2005-04-16 16:20:36 -06:00
|
|
|
/*
|
|
|
|
* Copyright (c) 2000, 2003 Silicon Graphics, Inc. All rights reserved.
|
|
|
|
* Copyright (c) 2001 Intel Corp.
|
|
|
|
* Copyright (c) 2001 Tony Luck <tony.luck@intel.com>
|
|
|
|
* Copyright (c) 2002 NEC Corp.
|
|
|
|
* Copyright (c) 2002 Kimio Suganuma <k-suganuma@da.jp.nec.com>
|
|
|
|
* Copyright (c) 2004 Silicon Graphics, Inc
|
|
|
|
* Russ Anderson <rja@sgi.com>
|
|
|
|
* Jesse Barnes <jbarnes@sgi.com>
|
|
|
|
* Jack Steiner <steiner@sgi.com>
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Platform initialization for Discontig Memory
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/mm.h>
|
2007-08-22 12:34:38 -06:00
|
|
|
#include <linux/nmi.h>
|
2005-04-16 16:20:36 -06:00
|
|
|
#include <linux/swap.h>
|
|
|
|
#include <linux/bootmem.h>
|
|
|
|
#include <linux/acpi.h>
|
|
|
|
#include <linux/efi.h>
|
|
|
|
#include <linux/nodemask.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 02:04:11 -06:00
|
|
|
#include <linux/slab.h>
|
2005-04-16 16:20:36 -06:00
|
|
|
#include <asm/pgalloc.h>
|
|
|
|
#include <asm/tlb.h>
|
|
|
|
#include <asm/meminit.h>
|
|
|
|
#include <asm/numa.h>
|
|
|
|
#include <asm/sections.h>
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Track per-node information needed to setup the boot memory allocator, the
|
|
|
|
* per-node areas, and the real VM.
|
|
|
|
*/
|
|
|
|
struct early_node_data {
|
|
|
|
struct ia64_node_data *node_data;
|
|
|
|
unsigned long pernode_addr;
|
|
|
|
unsigned long pernode_size;
|
|
|
|
unsigned long num_physpages;
|
2007-02-10 02:43:11 -07:00
|
|
|
#ifdef CONFIG_ZONE_DMA
|
2005-04-16 16:20:36 -06:00
|
|
|
unsigned long num_dma_physpages;
|
2007-02-10 02:43:11 -07:00
|
|
|
#endif
|
2005-04-16 16:20:36 -06:00
|
|
|
unsigned long min_pfn;
|
|
|
|
unsigned long max_pfn;
|
|
|
|
};
|
|
|
|
|
|
|
|
static struct early_node_data mem_data[MAX_NUMNODES] __initdata;
|
2005-06-30 10:52:00 -06:00
|
|
|
static nodemask_t memory_less_mask __initdata;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2007-10-17 00:27:27 -06:00
|
|
|
pg_data_t *pgdat_list[MAX_NUMNODES];
|
2006-06-27 03:53:38 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/*
|
|
|
|
* To prevent cache aliasing effects, align per-node structures so that they
|
|
|
|
* start at addresses that are strided by node number.
|
|
|
|
*/
|
2005-12-05 12:56:50 -07:00
|
|
|
#define MAX_NODE_ALIGN_OFFSET (32 * 1024 * 1024)
|
2005-04-16 16:20:36 -06:00
|
|
|
#define NODEDATA_ALIGN(addr, node) \
|
2005-12-05 12:56:50 -07:00
|
|
|
((((addr) + 1024*1024-1) & ~(1024*1024-1)) + \
|
|
|
|
(((node)*PERCPU_PAGE_SIZE) & (MAX_NODE_ALIGN_OFFSET - 1)))
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/**
|
|
|
|
* build_node_maps - callback to setup bootmem structs for each node
|
|
|
|
* @start: physical start of range
|
|
|
|
* @len: length of range
|
|
|
|
* @node: node where this range resides
|
|
|
|
*
|
|
|
|
* We allocate a struct bootmem_data for each piece of memory that we wish to
|
|
|
|
* treat as a virtually contiguous block (i.e. each node). Each such block
|
|
|
|
* must start on an %IA64_GRANULE_SIZE boundary, so we round the address down
|
|
|
|
* if necessary. Any non-existent pages will simply be part of the virtual
|
|
|
|
* memmap. We also update min_low_pfn and max_low_pfn here as we receive
|
|
|
|
* memory ranges from the caller.
|
|
|
|
*/
|
|
|
|
static int __init build_node_maps(unsigned long start, unsigned long len,
|
|
|
|
int node)
|
|
|
|
{
|
2008-07-23 22:28:09 -06:00
|
|
|
unsigned long spfn, epfn, end = start + len;
|
2008-07-23 22:26:55 -06:00
|
|
|
struct bootmem_data *bdp = &bootmem_node_data[node];
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
epfn = GRANULEROUNDUP(end) >> PAGE_SHIFT;
|
2008-07-23 22:28:09 -06:00
|
|
|
spfn = GRANULEROUNDDOWN(start) >> PAGE_SHIFT;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if (!bdp->node_low_pfn) {
|
2008-07-23 22:28:09 -06:00
|
|
|
bdp->node_min_pfn = spfn;
|
2005-04-16 16:20:36 -06:00
|
|
|
bdp->node_low_pfn = epfn;
|
|
|
|
} else {
|
2008-07-23 22:28:09 -06:00
|
|
|
bdp->node_min_pfn = min(spfn, bdp->node_min_pfn);
|
2005-04-16 16:20:36 -06:00
|
|
|
bdp->node_low_pfn = max(epfn, bdp->node_low_pfn);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2005-06-30 10:52:00 -06:00
|
|
|
* early_nr_cpus_node - return number of cpus on a given node
|
2005-04-16 16:20:36 -06:00
|
|
|
* @node: node to check
|
|
|
|
*
|
2005-06-30 10:52:00 -06:00
|
|
|
* Count the number of cpus on @node. We can't use nr_cpus_node() yet because
|
2005-04-16 16:20:36 -06:00
|
|
|
* acpi_boot_init() (which builds the node_to_cpu_mask array) hasn't been
|
2005-06-30 10:52:00 -06:00
|
|
|
* called yet. Note that node 0 will also count all non-existent cpus.
|
2005-04-16 16:20:36 -06:00
|
|
|
*/
|
2006-06-27 03:53:40 -06:00
|
|
|
static int __meminit early_nr_cpus_node(int node)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
int cpu, n = 0;
|
|
|
|
|
2008-04-03 14:17:13 -06:00
|
|
|
for_each_possible_early_cpu(cpu)
|
2005-04-16 16:20:36 -06:00
|
|
|
if (node == node_cpuid[cpu].nid)
|
2005-06-30 10:52:00 -06:00
|
|
|
n++;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
return n;
|
|
|
|
}
|
|
|
|
|
2005-06-30 10:52:00 -06:00
|
|
|
/**
|
|
|
|
* compute_pernodesize - compute size of pernode data
|
|
|
|
* @node: the node id.
