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(max_physmem_end);
 103 struct mem_detect_info __bootdata(mem_detect);
 104
 105 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
 106 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
 107 unsigned long __bootdata_preserved(__swsusp_reset_dma);
 108 unsigned long __bootdata_preserved(__stext_dma);
 109 unsigned long __bootdata_preserved(__etext_dma);
 110 unsigned long __bootdata_preserved(__sdma);
 111 unsigned long __bootdata_preserved(__edma);
 112 unsigned long __bootdata_preserved(__kaslr_offset);
 113
 114 unsigned long VMALLOC_START;
 115 EXPORT_SYMBOL(VMALLOC_START);
 116
 117 unsigned long VMALLOC_END;
 118 EXPORT_SYMBOL(VMALLOC_END);
 119
 120 struct page *vmemmap;
 121 EXPORT_SYMBOL(vmemmap);
 122
 123 unsigned long MODULES_VADDR;
 124 unsigned long MODULES_END;
 125
 126 /* An array with a pointer to the lowcore of every CPU. */
 127 struct lowcore *lowcore_ptr[NR_CPUS];
 128 EXPORT_SYMBOL(lowcore_ptr);
 129
 130 /*
 131  * This is set up by the setup-routine at boot-time
 132  * for S390 need to find out, what we have to setup
 133  * using address 0x10400 ...
 134  */
 135
 136 #include <asm/setup.h>
 137
 138 /*
 139  * condev= and conmode= setup parameter.
 140  */
 141
 142 static int __init condev_setup(char *str)
 143 {
 144         int vdev;
 145
 146         vdev = simple_strtoul(str, &str, 0);
 147         if (vdev >= 0 && vdev < 65536) {
 148                 console_devno = vdev;
 149                 console_irq = -1;
 150         }
 151         return 1;
 152 }
 153
 154 __setup("condev=", condev_setup);
 155
 156 static void __init set_preferred_console(void)
 157 {
 158         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 159                 add_preferred_console("ttyS", 0, NULL);
 160         else if (CONSOLE_IS_3270)
 161                 add_preferred_console("tty3270", 0, NULL);
 162         else if (CONSOLE_IS_VT220)
 163                 add_preferred_console("ttyS", 1, NULL);
 164         else if (CONSOLE_IS_HVC)
 165                 add_preferred_console("hvc", 0, NULL);
 166 }
 167
 168 static int __init conmode_setup(char *str)
 169 {
 170 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 171         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 172                 SET_CONSOLE_SCLP;
 173 #endif
 174 #if defined(CONFIG_TN3215_CONSOLE)
 175         if (strncmp(str, "3215", 5) == 0)
 176                 SET_CONSOLE_3215;
 177 #endif
 178 #if defined(CONFIG_TN3270_CONSOLE)
 179         if (strncmp(str, "3270", 5) == 0)
 180                 SET_CONSOLE_3270;
 181 #endif
 182         set_preferred_console();
 183         return 1;
 184 }
 185
 186 __setup("conmode=", conmode_setup);
 187
 188 static void __init conmode_default(void)
 189 {
 190         char query_buffer[1024];
 191         char *ptr;
 192
 193         if (MACHINE_IS_VM) {
 194                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 195                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 196                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 197                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 198                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 199                 ptr = strstr(query_buffer, "CONMODE");
 200                 /*
 201                  * Set the conmode to 3215 so that the device recognition
 202                  * will set the cu_type of the console to 3215. If the
 203                  * conmode is 3270 and we don't set it back then both
 204                  * 3215 and the 3270 driver will try to access the console
 205                  * device (3215 as console and 3270 as normal tty).
 206                  */
 207                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 208                 if (ptr == NULL) {
 209 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 210                         SET_CONSOLE_SCLP;
 211 #endif
 212                         return;
 213                 }
 214                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 215 #if defined(CONFIG_TN3270_CONSOLE)
 216                         SET_CONSOLE_3270;
 217 #elif defined(CONFIG_TN3215_CONSOLE)
 218                         SET_CONSOLE_3215;
 219 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 220                         SET_CONSOLE_SCLP;
 221 #endif
 222                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 223 #if defined(CONFIG_TN3215_CONSOLE)
 224                         SET_CONSOLE_3215;
 225 #elif defined(CONFIG_TN3270_CONSOLE)
 226                         SET_CONSOLE_3270;
 227 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 228                         SET_CONSOLE_SCLP;
 229 #endif
 230                 }
 231         } else if (MACHINE_IS_KVM) {
 232                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
 233                         SET_CONSOLE_VT220;
 234                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
 235                         SET_CONSOLE_SCLP;
 236                 else
 237                         SET_CONSOLE_HVC;
 238         } else {
 239 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 240                 SET_CONSOLE_SCLP;
 241 #endif
 242         }
 243         if (IS_ENABLED(CONFIG_VT) && IS_ENABLED(CONFIG_DUMMY_CONSOLE))
 244                 conswitchp = &dummy_con;
 245 }
 246
 247 #ifdef CONFIG_CRASH_DUMP
 248 static void __init setup_zfcpdump(void)
 249 {
 250         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 251                 return;
 252         if (OLDMEM_BASE)
 253                 return;
 254         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 255         console_loglevel = 2;
 256 }
 257 #else
 258 static inline void setup_zfcpdump(void) {}
 259 #endif /* CONFIG_CRASH_DUMP */
 260
 261  /*
 262  * Reboot, halt and power_off stubs. They just call _machine_restart,
 263  * _machine_halt or _machine_power_off.
