ia64/xen-unstable

annotate xen/arch/x86/setup.c @ 15740:db21f714d37f

Ensure multiboot modules are mapped when we copy them.
Signed-off-by: Keir Fraser <keir@xensource.com>
author kfraser@localhost.localdomain
date Thu Aug 09 15:16:51 2007 +0100 (2007-08-09)
parents f479595a3c5c
children f2649861d594
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>
iap10@6721 22 #include <public/version.h>
ack@13291 23 #ifdef CONFIG_COMPAT
ack@13291 24 #include <compat/platform.h>
ack@13291 25 #include <compat/xen.h>
ack@13291 26 #endif
kaf24@1452 27 #include <asm/bitops.h>
kaf24@1452 28 #include <asm/smp.h>
kaf24@1452 29 #include <asm/processor.h>
kaf24@1452 30 #include <asm/mpspec.h>
kaf24@1452 31 #include <asm/apic.h>
kaf24@1452 32 #include <asm/desc.h>
Tim@13909 33 #include <asm/paging.h>
kaf24@3344 34 #include <asm/e820.h>
kaf24@5536 35 #include <acm/acm_hooks.h>
ian@12677 36 #include <xen/kexec.h>
kfraser@15336 37 #include <asm/edd.h>
kaf24@3338 38
kfraser@15074 39 #if defined(CONFIG_X86_64)
kfraser@15597 40 #define BOOTSTRAP_DIRECTMAP_END (1UL << 32) /* 4GB */
kfraser@15074 41 #define maddr_to_bootstrap_virt(m) maddr_to_virt(m)
kfraser@15074 42 #else
kfraser@15597 43 #define BOOTSTRAP_DIRECTMAP_END (1UL << 30) /* 1GB */
kfraser@15074 44 #define maddr_to_bootstrap_virt(m) ((void *)(long)(m))
kfraser@15074 45 #endif
kfraser@15074 46
kaf24@5157 47 extern void dmi_scan_machine(void);
kaf24@5211 48 extern void generic_apic_probe(void);
kfraser@11971 49 extern void numa_initmem_init(unsigned long start_pfn, unsigned long end_pfn);
kaf24@5157 50
kfraser@15330 51 extern u16 boot_edid_caps;
kfraser@15330 52 extern u8 boot_edid_info[128];
kfraser@15330 53 extern struct boot_video_info boot_vid_info;
kfraser@15330 54
kaf24@3338 55 /*
kaf24@3338 56 * opt_xenheap_megabytes: Size of Xen heap in megabytes, excluding the
kaf24@8726 57 * page_info table and allocation bitmap.
kaf24@3338 58 */
kaf24@3338 59 static unsigned int opt_xenheap_megabytes = XENHEAP_DEFAULT_MB;
kaf24@4950 60 #if defined(CONFIG_X86_64)
kaf24@3338 61 integer_param("xenheap_megabytes", opt_xenheap_megabytes);
kaf24@3354 62 #endif
kaf24@1452 63
kaf24@5146 64 /* opt_nosmp: If true, secondary processors are ignored. */
kaf24@5900 65 static int opt_nosmp = 0;
kaf24@5146 66 boolean_param("nosmp", opt_nosmp);
kaf24@5146 67
kaf24@5146 68 /* maxcpus: maximum number of CPUs to activate. */
kaf24@5146 69 static unsigned int max_cpus = NR_CPUS;
shand@11156 70 integer_param("maxcpus", max_cpus);
kaf24@5146 71
kaf24@3334 72 /* opt_watchdog: If true, run a watchdog NMI on each processor. */
kaf24@3334 73 static int opt_watchdog = 0;
kaf24@3334 74 boolean_param("watchdog", opt_watchdog);
kaf24@3334 75
kaf24@4850 76 /* **** Linux config option: propagated to domain0. */
kaf24@4850 77 /* "acpi=off": Sisables both ACPI table parsing and interpreter. */
kaf24@4850 78 /* "acpi=force": Override the disable blacklist. */
kaf24@4850 79 /* "acpi=strict": Disables out-of-spec workarounds. */
kaf24@4850 80 /* "acpi=ht": Limit ACPI just to boot-time to enable HT. */
kaf24@4850 81 /* "acpi=noirq": Disables ACPI interrupt routing. */
kaf24@4850 82 static void parse_acpi_param(char *s);
kaf24@4850 83 custom_param("acpi", parse_acpi_param);
kaf24@4850 84
kaf24@4850 85 /* **** Linux config option: propagated to domain0. */
kaf24@4850 86 /* acpi_skip_timer_override: Skip IRQ0 overrides. */
kaf24@4850 87 extern int acpi_skip_timer_override;
kaf24@4850 88 boolean_param("acpi_skip_timer_override", acpi_skip_timer_override);
kaf24@4850 89
kaf24@4850 90 /* **** Linux config option: propagated to domain0. */
kaf24@4850 91 /* noapic: Disable IOAPIC setup. */
kaf24@4850 92 extern int skip_ioapic_setup;
kaf24@4850 93 boolean_param("noapic", skip_ioapic_setup);
kaf24@4850 94
kaf24@3594 95 int early_boot = 1;
kaf24@3594 96
kaf24@5146 97 cpumask_t cpu_present_map;
kaf24@5146 98
kfraser@15074 99 unsigned long xen_phys_start;
kfraser@15074 100
kaf24@5003 101 /* Limits of Xen heap, used to initialise the allocator. */
kaf24@5003 102 unsigned long xenheap_phys_start, xenheap_phys_end;
kaf24@3338 103
kaf24@2298 104 extern void arch_init_memory(void);
kaf24@1589 105 extern void init_IRQ(void);
kaf24@1589 106 extern void trap_init(void);
kaf24@5604 107 extern void early_time_init(void);
kaf24@5167 108 extern void early_cpu_init(void);
kaf24@1589 109
kaf24@8533 110 struct tss_struct init_tss[NR_CPUS];
kaf24@8533 111
kfraser@15490 112 char __attribute__ ((__section__(".bss.stack_aligned"))) cpu0_stack[STACK_SIZE];
kaf24@5011 113
kaf24@5214 114 struct cpuinfo_x86 boot_cpu_data = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
kaf24@1452 115
kaf24@5237 116 #if CONFIG_PAGING_LEVELS > 2
kaf24@1670 117 unsigned long mmu_cr4_features = X86_CR4_PSE | X86_CR4_PGE | X86_CR4_PAE;
kaf24@1670 118 #else
kaf24@5593 119 unsigned long mmu_cr4_features = X86_CR4_PSE;
kaf24@1670 120 #endif
kaf24@1452 121 EXPORT_SYMBOL(mmu_cr4_features);
kaf24@1452 122
kaf24@4818 123 int acpi_disabled;
kaf24@1452 124
kaf24@4850 125 int acpi_force;
kaf24@4850 126 char acpi_param[10] = "";
kfraser@15074 127 static void __init parse_acpi_param(char *s)
kaf24@4850 128 {
kaf24@4850 129 /* Save the parameter so it can be propagated to domain0. */
kfraser@13689 130 safe_strcpy(acpi_param, s);
kaf24@4850 131
kaf24@4850 132 /* Interpret the parameter for use within Xen. */
kaf24@4850 133 if ( !strcmp(s, "off") )
kaf24@4850 134 {
kaf24@4850 135 disable_acpi();
kaf24@4850 136 }
kaf24@4850 137 else if ( !strcmp(s, "force") )
kaf24@4850 138 {
kaf24@4850 139 acpi_force = 1;
kaf24@4850 140 acpi_ht = 1;
kaf24@4850 141 acpi_disabled = 0;
kaf24@4850 142 }
kaf24@4850 143 else if ( !strcmp(s, "strict") )
kaf24@4850 144 {
kaf24@4850 145 acpi_strict = 1;
kaf24@4850 146 }
kaf24@4850 147 else if ( !strcmp(s, "ht") )
kaf24@4850 148 {
kaf24@4850 149 if ( !acpi_force )
