arch/s390/kernel/setup.c

   1 /*
   2  *  arch/s390/kernel/setup.c
   3  *
   4  *  S390 version
   5  *    Copyright (C) IBM Corp. 1999,2010
   6  *    Author(s): Hartmut Penner (hp@de.ibm.com),
   7  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
   8  *
   9  *  Derived from "arch/i386/kernel/setup.c"
  10  *    Copyright (C) 1995, Linus Torvalds
  11  */
  12
  13 /*
  14  * This file handles the architecture-dependent parts of initialization
  15  */
  16
  17 #define KMSG_COMPONENT "setup"
  18 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  19
  20 #include <linux/errno.h>
  21 #include <linux/module.h>
  22 #include <linux/sched.h>
  23 #include <linux/kernel.h>
  24 #include <linux/mm.h>
  25 #include <linux/stddef.h>
  26 #include <linux/unistd.h>
  27 #include <linux/ptrace.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/ftrace.h>
  45 #include <linux/kexec.h>
  46 #include <linux/crash_dump.h>
  47 #include <linux/memory.h>
  48
  49 #include <asm/ipl.h>
  50 #include <asm/uaccess.h>
  51 #include <asm/system.h>
  52 #include <asm/smp.h>
  53 #include <asm/mmu_context.h>
  54 #include <asm/cpcmd.h>
  55 #include <asm/lowcore.h>
  56 #include <asm/irq.h>
  57 #include <asm/page.h>
  58 #include <asm/ptrace.h>
  59 #include <asm/sections.h>
  60 #include <asm/ebcdic.h>
  61 #include <asm/compat.h>
  62 #include <asm/kvm_virtio.h>
  63 #include <asm/diag.h>
  64
  65 long psw_kernel_bits    = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_ASC_PRIMARY |
  66                           PSW_MASK_EA | PSW_MASK_BA;
  67 long psw_user_bits      = PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT |
  68                           PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK |
  69                           PSW_MASK_PSTATE | PSW_ASC_HOME;
  70
  71 /*
  72  * User copy operations.
  73  */
  74 struct uaccess_ops uaccess;
  75 EXPORT_SYMBOL(uaccess);
  76
  77 /*
  78  * Machine setup..
  79  */
  80 unsigned int console_mode = 0;
  81 EXPORT_SYMBOL(console_mode);
  82
  83 unsigned int console_devno = -1;
  84 EXPORT_SYMBOL(console_devno);
  85
  86 unsigned int console_irq = -1;
  87 EXPORT_SYMBOL(console_irq);
  88
  89 unsigned long elf_hwcap = 0;
  90 char elf_platform[ELF_PLATFORM_SIZE];
  91
  92 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
  93
  94 int __initdata memory_end_set;
  95 unsigned long __initdata memory_end;
  96
  97 unsigned long VMALLOC_START;
  98 EXPORT_SYMBOL(VMALLOC_START);
  99
 100 unsigned long VMALLOC_END;
 101 EXPORT_SYMBOL(VMALLOC_END);
 102
 103 struct page *vmemmap;
 104 EXPORT_SYMBOL(vmemmap);
 105
 106 /* An array with a pointer to the lowcore of every CPU. */
 107 struct _lowcore *lowcore_ptr[NR_CPUS];
 108 EXPORT_SYMBOL(lowcore_ptr);
 109
 110 /*
 111  * This is set up by the setup-routine at boot-time
 112  * for S390 need to find out, what we have to setup
 113  * using address 0x10400 ...
 114  */
 115
 116 #include <asm/setup.h>
 117
 118 /*
 119  * condev= and conmode= setup parameter.