 264  */
 265
 266 void machine_restart(char *command)
 267 {
 268         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 269                 /*
 270                  * Only unblank the console if we are called in enabled
 271                  * context or a bust_spinlocks cleared the way for us.
 272                  */
 273                 console_unblank();
 274         _machine_restart(command);
 275 }
 276
 277 void machine_halt(void)
 278 {
 279         if (!in_interrupt() || oops_in_progress)
 280                 /*
 281                  * Only unblank the console if we are called in enabled
 282                  * context or a bust_spinlocks cleared the way for us.
 283                  */
 284                 console_unblank();
 285         _machine_halt();
 286 }
 287
 288 void machine_power_off(void)
 289 {
 290         if (!in_interrupt() || oops_in_progress)
 291                 /*
 292                  * Only unblank the console if we are called in enabled
 293                  * context or a bust_spinlocks cleared the way for us.
 294                  */
 295                 console_unblank();
 296         _machine_power_off();
 297 }
 298
 299 /*
 300  * Dummy power off function.
 301  */
 302 void (*pm_power_off)(void) = machine_power_off;
 303 EXPORT_SYMBOL_GPL(pm_power_off);
 304
 305 static int __init parse_vmalloc(char *arg)
 306 {
 307         if (!arg)
 308                 return -EINVAL;
 309         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
 310         return 0;
 311 }
 312 early_param("vmalloc", parse_vmalloc);
 313
 314 void *restart_stack __section(.data);
 315
 316 unsigned long stack_alloc(void)
 317 {
 318 #ifdef CONFIG_VMAP_STACK
 319         return (unsigned long)
 320                 __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
 321                                      VMALLOC_START, VMALLOC_END,
 322                                      THREADINFO_GFP,
 323                                      PAGE_KERNEL, 0, NUMA_NO_NODE,
 324                                      __builtin_return_address(0));
 325 #else
 326         return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 327 #endif
 328 }
 329
 330 void stack_free(unsigned long stack)
 331 {
 332 #ifdef CONFIG_VMAP_STACK
 333         vfree((void *) stack);
 334 #else
 335         free_pages(stack, THREAD_SIZE_ORDER);
 336 #endif
 337 }
 338
 339 int __init arch_early_irq_init(void)
 340 {
 341         unsigned long stack;
 342
 343         stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 344         if (!stack)
 345                 panic("Couldn't allocate async stack");
 346         S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
 347         return 0;
 348 }
 349
 350 static int __init async_stack_realloc(void)
 351 {
 352         unsigned long old, new;
 353
 354         old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
 355         new = stack_alloc();
 356         if (!new)
 357                 panic("Couldn't allocate async stack");
 358         S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
 359         free_pages(old, THREAD_SIZE_ORDER);
 360         return 0;
 361 }
 362 early_initcall(async_stack_realloc);
 363
 364 void __init arch_call_rest_init(void)
 365 {
 366         struct stack_frame *frame;
 367         unsigned long stack;
 368
 369         stack = stack_alloc();
 370         if (!stack)
 371                 panic("Couldn't allocate kernel stack");
 372         current->stack = (void *) stack;
 373 #ifdef CONFIG_VMAP_STACK
 374         current->stack_vm_area = (void *) stack;
 375 #endif
 376         set_task_stack_end_magic(current);
 377         stack += STACK_INIT_OFFSET;
 378         S390_lowcore.kernel_stack = stack;
 379         frame = (struct stack_frame *) stack;
 380         memset(frame, 0, sizeof(*frame));
 381         /* Branch to rest_init on the new stack, never returns */
 382         asm volatile(
 383                 "       la      15,0(%[_frame])\n"
 384                 "       jg      rest_init\n"
 385                 : : [_frame] "a" (frame));
 386 }
 387
 388 static void __init setup_lowcore_dat_off(void)
 389 {
 390         struct lowcore *lc;
 391
 392         /*
 393          * Setup lowcore for boot cpu
 394          */
 395         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 396         lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
 397         if (!lc)
 398                 panic("%s: Failed to allocate %zu bytes align=%zx\n",
 399                       __func__, sizeof(*lc), sizeof(*lc));
 400
 401         lc->restart_psw.mask = PSW_KERNEL_BITS;
 402         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 403         lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 404         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 405         lc->svc_new_psw.mask = PSW_KERNEL_BITS |
 406                 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 407         lc->svc_new_psw.addr = (unsigned long) system_call;
 408         lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 409         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 410         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 411         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 412         lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 413         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 414         lc->clock_comparator = clock_comparator_max;
 415         lc->nodat_stack = ((unsigned long) &init_thread_union)
 416                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 417         lc->current_task = (unsigned long)&init_task;
 418         lc->lpp = LPP_MAGIC;
 419         lc->machine_flags = S390_lowcore.machine_flags;
 420         lc->preempt_count = S390_lowcore.preempt_count;
