ia64/xen-unstable

annotate xen/arch/x86/setup.c @ 16888:76601c290fa9

x86: First 1MB of memory should be mapped with 4kB mappings to avoid
conflict with fixed-range MTRRs. While there, we now map the VGA hole
as uncacheable.
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
author Keir Fraser <keir.fraser@citrix.com>
date Fri Jan 25 16:02:20 2008 +0000 (2008-01-25)
parents cff4c8a1aa28
children ed8ab1a36b09
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;
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@16563 364 if ( !reserve_e820_ram(e820, 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@15082 382 #ifndef MEMORY_GUARD
keir@15082 383 init_xenheap_pages(__pa(__init_begin), __pa(__init_end));
keir@15082 384 #endif
keir@15082 385 memguard_guard_range(__init_begin, __init_end - __init_begin);
keir@15082 386 #if defined(CONFIG_X86_64)
keir@15082 387 /* Also zap the mapping in the 1:1 area. */
keir@15082 388 memguard_guard_range(__va(__pa(__init_begin)), __init_end - __init_begin);
keir@15082 389 #endif
keir@15082 390 printk("Freed %ldkB init memory.\n", (long)(__init_end-__init_begin)>>10);
keir@15082 391
keir@15082 392 startup_cpu_idle_loop();
keir@15082 393 }
keir@15082 394
kfraser@15796 395 static char * __init cmdline_cook(char *p)
kfraser@15796 396 {
kfraser@15796 397 p = p ? : "";
kfraser@15796 398 while ( *p == ' ' )
kfraser@15796 399 p++;
kfraser@15796 400 while ( (*p != ' ') && (*p != '\0') )
kfraser@15796 401 p++;
kfraser@15796 402 while ( *p == ' ' )
kfraser@15796 403 p++;
kfraser@15796 404 return p;
kfraser@15796 405 }
kfraser@15796 406
kfraser@15379 407 void __init __start_xen(unsigned long mbi_p)
kaf24@1452 408 {
kfraser@15293 409 char *memmap_type = NULL;
kfraser@15796 410 char *cmdline, *kextra;
kaf24@8457 411 unsigned long _initrd_start = 0, _initrd_len = 0;
kaf24@8457 412 unsigned int initrdidx = 1;
kfraser@15379 413 multiboot_info_t *mbi = __va(mbi_p);
kaf24@8457 414 module_t *mod = (module_t *)__va(mbi->mods_addr);
kaf24@6111 415 unsigned long nr_pages, modules_length;
kfraser@15293 416 int i, e820_warn = 0, bytes = 0;
kaf24@5776 417 struct ns16550_defaults ns16550 = {
kaf24@5776 418 .data_bits = 8,
kaf24@5776 419 .parity = 'n',
kaf24@5776 420 .stop_bits = 1
kaf24@5776 421 };
kaf24@3338 422
kfraser@12853 423 extern void early_page_fault(void);
kfraser@12853 424 set_intr_gate(TRAP_page_fault, &early_page_fault);
kfraser@12853 425
kaf24@3338 426 /* Parse the command-line options. */
kfraser@15796 427 cmdline = cmdline_cook((mbi->flags & MBI_CMDLINE) ?
kfraser@15796 428 __va(mbi->cmdline) : NULL);
kfraser@15426 429 if ( (kextra = strstr(cmdline, " -- ")) != NULL )
kfraser@15426 430 {
kfraser@15426 431 /*
kfraser@15426 432 * Options after ' -- ' separator belong to dom0.
kfraser@15426 433 * 1. Orphan dom0's options from Xen's command line.
kfraser@15426 434 * 2. Skip all but final leading space from dom0's options.
kfraser@15426 435 */
kfraser@15426 436 *kextra = '\0';
kfraser@15426 437 kextra += 3;
kfraser@15426 438 while ( kextra[1] == ' ' ) kextra++;
kfraser@15426 439 }
kaf24@9823 440 cmdline_parse(cmdline);
kaf24@3338 441
keir@15298 442 parse_video_info();
keir@15298 443
kaf24@8534 444 set_current((struct vcpu *)0xfffff000); /* debug sanity */
kfraser@11240 445 idle_vcpu[0] = current;
kaf24@8534 446 set_processor_id(0); /* needed early, for smp_processor_id() */
keir@16378 447 if ( cpu_has_efer )
keir@16378 448 rdmsrl(MSR_EFER, this_cpu(efer));
keir@16267 449 asm volatile ( "mov %%cr4,%0" : "=r" (this_cpu(cr4)) );
kaf24@3338 450
kaf24@5146 451 smp_prepare_boot_cpu();
kaf24@5146 452
kaf24@3338 453 /* We initialise the serial devices very early so we can get debugging. */
kaf24@5776 454 ns16550.io_base = 0x3f8;
kaf24@5776 455 ns16550.irq = 4;
kaf24@5776 456 ns16550_init(0, &ns16550);
kaf24@5776 457 ns16550.io_base = 0x2f8;
kaf24@5776 458 ns16550.irq = 3;
kaf24@5776 459 ns16550_init(1, &ns16550);
kaf24@5195 460 serial_init_preirq();
kaf24@3338 461
kaf24@3338 462 init_console();
kaf24@3338 463
kfraser@11947 464 printk("Command line: %s\n", cmdline);
kaf24@9823 465
kfraser@15330 466 printk("Video information:\n");
kfraser@15330 467
kfraser@15330 468 /* Print VGA display mode information. */
keir@15298 469 switch ( vga_console_info.video_type )
