ia64/xen-unstable

annotate xen/arch/x86/setup.c @ 15379:b1eb43f94a3a

x86/64: Avoid bogus mbi pointer into relocated Xen address space.
Prevent similar bugs in future by poisoning the relocated bottom
megabyte.
Signed-off-by: Keir Fraser <keir@xensource.com>
author kfraser@localhost.localdomain
date Mon Jun 18 16:48:05 2007 +0100 (2007-06-18)
parents 699f0c429620
children 499bab040137
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 {
kaf24@9067 175 unsigned long s = initial_images_start + PAGE_SIZE - 1;
kaf24@9067 176 unsigned long e = initial_images_end;
kaf24@9067 177 return ((e >> PAGE_SHIFT) - (s >> PAGE_SHIFT));
kaf24@9067 178 }
kaf24@9067 179
kfraser@15074 180 void __init discard_initial_images(void)
kaf24@9067 181 {
kaf24@9067 182 init_domheap_pages(initial_images_start, initial_images_end);
kaf24@9067 183 }
kaf24@9067 184
kaf24@9818 185 extern char __per_cpu_start[], __per_cpu_data_end[], __per_cpu_end[];
kaf24@9818 186
kfraser@11241 187 static void __init percpu_init_areas(void)
kaf24@9818 188 {
kaf24@9818 189 unsigned int i, data_size = __per_cpu_data_end - __per_cpu_start;
kfraser@15074 190 unsigned int first_unused;
kaf24@9818 191
kaf24@9818 192 BUG_ON(data_size > PERCPU_SIZE);
kaf24@9818 193
kfraser@15074 194 /* Initialise per-cpu data area for all possible secondary CPUs. */
kfraser@15074 195 for ( i = 1; (i < NR_CPUS) && cpu_possible(i); i++ )
kfraser@15074 196 memcpy(__per_cpu_start + (i << PERCPU_SHIFT),
kfraser@15074 197 __per_cpu_start,
kfraser@15074 198 data_size);
kaf24@9818 199 first_unused = i;
kaf24@9818 200
kfraser@14340 201 /* Check that there are no holes in cpu_possible_map. */
kaf24@9818 202 for ( ; i < NR_CPUS; i++ )
kfraser@14340 203 BUG_ON(cpu_possible(i));
kaf24@9818 204
kfraser@11241 205 #ifndef MEMORY_GUARD
kaf24@9818 206 init_xenheap_pages(__pa(__per_cpu_start) + (first_unused << PERCPU_SHIFT),
kaf24@9818 207 __pa(__per_cpu_end));
kfraser@11241 208 #endif
keir@15082 209 memguard_guard_range(&__per_cpu_start[first_unused << PERCPU_SHIFT],
keir@15082 210 (NR_CPUS - first_unused) << PERCPU_SHIFT);
keir@15082 211 #if defined(CONFIG_X86_64)
keir@15082 212 /* Also zap the mapping in the 1:1 area. */
keir@15082 213 memguard_guard_range(__va(__pa(__per_cpu_start)) +
keir@15082 214 (first_unused << PERCPU_SHIFT),
keir@15082 215 (NR_CPUS - first_unused) << PERCPU_SHIFT);
keir@15082 216 #endif
kaf24@9818 217 }
kaf24@9818 218
kfraser@11881 219 /* Fetch acm policy module from multiboot modules. */
kfraser@15074 220 static void __init extract_acm_policy(
kfraser@11881 221 multiboot_info_t *mbi,
kfraser@11881 222 unsigned int *initrdidx,
kfraser@11881 223 char **_policy_start,
kfraser@11881 224 unsigned long *_policy_len)
kfraser@11881 225 {
kfraser@11881 226 int i;
kfraser@11881 227 module_t *mod = (module_t *)__va(mbi->mods_addr);
kfraser@11881 228 unsigned long start, policy_len;
kfraser@11881 229 char *policy_start;
kfraser@11881 230
kfraser@11881 231 /*
kfraser@11881 232 * Try all modules and see whichever could be the binary policy.
kfraser@11881 233 * Adjust the initrdidx if module[1] is the binary policy.
kfraser@11881 234 */
kfraser@11881 235 for ( i = mbi->mods_count-1; i >= 1; i-- )
kfraser@11881 236 {
kfraser@11881 237 start = initial_images_start + (mod[i].mod_start-mod[0].mod_start);
kfraser@15074 238 policy_start = maddr_to_bootstrap_virt(start);
kfraser@11881 239 policy_len = mod[i].mod_end - mod[i].mod_start;
kfraser@11881 240 if ( acm_is_policy(policy_start, policy_len) )
kfraser@11881 241 {
kfraser@11881 242 printk("Policy len 0x%lx, start at %p - module %d.\n",
kfraser@11881 243 policy_len, policy_start, i);
kfraser@11881 244 *_policy_start = policy_start;
kfraser@11881 245 *_policy_len = policy_len;
kfraser@11881 246 if ( i == 1 )
kfraser@11881 247 *initrdidx = (mbi->mods_count > 2) ? 2 : 0;
kfraser@11881 248 break;
kfraser@11881 249 }
kfraser@11881 250 }
kfraser@11881 251 }
kfraser@11881 252
kfraser@11241 253 static void __init init_idle_domain(void)
kfraser@11240 254 {
kfraser@11240 255 struct domain *idle_domain;
kfraser@11240 256
kfraser@11240 257 /* Domain creation requires that scheduler structures are initialised. */
kfraser@11240 258 scheduler_init();
kfraser@11240 259
kfraser@14911 260 idle_domain = domain_create(IDLE_DOMAIN_ID, 0, 0);
kfraser@11240 261 if ( (idle_domain == NULL) || (alloc_vcpu(idle_domain, 0, 0) == NULL) )
kfraser@11240 262 BUG();
kfraser@11240 263
kfraser@11240 264 set_current(idle_domain->vcpu[0]);
kfraser@11240 265 idle_vcpu[0] = this_cpu(curr_vcpu) = current;
kfraser@11240 266
kfraser@11240 267 setup_idle_pagetable();
kfraser@11240 268 }
kfraser@11240 269
kfraser@15074 270 static void __init srat_detect_node(int cpu)
kfraser@11971 271 {
kfraser@11998 272 unsigned node;
kfraser@11998 273 u8 apicid = x86_cpu_to_apicid[cpu];
kfraser@11971 274
kfraser@11998 275 node = apicid_to_node[apicid];
kfraser@11998 276 if ( node == NUMA_NO_NODE )
kfraser@11998 277 node = 0;
kfraser@11998 278 numa_set_node(cpu, node);
kfraser@11971 279
kfraser@11998 280 if ( acpi_numa > 0 )
kfraser@11998 281 printk(KERN_INFO "CPU %d APIC %d -> Node %d\n", cpu, apicid, node);
kfraser@11971 282 }
kfraser@11971 283
