ia64/xen-unstable

annotate xen/arch/x86/setup.c @ 15489:0900fb1a3693

xen: Fix kexec_crash_area reservation/allocation.
Signed-off-by: Keir Fraser <keir@xensource.com>
author kfraser@localhost.localdomain
date Tue Jul 03 11:24:40 2007 +0100 (2007-07-03)
parents 296fd2598e00
children 83cbda5c1e1b
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@15074 40 #define BOOTSTRAP_DIRECTMAP_END (1UL << 32)
kfraser@15074 41 #define maddr_to_bootstrap_virt(m) maddr_to_virt(m)
kfraser@15074 42 #else
kfraser@15074 43 #define BOOTSTRAP_DIRECTMAP_END HYPERVISOR_VIRT_START
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@15074 112 char __attribute__ ((__section__(".bss.page_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@15074 285 static void __init move_memory(
kfraser@15074 286 unsigned long dst, unsigned long src_start, unsigned long src_end)
ian@12677 287 {
kfraser@15074 288 memmove(maddr_to_bootstrap_virt(dst),
kfraser@15074 289 maddr_to_bootstrap_virt(src_start),
ian@12677 290 src_end - src_start);
kfraser@15074 291 }
kfraser@15074 292
kfraser@15074 293 /* A temporary copy of the e820 map that we can mess with during bootstrap. */
kfraser@15074 294 static struct e820map __initdata boot_e820;
kfraser@15074 295
kfraser@15074 296 /* Reserve area (@s,@e) in the temporary bootstrap e820 map. */
kfraser@15489 297 static int __init reserve_in_boot_e820(unsigned long s, unsigned long e)
kfraser@15074 298 {
kfraser@15403 299 uint64_t rs, re;
kfraser@15074 300 int i;
kfraser@15074 301
kfraser@15074 302 for ( i = 0; i < boot_e820.nr_map; i++ )
kfraser@15074 303 {
kfraser@15074 304 /* Have we found the e820 region that includes the specified range? */
kfraser@15074 305 rs = boot_e820.map[i].addr;
kfraser@15403 306 re = rs + boot_e820.map[i].size;
kfraser@15489 307 if ( (s >= rs) && (e <= re) )
kfraser@15489 308 goto found;
kfraser@15489 309 }
kfraser@15074 310
kfraser@15489 311 return 0;
kfraser@15489 312
kfraser@15489 313 found:
kfraser@15489 314 /* Start fragment. */
kfraser@15489 315 boot_e820.map[i].size = s - rs;
kfraser@15489 316
kfraser@15489 317 /* End fragment. */
kfraser@15489 318 if ( e < re )
kfraser@15489 319 {
kfraser@15489 320 memmove(&boot_e820.map[i+1], &boot_e820.map[i],
kfraser@15489 321 (boot_e820.nr_map-i) * sizeof(boot_e820.map[0]));
kfraser@15489 322 boot_e820.nr_map++;
kfraser@15489 323 i++;
kfraser@15489 324 boot_e820.map[i].addr = e;
kfraser@15489 325 boot_e820.map[i].size = re - e;
kfraser@15074 326 }
kfraser@15489 327
kfraser@15489 328 return 1;
ian@12677 329 }
ian@12677 330
keir@15298 331 struct boot_video_info {
keir@15298 332 u8 orig_x; /* 0x00 */
keir@15298 333 u8 orig_y; /* 0x01 */
keir@15298 334 u8 orig_video_mode; /* 0x02 */
keir@15298 335 u8 orig_video_cols; /* 0x03 */
keir@15298 336 u8 orig_video_lines; /* 0x04 */
keir@15298 337 u8 orig_video_isVGA; /* 0x05 */
keir@15298 338 u16 orig_video_points; /* 0x06 */
keir@15298 339
keir@15298 340 /* VESA graphic mode -- linear frame buffer */
keir@15298 341 u32 capabilities; /* 0x08 */
keir@15298 342 u16 lfb_linelength; /* 0x0c */
keir@15298 343 u16 lfb_width; /* 0x0e */
keir@15298 344 u16 lfb_height; /* 0x10 */
keir@15298 345 u16 lfb_depth; /* 0x12 */
keir@15298 346 u32 lfb_base; /* 0x14 */
keir@15298 347 u32 lfb_size; /* 0x18 */
keir@15298 348 u8 red_size; /* 0x1c */
keir@15298 349 u8 red_pos; /* 0x1d */
keir@15298 350 u8 green_size; /* 0x1e */
keir@15298 351 u8 green_pos; /* 0x1f */
keir@15298 352 u8 blue_size; /* 0x20 */
keir@15298 353 u8 blue_pos; /* 0x21 */
keir@15298 354 u8 rsvd_size; /* 0x22 */
keir@15298 355 u8 rsvd_pos; /* 0x23 */
keir@15298 356 u16 vesapm_seg; /* 0x24 */
keir@15298 357 u16 vesapm_off; /* 0x26 */
keir@15298 358 };
keir@15298 359
keir@15298 360 static void __init parse_video_info(void)
keir@15298 361 {
keir@15298 362 struct boot_video_info *bvi = &bootsym(boot_vid_info);
keir@15298 363
keir@15298 364 if ( (bvi->orig_video_isVGA == 1) && (bvi->orig_video_mode == 3) )
keir@15298 365 {
keir@15298 366 vga_console_info.video_type = XEN_VGATYPE_TEXT_MODE_3;
keir@15298 367 vga_console_info.u.text_mode_3.font_height = bvi->orig_video_points;
keir@15298 368 vga_console_info.u.text_mode_3.cursor_x = bvi->orig_x;
keir@15298 369 vga_console_info.u.text_mode_3.cursor_y = bvi->orig_y;
keir@15298 370 vga_console_info.u.text_mode_3.rows = bvi->orig_video_lines;
keir@15298 371 vga_console_info.u.text_mode_3.columns = bvi->orig_video_cols;
