ia64/xen-unstable

annotate xen/arch/x86/setup.c @ 19055:845aa241e163

x86_64: Remove statically-partitioned Xen heap.
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
author Keir Fraser <keir.fraser@citrix.com>
date Fri Jan 16 15:12:12 2009 +0000 (2009-01-16)
parents d77f66f89c04
children 681af1946724
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@5146 57 /* opt_nosmp: If true, secondary processors are ignored. */
kaf24@5900 58 static int opt_nosmp = 0;
kaf24@5146 59 boolean_param("nosmp", opt_nosmp);
kaf24@5146 60
kaf24@5146 61 /* maxcpus: maximum number of CPUs to activate. */
kaf24@5146 62 static unsigned int max_cpus = NR_CPUS;
shand@11156 63 integer_param("maxcpus", max_cpus);
kaf24@5146 64
kaf24@3334 65 /* opt_watchdog: If true, run a watchdog NMI on each processor. */
kaf24@3334 66 static int opt_watchdog = 0;
kaf24@3334 67 boolean_param("watchdog", opt_watchdog);
kaf24@3334 68
kaf24@4850 69 /* **** Linux config option: propagated to domain0. */
kaf24@4850 70 /* "acpi=off": Sisables both ACPI table parsing and interpreter. */
kaf24@4850 71 /* "acpi=force": Override the disable blacklist. */
kaf24@4850 72 /* "acpi=strict": Disables out-of-spec workarounds. */
kaf24@4850 73 /* "acpi=ht": Limit ACPI just to boot-time to enable HT. */
kaf24@4850 74 /* "acpi=noirq": Disables ACPI interrupt routing. */
kaf24@4850 75 static void parse_acpi_param(char *s);
kaf24@4850 76 custom_param("acpi", parse_acpi_param);
kaf24@4850 77
kaf24@4850 78 /* **** Linux config option: propagated to domain0. */
kaf24@4850 79 /* acpi_skip_timer_override: Skip IRQ0 overrides. */
kaf24@4850 80 extern int acpi_skip_timer_override;
kaf24@4850 81 boolean_param("acpi_skip_timer_override", acpi_skip_timer_override);
kaf24@4850 82
kaf24@4850 83 /* **** Linux config option: propagated to domain0. */
kaf24@4850 84 /* noapic: Disable IOAPIC setup. */
kaf24@4850 85 extern int skip_ioapic_setup;
kaf24@4850 86 boolean_param("noapic", skip_ioapic_setup);
kaf24@4850 87
keir@17546 88 /* **** Linux config option: propagated to domain0. */
keir@17657 89 /* xen_cpuidle: xen control cstate. */
keir@18051 90 /*static*/ int xen_cpuidle;
keir@17657 91 boolean_param("cpuidle", xen_cpuidle);
keir@17546 92
kaf24@3594 93 int early_boot = 1;
kaf24@3594 94
kaf24@5146 95 cpumask_t cpu_present_map;
kaf24@5146 96
kfraser@15074 97 unsigned long xen_phys_start;
kfraser@15074 98
keir@19055 99 #ifdef CONFIG_X86_32
kaf24@5003 100 /* Limits of Xen heap, used to initialise the allocator. */
kaf24@5003 101 unsigned long xenheap_phys_start, xenheap_phys_end;
keir@19055 102 #endif
kaf24@3338 103
kaf24@2298 104 extern void arch_init_memory(void);
kaf24@1589 105 extern void init_IRQ(void);
kaf24@5604 106 extern void early_time_init(void);
kaf24@5167 107 extern void early_cpu_init(void);
kfraser@15747 108 extern void vesa_init(void);
kfraser@15747 109 extern void vesa_mtrr_init(void);
kaf24@1589 110
keir@18523 111 DEFINE_PER_CPU(struct desc_struct *, gdt_table) = boot_cpu_gdt_table;
keir@18523 112 #ifdef CONFIG_COMPAT
keir@18523 113 DEFINE_PER_CPU(struct desc_struct *, compat_gdt_table)
keir@18523 114 = boot_cpu_compat_gdt_table;
keir@18523 115 #endif
keir@18523 116
kaf24@8533 117 struct tss_struct init_tss[NR_CPUS];
kaf24@8533 118
kfraser@15490 119 char __attribute__ ((__section__(".bss.stack_aligned"))) cpu0_stack[STACK_SIZE];
kaf24@5011 120
keir@16144 121 struct cpuinfo_x86 boot_cpu_data = { 0, 0, 0, 0, -1 };
kaf24@1452 122
kaf24@1670 123 unsigned long mmu_cr4_features = X86_CR4_PSE | X86_CR4_PGE | X86_CR4_PAE;
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;
keir@17551 243 u32 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;
keir@16888 261 start = max_t(unsigned long, start & ~mask, 16UL << 20);
kfraser@15740 262 end = (end + mask) & ~mask;
keir@16888 263 if ( start >= end )
keir@16888 264 return;
kfraser@15740 265 if ( end > BOOTSTRAP_DIRECTMAP_END )
kfraser@15740 266 panic("Cannot access memory beyond end of "
kfraser@15740 267 "bootstrap direct-map area\n");
kfraser@15740 268 map_pages_to_xen(
kfraser@15740 269 (unsigned long)maddr_to_bootstrap_virt(start),
kfraser@15740 270 start >> PAGE_SHIFT, (end-start) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kfraser@15740 271 }
kfraser@15740 272
kfraser@15074 273 static void __init move_memory(
kfraser@15074 274 unsigned long dst, unsigned long src_start, unsigned long src_end)
ian@12677 275 {
kfraser@15740 276 bootstrap_map(src_start, src_end);
kfraser@15740 277 bootstrap_map(dst, dst + src_end - src_start);
kfraser@15074 278 memmove(maddr_to_bootstrap_virt(dst),
kfraser@15074 279 maddr_to_bootstrap_virt(src_start),
ian@12677 280 src_end - src_start);
kfraser@15074 281 }
kfraser@15074 282
kfraser@15074 283 /* A temporary copy of the e820 map that we can mess with during bootstrap. */
kfraser@15074 284 static struct e820map __initdata boot_e820;
kfraser@15074 285
keir@15298 286 struct boot_video_info {
keir@15298 287 u8 orig_x; /* 0x00 */
keir@15298 288 u8 orig_y; /* 0x01 */
keir@15298 289 u8 orig_video_mode; /* 0x02 */
keir@15298 290 u8 orig_video_cols; /* 0x03 */
keir@15298 291 u8 orig_video_lines; /* 0x04 */
keir@15298 292 u8 orig_video_isVGA; /* 0x05 */
keir@15298 293 u16 orig_video_points; /* 0x06 */
keir@15298 294
keir@15298 295 /* VESA graphic mode -- linear frame buffer */
keir@15298 296 u32 capabilities; /* 0x08 */
keir@15298 297 u16 lfb_linelength; /* 0x0c */
keir@15298 298 u16 lfb_width; /* 0x0e */
keir@15298 299 u16 lfb_height; /* 0x10 */
keir@15298 300 u16 lfb_depth; /* 0x12 */
keir@15298 301 u32 lfb_base; /* 0x14 */
keir@15298 302 u32 lfb_size; /* 0x18 */
keir@15298 303 u8 red_size; /* 0x1c */
keir@15298 304 u8 red_pos; /* 0x1d */
keir@15298 305 u8 green_size; /* 0x1e */