 120  */
 121
 122 static int __init condev_setup(char *str)
 123 {
 124         int vdev;
 125
 126         vdev = simple_strtoul(str, &str, 0);
 127         if (vdev >= 0 && vdev < 65536) {
 128                 console_devno = vdev;
 129                 console_irq = -1;
 130         }
 131         return 1;
 132 }
 133
 134 __setup("condev=", condev_setup);
 135
 136 static void __init set_preferred_console(void)
 137 {
 138         if (MACHINE_IS_KVM)
 139                 add_preferred_console("hvc", 0, NULL);
 140         else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 141                 add_preferred_console("ttyS", 0, NULL);
 142         else if (CONSOLE_IS_3270)
 143                 add_preferred_console("tty3270", 0, NULL);
 144 }
 145
 146 static int __init conmode_setup(char *str)
 147 {
 148 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 149         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 150                 SET_CONSOLE_SCLP;
 151 #endif
 152 #if defined(CONFIG_TN3215_CONSOLE)
 153         if (strncmp(str, "3215", 5) == 0)
 154                 SET_CONSOLE_3215;
 155 #endif
 156 #if defined(CONFIG_TN3270_CONSOLE)
 157         if (strncmp(str, "3270", 5) == 0)
 158                 SET_CONSOLE_3270;
 159 #endif
 160         set_preferred_console();
 161         return 1;
 162 }
 163
 164 __setup("conmode=", conmode_setup);
 165
 166 static void __init conmode_default(void)
 167 {
 168         char query_buffer[1024];
 169         char *ptr;
 170
 171         if (MACHINE_IS_VM) {
 172                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 173                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 174                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 175                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 176                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 177                 ptr = strstr(query_buffer, "CONMODE");
 178                 /*
 179                  * Set the conmode to 3215 so that the device recognition
 180                  * will set the cu_type of the console to 3215. If the
 181                  * conmode is 3270 and we don't set it back then both
 182                  * 3215 and the 3270 driver will try to access the console
 183                  * device (3215 as console and 3270 as normal tty).
 184                  */
 185                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 186                 if (ptr == NULL) {
 187 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 188                         SET_CONSOLE_SCLP;
 189 #endif
 190                         return;
 191                 }
 192                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 193 #if defined(CONFIG_TN3270_CONSOLE)
 194                         SET_CONSOLE_3270;
 195 #elif defined(CONFIG_TN3215_CONSOLE)
 196                         SET_CONSOLE_3215;
 197 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 198                         SET_CONSOLE_SCLP;
 199 #endif
 200                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 201 #if defined(CONFIG_TN3215_CONSOLE)
 202                         SET_CONSOLE_3215;
 203 #elif defined(CONFIG_TN3270_CONSOLE)
 204                         SET_CONSOLE_3270;
 205 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 206                         SET_CONSOLE_SCLP;
 207 #endif
 208                 }
 209         } else {
 210 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 211                 SET_CONSOLE_SCLP;
 212 #endif
 213         }
 214 }
 215
 216 #ifdef CONFIG_ZFCPDUMP
 217 static void __init setup_zfcpdump(unsigned int console_devno)
 218 {
 219         static char str[41];
 220
 221         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 222                 return;
 223         if (OLDMEM_BASE)
 224                 return;
 225         if (console_devno != -1)
 226                 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
 227                         ipl_info.data.fcp.dev_id.devno, console_devno);
 228         else
 229                 sprintf(str, " cio_ignore=all,!0.0.%04x",
 230                         ipl_info.data.fcp.dev_id.devno);
 231         strcat(boot_command_line, str);
 232         console_loglevel = 2;
 233 }
 234 #else
 235 static inline void setup_zfcpdump(unsigned int console_devno) {}
 236 #endif /* CONFIG_ZFCPDUMP */
 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
 281 static int __init early_parse_mem(char *p)
 282 {
 283         memory_end = memparse(p, &p);
 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 unsigned int user_mode = HOME_SPACE_MODE;
 299 EXPORT_SYMBOL_GPL(user_mode);
 300
 301 static int set_amode_primary(void)
 302 {
 303         psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
 304         psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
 305 #ifdef CONFIG_COMPAT
 306         psw32_user_bits =
 307                 (psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
 308 #endif
 309
 310         if (MACHINE_HAS_MVCOS) {
 311                 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
 312                 return 1;
 313         } else {
 314                 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
 315                 return 0;
 316         }
 317 }
 318
 319 /*
 320  * Switch kernel/user addressing modes?
