ia64/xen-unstable

annotate xen/arch/x86/setup.c @ 16274:afd768e9382e

Intel(R) Trusted Execution Technology (Intel(R) TXT) support for Xen.
Signed-off-by: Joseph Cihula <joseph.cihula@intel.com>
author Keir Fraser <keir@xensource.com>
date Tue Oct 30 10:13:43 2007 +0000 (2007-10-30)
parents 4034317507de
children a7f8ff1ca311
rev   line source
kaf24@1452 1 #include <xen/config.h>
kaf24@1452 2 #include <xen/init.h>
kaf24@1452 3 #include <xen/lib.h>
kaf24@1452 4 #include <xen/sched.h>
cl349@5247 5 #include <xen/domain.h>
kaf24@1452 6 #include <xen/serial.h>
kaf24@1506 7 #include <xen/softirq.h>
kaf24@1452 8 #include <xen/acpi.h>
kaf24@3338 9 #include <xen/console.h>
iap10@4287 10 #include <xen/serial.h>
kaf24@3338 11 #include <xen/trace.h>
kaf24@3338 12 #include <xen/multiboot.h>
kaf24@5356 13 #include <xen/domain_page.h>
kfraser@10890 14 #include <xen/version.h>
kaf24@9117 15 #include <xen/gdbstub.h>
kaf24@9818 16 #include <xen/percpu.h>
kfraser@11296 17 #include <xen/hypercall.h>
kfraser@11601 18 #include <xen/keyhandler.h>
kfraser@11971 19 #include <xen/numa.h>
kaf24@13662 20 #include <xen/rcupdate.h>
keir@15298 21 #include <xen/vga.h>
keir@15988 22 #include <xen/dmi.h>
iap10@6721 23 #include <public/version.h>
ack@13291 24 #ifdef CONFIG_COMPAT
ack@13291 25 #include <compat/platform.h>
ack@13291 26 #include <compat/xen.h>
ack@13291 27 #endif
kaf24@1452 28 #include <asm/bitops.h>
kaf24@1452 29 #include <asm/smp.h>
kaf24@1452 30 #include <asm/processor.h>
kaf24@1452 31 #include <asm/mpspec.h>
kaf24@1452 32 #include <asm/apic.h>
kaf24@1452 33 #include <asm/desc.h>
Tim@13909 34 #include <asm/paging.h>
kaf24@3344 35 #include <asm/e820.h>
kfraser@15819 36 #include <xsm/acm/acm_hooks.h>
ian@12677 37 #include <xen/kexec.h>
kfraser@15336 38 #include <asm/edd.h>
kfraser@15815 39 #include <xsm/xsm.h>
keir@16274 40 #include <asm/tboot.h>
kaf24@3338 41
kfraser@15074 42 #if defined(CONFIG_X86_64)
kfraser@15597 43 #define BOOTSTRAP_DIRECTMAP_END (1UL << 32) /* 4GB */
kfraser@15074 44 #define maddr_to_bootstrap_virt(m) maddr_to_virt(m)
kfraser@15074 45 #else
kfraser@15597 46 #define BOOTSTRAP_DIRECTMAP_END (1UL << 30) /* 1GB */
kfraser@15074 47 #define maddr_to_bootstrap_virt(m) ((void *)(long)(m))
kfraser@15074 48 #endif
kfraser@15074 49
kaf24@5211 50 extern void generic_apic_probe(void);
kfraser@11971 51 extern void numa_initmem_init(unsigned long start_pfn, unsigned long end_pfn);
kaf24@5157 52
kfraser@15330 53 extern u16 boot_edid_caps;
kfraser@15330 54 extern u8 boot_edid_info[128];
kfraser@15330 55 extern struct boot_video_info boot_vid_info;
kfraser@15330 56
kaf24@3338 57 /*
kaf24@3338 58 * opt_xenheap_megabytes: Size of Xen heap in megabytes, excluding the
kaf24@8726 59 * page_info table and allocation bitmap.
kaf24@3338 60 */
kaf24@3338 61 static unsigned int opt_xenheap_megabytes = XENHEAP_DEFAULT_MB;
kaf24@4950 62 #if defined(CONFIG_X86_64)
kaf24@3338 63 integer_param("xenheap_megabytes", opt_xenheap_megabytes);
kaf24@3354 64 #endif
kaf24@1452 65
kaf24@5146 66 /* opt_nosmp: If true, secondary processors are ignored. */
kaf24@5900 67 static int opt_nosmp = 0;
kaf24@5146 68 boolean_param("nosmp", opt_nosmp);
kaf24@5146 69
kaf24@5146 70 /* maxcpus: maximum number of CPUs to activate. */
kaf24@5146 71 static unsigned int max_cpus = NR_CPUS;
shand@11156 72 integer_param("maxcpus", max_cpus);
kaf24@5146 73
kaf24@3334 74 /* opt_watchdog: If true, run a watchdog NMI on each processor. */
kaf24@3334 75 static int opt_watchdog = 0;
kaf24@3334 76 boolean_param("watchdog", opt_watchdog);
kaf24@3334 77
kaf24@4850 78 /* **** Linux config option: propagated to domain0. */
kaf24@4850 79 /* "acpi=off": Sisables both ACPI table parsing and interpreter. */
kaf24@4850 80 /* "acpi=force": Override the disable blacklist. */
kaf24@4850 81 /* "acpi=strict": Disables out-of-spec workarounds. */
kaf24@4850 82 /* "acpi=ht": Limit ACPI just to boot-time to enable HT. */
kaf24@4850 83 /* "acpi=noirq": Disables ACPI interrupt routing. */
kaf24@4850 84 static void parse_acpi_param(char *s);
kaf24@4850 85 custom_param("acpi", parse_acpi_param);
kaf24@4850 86
kaf24@4850 87 /* **** Linux config option: propagated to domain0. */
kaf24@4850 88 /* acpi_skip_timer_override: Skip IRQ0 overrides. */
kaf24@4850 89 extern int acpi_skip_timer_override;
kaf24@4850 90 boolean_param("acpi_skip_timer_override", acpi_skip_timer_override);
kaf24@4850 91
kaf24@4850 92 /* **** Linux config option: propagated to domain0. */
kaf24@4850 93 /* noapic: Disable IOAPIC setup. */
kaf24@4850 94 extern int skip_ioapic_setup;
kaf24@4850 95 boolean_param("noapic", skip_ioapic_setup);
kaf24@4850 96
kaf24@3594 97 int early_boot = 1;
kaf24@3594 98
kaf24@5146 99 cpumask_t cpu_present_map;
kaf24@5146 100
kfraser@15074 101 unsigned long xen_phys_start;
kfraser@15074 102
kaf24@5003 103 /* Limits of Xen heap, used to initialise the allocator. */
kaf24@5003 104 unsigned long xenheap_phys_start, xenheap_phys_end;
kaf24@3338 105
kaf24@2298 106 extern void arch_init_memory(void);
kaf24@1589 107 extern void init_IRQ(void);
kaf24@5604 108 extern void early_time_init(void);
kaf24@5167 109 extern void early_cpu_init(void);
kfraser@15747 110 extern void vesa_init(void);
kfraser@15747 111 extern void vesa_mtrr_init(void);
kaf24@1589 112
kaf24@8533 113 struct tss_struct init_tss[NR_CPUS];
kaf24@8533 114
kfraser@15490 115 char __attribute__ ((__section__(".bss.stack_aligned"))) cpu0_stack[STACK_SIZE];
kaf24@5011 116
keir@16144 117 struct cpuinfo_x86 boot_cpu_data = { 0, 0, 0, 0, -1 };
