ia64/xen-unstable

annotate xen/arch/x86/setup.c @ 11240:80c5350a68f1

[XEN] Clean up some x86 bootstrap code. Replace some CPU iterators
with for_each_cpu() -- we want to ensure that per_cpu areas are
accessed only for cpus in cpu_possible_map.
Signed-off-by: Keir Fraser <keir@xensource.com>
author kfraser@localhost.localdomain
date Tue Aug 22 11:19:48 2006 +0100 (2006-08-22)
parents 0f917d63e960
children 56b05c672033
rev   line source
kaf24@1452 1
kaf24@1452 2 #include <xen/config.h>
kaf24@1452 3 #include <xen/init.h>
kaf24@1452 4 #include <xen/lib.h>
kaf24@1452 5 #include <xen/sched.h>
cl349@5247 6 #include <xen/domain.h>
kaf24@1452 7 #include <xen/serial.h>
kaf24@1506 8 #include <xen/softirq.h>
kaf24@1452 9 #include <xen/acpi.h>
kaf24@3338 10 #include <xen/console.h>
iap10@4287 11 #include <xen/serial.h>
kaf24@3338 12 #include <xen/trace.h>
kaf24@3338 13 #include <xen/multiboot.h>
kaf24@5356 14 #include <xen/domain_page.h>
kfraser@10890 15 #include <xen/version.h>
kaf24@9117 16 #include <xen/gdbstub.h>
kaf24@9818 17 #include <xen/percpu.h>
iap10@6721 18 #include <public/version.h>
kaf24@1452 19 #include <asm/bitops.h>
kaf24@1452 20 #include <asm/smp.h>
kaf24@1452 21 #include <asm/processor.h>
kaf24@1452 22 #include <asm/mpspec.h>
kaf24@1452 23 #include <asm/apic.h>
kaf24@1452 24 #include <asm/desc.h>
kaf24@3338 25 #include <asm/shadow.h>
kaf24@3344 26 #include <asm/e820.h>
kaf24@5536 27 #include <acm/acm_hooks.h>
kaf24@3338 28
kaf24@5157 29 extern void dmi_scan_machine(void);
kaf24@5211 30 extern void generic_apic_probe(void);
kaf24@5157 31
kaf24@3338 32 /*
kaf24@3338 33 * opt_xenheap_megabytes: Size of Xen heap in megabytes, excluding the
kaf24@8726 34 * page_info table and allocation bitmap.
kaf24@3338 35 */
kaf24@3338 36 static unsigned int opt_xenheap_megabytes = XENHEAP_DEFAULT_MB;
kaf24@4950 37 #if defined(CONFIG_X86_64)
kaf24@3338 38 integer_param("xenheap_megabytes", opt_xenheap_megabytes);
kaf24@3354 39 #endif
kaf24@1452 40
kaf24@5146 41 /* opt_nosmp: If true, secondary processors are ignored. */
kaf24@5900 42 static int opt_nosmp = 0;
kaf24@5146 43 boolean_param("nosmp", opt_nosmp);
kaf24@5146 44
kaf24@5146 45 /* maxcpus: maximum number of CPUs to activate. */
kaf24@5146 46 static unsigned int max_cpus = NR_CPUS;
shand@11156 47 integer_param("maxcpus", max_cpus);
kaf24@5146 48
kaf24@3334 49 /* opt_watchdog: If true, run a watchdog NMI on each processor. */
kaf24@3334 50 static int opt_watchdog = 0;
kaf24@3334 51 boolean_param("watchdog", opt_watchdog);
kaf24@3334 52
kaf24@4850 53 /* **** Linux config option: propagated to domain0. */
kaf24@4850 54 /* "acpi=off": Sisables both ACPI table parsing and interpreter. */
kaf24@4850 55 /* "acpi=force": Override the disable blacklist. */
kaf24@4850 56 /* "acpi=strict": Disables out-of-spec workarounds. */
kaf24@4850 57 /* "acpi=ht": Limit ACPI just to boot-time to enable HT. */
kaf24@4850 58 /* "acpi=noirq": Disables ACPI interrupt routing. */
kaf24@4850 59 static void parse_acpi_param(char *s);
kaf24@4850 60 custom_param("acpi", parse_acpi_param);
kaf24@4850 61
kaf24@4850 62 /* **** Linux config option: propagated to domain0. */
kaf24@4850 63 /* acpi_skip_timer_override: Skip IRQ0 overrides. */
kaf24@4850 64 extern int acpi_skip_timer_override;
kaf24@4850 65 boolean_param("acpi_skip_timer_override", acpi_skip_timer_override);
kaf24@4850 66
kaf24@4850 67 /* **** Linux config option: propagated to domain0. */
kaf24@4850 68 /* noapic: Disable IOAPIC setup. */
kaf24@4850 69 extern int skip_ioapic_setup;
