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/bootmem.h>
  38 #include <linux/root_dev.h>
  39 #include <linux/console.h>
  40 #include <linux/kernel_stat.h>
  41 #include <linux/dma-contiguous.h>
  42 #include <linux/device.h>
  43 #include <linux/notifier.h>
  44 #include <linux/pfn.h>
  45 #include <linux/ctype.h>
  46 #include <linux/reboot.h>
  47 #include <linux/topology.h>
  48 #include <linux/kexec.h>
  49 #include <linux/crash_dump.h>
  50 #include <linux/memory.h>
  51 #include <linux/compat.h>
  52
  53 #include <asm/ipl.h>
  54 #include <asm/facility.h>
  55 #include <asm/smp.h>
  56 #include <asm/mmu_context.h>
  57 #include <asm/cpcmd.h>
  58 #include <asm/lowcore.h>
  59 #include <asm/nmi.h>
  60 #include <asm/irq.h>
  61 #include <asm/page.h>
  62 #include <asm/ptrace.h>
  63 #include <asm/sections.h>
  64 #include <asm/ebcdic.h>
  65 #include <asm/diag.h>
  66 #include <asm/os_info.h>
  67 #include <asm/sclp.h>
  68 #include <asm/sysinfo.h>
  69 #include <asm/numa.h>
  70 #include <asm/alternative.h>
  71 #include <asm/nospec-branch.h>
  72 #include "entry.h"
  73
  74 /*
  75  * Machine setup..
  76  */
  77 unsigned int console_mode = 0;
  78 EXPORT_SYMBOL(console_mode);
  79
  80 unsigned int console_devno = -1;
  81 EXPORT_SYMBOL(console_devno);
  82
  83 unsigned int console_irq = -1;
  84 EXPORT_SYMBOL(console_irq);
  85
  86 unsigned long elf_hwcap __read_mostly = 0;
  87 char elf_platform[ELF_PLATFORM_SIZE];
  88
  89 unsigned long int_hwcap = 0;
  90
  91 int __initdata memory_end_set;
  92 unsigned long __initdata memory_end;
  93 unsigned long __initdata max_physmem_end;
  94
  95 unsigned long VMALLOC_START;
  96 EXPORT_SYMBOL(VMALLOC_START);
  97
  98 unsigned long VMALLOC_END;
  99 EXPORT_SYMBOL(VMALLOC_END);
 100
 101 struct page *vmemmap;
 102 EXPORT_SYMBOL(vmemmap);
 103
 104 unsigned long MODULES_VADDR;
 105 unsigned long MODULES_END;
 106
 107 /* An array with a pointer to the lowcore of every CPU. */
 108 struct lowcore *lowcore_ptr[NR_CPUS];
 109 EXPORT_SYMBOL(lowcore_ptr);
 110
 111 /*
 112  * This is set up by the setup-routine at boot-time
 113  * for S390 need to find out, what we have to setup
 114  * using address 0x10400 ...
 115  */
 116
 117 #include <asm/setup.h>
 118
 119 /*
 120  * condev= and conmode= setup parameter.