keir@15298 470 {
keir@15298 471 case XEN_VGATYPE_TEXT_MODE_3:
kfraser@15330 472 printk(" VGA is text mode %dx%d, font 8x%d\n",
keir@15298 473 vga_console_info.u.text_mode_3.columns,
keir@15298 474 vga_console_info.u.text_mode_3.rows,
keir@15298 475 vga_console_info.u.text_mode_3.font_height);
keir@15298 476 break;
keir@15298 477 case XEN_VGATYPE_VESA_LFB:
kfraser@15330 478 printk(" VGA is graphics mode %dx%d, %d bpp\n",
keir@15298 479 vga_console_info.u.vesa_lfb.width,
keir@15298 480 vga_console_info.u.vesa_lfb.height,
keir@15298 481 vga_console_info.u.vesa_lfb.bits_per_pixel);
keir@15298 482 break;
kfraser@15330 483 default:
kfraser@15330 484 printk(" No VGA detected\n");
kfraser@15330 485 break;
kfraser@15330 486 }
kfraser@15330 487
kfraser@15330 488 /* Print VBE/DDC EDID information. */
kfraser@15330 489 if ( bootsym(boot_edid_caps) != 0x1313 )
kfraser@15330 490 {
kfraser@15330 491 u16 caps = bootsym(boot_edid_caps);
kfraser@15330 492 printk(" VBE/DDC methods:%s%s%s; ",
kfraser@15330 493 (caps & 1) ? " V1" : "",
kfraser@15330 494 (caps & 2) ? " V2" : "",
kfraser@15330 495 !(caps & 3) ? " none" : "");
kfraser@15330 496 printk("EDID transfer time: %d seconds\n", caps >> 8);
kfraser@15330 497 if ( *(u32 *)bootsym(boot_edid_info) == 0x13131313 )
kfraser@15330 498 {
kfraser@15330 499 printk(" EDID info not retrieved because ");
kfraser@15330 500 if ( !(caps & 3) )
kfraser@15330 501 printk("no DDC retrieval method detected\n");
kfraser@15330 502 else if ( (caps >> 8) > 5 )
kfraser@15330 503 printk("takes longer than 5 seconds\n");
kfraser@15330 504 else
kfraser@15330 505 printk("of reasons unknown\n");
kfraser@15330 506 }
keir@15298 507 }
keir@15298 508
kfraser@15336 509 printk("Disc information:\n");
kfraser@15336 510 printk(" Found %d MBR signatures\n",
kfraser@15430 511 bootsym(boot_mbr_signature_nr));
kfraser@15336 512 printk(" Found %d EDD information structures\n",
kfraser@15336 513 bootsym(boot_edd_info_nr));
kfraser@15336 514
kaf24@3344 515 /* Check that we have at least one Multiboot module. */
kaf24@3344 516 if ( !(mbi->flags & MBI_MODULES) || (mbi->mods_count == 0) )
kfraser@15074 517 EARLY_FAIL("dom0 kernel not specified. "
kfraser@15074 518 "Check bootloader configuration.\n");
kaf24@5011 519
kaf24@5011 520 if ( ((unsigned long)cpu0_stack & (STACK_SIZE-1)) != 0 )
kfraser@15074 521 EARLY_FAIL("Misaligned CPU0 stack.\n");
kaf24@3338 522
kfraser@11618 523 /*
kfraser@11618 524 * Since there are some stubs getting built on the stacks which use
kfraser@11618 525 * direct calls/jumps, the heap must be confined to the lower 2G so
kfraser@11618 526 * that those branches can reach their targets.
kfraser@11618 527 */
kfraser@11618 528 if ( opt_xenheap_megabytes > 2048 )
kfraser@11618 529 opt_xenheap_megabytes = 2048;
kaf24@3338 530
kfraser@15293 531 if ( e820_raw_nr != 0 )
kfraser@15292 532 {
kfraser@15293 533 memmap_type = "Xen-e820";
kfraser@15292 534 }
kfraser@15293 535 else if ( bootsym(lowmem_kb) )
kfraser@15292 536 {
kfraser@15293 537 memmap_type = "Xen-e801";
kfraser@15292 538 e820_raw[0].addr = 0;
kfraser@15293 539 e820_raw[0].size = bootsym(lowmem_kb) << 10;
kfraser@15292 540 e820_raw[0].type = E820_RAM;
kfraser@15292 541 e820_raw[1].addr = 0x100000;
kfraser@15293 542 e820_raw[1].size = bootsym(highmem_kb) << 10;
kfraser@15292 543 e820_raw[1].type = E820_RAM;
kfraser@15292 544 e820_raw_nr = 2;
kfraser@15292 545 }
kfraser@15292 546 else if ( mbi->flags & MBI_MEMMAP )
kaf24@3344 547 {
kfraser@15293 548 memmap_type = "Multiboot-e820";
keir@15988 549 while ( (bytes < mbi->mmap_length) && (e820_raw_nr < E820MAX) )
kaf24@3344 550 {
kaf24@3344 551 memory_map_t *map = __va(mbi->mmap_addr + bytes);
kaf24@8402 552
kaf24@8402 553 /*
kaf24@8403 554 * This is a gross workaround for a BIOS bug. Some bootloaders do
kaf24@8402 555 * not write e820 map entries into pre-zeroed memory. This is
kaf24@8402 556 * okay if the BIOS fills in all fields of the map entry, but
kaf24@8402 557 * some broken BIOSes do not bother to write the high word of
kaf24@8402 558 * the length field if the length is smaller than 4GB. We
kaf24@8402 559 * detect and fix this by flagging sections below 4GB that
kaf24@8403 560 * appear to be larger than 4GB in size.