keir@15298 372 }
keir@15298 373 else if ( bvi->orig_video_isVGA == 0x23 )
keir@15298 374 {
keir@15298 375 vga_console_info.video_type = XEN_VGATYPE_VESA_LFB;
keir@15298 376 vga_console_info.u.vesa_lfb.width = bvi->lfb_width;
keir@15298 377 vga_console_info.u.vesa_lfb.height = bvi->lfb_height;
keir@15298 378 vga_console_info.u.vesa_lfb.bytes_per_line = bvi->lfb_linelength;
keir@15298 379 vga_console_info.u.vesa_lfb.bits_per_pixel = bvi->lfb_depth;
keir@15298 380 vga_console_info.u.vesa_lfb.lfb_base = bvi->lfb_base;
keir@15298 381 vga_console_info.u.vesa_lfb.lfb_size = bvi->lfb_size;
keir@15298 382 vga_console_info.u.vesa_lfb.red_pos = bvi->red_pos;
keir@15298 383 vga_console_info.u.vesa_lfb.red_size = bvi->red_size;
keir@15298 384 vga_console_info.u.vesa_lfb.green_pos = bvi->green_pos;
keir@15298 385 vga_console_info.u.vesa_lfb.green_size = bvi->green_size;
keir@15298 386 vga_console_info.u.vesa_lfb.blue_pos = bvi->blue_pos;
keir@15298 387 vga_console_info.u.vesa_lfb.blue_size = bvi->blue_size;
keir@15298 388 vga_console_info.u.vesa_lfb.rsvd_pos = bvi->rsvd_pos;
keir@15298 389 vga_console_info.u.vesa_lfb.rsvd_size = bvi->rsvd_size;
keir@15298 390 }
keir@15298 391 }
keir@15298 392
keir@15082 393 void init_done(void)
keir@15082 394 {
keir@15082 395 extern char __init_begin[], __init_end[];
keir@15082 396
keir@15082 397 /* Free (or page-protect) the init areas. */
keir@15082 398 #ifndef MEMORY_GUARD
keir@15082 399 init_xenheap_pages(__pa(__init_begin), __pa(__init_end));
keir@15082 400 #endif
keir@15082 401 memguard_guard_range(__init_begin, __init_end - __init_begin);
keir@15082 402 #if defined(CONFIG_X86_64)
keir@15082 403 /* Also zap the mapping in the 1:1 area. */
keir@15082 404 memguard_guard_range(__va(__pa(__init_begin)), __init_end - __init_begin);
keir@15082 405 #endif
keir@15082 406 printk("Freed %ldkB init memory.\n", (long)(__init_end-__init_begin)>>10);
keir@15082 407
keir@15082 408 startup_cpu_idle_loop();
keir@15082 409 }
keir@15082 410
kfraser@15379 411 void __init __start_xen(unsigned long mbi_p)
kaf24@1452 412 {
kfraser@15293 413 char *memmap_type = NULL;
kfraser@15426 414 char __cmdline[] = "", *cmdline = __cmdline, *kextra;
kaf24@8457 415 unsigned long _initrd_start = 0, _initrd_len = 0;
kaf24@8457 416 unsigned int initrdidx = 1;
kfraser@11881 417 char *_policy_start = NULL;
kfraser@11881 418 unsigned long _policy_len = 0;
kfraser@15379 419 multiboot_info_t *mbi = __va(mbi_p);
kaf24@8457 420 module_t *mod = (module_t *)__va(mbi->mods_addr);
kaf24@6111 421 unsigned long nr_pages, modules_length;
kfraser@15293 422 int i, e820_warn = 0, bytes = 0;
kaf24@5776 423 struct ns16550_defaults ns16550 = {
kaf24@5776 424 .data_bits = 8,
kaf24@5776 425 .parity = 'n',
kaf24@5776 426 .stop_bits = 1
kaf24@5776 427 };
kaf24@3338 428
kfraser@12853 429 extern void early_page_fault(void);
kfraser@12853 430 set_intr_gate(TRAP_page_fault, &early_page_fault);
kfraser@12853 431
kaf24@3338 432 /* Parse the command-line options. */
kaf24@3344 433 if ( (mbi->flags & MBI_CMDLINE) && (mbi->cmdline != 0) )
kaf24@9823 434 cmdline = __va(mbi->cmdline);
kfraser@15426 435 if ( (kextra = strstr(cmdline, " -- ")) != NULL )
kfraser@15426 436 {
kfraser@15426 437 /*
kfraser@15426 438 * Options after ' -- ' separator belong to dom0.
kfraser@15426 439 * 1. Orphan dom0's options from Xen's command line.
kfraser@15426 440 * 2. Skip all but final leading space from dom0's options.
kfraser@15426 441 */
kfraser@15426 442 *kextra = '\0';
kfraser@15426 443 kextra += 3;
kfraser@15426 444 while ( kextra[1] == ' ' ) kextra++;
kfraser@15426 445 }
kaf24@9823 446 cmdline_parse(cmdline);
kaf24@3338 447
keir@15298 448 parse_video_info();
keir@15298 449
kaf24@8534 450 set_current((struct vcpu *)0xfffff000); /* debug sanity */
kfraser@11240 451 idle_vcpu[0] = current;
kaf24@8534 452 set_processor_id(0); /* needed early, for smp_processor_id() */
kaf24@3338 453
kaf24@5146 454 smp_prepare_boot_cpu();
kaf24@5146 455
kaf24@3338 456 /* We initialise the serial devices very early so we can get debugging. */
kaf24@5776 457 ns16550.io_base = 0x3f8;
kaf24@5776 458 ns16550.irq = 4;
kaf24@5776 459 ns16550_init(0, &ns16550);
kaf24@5776 460 ns16550.io_base = 0x2f8;
kaf24@5776 461 ns16550.irq = 3;
kaf24@5776 462 ns16550_init(1, &ns16550);
kaf24@5195 463 serial_init_preirq();
kaf24@3338 464
kaf24@3338 465 init_console();
kaf24@3338 466
kfraser@11947 467 printk("Command line: %s\n", cmdline);
