arch/s390/kernel/setup.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  *  S390 version
   4  *    Copyright IBM Corp. 1999, 2012
   5  *    Author(s): Hartmut Penner (hp@de.ibm.com),
   6  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
   7  *
   8  *  Derived from "arch/i386/kernel/setup.c"
   9  *    Copyright (C) 1995, Linus Torvalds
  10  */
  11
  12 /*
  13  * This file handles the architecture-dependent parts of initialization
  14  */
  15
  16 #define KMSG_COMPONENT "setup"
  17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  18
  19 #include <linux/errno.h>
  20 #include <linux/export.h>
  21 #include <linux/sched.h>
  22 #include <linux/sched/task.h>
  23 #include <linux/cpu.h>
  24 #include <linux/kernel.h>
  25 #include <linux/memblock.h>
  26 #include <linux/mm.h>
  27 #include <linux/stddef.h>
  28 #include <linux/unistd.h>
  29 #include <linux/ptrace.h>
  30 #include <linux/random.h>
  31 #include <linux/user.h>
  32 #include <linux/tty.h>
  33 #include <linux/ioport.h>
  34 #include <linux/delay.h>
  35 #include <linux/init.h>
  36 #include <linux/initrd.h>
  37 #include <linux/root_dev.h>
  38 #include <linux/console.h>
  39 #include <linux/kernel_stat.h>
  40 #include <linux/dma-contiguous.h>
  41 #include <linux/device.h>
  42 #include <linux/notifier.h>
  43 #include <linux/pfn.h>
  44 #include <linux/ctype.h>
  45 #include <linux/reboot.h>
  46 #include <linux/topology.h>
  47 #include <linux/kexec.h>
  48 #include <linux/crash_dump.h>
  49 #include <linux/memory.h>
  50 #include <linux/compat.h>
  51 #include <linux/start_kernel.h>
  52
  53 #include <asm/boot_data.h>
  54 #include <asm/ipl.h>
  55 #include <asm/facility.h>
  56 #include <asm/smp.h>
  57 #include <asm/mmu_context.h>
  58 #include <asm/cpcmd.h>
  59 #include <asm/lowcore.h>
  60 #include <asm/nmi.h>
  61 #include <asm/irq.h>
  62 #include <asm/page.h>
  63 #include <asm/ptrace.h>
  64 #include <asm/sections.h>
  65 #include <asm/ebcdic.h>
  66 #include <asm/diag.h>
  67 #include <asm/os_info.h>
  68 #include <asm/sclp.h>
  69 #include <asm/stacktrace.h>
  70 #include <asm/sysinfo.h>
  71 #include <asm/numa.h>
  72 #include <asm/alternative.h>
  73 #include <asm/nospec-branch.h>
  74 #include <asm/mem_detect.h>
  75 #include <asm/uv.h>
  76 #include "entry.h"
  77
  78 /*
  79  * Machine setup..
  80  */
  81 unsigned int console_mode = 0;
  82 EXPORT_SYMBOL(console_mode);
  83
  84 unsigned int console_devno = -1;
  85 EXPORT_SYMBOL(console_devno);
  86
  87 unsigned int console_irq = -1;
  88 EXPORT_SYMBOL(console_irq);
  89
  90 unsigned long elf_hwcap __read_mostly = 0;
  91 char elf_platform[ELF_PLATFORM_SIZE];
  92
  93 unsigned long int_hwcap = 0;
  94
  95 #ifdef CONFIG_PROTECTED_VIRTUALIZATION_GUEST
  96 int __bootdata_preserved(prot_virt_guest);
  97 #endif
  98
  99 int __bootdata(noexec_disabled);
 100 int __bootdata(memory_end_set);
 101 unsigned long __bootdata(memory_end);
 102 unsigned long __bootdata(vmalloc_size);
 103 unsigned long __bootdata(max_physmem_end);
 104 struct mem_detect_info __bootdata(mem_detect);
 105
 106 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
 107 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
 108 unsigned long __bootdata_preserved(__swsusp_reset_dma);
 109 unsigned long __bootdata_preserved(__stext_dma);
 110 unsigned long __bootdata_preserved(__etext_dma);
 111 unsigned long __bootdata_preserved(__sdma);
 112 unsigned long __bootdata_preserved(__edma);
 113 unsigned long __bootdata_preserved(__kaslr_offset);
 114
 115 unsigned long VMALLOC_START;
 116 EXPORT_SYMBOL(VMALLOC_START);
 117
 118 unsigned long VMALLOC_END;
 119 EXPORT_SYMBOL(VMALLOC_END);
 120
 121 struct page *vmemmap;
 122 EXPORT_SYMBOL(vmemmap);
 123
 124 unsigned long MODULES_VADDR;
 125 unsigned long MODULES_END;
 126
 127 /* An array with a pointer to the lowcore of every CPU. */
 128 struct lowcore *lowcore_ptr[NR_CPUS];
 129 EXPORT_SYMBOL(lowcore_ptr);
 130
 131 /*
 132  * This is set up by the setup-routine at boot-time
 133  * for S390 need to find out, what we have to setup
 134  * using address 0x10400 ...
