2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* pSeries NUMA support
|
|
|
|
*
|
|
|
|
* Copyright (C) 2002 Anton Blanchard <anton@au.ibm.com>, IBM
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public License
|
|
|
|
* as published by the Free Software Foundation; either version
|
|
|
|
* 2 of the License, or (at your option) any later version.
|
|
|
|
*/
|
|
|
|
#include <linux/threads.h>
|
|
|
|
#include <linux/bootmem.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/mmzone.h>
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/nodemask.h>
|
|
|
|
#include <linux/cpu.h>
|
|
|
|
#include <linux/notifier.h>
|
2008-02-14 07:56:49 +07:00
|
|
|
#include <linux/lmb.h>
|
2008-02-14 07:37:49 +07:00
|
|
|
#include <linux/of.h>
|
2005-11-11 10:22:35 +07:00
|
|
|
#include <asm/sparsemem.h>
|
2008-02-14 07:56:49 +07:00
|
|
|
#include <asm/prom.h>
|
2005-10-31 09:07:02 +07:00
|
|
|
#include <asm/system.h>
|
2005-11-07 09:18:13 +07:00
|
|
|
#include <asm/smp.h>
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
static int numa_enabled = 1;
|
|
|
|
|
2008-02-01 11:57:31 +07:00
|
|
|
static char *cmdline __initdata;
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
static int numa_debug;
|
|
|
|
#define dbg(args...) if (numa_debug) { printk(KERN_INFO args); }
|
|
|
|
|
2005-11-11 10:22:35 +07:00
|
|
|
int numa_cpu_lookup_table[NR_CPUS];
|
2005-04-17 05:20:36 +07:00
|
|
|
cpumask_t numa_cpumask_lookup_table[MAX_NUMNODES];
|
|
|
|
struct pglist_data *node_data[MAX_NUMNODES];
|
2005-11-11 10:22:35 +07:00
|
|
|
|
|
|
|
EXPORT_SYMBOL(numa_cpu_lookup_table);
|
|
|
|
EXPORT_SYMBOL(numa_cpumask_lookup_table);
|
|
|
|
EXPORT_SYMBOL(node_data);
|
|
|
|
|
|
|
|
static bootmem_data_t __initdata plat_node_bdata[MAX_NUMNODES];
|
2005-04-17 05:20:36 +07:00
|
|
|
static int min_common_depth;
|
2005-12-06 03:06:42 +07:00
|
|
|
static int n_mem_addr_cells, n_mem_size_cells;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2008-02-01 11:57:31 +07:00
|
|
|
static int __cpuinit fake_numa_create_new_node(unsigned long end_pfn,
|
|
|
|
unsigned int *nid)
|
|
|
|
{
|
|
|
|
unsigned long long mem;
|
|
|
|
char *p = cmdline;
|
|
|
|
static unsigned int fake_nid;
|
|
|
|
static unsigned long long curr_boundary;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Modify node id, iff we started creating NUMA nodes
|
|
|
|
* We want to continue from where we left of the last time
|
|
|
|
*/
|
|
|
|
if (fake_nid)
|
|
|
|
*nid = fake_nid;
|
|
|
|
/*
|
|
|
|
* In case there are no more arguments to parse, the
|
|
|
|
* node_id should be the same as the last fake node id
|
|
|
|
* (we've handled this above).
|
|
|
|
*/
|
|
|
|
if (!p)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
mem = memparse(p, &p);
|
|
|
|
if (!mem)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (mem < curr_boundary)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
curr_boundary = mem;
|
|
|
|
|
|
|
|
if ((end_pfn << PAGE_SHIFT) > mem) {
|
|
|
|
/*
|
|
|
|
* Skip commas and spaces
|
|
|
|
*/
|
|
|
|
while (*p == ',' || *p == ' ' || *p == '\t')
|
|
|
|
p++;
|
|
|
|
|
|
|
|
cmdline = p;
|
|
|
|
fake_nid++;
|
|
|
|
*nid = fake_nid;
|
|
|
|
dbg("created new fake_node with id %d\n", fake_nid);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-03-21 07:35:15 +07:00
|
|
|
static void __cpuinit map_cpu_to_node(int cpu, int node)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
numa_cpu_lookup_table[cpu] = node;
|
2005-11-11 10:22:35 +07:00
|
|
|
|
2006-03-21 07:34:45 +07:00
|
|
|
dbg("adding cpu %d to node %d\n", cpu, node);
|
|
|
|
|
2005-11-11 10:22:35 +07:00
|
|
|
if (!(cpu_isset(cpu, numa_cpumask_lookup_table[node])))
|
2005-04-17 05:20:36 +07:00
|
|
|
cpu_set(cpu, numa_cpumask_lookup_table[node]);
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef CONFIG_HOTPLUG_CPU
|
|
|
|
static void unmap_cpu_from_node(unsigned long cpu)
|
|
|
|
{
|
|
|
|
int node = numa_cpu_lookup_table[cpu];
|
|
|
|
|
|
|
|
dbg("removing cpu %lu from node %d\n", cpu, node);
|
|
|
|
|
|
|
|
if (cpu_isset(cpu, numa_cpumask_lookup_table[node])) {
|
|
|
|
cpu_clear(cpu, numa_cpumask_lookup_table[node]);
|
|
|
|
} else {
|
|
|
|
printk(KERN_ERR "WARNING: cpu %lu not found in node %d\n",
|
|
|
|
cpu, node);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_HOTPLUG_CPU */
|
|
|
|
|
2006-03-21 07:35:15 +07:00
|
|
|
static struct device_node * __cpuinit find_cpu_node(unsigned int cpu)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
unsigned int hw_cpuid = get_hard_smp_processor_id(cpu);
|
|
|
|
struct device_node *cpu_node = NULL;
|
2006-07-12 12:35:54 +07:00
|
|
|
const unsigned int *interrupt_server, *reg;
|
2005-04-17 05:20:36 +07:00
|
|
|
int len;
|
|
|
|
|
|
|
|
while ((cpu_node = of_find_node_by_type(cpu_node, "cpu")) != NULL) {
|
|
|
|
/* Try interrupt server first */
|
2007-04-03 19:26:41 +07:00
|
|
|
interrupt_server = of_get_property(cpu_node,
|
2005-04-17 05:20:36 +07:00
|
|
|
"ibm,ppc-interrupt-server#s", &len);
|
|
|
|
|
|
|
|
len = len / sizeof(u32);
|
|
|
|
|
|
|
|
if (interrupt_server && (len > 0)) {
|
|
|
|
while (len--) {
|
|
|
|
if (interrupt_server[len] == hw_cpuid)
|
|
|
|
return cpu_node;
|
|
|
|
}
|
|
|
|
} else {
|
2007-04-03 19:26:41 +07:00
|
|
|
reg = of_get_property(cpu_node, "reg", &len);
|
2005-04-17 05:20:36 +07:00
|
|
|
if (reg && (len > 0) && (reg[0] == hw_cpuid))
|
|
|
|
return cpu_node;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* must hold reference to node during call */
|
2006-07-12 12:35:54 +07:00
|
|
|
static const int *of_get_associativity(struct device_node *dev)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2007-04-03 19:26:41 +07:00
|
|
|
return of_get_property(dev, "ibm,associativity", NULL);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2006-03-21 07:36:45 +07:00
|
|
|
/* Returns nid in the range [0..MAX_NUMNODES-1], or -1 if no useful numa
|
|
|
|
* info is found.
