arch/s390/kernel/setup.c

   1 /*
   2  *  arch/s390/kernel/setup.c
   3  *
   4  *  S390 version
   5  *    Copyright (C) IBM Corp. 1999,2012
   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/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/user.h>
  30 #include <linux/tty.h>
  31 #include <linux/ioport.h>
  32 #include <linux/delay.h>
  33 #include <linux/init.h>
  34 #include <linux/initrd.h>
  35 #include <linux/bootmem.h>
  36 #include <linux/root_dev.h>
  37 #include <linux/console.h>
  38 #include <linux/kernel_stat.h>
  39 #include <linux/device.h>
  40 #include <linux/notifier.h>
  41 #include <linux/pfn.h>
  42 #include <linux/ctype.h>
  43 #include <linux/reboot.h>
  44 #include <linux/topology.h>
  45 #include <linux/ftrace.h>
  46 #include <linux/kexec.h>
  47 #include <linux/crash_dump.h>
  48 #include <linux/memory.h>
  49 #include <linux/compat.h>
  50
  51 #include <asm/ipl.h>
  52 #include <asm/uaccess.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/irq.h>
  59 #include <asm/page.h>
  60 #include <asm/ptrace.h>
  61 #include <asm/sections.h>
  62 #include <asm/ebcdic.h>
  63 #include <asm/kvm_virtio.h>
  64 #include <asm/diag.h>
  65 #include <asm/os_info.h>
  66 #include <asm/sclp.h>
  67 #include "entry.h"
  68
  69 long psw_kernel_bits    = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_ASC_PRIMARY |
  70                           PSW_MASK_EA | PSW_MASK_BA;
  71 long psw_user_bits      = PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT |
  72                           PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK |
  73                           PSW_MASK_PSTATE | PSW_ASC_HOME;
  74
  75 /*
  76  * User copy operations.
  77  */
  78 struct uaccess_ops uaccess;
  79 EXPORT_SYMBOL(uaccess);
  80
  81 /*
  82  * Machine setup..
  83  */
  84 unsigned int console_mode = 0;
  85 EXPORT_SYMBOL(console_mode);
  86
  87 unsigned int console_devno = -1;
  88 EXPORT_SYMBOL(console_devno);
  89
  90 unsigned int console_irq = -1;
  91 EXPORT_SYMBOL(console_irq);
  92
  93 unsigned long elf_hwcap = 0;
  94 char elf_platform[ELF_PLATFORM_SIZE];
  95
  96 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
  97
  98 int __initdata memory_end_set;
  99 unsigned long __initdata memory_end;
 100
 101 unsigned long VMALLOC_START;
 102 EXPORT_SYMBOL(VMALLOC_START);
 103
 104 unsigned long VMALLOC_END;
 105 EXPORT_SYMBOL(VMALLOC_END);
 106
 107 struct page *vmemmap;
 108 EXPORT_SYMBOL(vmemmap);
 109
 110 /* An array with a pointer to the lowcore of every CPU. */
 111 struct _lowcore *lowcore_ptr[NR_CPUS];
 112 EXPORT_SYMBOL(lowcore_ptr);
 113
 114 /*
 115  * This is set up by the setup-routine at boot-time
 116  * for S390 need to find out, what we have to setup
 117  * using address 0x10400 ...
 118  */
 119
 120 #include <asm/setup.h>
 121
 122 /*
 123  * condev= and conmode= setup parameter.
 124  */
 125
 126 static int __init condev_setup(char *str)
 127 {
 128         int vdev;
 129
 130         vdev = simple_strtoul(str, &str, 0);
 131         if (vdev >= 0 && vdev < 65536) {
 132                 console_devno = vdev;
 133                 console_irq = -1;
 134         }
 135         return 1;
 136 }
 137
 138 __setup("condev=", condev_setup);
 139
 140 static void __init set_preferred_console(void)
 141 {
 142         if (MACHINE_IS_KVM) {
 143                 if (sclp_has_vt220())
 144                         add_preferred_console("ttyS", 1, NULL);
 145                 else if (sclp_has_linemode())
 146                         add_preferred_console("ttyS", 0, NULL);
 147                 else
 148                         add_preferred_console("hvc", 0, NULL);
 149         } else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 150                 add_preferred_console("ttyS", 0, NULL);
 151         else if (CONSOLE_IS_3270)
 152                 add_preferred_console("tty3270", 0, NULL);
 153 }
 154
 155 static int __init conmode_setup(char *str)
 156 {
 157 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 158         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 159                 SET_CONSOLE_SCLP;
 160 #endif
 161 #if defined(CONFIG_TN3215_CONSOLE)
 162         if (strncmp(str, "3215", 5) == 0)
 163                 SET_CONSOLE_3215;
 164 #endif
 165 #if defined(CONFIG_TN3270_CONSOLE)
 166         if (strncmp(str, "3270", 5) == 0)
 167                 SET_CONSOLE_3270;
 168 #endif
 169         set_preferred_console();
 170         return 1;
 171 }
 172
 173 __setup("conmode=", conmode_setup);
 174
 175 static void __init conmode_default(void)
 176 {
 177         char query_buffer[1024];
 178         char *ptr;
 179
 180         if (MACHINE_IS_VM) {
 181                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 182                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 183                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 184                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 185                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 186                 ptr = strstr(query_buffer, "CONMODE");
 187                 /*
 188                  * Set the conmode to 3215 so that the device recognition
 189                  * will set the cu_type of the console to 3215. If the
 190                  * conmode is 3270 and we don't set it back then both
 191                  * 3215 and the 3270 driver will try to access the console
 192                  * device (3215 as console and 3270 as normal tty).
