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/kernel.h>
  22 #include <linux/memblock.h>
  23 #include <linux/mm.h>
  24 #include <linux/stddef.h>
  25 #include <linux/unistd.h>
  26 #include <linux/ptrace.h>
  27 #include <linux/random.h>
  28 #include <linux/user.h>
  29 #include <linux/tty.h>
  30 #include <linux/ioport.h>
  31 #include <linux/delay.h>
  32 #include <linux/init.h>
  33 #include <linux/initrd.h>
  34 #include <linux/bootmem.h>
  35 #include <linux/root_dev.h>
  36 #include <linux/console.h>
  37 #include <linux/kernel_stat.h>
  38 #include <linux/device.h>
  39 #include <linux/notifier.h>
  40 #include <linux/pfn.h>
  41 #include <linux/ctype.h>
  42 #include <linux/reboot.h>
  43 #include <linux/topology.h>
  44 #include <linux/kexec.h>
  45 #include <linux/crash_dump.h>
  46 #include <linux/memory.h>
  47 #include <linux/compat.h>
  48
  49 #include <asm/ipl.h>
  50 #include <asm/facility.h>
  51 #include <asm/smp.h>
  52 #include <asm/mmu_context.h>
  53 #include <asm/cpcmd.h>
  54 #include <asm/lowcore.h>
  55 #include <asm/irq.h>
  56 #include <asm/page.h>
  57 #include <asm/ptrace.h>
  58 #include <asm/sections.h>
  59 #include <asm/ebcdic.h>
  60 #include <asm/kvm_virtio.h>
  61 #include <asm/diag.h>
  62 #include <asm/os_info.h>
  63 #include <asm/sclp.h>
  64 #include <asm/sysinfo.h>
  65 #include <asm/numa.h>
  66 #include "entry.h"
  67
  68 /*
  69  * Machine setup..
  70  */
  71 unsigned int console_mode = 0;
  72 EXPORT_SYMBOL(console_mode);
  73
  74 unsigned int console_devno = -1;
  75 EXPORT_SYMBOL(console_devno);
  76
  77 unsigned int console_irq = -1;
  78 EXPORT_SYMBOL(console_irq);
  79
  80 unsigned long elf_hwcap __read_mostly = 0;
  81 char elf_platform[ELF_PLATFORM_SIZE];
  82
  83 unsigned long int_hwcap = 0;
  84
  85 int __initdata memory_end_set;
  86 unsigned long __initdata memory_end;
  87 unsigned long __initdata max_physmem_end;
  88
  89 unsigned long VMALLOC_START;
  90 EXPORT_SYMBOL(VMALLOC_START);
  91
  92 unsigned long VMALLOC_END;
  93 EXPORT_SYMBOL(VMALLOC_END);
  94
  95 struct page *vmemmap;
  96 EXPORT_SYMBOL(vmemmap);
  97
  98 unsigned long MODULES_VADDR;
  99 unsigned long MODULES_END;
 100
 101 /* An array with a pointer to the lowcore of every CPU. */
 102 struct lowcore *lowcore_ptr[NR_CPUS];
 103 EXPORT_SYMBOL(lowcore_ptr);
 104
 105 /*
 106  * This is set up by the setup-routine at boot-time
 107  * for S390 need to find out, what we have to setup
 108  * using address 0x10400 ...
 109  */
 110
 111 #include <asm/setup.h>
 112
 113 /*
 114  * condev= and conmode= setup parameter.