|
|
|
|
*/
|
2006-06-27 03:53:40 -06:00
|
|
|
static unsigned long __meminit compute_pernodesize(int node)
|
2005-06-30 10:52:00 -06:00
|
|
|
{
|
|
|
|
unsigned long pernodesize = 0, cpus;
|
|
|
|
|
|
|
|
cpus = early_nr_cpus_node(node);
|
|
|
|
pernodesize += PERCPU_PAGE_SIZE * cpus;
|
|
|
|
pernodesize += node * L1_CACHE_BYTES;
|
|
|
|
pernodesize += L1_CACHE_ALIGN(sizeof(pg_data_t));
|
|
|
|
pernodesize += L1_CACHE_ALIGN(sizeof(struct ia64_node_data));
|
2008-04-03 14:17:12 -06:00
|
|
|
pernodesize += L1_CACHE_ALIGN(sizeof(pg_data_t));
|
2005-06-30 10:52:00 -06:00
|
|
|
pernodesize = PAGE_ALIGN(pernodesize);
|
|
|
|
return pernodesize;
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2005-07-06 19:18:10 -06:00
|
|
|
/**
|
|
|
|
* per_cpu_node_setup - setup per-cpu areas on each node
|
|
|
|
* @cpu_data: per-cpu area on this node
|
|
|
|
* @node: node to setup
|
|
|
|
*
|
|
|
|
* Copy the static per-cpu data into the region we just set aside and then
|
|
|
|
* setup __per_cpu_offset for each CPU on this node. Return a pointer to
|
|
|
|
* the end of the area.
|
|
|
|
*/
|
|
|
|
static void *per_cpu_node_setup(void *cpu_data, int node)
|
|
|
|
{
|
|
|
|
#ifdef CONFIG_SMP
|
|
|
|
int cpu;
|
|
|
|
|
2008-04-03 14:17:13 -06:00
|
|
|
for_each_possible_early_cpu(cpu) {
|
2009-10-01 22:28:56 -06:00
|
|
|
void *src = cpu == 0 ? __cpu0_per_cpu : __phys_per_cpu_start;
|
|
|
|
|
|
|
|
if (node != node_cpuid[cpu].nid)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
memcpy(__va(cpu_data), src, __per_cpu_end - __per_cpu_start);
|
|
|
|
__per_cpu_offset[cpu] = (char *)__va(cpu_data) -
|
|
|
|
__per_cpu_start;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* percpu area for cpu0 is moved from the __init area
|
|
|
|
* which is setup by head.S and used till this point.
|
|
|
|
* Update ar.k3. This move is ensures that percpu
|
|
|
|
* area for cpu0 is on the correct node and its
|
|
|
|
* virtual address isn't insanely far from other
|
|
|
|
* percpu areas which is important for congruent
|
|
|
|
* percpu allocator.
|
|
|
|
*/
|
|
|
|
if (cpu == 0)
|
|
|
|
ia64_set_kr(IA64_KR_PER_CPU_DATA,
|
|
|
|
(unsigned long)cpu_data -
|
|
|
|
(unsigned long)__per_cpu_start);
|
|
|
|
|
|
|
|
cpu_data += PERCPU_PAGE_SIZE;
|
2005-07-06 19:18:10 -06:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
return cpu_data;
|
|
|
|
}
|
|
|
|
|
2009-10-01 22:28:56 -06:00
|
|
|
#ifdef CONFIG_SMP
|
|
|
|
/**
|
|
|
|
* setup_per_cpu_areas - setup percpu areas
|
|
|
|
*
|
|
|
|
* Arch code has already allocated and initialized percpu areas. All
|
|
|
|
* this function has to do is to teach the determined layout to the
|
|
|
|
* dynamic percpu allocator, which happens to be more complex than
|
|
|
|
* creating whole new ones using helpers.
|
|
|
|
*/
|
|
|
|
void __init setup_per_cpu_areas(void)
|
|
|
|
{
|
|
|
|
struct pcpu_alloc_info *ai;
|
|
|
|
struct pcpu_group_info *uninitialized_var(gi);
|
|
|
|
unsigned int *cpu_map;
|
|
|
|
void *base;
|
|
|
|
unsigned long base_offset;
|
|
|
|
unsigned int cpu;
|
|
|
|
ssize_t static_size, reserved_size, dyn_size;
|
|
|
|
int node, prev_node, unit, nr_units, rc;
|
|
|
|
|
|
|
|
ai = pcpu_alloc_alloc_info(MAX_NUMNODES, nr_cpu_ids);
|
|
|
|
if (!ai)
|
|
|
|
panic("failed to allocate pcpu_alloc_info");
|
|
|
|
cpu_map = ai->groups[0].cpu_map;
|
|
|
|
|
|
|
|
/* determine base */
|
|
|
|
base = (void *)ULONG_MAX;
|
|
|
|
for_each_possible_cpu(cpu)
|
|
|
|
base = min(base,
|
|
|
|
(void *)(__per_cpu_offset[cpu] + __per_cpu_start));
|
|
|
|
base_offset = (void *)__per_cpu_start - base;
|
|
|
|
|
|
|
|
/* build cpu_map, units are grouped by node */
|
|
|
|
unit = 0;
|
|
|
|
for_each_node(node)
|
|
|
|
for_each_possible_cpu(cpu)
|
|
|
|
if (node == node_cpuid[cpu].nid)
|
|
|
|
cpu_map[unit++] = cpu;
|
|
|
|
nr_units = unit;
|
|
|
|
|
|
|
|
/* set basic parameters */
|
|
|
|
static_size = __per_cpu_end - __per_cpu_start;
|
|
|
|
reserved_size = PERCPU_MODULE_RESERVE;
|
|
|
|
dyn_size = PERCPU_PAGE_SIZE - static_size - reserved_size;
|
|
|
|
if (dyn_size < 0)
|
|
|
|
panic("percpu area overflow static=%zd reserved=%zd\n",
|
|
|
|
static_size, reserved_size);
|
|
|
|
|
|
|
|
ai->static_size = static_size;
|
|
|
|
ai->reserved_size = reserved_size;
|
|
|
|
ai->dyn_size = dyn_size;
|
|
|
|
ai->unit_size = PERCPU_PAGE_SIZE;
|
|
|
|
ai->atom_size = PAGE_SIZE;
|
|
|
|
ai->alloc_size = PERCPU_PAGE_SIZE;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* CPUs are put into groups according to node. Walk cpu_map
|
|
|
|
* and create new groups at node boundaries.
|
|
|
|
*/
|
|
|
|
prev_node = -1;
|
|
|
|
ai->nr_groups = 0;
|
|
|
|
for (unit = 0; unit < nr_units; unit++) {
|
|
|
|
cpu = cpu_map[unit];
|
|
|
|
node = node_cpuid[cpu].nid;
|
|
|
|
|
|
|
|
if (node == prev_node) {
|
|
|
|
gi->nr_units++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
prev_node = node;
|
|
|
|
|
|
|
|
gi = &ai->groups[ai->nr_groups++];
|
|
|
|
gi->nr_units = 1;
|
|
|
|
gi->base_offset = __per_cpu_offset[cpu] + base_offset;
|
|
|
|
gi->cpu_map = &cpu_map[unit];
|
|
|
|
}
|
|
|
|
|
|
|
|
rc = pcpu_setup_first_chunk(ai, base);
|
|
|
|
if (rc)
|
|
|
|
panic("failed to setup percpu area (err=%d)", rc);
|
|
|
|
|
|
|
|
pcpu_free_alloc_info(ai);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/**
|
2005-06-30 10:52:00 -06:00
|
|
|
* fill_pernode - initialize pernode data.