|
|
|
|
*/
|
2006-05-02 02:16:12 +07:00
|
|
|
static int of_node_to_nid_single(struct device_node *device)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2006-03-21 07:36:45 +07:00
|
|
|
int nid = -1;
|
2006-07-12 12:35:54 +07:00
|
|
|
const unsigned int *tmp;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
if (min_common_depth == -1)
|
2006-03-21 07:36:45 +07:00
|
|
|
goto out;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
tmp = of_get_associativity(device);
|
2006-03-21 07:36:45 +07:00
|
|
|
if (!tmp)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (tmp[0] >= min_common_depth)
|
2006-03-21 07:35:45 +07:00
|
|
|
nid = tmp[min_common_depth];
|
2006-03-21 07:36:15 +07:00
|
|
|
|
|
|
|
/* POWER4 LPAR uses 0xffff as invalid node */
|
2006-03-21 07:36:45 +07:00
|
|
|
if (nid == 0xffff || nid >= MAX_NUMNODES)
|
|
|
|
nid = -1;
|
|
|
|
out:
|
2006-03-21 07:35:45 +07:00
|
|
|
return nid;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2006-05-02 02:16:12 +07:00
|
|
|
/* Walk the device tree upwards, looking for an associativity id */
|
|
|
|
int of_node_to_nid(struct device_node *device)
|
|
|
|
{
|
|
|
|
struct device_node *tmp;
|
|
|
|
int nid = -1;
|
|
|
|
|
|
|
|
of_node_get(device);
|
|
|
|
while (device) {
|
|
|
|
nid = of_node_to_nid_single(device);
|
|
|
|
if (nid != -1)
|
|
|
|
break;
|
|
|
|
|
|
|
|
tmp = device;
|
|
|
|
device = of_get_parent(tmp);
|
|
|
|
of_node_put(tmp);
|
|
|
|
}
|
|
|
|
of_node_put(device);
|
|
|
|
|
|
|
|
return nid;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(of_node_to_nid);
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* In theory, the "ibm,associativity" property may contain multiple
|
|
|
|
* associativity lists because a resource may be multiply connected
|
|
|
|
* into the machine. This resource then has different associativity
|
|
|
|
* characteristics relative to its multiple connections. We ignore
|
|
|
|
* this for now. We also assume that all cpu and memory sets have
|
|
|
|
* their distances represented at a common level. This won't be
|
2007-02-18 01:23:03 +07:00
|
|
|
* true for hierarchical NUMA.
|
2005-04-17 05:20:36 +07:00
|
|
|
*
|
|
|
|
* In any case the ibm,associativity-reference-points should give
|
|
|
|
* the correct depth for a normal NUMA system.
|
|
|
|
*
|
|
|
|
* - Dave Hansen <haveblue@us.ibm.com>
|
|
|
|
*/
|
|
|
|
static int __init find_min_common_depth(void)
|
|
|
|
{
|
|
|
|
int depth;
|
2006-07-12 12:35:54 +07:00
|
|
|
const unsigned int *ref_points;
|
2005-04-17 05:20:36 +07:00
|
|
|
struct device_node *rtas_root;
|
|
|
|
unsigned int len;
|
|
|
|
|
|
|
|
rtas_root = of_find_node_by_path("/rtas");
|
|
|
|
|
|
|
|
if (!rtas_root)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* this property is 2 32-bit integers, each representing a level of
|
|
|
|
* depth in the associativity nodes. The first is for an SMP
|
|
|
|
* configuration (should be all 0's) and the second is for a normal
|
|
|
|
* NUMA configuration.
|
|
|
|
*/
|
2007-04-03 19:26:41 +07:00
|
|
|
ref_points = of_get_property(rtas_root,
|
2005-04-17 05:20:36 +07:00
|
|
|
"ibm,associativity-reference-points", &len);
|
|
|
|
|
|
|
|
if ((len >= 1) && ref_points) {
|
|
|
|
depth = ref_points[1];
|
|
|
|
} else {
|
2006-03-21 07:34:45 +07:00
|
|
|
dbg("NUMA: ibm,associativity-reference-points not found.\n");
|
2005-04-17 05:20:36 +07:00
|
|
|
depth = -1;
|
|
|
|
}
|
|
|
|
of_node_put(rtas_root);
|
|
|
|
|
|
|
|
return depth;
|
|
|
|
}
|
|
|
|
|
2005-12-01 04:47:23 +07:00
|
|
|
static void __init get_n_mem_cells(int *n_addr_cells, int *n_size_cells)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
struct device_node *memory = NULL;
|
|
|
|
|
|
|
|
memory = of_find_node_by_type(memory, "memory");
|
2005-12-05 11:50:39 +07:00
|
|
|
if (!memory)
|
2005-12-01 04:47:23 +07:00
|
|
|
panic("numa.c: No memory nodes found!");
|
2005-12-05 11:50:39 +07:00
|
|
|
|
2007-04-03 07:56:50 +07:00
|
|
|
*n_addr_cells = of_n_addr_cells(memory);
|
2007-04-03 07:57:48 +07:00
|
|
|
*n_size_cells = of_n_size_cells(memory);
|
2005-12-01 04:47:23 +07:00
|
|
|
of_node_put(memory);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2006-07-12 12:35:54 +07:00
|
|
|
static unsigned long __devinit read_n_cells(int n, const unsigned int **buf)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
unsigned long result = 0;
|
|
|
|
|
|
|
|
while (n--) {
|
|
|
|
result = (result << 32) | **buf;
|
|
|
|
(*buf)++;
|
|
|
|
}
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Figure out to which domain a cpu belongs and stick it there.