 121  */
 122
 123 static int __init condev_setup(char *str)
 124 {
 125         int vdev;
 126
 127         vdev = simple_strtoul(str, &str, 0);
 128         if (vdev >= 0 && vdev < 65536) {
 129                 console_devno = vdev;
 130                 console_irq = -1;
 131         }
 132         return 1;
 133 }
 134
 135 __setup("condev=", condev_setup);
 136
 137 static void __init set_preferred_console(void)
 138 {
 139         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 140                 add_preferred_console("ttyS", 0, NULL);
 141         else if (CONSOLE_IS_3270)
 142                 add_preferred_console("tty3270", 0, NULL);
 143         else if (CONSOLE_IS_VT220)
 144                 add_preferred_console("ttyS", 1, NULL);
 145         else if (CONSOLE_IS_HVC)
 146                 add_preferred_console("hvc", 0, NULL);
 147 }
 148
 149 static int __init conmode_setup(char *str)
 150 {
 151 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 152         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 153                 SET_CONSOLE_SCLP;
 154 #endif
 155 #if defined(CONFIG_TN3215_CONSOLE)
 156         if (strncmp(str, "3215", 5) == 0)
 157                 SET_CONSOLE_3215;
 158 #endif
 159 #if defined(CONFIG_TN3270_CONSOLE)
 160         if (strncmp(str, "3270", 5) == 0)
 161                 SET_CONSOLE_3270;
 162 #endif
 163         set_preferred_console();
 164         return 1;
 165 }
 166
 167 __setup("conmode=", conmode_setup);
 168
 169 static void __init conmode_default(void)
 170 {
 171         char query_buffer[1024];
 172         char *ptr;
 173
 174         if (MACHINE_IS_VM) {
 175                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 176                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 177                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 178                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 179                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 180                 ptr = strstr(query_buffer, "CONMODE");
 181                 /*
 182                  * Set the conmode to 3215 so that the device recognition
 183                  * will set the cu_type of the console to 3215. If the
 184                  * conmode is 3270 and we don't set it back then both
 185                  * 3215 and the 3270 driver will try to access the console
 186                  * device (3215 as console and 3270 as normal tty).
 187                  */
 188                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 189                 if (ptr == NULL) {
 190 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 191                         SET_CONSOLE_SCLP;
 192 #endif
 193                         return;
 194                 }
 195                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 196 #if defined(CONFIG_TN3270_CONSOLE)
 197                         SET_CONSOLE_3270;
 198 #elif defined(CONFIG_TN3215_CONSOLE)
 199                         SET_CONSOLE_3215;
 200 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 201                         SET_CONSOLE_SCLP;
 202 #endif
 203                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 204 #if defined(CONFIG_TN3215_CONSOLE)
 205                         SET_CONSOLE_3215;
 206 #elif defined(CONFIG_TN3270_CONSOLE)
 207                         SET_CONSOLE_3270;
 208 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 209                         SET_CONSOLE_SCLP;
 210 #endif
 211                 }
 212         } else if (MACHINE_IS_KVM) {
 213                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
 214                         SET_CONSOLE_VT220;
 215                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
 216                         SET_CONSOLE_SCLP;
 217                 else
 218                         SET_CONSOLE_HVC;
 219         } else {
 220 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 221                 SET_CONSOLE_SCLP;
 222 #endif
 223         }
 224 }
 225
 226 #ifdef CONFIG_CRASH_DUMP
 227 static void __init setup_zfcpdump(void)
 228 {
 229         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 230                 return;
 231         if (OLDMEM_BASE)
 232                 return;
 233         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 234         console_loglevel = 2;
 235 }
 236 #else
 237 static inline void setup_zfcpdump(void) {}
 238 #endif /* CONFIG_CRASH_DUMP */
 239
 240  /*
 241  * Reboot, halt and power_off stubs. They just call _machine_restart,
 242  * _machine_halt or _machine_power_off.
 243  */
 244
 245 void machine_restart(char *command)
 246 {
 247         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 248                 /*
 249                  * Only unblank the console if we are called in enabled
 250                  * context or a bust_spinlocks cleared the way for us.
 251                  */
 252                 console_unblank();
 253         _machine_restart(command);
 254 }
 255
 256 void machine_halt(void)
 257 {
 258         if (!in_interrupt() || oops_in_progress)
 259                 /*
 260                  * Only unblank the console if we are called in enabled
 261                  * context or a bust_spinlocks cleared the way for us.
 262                  */
 263                 console_unblank();
 264         _machine_halt();
 265 }
 266
 267 void machine_power_off(void)
 268 {
 269         if (!in_interrupt() || 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_power_off();
 276 }
 277
 278 /*
 279  * Dummy power off function.
