2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* 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>
|
2010-07-12 04:36:09 +00:00
|
|
|
#include <linux/memblock.h>
|
2008-02-14 00:37:49 +00:00
|
|
|
#include <linux/of.h>
|
powerpc/mm: Fix numa reserve bootmem page selection
Fix the powerpc NUMA reserve bootmem page selection logic.
commit 8f64e1f2d1e09267ac926e15090fd505c1c0cbcb (powerpc: Reserve
in bootmem lmb reserved regions that cross NUMA nodes) changed
the logic for how the powerpc LMB reserved regions were converted
to bootmen reserved regions. As the folowing discussion reports,
the new logic was not correct.
mark_reserved_regions_for_nid() goes through each LMB on the
system that specifies a reserved area. It searches for
active regions that intersect with that LMB and are on the
specified node. It attempts to bootmem-reserve only the area
where the active region and the reserved LMB intersect. We
can not reserve things on other nodes as they may not have
bootmem structures allocated, yet.
We base the size of the bootmem reservation on two possible
things. Normally, we just make the reservation start and
stop exactly at the start and end of the LMB.
However, the LMB reservations are not aware of NUMA nodes and
on occasion a single LMB may cross into several adjacent
active regions. Those may even be on different NUMA nodes
and will require separate calls to the bootmem reserve
functions. So, the bootmem reservation must be trimmed to
fit inside the current active region.
That's all fine and dandy, but we trim the reservation
in a page-aligned fashion. That's bad because we start the
reservation at a non-page-aligned address: physbase.
The reservation may only span 2 bytes, but that those bytes
may span two pfns and cause a reserve_size of 2*PAGE_SIZE.
Take the case where you reserve 0x2 bytes at 0x0fff and
where the active region ends at 0x1000. You'll jump into
that if() statment, but node_ar.end_pfn=0x1 and
start_pfn=0x0. You'll end up with a reserve_size=0x1000,
and then call
reserve_bootmem_node(node, physbase=0xfff, size=0x1000);
0x1000 may not be on the same node as 0xfff. Oops.
In almost all the vm code, end_<anything> is not inclusive.
If you have an end_pfn of 0x1234, page 0x1234 is not
included in the range. Using PFN_UP instead of the
(>> >> PAGE_SHIFT) will make this consistent with the other VM
code.
We also need to do math for the reserved size with physbase
instead of start_pfn. node_ar.end_pfn << PAGE_SHIFT is
*precisely* the end of the node. However,
(start_pfn << PAGE_SHIFT) is *NOT* precisely the beginning
of the reserved area. That is, of course, physbase.
If we don't use physbase here, the reserve_size can be
made too large.
From: Dave Hansen <dave@linux.vnet.ibm.com>
Tested-by: Geoff Levand <geoffrey.levand@am.sony.com> Tested on PS3.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2009-02-12 12:36:04 +00:00
|
|
|
#include <linux/pfn.h>
|
2010-12-01 12:31:15 +00:00
|
|
|
#include <linux/cpuset.h>
|
|
|
|
#include <linux/node.h>
|
2005-11-11 03:22:35 +00:00
|
|
|
#include <asm/sparsemem.h>
|
2008-02-14 00:56:49 +00:00
|
|
|
#include <asm/prom.h>
|
2005-10-31 02:07:02 +00:00
|
|
|
#include <asm/system.h>
|
2005-11-07 02:18:13 +00:00
|
|
|
#include <asm/smp.h>
|
2010-12-01 12:31:15 +00:00
|
|
|
#include <asm/firmware.h>
|
|
|
|
#include <asm/paca.h>
|
2010-12-17 22:07:47 +00:00
|
|
|
#include <asm/hvcall.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
static int numa_enabled = 1;
|
|
|
|
|
2008-02-01 04:57:31 +00:00
|
|
|
static char *cmdline __initdata;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
static int numa_debug;
|
|
|
|
#define dbg(args...) if (numa_debug) { printk(KERN_INFO args); }
|
|
|
|
|
2005-11-11 03:22:35 +00:00
|
|
|
int numa_cpu_lookup_table[NR_CPUS];
|
2010-04-26 15:32:43 +00:00
|
|
|
cpumask_var_t node_to_cpumask_map[MAX_NUMNODES];
|
2005-04-16 22:20:36 +00:00
|
|
|
struct pglist_data *node_data[MAX_NUMNODES];
|
2005-11-11 03:22:35 +00:00
|
|
|
|
|
|
|
EXPORT_SYMBOL(numa_cpu_lookup_table);
|
2010-04-26 15:32:43 +00:00
|
|
|
EXPORT_SYMBOL(node_to_cpumask_map);
|
2005-11-11 03:22:35 +00:00
|
|
|
EXPORT_SYMBOL(node_data);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
static int min_common_depth;
|
2005-12-05 20:06:42 +00:00
|
|
|
static int n_mem_addr_cells, n_mem_size_cells;
|
2010-05-16 20:22:31 +00:00
|
|
|
static int form1_affinity;
|
|
|
|
|
|
|
|
#define MAX_DISTANCE_REF_POINTS 4
|
|
|
|
static int distance_ref_points_depth;
|
|
|
|
static const unsigned int *distance_ref_points;
|
|
|
|
static int distance_lookup_table[MAX_NUMNODES][MAX_DISTANCE_REF_POINTS];
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2010-04-26 15:32:43 +00:00
|
|
|
/*
|
|
|
|
* Allocate node_to_cpumask_map based on number of available nodes
|
|
|
|
* Requires node_possible_map to be valid.
|
|
|
|
*
|
|
|
|
* Note: node_to_cpumask() is not valid until after this is done.
|
|
|
|
*/
|
|
|
|
static void __init setup_node_to_cpumask_map(void)
|
|
|
|
{
|
|
|
|
unsigned int node, num = 0;
|
|
|
|
|
|
|
|
/* setup nr_node_ids if not done yet */
|
|
|
|
if (nr_node_ids == MAX_NUMNODES) {
|
|
|
|
for_each_node_mask(node, node_possible_map)
|
|
|
|
num = node;
|
|
|
|
nr_node_ids = num + 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* allocate the map */
|
|
|
|
for (node = 0; node < nr_node_ids; node++)
|
|
|
|
alloc_bootmem_cpumask_var(&node_to_cpumask_map[node]);
|
|
|
|
|
|
|
|
/* cpumask_of_node() will now work */
|
|
|
|
dbg("Node to cpumask map for %d nodes\n", nr_node_ids);
|
|
|
|
}
|
|
|
|
|
2008-02-01 04:57:31 +00:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2008-10-09 10:18:40 +00:00
|
|
|
/*
|
|
|
|
* get_active_region_work_fn - A helper function for get_node_active_region
|
|
|
|
* Returns datax set to the start_pfn and end_pfn if they contain
|
|
|
|
* the initial value of datax->start_pfn between them
|
|
|
|
* @start_pfn: start page(inclusive) of region to check
|
|
|
|
* @end_pfn: end page(exclusive) of region to check
|
|
|
|
* @datax: comes in with ->start_pfn set to value to search for and
|
|
|
|
* goes out with active range if it contains it
|
|
|
|
* Returns 1 if search value is in range else 0
|
|
|
|
*/
|
|
|
|
static int __init get_active_region_work_fn(unsigned long start_pfn,
|
|
|
|
unsigned long end_pfn, void *datax)
|
|
|
|
{
|
|
|
|
struct node_active_region *data;
|
|
|
|
data = (struct node_active_region *)datax;
|
|
|
|
|
|
|
|
if (start_pfn <= data->start_pfn && end_pfn > data->start_pfn) {
|
|
|
|
data->start_pfn = start_pfn;
|
|
|
|
data->end_pfn = end_pfn;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* get_node_active_region - Return active region containing start_pfn
|
2008-10-16 18:59:43 +00:00
|
|
|
* Active range returned is empty if none found.
|
2008-10-09 10:18:40 +00:00
|
|
|
* @start_pfn: The page to return the region for.
|
|
|
|
* @node_ar: Returned set to the active region containing start_pfn
|
|
|
|
*/
|
|
|
|
static void __init get_node_active_region(unsigned long start_pfn,
|
|
|
|
struct node_active_region *node_ar)
|
|
|
|
{
|
|
|
|
int nid = early_pfn_to_nid(start_pfn);
|
|
|
|
|
|
|
|
node_ar->nid = nid;
|
|
|
|
node_ar->start_pfn = start_pfn;
|
2008-10-16 18:59:43 +00:00
|
|
|
node_ar->end_pfn = start_pfn;
|
2008-10-09 10:18:40 +00:00
|
|
|
work_with_active_regions(nid, get_active_region_work_fn, node_ar);
|
|
|
|
}
|
|
|
|
|
2010-12-17 22:07:47 +00:00
|
|
|
static void map_cpu_to_node(int cpu, int node)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
numa_cpu_lookup_table[cpu] = node;
|
2005-11-11 03:22:35 +00:00
|
|
|
|
2006-03-21 00:34:45 +00:00
|
|
|
dbg("adding cpu %d to node %d\n", cpu, node);
|
|
|
|
|
2010-04-26 15:32:43 +00:00
|
|
|
if (!(cpumask_test_cpu(cpu, node_to_cpumask_map[node])))
|
|
|
|
cpumask_set_cpu(cpu, node_to_cpumask_map[node]);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2010-12-17 22:07:47 +00:00
|
|
|
#if defined(CONFIG_HOTPLUG_CPU) || defined(CONFIG_PPC_SPLPAR)
|
2005-04-16 22:20:36 +00:00
|
|
|
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);
|
|
|
|
|
2010-04-26 15:32:43 +00:00
|
|
|
if (cpumask_test_cpu(cpu, node_to_cpumask_map[node])) {
|
|
|
|
cpumask_set_cpu(cpu, node_to_cpumask_map[node]);
|
2005-04-16 22:20:36 +00:00
|
|
|
} else {
|
|
|
|
printk(KERN_ERR "WARNING: cpu %lu not found in node %d\n",
|
|
|
|
cpu, node);
|
|
|
|
}
|
|
|
|
}
|
2010-12-17 22:07:47 +00:00
|
|
|
#endif /* CONFIG_HOTPLUG_CPU || CONFIG_PPC_SPLPAR */
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/* must hold reference to node during call */
|
2006-07-12 05:35:54 +00:00
|
|
|
static const int *of_get_associativity(struct device_node *dev)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2007-04-03 12:26:41 +00:00
|
|
|
return of_get_property(dev, "ibm,associativity", NULL);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
powerpc: Add support for dynamic reconfiguration memory in kexec/kdump kernels
Kdump kernel needs to use only those memory regions that it is allowed
to use (crashkernel, rtas, tce, etc.). Each of these regions have
their own sizes and are currently added under 'linux,usable-memory'
property under each memory@xxx node of the device tree.
The ibm,dynamic-memory property of ibm,dynamic-reconfiguration-memory
node (on POWER6) now stores in it the representation for most of the
logical memory blocks with the size of each memory block being a
constant (lmb_size). If one or more or part of the above mentioned
regions lie under one of the lmb from ibm,dynamic-memory property,
there is a need to identify those regions within the given lmb.