kaf24@4850 150 disable_acpi();
kaf24@4850 151 acpi_ht = 1;
kaf24@4850 152 }
kaf24@4850 153 else if ( !strcmp(s, "noirq") )
kaf24@4850 154 {
kaf24@4850 155 acpi_noirq_set();
kaf24@4850 156 }
kaf24@4850 157 }
kaf24@4850 158
kaf24@1452 159 static void __init do_initcalls(void)
kaf24@1452 160 {
kaf24@1452 161 initcall_t *call;
kaf24@1452 162 for ( call = &__initcall_start; call < &__initcall_end; call++ )
kaf24@1452 163 (*call)();
kaf24@1452 164 }
kaf24@1452 165
kfraser@15074 166 #define EARLY_FAIL(f, a...) do { \
kfraser@15074 167 printk( f , ## a ); \
kfraser@15074 168 for ( ; ; ) __asm__ __volatile__ ( "hlt" ); \
kfraser@15074 169 } while (0)
kaf24@8459 170
kfraser@15074 171 static unsigned long __initdata initial_images_start, initial_images_end;
kaf24@9067 172
kfraser@15074 173 unsigned long __init initial_images_nrpages(void)
kaf24@9067 174 {
kfraser@15489 175 ASSERT(!(initial_images_start & ~PAGE_MASK));
kfraser@15489 176 ASSERT(!(initial_images_end & ~PAGE_MASK));
kfraser@15489 177 return ((initial_images_end >> PAGE_SHIFT) -
kfraser@15489 178 (initial_images_start >> PAGE_SHIFT));
kaf24@9067 179 }
kaf24@9067 180
kfraser@15074 181 void __init discard_initial_images(void)
kaf24@9067 182 {
kaf24@9067 183 init_domheap_pages(initial_images_start, initial_images_end);
kaf24@9067 184 }
kaf24@9067 185
kaf24@9818 186 extern char __per_cpu_start[], __per_cpu_data_end[], __per_cpu_end[];
kaf24@9818 187
kfraser@11241 188 static void __init percpu_init_areas(void)
kaf24@9818 189 {
kaf24@9818 190 unsigned int i, data_size = __per_cpu_data_end - __per_cpu_start;
kfraser@15074 191 unsigned int first_unused;
kaf24@9818 192
kaf24@9818 193 BUG_ON(data_size > PERCPU_SIZE);
kaf24@9818 194
kfraser@15074 195 /* Initialise per-cpu data area for all possible secondary CPUs. */
kfraser@15074 196 for ( i = 1; (i < NR_CPUS) && cpu_possible(i); i++ )
kfraser@15074 197 memcpy(__per_cpu_start + (i << PERCPU_SHIFT),
kfraser@15074 198 __per_cpu_start,
kfraser@15074 199 data_size);
kaf24@9818 200 first_unused = i;
kaf24@9818 201
kfraser@14340 202 /* Check that there are no holes in cpu_possible_map. */
kaf24@9818 203 for ( ; i < NR_CPUS; i++ )
kfraser@14340 204 BUG_ON(cpu_possible(i));
kaf24@9818 205
kfraser@11241 206 #ifndef MEMORY_GUARD
kaf24@9818 207 init_xenheap_pages(__pa(__per_cpu_start) + (first_unused << PERCPU_SHIFT),
kaf24@9818 208 __pa(__per_cpu_end));
kfraser@11241 209 #endif
keir@15082 210 memguard_guard_range(&__per_cpu_start[first_unused << PERCPU_SHIFT],
keir@15082 211 (NR_CPUS - first_unused) << PERCPU_SHIFT);
keir@15082 212 #if defined(CONFIG_X86_64)
keir@15082 213 /* Also zap the mapping in the 1:1 area. */
keir@15082 214 memguard_guard_range(__va(__pa(__per_cpu_start)) +
keir@15082 215 (first_unused << PERCPU_SHIFT),
keir@15082 216 (NR_CPUS - first_unused) << PERCPU_SHIFT);
keir@15082 217 #endif
kaf24@9818 218 }
kaf24@9818 219
kfraser@11881 220 /* Fetch acm policy module from multiboot modules. */
kfraser@15074 221 static void __init extract_acm_policy(
kfraser@11881 222 multiboot_info_t *mbi,
kfraser@11881 223 unsigned int *initrdidx,
kfraser@11881 224 char **_policy_start,
kfraser@11881 225 unsigned long *_policy_len)
kfraser@11881 226 {
kfraser@11881 227 int i;
kfraser@11881 228 module_t *mod = (module_t *)__va(mbi->mods_addr);
kfraser@11881 229 unsigned long start, policy_len;
kfraser@11881 230 char *policy_start;
kfraser@11881 231
kfraser@11881 232 /*
kfraser@11881 233 * Try all modules and see whichever could be the binary policy.
kfraser@11881 234 * Adjust the initrdidx if module[1] is the binary policy.
kfraser@11881 235 */
kfraser@11881 236 for ( i = mbi->mods_count-1; i >= 1; i-- )
kfraser@11881 237 {
kfraser@11881 238 start = initial_images_start + (mod[i].mod_start-mod[0].mod_start);
kfraser@15074 239 policy_start = maddr_to_bootstrap_virt(start);
kfraser@11881 240 policy_len = mod[i].mod_end - mod[i].mod_start;
kfraser@11881 241 if ( acm_is_policy(policy_start, policy_len) )
kfraser@11881 242 {
kfraser@11881 243 printk("Policy len 0x%lx, start at %p - module %d.\n",
kfraser@11881 244 policy_len, policy_start, i);
kfraser@11881 245 *_policy_start = policy_start;
kfraser@11881 246 *_policy_len = policy_len;
kfraser@11881 247 if ( i == 1 )
kfraser@11881 248 *initrdidx = (mbi->mods_count > 2) ? 2 : 0;
kfraser@11881 249 break;
kfraser@11881 250 }
kfraser@11881 251 }
kfraser@11881 252 }
kfraser@11881 253
kfraser@11241 254 static void __init init_idle_domain(void)
kfraser@11240 255 {
kfraser@11240 256 struct domain *idle_domain;
kfraser@11240 257
kfraser@11240 258 /* Domain creation requires that scheduler structures are initialised. */
kfraser@11240 259 scheduler_init();
kfraser@11240 260
kfraser@14911 261 idle_domain = domain_create(IDLE_DOMAIN_ID, 0, 0);
kfraser@11240 262 if ( (idle_domain == NULL) || (alloc_vcpu(idle_domain, 0, 0) == NULL) )
kfraser@11240 263 BUG();
kfraser@11240 264
kfraser@11240 265 set_current(idle_domain->vcpu[0]);
kfraser@11240 266 idle_vcpu[0] = this_cpu(curr_vcpu) = current;
kfraser@11240 267
kfraser@11240 268 setup_idle_pagetable();
kfraser@11240 269 }
kfraser@11240 270
kfraser@15074 271 static void __init srat_detect_node(int cpu)
kfraser@11971 272 {
kfraser@11998 273 unsigned node;
kfraser@11998 274 u8 apicid = x86_cpu_to_apicid[cpu];
kfraser@11971 275
kfraser@11998 276 node = apicid_to_node[apicid];
kfraser@11998 277 if ( node == NUMA_NO_NODE )
kfraser@11998 278 node = 0;
kfraser@11998 279 numa_set_node(cpu, node);
kfraser@11971 280
kfraser@11998 281 if ( acpi_numa > 0 )
kfraser@11998 282 printk(KERN_INFO "CPU %d APIC %d -> Node %d\n", cpu, apicid, node);
kfraser@11971 283 }
kfraser@11971 284
kfraser@15740 285 /*
kfraser@15740 286 * Ensure a given physical memory range is present in the bootstrap mappings.
kfraser@15740 287 * Use superpage mappings to ensure that pagetable memory needn't be allocated.