 193                  */
 194                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 195                 if (ptr == NULL) {
 196 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 197                         SET_CONSOLE_SCLP;
 198 #endif
 199                         return;
 200                 }
 201                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 202 #if defined(CONFIG_TN3270_CONSOLE)
 203                         SET_CONSOLE_3270;
 204 #elif defined(CONFIG_TN3215_CONSOLE)
 205                         SET_CONSOLE_3215;
 206 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 207                         SET_CONSOLE_SCLP;
 208 #endif
 209                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 210 #if defined(CONFIG_TN3215_CONSOLE)
 211                         SET_CONSOLE_3215;
 212 #elif defined(CONFIG_TN3270_CONSOLE)
 213                         SET_CONSOLE_3270;
 214 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 215                         SET_CONSOLE_SCLP;
 216 #endif
 217                 }
 218         } else {
 219 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 220                 SET_CONSOLE_SCLP;
 221 #endif
 222         }
 223 }
 224
 225 #ifdef CONFIG_ZFCPDUMP
 226 static void __init setup_zfcpdump(unsigned int console_devno)
 227 {
 228         static char str[41];
 229
 230         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 231                 return;
 232         if (OLDMEM_BASE)
 233                 return;
 234         if (console_devno != -1)
 235                 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
 236                         ipl_info.data.fcp.dev_id.devno, console_devno);
 237         else
 238                 sprintf(str, " cio_ignore=all,!0.0.%04x",
 239                         ipl_info.data.fcp.dev_id.devno);
 240         strcat(boot_command_line, str);
 241         console_loglevel = 2;
 242 }
 243 #else
 244 static inline void setup_zfcpdump(unsigned int console_devno) {}
 245 #endif /* CONFIG_ZFCPDUMP */
 246
 247  /*
 248  * Reboot, halt and power_off stubs. They just call _machine_restart,
 249  * _machine_halt or _machine_power_off.
 250  */
 251
 252 void machine_restart(char *command)
 253 {
 254         if ((!in_interrupt() && !in_atomic()) || 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_restart(command);
 261 }
 262
 263 void machine_halt(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_halt();
 272 }
 273
 274 void machine_power_off(void)
 275 {
 276         if (!in_interrupt() || oops_in_progress)
 277                 /*
 278                  * Only unblank the console if we are called in enabled
 279                  * context or a bust_spinlocks cleared the way for us.
 280                  */
 281                 console_unblank();
 282         _machine_power_off();
 283 }
 284
 285 /*
 286  * Dummy power off function.
 287  */
 288 void (*pm_power_off)(void) = machine_power_off;
 289
 290 static int __init early_parse_mem(char *p)
 291 {
 292         memory_end = memparse(p, &p);
 293         memory_end_set = 1;
 294         return 0;
 295 }
 296 early_param("mem", early_parse_mem);
 297
 298 static int __init parse_vmalloc(char *arg)
 299 {
 300         if (!arg)
 301                 return -EINVAL;
 302         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
 303         return 0;
 304 }
 305 early_param("vmalloc", parse_vmalloc);
 306
 307 unsigned int user_mode = HOME_SPACE_MODE;
 308 EXPORT_SYMBOL_GPL(user_mode);
 309
 310 static int set_amode_primary(void)
 311 {
 312         psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
 313         psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
 314 #ifdef CONFIG_COMPAT
 315         psw32_user_bits =
 316                 (psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
 317 #endif
 318
 319         if (MACHINE_HAS_MVCOS) {
 320                 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
 321                 return 1;
 322         } else {
 323                 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
 324                 return 0;
 325         }
 326 }
 327
 328 /*
 329  * Switch kernel/user addressing modes?