kaf24@8402 561 */
kaf24@8403 562 if ( (map->base_addr_high == 0) && (map->length_high != 0) )
kaf24@8402 563 {
kfraser@15292 564 if ( !e820_warn )
kfraser@15292 565 {
kfraser@15292 566 printk("WARNING: Buggy e820 map detected and fixed "
kfraser@15292 567 "(truncated length fields).\n");
kfraser@15292 568 e820_warn = 1;
kfraser@15292 569 }
kaf24@8402 570 map->length_high = 0;
kaf24@8402 571 }
kaf24@8402 572
kaf24@3344 573 e820_raw[e820_raw_nr].addr =
kaf24@3344 574 ((u64)map->base_addr_high << 32) | (u64)map->base_addr_low;
kaf24@3344 575 e820_raw[e820_raw_nr].size =
kaf24@3344 576 ((u64)map->length_high << 32) | (u64)map->length_low;
kfraser@15799 577 e820_raw[e820_raw_nr].type = map->type;
kaf24@3344 578 e820_raw_nr++;
kaf24@8402 579
kaf24@3344 580 bytes += map->size + 4;
kaf24@3344 581 }
kaf24@3344 582 }
kaf24@3344 583 else if ( mbi->flags & MBI_MEMLIMITS )
kaf24@3344 584 {
kfraser@15293 585 memmap_type = "Multiboot-e801";
kaf24@3344 586 e820_raw[0].addr = 0;
kaf24@3344 587 e820_raw[0].size = mbi->mem_lower << 10;
kaf24@3344 588 e820_raw[0].type = E820_RAM;
kaf24@3354 589 e820_raw[1].addr = 0x100000;
kaf24@3354 590 e820_raw[1].size = mbi->mem_upper << 10;
kaf24@3354 591 e820_raw[1].type = E820_RAM;
kaf24@3344 592 e820_raw_nr = 2;
kaf24@3344 593 }
kaf24@3344 594 else
kaf24@3344 595 {
kfraser@15074 596 EARLY_FAIL("Bootloader provided no memory information.\n");
kaf24@3344 597 }
kaf24@3344 598
kaf24@13427 599 /* Sanitise the raw E820 map to produce a final clean version. */
kfraser@15293 600 max_page = init_e820(memmap_type, e820_raw, &e820_raw_nr);
kaf24@3338 601
keir@16563 602 /* Create a temporary copy of the E820 map. */
kfraser@15074 603 memcpy(&boot_e820, &e820, sizeof(e820));
keir@16563 604
keir@16563 605 /* Early kexec reservation (explicit static start address). */
keir@16563 606 kexec_reserve_area(&boot_e820);
kaf24@6111 607
kfraser@15074 608 /*
keir@15077 609 * Iterate backwards over all superpage-aligned RAM regions.
kfraser@15074 610 *
kfraser@15074 611 * We require superpage alignment because the boot allocator is not yet
kfraser@15074 612 * initialised. Hence we can only map superpages in the address range
kfraser@15074 613 * 0 to BOOTSTRAP_DIRECTMAP_END, as this is guaranteed not to require
kfraser@15074 614 * dynamic allocation of pagetables.
kfraser@15074 615 *
kfraser@15074 616 * As well as mapping superpages in that range, in preparation for
kfraser@15074 617 * initialising the boot allocator, we also look for a region to which
kfraser@15074 618 * we can relocate the dom0 kernel and other multiboot modules. Also, on
kfraser@15074 619 * x86/64, we relocate Xen to higher memory.
kfraser@15074 620 */
kfraser@15074 621 modules_length = mod[mbi->mods_count-1].mod_end - mod[0].mod_start;
keir@15077 622 for ( i = boot_e820.nr_map-1; i >= 0; i-- )
kfraser@15074 623 {
kfraser@15074 624 uint64_t s, e, mask = (1UL << L2_PAGETABLE_SHIFT) - 1;
kaf24@6134 625
keir@16563 626 /* Superpage-aligned chunks from 16MB to BOOTSTRAP_DIRECTMAP_END. */
kfraser@15074 627 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 628 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
keir@16563 629 s = max_t(uint64_t, s, 16 << 20);
kfraser@15074 630 e = min_t(uint64_t, e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 631 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kaf24@3354 632 continue;
kaf24@6111 633
kfraser@15074 634 /* Map the chunk. No memory will need to be allocated to do this. */
kfraser@15074 635 map_pages_to_xen(
kfraser@15074 636 (unsigned long)maddr_to_bootstrap_virt(s),
kfraser@15074 637 s >> PAGE_SHIFT, (e-s) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kaf24@6111 638
kfraser@14084 639 #if defined(CONFIG_X86_64)
kfraser@15074 640 /* Is the region suitable for relocating Xen? */
kfraser@15074 641 if ( !xen_phys_start && (((e-s) >> 20) >= opt_xenheap_megabytes) )
kaf24@5003 642 {
kfraser@15074 643 extern l2_pgentry_t l2_xenmap[];
kfraser@15074 644 l4_pgentry_t *pl4e;
kfraser@15074 645 l3_pgentry_t *pl3e;
kfraser@15074 646 l2_pgentry_t *pl2e;
keir@16888 647 int i, j, k;
kfraser@15074 648
kfraser@15074 649 /* Select relocation address. */
kfraser@15074 650 e = (e - (opt_xenheap_megabytes << 20)) & ~mask;
kfraser@15074 651 xen_phys_start = e;
kfraser@15292 652 bootsym(trampoline_xen_phys_start) = e;
kfraser@15074 653
kfraser@15074 654 /*
kfraser@15074 655 * Perform relocation to new physical address.
