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 unsigned int __bootdata_preserved(zlib_dfltcc_support);
 115 EXPORT_SYMBOL(zlib_dfltcc_support);
 116
 117 unsigned long VMALLOC_START;
 118 EXPORT_SYMBOL(VMALLOC_START);
 119
 120 unsigned long VMALLOC_END;
 121 EXPORT_SYMBOL(VMALLOC_END);
 122
 123 struct page *vmemmap;
 124 EXPORT_SYMBOL(vmemmap);
 125
 126 unsigned long MODULES_VADDR;
 127 unsigned long MODULES_END;
 128
 129 /* An array with a pointer to the lowcore of every CPU. */
 130 struct lowcore *lowcore_ptr[NR_CPUS];
 131 EXPORT_SYMBOL(lowcore_ptr);
 132
 133 /*
 134  * This is set up by the setup-routine at boot-time
 135  * for S390 need to find out, what we have to setup
 136  * using address 0x10400 ...
 137  */
 138
 139 #include <asm/setup.h>
 140
 141 /*
 142  * condev= and conmode= setup parameter.
 143  */
 144
 145 static int __init condev_setup(char *str)
 146 {
 147         int vdev;
 148
 149         vdev = simple_strtoul(str, &str, 0);
 150         if (vdev >= 0 && vdev < 65536) {
 151                 console_devno = vdev;
 152                 console_irq = -1;
 153         }
 154         return 1;
 155 }
 156
 157 __setup("condev=", condev_setup);
 158
 159 static void __init set_preferred_console(void)
 160 {
 161         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 162                 add_preferred_console("ttyS", 0, NULL);
 163         else if (CONSOLE_IS_3270)
 164                 add_preferred_console("tty3270", 0, NULL);
 165         else if (CONSOLE_IS_VT220)
 166                 add_preferred_console("ttyS", 1, NULL);
 167         else if (CONSOLE_IS_HVC)
 168                 add_preferred_console("hvc", 0, NULL);
 169 }
 170
 171 static int __init conmode_setup(char *str)
 172 {
 173 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 174         if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
 175                 SET_CONSOLE_SCLP;
 176 #endif
 177 #if defined(CONFIG_TN3215_CONSOLE)
 178         if (!strcmp(str, "3215"))
 179                 SET_CONSOLE_3215;
 180 #endif
 181 #if defined(CONFIG_TN3270_CONSOLE)
 182         if (!strcmp(str, "3270"))
 183                 SET_CONSOLE_3270;
 184 #endif
 185         set_preferred_console();
 186         return 1;
 187 }
 188
 189 __setup("conmode=", conmode_setup);
 190
 191 static void __init conmode_default(void)
 192 {
 193         char query_buffer[1024];
 194         char *ptr;
 195
 196         if (MACHINE_IS_VM) {
 197                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 198                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 199                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 200                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 201                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 202                 ptr = strstr(query_buffer, "CONMODE");
 203                 /*
 204                  * Set the conmode to 3215 so that the device recognition
 205                  * will set the cu_type of the console to 3215. If the
 206                  * conmode is 3270 and we don't set it back then both
 207                  * 3215 and the 3270 driver will try to access the console
 208                  * device (3215 as console and 3270 as normal tty).
 209                  */
 210                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 211                 if (ptr == NULL) {
 212 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 213                         SET_CONSOLE_SCLP;
 214 #endif
 215                         return;
 216                 }
 217                 if (str_has_prefix(ptr + 8, "3270")) {
 218 #if defined(CONFIG_TN3270_CONSOLE)
 219                         SET_CONSOLE_3270;
 220 #elif defined(CONFIG_TN3215_CONSOLE)
 221                         SET_CONSOLE_3215;
 222 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 223                         SET_CONSOLE_SCLP;
 224 #endif
 225                 } else if (str_has_prefix(ptr + 8, "3215")) {
 226 #if defined(CONFIG_TN3215_CONSOLE)
 227                         SET_CONSOLE_3215;
 228 #elif defined(CONFIG_TN3270_CONSOLE)
 229                         SET_CONSOLE_3270;
 230 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 231                         SET_CONSOLE_SCLP;
 232 #endif
 233                 }
 234         } else if (MACHINE_IS_KVM) {
 235                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
 236                         SET_CONSOLE_VT220;
 237                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
 238                         SET_CONSOLE_SCLP;
 239                 else
 240                         SET_CONSOLE_HVC;
 241         } else {
 242 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 243                 SET_CONSOLE_SCLP;
 244 #endif
 245         }
 246 }
 247
 248 #ifdef CONFIG_CRASH_DUMP
 249 static void __init setup_zfcpdump(void)
 250 {
 251         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 252                 return;
 253         if (OLDMEM_BASE)
 254                 return;
 255         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 256         console_loglevel = 2;
 257 }
 258 #else
 259 static inline void setup_zfcpdump(void) {}
 260 #endif /* CONFIG_CRASH_DUMP */
 261
 262  /*
 263  * Reboot, halt and power_off stubs. They just call _machine_restart,
 264  * _machine_halt or _machine_power_off.