 330  */
 331 static int __init early_parse_switch_amode(char *p)
 332 {
 333         user_mode = PRIMARY_SPACE_MODE;
 334         return 0;
 335 }
 336 early_param("switch_amode", early_parse_switch_amode);
 337
 338 static int __init early_parse_user_mode(char *p)
 339 {
 340         if (p && strcmp(p, "primary") == 0)
 341                 user_mode = PRIMARY_SPACE_MODE;
 342         else if (!p || strcmp(p, "home") == 0)
 343                 user_mode = HOME_SPACE_MODE;
 344         else
 345                 return 1;
 346         return 0;
 347 }
 348 early_param("user_mode", early_parse_user_mode);
 349
 350 static void setup_addressing_mode(void)
 351 {
 352         if (user_mode == PRIMARY_SPACE_MODE) {
 353                 if (set_amode_primary())
 354                         pr_info("Address spaces switched, "
 355                                 "mvcos available\n");
 356                 else
 357                         pr_info("Address spaces switched, "
 358                                 "mvcos not available\n");
 359         }
 360 }
 361
 362 void *restart_stack __attribute__((__section__(".data")));
 363
 364 static void __init setup_lowcore(void)
 365 {
 366         struct _lowcore *lc;
 367
 368         /*
 369          * Setup lowcore for boot cpu
 370          */
 371         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
 372         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 373         lc->restart_psw.mask = psw_kernel_bits;
 374         lc->restart_psw.addr =
 375                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
 376         lc->external_new_psw.mask = psw_kernel_bits |
 377                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 378         lc->external_new_psw.addr =
 379                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
 380         lc->svc_new_psw.mask = psw_kernel_bits |
 381                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 382         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
 383         lc->program_new_psw.mask = psw_kernel_bits |
 384                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 385         lc->program_new_psw.addr =
 386                 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
 387         lc->mcck_new_psw.mask = psw_kernel_bits;
 388         lc->mcck_new_psw.addr =
 389                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
 390         lc->io_new_psw.mask = psw_kernel_bits |
 391                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 392         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 393         lc->clock_comparator = -1ULL;
 394         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
 395         lc->async_stack = (unsigned long)
 396                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
 397         lc->panic_stack = (unsigned long)
 398                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
 399         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 400         lc->thread_info = (unsigned long) &init_thread_union;
 401         lc->machine_flags = S390_lowcore.machine_flags;
 402         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 403         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 404                MAX_FACILITY_BIT/8);
 405 #ifndef CONFIG_64BIT
 406         if (MACHINE_HAS_IEEE) {
 407                 lc->extended_save_area_addr = (__u32)
 408                         __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
 409                 /* enable extended save area */
 410                 __ctl_set_bit(14, 29);
 411         }
 412 #else
 413         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
 414 #endif
 415         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 416         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 417         lc->exit_timer = S390_lowcore.exit_timer;
 418         lc->user_timer = S390_lowcore.user_timer;
 419         lc->system_timer = S390_lowcore.system_timer;
 420         lc->steal_timer = S390_lowcore.steal_timer;
 421         lc->last_update_timer = S390_lowcore.last_update_timer;
 422         lc->last_update_clock = S390_lowcore.last_update_clock;
 423         lc->ftrace_func = S390_lowcore.ftrace_func;
 424
 425         restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
 426         restart_stack += ASYNC_SIZE;
 427
 428         /*
 429          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 430          * restart data to the absolute zero lowcore. This is necesary if
 431          * PSW restart is done on an offline CPU that has lowcore zero.
