ia64/xen-unstable

annotate xen/arch/x86/setup.c @ 18920:df32584505c2

x86: enable interrupts explicitly in __start_xen()

Instead of relying on smp_prepare_cpus() (via check_nmi_watchdog()) or
init_xen_time() (via init_platform_timer() -> plt_overflow())
implicitly enabling interrupts, enable them explicitly once safe to do
so (it may actually be possible to move this even further up, but I
don't think that would buy us much).

Signed-off-by: Jan Beulich <jbeulich@novell.com>

Also move spin_debug_enable() a bit higer. Moving it above
smp_prepare_cpus() didn't work for some reason though!

Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
author Keir Fraser <keir.fraser@citrix.com>
date Thu Dec 11 13:25:28 2008 +0000 (2008-12-11)
parents 8e18dd41c6c7
children d77f66f89c04
rev   line source
kaf24@1452 1 #include <xen/config.h>
kaf24@1452 2 #include <xen/init.h>
kaf24@1452 3 #include <xen/lib.h>
kaf24@1452 4 #include <xen/sched.h>
cl349@5247 5 #include <xen/domain.h>
kaf24@1452 6 #include <xen/serial.h>
kaf24@1506 7 #include <xen/softirq.h>
kaf24@1452 8 #include <xen/acpi.h>
kaf24@3338 9 #include <xen/console.h>
iap10@4287 10 #include <xen/serial.h>
kaf24@3338 11 #include <xen/trace.h>
kaf24@3338 12 #include <xen/multiboot.h>
kaf24@5356 13 #include <xen/domain_page.h>
kfraser@10890 14 #include <xen/version.h>
kaf24@9117 15 #include <xen/gdbstub.h>
kaf24@9818 16 #include <xen/percpu.h>
kfraser@11296 17 #include <xen/hypercall.h>
kfraser@11601 18 #include <xen/keyhandler.h>
kfraser@11971 19 #include <xen/numa.h>
kaf24@13662 20 #include <xen/rcupdate.h>
keir@15298 21 #include <xen/vga.h>
keir@15988 22 #include <xen/dmi.h>
iap10@6721 23 #include <public/version.h>
ack@13291 24 #ifdef CONFIG_COMPAT
ack@13291 25 #include <compat/platform.h>
ack@13291 26 #include <compat/xen.h>
ack@13291 27 #endif
kaf24@1452 28 #include <asm/bitops.h>
kaf24@1452 29 #include <asm/smp.h>
kaf24@1452 30 #include <asm/processor.h>
kaf24@1452 31 #include <asm/mpspec.h>
kaf24@1452 32 #include <asm/apic.h>
kaf24@1452 33 #include <asm/desc.h>
Tim@13909 34 #include <asm/paging.h>
kaf24@3344 35 #include <asm/e820.h>
kfraser@15819 36 #include <xsm/acm/acm_hooks.h>
ian@12677 37 #include <xen/kexec.h>
kfraser@15336 38 #include <asm/edd.h>
kfraser@15815 39 #include <xsm/xsm.h>
keir@16274 40 #include <asm/tboot.h>
kaf24@3338 41
kfraser@15074 42 #if defined(CONFIG_X86_64)
kfraser@15597 43 #define BOOTSTRAP_DIRECTMAP_END (1UL << 32) /* 4GB */
kfraser@15074 44 #define maddr_to_bootstrap_virt(m) maddr_to_virt(m)
kfraser@15074 45 #else
kfraser@15597 46 #define BOOTSTRAP_DIRECTMAP_END (1UL << 30) /* 1GB */
kfraser@15074 47 #define maddr_to_bootstrap_virt(m) ((void *)(long)(m))
kfraser@15074 48 #endif
kfraser@15074 49
kaf24@5211 50 extern void generic_apic_probe(void);
kfraser@11971 51 extern void numa_initmem_init(unsigned long start_pfn, unsigned long end_pfn);
kaf24@5157 52
kfraser@15330 53 extern u16 boot_edid_caps;
kfraser@15330 54 extern u8 boot_edid_info[128];
kfraser@15330 55 extern struct boot_video_info boot_vid_info;
kfraser@15330 56
kaf24@3338 57 /*
kaf24@3338 58 * opt_xenheap_megabytes: Size of Xen heap in megabytes, excluding the
kaf24@8726 59 * page_info table and allocation bitmap.
kaf24@3338 60 */
kaf24@3338 61 static unsigned int opt_xenheap_megabytes = XENHEAP_DEFAULT_MB;
kaf24@4950 62 #if defined(CONFIG_X86_64)
kaf24@3338 63 integer_param("xenheap_megabytes", opt_xenheap_megabytes);
kaf24@3354 64 #endif
kaf24@1452 65
kaf24@5146 66 /* opt_nosmp: If true, secondary processors are ignored. */
kaf24@5900 67 static int opt_nosmp = 0;
kaf24@5146 68 boolean_param("nosmp", opt_nosmp);
kaf24@5146 69
kaf24@5146 70 /* maxcpus: maximum number of CPUs to activate. */
kaf24@5146 71 static unsigned int max_cpus = NR_CPUS;
shand@11156 72 integer_param("maxcpus", max_cpus);
kaf24@5146 73
kaf24@3334 74 /* opt_watchdog: If true, run a watchdog NMI on each processor. */
kaf24@3334 75 static int opt_watchdog = 0;
kaf24@3334 76 boolean_param("watchdog", opt_watchdog);
kaf24@3334 77
kaf24@4850 78 /* **** Linux config option: propagated to domain0. */
kaf24@4850 79 /* "acpi=off": Sisables both ACPI table parsing and interpreter. */
kaf24@4850 80 /* "acpi=force": Override the disable blacklist. */
kaf24@4850 81 /* "acpi=strict": Disables out-of-spec workarounds. */
kaf24@4850 82 /* "acpi=ht": Limit ACPI just to boot-time to enable HT. */
kaf24@4850 83 /* "acpi=noirq": Disables ACPI interrupt routing. */
kaf24@4850 84 static void parse_acpi_param(char *s);
kaf24@4850 85 custom_param("acpi", parse_acpi_param);
kaf24@4850 86
kaf24@4850 87 /* **** Linux config option: propagated to domain0. */
kaf24@4850 88 /* acpi_skip_timer_override: Skip IRQ0 overrides. */
kaf24@4850 89 extern int acpi_skip_timer_override;
kaf24@4850 90 boolean_param("acpi_skip_timer_override", acpi_skip_timer_override);
kaf24@4850 91
kaf24@4850 92 /* **** Linux config option: propagated to domain0. */
kaf24@4850 93 /* noapic: Disable IOAPIC setup. */
kaf24@4850 94 extern int skip_ioapic_setup;
kaf24@4850 95 boolean_param("noapic", skip_ioapic_setup);
kaf24@4850 96
keir@17546 97 /* **** Linux config option: propagated to domain0. */
keir@17657 98 /* xen_cpuidle: xen control cstate. */
keir@18051 99 /*static*/ int xen_cpuidle;
keir@17657 100 boolean_param("cpuidle", xen_cpuidle);
keir@17546 101
kaf24@3594 102 int early_boot = 1;
kaf24@3594 103
kaf24@5146 104 cpumask_t cpu_present_map;
kaf24@5146 105
kfraser@15074 106 unsigned long xen_phys_start;
kfraser@15074 107
kaf24@5003 108 /* Limits of Xen heap, used to initialise the allocator. */
kaf24@5003 109 unsigned long xenheap_phys_start, xenheap_phys_end;
kaf24@3338 110
kaf24@2298 111 extern void arch_init_memory(void);
kaf24@1589 112 extern void init_IRQ(void);
kaf24@5604 113 extern void early_time_init(void);
kaf24@5167 114 extern void early_cpu_init(void);
kfraser@15747 115 extern void vesa_init(void);
kfraser@15747 116 extern void vesa_mtrr_init(void);
