linux/arch/arm64/kernel/setup.c
<<
>>
Prefs
   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/initrd.h>
  27#include <linux/console.h>
  28#include <linux/cache.h>
  29#include <linux/bootmem.h>
  30#include <linux/screen_info.h>
  31#include <linux/init.h>
  32#include <linux/kexec.h>
  33#include <linux/root_dev.h>
  34#include <linux/cpu.h>
  35#include <linux/interrupt.h>
  36#include <linux/smp.h>
  37#include <linux/fs.h>
  38#include <linux/proc_fs.h>
  39#include <linux/memblock.h>
  40#include <linux/of_fdt.h>
  41#include <linux/efi.h>
  42#include <linux/psci.h>
  43#include <linux/sched/task.h>
  44#include <linux/mm.h>
  45
  46#include <asm/acpi.h>
  47#include <asm/fixmap.h>
  48#include <asm/cpu.h>
  49#include <asm/cputype.h>
  50#include <asm/daifflags.h>
  51#include <asm/elf.h>
  52#include <asm/cpufeature.h>
  53#include <asm/cpu_ops.h>
  54#include <asm/kasan.h>
  55#include <asm/numa.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
  67static int num_standard_resources;
  68static struct resource *standard_resources;
  69
  70phys_addr_t __fdt_pointer __initdata;
  71
  72/*
  73 * Standard memory resources
  74 */
  75static struct resource mem_res[] = {
  76        {
  77                .name = "Kernel code",
  78                .start = 0,
  79                .end = 0,
  80                .flags = IORESOURCE_SYSTEM_RAM
  81        },
  82        {
  83                .name = "Kernel data",
  84                .start = 0,
  85                .end = 0,
  86                .flags = IORESOURCE_SYSTEM_RAM
  87        }
  88};
  89
  90#define kernel_code mem_res[0]
  91#define kernel_data mem_res[1]
  92
  93/*
  94 * The recorded values of x0 .. x3 upon kernel entry.
  95 */
  96u64 __cacheline_aligned boot_args[4];
  97
  98void __init smp_setup_processor_id(void)
  99{
 100        u64 mpidr = read_cpuid_mpidr() & MPIDR_HWID_BITMASK;
 101        cpu_logical_map(0) = mpidr;
 102
 103        /*
 104         * clear __my_cpu_offset on boot CPU to avoid hang caused by
 105         * using percpu variable early, for example, lockdep will
 106         * access percpu variable inside lock_release
 107         */
 108        set_my_cpu_offset(0);
 109        pr_info("Booting Linux on physical CPU 0x%010lx [0x%08x]\n",
 110                (unsigned long)mpidr, read_cpuid_id());
 111}
 112
 113bool arch_match_cpu_phys_id(int cpu, u64 phys_id)
 114{
 115        return phys_id == cpu_logical_map(cpu);
 116}
 117
 118struct mpidr_hash mpidr_hash;
 119/**
 120 * smp_build_mpidr_hash - Pre-compute shifts required at each affinity
 121 *                        level in order to build a linear index from an
 122 *                        MPIDR value. Resulting algorithm is a collision
 123 *                        free hash carried out through shifting and ORing
 124 */
 125static void __init smp_build_mpidr_hash(void)
 126{
 127        u32 i, affinity, fs[4], bits[4], ls;
 128        u64 mask = 0;
 129        /*
 130         * Pre-scan the list of MPIDRS and filter out bits that do
 131         * not contribute to affinity levels, ie they never toggle.
 132         */
 133        for_each_possible_cpu(i)
 134                mask |= (cpu_logical_map(i) ^ cpu_logical_map(0));
 135        pr_debug("mask of set bits %#llx\n", mask);
 136        /*
 137         * Find and stash the last and first bit set at all affinity levels to
 138         * check how many bits are required to represent them.
 139         */
 140        for (i = 0; i < 4; i++) {
 141                affinity = MPIDR_AFFINITY_LEVEL(mask, i);
 142                /*
 143                 * Find the MSB bit and LSB bits position
 144                 * to determine how many bits are required
 145                 * to express the affinity level.
 146                 */
 147                ls = fls(affinity);
 148                fs[i] = affinity ? ffs(affinity) - 1 : 0;
 149                bits[i] = ls - fs[i];
 150        }
 151        /*
 152         * An index can be created from the MPIDR_EL1 by isolating the
 153         * significant bits at each affinity level and by shifting
 154         * them in order to compress the 32 bits values space to a
 155         * compressed set of values. This is equivalent to hashing
 156         * the MPIDR_EL1 through shifting and ORing. It is a collision free
 157         * hash though not minimal since some levels might contain a number
 158         * of CPUs that is not an exact power of 2 and their bit
 159         * representation might contain holes, eg MPIDR_EL1[7:0] = {0x2, 0x80}.