 280  */
 281 void (*pm_power_off)(void) = machine_power_off;
 282 EXPORT_SYMBOL_GPL(pm_power_off);
 283
 284 static int __init early_parse_mem(char *p)
 285 {
 286         memory_end = memparse(p, &p);
 287         memory_end &= PAGE_MASK;
 288         memory_end_set = 1;
 289         return 0;
 290 }
 291 early_param("mem", early_parse_mem);
 292
 293 static int __init parse_vmalloc(char *arg)
 294 {
 295         if (!arg)
 296                 return -EINVAL;
 297         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
 298         return 0;
 299 }
 300 early_param("vmalloc", parse_vmalloc);
 301
 302 void *restart_stack __section(.data);
 303
 304 static void __init setup_lowcore(void)
 305 {
 306         struct lowcore *lc;
 307
 308         /*
 309          * Setup lowcore for boot cpu
 310          */
 311         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 312         lc = memblock_virt_alloc_low(sizeof(*lc), sizeof(*lc));
 313         lc->restart_psw.mask = PSW_KERNEL_BITS;
 314         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 315         lc->external_new_psw.mask = PSW_KERNEL_BITS |
 316                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 317         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 318         lc->svc_new_psw.mask = PSW_KERNEL_BITS |
 319                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 320         lc->svc_new_psw.addr = (unsigned long) system_call;
 321         lc->program_new_psw.mask = PSW_KERNEL_BITS |
 322                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 323         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 324         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 325         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 326         lc->io_new_psw.mask = PSW_KERNEL_BITS |
 327                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 328         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 329         lc->clock_comparator = clock_comparator_max;
 330         lc->kernel_stack = ((unsigned long) &init_thread_union)
 331                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 332         lc->async_stack = (unsigned long)
 333                 memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE)
 334                 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 335         lc->panic_stack = (unsigned long)
 336                 memblock_virt_alloc(PAGE_SIZE, PAGE_SIZE)
 337                 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 338         lc->current_task = (unsigned long)&init_task;
 339         lc->lpp = LPP_MAGIC;
 340         lc->machine_flags = S390_lowcore.machine_flags;
 341         lc->preempt_count = S390_lowcore.preempt_count;
 342         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 343         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 344                sizeof(lc->stfle_fac_list));
 345         memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
 346                sizeof(lc->alt_stfle_fac_list));
 347         nmi_alloc_boot_cpu(lc);
 348         vdso_alloc_boot_cpu(lc);
 349         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 350         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 351         lc->exit_timer = S390_lowcore.exit_timer;
 352         lc->user_timer = S390_lowcore.user_timer;
 353         lc->system_timer = S390_lowcore.system_timer;
 354         lc->steal_timer = S390_lowcore.steal_timer;
 355         lc->last_update_timer = S390_lowcore.last_update_timer;
 356         lc->last_update_clock = S390_lowcore.last_update_clock;
 357
 358         restart_stack = memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE);
 359         restart_stack += ASYNC_SIZE;
 360
 361         /*
 362          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 363          * restart data to the absolute zero lowcore. This is necessary if
 364          * PSW restart is done on an offline CPU that has lowcore zero.