 421         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 422         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 423                sizeof(lc->stfle_fac_list));
 424         memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
 425                sizeof(lc->alt_stfle_fac_list));
 426         nmi_alloc_boot_cpu(lc);
 427         vdso_alloc_boot_cpu(lc);
 428         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 429         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 430         lc->exit_timer = S390_lowcore.exit_timer;
 431         lc->user_timer = S390_lowcore.user_timer;
 432         lc->system_timer = S390_lowcore.system_timer;
 433         lc->steal_timer = S390_lowcore.steal_timer;
 434         lc->last_update_timer = S390_lowcore.last_update_timer;
 435         lc->last_update_clock = S390_lowcore.last_update_clock;
 436
 437         /*
 438          * Allocate the global restart stack which is the same for
 439          * all CPUs in cast *one* of them does a PSW restart.
 440          */
 441         restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 442         if (!restart_stack)
 443                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 444                       __func__, THREAD_SIZE, THREAD_SIZE);
 445         restart_stack += STACK_INIT_OFFSET;
 446
 447         /*
 448          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 449          * restart data to the absolute zero lowcore. This is necessary if
 450          * PSW restart is done on an offline CPU that has lowcore zero.
 451          */
 452         lc->restart_stack = (unsigned long) restart_stack;
 453         lc->restart_fn = (unsigned long) do_restart;
 454         lc->restart_data = 0;
 455         lc->restart_source = -1UL;
 456
 457         /* Setup absolute zero lowcore */
 458         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 459         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 460         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 461         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 462         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 463
 464         lc->spinlock_lockval = arch_spin_lockval(0);
 465         lc->spinlock_index = 0;
 466         arch_spin_lock_setup(0);
 467         lc->br_r1_trampoline = 0x07f1;  /* br %r1 */
 468
 469         set_prefix((u32)(unsigned long) lc);
 470         lowcore_ptr[0] = lc;
 471 }
 472
 473 static void __init setup_lowcore_dat_on(void)
 474 {
 475         __ctl_clear_bit(0, 28);
 476         S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
 477         S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
 478         S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
 479         S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
 480         __ctl_set_bit(0, 28);
 481 }
 482
 483 static struct resource code_resource = {
 484         .name  = "Kernel code",
 485         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 486 };
 487
 488 static struct resource data_resource = {
 489         .name = "Kernel data",
 490         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 491 };
 492
 493 static struct resource bss_resource = {
 494         .name = "Kernel bss",
 495         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 496 };
 497
 498 static struct resource __initdata *standard_resources[] = {
 499         &code_resource,
 500         &data_resource,
 501         &bss_resource,
 502 };
 503
 504 static void __init setup_resources(void)
 505 {
 506         struct resource *res, *std_res, *sub_res;
 507         struct memblock_region *reg;
 508         int j;
 509
 510         code_resource.start = (unsigned long) _text;
 511         code_resource.end = (unsigned long) _etext - 1;
 512         data_resource.start = (unsigned long) _etext;
 513         data_resource.end = (unsigned long) _edata - 1;
 514         bss_resource.start = (unsigned long) __bss_start;
 515         bss_resource.end = (unsigned long) __bss_stop - 1;
 516
 517         for_each_memblock(memory, reg) {
 518                 res = memblock_alloc(sizeof(*res), 8);
 519                 if (!res)
 520                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 521                               __func__, sizeof(*res), 8);
 522                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 523
 524                 res->name = "System RAM";
 525                 res->start = reg->base;
 526                 res->end = reg->base + reg->size - 1;
 527                 request_resource(&iomem_resource, res);
 528
 529                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 530                         std_res = standard_resources[j];
 531                         if (std_res->start < res->start ||
 532                             std_res->start > res->end)
 533                                 continue;
 534                         if (std_res->end > res->end) {
 535                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
 536                                 if (!sub_res)
 537                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 538                                               __func__, sizeof(*sub_res), 8);
 539                                 *sub_res = *std_res;
 540                                 sub_res->end = res->end;
 541                                 std_res->start = res->end + 1;
 542                                 request_resource(res, sub_res);
 543                         } else {
 544                                 request_resource(res, std_res);
 545                         }
 546                 }
 547         }
 548 #ifdef CONFIG_CRASH_DUMP
 549         /*
 550          * Re-add removed crash kernel memory as reserved memory. This makes
 551          * sure it will be mapped with the identity mapping and struct pages
 552          * will be created, so it can be resized later on.