 265  */
 266
 267 void machine_restart(char *command)
 268 {
 269         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 270                 /*
 271                  * Only unblank the console if we are called in enabled
 272                  * context or a bust_spinlocks cleared the way for us.
 273                  */
 274                 console_unblank();
 275         _machine_restart(command);
 276 }
 277
 278 void machine_halt(void)
 279 {
 280         if (!in_interrupt() || oops_in_progress)
 281                 /*
 282                  * Only unblank the console if we are called in enabled
 283                  * context or a bust_spinlocks cleared the way for us.
 284                  */
 285                 console_unblank();
 286         _machine_halt();
 287 }
 288
 289 void machine_power_off(void)
 290 {
 291         if (!in_interrupt() || oops_in_progress)
 292                 /*
 293                  * Only unblank the console if we are called in enabled
 294                  * context or a bust_spinlocks cleared the way for us.
 295                  */
 296                 console_unblank();
 297         _machine_power_off();
 298 }
 299
 300 /*
 301  * Dummy power off function.
 302  */
 303 void (*pm_power_off)(void) = machine_power_off;
 304 EXPORT_SYMBOL_GPL(pm_power_off);
 305
 306 void *restart_stack __section(.data);
 307
 308 unsigned long stack_alloc(void)
 309 {
 310 #ifdef CONFIG_VMAP_STACK
 311         return (unsigned long)
 312                 __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
 313                                      VMALLOC_START, VMALLOC_END,
 314                                      THREADINFO_GFP,
 315                                      PAGE_KERNEL, 0, NUMA_NO_NODE,
 316                                      __builtin_return_address(0));
 317 #else
 318         return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 319 #endif
 320 }
 321
 322 void stack_free(unsigned long stack)
 323 {
 324 #ifdef CONFIG_VMAP_STACK
 325         vfree((void *) stack);
 326 #else
 327         free_pages(stack, THREAD_SIZE_ORDER);
 328 #endif
 329 }
 330
 331 int __init arch_early_irq_init(void)
 332 {
 333         unsigned long stack;
 334
 335         stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 336         if (!stack)
 337                 panic("Couldn't allocate async stack");
 338         S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
 339         return 0;
 340 }
 341
 342 static int __init async_stack_realloc(void)
 343 {
 344         unsigned long old, new;
 345
 346         old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
 347         new = stack_alloc();
 348         if (!new)
 349                 panic("Couldn't allocate async stack");
 350         S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
 351         free_pages(old, THREAD_SIZE_ORDER);
 352         return 0;
 353 }
 354 early_initcall(async_stack_realloc);
 355
 356 void __init arch_call_rest_init(void)
 357 {
 358         unsigned long stack;
 359
 360         stack = stack_alloc();
 361         if (!stack)
 362                 panic("Couldn't allocate kernel stack");
 363         current->stack = (void *) stack;
 364 #ifdef CONFIG_VMAP_STACK
 365         current->stack_vm_area = (void *) stack;
 366 #endif
 367         set_task_stack_end_magic(current);
 368         stack += STACK_INIT_OFFSET;
 369         S390_lowcore.kernel_stack = stack;
 370         CALL_ON_STACK_NORETURN(rest_init, stack);
 371 }
 372
 373 static void __init setup_lowcore_dat_off(void)
 374 {
 375         struct lowcore *lc;
 376
 377         /*
 378          * Setup lowcore for boot cpu
 379          */
 380         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 381         lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
 382         if (!lc)
 383                 panic("%s: Failed to allocate %zu bytes align=%zx\n",
 384                       __func__, sizeof(*lc), sizeof(*lc));
 385
 386         lc->restart_psw.mask = PSW_KERNEL_BITS;
 387         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 388         lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 389         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 390         lc->svc_new_psw.mask = PSW_KERNEL_BITS |
 391                 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 392         lc->svc_new_psw.addr = (unsigned long) system_call;
 393         lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 394         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 395         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 396         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 397         lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 398         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 399         lc->clock_comparator = clock_comparator_max;
 400         lc->nodat_stack = ((unsigned long) &init_thread_union)
 401                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 402         lc->current_task = (unsigned long)&init_task;
 403         lc->lpp = LPP_MAGIC;
 404         lc->machine_flags = S390_lowcore.machine_flags;
 405         lc->preempt_count = S390_lowcore.preempt_count;
 406         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 407         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 408                sizeof(lc->stfle_fac_list));
 409         memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
 410                sizeof(lc->alt_stfle_fac_list));
 411         nmi_alloc_boot_cpu(lc);
 412         vdso_alloc_boot_cpu(lc);