This makes the kernel recognize a new 'linux,drconf-usable-memory'
property added by kexec-tools. Each entry in this property is of the
form of a count followed by that many (base, size) pairs for the above
mentioned regions. The number of cells in the count value is given by
the #size-cells property of the root node.
Signed-off-by: Chandru Siddalingappa <chandru@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-08-29 14:28:16 +00:00
|
|
|
/*
|
|
|
|
* Returns the property linux,drconf-usable-memory if
|
|
|
|
* it exists (the property exists only in kexec/kdump kernels,
|
|
|
|
* added by kexec-tools)
|
|
|
|
*/
|
|
|
|
static const u32 *of_get_usable_memory(struct device_node *memory)
|
|
|
|
{
|
|
|
|
const u32 *prop;
|
|
|
|
u32 len;
|
|
|
|
prop = of_get_property(memory, "linux,drconf-usable-memory", &len);
|
|
|
|
if (!prop || len < sizeof(unsigned int))
|
|
|
|
return 0;
|
|
|
|
return prop;
|
|
|
|
}
|
|
|
|
|
2010-05-16 20:22:31 +00:00
|
|
|
int __node_distance(int a, int b)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
int distance = LOCAL_DISTANCE;
|
|
|
|
|
|
|
|
if (!form1_affinity)
|
|
|
|
return distance;
|
|
|
|
|
|
|
|
for (i = 0; i < distance_ref_points_depth; i++) {
|
|
|
|
if (distance_lookup_table[a][i] == distance_lookup_table[b][i])
|
|
|
|
break;
|
|
|
|
|
|
|
|
/* Double the distance for each NUMA level */
|
|
|
|
distance *= 2;
|
|
|
|
}
|
|
|
|
|
|
|
|
return distance;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void initialize_distance_lookup_table(int nid,
|
|
|
|
const unsigned int *associativity)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (!form1_affinity)
|
|
|
|
return;
|
|
|
|
|
|
|
|
for (i = 0; i < distance_ref_points_depth; i++) {
|
|
|
|
distance_lookup_table[nid][i] =
|
|
|
|
associativity[distance_ref_points[i]];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2006-03-21 00:36:45 +00:00
|
|
|
/* Returns nid in the range [0..MAX_NUMNODES-1], or -1 if no useful numa
|
|
|
|
* info is found.
|
|
|
|
*/
|
2010-12-01 12:31:15 +00:00
|
|
|
static int associativity_to_nid(const unsigned int *associativity)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2006-03-21 00:36:45 +00:00
|
|
|
int nid = -1;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
if (min_common_depth == -1)
|
2006-03-21 00:36:45 +00:00
|
|
|
goto out;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2010-12-01 12:31:15 +00:00
|
|
|
if (associativity[0] >= min_common_depth)
|
|
|
|
nid = associativity[min_common_depth];
|
2006-03-21 00:36:15 +00:00
|
|
|
|
|
|
|
/* POWER4 LPAR uses 0xffff as invalid node */
|
2006-03-21 00:36:45 +00:00
|
|
|
if (nid == 0xffff || nid >= MAX_NUMNODES)
|
|
|
|
nid = -1;
|
2010-05-16 20:22:31 +00:00
|
|
|
|
2010-12-01 12:31:15 +00:00
|
|
|
if (nid > 0 && associativity[0] >= distance_ref_points_depth)
|
|
|
|
initialize_distance_lookup_table(nid, associativity);
|
2010-05-16 20:22:31 +00:00
|
|
|
|
2006-03-21 00:36:45 +00:00
|
|
|
out:
|
2006-03-21 00:35:45 +00:00
|
|
|
return nid;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2010-12-01 12:31:15 +00:00
|
|
|
/* Returns the nid associated with the given device tree node,
|
|
|
|
* or -1 if not found.
|
|
|
|
*/
|
|
|
|
static int of_node_to_nid_single(struct device_node *device)
|
|
|
|
{
|
|
|
|
int nid = -1;
|
|
|
|
const unsigned int *tmp;
|
|
|
|
|
|
|
|
tmp = of_get_associativity(device);
|
|
|
|
if (tmp)
|
|
|
|
nid = associativity_to_nid(tmp);
|
|
|
|
return nid;
|
|
|
|
}
|
|
|
|
|
2006-05-01 19:16:12 +00:00
|
|
|
/* 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);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
static int __init find_min_common_depth(void)
|
|
|
|
{
|
2010-05-16 20:22:31 +00:00
|
|
|
int depth;
|
2005-04-16 22:20:36 +00:00
|
|
|
struct device_node *rtas_root;
|
2010-05-16 20:28:35 +00:00
|
|
|
struct device_node *chosen;
|
|
|
|
const char *vec5;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
rtas_root = of_find_node_by_path("/rtas");
|
|
|
|
|
|
|
|
if (!rtas_root)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
/*
|
2010-05-16 20:22:31 +00:00
|
|
|
* This property is a set of 32-bit integers, each representing
|
|
|
|
* an index into the ibm,associativity nodes.
|
|
|
|
*
|
|
|
|
* With form 0 affinity the first integer is for an SMP configuration
|
|
|
|
* (should be all 0's) and the second is for a normal NUMA
|
|
|
|
* configuration. We have only one level of NUMA.
|
|
|
|
*
|
|
|
|
* With form 1 affinity the first integer is the most significant
|
|
|
|
* NUMA boundary and the following are progressively less significant
|
|
|
|
* boundaries. There can be more than one level of NUMA.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2010-05-16 20:22:31 +00:00
|
|
|
distance_ref_points = of_get_property(rtas_root,
|
|
|
|
"ibm,associativity-reference-points",
|
|
|
|
&distance_ref_points_depth);
|
|
|
|
|
|
|
|
if (!distance_ref_points) {
|
|
|
|
dbg("NUMA: ibm,associativity-reference-points not found.\n");
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
distance_ref_points_depth /= sizeof(int);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2010-05-16 20:28:35 +00:00
|
|
|
#define VEC5_AFFINITY_BYTE 5
|
|
|
|
#define VEC5_AFFINITY 0x80
|
|
|
|
chosen = of_find_node_by_path("/chosen");
|
|
|
|
if (chosen) {
|
|
|
|
vec5 = of_get_property(chosen, "ibm,architecture-vec-5", NULL);
|
|
|
|
if (vec5 && (vec5[VEC5_AFFINITY_BYTE] & VEC5_AFFINITY)) {
|
|
|
|
dbg("Using form 1 affinity\n");
|
2010-05-16 20:22:31 +00:00
|
|
|
form1_affinity = 1;
|
2010-05-16 20:28:35 +00:00
|
|
|
}
|
2010-04-07 15:33:44 +00:00
|
|
|
}
|
|
|
|
|
2010-05-16 20:22:31 +00:00
|
|
|
if (form1_affinity) {
|
|
|
|
depth = distance_ref_points[0];
|
2005-04-16 22:20:36 +00:00
|
|
|
} else {
|
2010-05-16 20:22:31 +00:00
|
|
|
if (distance_ref_points_depth < 2) {
|
|
|
|
printk(KERN_WARNING "NUMA: "
|
|
|
|
"short ibm,associativity-reference-points\n");
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
depth = distance_ref_points[1];
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2010-05-16 20:22:31 +00:00
|
|
|
/*
|
|
|
|
* Warn and cap if the hardware supports more than
|
|
|
|
* MAX_DISTANCE_REF_POINTS domains.
|
|
|
|
*/
|
|
|
|
if (distance_ref_points_depth > MAX_DISTANCE_REF_POINTS) {
|
|
|
|
printk(KERN_WARNING "NUMA: distance array capped at "
|
|
|
|
"%d entries\n", MAX_DISTANCE_REF_POINTS);
|
|
|
|
distance_ref_points_depth = MAX_DISTANCE_REF_POINTS;
|
|
|
|
}
|
|
|
|
|
|
|
|
of_node_put(rtas_root);
|
2005-04-16 22:20:36 +00:00
|
|
|
return depth;
|
2010-05-16 20:22:31 +00:00
|
|
|
|
|
|
|
err:
|
|
|
|
of_node_put(rtas_root);
|
|
|
|
return -1;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2005-11-30 21:47:23 +00:00
|
|
|
static void __init get_n_mem_cells(int *n_addr_cells, int *n_size_cells)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
struct device_node *memory = NULL;
|
|
|
|
|
|
|
|
memory = of_find_node_by_type(memory, "memory");
|
2005-12-05 04:50:39 +00:00
|
|
|
if (!memory)
|
2005-11-30 21:47:23 +00:00
|
|
|
panic("numa.c: No memory nodes found!");
|
2005-12-05 04:50:39 +00:00
|
|
|
|
2007-04-03 00:56:50 +00:00
|
|
|
*n_addr_cells = of_n_addr_cells(memory);
|
2007-04-03 00:57:48 +00:00
|
|
|
*n_size_cells = of_n_size_cells(memory);
|
2005-11-30 21:47:23 +00:00
|
|
|
of_node_put(memory);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2006-07-12 05:35:54 +00:00
|
|
|
static unsigned long __devinit read_n_cells(int n, const unsigned int **buf)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
unsigned long result = 0;
|
|
|
|
|
|
|
|
while (n--) {
|
|
|
|
result = (result << 32) | **buf;
|
|
|
|
(*buf)++;
|
|
|
|
}
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2008-07-03 03:35:54 +00:00
|
|
|
struct of_drconf_cell {
|
|
|
|
u64 base_addr;
|
|
|
|
u32 drc_index;
|
|
|
|
u32 reserved;
|
|
|
|
u32 aa_index;
|
|
|
|
u32 flags;
|
|
|
|
};
|
|
|
|
|
|
|
|
#define DRCONF_MEM_ASSIGNED 0x00000008
|
|
|
|
#define DRCONF_MEM_AI_INVALID 0x00000040
|
|
|
|
#define DRCONF_MEM_RESERVED 0x00000080
|
|
|
|
|
|
|
|
/*
|
2010-07-12 04:36:09 +00:00
|
|
|
* Read the next memblock list entry from the ibm,dynamic-memory property
|
2008-07-03 03:35:54 +00:00
|
|
|
* and return the information in the provided of_drconf_cell structure.
|
|
|
|
*/
|
|
|
|
static void read_drconf_cell(struct of_drconf_cell *drmem, const u32 **cellp)
|
|
|
|
{
|
|
|
|
const u32 *cp;
|
|
|
|
|
|
|
|
drmem->base_addr = read_n_cells(n_mem_addr_cells, cellp);
|
|
|
|
|
|
|
|
cp = *cellp;
|
|
|
|
drmem->drc_index = cp[0];
|
|
|
|
drmem->reserved = cp[1];
|
|
|
|
drmem->aa_index = cp[2];
|
|
|
|
drmem->flags = cp[3];
|
|
|
|
|
|
|
|
*cellp = cp + 4;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Retreive and validate the ibm,dynamic-memory property of the device tree.
|
|
|
|
*
|
2010-07-12 04:36:09 +00:00
|
|
|
* The layout of the ibm,dynamic-memory property is a number N of memblock
|
|
|
|
* list entries followed by N memblock list entries. Each memblock list entry
|
2008-07-03 03:35:54 +00:00
|
|
|
* contains information as layed out in the of_drconf_cell struct above.