kfraser@15074 284 static void __init move_memory(
kfraser@15074 285 unsigned long dst, unsigned long src_start, unsigned long src_end)
ian@12677 286 {
kfraser@15074 287 memmove(maddr_to_bootstrap_virt(dst),
kfraser@15074 288 maddr_to_bootstrap_virt(src_start),
ian@12677 289 src_end - src_start);
kfraser@15074 290 }
kfraser@15074 291
kfraser@15074 292 /* A temporary copy of the e820 map that we can mess with during bootstrap. */
kfraser@15074 293 static struct e820map __initdata boot_e820;
kfraser@15074 294
kfraser@15074 295 /* Reserve area (@s,@e) in the temporary bootstrap e820 map. */
kfraser@15074 296 static void __init reserve_in_boot_e820(unsigned long s, unsigned long e)
kfraser@15074 297 {
kfraser@15074 298 unsigned long rs, re;
kfraser@15074 299 int i;
kfraser@15074 300
kfraser@15074 301 for ( i = 0; i < boot_e820.nr_map; i++ )
kfraser@15074 302 {
kfraser@15074 303 /* Have we found the e820 region that includes the specified range? */
kfraser@15074 304 rs = boot_e820.map[i].addr;
kfraser@15074 305 re = boot_e820.map[i].addr + boot_e820.map[i].size;
kfraser@15074 306 if ( (s < rs) || (e > re) )
kfraser@15074 307 continue;
kfraser@15074 308
kfraser@15074 309 /* Start fragment. */
kfraser@15074 310 boot_e820.map[i].size = s - rs;
kfraser@15074 311
kfraser@15074 312 /* End fragment. */
kfraser@15074 313 if ( e < re )
kfraser@15074 314 {
kfraser@15074 315 memmove(&boot_e820.map[i+1], &boot_e820.map[i],
kfraser@15074 316 (boot_e820.nr_map-i) * sizeof(boot_e820.map[0]));
kfraser@15074 317 boot_e820.nr_map++;
kfraser@15074 318 i++;
kfraser@15074 319 boot_e820.map[i].addr = e;
kfraser@15074 320 boot_e820.map[i].size = re - e;
kfraser@15074 321 }
kfraser@15074 322 }
ian@12677 323 }
ian@12677 324
keir@15298 325 struct boot_video_info {
keir@15298 326 u8 orig_x; /* 0x00 */
keir@15298 327 u8 orig_y; /* 0x01 */
keir@15298 328 u8 orig_video_mode; /* 0x02 */
keir@15298 329 u8 orig_video_cols; /* 0x03 */
keir@15298 330 u8 orig_video_lines; /* 0x04 */
keir@15298 331 u8 orig_video_isVGA; /* 0x05 */
keir@15298 332 u16 orig_video_points; /* 0x06 */
keir@15298 333
keir@15298 334 /* VESA graphic mode -- linear frame buffer */
keir@15298 335 u32 capabilities; /* 0x08 */
keir@15298 336 u16 lfb_linelength; /* 0x0c */
keir@15298 337 u16 lfb_width; /* 0x0e */
keir@15298 338 u16 lfb_height; /* 0x10 */
keir@15298 339 u16 lfb_depth; /* 0x12 */
keir@15298 340 u32 lfb_base; /* 0x14 */
keir@15298 341 u32 lfb_size; /* 0x18 */
keir@15298 342 u8 red_size; /* 0x1c */
keir@15298 343 u8 red_pos; /* 0x1d */
keir@15298 344 u8 green_size; /* 0x1e */
keir@15298 345 u8 green_pos; /* 0x1f */
keir@15298 346 u8 blue_size; /* 0x20 */
keir@15298 347 u8 blue_pos; /* 0x21 */
keir@15298 348 u8 rsvd_size; /* 0x22 */
keir@15298 349 u8 rsvd_pos; /* 0x23 */
keir@15298 350 u16 vesapm_seg; /* 0x24 */
keir@15298 351 u16 vesapm_off; /* 0x26 */
keir@15298 352 };
keir@15298 353
keir@15298 354 static void __init parse_video_info(void)
keir@15298 355 {
keir@15298 356 struct boot_video_info *bvi = &bootsym(boot_vid_info);
keir@15298 357
keir@15298 358 if ( (bvi->orig_video_isVGA == 1) && (bvi->orig_video_mode == 3) )
keir@15298 359 {
keir@15298 360 vga_console_info.video_type = XEN_VGATYPE_TEXT_MODE_3;
keir@15298 361 vga_console_info.u.text_mode_3.font_height = bvi->orig_video_points;
keir@15298 362 vga_console_info.u.text_mode_3.cursor_x = bvi->orig_x;
keir@15298 363 vga_console_info.u.text_mode_3.cursor_y = bvi->orig_y;
keir@15298 364 vga_console_info.u.text_mode_3.rows = bvi->orig_video_lines;
keir@15298 365 vga_console_info.u.text_mode_3.columns = bvi->orig_video_cols;
keir@15298 366 }
keir@15298 367 else if ( bvi->orig_video_isVGA == 0x23 )
keir@15298 368 {
keir@15298 369 vga_console_info.video_type = XEN_VGATYPE_VESA_LFB;
keir@15298 370 vga_console_info.u.vesa_lfb.width = bvi->lfb_width;
keir@15298 371 vga_console_info.u.vesa_lfb.height = bvi->lfb_height;
keir@15298 372 vga_console_info.u.vesa_lfb.bytes_per_line = bvi->lfb_linelength;
keir@15298 373 vga_console_info.u.vesa_lfb.bits_per_pixel = bvi->lfb_depth;
keir@15298 374 vga_console_info.u.vesa_lfb.lfb_base = bvi->lfb_base;
keir@15298 375 vga_console_info.u.vesa_lfb.lfb_size = bvi->lfb_size;
keir@15298 376 vga_console_info.u.vesa_lfb.red_pos = bvi->red_pos;
keir@15298 377 vga_console_info.u.vesa_lfb.red_size = bvi->red_size;
keir@15298 378 vga_console_info.u.vesa_lfb.green_pos = bvi->green_pos;
keir@15298 379 vga_console_info.u.vesa_lfb.green_size = bvi->green_size;
keir@15298 380 vga_console_info.u.vesa_lfb.blue_pos = bvi->blue_pos;
keir@15298 381 vga_console_info.u.vesa_lfb.blue_size = bvi->blue_size;
keir@15298 382 vga_console_info.u.vesa_lfb.rsvd_pos = bvi->rsvd_pos;
keir@15298 383 vga_console_info.u.vesa_lfb.rsvd_size = bvi->rsvd_size;
keir@15298 384 }
keir@15298 385 }
keir@15298 386
keir@15082 387 void init_done(void)
keir@15082 388 {
keir@15082 389 extern char __init_begin[], __init_end[];
keir@15082 390
keir@15082 391 /* Free (or page-protect) the init areas. */
keir@15082 392 #ifndef MEMORY_GUARD
keir@15082 393 init_xenheap_pages(__pa(__init_begin), __pa(__init_end));
keir@15082 394 #endif
keir@15082 395 memguard_guard_range(__init_begin, __init_end - __init_begin);