 321  */
 322 static int __init early_parse_switch_amode(char *p)
 323 {
 324         user_mode = PRIMARY_SPACE_MODE;
 325         return 0;
 326 }
 327 early_param("switch_amode", early_parse_switch_amode);
 328
 329 static int __init early_parse_user_mode(char *p)
 330 {
 331         if (p && strcmp(p, "primary") == 0)
 332                 user_mode = PRIMARY_SPACE_MODE;
 333         else if (!p || strcmp(p, "home") == 0)
 334                 user_mode = HOME_SPACE_MODE;
 335         else
 336                 return 1;
 337         return 0;
 338 }
 339 early_param("user_mode", early_parse_user_mode);
 340
 341 static void setup_addressing_mode(void)
 342 {
 343         if (user_mode == PRIMARY_SPACE_MODE) {
 344                 if (set_amode_primary())
 345                         pr_info("Address spaces switched, "
 346                                 "mvcos available\n");
 347                 else
 348                         pr_info("Address spaces switched, "
 349                                 "mvcos not available\n");
 350         }
 351 }
 352
 353 static void __init
 354 setup_lowcore(void)
 355 {
 356         struct _lowcore *lc;
 357
 358         /*
 359          * Setup lowcore for boot cpu
 360          */
 361         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
 362         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 363         lc->restart_psw.mask = psw_kernel_bits;
 364         lc->restart_psw.addr =
 365                 PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
 366         lc->external_new_psw.mask = psw_kernel_bits |
 367                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 368         lc->external_new_psw.addr =
 369                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
 370         lc->svc_new_psw.mask = psw_kernel_bits |
 371                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 372         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
 373         lc->program_new_psw.mask = psw_kernel_bits |
 374                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 375         lc->program_new_psw.addr =
 376                 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
 377         lc->mcck_new_psw.mask = psw_kernel_bits;
 378         lc->mcck_new_psw.addr =
 379                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
 380         lc->io_new_psw.mask = psw_kernel_bits |
 381                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 382         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 383         lc->clock_comparator = -1ULL;
 384         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
 385         lc->async_stack = (unsigned long)
 386                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
 387         lc->panic_stack = (unsigned long)
 388                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
 389         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 390         lc->thread_info = (unsigned long) &init_thread_union;
 391         lc->machine_flags = S390_lowcore.machine_flags;
 392         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 393         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 394                MAX_FACILITY_BIT/8);
 395 #ifndef CONFIG_64BIT
 396         if (MACHINE_HAS_IEEE) {
 397                 lc->extended_save_area_addr = (__u32)
 398                         __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
 399                 /* enable extended save area */
 400                 __ctl_set_bit(14, 29);
 401         }
 402 #else
 403         lc->cmf_hpp = -1ULL;
 404         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
 405 #endif
 406         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 407         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 408         lc->exit_timer = S390_lowcore.exit_timer;
 409         lc->user_timer = S390_lowcore.user_timer;
 410         lc->system_timer = S390_lowcore.system_timer;
 411         lc->steal_timer = S390_lowcore.steal_timer;
 412         lc->last_update_timer = S390_lowcore.last_update_timer;
 413         lc->last_update_clock = S390_lowcore.last_update_clock;
 414         lc->ftrace_func = S390_lowcore.ftrace_func;
 415         set_prefix((u32)(unsigned long) lc);
 416         lowcore_ptr[0] = lc;
 417 }
 418
 419 static struct resource code_resource = {
 420         .name  = "Kernel code",
 421         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 422 };
 423
 424 static struct resource data_resource = {
 425         .name = "Kernel data",
 426         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 427 };
 428
 429 static struct resource bss_resource = {
 430         .name = "Kernel bss",
 431         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 432 };
 433
 434 static struct resource __initdata *standard_resources[] = {
 435         &code_resource,
 436         &data_resource,
 437         &bss_resource,
 438 };
 439
 440 static void __init setup_resources(void)
 441 {
 442         struct resource *res, *std_res, *sub_res;
 443         int i, j;
 444
 445         code_resource.start = (unsigned long) &_text;
 446         code_resource.end = (unsigned long) &_etext - 1;
