology.\n");
for (i = 0; i < num_nodes; i++) {
int nid, pxm;
nid = find_node_by_addr(fake_nodes[i].start);
if (nid == NUMA_NO_NODE)
continue;
pxm = node_to_pxm(nid);
if (pxm == PXM_INVAL)
continue;
fake_node_to_pxm_map[i] = pxm;
/*
* For each apicid_to_node mapping that exists for this real
* node, it must now point to the fake node ID.
*/
for (j = 0; j < MAX_LOCAL_APIC; j++)
if (apicid_to_node[j] == nid)
fake_apicid_to_node[j] = i;
}
for (i = 0; i < num_nodes; i++)
__acpi_map_pxm_to_node(fake_node_to_pxm_map[i], i);
memcpy(apicid_to_node, fake_apicid_to_node, sizeof(apicid_to_node));
nodes_clear(nodes_parsed);
for (i = 0; i < num_nodes; i++)
if (fake_nodes[i].start != fake_nodes[i].end)
node_set(i, nodes_parsed);
}
static int null_slit_node_compare(int a, int b)
{
return node_to_pxm(a) == node_to_pxm(b);
}
#else
static int null_slit_node_compare(int a, int b)
{
return a == b;
}
#endif /* CONFIG_NUMA_EMU */
int __node_distance(int a, int b)
{
int index;
if (!acpi_slit)
return null_slit_node_compare(a, b) ? LOCAL_DISTANCE :
REMOTE_DISTANCE;
index = acpi_slit->locality_count * node_to_pxm(a);
return acpi_slit->entry[index + node_to_pxm(b)];
}
EXPORT_SYMBOL(__node_distance);
#if defined(CONFIG_MEMORY_HOTPLUG_SPARSE) || defined(CONFIG_ACPI_HOTPLUG_MEMORY)
int memory_add_physaddr_to_nid(u64 start)
{
int i, ret = 0;
for_each_node(i)
if (nodes_add[i].start <= start && nodes_add[i].end > start)
ret = i;
return ret;
}
EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
#endif