 413         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 414         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 415         lc->exit_timer = S390_lowcore.exit_timer;
 416         lc->user_timer = S390_lowcore.user_timer;
 417         lc->system_timer = S390_lowcore.system_timer;
 418         lc->steal_timer = S390_lowcore.steal_timer;
 419         lc->last_update_timer = S390_lowcore.last_update_timer;
 420         lc->last_update_clock = S390_lowcore.last_update_clock;
 421
 422         /*
 423          * Allocate the global restart stack which is the same for
 424          * all CPUs in cast *one* of them does a PSW restart.
 425          */
 426         restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 427         if (!restart_stack)
 428                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 429                       __func__, THREAD_SIZE, THREAD_SIZE);
 430         restart_stack += STACK_INIT_OFFSET;
 431
 432         /*
 433          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 434          * restart data to the absolute zero lowcore. This is necessary if
 435          * PSW restart is done on an offline CPU that has lowcore zero.
 436          */
 437         lc->restart_stack = (unsigned long) restart_stack;
 438         lc->restart_fn = (unsigned long) do_restart;
 439         lc->restart_data = 0;
 440         lc->restart_source = -1UL;
 441
 442         /* Setup absolute zero lowcore */
 443         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 444         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 445         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 446         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 447         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 448
 449         lc->spinlock_lockval = arch_spin_lockval(0);
 450         lc->spinlock_index = 0;
 451         arch_spin_lock_setup(0);
 452         lc->br_r1_trampoline = 0x07f1;  /* br %r1 */
 453
 454         set_prefix((u32)(unsigned long) lc);
 455         lowcore_ptr[0] = lc;
 456 }
 457
 458 static void __init setup_lowcore_dat_on(void)
 459 {
 460         __ctl_clear_bit(0, 28);
 461         S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
 462         S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
 463         S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
 464         S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
 465         __ctl_set_bit(0, 28);
 466 }
 467
 468 static struct resource code_resource = {
 469         .name  = "Kernel code",
 470         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 471 };
 472
 473 static struct resource data_resource = {
 474         .name = "Kernel data",
 475         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 476 };
 477
 478 static struct resource bss_resource = {
 479         .name = "Kernel bss",
 480         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 481 };
 482
 483 static struct resource __initdata *standard_resources[] = {
 484         &code_resource,
 485         &data_resource,
 486         &bss_resource,
 487 };
 488
 489 static void __init setup_resources(void)
 490 {
 491         struct resource *res, *std_res, *sub_res;
 492         struct memblock_region *reg;
 493         int j;
 494
 495         code_resource.start = (unsigned long) _text;
 496         code_resource.end = (unsigned long) _etext - 1;
 497         data_resource.start = (unsigned long) _etext;
 498         data_resource.end = (unsigned long) _edata - 1;
 499         bss_resource.start = (unsigned long) __bss_start;
 500         bss_resource.end = (unsigned long) __bss_stop - 1;
 501
 502         for_each_memblock(memory, reg) {
 503                 res = memblock_alloc(sizeof(*res), 8);
 504                 if (!res)
 505                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 506                               __func__, sizeof(*res), 8);
 507                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 508
 509                 res->name = "System RAM";
 510                 res->start = reg->base;
 511                 res->end = reg->base + reg->size - 1;
 512                 request_resource(&iomem_resource, res);
 513
 514                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 515                         std_res = standard_resources[j];
 516                         if (std_res->start < res->start ||
 517                             std_res->start > res->end)
 518                                 continue;
 519                         if (std_res->end > res->end) {
 520                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
 521                                 if (!sub_res)
 522                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 523                                               __func__, sizeof(*sub_res), 8);
 524                                 *sub_res = *std_res;
 525                                 sub_res->end = res->end;
 526                                 std_res->start = res->end + 1;
 527                                 request_resource(res, sub_res);
 528                         } else {
 529                                 request_resource(res, std_res);
 530                         }
 531                 }
 532         }
 533 #ifdef CONFIG_CRASH_DUMP
 534         /*
 535          * Re-add removed crash kernel memory as reserved memory. This makes
 536          * sure it will be mapped with the identity mapping and struct pages
 537          * will be created, so it can be resized later on.