|
|
|
|
*/
|
|
|
|
static int of_get_drconf_memory(struct device_node *memory, const u32 **dm)
|
|
|
|
{
|
|
|
|
const u32 *prop;
|
|
|
|
u32 len, entries;
|
|
|
|
|
|
|
|
prop = of_get_property(memory, "ibm,dynamic-memory", &len);
|
|
|
|
if (!prop || len < sizeof(unsigned int))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
entries = *prop++;
|
|
|
|
|
|
|
|
/* Now that we know the number of entries, revalidate the size
|
|
|
|
* of the property read in to ensure we have everything
|
|
|
|
*/
|
|
|
|
if (len < (entries * (n_mem_addr_cells + 4) + 1) * sizeof(unsigned int))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
*dm = prop;
|
|
|
|
return entries;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2010-07-23 00:35:52 +00:00
|
|
|
* Retreive and validate the ibm,lmb-size property for drconf memory
|
2008-07-03 03:35:54 +00:00
|
|
|
* from the device tree.
|
|
|
|
*/
|
2010-07-23 00:35:52 +00:00
|
|
|
static u64 of_get_lmb_size(struct device_node *memory)
|
2008-07-03 03:35:54 +00:00
|
|
|
{
|
|
|
|
const u32 *prop;
|
|
|
|
u32 len;
|
|
|
|
|
2010-07-23 00:35:52 +00:00
|
|
|
prop = of_get_property(memory, "ibm,lmb-size", &len);
|
2008-07-03 03:35:54 +00:00
|
|
|
if (!prop || len < sizeof(unsigned int))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
return read_n_cells(n_mem_size_cells, &prop);
|
|
|
|
}
|
|
|
|
|
|
|
|
struct assoc_arrays {
|
|
|
|
u32 n_arrays;
|
|
|
|
u32 array_sz;
|
|
|
|
const u32 *arrays;
|
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Retreive and validate the list of associativity arrays for drconf
|
|
|
|
* memory from the ibm,associativity-lookup-arrays property of the
|
|
|
|
* device tree..
|
|
|
|
*
|
|
|
|
* The layout of the ibm,associativity-lookup-arrays property is a number N
|
|
|
|
* indicating the number of associativity arrays, followed by a number M
|
|
|
|
* indicating the size of each associativity array, followed by a list
|
|
|
|
* of N associativity arrays.
|
|
|
|
*/
|
|
|
|
static int of_get_assoc_arrays(struct device_node *memory,
|
|
|
|
struct assoc_arrays *aa)
|
|
|
|
{
|
|
|
|
const u32 *prop;
|
|
|
|
u32 len;
|
|
|
|
|
|
|
|
prop = of_get_property(memory, "ibm,associativity-lookup-arrays", &len);
|
|
|
|
if (!prop || len < 2 * sizeof(unsigned int))
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
aa->n_arrays = *prop++;
|
|
|
|
aa->array_sz = *prop++;
|
|
|
|
|
|
|
|
/* Now that we know the number of arrrays and size of each array,
|
|
|
|
* revalidate the size of the property read in.
|
|
|
|
*/
|
|
|
|
if (len < (aa->n_arrays * aa->array_sz + 2) * sizeof(unsigned int))
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
aa->arrays = prop;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This is like of_node_to_nid_single() for memory represented in the
|
|
|
|
* ibm,dynamic-reconfiguration-memory node.
|
|
|
|
*/
|
|
|
|
static int of_drconf_to_nid_single(struct of_drconf_cell *drmem,
|
|
|
|
struct assoc_arrays *aa)
|
|
|
|
{
|
|
|
|
int default_nid = 0;
|
|
|
|
int nid = default_nid;
|
|
|
|
int index;
|
|
|
|
|
|
|
|
if (min_common_depth > 0 && min_common_depth <= aa->array_sz &&
|
|
|
|
!(drmem->flags & DRCONF_MEM_AI_INVALID) &&
|
|
|
|
drmem->aa_index < aa->n_arrays) {
|
|
|
|
index = drmem->aa_index * aa->array_sz + min_common_depth - 1;
|
|
|
|
nid = aa->arrays[index];
|
|
|
|
|
|
|
|
if (nid == 0xffff || nid >= MAX_NUMNODES)
|
|
|
|
nid = default_nid;
|
|
|
|
}
|
|
|
|
|
|
|
|
return nid;
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* Figure out to which domain a cpu belongs and stick it there.
|
|
|
|
* Return the id of the domain used.
|
|
|
|
*/
|
2006-03-21 00:35:15 +00:00
|
|
|
static int __cpuinit numa_setup_cpu(unsigned long lcpu)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2006-03-21 00:35:45 +00:00
|
|
|
int nid = 0;
|
2009-01-08 02:19:45 +00:00
|
|
|
struct device_node *cpu = of_get_cpu_node(lcpu, NULL);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
if (!cpu) {
|
|
|
|
WARN_ON(1);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2006-05-01 19:16:12 +00:00
|
|
|
nid = of_node_to_nid_single(cpu);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-03-21 00:36:45 +00:00
|
|
|
if (nid < 0 || !node_online(nid))
|
2010-03-05 21:42:43 +00:00
|
|
|
nid = first_online_node;
|
2005-04-16 22:20:36 +00:00
|
|
|
out:
|
2006-03-21 00:35:45 +00:00
|
|
|
map_cpu_to_node(lcpu, nid);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
of_node_put(cpu);
|
|
|
|
|
2006-03-21 00:35:45 +00:00
|
|
|
return nid;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2006-06-27 09:54:09 +00:00
|
|
|
static int __cpuinit cpu_numa_callback(struct notifier_block *nfb,
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long action,
|
|
|
|
void *hcpu)
|
|
|
|
{
|
|
|
|
unsigned long lcpu = (unsigned long)hcpu;
|
|
|
|
int ret = NOTIFY_DONE;
|
|
|
|
|
|
|
|
switch (action) {
|
|
|
|
case CPU_UP_PREPARE:
|
2007-05-09 09:35:10 +00:00
|
|
|
case CPU_UP_PREPARE_FROZEN:
|
2006-03-21 00:37:15 +00:00
|
|
|
numa_setup_cpu(lcpu);
|
2005-04-16 22:20:36 +00:00
|
|
|
ret = NOTIFY_OK;
|
|
|
|
break;
|
|
|
|
#ifdef CONFIG_HOTPLUG_CPU
|
|
|
|
case CPU_DEAD:
|
2007-05-09 09:35:10 +00:00
|
|
|
case CPU_DEAD_FROZEN:
|
2005-04-16 22:20:36 +00:00
|
|
|
case CPU_UP_CANCELED:
|
2007-05-09 09:35:10 +00:00
|
|
|
case CPU_UP_CANCELED_FROZEN:
|
2005-04-16 22:20:36 +00:00
|
|
|
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 03:22:35 +00:00
|
|
|
static unsigned long __init numa_enforce_memory_limit(unsigned long start,
|
|
|
|
unsigned long size)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
/*
|
2010-07-12 04:36:09 +00:00
|
|
|
* We use memblock_end_of_DRAM() in here instead of memory_limit because
|
2005-04-16 22:20:36 +00:00
|
|
|
* we've already adjusted it for the limit and it takes care of
|
2008-10-20 15:37:04 +00:00
|
|
|
* having memory holes below the limit. Also, in the case of
|
|
|
|
* iommu_is_off, memory_limit is not set but is implicitly enforced.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
|
|
|
|
2010-07-12 04:36:09 +00:00
|
|
|
if (start + size <= memblock_end_of_DRAM())
|
2005-04-16 22:20:36 +00:00
|
|
|
return size;
|
|
|
|
|
2010-07-12 04:36:09 +00:00
|
|
|
if (start >= memblock_end_of_DRAM())
|
2005-04-16 22:20:36 +00:00
|
|
|
return 0;
|
|
|
|
|
2010-07-12 04:36:09 +00:00
|
|
|
return memblock_end_of_DRAM() - start;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
powerpc: Add support for dynamic reconfiguration memory in kexec/kdump kernels
Kdump kernel needs to use only those memory regions that it is allowed
to use (crashkernel, rtas, tce, etc.). Each of these regions have
their own sizes and are currently added under 'linux,usable-memory'
property under each memory@xxx node of the device tree.
The ibm,dynamic-memory property of ibm,dynamic-reconfiguration-memory
node (on POWER6) now stores in it the representation for most of the
logical memory blocks with the size of each memory block being a
constant (lmb_size). If one or more or part of the above mentioned
regions lie under one of the lmb from ibm,dynamic-memory property,
there is a need to identify those regions within the given lmb.
This makes the kernel recognize a new 'linux,drconf-usable-memory'
property added by kexec-tools. Each entry in this property is of the
form of a count followed by that many (base, size) pairs for the above
mentioned regions. The number of cells in the count value is given by
the #size-cells property of the root node.
Signed-off-by: Chandru Siddalingappa <chandru@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-08-29 14:28:16 +00:00
|
|
|
/*
|
|
|
|
* Reads the counter for a given entry in
|
|
|
|
* linux,drconf-usable-memory property
|
|
|
|
*/
|
|
|
|
static inline int __init read_usm_ranges(const u32 **usm)
|
|
|
|
{
|
|
|
|
/*
|
2010-07-23 00:35:52 +00:00
|
|
|
* For each lmb in ibm,dynamic-memory a corresponding
|
powerpc: Add support for dynamic reconfiguration memory in kexec/kdump kernels
Kdump kernel needs to use only those memory regions that it is allowed
to use (crashkernel, rtas, tce, etc.). Each of these regions have
their own sizes and are currently added under 'linux,usable-memory'
property under each memory@xxx node of the device tree.
The ibm,dynamic-memory property of ibm,dynamic-reconfiguration-memory
node (on POWER6) now stores in it the representation for most of the
logical memory blocks with the size of each memory block being a
constant (lmb_size). If one or more or part of the above mentioned
regions lie under one of the lmb from ibm,dynamic-memory property,
there is a need to identify those regions within the given lmb.
This makes the kernel recognize a new 'linux,drconf-usable-memory'
property added by kexec-tools. Each entry in this property is of the
form of a count followed by that many (base, size) pairs for the above
mentioned regions. The number of cells in the count value is given by
the #size-cells property of the root node.
Signed-off-by: Chandru Siddalingappa <chandru@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-08-29 14:28:16 +00:00
|
|
|
* entry in linux,drconf-usable-memory property contains
|
|
|
|
* a counter followed by that many (base, size) duple.
|
|
|
|
* read the counter from linux,drconf-usable-memory
|
|
|
|
*/
|
|
|
|
return read_n_cells(n_mem_size_cells, usm);
|
|
|
|
}
|
|
|
|
|
[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 11:27:42 +00:00
|
|
|
/*
|
|
|
|
* 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)
|
|
|
|
{
|
powerpc: Add support for dynamic reconfiguration memory in kexec/kdump kernels
Kdump kernel needs to use only those memory regions that it is allowed
to use (crashkernel, rtas, tce, etc.). Each of these regions have
their own sizes and are currently added under 'linux,usable-memory'
property under each memory@xxx node of the device tree.