kaf24@4850 70 boolean_param("noapic", skip_ioapic_setup);
kaf24@4850 71
kaf24@3594 72 int early_boot = 1;
kaf24@3594 73
kaf24@5146 74 cpumask_t cpu_present_map;
kaf24@5146 75
kaf24@5003 76 /* Limits of Xen heap, used to initialise the allocator. */
kaf24@5003 77 unsigned long xenheap_phys_start, xenheap_phys_end;
kaf24@3338 78
kaf24@2298 79 extern void arch_init_memory(void);
kaf24@1589 80 extern void init_IRQ(void);
kaf24@1589 81 extern void trap_init(void);
kaf24@5604 82 extern void early_time_init(void);
kaf24@5167 83 extern void initialize_keytable(void);
kaf24@5167 84 extern void early_cpu_init(void);
kaf24@1589 85
kaf24@8533 86 struct tss_struct init_tss[NR_CPUS];
kaf24@8533 87
kaf24@5011 88 extern unsigned long cpu0_stack[];
kaf24@5011 89
kaf24@5214 90 struct cpuinfo_x86 boot_cpu_data = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
kaf24@1452 91
kaf24@5237 92 #if CONFIG_PAGING_LEVELS > 2
kaf24@1670 93 unsigned long mmu_cr4_features = X86_CR4_PSE | X86_CR4_PGE | X86_CR4_PAE;
kaf24@1670 94 #else
kaf24@5593 95 unsigned long mmu_cr4_features = X86_CR4_PSE;
kaf24@1670 96 #endif
kaf24@1452 97 EXPORT_SYMBOL(mmu_cr4_features);
kaf24@1452 98
kaf24@4818 99 int acpi_disabled;
kaf24@1452 100
kaf24@4850 101 int acpi_force;
kaf24@4850 102 char acpi_param[10] = "";
kaf24@4850 103 static void parse_acpi_param(char *s)
kaf24@4850 104 {
kaf24@4850 105 /* Save the parameter so it can be propagated to domain0. */
kaf24@4850 106 strncpy(acpi_param, s, sizeof(acpi_param));
kaf24@4850 107 acpi_param[sizeof(acpi_param)-1] = '\0';
kaf24@4850 108
kaf24@4850 109 /* Interpret the parameter for use within Xen. */
kaf24@4850 110 if ( !strcmp(s, "off") )
kaf24@4850 111 {
kaf24@4850 112 disable_acpi();
kaf24@4850 113 }
kaf24@4850 114 else if ( !strcmp(s, "force") )
kaf24@4850 115 {
kaf24@4850 116 acpi_force = 1;
kaf24@4850 117 acpi_ht = 1;
kaf24@4850 118 acpi_disabled = 0;
kaf24@4850 119 }
kaf24@4850 120 else if ( !strcmp(s, "strict") )
kaf24@4850 121 {
kaf24@4850 122 acpi_strict = 1;
kaf24@4850 123 }
kaf24@4850 124 else if ( !strcmp(s, "ht") )
kaf24@4850 125 {
kaf24@4850 126 if ( !acpi_force )
kaf24@4850 127 disable_acpi();
kaf24@4850 128 acpi_ht = 1;
kaf24@4850 129 }
kaf24@4850 130 else if ( !strcmp(s, "noirq") )
kaf24@4850 131 {
kaf24@4850 132 acpi_noirq_set();
kaf24@4850 133 }
kaf24@4850 134 }
kaf24@4850 135
kaf24@1452 136 static void __init do_initcalls(void)
kaf24@1452 137 {
kaf24@1452 138 initcall_t *call;
kaf24@1452 139 for ( call = &__initcall_start; call < &__initcall_end; call++ )
kaf24@1452 140 (*call)();
kaf24@1452 141 }
kaf24@1452 142
kaf24@8459 143 #define EARLY_FAIL() for ( ; ; ) __asm__ __volatile__ ( "hlt" )
kaf24@8459 144
kaf24@8459 145 static struct e820entry e820_raw[E820MAX];
kaf24@8459 146
kaf24@9067 147 static unsigned long initial_images_start, initial_images_end;
kaf24@9067 148
kaf24@9067 149 unsigned long initial_images_nrpages(void)
kaf24@9067 150 {
kaf24@9067 151 unsigned long s = initial_images_start + PAGE_SIZE - 1;
kaf24@9067 152 unsigned long e = initial_images_end;
kaf24@9067 153 return ((e >> PAGE_SHIFT) - (s >> PAGE_SHIFT));
kaf24@9067 154 }
kaf24@9067 155
kaf24@9067 156 void discard_initial_images(void)
kaf24@9067 157 {
kaf24@9067 158 init_domheap_pages(initial_images_start, initial_images_end);
kaf24@9067 159 }
kaf24@9067 160
kaf24@9818 161 extern char __per_cpu_start[], __per_cpu_data_end[], __per_cpu_end[];
kaf24@9818 162
kaf24@9818 163 static void percpu_init_areas(void)
kaf24@9818 164 {
