arch/arm64/kernel/setup.c

   1 /*
   2  * Based on arch/arm/kernel/setup.c
   3  *
   4  * Copyright (C) 1995-2001 Russell King
   5  * Copyright (C) 2012 ARM Ltd.
   6  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License version 2 as
   9  * published by the Free Software Foundation.
  10  *
  11  * This program is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14  * GNU General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
  18  */
  19
  20 #include <linux/acpi.h>
  21 #include <linux/export.h>
  22 #include <linux/kernel.h>
  23 #include <linux/stddef.h>
  24 #include <linux/ioport.h>
  25 #include <linux/delay.h>
  26 #include <linux/utsname.h>
  27 #include <linux/initrd.h>
  28 #include <linux/console.h>
  29 #include <linux/cache.h>
  30 #include <linux/bootmem.h>
  31 #include <linux/screen_info.h>
  32 #include <linux/init.h>
  33 #include <linux/kexec.h>
  34 #include <linux/crash_dump.h>
  35 #include <linux/root_dev.h>
  36 #include <linux/cpu.h>
  37 #include <linux/interrupt.h>
  38 #include <linux/smp.h>
  39 #include <linux/fs.h>
  40 #include <linux/proc_fs.h>
  41 #include <linux/memblock.h>
  42 #include <linux/of_iommu.h>
  43 #include <linux/of_fdt.h>
  44 #include <linux/of_platform.h>
  45 #include <linux/efi.h>
  46 #include <linux/psci.h>
  47
  48 #include <asm/acpi.h>
  49 #include <asm/fixmap.h>
  50 #include <asm/cpu.h>
  51 #include <asm/cputype.h>
  52 #include <asm/elf.h>
  53 #include <asm/cpufeature.h>
  54 #include <asm/cpu_ops.h>
  55 #include <asm/kasan.h>
  56 #include <asm/sections.h>
  57 #include <asm/setup.h>
  58 #include <asm/smp_plat.h>
  59 #include <asm/cacheflush.h>
  60 #include <asm/tlbflush.h>
  61 #include <asm/traps.h>
  62 #include <asm/memblock.h>
  63 #include <asm/efi.h>
  64 #include <asm/xen/hypervisor.h>
  65 #include <asm/mmu_context.h>
  66
  67 phys_addr_t __fdt_pointer __initdata;
  68
  69 /*
  70  * Standard memory resources
  71  */
  72 static struct resource mem_res[] = {
  73         {
  74                 .name = "Kernel code",
  75                 .start = 0,
  76                 .end = 0,
  77                 .flags = IORESOURCE_SYSTEM_RAM
  78         },
  79         {
  80                 .name = "Kernel data",
  81                 .start = 0,
  82                 .end = 0,
  83                 .flags = IORESOURCE_SYSTEM_RAM
  84         }
  85 };
  86
  87 #define kernel_code mem_res[0]
  88 #define kernel_data mem_res[1]
  89
  90 /*
  91  * The recorded values of x0 .. x3 upon kernel entry.
  92  */
  93 u64 __cacheline_aligned boot_args[4];
  94
  95 void __init smp_setup_processor_id(void)
  96 {
  97         u64 mpidr = read_cpuid_mpidr() & MPIDR_HWID_BITMASK;
  98         cpu_logical_map(0) = mpidr;
  99
 100         /*
 101          * clear __my_cpu_offset on boot CPU to avoid hang caused by
 102          * using percpu variable early, for example, lockdep will
 103          * access percpu variable inside lock_release
 104          */
 105         set_my_cpu_offset(0);
 106         pr_info("Booting Linux on physical CPU 0x%lx\n", (unsigned long)mpidr);
 107 }
 108
 109 bool arch_match_cpu_phys_id(int cpu, u64 phys_id)
 110 {
 111         return phys_id == cpu_logical_map(cpu);
 112 }
 113
 114 struct mpidr_hash mpidr_hash;
 115 /**
 116  * smp_build_mpidr_hash - Pre-compute shifts required at each affinity
 117  *                        level in order to build a linear index from an
 118  *                        MPIDR value. Resulting algorithm is a collision
 119  *                        free hash carried out through shifting and ORing
 120  */
 121 static void __init smp_build_mpidr_hash(void)
 122 {
 123         u32 i, affinity, fs[4], bits[4], ls;
 124         u64 mask = 0;
 125         /*
 126          * Pre-scan the list of MPIDRS and filter out bits that do
 127          * not contribute to affinity levels, ie they never toggle.
