arch/s390/kernel/setup.c

   1 /*
   2  *  S390 version
   3  *    Copyright IBM Corp. 1999, 2012
   4  *    Author(s): Hartmut Penner (hp@de.ibm.com),
   5  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
   6  *
   7  *  Derived from "arch/i386/kernel/setup.c"
   8  *    Copyright (C) 1995, Linus Torvalds
   9  */
  10
  11 /*
  12  * This file handles the architecture-dependent parts of initialization
  13  */
  14
  15 #define KMSG_COMPONENT "setup"
  16 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  17
  18 #include <linux/errno.h>
  19 #include <linux/export.h>
  20 #include <linux/sched.h>
  21 #include <linux/sched/task.h>
  22 #include <linux/cpu.h>
  23 #include <linux/kernel.h>
  24 #include <linux/memblock.h>
  25 #include <linux/mm.h>
  26 #include <linux/stddef.h>
  27 #include <linux/unistd.h>
  28 #include <linux/ptrace.h>
  29 #include <linux/random.h>
  30 #include <linux/user.h>
  31 #include <linux/tty.h>
  32 #include <linux/ioport.h>
  33 #include <linux/delay.h>
  34 #include <linux/init.h>
  35 #include <linux/initrd.h>
  36 #include <linux/bootmem.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
  52 #include <asm/ipl.h>
  53 #include <asm/facility.h>
  54 #include <asm/smp.h>
  55 #include <asm/mmu_context.h>
  56 #include <asm/cpcmd.h>
  57 #include <asm/lowcore.h>
  58 #include <asm/nmi.h>
  59 #include <asm/irq.h>
  60 #include <asm/page.h>
  61 #include <asm/ptrace.h>
  62 #include <asm/sections.h>
  63 #include <asm/ebcdic.h>
  64 #include <asm/diag.h>
  65 #include <asm/os_info.h>
  66 #include <asm/sclp.h>
  67 #include <asm/sysinfo.h>
  68 #include <asm/numa.h>
  69 #include <asm/alternative.h>
  70 #include "entry.h"
  71
  72 /*
  73  * Machine setup..
  74  */
  75 unsigned int console_mode = 0;
  76 EXPORT_SYMBOL(console_mode);
  77
  78 unsigned int console_devno = -1;
  79 EXPORT_SYMBOL(console_devno);
  80
  81 unsigned int console_irq = -1;
  82 EXPORT_SYMBOL(console_irq);
  83
  84 unsigned long elf_hwcap __read_mostly = 0;
  85 char elf_platform[ELF_PLATFORM_SIZE];
  86
  87 unsigned long int_hwcap = 0;
  88
  89 int __initdata memory_end_set;
  90 unsigned long __initdata memory_end;
  91 unsigned long __initdata max_physmem_end;
  92
  93 unsigned long VMALLOC_START;
  94 EXPORT_SYMBOL(VMALLOC_START);
  95
  96 unsigned long VMALLOC_END;
  97 EXPORT_SYMBOL(VMALLOC_END);
  98
  99 struct page *vmemmap;
 100 EXPORT_SYMBOL(vmemmap);
 101
 102 unsigned long MODULES_VADDR;
 103 unsigned long MODULES_END;
 104
 105 /* An array with a pointer to the lowcore of every CPU. */
 106 struct lowcore *lowcore_ptr[NR_CPUS];
 107 EXPORT_SYMBOL(lowcore_ptr);
 108
 109 /*
 110  * This is set up by the setup-routine at boot-time
 111  * for S390 need to find out, what we have to setup
 112  * using address 0x10400 ...
 113  */
 114
 115 #include <asm/setup.h>
 116
 117 /*
 118  * condev= and conmode= setup parameter.