 365          */
 366         lc->restart_stack = (unsigned long) restart_stack;
 367         lc->restart_fn = (unsigned long) do_restart;
 368         lc->restart_data = 0;
 369         lc->restart_source = -1UL;
 370
 371         /* Setup absolute zero lowcore */
 372         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 373         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 374         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 375         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 376         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 377
 378 #ifdef CONFIG_SMP
 379         lc->spinlock_lockval = arch_spin_lockval(0);
 380         lc->spinlock_index = 0;
 381         arch_spin_lock_setup(0);
 382 #endif
 383         lc->br_r1_trampoline = 0x07f1;  /* br %r1 */
 384
 385         set_prefix((u32)(unsigned long) lc);
 386         lowcore_ptr[0] = lc;
 387 }
 388
 389 static struct resource code_resource = {
 390         .name  = "Kernel code",
 391         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 392 };
 393
 394 static struct resource data_resource = {
 395         .name = "Kernel data",
 396         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 397 };
 398
 399 static struct resource bss_resource = {
 400         .name = "Kernel bss",
 401         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 402 };
 403
 404 static struct resource __initdata *standard_resources[] = {
 405         &code_resource,
 406         &data_resource,
 407         &bss_resource,
 408 };
 409
 410 static void __init setup_resources(void)
 411 {
 412         struct resource *res, *std_res, *sub_res;
 413         struct memblock_region *reg;
 414         int j;
 415
 416         code_resource.start = (unsigned long) &_text;
 417         code_resource.end = (unsigned long) &_etext - 1;
 418         data_resource.start = (unsigned long) &_etext;
 419         data_resource.end = (unsigned long) &_edata - 1;
 420         bss_resource.start = (unsigned long) &__bss_start;
 421         bss_resource.end = (unsigned long) &__bss_stop - 1;
 422
 423         for_each_memblock(memory, reg) {
 424                 res = memblock_virt_alloc(sizeof(*res), 8);
 425                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 426
 427                 res->name = "System RAM";
 428                 res->start = reg->base;
 429                 res->end = reg->base + reg->size - 1;
 430                 request_resource(&iomem_resource, res);
 431
 432                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 433                         std_res = standard_resources[j];
 434                         if (std_res->start < res->start ||
 435                             std_res->start > res->end)
 436                                 continue;
 437                         if (std_res->end > res->end) {
 438                                 sub_res = memblock_virt_alloc(sizeof(*sub_res), 8);
 439                                 *sub_res = *std_res;
 440                                 sub_res->end = res->end;
 441                                 std_res->start = res->end + 1;
 442                                 request_resource(res, sub_res);
 443                         } else {
 444                                 request_resource(res, std_res);
 445                         }
 446                 }
 447         }
 448 #ifdef CONFIG_CRASH_DUMP
 449         /*
 450          * Re-add removed crash kernel memory as reserved memory. This makes
 451          * sure it will be mapped with the identity mapping and struct pages
 452          * will be created, so it can be resized later on.
 453          * However add it later since the crash kernel resource should not be
 454          * part of the System RAM resource.
 455          */
 456         if (crashk_res.end) {
 457                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 458                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 459                 insert_resource(&iomem_resource, &crashk_res);
 460         }
 461 #endif
 462 }
 463
 464 static void __init setup_memory_end(void)
 465 {
 466         unsigned long vmax, vmalloc_size, tmp;
 467
 468         /* Choose kernel address space layout: 2, 3, or 4 levels. */
 469         vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
 470         tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
 471         tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
 472         if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
 473                 vmax = _REGION2_SIZE; /* 3-level kernel page table */
 474         else
 475                 vmax = _REGION1_SIZE; /* 4-level kernel page table */
 476         /* module area is at the end of the kernel address space. */
 477         MODULES_END = vmax;
 478         MODULES_VADDR = MODULES_END - MODULES_LEN;
 479         VMALLOC_END = MODULES_VADDR;
 480         VMALLOC_START = vmax - vmalloc_size;
 481
 482         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 483         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 484         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 485         tmp = SECTION_ALIGN_UP(tmp);
 486         tmp = VMALLOC_START - tmp * sizeof(struct page);
 487         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 488         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 489         vmemmap = (struct page *) tmp;
 490
 491         /* Take care that memory_end is set and <= vmemmap */
 492         memory_end = min(memory_end ?: max_physmem_end, tmp);
 493         max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 494         memblock_remove(memory_end, ULONG_MAX);
 495
 496         pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
 497 }
 498
 499 #ifdef CONFIG_CRASH_DUMP
 500
 501 /*
 502  * When kdump is enabled, we have to ensure that no memory from
 503  * the area [0 - crashkernel memory size] and
 504  * [crashk_res.start - crashk_res.end] is set offline.