|
|
|
|
* Return the id of the domain used.
|
|
|
|
*/
|
2006-03-21 07:35:15 +07:00
|
|
|
static int __cpuinit numa_setup_cpu(unsigned long lcpu)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2006-03-21 07:35:45 +07:00
|
|
|
int nid = 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
struct device_node *cpu = find_cpu_node(lcpu);
|
|
|
|
|
|
|
|
if (!cpu) {
|
|
|
|
WARN_ON(1);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2006-05-02 02:16:12 +07:00
|
|
|
nid = of_node_to_nid_single(cpu);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2006-03-21 07:36:45 +07:00
|
|
|
if (nid < 0 || !node_online(nid))
|
|
|
|
nid = any_online_node(NODE_MASK_ALL);
|
2005-04-17 05:20:36 +07:00
|
|
|
out:
|
2006-03-21 07:35:45 +07:00
|
|
|
map_cpu_to_node(lcpu, nid);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
of_node_put(cpu);
|
|
|
|
|
2006-03-21 07:35:45 +07:00
|
|
|
return nid;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2006-06-27 16:54:09 +07:00
|
|
|
static int __cpuinit cpu_numa_callback(struct notifier_block *nfb,
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned long action,
|
|
|
|
void *hcpu)
|
|
|
|
{
|
|
|
|
unsigned long lcpu = (unsigned long)hcpu;
|
|
|
|
int ret = NOTIFY_DONE;
|
|
|
|
|
|
|
|
switch (action) {
|
|
|
|
case CPU_UP_PREPARE:
|
2007-05-09 16:35:10 +07:00
|
|
|
case CPU_UP_PREPARE_FROZEN:
|
2006-03-21 07:37:15 +07:00
|
|
|
numa_setup_cpu(lcpu);
|
2005-04-17 05:20:36 +07:00
|
|
|
ret = NOTIFY_OK;
|
|
|
|
break;
|
|
|
|
#ifdef CONFIG_HOTPLUG_CPU
|
|
|
|
case CPU_DEAD:
|
2007-05-09 16:35:10 +07:00
|
|
|
case CPU_DEAD_FROZEN:
|
2005-04-17 05:20:36 +07:00
|
|
|
case CPU_UP_CANCELED:
|
2007-05-09 16:35:10 +07:00
|
|
|
case CPU_UP_CANCELED_FROZEN:
|
2005-04-17 05:20:36 +07:00
|
|
|
unmap_cpu_from_node(lcpu);
|
|
|
|
break;
|
|
|
|
ret = NOTIFY_OK;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check and possibly modify a memory region to enforce the memory limit.
|
|
|
|
*
|
|
|
|
* Returns the size the region should have to enforce the memory limit.
|
|
|
|
* This will either be the original value of size, a truncated value,
|
|
|
|
* or zero. If the returned value of size is 0 the region should be
|
|
|
|
* discarded as it lies wholy above the memory limit.
|
|
|
|
*/
|
2005-11-11 10:22:35 +07:00
|
|
|
static unsigned long __init numa_enforce_memory_limit(unsigned long start,
|
|
|
|
unsigned long size)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
/*
|
|
|
|
* We use lmb_end_of_DRAM() in here instead of memory_limit because
|
|
|
|
* we've already adjusted it for the limit and it takes care of
|
|
|
|
* having memory holes below the limit.
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (! memory_limit)
|
|
|
|
return size;
|
|
|
|
|
|
|
|
if (start + size <= lmb_end_of_DRAM())
|
|
|
|
return size;
|
|
|
|
|
|
|
|
if (start >= lmb_end_of_DRAM())
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
return lmb_end_of_DRAM() - start;
|
|
|
|
}
|
|
|
|
|
[POWERPC] Support ibm,dynamic-reconfiguration-memory nodes
For PAPR partitions with large amounts of memory, the firmware has an
alternative, more compact representation for the information about the
memory in the partition and its NUMA associativity information. This
adds the code to the kernel to parse this alternative representation.
The other part of this patch is telling the firmware that we can
handle the alternative representation. There is however a subtlety
here, because the firmware will invoke a reboot if the memory
representation we request is different from the representation that
firmware is currently using. This is because firmware can't change
the representation on the fly. Further, some firmware versions used
on POWER5+ machines have a bug where this reboot leaves the machine
with an altered value of load-base, which will prevent any kernel
booting until it is reset to the normal value (0x4000). Because of
this bug, we do NOT set fake_elf.rpanote.new_mem_def = 1, and thus we
do not request the new representation on POWER5+ and earlier machines.
We do request the new representation on POWER6, which uses the
ibm,client-architecture-support call.
Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-11-29 18:27:42 +07:00
|
|
|
/*
|
|
|
|
* Extract NUMA information from the ibm,dynamic-reconfiguration-memory
|
|
|
|
* node. This assumes n_mem_{addr,size}_cells have been set.
|
|
|
|
*/
|
|
|
|
static void __init parse_drconf_memory(struct device_node *memory)
|
|
|
|
{
|
|
|
|
const unsigned int *lm, *dm, *aa;
|
|
|
|
unsigned int ls, ld, la;
|
|
|
|
unsigned int n, aam, aalen;
|
2007-08-01 08:34:38 +07:00
|
|
|
unsigned long lmb_size, size, start;
|
[POWERPC] Support ibm,dynamic-reconfiguration-memory nodes
For PAPR partitions with large amounts of memory, the firmware has an
alternative, more compact representation for the information about the
memory in the partition and its NUMA associativity information. This
adds the code to the kernel to parse this alternative representation.