 553          * However add it later since the crash kernel resource should not be
 554          * part of the System RAM resource.
 555          */
 556         if (crashk_res.end) {
 557                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 558                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 559                 insert_resource(&iomem_resource, &crashk_res);
 560         }
 561 #endif
 562 }
 563
 564 static void __init setup_memory_end(void)
 565 {
 566         unsigned long vmax, vmalloc_size, tmp;
 567
 568         /* Choose kernel address space layout: 3 or 4 levels. */
 569         vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
 570         if (IS_ENABLED(CONFIG_KASAN)) {
 571                 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
 572                            ? _REGION1_SIZE
 573                            : _REGION2_SIZE;
 574         } else {
 575                 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
 576                 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
 577                 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
 578                         vmax = _REGION2_SIZE; /* 3-level kernel page table */
 579                 else
 580                         vmax = _REGION1_SIZE; /* 4-level kernel page table */
 581         }
 582
 583         /* module area is at the end of the kernel address space. */
 584         MODULES_END = vmax;
 585         MODULES_VADDR = MODULES_END - MODULES_LEN;
 586         VMALLOC_END = MODULES_VADDR;
 587         VMALLOC_START = VMALLOC_END - vmalloc_size;
 588
 589         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 590         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 591         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 592         tmp = SECTION_ALIGN_UP(tmp);
 593         tmp = VMALLOC_START - tmp * sizeof(struct page);
 594         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 595         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 596         vmemmap = (struct page *) tmp;
 597
 598         /* Take care that memory_end is set and <= vmemmap */
 599         memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
 600 #ifdef CONFIG_KASAN
 601         /* fit in kasan shadow memory region between 1:1 and vmemmap */
 602         memory_end = min(memory_end, KASAN_SHADOW_START);
 603         vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
 604 #endif
 605         max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 606         memblock_remove(memory_end, ULONG_MAX);
 607
 608         pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
 609 }
 610
 611 #ifdef CONFIG_CRASH_DUMP
 612
 613 /*
 614  * When kdump is enabled, we have to ensure that no memory from
 615  * the area [0 - crashkernel memory size] and
 616  * [crashk_res.start - crashk_res.end] is set offline.