 432          */
 433         lc->restart_stack = (unsigned long) restart_stack;
 434         lc->restart_fn = (unsigned long) do_restart;
 435         lc->restart_data = 0;
 436         lc->restart_source = -1UL;
 437
 438         /* Setup absolute zero lowcore */
 439         memcpy_absolute(&S390_lowcore.restart_stack, &lc->restart_stack,
 440                         4 * sizeof(unsigned long));
 441         memcpy_absolute(&S390_lowcore.restart_psw, &lc->restart_psw,
 442                         sizeof(lc->restart_psw));
 443
 444         set_prefix((u32)(unsigned long) lc);
 445         lowcore_ptr[0] = lc;
 446 }
 447
 448 static struct resource code_resource = {
 449         .name  = "Kernel code",
 450         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 451 };
 452
 453 static struct resource data_resource = {
 454         .name = "Kernel data",
 455         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 456 };
 457
 458 static struct resource bss_resource = {
 459         .name = "Kernel bss",
 460         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 461 };
 462
 463 static struct resource __initdata *standard_resources[] = {
 464         &code_resource,
 465         &data_resource,
 466         &bss_resource,
 467 };
 468
 469 static void __init setup_resources(void)
 470 {
 471         struct resource *res, *std_res, *sub_res;
 472         int i, j;
 473
 474         code_resource.start = (unsigned long) &_text;
 475         code_resource.end = (unsigned long) &_etext - 1;
 476         data_resource.start = (unsigned long) &_etext;
 477         data_resource.end = (unsigned long) &_edata - 1;
 478         bss_resource.start = (unsigned long) &__bss_start;
 479         bss_resource.end = (unsigned long) &__bss_stop - 1;
 480
 481         for (i = 0; i < MEMORY_CHUNKS; i++) {
 482                 if (!memory_chunk[i].size)
 483                         continue;
 484                 if (memory_chunk[i].type == CHUNK_OLDMEM ||
 485                     memory_chunk[i].type == CHUNK_CRASHK)
 486                         continue;
 487                 res = alloc_bootmem_low(sizeof(*res));
 488                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
 489                 switch (memory_chunk[i].type) {
 490                 case CHUNK_READ_WRITE:
 491                 case CHUNK_CRASHK:
 492                         res->name = "System RAM";
 493                         break;
 494                 case CHUNK_READ_ONLY:
 495                         res->name = "System ROM";
 496                         res->flags |= IORESOURCE_READONLY;
 497                         break;
 498                 default:
 499                         res->name = "reserved";
 500                 }
 501                 res->start = memory_chunk[i].addr;
 502                 res->end = res->start + memory_chunk[i].size - 1;
 503                 request_resource(&iomem_resource, res);
 504
 505                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 506                         std_res = standard_resources[j];
 507                         if (std_res->start < res->start ||
 508                             std_res->start > res->end)
 509                                 continue;
 510                         if (std_res->end > res->end) {
 511                                 sub_res = alloc_bootmem_low(sizeof(*sub_res));
 512                                 *sub_res = *std_res;
 513                                 sub_res->end = res->end;
 514                                 std_res->start = res->end + 1;
 515                                 request_resource(res, sub_res);
 516                         } else {
 517                                 request_resource(res, std_res);
 518                         }
 519                 }
 520         }
 521 }
 522
 523 unsigned long real_memory_size;
 524 EXPORT_SYMBOL_GPL(real_memory_size);
 525
 526 static void __init setup_memory_end(void)
 527 {
 528         unsigned long vmax, vmalloc_size, tmp;
 529         int i;
 530
 531
 532 #ifdef CONFIG_ZFCPDUMP
 533         if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) {
 534                 memory_end = ZFCPDUMP_HSA_SIZE;
 535                 memory_end_set = 1;
 536         }
 537 #endif
 538         real_memory_size = 0;
 539         memory_end &= PAGE_MASK;
 540
 541         /*
 542          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 543          * extra checks that HOLES_IN_ZONE would require.
 544          */
 545         for (i = 0; i < MEMORY_CHUNKS; i++) {
 546                 unsigned long start, end;
 547                 struct mem_chunk *chunk;
 548                 unsigned long align;
 549
 550                 chunk = &memory_chunk[i];
 551                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
 552                 start = (chunk->addr + align - 1) & ~(align - 1);
 553                 end = (chunk->addr + chunk->size) & ~(align - 1);
 554                 if (start >= end)
 555                         memset(chunk, 0, sizeof(*chunk));
 556                 else {
 557                         chunk->addr = start;
 558                         chunk->size = end - start;
 559                 }
 560                 real_memory_size = max(real_memory_size,
 561                                        chunk->addr + chunk->size);
 562         }
 563
 564         /* Choose kernel address space layout: 2, 3, or 4 levels. */
 565 #ifdef CONFIG_64BIT
 566         vmalloc_size = VMALLOC_END ?: 128UL << 30;
 567         tmp = (memory_end ?: real_memory_size) / PAGE_SIZE;
 568         tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