 447         data_resource.start = (unsigned long) &_etext;
 448         data_resource.end = (unsigned long) &_edata - 1;
 449         bss_resource.start = (unsigned long) &__bss_start;
 450         bss_resource.end = (unsigned long) &__bss_stop - 1;
 451
 452         for (i = 0; i < MEMORY_CHUNKS; i++) {
 453                 if (!memory_chunk[i].size)
 454                         continue;
 455                 if (memory_chunk[i].type == CHUNK_OLDMEM ||
 456                     memory_chunk[i].type == CHUNK_CRASHK)
 457                         continue;
 458                 res = alloc_bootmem_low(sizeof(*res));
 459                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
 460                 switch (memory_chunk[i].type) {
 461                 case CHUNK_READ_WRITE:
 462                 case CHUNK_CRASHK:
 463                         res->name = "System RAM";
 464                         break;
 465                 case CHUNK_READ_ONLY:
 466                         res->name = "System ROM";
 467                         res->flags |= IORESOURCE_READONLY;
 468                         break;
 469                 default:
 470                         res->name = "reserved";
 471                 }
 472                 res->start = memory_chunk[i].addr;
 473                 res->end = res->start + memory_chunk[i].size - 1;
 474                 request_resource(&iomem_resource, res);
 475
 476                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 477                         std_res = standard_resources[j];
 478                         if (std_res->start < res->start ||
 479                             std_res->start > res->end)
 480                                 continue;
 481                         if (std_res->end > res->end) {
 482                                 sub_res = alloc_bootmem_low(sizeof(*sub_res));
 483                                 *sub_res = *std_res;
 484                                 sub_res->end = res->end;
 485                                 std_res->start = res->end + 1;
 486                                 request_resource(res, sub_res);
 487                         } else {
 488                                 request_resource(res, std_res);
 489                         }
 490                 }
 491         }
 492 }
 493
 494 unsigned long real_memory_size;
 495 EXPORT_SYMBOL_GPL(real_memory_size);
 496
 497 static void __init setup_memory_end(void)
 498 {
 499         unsigned long vmax, vmalloc_size, tmp;
 500         int i;
 501
 502
 503 #ifdef CONFIG_ZFCPDUMP
 504         if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) {
 505                 memory_end = ZFCPDUMP_HSA_SIZE;
 506                 memory_end_set = 1;
 507         }
 508 #endif
 509         real_memory_size = 0;
 510         memory_end &= PAGE_MASK;
 511
 512         /*
 513          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 514          * extra checks that HOLES_IN_ZONE would require.
 515          */
 516         for (i = 0; i < MEMORY_CHUNKS; i++) {
 517                 unsigned long start, end;
 518                 struct mem_chunk *chunk;
 519                 unsigned long align;
 520
 521                 chunk = &memory_chunk[i];
 522                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
 523                 start = (chunk->addr + align - 1) & ~(align - 1);
 524                 end = (chunk->addr + chunk->size) & ~(align - 1);
 525                 if (start >= end)
 526                         memset(chunk, 0, sizeof(*chunk));
 527                 else {
 528                         chunk->addr = start;
 529                         chunk->size = end - start;
 530                 }
 531                 real_memory_size = max(real_memory_size,
 532                                        chunk->addr + chunk->size);
 533         }
 534
 535         /* Choose kernel address space layout: 2, 3, or 4 levels. */
 536 #ifdef CONFIG_64BIT
 537         vmalloc_size = VMALLOC_END ?: 128UL << 30;
 538         tmp = (memory_end ?: real_memory_size) / PAGE_SIZE;
 539         tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
 540         if (tmp <= (1UL << 42))
 541                 vmax = 1UL << 42;       /* 3-level kernel page table */
 542         else
 543                 vmax = 1UL << 53;       /* 4-level kernel page table */
 544 #else
 545         vmalloc_size = VMALLOC_END ?: 96UL << 20;
 546         vmax = 1UL << 31;               /* 2-level kernel page table */
 547 #endif
 548         /* vmalloc area is at the end of the kernel address space. */
 549         VMALLOC_END = vmax;
 550         VMALLOC_START = vmax - vmalloc_size;
 551
 552         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 553         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 554         tmp = VMALLOC_START - tmp * sizeof(struct page);
 555         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 556         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 557         vmemmap = (struct page *) tmp;
 558
 559         /* Take care that memory_end is set and <= vmemmap */
 560         memory_end = min(memory_end ?: real_memory_size, tmp);
 561
 562         /* Fixup memory chunk array to fit into 0..memory_end */
 563         for (i = 0; i < MEMORY_CHUNKS; i++) {
 564                 struct mem_chunk *chunk = &memory_chunk[i];
