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 #ifdef CONFIG_SMP
 465         lc->spinlock_lockval = arch_spin_lockval(0);
 466         lc->spinlock_index = 0;
 467         arch_spin_lock_setup(0);
 468 #endif
 469         lc->br_r1_trampoline = 0x07f1;  /* br %r1 */
 470
 471         set_prefix((u32)(unsigned long) lc);
 472         lowcore_ptr[0] = lc;
 473 }
 474
 475 static void __init setup_lowcore_dat_on(void)
 476 {
 477         __ctl_clear_bit(0, 28);
 478         S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
 479         S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
 480         S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
 481         S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
 482         __ctl_set_bit(0, 28);
 483 }
 484
 485 static struct resource code_resource = {
 486         .name  = "Kernel code",
 487         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 488 };
 489
 490 static struct resource data_resource = {
 491         .name = "Kernel data",
 492         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 493 };
 494
 495 static struct resource bss_resource = {
 496         .name = "Kernel bss",
 497         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 498 };
 499
 500 static struct resource __initdata *standard_resources[] = {
 501         &code_resource,
 502         &data_resource,
 503         &bss_resource,
 504 };
 505
 506 static void __init setup_resources(void)
 507 {
 508         struct resource *res, *std_res, *sub_res;
 509         struct memblock_region *reg;
 510         int j;
 511
 512         code_resource.start = (unsigned long) _text;
 513         code_resource.end = (unsigned long) _etext - 1;
 514         data_resource.start = (unsigned long) _etext;
 515         data_resource.end = (unsigned long) _edata - 1;
 516         bss_resource.start = (unsigned long) __bss_start;
 517         bss_resource.end = (unsigned long) __bss_stop - 1;
 518
 519         for_each_memblock(memory, reg) {
 520                 res = memblock_alloc(sizeof(*res), 8);
 521                 if (!res)
 522                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 523                               __func__, sizeof(*res), 8);
 524                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 525
 526                 res->name = "System RAM";
 527                 res->start = reg->base;
 528                 res->end = reg->base + reg->size - 1;
 529                 request_resource(&iomem_resource, res);
 530
 531                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 532                         std_res = standard_resources[j];
 533                         if (std_res->start < res->start ||
 534                             std_res->start > res->end)
 535                                 continue;
 536                         if (std_res->end > res->end) {
 537                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
 538                                 if (!sub_res)
 539                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 540                                               __func__, sizeof(*sub_res), 8);
 541                                 *sub_res = *std_res;
 542                                 sub_res->end = res->end;
 543                                 std_res->start = res->end + 1;
 544                                 request_resource(res, sub_res);
 545                         } else {
 546                                 request_resource(res, std_res);
 547                         }
 548                 }
 549         }
 550 #ifdef CONFIG_CRASH_DUMP
 551         /*
 552          * Re-add removed crash kernel memory as reserved memory. This makes
 553          * sure it will be mapped with the identity mapping and struct pages
 554          * will be created, so it can be resized later on.
 555          * However add it later since the crash kernel resource should not be
 556          * part of the System RAM resource.
 557          */
 558         if (crashk_res.end) {
 559                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 560                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 561                 insert_resource(&iomem_resource, &crashk_res);
 562         }
 563 #endif
 564 }
 565
 566 static void __init setup_memory_end(void)
 567 {
 568         unsigned long vmax, vmalloc_size, tmp;
 569
 570         /* Choose kernel address space layout: 3 or 4 levels. */
 571         vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
 572         if (IS_ENABLED(CONFIG_KASAN)) {
 573                 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
 574                            ? _REGION1_SIZE
 575                            : _REGION2_SIZE;
 576         } else {
 577                 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
 578                 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
 579                 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
 580                         vmax = _REGION2_SIZE; /* 3-level kernel page table */
 581                 else
 582                         vmax = _REGION1_SIZE; /* 4-level kernel page table */
 583         }
 584
 585         /* module area is at the end of the kernel address space. */
 586         MODULES_END = vmax;
 587         MODULES_VADDR = MODULES_END - MODULES_LEN;
 588         VMALLOC_END = MODULES_VADDR;
 589         VMALLOC_START = VMALLOC_END - vmalloc_size;
 590
 591         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 592         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 593         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 594         tmp = SECTION_ALIGN_UP(tmp);
 595         tmp = VMALLOC_START - tmp * sizeof(struct page);
 596         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 597         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 598         vmemmap = (struct page *) tmp;
 599
 600         /* Take care that memory_end is set and <= vmemmap */
 601         memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
 602 #ifdef CONFIG_KASAN
 603         /* fit in kasan shadow memory region between 1:1 and vmemmap */
 604         memory_end = min(memory_end, KASAN_SHADOW_START);
 605         vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
 606 #endif
 607         max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 608         memblock_remove(memory_end, ULONG_MAX);
 609
 610         pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
 611 }
 612
 613 #ifdef CONFIG_CRASH_DUMP
 614
 615 /*
 616  * When kdump is enabled, we have to ensure that no memory from
 617  * the area [0 - crashkernel memory size] and
 618  * [crashk_res.start - crashk_res.end] is set offline.