kaf24@1452 118
kaf24@5237 119 #if CONFIG_PAGING_LEVELS > 2
kaf24@1670 120 unsigned long mmu_cr4_features = X86_CR4_PSE | X86_CR4_PGE | X86_CR4_PAE;
kaf24@1670 121 #else
kaf24@5593 122 unsigned long mmu_cr4_features = X86_CR4_PSE;
kaf24@1670 123 #endif
kaf24@1452 124 EXPORT_SYMBOL(mmu_cr4_features);
kaf24@1452 125
kaf24@4818 126 int acpi_disabled;
kaf24@1452 127
kaf24@4850 128 int acpi_force;
kaf24@4850 129 char acpi_param[10] = "";
kfraser@15074 130 static void __init parse_acpi_param(char *s)
kaf24@4850 131 {
kaf24@4850 132 /* Save the parameter so it can be propagated to domain0. */
kfraser@13689 133 safe_strcpy(acpi_param, s);
kaf24@4850 134
kaf24@4850 135 /* Interpret the parameter for use within Xen. */
kaf24@4850 136 if ( !strcmp(s, "off") )
kaf24@4850 137 {
kaf24@4850 138 disable_acpi();
kaf24@4850 139 }
kaf24@4850 140 else if ( !strcmp(s, "force") )
kaf24@4850 141 {
kaf24@4850 142 acpi_force = 1;
kaf24@4850 143 acpi_ht = 1;
kaf24@4850 144 acpi_disabled = 0;
kaf24@4850 145 }
kaf24@4850 146 else if ( !strcmp(s, "strict") )
kaf24@4850 147 {
kaf24@4850 148 acpi_strict = 1;
kaf24@4850 149 }
kaf24@4850 150 else if ( !strcmp(s, "ht") )
kaf24@4850 151 {
kaf24@4850 152 if ( !acpi_force )
kaf24@4850 153 disable_acpi();
kaf24@4850 154 acpi_ht = 1;
kaf24@4850 155 }
kaf24@4850 156 else if ( !strcmp(s, "noirq") )
kaf24@4850 157 {
kaf24@4850 158 acpi_noirq_set();
kaf24@4850 159 }
kaf24@4850 160 }
kaf24@4850 161
kaf24@1452 162 static void __init do_initcalls(void)
kaf24@1452 163 {
kaf24@1452 164 initcall_t *call;
kaf24@1452 165 for ( call = &__initcall_start; call < &__initcall_end; call++ )
kaf24@1452 166 (*call)();
kaf24@1452 167 }
kaf24@1452 168
kfraser@15074 169 #define EARLY_FAIL(f, a...) do { \
kfraser@15074 170 printk( f , ## a ); \
kfraser@15871 171 for ( ; ; ) halt(); \
kfraser@15074 172 } while (0)
kaf24@8459 173
kfraser@15074 174 static unsigned long __initdata initial_images_start, initial_images_end;
kaf24@9067 175
kfraser@15074 176 unsigned long __init initial_images_nrpages(void)
kaf24@9067 177 {
kfraser@15489 178 ASSERT(!(initial_images_start & ~PAGE_MASK));
kfraser@15489 179 ASSERT(!(initial_images_end & ~PAGE_MASK));
kfraser@15489 180 return ((initial_images_end >> PAGE_SHIFT) -
kfraser@15489 181 (initial_images_start >> PAGE_SHIFT));
kaf24@9067 182 }
kaf24@9067 183
kfraser@15074 184 void __init discard_initial_images(void)
kaf24@9067 185 {
kaf24@9067 186 init_domheap_pages(initial_images_start, initial_images_end);
kaf24@9067 187 }
kaf24@9067 188
kaf24@9818 189 extern char __per_cpu_start[], __per_cpu_data_end[], __per_cpu_end[];
kaf24@9818 190
kfraser@11241 191 static void __init percpu_init_areas(void)
kaf24@9818 192 {
kaf24@9818 193 unsigned int i, data_size = __per_cpu_data_end - __per_cpu_start;
kfraser@15074 194 unsigned int first_unused;
kaf24@9818 195
kaf24@9818 196 BUG_ON(data_size > PERCPU_SIZE);
kaf24@9818 197
kfraser@15074 198 /* Initialise per-cpu data area for all possible secondary CPUs. */
kfraser@15074 199 for ( i = 1; (i < NR_CPUS) && cpu_possible(i); i++ )
kfraser@15074 200 memcpy(__per_cpu_start + (i << PERCPU_SHIFT),
kfraser@15074 201 __per_cpu_start,
kfraser@15074 202 data_size);
kaf24@9818 203 first_unused = i;
kaf24@9818 204
kfraser@14340 205 /* Check that there are no holes in cpu_possible_map. */
kaf24@9818 206 for ( ; i < NR_CPUS; i++ )
kfraser@14340 207 BUG_ON(cpu_possible(i));
kaf24@9818 208
kfraser@11241 209 #ifndef MEMORY_GUARD
kaf24@9818 210 init_xenheap_pages(__pa(__per_cpu_start) + (first_unused << PERCPU_SHIFT),
kaf24@9818 211 __pa(__per_cpu_end));
kfraser@11241 212 #endif
keir@15082 213 memguard_guard_range(&__per_cpu_start[first_unused << PERCPU_SHIFT],
keir@15082 214 (NR_CPUS - first_unused) << PERCPU_SHIFT);
keir@15082 215 #if defined(CONFIG_X86_64)
keir@15082 216 /* Also zap the mapping in the 1:1 area. */
keir@15082 217 memguard_guard_range(__va(__pa(__per_cpu_start)) +
keir@15082 218 (first_unused << PERCPU_SHIFT),
keir@15082 219 (NR_CPUS - first_unused) << PERCPU_SHIFT);
keir@15082 220 #endif
kaf24@9818 221 }
kaf24@9818 222
kfraser@11241 223 static void __init init_idle_domain(void)
kfraser@11240 224 {
kfraser@11240 225 struct domain *idle_domain;
kfraser@11240 226
kfraser@11240 227 /* Domain creation requires that scheduler structures are initialised. */
kfraser@11240 228 scheduler_init();
kfraser@11240 229
kfraser@14911 230 idle_domain = domain_create(IDLE_DOMAIN_ID, 0, 0);
kfraser@11240 231 if ( (idle_domain == NULL) || (alloc_vcpu(idle_domain, 0, 0) == NULL) )
kfraser@11240 232 BUG();
kfraser@11240 233
kfraser@11240 234 set_current(idle_domain->vcpu[0]);
kfraser@11240 235 idle_vcpu[0] = this_cpu(curr_vcpu) = current;
kfraser@11240 236
kfraser@11240 237 setup_idle_pagetable();
kfraser@11240 238 }
kfraser@11240 239
kfraser@15074 240 static void __init srat_detect_node(int cpu)
kfraser@11971 241 {
kfraser@11998 242 unsigned node;
kfraser@11998 243 u8 apicid = x86_cpu_to_apicid[cpu];
kfraser@11971 244
kfraser@11998 245 node = apicid_to_node[apicid];
kfraser@11998 246 if ( node == NUMA_NO_NODE )
kfraser@11998 247 node = 0;
kfraser@11998 248 numa_set_node(cpu, node);
kfraser@11971 249
kfraser@11998 250 if ( acpi_numa > 0 )
kfraser@11998 251 printk(KERN_INFO "CPU %d APIC %d -> Node %d\n", cpu, apicid, node);
kfraser@11971 252 }
kfraser@11971 253
kfraser@15740 254 /*
kfraser@15740 255 * Ensure a given physical memory range is present in the bootstrap mappings.
kfraser@15740 256 * Use superpage mappings to ensure that pagetable memory needn't be allocated.
