ecryptfs: fixed testing of file descriptor flags
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / arch / x86 / mm / srat_32.c
blobf16434568a51da26ea8524ae11fa84f0c7405717
1 /*
2 * Some of the code in this file has been gleaned from the 64 bit
3 * discontigmem support code base.
5 * Copyright (C) 2002, IBM Corp.
7 * All rights reserved.
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful, but
15 * WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
17 * NON INFRINGEMENT. See the GNU General Public License for more
18 * details.
20 * You should have received a copy of the GNU General Public License
21 * along with this program; if not, write to the Free Software
22 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
24 * Send feedback to Pat Gaughen <gone@us.ibm.com>
26 #include <linux/mm.h>
27 #include <linux/bootmem.h>
28 #include <linux/memblock.h>
29 #include <linux/mmzone.h>
30 #include <linux/acpi.h>
31 #include <linux/nodemask.h>
32 #include <asm/srat.h>
33 #include <asm/topology.h>
34 #include <asm/smp.h>
35 #include <asm/e820.h>
38 * proximity macros and definitions
40 #define NODE_ARRAY_INDEX(x) ((x) / 8) /* 8 bits/char */
41 #define NODE_ARRAY_OFFSET(x) ((x) % 8) /* 8 bits/char */
42 #define BMAP_SET(bmap, bit) ((bmap)[NODE_ARRAY_INDEX(bit)] |= 1 << NODE_ARRAY_OFFSET(bit))
43 #define BMAP_TEST(bmap, bit) ((bmap)[NODE_ARRAY_INDEX(bit)] & (1 << NODE_ARRAY_OFFSET(bit)))
44 /* bitmap length; _PXM is at most 255 */
45 #define PXM_BITMAP_LEN (MAX_PXM_DOMAINS / 8)
46 static u8 __initdata pxm_bitmap[PXM_BITMAP_LEN]; /* bitmap of proximity domains */
48 #define MAX_CHUNKS_PER_NODE 3
49 #define MAXCHUNKS (MAX_CHUNKS_PER_NODE * MAX_NUMNODES)
50 struct node_memory_chunk_s {
51 unsigned long start_pfn;
52 unsigned long end_pfn;
53 u8 pxm; // proximity domain of node
54 u8 nid; // which cnode contains this chunk?
55 u8 bank; // which mem bank on this node
57 static struct node_memory_chunk_s __initdata node_memory_chunk[MAXCHUNKS];
59 static int __initdata num_memory_chunks; /* total number of memory chunks */
60 static u8 __initdata apicid_to_pxm[MAX_APICID];
62 int numa_off __initdata;
63 int acpi_numa __initdata;
65 static __init void bad_srat(void)
67 printk(KERN_ERR "SRAT: SRAT not used.\n");
68 acpi_numa = -1;
69 num_memory_chunks = 0;
72 static __init inline int srat_disabled(void)
74 return numa_off || acpi_numa < 0;
77 /* Identify CPU proximity domains */
78 void __init
79 acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *cpu_affinity)
81 if (srat_disabled())
82 return;
83 if (cpu_affinity->header.length !=
84 sizeof(struct acpi_srat_cpu_affinity)) {
85 bad_srat();
86 return;
89 if ((cpu_affinity->flags & ACPI_SRAT_CPU_ENABLED) == 0)
90 return; /* empty entry */
92 /* mark this node as "seen" in node bitmap */
93 BMAP_SET(pxm_bitmap, cpu_affinity->proximity_domain_lo);
95 /* don't need to check apic_id here, because it is always 8 bits */
96 apicid_to_pxm[cpu_affinity->apic_id] = cpu_affinity->proximity_domain_lo;
98 printk(KERN_DEBUG "CPU %02x in proximity domain %02x\n",
99 cpu_affinity->apic_id, cpu_affinity->proximity_domain_lo);
103 * Identify memory proximity domains and hot-remove capabilities.
104 * Fill node memory chunk list structure.
106 void __init
107 acpi_numa_memory_affinity_init(struct acpi_srat_mem_affinity *memory_affinity)
109 unsigned long long paddr, size;
110 unsigned long start_pfn, end_pfn;
111 u8 pxm;
112 struct node_memory_chunk_s *p, *q, *pend;
114 if (srat_disabled())
115 return;
116 if (memory_affinity->header.length !=
117 sizeof(struct acpi_srat_mem_affinity)) {
118 bad_srat();
119 return;
122 if ((memory_affinity->flags & ACPI_SRAT_MEM_ENABLED) == 0)
123 return; /* empty entry */
125 pxm = memory_affinity->proximity_domain & 0xff;
127 /* mark this node as "seen" in node bitmap */
128 BMAP_SET(pxm_bitmap, pxm);
130 /* calculate info for memory chunk structure */
131 paddr = memory_affinity->base_address;
132 size = memory_affinity->length;
134 start_pfn = paddr >> PAGE_SHIFT;
135 end_pfn = (paddr + size) >> PAGE_SHIFT;
138 if (num_memory_chunks >= MAXCHUNKS) {
139 printk(KERN_WARNING "Too many mem chunks in SRAT."