 119  */
 120
 121 static int __init condev_setup(char *str)
 122 {
 123         int vdev;
 124
 125         vdev = simple_strtoul(str, &str, 0);
 126         if (vdev >= 0 && vdev < 65536) {
 127                 console_devno = vdev;
 128                 console_irq = -1;
 129         }
 130         return 1;
 131 }
 132
 133 __setup("condev=", condev_setup);
 134
 135 static void __init set_preferred_console(void)
 136 {
 137         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 138                 add_preferred_console("ttyS", 0, NULL);
 139         else if (CONSOLE_IS_3270)
 140                 add_preferred_console("tty3270", 0, NULL);
 141         else if (CONSOLE_IS_VT220)
 142                 add_preferred_console("ttyS", 1, NULL);
 143         else if (CONSOLE_IS_HVC)
 144                 add_preferred_console("hvc", 0, NULL);
 145 }
 146
 147 static int __init conmode_setup(char *str)
 148 {
 149 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 150         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 151                 SET_CONSOLE_SCLP;
 152 #endif
 153 #if defined(CONFIG_TN3215_CONSOLE)
 154         if (strncmp(str, "3215", 5) == 0)
 155                 SET_CONSOLE_3215;
 156 #endif
 157 #if defined(CONFIG_TN3270_CONSOLE)
 158         if (strncmp(str, "3270", 5) == 0)
 159                 SET_CONSOLE_3270;
 160 #endif
 161         set_preferred_console();
 162         return 1;
 163 }
 164
 165 __setup("conmode=", conmode_setup);
 166
 167 static void __init conmode_default(void)
 168 {
 169         char query_buffer[1024];
 170         char *ptr;
 171
 172         if (MACHINE_IS_VM) {
 173                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 174                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 175                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 176                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 177                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 178                 ptr = strstr(query_buffer, "CONMODE");
 179                 /*
 180                  * Set the conmode to 3215 so that the device recognition
 181                  * will set the cu_type of the console to 3215. If the
 182                  * conmode is 3270 and we don't set it back then both
 183                  * 3215 and the 3270 driver will try to access the console
 184                  * device (3215 as console and 3270 as normal tty).
 185                  */
 186                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 187                 if (ptr == NULL) {
 188 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 189                         SET_CONSOLE_SCLP;
 190 #endif
 191                         return;
 192                 }
 193                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 194 #if defined(CONFIG_TN3270_CONSOLE)
 195                         SET_CONSOLE_3270;
 196 #elif defined(CONFIG_TN3215_CONSOLE)
 197                         SET_CONSOLE_3215;
 198 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 199                         SET_CONSOLE_SCLP;
 200 #endif
 201                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 202 #if defined(CONFIG_TN3215_CONSOLE)
 203                         SET_CONSOLE_3215;
 204 #elif defined(CONFIG_TN3270_CONSOLE)
 205                         SET_CONSOLE_3270;
 206 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 207                         SET_CONSOLE_SCLP;
 208 #endif
 209                 }
 210         } else if (MACHINE_IS_KVM) {
 211                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
 212                         SET_CONSOLE_VT220;
 213                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
 214                         SET_CONSOLE_SCLP;
 215                 else
 216                         SET_CONSOLE_HVC;
 217         } else {
 218 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 219                 SET_CONSOLE_SCLP;
 220 #endif
 221         }
 222 }
 223
 224 #ifdef CONFIG_CRASH_DUMP
 225 static void __init setup_zfcpdump(void)
 226 {
 227         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 228                 return;
 229         if (OLDMEM_BASE)
 230                 return;
 231         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 232         console_loglevel = 2;
 233 }
 234 #else
 235 static inline void setup_zfcpdump(void) {}
 236 #endif /* CONFIG_CRASH_DUMP */
 237
 238  /*
 239  * Reboot, halt and power_off stubs. They just call _machine_restart,
 240  * _machine_halt or _machine_power_off.
 241  */
 242
 243 void machine_restart(char *command)
 244 {
 245         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 246                 /*
 247                  * Only unblank the console if we are called in enabled
 248                  * context or a bust_spinlocks cleared the way for us.