kfraser@15074 656 * Before doing so we must sync static/global data with main memory
kfraser@15074 657 * with a barrier(). After this we must *not* modify static/global
kfraser@15074 658 * data until after we have switched to the relocated pagetables!
kfraser@15074 659 */
kfraser@15074 660 barrier();
kfraser@15074 661 move_memory(e, 0, __pa(&_end) - xen_phys_start);
kfraser@15074 662
kfraser@15379 663 /* Poison low 1MB to detect stray pointers to physical 0-1MB. */
kfraser@15379 664 memset(maddr_to_bootstrap_virt(e), 0x55, 1U<<20);
kfraser@15379 665
kfraser@15074 666 /* Walk initial pagetables, relocating page directory entries. */
kfraser@15074 667 pl4e = __va(__pa(idle_pg_table));
kfraser@15074 668 for ( i = 0 ; i < L4_PAGETABLE_ENTRIES; i++, pl4e++ )
kfraser@15074 669 {
kfraser@15074 670 if ( !(l4e_get_flags(*pl4e) & _PAGE_PRESENT) )
kfraser@15074 671 continue;
kfraser@15074 672 *pl4e = l4e_from_intpte(l4e_get_intpte(*pl4e) +
kfraser@15074 673 xen_phys_start);
kfraser@15074 674 pl3e = l4e_to_l3e(*pl4e);
kfraser@15074 675 for ( j = 0; j < L3_PAGETABLE_ENTRIES; j++, pl3e++ )
kfraser@15074 676 {
kfraser@15074 677 /* Not present or already relocated? */
kfraser@15074 678 if ( !(l3e_get_flags(*pl3e) & _PAGE_PRESENT) ||
kfraser@15074 679 (l3e_get_pfn(*pl3e) > 0x1000) )
kfraser@15074 680 continue;
kfraser@15074 681 *pl3e = l3e_from_intpte(l3e_get_intpte(*pl3e) +
kfraser@15074 682 xen_phys_start);
keir@16888 683 pl2e = l3e_to_l2e(*pl3e);
keir@16888 684 for ( k = 0; k < L2_PAGETABLE_ENTRIES; k++, pl2e++ )
keir@16888 685 {
keir@16888 686 /* Not present, PSE, or already relocated? */
keir@16888 687 if ( !(l2e_get_flags(*pl2e) & _PAGE_PRESENT) ||
keir@16888 688 (l2e_get_flags(*pl2e) & _PAGE_PSE) ||
keir@16888 689 (l2e_get_pfn(*pl2e) > 0x1000) )
keir@16888 690 continue;
keir@16888 691 *pl2e = l2e_from_intpte(l2e_get_intpte(*pl2e) +
keir@16888 692 xen_phys_start);
keir@16888 693 }
kfraser@15074 694 }
kfraser@15074 695 }
kfraser@15074 696
kfraser@15074 697 /* The only data mappings to be relocated are in the Xen area. */
kfraser@15074 698 pl2e = __va(__pa(l2_xenmap));
keir@16888 699 *pl2e++ = l2e_from_pfn(xen_phys_start >> PAGE_SHIFT,
keir@16888 700 PAGE_HYPERVISOR | _PAGE_PSE);
keir@16888 701 for ( i = 1; i < L2_PAGETABLE_ENTRIES; i++, pl2e++ )
kfraser@15074 702 {
kfraser@15074 703 if ( !(l2e_get_flags(*pl2e) & _PAGE_PRESENT) )
kfraser@15074 704 continue;
kfraser@15074 705 *pl2e = l2e_from_intpte(l2e_get_intpte(*pl2e) +
kfraser@15074 706 xen_phys_start);
kfraser@15074 707 }
kfraser@15074 708
kfraser@15074 709 /* Re-sync the stack and then switch to relocated pagetables. */
kfraser@15074 710 asm volatile (
kfraser@15074 711 "rep movsb ; " /* re-sync the stack */
kfraser@15074 712 "movq %%cr4,%%rsi ; "
kfraser@15074 713 "andb $0x7f,%%sil ; "
kfraser@15074 714 "movq %%rsi,%%cr4 ; " /* CR4.PGE == 0 */
kfraser@15074 715 "movq %0,%%cr3 ; " /* CR3 == new pagetables */
kfraser@15074 716 "orb $0x80,%%sil ; "
kfraser@15074 717 "movq %%rsi,%%cr4 " /* CR4.PGE == 1 */
kfraser@15074 718 : : "r" (__pa(idle_pg_table)), "S" (cpu0_stack),
kfraser@15074 719 "D" (__va(__pa(cpu0_stack))), "c" (STACK_SIZE) : "memory" );
kaf24@5003 720 }
kaf24@5003 721 #endif
keir@15077 722
keir@15077 723 /* Is the region suitable for relocating the multiboot modules? */
keir@15077 724 if ( !initial_images_start && (s < e) && ((e-s) >= modules_length) )
keir@15077 725 {
kfraser@15489 726 initial_images_end = e;
kfraser@15489 727 e = (e - modules_length) & PAGE_MASK;
keir@15077 728 initial_images_start = e;
keir@15077 729 move_memory(initial_images_start,
keir@15077 730 mod[0].mod_start, mod[mbi->mods_count-1].mod_end);
keir@15077 731 }
kfraser@15489 732
kfraser@15489 733 if ( !kexec_crash_area.start && (s < e) &&
kfraser@15489 734 ((e-s) >= kexec_crash_area.size) )
kfraser@15489 735 {
kfraser@15489 736 e = (e - kexec_crash_area.size) & PAGE_MASK;
kfraser@15489 737 kexec_crash_area.start = e;