kfraser@15740 257 */
kfraser@15740 258 static void __init bootstrap_map(unsigned long start, unsigned long end)
kfraser@15740 259 {
kfraser@15740 260 unsigned long mask = (1UL << L2_PAGETABLE_SHIFT) - 1;
kfraser@15740 261 start = start & ~mask;
kfraser@15740 262 end = (end + mask) & ~mask;
kfraser@15740 263 if ( end > BOOTSTRAP_DIRECTMAP_END )
kfraser@15740 264 panic("Cannot access memory beyond end of "
kfraser@15740 265 "bootstrap direct-map area\n");
kfraser@15740 266 map_pages_to_xen(
kfraser@15740 267 (unsigned long)maddr_to_bootstrap_virt(start),
kfraser@15740 268 start >> PAGE_SHIFT, (end-start) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kfraser@15740 269 }
kfraser@15740 270
kfraser@15074 271 static void __init move_memory(
kfraser@15074 272 unsigned long dst, unsigned long src_start, unsigned long src_end)
ian@12677 273 {
kfraser@15740 274 bootstrap_map(src_start, src_end);
kfraser@15740 275 bootstrap_map(dst, dst + src_end - src_start);
kfraser@15074 276 memmove(maddr_to_bootstrap_virt(dst),
kfraser@15074 277 maddr_to_bootstrap_virt(src_start),
ian@12677 278 src_end - src_start);
kfraser@15074 279 }
kfraser@15074 280
kfraser@15074 281 /* A temporary copy of the e820 map that we can mess with during bootstrap. */
kfraser@15074 282 static struct e820map __initdata boot_e820;
kfraser@15074 283
keir@15298 284 struct boot_video_info {
keir@15298 285 u8 orig_x; /* 0x00 */
keir@15298 286 u8 orig_y; /* 0x01 */
keir@15298 287 u8 orig_video_mode; /* 0x02 */
keir@15298 288 u8 orig_video_cols; /* 0x03 */
keir@15298 289 u8 orig_video_lines; /* 0x04 */
keir@15298 290 u8 orig_video_isVGA; /* 0x05 */
keir@15298 291 u16 orig_video_points; /* 0x06 */
keir@15298 292
keir@15298 293 /* VESA graphic mode -- linear frame buffer */
keir@15298 294 u32 capabilities; /* 0x08 */
keir@15298 295 u16 lfb_linelength; /* 0x0c */
keir@15298 296 u16 lfb_width; /* 0x0e */
keir@15298 297 u16 lfb_height; /* 0x10 */
keir@15298 298 u16 lfb_depth; /* 0x12 */
keir@15298 299 u32 lfb_base; /* 0x14 */
keir@15298 300 u32 lfb_size; /* 0x18 */
keir@15298 301 u8 red_size; /* 0x1c */
keir@15298 302 u8 red_pos; /* 0x1d */
keir@15298 303 u8 green_size; /* 0x1e */
keir@15298 304 u8 green_pos; /* 0x1f */
keir@15298 305 u8 blue_size; /* 0x20 */
keir@15298 306 u8 blue_pos; /* 0x21 */
keir@15298 307 u8 rsvd_size; /* 0x22 */
keir@15298 308 u8 rsvd_pos; /* 0x23 */
keir@15298 309 u16 vesapm_seg; /* 0x24 */
keir@15298 310 u16 vesapm_off; /* 0x26 */
keir@16124 311 u16 vesa_attrib; /* 0x28 */
keir@15298 312 };
keir@15298 313
keir@15298 314 static void __init parse_video_info(void)
keir@15298 315 {
keir@15298 316 struct boot_video_info *bvi = &bootsym(boot_vid_info);
keir@15298 317
keir@15298 318 if ( (bvi->orig_video_isVGA == 1) && (bvi->orig_video_mode == 3) )
keir@15298 319 {
keir@15298 320 vga_console_info.video_type = XEN_VGATYPE_TEXT_MODE_3;
keir@15298 321 vga_console_info.u.text_mode_3.font_height = bvi->orig_video_points;
keir@15298 322 vga_console_info.u.text_mode_3.cursor_x = bvi->orig_x;
keir@15298 323 vga_console_info.u.text_mode_3.cursor_y = bvi->orig_y;
keir@15298 324 vga_console_info.u.text_mode_3.rows = bvi->orig_video_lines;
keir@15298 325 vga_console_info.u.text_mode_3.columns = bvi->orig_video_cols;
keir@15298 326 }
keir@15298 327 else if ( bvi->orig_video_isVGA == 0x23 )
keir@15298 328 {
keir@15298 329 vga_console_info.video_type = XEN_VGATYPE_VESA_LFB;
keir@15298 330 vga_console_info.u.vesa_lfb.width = bvi->lfb_width;
keir@15298 331 vga_console_info.u.vesa_lfb.height = bvi->lfb_height;
keir@15298 332 vga_console_info.u.vesa_lfb.bytes_per_line = bvi->lfb_linelength;
keir@15298 333 vga_console_info.u.vesa_lfb.bits_per_pixel = bvi->lfb_depth;
keir@15298 334 vga_console_info.u.vesa_lfb.lfb_base = bvi->lfb_base;
keir@15298 335 vga_console_info.u.vesa_lfb.lfb_size = bvi->lfb_size;
keir@15298 336 vga_console_info.u.vesa_lfb.red_pos = bvi->red_pos;
keir@15298 337 vga_console_info.u.vesa_lfb.red_size = bvi->red_size;
keir@15298 338 vga_console_info.u.vesa_lfb.green_pos = bvi->green_pos;
keir@15298 339 vga_console_info.u.vesa_lfb.green_size = bvi->green_size;
keir@15298 340 vga_console_info.u.vesa_lfb.blue_pos = bvi->blue_pos;
keir@15298 341 vga_console_info.u.vesa_lfb.blue_size = bvi->blue_size;
keir@15298 342 vga_console_info.u.vesa_lfb.rsvd_pos = bvi->rsvd_pos;
keir@15298 343 vga_console_info.u.vesa_lfb.rsvd_size = bvi->rsvd_size;
keir@16124 344 vga_console_info.u.vesa_lfb.gbl_caps = bvi->capabilities;
keir@16124 345 vga_console_info.u.vesa_lfb.mode_attrs = bvi->vesa_attrib;
keir@15298 346 }
keir@15298 347 }
keir@15298 348
keir@15082 349 void init_done(void)
keir@15082 350 {
keir@15082 351 extern char __init_begin[], __init_end[];
keir@15082 352
keir@15082 353 /* Free (or page-protect) the init areas. */
keir@15082 354 #ifndef MEMORY_GUARD
keir@15082 355 init_xenheap_pages(__pa(__init_begin), __pa(__init_end));
keir@15082 356 #endif
keir@15082 357 memguard_guard_range(__init_begin, __init_end - __init_begin);
keir@15082 358 #if defined(CONFIG_X86_64)
keir@15082 359 /* Also zap the mapping in the 1:1 area. */
keir@15082 360 memguard_guard_range(__va(__pa(__init_begin)), __init_end - __init_begin);
keir@15082 361 #endif
keir@15082 362 printk("Freed %ldkB init memory.\n", (long)(__init_end-__init_begin)>>10);
keir@15082 363
keir@15082 364 startup_cpu_idle_loop();
keir@15082 365 }
keir@15082 366
kfraser@15796 367 static char * __init cmdline_cook(char *p)
kfraser@15796 368 {
kfraser@15796 369 p = p ? : "";
kfraser@15796 370 while ( *p == ' ' )
kfraser@15796 371 p++;
kfraser@15796 372 while ( (*p != ' ') && (*p != '\0') )
kfraser@15796 373 p++;
kfraser@15796 374 while ( *p == ' ' )
kfraser@15796 375 p++;
kfraser@15796 376 return p;
kfraser@15796 377 }
kfraser@15796 378
kfraser@15379 379 void __init __start_xen(unsigned long mbi_p)
kaf24@1452 380 {
kfraser@15293 381 char *memmap_type = NULL;
kfraser@15796 382 char *cmdline, *kextra;
kaf24@8457 383 unsigned long _initrd_start = 0, _initrd_len = 0;
kaf24@8457 384 unsigned int initrdidx = 1;
kfraser@15379 385 multiboot_info_t *mbi = __va(mbi_p);
kaf24@8457 386 module_t *mod = (module_t *)__va(mbi->mods_addr);
kaf24@6111 387 unsigned long nr_pages, modules_length;
kfraser@15293 388 int i, e820_warn = 0, bytes = 0;
kaf24@5776 389 struct ns16550_defaults ns16550 = {
kaf24@5776 390 .data_bits = 8,
kaf24@5776 391 .parity = 'n',
kaf24@5776 392 .stop_bits = 1
kaf24@5776 393 };
kaf24@3338 394
kfraser@12853 395 extern void early_page_fault(void);
kfraser@12853 396 set_intr_gate(TRAP_page_fault, &early_page_fault);
kfraser@12853 397
kaf24@3338 398 /* Parse the command-line options. */
kfraser@15796 399 cmdline = cmdline_cook((mbi->flags & MBI_CMDLINE) ?