keir@15298 306 u8 green_pos; /* 0x1f */
keir@15298 307 u8 blue_size; /* 0x20 */
keir@15298 308 u8 blue_pos; /* 0x21 */
keir@15298 309 u8 rsvd_size; /* 0x22 */
keir@15298 310 u8 rsvd_pos; /* 0x23 */
keir@15298 311 u16 vesapm_seg; /* 0x24 */
keir@15298 312 u16 vesapm_off; /* 0x26 */
keir@16124 313 u16 vesa_attrib; /* 0x28 */
keir@15298 314 };
keir@15298 315
keir@15298 316 static void __init parse_video_info(void)
keir@15298 317 {
keir@15298 318 struct boot_video_info *bvi = &bootsym(boot_vid_info);
keir@15298 319
keir@15298 320 if ( (bvi->orig_video_isVGA == 1) && (bvi->orig_video_mode == 3) )
keir@15298 321 {
keir@15298 322 vga_console_info.video_type = XEN_VGATYPE_TEXT_MODE_3;
keir@15298 323 vga_console_info.u.text_mode_3.font_height = bvi->orig_video_points;
keir@15298 324 vga_console_info.u.text_mode_3.cursor_x = bvi->orig_x;
keir@15298 325 vga_console_info.u.text_mode_3.cursor_y = bvi->orig_y;
keir@15298 326 vga_console_info.u.text_mode_3.rows = bvi->orig_video_lines;
keir@15298 327 vga_console_info.u.text_mode_3.columns = bvi->orig_video_cols;
keir@15298 328 }
keir@15298 329 else if ( bvi->orig_video_isVGA == 0x23 )
keir@15298 330 {
keir@15298 331 vga_console_info.video_type = XEN_VGATYPE_VESA_LFB;
keir@15298 332 vga_console_info.u.vesa_lfb.width = bvi->lfb_width;
keir@15298 333 vga_console_info.u.vesa_lfb.height = bvi->lfb_height;
keir@15298 334 vga_console_info.u.vesa_lfb.bytes_per_line = bvi->lfb_linelength;
keir@15298 335 vga_console_info.u.vesa_lfb.bits_per_pixel = bvi->lfb_depth;
keir@15298 336 vga_console_info.u.vesa_lfb.lfb_base = bvi->lfb_base;
keir@15298 337 vga_console_info.u.vesa_lfb.lfb_size = bvi->lfb_size;
keir@15298 338 vga_console_info.u.vesa_lfb.red_pos = bvi->red_pos;
keir@15298 339 vga_console_info.u.vesa_lfb.red_size = bvi->red_size;
keir@15298 340 vga_console_info.u.vesa_lfb.green_pos = bvi->green_pos;
keir@15298 341 vga_console_info.u.vesa_lfb.green_size = bvi->green_size;
keir@15298 342 vga_console_info.u.vesa_lfb.blue_pos = bvi->blue_pos;
keir@15298 343 vga_console_info.u.vesa_lfb.blue_size = bvi->blue_size;
keir@15298 344 vga_console_info.u.vesa_lfb.rsvd_pos = bvi->rsvd_pos;
keir@15298 345 vga_console_info.u.vesa_lfb.rsvd_size = bvi->rsvd_size;
keir@16124 346 vga_console_info.u.vesa_lfb.gbl_caps = bvi->capabilities;
keir@16124 347 vga_console_info.u.vesa_lfb.mode_attrs = bvi->vesa_attrib;
keir@15298 348 }
keir@15298 349 }
keir@15298 350
keir@16563 351 void __init kexec_reserve_area(struct e820map *e820)
keir@16563 352 {
keir@16563 353 unsigned long kdump_start = kexec_crash_area.start;
keir@16563 354 unsigned long kdump_size = kexec_crash_area.size;
keir@16563 355 static int is_reserved = 0;
keir@16563 356
keir@16563 357 kdump_size = (kdump_size + PAGE_SIZE - 1) & PAGE_MASK;
keir@16563 358
keir@16563 359 if ( (kdump_start == 0) || (kdump_size == 0) || is_reserved )
keir@16563 360 return;
keir@16563 361
keir@16563 362 is_reserved = 1;
keir@16563 363
keir@17674 364 if ( !reserve_e820_ram(e820, kdump_start, kdump_start + kdump_size) )
keir@16563 365 {
keir@16563 366 printk("Kdump: DISABLED (failed to reserve %luMB (%lukB) at 0x%lx)"
keir@16563 367 "\n", kdump_size >> 20, kdump_size >> 10, kdump_start);
keir@16563 368 kexec_crash_area.start = kexec_crash_area.size = 0;
keir@16563 369 }
keir@16563 370 else
keir@16563 371 {
keir@16563 372 printk("Kdump: %luMB (%lukB) at 0x%lx\n",
keir@16563 373 kdump_size >> 20, kdump_size >> 10, kdump_start);
keir@16563 374 }
keir@16563 375 }
keir@16563 376
keir@15082 377 void init_done(void)
keir@15082 378 {
keir@15082 379 extern char __init_begin[], __init_end[];
keir@15082 380
keir@15082 381 /* Free (or page-protect) the init areas. */
keir@18988 382 memset(__init_begin, 0xcc, __init_end - __init_begin); /* int3 poison */
keir@15082 383 #ifndef MEMORY_GUARD
keir@15082 384 init_xenheap_pages(__pa(__init_begin), __pa(__init_end));
keir@15082 385 #endif
keir@15082 386 memguard_guard_range(__init_begin, __init_end - __init_begin);
keir@15082 387 #if defined(CONFIG_X86_64)
keir@15082 388 /* Also zap the mapping in the 1:1 area. */
keir@15082 389 memguard_guard_range(__va(__pa(__init_begin)), __init_end - __init_begin);
keir@15082 390 #endif
keir@15082 391 printk("Freed %ldkB init memory.\n", (long)(__init_end-__init_begin)>>10);
keir@15082 392
keir@15082 393 startup_cpu_idle_loop();
keir@15082 394 }
keir@15082 395
kfraser@15796 396 static char * __init cmdline_cook(char *p)
kfraser@15796 397 {
kfraser@15796 398 p = p ? : "";
kfraser@15796 399 while ( *p == ' ' )
kfraser@15796 400 p++;
kfraser@15796 401 while ( (*p != ' ') && (*p != '\0') )
kfraser@15796 402 p++;
kfraser@15796 403 while ( *p == ' ' )
kfraser@15796 404 p++;
kfraser@15796 405 return p;
kfraser@15796 406 }
kfraser@15796 407
kfraser@15379 408 void __init __start_xen(unsigned long mbi_p)
kaf24@1452 409 {
kfraser@15293 410 char *memmap_type = NULL;
kfraser@15796 411 char *cmdline, *kextra;
kaf24@8457 412 unsigned long _initrd_start = 0, _initrd_len = 0;
kaf24@8457 413 unsigned int initrdidx = 1;
kfraser@15379 414 multiboot_info_t *mbi = __va(mbi_p);
kaf24@8457 415 module_t *mod = (module_t *)__va(mbi->mods_addr);
kaf24@6111 416 unsigned long nr_pages, modules_length;
keir@19055 417 unsigned long allocator_bitmap_end;
kfraser@15293 418 int i, e820_warn = 0, bytes = 0;
kaf24@5776 419 struct ns16550_defaults ns16550 = {
kaf24@5776 420 .data_bits = 8,
kaf24@5776 421 .parity = 'n',
kaf24@5776 422 .stop_bits = 1
kaf24@5776 423 };
kaf24@3338 424
kfraser@12853 425 extern void early_page_fault(void);
kfraser@12853 426 set_intr_gate(TRAP_page_fault, &early_page_fault);
kfraser@12853 427
kaf24@3338 428 /* Parse the command-line options. */
kfraser@15796 429 cmdline = cmdline_cook((mbi->flags & MBI_CMDLINE) ?
