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/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
  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%010lx [0x%08x]\n",
 107                 (unsigned long)mpidr, read_cpuid_id());
 108 }
 109
 110 bool arch_match_cpu_phys_id(int cpu, u64 phys_id)
 111 {
 112         return phys_id == cpu_logical_map(cpu);
 113 }
 114
 115 struct mpidr_hash mpidr_hash;
 116 /**
 117  * smp_build_mpidr_hash - Pre-compute shifts required at each affinity
 118  *                        level in order to build a linear index from an
 119  *                        MPIDR value. Resulting algorithm is a collision
 120  *                        free hash carried out through shifting and ORing
 121  */
 122 static void __init smp_build_mpidr_hash(void)
 123 {
 124         u32 i, affinity, fs[4], bits[4], ls;
 125         u64 mask = 0;
 126         /*
 127          * Pre-scan the list of MPIDRS and filter out bits that do
 128          * not contribute to affinity levels, ie they never toggle.
 129          */
 130         for_each_possible_cpu(i)
 131                 mask |= (cpu_logical_map(i) ^ cpu_logical_map(0));
 132         pr_debug("mask of set bits %#llx\n", mask);
 133         /*
 134          * Find and stash the last and first bit set at all affinity levels to
 135          * check how many bits are required to represent them.
 136          */
 137         for (i = 0; i < 4; i++) {
 138                 affinity = MPIDR_AFFINITY_LEVEL(mask, i);
 139                 /*
 140                  * Find the MSB bit and LSB bits position
 141                  * to determine how many bits are required
 142                  * to express the affinity level.
 143                  */
 144                 ls = fls(affinity);
 145                 fs[i] = affinity ? ffs(affinity) - 1 : 0;
 146                 bits[i] = ls - fs[i];
 147         }
 148         /*
 149          * An index can be created from the MPIDR_EL1 by isolating the
 150          * significant bits at each affinity level and by shifting
 151          * them in order to compress the 32 bits values space to a
 152          * compressed set of values. This is equivalent to hashing
 153          * the MPIDR_EL1 through shifting and ORing. It is a collision free
 154          * hash though not minimal since some levels might contain a number
 155          * of CPUs that is not an exact power of 2 and their bit
 156          * representation might contain holes, eg MPIDR_EL1[7:0] = {0x2, 0x80}.
 157          */
 158         mpidr_hash.shift_aff[0] = MPIDR_LEVEL_SHIFT(0) + fs[0];
 159         mpidr_hash.shift_aff[1] = MPIDR_LEVEL_SHIFT(1) + fs[1] - bits[0];
 160         mpidr_hash.shift_aff[2] = MPIDR_LEVEL_SHIFT(2) + fs[2] -
 161                                                 (bits[1] + bits[0]);
 162         mpidr_hash.shift_aff[3] = MPIDR_LEVEL_SHIFT(3) +
 163                                   fs[3] - (bits[2] + bits[1] + bits[0]);
 164         mpidr_hash.mask = mask;
 165         mpidr_hash.bits = bits[3] + bits[2] + bits[1] + bits[0];
 166         pr_debug("MPIDR hash: aff0[%u] aff1[%u] aff2[%u] aff3[%u] mask[%#llx] bits[%u]\n",
 167                 mpidr_hash.shift_aff[0],
 168                 mpidr_hash.shift_aff[1],
 169                 mpidr_hash.shift_aff[2],
 170                 mpidr_hash.shift_aff[3],
 171                 mpidr_hash.mask,
 172                 mpidr_hash.bits);
 173         /*
 174          * 4x is an arbitrary value used to warn on a hash table much bigger
 175          * than expected on most systems.
 176          */
 177         if (mpidr_hash_size() > 4 * num_possible_cpus())
 178                 pr_warn("Large number of MPIDR hash buckets detected\n");
 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         const char *name;
 185
 186         if (!dt_virt || !early_init_dt_scan(dt_virt)) {
 187                 pr_crit("\n"
 188                         "Error: invalid device tree blob at physical address %pa (virtual address 0x%p)\n"
 189                         "The dtb must be 8-byte aligned and must not exceed 2 MB in size\n"
 190                         "\nPlease check your bootloader.",
 191                         &dt_phys, dt_virt);
 192
 193                 while (true)
 194                         cpu_relax();
 195         }
 196
 197         name = of_flat_dt_get_machine_name();
 198         if (!name)
 199                 return;
 200
 201         pr_info("Machine model: %s\n", name);
 202         dump_stack_set_arch_desc("%s (DT)", name);
 203 }
 204
 205 static void __init request_standard_resources(void)
 206 {
 207         struct memblock_region *region;
 208         struct resource *res;
 209
 210         kernel_code.start   = __pa_symbol(_text);
 211         kernel_code.end     = __pa_symbol(__init_begin - 1);
 212         kernel_data.start   = __pa_symbol(_sdata);
 213         kernel_data.end     = __pa_symbol(_end - 1);
 214
 215         for_each_memblock(memory, region) {
 216                 res = alloc_bootmem_low(sizeof(*res));
 217                 if (memblock_is_nomap(region)) {
 218                         res->name  = "reserved";
 219                         res->flags = IORESOURCE_MEM;
 220                 } else {
 221                         res->name  = "System RAM";