kfraser@15740 288 */
kfraser@15740 289 static void __init bootstrap_map(unsigned long start, unsigned long end)
kfraser@15740 290 {
kfraser@15740 291 unsigned long mask = (1UL << L2_PAGETABLE_SHIFT) - 1;
kfraser@15740 292 start = start & ~mask;
kfraser@15740 293 end = (end + mask) & ~mask;
kfraser@15740 294 if ( end > BOOTSTRAP_DIRECTMAP_END )
kfraser@15740 295 panic("Cannot access memory beyond end of "
kfraser@15740 296 "bootstrap direct-map area\n");
kfraser@15740 297 map_pages_to_xen(
kfraser@15740 298 (unsigned long)maddr_to_bootstrap_virt(start),
kfraser@15740 299 start >> PAGE_SHIFT, (end-start) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kfraser@15740 300 }
kfraser@15740 301
kfraser@15074 302 static void __init move_memory(
kfraser@15074 303 unsigned long dst, unsigned long src_start, unsigned long src_end)
ian@12677 304 {
kfraser@15740 305 bootstrap_map(src_start, src_end);
kfraser@15740 306 bootstrap_map(dst, dst + src_end - src_start);
kfraser@15074 307 memmove(maddr_to_bootstrap_virt(dst),
kfraser@15074 308 maddr_to_bootstrap_virt(src_start),
ian@12677 309 src_end - src_start);
kfraser@15074 310 }
kfraser@15074 311
kfraser@15074 312 /* A temporary copy of the e820 map that we can mess with during bootstrap. */
kfraser@15074 313 static struct e820map __initdata boot_e820;
kfraser@15074 314
kfraser@15074 315 /* Reserve area (@s,@e) in the temporary bootstrap e820 map. */
kfraser@15489 316 static int __init reserve_in_boot_e820(unsigned long s, unsigned long e)
kfraser@15074 317 {
kfraser@15403 318 uint64_t rs, re;
kfraser@15074 319 int i;
kfraser@15074 320
kfraser@15074 321 for ( i = 0; i < boot_e820.nr_map; i++ )
kfraser@15074 322 {
kfraser@15074 323 /* Have we found the e820 region that includes the specified range? */
kfraser@15074 324 rs = boot_e820.map[i].addr;
kfraser@15403 325 re = rs + boot_e820.map[i].size;
kfraser@15489 326 if ( (s >= rs) && (e <= re) )
kfraser@15489 327 goto found;
kfraser@15489 328 }
kfraser@15074 329
kfraser@15489 330 return 0;
kfraser@15489 331
kfraser@15489 332 found:
kfraser@15489 333 /* Start fragment. */
kfraser@15489 334 boot_e820.map[i].size = s - rs;
kfraser@15489 335
kfraser@15489 336 /* End fragment. */
kfraser@15489 337 if ( e < re )
kfraser@15489 338 {
kfraser@15489 339 memmove(&boot_e820.map[i+1], &boot_e820.map[i],
kfraser@15489 340 (boot_e820.nr_map-i) * sizeof(boot_e820.map[0]));
kfraser@15489 341 boot_e820.nr_map++;
kfraser@15489 342 i++;
kfraser@15489 343 boot_e820.map[i].addr = e;
kfraser@15489 344 boot_e820.map[i].size = re - e;
kfraser@15074 345 }
kfraser@15489 346
kfraser@15489 347 return 1;
ian@12677 348 }
ian@12677 349
keir@15298 350 struct boot_video_info {
keir@15298 351 u8 orig_x; /* 0x00 */
keir@15298 352 u8 orig_y; /* 0x01 */
keir@15298 353 u8 orig_video_mode; /* 0x02 */
keir@15298 354 u8 orig_video_cols; /* 0x03 */
keir@15298 355 u8 orig_video_lines; /* 0x04 */
keir@15298 356 u8 orig_video_isVGA; /* 0x05 */
keir@15298 357 u16 orig_video_points; /* 0x06 */
keir@15298 358
keir@15298 359 /* VESA graphic mode -- linear frame buffer */
keir@15298 360 u32 capabilities; /* 0x08 */
keir@15298 361 u16 lfb_linelength; /* 0x0c */
keir@15298 362 u16 lfb_width; /* 0x0e */
keir@15298 363 u16 lfb_height; /* 0x10 */
keir@15298 364 u16 lfb_depth; /* 0x12 */
keir@15298 365 u32 lfb_base; /* 0x14 */
keir@15298 366 u32 lfb_size; /* 0x18 */
keir@15298 367 u8 red_size; /* 0x1c */
keir@15298 368 u8 red_pos; /* 0x1d */
keir@15298 369 u8 green_size; /* 0x1e */
keir@15298 370 u8 green_pos; /* 0x1f */
keir@15298 371 u8 blue_size; /* 0x20 */
keir@15298 372 u8 blue_pos; /* 0x21 */
keir@15298 373 u8 rsvd_size; /* 0x22 */
keir@15298 374 u8 rsvd_pos; /* 0x23 */
keir@15298 375 u16 vesapm_seg; /* 0x24 */
keir@15298 376 u16 vesapm_off; /* 0x26 */
keir@15298 377 };
keir@15298 378
keir@15298 379 static void __init parse_video_info(void)
keir@15298 380 {
keir@15298 381 struct boot_video_info *bvi = &bootsym(boot_vid_info);
keir@15298 382
keir@15298 383 if ( (bvi->orig_video_isVGA == 1) && (bvi->orig_video_mode == 3) )
keir@15298 384 {
keir@15298 385 vga_console_info.video_type = XEN_VGATYPE_TEXT_MODE_3;
keir@15298 386 vga_console_info.u.text_mode_3.font_height = bvi->orig_video_points;
keir@15298 387 vga_console_info.u.text_mode_3.cursor_x = bvi->orig_x;
keir@15298 388 vga_console_info.u.text_mode_3.cursor_y = bvi->orig_y;
keir@15298 389 vga_console_info.u.text_mode_3.rows = bvi->orig_video_lines;
keir@15298 390 vga_console_info.u.text_mode_3.columns = bvi->orig_video_cols;
keir@15298 391 }
keir@15298 392 else if ( bvi->orig_video_isVGA == 0x23 )
keir@15298 393 {
keir@15298 394 vga_console_info.video_type = XEN_VGATYPE_VESA_LFB;
keir@15298 395 vga_console_info.u.vesa_lfb.width = bvi->lfb_width;
keir@15298 396 vga_console_info.u.vesa_lfb.height = bvi->lfb_height;
keir@15298 397 vga_console_info.u.vesa_lfb.bytes_per_line = bvi->lfb_linelength;
keir@15298 398 vga_console_info.u.vesa_lfb.bits_per_pixel = bvi->lfb_depth;
keir@15298 399 vga_console_info.u.vesa_lfb.lfb_base = bvi->lfb_base;
keir@15298 400 vga_console_info.u.vesa_lfb.lfb_size = bvi->lfb_size;
keir@15298 401 vga_console_info.u.vesa_lfb.red_pos = bvi->red_pos;
keir@15298 402 vga_console_info.u.vesa_lfb.red_size = bvi->red_size;
keir@15298 403 vga_console_info.u.vesa_lfb.green_pos = bvi->green_pos;
keir@15298 404 vga_console_info.u.vesa_lfb.green_size = bvi->green_size;
keir@15298 405 vga_console_info.u.vesa_lfb.blue_pos = bvi->blue_pos;
keir@15298 406 vga_console_info.u.vesa_lfb.blue_size = bvi->blue_size;
keir@15298 407 vga_console_info.u.vesa_lfb.rsvd_pos = bvi->rsvd_pos;
keir@15298 408 vga_console_info.u.vesa_lfb.rsvd_size = bvi->rsvd_size;
keir@15298 409 }
keir@15298 410 }
keir@15298 411
keir@15082 412 void init_done(void)
keir@15082 413 {
keir@15082 414 extern char __init_begin[], __init_end[];
keir@15082 415
keir@15082 416 /* Free (or page-protect) the init areas. */
keir@15082 417 #ifndef MEMORY_GUARD
keir@15082 418 init_xenheap_pages(__pa(__init_begin), __pa(__init_end));
keir@15082 419 #endif
keir@15082 420 memguard_guard_range(__init_begin, __init_end - __init_begin);
keir@15082 421 #if defined(CONFIG_X86_64)
keir@15082 422 /* Also zap the mapping in the 1:1 area. */
keir@15082 423 memguard_guard_range(__va(__pa(__init_begin)), __init_end - __init_begin);
keir@15082 424 #endif
keir@15082 425 printk("Freed %ldkB init memory.\n", (long)(__init_end-__init_begin)>>10);
keir@15082 426
keir@15082 427 startup_cpu_idle_loop();
keir@15082 428 }
keir@15082 429
kfraser@15379 430 void __init __start_xen(unsigned long mbi_p)
kaf24@1452 431 {
kfraser@15293 432 char *memmap_type = NULL;
kfraser@15426 433 char __cmdline[] = "", *cmdline = __cmdline, *kextra;
kaf24@8457 434 unsigned long _initrd_start = 0, _initrd_len = 0;
kaf24@8457 435 unsigned int initrdidx = 1;
kfraser@11881 436 char *_policy_start = NULL;
kfraser@11881 437 unsigned long _policy_len = 0;
kfraser@15379 438 multiboot_info_t *mbi = __va(mbi_p);
kaf24@8457 439 module_t *mod = (module_t *)__va(mbi->mods_addr);
kaf24@6111 440 unsigned long nr_pages, modules_length;
kfraser@15293 441 int i, e820_warn = 0, bytes = 0;
kaf24@5776 442 struct ns16550_defaults ns16550 = {
kaf24@5776 443 .data_bits = 8,
kaf24@5776 444 .parity = 'n',
kaf24@5776 445 .stop_bits = 1
kaf24@5776 446 };
kaf24@3338 447
kfraser@12853 448 extern void early_page_fault(void);
kfraser@12853 449 set_intr_gate(TRAP_page_fault, &early_page_fault);
kfraser@12853 450
kaf24@3338 451 /* Parse the command-line options. */
kaf24@3344 452 if ( (mbi->flags & MBI_CMDLINE) && (mbi->cmdline != 0) )
kaf24@9823 453 cmdline = __va(mbi->cmdline);
kfraser@15426 454 if ( (kextra = strstr(cmdline, " -- ")) != NULL )
kfraser@15426 455 {
kfraser@15426 456 /*
kfraser@15426 457 * Options after ' -- ' separator belong to dom0.