kfraser@15796 400 __va(mbi->cmdline) : NULL);
kfraser@15426 401 if ( (kextra = strstr(cmdline, " -- ")) != NULL )
kfraser@15426 402 {
kfraser@15426 403 /*
kfraser@15426 404 * Options after ' -- ' separator belong to dom0.
kfraser@15426 405 * 1. Orphan dom0's options from Xen's command line.
kfraser@15426 406 * 2. Skip all but final leading space from dom0's options.
kfraser@15426 407 */
kfraser@15426 408 *kextra = '\0';
kfraser@15426 409 kextra += 3;
kfraser@15426 410 while ( kextra[1] == ' ' ) kextra++;
kfraser@15426 411 }
kaf24@9823 412 cmdline_parse(cmdline);
kaf24@3338 413
keir@15298 414 parse_video_info();
keir@15298 415
kaf24@8534 416 set_current((struct vcpu *)0xfffff000); /* debug sanity */
kfraser@11240 417 idle_vcpu[0] = current;
kaf24@8534 418 set_processor_id(0); /* needed early, for smp_processor_id() */
keir@16267 419 rdmsrl(MSR_EFER, this_cpu(efer));
keir@16267 420 asm volatile ( "mov %%cr4,%0" : "=r" (this_cpu(cr4)) );
kaf24@3338 421
kaf24@5146 422 smp_prepare_boot_cpu();
kaf24@5146 423
kaf24@3338 424 /* We initialise the serial devices very early so we can get debugging. */
kaf24@5776 425 ns16550.io_base = 0x3f8;
kaf24@5776 426 ns16550.irq = 4;
kaf24@5776 427 ns16550_init(0, &ns16550);
kaf24@5776 428 ns16550.io_base = 0x2f8;
kaf24@5776 429 ns16550.irq = 3;
kaf24@5776 430 ns16550_init(1, &ns16550);
kaf24@5195 431 serial_init_preirq();
kaf24@3338 432
kaf24@3338 433 init_console();
kaf24@3338 434
kfraser@11947 435 printk("Command line: %s\n", cmdline);
kaf24@9823 436
kfraser@15330 437 printk("Video information:\n");
kfraser@15330 438
kfraser@15330 439 /* Print VGA display mode information. */
keir@15298 440 switch ( vga_console_info.video_type )
keir@15298 441 {
keir@15298 442 case XEN_VGATYPE_TEXT_MODE_3:
kfraser@15330 443 printk(" VGA is text mode %dx%d, font 8x%d\n",
keir@15298 444 vga_console_info.u.text_mode_3.columns,
keir@15298 445 vga_console_info.u.text_mode_3.rows,
keir@15298 446 vga_console_info.u.text_mode_3.font_height);
keir@15298 447 break;
keir@15298 448 case XEN_VGATYPE_VESA_LFB:
kfraser@15330 449 printk(" VGA is graphics mode %dx%d, %d bpp\n",
keir@15298 450 vga_console_info.u.vesa_lfb.width,
keir@15298 451 vga_console_info.u.vesa_lfb.height,
keir@15298 452 vga_console_info.u.vesa_lfb.bits_per_pixel);
keir@15298 453 break;
kfraser@15330 454 default:
kfraser@15330 455 printk(" No VGA detected\n");
kfraser@15330 456 break;
kfraser@15330 457 }
kfraser@15330 458
kfraser@15330 459 /* Print VBE/DDC EDID information. */
kfraser@15330 460 if ( bootsym(boot_edid_caps) != 0x1313 )
kfraser@15330 461 {
kfraser@15330 462 u16 caps = bootsym(boot_edid_caps);
kfraser@15330 463 printk(" VBE/DDC methods:%s%s%s; ",
kfraser@15330 464 (caps & 1) ? " V1" : "",
kfraser@15330 465 (caps & 2) ? " V2" : "",
kfraser@15330 466 !(caps & 3) ? " none" : "");
kfraser@15330 467 printk("EDID transfer time: %d seconds\n", caps >> 8);
kfraser@15330 468 if ( *(u32 *)bootsym(boot_edid_info) == 0x13131313 )
kfraser@15330 469 {
kfraser@15330 470 printk(" EDID info not retrieved because ");
kfraser@15330 471 if ( !(caps & 3) )
kfraser@15330 472 printk("no DDC retrieval method detected\n");
kfraser@15330 473 else if ( (caps >> 8) > 5 )
kfraser@15330 474 printk("takes longer than 5 seconds\n");
kfraser@15330 475 else
kfraser@15330 476 printk("of reasons unknown\n");
kfraser@15330 477 }
keir@15298 478 }
keir@15298 479
kfraser@15336 480 printk("Disc information:\n");
kfraser@15336 481 printk(" Found %d MBR signatures\n",
kfraser@15430 482 bootsym(boot_mbr_signature_nr));
kfraser@15336 483 printk(" Found %d EDD information structures\n",
kfraser@15336 484 bootsym(boot_edd_info_nr));
kfraser@15336 485
kaf24@3344 486 /* Check that we have at least one Multiboot module. */
kaf24@3344 487 if ( !(mbi->flags & MBI_MODULES) || (mbi->mods_count == 0) )
kfraser@15074 488 EARLY_FAIL("dom0 kernel not specified. "
kfraser@15074 489 "Check bootloader configuration.\n");
kaf24@5011 490
kaf24@5011 491 if ( ((unsigned long)cpu0_stack & (STACK_SIZE-1)) != 0 )
kfraser@15074 492 EARLY_FAIL("Misaligned CPU0 stack.\n");
kaf24@3338 493
kfraser@11618 494 /*
kfraser@11618 495 * Since there are some stubs getting built on the stacks which use
kfraser@11618 496 * direct calls/jumps, the heap must be confined to the lower 2G so
kfraser@11618 497 * that those branches can reach their targets.
kfraser@11618 498 */
kfraser@11618 499 if ( opt_xenheap_megabytes > 2048 )
kfraser@11618 500 opt_xenheap_megabytes = 2048;
kaf24@3338 501
kfraser@15293 502 if ( e820_raw_nr != 0 )
kfraser@15292 503 {
kfraser@15293 504 memmap_type = "Xen-e820";
kfraser@15292 505 }
kfraser@15293 506 else if ( bootsym(lowmem_kb) )
kfraser@15292 507 {
kfraser@15293 508 memmap_type = "Xen-e801";
kfraser@15292 509 e820_raw[0].addr = 0;
kfraser@15293 510 e820_raw[0].size = bootsym(lowmem_kb) << 10;
kfraser@15292 511 e820_raw[0].type = E820_RAM;
kfraser@15292 512 e820_raw[1].addr = 0x100000;
kfraser@15293 513 e820_raw[1].size = bootsym(highmem_kb) << 10;
kfraser@15292 514 e820_raw[1].type = E820_RAM;
kfraser@15292 515 e820_raw_nr = 2;
kfraser@15292 516 }
kfraser@15292 517 else if ( mbi->flags & MBI_MEMMAP )
kaf24@3344 518 {
kfraser@15293 519 memmap_type = "Multiboot-e820";
keir@15988 520 while ( (bytes < mbi->mmap_length) && (e820_raw_nr < E820MAX) )
kaf24@3344 521 {
kaf24@3344 522 memory_map_t *map = __va(mbi->mmap_addr + bytes);
kaf24@8402 523
kaf24@8402 524 /*
kaf24@8403 525 * This is a gross workaround for a BIOS bug. Some bootloaders do
kaf24@8402 526 * not write e820 map entries into pre-zeroed memory. This is
kaf24@8402 527 * okay if the BIOS fills in all fields of the map entry, but
kaf24@8402 528 * some broken BIOSes do not bother to write the high word of
kaf24@8402 529 * the length field if the length is smaller than 4GB. We
kaf24@8402 530 * detect and fix this by flagging sections below 4GB that
kaf24@8403 531 * appear to be larger than 4GB in size.