 565
 566                 if (chunk->addr >= memory_end) {
 567                         memset(chunk, 0, sizeof(*chunk));
 568                         continue;
 569                 }
 570                 if (chunk->addr + chunk->size > memory_end)
 571                         chunk->size = memory_end - chunk->addr;
 572         }
 573 }
 574
 575 void *restart_stack __attribute__((__section__(".data")));
 576
 577 /*
 578  * Setup new PSW and allocate stack for PSW restart interrupt
 579  */
 580 static void __init setup_restart_psw(void)
 581 {
 582         psw_t psw;
 583
 584         restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
 585         restart_stack += ASYNC_SIZE;
 586
 587         /*
 588          * Setup restart PSW for absolute zero lowcore. This is necesary
 589          * if PSW restart is done on an offline CPU that has lowcore zero
 590          */
 591         psw.mask = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_EA | PSW_MASK_BA;
 592         psw.addr = PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
 593         copy_to_absolute_zero(&S390_lowcore.restart_psw, &psw, sizeof(psw));
 594 }
 595
 596 static void __init setup_vmcoreinfo(void)
 597 {
 598 #ifdef CONFIG_KEXEC
 599         unsigned long ptr = paddr_vmcoreinfo_note();
 600
 601         copy_to_absolute_zero(&S390_lowcore.vmcore_info, &ptr, sizeof(ptr));
 602 #endif
 603 }
 604
 605 #ifdef CONFIG_CRASH_DUMP
 606
 607 /*
 608  * Find suitable location for crashkernel memory
 609  */
 610 static unsigned long __init find_crash_base(unsigned long crash_size,
 611                                             char **msg)
 612 {
 613         unsigned long crash_base;
 614         struct mem_chunk *chunk;
 615         int i;
 616
 617         if (memory_chunk[0].size < crash_size) {
 618                 *msg = "first memory chunk must be at least crashkernel size";
 619                 return 0;
 620         }
 621         if (OLDMEM_BASE && crash_size == OLDMEM_SIZE)
 622                 return OLDMEM_BASE;
 623
 624         for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
 625                 chunk = &memory_chunk[i];
 626                 if (chunk->size == 0)
 627                         continue;
 628                 if (chunk->type != CHUNK_READ_WRITE)
 629                         continue;
 630                 if (chunk->size < crash_size)
 631                         continue;
 632                 crash_base = (chunk->addr + chunk->size) - crash_size;
 633                 if (crash_base < crash_size)
 634                         continue;
 635                 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
 636                         continue;
 637                 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
 638                         continue;
 639                 return crash_base;
 640         }
 641         *msg = "no suitable area found";
 642         return 0;
 643 }
 644
 645 /*
 646  * Check if crash_base and crash_size is valid
 647  */
 648 static int __init verify_crash_base(unsigned long crash_base,
 649                                     unsigned long crash_size,
 650                                     char **msg)
 651 {
 652         struct mem_chunk *chunk;
 653         int i;
 654
 655         /*
 656          * Because we do the swap to zero, we must have at least 'crash_size'
 657          * bytes free space before crash_base
 658          */
 659         if (crash_size > crash_base) {
 660                 *msg = "crashkernel offset must be greater than size";
 661                 return -EINVAL;
 662         }
 663
 664         /* First memory chunk must be at least crash_size */
 665         if (memory_chunk[0].size < crash_size) {
 666                 *msg = "first memory chunk must be at least crashkernel size";
 667                 return -EINVAL;
 668         }
 669         /* Check if we fit into the respective memory chunk */
 670         for (i = 0; i < MEMORY_CHUNKS; i++) {
 671                 chunk = &memory_chunk[i];
 672                 if (chunk->size == 0)
 673                         continue;
 674                 if (crash_base < chunk->addr)
 675                         continue;
 676                 if (crash_base >= chunk->addr + chunk->size)
 677                         continue;
 678                 /* we have found the memory chunk */
 679                 if (crash_base + crash_size > chunk->addr + chunk->size) {
 680                         *msg = "selected memory chunk is too small for "
 681                                 "crashkernel memory";
 682                         return -EINVAL;
 683                 }
 684                 return 0;
 685         }
 686         *msg = "invalid memory range specified";
 687         return -EINVAL;
 688 }
 689
 690 /*
 691  * Reserve kdump memory by creating a memory hole in the mem_chunk array
 692  */
 693 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
 694                                          int type)
 695 {
 696         create_mem_hole(memory_chunk, addr, size, type);
 697 }
 698
 699 /*
 700  * When kdump is enabled, we have to ensure that no memory from
 701  * the area [0 - crashkernel memory size] and
 702  * [crashk_res.start - crashk_res.end] is set offline.