 128          */
 129         for_each_possible_cpu(i)
 130                 mask |= (cpu_logical_map(i) ^ cpu_logical_map(0));
 131         pr_debug("mask of set bits %#llx\n", mask);
 132         /*
 133          * Find and stash the last and first bit set at all affinity levels to
 134          * check how many bits are required to represent them.
 135          */
 136         for (i = 0; i < 4; i++) {
 137                 affinity = MPIDR_AFFINITY_LEVEL(mask, i);
 138                 /*
 139                  * Find the MSB bit and LSB bits position
 140                  * to determine how many bits are required
 141                  * to express the affinity level.
 142                  */
 143                 ls = fls(affinity);
 144                 fs[i] = affinity ? ffs(affinity) - 1 : 0;
 145                 bits[i] = ls - fs[i];
 146         }
 147         /*
 148          * An index can be created from the MPIDR_EL1 by isolating the
 149          * significant bits at each affinity level and by shifting
 150          * them in order to compress the 32 bits values space to a
 151          * compressed set of values. This is equivalent to hashing
 152          * the MPIDR_EL1 through shifting and ORing. It is a collision free
 153          * hash though not minimal since some levels might contain a number
 154          * of CPUs that is not an exact power of 2 and their bit
 155          * representation might contain holes, eg MPIDR_EL1[7:0] = {0x2, 0x80}.
 156          */
 157         mpidr_hash.shift_aff[0] = MPIDR_LEVEL_SHIFT(0) + fs[0];
 158         mpidr_hash.shift_aff[1] = MPIDR_LEVEL_SHIFT(1) + fs[1] - bits[0];
 159         mpidr_hash.shift_aff[2] = MPIDR_LEVEL_SHIFT(2) + fs[2] -
 160                                                 (bits[1] + bits[0]);
 161         mpidr_hash.shift_aff[3] = MPIDR_LEVEL_SHIFT(3) +
 162                                   fs[3] - (bits[2] + bits[1] + bits[0]);
 163         mpidr_hash.mask = mask;
 164         mpidr_hash.bits = bits[3] + bits[2] + bits[1] + bits[0];
 165         pr_debug("MPIDR hash: aff0[%u] aff1[%u] aff2[%u] aff3[%u] mask[%#llx] bits[%u]\n",
 166                 mpidr_hash.shift_aff[0],
 167                 mpidr_hash.shift_aff[1],
 168                 mpidr_hash.shift_aff[2],
 169                 mpidr_hash.shift_aff[3],
 170                 mpidr_hash.mask,
 171                 mpidr_hash.bits);
 172         /*
 173          * 4x is an arbitrary value used to warn on a hash table much bigger
 174          * than expected on most systems.