kaf24@1589 117
keir@18523 118 DEFINE_PER_CPU(struct desc_struct *, gdt_table) = boot_cpu_gdt_table;
keir@18523 119 #ifdef CONFIG_COMPAT
keir@18523 120 DEFINE_PER_CPU(struct desc_struct *, compat_gdt_table)
keir@18523 121 = boot_cpu_compat_gdt_table;
keir@18523 122 #endif
keir@18523 123
kaf24@8533 124 struct tss_struct init_tss[NR_CPUS];
kaf24@8533 125
kfraser@15490 126 char __attribute__ ((__section__(".bss.stack_aligned"))) cpu0_stack[STACK_SIZE];
kaf24@5011 127
keir@16144 128 struct cpuinfo_x86 boot_cpu_data = { 0, 0, 0, 0, -1 };
kaf24@1452 129
kaf24@1670 130 unsigned long mmu_cr4_features = X86_CR4_PSE | X86_CR4_PGE | X86_CR4_PAE;
kaf24@1452 131 EXPORT_SYMBOL(mmu_cr4_features);
kaf24@1452 132
kaf24@4818 133 int acpi_disabled;
kaf24@1452 134
kaf24@4850 135 int acpi_force;
kaf24@4850 136 char acpi_param[10] = "";
kfraser@15074 137 static void __init parse_acpi_param(char *s)
kaf24@4850 138 {
kaf24@4850 139 /* Save the parameter so it can be propagated to domain0. */
kfraser@13689 140 safe_strcpy(acpi_param, s);
kaf24@4850 141
kaf24@4850 142 /* Interpret the parameter for use within Xen. */
kaf24@4850 143 if ( !strcmp(s, "off") )
kaf24@4850 144 {
kaf24@4850 145 disable_acpi();
kaf24@4850 146 }
kaf24@4850 147 else if ( !strcmp(s, "force") )
kaf24@4850 148 {
kaf24@4850 149 acpi_force = 1;
kaf24@4850 150 acpi_ht = 1;
kaf24@4850 151 acpi_disabled = 0;
kaf24@4850 152 }
kaf24@4850 153 else if ( !strcmp(s, "strict") )
kaf24@4850 154 {
kaf24@4850 155 acpi_strict = 1;
kaf24@4850 156 }
kaf24@4850 157 else if ( !strcmp(s, "ht") )
kaf24@4850 158 {
kaf24@4850 159 if ( !acpi_force )
kaf24@4850 160 disable_acpi();
kaf24@4850 161 acpi_ht = 1;
kaf24@4850 162 }
kaf24@4850 163 else if ( !strcmp(s, "noirq") )
kaf24@4850 164 {
kaf24@4850 165 acpi_noirq_set();
kaf24@4850 166 }
kaf24@4850 167 }
kaf24@4850 168
kaf24@1452 169 static void __init do_initcalls(void)
kaf24@1452 170 {
kaf24@1452 171 initcall_t *call;
kaf24@1452 172 for ( call = &__initcall_start; call < &__initcall_end; call++ )
kaf24@1452 173 (*call)();
kaf24@1452 174 }
kaf24@1452 175
kfraser@15074 176 #define EARLY_FAIL(f, a...) do { \
kfraser@15074 177 printk( f , ## a ); \
kfraser@15871 178 for ( ; ; ) halt(); \
kfraser@15074 179 } while (0)
kaf24@8459 180
kfraser@15074 181 static unsigned long __initdata initial_images_start, initial_images_end;
kaf24@9067 182
kfraser@15074 183 unsigned long __init initial_images_nrpages(void)
kaf24@9067 184 {
kfraser@15489 185 ASSERT(!(initial_images_start & ~PAGE_MASK));
kfraser@15489 186 ASSERT(!(initial_images_end & ~PAGE_MASK));
kfraser@15489 187 return ((initial_images_end >> PAGE_SHIFT) -
kfraser@15489 188 (initial_images_start >> PAGE_SHIFT));
kaf24@9067 189 }
kaf24@9067 190
kfraser@15074 191 void __init discard_initial_images(void)
kaf24@9067 192 {
kaf24@9067 193 init_domheap_pages(initial_images_start, initial_images_end);
kaf24@9067 194 }
kaf24@9067 195
kaf24@9818 196 extern char __per_cpu_start[], __per_cpu_data_end[], __per_cpu_end[];
kaf24@9818 197
kfraser@11241 198 static void __init percpu_init_areas(void)
kaf24@9818 199 {
kaf24@9818 200 unsigned int i, data_size = __per_cpu_data_end - __per_cpu_start;
kfraser@15074 201 unsigned int first_unused;
kaf24@9818 202
kaf24@9818 203 BUG_ON(data_size > PERCPU_SIZE);
kaf24@9818 204
kfraser@15074 205 /* Initialise per-cpu data area for all possible secondary CPUs. */
kfraser@15074 206 for ( i = 1; (i < NR_CPUS) && cpu_possible(i); i++ )
kfraser@15074 207 memcpy(__per_cpu_start + (i << PERCPU_SHIFT),
kfraser@15074 208 __per_cpu_start,
kfraser@15074 209 data_size);
kaf24@9818 210 first_unused = i;
kaf24@9818 211
kfraser@14340 212 /* Check that there are no holes in cpu_possible_map. */
kaf24@9818 213 for ( ; i < NR_CPUS; i++ )
kfraser@14340 214 BUG_ON(cpu_possible(i));
kaf24@9818 215
kfraser@11241 216 #ifndef MEMORY_GUARD
kaf24@9818 217 init_xenheap_pages(__pa(__per_cpu_start) + (first_unused << PERCPU_SHIFT),
kaf24@9818 218 __pa(__per_cpu_end));
kfraser@11241 219 #endif
keir@15082 220 memguard_guard_range(&__per_cpu_start[first_unused << PERCPU_SHIFT],
keir@15082 221 (NR_CPUS - first_unused) << PERCPU_SHIFT);
keir@15082 222 #if defined(CONFIG_X86_64)
keir@15082 223 /* Also zap the mapping in the 1:1 area. */
keir@15082 224 memguard_guard_range(__va(__pa(__per_cpu_start)) +
keir@15082 225 (first_unused << PERCPU_SHIFT),
keir@15082 226 (NR_CPUS - first_unused) << PERCPU_SHIFT);
keir@15082 227 #endif
kaf24@9818 228 }
kaf24@9818 229
kfraser@11241 230 static void __init init_idle_domain(void)
kfraser@11240 231 {
kfraser@11240 232 struct domain *idle_domain;
kfraser@11240 233
kfraser@11240 234 /* Domain creation requires that scheduler structures are initialised. */
kfraser@11240 235 scheduler_init();
kfraser@11240 236
kfraser@14911 237 idle_domain = domain_create(IDLE_DOMAIN_ID, 0, 0);
kfraser@11240 238 if ( (idle_domain == NULL) || (alloc_vcpu(idle_domain, 0, 0) == NULL) )
kfraser@11240 239 BUG();
kfraser@11240 240
kfraser@11240 241 set_current(idle_domain->vcpu[0]);
kfraser@11240 242 idle_vcpu[0] = this_cpu(curr_vcpu) = current;
kfraser@11240 243
kfraser@11240 244 setup_idle_pagetable();
kfraser@11240 245 }
kfraser@11240 246
kfraser@15074 247 static void __init srat_detect_node(int cpu)
kfraser@11971 248 {
kfraser@11998 249 unsigned node;
keir@17551 250 u32 apicid = x86_cpu_to_apicid[cpu];
kfraser@11971 251
kfraser@11998 252 node = apicid_to_node[apicid];
kfraser@11998 253 if ( node == NUMA_NO_NODE )
kfraser@11998 254 node = 0;
kfraser@11998 255 numa_set_node(cpu, node);
kfraser@11971 256
kfraser@11998 257 if ( acpi_numa > 0 )
kfraser@11998 258 printk(KERN_INFO "CPU %d APIC %d -> Node %d\n", cpu, apicid, node);
kfraser@11971 259 }
kfraser@11971 260
kfraser@15740 261 /*
kfraser@15740 262 * Ensure a given physical memory range is present in the bootstrap mappings.
kfraser@15740 263 * Use superpage mappings to ensure that pagetable memory needn't be allocated.