 249                  */
 250                 console_unblank();
 251         _machine_restart(command);
 252 }
 253
 254 void machine_halt(void)
 255 {
 256         if (!in_interrupt() || oops_in_progress)
 257                 /*
 258                  * Only unblank the console if we are called in enabled
 259                  * context or a bust_spinlocks cleared the way for us.
 260                  */
 261                 console_unblank();
 262         _machine_halt();
 263 }
 264
 265 void machine_power_off(void)
 266 {
 267         if (!in_interrupt() || oops_in_progress)
 268                 /*
 269                  * Only unblank the console if we are called in enabled
 270                  * context or a bust_spinlocks cleared the way for us.
 271                  */
 272                 console_unblank();
 273         _machine_power_off();
 274 }
 275
 276 /*
 277  * Dummy power off function.
 278  */
 279 void (*pm_power_off)(void) = machine_power_off;
 280 EXPORT_SYMBOL_GPL(pm_power_off);
 281
 282 static int __init early_parse_mem(char *p)
 283 {
 284         memory_end = memparse(p, &p);
 285         memory_end &= PAGE_MASK;
 286         memory_end_set = 1;
 287         return 0;
 288 }
 289 early_param("mem", early_parse_mem);
 290
 291 static int __init parse_vmalloc(char *arg)
 292 {
 293         if (!arg)
 294                 return -EINVAL;
 295         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
 296         return 0;
 297 }
 298 early_param("vmalloc", parse_vmalloc);
 299
 300 void *restart_stack __section(.data);
 301
 302 static void __init setup_lowcore(void)
 303 {
 304         struct lowcore *lc;
 305
 306         /*
 307          * Setup lowcore for boot cpu
 308          */
 309         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 310         lc = memblock_virt_alloc_low(sizeof(*lc), sizeof(*lc));
 311         lc->restart_psw.mask = PSW_KERNEL_BITS;
 312         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 313         lc->external_new_psw.mask = PSW_KERNEL_BITS |
 314                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 315         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 316         lc->svc_new_psw.mask = PSW_KERNEL_BITS |
 317                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 318         lc->svc_new_psw.addr = (unsigned long) system_call;
 319         lc->program_new_psw.mask = PSW_KERNEL_BITS |
 320                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 321         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 322         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 323         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 324         lc->io_new_psw.mask = PSW_KERNEL_BITS |
 325                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 326         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 327         lc->clock_comparator = clock_comparator_max;
 328         lc->kernel_stack = ((unsigned long) &init_thread_union)
 329                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 330         lc->async_stack = (unsigned long)
 331                 memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE)
 332                 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 333         lc->panic_stack = (unsigned long)
 334                 memblock_virt_alloc(PAGE_SIZE, PAGE_SIZE)
 335                 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 336         lc->current_task = (unsigned long)&init_task;
 337         lc->lpp = LPP_MAGIC;
 338         lc->machine_flags = S390_lowcore.machine_flags;
 339         lc->preempt_count = S390_lowcore.preempt_count;
 340         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 341         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 342                MAX_FACILITY_BIT/8);
 343         nmi_alloc_boot_cpu(lc);
 344         vdso_alloc_boot_cpu(lc);
 345         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 346         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 347         lc->exit_timer = S390_lowcore.exit_timer;
 348         lc->user_timer = S390_lowcore.user_timer;
 349         lc->system_timer = S390_lowcore.system_timer;
 350         lc->steal_timer = S390_lowcore.steal_timer;
 351         lc->last_update_timer = S390_lowcore.last_update_timer;
 352         lc->last_update_clock = S390_lowcore.last_update_clock;
 353
 354         restart_stack = memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE);
 355         restart_stack += ASYNC_SIZE;
 356
 357         /*
 358          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 359          * restart data to the absolute zero lowcore. This is necessary if
 360          * PSW restart is done on an offline CPU that has lowcore zero.