kfraser@15796 430 __va(mbi->cmdline) : NULL);
kfraser@15426 431 if ( (kextra = strstr(cmdline, " -- ")) != NULL )
kfraser@15426 432 {
kfraser@15426 433 /*
kfraser@15426 434 * Options after ' -- ' separator belong to dom0.
kfraser@15426 435 * 1. Orphan dom0's options from Xen's command line.
kfraser@15426 436 * 2. Skip all but final leading space from dom0's options.
kfraser@15426 437 */
kfraser@15426 438 *kextra = '\0';
kfraser@15426 439 kextra += 3;
kfraser@15426 440 while ( kextra[1] == ' ' ) kextra++;
kfraser@15426 441 }
kaf24@9823 442 cmdline_parse(cmdline);
kaf24@3338 443
keir@15298 444 parse_video_info();
keir@15298 445
kaf24@8534 446 set_current((struct vcpu *)0xfffff000); /* debug sanity */
keir@18790 447 idle_vcpu[0] = current;
kaf24@8534 448 set_processor_id(0); /* needed early, for smp_processor_id() */
keir@16378 449 if ( cpu_has_efer )
keir@16378 450 rdmsrl(MSR_EFER, this_cpu(efer));
keir@16267 451 asm volatile ( "mov %%cr4,%0" : "=r" (this_cpu(cr4)) );
kaf24@3338 452
kaf24@5146 453 smp_prepare_boot_cpu();
kaf24@5146 454
kaf24@3338 455 /* We initialise the serial devices very early so we can get debugging. */
kaf24@5776 456 ns16550.io_base = 0x3f8;
kaf24@5776 457 ns16550.irq = 4;
kaf24@5776 458 ns16550_init(0, &ns16550);
kaf24@5776 459 ns16550.io_base = 0x2f8;
kaf24@5776 460 ns16550.irq = 3;
kaf24@5776 461 ns16550_init(1, &ns16550);
kaf24@5195 462 serial_init_preirq();
kaf24@3338 463
kaf24@3338 464 init_console();
kaf24@3338 465
kfraser@11947 466 printk("Command line: %s\n", cmdline);
kaf24@9823 467
kfraser@15330 468 printk("Video information:\n");
kfraser@15330 469
kfraser@15330 470 /* Print VGA display mode information. */
keir@15298 471 switch ( vga_console_info.video_type )
keir@15298 472 {
keir@15298 473 case XEN_VGATYPE_TEXT_MODE_3:
kfraser@15330 474 printk(" VGA is text mode %dx%d, font 8x%d\n",
keir@15298 475 vga_console_info.u.text_mode_3.columns,
keir@15298 476 vga_console_info.u.text_mode_3.rows,
keir@15298 477 vga_console_info.u.text_mode_3.font_height);
keir@15298 478 break;
keir@15298 479 case XEN_VGATYPE_VESA_LFB:
kfraser@15330 480 printk(" VGA is graphics mode %dx%d, %d bpp\n",
keir@15298 481 vga_console_info.u.vesa_lfb.width,
keir@15298 482 vga_console_info.u.vesa_lfb.height,
keir@15298 483 vga_console_info.u.vesa_lfb.bits_per_pixel);
keir@15298 484 break;
kfraser@15330 485 default:
kfraser@15330 486 printk(" No VGA detected\n");
kfraser@15330 487 break;
kfraser@15330 488 }
kfraser@15330 489
kfraser@15330 490 /* Print VBE/DDC EDID information. */
kfraser@15330 491 if ( bootsym(boot_edid_caps) != 0x1313 )
kfraser@15330 492 {
kfraser@15330 493 u16 caps = bootsym(boot_edid_caps);
kfraser@15330 494 printk(" VBE/DDC methods:%s%s%s; ",
kfraser@15330 495 (caps & 1) ? " V1" : "",
kfraser@15330 496 (caps & 2) ? " V2" : "",
kfraser@15330 497 !(caps & 3) ? " none" : "");
kfraser@15330 498 printk("EDID transfer time: %d seconds\n", caps >> 8);
kfraser@15330 499 if ( *(u32 *)bootsym(boot_edid_info) == 0x13131313 )
kfraser@15330 500 {
kfraser@15330 501 printk(" EDID info not retrieved because ");
kfraser@15330 502 if ( !(caps & 3) )
kfraser@15330 503 printk("no DDC retrieval method detected\n");
kfraser@15330 504 else if ( (caps >> 8) > 5 )
kfraser@15330 505 printk("takes longer than 5 seconds\n");
kfraser@15330 506 else
kfraser@15330 507 printk("of reasons unknown\n");
kfraser@15330 508 }
keir@15298 509 }
keir@15298 510
kfraser@15336 511 printk("Disc information:\n");
kfraser@15336 512 printk(" Found %d MBR signatures\n",
kfraser@15430 513 bootsym(boot_mbr_signature_nr));
kfraser@15336 514 printk(" Found %d EDD information structures\n",
kfraser@15336 515 bootsym(boot_edd_info_nr));
kfraser@15336 516
kaf24@3344 517 /* Check that we have at least one Multiboot module. */
kaf24@3344 518 if ( !(mbi->flags & MBI_MODULES) || (mbi->mods_count == 0) )
kfraser@15074 519 EARLY_FAIL("dom0 kernel not specified. "
kfraser@15074 520 "Check bootloader configuration.\n");
kaf24@5011 521
kaf24@5011 522 if ( ((unsigned long)cpu0_stack & (STACK_SIZE-1)) != 0 )
kfraser@15074 523 EARLY_FAIL("Misaligned CPU0 stack.\n");