kfraser@15740 264 */
kfraser@15740 265 static void __init bootstrap_map(unsigned long start, unsigned long end)
kfraser@15740 266 {
kfraser@15740 267 unsigned long mask = (1UL << L2_PAGETABLE_SHIFT) - 1;
keir@16888 268 start = max_t(unsigned long, start & ~mask, 16UL << 20);
kfraser@15740 269 end = (end + mask) & ~mask;
keir@16888 270 if ( start >= end )
keir@16888 271 return;
kfraser@15740 272 if ( end > BOOTSTRAP_DIRECTMAP_END )
kfraser@15740 273 panic("Cannot access memory beyond end of "
kfraser@15740 274 "bootstrap direct-map area\n");
kfraser@15740 275 map_pages_to_xen(
kfraser@15740 276 (unsigned long)maddr_to_bootstrap_virt(start),
kfraser@15740 277 start >> PAGE_SHIFT, (end-start) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kfraser@15740 278 }
kfraser@15740 279
kfraser@15074 280 static void __init move_memory(
kfraser@15074 281 unsigned long dst, unsigned long src_start, unsigned long src_end)
ian@12677 282 {
kfraser@15740 283 bootstrap_map(src_start, src_end);
kfraser@15740 284 bootstrap_map(dst, dst + src_end - src_start);
kfraser@15074 285 memmove(maddr_to_bootstrap_virt(dst),
kfraser@15074 286 maddr_to_bootstrap_virt(src_start),
ian@12677 287 src_end - src_start);
kfraser@15074 288 }
kfraser@15074 289
kfraser@15074 290 /* A temporary copy of the e820 map that we can mess with during bootstrap. */
kfraser@15074 291 static struct e820map __initdata boot_e820;
kfraser@15074 292
keir@15298 293 struct boot_video_info {
keir@15298 294 u8 orig_x; /* 0x00 */
keir@15298 295 u8 orig_y; /* 0x01 */
keir@15298 296 u8 orig_video_mode; /* 0x02 */
keir@15298 297 u8 orig_video_cols; /* 0x03 */
keir@15298 298 u8 orig_video_lines; /* 0x04 */
keir@15298 299 u8 orig_video_isVGA; /* 0x05 */
keir@15298 300 u16 orig_video_points; /* 0x06 */
keir@15298 301
keir@15298 302 /* VESA graphic mode -- linear frame buffer */
keir@15298 303 u32 capabilities; /* 0x08 */
keir@15298 304 u16 lfb_linelength; /* 0x0c */
keir@15298 305 u16 lfb_width; /* 0x0e */
keir@15298 306 u16 lfb_height; /* 0x10 */
keir@15298 307 u16 lfb_depth; /* 0x12 */
keir@15298 308 u32 lfb_base; /* 0x14 */
keir@15298 309 u32 lfb_size; /* 0x18 */
keir@15298 310 u8 red_size; /* 0x1c */
keir@15298 311 u8 red_pos; /* 0x1d */
keir@15298 312 u8 green_size; /* 0x1e */
keir@15298 313 u8 green_pos; /* 0x1f */
keir@15298 314 u8 blue_size; /* 0x20 */
keir@15298 315 u8 blue_pos; /* 0x21 */
keir@15298 316 u8 rsvd_size; /* 0x22 */
keir@15298 317 u8 rsvd_pos; /* 0x23 */
keir@15298 318 u16 vesapm_seg; /* 0x24 */
keir@15298 319 u16 vesapm_off; /* 0x26 */
keir@16124 320 u16 vesa_attrib; /* 0x28 */
keir@15298 321 };
keir@15298 322
keir@15298 323 static void __init parse_video_info(void)
keir@15298 324 {
keir@15298 325 struct boot_video_info *bvi = &bootsym(boot_vid_info);
keir@15298 326
keir@15298 327 if ( (bvi->orig_video_isVGA == 1) && (bvi->orig_video_mode == 3) )
keir@15298 328 {
keir@15298 329 vga_console_info.video_type = XEN_VGATYPE_TEXT_MODE_3;
keir@15298 330 vga_console_info.u.text_mode_3.font_height = bvi->orig_video_points;
keir@15298 331 vga_console_info.u.text_mode_3.cursor_x = bvi->orig_x;
keir@15298 332 vga_console_info.u.text_mode_3.cursor_y = bvi->orig_y;
keir@15298 333 vga_console_info.u.text_mode_3.rows = bvi->orig_video_lines;
keir@15298 334 vga_console_info.u.text_mode_3.columns = bvi->orig_video_cols;
keir@15298 335 }
keir@15298 336 else if ( bvi->orig_video_isVGA == 0x23 )
keir@15298 337 {
keir@15298 338 vga_console_info.video_type = XEN_VGATYPE_VESA_LFB;
keir@15298 339 vga_console_info.u.vesa_lfb.width = bvi->lfb_width;
keir@15298 340 vga_console_info.u.vesa_lfb.height = bvi->lfb_height;
keir@15298 341 vga_console_info.u.vesa_lfb.bytes_per_line = bvi->lfb_linelength;
keir@15298 342 vga_console_info.u.vesa_lfb.bits_per_pixel = bvi->lfb_depth;
keir@15298 343 vga_console_info.u.vesa_lfb.lfb_base = bvi->lfb_base;
keir@15298 344 vga_console_info.u.vesa_lfb.lfb_size = bvi->lfb_size;
keir@15298 345 vga_console_info.u.vesa_lfb.red_pos = bvi->red_pos;
keir@15298 346 vga_console_info.u.vesa_lfb.red_size = bvi->red_size;
keir@15298 347 vga_console_info.u.vesa_lfb.green_pos = bvi->green_pos;
keir@15298 348 vga_console_info.u.vesa_lfb.green_size = bvi->green_size;
keir@15298 349 vga_console_info.u.vesa_lfb.blue_pos = bvi->blue_pos;
keir@15298 350 vga_console_info.u.vesa_lfb.blue_size = bvi->blue_size;
keir@15298 351 vga_console_info.u.vesa_lfb.rsvd_pos = bvi->rsvd_pos;
keir@15298 352 vga_console_info.u.vesa_lfb.rsvd_size = bvi->rsvd_size;
keir@16124 353 vga_console_info.u.vesa_lfb.gbl_caps = bvi->capabilities;
keir@16124 354 vga_console_info.u.vesa_lfb.mode_attrs = bvi->vesa_attrib;
keir@15298 355 }
keir@15298 356 }
keir@15298 357
keir@16563 358 void __init kexec_reserve_area(struct e820map *e820)
keir@16563 359 {
keir@16563 360 unsigned long kdump_start = kexec_crash_area.start;
keir@16563 361 unsigned long kdump_size = kexec_crash_area.size;
keir@16563 362 static int is_reserved = 0;
keir@16563 363
keir@16563 364 kdump_size = (kdump_size + PAGE_SIZE - 1) & PAGE_MASK;
keir@16563 365
keir@16563 366 if ( (kdump_start == 0) || (kdump_size == 0) || is_reserved )
keir@16563 367 return;
keir@16563 368
keir@16563 369 is_reserved = 1;
keir@16563 370
keir@17674 371 if ( !reserve_e820_ram(e820, kdump_start, kdump_start + kdump_size) )
keir@16563 372 {
keir@16563 373 printk("Kdump: DISABLED (failed to reserve %luMB (%lukB) at 0x%lx)"
keir@16563 374 "\n", kdump_size >> 20, kdump_size >> 10, kdump_start);
keir@16563 375 kexec_crash_area.start = kexec_crash_area.size = 0;
keir@16563 376 }
keir@16563 377 else
keir@16563 378 {
keir@16563 379 printk("Kdump: %luMB (%lukB) at 0x%lx\n",
keir@16563 380 kdump_size >> 20, kdump_size >> 10, kdump_start);
keir@16563 381 }
keir@16563 382 }
keir@16563 383
keir@15082 384 void init_done(void)
keir@15082 385 {
keir@15082 386 extern char __init_begin[], __init_end[];
keir@15082 387
keir@15082 388 /* Free (or page-protect) the init areas. */
keir@15082 389 #ifndef MEMORY_GUARD
keir@15082 390 init_xenheap_pages(__pa(__init_begin), __pa(__init_end));
keir@15082 391 #endif
keir@15082 392 memguard_guard_range(__init_begin, __init_end - __init_begin);
keir@15082 393 #if defined(CONFIG_X86_64)
keir@15082 394 /* Also zap the mapping in the 1:1 area. */
keir@15082 395 memguard_guard_range(__va(__pa(__init_begin)), __init_end - __init_begin);
keir@15082 396 #endif
keir@15082 397 printk("Freed %ldkB init memory.\n", (long)(__init_end-__init_begin)>>10);
keir@15082 398
keir@15082 399 startup_cpu_idle_loop();
keir@15082 400 }
keir@15082 401
kfraser@15796 402 static char * __init cmdline_cook(char *p)
kfraser@15796 403 {
kfraser@15796 404 p = p ? : "";
kfraser@15796 405 while ( *p == ' ' )
kfraser@15796 406 p++;
kfraser@15796 407 while ( (*p != ' ') && (*p != '\0') )
kfraser@15796 408 p++;
kfraser@15796 409 while ( *p == ' ' )
kfraser@15796 410 p++;
kfraser@15796 411 return p;
kfraser@15796 412 }
kfraser@15796 413
kfraser@15379 414 void __init __start_xen(unsigned long mbi_p)
kaf24@1452 415 {
kfraser@15293 416 char *memmap_type = NULL;
kfraser@15796 417 char *cmdline, *kextra;
kaf24@8457 418 unsigned long _initrd_start = 0, _initrd_len = 0;
kaf24@8457 419 unsigned int initrdidx = 1;
kfraser@15379 420 multiboot_info_t *mbi = __va(mbi_p);
kaf24@8457 421 module_t *mod = (module_t *)__va(mbi->mods_addr);
kaf24@6111 422 unsigned long nr_pages, modules_length;
kfraser@15293 423 int i, e820_warn = 0, bytes = 0;
kaf24@5776 424 struct ns16550_defaults ns16550 = {
kaf24@5776 425 .data_bits = 8,
kaf24@5776 426 .parity = 'n',
kaf24@5776 427 .stop_bits = 1
kaf24@5776 428 };
kaf24@3338 429
kfraser@12853 430 extern void early_page_fault(void);
kfraser@12853 431 set_intr_gate(TRAP_page_fault, &early_page_fault);
kfraser@12853 432
kaf24@3338 433 /* Parse the command-line options. */
kfraser@15796 434 cmdline = cmdline_cook((mbi->flags & MBI_CMDLINE) ?