 115  */
 116
 117 static int __init condev_setup(char *str)
 118 {
 119         int vdev;
 120
 121         vdev = simple_strtoul(str, &str, 0);
 122         if (vdev >= 0 && vdev < 65536) {
 123                 console_devno = vdev;
 124                 console_irq = -1;
 125         }
 126         return 1;
 127 }
 128
 129 __setup("condev=", condev_setup);
 130
 131 static void __init set_preferred_console(void)
 132 {
 133         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 134                 add_preferred_console("ttyS", 0, NULL);
 135         else if (CONSOLE_IS_3270)
 136                 add_preferred_console("tty3270", 0, NULL);
 137         else if (CONSOLE_IS_VT220)
 138                 add_preferred_console("ttyS", 1, NULL);
 139         else if (CONSOLE_IS_HVC)
 140                 add_preferred_console("hvc", 0, NULL);
 141 }
 142
 143 static int __init conmode_setup(char *str)
 144 {
 145 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 146         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 147                 SET_CONSOLE_SCLP;
 148 #endif
 149 #if defined(CONFIG_TN3215_CONSOLE)
 150         if (strncmp(str, "3215", 5) == 0)
 151                 SET_CONSOLE_3215;
 152 #endif
 153 #if defined(CONFIG_TN3270_CONSOLE)
 154         if (strncmp(str, "3270", 5) == 0)
 155                 SET_CONSOLE_3270;
 156 #endif
 157         set_preferred_console();
 158         return 1;
 159 }
 160
 161 __setup("conmode=", conmode_setup);
 162
 163 static void __init conmode_default(void)
 164 {
 165         char query_buffer[1024];
 166         char *ptr;
 167
 168         if (MACHINE_IS_VM) {
 169                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 170                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 171                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 172                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 173                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 174                 ptr = strstr(query_buffer, "CONMODE");
 175                 /*
 176                  * Set the conmode to 3215 so that the device recognition
 177                  * will set the cu_type of the console to 3215. If the
 178                  * conmode is 3270 and we don't set it back then both
 179                  * 3215 and the 3270 driver will try to access the console
 180                  * device (3215 as console and 3270 as normal tty).
 181                  */
 182                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 183                 if (ptr == NULL) {
 184 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 185                         SET_CONSOLE_SCLP;
 186 #endif
 187                         return;
 188                 }
 189                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 190 #if defined(CONFIG_TN3270_CONSOLE)
 191                         SET_CONSOLE_3270;
 192 #elif defined(CONFIG_TN3215_CONSOLE)
 193                         SET_CONSOLE_3215;
 194 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 195                         SET_CONSOLE_SCLP;
 196 #endif
 197                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 198 #if defined(CONFIG_TN3215_CONSOLE)
 199                         SET_CONSOLE_3215;
 200 #elif defined(CONFIG_TN3270_CONSOLE)
 201                         SET_CONSOLE_3270;
 202 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 203                         SET_CONSOLE_SCLP;
 204 #endif
 205                 }
 206         } else if (MACHINE_IS_KVM) {
 207                 if (sclp.has_vt220 &&
 208                     config_enabled(CONFIG_SCLP_VT220_CONSOLE))
 209                         SET_CONSOLE_VT220;
 210                 else if (sclp.has_linemode &&
 211                          config_enabled(CONFIG_SCLP_CONSOLE))
 212                         SET_CONSOLE_SCLP;
 213                 else
 214                         SET_CONSOLE_HVC;
 215         } else {
 216 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 217                 SET_CONSOLE_SCLP;
 218 #endif
 219         }
 220 }
 221
 222 #ifdef CONFIG_CRASH_DUMP
 223 static void __init setup_zfcpdump(void)
 224 {
 225         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 226                 return;
 227         if (OLDMEM_BASE)
 228                 return;
 229         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 230         console_loglevel = 2;
 231 }
 232 #else
 233 static inline void setup_zfcpdump(void) {}
 234 #endif /* CONFIG_CRASH_DUMP */
 235
 236  /*
 237  * Reboot, halt and power_off stubs. They just call _machine_restart,
 238  * _machine_halt or _machine_power_off.
 239  */
 240
 241 void machine_restart(char *command)
 242 {
 243         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 244                 /*
 245                  * Only unblank the console if we are called in enabled
 246                  * context or a bust_spinlocks cleared the way for us.
 247                  */
 248                 console_unblank();
 249         _machine_restart(command);
 250 }
 251
 252 void machine_halt(void)
 253 {
 254         if (!in_interrupt() || oops_in_progress)
 255                 /*
 256                  * Only unblank the console if we are called in enabled
 257                  * context or a bust_spinlocks cleared the way for us.
 258                  */
 259                 console_unblank();
 260         _machine_halt();
 261 }
 262
 263 void machine_power_off(void)
 264 {
 265         if (!in_interrupt() || oops_in_progress)
 266                 /*
 267                  * Only unblank the console if we are called in enabled
 268                  * context or a bust_spinlocks cleared the way for us.