kaf24@3338 524
kfraser@15293 525 if ( e820_raw_nr != 0 )
kfraser@15292 526 {
kfraser@15293 527 memmap_type = "Xen-e820";
kfraser@15292 528 }
kfraser@15293 529 else if ( bootsym(lowmem_kb) )
kfraser@15292 530 {
kfraser@15293 531 memmap_type = "Xen-e801";
kfraser@15292 532 e820_raw[0].addr = 0;
kfraser@15293 533 e820_raw[0].size = bootsym(lowmem_kb) << 10;
kfraser@15292 534 e820_raw[0].type = E820_RAM;
kfraser@15292 535 e820_raw[1].addr = 0x100000;
kfraser@15293 536 e820_raw[1].size = bootsym(highmem_kb) << 10;
kfraser@15292 537 e820_raw[1].type = E820_RAM;
kfraser@15292 538 e820_raw_nr = 2;
kfraser@15292 539 }
kfraser@15292 540 else if ( mbi->flags & MBI_MEMMAP )
kaf24@3344 541 {
kfraser@15293 542 memmap_type = "Multiboot-e820";
keir@15988 543 while ( (bytes < mbi->mmap_length) && (e820_raw_nr < E820MAX) )
kaf24@3344 544 {
kaf24@3344 545 memory_map_t *map = __va(mbi->mmap_addr + bytes);
kaf24@8402 546
kaf24@8402 547 /*
kaf24@8403 548 * This is a gross workaround for a BIOS bug. Some bootloaders do
kaf24@8402 549 * not write e820 map entries into pre-zeroed memory. This is
kaf24@8402 550 * okay if the BIOS fills in all fields of the map entry, but
kaf24@8402 551 * some broken BIOSes do not bother to write the high word of
kaf24@8402 552 * the length field if the length is smaller than 4GB. We
kaf24@8402 553 * detect and fix this by flagging sections below 4GB that
kaf24@8403 554 * appear to be larger than 4GB in size.
kaf24@8402 555 */
kaf24@8403 556 if ( (map->base_addr_high == 0) && (map->length_high != 0) )
kaf24@8402 557 {
kfraser@15292 558 if ( !e820_warn )
kfraser@15292 559 {
kfraser@15292 560 printk("WARNING: Buggy e820 map detected and fixed "
kfraser@15292 561 "(truncated length fields).\n");
kfraser@15292 562 e820_warn = 1;
kfraser@15292 563 }
kaf24@8402 564 map->length_high = 0;
kaf24@8402 565 }
kaf24@8402 566
kaf24@3344 567 e820_raw[e820_raw_nr].addr =
kaf24@3344 568 ((u64)map->base_addr_high << 32) | (u64)map->base_addr_low;
kaf24@3344 569 e820_raw[e820_raw_nr].size =
kaf24@3344 570 ((u64)map->length_high << 32) | (u64)map->length_low;
kfraser@15799 571 e820_raw[e820_raw_nr].type = map->type;
kaf24@3344 572 e820_raw_nr++;
kaf24@8402 573
kaf24@3344 574 bytes += map->size + 4;
kaf24@3344 575 }
kaf24@3344 576 }
kaf24@3344 577 else if ( mbi->flags & MBI_MEMLIMITS )
kaf24@3344 578 {
kfraser@15293 579 memmap_type = "Multiboot-e801";
kaf24@3344 580 e820_raw[0].addr = 0;
kaf24@3344 581 e820_raw[0].size = mbi->mem_lower << 10;
kaf24@3344 582 e820_raw[0].type = E820_RAM;
kaf24@3354 583 e820_raw[1].addr = 0x100000;
kaf24@3354 584 e820_raw[1].size = mbi->mem_upper << 10;
kaf24@3354 585 e820_raw[1].type = E820_RAM;
kaf24@3344 586 e820_raw_nr = 2;
kaf24@3344 587 }
kaf24@3344 588 else
kaf24@3344 589 {
kfraser@15074 590 EARLY_FAIL("Bootloader provided no memory information.\n");
kaf24@3344 591 }
kaf24@3344 592
kaf24@13427 593 /* Sanitise the raw E820 map to produce a final clean version. */
kfraser@15293 594 max_page = init_e820(memmap_type, e820_raw, &e820_raw_nr);
kaf24@3338 595
keir@16563 596 /* Create a temporary copy of the E820 map. */
kfraser@15074 597 memcpy(&boot_e820, &e820, sizeof(e820));
keir@16563 598
keir@16563 599 /* Early kexec reservation (explicit static start address). */
keir@16563 600 kexec_reserve_area(&boot_e820);
kaf24@6111 601
kfraser@15074 602 /*
keir@15077 603 * Iterate backwards over all superpage-aligned RAM regions.
kfraser@15074 604 *
kfraser@15074 605 * We require superpage alignment because the boot allocator is not yet
kfraser@15074 606 * initialised. Hence we can only map superpages in the address range
kfraser@15074 607 * 0 to BOOTSTRAP_DIRECTMAP_END, as this is guaranteed not to require
kfraser@15074 608 * dynamic allocation of pagetables.
kfraser@15074 609 *
kfraser@15074 610 * As well as mapping superpages in that range, in preparation for
kfraser@15074 611 * initialising the boot allocator, we also look for a region to which
kfraser@15074 612 * we can relocate the dom0 kernel and other multiboot modules. Also, on
kfraser@15074 613 * x86/64, we relocate Xen to higher memory.