kfraser@15426 458 * 1. Orphan dom0's options from Xen's command line.
kfraser@15426 459 * 2. Skip all but final leading space from dom0's options.
kfraser@15426 460 */
kfraser@15426 461 *kextra = '\0';
kfraser@15426 462 kextra += 3;
kfraser@15426 463 while ( kextra[1] == ' ' ) kextra++;
kfraser@15426 464 }
kaf24@9823 465 cmdline_parse(cmdline);
kaf24@3338 466
keir@15298 467 parse_video_info();
keir@15298 468
kaf24@8534 469 set_current((struct vcpu *)0xfffff000); /* debug sanity */
kfraser@11240 470 idle_vcpu[0] = current;
kaf24@8534 471 set_processor_id(0); /* needed early, for smp_processor_id() */
kaf24@3338 472
kaf24@5146 473 smp_prepare_boot_cpu();
kaf24@5146 474
kaf24@3338 475 /* We initialise the serial devices very early so we can get debugging. */
kaf24@5776 476 ns16550.io_base = 0x3f8;
kaf24@5776 477 ns16550.irq = 4;
kaf24@5776 478 ns16550_init(0, &ns16550);
kaf24@5776 479 ns16550.io_base = 0x2f8;
kaf24@5776 480 ns16550.irq = 3;
kaf24@5776 481 ns16550_init(1, &ns16550);
kaf24@5195 482 serial_init_preirq();
kaf24@3338 483
kaf24@3338 484 init_console();
kaf24@3338 485
kfraser@11947 486 printk("Command line: %s\n", cmdline);
kaf24@9823 487
kfraser@15330 488 printk("Video information:\n");
kfraser@15330 489
kfraser@15330 490 /* Print VGA display mode information. */
keir@15298 491 switch ( vga_console_info.video_type )
keir@15298 492 {
keir@15298 493 case XEN_VGATYPE_TEXT_MODE_3:
kfraser@15330 494 printk(" VGA is text mode %dx%d, font 8x%d\n",
keir@15298 495 vga_console_info.u.text_mode_3.columns,
keir@15298 496 vga_console_info.u.text_mode_3.rows,
keir@15298 497 vga_console_info.u.text_mode_3.font_height);
keir@15298 498 break;
keir@15298 499 case XEN_VGATYPE_VESA_LFB:
kfraser@15330 500 printk(" VGA is graphics mode %dx%d, %d bpp\n",
keir@15298 501 vga_console_info.u.vesa_lfb.width,
keir@15298 502 vga_console_info.u.vesa_lfb.height,
keir@15298 503 vga_console_info.u.vesa_lfb.bits_per_pixel);
keir@15298 504 break;
kfraser@15330 505 default:
kfraser@15330 506 printk(" No VGA detected\n");
kfraser@15330 507 break;
kfraser@15330 508 }
kfraser@15330 509
kfraser@15330 510 /* Print VBE/DDC EDID information. */
kfraser@15330 511 if ( bootsym(boot_edid_caps) != 0x1313 )
kfraser@15330 512 {
kfraser@15330 513 u16 caps = bootsym(boot_edid_caps);
kfraser@15330 514 printk(" VBE/DDC methods:%s%s%s; ",
kfraser@15330 515 (caps & 1) ? " V1" : "",
kfraser@15330 516 (caps & 2) ? " V2" : "",
kfraser@15330 517 !(caps & 3) ? " none" : "");
kfraser@15330 518 printk("EDID transfer time: %d seconds\n", caps >> 8);
kfraser@15330 519 if ( *(u32 *)bootsym(boot_edid_info) == 0x13131313 )
kfraser@15330 520 {
kfraser@15330 521 printk(" EDID info not retrieved because ");
kfraser@15330 522 if ( !(caps & 3) )
kfraser@15330 523 printk("no DDC retrieval method detected\n");
kfraser@15330 524 else if ( (caps >> 8) > 5 )
kfraser@15330 525 printk("takes longer than 5 seconds\n");
kfraser@15330 526 else
kfraser@15330 527 printk("of reasons unknown\n");
kfraser@15330 528 }
keir@15298 529 }
keir@15298 530
kfraser@15336 531 printk("Disc information:\n");
kfraser@15336 532 printk(" Found %d MBR signatures\n",
kfraser@15430 533 bootsym(boot_mbr_signature_nr));
kfraser@15336 534 printk(" Found %d EDD information structures\n",
kfraser@15336 535 bootsym(boot_edd_info_nr));
kfraser@15336 536
kaf24@3344 537 /* Check that we have at least one Multiboot module. */
kaf24@3344 538 if ( !(mbi->flags & MBI_MODULES) || (mbi->mods_count == 0) )
kfraser@15074 539 EARLY_FAIL("dom0 kernel not specified. "
kfraser@15074 540 "Check bootloader configuration.\n");
kaf24@5011 541
kaf24@5011 542 if ( ((unsigned long)cpu0_stack & (STACK_SIZE-1)) != 0 )
kfraser@15074 543 EARLY_FAIL("Misaligned CPU0 stack.\n");
kaf24@3338 544
kfraser@11618 545 /*
kfraser@11618 546 * Since there are some stubs getting built on the stacks which use
kfraser@11618 547 * direct calls/jumps, the heap must be confined to the lower 2G so
kfraser@11618 548 * that those branches can reach their targets.
kfraser@11618 549 */
kfraser@11618 550 if ( opt_xenheap_megabytes > 2048 )
kfraser@11618 551 opt_xenheap_megabytes = 2048;
kaf24@3338 552
kfraser@15293 553 if ( e820_raw_nr != 0 )
kfraser@15292 554 {
kfraser@15293 555 memmap_type = "Xen-e820";
kfraser@15292 556 }
kfraser@15293 557 else if ( bootsym(lowmem_kb) )
kfraser@15292 558 {
kfraser@15293 559 memmap_type = "Xen-e801";
kfraser@15292 560 e820_raw[0].addr = 0;
kfraser@15293 561 e820_raw[0].size = bootsym(lowmem_kb) << 10;
kfraser@15292 562 e820_raw[0].type = E820_RAM;
kfraser@15292 563 e820_raw[1].addr = 0x100000;
kfraser@15293 564 e820_raw[1].size = bootsym(highmem_kb) << 10;
kfraser@15292 565 e820_raw[1].type = E820_RAM;
kfraser@15292 566 e820_raw_nr = 2;
kfraser@15292 567 }
kfraser@15292 568 else if ( mbi->flags & MBI_MEMMAP )
kaf24@3344 569 {
kfraser@15293 570 memmap_type = "Multiboot-e820";
kaf24@3344 571 while ( bytes < mbi->mmap_length )
kaf24@3344 572 {
kaf24@3344 573 memory_map_t *map = __va(mbi->mmap_addr + bytes);
kaf24@8402 574
kaf24@8402 575 /*
kaf24@8403 576 * This is a gross workaround for a BIOS bug. Some bootloaders do
kaf24@8402 577 * not write e820 map entries into pre-zeroed memory. This is
kaf24@8402 578 * okay if the BIOS fills in all fields of the map entry, but
kaf24@8402 579 * some broken BIOSes do not bother to write the high word of
kaf24@8402 580 * the length field if the length is smaller than 4GB. We
kaf24@8402 581 * detect and fix this by flagging sections below 4GB that
kaf24@8403 582 * appear to be larger than 4GB in size.