kaf24@8402 532 */
kaf24@8403 533 if ( (map->base_addr_high == 0) && (map->length_high != 0) )
kaf24@8402 534 {
kfraser@15292 535 if ( !e820_warn )
kfraser@15292 536 {
kfraser@15292 537 printk("WARNING: Buggy e820 map detected and fixed "
kfraser@15292 538 "(truncated length fields).\n");
kfraser@15292 539 e820_warn = 1;
kfraser@15292 540 }
kaf24@8402 541 map->length_high = 0;
kaf24@8402 542 }
kaf24@8402 543
kaf24@3344 544 e820_raw[e820_raw_nr].addr =
kaf24@3344 545 ((u64)map->base_addr_high << 32) | (u64)map->base_addr_low;
kaf24@3344 546 e820_raw[e820_raw_nr].size =
kaf24@3344 547 ((u64)map->length_high << 32) | (u64)map->length_low;
kfraser@15799 548 e820_raw[e820_raw_nr].type = map->type;
kaf24@3344 549 e820_raw_nr++;
kaf24@8402 550
kaf24@3344 551 bytes += map->size + 4;
kaf24@3344 552 }
kaf24@3344 553 }
kaf24@3344 554 else if ( mbi->flags & MBI_MEMLIMITS )
kaf24@3344 555 {
kfraser@15293 556 memmap_type = "Multiboot-e801";
kaf24@3344 557 e820_raw[0].addr = 0;
kaf24@3344 558 e820_raw[0].size = mbi->mem_lower << 10;
kaf24@3344 559 e820_raw[0].type = E820_RAM;
kaf24@3354 560 e820_raw[1].addr = 0x100000;
kaf24@3354 561 e820_raw[1].size = mbi->mem_upper << 10;
kaf24@3354 562 e820_raw[1].type = E820_RAM;
kaf24@3344 563 e820_raw_nr = 2;
kaf24@3344 564 }
kaf24@3344 565 else
kaf24@3344 566 {
kfraser@15074 567 EARLY_FAIL("Bootloader provided no memory information.\n");
kaf24@3344 568 }
kaf24@3344 569
kaf24@13427 570 /* Sanitise the raw E820 map to produce a final clean version. */
kfraser@15293 571 max_page = init_e820(memmap_type, e820_raw, &e820_raw_nr);
kaf24@3338 572
kfraser@15074 573 /*
kfraser@15074 574 * Create a temporary copy of the E820 map. Truncate it to above 16MB
kfraser@15074 575 * as anything below that is already mapped and has a statically-allocated
kfraser@15074 576 * purpose.
kfraser@15074 577 */
kfraser@15074 578 memcpy(&boot_e820, &e820, sizeof(e820));
kfraser@15074 579 for ( i = 0; i < boot_e820.nr_map; i++ )
kaf24@3338 580 {
kfraser@15074 581 uint64_t s, e, min = 16 << 20; /* 16MB */
kfraser@15074 582 s = boot_e820.map[i].addr;
kfraser@15074 583 e = boot_e820.map[i].addr + boot_e820.map[i].size;
kfraser@15074 584 if ( s >= min )
kfraser@15074 585 continue;
kfraser@15074 586 if ( e > min )
kaf24@6111 587 {
kfraser@15074 588 boot_e820.map[i].addr = min;
kfraser@15074 589 boot_e820.map[i].size = e - min;
kaf24@6111 590 }
kfraser@15074 591 else
kfraser@15074 592 boot_e820.map[i].type = E820_RESERVED;
kaf24@3338 593 }
kaf24@6111 594
kfraser@15074 595 /*
keir@15077 596 * Iterate backwards over all superpage-aligned RAM regions.
kfraser@15074 597 *
kfraser@15074 598 * We require superpage alignment because the boot allocator is not yet
kfraser@15074 599 * initialised. Hence we can only map superpages in the address range
kfraser@15074 600 * 0 to BOOTSTRAP_DIRECTMAP_END, as this is guaranteed not to require
kfraser@15074 601 * dynamic allocation of pagetables.
kfraser@15074 602 *
kfraser@15074 603 * As well as mapping superpages in that range, in preparation for
kfraser@15074 604 * initialising the boot allocator, we also look for a region to which
kfraser@15074 605 * we can relocate the dom0 kernel and other multiboot modules. Also, on
kfraser@15074 606 * x86/64, we relocate Xen to higher memory.
kfraser@15074 607 */
kfraser@15074 608 modules_length = mod[mbi->mods_count-1].mod_end - mod[0].mod_start;
keir@15077 609 for ( i = boot_e820.nr_map-1; i >= 0; i-- )
kfraser@15074 610 {
kfraser@15074 611 uint64_t s, e, mask = (1UL << L2_PAGETABLE_SHIFT) - 1;
kaf24@6134 612
kfraser@15074 613 /* Superpage-aligned chunks up to BOOTSTRAP_DIRECTMAP_END, please. */
kfraser@15074 614 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 615 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
kfraser@15074 616 e = min_t(uint64_t, e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 617 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kaf24@3354 618 continue;
kaf24@6111 619
kfraser@15074 620 /* Map the chunk. No memory will need to be allocated to do this. */
kfraser@15074 621 map_pages_to_xen(
kfraser@15074 622 (unsigned long)maddr_to_bootstrap_virt(s),
kfraser@15074 623 s >> PAGE_SHIFT, (e-s) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kaf24@6111 624
kfraser@14084 625 #if defined(CONFIG_X86_64)
kfraser@15074 626 /* Is the region suitable for relocating Xen? */
kfraser@15074 627 if ( !xen_phys_start && (((e-s) >> 20) >= opt_xenheap_megabytes) )
kaf24@5003 628 {
kfraser@15074 629 extern l2_pgentry_t l2_xenmap[];
kfraser@15074 630 l4_pgentry_t *pl4e;
kfraser@15074 631 l3_pgentry_t *pl3e;
kfraser@15074 632 l2_pgentry_t *pl2e;
kfraser@15074 633 int i, j;
kfraser@15074 634
kfraser@15074 635 /* Select relocation address. */
kfraser@15074 636 e = (e - (opt_xenheap_megabytes << 20)) & ~mask;
kfraser@15074 637 xen_phys_start = e;
kfraser@15292 638 bootsym(trampoline_xen_phys_start) = e;
kfraser@15074 639
kfraser@15074 640 /*
kfraser@15074 641 * Perform relocation to new physical address.
kfraser@15074 642 * Before doing so we must sync static/global data with main memory
kfraser@15074 643 * with a barrier(). After this we must *not* modify static/global
kfraser@15074 644 * data until after we have switched to the relocated pagetables!