The other part of this patch is telling the firmware that we can
handle the alternative representation. There is however a subtlety
here, because the firmware will invoke a reboot if the memory
representation we request is different from the representation that
firmware is currently using. This is because firmware can't change
the representation on the fly. Further, some firmware versions used
on POWER5+ machines have a bug where this reboot leaves the machine
with an altered value of load-base, which will prevent any kernel
booting until it is reset to the normal value (0x4000). Because of
this bug, we do NOT set fake_elf.rpanote.new_mem_def = 1, and thus we
do not request the new representation on POWER5+ and earlier machines.
We do request the new representation on POWER6, which uses the
ibm,client-architecture-support call.
Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-11-29 18:27:42 +07:00
|
|
|
int nid, default_nid = 0;
|
2007-08-01 08:34:38 +07:00
|
|
|
unsigned int ai, flags;
|
[POWERPC] Support ibm,dynamic-reconfiguration-memory nodes
For PAPR partitions with large amounts of memory, the firmware has an
alternative, more compact representation for the information about the
memory in the partition and its NUMA associativity information. This
adds the code to the kernel to parse this alternative representation.
The other part of this patch is telling the firmware that we can
handle the alternative representation. There is however a subtlety
here, because the firmware will invoke a reboot if the memory
representation we request is different from the representation that
firmware is currently using. This is because firmware can't change
the representation on the fly. Further, some firmware versions used
on POWER5+ machines have a bug where this reboot leaves the machine
with an altered value of load-base, which will prevent any kernel
booting until it is reset to the normal value (0x4000). Because of
this bug, we do NOT set fake_elf.rpanote.new_mem_def = 1, and thus we
do not request the new representation on POWER5+ and earlier machines.
We do request the new representation on POWER6, which uses the
ibm,client-architecture-support call.
Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-11-29 18:27:42 +07:00
|
|
|
|
2007-04-03 19:26:41 +07:00
|
|
|
lm = of_get_property(memory, "ibm,lmb-size", &ls);
|
|
|
|
dm = of_get_property(memory, "ibm,dynamic-memory", &ld);
|
|
|
|
aa = of_get_property(memory, "ibm,associativity-lookup-arrays", &la);
|
[POWERPC] Support ibm,dynamic-reconfiguration-memory nodes
For PAPR partitions with large amounts of memory, the firmware has an
alternative, more compact representation for the information about the
memory in the partition and its NUMA associativity information. This
adds the code to the kernel to parse this alternative representation.
The other part of this patch is telling the firmware that we can
handle the alternative representation. There is however a subtlety
here, because the firmware will invoke a reboot if the memory
representation we request is different from the representation that
firmware is currently using. This is because firmware can't change
the representation on the fly. Further, some firmware versions used
on POWER5+ machines have a bug where this reboot leaves the machine
with an altered value of load-base, which will prevent any kernel
booting until it is reset to the normal value (0x4000). Because of
this bug, we do NOT set fake_elf.rpanote.new_mem_def = 1, and thus we
do not request the new representation on POWER5+ and earlier machines.
We do request the new representation on POWER6, which uses the
ibm,client-architecture-support call.
Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-11-29 18:27:42 +07:00
|
|
|
if (!lm || !dm || !aa ||
|
|
|
|
ls < sizeof(unsigned int) || ld < sizeof(unsigned int) ||
|
|
|
|
la < 2 * sizeof(unsigned int))
|
|
|
|
return;
|
|
|
|
|
|
|
|
lmb_size = read_n_cells(n_mem_size_cells, &lm);
|
|
|
|
n = *dm++; /* number of LMBs */
|
|
|
|
aam = *aa++; /* number of associativity lists */
|
|
|
|
aalen = *aa++; /* length of each associativity list */
|
|
|
|
if (ld < (n * (n_mem_addr_cells + 4) + 1) * sizeof(unsigned int) ||
|
|
|
|
la < (aam * aalen + 2) * sizeof(unsigned int))
|
|
|
|
return;
|
|
|
|
|
|
|
|
for (; n != 0; --n) {
|
|
|
|
start = read_n_cells(n_mem_addr_cells, &dm);
|
|
|
|
ai = dm[2];
|
|
|
|
flags = dm[3];
|
|
|
|
dm += 4;
|
|
|
|
/* 0x80 == reserved, 0x8 = assigned to us */
|
|
|
|
if ((flags & 0x80) || !(flags & 0x8))
|
|
|
|
continue;
|
|
|
|
nid = default_nid;
|
|
|
|
/* flags & 0x40 means associativity index is invalid */
|
|
|
|
if (min_common_depth > 0 && min_common_depth <= aalen &&
|
|
|
|
(flags & 0x40) == 0 && ai < aam) {
|
|
|
|
/* this is like of_node_to_nid_single */
|
|
|
|
nid = aa[ai * aalen + min_common_depth - 1];
|
|
|
|
if (nid == 0xffff || nid >= MAX_NUMNODES)
|
|
|
|
nid = default_nid;
|
|
|
|
}
|
2008-02-01 11:57:31 +07:00
|
|
|
|
|
|
|
fake_numa_create_new_node(((start + lmb_size) >> PAGE_SHIFT),
|
|
|
|
&nid);
|
[POWERPC] Support ibm,dynamic-reconfiguration-memory nodes
For PAPR partitions with large amounts of memory, the firmware has an
alternative, more compact representation for the information about the
memory in the partition and its NUMA associativity information. This
adds the code to the kernel to parse this alternative representation.
The other part of this patch is telling the firmware that we can
handle the alternative representation. There is however a subtlety
here, because the firmware will invoke a reboot if the memory
representation we request is different from the representation that
firmware is currently using. This is because firmware can't change
the representation on the fly. Further, some firmware versions used
on POWER5+ machines have a bug where this reboot leaves the machine
with an altered value of load-base, which will prevent any kernel
booting until it is reset to the normal value (0x4000). Because of
this bug, we do NOT set fake_elf.rpanote.new_mem_def = 1, and thus we
do not request the new representation on POWER5+ and earlier machines.