 617  */
 618 static int kdump_mem_notifier(struct notifier_block *nb,
 619                               unsigned long action, void *data)
 620 {
 621         struct memory_notify *arg = data;
 622
 623         if (action != MEM_GOING_OFFLINE)
 624                 return NOTIFY_OK;
 625         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 626                 return NOTIFY_BAD;
 627         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 628                 return NOTIFY_OK;
 629         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 630                 return NOTIFY_OK;
 631         return NOTIFY_BAD;
 632 }
 633
 634 static struct notifier_block kdump_mem_nb = {
 635         .notifier_call = kdump_mem_notifier,
 636 };
 637
 638 #endif
 639
 640 /*
 641  * Make sure that the area behind memory_end is protected
 642  */
 643 static void reserve_memory_end(void)
 644 {
 645         if (memory_end_set)
 646                 memblock_reserve(memory_end, ULONG_MAX);
 647 }
 648
 649 /*
 650  * Make sure that oldmem, where the dump is stored, is protected
 651  */
 652 static void reserve_oldmem(void)
 653 {
 654 #ifdef CONFIG_CRASH_DUMP
 655         if (OLDMEM_BASE)
 656                 /* Forget all memory above the running kdump system */
 657                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 658 #endif
 659 }
 660
 661 /*
 662  * Make sure that oldmem, where the dump is stored, is protected
 663  */
 664 static void remove_oldmem(void)
 665 {
 666 #ifdef CONFIG_CRASH_DUMP
 667         if (OLDMEM_BASE)
 668                 /* Forget all memory above the running kdump system */
 669                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 670 #endif
 671 }
 672
 673 /*
 674  * Reserve memory for kdump kernel to be loaded with kexec
 675  */
 676 static void __init reserve_crashkernel(void)
 677 {
 678 #ifdef CONFIG_CRASH_DUMP
 679         unsigned long long crash_base, crash_size;
 680         phys_addr_t low, high;
 681         int rc;
 682
 683         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 684                                &crash_base);
 685
 686         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 687         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 688         if (rc || crash_size == 0)
 689                 return;
 690
 691         if (memblock.memory.regions[0].size < crash_size) {
 692                 pr_info("crashkernel reservation failed: %s\n",
 693                         "first memory chunk must be at least crashkernel size");
 694                 return;
 695         }
 696
 697         low = crash_base ?: OLDMEM_BASE;
 698         high = low + crash_size;
 699         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 700                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 701                 crash_base = low;
 702         } else {
 703                 /* Find suitable area in free memory */
 704                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 705                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 706
 707                 if (crash_base && crash_base < low) {
 708                         pr_info("crashkernel reservation failed: %s\n",
 709                                 "crash_base too low");
 710                         return;
 711                 }
 712                 low = crash_base ?: low;
 713                 crash_base = memblock_find_in_range(low, high, crash_size,
 714                                                     KEXEC_CRASH_MEM_ALIGN);
 715         }
 716
 717         if (!crash_base) {
 718                 pr_info("crashkernel reservation failed: %s\n",
 719                         "no suitable area found");
 720                 return;
 721         }
 722
 723         if (register_memory_notifier(&kdump_mem_nb))
 724                 return;
 725
 726         if (!OLDMEM_BASE && MACHINE_IS_VM)
 727                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 728         crashk_res.start = crash_base;
 729         crashk_res.end = crash_base + crash_size - 1;
 730         memblock_remove(crash_base, crash_size);
 731         pr_info("Reserving %lluMB of memory at %lluMB "
 732                 "for crashkernel (System RAM: %luMB)\n",
 733                 crash_size >> 20, crash_base >> 20,
 734                 (unsigned long)memblock.memory.total_size >> 20);
 735         os_info_crashkernel_add(crash_base, crash_size);
 736 #endif
 737 }
 738
 739 /*
 740  * Reserve the initrd from being used by memblock
 741  */
 742 static void __init reserve_initrd(void)
 743 {
 744 #ifdef CONFIG_BLK_DEV_INITRD
 745         if (!INITRD_START || !INITRD_SIZE)
 746                 return;
 747         initrd_start = INITRD_START;
 748         initrd_end = initrd_start + INITRD_SIZE;
 749         memblock_reserve(INITRD_START, INITRD_SIZE);
 750 #endif
 751 }
 752
 753 /*
 754  * Reserve the memory area used to pass the certificate lists
 755  */
 756 static void __init reserve_certificate_list(void)
 757 {
 758         if (ipl_cert_list_addr)
 759                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
 760 }
 761
 762 static void __init reserve_mem_detect_info(void)
 763 {
 764         unsigned long start, size;
 765
 766         get_mem_detect_reserved(&start, &size);
 767         if (size)
 768                 memblock_reserve(start, size);
 769 }
 770
 771 static void __init free_mem_detect_info(void)
 772 {
 773         unsigned long start, size;
 774
 775         get_mem_detect_reserved(&start, &size);
 776         if (size)
 777                 memblock_free(start, size);
 778 }
 779
 780 static void __init memblock_physmem_add(phys_addr_t start, phys_addr_t size)
 781 {
 782         memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n",
 783                      start, start + size - 1);
 784         memblock_add_range(&memblock.memory, start, size, 0, 0);
 785         memblock_add_range(&memblock.physmem, start, size, 0, 0);
 786 }
 787
 788 static const char * __init get_mem_info_source(void)
 789 {
 790         switch (mem_detect.info_source) {
 791         case MEM_DETECT_SCLP_STOR_INFO:
 792                 return "sclp storage info";
 793         case MEM_DETECT_DIAG260:
 794                 return "diag260";
 795         case MEM_DETECT_SCLP_READ_INFO:
 796                 return "sclp read info";
 797         case MEM_DETECT_BIN_SEARCH:
 798                 return "binary search";
 799         }
 800         return "none";
 801 }
 802
 803 static void __init memblock_add_mem_detect_info(void)
 804 {
 805         unsigned long start, end;
 806         int i;
 807
 808         memblock_dbg("physmem info source: %s (%hhd)\n",
 809                      get_mem_info_source(), mem_detect.info_source);
 810         /* keep memblock lists close to the kernel */
 811         memblock_set_bottom_up(true);
 812         for_each_mem_detect_block(i, &start, &end)
 813                 memblock_physmem_add(start, end - start);
 814         memblock_set_bottom_up(false);
 815         memblock_dump_all();
 816 }
 817
 818 /*
 819  * Check for initrd being in usable memory
 820  */
 821 static void __init check_initrd(void)
 822 {
 823 #ifdef CONFIG_BLK_DEV_INITRD
 824         if (INITRD_START && INITRD_SIZE &&
 825             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 826                 pr_err("The initial RAM disk does not fit into the memory\n");
 827                 memblock_free(INITRD_START, INITRD_SIZE);
 828                 initrd_start = initrd_end = 0;
 829         }
 830 #endif
 831 }
 832
 833 /*
 834  * Reserve memory used for lowcore/command line/kernel image.