The ibm,dynamic-memory property of ibm,dynamic-reconfiguration-memory
node (on POWER6) now stores in it the representation for most of the
logical memory blocks with the size of each memory block being a
constant (lmb_size). If one or more or part of the above mentioned
regions lie under one of the lmb from ibm,dynamic-memory property,
there is a need to identify those regions within the given lmb.
This makes the kernel recognize a new 'linux,drconf-usable-memory'
property added by kexec-tools. Each entry in this property is of the
form of a count followed by that many (base, size) pairs for the above
mentioned regions. The number of cells in the count value is given by
the #size-cells property of the root node.
Signed-off-by: Chandru Siddalingappa <chandru@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-08-29 14:28:16 +00:00
|
|
|
const u32 *dm, *usm;
|
|
|
|
unsigned int n, rc, ranges, is_kexec_kdump = 0;
|
2010-07-23 00:35:52 +00:00
|
|
|
unsigned long lmb_size, base, size, sz;
|
2008-07-03 03:35:54 +00:00
|
|
|
int nid;
|
|
|
|
struct assoc_arrays aa;
|
|
|
|
|
|
|
|
n = of_get_drconf_memory(memory, &dm);
|
|
|
|
if (!n)
|
[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 11:27:42 +00:00
|
|
|
return;
|
|
|
|
|
2010-07-23 00:35:52 +00:00
|
|
|
lmb_size = of_get_lmb_size(memory);
|
|
|
|
if (!lmb_size)
|
2008-07-03 03:35:54 +00:00
|
|
|
return;
|
|
|
|
|
|
|
|
rc = of_get_assoc_arrays(memory, &aa);
|
|
|
|
if (rc)
|
[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 11:27:42 +00:00
|
|
|
return;
|
|
|
|
|
powerpc: Add support for dynamic reconfiguration memory in kexec/kdump kernels
Kdump kernel needs to use only those memory regions that it is allowed
to use (crashkernel, rtas, tce, etc.). Each of these regions have
their own sizes and are currently added under 'linux,usable-memory'
property under each memory@xxx node of the device tree.
The ibm,dynamic-memory property of ibm,dynamic-reconfiguration-memory
node (on POWER6) now stores in it the representation for most of the
logical memory blocks with the size of each memory block being a
constant (lmb_size). If one or more or part of the above mentioned
regions lie under one of the lmb from ibm,dynamic-memory property,
there is a need to identify those regions within the given lmb.
This makes the kernel recognize a new 'linux,drconf-usable-memory'
property added by kexec-tools. Each entry in this property is of the
form of a count followed by that many (base, size) pairs for the above
mentioned regions. The number of cells in the count value is given by
the #size-cells property of the root node.
Signed-off-by: Chandru Siddalingappa <chandru@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-08-29 14:28:16 +00:00
|
|
|
/* check if this is a kexec/kdump kernel */
|
|
|
|
usm = of_get_usable_memory(memory);
|
|
|
|
if (usm != NULL)
|
|
|
|
is_kexec_kdump = 1;
|
|
|
|
|
[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 11:27:42 +00:00
|
|
|
for (; n != 0; --n) {
|
2008-07-03 03:35:54 +00:00
|
|
|
struct of_drconf_cell drmem;
|
|
|
|
|
|
|
|
read_drconf_cell(&drmem, &dm);
|
|
|
|
|
|
|
|
/* skip this block if the reserved bit is set in flags (0x80)
|
|
|
|
or if the block is not assigned to this partition (0x8) */
|
|
|
|
if ((drmem.flags & DRCONF_MEM_RESERVED)
|
|
|
|
|| !(drmem.flags & DRCONF_MEM_ASSIGNED))
|
[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 11:27:42 +00:00
|
|
|
continue;
|
2008-02-01 04:57:31 +00:00
|
|
|
|
powerpc: Add support for dynamic reconfiguration memory in kexec/kdump kernels
Kdump kernel needs to use only those memory regions that it is allowed
to use (crashkernel, rtas, tce, etc.). Each of these regions have
their own sizes and are currently added under 'linux,usable-memory'
property under each memory@xxx node of the device tree.
The ibm,dynamic-memory property of ibm,dynamic-reconfiguration-memory
node (on POWER6) now stores in it the representation for most of the
logical memory blocks with the size of each memory block being a
constant (lmb_size). If one or more or part of the above mentioned
regions lie under one of the lmb from ibm,dynamic-memory property,
there is a need to identify those regions within the given lmb.
This makes the kernel recognize a new 'linux,drconf-usable-memory'
property added by kexec-tools. Each entry in this property is of the
form of a count followed by that many (base, size) pairs for the above
mentioned regions. The number of cells in the count value is given by
the #size-cells property of the root node.
Signed-off-by: Chandru Siddalingappa <chandru@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-08-29 14:28:16 +00:00
|
|
|
base = drmem.base_addr;
|
2010-07-23 00:35:52 +00:00
|
|
|
size = lmb_size;
|
powerpc: Add support for dynamic reconfiguration memory in kexec/kdump kernels
Kdump kernel needs to use only those memory regions that it is allowed
to use (crashkernel, rtas, tce, etc.). Each of these regions have
their own sizes and are currently added under 'linux,usable-memory'
property under each memory@xxx node of the device tree.
The ibm,dynamic-memory property of ibm,dynamic-reconfiguration-memory
node (on POWER6) now stores in it the representation for most of the
logical memory blocks with the size of each memory block being a
constant (lmb_size). If one or more or part of the above mentioned
regions lie under one of the lmb from ibm,dynamic-memory property,
there is a need to identify those regions within the given lmb.
This makes the kernel recognize a new 'linux,drconf-usable-memory'
property added by kexec-tools. Each entry in this property is of the
form of a count followed by that many (base, size) pairs for the above
mentioned regions. The number of cells in the count value is given by
the #size-cells property of the root node.
Signed-off-by: Chandru Siddalingappa <chandru@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-08-29 14:28:16 +00:00
|
|
|
ranges = 1;
|
2008-07-03 03:35:54 +00:00
|
|
|
|
powerpc: Add support for dynamic reconfiguration memory in kexec/kdump kernels
Kdump kernel needs to use only those memory regions that it is allowed
to use (crashkernel, rtas, tce, etc.). Each of these regions have
their own sizes and are currently added under 'linux,usable-memory'
property under each memory@xxx node of the device tree.
The ibm,dynamic-memory property of ibm,dynamic-reconfiguration-memory
node (on POWER6) now stores in it the representation for most of the
logical memory blocks with the size of each memory block being a
constant (lmb_size). If one or more or part of the above mentioned
regions lie under one of the lmb from ibm,dynamic-memory property,
there is a need to identify those regions within the given lmb.
This makes the kernel recognize a new 'linux,drconf-usable-memory'
property added by kexec-tools. Each entry in this property is of the
form of a count followed by that many (base, size) pairs for the above
mentioned regions. The number of cells in the count value is given by
the #size-cells property of the root node.
Signed-off-by: Chandru Siddalingappa <chandru@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-08-29 14:28:16 +00:00
|
|
|
if (is_kexec_kdump) {
|
|
|
|
ranges = read_usm_ranges(&usm);
|
|
|
|
if (!ranges) /* there are no (base, size) duple */
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
do {
|
|
|
|
if (is_kexec_kdump) {
|
|
|
|
base = read_n_cells(n_mem_addr_cells, &usm);
|
|
|
|
size = read_n_cells(n_mem_size_cells, &usm);
|
|
|
|
}
|
|
|
|
nid = of_drconf_to_nid_single(&drmem, &aa);
|
|
|
|
fake_numa_create_new_node(
|
|
|
|
((base + size) >> PAGE_SHIFT),
|
2008-07-03 03:35:54 +00:00
|
|
|
&nid);
|
powerpc: Add support for dynamic reconfiguration memory in kexec/kdump kernels
Kdump kernel needs to use only those memory regions that it is allowed
to use (crashkernel, rtas, tce, etc.). Each of these regions have
their own sizes and are currently added under 'linux,usable-memory'
property under each memory@xxx node of the device tree.
The ibm,dynamic-memory property of ibm,dynamic-reconfiguration-memory
node (on POWER6) now stores in it the representation for most of the
logical memory blocks with the size of each memory block being a
constant (lmb_size). If one or more or part of the above mentioned
regions lie under one of the lmb from ibm,dynamic-memory property,
there is a need to identify those regions within the given lmb.
This makes the kernel recognize a new 'linux,drconf-usable-memory'
property added by kexec-tools. Each entry in this property is of the
form of a count followed by that many (base, size) pairs for the above
mentioned regions. The number of cells in the count value is given by
the #size-cells property of the root node.
Signed-off-by: Chandru Siddalingappa <chandru@in.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-08-29 14:28:16 +00:00
|
|
|
node_set_online(nid);
|
|
|
|
sz = numa_enforce_memory_limit(base, size);
|
|
|
|
if (sz)
|
|
|
|
add_active_range(nid, base >> PAGE_SHIFT,
|
|
|
|
(base >> PAGE_SHIFT)
|
|
|
|
+ (sz >> PAGE_SHIFT));
|
|
|
|
} while (--ranges);
|
[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 11:27:42 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
static int __init parse_numa_properties(void)
|
|
|
|
{
|
|
|
|
struct device_node *cpu = NULL;
|
|
|
|
struct device_node *memory = NULL;
|
2006-03-21 00:36:45 +00:00
|
|
|
int default_nid = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
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 00:34:45 +00:00
|
|
|
dbg("NUMA associativity depth for CPU/Memory: %d\n", min_common_depth);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
2006-03-21 00:36:45 +00:00
|
|
|
* 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.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2006-03-21 00:36:45 +00:00
|
|
|
for_each_present_cpu(i) {
|
2006-03-21 00:35:45 +00:00
|
|
|
int nid;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2009-01-08 02:19:45 +00:00
|
|
|
cpu = of_get_cpu_node(i, NULL);
|
2006-03-21 00:36:45 +00:00
|
|
|
BUG_ON(!cpu);
|
2006-05-01 19:16:12 +00:00
|
|
|
nid = of_node_to_nid_single(cpu);
|
2006-03-21 00:36:45 +00:00
|
|
|
of_node_put(cpu);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-03-21 00:36:45 +00:00
|
|
|
/*
|
|
|
|
* 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);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2005-12-05 20:06:42 +00:00
|
|
|
get_n_mem_cells(&n_mem_addr_cells, &n_mem_size_cells);
|
2005-04-16 22:20:36 +00:00
|
|
|
memory = NULL;
|
|
|
|
while ((memory = of_find_node_by_type(memory, "memory")) != NULL) {
|
|
|
|
unsigned long start;
|
|
|
|
unsigned long size;
|
2006-03-21 00:35:45 +00:00
|
|
|
int nid;
|
2005-04-16 22:20:36 +00:00
|
|
|
int ranges;
|
2006-07-12 05:35:54 +00:00
|
|
|
const unsigned int *memcell_buf;
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned int len;
|
|
|
|
|
2007-04-03 12:26:41 +00:00
|
|
|
memcell_buf = of_get_property(memory,
|
2005-12-04 07:39:55 +00:00
|
|
|
"linux,usable-memory", &len);
|
|
|
|
if (!memcell_buf || len <= 0)
|
2007-04-03 12:26:41 +00:00
|
|
|
memcell_buf = of_get_property(memory, "reg", &len);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (!memcell_buf || len <= 0)
|
|
|
|
continue;
|
|
|
|
|
2005-12-13 07:01:21 +00:00
|
|
|
/* ranges in cell */
|
|
|
|
ranges = (len >> 2) / (n_mem_addr_cells + n_mem_size_cells);
|
2005-04-16 22:20:36 +00:00
|
|
|
new_range:
|
|
|
|
/* these are order-sensitive, and modify the buffer pointer */
|
2005-12-05 20:06:42 +00:00
|
|
|
start = read_n_cells(n_mem_addr_cells, &memcell_buf);
|
|
|
|
size = read_n_cells(n_mem_size_cells, &memcell_buf);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-03-21 00:36:45 +00:00
|
|
|
/*
|
|
|
|
* Assumption: either all memory nodes or none will
|
|
|
|
* have associativity properties. If none, then
|
|
|
|
* everything goes to default_nid.