 703  */
 704 static int kdump_mem_notifier(struct notifier_block *nb,
 705                               unsigned long action, void *data)
 706 {
 707         struct memory_notify *arg = data;
 708
 709         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 710                 return NOTIFY_BAD;
 711         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 712                 return NOTIFY_OK;
 713         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 714                 return NOTIFY_OK;
 715         return NOTIFY_BAD;
 716 }
 717
 718 static struct notifier_block kdump_mem_nb = {
 719         .notifier_call = kdump_mem_notifier,
 720 };
 721
 722 #endif
 723
 724 /*
 725  * Make sure that oldmem, where the dump is stored, is protected
 726  */
 727 static void reserve_oldmem(void)
 728 {
 729 #ifdef CONFIG_CRASH_DUMP
 730         if (!OLDMEM_BASE)
 731                 return;
 732
 733         reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
 734         reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
 735                               CHUNK_OLDMEM);
 736         if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
 737                 saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
 738         else
 739                 saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
 740 #endif
 741 }
 742
 743 /*
 744  * Reserve memory for kdump kernel to be loaded with kexec
 745  */
 746 static void __init reserve_crashkernel(void)
 747 {
 748 #ifdef CONFIG_CRASH_DUMP
 749         unsigned long long crash_base, crash_size;
 750         char *msg;
 751         int rc;
 752
 753         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 754                                &crash_base);
 755         if (rc || crash_size == 0)
 756                 return;
 757         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 758         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 759         if (register_memory_notifier(&kdump_mem_nb))
 760                 return;
 761         if (!crash_base)
 762                 crash_base = find_crash_base(crash_size, &msg);
 763         if (!crash_base) {
 764                 pr_info("crashkernel reservation failed: %s\n", msg);
 765                 unregister_memory_notifier(&kdump_mem_nb);
 766                 return;
 767         }
 768         if (verify_crash_base(crash_base, crash_size, &msg)) {
 769                 pr_info("crashkernel reservation failed: %s\n", msg);
 770                 unregister_memory_notifier(&kdump_mem_nb);
 771                 return;
 772         }
 773         if (!OLDMEM_BASE && MACHINE_IS_VM)
 774                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 775         crashk_res.start = crash_base;
 776         crashk_res.end = crash_base + crash_size - 1;
 777         insert_resource(&iomem_resource, &crashk_res);
 778         reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
 779         pr_info("Reserving %lluMB of memory at %lluMB "
 780                 "for crashkernel (System RAM: %luMB)\n",
 781                 crash_size >> 20, crash_base >> 20, memory_end >> 20);
 782 #endif
 783 }
 784
 785 static void __init
 786 setup_memory(void)
 787 {
 788         unsigned long bootmap_size;
 789         unsigned long start_pfn, end_pfn;
 790         int i;
 791
 792         /*
 793          * partially used pages are not usable - thus
 794          * we are rounding upwards:
 795          */
 796         start_pfn = PFN_UP(__pa(&_end));
 797         end_pfn = max_pfn = PFN_DOWN(memory_end);
 798
 799 #ifdef CONFIG_BLK_DEV_INITRD
 800         /*
 801          * Move the initrd in case the bitmap of the bootmem allocater
 802          * would overwrite it.
 803          */
 804
 805         if (INITRD_START && INITRD_SIZE) {
 806                 unsigned long bmap_size;
 807                 unsigned long start;
 808
 809                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
 810                 bmap_size = PFN_PHYS(bmap_size);
 811
 812                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
 813                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
 814
 815 #ifdef CONFIG_CRASH_DUMP
 816                         if (OLDMEM_BASE) {
 817                                 /* Move initrd behind kdump oldmem */
 818                                 if (start + INITRD_SIZE > OLDMEM_BASE &&
 819                                     start < OLDMEM_BASE + OLDMEM_SIZE)
 820                                         start = OLDMEM_BASE + OLDMEM_SIZE;
 821                         }
 822 #endif
 823                         if (start + INITRD_SIZE > memory_end) {
 824                                 pr_err("initrd extends beyond end of "
 825                                        "memory (0x%08lx > 0x%08lx) "
 826                                        "disabling initrd\n",
 827                                        start + INITRD_SIZE, memory_end);
 828                                 INITRD_START = INITRD_SIZE = 0;
 829                         } else {
 830                                 pr_info("Moving initrd (0x%08lx -> "
 831                                         "0x%08lx, size: %ld)\n",
 832                                         INITRD_START, start, INITRD_SIZE);
 833                                 memmove((void *) start, (void *) INITRD_START,
 834                                         INITRD_SIZE);
 835                                 INITRD_START = start;
 836                         }
 837                 }
 838         }
 839 #endif
 840
 841         /*
 842          * Initialize the boot-time allocator
 843          */
 844         bootmap_size = init_bootmem(start_pfn, end_pfn);
 845
 846         /*
 847          * Register RAM areas with the bootmem allocator.