 569         if (tmp <= (1UL << 42))
 570                 vmax = 1UL << 42;       /* 3-level kernel page table */
 571         else
 572                 vmax = 1UL << 53;       /* 4-level kernel page table */
 573 #else
 574         vmalloc_size = VMALLOC_END ?: 96UL << 20;
 575         vmax = 1UL << 31;               /* 2-level kernel page table */
 576 #endif
 577         /* vmalloc area is at the end of the kernel address space. */
 578         VMALLOC_END = vmax;
 579         VMALLOC_START = vmax - vmalloc_size;
 580
 581         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 582         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 583         tmp = VMALLOC_START - tmp * sizeof(struct page);
 584         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 585         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 586         vmemmap = (struct page *) tmp;
 587
 588         /* Take care that memory_end is set and <= vmemmap */
 589         memory_end = min(memory_end ?: real_memory_size, tmp);
 590
 591         /* Fixup memory chunk array to fit into 0..memory_end */
 592         for (i = 0; i < MEMORY_CHUNKS; i++) {
 593                 struct mem_chunk *chunk = &memory_chunk[i];
 594
 595                 if (chunk->addr >= memory_end) {
 596                         memset(chunk, 0, sizeof(*chunk));
 597                         continue;
 598                 }
 599                 if (chunk->addr + chunk->size > memory_end)
 600                         chunk->size = memory_end - chunk->addr;
 601         }
 602 }
 603
 604 static void __init setup_vmcoreinfo(void)
 605 {
 606 #ifdef CONFIG_KEXEC
 607         unsigned long ptr = paddr_vmcoreinfo_note();
 608
 609         memcpy_absolute(&S390_lowcore.vmcore_info, &ptr, sizeof(ptr));
 610 #endif
 611 }
 612
 613 #ifdef CONFIG_CRASH_DUMP
 614
 615 /*
 616  * Find suitable location for crashkernel memory
 617  */
 618 static unsigned long __init find_crash_base(unsigned long crash_size,
 619                                             char **msg)
 620 {
 621         unsigned long crash_base;
 622         struct mem_chunk *chunk;
 623         int i;
 624
 625         if (memory_chunk[0].size < crash_size) {
 626                 *msg = "first memory chunk must be at least crashkernel size";
 627                 return 0;
 628         }
 629         if (OLDMEM_BASE && crash_size == OLDMEM_SIZE)
 630                 return OLDMEM_BASE;
 631
 632         for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
 633                 chunk = &memory_chunk[i];
 634                 if (chunk->size == 0)
 635                         continue;
 636                 if (chunk->type != CHUNK_READ_WRITE)
 637                         continue;
 638                 if (chunk->size < crash_size)
 639                         continue;
 640                 crash_base = (chunk->addr + chunk->size) - crash_size;
 641                 if (crash_base < crash_size)
 642                         continue;
 643                 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
 644                         continue;
 645                 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
 646                         continue;
 647                 return crash_base;
 648         }
 649         *msg = "no suitable area found";
 650         return 0;
 651 }
 652
 653 /*
 654  * Check if crash_base and crash_size is valid
 655  */
 656 static int __init verify_crash_base(unsigned long crash_base,
 657                                     unsigned long crash_size,
 658                                     char **msg)
 659 {
 660         struct mem_chunk *chunk;
 661         int i;
 662
 663         /*
 664          * Because we do the swap to zero, we must have at least 'crash_size'
 665          * bytes free space before crash_base
 666          */
 667         if (crash_size > crash_base) {
 668                 *msg = "crashkernel offset must be greater than size";
 669                 return -EINVAL;
 670         }
 671
 672         /* First memory chunk must be at least crash_size */
 673         if (memory_chunk[0].size < crash_size) {
 674                 *msg = "first memory chunk must be at least crashkernel size";
 675                 return -EINVAL;
 676         }
 677         /* Check if we fit into the respective memory chunk */
 678         for (i = 0; i < MEMORY_CHUNKS; i++) {
 679                 chunk = &memory_chunk[i];
 680                 if (chunk->size == 0)
 681                         continue;
 682                 if (crash_base < chunk->addr)
 683                         continue;
 684                 if (crash_base >= chunk->addr + chunk->size)
 685                         continue;
 686                 /* we have found the memory chunk */
 687                 if (crash_base + crash_size > chunk->addr + chunk->size) {
 688                         *msg = "selected memory chunk is too small for "
 689                                 "crashkernel memory";
 690                         return -EINVAL;
 691                 }
 692                 return 0;
 693         }
 694         *msg = "invalid memory range specified";
 695         return -EINVAL;
 696 }
 697
 698 /*
 699  * Reserve kdump memory by creating a memory hole in the mem_chunk array
 700  */
 701 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
 702                                          int type)
 703 {
 704         create_mem_hole(memory_chunk, addr, size, type);
 705 }
 706
 707 /*
 708  * When kdump is enabled, we have to ensure that no memory from
 709  * the area [0 - crashkernel memory size] and
 710  * [crashk_res.start - crashk_res.end] is set offline.