kaf24@9823 468
kfraser@15330 469 printk("Video information:\n");
kfraser@15330 470
kfraser@15330 471 /* Print VGA display mode information. */
keir@15298 472 switch ( vga_console_info.video_type )
keir@15298 473 {
keir@15298 474 case XEN_VGATYPE_TEXT_MODE_3:
kfraser@15330 475 printk(" VGA is text mode %dx%d, font 8x%d\n",
keir@15298 476 vga_console_info.u.text_mode_3.columns,
keir@15298 477 vga_console_info.u.text_mode_3.rows,
keir@15298 478 vga_console_info.u.text_mode_3.font_height);
keir@15298 479 break;
keir@15298 480 case XEN_VGATYPE_VESA_LFB:
kfraser@15330 481 printk(" VGA is graphics mode %dx%d, %d bpp\n",
keir@15298 482 vga_console_info.u.vesa_lfb.width,
keir@15298 483 vga_console_info.u.vesa_lfb.height,
keir@15298 484 vga_console_info.u.vesa_lfb.bits_per_pixel);
keir@15298 485 break;
kfraser@15330 486 default:
kfraser@15330 487 printk(" No VGA detected\n");
kfraser@15330 488 break;
kfraser@15330 489 }
kfraser@15330 490
kfraser@15330 491 /* Print VBE/DDC EDID information. */
kfraser@15330 492 if ( bootsym(boot_edid_caps) != 0x1313 )
kfraser@15330 493 {
kfraser@15330 494 u16 caps = bootsym(boot_edid_caps);
kfraser@15330 495 printk(" VBE/DDC methods:%s%s%s; ",
kfraser@15330 496 (caps & 1) ? " V1" : "",
kfraser@15330 497 (caps & 2) ? " V2" : "",
kfraser@15330 498 !(caps & 3) ? " none" : "");
kfraser@15330 499 printk("EDID transfer time: %d seconds\n", caps >> 8);
kfraser@15330 500 if ( *(u32 *)bootsym(boot_edid_info) == 0x13131313 )
kfraser@15330 501 {
kfraser@15330 502 printk(" EDID info not retrieved because ");
kfraser@15330 503 if ( !(caps & 3) )
kfraser@15330 504 printk("no DDC retrieval method detected\n");
kfraser@15330 505 else if ( (caps >> 8) > 5 )
kfraser@15330 506 printk("takes longer than 5 seconds\n");
kfraser@15330 507 else
kfraser@15330 508 printk("of reasons unknown\n");
kfraser@15330 509 }
keir@15298 510 }
keir@15298 511
kfraser@15336 512 printk("Disc information:\n");
kfraser@15336 513 printk(" Found %d MBR signatures\n",
kfraser@15430 514 bootsym(boot_mbr_signature_nr));
kfraser@15336 515 printk(" Found %d EDD information structures\n",
kfraser@15336 516 bootsym(boot_edd_info_nr));
kfraser@15336 517
kaf24@3344 518 /* Check that we have at least one Multiboot module. */
kaf24@3344 519 if ( !(mbi->flags & MBI_MODULES) || (mbi->mods_count == 0) )
kfraser@15074 520 EARLY_FAIL("dom0 kernel not specified. "
kfraser@15074 521 "Check bootloader configuration.\n");
kaf24@5011 522
kaf24@5011 523 if ( ((unsigned long)cpu0_stack & (STACK_SIZE-1)) != 0 )
kfraser@15074 524 EARLY_FAIL("Misaligned CPU0 stack.\n");
kaf24@3338 525
kfraser@11618 526 /*
kfraser@11618 527 * Since there are some stubs getting built on the stacks which use
kfraser@11618 528 * direct calls/jumps, the heap must be confined to the lower 2G so
kfraser@11618 529 * that those branches can reach their targets.
kfraser@11618 530 */
kfraser@11618 531 if ( opt_xenheap_megabytes > 2048 )
kfraser@11618 532 opt_xenheap_megabytes = 2048;
kaf24@3338 533
kfraser@15293 534 if ( e820_raw_nr != 0 )
kfraser@15292 535 {
kfraser@15293 536 memmap_type = "Xen-e820";
kfraser@15292 537 }
kfraser@15293 538 else if ( bootsym(lowmem_kb) )
kfraser@15292 539 {
kfraser@15293 540 memmap_type = "Xen-e801";
kfraser@15292 541 e820_raw[0].addr = 0;
kfraser@15293 542 e820_raw[0].size = bootsym(lowmem_kb) << 10;
kfraser@15292 543 e820_raw[0].type = E820_RAM;
kfraser@15292 544 e820_raw[1].addr = 0x100000;
kfraser@15293 545 e820_raw[1].size = bootsym(highmem_kb) << 10;
kfraser@15292 546 e820_raw[1].type = E820_RAM;
kfraser@15292 547 e820_raw_nr = 2;
kfraser@15292 548 }
kfraser@15292 549 else if ( mbi->flags & MBI_MEMMAP )
kaf24@3344 550 {
kfraser@15293 551 memmap_type = "Multiboot-e820";
kaf24@3344 552 while ( bytes < mbi->mmap_length )
kaf24@3344 553 {
kaf24@3344 554 memory_map_t *map = __va(mbi->mmap_addr + bytes);
kaf24@8402 555
kaf24@8402 556 /*
kaf24@8403 557 * This is a gross workaround for a BIOS bug. Some bootloaders do
kaf24@8402 558 * not write e820 map entries into pre-zeroed memory. This is
kaf24@8402 559 * okay if the BIOS fills in all fields of the map entry, but
kaf24@8402 560 * some broken BIOSes do not bother to write the high word of
kaf24@8402 561 * the length field if the length is smaller than 4GB. We
kaf24@8402 562 * detect and fix this by flagging sections below 4GB that
kaf24@8403 563 * appear to be larger than 4GB in size.