|
|
|
|
* @node: the node id.
|
|
|
|
* @pernode: physical address of pernode data
|
|
|
|
* @pernodesize: size of the pernode data
|
2005-04-16 16:20:36 -06:00
|
|
|
*/
|
2005-06-30 10:52:00 -06:00
|
|
|
static void __init fill_pernode(int node, unsigned long pernode,
|
|
|
|
unsigned long pernodesize)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2005-06-30 10:52:00 -06:00
|
|
|
void *cpu_data;
|
2005-07-06 19:18:10 -06:00
|
|
|
int cpus = early_nr_cpus_node(node);
|
2008-07-23 22:26:55 -06:00
|
|
|
struct bootmem_data *bdp = &bootmem_node_data[node];
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2005-06-30 10:52:00 -06:00
|
|
|
mem_data[node].pernode_addr = pernode;
|
|
|
|
mem_data[node].pernode_size = pernodesize;
|
|
|
|
memset(__va(pernode), 0, pernodesize);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2005-06-30 10:52:00 -06:00
|
|
|
cpu_data = (void *)pernode;
|
|
|
|
pernode += PERCPU_PAGE_SIZE * cpus;
|
|
|
|
pernode += node * L1_CACHE_BYTES;
|
|
|
|
|
2006-06-27 03:53:38 -06:00
|
|
|
pgdat_list[node] = __va(pernode);
|
2005-06-30 10:52:00 -06:00
|
|
|
pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
|
|
|
|
|
|
|
|
mem_data[node].node_data = __va(pernode);
|
|
|
|
pernode += L1_CACHE_ALIGN(sizeof(struct ia64_node_data));
|
|
|
|
|
2006-06-27 03:53:38 -06:00
|
|
|
pgdat_list[node]->bdata = bdp;
|
2005-06-30 10:52:00 -06:00
|
|
|
pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
|
|
|
|
|
2005-07-06 19:18:10 -06:00
|
|
|
cpu_data = per_cpu_node_setup(cpu_data, node);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2005-06-30 10:52:00 -06:00
|
|
|
return;
|
|
|
|
}
|
2005-07-06 19:18:10 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/**
|
|
|
|
* find_pernode_space - allocate memory for memory map and per-node structures
|
|
|
|
* @start: physical start of range
|
|
|
|
* @len: length of range
|
|
|
|
* @node: node where this range resides
|
|
|
|
*
|
|
|
|
* This routine reserves space for the per-cpu data struct, the list of
|
|
|
|
* pg_data_ts and the per-node data struct. Each node will have something like
|
|
|
|
* the following in the first chunk of addr. space large enough to hold it.
|
|
|
|
*
|
|
|
|
* ________________________
|
|
|
|
* | |
|
|
|
|
* |~~~~~~~~~~~~~~~~~~~~~~~~| <-- NODEDATA_ALIGN(start, node) for the first
|
|
|
|
* | PERCPU_PAGE_SIZE * | start and length big enough
|
|
|
|
* | cpus_on_this_node | Node 0 will also have entries for all non-existent cpus.
|
|
|
|
* |------------------------|
|
|
|
|
* | local pg_data_t * |
|
|
|
|
* |------------------------|
|
|
|
|
* | local ia64_node_data |
|
|
|
|
* |------------------------|
|
|
|
|
* | ??? |
|
|
|
|
* |________________________|
|
|
|
|
*
|
|
|
|
* Once this space has been set aside, the bootmem maps are initialized. We
|
|
|
|
* could probably move the allocation of the per-cpu and ia64_node_data space
|
|
|
|
* outside of this function and use alloc_bootmem_node(), but doing it here
|
|
|
|
* is straightforward and we get the alignments we want so...
|
|
|
|
*/
|
|
|
|
static int __init find_pernode_space(unsigned long start, unsigned long len,
|
|
|
|
int node)
|
|
|
|
{
|
2008-07-23 22:28:09 -06:00
|
|
|
unsigned long spfn, epfn;
|
2005-04-16 16:20:36 -06:00
|
|
|
unsigned long pernodesize = 0, pernode, pages, mapsize;
|
2008-07-23 22:26:55 -06:00
|
|
|
struct bootmem_data *bdp = &bootmem_node_data[node];
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-07-23 22:28:09 -06:00
|
|
|
spfn = start >> PAGE_SHIFT;
|
2005-04-16 16:20:36 -06:00
|
|
|
epfn = (start + len) >> PAGE_SHIFT;
|
|
|
|
|
2008-07-23 22:28:09 -06:00
|
|
|
pages = bdp->node_low_pfn - bdp->node_min_pfn;
|
2005-04-16 16:20:36 -06:00
|
|
|
mapsize = bootmem_bootmap_pages(pages) << PAGE_SHIFT;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Make sure this memory falls within this node's usable memory
|
|
|
|
* since we may have thrown some away in build_maps().
|
|
|
|
*/
|
2008-07-23 22:28:09 -06:00
|
|
|
if (spfn < bdp->node_min_pfn || epfn > bdp->node_low_pfn)
|
2005-04-16 16:20:36 -06:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* Don't setup this node's local space twice... */
|
|
|
|
if (mem_data[node].pernode_addr)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Calculate total size needed, incl. what's necessary
|
|
|
|
* for good alignment and alias prevention.
|
|
|
|
*/
|
2005-06-30 10:52:00 -06:00
|
|
|
pernodesize = compute_pernodesize(node);
|
2005-04-16 16:20:36 -06:00
|
|
|
pernode = NODEDATA_ALIGN(start, node);
|
|
|
|
|
|
|
|
/* Is this range big enough for what we want to store here? */
|
2005-06-30 10:52:00 -06:00
|
|
|
if (start + len > (pernode + pernodesize + mapsize))
|
|
|
|
fill_pernode(node, pernode, pernodesize);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* free_node_bootmem - free bootmem allocator memory for use
|
|
|
|
* @start: physical start of range
|
|
|
|
* @len: length of range
|
|
|
|
* @node: node where this range resides
|
|
|
|
*
|
|
|
|
* Simply calls the bootmem allocator to free the specified ranged from
|
|
|
|
* the given pg_data_t's bdata struct. After this function has been called
|
|
|
|
* for all the entries in the EFI memory map, the bootmem allocator will
|
|
|
|
* be ready to service allocation requests.
|
|
|
|
*/
|
|
|
|
static int __init free_node_bootmem(unsigned long start, unsigned long len,
|
|
|
|
int node)
|
|
|
|
{
|
2006-06-27 03:53:38 -06:00
|
|
|
free_bootmem_node(pgdat_list[node], start, len);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* reserve_pernode_space - reserve memory for per-node space
|
|
|
|
*
|
|
|
|
* Reserve the space used by the bootmem maps & per-node space in the boot
|
|
|
|
* allocator so that when we actually create the real mem maps we don't
|
|
|
|
* use their memory.