 135  */
 136
 137 #include <asm/setup.h>
 138
 139 /*
 140  * condev= and conmode= setup parameter.
 141  */
 142
 143 static int __init condev_setup(char *str)
 144 {
 145         int vdev;
 146
 147         vdev = simple_strtoul(str, &str, 0);
 148         if (vdev >= 0 && vdev < 65536) {
 149                 console_devno = vdev;
 150                 console_irq = -1;
 151         }
 152         return 1;
 153 }
 154
 155 __setup("condev=", condev_setup);
 156
 157 static void __init set_preferred_console(void)
 158 {
 159         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 160                 add_preferred_console("ttyS", 0, NULL);
 161         else if (CONSOLE_IS_3270)
 162                 add_preferred_console("tty3270", 0, NULL);
 163         else if (CONSOLE_IS_VT220)
 164                 add_preferred_console("ttyS", 1, NULL);
 165         else if (CONSOLE_IS_HVC)
 166                 add_preferred_console("hvc", 0, NULL);
 167 }
 168
 169 static int __init conmode_setup(char *str)
 170 {
 171 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 172         if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
 173                 SET_CONSOLE_SCLP;
 174 #endif
 175 #if defined(CONFIG_TN3215_CONSOLE)
 176         if (!strcmp(str, "3215"))
 177                 SET_CONSOLE_3215;
 178 #endif
 179 #if defined(CONFIG_TN3270_CONSOLE)
 180         if (!strcmp(str, "3270"))
 181                 SET_CONSOLE_3270;
 182 #endif
 183         set_preferred_console();
 184         return 1;
 185 }
 186
 187 __setup("conmode=", conmode_setup);
 188
 189 static void __init conmode_default(void)
 190 {
 191         char query_buffer[1024];
 192         char *ptr;
 193
 194         if (MACHINE_IS_VM) {
 195                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 196                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 197                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 198                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 199                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 200                 ptr = strstr(query_buffer, "CONMODE");
 201                 /*
 202                  * Set the conmode to 3215 so that the device recognition
 203                  * will set the cu_type of the console to 3215. If the
 204                  * conmode is 3270 and we don't set it back then both
 205                  * 3215 and the 3270 driver will try to access the console
 206                  * device (3215 as console and 3270 as normal tty).
 207                  */
 208                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 209                 if (ptr == NULL) {
 210 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 211                         SET_CONSOLE_SCLP;
 212 #endif
 213                         return;
 214                 }
 215                 if (str_has_prefix(ptr + 8, "3270")) {
 216 #if defined(CONFIG_TN3270_CONSOLE)
 217                         SET_CONSOLE_3270;
 218 #elif defined(CONFIG_TN3215_CONSOLE)
 219                         SET_CONSOLE_3215;
 220 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 221                         SET_CONSOLE_SCLP;
 222 #endif
 223                 } else if (str_has_prefix(ptr + 8, "3215")) {
 224 #if defined(CONFIG_TN3215_CONSOLE)
 225                         SET_CONSOLE_3215;
 226 #elif defined(CONFIG_TN3270_CONSOLE)
 227                         SET_CONSOLE_3270;
 228 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 229                         SET_CONSOLE_SCLP;
 230 #endif
 231                 }
 232         } else if (MACHINE_IS_KVM) {
 233                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
 234                         SET_CONSOLE_VT220;
 235                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
 236                         SET_CONSOLE_SCLP;
 237                 else
 238                         SET_CONSOLE_HVC;
 239         } else {
 240 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 241                 SET_CONSOLE_SCLP;
 242 #endif
 243         }
 244 }
 245
 246 #ifdef CONFIG_CRASH_DUMP
 247 static void __init setup_zfcpdump(void)
 248 {
 249         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 250                 return;
 251         if (OLDMEM_BASE)
 252                 return;
 253         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 254         console_loglevel = 2;
 255 }
 256 #else
 257 static inline void setup_zfcpdump(void) {}
 258 #endif /* CONFIG_CRASH_DUMP */
 259
 260  /*
 261  * Reboot, halt and power_off stubs. They just call _machine_restart,
 262  * _machine_halt or _machine_power_off.
 263  */
 264
 265 void machine_restart(char *command)
 266 {
 267         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 268                 /*
 269                  * Only unblank the console if we are called in enabled
 270                  * context or a bust_spinlocks cleared the way for us.
 271                  */
 272                 console_unblank();
 273         _machine_restart(command);
 274 }
 275
 276 void machine_halt(void)
 277 {
 278         if (!in_interrupt() || oops_in_progress)
 279                 /*
 280                  * Only unblank the console if we are called in enabled
 281                  * context or a bust_spinlocks cleared the way for us.