kaf24@9818 165 unsigned int i, data_size = __per_cpu_data_end - __per_cpu_start;
kaf24@9818 166
kaf24@9818 167 BUG_ON(data_size > PERCPU_SIZE);
kaf24@9818 168
kaf24@9818 169 for ( i = 1; i < NR_CPUS; i++ )
kaf24@9818 170 memcpy(__per_cpu_start + (i << PERCPU_SHIFT),
kaf24@9818 171 __per_cpu_start,
kaf24@9818 172 data_size);
kaf24@9818 173 }
kaf24@9818 174
kaf24@9818 175 static void percpu_free_unused_areas(void)
kaf24@9818 176 {
kaf24@9818 177 unsigned int i, first_unused;
kaf24@9818 178
kaf24@9818 179 /* Find first unused CPU number. */
kaf24@9818 180 for ( i = 0; i < NR_CPUS; i++ )
kaf24@9818 181 if ( !cpu_online(i) )
kaf24@9818 182 break;
kaf24@9818 183 first_unused = i;
kaf24@9818 184
kaf24@9818 185 /* Check that there are no holes in cpu_online_map. */
kaf24@9818 186 for ( ; i < NR_CPUS; i++ )
kaf24@9818 187 BUG_ON(cpu_online(i));
kaf24@9818 188
kaf24@9818 189 init_xenheap_pages(__pa(__per_cpu_start) + (first_unused << PERCPU_SHIFT),
kaf24@9818 190 __pa(__per_cpu_end));
kaf24@9818 191 }
kaf24@9818 192
kfraser@11240 193 static void init_idle_domain(void)
kfraser@11240 194 {
kfraser@11240 195 struct domain *idle_domain;
kfraser@11240 196
kfraser@11240 197 /* Domain creation requires that scheduler structures are initialised. */
kfraser@11240 198 scheduler_init();
kfraser@11240 199
kfraser@11240 200 idle_domain = domain_create(IDLE_DOMAIN_ID);
kfraser@11240 201 if ( (idle_domain == NULL) || (alloc_vcpu(idle_domain, 0, 0) == NULL) )
kfraser@11240 202 BUG();
kfraser@11240 203
kfraser@11240 204 set_current(idle_domain->vcpu[0]);
kfraser@11240 205 idle_vcpu[0] = this_cpu(curr_vcpu) = current;
kfraser@11240 206
kfraser@11240 207 setup_idle_pagetable();
kfraser@11240 208 }
kfraser@11240 209
kaf24@8463 210 void __init __start_xen(multiboot_info_t *mbi)
kaf24@1452 211 {
kaf24@9823 212 char __cmdline[] = "", *cmdline = __cmdline;
kaf24@8457 213 unsigned long _initrd_start = 0, _initrd_len = 0;
kaf24@8457 214 unsigned int initrdidx = 1;
kaf24@8457 215 module_t *mod = (module_t *)__va(mbi->mods_addr);
kaf24@6111 216 unsigned long nr_pages, modules_length;
kaf24@8726 217 paddr_t s, e;
kaf24@8402 218 int i, e820_warn = 0, e820_raw_nr = 0, bytes = 0;
kaf24@5776 219 struct ns16550_defaults ns16550 = {
kaf24@5776 220 .data_bits = 8,
kaf24@5776 221 .parity = 'n',
kaf24@5776 222 .stop_bits = 1
kaf24@5776 223 };
kaf24@3338 224
kaf24@3338 225 /* Parse the command-line options. */
kaf24@3344 226 if ( (mbi->flags & MBI_CMDLINE) && (mbi->cmdline != 0) )
kaf24@9823 227 cmdline = __va(mbi->cmdline);
kaf24@9823 228 cmdline_parse(cmdline);
kaf24@3338 229
kaf24@8534 230 set_current((struct vcpu *)0xfffff000); /* debug sanity */
kfraser@11240 231 idle_vcpu[0] = current;
kaf24@8534 232 set_processor_id(0); /* needed early, for smp_processor_id() */
kaf24@3338 233
kaf24@5146 234 smp_prepare_boot_cpu();
kaf24@5146 235
kaf24@3338 236 /* We initialise the serial devices very early so we can get debugging. */
kaf24@5776 237 ns16550.io_base = 0x3f8;
kaf24@5776 238 ns16550.irq = 4;
kaf24@5776 239 ns16550_init(0, &ns16550);
kaf24@5776 240 ns16550.io_base = 0x2f8;
kaf24@5776 241 ns16550.irq = 3;
kaf24@5776 242 ns16550_init(1, &ns16550);
kaf24@5195 243 serial_init_preirq();
kaf24@3338 244
kaf24@3338 245 init_console();
kaf24@3338 246
kaf24@9823 247 printf("Command line: %s\n", cmdline);
kaf24@9823 248
kaf24@3344 249 /* Check that we have at least one Multiboot module. */
kaf24@3344 250 if ( !(mbi->flags & MBI_MODULES) || (mbi->mods_count == 0) )
kaf24@3338 251 {
kaf24@5542 252 printk("FATAL ERROR: dom0 kernel not specified."