kaf24@8402 583 */
kaf24@8403 584 if ( (map->base_addr_high == 0) && (map->length_high != 0) )
kaf24@8402 585 {
kfraser@15292 586 if ( !e820_warn )
kfraser@15292 587 {
kfraser@15292 588 printk("WARNING: Buggy e820 map detected and fixed "
kfraser@15292 589 "(truncated length fields).\n");
kfraser@15292 590 e820_warn = 1;
kfraser@15292 591 }
kaf24@8402 592 map->length_high = 0;
kaf24@8402 593 }
kaf24@8402 594
kaf24@3344 595 e820_raw[e820_raw_nr].addr =
kaf24@3344 596 ((u64)map->base_addr_high << 32) | (u64)map->base_addr_low;
kaf24@3344 597 e820_raw[e820_raw_nr].size =
kaf24@3344 598 ((u64)map->length_high << 32) | (u64)map->length_low;
kaf24@3344 599 e820_raw[e820_raw_nr].type =
kfraser@12226 600 (map->type > E820_NVS) ? E820_RESERVED : map->type;
kaf24@3344 601 e820_raw_nr++;
kaf24@8402 602
kaf24@3344 603 bytes += map->size + 4;
kaf24@3344 604 }
kaf24@3344 605 }
kaf24@3344 606 else if ( mbi->flags & MBI_MEMLIMITS )
kaf24@3344 607 {
kfraser@15293 608 memmap_type = "Multiboot-e801";
kaf24@3344 609 e820_raw[0].addr = 0;
kaf24@3344 610 e820_raw[0].size = mbi->mem_lower << 10;
kaf24@3344 611 e820_raw[0].type = E820_RAM;
kaf24@3354 612 e820_raw[1].addr = 0x100000;
kaf24@3354 613 e820_raw[1].size = mbi->mem_upper << 10;
kaf24@3354 614 e820_raw[1].type = E820_RAM;
kaf24@3344 615 e820_raw_nr = 2;
kaf24@3344 616 }
kaf24@3344 617 else
kaf24@3344 618 {
kfraser@15074 619 EARLY_FAIL("Bootloader provided no memory information.\n");
kaf24@3344 620 }
kaf24@3344 621
kaf24@13427 622 /* Ensure that all E820 RAM regions are page-aligned and -sized. */
kaf24@13427 623 for ( i = 0; i < e820_raw_nr; i++ )
kaf24@13427 624 {
kaf24@13427 625 uint64_t s, e;
kfraser@15292 626
kaf24@13427 627 if ( e820_raw[i].type != E820_RAM )
kaf24@13427 628 continue;
kaf24@13427 629 s = PFN_UP(e820_raw[i].addr);
kaf24@13427 630 e = PFN_DOWN(e820_raw[i].addr + e820_raw[i].size);
kaf24@13427 631 e820_raw[i].size = 0; /* discarded later */
kaf24@13427 632 if ( s < e )
kaf24@13427 633 {
kaf24@13427 634 e820_raw[i].addr = s << PAGE_SHIFT;
kaf24@13427 635 e820_raw[i].size = (e - s) << PAGE_SHIFT;
kaf24@13427 636 }
kaf24@13427 637 }
kaf24@13427 638
kaf24@13427 639 /* Sanitise the raw E820 map to produce a final clean version. */
kfraser@15293 640 max_page = init_e820(memmap_type, e820_raw, &e820_raw_nr);
kaf24@3338 641
kfraser@15074 642 /*
kfraser@15074 643 * Create a temporary copy of the E820 map. Truncate it to above 16MB
kfraser@15074 644 * as anything below that is already mapped and has a statically-allocated
kfraser@15074 645 * purpose.
kfraser@15074 646 */
kfraser@15074 647 memcpy(&boot_e820, &e820, sizeof(e820));
kfraser@15074 648 for ( i = 0; i < boot_e820.nr_map; i++ )
kaf24@3338 649 {
kfraser@15074 650 uint64_t s, e, min = 16 << 20; /* 16MB */
kfraser@15074 651 s = boot_e820.map[i].addr;
kfraser@15074 652 e = boot_e820.map[i].addr + boot_e820.map[i].size;
kfraser@15074 653 if ( s >= min )
kfraser@15074 654 continue;
kfraser@15074 655 if ( e > min )
kaf24@6111 656 {
kfraser@15074 657 boot_e820.map[i].addr = min;
kfraser@15074 658 boot_e820.map[i].size = e - min;
kaf24@6111 659 }
kfraser@15074 660 else
kfraser@15074 661 boot_e820.map[i].type = E820_RESERVED;
kaf24@3338 662 }
kaf24@6111 663
kfraser@15074 664 /*
keir@15077 665 * Iterate backwards over all superpage-aligned RAM regions.
kfraser@15074 666 *
kfraser@15074 667 * We require superpage alignment because the boot allocator is not yet
kfraser@15074 668 * initialised. Hence we can only map superpages in the address range
kfraser@15074 669 * 0 to BOOTSTRAP_DIRECTMAP_END, as this is guaranteed not to require
kfraser@15074 670 * dynamic allocation of pagetables.
kfraser@15074 671 *
kfraser@15074 672 * As well as mapping superpages in that range, in preparation for
kfraser@15074 673 * initialising the boot allocator, we also look for a region to which
kfraser@15074 674 * we can relocate the dom0 kernel and other multiboot modules. Also, on
kfraser@15074 675 * x86/64, we relocate Xen to higher memory.
kfraser@15074 676 */
kfraser@15074 677 modules_length = mod[mbi->mods_count-1].mod_end - mod[0].mod_start;
keir@15077 678 for ( i = boot_e820.nr_map-1; i >= 0; i-- )
kfraser@15074 679 {
kfraser@15074 680 uint64_t s, e, mask = (1UL << L2_PAGETABLE_SHIFT) - 1;
kaf24@6134 681
kfraser@15074 682 /* Superpage-aligned chunks up to BOOTSTRAP_DIRECTMAP_END, please. */
kfraser@15074 683 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 684 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
kfraser@15074 685 e = min_t(uint64_t, e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 686 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kaf24@3354 687 continue;
kaf24@6111 688
kfraser@15074 689 /* Map the chunk. No memory will need to be allocated to do this. */
kfraser@15074 690 map_pages_to_xen(
kfraser@15074 691 (unsigned long)maddr_to_bootstrap_virt(s),
kfraser@15074 692 s >> PAGE_SHIFT, (e-s) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kaf24@6111 693
kfraser@14084 694 #if defined(CONFIG_X86_64)
kfraser@15074 695 /* Is the region suitable for relocating Xen? */
kfraser@15074 696 if ( !xen_phys_start && (((e-s) >> 20) >= opt_xenheap_megabytes) )
kaf24@5003 697 {
kfraser@15074 698 extern l2_pgentry_t l2_xenmap[];
kfraser@15074 699 l4_pgentry_t *pl4e;
kfraser@15074 700 l3_pgentry_t *pl3e;
kfraser@15074 701 l2_pgentry_t *pl2e;
kfraser@15074 702 int i, j;
kfraser@15074 703
kfraser@15074 704 /* Select relocation address. */
kfraser@15074 705 e = (e - (opt_xenheap_megabytes << 20)) & ~mask;
kfraser@15074 706 xen_phys_start = e;
kfraser@15292 707 bootsym(trampoline_xen_phys_start) = e;
kfraser@15074 708
kfraser@15074 709 /*
kfraser@15074 710 * Perform relocation to new physical address.
kfraser@15074 711 * Before doing so we must sync static/global data with main memory
kfraser@15074 712 * with a barrier(). After this we must *not* modify static/global
kfraser@15074 713 * data until after we have switched to the relocated pagetables!