 282                  */
 283                 console_unblank();
 284         _machine_halt();
 285 }
 286
 287 void machine_power_off(void)
 288 {
 289         if (!in_interrupt() || oops_in_progress)
 290                 /*
 291                  * Only unblank the console if we are called in enabled
 292                  * context or a bust_spinlocks cleared the way for us.
 293                  */
 294                 console_unblank();
 295         _machine_power_off();
 296 }
 297
 298 /*
 299  * Dummy power off function.
 300  */
 301 void (*pm_power_off)(void) = machine_power_off;
 302 EXPORT_SYMBOL_GPL(pm_power_off);
 303
 304 void *restart_stack __section(.data);
 305
 306 unsigned long stack_alloc(void)
 307 {
 308 #ifdef CONFIG_VMAP_STACK
 309         return (unsigned long)
 310                 __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
 311                                      VMALLOC_START, VMALLOC_END,
 312                                      THREADINFO_GFP,
 313                                      PAGE_KERNEL, 0, NUMA_NO_NODE,
 314                                      __builtin_return_address(0));
 315 #else
 316         return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 317 #endif
 318 }
 319
 320 void stack_free(unsigned long stack)
 321 {
 322 #ifdef CONFIG_VMAP_STACK
 323         vfree((void *) stack);
 324 #else
 325         free_pages(stack, THREAD_SIZE_ORDER);
 326 #endif
 327 }
 328
 329 int __init arch_early_irq_init(void)
 330 {
 331         unsigned long stack;
 332
 333         stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 334         if (!stack)
 335                 panic("Couldn't allocate async stack");
 336         S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
 337         return 0;
 338 }
 339
 340 static int __init async_stack_realloc(void)
 341 {
 342         unsigned long old, new;
 343
 344         old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
 345         new = stack_alloc();
 346         if (!new)
 347                 panic("Couldn't allocate async stack");
 348         S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
 349         free_pages(old, THREAD_SIZE_ORDER);
 350         return 0;
 351 }
 352 early_initcall(async_stack_realloc);
 353
 354 void __init arch_call_rest_init(void)
 355 {
 356         unsigned long stack;
 357
 358         stack = stack_alloc();
 359         if (!stack)
 360                 panic("Couldn't allocate kernel stack");
 361         current->stack = (void *) stack;
 362 #ifdef CONFIG_VMAP_STACK
 363         current->stack_vm_area = (void *) stack;
 364 #endif
 365         set_task_stack_end_magic(current);
 366         stack += STACK_INIT_OFFSET;
 367         S390_lowcore.kernel_stack = stack;
 368         CALL_ON_STACK_NORETURN(rest_init, stack);
 369 }
 370
 371 static void __init setup_lowcore_dat_off(void)
 372 {
 373         struct lowcore *lc;
 374
 375         /*
 376          * Setup lowcore for boot cpu
 377          */
 378         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 379         lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
 380         if (!lc)
 381                 panic("%s: Failed to allocate %zu bytes align=%zx\n",
 382                       __func__, sizeof(*lc), sizeof(*lc));
 383
 384         lc->restart_psw.mask = PSW_KERNEL_BITS;
 385         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 386         lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 387         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 388         lc->svc_new_psw.mask = PSW_KERNEL_BITS |
 389                 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 390         lc->svc_new_psw.addr = (unsigned long) system_call;
 391         lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 392         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 393         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 394         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 395         lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 396         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 397         lc->clock_comparator = clock_comparator_max;
 398         lc->nodat_stack = ((unsigned long) &init_thread_union)
 399                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 400         lc->current_task = (unsigned long)&init_task;
 401         lc->lpp = LPP_MAGIC;
 402         lc->machine_flags = S390_lowcore.machine_flags;
 403         lc->preempt_count = S390_lowcore.preempt_count;
 404         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 405         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 406                sizeof(lc->stfle_fac_list));
 407         memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
 408                sizeof(lc->alt_stfle_fac_list));
 409         nmi_alloc_boot_cpu(lc);
 410         vdso_alloc_boot_cpu(lc);
 411         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 412         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 413         lc->exit_timer = S390_lowcore.exit_timer;
 414         lc->user_timer = S390_lowcore.user_timer;
 415         lc->system_timer = S390_lowcore.system_timer;
 416         lc->steal_timer = S390_lowcore.steal_timer;
 417         lc->last_update_timer = S390_lowcore.last_update_timer;
 418         lc->last_update_clock = S390_lowcore.last_update_clock;
 419
 420         /*
 421          * Allocate the global restart stack which is the same for
 422          * all CPUs in cast *one* of them does a PSW restart.
 423          */
 424         restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 425         if (!restart_stack)
 426                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 427                       __func__, THREAD_SIZE, THREAD_SIZE);
 428         restart_stack += STACK_INIT_OFFSET;
 429
 430         /*
 431          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 432          * restart data to the absolute zero lowcore. This is necessary if
 433          * PSW restart is done on an offline CPU that has lowcore zero.