kaf24@8402 564 */
kaf24@8403 565 if ( (map->base_addr_high == 0) && (map->length_high != 0) )
kaf24@8402 566 {
kfraser@15292 567 if ( !e820_warn )
kfraser@15292 568 {
kfraser@15292 569 printk("WARNING: Buggy e820 map detected and fixed "
kfraser@15292 570 "(truncated length fields).\n");
kfraser@15292 571 e820_warn = 1;
kfraser@15292 572 }
kaf24@8402 573 map->length_high = 0;
kaf24@8402 574 }
kaf24@8402 575
kaf24@3344 576 e820_raw[e820_raw_nr].addr =
kaf24@3344 577 ((u64)map->base_addr_high << 32) | (u64)map->base_addr_low;
kaf24@3344 578 e820_raw[e820_raw_nr].size =
kaf24@3344 579 ((u64)map->length_high << 32) | (u64)map->length_low;
kaf24@3344 580 e820_raw[e820_raw_nr].type =
kfraser@12226 581 (map->type > E820_NVS) ? E820_RESERVED : map->type;
kaf24@3344 582 e820_raw_nr++;
kaf24@8402 583
kaf24@3344 584 bytes += map->size + 4;
kaf24@3344 585 }
kaf24@3344 586 }
kaf24@3344 587 else if ( mbi->flags & MBI_MEMLIMITS )
kaf24@3344 588 {
kfraser@15293 589 memmap_type = "Multiboot-e801";
kaf24@3344 590 e820_raw[0].addr = 0;
kaf24@3344 591 e820_raw[0].size = mbi->mem_lower << 10;
kaf24@3344 592 e820_raw[0].type = E820_RAM;
kaf24@3354 593 e820_raw[1].addr = 0x100000;
kaf24@3354 594 e820_raw[1].size = mbi->mem_upper << 10;
kaf24@3354 595 e820_raw[1].type = E820_RAM;
kaf24@3344 596 e820_raw_nr = 2;
kaf24@3344 597 }
kaf24@3344 598 else
kaf24@3344 599 {
kfraser@15074 600 EARLY_FAIL("Bootloader provided no memory information.\n");
kaf24@3344 601 }
kaf24@3344 602
kaf24@13427 603 /* Ensure that all E820 RAM regions are page-aligned and -sized. */
kaf24@13427 604 for ( i = 0; i < e820_raw_nr; i++ )
kaf24@13427 605 {
kaf24@13427 606 uint64_t s, e;
kfraser@15292 607
kaf24@13427 608 if ( e820_raw[i].type != E820_RAM )
kaf24@13427 609 continue;
kaf24@13427 610 s = PFN_UP(e820_raw[i].addr);
kaf24@13427 611 e = PFN_DOWN(e820_raw[i].addr + e820_raw[i].size);
kaf24@13427 612 e820_raw[i].size = 0; /* discarded later */
kaf24@13427 613 if ( s < e )
kaf24@13427 614 {
kaf24@13427 615 e820_raw[i].addr = s << PAGE_SHIFT;
kaf24@13427 616 e820_raw[i].size = (e - s) << PAGE_SHIFT;
kaf24@13427 617 }
kaf24@13427 618 }
kaf24@13427 619
kaf24@13427 620 /* Sanitise the raw E820 map to produce a final clean version. */
kfraser@15293 621 max_page = init_e820(memmap_type, e820_raw, &e820_raw_nr);
kaf24@3338 622
kfraser@15074 623 /*
kfraser@15074 624 * Create a temporary copy of the E820 map. Truncate it to above 16MB
kfraser@15074 625 * as anything below that is already mapped and has a statically-allocated
kfraser@15074 626 * purpose.
kfraser@15074 627 */
kfraser@15074 628 memcpy(&boot_e820, &e820, sizeof(e820));
kfraser@15074 629 for ( i = 0; i < boot_e820.nr_map; i++ )
kaf24@3338 630 {
kfraser@15074 631 uint64_t s, e, min = 16 << 20; /* 16MB */
kfraser@15074 632 s = boot_e820.map[i].addr;
kfraser@15074 633 e = boot_e820.map[i].addr + boot_e820.map[i].size;
kfraser@15074 634 if ( s >= min )
kfraser@15074 635 continue;
kfraser@15074 636 if ( e > min )
kaf24@6111 637 {
kfraser@15074 638 boot_e820.map[i].addr = min;
kfraser@15074 639 boot_e820.map[i].size = e - min;
kaf24@6111 640 }
kfraser@15074 641 else
kfraser@15074 642 boot_e820.map[i].type = E820_RESERVED;
kaf24@3338 643 }
kaf24@6111 644
kfraser@15074 645 /*
keir@15077 646 * Iterate backwards over all superpage-aligned RAM regions.
kfraser@15074 647 *
kfraser@15074 648 * We require superpage alignment because the boot allocator is not yet
kfraser@15074 649 * initialised. Hence we can only map superpages in the address range
kfraser@15074 650 * 0 to BOOTSTRAP_DIRECTMAP_END, as this is guaranteed not to require
kfraser@15074 651 * dynamic allocation of pagetables.
kfraser@15074 652 *
kfraser@15074 653 * As well as mapping superpages in that range, in preparation for
kfraser@15074 654 * initialising the boot allocator, we also look for a region to which
kfraser@15074 655 * we can relocate the dom0 kernel and other multiboot modules. Also, on
kfraser@15074 656 * x86/64, we relocate Xen to higher memory.
kfraser@15074 657 */
kfraser@15074 658 modules_length = mod[mbi->mods_count-1].mod_end - mod[0].mod_start;
keir@15077 659 for ( i = boot_e820.nr_map-1; i >= 0; i-- )
kfraser@15074 660 {
kfraser@15074 661 uint64_t s, e, mask = (1UL << L2_PAGETABLE_SHIFT) - 1;
kaf24@6134 662
kfraser@15074 663 /* Superpage-aligned chunks up to BOOTSTRAP_DIRECTMAP_END, please. */
kfraser@15074 664 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 665 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
kfraser@15074 666 e = min_t(uint64_t, e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 667 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kaf24@3354 668 continue;
kaf24@6111 669
kfraser@15074 670 /* Map the chunk. No memory will need to be allocated to do this. */
kfraser@15074 671 map_pages_to_xen(
kfraser@15074 672 (unsigned long)maddr_to_bootstrap_virt(s),
kfraser@15074 673 s >> PAGE_SHIFT, (e-s) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kaf24@6111 674
kfraser@14084 675 #if defined(CONFIG_X86_64)
kfraser@15074 676 /* Is the region suitable for relocating Xen? */
kfraser@15074 677 if ( !xen_phys_start && (((e-s) >> 20) >= opt_xenheap_megabytes) )
kaf24@5003 678 {
kfraser@15074 679 extern l2_pgentry_t l2_xenmap[];
kfraser@15074 680 l4_pgentry_t *pl4e;
kfraser@15074 681 l3_pgentry_t *pl3e;
kfraser@15074 682 l2_pgentry_t *pl2e;
kfraser@15074 683 int i, j;
kfraser@15074 684
kfraser@15074 685 /* Select relocation address. */
kfraser@15074 686 e = (e - (opt_xenheap_megabytes << 20)) & ~mask;
kfraser@15074 687 xen_phys_start = e;
kfraser@15292 688 bootsym(trampoline_xen_phys_start) = e;
kfraser@15074 689
kfraser@15074 690 /*
kfraser@15074 691 * Perform relocation to new physical address.
kfraser@15074 692 * Before doing so we must sync static/global data with main memory
kfraser@15074 693 * with a barrier(). After this we must *not* modify static/global
kfraser@15074 694 * data until after we have switched to the relocated pagetables!