kfraser@15074 714 */
kfraser@15074 715 barrier();
kfraser@15074 716 move_memory(e, 0, __pa(&_end) - xen_phys_start);
kfraser@15074 717
kfraser@15379 718 /* Poison low 1MB to detect stray pointers to physical 0-1MB. */
kfraser@15379 719 memset(maddr_to_bootstrap_virt(e), 0x55, 1U<<20);
kfraser@15379 720
kfraser@15074 721 /* Walk initial pagetables, relocating page directory entries. */
kfraser@15074 722 pl4e = __va(__pa(idle_pg_table));
kfraser@15074 723 for ( i = 0 ; i < L4_PAGETABLE_ENTRIES; i++, pl4e++ )
kfraser@15074 724 {
kfraser@15074 725 if ( !(l4e_get_flags(*pl4e) & _PAGE_PRESENT) )
kfraser@15074 726 continue;
kfraser@15074 727 *pl4e = l4e_from_intpte(l4e_get_intpte(*pl4e) +
kfraser@15074 728 xen_phys_start);
kfraser@15074 729 pl3e = l4e_to_l3e(*pl4e);
kfraser@15074 730 for ( j = 0; j < L3_PAGETABLE_ENTRIES; j++, pl3e++ )
kfraser@15074 731 {
kfraser@15074 732 /* Not present or already relocated? */
kfraser@15074 733 if ( !(l3e_get_flags(*pl3e) & _PAGE_PRESENT) ||
kfraser@15074 734 (l3e_get_pfn(*pl3e) > 0x1000) )
kfraser@15074 735 continue;
kfraser@15074 736 *pl3e = l3e_from_intpte(l3e_get_intpte(*pl3e) +
kfraser@15074 737 xen_phys_start);
kfraser@15074 738 }
kfraser@15074 739 }
kfraser@15074 740
kfraser@15074 741 /* The only data mappings to be relocated are in the Xen area. */
kfraser@15074 742 pl2e = __va(__pa(l2_xenmap));
kfraser@15074 743 for ( i = 0; i < L2_PAGETABLE_ENTRIES; i++, pl2e++ )
kfraser@15074 744 {
kfraser@15074 745 if ( !(l2e_get_flags(*pl2e) & _PAGE_PRESENT) )
kfraser@15074 746 continue;
kfraser@15074 747 *pl2e = l2e_from_intpte(l2e_get_intpte(*pl2e) +
kfraser@15074 748 xen_phys_start);
kfraser@15074 749 }
kfraser@15074 750
kfraser@15074 751 /* Re-sync the stack and then switch to relocated pagetables. */
kfraser@15074 752 asm volatile (
kfraser@15074 753 "rep movsb ; " /* re-sync the stack */
kfraser@15074 754 "movq %%cr4,%%rsi ; "
kfraser@15074 755 "andb $0x7f,%%sil ; "
kfraser@15074 756 "movq %%rsi,%%cr4 ; " /* CR4.PGE == 0 */
kfraser@15074 757 "movq %0,%%cr3 ; " /* CR3 == new pagetables */
kfraser@15074 758 "orb $0x80,%%sil ; "
kfraser@15074 759 "movq %%rsi,%%cr4 " /* CR4.PGE == 1 */
kfraser@15074 760 : : "r" (__pa(idle_pg_table)), "S" (cpu0_stack),
kfraser@15074 761 "D" (__va(__pa(cpu0_stack))), "c" (STACK_SIZE) : "memory" );
kaf24@5003 762 }
kaf24@5003 763 #endif
keir@15077 764
keir@15077 765 /* Is the region suitable for relocating the multiboot modules? */
keir@15077 766 if ( !initial_images_start && (s < e) && ((e-s) >= modules_length) )
keir@15077 767 {
kfraser@15489 768 initial_images_end = e;
kfraser@15489 769 e = (e - modules_length) & PAGE_MASK;
keir@15077 770 initial_images_start = e;
keir@15077 771 move_memory(initial_images_start,
keir@15077 772 mod[0].mod_start, mod[mbi->mods_count-1].mod_end);
keir@15077 773 }
kfraser@15489 774
kfraser@15489 775 if ( !kexec_crash_area.start && (s < e) &&
kfraser@15489 776 ((e-s) >= kexec_crash_area.size) )
kfraser@15489 777 {
kfraser@15489 778 e = (e - kexec_crash_area.size) & PAGE_MASK;
kfraser@15489 779 kexec_crash_area.start = e;
kfraser@15489 780 }
kaf24@3354 781 }
kaf24@3354 782
kfraser@15074 783 if ( !initial_images_start )
kfraser@15074 784 EARLY_FAIL("Not enough memory to relocate the dom0 kernel image.\n");
kfraser@15074 785 reserve_in_boot_e820(initial_images_start, initial_images_end);
kfraser@15074 786
kfraser@15074 787 /*
kfraser@15074 788 * With modules (and Xen itself, on x86/64) relocated out of the way, we
kfraser@15074 789 * can now initialise the boot allocator with some memory.
kfraser@15074 790 */
kfraser@15074 791 xenheap_phys_start = init_boot_allocator(__pa(&_end));
kfraser@15074 792 xenheap_phys_end = opt_xenheap_megabytes << 20;
kfraser@15074 793 #if defined(CONFIG_X86_64)
kfraser@15074 794 if ( !xen_phys_start )
kfraser@15074 795 EARLY_FAIL("Not enough memory to relocate Xen.\n");
kfraser@15074 796 xenheap_phys_end += xen_phys_start;
kfraser@15074 797 reserve_in_boot_e820(xen_phys_start,
kfraser@15074 798 xen_phys_start + (opt_xenheap_megabytes<<20));
kfraser@15074 799 init_boot_pages(1<<20, 16<<20); /* Initial seed: 15MB */
kfraser@15074 800 #else
kfraser@15074 801 init_boot_pages(xenheap_phys_end, 16<<20); /* Initial seed: 4MB */
kfraser@15074 802 #endif
kfraser@15074 803
kfraser@15489 804 if ( kexec_crash_area.size != 0 )
kfraser@15489 805 {
kfraser@15489 806 unsigned long kdump_start = kexec_crash_area.start;
kfraser@15489 807 unsigned long kdump_size = kexec_crash_area.size;
kfraser@15489 808
kfraser@15489 809 kdump_size = (kdump_size + PAGE_SIZE - 1) & PAGE_MASK;
kfraser@15489 810
kfraser@15489 811 if ( !reserve_in_boot_e820(kdump_start, kdump_size) )
kfraser@15489 812 {
kfraser@15489 813 printk("Kdump: DISABLED (failed to reserve %luMB (%lukB) at 0x%lx)"
kfraser@15489 814 "\n", kdump_size >> 20, kdump_size >> 10, kdump_start);
kfraser@15489 815 kexec_crash_area.start = kexec_crash_area.size = 0;
kfraser@15489 816 }
kfraser@15489 817 else
kfraser@15489 818 {
kfraser@15489 819 printk("Kdump: %luMB (%lukB) at 0x%lx\n",
kfraser@15489 820 kdump_size >> 20, kdump_size >> 10, kdump_start);
kfraser@15489 821 }
kfraser@15489 822 }
kfraser@15489 823
kfraser@15074 824 /*
kfraser@15074 825 * With the boot allocator now seeded, we can walk every RAM region and
kfraser@15074 826 * map it in its entirety (on x86/64, at least) and notify it to the
kfraser@15074 827 * boot allocator.