keir@15082 396 #if defined(CONFIG_X86_64)
keir@15082 397 /* Also zap the mapping in the 1:1 area. */
keir@15082 398 memguard_guard_range(__va(__pa(__init_begin)), __init_end - __init_begin);
keir@15082 399 #endif
keir@15082 400 printk("Freed %ldkB init memory.\n", (long)(__init_end-__init_begin)>>10);
keir@15082 401
keir@15082 402 startup_cpu_idle_loop();
keir@15082 403 }
keir@15082 404
kfraser@15379 405 void __init __start_xen(unsigned long mbi_p)
kaf24@1452 406 {
kfraser@15293 407 char *memmap_type = NULL;
kaf24@9823 408 char __cmdline[] = "", *cmdline = __cmdline;
kaf24@8457 409 unsigned long _initrd_start = 0, _initrd_len = 0;
kaf24@8457 410 unsigned int initrdidx = 1;
kfraser@11881 411 char *_policy_start = NULL;
kfraser@11881 412 unsigned long _policy_len = 0;
kfraser@15379 413 multiboot_info_t *mbi = __va(mbi_p);
kaf24@8457 414 module_t *mod = (module_t *)__va(mbi->mods_addr);
kaf24@6111 415 unsigned long nr_pages, modules_length;
kfraser@15293 416 int i, e820_warn = 0, bytes = 0;
kaf24@5776 417 struct ns16550_defaults ns16550 = {
kaf24@5776 418 .data_bits = 8,
kaf24@5776 419 .parity = 'n',
kaf24@5776 420 .stop_bits = 1
kaf24@5776 421 };
kaf24@3338 422
kfraser@12853 423 extern void early_page_fault(void);
kfraser@12853 424 set_intr_gate(TRAP_page_fault, &early_page_fault);
kfraser@12853 425
kaf24@3338 426 /* Parse the command-line options. */
kaf24@3344 427 if ( (mbi->flags & MBI_CMDLINE) && (mbi->cmdline != 0) )
kaf24@9823 428 cmdline = __va(mbi->cmdline);
kaf24@9823 429 cmdline_parse(cmdline);
kaf24@3338 430
keir@15298 431 parse_video_info();
keir@15298 432
kaf24@8534 433 set_current((struct vcpu *)0xfffff000); /* debug sanity */
kfraser@11240 434 idle_vcpu[0] = current;
kaf24@8534 435 set_processor_id(0); /* needed early, for smp_processor_id() */
kaf24@3338 436
kaf24@5146 437 smp_prepare_boot_cpu();
kaf24@5146 438
kaf24@3338 439 /* We initialise the serial devices very early so we can get debugging. */
kaf24@5776 440 ns16550.io_base = 0x3f8;
kaf24@5776 441 ns16550.irq = 4;
kaf24@5776 442 ns16550_init(0, &ns16550);
kaf24@5776 443 ns16550.io_base = 0x2f8;
kaf24@5776 444 ns16550.irq = 3;
kaf24@5776 445 ns16550_init(1, &ns16550);
kaf24@5195 446 serial_init_preirq();
kaf24@3338 447
kaf24@3338 448 init_console();
kaf24@3338 449
kfraser@11947 450 printk("Command line: %s\n", cmdline);
kaf24@9823 451
kfraser@15330 452 printk("Video information:\n");
kfraser@15330 453
kfraser@15330 454 /* Print VGA display mode information. */
keir@15298 455 switch ( vga_console_info.video_type )
keir@15298 456 {
keir@15298 457 case XEN_VGATYPE_TEXT_MODE_3:
kfraser@15330 458 printk(" VGA is text mode %dx%d, font 8x%d\n",
keir@15298 459 vga_console_info.u.text_mode_3.columns,
keir@15298 460 vga_console_info.u.text_mode_3.rows,
keir@15298 461 vga_console_info.u.text_mode_3.font_height);
keir@15298 462 break;
keir@15298 463 case XEN_VGATYPE_VESA_LFB:
kfraser@15330 464 printk(" VGA is graphics mode %dx%d, %d bpp\n",
keir@15298 465 vga_console_info.u.vesa_lfb.width,
keir@15298 466 vga_console_info.u.vesa_lfb.height,
keir@15298 467 vga_console_info.u.vesa_lfb.bits_per_pixel);
keir@15298 468 break;
kfraser@15330 469 default:
kfraser@15330 470 printk(" No VGA detected\n");
kfraser@15330 471 break;
kfraser@15330 472 }
kfraser@15330 473
kfraser@15330 474 /* Print VBE/DDC EDID information. */
kfraser@15330 475 if ( bootsym(boot_edid_caps) != 0x1313 )
kfraser@15330 476 {
kfraser@15330 477 u16 caps = bootsym(boot_edid_caps);
kfraser@15330 478 printk(" VBE/DDC methods:%s%s%s; ",
kfraser@15330 479 (caps & 1) ? " V1" : "",
kfraser@15330 480 (caps & 2) ? " V2" : "",
kfraser@15330 481 !(caps & 3) ? " none" : "");
kfraser@15330 482 printk("EDID transfer time: %d seconds\n", caps >> 8);
kfraser@15330 483 if ( *(u32 *)bootsym(boot_edid_info) == 0x13131313 )
kfraser@15330 484 {
kfraser@15330 485 printk(" EDID info not retrieved because ");
kfraser@15330 486 if ( !(caps & 3) )
kfraser@15330 487 printk("no DDC retrieval method detected\n");
kfraser@15330 488 else if ( (caps >> 8) > 5 )
kfraser@15330 489 printk("takes longer than 5 seconds\n");
kfraser@15330 490 else
kfraser@15330 491 printk("of reasons unknown\n");
kfraser@15330 492 }
keir@15298 493 }
keir@15298 494
kfraser@15336 495 printk("Disc information:\n");
kfraser@15336 496 printk(" Found %d MBR signatures\n",
kfraser@15336 497 bootsym(boot_edd_signature_nr));
kfraser@15336 498 printk(" Found %d EDD information structures\n",
kfraser@15336 499 bootsym(boot_edd_info_nr));
kfraser@15336 500
kaf24@3344 501 /* Check that we have at least one Multiboot module. */
kaf24@3344 502 if ( !(mbi->flags & MBI_MODULES) || (mbi->mods_count == 0) )
kfraser@15074 503 EARLY_FAIL("dom0 kernel not specified. "
kfraser@15074 504 "Check bootloader configuration.\n");
kaf24@5011 505
kaf24@5011 506 if ( ((unsigned long)cpu0_stack & (STACK_SIZE-1)) != 0 )
kfraser@15074 507 EARLY_FAIL("Misaligned CPU0 stack.\n");
kaf24@3338 508
kfraser@11618 509 /*
kfraser@11618 510 * Since there are some stubs getting built on the stacks which use
kfraser@11618 511 * direct calls/jumps, the heap must be confined to the lower 2G so
kfraser@11618 512 * that those branches can reach their targets.