 711  */
 712 static int kdump_mem_notifier(struct notifier_block *nb,
 713                               unsigned long action, void *data)
 714 {
 715         struct memory_notify *arg = data;
 716
 717         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 718                 return NOTIFY_BAD;
 719         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 720                 return NOTIFY_OK;
 721         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 722                 return NOTIFY_OK;
 723         return NOTIFY_BAD;
 724 }
 725
 726 static struct notifier_block kdump_mem_nb = {
 727         .notifier_call = kdump_mem_notifier,
 728 };
 729
 730 #endif
 731
 732 /*
 733  * Make sure that oldmem, where the dump is stored, is protected
 734  */
 735 static void reserve_oldmem(void)
 736 {
 737 #ifdef CONFIG_CRASH_DUMP
 738         if (!OLDMEM_BASE)
 739                 return;
 740
 741         reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
 742         reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
 743                               CHUNK_OLDMEM);
 744         if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
 745                 saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
 746         else
 747                 saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
 748 #endif
 749 }
 750
 751 /*
 752  * Reserve memory for kdump kernel to be loaded with kexec
 753  */
 754 static void __init reserve_crashkernel(void)
 755 {
 756 #ifdef CONFIG_CRASH_DUMP
 757         unsigned long long crash_base, crash_size;
 758         char *msg = NULL;
 759         int rc;
 760
 761         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 762                                &crash_base);
 763         if (rc || crash_size == 0)
 764                 return;
 765         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 766         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 767         if (register_memory_notifier(&kdump_mem_nb))
 768                 return;
 769         if (!crash_base)
 770                 crash_base = find_crash_base(crash_size, &msg);
 771         if (!crash_base) {
 772                 pr_info("crashkernel reservation failed: %s\n", msg);
 773                 unregister_memory_notifier(&kdump_mem_nb);
 774                 return;
 775         }
 776         if (verify_crash_base(crash_base, crash_size, &msg)) {
 777                 pr_info("crashkernel reservation failed: %s\n", msg);
 778                 unregister_memory_notifier(&kdump_mem_nb);
 779                 return;
 780         }
 781         if (!OLDMEM_BASE && MACHINE_IS_VM)
 782                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 783         crashk_res.start = crash_base;
 784         crashk_res.end = crash_base + crash_size - 1;
 785         insert_resource(&iomem_resource, &crashk_res);
 786         reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
 787         pr_info("Reserving %lluMB of memory at %lluMB "
 788                 "for crashkernel (System RAM: %luMB)\n",
 789                 crash_size >> 20, crash_base >> 20, memory_end >> 20);
 790         os_info_crashkernel_add(crash_base, crash_size);
 791 #endif
 792 }
 793
 794 static void __init setup_memory(void)
 795 {
 796         unsigned long bootmap_size;
 797         unsigned long start_pfn, end_pfn;
 798         int i;
 799
 800         /*
 801          * partially used pages are not usable - thus
 802          * we are rounding upwards:
 803          */
 804         start_pfn = PFN_UP(__pa(&_end));
 805         end_pfn = max_pfn = PFN_DOWN(memory_end);
 806
 807 #ifdef CONFIG_BLK_DEV_INITRD
 808         /*
 809          * Move the initrd in case the bitmap of the bootmem allocater
 810          * would overwrite it.
 811          */
 812
 813         if (INITRD_START && INITRD_SIZE) {
 814                 unsigned long bmap_size;
 815                 unsigned long start;
 816
 817                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
 818                 bmap_size = PFN_PHYS(bmap_size);
 819
 820                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
 821                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
 822
 823 #ifdef CONFIG_CRASH_DUMP
 824                         if (OLDMEM_BASE) {
 825                                 /* Move initrd behind kdump oldmem */
 826                                 if (start + INITRD_SIZE > OLDMEM_BASE &&
 827                                     start < OLDMEM_BASE + OLDMEM_SIZE)
 828                                         start = OLDMEM_BASE + OLDMEM_SIZE;
 829                         }
 830 #endif
 831                         if (start + INITRD_SIZE > memory_end) {
 832                                 pr_err("initrd extends beyond end of "
 833                                        "memory (0x%08lx > 0x%08lx) "
 834                                        "disabling initrd\n",
 835                                        start + INITRD_SIZE, memory_end);
 836                                 INITRD_START = INITRD_SIZE = 0;
 837                         } else {
 838                                 pr_info("Moving initrd (0x%08lx -> "
 839                                         "0x%08lx, size: %ld)\n",
 840                                         INITRD_START, start, INITRD_SIZE);
 841                                 memmove((void *) start, (void *) INITRD_START,
 842                                         INITRD_SIZE);
 843                                 INITRD_START = start;
 844                         }
 845                 }
 846         }
 847 #endif
 848
 849         /*
 850          * Initialize the boot-time allocator
 851          */
 852         bootmap_size = init_bootmem(start_pfn, end_pfn);
 853
 854         /*
 855          * Register RAM areas with the bootmem allocator.