kfraser@15796 435 __va(mbi->cmdline) : NULL);
kfraser@15426 436 if ( (kextra = strstr(cmdline, " -- ")) != NULL )
kfraser@15426 437 {
kfraser@15426 438 /*
kfraser@15426 439 * Options after ' -- ' separator belong to dom0.
kfraser@15426 440 * 1. Orphan dom0's options from Xen's command line.
kfraser@15426 441 * 2. Skip all but final leading space from dom0's options.
kfraser@15426 442 */
kfraser@15426 443 *kextra = '\0';
kfraser@15426 444 kextra += 3;
kfraser@15426 445 while ( kextra[1] == ' ' ) kextra++;
kfraser@15426 446 }
kaf24@9823 447 cmdline_parse(cmdline);
kaf24@3338 448
keir@15298 449 parse_video_info();
keir@15298 450
kaf24@8534 451 set_current((struct vcpu *)0xfffff000); /* debug sanity */
keir@18790 452 idle_vcpu[0] = current;
kaf24@8534 453 set_processor_id(0); /* needed early, for smp_processor_id() */
keir@16378 454 if ( cpu_has_efer )
keir@16378 455 rdmsrl(MSR_EFER, this_cpu(efer));
keir@16267 456 asm volatile ( "mov %%cr4,%0" : "=r" (this_cpu(cr4)) );
kaf24@3338 457
kaf24@5146 458 smp_prepare_boot_cpu();
kaf24@5146 459
kaf24@3338 460 /* We initialise the serial devices very early so we can get debugging. */
kaf24@5776 461 ns16550.io_base = 0x3f8;
kaf24@5776 462 ns16550.irq = 4;
kaf24@5776 463 ns16550_init(0, &ns16550);
kaf24@5776 464 ns16550.io_base = 0x2f8;
kaf24@5776 465 ns16550.irq = 3;
kaf24@5776 466 ns16550_init(1, &ns16550);
kaf24@5195 467 serial_init_preirq();
kaf24@3338 468
kaf24@3338 469 init_console();
kaf24@3338 470
kfraser@11947 471 printk("Command line: %s\n", cmdline);
kaf24@9823 472
kfraser@15330 473 printk("Video information:\n");
kfraser@15330 474
kfraser@15330 475 /* Print VGA display mode information. */
keir@15298 476 switch ( vga_console_info.video_type )
keir@15298 477 {
keir@15298 478 case XEN_VGATYPE_TEXT_MODE_3:
kfraser@15330 479 printk(" VGA is text mode %dx%d, font 8x%d\n",
keir@15298 480 vga_console_info.u.text_mode_3.columns,
keir@15298 481 vga_console_info.u.text_mode_3.rows,
keir@15298 482 vga_console_info.u.text_mode_3.font_height);
keir@15298 483 break;
keir@15298 484 case XEN_VGATYPE_VESA_LFB:
kfraser@15330 485 printk(" VGA is graphics mode %dx%d, %d bpp\n",
keir@15298 486 vga_console_info.u.vesa_lfb.width,
keir@15298 487 vga_console_info.u.vesa_lfb.height,
keir@15298 488 vga_console_info.u.vesa_lfb.bits_per_pixel);
keir@15298 489 break;
kfraser@15330 490 default:
kfraser@15330 491 printk(" No VGA detected\n");
kfraser@15330 492 break;
kfraser@15330 493 }
kfraser@15330 494
kfraser@15330 495 /* Print VBE/DDC EDID information. */
kfraser@15330 496 if ( bootsym(boot_edid_caps) != 0x1313 )
kfraser@15330 497 {
kfraser@15330 498 u16 caps = bootsym(boot_edid_caps);
kfraser@15330 499 printk(" VBE/DDC methods:%s%s%s; ",
kfraser@15330 500 (caps & 1) ? " V1" : "",
kfraser@15330 501 (caps & 2) ? " V2" : "",
kfraser@15330 502 !(caps & 3) ? " none" : "");
kfraser@15330 503 printk("EDID transfer time: %d seconds\n", caps >> 8);
kfraser@15330 504 if ( *(u32 *)bootsym(boot_edid_info) == 0x13131313 )
kfraser@15330 505 {
kfraser@15330 506 printk(" EDID info not retrieved because ");
kfraser@15330 507 if ( !(caps & 3) )
kfraser@15330 508 printk("no DDC retrieval method detected\n");
kfraser@15330 509 else if ( (caps >> 8) > 5 )
kfraser@15330 510 printk("takes longer than 5 seconds\n");
kfraser@15330 511 else
kfraser@15330 512 printk("of reasons unknown\n");
kfraser@15330 513 }
keir@15298 514 }
keir@15298 515
kfraser@15336 516 printk("Disc information:\n");
kfraser@15336 517 printk(" Found %d MBR signatures\n",
kfraser@15430 518 bootsym(boot_mbr_signature_nr));
kfraser@15336 519 printk(" Found %d EDD information structures\n",
kfraser@15336 520 bootsym(boot_edd_info_nr));
kfraser@15336 521
kaf24@3344 522 /* Check that we have at least one Multiboot module. */
kaf24@3344 523 if ( !(mbi->flags & MBI_MODULES) || (mbi->mods_count == 0) )
kfraser@15074 524 EARLY_FAIL("dom0 kernel not specified. "
kfraser@15074 525 "Check bootloader configuration.\n");
kaf24@5011 526
kaf24@5011 527 if ( ((unsigned long)cpu0_stack & (STACK_SIZE-1)) != 0 )
kfraser@15074 528 EARLY_FAIL("Misaligned CPU0 stack.\n");
kaf24@3338 529
kfraser@15293 530 if ( e820_raw_nr != 0 )
kfraser@15292 531 {
kfraser@15293 532 memmap_type = "Xen-e820";
kfraser@15292 533 }
kfraser@15293 534 else if ( bootsym(lowmem_kb) )
kfraser@15292 535 {
kfraser@15293 536 memmap_type = "Xen-e801";
kfraser@15292 537 e820_raw[0].addr = 0;
kfraser@15293 538 e820_raw[0].size = bootsym(lowmem_kb) << 10;
kfraser@15292 539 e820_raw[0].type = E820_RAM;
kfraser@15292 540 e820_raw[1].addr = 0x100000;
kfraser@15293 541 e820_raw[1].size = bootsym(highmem_kb) << 10;
kfraser@15292 542 e820_raw[1].type = E820_RAM;
kfraser@15292 543 e820_raw_nr = 2;
kfraser@15292 544 }
kfraser@15292 545 else if ( mbi->flags & MBI_MEMMAP )
kaf24@3344 546 {
kfraser@15293 547 memmap_type = "Multiboot-e820";
keir@15988 548 while ( (bytes < mbi->mmap_length) && (e820_raw_nr < E820MAX) )
kaf24@3344 549 {
kaf24@3344 550 memory_map_t *map = __va(mbi->mmap_addr + bytes);
kaf24@8402 551
kaf24@8402 552 /*
kaf24@8403 553 * This is a gross workaround for a BIOS bug. Some bootloaders do
kaf24@8402 554 * not write e820 map entries into pre-zeroed memory. This is
kaf24@8402 555 * okay if the BIOS fills in all fields of the map entry, but
kaf24@8402 556 * some broken BIOSes do not bother to write the high word of
kaf24@8402 557 * the length field if the length is smaller than 4GB. We
kaf24@8402 558 * detect and fix this by flagging sections below 4GB that
kaf24@8403 559 * appear to be larger than 4GB in size.