|
|
|
|
*/
|
2006-05-01 19:16:12 +00:00
|
|
|
nid = of_node_to_nid_single(memory);
|
2006-03-21 00:36:45 +00:00
|
|
|
if (nid < 0)
|
|
|
|
nid = default_nid;
|
2008-02-01 04:57:31 +00:00
|
|
|
|
|
|
|
fake_numa_create_new_node(((start + size) >> PAGE_SHIFT), &nid);
|
2006-03-21 00:36:45 +00:00
|
|
|
node_set_online(nid);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2005-11-11 03:22:35 +00:00
|
|
|
if (!(size = numa_enforce_memory_limit(start, size))) {
|
2005-04-16 22:20:36 +00:00
|
|
|
if (--ranges)
|
|
|
|
goto new_range;
|
|
|
|
else
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2006-09-27 08:49:49 +00:00
|
|
|
add_active_range(nid, start >> PAGE_SHIFT,
|
|
|
|
(start >> PAGE_SHIFT) + (size >> PAGE_SHIFT));
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
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 11:27:42 +00:00
|
|
|
/*
|
2010-07-12 04:36:09 +00:00
|
|
|
* Now do the same thing for each MEMBLOCK listed in the ibm,dynamic-memory
|
[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 11:27:42 +00:00
|
|
|
* property in the ibm,dynamic-reconfiguration-memory node.
|
|
|
|
*/
|
|
|
|
memory = of_find_node_by_path("/ibm,dynamic-reconfiguration-memory");
|
|
|
|
if (memory)
|
|
|
|
parse_drconf_memory(memory);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __init setup_nonnuma(void)
|
|
|
|
{
|
2010-07-12 04:36:09 +00:00
|
|
|
unsigned long top_of_ram = memblock_end_of_DRAM();
|
|
|
|
unsigned long total_ram = memblock_phys_mem_size();
|
2006-09-27 08:49:49 +00:00
|
|
|
unsigned long start_pfn, end_pfn;
|
2010-08-04 03:43:53 +00:00
|
|
|
unsigned int nid = 0;
|
|
|
|
struct memblock_region *reg;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-04-12 20:25:01 +00:00
|
|
|
printk(KERN_DEBUG "Top of RAM: 0x%lx, Total RAM: 0x%lx\n",
|
2005-04-16 22:20:36 +00:00
|
|
|
top_of_ram, total_ram);
|
2006-04-12 20:25:01 +00:00
|
|
|
printk(KERN_DEBUG "Memory hole size: %ldMB\n",
|
2005-04-16 22:20:36 +00:00
|
|
|
(top_of_ram - total_ram) >> 20);
|
|
|
|
|
2010-08-04 03:43:53 +00:00
|
|
|
for_each_memblock(memory, reg) {
|
2010-10-12 21:07:09 +00:00
|
|
|
start_pfn = memblock_region_memory_base_pfn(reg);
|
|
|
|
end_pfn = memblock_region_memory_end_pfn(reg);
|
2008-02-01 04:57:31 +00:00
|
|
|
|
|
|
|
fake_numa_create_new_node(end_pfn, &nid);
|
|
|
|
add_active_range(nid, start_pfn, end_pfn);
|
|
|
|
node_set_online(nid);
|
2006-09-27 08:49:49 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2005-12-12 19:56:47 +00:00
|
|
|
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) {
|
2006-04-12 20:25:01 +00:00
|
|
|
printk(KERN_DEBUG "Node %d CPUs:", node);
|
2005-12-12 19:56:47 +00:00
|
|
|
|
|
|
|
count = 0;
|
|
|
|
/*
|
|
|
|
* If we used a CPU iterator here we would miss printing
|
|
|
|
* the holes in the cpumap.
|
|
|
|
*/
|
2010-04-26 15:32:43 +00:00
|
|
|
for (cpu = 0; cpu < nr_cpu_ids; cpu++) {
|
|
|
|
if (cpumask_test_cpu(cpu,
|
|
|
|
node_to_cpumask_map[node])) {
|
2005-12-12 19:56:47 +00:00
|
|
|
if (count == 0)
|
|
|
|
printk(" %u", cpu);
|
|
|
|
++count;
|
|
|
|
} else {
|
|
|
|
if (count > 1)
|
|
|
|
printk("-%u", cpu - 1);
|
|
|
|
count = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (count > 1)
|
2010-04-26 15:32:43 +00:00
|
|
|
printk("-%u", nr_cpu_ids - 1);
|
2005-12-12 19:56:47 +00:00
|
|
|
printk("\n");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __init dump_numa_memory_topology(void)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
unsigned int node;
|
|
|
|
unsigned int count;
|
|
|
|
|
|
|
|
if (min_common_depth == -1 || !numa_enabled)
|
|
|
|
return;
|
|
|
|
|
|
|
|
for_each_online_node(node) {
|
|
|
|
unsigned long i;
|
|
|
|
|
2006-04-12 20:25:01 +00:00
|
|
|
printk(KERN_DEBUG "Node %d Memory:", node);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
count = 0;
|
|
|
|
|
2010-07-12 04:36:09 +00:00
|
|
|
for (i = 0; i < memblock_end_of_DRAM();
|
2005-11-11 03:22:35 +00:00
|
|
|
i += (1 << SECTION_SIZE_BITS)) {
|
|
|
|
if (early_pfn_to_nid(i >> PAGE_SHIFT) == node) {
|
2005-04-16 22:20:36 +00:00
|
|
|
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");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2010-07-12 04:36:09 +00:00
|
|
|
* Allocate some memory, satisfying the memblock or bootmem allocator where
|
2005-04-16 22:20:36 +00:00
|
|
|
* required. nid is the preferred node and end is the physical address of
|
|
|
|
* the highest address in the node.
|
|
|
|
*
|
2008-12-09 08:21:35 +00:00
|
|
|
* Returns the virtual address of the memory.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2008-12-09 08:21:36 +00:00
|
|
|
static void __init *careful_zallocation(int nid, unsigned long size,
|
2005-11-11 03:22:35 +00:00
|
|
|
unsigned long align,
|
|
|
|
unsigned long end_pfn)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2008-12-09 08:21:35 +00:00
|
|
|
void *ret;
|
2005-11-11 03:22:35 +00:00
|
|
|
int new_nid;
|
2008-12-09 08:21:35 +00:00
|
|
|
unsigned long ret_paddr;
|
|
|
|
|
2010-07-12 04:36:09 +00:00
|
|
|
ret_paddr = __memblock_alloc_base(size, align, end_pfn << PAGE_SHIFT);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/* retry over all memory */
|
2008-12-09 08:21:35 +00:00
|
|
|
if (!ret_paddr)
|
2010-07-12 04:36:09 +00:00
|
|
|
ret_paddr = __memblock_alloc_base(size, align, memblock_end_of_DRAM());
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-12-09 08:21:35 +00:00
|
|
|
if (!ret_paddr)
|
2008-12-09 08:21:33 +00:00
|
|
|
panic("numa.c: cannot allocate %lu bytes for node %d",
|
2005-04-16 22:20:36 +00:00
|
|
|
size, nid);
|
|
|
|
|
2008-12-09 08:21:35 +00:00
|
|
|
ret = __va(ret_paddr);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
2008-12-09 08:21:32 +00:00
|
|
|
* We initialize the nodes in numeric order: 0, 1, 2...
|
2010-07-12 04:36:09 +00:00
|
|
|
* and hand over control from the MEMBLOCK allocator to the
|
2008-12-09 08:21:32 +00:00
|
|
|
* bootmem allocator. If this function is called for
|
|
|
|
* node 5, then we know that all nodes <5 are using the
|
2010-07-12 04:36:09 +00:00
|
|
|
* bootmem allocator instead of the MEMBLOCK allocator.
|
2008-12-09 08:21:32 +00:00
|
|
|
*
|
|
|
|
* So, check the nid from which this allocation came
|
|
|
|
* and double check to see if we need to use bootmem
|
2010-07-12 04:36:09 +00:00
|
|
|
* instead of the MEMBLOCK. We don't free the MEMBLOCK memory
|
2008-12-09 08:21:32 +00:00
|
|
|
* since it would be useless.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2008-12-09 08:21:35 +00:00
|
|
|
new_nid = early_pfn_to_nid(ret_paddr >> PAGE_SHIFT);
|
2005-11-11 03:22:35 +00:00
|
|
|
if (new_nid < nid) {
|
2008-12-09 08:21:35 +00:00
|
|
|
ret = __alloc_bootmem_node(NODE_DATA(new_nid),
|
2005-04-16 22:20:36 +00:00
|
|
|
size, align, 0);
|
|
|
|
|
2008-12-09 08:21:35 +00:00
|
|
|
dbg("alloc_bootmem %p %lx\n", ret, size);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2008-12-09 08:21:36 +00:00
|
|
|
memset(ret, 0, size);
|
2008-12-09 08:21:35 +00:00
|
|
|
return ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2006-06-27 09:54:09 +00:00
|
|
|
static struct notifier_block __cpuinitdata ppc64_numa_nb = {
|
|
|
|
.notifier_call = cpu_numa_callback,
|
|
|
|
.priority = 1 /* Must run before sched domains notifier. */
|
|
|
|
};
|
|
|
|
|
2008-11-24 12:02:35 +00:00
|
|
|
static void mark_reserved_regions_for_nid(int nid)
|
|
|
|
{
|
|
|
|
struct pglist_data *node = NODE_DATA(nid);
|
2010-08-04 03:43:53 +00:00
|
|
|
struct memblock_region *reg;
|
2008-11-24 12:02:35 +00:00
|
|
|
|
2010-08-04 03:43:53 +00:00
|
|
|
for_each_memblock(reserved, reg) {
|
|
|
|
unsigned long physbase = reg->base;
|
|
|
|
unsigned long size = reg->size;
|
2008-11-24 12:02:35 +00:00
|
|
|
unsigned long start_pfn = physbase >> PAGE_SHIFT;
|
powerpc/mm: Fix numa reserve bootmem page selection
Fix the powerpc NUMA reserve bootmem page selection logic.
commit 8f64e1f2d1e09267ac926e15090fd505c1c0cbcb (powerpc: Reserve
in bootmem lmb reserved regions that cross NUMA nodes) changed
the logic for how the powerpc LMB reserved regions were converted
to bootmen reserved regions. As the folowing discussion reports,
the new logic was not correct.
mark_reserved_regions_for_nid() goes through each LMB on the
system that specifies a reserved area. It searches for
active regions that intersect with that LMB and are on the
specified node. It attempts to bootmem-reserve only the area
where the active region and the reserved LMB intersect. We
can not reserve things on other nodes as they may not have
bootmem structures allocated, yet.