 434          */
 435         lc->restart_stack = (unsigned long) restart_stack;
 436         lc->restart_fn = (unsigned long) do_restart;
 437         lc->restart_data = 0;
 438         lc->restart_source = -1UL;
 439
 440         /* Setup absolute zero lowcore */
 441         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 442         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 443         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 444         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 445         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 446
 447         lc->spinlock_lockval = arch_spin_lockval(0);
 448         lc->spinlock_index = 0;
 449         arch_spin_lock_setup(0);
 450         lc->br_r1_trampoline = 0x07f1;  /* br %r1 */
 451
 452         set_prefix((u32)(unsigned long) lc);
 453         lowcore_ptr[0] = lc;
 454 }
 455
 456 static void __init setup_lowcore_dat_on(void)
 457 {
 458         __ctl_clear_bit(0, 28);
 459         S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
 460         S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
 461         S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
 462         S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
 463         __ctl_set_bit(0, 28);
 464 }
 465
 466 static struct resource code_resource = {
 467         .name  = "Kernel code",
 468         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 469 };
 470
 471 static struct resource data_resource = {
 472         .name = "Kernel data",
 473         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 474 };
 475
 476 static struct resource bss_resource = {
 477         .name = "Kernel bss",
 478         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 479 };
 480
 481 static struct resource __initdata *standard_resources[] = {
 482         &code_resource,
 483         &data_resource,
 484         &bss_resource,
 485 };
 486
 487 static void __init setup_resources(void)
 488 {
 489         struct resource *res, *std_res, *sub_res;
 490         struct memblock_region *reg;
 491         int j;
 492
 493         code_resource.start = (unsigned long) _text;
 494         code_resource.end = (unsigned long) _etext - 1;
 495         data_resource.start = (unsigned long) _etext;
 496         data_resource.end = (unsigned long) _edata - 1;
 497         bss_resource.start = (unsigned long) __bss_start;
 498         bss_resource.end = (unsigned long) __bss_stop - 1;
 499
 500         for_each_memblock(memory, reg) {
 501                 res = memblock_alloc(sizeof(*res), 8);
 502                 if (!res)
 503                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 504                               __func__, sizeof(*res), 8);
 505                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 506
 507                 res->name = "System RAM";
 508                 res->start = reg->base;
 509                 res->end = reg->base + reg->size - 1;
 510                 request_resource(&iomem_resource, res);
 511
 512                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 513                         std_res = standard_resources[j];
 514                         if (std_res->start < res->start ||
 515                             std_res->start > res->end)
 516                                 continue;
 517                         if (std_res->end > res->end) {
 518                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
 519                                 if (!sub_res)
 520                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 521                                               __func__, sizeof(*sub_res), 8);
 522                                 *sub_res = *std_res;
 523                                 sub_res->end = res->end;
 524                                 std_res->start = res->end + 1;
 525                                 request_resource(res, sub_res);
 526                         } else {
 527                                 request_resource(res, std_res);
 528                         }
 529                 }
 530         }
 531 #ifdef CONFIG_CRASH_DUMP
 532         /*
 533          * Re-add removed crash kernel memory as reserved memory. This makes
 534          * sure it will be mapped with the identity mapping and struct pages
 535          * will be created, so it can be resized later on.
 536          * However add it later since the crash kernel resource should not be
 537          * part of the System RAM resource.
 538          */
 539         if (crashk_res.end) {
 540                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 541                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 542                 insert_resource(&iomem_resource, &crashk_res);
 543         }
 544 #endif
 545 }
 546
 547 static void __init setup_memory_end(void)
 548 {
 549         unsigned long vmax, tmp;
 550
 551         /* Choose kernel address space layout: 3 or 4 levels. */
 552         if (IS_ENABLED(CONFIG_KASAN)) {
 553                 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
 554                            ? _REGION1_SIZE
 555                            : _REGION2_SIZE;
 556         } else {
 557                 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
 558                 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
 559                 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
 560                         vmax = _REGION2_SIZE; /* 3-level kernel page table */
 561                 else
 562                         vmax = _REGION1_SIZE; /* 4-level kernel page table */
 563         }
 564
 565         /* module area is at the end of the kernel address space. */
 566         MODULES_END = vmax;
 567         MODULES_VADDR = MODULES_END - MODULES_LEN;
 568         VMALLOC_END = MODULES_VADDR;
 569         VMALLOC_START = VMALLOC_END - vmalloc_size;
 570
 571         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 572         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 573         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 574         tmp = SECTION_ALIGN_UP(tmp);
 575         tmp = VMALLOC_START - tmp * sizeof(struct page);
 576         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 577         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 578         vmemmap = (struct page *) tmp;
 579
 580         /* Take care that memory_end is set and <= vmemmap */
 581         memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
 582 #ifdef CONFIG_KASAN
 583         /* fit in kasan shadow memory region between 1:1 and vmemmap */
 584         memory_end = min(memory_end, KASAN_SHADOW_START);
 585         vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
 586 #endif
 587         max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 588         memblock_remove(memory_end, ULONG_MAX);
 589
 590         pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
 591 }
 592
 593 #ifdef CONFIG_CRASH_DUMP
 594
 595 /*
 596  * When kdump is enabled, we have to ensure that no memory from
 597  * the area [0 - crashkernel memory size] and
 598  * [crashk_res.start - crashk_res.end] is set offline.