kaf24@8402 560 */
kaf24@8403 561 if ( (map->base_addr_high == 0) && (map->length_high != 0) )
kaf24@8402 562 {
kfraser@15292 563 if ( !e820_warn )
kfraser@15292 564 {
kfraser@15292 565 printk("WARNING: Buggy e820 map detected and fixed "
kfraser@15292 566 "(truncated length fields).\n");
kfraser@15292 567 e820_warn = 1;
kfraser@15292 568 }
kaf24@8402 569 map->length_high = 0;
kaf24@8402 570 }
kaf24@8402 571
kaf24@3344 572 e820_raw[e820_raw_nr].addr =
kaf24@3344 573 ((u64)map->base_addr_high << 32) | (u64)map->base_addr_low;
kaf24@3344 574 e820_raw[e820_raw_nr].size =
kaf24@3344 575 ((u64)map->length_high << 32) | (u64)map->length_low;
kfraser@15799 576 e820_raw[e820_raw_nr].type = map->type;
kaf24@3344 577 e820_raw_nr++;
kaf24@8402 578
kaf24@3344 579 bytes += map->size + 4;
kaf24@3344 580 }
kaf24@3344 581 }
kaf24@3344 582 else if ( mbi->flags & MBI_MEMLIMITS )
kaf24@3344 583 {
kfraser@15293 584 memmap_type = "Multiboot-e801";
kaf24@3344 585 e820_raw[0].addr = 0;
kaf24@3344 586 e820_raw[0].size = mbi->mem_lower << 10;
kaf24@3344 587 e820_raw[0].type = E820_RAM;
kaf24@3354 588 e820_raw[1].addr = 0x100000;
kaf24@3354 589 e820_raw[1].size = mbi->mem_upper << 10;
kaf24@3354 590 e820_raw[1].type = E820_RAM;
kaf24@3344 591 e820_raw_nr = 2;
kaf24@3344 592 }
kaf24@3344 593 else
kaf24@3344 594 {
kfraser@15074 595 EARLY_FAIL("Bootloader provided no memory information.\n");
kaf24@3344 596 }
kaf24@3344 597
kaf24@13427 598 /* Sanitise the raw E820 map to produce a final clean version. */
kfraser@15293 599 max_page = init_e820(memmap_type, e820_raw, &e820_raw_nr);
kaf24@3338 600
keir@18131 601 #ifdef CONFIG_X86_64
keir@18131 602 /*
keir@18131 603 * On x86/64 we are able to account for the allocation bitmap
keir@18131 604 * (allocated in common/page_alloc.c:init_boot_allocator()) stealing
keir@18131 605 * from the Xen heap. Here we make the Xen heap appropriately larger.
keir@18131 606 */
keir@18131 607 opt_xenheap_megabytes += (max_page / 8) >> 20;
keir@18131 608 #endif
keir@18131 609
keir@18131 610 /*
keir@18131 611 * Since there are some stubs getting built on the stacks which use
keir@18131 612 * direct calls/jumps, the heap must be confined to the lower 2G so
keir@18131 613 * that those branches can reach their targets.
keir@18131 614 */
keir@18131 615 if ( opt_xenheap_megabytes > 2048 )
keir@18131 616 opt_xenheap_megabytes = 2048;
keir@18131 617
keir@16563 618 /* Create a temporary copy of the E820 map. */
kfraser@15074 619 memcpy(&boot_e820, &e820, sizeof(e820));
keir@16563 620
keir@16563 621 /* Early kexec reservation (explicit static start address). */
keir@16563 622 kexec_reserve_area(&boot_e820);
kaf24@6111 623
kfraser@15074 624 /*
keir@15077 625 * Iterate backwards over all superpage-aligned RAM regions.
kfraser@15074 626 *
kfraser@15074 627 * We require superpage alignment because the boot allocator is not yet
kfraser@15074 628 * initialised. Hence we can only map superpages in the address range
kfraser@15074 629 * 0 to BOOTSTRAP_DIRECTMAP_END, as this is guaranteed not to require
kfraser@15074 630 * dynamic allocation of pagetables.
kfraser@15074 631 *
kfraser@15074 632 * As well as mapping superpages in that range, in preparation for
kfraser@15074 633 * initialising the boot allocator, we also look for a region to which
kfraser@15074 634 * we can relocate the dom0 kernel and other multiboot modules. Also, on
kfraser@15074 635 * x86/64, we relocate Xen to higher memory.
kfraser@15074 636 */
kfraser@15074 637 modules_length = mod[mbi->mods_count-1].mod_end - mod[0].mod_start;
keir@15077 638 for ( i = boot_e820.nr_map-1; i >= 0; i-- )
kfraser@15074 639 {
kfraser@15074 640 uint64_t s, e, mask = (1UL << L2_PAGETABLE_SHIFT) - 1;
kaf24@6134 641
keir@16563 642 /* Superpage-aligned chunks from 16MB to BOOTSTRAP_DIRECTMAP_END. */
kfraser@15074 643 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 644 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
keir@16563 645 s = max_t(uint64_t, s, 16 << 20);
kfraser@15074 646 e = min_t(uint64_t, e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 647 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kaf24@3354 648 continue;
kaf24@6111 649
kfraser@15074 650 /* Map the chunk. No memory will need to be allocated to do this. */
kfraser@15074 651 map_pages_to_xen(
kfraser@15074 652 (unsigned long)maddr_to_bootstrap_virt(s),
kfraser@15074 653 s >> PAGE_SHIFT, (e-s) >> PAGE_SHIFT, PAGE_HYPERVISOR);
kaf24@6111 654
kfraser@14084 655 #if defined(CONFIG_X86_64)
kfraser@15074 656 /* Is the region suitable for relocating Xen? */
kfraser@15074 657 if ( !xen_phys_start && (((e-s) >> 20) >= opt_xenheap_megabytes) )
kaf24@5003 658 {
kfraser@15074 659 extern l2_pgentry_t l2_xenmap[];
kfraser@15074 660 l4_pgentry_t *pl4e;
kfraser@15074 661 l3_pgentry_t *pl3e;
kfraser@15074 662 l2_pgentry_t *pl2e;
keir@16888 663 int i, j, k;
kfraser@15074 664
kfraser@15074 665 /* Select relocation address. */
kfraser@15074 666 e = (e - (opt_xenheap_megabytes << 20)) & ~mask;
kfraser@15074 667 xen_phys_start = e;
kfraser@15292 668 bootsym(trampoline_xen_phys_start) = e;
kfraser@15074 669
kfraser@15074 670 /*
kfraser@15074 671 * Perform relocation to new physical address.
kfraser@15074 672 * Before doing so we must sync static/global data with main memory
kfraser@15074 673 * with a barrier(). After this we must *not* modify static/global
kfraser@15074 674 * data until after we have switched to the relocated pagetables!