kaf24@5542 253 " Check bootloader configuration.\n");
kaf24@5011 254 EARLY_FAIL();
kaf24@5011 255 }
kaf24@5011 256
kaf24@5011 257 if ( ((unsigned long)cpu0_stack & (STACK_SIZE-1)) != 0 )
kaf24@5011 258 {
kaf24@5011 259 printk("FATAL ERROR: Misaligned CPU0 stack.\n");
kaf24@5011 260 EARLY_FAIL();
kaf24@3338 261 }
kaf24@3338 262
kaf24@9818 263 percpu_init_areas();
kaf24@9818 264
kaf24@3338 265 xenheap_phys_end = opt_xenheap_megabytes << 20;
kaf24@3338 266
kaf24@3344 267 if ( mbi->flags & MBI_MEMMAP )
kaf24@3344 268 {
kaf24@3344 269 while ( bytes < mbi->mmap_length )
kaf24@3344 270 {
kaf24@3344 271 memory_map_t *map = __va(mbi->mmap_addr + bytes);
kaf24@8402 272
kaf24@8402 273 /*
kaf24@8403 274 * This is a gross workaround for a BIOS bug. Some bootloaders do
kaf24@8402 275 * not write e820 map entries into pre-zeroed memory. This is
kaf24@8402 276 * okay if the BIOS fills in all fields of the map entry, but
kaf24@8402 277 * some broken BIOSes do not bother to write the high word of
kaf24@8402 278 * the length field if the length is smaller than 4GB. We
kaf24@8402 279 * detect and fix this by flagging sections below 4GB that
kaf24@8403 280 * appear to be larger than 4GB in size.
kaf24@8402 281 */
kaf24@8403 282 if ( (map->base_addr_high == 0) && (map->length_high != 0) )
kaf24@8402 283 {
kaf24@8402 284 e820_warn = 1;
kaf24@8402 285 map->length_high = 0;
kaf24@8402 286 }
kaf24@8402 287
kaf24@3344 288 e820_raw[e820_raw_nr].addr =
kaf24@3344 289 ((u64)map->base_addr_high << 32) | (u64)map->base_addr_low;
kaf24@3344 290 e820_raw[e820_raw_nr].size =
kaf24@3344 291 ((u64)map->length_high << 32) | (u64)map->length_low;
kaf24@3344 292 e820_raw[e820_raw_nr].type =
kaf24@3346 293 (map->type > E820_SHARED_PAGE) ? E820_RESERVED : map->type;
kaf24@3344 294 e820_raw_nr++;
kaf24@8402 295
kaf24@3344 296 bytes += map->size + 4;
kaf24@3344 297 }
kaf24@3344 298 }
kaf24@3344 299 else if ( mbi->flags & MBI_MEMLIMITS )
kaf24@3344 300 {
kaf24@3344 301 e820_raw[0].addr = 0;
kaf24@3344 302 e820_raw[0].size = mbi->mem_lower << 10;
kaf24@3344 303 e820_raw[0].type = E820_RAM;
kaf24@3354 304 e820_raw[1].addr = 0x100000;
kaf24@3354 305 e820_raw[1].size = mbi->mem_upper << 10;
kaf24@3354 306 e820_raw[1].type = E820_RAM;
kaf24@3344 307 e820_raw_nr = 2;
kaf24@3344 308 }
kaf24@3344 309 else
kaf24@3344 310 {
kaf24@3344 311 printk("FATAL ERROR: Bootloader provided no memory information.\n");
kaf24@3344 312 for ( ; ; ) ;
kaf24@3344 313 }
kaf24@3344 314
kaf24@8402 315 if ( e820_warn )
kaf24@8402 316 printk("WARNING: Buggy e820 map detected and fixed "
kaf24@8402 317 "(truncated length fields).\n");
kaf24@8402 318
kaf24@4950 319 max_page = init_e820(e820_raw, &e820_raw_nr);
kaf24@3338 320
kaf24@6111 321 modules_length = mod[mbi->mods_count-1].mod_end - mod[0].mod_start;