140 " Ignoring %lld MBytes at %llx\n",
141 size/(1024*1024), paddr);
142 return;
145 /* Insertion sort based on base address */
146 pend = &node_memory_chunk[num_memory_chunks];
147 for (p = &node_memory_chunk[0]; p < pend; p++) {
148 if (start_pfn < p->start_pfn)
149 break;
151 if (p < pend) {
152 for (q = pend; q >= p; q--)
153 *(q + 1) = *q;
155 p->start_pfn = start_pfn;
156 p->end_pfn = end_pfn;
157 p->pxm = pxm;
159 num_memory_chunks++;
161 printk(KERN_DEBUG "Memory range %08lx to %08lx"
162 " in proximity domain %02x %s\n",
163 start_pfn, end_pfn,
164 pxm,
165 ((memory_affinity->flags & ACPI_SRAT_MEM_HOT_PLUGGABLE) ?
166 "enabled and removable" : "enabled" ) );
169 /* Callback for SLIT parsing */
170 void __init acpi_numa_slit_init(struct acpi_table_slit *slit)
174 void acpi_numa_arch_fixup(void)
178 * The SRAT table always lists ascending addresses, so can always
179 * assume that the first "start" address that you see is the real
180 * start of the node, and that the current "end" address is after
181 * the previous one.
183 static __init int node_read_chunk(int nid, struct node_memory_chunk_s *memory_chunk)
186 * Only add present memory as told by the e820.
187 * There is no guarantee from the SRAT that the memory it
188 * enumerates is present at boot time because it represents
189 * *possible* memory hotplug areas the same as normal RAM.
191 if (memory_chunk->start_pfn >= max_pfn) {
192 printk(KERN_INFO "Ignoring SRAT pfns: %08lx - %08lx\n",
193 memory_chunk->start_pfn, memory_chunk->end_pfn);
194 return -1;
196 if (memory_chunk->nid != nid)
197 return -1;
199 if (!node_has_online_mem(nid))
200 node_start_pfn[nid] = memory_chunk->start_pfn;
202 if (node_start_pfn[nid] > memory_chunk->start_pfn)
203 node_start_pfn[nid] = memory_chunk->start_pfn;
205 if (node_end_pfn[nid] < memory_chunk->end_pfn)
206 node_end_pfn[nid] = memory_chunk->end_pfn;
208 return 0;
211 int __init get_memcfg_from_srat(void)
213 int i, j, nid;
216 if (srat_disabled())
217 goto out_fail;
219 if (num_memory_chunks == 0) {
220 printk(KERN_DEBUG
221 "could not find any ACPI SRAT memory areas.\n");
222 goto out_fail;
225 /* Calculate total number of nodes in system from PXM bitmap and create
226 * a set of sequential node IDs starting at zero. (ACPI doesn't seem
227 * to specify the range of _PXM values.)
230 * MCD - we no longer HAVE to number nodes sequentially. PXM domain
231 * numbers could go as high as 256, and MAX_NUMNODES for i386 is typically
232 * 32, so we will continue numbering them in this manner until MAX_NUMNODES
233 * approaches MAX_PXM_DOMAINS for i386.
235 nodes_clear(node_online_map);
236 for (i = 0; i < MAX_PXM_DOMAINS; i++) {
237 if (BMAP_TEST(pxm_bitmap, i)) {
238 int nid = acpi_map_pxm_to_node(i);
239 node_set_online(nid);
242 BUG_ON(num_online_nodes() == 0);
244 /* set cnode id in memory chunk structure */
245 for (i = 0; i < num_memory_chunks; i++)
246 node_memory_chunk[i].nid = pxm_to_node(node_memory_chunk[i].pxm);
248 printk(KERN_DEBUG "pxm bitmap: ");
249 for (i = 0; i < sizeof(pxm_bitmap); i++) {
250 printk(KERN_CONT "%02x ", pxm_bitmap[i]);
252 printk(KERN_CONT "\n");
253 printk(KERN_DEBUG "Number of logical nodes in system = %d\n",
254 num_online_nodes());
255 printk(KERN_DEBUG "Number of memory chunks in system = %d\n",
256 num_memory_chunks);
258 for (i = 0; i < MAX_APICID; i++)
259 apicid_2_node[i] = pxm_to_node(apicid_to_pxm[i]);
261 for (j = 0; j < num_memory_chunks; j++){
262 struct node_memory_chunk_s * chunk = &node_memory_chunk[j];
263 printk(KERN_DEBUG
264 "chunk %d nid %d start_pfn %08lx end_pfn %08lx\n",
265 j, chunk->nid, chunk->start_pfn, chunk->end_pfn);
266 if (node_read_chunk(chunk->nid, chunk))
267 continue;
269 memblock_x86_register_active_regions(chunk->nid, chunk->start_pfn,
270 min(chunk->end_pfn, max_pfn));
272 /* for out of order entries in SRAT */
273 sort_node_map();
275 for_each_online_node(nid) {
276 unsigned long start = node_start_pfn[nid];
277 unsigned long end = min(node_end_pfn[nid], max_pfn);
279 memory_present(nid, start, end);
280 node_remap_size[nid] = node_memmap_size_bytes(nid, start, end);
282 return 1;
283 out_fail:
284 printk(KERN_DEBUG "failed to get NUMA memory information from SRAT"
285 " table\n");
286 return 0;