 619  */
 620 static int kdump_mem_notifier(struct notifier_block *nb,
 621                               unsigned long action, void *data)
 622 {
 623         struct memory_notify *arg = data;
 624
 625         if (action != MEM_GOING_OFFLINE)
 626                 return NOTIFY_OK;
 627         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 628                 return NOTIFY_BAD;
 629         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 630                 return NOTIFY_OK;
 631         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 632                 return NOTIFY_OK;
 633         return NOTIFY_BAD;
 634 }
 635
 636 static struct notifier_block kdump_mem_nb = {
 637         .notifier_call = kdump_mem_notifier,
 638 };
 639
 640 #endif
 641
 642 /*
 643  * Make sure that the area behind memory_end is protected
 644  */
 645 static void reserve_memory_end(void)
 646 {
 647         if (memory_end_set)
 648                 memblock_reserve(memory_end, ULONG_MAX);
 649 }
 650
 651 /*
 652  * Make sure that oldmem, where the dump is stored, is protected
 653  */
 654 static void reserve_oldmem(void)
 655 {
 656 #ifdef CONFIG_CRASH_DUMP
 657         if (OLDMEM_BASE)
 658                 /* Forget all memory above the running kdump system */
 659                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 660 #endif
 661 }
 662
 663 /*
 664  * Make sure that oldmem, where the dump is stored, is protected
 665  */
 666 static void remove_oldmem(void)
 667 {
 668 #ifdef CONFIG_CRASH_DUMP
 669         if (OLDMEM_BASE)
 670                 /* Forget all memory above the running kdump system */
 671                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 672 #endif
 673 }
 674
 675 /*
 676  * Reserve memory for kdump kernel to be loaded with kexec
 677  */
 678 static void __init reserve_crashkernel(void)
 679 {
 680 #ifdef CONFIG_CRASH_DUMP
 681         unsigned long long crash_base, crash_size;
 682         phys_addr_t low, high;
 683         int rc;
 684
 685         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 686                                &crash_base);
 687
 688         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 689         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 690         if (rc || crash_size == 0)
 691                 return;
 692
 693         if (memblock.memory.regions[0].size < crash_size) {
 694                 pr_info("crashkernel reservation failed: %s\n",
 695                         "first memory chunk must be at least crashkernel size");
 696                 return;
 697         }
 698
 699         low = crash_base ?: OLDMEM_BASE;
 700         high = low + crash_size;
 701         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 702                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 703                 crash_base = low;
 704         } else {
 705                 /* Find suitable area in free memory */
 706                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 707                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 708
 709                 if (crash_base && crash_base < low) {
 710                         pr_info("crashkernel reservation failed: %s\n",
 711                                 "crash_base too low");
 712                         return;
 713                 }
 714                 low = crash_base ?: low;
 715                 crash_base = memblock_find_in_range(low, high, crash_size,
 716                                                     KEXEC_CRASH_MEM_ALIGN);
 717         }
 718
 719         if (!crash_base) {
 720                 pr_info("crashkernel reservation failed: %s\n",
 721                         "no suitable area found");
 722                 return;
 723         }
 724
 725         if (register_memory_notifier(&kdump_mem_nb))