kfraser@15074 645 */
kfraser@15074 646 barrier();
kfraser@15074 647 move_memory(e, 0, __pa(&_end) - xen_phys_start);
kfraser@15074 648
kfraser@15379 649 /* Poison low 1MB to detect stray pointers to physical 0-1MB. */
kfraser@15379 650 memset(maddr_to_bootstrap_virt(e), 0x55, 1U<<20);
kfraser@15379 651
kfraser@15074 652 /* Walk initial pagetables, relocating page directory entries. */
kfraser@15074 653 pl4e = __va(__pa(idle_pg_table));
kfraser@15074 654 for ( i = 0 ; i < L4_PAGETABLE_ENTRIES; i++, pl4e++ )
kfraser@15074 655 {
kfraser@15074 656 if ( !(l4e_get_flags(*pl4e) & _PAGE_PRESENT) )
kfraser@15074 657 continue;
kfraser@15074 658 *pl4e = l4e_from_intpte(l4e_get_intpte(*pl4e) +
kfraser@15074 659 xen_phys_start);
kfraser@15074 660 pl3e = l4e_to_l3e(*pl4e);
kfraser@15074 661 for ( j = 0; j < L3_PAGETABLE_ENTRIES; j++, pl3e++ )
kfraser@15074 662 {
kfraser@15074 663 /* Not present or already relocated? */
kfraser@15074 664 if ( !(l3e_get_flags(*pl3e) & _PAGE_PRESENT) ||
kfraser@15074 665 (l3e_get_pfn(*pl3e) > 0x1000) )
kfraser@15074 666 continue;
kfraser@15074 667 *pl3e = l3e_from_intpte(l3e_get_intpte(*pl3e) +
kfraser@15074 668 xen_phys_start);
kfraser@15074 669 }
kfraser@15074 670 }
kfraser@15074 671
kfraser@15074 672 /* The only data mappings to be relocated are in the Xen area. */
kfraser@15074 673 pl2e = __va(__pa(l2_xenmap));
kfraser@15074 674 for ( i = 0; i < L2_PAGETABLE_ENTRIES; i++, pl2e++ )
kfraser@15074 675 {
kfraser@15074 676 if ( !(l2e_get_flags(*pl2e) & _PAGE_PRESENT) )
kfraser@15074 677 continue;
kfraser@15074 678 *pl2e = l2e_from_intpte(l2e_get_intpte(*pl2e) +
kfraser@15074 679 xen_phys_start);
kfraser@15074 680 }
kfraser@15074 681
kfraser@15074 682 /* Re-sync the stack and then switch to relocated pagetables. */
kfraser@15074 683 asm volatile (
kfraser@15074 684 "rep movsb ; " /* re-sync the stack */
kfraser@15074 685 "movq %%cr4,%%rsi ; "
kfraser@15074 686 "andb $0x7f,%%sil ; "
kfraser@15074 687 "movq %%rsi,%%cr4 ; " /* CR4.PGE == 0 */
kfraser@15074 688 "movq %0,%%cr3 ; " /* CR3 == new pagetables */
kfraser@15074 689 "orb $0x80,%%sil ; "
kfraser@15074 690 "movq %%rsi,%%cr4 " /* CR4.PGE == 1 */
kfraser@15074 691 : : "r" (__pa(idle_pg_table)), "S" (cpu0_stack),
kfraser@15074 692 "D" (__va(__pa(cpu0_stack))), "c" (STACK_SIZE) : "memory" );
kaf24@5003 693 }
kaf24@5003 694 #endif
keir@15077 695
keir@15077 696 /* Is the region suitable for relocating the multiboot modules? */
keir@15077 697 if ( !initial_images_start && (s < e) && ((e-s) >= modules_length) )
keir@15077 698 {
kfraser@15489 699 initial_images_end = e;
kfraser@15489 700 e = (e - modules_length) & PAGE_MASK;
keir@15077 701 initial_images_start = e;
keir@15077 702 move_memory(initial_images_start,
keir@15077 703 mod[0].mod_start, mod[mbi->mods_count-1].mod_end);
keir@15077 704 }
kfraser@15489 705
kfraser@15489 706 if ( !kexec_crash_area.start && (s < e) &&
kfraser@15489 707 ((e-s) >= kexec_crash_area.size) )
kfraser@15489 708 {
kfraser@15489 709 e = (e - kexec_crash_area.size) & PAGE_MASK;
kfraser@15489 710 kexec_crash_area.start = e;
kfraser@15489 711 }
kaf24@3354 712 }
kaf24@3354 713
kfraser@15074 714 if ( !initial_images_start )
kfraser@15074 715 EARLY_FAIL("Not enough memory to relocate the dom0 kernel image.\n");
keir@15988 716 reserve_e820_ram(&boot_e820, initial_images_start, initial_images_end);
kfraser@15074 717
kfraser@15074 718 /*
kfraser@15074 719 * With modules (and Xen itself, on x86/64) relocated out of the way, we
kfraser@15074 720 * can now initialise the boot allocator with some memory.
kfraser@15074 721 */
kfraser@15074 722 xenheap_phys_start = init_boot_allocator(__pa(&_end));
kfraser@15074 723 xenheap_phys_end = opt_xenheap_megabytes << 20;
kfraser@15074 724 #if defined(CONFIG_X86_64)
kfraser@15074 725 if ( !xen_phys_start )
kfraser@15074 726 EARLY_FAIL("Not enough memory to relocate Xen.\n");
kfraser@15074 727 xenheap_phys_end += xen_phys_start;
keir@15988 728 reserve_e820_ram(&boot_e820, xen_phys_start,
keir@15988 729 xen_phys_start + (opt_xenheap_megabytes<<20));
kfraser@15074 730 init_boot_pages(1<<20, 16<<20); /* Initial seed: 15MB */
kfraser@15074 731 #else
kfraser@15074 732 init_boot_pages(xenheap_phys_end, 16<<20); /* Initial seed: 4MB */
kfraser@15074 733 #endif
kfraser@15074 734
kfraser@15489 735 if ( kexec_crash_area.size != 0 )
kfraser@15489 736 {
kfraser@15489 737 unsigned long kdump_start = kexec_crash_area.start;
kfraser@15489 738 unsigned long kdump_size = kexec_crash_area.size;
kfraser@15489 739
kfraser@15489 740 kdump_size = (kdump_size + PAGE_SIZE - 1) & PAGE_MASK;
kfraser@15489 741
keir@15988 742 if ( !reserve_e820_ram(&boot_e820, kdump_start, kdump_size) )
kfraser@15489 743 {
kfraser@15489 744 printk("Kdump: DISABLED (failed to reserve %luMB (%lukB) at 0x%lx)"
kfraser@15489 745 "\n", kdump_size >> 20, kdump_size >> 10, kdump_start);
kfraser@15489 746 kexec_crash_area.start = kexec_crash_area.size = 0;
kfraser@15489 747 }
kfraser@15489 748 else
kfraser@15489 749 {
kfraser@15489 750 printk("Kdump: %luMB (%lukB) at 0x%lx\n",
kfraser@15489 751 kdump_size >> 20, kdump_size >> 10, kdump_start);
kfraser@15489 752 }
kfraser@15489 753 }
kfraser@15489 754
kfraser@15074 755 /*
kfraser@15074 756 * With the boot allocator now seeded, we can walk every RAM region and
kfraser@15074 757 * map it in its entirety (on x86/64, at least) and notify it to the
kfraser@15074 758 * boot allocator.