 835  */
 836 static void __init reserve_kernel(void)
 837 {
 838         unsigned long start_pfn = PFN_UP(__pa(_end));
 839
 840         memblock_reserve(0, HEAD_END);
 841         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 842                          - (unsigned long)_stext);
 843         memblock_reserve(__sdma, __edma - __sdma);
 844 }
 845
 846 static void __init setup_memory(void)
 847 {
 848         struct memblock_region *reg;
 849
 850         /*
 851          * Init storage key for present memory
 852          */
 853         for_each_memblock(memory, reg) {
 854                 storage_key_init_range(reg->base, reg->base + reg->size);
 855         }
 856         psw_set_key(PAGE_DEFAULT_KEY);
 857
 858         /* Only cosmetics */
 859         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 860 }
 861
 862 /*
 863  * Setup hardware capabilities.
 864  */
 865 static int __init setup_hwcaps(void)
 866 {
 867         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 868         struct cpuid cpu_id;
 869         int i;
 870
 871         /*
 872          * The store facility list bits numbers as found in the principles
 873          * of operation are numbered with bit 1UL<<31 as number 0 to
 874          * bit 1UL<<0 as number 31.
 875          *   Bit 0: instructions named N3, "backported" to esa-mode
 876          *   Bit 2: z/Architecture mode is active
 877          *   Bit 7: the store-facility-list-extended facility is installed
 878          *   Bit 17: the message-security assist is installed
 879          *   Bit 19: the long-displacement facility is installed
 880          *   Bit 21: the extended-immediate facility is installed
 881          *   Bit 22: extended-translation facility 3 is installed
 882          *   Bit 30: extended-translation facility 3 enhancement facility
 883          * These get translated to:
 884          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 885          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 886          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 887          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 888          */
 889         for (i = 0; i < 6; i++)
 890                 if (test_facility(stfl_bits[i]))
 891                         elf_hwcap |= 1UL << i;
 892
 893         if (test_facility(22) && test_facility(30))
 894                 elf_hwcap |= HWCAP_S390_ETF3EH;
 895
 896         /*
 897          * Check for additional facilities with store-facility-list-extended.
 898          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 899          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 900          * as stored by stfl, bits 32-xxx contain additional facilities.
 901          * How many facility words are stored depends on the number of
 902          * doublewords passed to the instruction. The additional facilities
 903          * are:
 904          *   Bit 42: decimal floating point facility is installed
 905          *   Bit 44: perform floating point operation facility is installed
 906          * translated to:
 907          *   HWCAP_S390_DFP bit 6 (42 && 44).
 908          */
 909         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 910                 elf_hwcap |= HWCAP_S390_DFP;
 911
 912         /*
 913          * Huge page support HWCAP_S390_HPAGE is bit 7.
 914          */
 915         if (MACHINE_HAS_EDAT1)
 916                 elf_hwcap |= HWCAP_S390_HPAGE;
 917
 918         /*
 919          * 64-bit register support for 31-bit processes
 920          * HWCAP_S390_HIGH_GPRS is bit 9.
 921          */
 922         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 923
 924         /*
 925          * Transactional execution support HWCAP_S390_TE is bit 10.
 926          */
 927         if (MACHINE_HAS_TE)
 928                 elf_hwcap |= HWCAP_S390_TE;
 929
 930         /*
 931          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 932          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 933          * instead of facility bit 129.