kfraser@15074 675 */
kfraser@15074 676 barrier();
kfraser@15074 677 move_memory(e, 0, __pa(&_end) - xen_phys_start);
kfraser@15074 678
kfraser@15379 679 /* Poison low 1MB to detect stray pointers to physical 0-1MB. */
kfraser@15379 680 memset(maddr_to_bootstrap_virt(e), 0x55, 1U<<20);
kfraser@15379 681
kfraser@15074 682 /* Walk initial pagetables, relocating page directory entries. */
kfraser@15074 683 pl4e = __va(__pa(idle_pg_table));
kfraser@15074 684 for ( i = 0 ; i < L4_PAGETABLE_ENTRIES; i++, pl4e++ )
kfraser@15074 685 {
kfraser@15074 686 if ( !(l4e_get_flags(*pl4e) & _PAGE_PRESENT) )
kfraser@15074 687 continue;
kfraser@15074 688 *pl4e = l4e_from_intpte(l4e_get_intpte(*pl4e) +
kfraser@15074 689 xen_phys_start);
kfraser@15074 690 pl3e = l4e_to_l3e(*pl4e);
kfraser@15074 691 for ( j = 0; j < L3_PAGETABLE_ENTRIES; j++, pl3e++ )
kfraser@15074 692 {
keir@16921 693 /* Not present, 1GB mapping, or already relocated? */
kfraser@15074 694 if ( !(l3e_get_flags(*pl3e) & _PAGE_PRESENT) ||
keir@16921 695 (l3e_get_flags(*pl3e) & _PAGE_PSE) ||
kfraser@15074 696 (l3e_get_pfn(*pl3e) > 0x1000) )
kfraser@15074 697 continue;
kfraser@15074 698 *pl3e = l3e_from_intpte(l3e_get_intpte(*pl3e) +
kfraser@15074 699 xen_phys_start);
keir@16888 700 pl2e = l3e_to_l2e(*pl3e);
keir@16888 701 for ( k = 0; k < L2_PAGETABLE_ENTRIES; k++, pl2e++ )
keir@16888 702 {
keir@16888 703 /* Not present, PSE, or already relocated? */
keir@16888 704 if ( !(l2e_get_flags(*pl2e) & _PAGE_PRESENT) ||
keir@16888 705 (l2e_get_flags(*pl2e) & _PAGE_PSE) ||
keir@16888 706 (l2e_get_pfn(*pl2e) > 0x1000) )
keir@16888 707 continue;
keir@16888 708 *pl2e = l2e_from_intpte(l2e_get_intpte(*pl2e) +
keir@16888 709 xen_phys_start);
keir@16888 710 }
kfraser@15074 711 }
kfraser@15074 712 }
kfraser@15074 713
kfraser@15074 714 /* The only data mappings to be relocated are in the Xen area. */
kfraser@15074 715 pl2e = __va(__pa(l2_xenmap));
keir@16888 716 *pl2e++ = l2e_from_pfn(xen_phys_start >> PAGE_SHIFT,
keir@16888 717 PAGE_HYPERVISOR | _PAGE_PSE);
keir@16888 718 for ( i = 1; i < L2_PAGETABLE_ENTRIES; i++, pl2e++ )
kfraser@15074 719 {
kfraser@15074 720 if ( !(l2e_get_flags(*pl2e) & _PAGE_PRESENT) )
kfraser@15074 721 continue;
kfraser@15074 722 *pl2e = l2e_from_intpte(l2e_get_intpte(*pl2e) +
kfraser@15074 723 xen_phys_start);
kfraser@15074 724 }
kfraser@15074 725
kfraser@15074 726 /* Re-sync the stack and then switch to relocated pagetables. */
kfraser@15074 727 asm volatile (
kfraser@15074 728 "rep movsb ; " /* re-sync the stack */
kfraser@15074 729 "movq %%cr4,%%rsi ; "
kfraser@15074 730 "andb $0x7f,%%sil ; "
kfraser@15074 731 "movq %%rsi,%%cr4 ; " /* CR4.PGE == 0 */
kfraser@15074 732 "movq %0,%%cr3 ; " /* CR3 == new pagetables */
kfraser@15074 733 "orb $0x80,%%sil ; "
kfraser@15074 734 "movq %%rsi,%%cr4 " /* CR4.PGE == 1 */
kfraser@15074 735 : : "r" (__pa(idle_pg_table)), "S" (cpu0_stack),
kfraser@15074 736 "D" (__va(__pa(cpu0_stack))), "c" (STACK_SIZE) : "memory" );
kaf24@5003 737 }
kaf24@5003 738 #endif
keir@15077 739
keir@15077 740 /* Is the region suitable for relocating the multiboot modules? */
keir@15077 741 if ( !initial_images_start && (s < e) && ((e-s) >= modules_length) )
keir@15077 742 {
kfraser@15489 743 initial_images_end = e;
kfraser@15489 744 e = (e - modules_length) & PAGE_MASK;
keir@15077 745 initial_images_start = e;
keir@15077 746 move_memory(initial_images_start,
keir@15077 747 mod[0].mod_start, mod[mbi->mods_count-1].mod_end);
keir@15077 748 }
kfraser@15489 749
kfraser@15489 750 if ( !kexec_crash_area.start && (s < e) &&
kfraser@15489 751 ((e-s) >= kexec_crash_area.size) )
kfraser@15489 752 {
kfraser@15489 753 e = (e - kexec_crash_area.size) & PAGE_MASK;
kfraser@15489 754 kexec_crash_area.start = e;
kfraser@15489 755 }
kaf24@3354 756 }
kaf24@3354 757
kfraser@15074 758 if ( !initial_images_start )
kfraser@15074 759 EARLY_FAIL("Not enough memory to relocate the dom0 kernel image.\n");
keir@15988 760 reserve_e820_ram(&boot_e820, initial_images_start, initial_images_end);
kfraser@15074 761
keir@16563 762 /* Initialise Xen heap and boot heap. */
kfraser@15074 763 xenheap_phys_start = init_boot_allocator(__pa(&_end));
kfraser@15074 764 xenheap_phys_end = opt_xenheap_megabytes << 20;
kfraser@15074 765 #if defined(CONFIG_X86_64)
kfraser@15074 766 if ( !xen_phys_start )
kfraser@15074 767 EARLY_FAIL("Not enough memory to relocate Xen.\n");
kfraser@15074 768 xenheap_phys_end += xen_phys_start;
keir@15988 769 reserve_e820_ram(&boot_e820, xen_phys_start,
keir@15988 770 xen_phys_start + (opt_xenheap_megabytes<<20));
kfraser@15074 771 #endif
kfraser@15074 772
keir@16563 773 /* Late kexec reservation (dynamic start address). */
keir@16563 774 kexec_reserve_area(&boot_e820);
kfraser@15489 775
kfraser@15074 776 /*
keir@16624 777 * With the boot allocator now initialised, we can walk every RAM region
keir@16624 778 * and map it in its entirety (on x86/64, at least) and notify it to the
kfraser@15074 779 * boot allocator.