 269                  */
 270                 console_unblank();
 271         _machine_power_off();
 272 }
 273
 274 /*
 275  * Dummy power off function.
 276  */
 277 void (*pm_power_off)(void) = machine_power_off;
 278 EXPORT_SYMBOL_GPL(pm_power_off);
 279
 280 static int __init early_parse_mem(char *p)
 281 {
 282         memory_end = memparse(p, &p);
 283         memory_end &= PAGE_MASK;
 284         memory_end_set = 1;
 285         return 0;
 286 }
 287 early_param("mem", early_parse_mem);
 288
 289 static int __init parse_vmalloc(char *arg)
 290 {
 291         if (!arg)
 292                 return -EINVAL;
 293         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
 294         return 0;
 295 }
 296 early_param("vmalloc", parse_vmalloc);
 297
 298 void *restart_stack __section(.data);
 299
 300 static void __init setup_lowcore(void)
 301 {
 302         struct lowcore *lc;
 303
 304         /*
 305          * Setup lowcore for boot cpu
 306          */
 307         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * 4096);
 308         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 309         lc->restart_psw.mask = PSW_KERNEL_BITS;
 310         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 311         lc->external_new_psw.mask = PSW_KERNEL_BITS |
 312                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 313         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 314         lc->svc_new_psw.mask = PSW_KERNEL_BITS |
 315                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 316         lc->svc_new_psw.addr = (unsigned long) system_call;
 317         lc->program_new_psw.mask = PSW_KERNEL_BITS |
 318                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 319         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 320         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 321         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 322         lc->io_new_psw.mask = PSW_KERNEL_BITS |
 323                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 324         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 325         lc->clock_comparator = -1ULL;
 326         lc->kernel_stack = ((unsigned long) &init_thread_union)
 327                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 328         lc->async_stack = (unsigned long)
 329                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
 330                 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 331         lc->panic_stack = (unsigned long)
 332                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
 333                 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 334         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 335         lc->thread_info = (unsigned long) &init_thread_union;
 336         lc->lpp = LPP_MAGIC;
 337         lc->machine_flags = S390_lowcore.machine_flags;
 338         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 339         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 340                MAX_FACILITY_BIT/8);
 341         if (MACHINE_HAS_VX)
 342                 lc->vector_save_area_addr =
 343                         (unsigned long) &lc->vector_save_area;
 344         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
 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 = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
 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 #endif
 377
 378         set_prefix((u32)(unsigned long) lc);
 379         lowcore_ptr[0] = lc;
 380 }
 381
 382 static struct resource code_resource = {
 383         .name  = "Kernel code",
 384         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 385 };
 386
 387 static struct resource data_resource = {
 388         .name = "Kernel data",
 389         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 390 };
 391
 392 static struct resource bss_resource = {
 393         .name = "Kernel bss",
 394         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 395 };
 396
 397 static struct resource __initdata *standard_resources[] = {
 398         &code_resource,
 399         &data_resource,
 400         &bss_resource,
 401 };
 402
 403 static void __init setup_resources(void)
 404 {
 405         struct resource *res, *std_res, *sub_res;
 406         struct memblock_region *reg;
 407         int j;
 408
 409         code_resource.start = (unsigned long) &_text;
 410         code_resource.end = (unsigned long) &_etext - 1;
 411         data_resource.start = (unsigned long) &_etext;
 412         data_resource.end = (unsigned long) &_edata - 1;
 413         bss_resource.start = (unsigned long) &__bss_start;
 414         bss_resource.end = (unsigned long) &__bss_stop - 1;
 415
 416         for_each_memblock(memory, reg) {
 417                 res = alloc_bootmem_low(sizeof(*res));
 418                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 419
 420                 res->name = "System RAM";
 421                 res->start = reg->base;
 422                 res->end = reg->base + reg->size - 1;
 423                 request_resource(&iomem_resource, res);
 424
 425                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 426                         std_res = standard_resources[j];
 427                         if (std_res->start < res->start ||
 428                             std_res->start > res->end)
 429                                 continue;
 430                         if (std_res->end > res->end) {
 431                                 sub_res = alloc_bootmem_low(sizeof(*sub_res));
 432                                 *sub_res = *std_res;
 433                                 sub_res->end = res->end;
 434                                 std_res->start = res->end + 1;
 435                                 request_resource(res, sub_res);
 436                         } else {
 437                                 request_resource(res, std_res);
 438                         }
 439                 }
 440         }
 441 #ifdef CONFIG_CRASH_DUMP
 442         /*
 443          * Re-add removed crash kernel memory as reserved memory. This makes
 444          * sure it will be mapped with the identity mapping and struct pages
 445          * will be created, so it can be resized later on.