 934          */
 935         if (MACHINE_HAS_VX) {
 936                 elf_hwcap |= HWCAP_S390_VXRS;
 937                 if (test_facility(134))
 938                         elf_hwcap |= HWCAP_S390_VXRS_EXT;
 939                 if (test_facility(135))
 940                         elf_hwcap |= HWCAP_S390_VXRS_BCD;
 941                 if (test_facility(148))
 942                         elf_hwcap |= HWCAP_S390_VXRS_EXT2;
 943                 if (test_facility(152))
 944                         elf_hwcap |= HWCAP_S390_VXRS_PDE;
 945         }
 946         if (test_facility(150))
 947                 elf_hwcap |= HWCAP_S390_SORT;
 948         if (test_facility(151))
 949                 elf_hwcap |= HWCAP_S390_DFLT;
 950
 951         /*
 952          * Guarded storage support HWCAP_S390_GS is bit 12.
 953          */
 954         if (MACHINE_HAS_GS)
 955                 elf_hwcap |= HWCAP_S390_GS;
 956
 957         get_cpu_id(&cpu_id);
 958         add_device_randomness(&cpu_id, sizeof(cpu_id));
 959         switch (cpu_id.machine) {
 960         case 0x2064:
 961         case 0x2066:
 962         default:        /* Use "z900" as default for 64 bit kernels. */
 963                 strcpy(elf_platform, "z900");
 964                 break;
 965         case 0x2084:
 966         case 0x2086:
 967                 strcpy(elf_platform, "z990");
 968                 break;
 969         case 0x2094:
 970         case 0x2096:
 971                 strcpy(elf_platform, "z9-109");
 972                 break;
 973         case 0x2097:
 974         case 0x2098:
 975                 strcpy(elf_platform, "z10");
 976                 break;
 977         case 0x2817:
 978         case 0x2818:
 979                 strcpy(elf_platform, "z196");
 980                 break;
 981         case 0x2827:
 982         case 0x2828:
 983                 strcpy(elf_platform, "zEC12");
 984                 break;
 985         case 0x2964:
 986         case 0x2965:
 987                 strcpy(elf_platform, "z13");
 988                 break;
 989         case 0x3906:
 990         case 0x3907:
 991                 strcpy(elf_platform, "z14");
 992                 break;
 993         }
 994
 995         /*
 996          * Virtualization support HWCAP_INT_SIE is bit 0.
 997          */
 998         if (sclp.has_sief2)
 999                 int_hwcap |= HWCAP_INT_SIE;
1000
1001         return 0;
1002 }
1003 arch_initcall(setup_hwcaps);
1004
1005 /*
1006  * Add system information as device randomness
1007  */
1008 static void __init setup_randomness(void)
1009 {
1010         struct sysinfo_3_2_2 *vmms;
1011
1012         vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
1013                                                             PAGE_SIZE);
1014         if (!vmms)
1015                 panic("Failed to allocate memory for sysinfo structure\n");
1016
1017         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1018                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1019         memblock_free((unsigned long) vmms, PAGE_SIZE);
1020 }
1021
1022 /*
1023  * Find the correct size for the task_struct. This depends on
1024  * the size of the struct fpu at the end of the thread_struct
1025  * which is embedded in the task_struct.
1026  */
1027 static void __init setup_task_size(void)
1028 {
1029         int task_size = sizeof(struct task_struct);
1030
1031         if (!MACHINE_HAS_VX) {
1032                 task_size -= sizeof(__vector128) * __NUM_VXRS;
1033                 task_size += sizeof(freg_t) * __NUM_FPRS;
1034         }
1035         arch_task_struct_size = task_size;
1036 }
1037
1038 /*
1039  * Issue diagnose 318 to set the control program name and
1040  * version codes.
1041  */
1042 static void __init setup_control_program_code(void)
1043 {
1044         union diag318_info diag318_info = {
1045                 .cpnc = CPNC_LINUX,
1046                 .cpvc_linux = 0,
1047                 .cpvc_distro = {0},
1048         };
1049
1050         if (!sclp.has_diag318)
1051                 return;
1052
1053         diag_stat_inc(DIAG_STAT_X318);
1054         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1055 }
1056
1057 /*
1058  * Print the component list from the IPL report
1059  */
1060 static void __init log_component_list(void)
1061 {
1062         struct ipl_rb_component_entry *ptr, *end;
1063         char *str;
1064
1065         if (!early_ipl_comp_list_addr)
1066                 return;
1067         if (ipl_block.hdr.flags & IPL_PL_FLAG_IPLSR)
1068                 pr_info("Linux is running with Secure-IPL enabled\n");
1069         else
1070                 pr_info("Linux is running with Secure-IPL disabled\n");
1071         ptr = (void *) early_ipl_comp_list_addr;
1072         end = (void *) ptr + early_ipl_comp_list_size;
1073         pr_info("The IPL report contains the following components:\n");
1074         while (ptr < end) {
1075                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1076                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1077                                 str = "signed, verified";
1078                         else
1079                                 str = "signed, verification failed";
1080                 } else {
1081                         str = "not signed";
1082                 }
1083                 pr_info("%016llx - %016llx (%s)\n",
1084                         ptr->addr, ptr->addr + ptr->len, str);
1085                 ptr++;
1086         }
1087 }
1088
1089 /*
1090  * Setup function called from init/main.c just after the banner
1091  * was printed.