kfraser@15074 695 */
kfraser@15074 696 barrier();
kfraser@15074 697 move_memory(e, 0, __pa(&_end) - xen_phys_start);
kfraser@15074 698
kfraser@15379 699 /* Poison low 1MB to detect stray pointers to physical 0-1MB. */
kfraser@15379 700 memset(maddr_to_bootstrap_virt(e), 0x55, 1U<<20);
kfraser@15379 701
kfraser@15074 702 /* Walk initial pagetables, relocating page directory entries. */
kfraser@15074 703 pl4e = __va(__pa(idle_pg_table));
kfraser@15074 704 for ( i = 0 ; i < L4_PAGETABLE_ENTRIES; i++, pl4e++ )
kfraser@15074 705 {
kfraser@15074 706 if ( !(l4e_get_flags(*pl4e) & _PAGE_PRESENT) )
kfraser@15074 707 continue;
kfraser@15074 708 *pl4e = l4e_from_intpte(l4e_get_intpte(*pl4e) +
kfraser@15074 709 xen_phys_start);
kfraser@15074 710 pl3e = l4e_to_l3e(*pl4e);
kfraser@15074 711 for ( j = 0; j < L3_PAGETABLE_ENTRIES; j++, pl3e++ )
kfraser@15074 712 {
kfraser@15074 713 /* Not present or already relocated? */
kfraser@15074 714 if ( !(l3e_get_flags(*pl3e) & _PAGE_PRESENT) ||
kfraser@15074 715 (l3e_get_pfn(*pl3e) > 0x1000) )
kfraser@15074 716 continue;
kfraser@15074 717 *pl3e = l3e_from_intpte(l3e_get_intpte(*pl3e) +
kfraser@15074 718 xen_phys_start);
kfraser@15074 719 }
kfraser@15074 720 }
kfraser@15074 721
kfraser@15074 722 /* The only data mappings to be relocated are in the Xen area. */
kfraser@15074 723 pl2e = __va(__pa(l2_xenmap));
kfraser@15074 724 for ( i = 0; i < L2_PAGETABLE_ENTRIES; i++, pl2e++ )
kfraser@15074 725 {
kfraser@15074 726 if ( !(l2e_get_flags(*pl2e) & _PAGE_PRESENT) )
kfraser@15074 727 continue;
kfraser@15074 728 *pl2e = l2e_from_intpte(l2e_get_intpte(*pl2e) +
kfraser@15074 729 xen_phys_start);
kfraser@15074 730 }
kfraser@15074 731
kfraser@15074 732 /* Re-sync the stack and then switch to relocated pagetables. */
kfraser@15074 733 asm volatile (
kfraser@15074 734 "rep movsb ; " /* re-sync the stack */
kfraser@15074 735 "movq %%cr4,%%rsi ; "
kfraser@15074 736 "andb $0x7f,%%sil ; "
kfraser@15074 737 "movq %%rsi,%%cr4 ; " /* CR4.PGE == 0 */
kfraser@15074 738 "movq %0,%%cr3 ; " /* CR3 == new pagetables */
kfraser@15074 739 "orb $0x80,%%sil ; "
kfraser@15074 740 "movq %%rsi,%%cr4 " /* CR4.PGE == 1 */
kfraser@15074 741 : : "r" (__pa(idle_pg_table)), "S" (cpu0_stack),
kfraser@15074 742 "D" (__va(__pa(cpu0_stack))), "c" (STACK_SIZE) : "memory" );
kaf24@5003 743 }
kaf24@5003 744 #endif
keir@15077 745
keir@15077 746 /* Is the region suitable for relocating the multiboot modules? */
keir@15077 747 if ( !initial_images_start && (s < e) && ((e-s) >= modules_length) )
keir@15077 748 {
kfraser@15489 749 initial_images_end = e;
kfraser@15489 750 e = (e - modules_length) & PAGE_MASK;
keir@15077 751 initial_images_start = e;
keir@15077 752 move_memory(initial_images_start,
keir@15077 753 mod[0].mod_start, mod[mbi->mods_count-1].mod_end);
keir@15077 754 }
kfraser@15489 755
kfraser@15489 756 if ( !kexec_crash_area.start && (s < e) &&
kfraser@15489 757 ((e-s) >= kexec_crash_area.size) )
kfraser@15489 758 {
kfraser@15489 759 e = (e - kexec_crash_area.size) & PAGE_MASK;
kfraser@15489 760 kexec_crash_area.start = e;
kfraser@15489 761 }
kaf24@3354 762 }
kaf24@3354 763
kfraser@15074 764 if ( !initial_images_start )
kfraser@15074 765 EARLY_FAIL("Not enough memory to relocate the dom0 kernel image.\n");
kfraser@15074 766 reserve_in_boot_e820(initial_images_start, initial_images_end);
kfraser@15074 767
kfraser@15074 768 /*
kfraser@15074 769 * With modules (and Xen itself, on x86/64) relocated out of the way, we
kfraser@15074 770 * can now initialise the boot allocator with some memory.
kfraser@15074 771 */
kfraser@15074 772 xenheap_phys_start = init_boot_allocator(__pa(&_end));
kfraser@15074 773 xenheap_phys_end = opt_xenheap_megabytes << 20;
kfraser@15074 774 #if defined(CONFIG_X86_64)
kfraser@15074 775 if ( !xen_phys_start )
kfraser@15074 776 EARLY_FAIL("Not enough memory to relocate Xen.\n");
kfraser@15074 777 xenheap_phys_end += xen_phys_start;
kfraser@15074 778 reserve_in_boot_e820(xen_phys_start,
kfraser@15074 779 xen_phys_start + (opt_xenheap_megabytes<<20));
kfraser@15074 780 init_boot_pages(1<<20, 16<<20); /* Initial seed: 15MB */
kfraser@15074 781 #else
kfraser@15074 782 init_boot_pages(xenheap_phys_end, 16<<20); /* Initial seed: 4MB */
kfraser@15074 783 #endif
kfraser@15074 784
kfraser@15489 785 if ( kexec_crash_area.size != 0 )
kfraser@15489 786 {
kfraser@15489 787 unsigned long kdump_start = kexec_crash_area.start;
kfraser@15489 788 unsigned long kdump_size = kexec_crash_area.size;
kfraser@15489 789
kfraser@15489 790 kdump_size = (kdump_size + PAGE_SIZE - 1) & PAGE_MASK;
kfraser@15489 791
kfraser@15489 792 if ( !reserve_in_boot_e820(kdump_start, kdump_size) )
kfraser@15489 793 {
kfraser@15489 794 printk("Kdump: DISABLED (failed to reserve %luMB (%lukB) at 0x%lx)"
kfraser@15489 795 "\n", kdump_size >> 20, kdump_size >> 10, kdump_start);
kfraser@15489 796 kexec_crash_area.start = kexec_crash_area.size = 0;
kfraser@15489 797 }
kfraser@15489 798 else
kfraser@15489 799 {
kfraser@15489 800 printk("Kdump: %luMB (%lukB) at 0x%lx\n",
kfraser@15489 801 kdump_size >> 20, kdump_size >> 10, kdump_start);
kfraser@15489 802 }
kfraser@15489 803 }
kfraser@15489 804
kfraser@15074 805 /*
kfraser@15074 806 * With the boot allocator now seeded, we can walk every RAM region and
kfraser@15074 807 * map it in its entirety (on x86/64, at least) and notify it to the
kfraser@15074 808 * boot allocator.