kfraser@15074 780 */
kfraser@15074 781 for ( i = 0; i < boot_e820.nr_map; i++ )
kfraser@15074 782 {
keir@16563 783 uint64_t s, e, map_s, map_e, mask = PAGE_SIZE - 1;
kfraser@15074 784
kfraser@15074 785 /* Only page alignment required now. */
kfraser@15074 786 s = (boot_e820.map[i].addr + mask) & ~mask;
kfraser@15074 787 e = (boot_e820.map[i].addr + boot_e820.map[i].size) & ~mask;
keir@16563 788 #if defined(CONFIG_X86_32)
keir@16563 789 s = max_t(uint64_t, s, xenheap_phys_end);
keir@16563 790 #else
keir@16563 791 s = max_t(uint64_t, s, 1<<20);
keir@16563 792 #endif
kfraser@15074 793 if ( (boot_e820.map[i].type != E820_RAM) || (s >= e) )
kfraser@15074 794 continue;
kfraser@15074 795
keir@16563 796 /* Need to create mappings above 16MB. */
keir@16563 797 map_s = max_t(uint64_t, s, 16<<20);
kfraser@15074 798 map_e = e;
keir@16563 799 #if defined(CONFIG_X86_32) /* mappings are truncated on x86_32 */
kfraser@15074 800 map_e = min_t(uint64_t, map_e, BOOTSTRAP_DIRECTMAP_END);
kfraser@15074 801 #endif
keir@16563 802
keir@16563 803 /* Pass mapped memory to allocator /before/ creating new mappings. */
keir@16624 804 init_boot_pages(s, min_t(uint64_t, map_s, e));
keir@16563 805
keir@16563 806 /* Create new mappings /before/ passing memory to the allocator. */
keir@16563 807 if ( map_s < map_e )
kfraser@15074 808 map_pages_to_xen(
keir@16563 809 (unsigned long)maddr_to_bootstrap_virt(map_s),
keir@16563 810 map_s >> PAGE_SHIFT, (map_e-map_s) >> PAGE_SHIFT,
keir@16563 811 PAGE_HYPERVISOR);
kfraser@15074 812
keir@16563 813 /* Pass remainder of this memory chunk to the allocator. */
keir@16624 814 init_boot_pages(map_s, e);
kfraser@15074 815 }
kfraser@15074 816
kaf24@5003 817 memguard_init();
kaf24@4950 818
kfraser@15074 819 nr_pages = 0;
kfraser@15074 820 for ( i = 0; i < e820.nr_map; i++ )
kfraser@15074 821 if ( e820.map[i].type == E820_RAM )
kfraser@15074 822 nr_pages += e820.map[i].size >> PAGE_SHIFT;
ian@12681 823 printk("System RAM: %luMB (%lukB)\n",
kaf24@3354 824 nr_pages >> (20 - PAGE_SHIFT),
kaf24@3354 825 nr_pages << (PAGE_SHIFT - 10));
kaf24@7220 826 total_pages = nr_pages;
kaf24@3354 827
kfraser@11296 828 /* Sanity check for unwanted bloat of certain hypercall structures. */
kfraser@11296 829 BUILD_BUG_ON(sizeof(((struct xen_platform_op *)0)->u) !=
kfraser@11296 830 sizeof(((struct xen_platform_op *)0)->u.pad));
kfraser@11296 831 BUILD_BUG_ON(sizeof(((struct xen_domctl *)0)->u) !=
kfraser@11296 832 sizeof(((struct xen_domctl *)0)->u.pad));
kfraser@11296 833 BUILD_BUG_ON(sizeof(((struct xen_sysctl *)0)->u) !=
kfraser@11296 834 sizeof(((struct xen_sysctl *)0)->u.pad));
kaf24@7388 835
kaf24@9878 836 BUILD_BUG_ON(sizeof(start_info_t) > PAGE_SIZE);
kaf24@9878 837 BUILD_BUG_ON(sizeof(shared_info_t) > PAGE_SIZE);
ack@13292 838 BUILD_BUG_ON(sizeof(struct vcpu_info) != 64);
kaf24@7744 839
ack@13291 840 #ifdef CONFIG_COMPAT
ack@13291 841 BUILD_BUG_ON(sizeof(((struct compat_platform_op *)0)->u) !=
ack@13291 842 sizeof(((struct compat_platform_op *)0)->u.pad));
ack@13291 843 BUILD_BUG_ON(sizeof(start_info_compat_t) > PAGE_SIZE);
ack@13292 844 BUILD_BUG_ON(sizeof(struct compat_vcpu_info) != 64);
ack@13291 845 #endif
ack@13291 846
kfraser@10492 847 /* Check definitions in public headers match internal defs. */
kaf24@9878 848 BUILD_BUG_ON(__HYPERVISOR_VIRT_START != HYPERVISOR_VIRT_START);
kaf24@8521 849 #ifdef HYPERVISOR_VIRT_END
kaf24@9878 850 BUILD_BUG_ON(__HYPERVISOR_VIRT_END != HYPERVISOR_VIRT_END);
kaf24@8521 851 #endif
kfraser@10492 852 BUILD_BUG_ON(MACH2PHYS_VIRT_START != RO_MPT_VIRT_START);
kfraser@10492 853 BUILD_BUG_ON(MACH2PHYS_VIRT_END != RO_MPT_VIRT_END);
kaf24@8521 854
kaf24@3354 855 init_frametable();
kaf24@3338 856
kfraser@11971 857 acpi_boot_table_init();
kfraser@11971 858
kfraser@11971 859 acpi_numa_init();
kfraser@11971 860
kfraser@11971 861 numa_initmem_init(0, max_page);
kfraser@11971 862
kaf24@6111 863 /* Initialise the Xen heap, skipping RAM holes. */
kfraser@15074 864 init_xenheap_pages(xenheap_phys_start, xenheap_phys_end);
kfraser@15074 865 nr_pages = (xenheap_phys_end - xenheap_phys_start) >> PAGE_SHIFT;
kfraser@15074 866 #ifdef __x86_64__
kfraser@15074 867 init_xenheap_pages(xen_phys_start, __pa(&_start));
kfraser@15074 868 nr_pages += (__pa(&_start) - xen_phys_start) >> PAGE_SHIFT;
kfraser@15747 869 vesa_init();
kfraser@15074 870 #endif
kfraser@15074 871 xenheap_phys_start = xen_phys_start;
kaf24@6111 872 printk("Xen heap: %luMB (%lukB)\n",
kaf24@6111 873 nr_pages >> (20 - PAGE_SHIFT),
kaf24@6111 874 nr_pages << (PAGE_SHIFT - 10));
kaf24@3338 875
keir@14680 876 end_boot_allocator();
keir@14680 877
kaf24@3594 878 early_boot = 0;
kaf24@3338 879
keir@17444 880 softirq_init();
keir@17444 881
kaf24@8459 882 early_cpu_init();
kaf24@8459 883
kaf24@8459 884 paging_init();
kaf24@8459 885
keir@16274 886 tboot_probe();
keir@16274 887
kaf24@8459 888 /* Unmap the first page of CPU0's stack. */
kaf24@8459 889 memguard_guard_stack(cpu0_stack);
kaf24@8459 890
kaf24@8459 891 open_softirq(NEW_TLBFLUSH_CLOCK_PERIOD_SOFTIRQ, new_tlbflush_clock_period);
kaf24@8459 892
kaf24@8459 893 if ( opt_watchdog )
kaf24@8459 894 nmi_watchdog = NMI_LOCAL_APIC;
kaf24@8459 895
kaf24@8459 896 sort_exception_tables();
kaf24@8459 897
kaf24@8459 898 find_smp_config();
kaf24@8459 899
kaf24@8459 900 dmi_scan_machine();
kaf24@8459 901
kaf24@8459 902 generic_apic_probe();
kaf24@8459 903
keir@17552 904 if ( x2apic_is_available() )
keir@17552 905 enable_x2apic();
keir@17552 906
kaf24@8459 907 acpi_boot_init();
kaf24@8459 908
kfraser@11971 909 init_cpu_to_node();
kfraser@11971 910
kfraser@11241 911 if ( smp_found_config )
kaf24@8459 912 get_smp_config();
kaf24@8459 913
keir@15083 914 #ifdef CONFIG_X86_64
keir@15083 915 /* Low mappings were only needed for some BIOS table parsing. */
keir@15083 916 zap_low_mappings();
keir@15083 917 #endif
keir@15083 918
kaf24@8459 919 init_apic_mappings();
kaf24@8459 920
kaf24@8459 921 init_IRQ();
kaf24@8459 922
kfraser@11241 923 percpu_init_areas();
kfraser@11241 924
kfraser@15815 925 xsm_init(&initrdidx, mbi, initial_images_start);
kfraser@15815 926
kfraser@11240 927 init_idle_domain();
kfraser@11240 928
kaf24@8459 929 trap_init();
kaf24@8459 930
kaf24@13662 931 rcu_init();
kaf24@13662 932
kaf24@8586 933 timer_init();
kaf24@8459 934
kaf24@8459 935 early_time_init();
kaf24@8459 936
kaf24@8459 937 arch_init_memory();
kaf24@8459 938
kaf24@8459 939 identify_cpu(&boot_cpu_data);
kaf24@8459 940 if ( cpu_has_fxsr )
kaf24@8459 941 set_in_cr4(X86_CR4_OSFXSR);
kaf24@8459 942 if ( cpu_has_xmm )
kaf24@8459 943 set_in_cr4(X86_CR4_OSXMMEXCPT);
keir@18920 944
keir@18920 945 local_irq_enable();
keir@18920 946
kfraser@15747 947 #ifdef CONFIG_X86_64
kfraser@15747 948 vesa_mtrr_init();
kfraser@15747 949 #endif
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
keir@18920 956 spin_debug_enable();
keir@18920 957
kaf24@8459 958 /*
kaf24@8459 959 * Initialise higher-level timer functions. We do this fairly late
kaf24@8459 960 * (post-SMP) because the time bases and scale factors need to be updated
kaf24@8459 961 * regularly, and SMP initialisation can cause a long delay with
kaf24@8459 962 * interrupts not yet enabled.