kaf24@3338 322
kaf24@6111 323 /* Find a large enough RAM extent to stash the DOM0 modules. */
kaf24@6111 324 for ( i = 0; ; i++ )
kaf24@3338 325 {
kaf24@6111 326 if ( i == e820.nr_map )
kaf24@6111 327 {
kaf24@6111 328 printk("Not enough memory to stash the DOM0 kernel image.\n");
kaf24@6111 329 for ( ; ; ) ;
kaf24@6111 330 }
kaf24@6134 331
kaf24@6134 332 if ( (e820.map[i].type == E820_RAM) &&
kaf24@6134 333 (e820.map[i].size >= modules_length) &&
kaf24@6134 334 ((e820.map[i].addr + e820.map[i].size) >=
kaf24@6134 335 (xenheap_phys_end + modules_length)) )
kaf24@6134 336 break;
kaf24@3338 337 }
kaf24@6111 338
kaf24@6134 339 /* Stash as near as possible to the beginning of the RAM extent. */
kaf24@6134 340 initial_images_start = e820.map[i].addr;
kaf24@6134 341 if ( initial_images_start < xenheap_phys_end )
kaf24@6134 342 initial_images_start = xenheap_phys_end;
kaf24@6134 343 initial_images_end = initial_images_start + modules_length;
kaf24@6134 344
kaf24@4950 345 #if defined(CONFIG_X86_32)
kaf24@3338 346 memmove((void *)initial_images_start, /* use low mapping */
kaf24@3338 347 (void *)mod[0].mod_start, /* use low mapping */
kaf24@3338 348 mod[mbi->mods_count-1].mod_end - mod[0].mod_start);
kaf24@4950 349 #elif defined(CONFIG_X86_64)
kaf24@3338 350 memmove(__va(initial_images_start),
kaf24@3338 351 __va(mod[0].mod_start),
kaf24@3338 352 mod[mbi->mods_count-1].mod_end - mod[0].mod_start);
kaf24@3338 353 #endif
kaf24@3338 354
kaf24@3354 355 /* Initialise boot-time allocator with all RAM situated after modules. */
kaf24@5003 356 xenheap_phys_start = init_boot_allocator(__pa(&_end));
kaf24@6111 357 nr_pages = 0;
kaf24@3354 358 for ( i = 0; i < e820.nr_map; i++ )
kaf24@3354 359 {
kaf24@3354 360 if ( e820.map[i].type != E820_RAM )
kaf24@3354 361 continue;
kaf24@6111 362
kaf24@3354 363 nr_pages += e820.map[i].size >> PAGE_SHIFT;
kaf24@6111 364
kaf24@6111 365 /* Initialise boot heap, skipping Xen heap and dom0 modules. */
kaf24@6111 366 s = e820.map[i].addr;
kaf24@6111 367 e = s + e820.map[i].size;
kaf24@6111 368 if ( s < xenheap_phys_end )
kaf24@6111 369 s = xenheap_phys_end;
kaf24@6111 370 if ( (s < initial_images_end) && (e > initial_images_start) )
kaf24@6111 371 s = initial_images_end;
kaf24@6111 372 init_boot_pages(s, e);
kaf24@6111 373
kaf24@5003 374 #if defined (CONFIG_X86_64)
kaf24@5003 375 /*
kaf24@5003 376 * x86/64 maps all registered RAM. Points to note:
kaf24@9783 377 * 1. The initial pagetable already maps low 1GB, so skip that.
kaf24@5003 378 * 2. We must map *only* RAM areas, taking care to avoid I/O holes.
kaf24@5003 379 * Failure to do this can cause coherency problems and deadlocks
kaf24@5003 380 * due to cache-attribute mismatches (e.g., AMD/AGP Linux bug).