 361          */
 362         lc->restart_stack = (unsigned long) restart_stack;
 363         lc->restart_fn = (unsigned long) do_restart;
 364         lc->restart_data = 0;
 365         lc->restart_source = -1UL;
 366
 367         /* Setup absolute zero lowcore */
 368         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 369         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 370         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 371         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 372         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 373
 374 #ifdef CONFIG_SMP
 375         lc->spinlock_lockval = arch_spin_lockval(0);
 376         lc->spinlock_index = 0;
 377         arch_spin_lock_setup(0);
 378 #endif
 379
 380         set_prefix((u32)(unsigned long) lc);
 381         lowcore_ptr[0] = lc;
 382 }
 383
 384 static struct resource code_resource = {
 385         .name  = "Kernel code",
 386         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 387 };
 388
 389 static struct resource data_resource = {
 390         .name = "Kernel data",
 391         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 392 };
 393
 394 static struct resource bss_resource = {
 395         .name = "Kernel bss",
 396         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 397 };
 398
 399 static struct resource __initdata *standard_resources[] = {
 400         &code_resource,
 401         &data_resource,
 402         &bss_resource,
 403 };
 404
 405 static void __init setup_resources(void)
 406 {
 407         struct resource *res, *std_res, *sub_res;
 408         struct memblock_region *reg;
 409         int j;
 410
 411         code_resource.start = (unsigned long) &_text;
 412         code_resource.end = (unsigned long) &_etext - 1;
 413         data_resource.start = (unsigned long) &_etext;
 414         data_resource.end = (unsigned long) &_edata - 1;
 415         bss_resource.start = (unsigned long) &__bss_start;
 416         bss_resource.end = (unsigned long) &__bss_stop - 1;
 417
 418         for_each_memblock(memory, reg) {
 419                 res = memblock_virt_alloc(sizeof(*res), 8);
 420                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 421
 422                 res->name = "System RAM";
 423                 res->start = reg->base;
 424                 res->end = reg->base + reg->size - 1;
 425                 request_resource(&iomem_resource, res);
 426
 427                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 428                         std_res = standard_resources[j];
 429                         if (std_res->start < res->start ||
 430                             std_res->start > res->end)
 431                                 continue;
 432                         if (std_res->end > res->end) {
 433                                 sub_res = memblock_virt_alloc(sizeof(*sub_res), 8);
 434                                 *sub_res = *std_res;
 435                                 sub_res->end = res->end;
 436                                 std_res->start = res->end + 1;
 437                                 request_resource(res, sub_res);
 438                         } else {
 439                                 request_resource(res, std_res);
 440                         }
 441                 }
 442         }
 443 #ifdef CONFIG_CRASH_DUMP
 444         /*
 445          * Re-add removed crash kernel memory as reserved memory. This makes
 446          * sure it will be mapped with the identity mapping and struct pages
 447          * will be created, so it can be resized later on.
 448          * However add it later since the crash kernel resource should not be
 449          * part of the System RAM resource.
 450          */
 451         if (crashk_res.end) {
 452                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 453                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 454                 insert_resource(&iomem_resource, &crashk_res);
 455         }
 456 #endif
 457 }
 458
 459 static void __init setup_memory_end(void)
 460 {
 461         unsigned long vmax, vmalloc_size, tmp;
 462
 463         /* Choose kernel address space layout: 2, 3, or 4 levels. */
 464         vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
 465         tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
 466         tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
 467         if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
 468                 vmax = _REGION2_SIZE; /* 3-level kernel page table */
 469         else
 470                 vmax = _REGION1_SIZE; /* 4-level kernel page table */
 471         /* module area is at the end of the kernel address space. */
 472         MODULES_END = vmax;
 473         MODULES_VADDR = MODULES_END - MODULES_LEN;
 474         VMALLOC_END = MODULES_VADDR;
 475         VMALLOC_START = vmax - vmalloc_size;
 476
 477         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 478         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 479         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 480         tmp = SECTION_ALIGN_UP(tmp);
 481         tmp = VMALLOC_START - tmp * sizeof(struct page);
 482         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 483         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 484         vmemmap = (struct page *) tmp;
 485
 486         /* Take care that memory_end is set and <= vmemmap */
 487         memory_end = min(memory_end ?: max_physmem_end, tmp);
 488         max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 489         memblock_remove(memory_end, ULONG_MAX);
 490
 491         pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
 492 }
 493
 494 #ifdef CONFIG_CRASH_DUMP
 495
 496 /*
 497  * When kdump is enabled, we have to ensure that no memory from
 498  * the area [0 - crashkernel memory size] and
 499  * [crashk_res.start - crashk_res.end] is set offline.