 599  */
 600 static int kdump_mem_notifier(struct notifier_block *nb,
 601                               unsigned long action, void *data)
 602 {
 603         struct memory_notify *arg = data;
 604
 605         if (action != MEM_GOING_OFFLINE)
 606                 return NOTIFY_OK;
 607         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 608                 return NOTIFY_BAD;
 609         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 610                 return NOTIFY_OK;
 611         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 612                 return NOTIFY_OK;
 613         return NOTIFY_BAD;
 614 }
 615
 616 static struct notifier_block kdump_mem_nb = {
 617         .notifier_call = kdump_mem_notifier,
 618 };
 619
 620 #endif
 621
 622 /*
 623  * Make sure that the area behind memory_end is protected
 624  */
 625 static void reserve_memory_end(void)
 626 {
 627         if (memory_end_set)
 628                 memblock_reserve(memory_end, ULONG_MAX);
 629 }
 630
 631 /*
 632  * Make sure that oldmem, where the dump is stored, is protected
 633  */
 634 static void reserve_oldmem(void)
 635 {
 636 #ifdef CONFIG_CRASH_DUMP
 637         if (OLDMEM_BASE)
 638                 /* Forget all memory above the running kdump system */
 639                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 640 #endif
 641 }
 642
 643 /*
 644  * Make sure that oldmem, where the dump is stored, is protected
 645  */
 646 static void remove_oldmem(void)
 647 {
 648 #ifdef CONFIG_CRASH_DUMP
 649         if (OLDMEM_BASE)
 650                 /* Forget all memory above the running kdump system */
 651                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 652 #endif
 653 }
 654
 655 /*
 656  * Reserve memory for kdump kernel to be loaded with kexec
 657  */
 658 static void __init reserve_crashkernel(void)
 659 {
 660 #ifdef CONFIG_CRASH_DUMP
 661         unsigned long long crash_base, crash_size;
 662         phys_addr_t low, high;
 663         int rc;
 664
 665         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 666                                &crash_base);
 667
 668         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 669         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 670         if (rc || crash_size == 0)
 671                 return;
 672
 673         if (memblock.memory.regions[0].size < crash_size) {
 674                 pr_info("crashkernel reservation failed: %s\n",
 675                         "first memory chunk must be at least crashkernel size");
 676                 return;
 677         }
 678
 679         low = crash_base ?: OLDMEM_BASE;
 680         high = low + crash_size;
 681         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 682                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 683                 crash_base = low;
 684         } else {
 685                 /* Find suitable area in free memory */
 686                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 687                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 688
 689                 if (crash_base && crash_base < low) {
 690                         pr_info("crashkernel reservation failed: %s\n",
 691                                 "crash_base too low");
 692                         return;
 693                 }
 694                 low = crash_base ?: low;
 695                 crash_base = memblock_find_in_range(low, high, crash_size,
 696                                                     KEXEC_CRASH_MEM_ALIGN);
 697         }
 698
 699         if (!crash_base) {
 700                 pr_info("crashkernel reservation failed: %s\n",
 701                         "no suitable area found");
 702                 return;
 703         }
 704
 705         if (register_memory_notifier(&kdump_mem_nb))
 706                 return;
 707
 708         if (!OLDMEM_BASE && MACHINE_IS_VM)
 709                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 710         crashk_res.start = crash_base;
 711         crashk_res.end = crash_base + crash_size - 1;
 712         memblock_remove(crash_base, crash_size);
 713         pr_info("Reserving %lluMB of memory at %lluMB "
 714                 "for crashkernel (System RAM: %luMB)\n",
 715                 crash_size >> 20, crash_base >> 20,
 716                 (unsigned long)memblock.memory.total_size >> 20);
 717         os_info_crashkernel_add(crash_base, crash_size);
 718 #endif
 719 }
 720
 721 /*
 722  * Reserve the initrd from being used by memblock
 723  */
 724 static void __init reserve_initrd(void)
 725 {
 726 #ifdef CONFIG_BLK_DEV_INITRD
 727         if (!INITRD_START || !INITRD_SIZE)
 728                 return;
 729         initrd_start = INITRD_START;
 730         initrd_end = initrd_start + INITRD_SIZE;
 731         memblock_reserve(INITRD_START, INITRD_SIZE);
 732 #endif
 733 }
 734
 735 /*
 736  * Reserve the memory area used to pass the certificate lists
 737  */
 738 static void __init reserve_certificate_list(void)
 739 {
 740         if (ipl_cert_list_addr)
 741                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
 742 }
 743
 744 static void __init reserve_mem_detect_info(void)
 745 {
 746         unsigned long start, size;
 747
 748         get_mem_detect_reserved(&start, &size);
 749         if (size)