We base the size of the bootmem reservation on two possible
things. Normally, we just make the reservation start and
stop exactly at the start and end of the LMB.
However, the LMB reservations are not aware of NUMA nodes and
on occasion a single LMB may cross into several adjacent
active regions. Those may even be on different NUMA nodes
and will require separate calls to the bootmem reserve
functions. So, the bootmem reservation must be trimmed to
fit inside the current active region.
That's all fine and dandy, but we trim the reservation
in a page-aligned fashion. That's bad because we start the
reservation at a non-page-aligned address: physbase.
The reservation may only span 2 bytes, but that those bytes
may span two pfns and cause a reserve_size of 2*PAGE_SIZE.
Take the case where you reserve 0x2 bytes at 0x0fff and
where the active region ends at 0x1000. You'll jump into
that if() statment, but node_ar.end_pfn=0x1 and
start_pfn=0x0. You'll end up with a reserve_size=0x1000,
and then call
reserve_bootmem_node(node, physbase=0xfff, size=0x1000);
0x1000 may not be on the same node as 0xfff. Oops.
In almost all the vm code, end_<anything> is not inclusive.
If you have an end_pfn of 0x1234, page 0x1234 is not
included in the range. Using PFN_UP instead of the
(>> >> PAGE_SHIFT) will make this consistent with the other VM
code.
We also need to do math for the reserved size with physbase
instead of start_pfn. node_ar.end_pfn << PAGE_SHIFT is
*precisely* the end of the node. However,
(start_pfn << PAGE_SHIFT) is *NOT* precisely the beginning
of the reserved area. That is, of course, physbase.
If we don't use physbase here, the reserve_size can be
made too large.
From: Dave Hansen <dave@linux.vnet.ibm.com>
Tested-by: Geoff Levand <geoffrey.levand@am.sony.com> Tested on PS3.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2009-02-12 12:36:04 +00:00
|
|
|
unsigned long end_pfn = PFN_UP(physbase + size);
|
2008-11-24 12:02:35 +00:00
|
|
|
struct node_active_region node_ar;
|
|
|
|
unsigned long node_end_pfn = node->node_start_pfn +
|
|
|
|
node->node_spanned_pages;
|
|
|
|
|
|
|
|
/*
|
2010-07-12 04:36:09 +00:00
|
|
|
* Check to make sure that this memblock.reserved area is
|
2008-11-24 12:02:35 +00:00
|
|
|
* within the bounds of the node that we care about.
|
|
|
|
* Checking the nid of the start and end points is not
|
|
|
|
* sufficient because the reserved area could span the
|
|
|
|
* entire node.
|
|
|
|
*/
|
|
|
|
if (end_pfn <= node->node_start_pfn ||
|
|
|
|
start_pfn >= node_end_pfn)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
get_node_active_region(start_pfn, &node_ar);
|
|
|
|
while (start_pfn < end_pfn &&
|
|
|
|
node_ar.start_pfn < node_ar.end_pfn) {
|
|
|
|
unsigned long reserve_size = size;
|
|
|
|
/*
|
|
|
|
* if reserved region extends past active region
|
|
|
|
* then trim size to active region
|
|
|
|
*/
|
|
|
|
if (end_pfn > node_ar.end_pfn)
|
|
|
|
reserve_size = (node_ar.end_pfn << PAGE_SHIFT)
|
powerpc/mm: Fix numa reserve bootmem page selection
Fix the powerpc NUMA reserve bootmem page selection logic.
commit 8f64e1f2d1e09267ac926e15090fd505c1c0cbcb (powerpc: Reserve
in bootmem lmb reserved regions that cross NUMA nodes) changed
the logic for how the powerpc LMB reserved regions were converted
to bootmen reserved regions. As the folowing discussion reports,
the new logic was not correct.
mark_reserved_regions_for_nid() goes through each LMB on the
system that specifies a reserved area. It searches for
active regions that intersect with that LMB and are on the
specified node. It attempts to bootmem-reserve only the area
where the active region and the reserved LMB intersect. We
can not reserve things on other nodes as they may not have
bootmem structures allocated, yet.
We base the size of the bootmem reservation on two possible
things. Normally, we just make the reservation start and
stop exactly at the start and end of the LMB.
However, the LMB reservations are not aware of NUMA nodes and
on occasion a single LMB may cross into several adjacent
active regions. Those may even be on different NUMA nodes
and will require separate calls to the bootmem reserve
functions. So, the bootmem reservation must be trimmed to
fit inside the current active region.
That's all fine and dandy, but we trim the reservation
in a page-aligned fashion. That's bad because we start the
reservation at a non-page-aligned address: physbase.
The reservation may only span 2 bytes, but that those bytes
may span two pfns and cause a reserve_size of 2*PAGE_SIZE.
Take the case where you reserve 0x2 bytes at 0x0fff and
where the active region ends at 0x1000. You'll jump into
that if() statment, but node_ar.end_pfn=0x1 and
start_pfn=0x0. You'll end up with a reserve_size=0x1000,
and then call
reserve_bootmem_node(node, physbase=0xfff, size=0x1000);
0x1000 may not be on the same node as 0xfff. Oops.
In almost all the vm code, end_<anything> is not inclusive.
If you have an end_pfn of 0x1234, page 0x1234 is not
included in the range. Using PFN_UP instead of the
(>> >> PAGE_SHIFT) will make this consistent with the other VM
code.
We also need to do math for the reserved size with physbase
instead of start_pfn. node_ar.end_pfn << PAGE_SHIFT is
*precisely* the end of the node. However,
(start_pfn << PAGE_SHIFT) is *NOT* precisely the beginning
of the reserved area. That is, of course, physbase.
If we don't use physbase here, the reserve_size can be
made too large.
From: Dave Hansen <dave@linux.vnet.ibm.com>
Tested-by: Geoff Levand <geoffrey.levand@am.sony.com> Tested on PS3.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2009-02-12 12:36:04 +00:00
|
|
|
- physbase;
|
2008-12-11 08:36:06 +00:00
|
|
|
/*
|
|
|
|
* Only worry about *this* node, others may not
|
|
|
|
* yet have valid NODE_DATA().
|
|
|
|
*/
|
|
|
|
if (node_ar.nid == nid) {
|
|
|
|
dbg("reserve_bootmem %lx %lx nid=%d\n",
|
|
|
|
physbase, reserve_size, node_ar.nid);
|
|
|
|
reserve_bootmem_node(NODE_DATA(node_ar.nid),
|
|
|
|
physbase, reserve_size,
|
|
|
|
BOOTMEM_DEFAULT);
|
|
|
|
}
|
2008-11-24 12:02:35 +00:00
|
|
|
/*
|
|
|
|
* if reserved region is contained in the active region
|
|
|
|
* then done.
|
|
|
|
*/
|
|
|
|
if (end_pfn <= node_ar.end_pfn)
|
|
|
|
break;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* reserved region extends past the active region
|
|
|
|
* get next active region that contains this
|
|
|
|
* reserved region
|
|
|
|
*/
|
|
|
|
start_pfn = node_ar.end_pfn;
|
|
|
|
physbase = start_pfn << PAGE_SHIFT;
|
|
|
|
size = size - reserve_size;
|
|
|
|
get_node_active_region(start_pfn, &node_ar);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
void __init do_init_bootmem(void)
|
|
|
|
{
|
|
|
|
int nid;
|
|
|
|
|
|
|
|
min_low_pfn = 0;
|
2010-07-12 04:36:09 +00:00
|
|
|
max_low_pfn = memblock_end_of_DRAM() >> PAGE_SHIFT;
|
2005-04-16 22:20:36 +00:00
|
|
|
max_pfn = max_low_pfn;
|
|
|
|
|
|
|
|
if (parse_numa_properties())
|
|
|
|
setup_nonnuma();
|
|
|
|
else
|
2005-12-12 19:56:47 +00:00
|
|
|
dump_numa_memory_topology();
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
for_each_online_node(nid) {
|
2006-09-27 08:49:49 +00:00
|
|
|
unsigned long start_pfn, end_pfn;
|
2008-12-09 08:21:35 +00:00
|
|
|
void *bootmem_vaddr;
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long bootmap_pages;
|
|
|
|
|
2006-09-27 08:49:49 +00:00
|
|
|
get_pfn_range_for_nid(nid, &start_pfn, &end_pfn);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-11-24 12:02:35 +00:00
|
|
|
/*
|
|
|
|
* Allocate the node structure node local if possible
|
|
|
|
*
|
|
|
|
* Be careful moving this around, as it relies on all
|
|
|
|
* previous nodes' bootmem to be initialized and have
|
|
|
|
* all reserved areas marked.
|
|
|
|
*/
|
2008-12-09 08:21:36 +00:00
|
|
|
NODE_DATA(nid) = careful_zallocation(nid,
|
2005-04-16 22:20:36 +00:00
|
|
|
sizeof(struct pglist_data),
|
2005-11-11 03:22:35 +00:00
|
|
|
SMP_CACHE_BYTES, end_pfn);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
dbg("node %d\n", nid);
|
|
|
|
dbg("NODE_DATA() = %p\n", NODE_DATA(nid));
|
|
|
|
|
2008-07-24 04:26:55 +00:00
|
|
|
NODE_DATA(nid)->bdata = &bootmem_node_data[nid];
|
2005-11-11 03:22:35 +00:00
|
|
|
NODE_DATA(nid)->node_start_pfn = start_pfn;
|
|
|
|
NODE_DATA(nid)->node_spanned_pages = end_pfn - start_pfn;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
if (NODE_DATA(nid)->node_spanned_pages == 0)
|
|
|
|
continue;
|
|
|
|
|
2005-11-11 03:22:35 +00:00
|
|
|
dbg("start_paddr = %lx\n", start_pfn << PAGE_SHIFT);
|
|
|
|
dbg("end_paddr = %lx\n", end_pfn << PAGE_SHIFT);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2005-11-11 03:22:35 +00:00
|
|
|
bootmap_pages = bootmem_bootmap_pages(end_pfn - start_pfn);
|
2008-12-09 08:21:36 +00:00
|
|
|
bootmem_vaddr = careful_zallocation(nid,
|
2005-11-11 03:22:35 +00:00
|
|
|
bootmap_pages << PAGE_SHIFT,
|
|
|
|
PAGE_SIZE, end_pfn);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-12-09 08:21:35 +00:00
|
|
|
dbg("bootmap_vaddr = %p\n", bootmem_vaddr);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-12-09 08:21:35 +00:00
|
|
|
init_bootmem_node(NODE_DATA(nid),
|
|
|
|
__pa(bootmem_vaddr) >> PAGE_SHIFT,
|
2005-11-11 03:22:35 +00:00
|
|
|
start_pfn, end_pfn);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-09-27 08:49:49 +00:00
|
|
|
free_bootmem_with_active_regions(nid, end_pfn);
|
2008-11-24 12:02:35 +00:00
|
|
|
/*
|
|
|
|
* Be very careful about moving this around. Future
|
2008-12-09 08:21:36 +00:00
|
|
|
* calls to careful_zallocation() depend on this getting
|
2008-11-24 12:02:35 +00:00
|
|
|
* done correctly.