kfraser@15489 738 }
kaf24@3354 739 }
kaf24@3354 740
kfraser@15074 741 if ( !initial_images_start )
kfraser@15074 742 EARLY_FAIL("Not enough memory to relocate the dom0 kernel image.\n");
keir@15988 743 reserve_e820_ram(&boot_e820, initial_images_start, initial_images_end);
kfraser@15074 744
keir@16563 745 /* Initialise Xen heap and boot heap. */
kfraser@15074 746 xenheap_phys_start = init_boot_allocator(__pa(&_end));
kfraser@15074 747 xenheap_phys_end = opt_xenheap_megabytes << 20;
kfraser@15074 748 #if defined(CONFIG_X86_64)
kfraser@15074 749 if ( !xen_phys_start )
kfraser@15074 750 EARLY_FAIL("Not enough memory to relocate Xen.\n");
kfraser@15074 751 xenheap_phys_end += xen_phys_start;
keir@15988 752 reserve_e820_ram(&boot_e820, xen_phys_start,
keir@15988 753 xen_phys_start + (opt_xenheap_megabytes<<20));
kfraser@15074 754 #endif
kfraser@15074 755
keir@16563 756 /* Late kexec reservation (dynamic start address). */
keir@16563 757 kexec_reserve_area(&boot_e820);
kfraser@15489 758
kfraser@15074 759 /*
keir@16624 760 * With the boot allocator now initialised, we can walk every RAM region
keir@16624 761 * and map it in its entirety (on x86/64, at least) and notify it to the
kfraser@15074 762 * boot allocator.
kfraser@15074 763 */
kfraser@15074 764 for ( i = 0; i < boot_e820.nr_map; i++ )
kfraser@15074 765 {
keir@16563 766 uint64_t s, e, map_s, map_e, mask = PAGE_SIZE - 1;
kfraser@15074 767
kfraser@15074 768 /* Only page alignment required now. */
kfraser@15074 769 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 770 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
keir@16563 771 #if defined(CONFIG_X86_32)
keir@16563 772 s = max_t(uint64_t, s, xenheap_phys_end);
keir@16563 773 #else
keir@16563 774 s = max_t(uint64_t, s, 1<<20);
keir@16563 775 #endif
kfraser@15074 776 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kfraser@15074 777 continue;
kfraser@15074 778
keir@16563 779 /* Need to create mappings above 16MB. */
keir@16563 780 map_s = max_t(uint64_t, s, 16<<20);
kfraser@15074 781 map_e = e;
keir@16563 782 #if defined(CONFIG_X86_32) /* mappings are truncated on x86_32 */
kfraser@15074 783 map_e = min_t(uint64_t, map_e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 784 #endif
keir@16563 785
keir@16563 786 /* Pass mapped memory to allocator /before/ creating new mappings. */
keir@16624 787 init_boot_pages(s, min_t(uint64_t, map_s, e));
keir@16563 788
keir@16563 789 /* Create new mappings /before/ passing memory to the allocator. */
keir@16563 790 if ( map_s < map_e )
kfraser@15074 791 map_pages_to_xen(
keir@16563 792 (unsigned long)maddr_to_bootstrap_virt(map_s),
keir@16563 793 map_s >> PAGE_SHIFT, (map_e-map_s) >> PAGE_SHIFT,
keir@16563 794 PAGE_HYPERVISOR);
kfraser@15074 795
keir@16563 796 /* Pass remainder of this memory chunk to the allocator. */
keir@16624 797 init_boot_pages(map_s, e);
kfraser@15074 798 }
kfraser@15074 799
kaf24@5003 800 memguard_init();
kaf24@4950 801
kfraser@15074 802 nr_pages = 0;
kfraser@15074 803 for ( i = 0; i < e820.nr_map; i++ )
kfraser@15074 804 if ( e820.map[i].type == E820_RAM )
kfraser@15074 805 nr_pages += e820.map[i].size >> PAGE_SHIFT;
ian@12681 806 printk("System RAM: %luMB (%lukB)\n",
kaf24@3354 807 nr_pages >> (20 - PAGE_SHIFT),
kaf24@3354 808 nr_pages << (PAGE_SHIFT - 10));
kaf24@7220 809 total_pages = nr_pages;
kaf24@3354 810
kfraser@11296 811 /* Sanity check for unwanted bloat of certain hypercall structures. */
kfraser@11296 812 BUILD_BUG_ON(sizeof(((struct xen_platform_op *)0)->u) !=
kfraser@11296 813 sizeof(((struct xen_platform_op *)0)->u.pad));
kfraser@11296 814 BUILD_BUG_ON(sizeof(((struct xen_domctl *)0)->u) !=
kfraser@11296 815 sizeof(((struct xen_domctl *)0)->u.pad));
kfraser@11296 816 BUILD_BUG_ON(sizeof(((struct xen_sysctl *)0)->u) !=
kfraser@11296 817 sizeof(((struct xen_sysctl *)0)->u.pad));
kaf24@7388 818
kaf24@9878 819 BUILD_BUG_ON(sizeof(start_info_t) > PAGE_SIZE);