We do request the new representation on POWER6, which uses the
ibm,client-architecture-support call.
Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-11-29 18:27:42 +07:00
|
|
|
node_set_online(nid);
|
|
|
|
|
|
|
|
size = numa_enforce_memory_limit(start, lmb_size);
|
|
|
|
if (!size)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
add_active_range(nid, start >> PAGE_SHIFT,
|
|
|
|
(start >> PAGE_SHIFT) + (size >> PAGE_SHIFT));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
static int __init parse_numa_properties(void)
|
|
|
|
{
|
|
|
|
struct device_node *cpu = NULL;
|
|
|
|
struct device_node *memory = NULL;
|
2006-03-21 07:36:45 +07:00
|
|
|
int default_nid = 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned long i;
|
|
|
|
|
|
|
|
if (numa_enabled == 0) {
|
|
|
|
printk(KERN_WARNING "NUMA disabled by user\n");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
min_common_depth = find_min_common_depth();
|
|
|
|
|
|
|
|
if (min_common_depth < 0)
|
|
|
|
return min_common_depth;
|
|
|
|
|
2006-03-21 07:34:45 +07:00
|
|
|
dbg("NUMA associativity depth for CPU/Memory: %d\n", min_common_depth);
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
2006-03-21 07:36:45 +07:00
|
|
|
* Even though we connect cpus to numa domains later in SMP
|
|
|
|
* init, we need to know the node ids now. This is because
|
|
|
|
* each node to be onlined must have NODE_DATA etc backing it.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
2006-03-21 07:36:45 +07:00
|
|
|
for_each_present_cpu(i) {
|
2006-03-21 07:35:45 +07:00
|
|
|
int nid;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
cpu = find_cpu_node(i);
|
2006-03-21 07:36:45 +07:00
|
|
|
BUG_ON(!cpu);
|
2006-05-02 02:16:12 +07:00
|
|
|
nid = of_node_to_nid_single(cpu);
|
2006-03-21 07:36:45 +07:00
|
|
|
of_node_put(cpu);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2006-03-21 07:36:45 +07:00
|
|
|
/*
|
|
|
|
* Don't fall back to default_nid yet -- we will plug
|
|
|
|
* cpus into nodes once the memory scan has discovered
|
|
|
|
* the topology.
|
|
|
|
*/
|
|
|
|
if (nid < 0)
|
|
|
|
continue;
|
|
|
|
node_set_online(nid);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2005-12-06 03:06:42 +07:00
|
|
|
get_n_mem_cells(&n_mem_addr_cells, &n_mem_size_cells);
|
2005-04-17 05:20:36 +07:00
|
|
|
memory = NULL;
|
|
|
|
while ((memory = of_find_node_by_type(memory, "memory")) != NULL) {
|
|
|
|
unsigned long start;
|
|
|
|
unsigned long size;
|
2006-03-21 07:35:45 +07:00
|
|
|
int nid;
|
2005-04-17 05:20:36 +07:00
|
|
|
int ranges;
|
2006-07-12 12:35:54 +07:00
|
|
|
const unsigned int *memcell_buf;
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned int len;
|
|
|
|
|
2007-04-03 19:26:41 +07:00
|
|
|
memcell_buf = of_get_property(memory,
|
2005-12-04 14:39:55 +07:00
|
|
|
"linux,usable-memory", &len);
|
|
|
|
if (!memcell_buf || len <= 0)
|
2007-04-03 19:26:41 +07:00
|
|
|
memcell_buf = of_get_property(memory, "reg", &len);
|
2005-04-17 05:20:36 +07:00
|
|
|
if (!memcell_buf || len <= 0)
|
|
|
|
continue;
|
|
|
|
|
2005-12-13 14:01:21 +07:00
|
|
|
/* ranges in cell */
|
|
|
|
ranges = (len >> 2) / (n_mem_addr_cells + n_mem_size_cells);
|
2005-04-17 05:20:36 +07:00
|
|
|
new_range:
|
|
|
|
/* these are order-sensitive, and modify the buffer pointer */
|
2005-12-06 03:06:42 +07:00
|
|
|
start = read_n_cells(n_mem_addr_cells, &memcell_buf);
|
|
|
|
size = read_n_cells(n_mem_size_cells, &memcell_buf);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2006-03-21 07:36:45 +07:00
|
|
|
/*
|
|
|
|
* Assumption: either all memory nodes or none will
|
|
|
|
* have associativity properties. If none, then
|
|
|
|
* everything goes to default_nid.
|
|
|
|
*/
|
2006-05-02 02:16:12 +07:00
|
|
|
nid = of_node_to_nid_single(memory);
|
2006-03-21 07:36:45 +07:00
|
|
|
if (nid < 0)
|
|
|
|
nid = default_nid;
|
2008-02-01 11:57:31 +07:00
|
|
|
|
|
|
|
fake_numa_create_new_node(((start + size) >> PAGE_SHIFT), &nid);
|
2006-03-21 07:36:45 +07:00
|
|
|
node_set_online(nid);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2005-11-11 10:22:35 +07:00
|
|
|
if (!(size = numa_enforce_memory_limit(start, size))) {
|
2005-04-17 05:20:36 +07:00
|
|
|
if (--ranges)
|
|
|
|
goto new_range;
|
|
|
|
else
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2006-09-27 15:49:49 +07:00
|
|
|
add_active_range(nid, start >> PAGE_SHIFT,
|
|
|
|
(start >> PAGE_SHIFT) + (size >> PAGE_SHIFT));
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
if (--ranges)
|
|
|
|
goto new_range;
|
|
|
|
}
|
|
|
|
|
[POWERPC] Support ibm,dynamic-reconfiguration-memory nodes
For PAPR partitions with large amounts of memory, the firmware has an
alternative, more compact representation for the information about the
memory in the partition and its NUMA associativity information. This
adds the code to the kernel to parse this alternative representation.