|
|
|
|
*/
|
|
|
|
static void __init reserve_pernode_space(void)
|
|
|
|
{
|
|
|
|
unsigned long base, size, pages;
|
|
|
|
struct bootmem_data *bdp;
|
|
|
|
int node;
|
|
|
|
|
|
|
|
for_each_online_node(node) {
|
2006-06-27 03:53:38 -06:00
|
|
|
pg_data_t *pdp = pgdat_list[node];
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2005-06-30 10:52:00 -06:00
|
|
|
if (node_isset(node, memory_less_mask))
|
|
|
|
continue;
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
bdp = pdp->bdata;
|
|
|
|
|
|
|
|
/* First the bootmem_map itself */
|
2008-07-23 22:28:09 -06:00
|
|
|
pages = bdp->node_low_pfn - bdp->node_min_pfn;
|
2005-04-16 16:20:36 -06:00
|
|
|
size = bootmem_bootmap_pages(pages) << PAGE_SHIFT;
|
|
|
|
base = __pa(bdp->node_bootmem_map);
|
2008-02-07 01:15:17 -07:00
|
|
|
reserve_bootmem_node(pdp, base, size, BOOTMEM_DEFAULT);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/* Now the per-node space */
|
|
|
|
size = mem_data[node].pernode_size;
|
|
|
|
base = __pa(mem_data[node].pernode_addr);
|
2008-02-07 01:15:17 -07:00
|
|
|
reserve_bootmem_node(pdp, base, size, BOOTMEM_DEFAULT);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
[PATCH] pgdat allocation and update for ia64 of memory hotplug: update pgdat address array
This is to refresh node_data[] array for ia64. As I mentioned previous
patches, ia64 has copies of information of pgdat address array on each node as
per node data.
At v2 of node_add, this function used stop_machine_run() to update them. (I
wished that they were copied safety as much as possible.) But, in this patch,
this arrays are just copied simply, and set node_online_map bit after
completion of pgdat initialization.
So, kernel must touch NODE_DATA() macro after checking node_online_map().
(Current code has already done it.) This is more simple way for just
hot-add.....
Note : It will be problem when hot-remove will occur,
because, even if online_map bit is set, kernel may
touch NODE_DATA() due to race condition. :-(
Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-27 03:53:39 -06:00
|
|
|
static void __meminit scatter_node_data(void)
|
|
|
|
{
|
|
|
|
pg_data_t **dst;
|
|
|
|
int node;
|
|
|
|
|
2006-07-04 03:57:51 -06:00
|
|
|
/*
|
|
|
|
* for_each_online_node() can't be used at here.
|
|
|
|
* node_online_map is not set for hot-added nodes at this time,
|
|
|
|
* because we are halfway through initialization of the new node's
|
|
|
|
* structures. If for_each_online_node() is used, a new node's
|
2007-05-11 15:55:43 -06:00
|
|
|
* pg_data_ptrs will be not initialized. Instead of using it,
|
2006-07-04 03:57:51 -06:00
|
|
|
* pgdat_list[] is checked.
|
|
|
|
*/
|
|
|
|
for_each_node(node) {
|
|
|
|
if (pgdat_list[node]) {
|
|
|
|
dst = LOCAL_DATA_ADDR(pgdat_list[node])->pg_data_ptrs;
|
|
|
|
memcpy(dst, pgdat_list, sizeof(pgdat_list));
|
|
|
|
}
|
[PATCH] pgdat allocation and update for ia64 of memory hotplug: update pgdat address array
This is to refresh node_data[] array for ia64. As I mentioned previous
patches, ia64 has copies of information of pgdat address array on each node as
per node data.
At v2 of node_add, this function used stop_machine_run() to update them. (I
wished that they were copied safety as much as possible.) But, in this patch,
this arrays are just copied simply, and set node_online_map bit after
completion of pgdat initialization.
So, kernel must touch NODE_DATA() macro after checking node_online_map().
(Current code has already done it.) This is more simple way for just
hot-add.....
Note : It will be problem when hot-remove will occur,
because, even if online_map bit is set, kernel may
touch NODE_DATA() due to race condition. :-(
Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-27 03:53:39 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/**
|
|
|
|
* initialize_pernode_data - fixup per-cpu & per-node pointers
|
|
|
|
*
|
|
|
|
* Each node's per-node area has a copy of the global pg_data_t list, so
|
|
|
|
* we copy that to each node here, as well as setting the per-cpu pointer
|
|
|
|
* to the local node data structure. The active_cpus field of the per-node
|
|
|
|
* structure gets setup by the platform_cpu_init() function later.
|
|
|
|
*/
|
|
|
|
static void __init initialize_pernode_data(void)
|
|
|
|
{
|
2005-07-06 19:18:10 -06:00
|
|
|
int cpu, node;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
[PATCH] pgdat allocation and update for ia64 of memory hotplug: update pgdat address array
This is to refresh node_data[] array for ia64. As I mentioned previous
patches, ia64 has copies of information of pgdat address array on each node as
per node data.
At v2 of node_add, this function used stop_machine_run() to update them. (I
wished that they were copied safety as much as possible.) But, in this patch,
this arrays are just copied simply, and set node_online_map bit after
completion of pgdat initialization.
So, kernel must touch NODE_DATA() macro after checking node_online_map().
(Current code has already done it.) This is more simple way for just
hot-add.....
Note : It will be problem when hot-remove will occur,
because, even if online_map bit is set, kernel may
touch NODE_DATA() due to race condition. :-(
Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-27 03:53:39 -06:00
|
|
|
scatter_node_data();
|
|
|
|
|
2005-07-06 19:18:10 -06:00
|
|
|
#ifdef CONFIG_SMP
|
2005-04-16 16:20:36 -06:00
|
|
|
/* Set the node_data pointer for each per-cpu struct */
|
2008-04-03 14:17:13 -06:00
|
|
|
for_each_possible_early_cpu(cpu) {
|
2005-04-16 16:20:36 -06:00
|
|
|
node = node_cpuid[cpu].nid;
|
2009-10-29 07:34:14 -06:00
|
|
|
per_cpu(ia64_cpu_info, cpu).node_data =
|
|
|
|
mem_data[node].node_data;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2005-07-06 19:18:10 -06:00
|
|
|
#else
|
|
|
|
{
|
|
|
|
struct cpuinfo_ia64 *cpu0_cpu_info;
|
|
|
|
cpu = 0;
|
|
|
|
node = node_cpuid[cpu].nid;
|
|
|
|
cpu0_cpu_info = (struct cpuinfo_ia64 *)(__phys_per_cpu_start +
|
2009-10-29 07:34:15 -06:00
|
|
|
((char *)&ia64_cpu_info - __per_cpu_start));
|
2005-07-06 19:18:10 -06:00
|
|
|
cpu0_cpu_info->node_data = mem_data[node].node_data;
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_SMP */
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2005-06-30 10:52:00 -06:00
|
|
|
/**
|
|
|
|
* memory_less_node_alloc - * attempt to allocate memory on the best NUMA slit
|
|
|
|
* node but fall back to any other node when __alloc_bootmem_node fails
|
|
|
|
* for best.