 446          * However add it later since the crash kernel resource should not be
 447          * part of the System RAM resource.
 448          */
 449         if (crashk_res.end) {
 450                 memblock_add(crashk_res.start, resource_size(&crashk_res));
 451                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 452                 insert_resource(&iomem_resource, &crashk_res);
 453         }
 454 #endif
 455 }
 456
 457 static void __init setup_memory_end(void)
 458 {
 459         unsigned long vmax, vmalloc_size, tmp;
 460
 461         /* Choose kernel address space layout: 2, 3, or 4 levels. */
 462         vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
 463         tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
 464         tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
 465         if (tmp + vmalloc_size + MODULES_LEN <= (1UL << 42))
 466                 vmax = 1UL << 42;       /* 3-level kernel page table */
 467         else
 468                 vmax = 1UL << 53;       /* 4-level kernel page table */
 469         /* module area is at the end of the kernel address space. */
 470         MODULES_END = vmax;
 471         MODULES_VADDR = MODULES_END - MODULES_LEN;
 472         VMALLOC_END = MODULES_VADDR;
 473         VMALLOC_START = vmax - vmalloc_size;
 474
 475         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 476         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 477         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 478         tmp = SECTION_ALIGN_UP(tmp);
 479         tmp = VMALLOC_START - tmp * sizeof(struct page);
 480         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 481         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 482         vmemmap = (struct page *) tmp;
 483
 484         /* Take care that memory_end is set and <= vmemmap */
 485         memory_end = min(memory_end ?: max_physmem_end, tmp);
 486         max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 487         memblock_remove(memory_end, ULONG_MAX);
 488
 489         pr_notice("Max memory size: %luMB\n", memory_end >> 20);
 490 }
 491
 492 static void __init setup_vmcoreinfo(void)
 493 {
 494         mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
 495 }
 496
 497 #ifdef CONFIG_CRASH_DUMP
 498
 499 /*
 500  * When kdump is enabled, we have to ensure that no memory from
 501  * the area [0 - crashkernel memory size] and
 502  * [crashk_res.start - crashk_res.end] is set offline.
 503  */
 504 static int kdump_mem_notifier(struct notifier_block *nb,
 505                               unsigned long action, void *data)
 506 {
 507         struct memory_notify *arg = data;
 508
 509         if (action != MEM_GOING_OFFLINE)
 510                 return NOTIFY_OK;
 511         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 512                 return NOTIFY_BAD;
 513         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 514                 return NOTIFY_OK;
 515         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 516                 return NOTIFY_OK;
 517         return NOTIFY_BAD;
 518 }
 519
 520 static struct notifier_block kdump_mem_nb = {
 521         .notifier_call = kdump_mem_notifier,
 522 };
 523
 524 #endif
 525
 526 /*
 527  * Make sure that the area behind memory_end is protected
 528  */
 529 static void reserve_memory_end(void)
 530 {
 531 #ifdef CONFIG_CRASH_DUMP
 532         if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
 533             !OLDMEM_BASE && sclp.hsa_size) {
 534                 memory_end = sclp.hsa_size;
 535                 memory_end &= PAGE_MASK;
 536                 memory_end_set = 1;
 537         }
 538 #endif
 539         if (!memory_end_set)
 540                 return;
 541         memblock_reserve(memory_end, ULONG_MAX);
 542 }
 543
 544 /*
 545  * Make sure that oldmem, where the dump is stored, is protected
 546  */
 547 static void reserve_oldmem(void)
 548 {
 549 #ifdef CONFIG_CRASH_DUMP
 550         if (OLDMEM_BASE)
 551                 /* Forget all memory above the running kdump system */
 552                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 553 #endif
 554 }
 555
 556 /*
 557  * Make sure that oldmem, where the dump is stored, is protected
 558  */
 559 static void remove_oldmem(void)
 560 {
 561 #ifdef CONFIG_CRASH_DUMP
 562         if (OLDMEM_BASE)
 563                 /* Forget all memory above the running kdump system */
 564                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 565 #endif
 566 }
 567
 568 /*
 569  * Reserve memory for kdump kernel to be loaded with kexec
 570  */
 571 static void __init reserve_crashkernel(void)