 505  */
 506 static int kdump_mem_notifier(struct notifier_block *nb,
 507                               unsigned long action, void *data)
 508 {
 509         struct memory_notify *arg = data;
 510
 511         if (action != MEM_GOING_OFFLINE)
 512                 return NOTIFY_OK;
 513         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 514                 return NOTIFY_BAD;
 515         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 516                 return NOTIFY_OK;
 517         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 518                 return NOTIFY_OK;
 519         return NOTIFY_BAD;
 520 }
 521
 522 static struct notifier_block kdump_mem_nb = {
 523         .notifier_call = kdump_mem_notifier,
 524 };
 525
 526 #endif
 527
 528 /*
 529  * Make sure that the area behind memory_end is protected
 530  */
 531 static void reserve_memory_end(void)
 532 {
 533 #ifdef CONFIG_CRASH_DUMP
 534         if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
 535             !OLDMEM_BASE && sclp.hsa_size) {
 536                 memory_end = sclp.hsa_size;
 537                 memory_end &= PAGE_MASK;
 538                 memory_end_set = 1;
 539         }
 540 #endif
 541         if (!memory_end_set)
 542                 return;
 543         memblock_reserve(memory_end, ULONG_MAX);
 544 }
 545
 546 /*
 547  * Make sure that oldmem, where the dump is stored, is protected
 548  */
 549 static void reserve_oldmem(void)
 550 {
 551 #ifdef CONFIG_CRASH_DUMP
 552         if (OLDMEM_BASE)
 553                 /* Forget all memory above the running kdump system */
 554                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 555 #endif
 556 }
 557
 558 /*
 559  * Make sure that oldmem, where the dump is stored, is protected
 560  */
 561 static void remove_oldmem(void)
 562 {
 563 #ifdef CONFIG_CRASH_DUMP
 564         if (OLDMEM_BASE)
 565                 /* Forget all memory above the running kdump system */
 566                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 567 #endif
 568 }
 569
 570 /*
 571  * Reserve memory for kdump kernel to be loaded with kexec
 572  */
 573 static void __init reserve_crashkernel(void)
 574 {
 575 #ifdef CONFIG_CRASH_DUMP
 576         unsigned long long crash_base, crash_size;
 577         phys_addr_t low, high;
 578         int rc;
 579
 580         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 581                                &crash_base);
 582
 583         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 584         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 585         if (rc || crash_size == 0)
 586                 return;
 587
 588         if (memblock.memory.regions[0].size < crash_size) {
 589                 pr_info("crashkernel reservation failed: %s\n",
 590                         "first memory chunk must be at least crashkernel size");
 591                 return;
 592         }
 593
 594         low = crash_base ?: OLDMEM_BASE;
 595         high = low + crash_size;
 596         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 597                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 598                 crash_base = low;
 599         } else {
 600                 /* Find suitable area in free memory */
 601                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 602                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 603
 604                 if (crash_base && crash_base < low) {
 605                         pr_info("crashkernel reservation failed: %s\n",
 606                                 "crash_base too low");
 607                         return;
 608                 }
 609                 low = crash_base ?: low;
 610                 crash_base = memblock_find_in_range(low, high, crash_size,
 611                                                     KEXEC_CRASH_MEM_ALIGN);
 612         }
 613
 614         if (!crash_base) {
 615                 pr_info("crashkernel reservation failed: %s\n",
 616                         "no suitable area found");
 617                 return;
 618         }
 619
 620         if (register_memory_notifier(&kdump_mem_nb))
 621                 return;
 622
 623         if (!OLDMEM_BASE && MACHINE_IS_VM)