|
|
|
|
* @nid: node id
|
|
|
|
* @pernodesize: size of this node's pernode data
|
|
|
|
*/
|
2005-10-29 15:23:05 -06:00
|
|
|
static void __init *memory_less_node_alloc(int nid, unsigned long pernodesize)
|
2005-06-30 10:52:00 -06:00
|
|
|
{
|
|
|
|
void *ptr = NULL;
|
|
|
|
u8 best = 0xff;
|
2005-10-29 15:23:05 -06:00
|
|
|
int bestnode = -1, node, anynode = 0;
|
2005-06-30 10:52:00 -06:00
|
|
|
|
|
|
|
for_each_online_node(node) {
|
|
|
|
if (node_isset(node, memory_less_mask))
|
|
|
|
continue;
|
|
|
|
else if (node_distance(nid, node) < best) {
|
|
|
|
best = node_distance(nid, node);
|
|
|
|
bestnode = node;
|
|
|
|
}
|
2005-10-29 15:23:05 -06:00
|
|
|
anynode = node;
|
2005-06-30 10:52:00 -06:00
|
|
|
}
|
|
|
|
|
2005-10-29 15:23:05 -06:00
|
|
|
if (bestnode == -1)
|
|
|
|
bestnode = anynode;
|
|
|
|
|
2006-06-27 03:53:38 -06:00
|
|
|
ptr = __alloc_bootmem_node(pgdat_list[bestnode], pernodesize,
|
2005-10-29 15:23:05 -06:00
|
|
|
PERCPU_PAGE_SIZE, __pa(MAX_DMA_ADDRESS));
|
2005-06-30 10:52:00 -06:00
|
|
|
|
|
|
|
return ptr;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* memory_less_nodes - allocate and initialize CPU only nodes pernode
|
|
|
|
* information.
|
|
|
|
*/
|
|
|
|
static void __init memory_less_nodes(void)
|
|
|
|
{
|
|
|
|
unsigned long pernodesize;
|
|
|
|
void *pernode;
|
|
|
|
int node;
|
|
|
|
|
|
|
|
for_each_node_mask(node, memory_less_mask) {
|
|
|
|
pernodesize = compute_pernodesize(node);
|
2005-10-29 15:23:05 -06:00
|
|
|
pernode = memory_less_node_alloc(node, pernodesize);
|
2005-06-30 10:52:00 -06:00
|
|
|
fill_pernode(node, __pa(pernode), pernodesize);
|
|
|
|
}
|
|
|
|
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/**
|
|
|
|
* find_memory - walk the EFI memory map and setup the bootmem allocator
|
|
|
|
*
|
|
|
|
* Called early in boot to setup the bootmem allocator, and to
|
|
|
|
* allocate the per-cpu and per-node structures.
|
|
|
|
*/
|
|
|
|
void __init find_memory(void)
|
|
|
|
{
|
|
|
|
int node;
|
|
|
|
|
|
|
|
reserve_memory();
|
|
|
|
|
|
|
|
if (num_online_nodes() == 0) {
|
|
|
|
printk(KERN_ERR "node info missing!\n");
|
|
|
|
node_set_online(0);
|
|
|
|
}
|
|
|
|
|
2005-06-30 10:52:00 -06:00
|
|
|
nodes_or(memory_less_mask, memory_less_mask, node_online_map);
|
2005-04-16 16:20:36 -06:00
|
|
|
min_low_pfn = -1;
|
|
|
|
max_low_pfn = 0;
|
|
|
|
|
|
|
|
/* These actually end up getting called by call_pernode_memory() */
|
|
|
|
efi_memmap_walk(filter_rsvd_memory, build_node_maps);
|
|
|
|
efi_memmap_walk(filter_rsvd_memory, find_pernode_space);
|
[IA64] min_low_pfn and max_low_pfn calculation fix
We have seen bad_pte_print when testing crashdump on an SN machine in
recent 2.6.20 kernel. There are tons of bad pte print (pfn < max_low_pfn)
reports when the crash kernel boots up, all those reported bad pages
are inside initmem range; That is because if the crash kernel code and
data happens to be at the beginning of the 1st node. build_node_maps in
discontig.c will bypass reserved regions with filter_rsvd_memory. Since
min_low_pfn is calculated in build_node_map, so in this case, min_low_pfn
will be greater than kernel code and data.
Because pages inside initmem are freed and reused later, we saw
pfn_valid check fail on those pages.
I think this theoretically happen on a normal kernel. When I check
min_low_pfn and max_low_pfn calculation in contig.c and discontig.c.
I found more issues than this.
1. min_low_pfn and max_low_pfn calculation is inconsistent between
contig.c and discontig.c,
min_low_pfn is calculated as the first page number of boot memmap in
contig.c (Why? Though this may work at the most of the time, I don't
think it is the right logic). It is calculated as the lowest physical
memory page number bypass reserved regions in discontig.c.
max_low_pfn is calculated include reserved regions in contig.c. It is
calculated exclude reserved regions in discontig.c.
2. If kernel code and data region is happen to be at the begin or the
end of physical memory, when min_low_pfn and max_low_pfn calculation is
bypassed kernel code and data, pages in initmem will report bad.
3. initrd is also in reserved regions, if it is at the begin or at the
end of physical memory, kernel will refuse to reuse the memory. Because
the virt_addr_valid check in free_initrd_mem.
So it is better to fix and clean up those issues.
Calculate min_low_pfn and max_low_pfn in a consistent way.