The other part of this patch is telling the firmware that we can
handle the alternative representation. There is however a subtlety
here, because the firmware will invoke a reboot if the memory
representation we request is different from the representation that
firmware is currently using. This is because firmware can't change
the representation on the fly. Further, some firmware versions used
on POWER5+ machines have a bug where this reboot leaves the machine
with an altered value of load-base, which will prevent any kernel
booting until it is reset to the normal value (0x4000). Because of
this bug, we do NOT set fake_elf.rpanote.new_mem_def = 1, and thus we
do not request the new representation on POWER5+ and earlier machines.
We do request the new representation on POWER6, which uses the
ibm,client-architecture-support call.
Signed-off-by: Paul Mackerras <paulus@samba.org>
2006-11-29 18:27:42 +07:00
|
|
|
/*
|
|
|
|
* Now do the same thing for each LMB listed in the ibm,dynamic-memory
|
|
|
|
* property in the ibm,dynamic-reconfiguration-memory node.
|
|
|
|
*/
|
|
|
|
memory = of_find_node_by_path("/ibm,dynamic-reconfiguration-memory");
|
|
|
|
if (memory)
|
|
|
|
parse_drconf_memory(memory);
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __init setup_nonnuma(void)
|
|
|
|
{
|
|
|
|
unsigned long top_of_ram = lmb_end_of_DRAM();
|
|
|
|
unsigned long total_ram = lmb_phys_mem_size();
|
2006-09-27 15:49:49 +07:00
|
|
|
unsigned long start_pfn, end_pfn;
|
2008-02-01 11:57:31 +07:00
|
|
|
unsigned int i, nid = 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2006-04-13 03:25:01 +07:00
|
|
|
printk(KERN_DEBUG "Top of RAM: 0x%lx, Total RAM: 0x%lx\n",
|
2005-04-17 05:20:36 +07:00
|
|
|
top_of_ram, total_ram);
|
2006-04-13 03:25:01 +07:00
|
|
|
printk(KERN_DEBUG "Memory hole size: %ldMB\n",
|
2005-04-17 05:20:36 +07:00
|
|
|
(top_of_ram - total_ram) >> 20);
|
|
|
|
|
2006-09-27 15:49:49 +07:00
|
|
|
for (i = 0; i < lmb.memory.cnt; ++i) {
|
|
|
|
start_pfn = lmb.memory.region[i].base >> PAGE_SHIFT;
|
|
|
|
end_pfn = start_pfn + lmb_size_pages(&lmb.memory, i);
|
2008-02-01 11:57:31 +07:00
|
|
|
|
|
|
|
fake_numa_create_new_node(end_pfn, &nid);
|
|
|
|
add_active_range(nid, start_pfn, end_pfn);
|
|
|
|
node_set_online(nid);
|
2006-09-27 15:49:49 +07:00
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2005-12-13 02:56:47 +07:00
|
|
|
void __init dump_numa_cpu_topology(void)
|
|
|
|
{
|
|
|
|
unsigned int node;
|
|
|
|
unsigned int cpu, count;
|
|
|
|
|
|
|
|
if (min_common_depth == -1 || !numa_enabled)
|
|
|
|
return;
|
|
|
|
|
|
|
|
for_each_online_node(node) {
|
2006-04-13 03:25:01 +07:00
|
|
|
printk(KERN_DEBUG "Node %d CPUs:", node);
|
2005-12-13 02:56:47 +07:00
|
|
|
|
|
|
|
count = 0;
|
|
|
|
/*
|
|
|
|
* If we used a CPU iterator here we would miss printing
|
|
|
|
* the holes in the cpumap.
|
|
|
|
*/
|
|
|
|
for (cpu = 0; cpu < NR_CPUS; cpu++) {
|
|
|
|
if (cpu_isset(cpu, numa_cpumask_lookup_table[node])) {
|
|
|
|
if (count == 0)
|
|
|
|
printk(" %u", cpu);
|
|
|
|
++count;
|
|
|
|
} else {
|
|
|
|
if (count > 1)
|
|
|
|
printk("-%u", cpu - 1);
|
|
|
|
count = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (count > 1)
|
|
|
|
printk("-%u", NR_CPUS - 1);
|
|
|
|
printk("\n");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __init dump_numa_memory_topology(void)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
unsigned int node;
|
|
|
|
unsigned int count;
|
|
|
|
|
|
|
|
if (min_common_depth == -1 || !numa_enabled)
|
|
|
|
return;
|
|
|
|
|
|
|
|
for_each_online_node(node) {
|
|
|
|
unsigned long i;
|
|
|
|
|
2006-04-13 03:25:01 +07:00
|
|
|
printk(KERN_DEBUG "Node %d Memory:", node);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
count = 0;
|
|
|
|
|
2005-11-11 10:22:35 +07:00
|
|
|
for (i = 0; i < lmb_end_of_DRAM();
|
|
|
|
i += (1 << SECTION_SIZE_BITS)) {
|
|
|
|
if (early_pfn_to_nid(i >> PAGE_SHIFT) == node) {
|
2005-04-17 05:20:36 +07:00
|
|
|
if (count == 0)
|
|
|
|
printk(" 0x%lx", i);
|
|
|
|
++count;
|
|
|
|
} else {
|
|
|
|
if (count > 0)
|
|
|
|
printk("-0x%lx", i);
|
|
|
|
count = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (count > 0)
|
|
|
|
printk("-0x%lx", i);
|
|
|
|
printk("\n");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocate some memory, satisfying the lmb or bootmem allocator where
|
|
|
|
* required. nid is the preferred node and end is the physical address of
|
|
|
|
* the highest address in the node.
|
|
|
|
*
|
|
|
|
* Returns the physical address of the memory.
|
|
|
|
*/
|
2005-11-11 10:22:35 +07:00
|
|
|
static void __init *careful_allocation(int nid, unsigned long size,
|
|
|
|
unsigned long align,
|
|
|
|
unsigned long end_pfn)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2005-11-11 10:22:35 +07:00
|
|
|
int new_nid;
|
2006-01-25 15:31:28 +07:00
|
|
|
unsigned long ret = __lmb_alloc_base(size, align, end_pfn << PAGE_SHIFT);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/* retry over all memory */
|
|
|
|
if (!ret)
|
2006-01-25 15:31:28 +07:00
|
|
|
ret = __lmb_alloc_base(size, align, lmb_end_of_DRAM());
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
if (!ret)
|
|
|
|
panic("numa.c: cannot allocate %lu bytes on node %d",
|
|
|
|
size, nid);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the memory came from a previously allocated node, we must
|
|
|
|
* retry with the bootmem allocator.