kfraser@15074 614 */
kfraser@15074 615 modules_length = mod[mbi->mods_count-1].mod_end - mod[0].mod_start;
keir@15077 616 for ( i = boot_e820.nr_map-1; i >= 0; i-- )
kfraser@15074 617 {
kfraser@15074 618 uint64_t s, e, mask = (1UL << L2_PAGETABLE_SHIFT) - 1;
kaf24@6134 619
keir@16563 620 /* Superpage-aligned chunks from 16MB to BOOTSTRAP_DIRECTMAP_END. */
kfraser@15074 621 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 622 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
keir@16563 623 s = max_t(uint64_t, s, 16 << 20);
kfraser@15074 624 e = min_t(uint64_t, e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 625 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kaf24@3354 626 continue;
kaf24@6111 627
kfraser@15074 628 /* Map the chunk. No memory will need to be allocated to do this. */
kfraser@15074 629 map_pages_to_xen(
kfraser@15074 630 (unsigned long)maddr_to_bootstrap_virt(s),
kfraser@15074 631 s >> PAGE_SHIFT, (e-s) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kaf24@6111 632
kfraser@14084 633 #if defined(CONFIG_X86_64)
keir@19055 634 #define reloc_size ((__pa(&_end) + mask) & ~mask)
kfraser@15074 635 /* Is the region suitable for relocating Xen? */
keir@19055 636 if ( !xen_phys_start && ((e-s) >= reloc_size) )
kaf24@5003 637 {
kfraser@15074 638 extern l2_pgentry_t l2_xenmap[];
kfraser@15074 639 l4_pgentry_t *pl4e;
kfraser@15074 640 l3_pgentry_t *pl3e;
kfraser@15074 641 l2_pgentry_t *pl2e;
keir@16888 642 int i, j, k;
kfraser@15074 643
kfraser@15074 644 /* Select relocation address. */
keir@19055 645 e -= reloc_size;
kfraser@15074 646 xen_phys_start = e;
kfraser@15292 647 bootsym(trampoline_xen_phys_start) = e;
kfraser@15074 648
kfraser@15074 649 /*
kfraser@15074 650 * Perform relocation to new physical address.
kfraser@15074 651 * Before doing so we must sync static/global data with main memory
kfraser@15074 652 * with a barrier(). After this we must *not* modify static/global
kfraser@15074 653 * data until after we have switched to the relocated pagetables!
kfraser@15074 654 */
kfraser@15074 655 barrier();
kfraser@15074 656 move_memory(e, 0, __pa(&_end) - xen_phys_start);
kfraser@15074 657
kfraser@15379 658 /* Poison low 1MB to detect stray pointers to physical 0-1MB. */
kfraser@15379 659 memset(maddr_to_bootstrap_virt(e), 0x55, 1U<<20);
kfraser@15379 660
kfraser@15074 661 /* Walk initial pagetables, relocating page directory entries. */
kfraser@15074 662 pl4e = __va(__pa(idle_pg_table));
kfraser@15074 663 for ( i = 0 ; i < L4_PAGETABLE_ENTRIES; i++, pl4e++ )
kfraser@15074 664 {
kfraser@15074 665 if ( !(l4e_get_flags(*pl4e) & _PAGE_PRESENT) )
kfraser@15074 666 continue;
kfraser@15074 667 *pl4e = l4e_from_intpte(l4e_get_intpte(*pl4e) +
kfraser@15074 668 xen_phys_start);
kfraser@15074 669 pl3e = l4e_to_l3e(*pl4e);
kfraser@15074 670 for ( j = 0; j < L3_PAGETABLE_ENTRIES; j++, pl3e++ )
kfraser@15074 671 {
keir@16921 672 /* Not present, 1GB mapping, or already relocated? */
kfraser@15074 673 if ( !(l3e_get_flags(*pl3e) & _PAGE_PRESENT) ||
keir@16921 674 (l3e_get_flags(*pl3e) & _PAGE_PSE) ||
kfraser@15074 675 (l3e_get_pfn(*pl3e) > 0x1000) )
kfraser@15074 676 continue;
kfraser@15074 677 *pl3e = l3e_from_intpte(l3e_get_intpte(*pl3e) +
kfraser@15074 678 xen_phys_start);
keir@16888 679 pl2e = l3e_to_l2e(*pl3e);
keir@16888 680 for ( k = 0; k < L2_PAGETABLE_ENTRIES; k++, pl2e++ )
keir@16888 681 {
keir@16888 682 /* Not present, PSE, or already relocated? */
keir@16888 683 if ( !(l2e_get_flags(*pl2e) & _PAGE_PRESENT) ||
keir@16888 684 (l2e_get_flags(*pl2e) & _PAGE_PSE) ||
keir@16888 685 (l2e_get_pfn(*pl2e) > 0x1000) )
keir@16888 686 continue;
keir@16888 687 *pl2e = l2e_from_intpte(l2e_get_intpte(*pl2e) +
keir@16888 688 xen_phys_start);
keir@16888 689 }
kfraser@15074 690 }
kfraser@15074 691 }
kfraser@15074 692
kfraser@15074 693 /* The only data mappings to be relocated are in the Xen area. */
kfraser@15074 694 pl2e = __va(__pa(l2_xenmap));
keir@16888 695 *pl2e++ = l2e_from_pfn(xen_phys_start >> PAGE_SHIFT,
keir@16888 696 PAGE_HYPERVISOR | _PAGE_PSE);
keir@16888 697 for ( i = 1; i < L2_PAGETABLE_ENTRIES; i++, pl2e++ )
kfraser@15074 698 {
kfraser@15074 699 if ( !(l2e_get_flags(*pl2e) & _PAGE_PRESENT) )
kfraser@15074 700 continue;
kfraser@15074 701 *pl2e = l2e_from_intpte(l2e_get_intpte(*pl2e) +
kfraser@15074 702 xen_phys_start);
kfraser@15074 703 }
kfraser@15074 704
kfraser@15074 705 /* Re-sync the stack and then switch to relocated pagetables. */
kfraser@15074 706 asm volatile (
kfraser@15074 707 "rep movsb ; " /* re-sync the stack */
kfraser@15074 708 "movq %%cr4,%%rsi ; "
kfraser@15074 709 "andb $0x7f,%%sil ; "
kfraser@15074 710 "movq %%rsi,%%cr4 ; " /* CR4.PGE == 0 */
kfraser@15074 711 "movq %0,%%cr3 ; " /* CR3 == new pagetables */
kfraser@15074 712 "orb $0x80,%%sil ; "
kfraser@15074 713 "movq %%rsi,%%cr4 " /* CR4.PGE == 1 */
kfraser@15074 714 : : "r" (__pa(idle_pg_table)), "S" (cpu0_stack),
kfraser@15074 715 "D" (__va(__pa(cpu0_stack))), "c" (STACK_SIZE) : "memory" );
kaf24@5003 716 }
kaf24@5003 717 #endif
keir@15077 718
keir@15077 719 /* Is the region suitable for relocating the multiboot modules? */