 856          */
 857
 858         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
 859                 unsigned long start_chunk, end_chunk, pfn;
 860
 861                 if (memory_chunk[i].type != CHUNK_READ_WRITE &&
 862                     memory_chunk[i].type != CHUNK_CRASHK)
 863                         continue;
 864                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
 865                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
 866                 end_chunk = min(end_chunk, end_pfn);
 867                 if (start_chunk >= end_chunk)
 868                         continue;
 869                 memblock_add_node(PFN_PHYS(start_chunk),
 870                                   PFN_PHYS(end_chunk - start_chunk), 0);
 871                 pfn = max(start_chunk, start_pfn);
 872                 for (; pfn < end_chunk; pfn++)
 873                         page_set_storage_key(PFN_PHYS(pfn),
 874                                              PAGE_DEFAULT_KEY, 0);
 875         }
 876
 877         psw_set_key(PAGE_DEFAULT_KEY);
 878
 879         free_bootmem_with_active_regions(0, max_pfn);
 880
 881         /*
 882          * Reserve memory used for lowcore/command line/kernel image.
 883          */
 884         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
 885         reserve_bootmem((unsigned long)_stext,
 886                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
 887                         BOOTMEM_DEFAULT);
 888         /*
 889          * Reserve the bootmem bitmap itself as well. We do this in two
 890          * steps (first step was init_bootmem()) because this catches
 891          * the (very unlikely) case of us accidentally initializing the
 892          * bootmem allocator with an invalid RAM area.
 893          */
 894         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
 895                         BOOTMEM_DEFAULT);
 896
 897 #ifdef CONFIG_CRASH_DUMP
 898         if (crashk_res.start)
 899                 reserve_bootmem(crashk_res.start,
 900                                 crashk_res.end - crashk_res.start + 1,
 901                                 BOOTMEM_DEFAULT);
 902         if (is_kdump_kernel())
 903                 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
 904                                 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
 905 #endif
 906 #ifdef CONFIG_BLK_DEV_INITRD
 907         if (INITRD_START && INITRD_SIZE) {
 908                 if (INITRD_START + INITRD_SIZE <= memory_end) {
 909                         reserve_bootmem(INITRD_START, INITRD_SIZE,
 910                                         BOOTMEM_DEFAULT);
 911                         initrd_start = INITRD_START;
 912                         initrd_end = initrd_start + INITRD_SIZE;
 913                 } else {
 914                         pr_err("initrd extends beyond end of "
 915                                "memory (0x%08lx > 0x%08lx) "
 916                                "disabling initrd\n",
 917                                initrd_start + INITRD_SIZE, memory_end);
 918                         initrd_start = initrd_end = 0;
 919                 }
 920         }
 921 #endif
 922 }
 923
 924 /*
 925  * Setup hardware capabilities.
 926  */
 927 static void __init setup_hwcaps(void)
 928 {
 929         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 930         struct cpuid cpu_id;
 931         int i;
 932
 933         /*
 934          * The store facility list bits numbers as found in the principles
 935          * of operation are numbered with bit 1UL<<31 as number 0 to
 936          * bit 1UL<<0 as number 31.
 937          *   Bit 0: instructions named N3, "backported" to esa-mode
 938          *   Bit 2: z/Architecture mode is active
 939          *   Bit 7: the store-facility-list-extended facility is installed
 940          *   Bit 17: the message-security assist is installed
 941          *   Bit 19: the long-displacement facility is installed
 942          *   Bit 21: the extended-immediate facility is installed
 943          *   Bit 22: extended-translation facility 3 is installed
 944          *   Bit 30: extended-translation facility 3 enhancement facility
 945          * These get translated to:
 946          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 947          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 948          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 949          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 950          */
 951         for (i = 0; i < 6; i++)
 952                 if (test_facility(stfl_bits[i]))
 953                         elf_hwcap |= 1UL << i;
 954
 955         if (test_facility(22) && test_facility(30))
 956                 elf_hwcap |= HWCAP_S390_ETF3EH;
 957
 958         /*
 959          * Check for additional facilities with store-facility-list-extended.
 960          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 961          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 962          * as stored by stfl, bits 32-xxx contain additional facilities.