kfraser@11618 513 */
kfraser@11618 514 if ( opt_xenheap_megabytes > 2048 )
kfraser@11618 515 opt_xenheap_megabytes = 2048;
kaf24@3338 516
kfraser@15293 517 if ( e820_raw_nr != 0 )
kfraser@15292 518 {
kfraser@15293 519 memmap_type = "Xen-e820";
kfraser@15292 520 }
kfraser@15293 521 else if ( bootsym(lowmem_kb) )
kfraser@15292 522 {
kfraser@15293 523 memmap_type = "Xen-e801";
kfraser@15292 524 e820_raw[0].addr = 0;
kfraser@15293 525 e820_raw[0].size = bootsym(lowmem_kb) << 10;
kfraser@15292 526 e820_raw[0].type = E820_RAM;
kfraser@15292 527 e820_raw[1].addr = 0x100000;
kfraser@15293 528 e820_raw[1].size = bootsym(highmem_kb) << 10;
kfraser@15292 529 e820_raw[1].type = E820_RAM;
kfraser@15292 530 e820_raw_nr = 2;
kfraser@15292 531 }
kfraser@15292 532 else if ( mbi->flags & MBI_MEMMAP )
kaf24@3344 533 {
kfraser@15293 534 memmap_type = "Multiboot-e820";
kaf24@3344 535 while ( bytes < mbi->mmap_length )
kaf24@3344 536 {
kaf24@3344 537 memory_map_t *map = __va(mbi->mmap_addr + bytes);
kaf24@8402 538
kaf24@8402 539 /*
kaf24@8403 540 * This is a gross workaround for a BIOS bug. Some bootloaders do
kaf24@8402 541 * not write e820 map entries into pre-zeroed memory. This is
kaf24@8402 542 * okay if the BIOS fills in all fields of the map entry, but
kaf24@8402 543 * some broken BIOSes do not bother to write the high word of
kaf24@8402 544 * the length field if the length is smaller than 4GB. We
kaf24@8402 545 * detect and fix this by flagging sections below 4GB that
kaf24@8403 546 * appear to be larger than 4GB in size.
kaf24@8402 547 */
kaf24@8403 548 if ( (map->base_addr_high == 0) && (map->length_high != 0) )
kaf24@8402 549 {
kfraser@15292 550 if ( !e820_warn )
kfraser@15292 551 {
kfraser@15292 552 printk("WARNING: Buggy e820 map detected and fixed "
kfraser@15292 553 "(truncated length fields).\n");
kfraser@15292 554 e820_warn = 1;
kfraser@15292 555 }
kaf24@8402 556 map->length_high = 0;
kaf24@8402 557 }
kaf24@8402 558
kaf24@3344 559 e820_raw[e820_raw_nr].addr =
kaf24@3344 560 ((u64)map->base_addr_high << 32) | (u64)map->base_addr_low;
kaf24@3344 561 e820_raw[e820_raw_nr].size =
kaf24@3344 562 ((u64)map->length_high << 32) | (u64)map->length_low;
kaf24@3344 563 e820_raw[e820_raw_nr].type =
kfraser@12226 564 (map->type > E820_NVS) ? E820_RESERVED : map->type;
kaf24@3344 565 e820_raw_nr++;
kaf24@8402 566
kaf24@3344 567 bytes += map->size + 4;
kaf24@3344 568 }
kaf24@3344 569 }
kaf24@3344 570 else if ( mbi->flags & MBI_MEMLIMITS )
kaf24@3344 571 {
kfraser@15293 572 memmap_type = "Multiboot-e801";
kaf24@3344 573 e820_raw[0].addr = 0;
kaf24@3344 574 e820_raw[0].size = mbi->mem_lower << 10;
kaf24@3344 575 e820_raw[0].type = E820_RAM;
kaf24@3354 576 e820_raw[1].addr = 0x100000;
kaf24@3354 577 e820_raw[1].size = mbi->mem_upper << 10;
kaf24@3354 578 e820_raw[1].type = E820_RAM;
kaf24@3344 579 e820_raw_nr = 2;
kaf24@3344 580 }
kaf24@3344 581 else
kaf24@3344 582 {
kfraser@15074 583 EARLY_FAIL("Bootloader provided no memory information.\n");
kaf24@3344 584 }
kaf24@3344 585
kaf24@13427 586 /* Ensure that all E820 RAM regions are page-aligned and -sized. */
kaf24@13427 587 for ( i = 0; i < e820_raw_nr; i++ )
kaf24@13427 588 {
kaf24@13427 589 uint64_t s, e;
kfraser@15292 590
kaf24@13427 591 if ( e820_raw[i].type != E820_RAM )
kaf24@13427 592 continue;
kaf24@13427 593 s = PFN_UP(e820_raw[i].addr);
kaf24@13427 594 e = PFN_DOWN(e820_raw[i].addr + e820_raw[i].size);
kaf24@13427 595 e820_raw[i].size = 0; /* discarded later */
kaf24@13427 596 if ( s < e )
kaf24@13427 597 {
kaf24@13427 598 e820_raw[i].addr = s << PAGE_SHIFT;
kaf24@13427 599 e820_raw[i].size = (e - s) << PAGE_SHIFT;
kaf24@13427 600 }
kaf24@13427 601 }
kaf24@13427 602
kaf24@13427 603 /* Sanitise the raw E820 map to produce a final clean version. */
kfraser@15293 604 max_page = init_e820(memmap_type, e820_raw, &e820_raw_nr);
kaf24@3338 605
kfraser@15074 606 /*
kfraser@15074 607 * Create a temporary copy of the E820 map. Truncate it to above 16MB
kfraser@15074 608 * as anything below that is already mapped and has a statically-allocated
kfraser@15074 609 * purpose.
kfraser@15074 610 */
kfraser@15074 611 memcpy(&boot_e820, &e820, sizeof(e820));
kfraser@15074 612 for ( i = 0; i < boot_e820.nr_map; i++ )
kaf24@3338 613 {
kfraser@15074 614 uint64_t s, e, min = 16 << 20; /* 16MB */
kfraser@15074 615 s = boot_e820.map[i].addr;
kfraser@15074 616 e = boot_e820.map[i].addr + boot_e820.map[i].size;
kfraser@15074 617 if ( s >= min )
kfraser@15074 618 continue;
kfraser@15074 619 if ( e > min )
kaf24@6111 620 {
kfraser@15074 621 boot_e820.map[i].addr = min;
kfraser@15074 622 boot_e820.map[i].size = e - min;
kaf24@6111 623 }
kfraser@15074 624 else
kfraser@15074 625 boot_e820.map[i].type = E820_RESERVED;
kaf24@3338 626 }
kaf24@6111 627
kfraser@15074 628 /*
keir@15077 629 * Iterate backwards over all superpage-aligned RAM regions.
kfraser@15074 630 *
kfraser@15074 631 * We require superpage alignment because the boot allocator is not yet
kfraser@15074 632 * initialised. Hence we can only map superpages in the address range
kfraser@15074 633 * 0 to BOOTSTRAP_DIRECTMAP_END, as this is guaranteed not to require
kfraser@15074 634 * dynamic allocation of pagetables.
kfraser@15074 635 *
kfraser@15074 636 * As well as mapping superpages in that range, in preparation for
kfraser@15074 637 * initialising the boot allocator, we also look for a region to which
kfraser@15074 638 * we can relocate the dom0 kernel and other multiboot modules. Also, on
kfraser@15074 639 * x86/64, we relocate Xen to higher memory.