Signed-off-by: Zou Nan hai <nanhai.zou@intel.com>
Acked-by: Jay Lan <jlan@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
2007-03-20 14:41:57 -06:00
|
|
|
efi_memmap_walk(find_max_min_low_pfn, NULL);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2005-06-30 10:52:00 -06:00
|
|
|
for_each_online_node(node)
|
2008-07-23 22:26:55 -06:00
|
|
|
if (bootmem_node_data[node].node_low_pfn) {
|
2005-06-30 10:52:00 -06:00
|
|
|
node_clear(node, memory_less_mask);
|
|
|
|
mem_data[node].min_pfn = ~0UL;
|
|
|
|
}
|
2007-01-30 03:11:09 -07:00
|
|
|
|
2008-04-11 16:21:35 -06:00
|
|
|
efi_memmap_walk(filter_memory, register_active_ranges);
|
2007-01-30 03:11:09 -07:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/*
|
|
|
|
* Initialize the boot memory maps in reverse order since that's
|
|
|
|
* what the bootmem allocator expects
|
|
|
|
*/
|
|
|
|
for (node = MAX_NUMNODES - 1; node >= 0; node--) {
|
|
|
|
unsigned long pernode, pernodesize, map;
|
|
|
|
struct bootmem_data *bdp;
|
|
|
|
|
|
|
|
if (!node_online(node))
|
|
|
|
continue;
|
2005-06-30 10:52:00 -06:00
|
|
|
else if (node_isset(node, memory_less_mask))
|
|
|
|
continue;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-07-23 22:26:55 -06:00
|
|
|
bdp = &bootmem_node_data[node];
|
2005-04-16 16:20:36 -06:00
|
|
|
pernode = mem_data[node].pernode_addr;
|
|
|
|
pernodesize = mem_data[node].pernode_size;
|
|
|
|
map = pernode + pernodesize;
|
|
|
|
|
2006-06-27 03:53:38 -06:00
|
|
|
init_bootmem_node(pgdat_list[node],
|
2005-04-16 16:20:36 -06:00
|
|
|
map>>PAGE_SHIFT,
|
2008-07-23 22:28:09 -06:00
|
|
|
bdp->node_min_pfn,
|
2005-04-16 16:20:36 -06:00
|
|
|
bdp->node_low_pfn);
|
|
|
|
}
|
|
|
|
|
|
|
|
efi_memmap_walk(filter_rsvd_memory, free_node_bootmem);
|
|
|
|
|
|
|
|
reserve_pernode_space();
|
2005-06-30 10:52:00 -06:00
|
|
|
memory_less_nodes();
|
2005-04-16 16:20:36 -06:00
|
|
|
initialize_pernode_data();
|
|
|
|
|
|
|
|
max_pfn = max_low_pfn;
|
|
|
|
|
|
|
|
find_initrd();
|
|
|
|
}
|
|
|
|
|
2005-07-06 19:18:10 -06:00
|
|
|
#ifdef CONFIG_SMP
|
2005-04-16 16:20:36 -06:00
|
|
|
/**
|
|
|
|
* per_cpu_init - setup per-cpu variables
|
|
|
|
*
|
|
|
|
* find_pernode_space() does most of this already, we just need to set
|
|
|
|
* local_per_cpu_offset
|
|
|
|
*/
|
2006-03-12 10:00:13 -07:00
|
|
|
void __cpuinit *per_cpu_init(void)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
int cpu;
|
2005-11-11 15:32:40 -07:00
|
|
|
static int first_time = 1;
|
|
|
|
|
|
|
|
if (first_time) {
|
|
|
|
first_time = 0;
|
2008-04-03 14:17:13 -06:00
|
|
|
for_each_possible_early_cpu(cpu)
|
2005-11-11 15:32:40 -07:00
|
|
|
per_cpu(local_per_cpu_offset, cpu) = __per_cpu_offset[cpu];
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
return __per_cpu_start + __per_cpu_offset[smp_processor_id()];
|
|
|
|
}
|
2005-07-06 19:18:10 -06:00
|
|
|
#endif /* CONFIG_SMP */
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/**
|
|
|
|
* show_mem - give short summary of memory stats
|
|
|
|
*
|
|
|
|
* Shows a simple page count of reserved and used pages in the system.
|
|
|
|
* For discontig machines, it does this on a per-pgdat basis.
|
|
|
|
*/
|
|
|
|
void show_mem(void)
|
|
|
|
{
|
|
|
|
int i, total_reserved = 0;
|
|
|
|
int total_shared = 0, total_cached = 0;
|
|
|
|
unsigned long total_present = 0;
|
|
|
|
pg_data_t *pgdat;
|
|
|
|
|
2006-09-13 06:43:42 -06:00
|
|
|
printk(KERN_INFO "Mem-info:\n");
|
2005-04-16 16:20:36 -06:00
|
|
|
show_free_areas();
|
2006-09-14 05:10:30 -06:00
|
|
|
printk(KERN_INFO "Node memory in pages:\n");
|
2006-03-27 02:15:59 -07:00
|
|
|
for_each_online_pgdat(pgdat) {
|
2005-10-29 19:16:52 -06:00
|
|
|
unsigned long present;
|
|
|
|
unsigned long flags;
|
2005-04-16 16:20:36 -06:00
|
|
|
int shared = 0, cached = 0, reserved = 0;
|
2005-10-29 19:16:52 -06:00
|
|
|
|
|
|
|
pgdat_resize_lock(pgdat, &flags);
|
|
|
|
present = pgdat->node_present_pages;
|
2005-04-16 16:20:36 -06:00
|
|
|
for(i = 0; i < pgdat->node_spanned_pages; i++) {
|
2005-10-04 13:13:57 -06:00
|
|
|
struct page *page;
|
2007-08-22 12:34:38 -06:00
|
|
|
if (unlikely(i % MAX_ORDER_NR_PAGES == 0))
|
|
|
|
touch_nmi_watchdog();
|
2005-10-04 13:13:57 -06:00
|
|
|
if (pfn_valid(pgdat->node_start_pfn + i))
|
|
|
|
page = pfn_to_page(pgdat->node_start_pfn + i);
|
2006-04-13 16:34:45 -06:00
|
|
|
else {
|
2006-06-28 10:55:43 -06:00
|
|
|
i = vmemmap_find_next_valid_pfn(pgdat->node_id,
|
|
|
|
i) - 1;
|
2005-04-16 16:20:36 -06:00
|
|
|
continue;
|
2006-04-13 16:34:45 -06:00
|
|
|
}
|
[PATCH] remove non-DISCONTIG use of pgdat->node_mem_map
This patch effectively eliminates direct use of pgdat->node_mem_map outside
of the DISCONTIG code. On a flat memory system, these fields aren't
currently used, neither are they on a sparsemem system.
There was also a node_mem_map(nid) macro on many architectures. Its use
along with the use of ->node_mem_map itself was not consistent. It has
been removed in favor of two new, more explicit, arch-independent macros:
pgdat_page_nr(pgdat, pagenr)
nid_page_nr(nid, pagenr)
I called them "pgdat" and "nid" because we overload the term "node" to mean
"NUMA node", "DISCONTIG node" or "pg_data_t" in very confusing ways. I
believe the newer names are much clearer.
These macros can be overridden in the sparsemem case with a theoretically
slower operation using node_start_pfn and pfn_to_page(), instead. We could
make this the only behavior if people want, but I don't want to change too
much at once. One thing at a time.
This patch removes more code than it adds.
Compile tested on alpha, alpha discontig, arm, arm-discontig, i386, i386
generic, NUMAQ, Summit, ppc64, ppc64 discontig, and x86_64. Full list
here: http://sr71.net/patches/2.6.12/2.6.12-rc1-mhp2/configs/
Boot tested on NUMAQ, x86 SMP and ppc64 power4/5 LPARs.
Signed-off-by: Dave Hansen <haveblue@us.ibm.com>
Signed-off-by: Martin J. Bligh <mbligh@aracnet.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 01:07:37 -06:00
|
|
|
if (PageReserved(page))
|
2005-04-16 16:20:36 -06:00
|
|
|
reserved++;
|
[PATCH] remove non-DISCONTIG use of pgdat->node_mem_map
This patch effectively eliminates direct use of pgdat->node_mem_map outside
of the DISCONTIG code. On a flat memory system, these fields aren't
currently used, neither are they on a sparsemem system.