 500  */
 501 static int kdump_mem_notifier(struct notifier_block *nb,
 502                               unsigned long action, void *data)
 503 {
 504         struct memory_notify *arg = data;
 505
 506         if (action != MEM_GOING_OFFLINE)
 507                 return NOTIFY_OK;
 508         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 509                 return NOTIFY_BAD;
 510         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 511                 return NOTIFY_OK;
 512         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 513                 return NOTIFY_OK;
 514         return NOTIFY_BAD;
 515 }
 516
 517 static struct notifier_block kdump_mem_nb = {
 518         .notifier_call = kdump_mem_notifier,
 519 };
 520
 521 #endif
 522
 523 /*
 524  * Make sure that the area behind memory_end is protected
 525  */
 526 static void reserve_memory_end(void)
 527 {
 528 #ifdef CONFIG_CRASH_DUMP
 529         if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
 530             !OLDMEM_BASE && sclp.hsa_size) {
 531                 memory_end = sclp.hsa_size;
 532                 memory_end &= PAGE_MASK;
 533                 memory_end_set = 1;
 534         }
 535 #endif
 536         if (!memory_end_set)
 537                 return;
 538         memblock_reserve(memory_end, ULONG_MAX);
 539 }
 540
 541 /*
 542  * Make sure that oldmem, where the dump is stored, is protected
 543  */
 544 static void reserve_oldmem(void)
 545 {
 546 #ifdef CONFIG_CRASH_DUMP
 547         if (OLDMEM_BASE)
 548                 /* Forget all memory above the running kdump system */
 549                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 550 #endif
 551 }
 552
 553 /*
 554  * Make sure that oldmem, where the dump is stored, is protected
 555  */
 556 static void remove_oldmem(void)
 557 {
 558 #ifdef CONFIG_CRASH_DUMP
 559         if (OLDMEM_BASE)
 560                 /* Forget all memory above the running kdump system */
 561                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 562 #endif
 563 }
 564
 565 /*
 566  * Reserve memory for kdump kernel to be loaded with kexec
 567  */
 568 static void __init reserve_crashkernel(void)
 569 {
 570 #ifdef CONFIG_CRASH_DUMP
 571         unsigned long long crash_base, crash_size;
 572         phys_addr_t low, high;
 573         int rc;
 574
 575         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 576                                &crash_base);
 577
 578         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 579         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 580         if (rc || crash_size == 0)
 581                 return;
 582
 583         if (memblock.memory.regions[0].size < crash_size) {
 584                 pr_info("crashkernel reservation failed: %s\n",
 585                         "first memory chunk must be at least crashkernel size");
 586                 return;
 587         }
 588
 589         low = crash_base ?: OLDMEM_BASE;
 590         high = low + crash_size;
 591         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 592                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 593                 crash_base = low;
 594         } else {
 595                 /* Find suitable area in free memory */
 596                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 597                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 598
 599                 if (crash_base && crash_base < low) {
 600                         pr_info("crashkernel reservation failed: %s\n",
 601                                 "crash_base too low");
 602                         return;
 603                 }
 604                 low = crash_base ?: low;
 605                 crash_base = memblock_find_in_range(low, high, crash_size,
 606                                                     KEXEC_CRASH_MEM_ALIGN);
 607         }
 608
 609         if (!crash_base) {
 610                 pr_info("crashkernel reservation failed: %s\n",
 611                         "no suitable area found");
 612                 return;
 613         }
 614
 615         if (register_memory_notifier(&kdump_mem_nb))
 616                 return;
 617
 618         if (!OLDMEM_BASE && MACHINE_IS_VM)