kaf24@8459 963 */
kaf24@8459 964 init_xen_time();
kaf24@8459 965
kaf24@8459 966 initialize_keytable();
kaf24@8459 967
kaf24@8459 968 serial_init_postirq();
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
kaf24@8459 996 /* Create initial domain 0. */
kfraser@14911 997 dom0 = domain_create(0, 0, DOM0_SSIDREF);
kfraser@10655 998 if ( (dom0 == NULL) || (alloc_vcpu(dom0, 0, 0) == NULL) )
kaf24@8459 999 panic("Error creating domain 0\n");
kaf24@8459 1000
kfraser@12210 1001 dom0->is_privileged = 1;
keir@16856 1002 dom0->target = NULL;
kfraser@12210 1003
kaf24@8459 1004 /* Grab the DOM0 command line. */
kaf24@8459 1005 cmdline = (char *)(mod[0].string ? __va(mod[0].string) : NULL);
kfraser@15426 1006 if ( (cmdline != NULL) || (kextra != NULL) )
kaf24@8459 1007 {
kaf24@8459 1008 static char dom0_cmdline[MAX_GUEST_CMDLINE];
kaf24@8459 1009
kfraser@15796 1010 cmdline = cmdline_cook(cmdline);
kfraser@15796 1011 safe_strcpy(dom0_cmdline, cmdline);
kaf24@8459 1012
kfraser@15426 1013 if ( kextra != NULL )
kfraser@15426 1014 /* kextra always includes exactly one leading space. */
kfraser@15426 1015 safe_strcat(dom0_cmdline, kextra);
kfraser@15426 1016
kaf24@8459 1017 /* Append any extra parameters. */
kfraser@13691 1018 if ( skip_ioapic_setup && !strstr(dom0_cmdline, "noapic") )
kfraser@13691 1019 safe_strcat(dom0_cmdline, " noapic");
kaf24@8459 1020 if ( acpi_skip_timer_override &&
kfraser@13691 1021 !strstr(dom0_cmdline, "acpi_skip_timer_override") )
kfraser@13691 1022 safe_strcat(dom0_cmdline, " acpi_skip_timer_override");
keir@16165 1023 if ( (strlen(acpi_param) == 0) && acpi_disabled )
keir@16165 1024 {
keir@16165 1025 printk("ACPI is disabled, notifying Domain 0 (acpi=off)\n");
keir@16165 1026 safe_strcpy(acpi_param, "off");
keir@16165 1027 }
kfraser@13691 1028 if ( (strlen(acpi_param) != 0) && !strstr(dom0_cmdline, "acpi=") )
kaf24@8459 1029 {
kfraser@13691 1030 safe_strcat(dom0_cmdline, " acpi=");
kfraser@13691 1031 safe_strcat(dom0_cmdline, acpi_param);
kaf24@8459 1032 }
kfraser@13691 1033
kfraser@13691 1034 cmdline = dom0_cmdline;
kaf24@8459 1035 }
kaf24@8459 1036
kaf24@8459 1037 if ( (initrdidx > 0) && (initrdidx < mbi->mods_count) )
kaf24@8459 1038 {
kaf24@8459 1039 _initrd_start = initial_images_start +
kaf24@8459 1040 (mod[initrdidx].mod_start - mod[0].mod_start);
kaf24@8459 1041 _initrd_len = mod[initrdidx].mod_end - mod[initrdidx].mod_start;
kaf24@8459 1042 }
kaf24@8459 1043
keir@18180 1044 if ( xen_cpuidle )
keir@18180 1045 xen_processor_pmbits |= XEN_PROCESSOR_PM_CX;
keir@18180 1046
kaf24@8459 1047 /*
kaf24@8459 1048 * We're going to setup domain0 using the module(s) that we stashed safely
kaf24@8459 1049 * above our heap. The second module, if present, is an initrd ramdisk.
kaf24@8459 1050 */
kaf24@8459 1051 if ( construct_dom0(dom0,
kaf24@8459 1052 initial_images_start,
kaf24@8459 1053 mod[0].mod_end-mod[0].mod_start,
kaf24@8459 1054 _initrd_start,
kaf24@8459 1055 _initrd_len,
kaf24@8459 1056 cmdline) != 0)
kaf24@8459 1057 panic("Could not set up DOM0 guest OS\n");
kaf24@8459 1058
kaf24@8459 1059 /* Scrub RAM that is still free and so may go to an unprivileged domain. */
kaf24@8459 1060 scrub_heap_pages();
kaf24@8459 1061
kaf24@8459 1062 init_trace_bufs();
kaf24@8459 1063
kaf24@10502 1064 console_endboot();
kaf24@8459 1065
kaf24@8459 1066 /* Hide UART from DOM0 if we're using it */
kaf24@8459 1067 serial_endboot();
kaf24@8459 1068
kaf24@8459 1069 domain_unpause_by_systemcontroller(dom0);
kaf24@8459 1070
keir@15082 1071 reset_stack_and_jump(init_done);
kaf24@8459 1072 }
kaf24@8459 1073
ian@13763 1074 void arch_get_xen_caps(xen_capabilities_info_t *info)
iap10@6721 1075 {
kfraser@14997 1076 /* Interface name is always xen-3.0-* for Xen-3.x. */
kfraser@14997 1077 int major = 3, minor = 0;
keir@13754 1078 char s[32];
keir@13754 1079
ian@13763 1080 (*info)[0] = '\0';
iap10@6721 1081
keir@17618 1082 #if defined(CONFIG_X86_32)
kaf24@6725 1083
keir@13754 1084 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1085 safe_strcat(*info, s);
kaf24@6725 1086 if ( hvm_enabled )
iap10@6721 1087 {
keir@13754 1088 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1089 safe_strcat(*info, s);
keir@13754 1090 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1091 safe_strcat(*info, s);
iap10@6721 1092 }
iap10@6721 1093
kaf24@6725 1094 #elif defined(CONFIG_X86_64)
iap10@6721 1095
keir@13754 1096 snprintf(s, sizeof(s), "xen-%d.%d-x86_64 ", major, minor);
ian@13763 1097 safe_strcat(*info, s);
ack@13288 1098 #ifdef CONFIG_COMPAT
keir@13754 1099 snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
ian@13763 1100 safe_strcat(*info, s);
ack@13288 1101 #endif
kaf24@6725 1102 if ( hvm_enabled )
iap10@6721 1103 {
keir@13754 1104 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32 ", major, minor);
ian@13763 1105 safe_strcat(*info, s);
keir@13754 1106 snprintf(s, sizeof(s), "hvm-%d.%d-x86_32p ", major, minor);
ian@13763 1107 safe_strcat(*info, s);
keir@13754 1108 snprintf(s, sizeof(s), "hvm-%d.%d-x86_64 ", major, minor);
ian@13763 1109 safe_strcat(*info, s);
iap10@6721 1110 }
kaf24@6725 1111
iap10@6721 1112 #endif
iap10@6721 1113 }
iap10@6721 1114
keir@17738 1115 int xen_in_range(paddr_t start, paddr_t end)
keir@17729 1116 {
keir@17738 1117 start = max_t(paddr_t, start, xenheap_phys_start);
keir@17738 1118 end = min_t(paddr_t, end, xenheap_phys_end);
keir@17729 1119
keir@17729 1120 return start < end;
keir@17729 1121 }
keir@17729 1122
kaf24@3914 1123 /*
kaf24@3914 1124 * Local variables:
kaf24@3914 1125 * mode: C
kaf24@3914 1126 * c-set-style: "BSD"
kaf24@3914 1127 * c-basic-offset: 4
kaf24@3914 1128 * tab-width: 4
kaf24@3914 1129 * indent-tabs-mode: nil
kaf24@3988 1130 * End:
kaf24@3914 1131 */