 538          * However add it later since the crash kernel resource should not be
 539          * part of the System RAM resource.
 540          */
 541         if (crashk_res.end) {
 542                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 543                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 544                 insert_resource(&iomem_resource, &crashk_res);
 545         }
 546 #endif
 547 }
 548
 549 static void __init setup_memory_end(void)
 550 {
 551         unsigned long vmax, tmp;
 552
 553         /* Choose kernel address space layout: 3 or 4 levels. */
 554         if (IS_ENABLED(CONFIG_KASAN)) {
 555                 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
 556                            ? _REGION1_SIZE
 557                            : _REGION2_SIZE;
 558         } else {
 559                 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
 560                 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
 561                 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
 562                         vmax = _REGION2_SIZE; /* 3-level kernel page table */
 563                 else
 564                         vmax = _REGION1_SIZE; /* 4-level kernel page table */
 565         }
 566
 567         /* module area is at the end of the kernel address space. */
 568         MODULES_END = vmax;
 569         MODULES_VADDR = MODULES_END - MODULES_LEN;
 570         VMALLOC_END = MODULES_VADDR;
 571         VMALLOC_START = VMALLOC_END - vmalloc_size;
 572
 573         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 574         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 575         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 576         tmp = SECTION_ALIGN_UP(tmp);
 577         tmp = VMALLOC_START - tmp * sizeof(struct page);
 578         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 579         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 580         vmemmap = (struct page *) tmp;
 581
 582         /* Take care that memory_end is set and <= vmemmap */
 583         memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
 584 #ifdef CONFIG_KASAN
 585         /* fit in kasan shadow memory region between 1:1 and vmemmap */
 586         memory_end = min(memory_end, KASAN_SHADOW_START);
 587         vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
 588 #endif
 589         max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 590         memblock_remove(memory_end, ULONG_MAX);
 591
 592         pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
 593 }
 594
 595 #ifdef CONFIG_CRASH_DUMP
 596
 597 /*
 598  * When kdump is enabled, we have to ensure that no memory from
 599  * the area [0 - crashkernel memory size] and
 600  * [crashk_res.start - crashk_res.end] is set offline.
 601  */
 602 static int kdump_mem_notifier(struct notifier_block *nb,
 603                               unsigned long action, void *data)
 604 {
 605         struct memory_notify *arg = data;
 606
 607         if (action != MEM_GOING_OFFLINE)
 608                 return NOTIFY_OK;
 609         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 610                 return NOTIFY_BAD;
 611         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 612                 return NOTIFY_OK;
 613         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 614                 return NOTIFY_OK;
 615         return NOTIFY_BAD;
 616 }
 617
 618 static struct notifier_block kdump_mem_nb = {
 619         .notifier_call = kdump_mem_notifier,
 620 };
 621
 622 #endif
 623
 624 /*
 625  * Make sure that the area behind memory_end is protected
 626  */
 627 static void reserve_memory_end(void)
 628 {
 629         if (memory_end_set)
 630                 memblock_reserve(memory_end, ULONG_MAX);
 631 }
 632
 633 /*
 634  * Make sure that oldmem, where the dump is stored, is protected
 635  */
 636 static void reserve_oldmem(void)
 637 {
 638 #ifdef CONFIG_CRASH_DUMP
 639         if (OLDMEM_BASE)
 640                 /* Forget all memory above the running kdump system */
 641                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 642 #endif
 643 }
 644
 645 /*
 646  * Make sure that oldmem, where the dump is stored, is protected
 647  */
 648 static void remove_oldmem(void)
 649 {
 650 #ifdef CONFIG_CRASH_DUMP
 651         if (OLDMEM_BASE)
 652                 /* Forget all memory above the running kdump system */
 653                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 654 #endif
 655 }
 656
 657 /*
 658  * Reserve memory for kdump kernel to be loaded with kexec
 659  */
 660 static void __init reserve_crashkernel(void)
 661 {
 662 #ifdef CONFIG_CRASH_DUMP
 663         unsigned long long crash_base, crash_size;
 664         phys_addr_t low, high;
 665         int rc;
 666
 667         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 668                                &crash_base);
 669
 670         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 671         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 672         if (rc || crash_size == 0)