 750                 memblock_reserve(start, size);
 751 }
 752
 753 static void __init free_mem_detect_info(void)
 754 {
 755         unsigned long start, size;
 756
 757         get_mem_detect_reserved(&start, &size);
 758         if (size)
 759                 memblock_free(start, size);
 760 }
 761
 762 static const char * __init get_mem_info_source(void)
 763 {
 764         switch (mem_detect.info_source) {
 765         case MEM_DETECT_SCLP_STOR_INFO:
 766                 return "sclp storage info";
 767         case MEM_DETECT_DIAG260:
 768                 return "diag260";
 769         case MEM_DETECT_SCLP_READ_INFO:
 770                 return "sclp read info";
 771         case MEM_DETECT_BIN_SEARCH:
 772                 return "binary search";
 773         }
 774         return "none";
 775 }
 776
 777 static void __init memblock_add_mem_detect_info(void)
 778 {
 779         unsigned long start, end;
 780         int i;
 781
 782         memblock_dbg("physmem info source: %s (%hhd)\n",
 783                      get_mem_info_source(), mem_detect.info_source);
 784         /* keep memblock lists close to the kernel */
 785         memblock_set_bottom_up(true);
 786         for_each_mem_detect_block(i, &start, &end) {
 787                 memblock_add(start, end - start);
 788                 memblock_physmem_add(start, end - start);
 789         }
 790         memblock_set_bottom_up(false);
 791         memblock_dump_all();
 792 }
 793
 794 /*
 795  * Check for initrd being in usable memory
 796  */
 797 static void __init check_initrd(void)
 798 {
 799 #ifdef CONFIG_BLK_DEV_INITRD
 800         if (INITRD_START && INITRD_SIZE &&
 801             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 802                 pr_err("The initial RAM disk does not fit into the memory\n");
 803                 memblock_free(INITRD_START, INITRD_SIZE);
 804                 initrd_start = initrd_end = 0;
 805         }
 806 #endif
 807 }
 808
 809 /*
 810  * Reserve memory used for lowcore/command line/kernel image.
 811  */
 812 static void __init reserve_kernel(void)
 813 {
 814         unsigned long start_pfn = PFN_UP(__pa(_end));
 815
 816         memblock_reserve(0, HEAD_END);
 817         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 818                          - (unsigned long)_stext);
 819         memblock_reserve(__sdma, __edma - __sdma);
 820 }
 821
 822 static void __init setup_memory(void)
 823 {
 824         struct memblock_region *reg;
 825
 826         /*
 827          * Init storage key for present memory
 828          */
 829         for_each_memblock(memory, reg) {
 830                 storage_key_init_range(reg->base, reg->base + reg->size);
 831         }
 832         psw_set_key(PAGE_DEFAULT_KEY);
 833
 834         /* Only cosmetics */
 835         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 836 }
 837
 838 /*
 839  * Setup hardware capabilities.
 840  */
 841 static int __init setup_hwcaps(void)
 842 {
 843         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 844         struct cpuid cpu_id;
 845         int i;
 846
 847         /*
 848          * The store facility list bits numbers as found in the principles
 849          * of operation are numbered with bit 1UL<<31 as number 0 to
 850          * bit 1UL<<0 as number 31.
 851          *   Bit 0: instructions named N3, "backported" to esa-mode
 852          *   Bit 2: z/Architecture mode is active
 853          *   Bit 7: the store-facility-list-extended facility is installed
 854          *   Bit 17: the message-security assist is installed
 855          *   Bit 19: the long-displacement facility is installed
 856          *   Bit 21: the extended-immediate facility is installed
 857          *   Bit 22: extended-translation facility 3 is installed
 858          *   Bit 30: extended-translation facility 3 enhancement facility
 859          * These get translated to:
 860          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 861          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 862          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 863          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 864          */
 865         for (i = 0; i < 6; i++)
 866                 if (test_facility(stfl_bits[i]))
 867                         elf_hwcap |= 1UL << i;
 868
 869         if (test_facility(22) && test_facility(30))
 870                 elf_hwcap |= HWCAP_S390_ETF3EH;
 871
 872         /*
 873          * Check for additional facilities with store-facility-list-extended.
 874          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 875          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 876          * as stored by stfl, bits 32-xxx contain additional facilities.
 877          * How many facility words are stored depends on the number of
 878          * doublewords passed to the instruction. The additional facilities
 879          * are:
 880          *   Bit 42: decimal floating point facility is installed
 881          *   Bit 44: perform floating point operation facility is installed
 882          * translated to:
 883          *   HWCAP_S390_DFP bit 6 (42 && 44).
 884          */
 885         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 886                 elf_hwcap |= HWCAP_S390_DFP;
 887
 888         /*
 889          * Huge page support HWCAP_S390_HPAGE is bit 7.