 848          */
 849
 850         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
 851                 unsigned long start_chunk, end_chunk, pfn;
 852
 853                 if (memory_chunk[i].type != CHUNK_READ_WRITE &&
 854                     memory_chunk[i].type != CHUNK_CRASHK)
 855                         continue;
 856                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
 857                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
 858                 end_chunk = min(end_chunk, end_pfn);
 859                 if (start_chunk >= end_chunk)
 860                         continue;
 861                 add_active_range(0, start_chunk, end_chunk);
 862                 pfn = max(start_chunk, start_pfn);
 863                 for (; pfn < end_chunk; pfn++)
 864                         page_set_storage_key(PFN_PHYS(pfn),
 865                                              PAGE_DEFAULT_KEY, 0);
 866         }
 867
 868         psw_set_key(PAGE_DEFAULT_KEY);
 869
 870         free_bootmem_with_active_regions(0, max_pfn);
 871
 872         /*
 873          * Reserve memory used for lowcore/command line/kernel image.
 874          */
 875         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
 876         reserve_bootmem((unsigned long)_stext,
 877                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
 878                         BOOTMEM_DEFAULT);
 879         /*
 880          * Reserve the bootmem bitmap itself as well. We do this in two
 881          * steps (first step was init_bootmem()) because this catches
 882          * the (very unlikely) case of us accidentally initializing the
 883          * bootmem allocator with an invalid RAM area.
 884          */
 885         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
 886                         BOOTMEM_DEFAULT);
 887
 888 #ifdef CONFIG_CRASH_DUMP
 889         if (crashk_res.start)
 890                 reserve_bootmem(crashk_res.start,
 891                                 crashk_res.end - crashk_res.start + 1,
 892                                 BOOTMEM_DEFAULT);
 893         if (is_kdump_kernel())
 894                 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
 895                                 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
 896 #endif
 897 #ifdef CONFIG_BLK_DEV_INITRD
 898         if (INITRD_START && INITRD_SIZE) {
 899                 if (INITRD_START + INITRD_SIZE <= memory_end) {
 900                         reserve_bootmem(INITRD_START, INITRD_SIZE,
 901                                         BOOTMEM_DEFAULT);
 902                         initrd_start = INITRD_START;
 903                         initrd_end = initrd_start + INITRD_SIZE;
 904                 } else {
 905                         pr_err("initrd extends beyond end of "
 906                                "memory (0x%08lx > 0x%08lx) "
 907                                "disabling initrd\n",
 908                                initrd_start + INITRD_SIZE, memory_end);
 909                         initrd_start = initrd_end = 0;
 910                 }
 911         }
 912 #endif
 913 }
 914
 915 /*
 916  * Setup hardware capabilities.
 917  */
 918 static void __init setup_hwcaps(void)
 919 {
 920         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 921         struct cpuid cpu_id;
 922         int i;
 923
 924         /*
 925          * The store facility list bits numbers as found in the principles
 926          * of operation are numbered with bit 1UL<<31 as number 0 to
 927          * bit 1UL<<0 as number 31.
 928          *   Bit 0: instructions named N3, "backported" to esa-mode
 929          *   Bit 2: z/Architecture mode is active
 930          *   Bit 7: the store-facility-list-extended facility is installed
 931          *   Bit 17: the message-security assist is installed
 932          *   Bit 19: the long-displacement facility is installed
 933          *   Bit 21: the extended-immediate facility is installed
 934          *   Bit 22: extended-translation facility 3 is installed
 935          *   Bit 30: extended-translation facility 3 enhancement facility
 936          * These get translated to:
 937          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 938          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 939          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 940          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 941          */
 942         for (i = 0; i < 6; i++)
 943                 if (test_facility(stfl_bits[i]))
 944                         elf_hwcap |= 1UL << i;
 945
 946         if (test_facility(22) && test_facility(30))
 947                 elf_hwcap |= HWCAP_S390_ETF3EH;
 948
 949         /*
 950          * Check for additional facilities with store-facility-list-extended.
 951          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 952          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 953          * as stored by stfl, bits 32-xxx contain additional facilities.
 954          * How many facility words are stored depends on the number of
 955          * doublewords passed to the instruction. The additional facilities
 956          * are:
 957          *   Bit 42: decimal floating point facility is installed
 958          *   Bit 44: perform floating point operation facility is installed
 959          * translated to:
 960          *   HWCAP_S390_DFP bit 6 (42 && 44).