kfraser@15074 809 */
kfraser@15074 810 for ( i = 0; i < boot_e820.nr_map; i++ )
kfraser@15074 811 {
kfraser@15074 812 uint64_t s, e, map_e, mask = PAGE_SIZE - 1;
kfraser@15074 813
kfraser@15074 814 /* Only page alignment required now. */
kfraser@15074 815 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 816 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
kfraser@15074 817 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kfraser@15074 818 continue;
kfraser@15074 819
kfraser@15074 820 /* Perform the mapping (truncated in 32-bit mode). */
kfraser@15074 821 map_e = e;
kfraser@15074 822 #if defined(CONFIG_X86_32)
kfraser@15074 823 map_e = min_t(uint64_t, map_e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 824 #endif
kfraser@15074 825 if ( s < map_e )
kfraser@15074 826 map_pages_to_xen(
kfraser@15074 827 (unsigned long)maddr_to_bootstrap_virt(s),
kfraser@15074 828 s >> PAGE_SHIFT, (map_e-s) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kfraser@15074 829
kfraser@15074 830 init_boot_pages(s, e);
kfraser@15074 831 }
kfraser@15074 832
kaf24@5003 833 memguard_init();
kaf24@4950 834
kfraser@15074 835 nr_pages = 0;
kfraser@15074 836 for ( i = 0; i < e820.nr_map; i++ )
kfraser@15074 837 if ( e820.map[i].type == E820_RAM )
kfraser@15074 838 nr_pages += e820.map[i].size >> PAGE_SHIFT;
ian@12681 839 printk("System RAM: %luMB (%lukB)\n",
kaf24@3354 840 nr_pages >> (20 - PAGE_SHIFT),
kaf24@3354 841 nr_pages << (PAGE_SHIFT - 10));
kaf24@7220 842 total_pages = nr_pages;
kaf24@3354 843
kfraser@11296 844 /* Sanity check for unwanted bloat of certain hypercall structures. */
kfraser@11296 845 BUILD_BUG_ON(sizeof(((struct xen_platform_op *)0)->u) !=
kfraser@11296 846 sizeof(((struct xen_platform_op *)0)->u.pad));
kfraser@11296 847 BUILD_BUG_ON(sizeof(((struct xen_domctl *)0)->u) !=
kfraser@11296 848 sizeof(((struct xen_domctl *)0)->u.pad));
kfraser@11296 849 BUILD_BUG_ON(sizeof(((struct xen_sysctl *)0)->u) !=
kfraser@11296 850 sizeof(((struct xen_sysctl *)0)->u.pad));
kaf24@7388 851
kaf24@9878 852 BUILD_BUG_ON(sizeof(start_info_t) > PAGE_SIZE);
kaf24@9878 853 BUILD_BUG_ON(sizeof(shared_info_t) > PAGE_SIZE);
ack@13292 854 BUILD_BUG_ON(sizeof(struct vcpu_info) != 64);
kaf24@7744 855
ack@13291 856 #ifdef CONFIG_COMPAT
ack@13291 857 BUILD_BUG_ON(sizeof(((struct compat_platform_op *)0)->u) !=
ack@13291 858 sizeof(((struct compat_platform_op *)0)->u.pad));
ack@13291 859 BUILD_BUG_ON(sizeof(start_info_compat_t) > PAGE_SIZE);
ack@13292 860 BUILD_BUG_ON(sizeof(struct compat_vcpu_info) != 64);
ack@13291 861 #endif
ack@13291 862
kfraser@10492 863 /* Check definitions in public headers match internal defs. */
kaf24@9878 864 BUILD_BUG_ON(__HYPERVISOR_VIRT_START != HYPERVISOR_VIRT_START);
kaf24@8521 865 #ifdef HYPERVISOR_VIRT_END
kaf24@9878 866 BUILD_BUG_ON(__HYPERVISOR_VIRT_END != HYPERVISOR_VIRT_END);
kaf24@8521 867 #endif
kfraser@10492 868 BUILD_BUG_ON(MACH2PHYS_VIRT_START != RO_MPT_VIRT_START);
kfraser@10492 869 BUILD_BUG_ON(MACH2PHYS_VIRT_END != RO_MPT_VIRT_END);
kaf24@8521 870
kaf24@3354 871 init_frametable();
kaf24@3338 872
kfraser@11971 873 acpi_boot_table_init();
kfraser@11971 874
kfraser@11971 875 acpi_numa_init();
kfraser@11971 876
kfraser@11971 877 numa_initmem_init(0, max_page);
kfraser@11971 878
kaf24@6111 879 /* Initialise the Xen heap, skipping RAM holes. */
kfraser@15074 880 init_xenheap_pages(xenheap_phys_start, xenheap_phys_end);
kfraser@15074 881 nr_pages = (xenheap_phys_end - xenheap_phys_start) >> PAGE_SHIFT;
kfraser@15074 882 #ifdef __x86_64__
kfraser@15074 883 init_xenheap_pages(xen_phys_start, __pa(&_start));
kfraser@15074 884 nr_pages += (__pa(&_start) - xen_phys_start) >> PAGE_SHIFT;
kfraser@15074 885 #endif
kfraser@15074 886 xenheap_phys_start = xen_phys_start;
kaf24@6111 887 printk("Xen heap: %luMB (%lukB)\n",
kaf24@6111 888 nr_pages >> (20 - PAGE_SHIFT),
kaf24@6111 889 nr_pages << (PAGE_SHIFT - 10));
kaf24@3338 890
keir@14680 891 end_boot_allocator();
keir@14680 892
kaf24@3594 893 early_boot = 0;
kaf24@3338 894
kaf24@8459 895 early_cpu_init();
kaf24@8459 896
kaf24@8459 897 paging_init();
kaf24@8459 898
kaf24@8459 899 /* Unmap the first page of CPU0's stack. */
kaf24@8459 900 memguard_guard_stack(cpu0_stack);
kaf24@8459 901
kaf24@8459 902 open_softirq(NEW_TLBFLUSH_CLOCK_PERIOD_SOFTIRQ, new_tlbflush_clock_period);
kaf24@8459 903
kaf24@8459 904 if ( opt_watchdog )