 726                 return;
 727
 728         if (!OLDMEM_BASE && MACHINE_IS_VM)
 729                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 730         crashk_res.start = crash_base;
 731         crashk_res.end = crash_base + crash_size - 1;
 732         memblock_remove(crash_base, crash_size);
 733         pr_info("Reserving %lluMB of memory at %lluMB "
 734                 "for crashkernel (System RAM: %luMB)\n",
 735                 crash_size >> 20, crash_base >> 20,
 736                 (unsigned long)memblock.memory.total_size >> 20);
 737         os_info_crashkernel_add(crash_base, crash_size);
 738 #endif
 739 }
 740
 741 /*
 742  * Reserve the initrd from being used by memblock
 743  */
 744 static void __init reserve_initrd(void)
 745 {
 746 #ifdef CONFIG_BLK_DEV_INITRD
 747         if (!INITRD_START || !INITRD_SIZE)
 748                 return;
 749         initrd_start = INITRD_START;
 750         initrd_end = initrd_start + INITRD_SIZE;
 751         memblock_reserve(INITRD_START, INITRD_SIZE);
 752 #endif
 753 }
 754
 755 /*
 756  * Reserve the memory area used to pass the certificate lists
 757  */
 758 static void __init reserve_certificate_list(void)
 759 {
 760         if (ipl_cert_list_addr)
 761                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
 762 }
 763
 764 static void __init reserve_mem_detect_info(void)
 765 {
 766         unsigned long start, size;
 767
 768         get_mem_detect_reserved(&start, &size);
 769         if (size)
 770                 memblock_reserve(start, size);
 771 }
 772
 773 static void __init free_mem_detect_info(void)
 774 {
 775         unsigned long start, size;
 776
 777         get_mem_detect_reserved(&start, &size);
 778         if (size)
 779                 memblock_free(start, size);
 780 }
 781
 782 static void __init memblock_physmem_add(phys_addr_t start, phys_addr_t size)
 783 {
 784         memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n",
 785                      start, start + size - 1);
 786         memblock_add_range(&memblock.memory, start, size, 0, 0);
 787         memblock_add_range(&memblock.physmem, start, size, 0, 0);
 788 }
 789
 790 static const char * __init get_mem_info_source(void)
 791 {
 792         switch (mem_detect.info_source) {
 793         case MEM_DETECT_SCLP_STOR_INFO:
 794                 return "sclp storage info";
 795         case MEM_DETECT_DIAG260:
 796                 return "diag260";
 797         case MEM_DETECT_SCLP_READ_INFO:
 798                 return "sclp read info";
 799         case MEM_DETECT_BIN_SEARCH:
 800                 return "binary search";
 801         }
 802         return "none";
 803 }
 804
 805 static void __init memblock_add_mem_detect_info(void)
 806 {
 807         unsigned long start, end;
 808         int i;
 809
 810         memblock_dbg("physmem info source: %s (%hhd)\n",
 811                      get_mem_info_source(), mem_detect.info_source);
 812         /* keep memblock lists close to the kernel */
 813         memblock_set_bottom_up(true);
 814         for_each_mem_detect_block(i, &start, &end)
 815                 memblock_physmem_add(start, end - start);
 816         memblock_set_bottom_up(false);
 817         memblock_dump_all();
 818 }
 819
 820 /*
 821  * Check for initrd being in usable memory
 822  */
 823 static void __init check_initrd(void)
 824 {
 825 #ifdef CONFIG_BLK_DEV_INITRD
 826         if (INITRD_START && INITRD_SIZE &&
 827             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 828                 pr_err("The initial RAM disk does not fit into the memory\n");
 829                 memblock_free(INITRD_START, INITRD_SIZE);
 830                 initrd_start = initrd_end = 0;
 831         }
 832 #endif
 833 }
 834
 835 /*
 836  * Reserve memory used for lowcore/command line/kernel image.