 160         */
 161        mpidr_hash.shift_aff[0] = MPIDR_LEVEL_SHIFT(0) + fs[0];
 162        mpidr_hash.shift_aff[1] = MPIDR_LEVEL_SHIFT(1) + fs[1] - bits[0];
 163        mpidr_hash.shift_aff[2] = MPIDR_LEVEL_SHIFT(2) + fs[2] -
 164                                                (bits[1] + bits[0]);
 165        mpidr_hash.shift_aff[3] = MPIDR_LEVEL_SHIFT(3) +
 166                                  fs[3] - (bits[2] + bits[1] + bits[0]);
 167        mpidr_hash.mask = mask;
 168        mpidr_hash.bits = bits[3] + bits[2] + bits[1] + bits[0];
 169        pr_debug("MPIDR hash: aff0[%u] aff1[%u] aff2[%u] aff3[%u] mask[%#llx] bits[%u]\n",
 170                mpidr_hash.shift_aff[0],
 171                mpidr_hash.shift_aff[1],
 172                mpidr_hash.shift_aff[2],
 173                mpidr_hash.shift_aff[3],
 174                mpidr_hash.mask,
 175                mpidr_hash.bits);
 176        /*
 177         * 4x is an arbitrary value used to warn on a hash table much bigger
 178         * than expected on most systems.
 179         */
 180        if (mpidr_hash_size() > 4 * num_possible_cpus())
 181                pr_warn("Large number of MPIDR hash buckets detected\n");
 182}
 183
 184static void __init setup_machine_fdt(phys_addr_t dt_phys)
 185{
 186        void *dt_virt = fixmap_remap_fdt(dt_phys);
 187        const char *name;
 188
 189        if (!dt_virt || !early_init_dt_scan(dt_virt)) {
 190                pr_crit("\n"
 191                        "Error: invalid device tree blob at physical address %pa (virtual address 0x%p)\n"
 192                        "The dtb must be 8-byte aligned and must not exceed 2 MB in size\n"
 193                        "\nPlease check your bootloader.",
 194                        &dt_phys, dt_virt);
 195
 196                while (true)
 197                        cpu_relax();
 198        }
 199
 200        name = of_flat_dt_get_machine_name();
 201        if (!name)
 202                return;
 203
 204        pr_info("Machine model: %s\n", name);
 205        dump_stack_set_arch_desc("%s (DT)", name);
 206}
 207
 208static void __init request_standard_resources(void)
 209{
 210        struct memblock_region *region;
 211        struct resource *res;
 212        unsigned long i = 0;
 213
 214        kernel_code.start   = __pa_symbol(_text);
 215        kernel_code.end     = __pa_symbol(__init_begin - 1);
 216        kernel_data.start   = __pa_symbol(_sdata);
 217        kernel_data.end     = __pa_symbol(_end - 1);
 218
 219        num_standard_resources = memblock.memory.cnt;
 220        standard_resources = alloc_bootmem_low(num_standard_resources *
 221                                               sizeof(*standard_resources));
 222
 223        for_each_memblock(memory, region) {
 224                res = &standard_resources[i++];
 225                if (memblock_is_nomap(region)) {
 226                        res->name  = "reserved";
 227                        res->flags = IORESOURCE_MEM;
 228                } else {
 229                        res->name  = "System RAM";
 230                        res->flags = IORESOURCE_SYSTEM_RAM | IORESOURCE_BUSY;
 231                }
 232                res->start = __pfn_to_phys(memblock_region_memory_base_pfn(region));
 233                res->end = __pfn_to_phys(memblock_region_memory_end_pfn(region)) - 1;
 234
 235                request_resource(&iomem_resource, res);
 236
 237                if (kernel_code.start >= res->start &&
 238                    kernel_code.end <= res->end)
 239                        request_resource(res, &kernel_code);
 240                if (kernel_data.start >= res->start &&
 241                    kernel_data.end <= res->end)
 242                        request_resource(res, &kernel_data);
 243#ifdef CONFIG_KEXEC_CORE
 244                /* Userspace will find "Crash kernel" region in /proc/iomem. */
 245                if (crashk_res.end && crashk_res.start >= res->start &&
 246                    crashk_res.end <= res->end)
 247                        request_resource(res, &crashk_res);
 248#endif
 249        }
 250}
 251
 252static int __init reserve_memblock_reserved_regions(void)
 253{
 254        u64 i, j;
 255
 256        for (i = 0; i < num_standard_resources; ++i) {
 257                struct resource *mem = &standard_resources[i];
 258                phys_addr_t r_start, r_end, mem_size = resource_size(mem);
 259
 260                if (!memblock_is_region_reserved(mem->start, mem_size))
 261                        continue;
 262
 263                for_each_reserved_mem_region(j, &r_start, &r_end) {
 264                        resource_size_t start, end;
 265
 266                        start = max(PFN_PHYS(PFN_DOWN(r_start)), mem->start);
 267                        end = min(PFN_PHYS(PFN_UP(r_end)) - 1, mem->end);
 268
 269                        if (start > mem->end || end < mem->start)
 270                                continue;
 271
 272                        reserve_region_with_split(mem, start, end, "reserved");
 273                }
 274        }
 275
 276        return 0;
 277}
 278arch_initcall(reserve_memblock_reserved_regions);
 279
 280u64 __cpu_logical_map[NR_CPUS] = { [0 ... NR_CPUS-1] = INVALID_HWID };
 281
 282void __init setup_arch(char **cmdline_p)
 283{
 284        init_mm.start_code = (unsigned long) _text;
 285        init_mm.end_code   = (unsigned long) _etext;
 286        init_mm.end_data   = (unsigned long) _edata;
 287        init_mm.brk        = (unsigned long) _end;
 288
 289        *cmdline_p = boot_command_line;
 290
 291        early_fixmap_init();
 292        early_ioremap_init();
 293
 294        setup_machine_fdt(__fdt_pointer);
 295
 296        parse_early_param();
 297
 298        /*
 299         * Unmask asynchronous aborts and fiq after bringing up possible
 300         * earlycon. (Report possible System Errors once we can report this
 301         * occurred).