kaf24@8459 905 nmi_watchdog = NMI_LOCAL_APIC;
kaf24@8459 906
kaf24@8459 907 sort_exception_tables();
kaf24@8459 908
kaf24@8459 909 find_smp_config();
kaf24@8459 910
kaf24@8459 911 dmi_scan_machine();
kaf24@8459 912
kaf24@8459 913 generic_apic_probe();
kaf24@8459 914
kaf24@8459 915 acpi_boot_init();
kaf24@8459 916
kfraser@11971 917 init_cpu_to_node();
kfraser@11971 918
kfraser@11241 919 if ( smp_found_config )
kaf24@8459 920 get_smp_config();
kaf24@8459 921
keir@15083 922 #ifdef CONFIG_X86_64
keir@15083 923 /* Low mappings were only needed for some BIOS table parsing. */
keir@15083 924 zap_low_mappings();
keir@15083 925 #endif
keir@15083 926
kaf24@8459 927 init_apic_mappings();
kaf24@8459 928
kaf24@8459 929 init_IRQ();
kaf24@8459 930
kfraser@11241 931 percpu_init_areas();
kfraser@11241 932
kfraser@11240 933 init_idle_domain();
kfraser@11240 934
kaf24@8459 935 trap_init();
kaf24@8459 936
kaf24@13662 937 rcu_init();
kaf24@13662 938
kaf24@8586 939 timer_init();
kaf24@8459 940
kaf24@8459 941 early_time_init();
kaf24@8459 942
kaf24@8459 943 arch_init_memory();
kaf24@8459 944
kaf24@8459 945 identify_cpu(&boot_cpu_data);
kaf24@8459 946 if ( cpu_has_fxsr )
kaf24@8459 947 set_in_cr4(X86_CR4_OSFXSR);
kaf24@8459 948 if ( cpu_has_xmm )
kaf24@8459 949 set_in_cr4(X86_CR4_OSXMMEXCPT);
kaf24@8459 950
kaf24@8459 951 if ( opt_nosmp )
kaf24@8459 952 max_cpus = 0;
kaf24@8459 953
kaf24@8459 954 smp_prepare_cpus(max_cpus);
kaf24@8459 955
kaf24@8459 956 /*
kaf24@8459 957 * Initialise higher-level timer functions. We do this fairly late
kaf24@8459 958 * (post-SMP) because the time bases and scale factors need to be updated
kaf24@8459 959 * regularly, and SMP initialisation can cause a long delay with
kaf24@8459 960 * interrupts not yet enabled.
kaf24@8459 961 */
kaf24@8459 962 init_xen_time();
kaf24@8459 963
kaf24@8459 964 initialize_keytable();
kaf24@8459 965
kaf24@8459 966 serial_init_postirq();
kaf24@8459 967
kaf24@8459 968 BUG_ON(!local_irq_is_enabled());
kaf24@8459 969
kaf24@8459 970 for_each_present_cpu ( i )
kaf24@8459 971 {
kaf24@8459 972 if ( num_online_cpus() >= max_cpus )
kaf24@8459 973 break;
kaf24@8459 974 if ( !cpu_online(i) )
kaf24@13662 975 {
kaf24@13662 976 rcu_online_cpu(i);
kaf24@8459 977 __cpu_up(i);
kaf24@13662 978 }
kfraser@11971 979
kfraser@11998 980 /* Set up cpu_to_node[]. */
kfraser@11971 981 srat_detect_node(i);
kfraser@11998 982 /* Set up node_to_cpumask based on cpu_to_node[]. */
kfraser@11971 983 numa_add_cpu(i);
kaf24@8459 984 }
kaf24@8459 985
kaf24@8459 986 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
kaf24@8459 987 smp_cpus_done(max_cpus);
kaf24@8459 988
kaf24@9117 989 initialise_gdb(); /* could be moved earlier */
kaf24@9117 990
kaf24@8459 991 do_initcalls();
kaf24@8459 992
kaf24@8594 993 if ( opt_watchdog )
kaf24@8594 994 watchdog_enable();
kaf24@8459 995
kfraser@11881 996 /* Extract policy from multiboot. */
kfraser@11881 997 extract_acm_policy(mbi, &initrdidx, &_policy_start, &_policy_len);
kfraser@11881 998
kaf24@8459 999 /* initialize access control security module */
kfraser@11881 1000 acm_init(_policy_start, _policy_len);
kaf24@8459 1001
kaf24@8459 1002 /* Create initial domain 0. */
kfraser@14911 1003 dom0 = domain_create(0, 0, DOM0_SSIDREF);
kfraser@10655 1004 if ( (dom0 == NULL) || (alloc_vcpu(dom0, 0, 0) == NULL) )
kaf24@8459 1005 panic("Error creating domain 0\n");
kaf24@8459 1006
kfraser@12210 1007 dom0->is_privileged = 1;
kfraser@12210 1008
kaf24@8459 1009 /* Grab the DOM0 command line. */
kaf24@8459 1010 cmdline = (char *)(mod[0].string ? __va(mod[0].string) : NULL);
kfraser@15426 1011 if ( (cmdline != NULL) || (kextra != NULL) )
kaf24@8459 1012 {
kaf24@8459 1013 static char dom0_cmdline[MAX_GUEST_CMDLINE];
kaf24@8459 1014
kfraser@15426 1015 dom0_cmdline[0] = '\0';
kfraser@15426 1016
kfraser@15426 1017 if ( cmdline != NULL )
kaf24@8459 1018 {
kfraser@15426 1019 /* Skip past the image name and copy to a local buffer. */
kaf24@8459 1020 while ( *cmdline == ' ' ) cmdline++;
kfraser@15426 1021 if ( (cmdline = strchr(cmdline, ' ')) != NULL )
kfraser@15426 1022 {
kfraser@15426 1023 while ( *cmdline == ' ' ) cmdline++;
kfraser@15426 1024 safe_strcpy(dom0_cmdline, cmdline);
kfraser@15426 1025 }
kaf24@8459 1026 }
kaf24@8459 1027
kfraser@15426 1028 if ( kextra != NULL )
kfraser@15426 1029 /* kextra always includes exactly one leading space. */