kaf24@9878 820 BUILD_BUG_ON(sizeof(shared_info_t) > PAGE_SIZE);
ack@13292 821 BUILD_BUG_ON(sizeof(struct vcpu_info) != 64);
kaf24@7744 822
ack@13291 823 #ifdef CONFIG_COMPAT
ack@13291 824 BUILD_BUG_ON(sizeof(((struct compat_platform_op *)0)->u) !=
ack@13291 825 sizeof(((struct compat_platform_op *)0)->u.pad));
ack@13291 826 BUILD_BUG_ON(sizeof(start_info_compat_t) > PAGE_SIZE);
ack@13292 827 BUILD_BUG_ON(sizeof(struct compat_vcpu_info) != 64);
ack@13291 828 #endif
ack@13291 829
kfraser@10492 830 /* Check definitions in public headers match internal defs. */
kaf24@9878 831 BUILD_BUG_ON(__HYPERVISOR_VIRT_START != HYPERVISOR_VIRT_START);
kaf24@8521 832 #ifdef HYPERVISOR_VIRT_END
kaf24@9878 833 BUILD_BUG_ON(__HYPERVISOR_VIRT_END != HYPERVISOR_VIRT_END);
kaf24@8521 834 #endif
kfraser@10492 835 BUILD_BUG_ON(MACH2PHYS_VIRT_START != RO_MPT_VIRT_START);
kfraser@10492 836 BUILD_BUG_ON(MACH2PHYS_VIRT_END != RO_MPT_VIRT_END);
kaf24@8521 837
kaf24@3354 838 init_frametable();
kaf24@3338 839
kfraser@11971 840 acpi_boot_table_init();
kfraser@11971 841
kfraser@11971 842 acpi_numa_init();
kfraser@11971 843
kfraser@11971 844 numa_initmem_init(0, max_page);
kfraser@11971 845
kaf24@6111 846 /* Initialise the Xen heap, skipping RAM holes. */
kfraser@15074 847 init_xenheap_pages(xenheap_phys_start, xenheap_phys_end);
kfraser@15074 848 nr_pages = (xenheap_phys_end - xenheap_phys_start) >> PAGE_SHIFT;
kfraser@15074 849 #ifdef __x86_64__
kfraser@15074 850 init_xenheap_pages(xen_phys_start, __pa(&_start));
kfraser@15074 851 nr_pages += (__pa(&_start) - xen_phys_start) >> PAGE_SHIFT;
kfraser@15747 852 vesa_init();
kfraser@15074 853 #endif
kfraser@15074 854 xenheap_phys_start = xen_phys_start;
kaf24@6111 855 printk("Xen heap: %luMB (%lukB)\n",
kaf24@6111 856 nr_pages >> (20 - PAGE_SHIFT),
kaf24@6111 857 nr_pages << (PAGE_SHIFT - 10));
kaf24@3338 858
keir@14680 859 end_boot_allocator();
keir@14680 860
kaf24@3594 861 early_boot = 0;
kaf24@3338 862
kaf24@8459 863 early_cpu_init();
kaf24@8459 864
kaf24@8459 865 paging_init();
kaf24@8459 866
keir@16274 867 tboot_probe();
keir@16274 868
kaf24@8459 869 /* Unmap the first page of CPU0's stack. */
kaf24@8459 870 memguard_guard_stack(cpu0_stack);
kaf24@8459 871
kaf24@8459 872 open_softirq(NEW_TLBFLUSH_CLOCK_PERIOD_SOFTIRQ, new_tlbflush_clock_period);
kaf24@8459 873
kaf24@8459 874 if ( opt_watchdog )
kaf24@8459 875 nmi_watchdog = NMI_LOCAL_APIC;
kaf24@8459 876
kaf24@8459 877 sort_exception_tables();
kaf24@8459 878
kaf24@8459 879 find_smp_config();
kaf24@8459 880
kaf24@8459 881 dmi_scan_machine();
kaf24@8459 882
kaf24@8459 883 generic_apic_probe();
kaf24@8459 884
kaf24@8459 885 acpi_boot_init();
kaf24@8459 886
kfraser@11971 887 init_cpu_to_node();
kfraser@11971 888
kfraser@11241 889 if ( smp_found_config )
kaf24@8459 890 get_smp_config();
kaf24@8459 891
keir@15083 892 #ifdef CONFIG_X86_64
keir@15083 893 /* Low mappings were only needed for some BIOS table parsing. */
keir@15083 894 zap_low_mappings();
keir@15083 895 #endif
keir@15083 896
kaf24@8459 897 init_apic_mappings();
kaf24@8459 898
kaf24@8459 899 init_IRQ();
kaf24@8459 900
kfraser@11241 901 percpu_init_areas();
kfraser@11241 902
kfraser@15815 903 xsm_init(&initrdidx, mbi, initial_images_start);
kfraser@15815 904
kfraser@11240 905 init_idle_domain();
kfraser@11240 906
kaf24@8459 907 trap_init();
kaf24@8459 908
kaf24@13662 909 rcu_init();
kaf24@13662 910
kaf24@8586 911 timer_init();
kaf24@8459 912
kaf24@8459 913 early_time_init();
kaf24@8459 914
kaf24@8459 915 arch_init_memory();
kaf24@8459 916
kaf24@8459 917 identify_cpu(&boot_cpu_data);
kaf24@8459 918 if ( cpu_has_fxsr )
kaf24@8459 919 set_in_cr4(X86_CR4_OSFXSR);
kaf24@8459 920 if ( cpu_has_xmm )
kaf24@8459 921 set_in_cr4(X86_CR4_OSXMMEXCPT);
kfraser@15747 922 #ifdef CONFIG_X86_64
kfraser@15747 923 vesa_mtrr_init();
kfraser@15747 924 #endif
kaf24@8459 925
kaf24@8459 926 if ( opt_nosmp )
kaf24@8459 927 max_cpus = 0;
kaf24@8459 928