|
|
|
|
*/
|
|
|
|
mark_reserved_regions_for_nid(nid);
|
2008-10-09 10:18:40 +00:00
|
|
|
sparse_memory_present_with_active_regions(nid);
|
2008-11-24 12:02:35 +00:00
|
|
|
}
|
2009-06-02 21:16:38 +00:00
|
|
|
|
|
|
|
init_bootmem_done = 1;
|
2010-04-26 15:32:43 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Now bootmem is initialised we can create the node to cpumask
|
|
|
|
* lookup tables and setup the cpu callback to populate them.
|
|
|
|
*/
|
|
|
|
setup_node_to_cpumask_map();
|
|
|
|
|
|
|
|
register_cpu_notifier(&ppc64_numa_nb);
|
|
|
|
cpu_numa_callback(&ppc64_numa_nb, CPU_UP_PREPARE,
|
|
|
|
(void *)(unsigned long)boot_cpuid);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void __init paging_init(void)
|
|
|
|
{
|
2006-10-11 08:20:39 +00:00
|
|
|
unsigned long max_zone_pfns[MAX_NR_ZONES];
|
|
|
|
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
|
2010-07-12 04:36:09 +00:00
|
|
|
max_zone_pfns[ZONE_DMA] = memblock_end_of_DRAM() >> PAGE_SHIFT;
|
2006-09-27 08:49:49 +00:00
|
|
|
free_area_init_nodes(max_zone_pfns);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
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 04:57:31 +00:00
|
|
|
p = strstr(p, "fake=");
|
|
|
|
if (p)
|
|
|
|
cmdline = p + strlen("fake=");
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
early_param("numa", early_numa);
|
2005-12-05 20:06:42 +00:00
|
|
|
|
|
|
|
#ifdef CONFIG_MEMORY_HOTPLUG
|
2008-07-03 03:25:08 +00:00
|
|
|
/*
|
2009-02-17 08:08:30 +00:00
|
|
|
* Find the node associated with a hot added memory section for
|
|
|
|
* memory represented in the device tree by the property
|
|
|
|
* ibm,dynamic-reconfiguration-memory/ibm,dynamic-memory.
|
2008-07-03 03:25:08 +00:00
|
|
|
*/
|
|
|
|
static int hot_add_drconf_scn_to_nid(struct device_node *memory,
|
|
|
|
unsigned long scn_addr)
|
|
|
|
{
|
|
|
|
const u32 *dm;
|
2009-02-17 08:08:30 +00:00
|
|
|
unsigned int drconf_cell_cnt, rc;
|
2010-07-23 00:35:52 +00:00
|
|
|
unsigned long lmb_size;
|
2008-07-03 03:25:08 +00:00
|
|
|
struct assoc_arrays aa;
|
2009-02-17 08:08:30 +00:00
|
|
|
int nid = -1;
|
2008-07-03 03:25:08 +00:00
|
|
|
|
2009-02-17 08:08:30 +00:00
|
|
|
drconf_cell_cnt = of_get_drconf_memory(memory, &dm);
|
|
|
|
if (!drconf_cell_cnt)
|
|
|
|
return -1;
|
2008-07-03 03:25:08 +00:00
|
|
|
|
2010-07-23 00:35:52 +00:00
|
|
|
lmb_size = of_get_lmb_size(memory);
|
|
|
|
if (!lmb_size)
|
2009-02-17 08:08:30 +00:00
|
|
|
return -1;
|
2008-07-03 03:25:08 +00:00
|
|
|
|
|
|
|
rc = of_get_assoc_arrays(memory, &aa);
|
|
|
|
if (rc)
|
2009-02-17 08:08:30 +00:00
|
|
|
return -1;
|
2008-07-03 03:25:08 +00:00
|
|
|
|
2009-02-17 08:08:30 +00:00
|
|
|
for (; drconf_cell_cnt != 0; --drconf_cell_cnt) {
|
2008-07-03 03:25:08 +00:00
|
|
|
struct of_drconf_cell drmem;
|
|
|
|
|
|
|
|
read_drconf_cell(&drmem, &dm);
|
|
|
|
|
|
|
|
/* skip this block if it is reserved or not assigned to
|
|
|
|
* this partition */
|
|
|
|
if ((drmem.flags & DRCONF_MEM_RESERVED)
|
|
|
|
|| !(drmem.flags & DRCONF_MEM_ASSIGNED))
|
|
|
|
continue;
|
|
|
|
|
2009-02-17 08:08:30 +00:00
|
|
|
if ((scn_addr < drmem.base_addr)
|
2010-07-23 00:35:52 +00:00
|
|
|
|| (scn_addr >= (drmem.base_addr + lmb_size)))
|
2009-02-17 08:08:30 +00:00
|
|
|
continue;
|
|
|
|
|
2008-07-03 03:25:08 +00:00
|
|
|
nid = of_drconf_to_nid_single(&drmem, &aa);
|
2009-02-17 08:08:30 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return nid;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Find the node associated with a hot added memory section for memory
|
|
|
|
* represented in the device tree as a node (i.e. memory@XXXX) for
|
2010-07-12 04:36:09 +00:00
|
|
|
* each memblock.
|
2009-02-17 08:08:30 +00:00
|
|
|
*/
|
|
|
|
int hot_add_node_scn_to_nid(unsigned long scn_addr)
|
|
|
|
{
|
|
|
|
struct device_node *memory = NULL;
|
|
|
|
int nid = -1;
|
|
|
|
|
|
|
|
while ((memory = of_find_node_by_type(memory, "memory")) != NULL) {
|
|
|
|
unsigned long start, size;
|
|
|
|
int ranges;
|
|
|
|
const unsigned int *memcell_buf;
|
|
|
|
unsigned int len;
|
|
|
|
|
|
|
|
memcell_buf = of_get_property(memory, "reg", &len);
|
|
|
|
if (!memcell_buf || len <= 0)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
/* ranges in cell */
|
|
|
|
ranges = (len >> 2) / (n_mem_addr_cells + n_mem_size_cells);
|
|
|
|
|
|
|
|
while (ranges--) {
|
|
|
|
start = read_n_cells(n_mem_addr_cells, &memcell_buf);
|
|
|
|
size = read_n_cells(n_mem_size_cells, &memcell_buf);
|
|
|
|
|
|
|
|
if ((scn_addr < start) || (scn_addr >= (start + size)))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
nid = of_node_to_nid_single(memory);
|
|
|
|
break;
|
|
|
|
}
|
2008-07-03 03:25:08 +00:00
|
|
|
|
2009-02-17 08:08:30 +00:00
|
|
|
of_node_put(memory);
|
|
|
|
if (nid >= 0)
|
|
|
|
break;
|
2008-07-03 03:25:08 +00:00
|
|
|
}
|
|
|
|
|
2009-02-17 08:08:30 +00:00
|
|
|
return nid;
|
2008-07-03 03:25:08 +00:00
|
|
|
}
|
|
|
|
|
2005-12-05 20:06:42 +00:00
|
|
|
/*
|
|
|
|
* Find the node associated with a hot added memory section. Section
|
2010-07-12 04:36:09 +00:00
|
|
|
* corresponds to a SPARSEMEM section, not an MEMBLOCK. It is assumed that
|
|
|
|
* sections are fully contained within a single MEMBLOCK.
|
2005-12-05 20:06:42 +00:00
|
|
|
*/
|
|
|
|
int hot_add_scn_to_nid(unsigned long scn_addr)
|
|
|
|
{
|
|
|
|
struct device_node *memory = NULL;
|
2009-02-17 08:08:30 +00:00
|
|
|
int nid, found = 0;
|
2005-12-05 20:06:42 +00:00
|
|
|
|
|
|
|
if (!numa_enabled || (min_common_depth < 0))
|
2010-03-05 21:42:43 +00:00
|
|
|
return first_online_node;
|
2008-07-03 03:25:08 +00:00
|
|
|
|
|
|
|
memory = of_find_node_by_path("/ibm,dynamic-reconfiguration-memory");
|
|
|
|
if (memory) {
|
|
|
|
nid = hot_add_drconf_scn_to_nid(memory, scn_addr);
|
|
|
|
of_node_put(memory);
|
2009-02-17 08:08:30 +00:00
|
|
|
} else {
|
|
|
|
nid = hot_add_node_scn_to_nid(scn_addr);
|
2008-07-03 03:25:08 +00:00
|
|
|
}
|
2005-12-05 20:06:42 +00:00
|
|
|
|
2009-02-17 08:08:30 +00:00
|
|
|
if (nid < 0 || !node_online(nid))
|
2010-03-05 21:42:43 +00:00
|
|
|
nid = first_online_node;
|
2005-12-05 20:06:42 +00:00
|
|
|
|
2009-02-17 08:08:30 +00:00
|
|
|
if (NODE_DATA(nid)->node_spanned_pages)
|
|
|
|
return nid;
|
2005-12-05 20:06:42 +00:00
|
|
|
|
2009-02-17 08:08:30 +00:00
|
|
|
for_each_online_node(nid) {
|
|
|
|
if (NODE_DATA(nid)->node_spanned_pages) {
|
|
|
|
found = 1;
|
|
|
|
break;
|
2005-12-05 20:06:42 +00:00
|
|
|
}
|
|
|
|
}
|
2009-02-17 08:08:30 +00:00
|
|
|
|
|
|
|
BUG_ON(!found);
|
|
|
|
return nid;
|
2005-12-05 20:06:42 +00:00
|
|
|
}
|
2009-02-17 08:08:30 +00:00
|
|
|
|
2010-10-26 17:35:12 +00:00
|
|
|
static u64 hot_add_drconf_memory_max(void)
|
|
|
|
{
|
|
|
|
struct device_node *memory = NULL;
|
|
|
|
unsigned int drconf_cell_cnt = 0;
|
|
|
|
u64 lmb_size = 0;
|
|
|
|
const u32 *dm = 0;
|
|
|
|
|
|
|
|
memory = of_find_node_by_path("/ibm,dynamic-reconfiguration-memory");
|
|
|
|
if (memory) {
|
|
|
|
drconf_cell_cnt = of_get_drconf_memory(memory, &dm);
|
|
|
|
lmb_size = of_get_lmb_size(memory);
|
|
|
|
of_node_put(memory);
|
|
|
|
}
|
|
|
|
return lmb_size * drconf_cell_cnt;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* memory_hotplug_max - return max address of memory that may be added
|
|
|
|
*
|
|
|
|
* This is currently only used on systems that support drconfig memory
|
|
|
|
* hotplug.