 222                         res->flags = IORESOURCE_SYSTEM_RAM | IORESOURCE_BUSY;
 223                 }
 224                 res->start = __pfn_to_phys(memblock_region_memory_base_pfn(region));
 225                 res->end = __pfn_to_phys(memblock_region_memory_end_pfn(region)) - 1;
 226
 227                 request_resource(&iomem_resource, res);
 228
 229                 if (kernel_code.start >= res->start &&
 230                     kernel_code.end <= res->end)
 231                         request_resource(res, &kernel_code);
 232                 if (kernel_data.start >= res->start &&
 233                     kernel_data.end <= res->end)
 234                         request_resource(res, &kernel_data);
 235 #ifdef CONFIG_KEXEC_CORE
 236                 /* Userspace will find "Crash kernel" region in /proc/iomem. */
 237                 if (crashk_res.end && crashk_res.start >= res->start &&
 238                     crashk_res.end <= res->end)
 239                         request_resource(res, &crashk_res);
 240 #endif
 241         }
 242 }
 243
 244 u64 __cpu_logical_map[NR_CPUS] = { [0 ... NR_CPUS-1] = INVALID_HWID };
 245
 246 void __init setup_arch(char **cmdline_p)
 247 {
 248         init_mm.start_code = (unsigned long) _text;
 249         init_mm.end_code   = (unsigned long) _etext;
 250         init_mm.end_data   = (unsigned long) _edata;
 251         init_mm.brk        = (unsigned long) _end;
 252
 253         *cmdline_p = boot_command_line;
 254
 255         early_fixmap_init();
 256         early_ioremap_init();
 257
 258         setup_machine_fdt(__fdt_pointer);
 259
 260         parse_early_param();
 261
 262         /*
 263          * Unmask asynchronous aborts and fiq after bringing up possible
 264          * earlycon. (Report possible System Errors once we can report this
 265          * occurred).
 266          */
 267         local_daif_restore(DAIF_PROCCTX_NOIRQ);
 268
 269         /*
 270          * TTBR0 is only used for the identity mapping at this stage. Make it
 271          * point to zero page to avoid speculatively fetching new entries.
 272          */
 273         cpu_uninstall_idmap();
 274
 275         xen_early_init();
 276         efi_init();
 277         arm64_memblock_init();
 278
 279         paging_init();
 280
 281         acpi_table_upgrade();
 282
 283         /* Parse the ACPI tables for possible boot-time configuration */
 284         acpi_boot_table_init();
 285
 286         if (acpi_disabled)
 287                 unflatten_device_tree();
 288
 289         bootmem_init();
 290
 291         kasan_init();
 292
 293         request_standard_resources();
 294
 295         early_ioremap_reset();
 296
 297         if (acpi_disabled)
 298                 psci_dt_init();
 299         else
 300                 psci_acpi_init();
 301
 302         cpu_read_bootcpu_ops();
 303         smp_init_cpus();
 304         smp_build_mpidr_hash();
 305
 306 #ifdef CONFIG_ARM64_SW_TTBR0_PAN
 307         /*
 308          * Make sure init_thread_info.ttbr0 always generates translation
 309          * faults in case uaccess_enable() is inadvertently called by the init
 310          * thread.
 311          */
 312         init_task.thread_info.ttbr0 = __pa_symbol(empty_zero_page);
 313 #endif
 314
 315 #ifdef CONFIG_VT
 316 #if defined(CONFIG_VGA_CONSOLE)
 317         conswitchp = &vga_con;
 318 #elif defined(CONFIG_DUMMY_CONSOLE)
 319         conswitchp = &dummy_con;
 320 #endif
 321 #endif
 322         if (boot_args[1] || boot_args[2] || boot_args[3]) {
 323                 pr_err("WARNING: x1-x3 nonzero in violation of boot protocol:\n"
 324                         "\tx1: %016llx\n\tx2: %016llx\n\tx3: %016llx\n"
 325                         "This indicates a broken bootloader or old kernel\n",
 326                         boot_args[1], boot_args[2], boot_args[3]);
 327         }
 328 }
 329
 330 static int __init topology_init(void)
 331 {
 332         int i;
 333
 334         for_each_online_node(i)
 335                 register_one_node(i);
 336
 337         for_each_possible_cpu(i) {
 338                 struct cpu *cpu = &per_cpu(cpu_data.cpu, i);
 339                 cpu->hotpluggable = 1;
 340                 register_cpu(cpu, i);
 341         }
 342
 343         return 0;
 344 }
 345 subsys_initcall(topology_init);
 346
 347 /*
 348  * Dump out kernel offset information on panic.
 349  */
 350 static int dump_kernel_offset(struct notifier_block *self, unsigned long v,
 351                               void *p)
 352 {
 353         const unsigned long offset = kaslr_offset();
 354
 355         if (IS_ENABLED(CONFIG_RANDOMIZE_BASE) && offset > 0) {
 356                 pr_emerg("Kernel Offset: 0x%lx from 0x%lx\n",
 357                          offset, KIMAGE_VADDR);
 358         } else {
 359                 pr_emerg("Kernel Offset: disabled\n");
 360         }
 361         return 0;
 362 }
 363
 364 static struct notifier_block kernel_offset_notifier = {
 365         .notifier_call = dump_kernel_offset
 366 };
 367
 368 static int __init register_kernel_offset_dumper(void)
 369 {
 370         atomic_notifier_chain_register(&panic_notifier_list,
 371                                        &kernel_offset_notifier);
 372         return 0;
 373 }
 374 __initcall(register_kernel_offset_dumper);