kfraser@15074 640 */
kfraser@15074 641 modules_length = mod[mbi->mods_count-1].mod_end - mod[0].mod_start;
keir@15077 642 for ( i = boot_e820.nr_map-1; i >= 0; i-- )
kfraser@15074 643 {
kfraser@15074 644 uint64_t s, e, mask = (1UL << L2_PAGETABLE_SHIFT) - 1;
kaf24@6134 645
kfraser@15074 646 /* Superpage-aligned chunks up to BOOTSTRAP_DIRECTMAP_END, please. */
kfraser@15074 647 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 648 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
kfraser@15074 649 e = min_t(uint64_t, e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 650 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kaf24@3354 651 continue;
kaf24@6111 652
kfraser@15074 653 /* Map the chunk. No memory will need to be allocated to do this. */
kfraser@15074 654 map_pages_to_xen(
kfraser@15074 655 (unsigned long)maddr_to_bootstrap_virt(s),
kfraser@15074 656 s >> PAGE_SHIFT, (e-s) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kaf24@6111 657
kfraser@14084 658 #if defined(CONFIG_X86_64)
kfraser@15074 659 /* Is the region suitable for relocating Xen? */
kfraser@15074 660 if ( !xen_phys_start && (((e-s) >> 20) >= opt_xenheap_megabytes) )
kaf24@5003 661 {
kfraser@15074 662 extern l2_pgentry_t l2_xenmap[];
kfraser@15074 663 l4_pgentry_t *pl4e;
kfraser@15074 664 l3_pgentry_t *pl3e;
kfraser@15074 665 l2_pgentry_t *pl2e;
kfraser@15074 666 int i, j;
kfraser@15074 667
kfraser@15074 668 /* Select relocation address. */
kfraser@15074 669 e = (e - (opt_xenheap_megabytes << 20)) & ~mask;
kfraser@15074 670 xen_phys_start = e;
kfraser@15292 671 bootsym(trampoline_xen_phys_start) = e;
kfraser@15074 672
kfraser@15074 673 /*
kfraser@15074 674 * Perform relocation to new physical address.
kfraser@15074 675 * Before doing so we must sync static/global data with main memory
kfraser@15074 676 * with a barrier(). After this we must *not* modify static/global
kfraser@15074 677 * data until after we have switched to the relocated pagetables!
kfraser@15074 678 */
kfraser@15074 679 barrier();
kfraser@15074 680 move_memory(e, 0, __pa(&_end) - xen_phys_start);
kfraser@15074 681
kfraser@15379 682 /* Poison low 1MB to detect stray pointers to physical 0-1MB. */
kfraser@15379 683 memset(maddr_to_bootstrap_virt(e), 0x55, 1U<<20);
kfraser@15379 684
kfraser@15074 685 /* Walk initial pagetables, relocating page directory entries. */
kfraser@15074 686 pl4e = __va(__pa(idle_pg_table));
kfraser@15074 687 for ( i = 0 ; i < L4_PAGETABLE_ENTRIES; i++, pl4e++ )
kfraser@15074 688 {
kfraser@15074 689 if ( !(l4e_get_flags(*pl4e) & _PAGE_PRESENT) )
kfraser@15074 690 continue;
kfraser@15074 691 *pl4e = l4e_from_intpte(l4e_get_intpte(*pl4e) +
kfraser@15074 692 xen_phys_start);
kfraser@15074 693 pl3e = l4e_to_l3e(*pl4e);
kfraser@15074 694 for ( j = 0; j < L3_PAGETABLE_ENTRIES; j++, pl3e++ )
kfraser@15074 695 {
kfraser@15074 696 /* Not present or already relocated? */
kfraser@15074 697 if ( !(l3e_get_flags(*pl3e) & _PAGE_PRESENT) ||
kfraser@15074 698 (l3e_get_pfn(*pl3e) > 0x1000) )
kfraser@15074 699 continue;
kfraser@15074 700 *pl3e = l3e_from_intpte(l3e_get_intpte(*pl3e) +
kfraser@15074 701 xen_phys_start);
kfraser@15074 702 }
kfraser@15074 703 }
kfraser@15074 704
kfraser@15074 705 /* The only data mappings to be relocated are in the Xen area. */
kfraser@15074 706 pl2e = __va(__pa(l2_xenmap));
kfraser@15074 707 for ( i = 0; i < L2_PAGETABLE_ENTRIES; i++, pl2e++ )
kfraser@15074 708 {
kfraser@15074 709 if ( !(l2e_get_flags(*pl2e) & _PAGE_PRESENT) )
kfraser@15074 710 continue;
kfraser@15074 711 *pl2e = l2e_from_intpte(l2e_get_intpte(*pl2e) +
kfraser@15074 712 xen_phys_start);
kfraser@15074 713 }
kfraser@15074 714
kfraser@15074 715 /* Re-sync the stack and then switch to relocated pagetables. */
kfraser@15074 716 asm volatile (
kfraser@15074 717 "rep movsb ; " /* re-sync the stack */
kfraser@15074 718 "movq %%cr4,%%rsi ; "
kfraser@15074 719 "andb $0x7f,%%sil ; "
kfraser@15074 720 "movq %%rsi,%%cr4 ; " /* CR4.PGE == 0 */
kfraser@15074 721 "movq %0,%%cr3 ; " /* CR3 == new pagetables */
kfraser@15074 722 "orb $0x80,%%sil ; "
kfraser@15074 723 "movq %%rsi,%%cr4 " /* CR4.PGE == 1 */
kfraser@15074 724 : : "r" (__pa(idle_pg_table)), "S" (cpu0_stack),
kfraser@15074 725 "D" (__va(__pa(cpu0_stack))), "c" (STACK_SIZE) : "memory" );
kaf24@5003 726 }
kaf24@5003 727 #endif
keir@15077 728
keir@15077 729 /* Is the region suitable for relocating the multiboot modules? */
keir@15077 730 if ( !initial_images_start && (s < e) && ((e-s) >= modules_length) )
keir@15077 731 {
keir@15077 732 e -= modules_length;
keir@15077 733 initial_images_start = e;
keir@15077 734 initial_images_end = initial_images_start + modules_length;
keir@15077 735 move_memory(initial_images_start,
keir@15077 736 mod[0].mod_start, mod[mbi->mods_count-1].mod_end);
keir@15077 737 }
kaf24@3354 738 }
kaf24@3354 739
kfraser@15074 740 if ( !initial_images_start )
kfraser@15074 741 EARLY_FAIL("Not enough memory to relocate the dom0 kernel image.\n");
kfraser@15074 742 reserve_in_boot_e820(initial_images_start, initial_images_end);
kfraser@15074 743
kfraser@15074 744 /*
kfraser@15074 745 * With modules (and Xen itself, on x86/64) relocated out of the way, we
kfraser@15074 746 * can now initialise the boot allocator with some memory.