|
|
|
|
*/
|
|
|
|
u64 memory_hotplug_max(void)
|
|
|
|
{
|
|
|
|
return max(hot_add_drconf_memory_max(), memblock_end_of_DRAM());
|
|
|
|
}
|
2005-12-05 20:06:42 +00:00
|
|
|
#endif /* CONFIG_MEMORY_HOTPLUG */
|
2010-12-01 12:31:15 +00:00
|
|
|
|
2011-01-20 19:00:51 +00:00
|
|
|
/* Virtual Processor Home Node (VPHN) support */
|
2010-12-17 22:07:47 +00:00
|
|
|
#ifdef CONFIG_PPC_SPLPAR
|
2010-12-01 12:31:15 +00:00
|
|
|
#define VPHN_NR_CHANGE_CTRS (8)
|
|
|
|
static u8 vphn_cpu_change_counts[NR_CPUS][VPHN_NR_CHANGE_CTRS];
|
|
|
|
static cpumask_t cpu_associativity_changes_mask;
|
|
|
|
static int vphn_enabled;
|
|
|
|
static void set_topology_timer(void);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Store the current values of the associativity change counters in the
|
|
|
|
* hypervisor.
|
|
|
|
*/
|
|
|
|
static void setup_cpu_associativity_change_counters(void)
|
|
|
|
{
|
2011-01-20 19:01:35 +00:00
|
|
|
int cpu;
|
2010-12-01 12:31:15 +00:00
|
|
|
|
|
|
|
for_each_possible_cpu(cpu) {
|
2011-01-20 19:01:35 +00:00
|
|
|
int i;
|
2010-12-01 12:31:15 +00:00
|
|
|
u8 *counts = vphn_cpu_change_counts[cpu];
|
|
|
|
volatile u8 *hypervisor_counts = lppaca[cpu].vphn_assoc_counts;
|
|
|
|
|
2011-01-20 19:01:13 +00:00
|
|
|
for (i = 0; i < VPHN_NR_CHANGE_CTRS; i++)
|
2010-12-01 12:31:15 +00:00
|
|
|
counts[i] = hypervisor_counts[i];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The hypervisor maintains a set of 8 associativity change counters in
|
|
|
|
* the VPA of each cpu that correspond to the associativity levels in the
|
|
|
|
* ibm,associativity-reference-points property. When an associativity
|
|
|
|
* level changes, the corresponding counter is incremented.
|
|
|
|
*
|
|
|
|
* Set a bit in cpu_associativity_changes_mask for each cpu whose home
|
|
|
|
* node associativity levels have changed.
|
|
|
|
*
|
|
|
|
* Returns the number of cpus with unhandled associativity changes.
|
|
|
|
*/
|
|
|
|
static int update_cpu_associativity_changes_mask(void)
|
|
|
|
{
|
2011-01-20 19:01:35 +00:00
|
|
|
int cpu, nr_cpus = 0;
|
2010-12-01 12:31:15 +00:00
|
|
|
cpumask_t *changes = &cpu_associativity_changes_mask;
|
|
|
|
|
|
|
|
cpumask_clear(changes);
|
|
|
|
|
|
|
|
for_each_possible_cpu(cpu) {
|
|
|
|
int i, changed = 0;
|
|
|
|
u8 *counts = vphn_cpu_change_counts[cpu];
|
|
|
|
volatile u8 *hypervisor_counts = lppaca[cpu].vphn_assoc_counts;
|
|
|
|
|
|
|
|
for (i = 0; i < VPHN_NR_CHANGE_CTRS; i++) {
|
|
|
|
if (hypervisor_counts[i] > counts[i]) {
|
|
|
|
counts[i] = hypervisor_counts[i];
|
|
|
|
changed = 1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (changed) {
|
|
|
|
cpumask_set_cpu(cpu, changes);
|
|
|
|
nr_cpus++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return nr_cpus;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* 6 64-bit registers unpacked into 12 32-bit associativity values */
|
|
|
|
#define VPHN_ASSOC_BUFSIZE (6*sizeof(u64)/sizeof(u32))
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Convert the associativity domain numbers returned from the hypervisor
|
|
|
|
* to the sequence they would appear in the ibm,associativity property.
|
|
|
|
*/
|
|
|
|
static int vphn_unpack_associativity(const long *packed, unsigned int *unpacked)
|
|
|
|
{
|
2011-01-20 19:01:35 +00:00
|
|
|
int i, nr_assoc_doms = 0;
|
2010-12-01 12:31:15 +00:00
|
|
|
const u16 *field = (const u16*) packed;
|
|
|
|
|
|
|
|
#define VPHN_FIELD_UNUSED (0xffff)
|
|
|
|
#define VPHN_FIELD_MSB (0x8000)
|
|
|
|
#define VPHN_FIELD_MASK (~VPHN_FIELD_MSB)
|
|
|
|
|
|
|
|
for (i = 0; i < VPHN_ASSOC_BUFSIZE; i++) {
|
|
|
|
if (*field == VPHN_FIELD_UNUSED) {
|
|
|
|
/* All significant fields processed, and remaining
|
|
|
|
* fields contain the reserved value of all 1's.
|
|
|
|
* Just store them.
|
|
|
|
*/
|
|
|
|
unpacked[i] = *((u32*)field);
|
|
|
|
field += 2;
|
2011-01-20 19:01:13 +00:00
|
|
|
} else if (*field & VPHN_FIELD_MSB) {
|
2010-12-01 12:31:15 +00:00
|
|
|
/* Data is in the lower 15 bits of this field */
|
|
|
|
unpacked[i] = *field & VPHN_FIELD_MASK;
|
|
|
|
field++;
|
|
|
|
nr_assoc_doms++;
|
2011-01-20 19:01:13 +00:00
|
|
|
} else {
|
2010-12-01 12:31:15 +00:00
|
|
|
/* Data is in the lower 15 bits of this field
|
|
|
|
* concatenated with the next 16 bit field
|
|
|
|
*/
|
|
|
|
unpacked[i] = *((u32*)field);
|
|
|
|
field += 2;
|
|
|
|
nr_assoc_doms++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return nr_assoc_doms;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Retrieve the new associativity information for a virtual processor's
|
|
|
|
* home node.
|
|
|
|
*/
|
|
|
|
static long hcall_vphn(unsigned long cpu, unsigned int *associativity)
|
|
|
|
{
|
2011-01-20 19:01:35 +00:00
|
|
|
long rc;
|
2010-12-01 12:31:15 +00:00
|
|
|
long retbuf[PLPAR_HCALL9_BUFSIZE] = {0};
|
|
|
|
u64 flags = 1;
|
|
|
|
int hwcpu = get_hard_smp_processor_id(cpu);
|
|
|
|
|
|
|
|
rc = plpar_hcall9(H_HOME_NODE_ASSOCIATIVITY, retbuf, flags, hwcpu);
|
|
|
|
vphn_unpack_associativity(retbuf, associativity);
|
|
|
|
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
static long vphn_get_associativity(unsigned long cpu,
|
|
|
|
unsigned int *associativity)
|
|
|
|
{
|
2011-01-20 19:01:35 +00:00
|
|
|
long rc;
|
2010-12-01 12:31:15 +00:00
|
|
|
|
|
|
|
rc = hcall_vphn(cpu, associativity);
|
|
|
|
|
|
|
|
switch (rc) {
|
|
|
|
case H_FUNCTION:
|
|
|
|
printk(KERN_INFO
|
|
|
|
"VPHN is not supported. Disabling polling...\n");
|
|
|
|
stop_topology_update();
|
|
|
|
break;
|
|
|
|
case H_HARDWARE:
|
|
|
|
printk(KERN_ERR
|
|
|
|
"hcall_vphn() experienced a hardware fault "
|
|
|
|
"preventing VPHN. Disabling polling...\n");
|
|
|
|
stop_topology_update();
|
|
|
|
}
|
|
|
|
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update the node maps and sysfs entries for each cpu whose home node
|
|
|
|
* has changed.
|
|
|
|
*/
|
|
|
|
int arch_update_cpu_topology(void)
|
|
|
|
{
|
2011-01-20 19:01:35 +00:00
|
|
|
int cpu, nid, old_nid;
|
2010-12-01 12:31:15 +00:00
|
|
|
unsigned int associativity[VPHN_ASSOC_BUFSIZE] = {0};
|
2011-01-20 19:01:35 +00:00
|
|
|
struct sys_device *sysdev;
|
2010-12-01 12:31:15 +00:00
|
|
|
|
|
|
|
for_each_cpu_mask(cpu, cpu_associativity_changes_mask) {
|
|
|
|
vphn_get_associativity(cpu, associativity);
|
|
|
|
nid = associativity_to_nid(associativity);
|
|
|
|
|
|
|
|
if (nid < 0 || !node_online(nid))
|
|
|
|
nid = first_online_node;
|
|
|
|
|
|
|
|
old_nid = numa_cpu_lookup_table[cpu];
|
|
|
|
|
|
|
|
/* Disable hotplug while we update the cpu
|
|
|
|
* masks and sysfs.
|
|
|
|
*/
|
|
|
|
get_online_cpus();
|
|
|
|
unregister_cpu_under_node(cpu, old_nid);
|
|
|
|
unmap_cpu_from_node(cpu);
|
|
|
|
map_cpu_to_node(cpu, nid);
|
|
|
|
register_cpu_under_node(cpu, nid);
|
|
|
|
put_online_cpus();
|
|
|
|
|
|
|
|
sysdev = get_cpu_sysdev(cpu);
|
|
|
|
if (sysdev)
|
|
|
|
kobject_uevent(&sysdev->kobj, KOBJ_CHANGE);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void topology_work_fn(struct work_struct *work)
|
|
|
|
{
|
|
|
|
rebuild_sched_domains();
|
|
|
|
}
|
|
|
|
static DECLARE_WORK(topology_work, topology_work_fn);
|
|
|
|
|
|
|
|
void topology_schedule_update(void)
|
|
|
|
{
|
|
|
|
schedule_work(&topology_work);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void topology_timer_fn(unsigned long ignored)
|
|
|
|
{
|
|
|
|
if (!vphn_enabled)
|
|
|
|
return;
|
|
|
|
if (update_cpu_associativity_changes_mask() > 0)
|
|
|
|
topology_schedule_update();
|
|
|
|
set_topology_timer();
|
|
|
|
}
|
|
|
|
static struct timer_list topology_timer =
|
|
|
|
TIMER_INITIALIZER(topology_timer_fn, 0, 0);
|
|
|
|
|
|
|
|
static void set_topology_timer(void)
|
|
|
|
{
|
|
|
|
topology_timer.data = 0;
|
|
|
|
topology_timer.expires = jiffies + 60 * HZ;
|
|
|
|
add_timer(&topology_timer);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Start polling for VPHN associativity changes.
|
|
|
|
*/
|
|
|
|
int start_topology_update(void)
|
|
|
|
{
|
|
|
|
int rc = 0;
|
|
|
|
|
|
|
|
if (firmware_has_feature(FW_FEATURE_VPHN)) {
|
|
|
|
vphn_enabled = 1;
|
|
|
|
setup_cpu_associativity_change_counters();
|
|
|
|
init_timer_deferrable(&topology_timer);
|
|
|
|
set_topology_timer();
|
|
|
|
rc = 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
__initcall(start_topology_update);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Disable polling for VPHN associativity changes.
|
|
|
|
*/
|
|
|
|
int stop_topology_update(void)
|
|
|
|
{
|
|
|
|
vphn_enabled = 0;
|
|
|
|
return del_timer_sync(&topology_timer);
|
|
|
|
}
|
2010-12-17 22:07:47 +00:00
|
|
|
#endif /* CONFIG_PPC_SPLPAR */
|