kfraser@15074 759 */
kfraser@15074 760 for ( i = 0; i < boot_e820.nr_map; i++ )
kfraser@15074 761 {
kfraser@15074 762 uint64_t s, e, map_e, mask = PAGE_SIZE - 1;
kfraser@15074 763
kfraser@15074 764 /* Only page alignment required now. */
kfraser@15074 765 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 766 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
kfraser@15074 767 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kfraser@15074 768 continue;
kfraser@15074 769
kfraser@15074 770 /* Perform the mapping (truncated in 32-bit mode). */
kfraser@15074 771 map_e = e;
kfraser@15074 772 #if defined(CONFIG_X86_32)
kfraser@15074 773 map_e = min_t(uint64_t, map_e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 774 #endif
kfraser@15074 775 if ( s < map_e )
kfraser@15074 776 map_pages_to_xen(
kfraser@15074 777 (unsigned long)maddr_to_bootstrap_virt(s),
kfraser@15074 778 s >> PAGE_SHIFT, (map_e-s) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kfraser@15074 779
kfraser@15074 780 init_boot_pages(s, e);
kfraser@15074 781 }
kfraser@15074 782
kaf24@5003 783 memguard_init();
kaf24@4950 784
kfraser@15074 785 nr_pages = 0;
kfraser@15074 786 for ( i = 0; i < e820.nr_map; i++ )
kfraser@15074 787 if ( e820.map[i].type == E820_RAM )
kfraser@15074 788 nr_pages += e820.map[i].size >> PAGE_SHIFT;
ian@12681 789 printk("System RAM: %luMB (%lukB)\n",
kaf24@3354 790 nr_pages >> (20 - PAGE_SHIFT),
kaf24@3354 791 nr_pages << (PAGE_SHIFT - 10));
kaf24@7220 792 total_pages = nr_pages;
kaf24@3354 793
kfraser@11296 794 /* Sanity check for unwanted bloat of certain hypercall structures. */
kfraser@11296 795 BUILD_BUG_ON(sizeof(((struct xen_platform_op *)0)->u) !=
kfraser@11296 796 sizeof(((struct xen_platform_op *)0)->u.pad));
kfraser@11296 797 BUILD_BUG_ON(sizeof(((struct xen_domctl *)0)->u) !=
kfraser@11296 798 sizeof(((struct xen_domctl *)0)->u.pad));
kfraser@11296 799 BUILD_BUG_ON(sizeof(((struct xen_sysctl *)0)->u) !=
kfraser@11296 800 sizeof(((struct xen_sysctl *)0)->u.pad));
kaf24@7388 801
kaf24@9878 802 BUILD_BUG_ON(sizeof(start_info_t) > PAGE_SIZE);
kaf24@9878 803 BUILD_BUG_ON(sizeof(shared_info_t) > PAGE_SIZE);
ack@13292 804 BUILD_BUG_ON(sizeof(struct vcpu_info) != 64);
kaf24@7744 805
ack@13291 806 #ifdef CONFIG_COMPAT
ack@13291 807 BUILD_BUG_ON(sizeof(((struct compat_platform_op *)0)->u) !=
ack@13291 808 sizeof(((struct compat_platform_op *)0)->u.pad));
ack@13291 809 BUILD_BUG_ON(sizeof(start_info_compat_t) > PAGE_SIZE);
ack@13292 810 BUILD_BUG_ON(sizeof(struct compat_vcpu_info) != 64);
ack@13291 811 #endif
ack@13291 812
kfraser@10492 813 /* Check definitions in public headers match internal defs. */
kaf24@9878 814 BUILD_BUG_ON(__HYPERVISOR_VIRT_START != HYPERVISOR_VIRT_START);
kaf24@8521 815 #ifdef HYPERVISOR_VIRT_END
kaf24@9878 816 BUILD_BUG_ON(__HYPERVISOR_VIRT_END != HYPERVISOR_VIRT_END);
kaf24@8521 817 #endif
kfraser@10492 818 BUILD_BUG_ON(MACH2PHYS_VIRT_START != RO_MPT_VIRT_START);
kfraser@10492 819 BUILD_BUG_ON(MACH2PHYS_VIRT_END != RO_MPT_VIRT_END);
kaf24@8521 820
kaf24@3354 821 init_frametable();
kaf24@3338 822
kfraser@11971 823 acpi_boot_table_init();
kfraser@11971 824
kfraser@11971 825 acpi_numa_init();
kfraser@11971 826
kfraser@11971 827 numa_initmem_init(0, max_page);
kfraser@11971 828
kaf24@6111 829 /* Initialise the Xen heap, skipping RAM holes. */
kfraser@15074 830 init_xenheap_pages(xenheap_phys_start, xenheap_phys_end);
kfraser@15074 831 nr_pages = (xenheap_phys_end - xenheap_phys_start) >> PAGE_SHIFT;
kfraser@15074 832 #ifdef __x86_64__
kfraser@15074 833 init_xenheap_pages(xen_phys_start, __pa(&_start));
kfraser@15074 834 nr_pages += (__pa(&_start) - xen_phys_start) >> PAGE_SHIFT;
kfraser@15747 835 vesa_init();
kfraser@15074 836 #endif
kfraser@15074 837 xenheap_phys_start = xen_phys_start;
kaf24@6111 838 printk("Xen heap: %luMB (%lukB)\n",
kaf24@6111 839 nr_pages >> (20 - PAGE_SHIFT),
kaf24@6111 840 nr_pages << (PAGE_SHIFT - 10));
kaf24@3338 841
keir@14680 842 end_boot_allocator();
keir@14680 843
kaf24@3594 844 early_boot = 0;
kaf24@3338 845
kaf24@8459 846 early_cpu_init();
kaf24@8459 847
kaf24@8459 848 paging_init();
kaf24@8459 849
keir@16274 850 tboot_probe();
keir@16274 851
kaf24@8459 852 /* Unmap the first page of CPU0's stack. */
kaf24@8459 853 memguard_guard_stack(cpu0_stack);
kaf24@8459 854
kaf24@8459 855 open_softirq(NEW_TLBFLUSH_CLOCK_PERIOD_SOFTIRQ, new_tlbflush_clock_period);
kaf24@8459 856
kaf24@8459 857 if ( opt_watchdog )
kaf24@8459 858 nmi_watchdog = NMI_LOCAL_APIC;
kaf24@8459 859
kaf24@8459 860 sort_exception_tables();
kaf24@8459 861
kaf24@8459 862 find_smp_config();
kaf24@8459 863
kaf24@8459 864 dmi_scan_machine();
kaf24@8459 865
kaf24@8459 866 generic_apic_probe();
kaf24@8459 867
kaf24@8459 868 acpi_boot_init();
kaf24@8459 869
kfraser@11971 870 init_cpu_to_node();
kfraser@11971 871
kfraser@11241 872 if ( smp_found_config )
kaf24@8459 873 get_smp_config();
kaf24@8459 874
keir@15083 875 #ifdef CONFIG_X86_64
keir@15083 876 /* Low mappings were only needed for some BIOS table parsing. */
keir@15083 877 zap_low_mappings();
keir@15083 878 #endif
keir@15083 879
kaf24@8459 880 init_apic_mappings();
kaf24@8459 881
kaf24@8459 882 init_IRQ();
kaf24@8459 883
kfraser@11241 884 percpu_init_areas();
kfraser@11241 885
kfraser@15815 886 xsm_init(&initrdidx, mbi, initial_images_start);
kfraser@15815 887
kfraser@11240 888 init_idle_domain();
kfraser@11240 889
kaf24@8459 890 trap_init();
kaf24@8459 891
kaf24@13662 892 rcu_init();
kaf24@13662 893
kaf24@8586 894 timer_init();
kaf24@8459 895
kaf24@8459 896 early_time_init();
kaf24@8459 897
kaf24@8459 898 arch_init_memory();
kaf24@8459 899
kaf24@8459 900 identify_cpu(&boot_cpu_data);
kaf24@8459 901 if ( cpu_has_fxsr )
kaf24@8459 902 set_in_cr4(X86_CR4_OSFXSR);
kaf24@8459 903 if ( cpu_has_xmm )
kaf24@8459 904 set_in_cr4(X86_CR4_OSXMMEXCPT);
kfraser@15747 905 #ifdef CONFIG_X86_64
kfraser@15747 906 vesa_mtrr_init();
kfraser@15747 907 #endif
kaf24@8459 908
kaf24@8459 909 if ( opt_nosmp )
kaf24@8459 910 max_cpus = 0;
kaf24@8459 911
kaf24@8459 912 smp_prepare_cpus(max_cpus);
kaf24@8459 913
kaf24@8459 914 /*
kaf24@8459 915 * Initialise higher-level timer functions. We do this fairly late
kaf24@8459 916 * (post-SMP) because the time bases and scale factors need to be updated
kaf24@8459 917 * regularly, and SMP initialisation can cause a long delay with
kaf24@8459 918 * interrupts not yet enabled.