 175          */
 176         if (mpidr_hash_size() > 4 * num_possible_cpus())
 177                 pr_warn("Large number of MPIDR hash buckets detected\n");
 178         __flush_dcache_area(&mpidr_hash, sizeof(struct mpidr_hash));
 179 }
 180
 181 static void __init setup_machine_fdt(phys_addr_t dt_phys)
 182 {
 183         void *dt_virt = fixmap_remap_fdt(dt_phys);
 184
 185         if (!dt_virt || !early_init_dt_scan(dt_virt)) {
 186                 pr_crit("\n"
 187                         "Error: invalid device tree blob at physical address %pa (virtual address 0x%p)\n"
 188                         "The dtb must be 8-byte aligned and must not exceed 2 MB in size\n"
 189                         "\nPlease check your bootloader.",
 190                         &dt_phys, dt_virt);
 191
 192                 while (true)
 193                         cpu_relax();
 194         }
 195
 196         dump_stack_set_arch_desc("%s (DT)", of_flat_dt_get_machine_name());
 197 }
 198
 199 static void __init request_standard_resources(void)
 200 {
 201         struct memblock_region *region;
 202         struct resource *res;
 203
 204         kernel_code.start   = virt_to_phys(_text);
 205         kernel_code.end     = virt_to_phys(_etext - 1);
 206         kernel_data.start   = virt_to_phys(_sdata);
 207         kernel_data.end     = virt_to_phys(_end - 1);
 208
 209         for_each_memblock(memory, region) {
 210                 res = alloc_bootmem_low(sizeof(*res));
 211                 res->name  = "System RAM";
 212                 res->start = __pfn_to_phys(memblock_region_memory_base_pfn(region));
 213                 res->end = __pfn_to_phys(memblock_region_memory_end_pfn(region)) - 1;
 214                 res->flags = IORESOURCE_SYSTEM_RAM | IORESOURCE_BUSY;
 215
 216                 request_resource(&iomem_resource, res);
 217
 218                 if (kernel_code.start >= res->start &&
 219                     kernel_code.end <= res->end)
 220                         request_resource(res, &kernel_code);
 221                 if (kernel_data.start >= res->start &&
 222                     kernel_data.end <= res->end)
 223                         request_resource(res, &kernel_data);
 224         }
 225 }
 226
 227 #ifdef CONFIG_BLK_DEV_INITRD
 228 /*
 229  * Relocate initrd if it is not completely within the linear mapping.
 230  * This would be the case if mem= cuts out all or part of it.
 231  */
 232 static void __init relocate_initrd(void)
 233 {
 234         phys_addr_t orig_start = __virt_to_phys(initrd_start);
 235         phys_addr_t orig_end = __virt_to_phys(initrd_end);
 236         phys_addr_t ram_end = memblock_end_of_DRAM();
 237         phys_addr_t new_start;
 238         unsigned long size, to_free = 0;
 239         void *dest;
 240
 241         if (orig_end <= ram_end)
 242                 return;
 243
 244         /*
 245          * Any of the original initrd which overlaps the linear map should
 246          * be freed after relocating.
 247          */
 248         if (orig_start < ram_end)
 249                 to_free = ram_end - orig_start;
 250
 251         size = orig_end - orig_start;
 252         if (!size)
 253                 return;
 254
 255         /* initrd needs to be relocated completely inside linear mapping */
 256         new_start = memblock_find_in_range(0, PFN_PHYS(max_pfn),
 257                                            size, PAGE_SIZE);
 258         if (!new_start)
 259                 panic("Cannot relocate initrd of size %ld\n", size);
 260         memblock_reserve(new_start, size);
 261
 262         initrd_start = __phys_to_virt(new_start);
 263         initrd_end   = initrd_start + size;
 264
 265         pr_info("Moving initrd from [%llx-%llx] to [%llx-%llx]\n",
 266                 orig_start, orig_start + size - 1,
 267                 new_start, new_start + size - 1);
 268
 269         dest = (void *)initrd_start;
 270
 271         if (to_free) {
 272                 memcpy(dest, (void *)__phys_to_virt(orig_start), to_free);
 273                 dest += to_free;
 274         }
 275
 276         copy_from_early_mem(dest, orig_start + to_free, size - to_free);
 277
 278         if (to_free) {
 279                 pr_info("Freeing original RAMDISK from [%llx-%llx]\n",
 280                         orig_start, orig_start + to_free - 1);
 281                 memblock_free(orig_start, to_free);
 282         }
 283 }
 284 #else
 285 static inline void __init relocate_initrd(void)
 286 {
 287 }
 288 #endif
 289
 290 u64 __cpu_logical_map[NR_CPUS] = { [0 ... NR_CPUS-1] = INVALID_HWID };
 291
 292 void __init setup_arch(char **cmdline_p)
 293 {
 294         pr_info("Boot CPU: AArch64 Processor [%08x]\n", read_cpuid_id());
 295
 296         sprintf(init_utsname()->machine, ELF_PLATFORM);
 297         init_mm.start_code = (unsigned long) _text;
 298         init_mm.end_code   = (unsigned long) _etext;
 299         init_mm.end_data   = (unsigned long) _edata;
 300         init_mm.brk        = (unsigned long) _end;
 301
 302         *cmdline_p = boot_command_line;
 303
 304         early_fixmap_init();
 305         early_ioremap_init();
 306
 307         setup_machine_fdt(__fdt_pointer);
 308
 309         parse_early_param();
 310
 311         /*
 312          *  Unmask asynchronous aborts after bringing up possible earlycon.