keir@15077 720 if ( !initial_images_start && (s < e) && ((e-s) >= modules_length) )
keir@15077 721 {
kfraser@15489 722 initial_images_end = e;
kfraser@15489 723 e = (e - modules_length) & PAGE_MASK;
keir@15077 724 initial_images_start = e;
keir@15077 725 move_memory(initial_images_start,
keir@15077 726 mod[0].mod_start, mod[mbi->mods_count-1].mod_end);
keir@15077 727 }
kfraser@15489 728
kfraser@15489 729 if ( !kexec_crash_area.start && (s < e) &&
kfraser@15489 730 ((e-s) >= kexec_crash_area.size) )
kfraser@15489 731 {
kfraser@15489 732 e = (e - kexec_crash_area.size) & PAGE_MASK;
kfraser@15489 733 kexec_crash_area.start = e;
kfraser@15489 734 }
kaf24@3354 735 }
kaf24@3354 736
kfraser@15074 737 if ( !initial_images_start )
kfraser@15074 738 EARLY_FAIL("Not enough memory to relocate the dom0 kernel image.\n");
keir@15988 739 reserve_e820_ram(&boot_e820, initial_images_start, initial_images_end);
kfraser@15074 740
keir@19055 741 /* Initialise boot heap. */
keir@19055 742 allocator_bitmap_end = init_boot_allocator(__pa(&_end));
keir@19055 743 #if defined(CONFIG_X86_32)
keir@19055 744 xenheap_phys_start = allocator_bitmap_end;
keir@19055 745 xenheap_phys_end = DIRECTMAP_MBYTES << 20;
keir@19055 746 #else
kfraser@15074 747 if ( !xen_phys_start )
kfraser@15074 748 EARLY_FAIL("Not enough memory to relocate Xen.\n");
keir@19055 749 reserve_e820_ram(&boot_e820, __pa(&_start), allocator_bitmap_end);
kfraser@15074 750 #endif
kfraser@15074 751
keir@16563 752 /* Late kexec reservation (dynamic start address). */
keir@16563 753 kexec_reserve_area(&boot_e820);
kfraser@15489 754
kfraser@15074 755 /*
keir@16624 756 * With the boot allocator now initialised, we can walk every RAM region
keir@16624 757 * and 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 {
keir@16563 762 uint64_t s, e, map_s, 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;
keir@16563 767 #if defined(CONFIG_X86_32)
keir@16563 768 s = max_t(uint64_t, s, xenheap_phys_end);
keir@16563 769 #else
keir@16563 770 s = max_t(uint64_t, s, 1<<20);
keir@16563 771 #endif
kfraser@15074 772 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kfraser@15074 773 continue;
kfraser@15074 774
keir@16563 775 /* Need to create mappings above 16MB. */
keir@16563 776 map_s = max_t(uint64_t, s, 16<<20);
kfraser@15074 777 map_e = e;
keir@16563 778 #if defined(CONFIG_X86_32) /* mappings are truncated on x86_32 */
kfraser@15074 779 map_e = min_t(uint64_t, map_e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 780 #endif
keir@16563 781
keir@16563 782 /* Pass mapped memory to allocator /before/ creating new mappings. */
keir@16624 783 init_boot_pages(s, min_t(uint64_t, map_s, e));
keir@16563 784
keir@16563 785 /* Create new mappings /before/ passing memory to the allocator. */
keir@16563 786 if ( map_s < map_e )
kfraser@15074 787 map_pages_to_xen(
keir@16563 788 (unsigned long)maddr_to_bootstrap_virt(map_s),
keir@16563 789 map_s >> PAGE_SHIFT, (map_e-map_s) >> PAGE_SHIFT,
keir@16563 790 PAGE_HYPERVISOR);
kfraser@15074 791
keir@16563 792 /* Pass remainder of this memory chunk to the allocator. */
keir@16624 793 init_boot_pages(map_s, e);
kfraser@15074 794 }
kfraser@15074 795
kaf24@5003 796 memguard_init();
kaf24@4950 797
kfraser@15074 798 nr_pages = 0;
kfraser@15074 799 for ( i = 0; i < e820.nr_map; i++ )
kfraser@15074 800 if ( e820.map[i].type == E820_RAM )
kfraser@15074 801 nr_pages += e820.map[i].size >> PAGE_SHIFT;
ian@12681 802 printk("System RAM: %luMB (%lukB)\n",
kaf24@3354 803 nr_pages >> (20 - PAGE_SHIFT),
kaf24@3354 804 nr_pages << (PAGE_SHIFT - 10));
kaf24@7220 805 total_pages = nr_pages;
kaf24@3354 806
kfraser@11296 807 /* Sanity check for unwanted bloat of certain hypercall structures. */
kfraser@11296 808 BUILD_BUG_ON(sizeof(((struct xen_platform_op *)0)->u) !=
kfraser@11296 809 sizeof(((struct xen_platform_op *)0)->u.pad));
kfraser@11296 810 BUILD_BUG_ON(sizeof(((struct xen_domctl *)0)->u) !=
kfraser@11296 811 sizeof(((struct xen_domctl *)0)->u.pad));
kfraser@11296 812 BUILD_BUG_ON(sizeof(((struct xen_sysctl *)0)->u) !=
kfraser@11296 813 sizeof(((struct xen_sysctl *)0)->u.pad));
kaf24@7388 814
kaf24@9878 815 BUILD_BUG_ON(sizeof(start_info_t) > PAGE_SIZE);
kaf24@9878 816 BUILD_BUG_ON(sizeof(shared_info_t) > PAGE_SIZE);
ack@13292 817 BUILD_BUG_ON(sizeof(struct vcpu_info) != 64);
kaf24@7744 818
ack@13291 819 #ifdef CONFIG_COMPAT
ack@13291 820 BUILD_BUG_ON(sizeof(((struct compat_platform_op *)0)->u) !=
ack@13291 821 sizeof(((struct compat_platform_op *)0)->u.pad));
ack@13291 822 BUILD_BUG_ON(sizeof(start_info_compat_t) > PAGE_SIZE);
ack@13292 823 BUILD_BUG_ON(sizeof(struct compat_vcpu_info) != 64);
ack@13291 824 #endif
ack@13291 825
kfraser@10492 826 /* Check definitions in public headers match internal defs. */
kaf24@9878 827 BUILD_BUG_ON(__HYPERVISOR_VIRT_START != HYPERVISOR_VIRT_START);
kaf24@8521 828 #ifdef HYPERVISOR_VIRT_END
kaf24@9878 829 BUILD_BUG_ON(__HYPERVISOR_VIRT_END != HYPERVISOR_VIRT_END);
kaf24@8521 830 #endif
kfraser@10492 831 BUILD_BUG_ON(MACH2PHYS_VIRT_START != RO_MPT_VIRT_START);