kaf24@8459 919 */
kaf24@8459 920 init_xen_time();
kaf24@8459 921
kaf24@8459 922 initialize_keytable();
kaf24@8459 923
kaf24@8459 924 serial_init_postirq();
kaf24@8459 925
kaf24@8459 926 BUG_ON(!local_irq_is_enabled());
kaf24@8459 927
kaf24@8459 928 for_each_present_cpu ( i )
kaf24@8459 929 {
kaf24@8459 930 if ( num_online_cpus() >= max_cpus )
kaf24@8459 931 break;
kaf24@8459 932 if ( !cpu_online(i) )
kaf24@13662 933 {
kaf24@13662 934 rcu_online_cpu(i);
kaf24@8459 935 __cpu_up(i);
kaf24@13662 936 }
kfraser@11971 937
kfraser@11998 938 /* Set up cpu_to_node[]. */
kfraser@11971 939 srat_detect_node(i);
kfraser@11998 940 /* Set up node_to_cpumask based on cpu_to_node[]. */
kfraser@11971 941 numa_add_cpu(i);
kaf24@8459 942 }
kaf24@8459 943
kaf24@8459 944 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
kaf24@8459 945 smp_cpus_done(max_cpus);
kaf24@8459 946
kaf24@9117 947 initialise_gdb(); /* could be moved earlier */
kaf24@9117 948
kaf24@8459 949 do_initcalls();
kaf24@8459 950
kaf24@8594 951 if ( opt_watchdog )
kaf24@8594 952 watchdog_enable();
kaf24@8459 953
kaf24@8459 954 /* Create initial domain 0. */
kfraser@14911 955 dom0 = domain_create(0, 0, DOM0_SSIDREF);
kfraser@10655 956 if ( (dom0 == NULL) || (alloc_vcpu(dom0, 0, 0) == NULL) )
kaf24@8459 957 panic("Error creating domain 0\n");
kaf24@8459 958
kfraser@12210 959 dom0->is_privileged = 1;
kfraser@12210 960
kaf24@8459 961 /* Grab the DOM0 command line. */
kaf24@8459 962 cmdline = (char *)(mod[0].string ? __va(mod[0].string) : NULL);
kfraser@15426 963 if ( (cmdline != NULL) || (kextra != NULL) )
kaf24@8459 964 {
kaf24@8459 965 static char dom0_cmdline[MAX_GUEST_CMDLINE];
kaf24@8459 966
kfraser@15796 967 cmdline = cmdline_cook(cmdline);
kfraser@15796 968 safe_strcpy(dom0_cmdline, cmdline);
kaf24@8459 969
kfraser@15426 970 if ( kextra != NULL )
kfraser@15426 971 /* kextra always includes exactly one leading space. */
kfraser@15426 972 safe_strcat(dom0_cmdline, kextra);
kfraser@15426 973
kaf24@8459 974 /* Append any extra parameters. */
kfraser@13691 975 if ( skip_ioapic_setup && !strstr(dom0_cmdline, "noapic") )
kfraser@13691 976 safe_strcat(dom0_cmdline, " noapic");
kaf24@8459 977 if ( acpi_skip_timer_override &&
kfraser@13691 978 !strstr(dom0_cmdline, "acpi_skip_timer_override") )
kfraser@13691 979 safe_strcat(dom0_cmdline, " acpi_skip_timer_override");
keir@16165 980 if ( (strlen(acpi_param) == 0) && acpi_disabled )
keir@16165 981 {
keir@16165 982 printk("ACPI is disabled, notifying Domain 0 (acpi=off)\n");
keir@16165 983 safe_strcpy(acpi_param, "off");
keir@16165 984 }
kfraser@13691 985 if ( (strlen(acpi_param) != 0) && !strstr(dom0_cmdline, "acpi=") )
kaf24@8459 986 {
kfraser@13691 987 safe_strcat(dom0_cmdline, " acpi=");
kfraser@13691 988 safe_strcat(dom0_cmdline, acpi_param);
kaf24@8459 989 }
kfraser@13691 990
kfraser@13691 991 cmdline = dom0_cmdline;
kaf24@8459 992 }
kaf24@8459 993
kaf24@8459 994 if ( (initrdidx > 0) && (initrdidx < mbi->mods_count) )
kaf24@8459 995 {
kaf24@8459 996 _initrd_start = initial_images_start +
kaf24@8459 997 (mod[initrdidx].mod_start - mod[0].mod_start);
kaf24@8459 998 _initrd_len = mod[initrdidx].mod_end - mod[initrdidx].mod_start;
kaf24@8459 999 }
kaf24@8459 1000
kfraser@15929 1001 iommu_setup();
kfraser@15929 1002
keir@15957 1003 amd_iommu_detect();
keir@15957 1004
kaf24@8459 1005 /*
kaf24@8459 1006 * We're going to setup domain0 using the module(s) that we stashed safely
kaf24@8459 1007 * above our heap. The second module, if present, is an initrd ramdisk.
kaf24@8459 1008 */
kaf24@8459 1009 if ( construct_dom0(dom0,
kaf24@8459 1010 initial_images_start,
kaf24@8459 1011 mod[0].mod_end-mod[0].mod_start,
kaf24@8459 1012 _initrd_start,
kaf24@8459 1013 _initrd_len,
kaf24@8459 1014 cmdline) != 0)
kaf24@8459 1015 panic("Could not set up DOM0 guest OS\n");
kaf24@8459 1016
kaf24@8459 1017 /* Scrub RAM that is still free and so may go to an unprivileged domain. */
kaf24@8459 1018 scrub_heap_pages();
kaf24@8459 1019
kaf24@8459 1020 init_trace_bufs();
kaf24@8459 1021
kaf24@10502 1022 console_endboot();
kaf24@8459 1023
kaf24@8459 1024 /* Hide UART from DOM0 if we're using it */
kaf24@8459 1025 serial_endboot();
kaf24@8459 1026
kaf24@8459 1027 domain_unpause_by_systemcontroller(dom0);
kaf24@8459 1028
keir@15082 1029 reset_stack_and_jump(init_done);
kaf24@8459 1030 }
kaf24@8459 1031
ian@13763 1032 void arch_get_xen_caps(xen_capabilities_info_t *info)
iap10@6721 1033 {
kfraser@14997 1034 /* Interface name is always xen-3.0-* for Xen-3.x. */
kfraser@14997 1035 int major = 3, minor = 0;
keir@13754 1036 char s[32];
keir@13754 1037
ian@13763 1038 (*info)[0] = '\0';
iap10@6721 1039
kaf24@6725 1040 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
kaf24@6725 1041
keir@13754 1042 snprintf(s, sizeof(s), "xen-%d.%d-x86_32 ", major, minor);
ian@13763 1043 safe_strcat(*info, s);
keir@13754 1044 if ( hvm_enabled )
keir@13754 1045 {
keir@13754 1046 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1047 safe_strcat(*info, s);
kfraser@13685 1048 }
kaf24@6725 1049
kaf24@6725 1050 #elif defined(CONFIG_X86_32) && defined(CONFIG_X86_PAE)
kaf24@6725 1051
keir@13754 1052 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1053 safe_strcat(*info, s);
kaf24@6725 1054 if ( hvm_enabled )
iap10@6721 1055 {
keir@13754 1056 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1057 safe_strcat(*info, s);
keir@13754 1058 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1059 safe_strcat(*info, s);
iap10@6721 1060 }
iap10@6721 1061
kaf24@6725 1062 #elif defined(CONFIG_X86_64)
iap10@6721 1063
keir@13754 1064 snprintf(s, sizeof(s), "xen-%d.%d-x86_64 ", major, minor);
ian@13763 1065 safe_strcat(*info, s);
ack@13288 1066 #ifdef CONFIG_COMPAT
keir@13754 1067 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1068 safe_strcat(*info, s);
ack@13288 1069 #endif
kaf24@6725 1070 if ( hvm_enabled )
iap10@6721 1071 {
keir@13754 1072 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1073 safe_strcat(*info, s);
keir@13754 1074 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1075 safe_strcat(*info, s);
keir@13754 1076 snprintf(s, sizeof(s), "hvm-%d.%d-x86_64 ", major, minor);
ian@13763 1077 safe_strcat(*info, s);
iap10@6721 1078 }
kaf24@6725 1079
iap10@6721 1080 #endif
iap10@6721 1081 }
iap10@6721 1082
kaf24@3914 1083 /*
kaf24@3914 1084 * Local variables:
kaf24@3914 1085 * mode: C
kaf24@3914 1086 * c-set-style: "BSD"
kaf24@3914 1087 * c-basic-offset: 4
kaf24@3914 1088 * tab-width: 4
kaf24@3914 1089 * indent-tabs-mode: nil
kaf24@3988 1090 * End:
kaf24@3914 1091 */