 619                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 620         crashk_res.start = crash_base;
 621         crashk_res.end = crash_base + crash_size - 1;
 622         memblock_remove(crash_base, crash_size);
 623         pr_info("Reserving %lluMB of memory at %lluMB "
 624                 "for crashkernel (System RAM: %luMB)\n",
 625                 crash_size >> 20, crash_base >> 20,
 626                 (unsigned long)memblock.memory.total_size >> 20);
 627         os_info_crashkernel_add(crash_base, crash_size);
 628 #endif
 629 }
 630
 631 /*
 632  * Reserve the initrd from being used by memblock
 633  */
 634 static void __init reserve_initrd(void)
 635 {
 636 #ifdef CONFIG_BLK_DEV_INITRD
 637         if (!INITRD_START || !INITRD_SIZE)
 638                 return;
 639         initrd_start = INITRD_START;
 640         initrd_end = initrd_start + INITRD_SIZE;
 641         memblock_reserve(INITRD_START, INITRD_SIZE);
 642 #endif
 643 }
 644
 645 /*
 646  * Check for initrd being in usable memory
 647  */
 648 static void __init check_initrd(void)
 649 {
 650 #ifdef CONFIG_BLK_DEV_INITRD
 651         if (INITRD_START && INITRD_SIZE &&
 652             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 653                 pr_err("The initial RAM disk does not fit into the memory\n");
 654                 memblock_free(INITRD_START, INITRD_SIZE);
 655                 initrd_start = initrd_end = 0;
 656         }
 657 #endif
 658 }
 659
 660 /*
 661  * Reserve memory used for lowcore/command line/kernel image.
 662  */
 663 static void __init reserve_kernel(void)
 664 {
 665         unsigned long start_pfn = PFN_UP(__pa(&_end));
 666
 667 #ifdef CONFIG_DMA_API_DEBUG
 668         /*
 669          * DMA_API_DEBUG code stumbles over addresses from the
 670          * range [_ehead, _stext]. Mark the memory as reserved
 671          * so it is not used for CONFIG_DMA_API_DEBUG=y.
 672          */
 673         memblock_reserve(0, PFN_PHYS(start_pfn));
 674 #else
 675         memblock_reserve(0, (unsigned long)_ehead);
 676         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 677                          - (unsigned long)_stext);
 678 #endif
 679 }
 680
 681 static void __init setup_memory(void)
 682 {
 683         struct memblock_region *reg;
 684
 685         /*
 686          * Init storage key for present memory
 687          */
 688         for_each_memblock(memory, reg) {
 689                 storage_key_init_range(reg->base, reg->base + reg->size);
 690         }
 691         psw_set_key(PAGE_DEFAULT_KEY);
 692
 693         /* Only cosmetics */
 694         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 695 }
 696
 697 /*
 698  * Setup hardware capabilities.
 699  */
 700 static int __init setup_hwcaps(void)
 701 {
 702         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 703         struct cpuid cpu_id;
 704         int i;
 705
 706         /*
 707          * The store facility list bits numbers as found in the principles
 708          * of operation are numbered with bit 1UL<<31 as number 0 to
 709          * bit 1UL<<0 as number 31.
 710          *   Bit 0: instructions named N3, "backported" to esa-mode
 711          *   Bit 2: z/Architecture mode is active
 712          *   Bit 7: the store-facility-list-extended facility is installed
 713          *   Bit 17: the message-security assist is installed
 714          *   Bit 19: the long-displacement facility is installed
 715          *   Bit 21: the extended-immediate facility is installed
 716          *   Bit 22: extended-translation facility 3 is installed
 717          *   Bit 30: extended-translation facility 3 enhancement facility
 718          * These get translated to:
 719          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 720          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 721          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 722          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 723          */
 724         for (i = 0; i < 6; i++)
 725                 if (test_facility(stfl_bits[i]))
 726                         elf_hwcap |= 1UL << i;
 727
 728         if (test_facility(22) && test_facility(30))
 729                 elf_hwcap |= HWCAP_S390_ETF3EH;
 730
 731         /*
 732          * Check for additional facilities with store-facility-list-extended.