 572 {
 573 #ifdef CONFIG_CRASH_DUMP
 574         unsigned long long crash_base, crash_size;
 575         phys_addr_t low, high;
 576         int rc;
 577
 578         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 579                                &crash_base);
 580
 581         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 582         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 583         if (rc || crash_size == 0)
 584                 return;
 585
 586         if (memblock.memory.regions[0].size < crash_size) {
 587                 pr_info("crashkernel reservation failed: %s\n",
 588                         "first memory chunk must be at least crashkernel size");
 589                 return;
 590         }
 591
 592         low = crash_base ?: OLDMEM_BASE;
 593         high = low + crash_size;
 594         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 595                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 596                 crash_base = low;
 597         } else {
 598                 /* Find suitable area in free memory */
 599                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 600                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 601
 602                 if (crash_base && crash_base < low) {
 603                         pr_info("crashkernel reservation failed: %s\n",
 604                                 "crash_base too low");
 605                         return;
 606                 }
 607                 low = crash_base ?: low;
 608                 crash_base = memblock_find_in_range(low, high, crash_size,
 609                                                     KEXEC_CRASH_MEM_ALIGN);
 610         }
 611
 612         if (!crash_base) {
 613                 pr_info("crashkernel reservation failed: %s\n",
 614                         "no suitable area found");
 615                 return;
 616         }
 617
 618         if (register_memory_notifier(&kdump_mem_nb))
 619                 return;
 620
 621         if (!OLDMEM_BASE && MACHINE_IS_VM)
 622                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 623         crashk_res.start = crash_base;
 624         crashk_res.end = crash_base + crash_size - 1;
 625         memblock_remove(crash_base, crash_size);
 626         pr_info("Reserving %lluMB of memory at %lluMB "
 627                 "for crashkernel (System RAM: %luMB)\n",
 628                 crash_size >> 20, crash_base >> 20,
 629                 (unsigned long)memblock.memory.total_size >> 20);
 630         os_info_crashkernel_add(crash_base, crash_size);
 631 #endif
 632 }
 633
 634 /*
 635  * Reserve the initrd from being used by memblock
 636  */
 637 static void __init reserve_initrd(void)
 638 {
 639 #ifdef CONFIG_BLK_DEV_INITRD
 640         initrd_start = INITRD_START;
 641         initrd_end = initrd_start + INITRD_SIZE;
 642         memblock_reserve(INITRD_START, INITRD_SIZE);
 643 #endif
 644 }
 645
 646 /*
 647  * Check for initrd being in usable memory
 648  */
 649 static void __init check_initrd(void)
 650 {
 651 #ifdef CONFIG_BLK_DEV_INITRD
 652         if (INITRD_START && INITRD_SIZE &&
 653             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 654                 pr_err("initrd does not fit memory.\n");
 655                 memblock_free(INITRD_START, INITRD_SIZE);
 656                 initrd_start = initrd_end = 0;
 657         }
 658 #endif
 659 }
 660
 661 /*
 662  * Reserve memory used for lowcore/command line/kernel image.
 663  */
 664 static void __init reserve_kernel(void)
 665 {
 666         unsigned long start_pfn = PFN_UP(__pa(&_end));
 667
 668 #ifdef CONFIG_DMA_API_DEBUG
 669         /*
 670          * DMA_API_DEBUG code stumbles over addresses from the
 671          * range [_ehead, _stext]. Mark the memory as reserved
 672          * so it is not used for CONFIG_DMA_API_DEBUG=y.
 673          */
 674         memblock_reserve(0, PFN_PHYS(start_pfn));
 675 #else
 676         memblock_reserve(0, (unsigned long)_ehead);
 677         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 678                          - (unsigned long)_stext);
 679 #endif
 680 }
 681
 682 static void __init setup_memory(void)
 683 {
 684         struct memblock_region *reg;
 685
 686         /*
 687          * Init storage key for present memory
 688          */
 689         for_each_memblock(memory, reg) {
 690                 storage_key_init_range(reg->base, reg->base + reg->size);
 691         }
 692         psw_set_key(PAGE_DEFAULT_KEY);
 693
 694         /* Only cosmetics */
 695         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 696 }
 697
 698 /*
 699  * Setup hardware capabilities.