 673                 return;
 674
 675         if (memblock.memory.regions[0].size < crash_size) {
 676                 pr_info("crashkernel reservation failed: %s\n",
 677                         "first memory chunk must be at least crashkernel size");
 678                 return;
 679         }
 680
 681         low = crash_base ?: OLDMEM_BASE;
 682         high = low + crash_size;
 683         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 684                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 685                 crash_base = low;
 686         } else {
 687                 /* Find suitable area in free memory */
 688                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 689                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 690
 691                 if (crash_base && crash_base < low) {
 692                         pr_info("crashkernel reservation failed: %s\n",
 693                                 "crash_base too low");
 694                         return;
 695                 }
 696                 low = crash_base ?: low;
 697                 crash_base = memblock_find_in_range(low, high, crash_size,
 698                                                     KEXEC_CRASH_MEM_ALIGN);
 699         }
 700
 701         if (!crash_base) {
 702                 pr_info("crashkernel reservation failed: %s\n",
 703                         "no suitable area found");
 704                 return;
 705         }
 706
 707         if (register_memory_notifier(&kdump_mem_nb))
 708                 return;
 709
 710         if (!OLDMEM_BASE && MACHINE_IS_VM)
 711                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 712         crashk_res.start = crash_base;
 713         crashk_res.end = crash_base + crash_size - 1;
 714         memblock_remove(crash_base, crash_size);
 715         pr_info("Reserving %lluMB of memory at %lluMB "
 716                 "for crashkernel (System RAM: %luMB)\n",
 717                 crash_size >> 20, crash_base >> 20,
 718                 (unsigned long)memblock.memory.total_size >> 20);
 719         os_info_crashkernel_add(crash_base, crash_size);
 720 #endif
 721 }
 722
 723 /*
 724  * Reserve the initrd from being used by memblock
 725  */
 726 static void __init reserve_initrd(void)
 727 {
 728 #ifdef CONFIG_BLK_DEV_INITRD
 729         if (!INITRD_START || !INITRD_SIZE)
 730                 return;
 731         initrd_start = INITRD_START;
 732         initrd_end = initrd_start + INITRD_SIZE;
 733         memblock_reserve(INITRD_START, INITRD_SIZE);
 734 #endif
 735 }
 736
 737 /*
 738  * Reserve the memory area used to pass the certificate lists
 739  */
 740 static void __init reserve_certificate_list(void)
 741 {
 742         if (ipl_cert_list_addr)
 743                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
 744 }
 745
 746 static void __init reserve_mem_detect_info(void)
 747 {
 748         unsigned long start, size;
 749
 750         get_mem_detect_reserved(&start, &size);
 751         if (size)
 752                 memblock_reserve(start, size);
 753 }
 754
 755 static void __init free_mem_detect_info(void)
 756 {
 757         unsigned long start, size;
 758
 759         get_mem_detect_reserved(&start, &size);
 760         if (size)
 761                 memblock_free(start, size);
 762 }
 763
 764 static const char * __init get_mem_info_source(void)
 765 {
 766         switch (mem_detect.info_source) {
 767         case MEM_DETECT_SCLP_STOR_INFO:
 768                 return "sclp storage info";
 769         case MEM_DETECT_DIAG260:
 770                 return "diag260";
 771         case MEM_DETECT_SCLP_READ_INFO:
 772                 return "sclp read info";
 773         case MEM_DETECT_BIN_SEARCH:
 774                 return "binary search";
 775         }
 776         return "none";
 777 }
 778
 779 static void __init memblock_add_mem_detect_info(void)
 780 {
 781         unsigned long start, end;
 782         int i;
 783
 784         memblock_dbg("physmem info source: %s (%hhd)\n",
 785                      get_mem_info_source(), mem_detect.info_source);
 786         /* keep memblock lists close to the kernel */
 787         memblock_set_bottom_up(true);
 788         for_each_mem_detect_block(i, &start, &end) {
 789                 memblock_add(start, end - start);
 790                 memblock_physmem_add(start, end - start);
 791         }
 792         memblock_set_bottom_up(false);
 793         memblock_dump_all();
 794 }
 795
 796 /*
 797  * Check for initrd being in usable memory
 798  */
 799 static void __init check_initrd(void)
 800 {
 801 #ifdef CONFIG_BLK_DEV_INITRD
 802         if (INITRD_START && INITRD_SIZE &&
 803             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 804                 pr_err("The initial RAM disk does not fit into the memory\n");
 805                 memblock_free(INITRD_START, INITRD_SIZE);
 806                 initrd_start = initrd_end = 0;
 807         }
 808 #endif
 809 }
 810
 811 /*
 812  * Reserve memory used for lowcore/command line/kernel image.