There was also a node_mem_map(nid) macro on many architectures. Its use
along with the use of ->node_mem_map itself was not consistent. It has
been removed in favor of two new, more explicit, arch-independent macros:
pgdat_page_nr(pgdat, pagenr)
nid_page_nr(nid, pagenr)
I called them "pgdat" and "nid" because we overload the term "node" to mean
"NUMA node", "DISCONTIG node" or "pg_data_t" in very confusing ways. I
believe the newer names are much clearer.
These macros can be overridden in the sparsemem case with a theoretically
slower operation using node_start_pfn and pfn_to_page(), instead. We could
make this the only behavior if people want, but I don't want to change too
much at once. One thing at a time.
This patch removes more code than it adds.
Compile tested on alpha, alpha discontig, arm, arm-discontig, i386, i386
generic, NUMAQ, Summit, ppc64, ppc64 discontig, and x86_64. Full list
here: http://sr71.net/patches/2.6.12/2.6.12-rc1-mhp2/configs/
Boot tested on NUMAQ, x86 SMP and ppc64 power4/5 LPARs.
Signed-off-by: Dave Hansen <haveblue@us.ibm.com>
Signed-off-by: Martin J. Bligh <mbligh@aracnet.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 01:07:37 -06:00
|
|
|
else if (PageSwapCache(page))
|
2005-04-16 16:20:36 -06:00
|
|
|
cached++;
|
[PATCH] remove non-DISCONTIG use of pgdat->node_mem_map
This patch effectively eliminates direct use of pgdat->node_mem_map outside
of the DISCONTIG code. On a flat memory system, these fields aren't
currently used, neither are they on a sparsemem system.
There was also a node_mem_map(nid) macro on many architectures. Its use
along with the use of ->node_mem_map itself was not consistent. It has
been removed in favor of two new, more explicit, arch-independent macros:
pgdat_page_nr(pgdat, pagenr)
nid_page_nr(nid, pagenr)
I called them "pgdat" and "nid" because we overload the term "node" to mean
"NUMA node", "DISCONTIG node" or "pg_data_t" in very confusing ways. I
believe the newer names are much clearer.
These macros can be overridden in the sparsemem case with a theoretically
slower operation using node_start_pfn and pfn_to_page(), instead. We could
make this the only behavior if people want, but I don't want to change too
much at once. One thing at a time.
This patch removes more code than it adds.
Compile tested on alpha, alpha discontig, arm, arm-discontig, i386, i386
generic, NUMAQ, Summit, ppc64, ppc64 discontig, and x86_64. Full list
here: http://sr71.net/patches/2.6.12/2.6.12-rc1-mhp2/configs/
Boot tested on NUMAQ, x86 SMP and ppc64 power4/5 LPARs.
Signed-off-by: Dave Hansen <haveblue@us.ibm.com>
Signed-off-by: Martin J. Bligh <mbligh@aracnet.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 01:07:37 -06:00
|
|
|
else if (page_count(page))
|
|
|
|
shared += page_count(page)-1;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2005-10-29 19:16:52 -06:00
|
|
|
pgdat_resize_unlock(pgdat, &flags);
|
2005-04-16 16:20:36 -06:00
|
|
|
total_present += present;
|
|
|
|
total_reserved += reserved;
|
|
|
|
total_cached += cached;
|
|
|
|
total_shared += shared;
|
2006-09-14 05:10:30 -06:00
|
|
|
printk(KERN_INFO "Node %4d: RAM: %11ld, rsvd: %8d, "
|
|
|
|
"shrd: %10d, swpd: %10d\n", pgdat->node_id,
|
|
|
|
present, reserved, shared, cached);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2006-09-13 06:43:42 -06:00
|
|
|
printk(KERN_INFO "%ld pages of RAM\n", total_present);
|
|
|
|
printk(KERN_INFO "%d reserved pages\n", total_reserved);
|
|
|
|
printk(KERN_INFO "%d pages shared\n", total_shared);
|
|
|
|
printk(KERN_INFO "%d pages swap cached\n", total_cached);
|
|
|
|
printk(KERN_INFO "Total of %ld pages in page table cache\n",
|
2007-05-10 23:42:53 -06:00
|
|
|
quicklist_total_size());
|
2006-09-13 06:43:42 -06:00
|
|
|
printk(KERN_INFO "%d free buffer pages\n", nr_free_buffer_pages());
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* call_pernode_memory - use SRAT to call callback functions with node info
|
|
|
|
* @start: physical start of range
|
|
|
|
* @len: length of range
|
|
|
|
* @arg: function to call for each range
|
|
|
|
*
|
|
|
|
* efi_memmap_walk() knows nothing about layout of memory across nodes. Find
|
|
|
|
* out to which node a block of memory belongs. Ignore memory that we cannot
|
|
|
|
* identify, and split blocks that run across multiple nodes.
|
|
|
|
*
|
|
|
|
* Take this opportunity to round the start address up and the end address
|
|
|
|
* down to page boundaries.
|
|
|
|
*/
|
|
|
|
void call_pernode_memory(unsigned long start, unsigned long len, void *arg)
|
|
|
|
{
|
|
|
|
unsigned long rs, re, end = start + len;
|
|
|
|
void (*func)(unsigned long, unsigned long, int);
|
|
|
|
int i;
|
|
|
|
|
|
|
|
start = PAGE_ALIGN(start);
|
|
|
|
end &= PAGE_MASK;
|
|
|
|
if (start >= end)
|
|
|
|
return;
|
|
|
|
|
|
|
|
func = arg;
|
|
|
|
|
|
|
|
if (!num_node_memblks) {
|
|
|
|
/* No SRAT table, so assume one node (node 0) */
|
|
|
|
if (start < end)
|
|
|
|
(*func)(start, end - start, 0);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < num_node_memblks; i++) {
|
|
|
|
rs = max(start, node_memblk[i].start_paddr);
|
|
|
|
re = min(end, node_memblk[i].start_paddr +
|
|
|
|
node_memblk[i].size);
|
|
|
|
|
|
|
|
if (rs < re)
|
|
|
|
(*func)(rs, re - rs, node_memblk[i].nid);
|
|
|
|
|
|
|
|
if (re == end)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* count_node_pages - callback to build per-node memory info structures
|
|
|
|
* @start: physical start of range
|
|
|
|
* @len: length of range
|
|
|
|
* @node: node where this range resides
|
|
|
|
*
|
|
|
|
* Each node has it's own number of physical pages, DMAable pages, start, and
|
|
|
|
* end page frame number. This routine will be called by call_pernode_memory()
|
|
|
|
* for each piece of usable memory and will setup these values for each node.
|
|
|
|
* Very similar to build_maps().