kaf24@8459 929 smp_prepare_cpus(max_cpus);
kaf24@8459 930
kaf24@8459 931 /*
kaf24@8459 932 * Initialise higher-level timer functions. We do this fairly late
kaf24@8459 933 * (post-SMP) because the time bases and scale factors need to be updated
kaf24@8459 934 * regularly, and SMP initialisation can cause a long delay with
kaf24@8459 935 * interrupts not yet enabled.
kaf24@8459 936 */
kaf24@8459 937 init_xen_time();
kaf24@8459 938
kaf24@8459 939 initialize_keytable();
kaf24@8459 940
kaf24@8459 941 serial_init_postirq();
kaf24@8459 942
kaf24@8459 943 BUG_ON(!local_irq_is_enabled());
kaf24@8459 944
kaf24@8459 945 for_each_present_cpu ( i )
kaf24@8459 946 {
kaf24@8459 947 if ( num_online_cpus() >= max_cpus )
kaf24@8459 948 break;
kaf24@8459 949 if ( !cpu_online(i) )
kaf24@13662 950 {
kaf24@13662 951 rcu_online_cpu(i);
kaf24@8459 952 __cpu_up(i);
kaf24@13662 953 }
kfraser@11971 954
kfraser@11998 955 /* Set up cpu_to_node[]. */
kfraser@11971 956 srat_detect_node(i);
kfraser@11998 957 /* Set up node_to_cpumask based on cpu_to_node[]. */
kfraser@11971 958 numa_add_cpu(i);
kaf24@8459 959 }
kaf24@8459 960
kaf24@8459 961 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
kaf24@8459 962 smp_cpus_done(max_cpus);
kaf24@8459 963
kaf24@9117 964 initialise_gdb(); /* could be moved earlier */
kaf24@9117 965
kaf24@8459 966 do_initcalls();
kaf24@8459 967
kaf24@8594 968 if ( opt_watchdog )
kaf24@8594 969 watchdog_enable();
kaf24@8459 970
kaf24@8459 971 /* Create initial domain 0. */
kfraser@14911 972 dom0 = domain_create(0, 0, DOM0_SSIDREF);
kfraser@10655 973 if ( (dom0 == NULL) || (alloc_vcpu(dom0, 0, 0) == NULL) )
kaf24@8459 974 panic("Error creating domain 0\n");
kaf24@8459 975
kfraser@12210 976 dom0->is_privileged = 1;
keir@16856 977 dom0->target = NULL;
kfraser@12210 978
kaf24@8459 979 /* Grab the DOM0 command line. */
kaf24@8459 980 cmdline = (char *)(mod[0].string ? __va(mod[0].string) : NULL);
kfraser@15426 981 if ( (cmdline != NULL) || (kextra != NULL) )
kaf24@8459 982 {
kaf24@8459 983 static char dom0_cmdline[MAX_GUEST_CMDLINE];
kaf24@8459 984
kfraser@15796 985 cmdline = cmdline_cook(cmdline);
kfraser@15796 986 safe_strcpy(dom0_cmdline, cmdline);
kaf24@8459 987
kfraser@15426 988 if ( kextra != NULL )
kfraser@15426 989 /* kextra always includes exactly one leading space. */
kfraser@15426 990 safe_strcat(dom0_cmdline, kextra);
kfraser@15426 991
kaf24@8459 992 /* Append any extra parameters. */
kfraser@13691 993 if ( skip_ioapic_setup && !strstr(dom0_cmdline, "noapic") )
kfraser@13691 994 safe_strcat(dom0_cmdline, " noapic");
kaf24@8459 995 if ( acpi_skip_timer_override &&
kfraser@13691 996 !strstr(dom0_cmdline, "acpi_skip_timer_override") )
kfraser@13691 997 safe_strcat(dom0_cmdline, " acpi_skip_timer_override");
keir@16165 998 if ( (strlen(acpi_param) == 0) && acpi_disabled )
keir@16165 999 {
keir@16165 1000 printk("ACPI is disabled, notifying Domain 0 (acpi=off)\n");
keir@16165 1001 safe_strcpy(acpi_param, "off");
keir@16165 1002 }
kfraser@13691 1003 if ( (strlen(acpi_param) != 0) && !strstr(dom0_cmdline, "acpi=") )
kaf24@8459 1004 {
kfraser@13691 1005 safe_strcat(dom0_cmdline, " acpi=");
kfraser@13691 1006 safe_strcat(dom0_cmdline, acpi_param);
kaf24@8459 1007 }
kfraser@13691 1008
kfraser@13691 1009 cmdline = dom0_cmdline;
kaf24@8459 1010 }
kaf24@8459 1011
kaf24@8459 1012 if ( (initrdidx > 0) && (initrdidx < mbi->mods_count) )
kaf24@8459 1013 {
kaf24@8459 1014 _initrd_start = initial_images_start +
kaf24@8459 1015 (mod[initrdidx].mod_start - mod[0].mod_start);
kaf24@8459 1016 _initrd_len = mod[initrdidx].mod_end - mod[initrdidx].mod_start;
kaf24@8459 1017 }
kaf24@8459 1018
kfraser@15929 1019 iommu_setup();
kfraser@15929 1020
keir@15957 1021 amd_iommu_detect();
keir@15957 1022
kaf24@8459 1023 /*
kaf24@8459 1024 * We're going to setup domain0 using the module(s) that we stashed safely
kaf24@8459 1025 * above our heap. The second module, if present, is an initrd ramdisk.