 813  */
 814 static void __init reserve_kernel(void)
 815 {
 816         unsigned long start_pfn = PFN_UP(__pa(_end));
 817
 818         memblock_reserve(0, HEAD_END);
 819         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 820                          - (unsigned long)_stext);
 821         memblock_reserve(__sdma, __edma - __sdma);
 822 }
 823
 824 static void __init setup_memory(void)
 825 {
 826         struct memblock_region *reg;
 827
 828         /*
 829          * Init storage key for present memory
 830          */
 831         for_each_memblock(memory, reg) {
 832                 storage_key_init_range(reg->base, reg->base + reg->size);
 833         }
 834         psw_set_key(PAGE_DEFAULT_KEY);
 835
 836         /* Only cosmetics */
 837         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 838 }
 839
 840 /*
 841  * Setup hardware capabilities.
 842  */
 843 static int __init setup_hwcaps(void)
 844 {
 845         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 846         struct cpuid cpu_id;
 847         int i;
 848
 849         /*
 850          * The store facility list bits numbers as found in the principles
 851          * of operation are numbered with bit 1UL<<31 as number 0 to
 852          * bit 1UL<<0 as number 31.
 853          *   Bit 0: instructions named N3, "backported" to esa-mode
 854          *   Bit 2: z/Architecture mode is active
 855          *   Bit 7: the store-facility-list-extended facility is installed
 856          *   Bit 17: the message-security assist is installed
 857          *   Bit 19: the long-displacement facility is installed
 858          *   Bit 21: the extended-immediate facility is installed
 859          *   Bit 22: extended-translation facility 3 is installed
 860          *   Bit 30: extended-translation facility 3 enhancement facility
 861          * These get translated to:
 862          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 863          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 864          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 865          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 866          */
 867         for (i = 0; i < 6; i++)
 868                 if (test_facility(stfl_bits[i]))
 869                         elf_hwcap |= 1UL << i;
 870
 871         if (test_facility(22) && test_facility(30))
 872                 elf_hwcap |= HWCAP_S390_ETF3EH;
 873
 874         /*
 875          * Check for additional facilities with store-facility-list-extended.
 876          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 877          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 878          * as stored by stfl, bits 32-xxx contain additional facilities.
 879          * How many facility words are stored depends on the number of
 880          * doublewords passed to the instruction. The additional facilities
 881          * are:
 882          *   Bit 42: decimal floating point facility is installed
 883          *   Bit 44: perform floating point operation facility is installed
 884          * translated to:
 885          *   HWCAP_S390_DFP bit 6 (42 && 44).
 886          */
 887         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 888                 elf_hwcap |= HWCAP_S390_DFP;
 889
 890         /*
 891          * Huge page support HWCAP_S390_HPAGE is bit 7.
 892          */
 893         if (MACHINE_HAS_EDAT1)
 894                 elf_hwcap |= HWCAP_S390_HPAGE;
 895
 896         /*
 897          * 64-bit register support for 31-bit processes
 898          * HWCAP_S390_HIGH_GPRS is bit 9.
 899          */
 900         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 901
 902         /*
 903          * Transactional execution support HWCAP_S390_TE is bit 10.
 904          */
 905         if (MACHINE_HAS_TE)
 906                 elf_hwcap |= HWCAP_S390_TE;
 907
 908         /*
 909          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 910          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 911          * instead of facility bit 129.