 624                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 625         crashk_res.start = crash_base;
 626         crashk_res.end = crash_base + crash_size - 1;
 627         memblock_remove(crash_base, crash_size);
 628         pr_info("Reserving %lluMB of memory at %lluMB "
 629                 "for crashkernel (System RAM: %luMB)\n",
 630                 crash_size >> 20, crash_base >> 20,
 631                 (unsigned long)memblock.memory.total_size >> 20);
 632         os_info_crashkernel_add(crash_base, crash_size);
 633 #endif
 634 }
 635
 636 /*
 637  * Reserve the initrd from being used by memblock
 638  */
 639 static void __init reserve_initrd(void)
 640 {
 641 #ifdef CONFIG_BLK_DEV_INITRD
 642         if (!INITRD_START || !INITRD_SIZE)
 643                 return;
 644         initrd_start = INITRD_START;
 645         initrd_end = initrd_start + INITRD_SIZE;
 646         memblock_reserve(INITRD_START, INITRD_SIZE);
 647 #endif
 648 }
 649
 650 /*
 651  * Check for initrd being in usable memory
 652  */
 653 static void __init check_initrd(void)
 654 {
 655 #ifdef CONFIG_BLK_DEV_INITRD
 656         if (INITRD_START && INITRD_SIZE &&
 657             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 658                 pr_err("The initial RAM disk does not fit into the memory\n");
 659                 memblock_free(INITRD_START, INITRD_SIZE);
 660                 initrd_start = initrd_end = 0;
 661         }
 662 #endif
 663 }
 664
 665 /*
 666  * Reserve memory used for lowcore/command line/kernel image.
 667  */
 668 static void __init reserve_kernel(void)
 669 {
 670         unsigned long start_pfn = PFN_UP(__pa(&_end));
 671
 672 #ifdef CONFIG_DMA_API_DEBUG
 673         /*
 674          * DMA_API_DEBUG code stumbles over addresses from the
 675          * range [_ehead, _stext]. Mark the memory as reserved
 676          * so it is not used for CONFIG_DMA_API_DEBUG=y.
 677          */
 678         memblock_reserve(0, PFN_PHYS(start_pfn));
 679 #else
 680         memblock_reserve(0, (unsigned long)_ehead);
 681         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 682                          - (unsigned long)_stext);
 683 #endif
 684 }
 685
 686 static void __init setup_memory(void)
 687 {
 688         struct memblock_region *reg;
 689
 690         /*
 691          * Init storage key for present memory
 692          */
 693         for_each_memblock(memory, reg) {
 694                 storage_key_init_range(reg->base, reg->base + reg->size);
 695         }
 696         psw_set_key(PAGE_DEFAULT_KEY);
 697
 698         /* Only cosmetics */
 699         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 700 }
 701
 702 /*
 703  * Setup hardware capabilities.
 704  */
 705 static int __init setup_hwcaps(void)
 706 {
 707         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 708         struct cpuid cpu_id;
 709         int i;
 710
 711         /*
 712          * The store facility list bits numbers as found in the principles
 713          * of operation are numbered with bit 1UL<<31 as number 0 to
 714          * bit 1UL<<0 as number 31.
 715          *   Bit 0: instructions named N3, "backported" to esa-mode
 716          *   Bit 2: z/Architecture mode is active
 717          *   Bit 7: the store-facility-list-extended facility is installed
 718          *   Bit 17: the message-security assist is installed
 719          *   Bit 19: the long-displacement facility is installed
 720          *   Bit 21: the extended-immediate facility is installed
 721          *   Bit 22: extended-translation facility 3 is installed
 722          *   Bit 30: extended-translation facility 3 enhancement facility
 723          * These get translated to:
 724          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 725          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 726          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 727          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 728          */
 729         for (i = 0; i < 6; i++)
 730                 if (test_facility(stfl_bits[i]))
 731                         elf_hwcap |= 1UL << i;
 732
 733         if (test_facility(22) && test_facility(30))
 734                 elf_hwcap |= HWCAP_S390_ETF3EH;
 735
 736         /*
 737          * Check for additional facilities with store-facility-list-extended.