kfraser@10492 832 BUILD_BUG_ON(MACH2PHYS_VIRT_END != RO_MPT_VIRT_END);
kaf24@8521 833
kaf24@3354 834 init_frametable();
kaf24@3338 835
kfraser@11971 836 acpi_boot_table_init();
kfraser@11971 837
kfraser@11971 838 acpi_numa_init();
kfraser@11971 839
kfraser@11971 840 numa_initmem_init(0, max_page);
kfraser@11971 841
keir@19055 842 #if defined(CONFIG_X86_32)
keir@19055 843 /* Initialise the Xen heap. */
kfraser@15074 844 init_xenheap_pages(xenheap_phys_start, xenheap_phys_end);
kfraser@15074 845 nr_pages = (xenheap_phys_end - xenheap_phys_start) >> PAGE_SHIFT;
kfraser@15074 846 xenheap_phys_start = xen_phys_start;
kaf24@6111 847 printk("Xen heap: %luMB (%lukB)\n",
kaf24@6111 848 nr_pages >> (20 - PAGE_SHIFT),
kaf24@6111 849 nr_pages << (PAGE_SHIFT - 10));
keir@19055 850 #endif
kaf24@3338 851
keir@14680 852 end_boot_allocator();
keir@19055 853 early_boot = 0;
keir@14680 854
keir@19055 855 #if defined(CONFIG_X86_64)
keir@19055 856 vesa_init();
keir@19055 857 #endif
kaf24@3338 858
keir@17444 859 softirq_init();
keir@17444 860
kaf24@8459 861 early_cpu_init();
kaf24@8459 862
kaf24@8459 863 paging_init();
kaf24@8459 864
keir@16274 865 tboot_probe();
keir@16274 866
kaf24@8459 867 /* Unmap the first page of CPU0's stack. */
kaf24@8459 868 memguard_guard_stack(cpu0_stack);
kaf24@8459 869
kaf24@8459 870 open_softirq(NEW_TLBFLUSH_CLOCK_PERIOD_SOFTIRQ, new_tlbflush_clock_period);
kaf24@8459 871
kaf24@8459 872 if ( opt_watchdog )
kaf24@8459 873 nmi_watchdog = NMI_LOCAL_APIC;
kaf24@8459 874
kaf24@8459 875 sort_exception_tables();
kaf24@8459 876
kaf24@8459 877 find_smp_config();
kaf24@8459 878
kaf24@8459 879 dmi_scan_machine();
kaf24@8459 880
kaf24@8459 881 generic_apic_probe();
kaf24@8459 882
keir@17552 883 if ( x2apic_is_available() )
keir@17552 884 enable_x2apic();
keir@17552 885
kaf24@8459 886 acpi_boot_init();
kaf24@8459 887
kfraser@11971 888 init_cpu_to_node();
kfraser@11971 889
kfraser@11241 890 if ( smp_found_config )
kaf24@8459 891 get_smp_config();
kaf24@8459 892
keir@15083 893 #ifdef CONFIG_X86_64
keir@15083 894 /* Low mappings were only needed for some BIOS table parsing. */
keir@15083 895 zap_low_mappings();
keir@15083 896 #endif
keir@15083 897
kaf24@8459 898 init_apic_mappings();
kaf24@8459 899
kaf24@8459 900 init_IRQ();
kaf24@8459 901
kfraser@11241 902 percpu_init_areas();
kfraser@11241 903
kfraser@15815 904 xsm_init(&initrdidx, mbi, initial_images_start);
kfraser@15815 905
kfraser@11240 906 init_idle_domain();
kfraser@11240 907
kaf24@8459 908 trap_init();
kaf24@8459 909
kaf24@13662 910 rcu_init();
kaf24@13662 911
kaf24@8586 912 timer_init();
kaf24@8459 913
kaf24@8459 914 early_time_init();
kaf24@8459 915
kaf24@8459 916 arch_init_memory();
kaf24@8459 917
kaf24@8459 918 identify_cpu(&boot_cpu_data);
kaf24@8459 919 if ( cpu_has_fxsr )
kaf24@8459 920 set_in_cr4(X86_CR4_OSFXSR);
kaf24@8459 921 if ( cpu_has_xmm )
kaf24@8459 922 set_in_cr4(X86_CR4_OSXMMEXCPT);
keir@18920 923
keir@18920 924 local_irq_enable();
keir@18920 925
kfraser@15747 926 #ifdef CONFIG_X86_64
kfraser@15747 927 vesa_mtrr_init();
kfraser@15747 928 #endif
kaf24@8459 929
kaf24@8459 930 if ( opt_nosmp )
kaf24@8459 931 max_cpus = 0;
kaf24@8459 932
kaf24@8459 933 smp_prepare_cpus(max_cpus);
kaf24@8459 934
keir@18920 935 spin_debug_enable();
keir@18920 936
kaf24@8459 937 /*
kaf24@8459 938 * Initialise higher-level timer functions. We do this fairly late
kaf24@8459 939 * (post-SMP) because the time bases and scale factors need to be updated
kaf24@8459 940 * regularly, and SMP initialisation can cause a long delay with
kaf24@8459 941 * interrupts not yet enabled.
kaf24@8459 942 */
kaf24@8459 943 init_xen_time();
kaf24@8459 944
kaf24@8459 945 initialize_keytable();
kaf24@8459 946
kaf24@8459 947 serial_init_postirq();
kaf24@8459 948
kaf24@8459 949 for_each_present_cpu ( i )
kaf24@8459 950 {
kaf24@8459 951 if ( num_online_cpus() >= max_cpus )
kaf24@8459 952 break;
kaf24@8459 953 if ( !cpu_online(i) )
kaf24@13662 954 {
kaf24@13662 955 rcu_online_cpu(i);
kaf24@8459 956 __cpu_up(i);
kaf24@13662 957 }
kfraser@11971 958
kfraser@11998 959 /* Set up cpu_to_node[]. */
kfraser@11971 960 srat_detect_node(i);
kfraser@11998 961 /* Set up node_to_cpumask based on cpu_to_node[]. */
kfraser@11971 962 numa_add_cpu(i);
kaf24@8459 963 }
kaf24@8459 964
kaf24@8459 965 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
kaf24@8459 966 smp_cpus_done(max_cpus);
kaf24@8459 967
kaf24@9117 968 initialise_gdb(); /* could be moved earlier */
kaf24@9117 969
kaf24@8459 970 do_initcalls();
kaf24@8459 971
kaf24@8594 972 if ( opt_watchdog )
kaf24@8594 973 watchdog_enable();
kaf24@8459 974
kaf24@8459 975 /* Create initial domain 0. */
kfraser@14911 976 dom0 = domain_create(0, 0, DOM0_SSIDREF);
kfraser@10655 977 if ( (dom0 == NULL) || (alloc_vcpu(dom0, 0, 0) == NULL) )
kaf24@8459 978 panic("Error creating domain 0\n");
kaf24@8459 979
kfraser@12210 980 dom0->is_privileged = 1;
keir@16856 981 dom0->target = NULL;
kfraser@12210 982
kaf24@8459 983 /* Grab the DOM0 command line. */
kaf24@8459 984 cmdline = (char *)(mod[0].string ? __va(mod[0].string) : NULL);