 963          * How many facility words are stored depends on the number of
 964          * doublewords passed to the instruction. The additional facilities
 965          * are:
 966          *   Bit 42: decimal floating point facility is installed
 967          *   Bit 44: perform floating point operation facility is installed
 968          * translated to:
 969          *   HWCAP_S390_DFP bit 6 (42 && 44).
 970          */
 971         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 972                 elf_hwcap |= HWCAP_S390_DFP;
 973
 974         /*
 975          * Huge page support HWCAP_S390_HPAGE is bit 7.
 976          */
 977         if (MACHINE_HAS_HPAGE)
 978                 elf_hwcap |= HWCAP_S390_HPAGE;
 979
 980         /*
 981          * 64-bit register support for 31-bit processes
 982          * HWCAP_S390_HIGH_GPRS is bit 9.
 983          */
 984         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 985
 986         get_cpu_id(&cpu_id);
 987         switch (cpu_id.machine) {
 988         case 0x9672:
 989 #if !defined(CONFIG_64BIT)
 990         default:        /* Use "g5" as default for 31 bit kernels. */
 991 #endif
 992                 strcpy(elf_platform, "g5");
 993                 break;
 994         case 0x2064:
 995         case 0x2066:
 996 #if defined(CONFIG_64BIT)
 997         default:        /* Use "z900" as default for 64 bit kernels. */
 998 #endif
 999                 strcpy(elf_platform, "z900");
1000                 break;
1001         case 0x2084:
1002         case 0x2086:
1003                 strcpy(elf_platform, "z990");
1004                 break;
1005         case 0x2094:
1006         case 0x2096:
1007                 strcpy(elf_platform, "z9-109");
1008                 break;
1009         case 0x2097:
1010         case 0x2098:
1011                 strcpy(elf_platform, "z10");
1012                 break;
1013         case 0x2817:
1014         case 0x2818:
1015                 strcpy(elf_platform, "z196");
1016                 break;
1017         }
1018 }
1019
1020 /*
1021  * Setup function called from init/main.c just after the banner
1022  * was printed.
1023  */
1024
1025 void __init setup_arch(char **cmdline_p)
1026 {
1027         /*
1028          * print what head.S has found out about the machine
1029          */
1030 #ifndef CONFIG_64BIT
1031         if (MACHINE_IS_VM)
1032                 pr_info("Linux is running as a z/VM "
1033                         "guest operating system in 31-bit mode\n");
1034         else if (MACHINE_IS_LPAR)
1035                 pr_info("Linux is running natively in 31-bit mode\n");
1036         if (MACHINE_HAS_IEEE)
1037                 pr_info("The hardware system has IEEE compatible "
1038                         "floating point units\n");
1039         else
1040                 pr_info("The hardware system has no IEEE compatible "
1041                         "floating point units\n");
1042 #else /* CONFIG_64BIT */
1043         if (MACHINE_IS_VM)
1044                 pr_info("Linux is running as a z/VM "
1045                         "guest operating system in 64-bit mode\n");
1046         else if (MACHINE_IS_KVM)
1047                 pr_info("Linux is running under KVM in 64-bit mode\n");
1048         else if (MACHINE_IS_LPAR)
1049                 pr_info("Linux is running natively in 64-bit mode\n");
1050 #endif /* CONFIG_64BIT */
1051
1052         /* Have one command line that is parsed and saved in /proc/cmdline */
1053         /* boot_command_line has been already set up in early.c */
1054         *cmdline_p = boot_command_line;
1055
1056         ROOT_DEV = Root_RAM0;
1057
1058         init_mm.start_code = PAGE_OFFSET;
1059         init_mm.end_code = (unsigned long) &_etext;
1060         init_mm.end_data = (unsigned long) &_edata;
1061         init_mm.brk = (unsigned long) &_end;
1062
1063         if (MACHINE_HAS_MVCOS)
1064                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1065         else
1066                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1067
1068         parse_early_param();
1069
1070         os_info_init();
1071         setup_ipl();
1072         setup_memory_end();
1073         setup_addressing_mode();
1074         reserve_oldmem();
1075         reserve_crashkernel();
1076         setup_memory();
1077         setup_resources();
1078         setup_vmcoreinfo();
1079         setup_lowcore();
1080
1081         cpu_init();
1082         s390_init_cpu_topology();
1083
1084         /*
1085          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1086          */
1087         setup_hwcaps();
1088
1089         /*
1090          * Create kernel page tables and switch to virtual addressing.
1091          */
1092         paging_init();
1093
1094         /* Setup default console */
1095         conmode_default();
1096         set_preferred_console();
1097
1098         /* Setup zfcpdump support */
1099         setup_zfcpdump(console_devno);
1100 }