kaf24@5003 381 */
kaf24@5003 382 {
kaf24@5004 383 /* Calculate page-frame range, discarding partial frames. */
kaf24@5004 384 unsigned long start, end;
kaf24@9783 385 unsigned long init_mapped = 1UL << (30 - PAGE_SHIFT); /* 1GB */
kaf24@5005 386 start = PFN_UP(e820.map[i].addr);
kaf24@5005 387 end = PFN_DOWN(e820.map[i].addr + e820.map[i].size);
kaf24@9800 388 /* Clip the range to exclude what the bootstrapper initialised. */
kaf24@9783 389 if ( end < init_mapped )
kaf24@5004 390 continue;
kaf24@9783 391 if ( start < init_mapped )
kaf24@9783 392 start = init_mapped;
kaf24@5004 393 /* Request the mapping. */
kaf24@5003 394 map_pages_to_xen(
kaf24@5004 395 PAGE_OFFSET + (start << PAGE_SHIFT),
kaf24@5004 396 start, end-start, PAGE_HYPERVISOR);
kaf24@5003 397 }
kaf24@5003 398 #endif
kaf24@3354 399 }
kaf24@3354 400
kaf24@5003 401 memguard_init();
kaf24@4950 402
kaf24@3354 403 printk("System RAM: %luMB (%lukB)\n",
kaf24@3354 404 nr_pages >> (20 - PAGE_SHIFT),
kaf24@3354 405 nr_pages << (PAGE_SHIFT - 10));
kaf24@7220 406 total_pages = nr_pages;
kaf24@3354 407
kaf24@8679 408 /* Sanity check for unwanted bloat of dom0_op structure. */
kaf24@9878 409 BUILD_BUG_ON(sizeof(((struct dom0_op *)0)->u) !=
kaf24@9878 410 sizeof(((struct dom0_op *)0)->u.pad));
kaf24@7388 411
kaf24@9878 412 BUILD_BUG_ON(sizeof(start_info_t) > PAGE_SIZE);
kaf24@9878 413 BUILD_BUG_ON(sizeof(shared_info_t) > PAGE_SIZE);
kaf24@9878 414 BUILD_BUG_ON(sizeof(vcpu_info_t) != 64);
kaf24@7744 415
kfraser@10492 416 /* Check definitions in public headers match internal defs. */
kaf24@9878 417 BUILD_BUG_ON(__HYPERVISOR_VIRT_START != HYPERVISOR_VIRT_START);
kaf24@8521 418 #ifdef HYPERVISOR_VIRT_END
kaf24@9878 419 BUILD_BUG_ON(__HYPERVISOR_VIRT_END != HYPERVISOR_VIRT_END);
kaf24@8521 420 #endif
kfraser@10492 421 BUILD_BUG_ON(MACH2PHYS_VIRT_START != RO_MPT_VIRT_START);
kfraser@10492 422 BUILD_BUG_ON(MACH2PHYS_VIRT_END != RO_MPT_VIRT_END);
kaf24@8521 423
kaf24@3354 424 init_frametable();
kaf24@3338 425
kaf24@3354 426 end_boot_allocator();
kaf24@3354 427
kaf24@6111 428 /* Initialise the Xen heap, skipping RAM holes. */
kaf24@6111 429 nr_pages = 0;
kaf24@6111 430 for ( i = 0; i < e820.nr_map; i++ )
kaf24@6111 431 {
kaf24@6111 432 if ( e820.map[i].type != E820_RAM )
kaf24@6111 433 continue;
kaf24@6111 434
kaf24@6111 435 s = e820.map[i].addr;
kaf24@6111 436 e = s + e820.map[i].size;
kaf24@6111 437 if ( s < xenheap_phys_start )
kaf24@6111 438 s = xenheap_phys_start;
kaf24@6111 439 if ( e > xenheap_phys_end )
kaf24@6111 440 e = xenheap_phys_end;
kaf24@6111 441
kaf24@6111 442 if ( s < e )
kaf24@6111 443 {
kaf24@6111 444 nr_pages += (e - s) >> PAGE_SHIFT;
kaf24@6111 445 init_xenheap_pages(s, e);
kaf24@6111 446 }
kaf24@6111 447 }
kaf24@6111 448
kaf24@6111 449 printk("Xen heap: %luMB (%lukB)\n",
kaf24@6111 450 nr_pages >> (20 - PAGE_SHIFT),
kaf24@6111 451 nr_pages << (PAGE_SHIFT - 10));
kaf24@3338 452
kaf24@3594 453 early_boot = 0;
kaf24@3338 454
kaf24@8459 455 early_cpu_init();
kaf24@8459 456
kaf24@8459 457 paging_init();
kaf24@8459 458
kaf24@8459 459 /* Unmap the first page of CPU0's stack. */
kaf24@8459 460 memguard_guard_stack(cpu0_stack);
kaf24@8459 461
kaf24@8459 462 open_softirq(NEW_TLBFLUSH_CLOCK_PERIOD_SOFTIRQ, new_tlbflush_clock_period);
kaf24@8459 463
kaf24@8459 464 if ( opt_watchdog )
kaf24@8459 465 nmi_watchdog = NMI_LOCAL_APIC;
kaf24@8459 466
kaf24@8459 467 sort_exception_tables();
kaf24@8459 468
kaf24@8459 469 find_smp_config();
kaf24@8459 470
kaf24@8459 471 smp_alloc_memory();
kaf24@8459 472
kaf24@8459 473 dmi_scan_machine();
kaf24@8459 474
kaf24@8459 475 generic_apic_probe();
kaf24@8459 476
kaf24@8459 477 acpi_boot_table_init();
kaf24@8459 478 acpi_boot_init();
kaf24@8459 479
kaf24@8459 480 if ( smp_found_config )
kaf24@8459 481 get_smp_config();
kaf24@8459 482
kaf24@8459 483 init_apic_mappings();
kaf24@8459 484
kaf24@8459 485 init_IRQ();
kaf24@8459 486
kfraser@11240 487 init_idle_domain();
kfraser@11240 488
kaf24@8459 489 trap_init();
kaf24@8459 490
kaf24@8586 491 timer_init();
kaf24@8459 492
kaf24@8459 493 early_time_init();
kaf24@8459 494
kaf24@8459 495 arch_init_memory();
kaf24@8459 496
kaf24@8459 497 identify_cpu(&boot_cpu_data);
kaf24@8459 498 if ( cpu_has_fxsr )
kaf24@8459 499 set_in_cr4(X86_CR4_OSFXSR);
kaf24@8459 500 if ( cpu_has_xmm )
kaf24@8459 501 set_in_cr4(X86_CR4_OSXMMEXCPT);
kaf24@8459 502
kaf24@8459 503 if ( opt_nosmp )
kaf24@8459 504 max_cpus = 0;
kaf24@8459 505
kaf24@8459 506 smp_prepare_cpus(max_cpus);
kaf24@8459 507
kaf24@8459 508 /*
kaf24@8459 509 * Initialise higher-level timer functions. We do this fairly late
kaf24@8459 510 * (post-SMP) because the time bases and scale factors need to be updated
kaf24@8459 511 * regularly, and SMP initialisation can cause a long delay with
kaf24@8459 512 * interrupts not yet enabled.