 837  */
 838 static void __init reserve_kernel(void)
 839 {
 840         unsigned long start_pfn = PFN_UP(__pa(_end));
 841
 842         memblock_reserve(0, HEAD_END);
 843         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 844                          - (unsigned long)_stext);
 845         memblock_reserve(__sdma, __edma - __sdma);
 846 }
 847
 848 static void __init setup_memory(void)
 849 {
 850         struct memblock_region *reg;
 851
 852         /*
 853          * Init storage key for present memory
 854          */
 855         for_each_memblock(memory, reg) {
 856                 storage_key_init_range(reg->base, reg->base + reg->size);
 857         }
 858         psw_set_key(PAGE_DEFAULT_KEY);
 859
 860         /* Only cosmetics */
 861         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 862 }
 863
 864 /*
 865  * Setup hardware capabilities.
 866  */
 867 static int __init setup_hwcaps(void)
 868 {
 869         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 870         struct cpuid cpu_id;
 871         int i;
 872
 873         /*
 874          * The store facility list bits numbers as found in the principles
 875          * of operation are numbered with bit 1UL<<31 as number 0 to
 876          * bit 1UL<<0 as number 31.
 877          *   Bit 0: instructions named N3, "backported" to esa-mode
 878          *   Bit 2: z/Architecture mode is active
 879          *   Bit 7: the store-facility-list-extended facility is installed
 880          *   Bit 17: the message-security assist is installed
 881          *   Bit 19: the long-displacement facility is installed
 882          *   Bit 21: the extended-immediate facility is installed
 883          *   Bit 22: extended-translation facility 3 is installed
 884          *   Bit 30: extended-translation facility 3 enhancement facility
 885          * These get translated to:
 886          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 887          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 888          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 889          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 890          */
 891         for (i = 0; i < 6; i++)
 892                 if (test_facility(stfl_bits[i]))
 893                         elf_hwcap |= 1UL << i;
 894
 895         if (test_facility(22) && test_facility(30))
 896                 elf_hwcap |= HWCAP_S390_ETF3EH;
 897
 898         /*
 899          * Check for additional facilities with store-facility-list-extended.
 900          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 901          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 902          * as stored by stfl, bits 32-xxx contain additional facilities.
 903          * How many facility words are stored depends on the number of
 904          * doublewords passed to the instruction. The additional facilities
 905          * are:
 906          *   Bit 42: decimal floating point facility is installed
 907          *   Bit 44: perform floating point operation facility is installed
 908          * translated to:
 909          *   HWCAP_S390_DFP bit 6 (42 && 44).
 910          */
 911         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 912                 elf_hwcap |= HWCAP_S390_DFP;
 913
 914         /*
 915          * Huge page support HWCAP_S390_HPAGE is bit 7.
 916          */
 917         if (MACHINE_HAS_EDAT1)
 918                 elf_hwcap |= HWCAP_S390_HPAGE;
 919
 920         /*
 921          * 64-bit register support for 31-bit processes
 922          * HWCAP_S390_HIGH_GPRS is bit 9.
 923          */
 924         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 925
 926         /*
 927          * Transactional execution support HWCAP_S390_TE is bit 10.
 928          */
 929         if (MACHINE_HAS_TE)
 930                 elf_hwcap |= HWCAP_S390_TE;
 931
 932         /*
 933          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 934          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 935          * instead of facility bit 129.
 936          */
 937         if (MACHINE_HAS_VX) {
 938                 elf_hwcap |= HWCAP_S390_VXRS;
 939                 if (test_facility(134))
 940                         elf_hwcap |= HWCAP_S390_VXRS_EXT;
 941                 if (test_facility(135))
 942                         elf_hwcap |= HWCAP_S390_VXRS_BCD;
 943                 if (test_facility(148))
 944                         elf_hwcap |= HWCAP_S390_VXRS_EXT2;
 945                 if (test_facility(152))
 946                         elf_hwcap |= HWCAP_S390_VXRS_PDE;
 947         }
 948         if (test_facility(150))
 949                 elf_hwcap |= HWCAP_S390_SORT;
 950         if (test_facility(151))
 951                 elf_hwcap |= HWCAP_S390_DFLT;
 952
 953         /*
 954          * Guarded storage support HWCAP_S390_GS is bit 12.