 733          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 734          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 735          * as stored by stfl, bits 32-xxx contain additional facilities.
 736          * How many facility words are stored depends on the number of
 737          * doublewords passed to the instruction. The additional facilities
 738          * are:
 739          *   Bit 42: decimal floating point facility is installed
 740          *   Bit 44: perform floating point operation facility is installed
 741          * translated to:
 742          *   HWCAP_S390_DFP bit 6 (42 && 44).
 743          */
 744         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 745                 elf_hwcap |= HWCAP_S390_DFP;
 746
 747         /*
 748          * Huge page support HWCAP_S390_HPAGE is bit 7.
 749          */
 750         if (MACHINE_HAS_EDAT1)
 751                 elf_hwcap |= HWCAP_S390_HPAGE;
 752
 753         /*
 754          * 64-bit register support for 31-bit processes
 755          * HWCAP_S390_HIGH_GPRS is bit 9.
 756          */
 757         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 758
 759         /*
 760          * Transactional execution support HWCAP_S390_TE is bit 10.
 761          */
 762         if (MACHINE_HAS_TE)
 763                 elf_hwcap |= HWCAP_S390_TE;
 764
 765         /*
 766          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 767          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 768          * instead of facility bit 129.
 769          */
 770         if (MACHINE_HAS_VX) {
 771                 elf_hwcap |= HWCAP_S390_VXRS;
 772                 if (test_facility(134))
 773                         elf_hwcap |= HWCAP_S390_VXRS_EXT;
 774                 if (test_facility(135))
 775                         elf_hwcap |= HWCAP_S390_VXRS_BCD;
 776         }
 777
 778         /*
 779          * Guarded storage support HWCAP_S390_GS is bit 12.
 780          */
 781         if (MACHINE_HAS_GS)
 782                 elf_hwcap |= HWCAP_S390_GS;
 783
 784         get_cpu_id(&cpu_id);
 785         add_device_randomness(&cpu_id, sizeof(cpu_id));
 786         switch (cpu_id.machine) {
 787         case 0x2064:
 788         case 0x2066:
 789         default:        /* Use "z900" as default for 64 bit kernels. */
 790                 strcpy(elf_platform, "z900");
 791                 break;
 792         case 0x2084:
 793         case 0x2086:
 794                 strcpy(elf_platform, "z990");
 795                 break;
 796         case 0x2094:
 797         case 0x2096:
 798                 strcpy(elf_platform, "z9-109");
 799                 break;
 800         case 0x2097:
 801         case 0x2098:
 802                 strcpy(elf_platform, "z10");
 803                 break;
 804         case 0x2817:
 805         case 0x2818:
 806                 strcpy(elf_platform, "z196");
 807                 break;
 808         case 0x2827:
 809         case 0x2828:
 810                 strcpy(elf_platform, "zEC12");
 811                 break;
 812         case 0x2964:
 813         case 0x2965:
 814                 strcpy(elf_platform, "z13");
 815                 break;
 816         case 0x3906:
 817                 strcpy(elf_platform, "z14");
 818                 break;
 819         }
 820
 821         /*
 822          * Virtualization support HWCAP_INT_SIE is bit 0.
 823          */
 824         if (sclp.has_sief2)
 825                 int_hwcap |= HWCAP_INT_SIE;
 826
 827         return 0;
 828 }
 829 arch_initcall(setup_hwcaps);
 830
 831 /*
 832  * Add system information as device randomness
 833  */
 834 static void __init setup_randomness(void)
 835 {
 836         struct sysinfo_3_2_2 *vmms;
 837
 838         vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE);
 839         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 840                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
 841         memblock_free((unsigned long) vmms, PAGE_SIZE);
 842 }
 843
 844 /*
 845  * Find the correct size for the task_struct. This depends on
 846  * the size of the struct fpu at the end of the thread_struct
 847  * which is embedded in the task_struct.