 912          */
 913         if (MACHINE_HAS_VX) {
 914                 elf_hwcap |= HWCAP_S390_VXRS;
 915                 if (test_facility(134))
 916                         elf_hwcap |= HWCAP_S390_VXRS_EXT;
 917                 if (test_facility(135))
 918                         elf_hwcap |= HWCAP_S390_VXRS_BCD;
 919                 if (test_facility(148))
 920                         elf_hwcap |= HWCAP_S390_VXRS_EXT2;
 921                 if (test_facility(152))
 922                         elf_hwcap |= HWCAP_S390_VXRS_PDE;
 923         }
 924         if (test_facility(150))
 925                 elf_hwcap |= HWCAP_S390_SORT;
 926         if (test_facility(151))
 927                 elf_hwcap |= HWCAP_S390_DFLT;
 928
 929         /*
 930          * Guarded storage support HWCAP_S390_GS is bit 12.
 931          */
 932         if (MACHINE_HAS_GS)
 933                 elf_hwcap |= HWCAP_S390_GS;
 934
 935         get_cpu_id(&cpu_id);
 936         add_device_randomness(&cpu_id, sizeof(cpu_id));
 937         switch (cpu_id.machine) {
 938         case 0x2064:
 939         case 0x2066:
 940         default:        /* Use "z900" as default for 64 bit kernels. */
 941                 strcpy(elf_platform, "z900");
 942                 break;
 943         case 0x2084:
 944         case 0x2086:
 945                 strcpy(elf_platform, "z990");
 946                 break;
 947         case 0x2094:
 948         case 0x2096:
 949                 strcpy(elf_platform, "z9-109");
 950                 break;
 951         case 0x2097:
 952         case 0x2098:
 953                 strcpy(elf_platform, "z10");
 954                 break;
 955         case 0x2817:
 956         case 0x2818:
 957                 strcpy(elf_platform, "z196");
 958                 break;
 959         case 0x2827:
 960         case 0x2828:
 961                 strcpy(elf_platform, "zEC12");
 962                 break;
 963         case 0x2964:
 964         case 0x2965:
 965                 strcpy(elf_platform, "z13");
 966                 break;
 967         case 0x3906:
 968         case 0x3907:
 969                 strcpy(elf_platform, "z14");
 970                 break;
 971         case 0x8561:
 972         case 0x8562:
 973                 strcpy(elf_platform, "z15");
 974                 break;
 975         }
 976
 977         /*
 978          * Virtualization support HWCAP_INT_SIE is bit 0.
 979          */
 980         if (sclp.has_sief2)
 981                 int_hwcap |= HWCAP_INT_SIE;
 982
 983         return 0;
 984 }
 985 arch_initcall(setup_hwcaps);
 986
 987 /*
 988  * Add system information as device randomness
 989  */
 990 static void __init setup_randomness(void)
 991 {
 992         struct sysinfo_3_2_2 *vmms;
 993
 994         vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
 995                                                             PAGE_SIZE);
 996         if (!vmms)
 997                 panic("Failed to allocate memory for sysinfo structure\n");
 998
 999         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1000                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1001         memblock_free((unsigned long) vmms, PAGE_SIZE);
1002 }
1003
1004 /*
1005  * Find the correct size for the task_struct. This depends on
1006  * the size of the struct fpu at the end of the thread_struct
1007  * which is embedded in the task_struct.
1008  */
1009 static void __init setup_task_size(void)
1010 {
1011         int task_size = sizeof(struct task_struct);
1012
1013         if (!MACHINE_HAS_VX) {
1014                 task_size -= sizeof(__vector128) * __NUM_VXRS;
1015                 task_size += sizeof(freg_t) * __NUM_FPRS;
1016         }
1017         arch_task_struct_size = task_size;
1018 }
1019
1020 /*
1021  * Issue diagnose 318 to set the control program name and
1022  * version codes.
1023  */
1024 static void __init setup_control_program_code(void)
1025 {
1026         union diag318_info diag318_info = {
1027                 .cpnc = CPNC_LINUX,
1028                 .cpvc_linux = 0,
1029                 .cpvc_distro = {0},
1030         };
1031
1032         if (!sclp.has_diag318)
1033                 return;
1034
1035         diag_stat_inc(DIAG_STAT_X318);
1036         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1037 }
1038
1039 /*
1040  * Print the component list from the IPL report
1041  */
1042 static void __init log_component_list(void)
1043 {
1044         struct ipl_rb_component_entry *ptr, *end;
1045         char *str;
1046
1047         if (!early_ipl_comp_list_addr)
1048                 return;
1049         if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1050                 pr_info("Linux is running with Secure-IPL enabled\n");
1051         else
1052                 pr_info("Linux is running with Secure-IPL disabled\n");
1053         ptr = (void *) early_ipl_comp_list_addr;
1054         end = (void *) ptr + early_ipl_comp_list_size;
1055         pr_info("The IPL report contains the following components:\n");
1056         while (ptr < end) {
1057                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1058                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1059                                 str = "signed, verified";
1060                         else
1061                                 str = "signed, verification failed";
1062                 } else {
1063                         str = "not signed";
1064                 }
1065                 pr_info("%016llx - %016llx (%s)\n",
1066                         ptr->addr, ptr->addr + ptr->len, str);
1067                 ptr++;
1068         }
1069 }
1070
1071 /*
1072  * Setup function called from init/main.c just after the banner
1073  * was printed.