kaf24@8459 1026 */
kaf24@8459 1027 if ( construct_dom0(dom0,
kaf24@8459 1028 initial_images_start,
kaf24@8459 1029 mod[0].mod_end-mod[0].mod_start,
kaf24@8459 1030 _initrd_start,
kaf24@8459 1031 _initrd_len,
kaf24@8459 1032 cmdline) != 0)
kaf24@8459 1033 panic("Could not set up DOM0 guest OS\n");
kaf24@8459 1034
kaf24@8459 1035 /* Scrub RAM that is still free and so may go to an unprivileged domain. */
kaf24@8459 1036 scrub_heap_pages();
kaf24@8459 1037
kaf24@8459 1038 init_trace_bufs();
kaf24@8459 1039
kaf24@10502 1040 console_endboot();
kaf24@8459 1041
kaf24@8459 1042 /* Hide UART from DOM0 if we're using it */
kaf24@8459 1043 serial_endboot();
kaf24@8459 1044
kaf24@8459 1045 domain_unpause_by_systemcontroller(dom0);
kaf24@8459 1046
keir@15082 1047 reset_stack_and_jump(init_done);
kaf24@8459 1048 }
kaf24@8459 1049
ian@13763 1050 void arch_get_xen_caps(xen_capabilities_info_t *info)
iap10@6721 1051 {
kfraser@14997 1052 /* Interface name is always xen-3.0-* for Xen-3.x. */
kfraser@14997 1053 int major = 3, minor = 0;
keir@13754 1054 char s[32];
keir@13754 1055
ian@13763 1056 (*info)[0] = '\0';
iap10@6721 1057
kaf24@6725 1058 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
kaf24@6725 1059
keir@13754 1060 snprintf(s, sizeof(s), "xen-%d.%d-x86_32 ", major, minor);
ian@13763 1061 safe_strcat(*info, s);
keir@13754 1062 if ( hvm_enabled )
keir@13754 1063 {
keir@13754 1064 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1065 safe_strcat(*info, s);
kfraser@13685 1066 }
kaf24@6725 1067
kaf24@6725 1068 #elif defined(CONFIG_X86_32) && defined(CONFIG_X86_PAE)
kaf24@6725 1069
keir@13754 1070 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1071 safe_strcat(*info, s);
kaf24@6725 1072 if ( hvm_enabled )
iap10@6721 1073 {
keir@13754 1074 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1075 safe_strcat(*info, s);
keir@13754 1076 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1077 safe_strcat(*info, s);
iap10@6721 1078 }
iap10@6721 1079
kaf24@6725 1080 #elif defined(CONFIG_X86_64)
iap10@6721 1081
keir@13754 1082 snprintf(s, sizeof(s), "xen-%d.%d-x86_64 ", major, minor);
ian@13763 1083 safe_strcat(*info, s);
ack@13288 1084 #ifdef CONFIG_COMPAT
keir@13754 1085 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1086 safe_strcat(*info, s);
ack@13288 1087 #endif
kaf24@6725 1088 if ( hvm_enabled )
iap10@6721 1089 {
keir@13754 1090 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1091 safe_strcat(*info, s);
keir@13754 1092 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1093 safe_strcat(*info, s);
keir@13754 1094 snprintf(s, sizeof(s), "hvm-%d.%d-x86_64 ", major, minor);
ian@13763 1095 safe_strcat(*info, s);
iap10@6721 1096 }
kaf24@6725 1097
iap10@6721 1098 #endif
iap10@6721 1099 }
iap10@6721 1100
kaf24@3914 1101 /*
kaf24@3914 1102 * Local variables:
kaf24@3914 1103 * mode: C
kaf24@3914 1104 * c-set-style: "BSD"
kaf24@3914 1105 * c-basic-offset: 4
kaf24@3914 1106 * tab-width: 4
kaf24@3914 1107 * indent-tabs-mode: nil
kaf24@3988 1108 * End:
kaf24@3914 1109 */