kfraser@15074 747 */
kfraser@15074 748 xenheap_phys_start = init_boot_allocator(__pa(&_end));
kfraser@15074 749 xenheap_phys_end = opt_xenheap_megabytes << 20;
kfraser@15074 750 #if defined(CONFIG_X86_64)
kfraser@15074 751 if ( !xen_phys_start )
kfraser@15074 752 EARLY_FAIL("Not enough memory to relocate Xen.\n");
kfraser@15074 753 xenheap_phys_end += xen_phys_start;
kfraser@15074 754 reserve_in_boot_e820(xen_phys_start,
kfraser@15074 755 xen_phys_start + (opt_xenheap_megabytes<<20));
kfraser@15074 756 init_boot_pages(1<<20, 16<<20); /* Initial seed: 15MB */
kfraser@15074 757 #else
kfraser@15074 758 init_boot_pages(xenheap_phys_end, 16<<20); /* Initial seed: 4MB */
kfraser@15074 759 #endif
kfraser@15074 760
kfraser@15074 761 /*
kfraser@15074 762 * With the boot allocator now seeded, we can walk every RAM region and
kfraser@15074 763 * map it in its entirety (on x86/64, at least) and notify it to the
kfraser@15074 764 * boot allocator.
kfraser@15074 765 */
kfraser@15074 766 for ( i = 0; i < boot_e820.nr_map; i++ )
kfraser@15074 767 {
kfraser@15074 768 uint64_t s, e, map_e, mask = PAGE_SIZE - 1;
kfraser@15074 769
kfraser@15074 770 /* Only page alignment required now. */
kfraser@15074 771 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 772 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
kfraser@15074 773 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kfraser@15074 774 continue;
kfraser@15074 775
kfraser@15074 776 /* Perform the mapping (truncated in 32-bit mode). */
kfraser@15074 777 map_e = e;
kfraser@15074 778 #if defined(CONFIG_X86_32)
kfraser@15074 779 map_e = min_t(uint64_t, map_e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 780 #endif
kfraser@15074 781 if ( s < map_e )
kfraser@15074 782 map_pages_to_xen(
kfraser@15074 783 (unsigned long)maddr_to_bootstrap_virt(s),
kfraser@15074 784 s >> PAGE_SHIFT, (map_e-s) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kfraser@15074 785
kfraser@15074 786 init_boot_pages(s, e);
kfraser@15074 787 }
kfraser@15074 788
kfraser@15074 789 if ( (kexec_crash_area.size > 0) && (kexec_crash_area.start > 0) )
ian@12681 790 {
ian@12677 791 unsigned long kdump_start, kdump_size, k;
ian@12677 792
ian@12681 793 /* Mark images pages as free for now. */
ian@12677 794 init_boot_pages(initial_images_start, initial_images_end);
ian@12677 795
ian@12713 796 kdump_start = kexec_crash_area.start;
ian@12713 797 kdump_size = kexec_crash_area.size;
ian@12677 798
ian@12681 799 printk("Kdump: %luMB (%lukB) at 0x%lx\n",
ian@12677 800 kdump_size >> 20,
ian@12677 801 kdump_size >> 10,
ian@12677 802 kdump_start);
ian@12677 803
ian@12681 804 if ( (kdump_start & ~PAGE_MASK) || (kdump_size & ~PAGE_MASK) )
ian@12677 805 panic("Kdump parameters not page aligned\n");
ian@12677 806
ian@12677 807 kdump_start >>= PAGE_SHIFT;
ian@12677 808 kdump_size >>= PAGE_SHIFT;
ian@12677 809
kfraser@12853 810 /* Allocate pages for Kdump memory area. */
kfraser@14083 811 if ( !reserve_boot_pages(kdump_start, kdump_size) )
ian@12677 812 panic("Unable to reserve Kdump memory\n");
ian@12677 813
kfraser@12853 814 /* Allocate pages for relocated initial images. */
ian@12677 815 k = ((initial_images_end - initial_images_start) & ~PAGE_MASK) ? 1 : 0;
ian@12677 816 k += (initial_images_end - initial_images_start) >> PAGE_SHIFT;
ian@12677 817
kfraser@14084 818 #if defined(CONFIG_X86_32)
kfraser@14084 819 /* Must allocate within bootstrap 1:1 limits. */
kfraser@15074 820 k = alloc_boot_low_pages(k, 1); /* 0x0 - BOOTSTRAP_DIRECTMAP_END */
kfraser@14084 821 #else
kfraser@14084 822 k = alloc_boot_pages(k, 1);
kfraser@14084 823 #endif
kfraser@12853 824 if ( k == 0 )
ian@12677 825 panic("Unable to allocate initial images memory\n");
ian@12677 826
ian@12677 827 move_memory(k << PAGE_SHIFT, initial_images_start, initial_images_end);
ian@12677 828
ian@12677 829 initial_images_end -= initial_images_start;
ian@12677 830 initial_images_start = k << PAGE_SHIFT;
ian@12677 831 initial_images_end += initial_images_start;
ian@12681 832 }
ian@12677 833
kaf24@5003 834 memguard_init();
kaf24@4950 835
kfraser@15074 836 nr_pages = 0;
kfraser@15074 837 for ( i = 0; i < e820.nr_map; i++ )
kfraser@15074 838 if ( e820.map[i].type == E820_RAM )
kfraser@15074 839 nr_pages += e820.map[i].size >> PAGE_SHIFT;
ian@12681 840 printk("System RAM: %luMB (%lukB)\n",
kaf24@3354 841 nr_pages >> (20 - PAGE_SHIFT),
kaf24@3354 842 nr_pages << (PAGE_SHIFT - 10));
kaf24@7220 843 total_pages = nr_pages;
kaf24@3354 844
kfraser@11296 845 /* Sanity check for unwanted bloat of certain hypercall structures. */