kfraser@15426 985 if ( (cmdline != NULL) || (kextra != NULL) )
kaf24@8459 986 {
kaf24@8459 987 static char dom0_cmdline[MAX_GUEST_CMDLINE];
kaf24@8459 988
kfraser@15796 989 cmdline = cmdline_cook(cmdline);
kfraser@15796 990 safe_strcpy(dom0_cmdline, cmdline);
kaf24@8459 991
kfraser@15426 992 if ( kextra != NULL )
kfraser@15426 993 /* kextra always includes exactly one leading space. */
kfraser@15426 994 safe_strcat(dom0_cmdline, kextra);
kfraser@15426 995
kaf24@8459 996 /* Append any extra parameters. */
kfraser@13691 997 if ( skip_ioapic_setup && !strstr(dom0_cmdline, "noapic") )
kfraser@13691 998 safe_strcat(dom0_cmdline, " noapic");
kaf24@8459 999 if ( acpi_skip_timer_override &&
kfraser@13691 1000 !strstr(dom0_cmdline, "acpi_skip_timer_override") )
kfraser@13691 1001 safe_strcat(dom0_cmdline, " acpi_skip_timer_override");
keir@16165 1002 if ( (strlen(acpi_param) == 0) && acpi_disabled )
keir@16165 1003 {
keir@16165 1004 printk("ACPI is disabled, notifying Domain 0 (acpi=off)\n");
keir@16165 1005 safe_strcpy(acpi_param, "off");
keir@16165 1006 }
kfraser@13691 1007 if ( (strlen(acpi_param) != 0) && !strstr(dom0_cmdline, "acpi=") )
kaf24@8459 1008 {
kfraser@13691 1009 safe_strcat(dom0_cmdline, " acpi=");
kfraser@13691 1010 safe_strcat(dom0_cmdline, acpi_param);
kaf24@8459 1011 }
kfraser@13691 1012
kfraser@13691 1013 cmdline = dom0_cmdline;
kaf24@8459 1014 }
kaf24@8459 1015
kaf24@8459 1016 if ( (initrdidx > 0) && (initrdidx < mbi->mods_count) )
kaf24@8459 1017 {
kaf24@8459 1018 _initrd_start = initial_images_start +
kaf24@8459 1019 (mod[initrdidx].mod_start - mod[0].mod_start);
kaf24@8459 1020 _initrd_len = mod[initrdidx].mod_end - mod[initrdidx].mod_start;
kaf24@8459 1021 }
kaf24@8459 1022
keir@18180 1023 if ( xen_cpuidle )
keir@18180 1024 xen_processor_pmbits |= XEN_PROCESSOR_PM_CX;
keir@18180 1025
kaf24@8459 1026 /*
kaf24@8459 1027 * We're going to setup domain0 using the module(s) that we stashed safely
kaf24@8459 1028 * above our heap. The second module, if present, is an initrd ramdisk.
kaf24@8459 1029 */
kaf24@8459 1030 if ( construct_dom0(dom0,
kaf24@8459 1031 initial_images_start,
kaf24@8459 1032 mod[0].mod_end-mod[0].mod_start,
kaf24@8459 1033 _initrd_start,
kaf24@8459 1034 _initrd_len,
kaf24@8459 1035 cmdline) != 0)
kaf24@8459 1036 panic("Could not set up DOM0 guest OS\n");
kaf24@8459 1037
kaf24@8459 1038 /* Scrub RAM that is still free and so may go to an unprivileged domain. */
kaf24@8459 1039 scrub_heap_pages();
kaf24@8459 1040
kaf24@8459 1041 init_trace_bufs();
kaf24@8459 1042
kaf24@10502 1043 console_endboot();
kaf24@8459 1044
kaf24@8459 1045 /* Hide UART from DOM0 if we're using it */
kaf24@8459 1046 serial_endboot();
kaf24@8459 1047
kaf24@8459 1048 domain_unpause_by_systemcontroller(dom0);
kaf24@8459 1049
keir@15082 1050 reset_stack_and_jump(init_done);
kaf24@8459 1051 }
kaf24@8459 1052
ian@13763 1053 void arch_get_xen_caps(xen_capabilities_info_t *info)
iap10@6721 1054 {
kfraser@14997 1055 /* Interface name is always xen-3.0-* for Xen-3.x. */
kfraser@14997 1056 int major = 3, minor = 0;
keir@13754 1057 char s[32];
keir@13754 1058
ian@13763 1059 (*info)[0] = '\0';
iap10@6721 1060
keir@17618 1061 #if defined(CONFIG_X86_32)
kaf24@6725 1062
keir@13754 1063 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1064 safe_strcat(*info, s);
kaf24@6725 1065 if ( hvm_enabled )
iap10@6721 1066 {
keir@13754 1067 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1068 safe_strcat(*info, s);
keir@13754 1069 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1070 safe_strcat(*info, s);
iap10@6721 1071 }
iap10@6721 1072
kaf24@6725 1073 #elif defined(CONFIG_X86_64)
iap10@6721 1074
keir@13754 1075 snprintf(s, sizeof(s), "xen-%d.%d-x86_64 ", major, minor);
ian@13763 1076 safe_strcat(*info, s);
ack@13288 1077 #ifdef CONFIG_COMPAT
keir@13754 1078 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1079 safe_strcat(*info, s);
ack@13288 1080 #endif
kaf24@6725 1081 if ( hvm_enabled )
iap10@6721 1082 {
keir@13754 1083 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1084 safe_strcat(*info, s);
keir@13754 1085 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1086 safe_strcat(*info, s);
keir@13754 1087 snprintf(s, sizeof(s), "hvm-%d.%d-x86_64 ", major, minor);
ian@13763 1088 safe_strcat(*info, s);
iap10@6721 1089 }
kaf24@6725 1090
iap10@6721 1091 #endif
iap10@6721 1092 }
iap10@6721 1093
keir@17738 1094 int xen_in_range(paddr_t start, paddr_t end)
keir@17729 1095 {
keir@19055 1096 #if defined(CONFIG_X86_32)
keir@19055 1097 paddr_t xs = xenheap_phys_start;
keir@19055 1098 paddr_t xe = xenheap_phys_end;
keir@19055 1099 #else
keir@19055 1100 paddr_t xs = __pa(&_start);
keir@19055 1101 paddr_t xe = __pa(&_end);
keir@19055 1102 #endif
keir@19055 1103
keir@19055 1104 return (start < xe) && (end > xs);
keir@17729 1105 }
keir@17729 1106
kaf24@3914 1107 /*
kaf24@3914 1108 * Local variables:
kaf24@3914 1109 * mode: C
kaf24@3914 1110 * c-set-style: "BSD"
kaf24@3914 1111 * c-basic-offset: 4
kaf24@3914 1112 * tab-width: 4
kaf24@3914 1113 * indent-tabs-mode: nil
kaf24@3988 1114 * End:
kaf24@3914 1115 */