kfraser@15074 828 */
kfraser@15074 829 for ( i = 0; i < boot_e820.nr_map; i++ )
kfraser@15074 830 {
kfraser@15074 831 uint64_t s, e, map_e, mask = PAGE_SIZE - 1;
kfraser@15074 832
kfraser@15074 833 /* Only page alignment required now. */
kfraser@15074 834 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 835 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
kfraser@15074 836 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kfraser@15074 837 continue;
kfraser@15074 838
kfraser@15074 839 /* Perform the mapping (truncated in 32-bit mode). */
kfraser@15074 840 map_e = e;
kfraser@15074 841 #if defined(CONFIG_X86_32)
kfraser@15074 842 map_e = min_t(uint64_t, map_e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 843 #endif
kfraser@15074 844 if ( s < map_e )
kfraser@15074 845 map_pages_to_xen(
kfraser@15074 846 (unsigned long)maddr_to_bootstrap_virt(s),
kfraser@15074 847 s >> PAGE_SHIFT, (map_e-s) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kfraser@15074 848
kfraser@15074 849 init_boot_pages(s, e);
kfraser@15074 850 }
kfraser@15074 851
kaf24@5003 852 memguard_init();
kaf24@4950 853
kfraser@15074 854 nr_pages = 0;
kfraser@15074 855 for ( i = 0; i < e820.nr_map; i++ )
kfraser@15074 856 if ( e820.map[i].type == E820_RAM )
kfraser@15074 857 nr_pages += e820.map[i].size >> PAGE_SHIFT;
ian@12681 858 printk("System RAM: %luMB (%lukB)\n",
kaf24@3354 859 nr_pages >> (20 - PAGE_SHIFT),
kaf24@3354 860 nr_pages << (PAGE_SHIFT - 10));
kaf24@7220 861 total_pages = nr_pages;
kaf24@3354 862
kfraser@11296 863 /* Sanity check for unwanted bloat of certain hypercall structures. */
kfraser@11296 864 BUILD_BUG_ON(sizeof(((struct xen_platform_op *)0)->u) !=
kfraser@11296 865 sizeof(((struct xen_platform_op *)0)->u.pad));
kfraser@11296 866 BUILD_BUG_ON(sizeof(((struct xen_domctl *)0)->u) !=
kfraser@11296 867 sizeof(((struct xen_domctl *)0)->u.pad));
kfraser@11296 868 BUILD_BUG_ON(sizeof(((struct xen_sysctl *)0)->u) !=
kfraser@11296 869 sizeof(((struct xen_sysctl *)0)->u.pad));
kaf24@7388 870
kaf24@9878 871 BUILD_BUG_ON(sizeof(start_info_t) > PAGE_SIZE);
kaf24@9878 872 BUILD_BUG_ON(sizeof(shared_info_t) > PAGE_SIZE);
ack@13292 873 BUILD_BUG_ON(sizeof(struct vcpu_info) != 64);
kaf24@7744 874
ack@13291 875 #ifdef CONFIG_COMPAT
ack@13291 876 BUILD_BUG_ON(sizeof(((struct compat_platform_op *)0)->u) !=
ack@13291 877 sizeof(((struct compat_platform_op *)0)->u.pad));
ack@13291 878 BUILD_BUG_ON(sizeof(start_info_compat_t) > PAGE_SIZE);
ack@13292 879 BUILD_BUG_ON(sizeof(struct compat_vcpu_info) != 64);
ack@13291 880 #endif
ack@13291 881
kfraser@10492 882 /* Check definitions in public headers match internal defs. */
kaf24@9878 883 BUILD_BUG_ON(__HYPERVISOR_VIRT_START != HYPERVISOR_VIRT_START);
kaf24@8521 884 #ifdef HYPERVISOR_VIRT_END
kaf24@9878 885 BUILD_BUG_ON(__HYPERVISOR_VIRT_END != HYPERVISOR_VIRT_END);
kaf24@8521 886 #endif
kfraser@10492 887 BUILD_BUG_ON(MACH2PHYS_VIRT_START != RO_MPT_VIRT_START);
kfraser@10492 888 BUILD_BUG_ON(MACH2PHYS_VIRT_END != RO_MPT_VIRT_END);
kaf24@8521 889
kaf24@3354 890 init_frametable();
kaf24@3338 891
kfraser@11971 892 acpi_boot_table_init();
kfraser@11971 893
kfraser@11971 894 acpi_numa_init();
kfraser@11971 895
kfraser@11971 896 numa_initmem_init(0, max_page);
kfraser@11971 897
kaf24@6111 898 /* Initialise the Xen heap, skipping RAM holes. */
kfraser@15074 899 init_xenheap_pages(xenheap_phys_start, xenheap_phys_end);
kfraser@15074 900 nr_pages = (xenheap_phys_end - xenheap_phys_start) >> PAGE_SHIFT;
kfraser@15074 901 #ifdef __x86_64__
kfraser@15074 902 init_xenheap_pages(xen_phys_start, __pa(&_start));
kfraser@15074 903 nr_pages += (__pa(&_start) - xen_phys_start) >> PAGE_SHIFT;
kfraser@15074 904 #endif
kfraser@15074 905 xenheap_phys_start = xen_phys_start;
kaf24@6111 906 printk("Xen heap: %luMB (%lukB)\n",
kaf24@6111 907 nr_pages >> (20 - PAGE_SHIFT),
kaf24@6111 908 nr_pages << (PAGE_SHIFT - 10));
kaf24@3338 909
keir@14680 910 end_boot_allocator();
keir@14680 911
kaf24@3594 912 early_boot = 0;
kaf24@3338 913
kaf24@8459 914 early_cpu_init();
kaf24@8459 915
kaf24@8459 916 paging_init();
kaf24@8459 917
kaf24@8459 918 /* Unmap the first page of CPU0's stack. */
kaf24@8459 919 memguard_guard_stack(cpu0_stack);
kaf24@8459 920
kaf24@8459 921 open_softirq(NEW_TLBFLUSH_CLOCK_PERIOD_SOFTIRQ, new_tlbflush_clock_period);
kaf24@8459 922
kaf24@8459 923 if ( opt_watchdog )
kaf24@8459 924 nmi_watchdog = NMI_LOCAL_APIC;
kaf24@8459 925
kaf24@8459 926 sort_exception_tables();
kaf24@8459 927
kaf24@8459 928 find_smp_config();
kaf24@8459 929
kaf24@8459 930 dmi_scan_machine();
kaf24@8459 931
kaf24@8459 932 generic_apic_probe();
kaf24@8459 933
kaf24@8459 934 acpi_boot_init();
kaf24@8459 935
kfraser@11971 936 init_cpu_to_node();
kfraser@11971 937
kfraser@11241 938 if ( smp_found_config )
kaf24@8459 939 get_smp_config();
kaf24@8459 940
keir@15083 941 #ifdef CONFIG_X86_64
keir@15083 942 /* Low mappings were only needed for some BIOS table parsing. */
keir@15083 943 zap_low_mappings();
keir@15083 944 #endif
keir@15083 945
kaf24@8459 946 init_apic_mappings();
kaf24@8459 947
kaf24@8459 948 init_IRQ();
kaf24@8459 949
kfraser@11241 950 percpu_init_areas();
kfraser@11241 951
kfraser@11240 952 init_idle_domain();
kfraser@11240 953
kaf24@8459 954 trap_init();
kaf24@8459 955
kaf24@13662 956 rcu_init();
kaf24@13662 957
kaf24@8586 958 timer_init();
kaf24@8459 959
kaf24@8459 960 early_time_init();
kaf24@8459 961
kaf24@8459 962 arch_init_memory();
kaf24@8459 963
kaf24@8459 964 identify_cpu(&boot_cpu_data);
kaf24@8459 965 if ( cpu_has_fxsr )
kaf24@8459 966 set_in_cr4(X86_CR4_OSFXSR);
kaf24@8459 967 if ( cpu_has_xmm )
kaf24@8459 968 set_in_cr4(X86_CR4_OSXMMEXCPT);
kaf24@8459 969
kaf24@8459 970 if ( opt_nosmp )
kaf24@8459 971 max_cpus = 0;
kaf24@8459 972
kaf24@8459 973 smp_prepare_cpus(max_cpus);
kaf24@8459 974
kaf24@8459 975 /*
kaf24@8459 976 * Initialise higher-level timer functions. We do this fairly late
kaf24@8459 977 * (post-SMP) because the time bases and scale factors need to be updated
kaf24@8459 978 * regularly, and SMP initialisation can cause a long delay with
kaf24@8459 979 * interrupts not yet enabled.