 955          */
 956         if (MACHINE_HAS_GS)
 957                 elf_hwcap |= HWCAP_S390_GS;
 958
 959         get_cpu_id(&cpu_id);
 960         add_device_randomness(&cpu_id, sizeof(cpu_id));
 961         switch (cpu_id.machine) {
 962         case 0x2064:
 963         case 0x2066:
 964         default:        /* Use "z900" as default for 64 bit kernels. */
 965                 strcpy(elf_platform, "z900");
 966                 break;
 967         case 0x2084:
 968         case 0x2086:
 969                 strcpy(elf_platform, "z990");
 970                 break;
 971         case 0x2094:
 972         case 0x2096:
 973                 strcpy(elf_platform, "z9-109");
 974                 break;
 975         case 0x2097:
 976         case 0x2098:
 977                 strcpy(elf_platform, "z10");
 978                 break;
 979         case 0x2817:
 980         case 0x2818:
 981                 strcpy(elf_platform, "z196");
 982                 break;
 983         case 0x2827:
 984         case 0x2828:
 985                 strcpy(elf_platform, "zEC12");
 986                 break;
 987         case 0x2964:
 988         case 0x2965:
 989                 strcpy(elf_platform, "z13");
 990                 break;
 991         case 0x3906:
 992         case 0x3907:
 993                 strcpy(elf_platform, "z14");
 994                 break;
 995         }
 996
 997         /*
 998          * Virtualization support HWCAP_INT_SIE is bit 0.
 999          */
1000         if (sclp.has_sief2)
1001                 int_hwcap |= HWCAP_INT_SIE;
1002
1003         return 0;
1004 }
1005 arch_initcall(setup_hwcaps);
1006
1007 /*
1008  * Add system information as device randomness
1009  */
1010 static void __init setup_randomness(void)
1011 {
1012         struct sysinfo_3_2_2 *vmms;
1013
1014         vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
1015                                                             PAGE_SIZE);
1016         if (!vmms)
1017                 panic("Failed to allocate memory for sysinfo structure\n");
1018
1019         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1020                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1021         memblock_free((unsigned long) vmms, PAGE_SIZE);
1022 }
1023
1024 /*
1025  * Find the correct size for the task_struct. This depends on
1026  * the size of the struct fpu at the end of the thread_struct
1027  * which is embedded in the task_struct.
1028  */
1029 static void __init setup_task_size(void)
1030 {
1031         int task_size = sizeof(struct task_struct);
1032
1033         if (!MACHINE_HAS_VX) {
1034                 task_size -= sizeof(__vector128) * __NUM_VXRS;
1035                 task_size += sizeof(freg_t) * __NUM_FPRS;
1036         }
1037         arch_task_struct_size = task_size;
1038 }
1039
1040 /*
1041  * Issue diagnose 318 to set the control program name and
1042  * version codes.
1043  */
1044 static void __init setup_control_program_code(void)
1045 {
1046         union diag318_info diag318_info = {
1047                 .cpnc = CPNC_LINUX,
1048                 .cpvc_linux = 0,
1049                 .cpvc_distro = {0},
1050         };
1051
1052         if (!sclp.has_diag318)
1053                 return;
1054
1055         diag_stat_inc(DIAG_STAT_X318);
1056         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1057 }
1058
1059 /*
1060  * Print the component list from the IPL report
1061  */
1062 static void __init log_component_list(void)
1063 {
1064         struct ipl_rb_component_entry *ptr, *end;
1065         char *str;
1066
1067         if (!early_ipl_comp_list_addr)
1068                 return;
1069         if (ipl_block.hdr.flags & IPL_PL_FLAG_IPLSR)
1070                 pr_info("Linux is running with Secure-IPL enabled\n");
1071         else
1072                 pr_info("Linux is running with Secure-IPL disabled\n");
1073         ptr = (void *) early_ipl_comp_list_addr;
1074         end = (void *) ptr + early_ipl_comp_list_size;
1075         pr_info("The IPL report contains the following components:\n");
1076         while (ptr < end) {
1077                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1078                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1079                                 str = "signed, verified";
1080                         else
1081                                 str = "signed, verification failed";
1082                 } else {
1083                         str = "not signed";
1084                 }
1085                 pr_info("%016llx - %016llx (%s)\n",
1086                         ptr->addr, ptr->addr + ptr->len, str);
1087                 ptr++;
1088         }
1089 }
1090
1091 /*
1092  * Setup function called from init/main.c just after the banner
1093  * was printed.