 848  */
 849 static void __init setup_task_size(void)
 850 {
 851         int task_size = sizeof(struct task_struct);
 852
 853         if (!MACHINE_HAS_VX) {
 854                 task_size -= sizeof(__vector128) * __NUM_VXRS;
 855                 task_size += sizeof(freg_t) * __NUM_FPRS;
 856         }
 857         arch_task_struct_size = task_size;
 858 }
 859
 860 /*
 861  * Setup function called from init/main.c just after the banner
 862  * was printed.
 863  */
 864
 865 void __init setup_arch(char **cmdline_p)
 866 {
 867         /*
 868          * print what head.S has found out about the machine
 869          */
 870         if (MACHINE_IS_VM)
 871                 pr_info("Linux is running as a z/VM "
 872                         "guest operating system in 64-bit mode\n");
 873         else if (MACHINE_IS_KVM)
 874                 pr_info("Linux is running under KVM in 64-bit mode\n");
 875         else if (MACHINE_IS_LPAR)
 876                 pr_info("Linux is running natively in 64-bit mode\n");
 877
 878         /* Have one command line that is parsed and saved in /proc/cmdline */
 879         /* boot_command_line has been already set up in early.c */
 880         *cmdline_p = boot_command_line;
 881
 882         ROOT_DEV = Root_RAM0;
 883
 884         /* Is init_mm really needed? */
 885         init_mm.start_code = PAGE_OFFSET;
 886         init_mm.end_code = (unsigned long) &_etext;
 887         init_mm.end_data = (unsigned long) &_edata;
 888         init_mm.brk = (unsigned long) &_end;
 889
 890         parse_early_param();
 891 #ifdef CONFIG_CRASH_DUMP
 892         /* Deactivate elfcorehdr= kernel parameter */
 893         elfcorehdr_addr = ELFCORE_ADDR_MAX;
 894 #endif
 895
 896         os_info_init();
 897         setup_ipl();
 898         setup_task_size();
 899
 900         /* Do some memory reservations *before* memory is added to memblock */
 901         reserve_memory_end();
 902         reserve_oldmem();
 903         reserve_kernel();
 904         reserve_initrd();
 905         memblock_allow_resize();
 906
 907         /* Get information about *all* installed memory */
 908         detect_memory_memblock();
 909
 910         remove_oldmem();
 911
 912         /*
 913          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 914          * extra checks that HOLES_IN_ZONE would require.
 915          *
 916          * Is this still required?
 917          */
 918         memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
 919
 920         setup_memory_end();
 921         setup_memory();
 922         dma_contiguous_reserve(memory_end);
 923         vmcp_cma_reserve();
 924
 925         check_initrd();
 926         reserve_crashkernel();
 927 #ifdef CONFIG_CRASH_DUMP
 928         /*
 929          * Be aware that smp_save_dump_cpus() triggers a system reset.
 930          * Therefore CPU and device initialization should be done afterwards.
 931          */
 932         smp_save_dump_cpus();
 933 #endif
 934
 935         setup_resources();
 936         setup_lowcore();
 937         smp_fill_possible_mask();
 938         cpu_detect_mhz_feature();
 939         cpu_init();
 940         numa_setup();
 941         smp_detect_cpus();
 942         topology_init_early();
 943
 944         /*
 945          * Create kernel page tables and switch to virtual addressing.
 946          */
 947         paging_init();
 948
 949         /* Setup default console */
 950         conmode_default();
 951         set_preferred_console();
 952
 953         apply_alternative_instructions();
 954
 955         /* Setup zfcpdump support */
 956         setup_zfcpdump();
 957
 958         /* Add system specific data to the random pool */
 959         setup_randomness();
 960 }