 700  */
 701 static int __init setup_hwcaps(void)
 702 {
 703         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 704         struct cpuid cpu_id;
 705         int i;
 706
 707         /*
 708          * The store facility list bits numbers as found in the principles
 709          * of operation are numbered with bit 1UL<<31 as number 0 to
 710          * bit 1UL<<0 as number 31.
 711          *   Bit 0: instructions named N3, "backported" to esa-mode
 712          *   Bit 2: z/Architecture mode is active
 713          *   Bit 7: the store-facility-list-extended facility is installed
 714          *   Bit 17: the message-security assist is installed
 715          *   Bit 19: the long-displacement facility is installed
 716          *   Bit 21: the extended-immediate facility is installed
 717          *   Bit 22: extended-translation facility 3 is installed
 718          *   Bit 30: extended-translation facility 3 enhancement facility
 719          * These get translated to:
 720          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 721          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 722          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 723          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 724          */
 725         for (i = 0; i < 6; i++)
 726                 if (test_facility(stfl_bits[i]))
 727                         elf_hwcap |= 1UL << i;
 728
 729         if (test_facility(22) && test_facility(30))
 730                 elf_hwcap |= HWCAP_S390_ETF3EH;
 731
 732         /*
 733          * Check for additional facilities with store-facility-list-extended.
 734          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 735          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 736          * as stored by stfl, bits 32-xxx contain additional facilities.
 737          * How many facility words are stored depends on the number of
 738          * doublewords passed to the instruction. The additional facilities
 739          * are:
 740          *   Bit 42: decimal floating point facility is installed
 741          *   Bit 44: perform floating point operation facility is installed
 742          * translated to:
 743          *   HWCAP_S390_DFP bit 6 (42 && 44).
 744          */
 745         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 746                 elf_hwcap |= HWCAP_S390_DFP;
 747
 748         /*
 749          * Huge page support HWCAP_S390_HPAGE is bit 7.
 750          */
 751         if (MACHINE_HAS_HPAGE)
 752                 elf_hwcap |= HWCAP_S390_HPAGE;
 753
 754         /*
 755          * 64-bit register support for 31-bit processes
 756          * HWCAP_S390_HIGH_GPRS is bit 9.
 757          */
 758         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 759
 760         /*
 761          * Transactional execution support HWCAP_S390_TE is bit 10.
 762          */
 763         if (test_facility(50) && test_facility(73))
 764                 elf_hwcap |= HWCAP_S390_TE;
 765
 766         /*
 767          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 768          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 769          * instead of facility bit 129.
 770          */
 771         if (MACHINE_HAS_VX)
 772                 elf_hwcap |= HWCAP_S390_VXRS;
 773         get_cpu_id(&cpu_id);
 774         add_device_randomness(&cpu_id, sizeof(cpu_id));
 775         switch (cpu_id.machine) {
 776         case 0x2064:
 777         case 0x2066:
 778         default:        /* Use "z900" as default for 64 bit kernels. */
 779                 strcpy(elf_platform, "z900");
 780                 break;
 781         case 0x2084:
 782         case 0x2086:
 783                 strcpy(elf_platform, "z990");
 784                 break;
 785         case 0x2094:
 786         case 0x2096:
 787                 strcpy(elf_platform, "z9-109");
 788                 break;
 789         case 0x2097:
 790         case 0x2098:
 791                 strcpy(elf_platform, "z10");
 792                 break;
 793         case 0x2817:
 794         case 0x2818:
 795                 strcpy(elf_platform, "z196");
 796                 break;
 797         case 0x2827:
 798         case 0x2828:
 799                 strcpy(elf_platform, "zEC12");
 800                 break;
 801         case 0x2964:
 802         case 0x2965:
 803                 strcpy(elf_platform, "z13");
 804                 break;
 805         }
 806
 807         /*
 808          * Virtualization support HWCAP_INT_SIE is bit 0.