 302         */
 303        local_daif_restore(DAIF_PROCCTX_NOIRQ);
 304
 305        /*
 306         * TTBR0 is only used for the identity mapping at this stage. Make it
 307         * point to zero page to avoid speculatively fetching new entries.
 308         */
 309        cpu_uninstall_idmap();
 310
 311        xen_early_init();
 312        efi_init();
 313        arm64_memblock_init();
 314
 315        paging_init();
 316        efi_apply_persistent_mem_reservations();
 317
 318        acpi_table_upgrade();
 319
 320        /* Parse the ACPI tables for possible boot-time configuration */
 321        acpi_boot_table_init();
 322
 323        if (acpi_disabled)
 324                unflatten_device_tree();
 325
 326        bootmem_init();
 327
 328        kasan_init();
 329
 330        request_standard_resources();
 331
 332        early_ioremap_reset();
 333
 334        if (acpi_disabled)
 335                psci_dt_init();
 336        else
 337                psci_acpi_init();
 338
 339        cpu_read_bootcpu_ops();
 340        smp_init_cpus();
 341        smp_build_mpidr_hash();
 342
 343#ifdef CONFIG_ARM64_SW_TTBR0_PAN
 344        /*
 345         * Make sure init_thread_info.ttbr0 always generates translation
 346         * faults in case uaccess_enable() is inadvertently called by the init
 347         * thread.
 348         */
 349        init_task.thread_info.ttbr0 = __pa_symbol(empty_zero_page);
 350#endif
 351
 352#ifdef CONFIG_VT
 353#if defined(CONFIG_VGA_CONSOLE)
 354        conswitchp = &vga_con;
 355#elif defined(CONFIG_DUMMY_CONSOLE)
 356        conswitchp = &dummy_con;
 357#endif
 358#endif
 359        if (boot_args[1] || boot_args[2] || boot_args[3]) {
 360                pr_err("WARNING: x1-x3 nonzero in violation of boot protocol:\n"
 361                        "\tx1: %016llx\n\tx2: %016llx\n\tx3: %016llx\n"
 362                        "This indicates a broken bootloader or old kernel\n",
 363                        boot_args[1], boot_args[2], boot_args[3]);
 364        }
 365}
 366
 367static int __init topology_init(void)
 368{
 369        int i;
 370
 371        for_each_online_node(i)
 372                register_one_node(i);
 373
 374        for_each_possible_cpu(i) {
 375                struct cpu *cpu = &per_cpu(cpu_data.cpu, i);
 376                cpu->hotpluggable = 1;
 377                register_cpu(cpu, i);
 378        }
 379
 380        return 0;
 381}
 382subsys_initcall(topology_init);
 383
 384/*
 385 * Dump out kernel offset information on panic.
 386 */
 387static int dump_kernel_offset(struct notifier_block *self, unsigned long v,
 388                              void *p)
 389{
 390        const unsigned long offset = kaslr_offset();
 391
 392        if (IS_ENABLED(CONFIG_RANDOMIZE_BASE) && offset > 0) {
 393                pr_emerg("Kernel Offset: 0x%lx from 0x%lx\n",
 394                         offset, KIMAGE_VADDR);
 395        } else {
 396                pr_emerg("Kernel Offset: disabled\n");
 397        }
 398        return 0;
 399}
 400
 401static struct notifier_block kernel_offset_notifier = {
 402        .notifier_call = dump_kernel_offset
 403};
 404
 405static int __init register_kernel_offset_dumper(void)
 406{
 407        atomic_notifier_chain_register(&panic_notifier_list,
 408                                       &kernel_offset_notifier);
 409        return 0;
 410}
 411__initcall(register_kernel_offset_dumper);
 412