 961          */
 962         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 963                 elf_hwcap |= HWCAP_S390_DFP;
 964
 965         /*
 966          * Huge page support HWCAP_S390_HPAGE is bit 7.
 967          */
 968         if (MACHINE_HAS_HPAGE)
 969                 elf_hwcap |= HWCAP_S390_HPAGE;
 970
 971         /*
 972          * 64-bit register support for 31-bit processes
 973          * HWCAP_S390_HIGH_GPRS is bit 9.
 974          */
 975         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 976
 977         get_cpu_id(&cpu_id);
 978         switch (cpu_id.machine) {
 979         case 0x9672:
 980 #if !defined(CONFIG_64BIT)
 981         default:        /* Use "g5" as default for 31 bit kernels. */
 982 #endif
 983                 strcpy(elf_platform, "g5");
 984                 break;
 985         case 0x2064:
 986         case 0x2066:
 987 #if defined(CONFIG_64BIT)
 988         default:        /* Use "z900" as default for 64 bit kernels. */
 989 #endif
 990                 strcpy(elf_platform, "z900");
 991                 break;
 992         case 0x2084:
 993         case 0x2086:
 994                 strcpy(elf_platform, "z990");
 995                 break;
 996         case 0x2094:
 997         case 0x2096:
 998                 strcpy(elf_platform, "z9-109");
 999                 break;
1000         case 0x2097:
1001         case 0x2098:
1002                 strcpy(elf_platform, "z10");
1003                 break;
1004         case 0x2817:
1005         case 0x2818:
1006                 strcpy(elf_platform, "z196");
1007                 break;
1008         }
1009 }
1010
1011 /*
1012  * Setup function called from init/main.c just after the banner
1013  * was printed.
1014  */
1015
1016 void __init
1017 setup_arch(char **cmdline_p)
1018 {
1019         /*
1020          * print what head.S has found out about the machine
1021          */
1022 #ifndef CONFIG_64BIT
1023         if (MACHINE_IS_VM)
1024                 pr_info("Linux is running as a z/VM "
1025                         "guest operating system in 31-bit mode\n");
1026         else if (MACHINE_IS_LPAR)
1027                 pr_info("Linux is running natively in 31-bit mode\n");
1028         if (MACHINE_HAS_IEEE)
1029                 pr_info("The hardware system has IEEE compatible "
1030                         "floating point units\n");
1031         else
1032                 pr_info("The hardware system has no IEEE compatible "
1033                         "floating point units\n");
1034 #else /* CONFIG_64BIT */
1035         if (MACHINE_IS_VM)
1036                 pr_info("Linux is running as a z/VM "
1037                         "guest operating system in 64-bit mode\n");
1038         else if (MACHINE_IS_KVM)
1039                 pr_info("Linux is running under KVM in 64-bit mode\n");
1040         else if (MACHINE_IS_LPAR)
1041                 pr_info("Linux is running natively in 64-bit mode\n");
1042 #endif /* CONFIG_64BIT */
1043
1044         /* Have one command line that is parsed and saved in /proc/cmdline */
1045         /* boot_command_line has been already set up in early.c */
1046         *cmdline_p = boot_command_line;
1047
1048         ROOT_DEV = Root_RAM0;
1049
1050         init_mm.start_code = PAGE_OFFSET;
1051         init_mm.end_code = (unsigned long) &_etext;
1052         init_mm.end_data = (unsigned long) &_edata;
1053         init_mm.brk = (unsigned long) &_end;
1054
1055         if (MACHINE_HAS_MVCOS)
1056                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1057         else
1058                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1059
1060         parse_early_param();
1061
1062         setup_ipl();
1063         setup_memory_end();
1064         setup_addressing_mode();
1065         reserve_oldmem();
1066         reserve_crashkernel();
1067         setup_memory();
1068         setup_resources();
1069         setup_vmcoreinfo();
1070         setup_restart_psw();
1071         setup_lowcore();
1072
1073         cpu_init();
1074         s390_init_cpu_topology();
1075
1076         /*
1077          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1078          */
1079         setup_hwcaps();
1080
1081         /*
1082          * Create kernel page tables and switch to virtual addressing.
1083          */
1084         paging_init();
1085
1086         /* Setup default console */
1087         conmode_default();
1088         set_preferred_console();
1089
1090         /* Setup zfcpdump support */
1091         setup_zfcpdump(console_devno);
1092 }