 313          * (Report possible System Errors once we can report this occurred)
 314          */
 315         local_async_enable();
 316
 317         /*
 318          * TTBR0 is only used for the identity mapping at this stage. Make it
 319          * point to zero page to avoid speculatively fetching new entries.
 320          */
 321         cpu_uninstall_idmap();
 322
 323         efi_init();
 324         arm64_memblock_init();
 325
 326         /* Parse the ACPI tables for possible boot-time configuration */
 327         acpi_boot_table_init();
 328
 329         paging_init();
 330         relocate_initrd();
 331
 332         kasan_init();
 333
 334         request_standard_resources();
 335
 336         early_ioremap_reset();
 337
 338         if (acpi_disabled) {
 339                 unflatten_device_tree();
 340                 psci_dt_init();
 341         } else {
 342                 psci_acpi_init();
 343         }
 344         xen_early_init();
 345
 346         cpu_read_bootcpu_ops();
 347         smp_init_cpus();
 348         smp_build_mpidr_hash();
 349
 350 #ifdef CONFIG_VT
 351 #if defined(CONFIG_VGA_CONSOLE)
 352         conswitchp = &vga_con;
 353 #elif defined(CONFIG_DUMMY_CONSOLE)
 354         conswitchp = &dummy_con;
 355 #endif
 356 #endif
 357         if (boot_args[1] || boot_args[2] || boot_args[3]) {
 358                 pr_err("WARNING: x1-x3 nonzero in violation of boot protocol:\n"
 359                         "\tx1: %016llx\n\tx2: %016llx\n\tx3: %016llx\n"
 360                         "This indicates a broken bootloader or old kernel\n",
 361                         boot_args[1], boot_args[2], boot_args[3]);
 362         }
 363 }
 364
 365 static int __init arm64_device_init(void)
 366 {
 367         if (of_have_populated_dt()) {
 368                 of_iommu_init();
 369                 of_platform_populate(NULL, of_default_bus_match_table,
 370                                      NULL, NULL);
 371         } else if (acpi_disabled) {
 372                 pr_crit("Device tree not populated\n");
 373         }
 374         return 0;
 375 }
 376 arch_initcall_sync(arm64_device_init);
 377
 378 static int __init topology_init(void)
 379 {
 380         int i;
 381
 382         for_each_possible_cpu(i) {
 383                 struct cpu *cpu = &per_cpu(cpu_data.cpu, i);
 384                 cpu->hotpluggable = 1;
 385                 register_cpu(cpu, i);
 386         }
 387
 388         return 0;
 389 }
 390 subsys_initcall(topology_init);
 391
 392 /*
 393  * Dump out kernel offset information on panic.
 394  */
 395 static int dump_kernel_offset(struct notifier_block *self, unsigned long v,
 396                               void *p)
 397 {
 398         u64 const kaslr_offset = kimage_vaddr - KIMAGE_VADDR;
 399
 400         if (IS_ENABLED(CONFIG_RANDOMIZE_BASE) && kaslr_offset > 0) {
 401                 pr_emerg("Kernel Offset: 0x%llx from 0x%lx\n",
 402                          kaslr_offset, KIMAGE_VADDR);
 403         } else {
 404                 pr_emerg("Kernel Offset: disabled\n");
 405         }
 406         return 0;
 407 }
 408
 409 static struct notifier_block kernel_offset_notifier = {
 410         .notifier_call = dump_kernel_offset
 411 };
 412
 413 static int __init register_kernel_offset_dumper(void)
 414 {
 415         atomic_notifier_chain_register(&panic_notifier_list,
 416                                        &kernel_offset_notifier);
 417         return 0;
 418 }
 419 __initcall(register_kernel_offset_dumper);