 738          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 739          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 740          * as stored by stfl, bits 32-xxx contain additional facilities.
 741          * How many facility words are stored depends on the number of
 742          * doublewords passed to the instruction. The additional facilities
 743          * are:
 744          *   Bit 42: decimal floating point facility is installed
 745          *   Bit 44: perform floating point operation facility is installed
 746          * translated to:
 747          *   HWCAP_S390_DFP bit 6 (42 && 44).
 748          */
 749         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 750                 elf_hwcap |= HWCAP_S390_DFP;
 751
 752         /*
 753          * Huge page support HWCAP_S390_HPAGE is bit 7.
 754          */
 755         if (MACHINE_HAS_EDAT1)
 756                 elf_hwcap |= HWCAP_S390_HPAGE;
 757
 758         /*
 759          * 64-bit register support for 31-bit processes
 760          * HWCAP_S390_HIGH_GPRS is bit 9.
 761          */
 762         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 763
 764         /*
 765          * Transactional execution support HWCAP_S390_TE is bit 10.
 766          */
 767         if (MACHINE_HAS_TE)
 768                 elf_hwcap |= HWCAP_S390_TE;
 769
 770         /*
 771          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 772          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 773          * instead of facility bit 129.
 774          */
 775         if (MACHINE_HAS_VX) {
 776                 elf_hwcap |= HWCAP_S390_VXRS;
 777                 if (test_facility(134))
 778                         elf_hwcap |= HWCAP_S390_VXRS_EXT;
 779                 if (test_facility(135))
 780                         elf_hwcap |= HWCAP_S390_VXRS_BCD;
 781         }
 782
 783         /*
 784          * Guarded storage support HWCAP_S390_GS is bit 12.
 785          */
 786         if (MACHINE_HAS_GS)
 787                 elf_hwcap |= HWCAP_S390_GS;
 788
 789         get_cpu_id(&cpu_id);
 790         add_device_randomness(&cpu_id, sizeof(cpu_id));
 791         switch (cpu_id.machine) {
 792         case 0x2064:
 793         case 0x2066:
 794         default:        /* Use "z900" as default for 64 bit kernels. */
 795                 strcpy(elf_platform, "z900");
 796                 break;
 797         case 0x2084:
 798         case 0x2086:
 799                 strcpy(elf_platform, "z990");
 800                 break;
 801         case 0x2094:
 802         case 0x2096:
 803                 strcpy(elf_platform, "z9-109");
 804                 break;
 805         case 0x2097:
 806         case 0x2098:
 807                 strcpy(elf_platform, "z10");
 808                 break;
 809         case 0x2817:
 810         case 0x2818:
 811                 strcpy(elf_platform, "z196");
 812                 break;
 813         case 0x2827:
 814         case 0x2828:
 815                 strcpy(elf_platform, "zEC12");
 816                 break;
 817         case 0x2964:
 818         case 0x2965:
 819                 strcpy(elf_platform, "z13");
 820                 break;
 821         case 0x3906:
 822                 strcpy(elf_platform, "z14");
 823                 break;
 824         }
 825
 826         /*
 827          * Virtualization support HWCAP_INT_SIE is bit 0.
 828          */
 829         if (sclp.has_sief2)
 830                 int_hwcap |= HWCAP_INT_SIE;
 831
 832         return 0;
 833 }
 834 arch_initcall(setup_hwcaps);
 835
 836 /*
 837  * Add system information as device randomness
 838  */
 839 static void __init setup_randomness(void)
 840 {
 841         struct sysinfo_3_2_2 *vmms;
 842
 843         vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE);
 844         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 845                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
 846         memblock_free((unsigned long) vmms, PAGE_SIZE);
 847 }
 848
 849 /*
 850  * Find the correct size for the task_struct. This depends on
 851  * the size of the struct fpu at the end of the thread_struct
 852  * which is embedded in the task_struct.