1092  */
1093
1094 void __init setup_arch(char **cmdline_p)
1095 {
1096         /*
1097          * print what head.S has found out about the machine
1098          */
1099         if (MACHINE_IS_VM)
1100                 pr_info("Linux is running as a z/VM "
1101                         "guest operating system in 64-bit mode\n");
1102         else if (MACHINE_IS_KVM)
1103                 pr_info("Linux is running under KVM in 64-bit mode\n");
1104         else if (MACHINE_IS_LPAR)
1105                 pr_info("Linux is running natively in 64-bit mode\n");
1106         else
1107                 pr_info("Linux is running as a guest in 64-bit mode\n");
1108
1109         log_component_list();
1110
1111         /* Have one command line that is parsed and saved in /proc/cmdline */
1112         /* boot_command_line has been already set up in early.c */
1113         *cmdline_p = boot_command_line;
1114
1115         ROOT_DEV = Root_RAM0;
1116
1117         /* Is init_mm really needed? */
1118         init_mm.start_code = PAGE_OFFSET;
1119         init_mm.end_code = (unsigned long) _etext;
1120         init_mm.end_data = (unsigned long) _edata;
1121         init_mm.brk = (unsigned long) _end;
1122
1123         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1124                 nospec_auto_detect();
1125
1126         parse_early_param();
1127 #ifdef CONFIG_CRASH_DUMP
1128         /* Deactivate elfcorehdr= kernel parameter */
1129         elfcorehdr_addr = ELFCORE_ADDR_MAX;
1130 #endif
1131
1132         os_info_init();
1133         setup_ipl();
1134         setup_task_size();
1135         setup_control_program_code();
1136
1137         /* Do some memory reservations *before* memory is added to memblock */
1138         reserve_memory_end();
1139         reserve_oldmem();
1140         reserve_kernel();
1141         reserve_initrd();
1142         reserve_certificate_list();
1143         reserve_mem_detect_info();
1144         memblock_allow_resize();
1145
1146         /* Get information about *all* installed memory */
1147         memblock_add_mem_detect_info();
1148
1149         free_mem_detect_info();
1150         remove_oldmem();
1151
1152         /*
1153          * Make sure all chunks are MAX_ORDER aligned so we don't need the
1154          * extra checks that HOLES_IN_ZONE would require.
1155          *
1156          * Is this still required?
1157          */
1158         memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
1159
1160         setup_memory_end();
1161         setup_memory();
1162         dma_contiguous_reserve(memory_end);
1163         vmcp_cma_reserve();
1164
1165         check_initrd();
1166         reserve_crashkernel();
1167 #ifdef CONFIG_CRASH_DUMP
1168         /*
1169          * Be aware that smp_save_dump_cpus() triggers a system reset.
1170          * Therefore CPU and device initialization should be done afterwards.
1171          */
1172         smp_save_dump_cpus();
1173 #endif
1174
1175         setup_resources();
1176         setup_lowcore_dat_off();
1177         smp_fill_possible_mask();
1178         cpu_detect_mhz_feature();
1179         cpu_init();
1180         numa_setup();
1181         smp_detect_cpus();
1182         topology_init_early();
1183
1184         /*
1185          * Create kernel page tables and switch to virtual addressing.
1186          */
1187         paging_init();
1188
1189         /*
1190          * After paging_init created the kernel page table, the new PSWs
1191          * in lowcore can now run with DAT enabled.
1192          */
1193         setup_lowcore_dat_on();
1194
1195         /* Setup default console */
1196         conmode_default();
1197         set_preferred_console();
1198
1199         apply_alternative_instructions();
1200         if (IS_ENABLED(CONFIG_EXPOLINE))
1201                 nospec_init_branches();
1202
1203         /* Setup zfcpdump support */
1204         setup_zfcpdump();
1205
1206         /* Add system specific data to the random pool */
1207         setup_randomness();
1208 }