1094  */
1095
1096 void __init setup_arch(char **cmdline_p)
1097 {
1098         /*
1099          * print what head.S has found out about the machine
1100          */
1101         if (MACHINE_IS_VM)
1102                 pr_info("Linux is running as a z/VM "
1103                         "guest operating system in 64-bit mode\n");
1104         else if (MACHINE_IS_KVM)
1105                 pr_info("Linux is running under KVM in 64-bit mode\n");
1106         else if (MACHINE_IS_LPAR)
1107                 pr_info("Linux is running natively in 64-bit mode\n");
1108         else
1109                 pr_info("Linux is running as a guest in 64-bit mode\n");
1110
1111         log_component_list();
1112
1113         /* Have one command line that is parsed and saved in /proc/cmdline */
1114         /* boot_command_line has been already set up in early.c */
1115         *cmdline_p = boot_command_line;
1116
1117         ROOT_DEV = Root_RAM0;
1118
1119         /* Is init_mm really needed? */
1120         init_mm.start_code = PAGE_OFFSET;
1121         init_mm.end_code = (unsigned long) _etext;
1122         init_mm.end_data = (unsigned long) _edata;
1123         init_mm.brk = (unsigned long) _end;
1124
1125         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1126                 nospec_auto_detect();
1127
1128         parse_early_param();
1129 #ifdef CONFIG_CRASH_DUMP
1130         /* Deactivate elfcorehdr= kernel parameter */
1131         elfcorehdr_addr = ELFCORE_ADDR_MAX;
1132 #endif
1133
1134         os_info_init();
1135         setup_ipl();
1136         setup_task_size();
1137         setup_control_program_code();
1138
1139         /* Do some memory reservations *before* memory is added to memblock */
1140         reserve_memory_end();
1141         reserve_oldmem();
1142         reserve_kernel();
1143         reserve_initrd();
1144         reserve_certificate_list();
1145         reserve_mem_detect_info();
1146         memblock_allow_resize();
1147
1148         /* Get information about *all* installed memory */
1149         memblock_add_mem_detect_info();
1150
1151         free_mem_detect_info();
1152         remove_oldmem();
1153
1154         /*
1155          * Make sure all chunks are MAX_ORDER aligned so we don't need the
1156          * extra checks that HOLES_IN_ZONE would require.
1157          *
1158          * Is this still required?
1159          */
1160         memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
1161
1162         setup_memory_end();
1163         setup_memory();
1164         dma_contiguous_reserve(memory_end);
1165         vmcp_cma_reserve();
1166
1167         check_initrd();
1168         reserve_crashkernel();
1169 #ifdef CONFIG_CRASH_DUMP
1170         /*
1171          * Be aware that smp_save_dump_cpus() triggers a system reset.
1172          * Therefore CPU and device initialization should be done afterwards.
1173          */
1174         smp_save_dump_cpus();
1175 #endif
1176
1177         setup_resources();
1178         setup_lowcore_dat_off();
1179         smp_fill_possible_mask();
1180         cpu_detect_mhz_feature();
1181         cpu_init();
1182         numa_setup();
1183         smp_detect_cpus();
1184         topology_init_early();
1185
1186         /*
1187          * Create kernel page tables and switch to virtual addressing.
1188          */
1189         paging_init();
1190
1191         /*
1192          * After paging_init created the kernel page table, the new PSWs
1193          * in lowcore can now run with DAT enabled.
1194          */
1195         setup_lowcore_dat_on();
1196
1197         /* Setup default console */
1198         conmode_default();
1199         set_preferred_console();
1200
1201         apply_alternative_instructions();
1202         if (IS_ENABLED(CONFIG_EXPOLINE))
1203                 nospec_init_branches();
1204
1205         /* Setup zfcpdump support */
1206         setup_zfcpdump();
1207
1208         /* Add system specific data to the random pool */
1209         setup_randomness();
1210 }