 809          */
 810         if (sclp.has_sief2)
 811                 int_hwcap |= HWCAP_INT_SIE;
 812
 813         return 0;
 814 }
 815 arch_initcall(setup_hwcaps);
 816
 817 /*
 818  * Add system information as device randomness
 819  */
 820 static void __init setup_randomness(void)
 821 {
 822         struct sysinfo_3_2_2 *vmms;
 823
 824         vmms = (struct sysinfo_3_2_2 *) alloc_page(GFP_KERNEL);
 825         if (vmms && stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 826                 add_device_randomness(&vmms, vmms->count);
 827         free_page((unsigned long) vmms);
 828 }
 829
 830 /*
 831  * Find the correct size for the task_struct. This depends on
 832  * the size of the struct fpu at the end of the thread_struct
 833  * which is embedded in the task_struct.
 834  */
 835 static void __init setup_task_size(void)
 836 {
 837         int task_size = sizeof(struct task_struct);
 838
 839         if (!MACHINE_HAS_VX) {
 840                 task_size -= sizeof(__vector128) * __NUM_VXRS;
 841                 task_size += sizeof(freg_t) * __NUM_FPRS;
 842         }
 843         arch_task_struct_size = task_size;
 844 }
 845
 846 /*
 847  * Setup function called from init/main.c just after the banner
 848  * was printed.
 849  */
 850
 851 void __init setup_arch(char **cmdline_p)
 852 {
 853         /*
 854          * print what head.S has found out about the machine
 855          */
 856         if (MACHINE_IS_VM)
 857                 pr_info("Linux is running as a z/VM "
 858                         "guest operating system in 64-bit mode\n");
 859         else if (MACHINE_IS_KVM)
 860                 pr_info("Linux is running under KVM in 64-bit mode\n");
 861         else if (MACHINE_IS_LPAR)
 862                 pr_info("Linux is running natively in 64-bit mode\n");
 863
 864         /* Have one command line that is parsed and saved in /proc/cmdline */
 865         /* boot_command_line has been already set up in early.c */
 866         *cmdline_p = boot_command_line;
 867
 868         ROOT_DEV = Root_RAM0;
 869
 870         /* Is init_mm really needed? */
 871         init_mm.start_code = PAGE_OFFSET;
 872         init_mm.end_code = (unsigned long) &_etext;
 873         init_mm.end_data = (unsigned long) &_edata;
 874         init_mm.brk = (unsigned long) &_end;
 875
 876         parse_early_param();
 877 #ifdef CONFIG_CRASH_DUMP
 878         /* Deactivate elfcorehdr= kernel parameter */
 879         elfcorehdr_addr = ELFCORE_ADDR_MAX;
 880 #endif
 881
 882         os_info_init();
 883         setup_ipl();
 884         setup_task_size();
 885
 886         /* Do some memory reservations *before* memory is added to memblock */
 887         reserve_memory_end();
 888         reserve_oldmem();
 889         reserve_kernel();
 890         reserve_initrd();
 891         memblock_allow_resize();
 892
 893         /* Get information about *all* installed memory */
 894         detect_memory_memblock();
 895
 896         remove_oldmem();
 897
 898         /*
 899          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 900          * extra checks that HOLES_IN_ZONE would require.
 901          *
 902          * Is this still required?
 903          */
 904         memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
 905
 906         setup_memory_end();
 907         setup_memory();
 908
 909         check_initrd();
 910         reserve_crashkernel();
 911 #ifdef CONFIG_CRASH_DUMP
 912         /*
 913          * Be aware that smp_save_dump_cpus() triggers a system reset.
 914          * Therefore CPU and device initialization should be done afterwards.
 915          */
 916         smp_save_dump_cpus();
 917 #endif
 918
 919         setup_resources();
 920         setup_vmcoreinfo();
 921         setup_lowcore();
 922         smp_fill_possible_mask();
 923         cpu_detect_mhz_feature();
 924         cpu_init();
 925         numa_setup();
 926
 927         /*
 928          * Create kernel page tables and switch to virtual addressing.
 929          */
 930         paging_init();
 931
 932         /* Setup default console */
 933         conmode_default();
 934         set_preferred_console();
 935
 936         /* Setup zfcpdump support */
 937         setup_zfcpdump();
 938
 939         /* Add system specific data to the random pool */
 940         setup_randomness();
 941 }