kaf24@8459 980 */
kaf24@8459 981 init_xen_time();
kaf24@8459 982
kaf24@8459 983 initialize_keytable();
kaf24@8459 984
kaf24@8459 985 serial_init_postirq();
kaf24@8459 986
kaf24@8459 987 BUG_ON(!local_irq_is_enabled());
kaf24@8459 988
kaf24@8459 989 for_each_present_cpu ( i )
kaf24@8459 990 {
kaf24@8459 991 if ( num_online_cpus() >= max_cpus )
kaf24@8459 992 break;
kaf24@8459 993 if ( !cpu_online(i) )
kaf24@13662 994 {
kaf24@13662 995 rcu_online_cpu(i);
kaf24@8459 996 __cpu_up(i);
kaf24@13662 997 }
kfraser@11971 998
kfraser@11998 999 /* Set up cpu_to_node[]. */
kfraser@11971 1000 srat_detect_node(i);
kfraser@11998 1001 /* Set up node_to_cpumask based on cpu_to_node[]. */
kfraser@11971 1002 numa_add_cpu(i);
kaf24@8459 1003 }
kaf24@8459 1004
kaf24@8459 1005 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
kaf24@8459 1006 smp_cpus_done(max_cpus);
kaf24@8459 1007
kaf24@9117 1008 initialise_gdb(); /* could be moved earlier */
kaf24@9117 1009
kaf24@8459 1010 do_initcalls();
kaf24@8459 1011
kaf24@8594 1012 if ( opt_watchdog )
kaf24@8594 1013 watchdog_enable();
kaf24@8459 1014
kfraser@11881 1015 /* Extract policy from multiboot. */
kfraser@11881 1016 extract_acm_policy(mbi, &initrdidx, &_policy_start, &_policy_len);
kfraser@11881 1017
kaf24@8459 1018 /* initialize access control security module */
kfraser@11881 1019 acm_init(_policy_start, _policy_len);
kaf24@8459 1020
kaf24@8459 1021 /* Create initial domain 0. */
kfraser@14911 1022 dom0 = domain_create(0, 0, DOM0_SSIDREF);
kfraser@10655 1023 if ( (dom0 == NULL) || (alloc_vcpu(dom0, 0, 0) == NULL) )
kaf24@8459 1024 panic("Error creating domain 0\n");
kaf24@8459 1025
kfraser@12210 1026 dom0->is_privileged = 1;
kfraser@12210 1027
kaf24@8459 1028 /* Grab the DOM0 command line. */
kaf24@8459 1029 cmdline = (char *)(mod[0].string ? __va(mod[0].string) : NULL);
kfraser@15426 1030 if ( (cmdline != NULL) || (kextra != NULL) )
kaf24@8459 1031 {
kaf24@8459 1032 static char dom0_cmdline[MAX_GUEST_CMDLINE];
kaf24@8459 1033
kfraser@15426 1034 dom0_cmdline[0] = '\0';
kfraser@15426 1035
kfraser@15426 1036 if ( cmdline != NULL )
kaf24@8459 1037 {
kfraser@15426 1038 /* Skip past the image name and copy to a local buffer. */
kaf24@8459 1039 while ( *cmdline == ' ' ) cmdline++;
kfraser@15426 1040 if ( (cmdline = strchr(cmdline, ' ')) != NULL )
kfraser@15426 1041 {
kfraser@15426 1042 while ( *cmdline == ' ' ) cmdline++;
kfraser@15426 1043 safe_strcpy(dom0_cmdline, cmdline);
kfraser@15426 1044 }
kaf24@8459 1045 }
kaf24@8459 1046
kfraser@15426 1047 if ( kextra != NULL )
kfraser@15426 1048 /* kextra always includes exactly one leading space. */
kfraser@15426 1049 safe_strcat(dom0_cmdline, kextra);
kfraser@15426 1050
kaf24@8459 1051 /* Append any extra parameters. */
kfraser@13691 1052 if ( skip_ioapic_setup && !strstr(dom0_cmdline, "noapic") )
kfraser@13691 1053 safe_strcat(dom0_cmdline, " noapic");
kaf24@8459 1054 if ( acpi_skip_timer_override &&
kfraser@13691 1055 !strstr(dom0_cmdline, "acpi_skip_timer_override") )
kfraser@13691 1056 safe_strcat(dom0_cmdline, " acpi_skip_timer_override");
kfraser@13691 1057 if ( (strlen(acpi_param) != 0) && !strstr(dom0_cmdline, "acpi=") )
kaf24@8459 1058 {
kfraser@13691 1059 safe_strcat(dom0_cmdline, " acpi=");
kfraser@13691 1060 safe_strcat(dom0_cmdline, acpi_param);
kaf24@8459 1061 }
kfraser@13691 1062
kfraser@13691 1063 cmdline = dom0_cmdline;
kaf24@8459 1064 }
kaf24@8459 1065
kaf24@8459 1066 if ( (initrdidx > 0) && (initrdidx < mbi->mods_count) )
kaf24@8459 1067 {
kaf24@8459 1068 _initrd_start = initial_images_start +
kaf24@8459 1069 (mod[initrdidx].mod_start - mod[0].mod_start);
kaf24@8459 1070 _initrd_len = mod[initrdidx].mod_end - mod[initrdidx].mod_start;
kaf24@8459 1071 }
kaf24@8459 1072
kaf24@8459 1073 /*
kaf24@8459 1074 * We're going to setup domain0 using the module(s) that we stashed safely
kaf24@8459 1075 * above our heap. The second module, if present, is an initrd ramdisk.
kaf24@8459 1076 */
kaf24@8459 1077 if ( construct_dom0(dom0,
kaf24@8459 1078 initial_images_start,
kaf24@8459 1079 mod[0].mod_end-mod[0].mod_start,
kaf24@8459 1080 _initrd_start,
kaf24@8459 1081 _initrd_len,
kaf24@8459 1082 cmdline) != 0)
kaf24@8459 1083 panic("Could not set up DOM0 guest OS\n");
kaf24@8459 1084
kaf24@8459 1085 /* Scrub RAM that is still free and so may go to an unprivileged domain. */
kaf24@8459 1086 scrub_heap_pages();
kaf24@8459 1087
kaf24@8459 1088 init_trace_bufs();
kaf24@8459 1089
kaf24@10502 1090 console_endboot();
kaf24@8459 1091
kaf24@8459 1092 /* Hide UART from DOM0 if we're using it */
kaf24@8459 1093 serial_endboot();
kaf24@8459 1094
kaf24@8459 1095 domain_unpause_by_systemcontroller(dom0);
kaf24@8459 1096
keir@15082 1097 reset_stack_and_jump(init_done);
kaf24@8459 1098 }
kaf24@8459 1099
ian@13763 1100 void arch_get_xen_caps(xen_capabilities_info_t *info)
iap10@6721 1101 {
kfraser@14997 1102 /* Interface name is always xen-3.0-* for Xen-3.x. */
kfraser@14997 1103 int major = 3, minor = 0;
keir@13754 1104 char s[32];
keir@13754 1105
ian@13763 1106 (*info)[0] = '\0';
iap10@6721 1107
kaf24@6725 1108 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
kaf24@6725 1109
keir@13754 1110 snprintf(s, sizeof(s), "xen-%d.%d-x86_32 ", major, minor);
ian@13763 1111 safe_strcat(*info, s);
keir@13754 1112 if ( hvm_enabled )
keir@13754 1113 {
keir@13754 1114 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1115 safe_strcat(*info, s);
kfraser@13685 1116 }
kaf24@6725 1117
kaf24@6725 1118 #elif defined(CONFIG_X86_32) && defined(CONFIG_X86_PAE)
kaf24@6725 1119
keir@13754 1120 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1121 safe_strcat(*info, s);
kaf24@6725 1122 if ( hvm_enabled )
iap10@6721 1123 {
keir@13754 1124 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1125 safe_strcat(*info, s);
keir@13754 1126 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1127 safe_strcat(*info, s);
iap10@6721 1128 }
iap10@6721 1129
kaf24@6725 1130 #elif defined(CONFIG_X86_64)
iap10@6721 1131
keir@13754 1132 snprintf(s, sizeof(s), "xen-%d.%d-x86_64 ", major, minor);
ian@13763 1133 safe_strcat(*info, s);
ack@13288 1134 #ifdef CONFIG_COMPAT
keir@13754 1135 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1136 safe_strcat(*info, s);
ack@13288 1137 #endif
kaf24@6725 1138 if ( hvm_enabled )
iap10@6721 1139 {
keir@13754 1140 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1141 safe_strcat(*info, s);
keir@13754 1142 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1143 safe_strcat(*info, s);
keir@13754 1144 snprintf(s, sizeof(s), "hvm-%d.%d-x86_64 ", major, minor);
ian@13763 1145 safe_strcat(*info, s);
iap10@6721 1146 }
kaf24@6725 1147
iap10@6721 1148 #endif
iap10@6721 1149 }
iap10@6721 1150
kaf24@3914 1151 /*
kaf24@3914 1152 * Local variables:
kaf24@3914 1153 * mode: C
kaf24@3914 1154 * c-set-style: "BSD"
kaf24@3914 1155 * c-basic-offset: 4
kaf24@3914 1156 * tab-width: 4
kaf24@3914 1157 * indent-tabs-mode: nil
kaf24@3988 1158 * End:
kaf24@3914 1159 */