kfraser@15426 1030 safe_strcat(dom0_cmdline, kextra);
kfraser@15426 1031
kaf24@8459 1032 /* Append any extra parameters. */
kfraser@13691 1033 if ( skip_ioapic_setup && !strstr(dom0_cmdline, "noapic") )
kfraser@13691 1034 safe_strcat(dom0_cmdline, " noapic");
kaf24@8459 1035 if ( acpi_skip_timer_override &&
kfraser@13691 1036 !strstr(dom0_cmdline, "acpi_skip_timer_override") )
kfraser@13691 1037 safe_strcat(dom0_cmdline, " acpi_skip_timer_override");
kfraser@13691 1038 if ( (strlen(acpi_param) != 0) && !strstr(dom0_cmdline, "acpi=") )
kaf24@8459 1039 {
kfraser@13691 1040 safe_strcat(dom0_cmdline, " acpi=");
kfraser@13691 1041 safe_strcat(dom0_cmdline, acpi_param);
kaf24@8459 1042 }
kfraser@13691 1043
kfraser@13691 1044 cmdline = dom0_cmdline;
kaf24@8459 1045 }
kaf24@8459 1046
kaf24@8459 1047 if ( (initrdidx > 0) && (initrdidx < mbi->mods_count) )
kaf24@8459 1048 {
kaf24@8459 1049 _initrd_start = initial_images_start +
kaf24@8459 1050 (mod[initrdidx].mod_start - mod[0].mod_start);
kaf24@8459 1051 _initrd_len = mod[initrdidx].mod_end - mod[initrdidx].mod_start;
kaf24@8459 1052 }
kaf24@8459 1053
kaf24@8459 1054 /*
kaf24@8459 1055 * We're going to setup domain0 using the module(s) that we stashed safely
kaf24@8459 1056 * above our heap. The second module, if present, is an initrd ramdisk.
kaf24@8459 1057 */
kaf24@8459 1058 if ( construct_dom0(dom0,
kaf24@8459 1059 initial_images_start,
kaf24@8459 1060 mod[0].mod_end-mod[0].mod_start,
kaf24@8459 1061 _initrd_start,
kaf24@8459 1062 _initrd_len,
kaf24@8459 1063 cmdline) != 0)
kaf24@8459 1064 panic("Could not set up DOM0 guest OS\n");
kaf24@8459 1065
kaf24@8459 1066 /* Scrub RAM that is still free and so may go to an unprivileged domain. */
kaf24@8459 1067 scrub_heap_pages();
kaf24@8459 1068
kaf24@8459 1069 init_trace_bufs();
kaf24@8459 1070
kaf24@10502 1071 console_endboot();
kaf24@8459 1072
kaf24@8459 1073 /* Hide UART from DOM0 if we're using it */
kaf24@8459 1074 serial_endboot();
kaf24@8459 1075
kaf24@8459 1076 domain_unpause_by_systemcontroller(dom0);
kaf24@8459 1077
keir@15082 1078 reset_stack_and_jump(init_done);
kaf24@8459 1079 }
kaf24@8459 1080
ian@13763 1081 void arch_get_xen_caps(xen_capabilities_info_t *info)
iap10@6721 1082 {
kfraser@14997 1083 /* Interface name is always xen-3.0-* for Xen-3.x. */
kfraser@14997 1084 int major = 3, minor = 0;
keir@13754 1085 char s[32];
keir@13754 1086
ian@13763 1087 (*info)[0] = '\0';
iap10@6721 1088
kaf24@6725 1089 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
kaf24@6725 1090
keir@13754 1091 snprintf(s, sizeof(s), "xen-%d.%d-x86_32 ", major, minor);
ian@13763 1092 safe_strcat(*info, s);
keir@13754 1093 if ( hvm_enabled )
keir@13754 1094 {
keir@13754 1095 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1096 safe_strcat(*info, s);
kfraser@13685 1097 }
kaf24@6725 1098
kaf24@6725 1099 #elif defined(CONFIG_X86_32) && defined(CONFIG_X86_PAE)
kaf24@6725 1100
keir@13754 1101 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1102 safe_strcat(*info, s);
kaf24@6725 1103 if ( hvm_enabled )
iap10@6721 1104 {
keir@13754 1105 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1106 safe_strcat(*info, s);
keir@13754 1107 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1108 safe_strcat(*info, s);
iap10@6721 1109 }
iap10@6721 1110
kaf24@6725 1111 #elif defined(CONFIG_X86_64)
iap10@6721 1112
keir@13754 1113 snprintf(s, sizeof(s), "xen-%d.%d-x86_64 ", major, minor);
ian@13763 1114 safe_strcat(*info, s);
ack@13288 1115 #ifdef CONFIG_COMPAT
keir@13754 1116 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1117 safe_strcat(*info, s);
ack@13288 1118 #endif
kaf24@6725 1119 if ( hvm_enabled )
iap10@6721 1120 {
keir@13754 1121 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1122 safe_strcat(*info, s);
keir@13754 1123 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1124 safe_strcat(*info, s);
keir@13754 1125 snprintf(s, sizeof(s), "hvm-%d.%d-x86_64 ", major, minor);
ian@13763 1126 safe_strcat(*info, s);
iap10@6721 1127 }
kaf24@6725 1128
iap10@6721 1129 #endif
iap10@6721 1130 }
iap10@6721 1131
kaf24@3914 1132 /*
kaf24@3914 1133 * Local variables:
kaf24@3914 1134 * mode: C
kaf24@3914 1135 * c-set-style: "BSD"
kaf24@3914 1136 * c-basic-offset: 4
kaf24@3914 1137 * tab-width: 4
kaf24@3914 1138 * indent-tabs-mode: nil
kaf24@3988 1139 * End:
kaf24@3914 1140 */