1074  */
1075
1076 void __init setup_arch(char **cmdline_p)
1077 {
1078         /*
1079          * print what head.S has found out about the machine
1080          */
1081         if (MACHINE_IS_VM)
1082                 pr_info("Linux is running as a z/VM "
1083                         "guest operating system in 64-bit mode\n");
1084         else if (MACHINE_IS_KVM)
1085                 pr_info("Linux is running under KVM in 64-bit mode\n");
1086         else if (MACHINE_IS_LPAR)
1087                 pr_info("Linux is running natively in 64-bit mode\n");
1088         else
1089                 pr_info("Linux is running as a guest in 64-bit mode\n");
1090
1091         log_component_list();
1092
1093         /* Have one command line that is parsed and saved in /proc/cmdline */
1094         /* boot_command_line has been already set up in early.c */
1095         *cmdline_p = boot_command_line;
1096
1097         ROOT_DEV = Root_RAM0;
1098
1099         init_mm.start_code = (unsigned long) _text;
1100         init_mm.end_code = (unsigned long) _etext;
1101         init_mm.end_data = (unsigned long) _edata;
1102         init_mm.brk = (unsigned long) _end;
1103
1104         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1105                 nospec_auto_detect();
1106
1107         parse_early_param();
1108 #ifdef CONFIG_CRASH_DUMP
1109         /* Deactivate elfcorehdr= kernel parameter */
1110         elfcorehdr_addr = ELFCORE_ADDR_MAX;
1111 #endif
1112
1113         os_info_init();
1114         setup_ipl();
1115         setup_task_size();
1116         setup_control_program_code();
1117
1118         /* Do some memory reservations *before* memory is added to memblock */
1119         reserve_memory_end();
1120         reserve_oldmem();
1121         reserve_kernel();
1122         reserve_initrd();
1123         reserve_certificate_list();
1124         reserve_mem_detect_info();
1125         memblock_allow_resize();
1126
1127         /* Get information about *all* installed memory */
1128         memblock_add_mem_detect_info();
1129
1130         free_mem_detect_info();
1131         remove_oldmem();
1132
1133         /*
1134          * Make sure all chunks are MAX_ORDER aligned so we don't need the
1135          * extra checks that HOLES_IN_ZONE would require.
1136          *
1137          * Is this still required?
1138          */
1139         memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
1140
1141         setup_memory_end();
1142         setup_memory();
1143         dma_contiguous_reserve(memory_end);
1144         vmcp_cma_reserve();
1145
1146         check_initrd();
1147         reserve_crashkernel();
1148 #ifdef CONFIG_CRASH_DUMP
1149         /*
1150          * Be aware that smp_save_dump_cpus() triggers a system reset.
1151          * Therefore CPU and device initialization should be done afterwards.
1152          */
1153         smp_save_dump_cpus();
1154 #endif
1155
1156         setup_resources();
1157         setup_lowcore_dat_off();
1158         smp_fill_possible_mask();
1159         cpu_detect_mhz_feature();
1160         cpu_init();
1161         numa_setup();
1162         smp_detect_cpus();
1163         topology_init_early();
1164
1165         /*
1166          * Create kernel page tables and switch to virtual addressing.
1167          */
1168         paging_init();
1169
1170         /*
1171          * After paging_init created the kernel page table, the new PSWs
1172          * in lowcore can now run with DAT enabled.
1173          */
1174         setup_lowcore_dat_on();
1175
1176         /* Setup default console */
1177         conmode_default();
1178         set_preferred_console();
1179
1180         apply_alternative_instructions();
1181         if (IS_ENABLED(CONFIG_EXPOLINE))
1182                 nospec_init_branches();
1183
1184         /* Setup zfcpdump support */
1185         setup_zfcpdump();
1186
1187         /* Add system specific data to the random pool */
1188         setup_randomness();
1189 }