kaf24@8459 513 */
kaf24@8459 514 init_xen_time();
kaf24@8459 515
kaf24@8459 516 initialize_keytable();
kaf24@8459 517
kaf24@8459 518 serial_init_postirq();
kaf24@8459 519
kaf24@8459 520 BUG_ON(!local_irq_is_enabled());
kaf24@8459 521
kaf24@8459 522 for_each_present_cpu ( i )
kaf24@8459 523 {
kaf24@8459 524 if ( num_online_cpus() >= max_cpus )
kaf24@8459 525 break;
kaf24@8459 526 if ( !cpu_online(i) )
kaf24@8459 527 __cpu_up(i);
kaf24@8459 528 }
kaf24@8459 529
kaf24@8459 530 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
kaf24@8459 531 smp_cpus_done(max_cpus);
kaf24@8459 532
kaf24@9818 533 percpu_free_unused_areas();
kaf24@9818 534
kaf24@9117 535 initialise_gdb(); /* could be moved earlier */
kaf24@9117 536
kaf24@8459 537 do_initcalls();
kaf24@8459 538
kaf24@8459 539 schedulers_start();
kaf24@8459 540
kaf24@8594 541 if ( opt_watchdog )
kaf24@8594 542 watchdog_enable();
kaf24@8459 543
kaf24@8459 544 /* initialize access control security module */
kaf24@8459 545 acm_init(&initrdidx, mbi, initial_images_start);
kaf24@8459 546
kaf24@8459 547 /* Create initial domain 0. */
kfraser@10655 548 dom0 = domain_create(0);
kfraser@10655 549 if ( (dom0 == NULL) || (alloc_vcpu(dom0, 0, 0) == NULL) )
kaf24@8459 550 panic("Error creating domain 0\n");
kaf24@8459 551
kaf24@8459 552 set_bit(_DOMF_privileged, &dom0->domain_flags);
kaf24@8459 553 /* post-create hooks sets security label */
kaf24@8459 554 acm_post_domain0_create(dom0->domain_id);
kaf24@8459 555
kaf24@8459 556 /* Grab the DOM0 command line. */
kaf24@8459 557 cmdline = (char *)(mod[0].string ? __va(mod[0].string) : NULL);
kaf24@8459 558 if ( cmdline != NULL )
kaf24@8459 559 {
kaf24@8459 560 static char dom0_cmdline[MAX_GUEST_CMDLINE];
kaf24@8459 561
kaf24@8459 562 /* Skip past the image name and copy to a local buffer. */
kaf24@8459 563 while ( *cmdline == ' ' ) cmdline++;
kaf24@8459 564 if ( (cmdline = strchr(cmdline, ' ')) != NULL )
kaf24@8459 565 {
kaf24@8459 566 while ( *cmdline == ' ' ) cmdline++;
kaf24@8459 567 strcpy(dom0_cmdline, cmdline);
kaf24@8459 568 }
kaf24@8459 569
kaf24@8459 570 cmdline = dom0_cmdline;
kaf24@8459 571
kaf24@8459 572 /* Append any extra parameters. */
kaf24@8459 573 if ( skip_ioapic_setup && !strstr(cmdline, "noapic") )
kaf24@8459 574 strcat(cmdline, " noapic");
kaf24@8459 575 if ( acpi_skip_timer_override &&
kaf24@8459 576 !strstr(cmdline, "acpi_skip_timer_override") )
kaf24@8459 577 strcat(cmdline, " acpi_skip_timer_override");
kaf24@8459 578 if ( (strlen(acpi_param) != 0) && !strstr(cmdline, "acpi=") )
kaf24@8459 579 {
kaf24@8459 580 strcat(cmdline, " acpi=");
kaf24@8459 581 strcat(cmdline, acpi_param);
kaf24@8459 582 }
kaf24@8459 583 }
kaf24@8459 584
kaf24@8459 585 if ( (initrdidx > 0) && (initrdidx < mbi->mods_count) )
kaf24@8459 586 {
kaf24@8459 587 _initrd_start = initial_images_start +
kaf24@8459 588 (mod[initrdidx].mod_start - mod[0].mod_start);
kaf24@8459 589 _initrd_len = mod[initrdidx].mod_end - mod[initrdidx].mod_start;
kaf24@8459 590 }
kaf24@8459 591
kaf24@8459 592 /*
kaf24@8459 593 * We're going to setup domain0 using the module(s) that we stashed safely
kaf24@8459 594 * above our heap. The second module, if present, is an initrd ramdisk.