kfraser@11296 846 BUILD_BUG_ON(sizeof(((struct xen_platform_op *)0)->u) !=
kfraser@11296 847 sizeof(((struct xen_platform_op *)0)->u.pad));
kfraser@11296 848 BUILD_BUG_ON(sizeof(((struct xen_domctl *)0)->u) !=
kfraser@11296 849 sizeof(((struct xen_domctl *)0)->u.pad));
kfraser@11296 850 BUILD_BUG_ON(sizeof(((struct xen_sysctl *)0)->u) !=
kfraser@11296 851 sizeof(((struct xen_sysctl *)0)->u.pad));
kaf24@7388 852
kaf24@9878 853 BUILD_BUG_ON(sizeof(start_info_t) > PAGE_SIZE);
kaf24@9878 854 BUILD_BUG_ON(sizeof(shared_info_t) > PAGE_SIZE);
ack@13292 855 BUILD_BUG_ON(sizeof(struct vcpu_info) != 64);
kaf24@7744 856
ack@13291 857 #ifdef CONFIG_COMPAT
ack@13291 858 BUILD_BUG_ON(sizeof(((struct compat_platform_op *)0)->u) !=
ack@13291 859 sizeof(((struct compat_platform_op *)0)->u.pad));
ack@13291 860 BUILD_BUG_ON(sizeof(start_info_compat_t) > PAGE_SIZE);
ack@13292 861 BUILD_BUG_ON(sizeof(struct compat_vcpu_info) != 64);
ack@13291 862 #endif
ack@13291 863
kfraser@10492 864 /* Check definitions in public headers match internal defs. */
kaf24@9878 865 BUILD_BUG_ON(__HYPERVISOR_VIRT_START != HYPERVISOR_VIRT_START);
kaf24@8521 866 #ifdef HYPERVISOR_VIRT_END
kaf24@9878 867 BUILD_BUG_ON(__HYPERVISOR_VIRT_END != HYPERVISOR_VIRT_END);
kaf24@8521 868 #endif
kfraser@10492 869 BUILD_BUG_ON(MACH2PHYS_VIRT_START != RO_MPT_VIRT_START);
kfraser@10492 870 BUILD_BUG_ON(MACH2PHYS_VIRT_END != RO_MPT_VIRT_END);
kaf24@8521 871
kaf24@3354 872 init_frametable();
kaf24@3338 873
kfraser@11971 874 acpi_boot_table_init();
kfraser@11971 875
kfraser@11971 876 acpi_numa_init();
kfraser@11971 877
kfraser@11971 878 numa_initmem_init(0, max_page);
kfraser@11971 879
kaf24@6111 880 /* Initialise the Xen heap, skipping RAM holes. */
kfraser@15074 881 init_xenheap_pages(xenheap_phys_start, xenheap_phys_end);
kfraser@15074 882 nr_pages = (xenheap_phys_end - xenheap_phys_start) >> PAGE_SHIFT;
kfraser@15074 883 #ifdef __x86_64__
kfraser@15074 884 init_xenheap_pages(xen_phys_start, __pa(&_start));
kfraser@15074 885 nr_pages += (__pa(&_start) - xen_phys_start) >> PAGE_SHIFT;
kfraser@15074 886 #endif
kfraser@15074 887 xenheap_phys_start = xen_phys_start;
kaf24@6111 888 printk("Xen heap: %luMB (%lukB)\n",
kaf24@6111 889 nr_pages >> (20 - PAGE_SHIFT),
kaf24@6111 890 nr_pages << (PAGE_SHIFT - 10));
kaf24@3338 891
keir@14680 892 end_boot_allocator();
keir@14680 893
kaf24@3594 894 early_boot = 0;
kaf24@3338 895
kaf24@8459 896 early_cpu_init();
kaf24@8459 897
kaf24@8459 898 paging_init();
kaf24@8459 899
kaf24@8459 900 /* Unmap the first page of CPU0's stack. */
kaf24@8459 901 memguard_guard_stack(cpu0_stack);
kaf24@8459 902
kaf24@8459 903 open_softirq(NEW_TLBFLUSH_CLOCK_PERIOD_SOFTIRQ, new_tlbflush_clock_period);
kaf24@8459 904
kaf24@8459 905 if ( opt_watchdog )
kaf24@8459 906 nmi_watchdog = NMI_LOCAL_APIC;
kaf24@8459 907
kaf24@8459 908 sort_exception_tables();
kaf24@8459 909
kaf24@8459 910 find_smp_config();
kaf24@8459 911
kaf24@8459 912 dmi_scan_machine();
kaf24@8459 913
kaf24@8459 914 generic_apic_probe();
kaf24@8459 915
kaf24@8459 916 acpi_boot_init();
kaf24@8459 917
kfraser@11971 918 init_cpu_to_node();
kfraser@11971 919
kfraser@11241 920 if ( smp_found_config )
kaf24@8459 921 get_smp_config();
kaf24@8459 922
keir@15083 923 #ifdef CONFIG_X86_64
keir@15083 924 /* Low mappings were only needed for some BIOS table parsing. */
keir@15083 925 zap_low_mappings();
keir@15083 926 #endif
keir@15083 927
kaf24@8459 928 init_apic_mappings();
kaf24@8459 929
kaf24@8459 930 init_IRQ();
kaf24@8459 931
kfraser@11241 932 percpu_init_areas();
kfraser@11241 933
kfraser@11240 934 init_idle_domain();
kfraser@11240 935
kaf24@8459 936 trap_init();
kaf24@8459 937
kaf24@13662 938 rcu_init();
kaf24@13662 939
kaf24@8586 940 timer_init();
kaf24@8459 941
kaf24@8459 942 early_time_init();
kaf24@8459 943
kaf24@8459 944 arch_init_memory();
kaf24@8459 945
kaf24@8459 946 identify_cpu(&boot_cpu_data);
kaf24@8459 947 if ( cpu_has_fxsr )
kaf24@8459 948 set_in_cr4(X86_CR4_OSFXSR);
kaf24@8459 949 if ( cpu_has_xmm )
kaf24@8459 950 set_in_cr4(X86_CR4_OSXMMEXCPT);
kaf24@8459 951
kaf24@8459 952 if ( opt_nosmp )
kaf24@8459 953 max_cpus = 0;
kaf24@8459 954
kaf24@8459 955 smp_prepare_cpus(max_cpus);
kaf24@8459 956
kaf24@8459 957 /*
kaf24@8459 958 * Initialise higher-level timer functions. We do this fairly late
kaf24@8459 959 * (post-SMP) because the time bases and scale factors need to be updated
kaf24@8459 960 * regularly, and SMP initialisation can cause a long delay with
kaf24@8459 961 * interrupts not yet enabled.