 890          */
 891         if (MACHINE_HAS_EDAT1)
 892                 elf_hwcap |= HWCAP_S390_HPAGE;
 893
 894         /*
 895          * 64-bit register support for 31-bit processes
 896          * HWCAP_S390_HIGH_GPRS is bit 9.
 897          */
 898         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 899
 900         /*
 901          * Transactional execution support HWCAP_S390_TE is bit 10.
 902          */
 903         if (MACHINE_HAS_TE)
 904                 elf_hwcap |= HWCAP_S390_TE;
 905
 906         /*
 907          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 908          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 909          * instead of facility bit 129.
 910          */
 911         if (MACHINE_HAS_VX) {
 912                 elf_hwcap |= HWCAP_S390_VXRS;
 913                 if (test_facility(134))
 914                         elf_hwcap |= HWCAP_S390_VXRS_EXT;
 915                 if (test_facility(135))
 916                         elf_hwcap |= HWCAP_S390_VXRS_BCD;
 917                 if (test_facility(148))
 918                         elf_hwcap |= HWCAP_S390_VXRS_EXT2;
 919                 if (test_facility(152))
 920                         elf_hwcap |= HWCAP_S390_VXRS_PDE;
 921         }
 922         if (test_facility(150))
 923                 elf_hwcap |= HWCAP_S390_SORT;
 924         if (test_facility(151))
 925                 elf_hwcap |= HWCAP_S390_DFLT;
 926
 927         /*
 928          * Guarded storage support HWCAP_S390_GS is bit 12.
 929          */
 930         if (MACHINE_HAS_GS)
 931                 elf_hwcap |= HWCAP_S390_GS;
 932
 933         get_cpu_id(&cpu_id);
 934         add_device_randomness(&cpu_id, sizeof(cpu_id));
 935         switch (cpu_id.machine) {
 936         case 0x2064:
 937         case 0x2066:
 938         default:        /* Use "z900" as default for 64 bit kernels. */
 939                 strcpy(elf_platform, "z900");
 940                 break;
 941         case 0x2084:
 942         case 0x2086:
 943                 strcpy(elf_platform, "z990");
 944                 break;
 945         case 0x2094:
 946         case 0x2096:
 947                 strcpy(elf_platform, "z9-109");
 948                 break;
 949         case 0x2097:
 950         case 0x2098:
 951                 strcpy(elf_platform, "z10");
 952                 break;
 953         case 0x2817:
 954         case 0x2818:
 955                 strcpy(elf_platform, "z196");
 956                 break;
 957         case 0x2827:
 958         case 0x2828:
 959                 strcpy(elf_platform, "zEC12");
 960                 break;
 961         case 0x2964:
 962         case 0x2965:
 963                 strcpy(elf_platform, "z13");
 964                 break;
 965         case 0x3906:
 966         case 0x3907:
 967                 strcpy(elf_platform, "z14");
 968                 break;
 969         case 0x8561:
 970         case 0x8562:
 971                 strcpy(elf_platform, "z15");
 972                 break;
 973         }
 974
 975         /*
 976          * Virtualization support HWCAP_INT_SIE is bit 0.
 977          */
 978         if (sclp.has_sief2)
 979                 int_hwcap |= HWCAP_INT_SIE;
 980
 981         return 0;
 982 }
 983 arch_initcall(setup_hwcaps);
 984
 985 /*
 986  * Add system information as device randomness
 987  */
 988 static void __init setup_randomness(void)
 989 {
 990         struct sysinfo_3_2_2 *vmms;
 991
 992         vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
 993                                                             PAGE_SIZE);
 994         if (!vmms)
 995                 panic("Failed to allocate memory for sysinfo structure\n");
 996
 997         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 998                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
 999         memblock_free((unsigned long) vmms, PAGE_SIZE);
1000 }
1001
1002 /*
1003  * Find the correct size for the task_struct. This depends on
1004  * the size of the struct fpu at the end of the thread_struct
1005  * which is embedded in the task_struct.
1006  */
1007 static void __init setup_task_size(void)
1008 {
1009         int task_size = sizeof(struct task_struct);
1010
1011         if (!MACHINE_HAS_VX) {
1012                 task_size -= sizeof(__vector128) * __NUM_VXRS;
1013                 task_size += sizeof(freg_t) * __NUM_FPRS;
1014         }
1015         arch_task_struct_size = task_size;
1016 }
1017
1018 /*
1019  * Issue diagnose 318 to set the control program name and
1020  * version codes.