|
|
|
|
*/
|
|
|
|
static __init int count_node_pages(unsigned long start, unsigned long len, int node)
|
|
|
|
{
|
|
|
|
unsigned long end = start + len;
|
|
|
|
|
|
|
|
mem_data[node].num_physpages += len >> PAGE_SHIFT;
|
2007-02-10 02:43:11 -07:00
|
|
|
#ifdef CONFIG_ZONE_DMA
|
2005-04-16 16:20:36 -06:00
|
|
|
if (start <= __pa(MAX_DMA_ADDRESS))
|
|
|
|
mem_data[node].num_dma_physpages +=
|
|
|
|
(min(end, __pa(MAX_DMA_ADDRESS)) - start) >>PAGE_SHIFT;
|
2007-02-10 02:43:11 -07:00
|
|
|
#endif
|
2005-04-16 16:20:36 -06:00
|
|
|
start = GRANULEROUNDDOWN(start);
|
|
|
|
end = GRANULEROUNDUP(end);
|
|
|
|
mem_data[node].max_pfn = max(mem_data[node].max_pfn,
|
|
|
|
end >> PAGE_SHIFT);
|
|
|
|
mem_data[node].min_pfn = min(mem_data[node].min_pfn,
|
|
|
|
start >> PAGE_SHIFT);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* paging_init - setup page tables
|
|
|
|
*
|
|
|
|
* paging_init() sets up the page tables for each node of the system and frees
|
|
|
|
* the bootmem allocator memory for general use.
|
|
|
|
*/
|
|
|
|
void __init paging_init(void)
|
|
|
|
{
|
|
|
|
unsigned long max_dma;
|
|
|
|
unsigned long pfn_offset = 0;
|
2006-09-27 02:49:54 -06:00
|
|
|
unsigned long max_pfn = 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
int node;
|
2006-09-27 02:49:54 -06:00
|
|
|
unsigned long max_zone_pfns[MAX_NR_ZONES];
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
max_dma = virt_to_phys((void *) MAX_DMA_ADDRESS) >> PAGE_SHIFT;
|
|
|
|
|
|
|
|
efi_memmap_walk(filter_rsvd_memory, count_node_pages);
|
|
|
|
|
2007-02-05 17:20:08 -07:00
|
|
|
sparse_memory_present_with_active_regions(MAX_NUMNODES);
|
|
|
|
sparse_init();
|
|
|
|
|
2005-10-04 13:13:57 -06:00
|
|
|
#ifdef CONFIG_VIRTUAL_MEM_MAP
|
2009-10-01 22:28:55 -06:00
|
|
|
VMALLOC_END -= PAGE_ALIGN(ALIGN(max_low_pfn, MAX_ORDER_NR_PAGES) *
|
2006-06-28 10:54:55 -06:00
|
|
|
sizeof(struct page));
|
2009-10-01 22:28:55 -06:00
|
|
|
vmem_map = (struct page *) VMALLOC_END;
|
2005-06-30 10:52:00 -06:00
|
|
|
efi_memmap_walk(create_mem_map_page_table, NULL);
|
|
|
|
printk("Virtual mem_map starts at 0x%p\n", vmem_map);
|
2005-10-04 13:13:57 -06:00
|
|
|
#endif
|
2005-06-30 10:52:00 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
for_each_online_node(node) {
|
|
|
|
num_physpages += mem_data[node].num_physpages;
|
|
|
|
pfn_offset = mem_data[node].min_pfn;
|
|
|
|
|
2005-10-04 13:13:57 -06:00
|
|
|
#ifdef CONFIG_VIRTUAL_MEM_MAP
|
2005-04-16 16:20:36 -06:00
|
|
|
NODE_DATA(node)->node_mem_map = vmem_map + pfn_offset;
|
2005-10-04 13:13:57 -06:00
|
|
|
#endif
|
2006-09-27 02:49:54 -06:00
|
|
|
if (mem_data[node].max_pfn > max_pfn)
|
|
|
|
max_pfn = mem_data[node].max_pfn;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2006-10-11 02:20:39 -06:00
|
|
|
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
|
2007-02-10 02:43:11 -07:00
|
|
|
#ifdef CONFIG_ZONE_DMA
|
2006-09-27 02:49:54 -06:00
|
|
|
max_zone_pfns[ZONE_DMA] = max_dma;
|
2007-02-10 02:43:11 -07:00
|
|
|
#endif
|
2006-09-27 02:49:54 -06:00
|
|
|
max_zone_pfns[ZONE_NORMAL] = max_pfn;
|
|
|
|
free_area_init_nodes(max_zone_pfns);
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
zero_page_memmap_ptr = virt_to_page(ia64_imva(empty_zero_page));
|
|
|
|
}
|
[PATCH] pgdat allocation and update for ia64 of memory hotplug: update pgdat address array
This is to refresh node_data[] array for ia64. As I mentioned previous
patches, ia64 has copies of information of pgdat address array on each node as
per node data.
At v2 of node_add, this function used stop_machine_run() to update them. (I
wished that they were copied safety as much as possible.) But, in this patch,
this arrays are just copied simply, and set node_online_map bit after
completion of pgdat initialization.
So, kernel must touch NODE_DATA() macro after checking node_online_map().
(Current code has already done it.) This is more simple way for just
hot-add.....
Note : It will be problem when hot-remove will occur,
because, even if online_map bit is set, kernel may
touch NODE_DATA() due to race condition. :-(
Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-27 03:53:39 -06:00
|
|
|
|
2007-05-08 01:23:07 -06:00
|
|
|
#ifdef CONFIG_MEMORY_HOTPLUG
|
2006-06-27 03:53:40 -06:00
|
|
|
pg_data_t *arch_alloc_nodedata(int nid)
|
|
|
|
{
|
|
|
|
unsigned long size = compute_pernodesize(nid);
|
|
|
|
|
|
|
|
return kzalloc(size, GFP_KERNEL);
|
|
|
|
}
|
|
|
|
|
|
|
|
void arch_free_nodedata(pg_data_t *pgdat)
|
|
|
|
{
|
|
|
|
kfree(pgdat);
|
|
|
|
}
|
|
|
|
|
[PATCH] pgdat allocation and update for ia64 of memory hotplug: update pgdat address array
This is to refresh node_data[] array for ia64. As I mentioned previous
patches, ia64 has copies of information of pgdat address array on each node as
per node data.
At v2 of node_add, this function used stop_machine_run() to update them. (I
wished that they were copied safety as much as possible.) But, in this patch,
this arrays are just copied simply, and set node_online_map bit after
completion of pgdat initialization.
So, kernel must touch NODE_DATA() macro after checking node_online_map().
(Current code has already done it.) This is more simple way for just
hot-add.....
Note : It will be problem when hot-remove will occur,
because, even if online_map bit is set, kernel may
touch NODE_DATA() due to race condition. :-(
Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-27 03:53:39 -06:00
|
|
|
void arch_refresh_nodedata(int update_node, pg_data_t *update_pgdat)
|
|
|
|
{
|
|
|
|
pgdat_list[update_node] = update_pgdat;
|
|
|
|
scatter_node_data();
|
|
|
|
}
|
2007-05-08 01:23:07 -06:00
|
|
|
#endif
|
2007-10-16 02:24:15 -06:00
|
|
|
|
|
|
|
#ifdef CONFIG_SPARSEMEM_VMEMMAP
|
|
|
|
int __meminit vmemmap_populate(struct page *start_page,
|
|
|
|
unsigned long size, int node)
|
|
|
|
{
|
|
|
|
return vmemmap_populate_basepages(start_page, size, node);
|
|
|
|
}
|
|
|
|
#endif
|