kaf24@8459 962 */
kaf24@8459 963 init_xen_time();
kaf24@8459 964
kaf24@8459 965 initialize_keytable();
kaf24@8459 966
kaf24@8459 967 serial_init_postirq();
kaf24@8459 968
kaf24@8459 969 BUG_ON(!local_irq_is_enabled());
kaf24@8459 970
kaf24@8459 971 for_each_present_cpu ( i )
kaf24@8459 972 {
kaf24@8459 973 if ( num_online_cpus() >= max_cpus )
kaf24@8459 974 break;
kaf24@8459 975 if ( !cpu_online(i) )
kaf24@13662 976 {
kaf24@13662 977 rcu_online_cpu(i);
kaf24@8459 978 __cpu_up(i);
kaf24@13662 979 }
kfraser@11971 980
kfraser@11998 981 /* Set up cpu_to_node[]. */
kfraser@11971 982 srat_detect_node(i);
kfraser@11998 983 /* Set up node_to_cpumask based on cpu_to_node[]. */
kfraser@11971 984 numa_add_cpu(i);
kaf24@8459 985 }
kaf24@8459 986
kaf24@8459 987 printk("Brought up %ld CPUs\n", (long)num_online_cpus());
kaf24@8459 988 smp_cpus_done(max_cpus);
kaf24@8459 989
kaf24@9117 990 initialise_gdb(); /* could be moved earlier */
kaf24@9117 991
kaf24@8459 992 do_initcalls();
kaf24@8459 993
kaf24@8594 994 if ( opt_watchdog )
kaf24@8594 995 watchdog_enable();
kaf24@8459 996
kfraser@11881 997 /* Extract policy from multiboot. */
kfraser@11881 998 extract_acm_policy(mbi, &initrdidx, &_policy_start, &_policy_len);
kfraser@11881 999
kaf24@8459 1000 /* initialize access control security module */
kfraser@11881 1001 acm_init(_policy_start, _policy_len);
kaf24@8459 1002
kaf24@8459 1003 /* Create initial domain 0. */
kfraser@14911 1004 dom0 = domain_create(0, 0, DOM0_SSIDREF);
kfraser@10655 1005 if ( (dom0 == NULL) || (alloc_vcpu(dom0, 0, 0) == NULL) )
kaf24@8459 1006 panic("Error creating domain 0\n");
kaf24@8459 1007
kfraser@12210 1008 dom0->is_privileged = 1;
kfraser@12210 1009
kaf24@8459 1010 /* Grab the DOM0 command line. */
kaf24@8459 1011 cmdline = (char *)(mod[0].string ? __va(mod[0].string) : NULL);
kaf24@8459 1012 if ( cmdline != NULL )
kaf24@8459 1013 {
kaf24@8459 1014 static char dom0_cmdline[MAX_GUEST_CMDLINE];
kaf24@8459 1015
kaf24@8459 1016 /* Skip past the image name and copy to a local buffer. */
kaf24@8459 1017 while ( *cmdline == ' ' ) cmdline++;
kaf24@8459 1018 if ( (cmdline = strchr(cmdline, ' ')) != NULL )
kaf24@8459 1019 {
kaf24@8459 1020 while ( *cmdline == ' ' ) cmdline++;
kfraser@13689 1021 safe_strcpy(dom0_cmdline, cmdline);
kaf24@8459 1022 }
kaf24@8459 1023
kaf24@8459 1024 /* Append any extra parameters. */
kfraser@13691 1025 if ( skip_ioapic_setup && !strstr(dom0_cmdline, "noapic") )
kfraser@13691 1026 safe_strcat(dom0_cmdline, " noapic");
kaf24@8459 1027 if ( acpi_skip_timer_override &&
kfraser@13691 1028 !strstr(dom0_cmdline, "acpi_skip_timer_override") )
kfraser@13691 1029 safe_strcat(dom0_cmdline, " acpi_skip_timer_override");
kfraser@13691 1030 if ( (strlen(acpi_param) != 0) && !strstr(dom0_cmdline, "acpi=") )
kaf24@8459 1031 {
kfraser@13691 1032 safe_strcat(dom0_cmdline, " acpi=");
kfraser@13691 1033 safe_strcat(dom0_cmdline, acpi_param);
kaf24@8459 1034 }
kfraser@13691 1035
kfraser@13691 1036 cmdline = dom0_cmdline;
kaf24@8459 1037 }
kaf24@8459 1038
kaf24@8459 1039 if ( (initrdidx > 0) && (initrdidx < mbi->mods_count) )
kaf24@8459 1040 {
kaf24@8459 1041 _initrd_start = initial_images_start +
kaf24@8459 1042 (mod[initrdidx].mod_start - mod[0].mod_start);
kaf24@8459 1043 _initrd_len = mod[initrdidx].mod_end - mod[initrdidx].mod_start;
kaf24@8459 1044 }
kaf24@8459 1045
kaf24@8459 1046 /*
kaf24@8459 1047 * We're going to setup domain0 using the module(s) that we stashed safely
kaf24@8459 1048 * above our heap. The second module, if present, is an initrd ramdisk.
kaf24@8459 1049 */
kaf24@8459 1050 if ( construct_dom0(dom0,
kaf24@8459 1051 initial_images_start,
kaf24@8459 1052 mod[0].mod_end-mod[0].mod_start,
kaf24@8459 1053 _initrd_start,
kaf24@8459 1054 _initrd_len,
kaf24@8459 1055 cmdline) != 0)
kaf24@8459 1056 panic("Could not set up DOM0 guest OS\n");
kaf24@8459 1057
kaf24@8459 1058 /* Scrub RAM that is still free and so may go to an unprivileged domain. */
kaf24@8459 1059 scrub_heap_pages();
kaf24@8459 1060
kaf24@8459 1061 init_trace_bufs();
kaf24@8459 1062
kaf24@10502 1063 console_endboot();
kaf24@8459 1064
kaf24@8459 1065 /* Hide UART from DOM0 if we're using it */
kaf24@8459 1066 serial_endboot();
kaf24@8459 1067
kaf24@8459 1068 domain_unpause_by_systemcontroller(dom0);
kaf24@8459 1069
keir@15082 1070 reset_stack_and_jump(init_done);
kaf24@8459 1071 }
kaf24@8459 1072
ian@13763 1073 void arch_get_xen_caps(xen_capabilities_info_t *info)
iap10@6721 1074 {
kfraser@14997 1075 /* Interface name is always xen-3.0-* for Xen-3.x. */
kfraser@14997 1076 int major = 3, minor = 0;
keir@13754 1077 char s[32];
keir@13754 1078
ian@13763 1079 (*info)[0] = '\0';
iap10@6721 1080
kaf24@6725 1081 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
kaf24@6725 1082
keir@13754 1083 snprintf(s, sizeof(s), "xen-%d.%d-x86_32 ", major, minor);
ian@13763 1084 safe_strcat(*info, s);
keir@13754 1085 if ( hvm_enabled )
keir@13754 1086 {
keir@13754 1087 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1088 safe_strcat(*info, s);
kfraser@13685 1089 }
kaf24@6725 1090
kaf24@6725 1091 #elif defined(CONFIG_X86_32) && defined(CONFIG_X86_PAE)
kaf24@6725 1092
keir@13754 1093 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1094 safe_strcat(*info, s);
kaf24@6725 1095 if ( hvm_enabled )
iap10@6721 1096 {
keir@13754 1097 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1098 safe_strcat(*info, s);
keir@13754 1099 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1100 safe_strcat(*info, s);
iap10@6721 1101 }
iap10@6721 1102
kaf24@6725 1103 #elif defined(CONFIG_X86_64)
iap10@6721 1104
keir@13754 1105 snprintf(s, sizeof(s), "xen-%d.%d-x86_64 ", major, minor);
ian@13763 1106 safe_strcat(*info, s);
ack@13288 1107 #ifdef CONFIG_COMPAT
keir@13754 1108 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1109 safe_strcat(*info, s);
ack@13288 1110 #endif
kaf24@6725 1111 if ( hvm_enabled )
iap10@6721 1112 {
keir@13754 1113 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1114 safe_strcat(*info, s);
keir@13754 1115 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1116 safe_strcat(*info, s);
keir@13754 1117 snprintf(s, sizeof(s), "hvm-%d.%d-x86_64 ", major, minor);
ian@13763 1118 safe_strcat(*info, s);
iap10@6721 1119 }
kaf24@6725 1120
iap10@6721 1121 #endif
iap10@6721 1122 }
iap10@6721 1123
kaf24@3914 1124 /*
kaf24@3914 1125 * Local variables:
kaf24@3914 1126 * mode: C
kaf24@3914 1127 * c-set-style: "BSD"
kaf24@3914 1128 * c-basic-offset: 4
kaf24@3914 1129 * tab-width: 4
kaf24@3914 1130 * indent-tabs-mode: nil
kaf24@3988 1131 * End:
kaf24@3914 1132 */