 853  */
 854 static void __init setup_task_size(void)
 855 {
 856         int task_size = sizeof(struct task_struct);
 857
 858         if (!MACHINE_HAS_VX) {
 859                 task_size -= sizeof(__vector128) * __NUM_VXRS;
 860                 task_size += sizeof(freg_t) * __NUM_FPRS;
 861         }
 862         arch_task_struct_size = task_size;
 863 }
 864
 865 /*
 866  * Setup function called from init/main.c just after the banner
 867  * was printed.
 868  */
 869
 870 void __init setup_arch(char **cmdline_p)
 871 {
 872         /*
 873          * print what head.S has found out about the machine
 874          */
 875         if (MACHINE_IS_VM)
 876                 pr_info("Linux is running as a z/VM "
 877                         "guest operating system in 64-bit mode\n");
 878         else if (MACHINE_IS_KVM)
 879                 pr_info("Linux is running under KVM in 64-bit mode\n");
 880         else if (MACHINE_IS_LPAR)
 881                 pr_info("Linux is running natively in 64-bit mode\n");
 882
 883         /* Have one command line that is parsed and saved in /proc/cmdline */
 884         /* boot_command_line has been already set up in early.c */
 885         *cmdline_p = boot_command_line;
 886
 887         ROOT_DEV = Root_RAM0;
 888
 889         /* Is init_mm really needed? */
 890         init_mm.start_code = PAGE_OFFSET;
 891         init_mm.end_code = (unsigned long) &_etext;
 892         init_mm.end_data = (unsigned long) &_edata;
 893         init_mm.brk = (unsigned long) &_end;
 894
 895         parse_early_param();
 896 #ifdef CONFIG_CRASH_DUMP
 897         /* Deactivate elfcorehdr= kernel parameter */
 898         elfcorehdr_addr = ELFCORE_ADDR_MAX;
 899 #endif
 900
 901         os_info_init();
 902         setup_ipl();
 903         setup_task_size();
 904
 905         /* Do some memory reservations *before* memory is added to memblock */
 906         reserve_memory_end();
 907         reserve_oldmem();
 908         reserve_kernel();
 909         reserve_initrd();
 910         memblock_allow_resize();
 911
 912         /* Get information about *all* installed memory */
 913         detect_memory_memblock();
 914
 915         remove_oldmem();
 916
 917         /*
 918          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 919          * extra checks that HOLES_IN_ZONE would require.
 920          *
 921          * Is this still required?
 922          */
 923         memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
 924
 925         setup_memory_end();
 926         setup_memory();
 927         dma_contiguous_reserve(memory_end);
 928         vmcp_cma_reserve();
 929
 930         check_initrd();
 931         reserve_crashkernel();
 932 #ifdef CONFIG_CRASH_DUMP
 933         /*
 934          * Be aware that smp_save_dump_cpus() triggers a system reset.
 935          * Therefore CPU and device initialization should be done afterwards.
 936          */
 937         smp_save_dump_cpus();
 938 #endif
 939
 940         setup_resources();
 941         setup_lowcore();
 942         smp_fill_possible_mask();
 943         cpu_detect_mhz_feature();
 944         cpu_init();
 945         numa_setup();
 946         smp_detect_cpus();
 947         topology_init_early();
 948
 949         /*
 950          * Create kernel page tables and switch to virtual addressing.
 951          */
 952         paging_init();
 953
 954         /* Setup default console */
 955         conmode_default();
 956         set_preferred_console();
 957
 958         apply_alternative_instructions();
 959         if (IS_ENABLED(CONFIG_EXPOLINE))
 960                 nospec_init_branches();
 961
 962         /* Setup zfcpdump support */
 963         setup_zfcpdump();
 964
 965         /* Add system specific data to the random pool */
 966         setup_randomness();
 967 }