kaf24@8459 595 */
kaf24@8459 596 if ( construct_dom0(dom0,
kaf24@8459 597 initial_images_start,
kaf24@8459 598 mod[0].mod_end-mod[0].mod_start,
kaf24@8459 599 _initrd_start,
kaf24@8459 600 _initrd_len,
kaf24@8459 601 cmdline) != 0)
kaf24@8459 602 panic("Could not set up DOM0 guest OS\n");
kaf24@8459 603
kaf24@8459 604 /* Scrub RAM that is still free and so may go to an unprivileged domain. */
kaf24@8459 605 scrub_heap_pages();
kaf24@8459 606
kaf24@8459 607 init_trace_bufs();
kaf24@8459 608
kaf24@10502 609 console_endboot();
kaf24@8459 610
kaf24@8459 611 /* Hide UART from DOM0 if we're using it */
kaf24@8459 612 serial_endboot();
kaf24@8459 613
kaf24@8459 614 domain_unpause_by_systemcontroller(dom0);
kaf24@8459 615
kaf24@8459 616 startup_cpu_idle_loop();
kaf24@8459 617 }
kaf24@8459 618
kaf24@6725 619 void arch_get_xen_caps(xen_capabilities_info_t info)
iap10@6721 620 {
kaf24@6725 621 char *p = info;
kfraser@10890 622 int major = xen_major_version();
kfraser@10890 623 int minor = xen_minor_version();
iap10@6721 624
kaf24@6725 625 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
kaf24@6725 626
kfraser@10890 627 p += sprintf(p, "xen-%d.%d-x86_32 ", major, minor);
kaf24@6725 628 if ( hvm_enabled )
kfraser@10890 629 p += sprintf(p, "hvm-%d.%d-x86_32 ", major, minor);
kaf24@6725 630
kaf24@6725 631 #elif defined(CONFIG_X86_32) && defined(CONFIG_X86_PAE)
kaf24@6725 632
kfraser@10890 633 p += sprintf(p, "xen-%d.%d-x86_32p ", major, minor);
kaf24@6725 634 if ( hvm_enabled )
iap10@6721 635 {
kfraser@10890 636 p += sprintf(p, "hvm-%d.%d-x86_32 ", major, minor);
kfraser@10890 637 p += sprintf(p, "hvm-%d.%d-x86_32p ", major, minor);
iap10@6721 638 }
iap10@6721 639
kaf24@6725 640 #elif defined(CONFIG_X86_64)
iap10@6721 641
kfraser@10890 642 p += sprintf(p, "xen-%d.%d-x86_64 ", major, minor);
kaf24@6725 643 if ( hvm_enabled )
iap10@6721 644 {
kfraser@10890 645 p += sprintf(p, "hvm-%d.%d-x86_32 ", major, minor);
kfraser@10890 646 p += sprintf(p, "hvm-%d.%d-x86_32p ", major, minor);
kfraser@10890 647 p += sprintf(p, "hvm-%d.%d-x86_64 ", major, minor);
iap10@6721 648 }
kaf24@6725 649
kaf24@6725 650 #else
kaf24@6725 651
kaf24@6725 652 p++;
kaf24@6725 653
iap10@6721 654 #endif
kaf24@8246 655
kaf24@6725 656 *(p-1) = 0;
iap10@6721 657
shand@6734 658 BUG_ON((p - info) > sizeof(xen_capabilities_info_t));
iap10@6721 659 }
iap10@6721 660
kaf24@3914 661 /*
kaf24@3914 662 * Local variables:
kaf24@3914 663 * mode: C
kaf24@3914 664 * c-set-style: "BSD"
kaf24@3914 665 * c-basic-offset: 4
kaf24@3914 666 * tab-width: 4
kaf24@3914 667 * indent-tabs-mode: nil
kaf24@3988 668 * End:
kaf24@3914 669 */