1021  */
1022 static void __init setup_control_program_code(void)
1023 {
1024         union diag318_info diag318_info = {
1025                 .cpnc = CPNC_LINUX,
1026                 .cpvc_linux = 0,
1027                 .cpvc_distro = {0},
1028         };
1029
1030         if (!sclp.has_diag318)
1031                 return;
1032
1033         diag_stat_inc(DIAG_STAT_X318);
1034         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1035 }
1036
1037 /*
1038  * Print the component list from the IPL report
1039  */
1040 static void __init log_component_list(void)
1041 {
1042         struct ipl_rb_component_entry *ptr, *end;
1043         char *str;
1044
1045         if (!early_ipl_comp_list_addr)
1046                 return;
1047         if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1048                 pr_info("Linux is running with Secure-IPL enabled\n");
1049         else
1050                 pr_info("Linux is running with Secure-IPL disabled\n");
1051         ptr = (void *) early_ipl_comp_list_addr;
1052         end = (void *) ptr + early_ipl_comp_list_size;
1053         pr_info("The IPL report contains the following components:\n");
1054         while (ptr < end) {
1055                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1056                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1057                                 str = "signed, verified";
1058                         else
1059                                 str = "signed, verification failed";
1060                 } else {
1061                         str = "not signed";
1062                 }
1063                 pr_info("%016llx - %016llx (%s)\n",
1064                         ptr->addr, ptr->addr + ptr->len, str);
1065                 ptr++;
1066         }
1067 }
1068
1069 /*
1070  * Setup function called from init/main.c just after the banner
1071  * was printed.
1072  */
1073
1074 void __init setup_arch(char **cmdline_p)
1075 {
1076         /*
1077          * print what head.S has found out about the machine
1078          */
1079         if (MACHINE_IS_VM)
1080                 pr_info("Linux is running as a z/VM "
1081                         "guest operating system in 64-bit mode\n");
1082         else if (MACHINE_IS_KVM)
1083                 pr_info("Linux is running under KVM in 64-bit mode\n");
1084         else if (MACHINE_IS_LPAR)
1085                 pr_info("Linux is running natively in 64-bit mode\n");
1086         else
1087                 pr_info("Linux is running as a guest in 64-bit mode\n");
1088
1089         log_component_list();
1090
1091         /* Have one command line that is parsed and saved in /proc/cmdline */
1092         /* boot_command_line has been already set up in early.c */
1093         *cmdline_p = boot_command_line;
1094
1095         ROOT_DEV = Root_RAM0;
1096
1097         init_mm.start_code = (unsigned long) _text;
1098         init_mm.end_code = (unsigned long) _etext;
1099         init_mm.end_data = (unsigned long) _edata;
1100         init_mm.brk = (unsigned long) _end;
1101
1102         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1103                 nospec_auto_detect();
1104
1105         parse_early_param();
1106 #ifdef CONFIG_CRASH_DUMP
1107         /* Deactivate elfcorehdr= kernel parameter */
1108         elfcorehdr_addr = ELFCORE_ADDR_MAX;
1109 #endif
1110
1111         os_info_init();
1112         setup_ipl();
1113         setup_task_size();
1114         setup_control_program_code();
1115
1116         /* Do some memory reservations *before* memory is added to memblock */
1117         reserve_memory_end();
1118         reserve_oldmem();
1119         reserve_kernel();
1120         reserve_initrd();
1121         reserve_certificate_list();
1122         reserve_mem_detect_info();
1123         memblock_allow_resize();
1124
1125         /* Get information about *all* installed memory */
1126         memblock_add_mem_detect_info();
1127
1128         free_mem_detect_info();
1129         remove_oldmem();
1130
1131         /*
1132          * Make sure all chunks are MAX_ORDER aligned so we don't need the
1133          * extra checks that HOLES_IN_ZONE would require.
1134          *
1135          * Is this still required?
1136          */
1137         memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
1138
1139         setup_memory_end();
1140         setup_memory();
1141         dma_contiguous_reserve(memory_end);
1142         vmcp_cma_reserve();
1143
1144         check_initrd();
1145         reserve_crashkernel();
1146 #ifdef CONFIG_CRASH_DUMP
1147         /*
1148          * Be aware that smp_save_dump_cpus() triggers a system reset.
1149          * Therefore CPU and device initialization should be done afterwards.
1150          */
1151         smp_save_dump_cpus();
1152 #endif
1153
1154         setup_resources();
1155         setup_lowcore_dat_off();
1156         smp_fill_possible_mask();
1157         cpu_detect_mhz_feature();
1158         cpu_init();
1159         numa_setup();
1160         smp_detect_cpus();
1161         topology_init_early();
1162
1163         /*
1164          * Create kernel page tables and switch to virtual addressing.
1165          */
1166         paging_init();
1167
1168         /*
1169          * After paging_init created the kernel page table, the new PSWs
1170          * in lowcore can now run with DAT enabled.
1171          */
1172         setup_lowcore_dat_on();
1173
1174         /* Setup default console */
1175         conmode_default();
1176         set_preferred_console();
1177
1178         apply_alternative_instructions();
1179         if (IS_ENABLED(CONFIG_EXPOLINE))
1180                 nospec_init_branches();
1181
1182         /* Setup zfcpdump support */
1183         setup_zfcpdump();
1184
1185         /* Add system specific data to the random pool */
1186         setup_randomness();
1187 }