|
|
|
|
*/
|
2005-11-11 10:22:35 +07:00
|
|
|
new_nid = early_pfn_to_nid(ret >> PAGE_SHIFT);
|
|
|
|
if (new_nid < nid) {
|
|
|
|
ret = (unsigned long)__alloc_bootmem_node(NODE_DATA(new_nid),
|
2005-04-17 05:20:36 +07:00
|
|
|
size, align, 0);
|
|
|
|
|
|
|
|
if (!ret)
|
|
|
|
panic("numa.c: cannot allocate %lu bytes on node %d",
|
2005-11-11 10:22:35 +07:00
|
|
|
size, new_nid);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2005-11-11 10:22:35 +07:00
|
|
|
ret = __pa(ret);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
dbg("alloc_bootmem %lx %lx\n", ret, size);
|
|
|
|
}
|
|
|
|
|
2005-11-11 10:22:35 +07:00
|
|
|
return (void *)ret;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2006-06-27 16:54:09 +07:00
|
|
|
static struct notifier_block __cpuinitdata ppc64_numa_nb = {
|
|
|
|
.notifier_call = cpu_numa_callback,
|
|
|
|
.priority = 1 /* Must run before sched domains notifier. */
|
|
|
|
};
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
void __init do_init_bootmem(void)
|
|
|
|
{
|
|
|
|
int nid;
|
2005-11-11 10:22:35 +07:00
|
|
|
unsigned int i;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
min_low_pfn = 0;
|
|
|
|
max_low_pfn = lmb_end_of_DRAM() >> PAGE_SHIFT;
|
|
|
|
max_pfn = max_low_pfn;
|
|
|
|
|
|
|
|
if (parse_numa_properties())
|
|
|
|
setup_nonnuma();
|
|
|
|
else
|
2005-12-13 02:56:47 +07:00
|
|
|
dump_numa_memory_topology();
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
register_cpu_notifier(&ppc64_numa_nb);
|
2006-03-21 07:37:15 +07:00
|
|
|
cpu_numa_callback(&ppc64_numa_nb, CPU_UP_PREPARE,
|
|
|
|
(void *)(unsigned long)boot_cpuid);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
for_each_online_node(nid) {
|
2006-09-27 15:49:49 +07:00
|
|
|
unsigned long start_pfn, end_pfn;
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned long bootmem_paddr;
|
|
|
|
unsigned long bootmap_pages;
|
|
|
|
|
2006-09-27 15:49:49 +07:00
|
|
|
get_pfn_range_for_nid(nid, &start_pfn, &end_pfn);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/* Allocate the node structure node local if possible */
|
2005-11-11 10:22:35 +07:00
|
|
|
NODE_DATA(nid) = careful_allocation(nid,
|
2005-04-17 05:20:36 +07:00
|
|
|
sizeof(struct pglist_data),
|
2005-11-11 10:22:35 +07:00
|
|
|
SMP_CACHE_BYTES, end_pfn);
|
|
|
|
NODE_DATA(nid) = __va(NODE_DATA(nid));
|
2005-04-17 05:20:36 +07:00
|
|
|
memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
|
|
|
|
|
|
|
|
dbg("node %d\n", nid);
|
|
|
|
dbg("NODE_DATA() = %p\n", NODE_DATA(nid));
|
|
|
|
|
|
|
|
NODE_DATA(nid)->bdata = &plat_node_bdata[nid];
|
2005-11-11 10:22:35 +07:00
|
|
|
NODE_DATA(nid)->node_start_pfn = start_pfn;
|
|
|
|
NODE_DATA(nid)->node_spanned_pages = end_pfn - start_pfn;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
if (NODE_DATA(nid)->node_spanned_pages == 0)
|
|
|
|
continue;
|
|
|
|
|
2005-11-11 10:22:35 +07:00
|
|
|
dbg("start_paddr = %lx\n", start_pfn << PAGE_SHIFT);
|
|
|
|
dbg("end_paddr = %lx\n", end_pfn << PAGE_SHIFT);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2005-11-11 10:22:35 +07:00
|
|
|
bootmap_pages = bootmem_bootmap_pages(end_pfn - start_pfn);
|
|
|
|
bootmem_paddr = (unsigned long)careful_allocation(nid,
|
|
|
|
bootmap_pages << PAGE_SHIFT,
|
|
|
|
PAGE_SIZE, end_pfn);
|
|
|
|
memset(__va(bootmem_paddr), 0, bootmap_pages << PAGE_SHIFT);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
dbg("bootmap_paddr = %lx\n", bootmem_paddr);
|
|
|
|
|
|
|
|
init_bootmem_node(NODE_DATA(nid), bootmem_paddr >> PAGE_SHIFT,
|
2005-11-11 10:22:35 +07:00
|
|
|
start_pfn, end_pfn);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2006-09-27 15:49:49 +07:00
|
|
|
free_bootmem_with_active_regions(nid, end_pfn);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2005-11-11 10:22:35 +07:00
|
|
|
/* Mark reserved regions on this node */
|
2005-04-17 05:20:36 +07:00
|
|
|
for (i = 0; i < lmb.reserved.cnt; i++) {
|
2005-08-03 17:21:26 +07:00
|
|
|
unsigned long physbase = lmb.reserved.region[i].base;
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned long size = lmb.reserved.region[i].size;
|
2005-11-11 10:22:35 +07:00
|
|
|
unsigned long start_paddr = start_pfn << PAGE_SHIFT;
|
|
|
|
unsigned long end_paddr = end_pfn << PAGE_SHIFT;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2005-11-11 10:22:35 +07:00
|
|
|
if (early_pfn_to_nid(physbase >> PAGE_SHIFT) != nid &&
|
|
|
|
early_pfn_to_nid((physbase+size-1) >> PAGE_SHIFT) != nid)
|
2005-04-17 05:20:36 +07:00
|
|
|
continue;
|
|
|
|
|
|
|
|
if (physbase < end_paddr &&
|
|
|
|
(physbase+size) > start_paddr) {
|
|
|
|
/* overlaps */
|
|
|
|
if (physbase < start_paddr) {
|
|
|
|
size -= start_paddr - physbase;
|
|
|
|
physbase = start_paddr;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (size > end_paddr - physbase)
|
|
|
|
size = end_paddr - physbase;
|
|
|
|
|
|
|
|
dbg("reserve_bootmem %lx %lx\n", physbase,
|
|
|
|
size);
|
|
|
|
reserve_bootmem_node(NODE_DATA(nid), physbase,
|
2008-02-07 15:15:17 +07:00
|
|
|
size, BOOTMEM_DEFAULT);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
}
|
2005-09-04 05:54:26 +07:00
|
|
|
|
2006-09-27 15:49:49 +07:00
|
|
|
sparse_memory_present_with_active_regions(nid);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void __init paging_init(void)
|
|
|
|
{
|
2006-10-11 15:20:39 +07:00
|
|
|
unsigned long max_zone_pfns[MAX_NR_ZONES];
|
|
|
|
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
|
|
|
|
max_zone_pfns[ZONE_DMA] = lmb_end_of_DRAM() >> PAGE_SHIFT;
|
2006-09-27 15:49:49 +07:00
|
|
|
free_area_init_nodes(max_zone_pfns);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
static int __init early_numa(char *p)
|
|
|
|
{
|
|
|
|
if (!p)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (strstr(p, "off"))
|
|
|
|
numa_enabled = 0;
|
|
|
|
|
|
|
|
if (strstr(p, "debug"))
|
|
|
|
numa_debug = 1;
|
|
|
|
|
2008-02-01 11:57:31 +07:00
|
|
|
p = strstr(p, "fake=");
|
|
|
|
if (p)
|
|
|
|
cmdline = p + strlen("fake=");
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
early_param("numa", early_numa);
|
2005-12-06 03:06:42 +07:00
|
|
|
|
|
|
|
#ifdef CONFIG_MEMORY_HOTPLUG
|
|
|
|
/*
|
|
|
|
* Find the node associated with a hot added memory section. Section
|
|
|
|
* corresponds to a SPARSEMEM section, not an LMB. It is assumed that
|
|
|
|
* sections are fully contained within a single LMB.
|
|
|
|
*/
|
|
|
|
int hot_add_scn_to_nid(unsigned long scn_addr)
|
|
|
|
{
|
|
|
|
struct device_node *memory = NULL;
|
2005-12-17 05:30:35 +07:00
|
|
|
nodemask_t nodes;
|
2006-03-21 07:36:45 +07:00
|
|
|
int default_nid = any_online_node(NODE_MASK_ALL);
|
2006-03-24 17:34:46 +07:00
|
|
|
int nid;
|
2005-12-06 03:06:42 +07:00
|
|
|
|
|
|
|
if (!numa_enabled || (min_common_depth < 0))
|
2006-03-21 07:36:45 +07:00
|
|
|
return default_nid;
|
2005-12-06 03:06:42 +07:00
|
|
|
|
|
|
|
while ((memory = of_find_node_by_type(memory, "memory")) != NULL) {
|
|
|
|
unsigned long start, size;
|
2005-12-17 05:30:35 +07:00
|
|
|
int ranges;
|
2006-07-12 12:35:54 +07:00
|
|
|
const unsigned int *memcell_buf;
|
2005-12-06 03:06:42 +07:00
|
|
|
unsigned int len;
|
|
|
|
|
2007-04-03 19:26:41 +07:00
|
|
|
memcell_buf = of_get_property(memory, "reg", &len);
|
2005-12-06 03:06:42 +07:00
|
|
|
if (!memcell_buf || len <= 0)
|
|
|
|
continue;
|
|
|
|
|
2005-12-13 14:01:21 +07:00
|
|
|
/* ranges in cell */
|
|
|
|
ranges = (len >> 2) / (n_mem_addr_cells + n_mem_size_cells);
|
2005-12-06 03:06:42 +07:00
|
|
|
ha_new_range:
|
|
|
|
start = read_n_cells(n_mem_addr_cells, &memcell_buf);
|
|
|
|
size = read_n_cells(n_mem_size_cells, &memcell_buf);
|
2006-05-02 02:16:12 +07:00
|
|
|
nid = of_node_to_nid_single(memory);
|
2005-12-06 03:06:42 +07:00
|
|
|
|
|
|
|
/* Domains not present at boot default to 0 */
|
2006-03-21 07:36:45 +07:00
|
|
|
if (nid < 0 || !node_online(nid))
|
|
|
|
nid = default_nid;
|
2005-12-06 03:06:42 +07:00
|
|
|
|
|
|
|
if ((scn_addr >= start) && (scn_addr < (start + size))) {
|
|
|
|
of_node_put(memory);
|
2006-03-21 07:35:45 +07:00
|
|
|
goto got_nid;
|
2005-12-06 03:06:42 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
if (--ranges) /* process all ranges in cell */
|
|
|
|
goto ha_new_range;
|
|
|
|
}
|
|
|
|
BUG(); /* section address should be found above */
|
2006-03-24 17:34:46 +07:00
|
|
|
return 0;
|
2005-12-17 05:30:35 +07:00
|
|
|
|
|
|
|
/* Temporary code to ensure that returned node is not empty */
|
2006-03-21 07:35:45 +07:00
|
|
|
got_nid:
|
2005-12-17 05:30:35 +07:00
|
|
|
nodes_setall(nodes);
|
2006-03-21 07:35:45 +07:00
|
|
|
while (NODE_DATA(nid)->node_spanned_pages == 0) {
|
|
|
|
node_clear(nid, nodes);
|
|
|
|
nid = any_online_node(nodes);
|
2005-12-17 05:30:35 +07:00
|
|
|
}
|
2006-03-21 07:35:45 +07:00
|
|
|
return nid;
|
2005-12-06 03:06:42 +07:00
|
|
|
}
|
|
|
|
#endif /* CONFIG_MEMORY_HOTPLUG */
|