arch/s390/kernel/setup.c

   1 /*
   2  *  arch/s390/kernel/setup.c
   3  *
   4  *  S390 version
   5  *    Copyright (C) 1999,2000 IBM Deutschland Entwicklung GmbH, IBM Corporation
   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 #include <linux/errno.h>
  18 #include <linux/module.h>
  19 #include <linux/sched.h>
  20 #include <linux/kernel.h>
  21 #include <linux/mm.h>
  22 #include <linux/stddef.h>
  23 #include <linux/unistd.h>
  24 #include <linux/ptrace.h>
  25 #include <linux/slab.h>
  26 #include <linux/user.h>
  27 #include <linux/tty.h>
  28 #include <linux/ioport.h>
  29 #include <linux/delay.h>
  30 #include <linux/init.h>
  31 #include <linux/initrd.h>
  32 #include <linux/bootmem.h>
  33 #include <linux/root_dev.h>
  34 #include <linux/console.h>
  35 #include <linux/seq_file.h>
  36 #include <linux/kernel_stat.h>
  37 #include <linux/device.h>
  38 #include <linux/notifier.h>
  39 #include <linux/pfn.h>
  40 #include <linux/ctype.h>
  41 #include <linux/reboot.h>
  42 #include <linux/topology.h>
  43
  44 #include <asm/ipl.h>
  45 #include <asm/uaccess.h>
  46 #include <asm/system.h>
  47 #include <asm/smp.h>
  48 #include <asm/mmu_context.h>
  49 #include <asm/cpcmd.h>
  50 #include <asm/lowcore.h>
  51 #include <asm/irq.h>
  52 #include <asm/page.h>
  53 #include <asm/ptrace.h>
  54 #include <asm/sections.h>
  55 #include <asm/ebcdic.h>
  56 #include <asm/compat.h>
  57
  58 long psw_kernel_bits    = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_PRIMARY |
  59                            PSW_MASK_MCHECK | PSW_DEFAULT_KEY);
  60 long psw_user_bits      = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
  61                            PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
  62                            PSW_MASK_PSTATE | PSW_DEFAULT_KEY);
  63
  64 /*
  65  * User copy operations.
  66  */
  67 struct uaccess_ops uaccess;
  68 EXPORT_SYMBOL(uaccess);
  69
  70 /*
  71  * Machine setup..
  72  */
  73 unsigned int console_mode = 0;
  74 unsigned int console_devno = -1;
  75 unsigned int console_irq = -1;
  76 unsigned long machine_flags = 0;
  77 unsigned long elf_hwcap = 0;
  78 char elf_platform[ELF_PLATFORM_SIZE];
  79
  80 struct mem_chunk __meminitdata memory_chunk[MEMORY_CHUNKS];
  81 volatile int __cpu_logical_map[NR_CPUS]; /* logical cpu to cpu address */
  82 static unsigned long __initdata memory_end;
  83
  84 /*
  85  * This is set up by the setup-routine at boot-time
  86  * for S390 need to find out, what we have to setup
  87  * using address 0x10400 ...
  88  */
  89
  90 #include <asm/setup.h>
  91
  92 static struct resource code_resource = {
  93         .name  = "Kernel code",
  94         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
  95 };
  96
  97 static struct resource data_resource = {
  98         .name = "Kernel data",
  99         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 100 };
 101
 102 /*
 103  * cpu_init() initializes state that is per-CPU.
 104  */
 105 void __cpuinit cpu_init(void)
 106 {
 107         int addr = hard_smp_processor_id();
 108
 109         /*
 110          * Store processor id in lowcore (used e.g. in timer_interrupt)
 111          */
 112         get_cpu_id(&S390_lowcore.cpu_data.cpu_id);
 113         S390_lowcore.cpu_data.cpu_addr = addr;
 114
 115         /*
 116          * Force FPU initialization:
 117          */
 118         clear_thread_flag(TIF_USEDFPU);
 119         clear_used_math();
 120
 121         atomic_inc(&init_mm.mm_count);
 122         current->active_mm = &init_mm;
 123         if (current->mm)
 124                 BUG();
 125         enter_lazy_tlb(&init_mm, current);
 126 }
 127
 128 /*
 129  * condev= and conmode= setup parameter.
 130  */
 131
 132 static int __init condev_setup(char *str)
 133 {
 134         int vdev;
 135
 136         vdev = simple_strtoul(str, &str, 0);
 137         if (vdev >= 0 && vdev < 65536) {
 138                 console_devno = vdev;
 139                 console_irq = -1;
 140         }
 141         return 1;
 142 }
 143
 144 __setup("condev=", condev_setup);
 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         return 1;
 161 }
 162
 163 __setup("conmode=", conmode_setup);
 164
 165 static void __init conmode_default(void)
 166 {
 167         char query_buffer[1024];
 168         char *ptr;
 169
 170         if (MACHINE_IS_VM) {
 171                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 172                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 173                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 174                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 175                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 176                 ptr = strstr(query_buffer, "CONMODE");
 177                 /*
 178                  * Set the conmode to 3215 so that the device recognition
 179                  * will set the cu_type of the console to 3215. If the
 180                  * conmode is 3270 and we don't set it back then both
 181                  * 3215 and the 3270 driver will try to access the console
 182                  * device (3215 as console and 3270 as normal tty).
 183                  */
 184                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 185                 if (ptr == NULL) {
 186 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 187                         SET_CONSOLE_SCLP;
 188 #endif
 189                         return;
 190                 }
 191                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 192 #if defined(CONFIG_TN3270_CONSOLE)
 193                         SET_CONSOLE_3270;
 194 #elif defined(CONFIG_TN3215_CONSOLE)
 195                         SET_CONSOLE_3215;
 196 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 197                         SET_CONSOLE_SCLP;
 198 #endif
 199                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 200 #if defined(CONFIG_TN3215_CONSOLE)
 201                         SET_CONSOLE_3215;
 202 #elif defined(CONFIG_TN3270_CONSOLE)
 203                         SET_CONSOLE_3270;
 204 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 205                         SET_CONSOLE_SCLP;
 206 #endif
 207                 }
 208         } else if (MACHINE_IS_P390) {
 209 #if defined(CONFIG_TN3215_CONSOLE)
 210                 SET_CONSOLE_3215;
 211 #elif defined(CONFIG_TN3270_CONSOLE)
 212                 SET_CONSOLE_3270;
 213 #endif
 214         } else {
 215 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 216                 SET_CONSOLE_SCLP;
 217 #endif
 218         }
 219 }
 220
 221 #if defined(CONFIG_ZFCPDUMP) || defined(CONFIG_ZFCPDUMP_MODULE)
 222 static void __init setup_zfcpdump(unsigned int console_devno)
 223 {
 224         static char str[64];
 225
 226         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 227                 return;
 228         if (console_devno != -1)
 229                 sprintf(str, "cio_ignore=all,!0.0.%04x,!0.0.%04x",
 230                         ipl_info.data.fcp.dev_id.devno, console_devno);
 231         else
 232                 sprintf(str, "cio_ignore=all,!0.0.%04x",
 233                         ipl_info.data.fcp.dev_id.devno);
 234         strcat(COMMAND_LINE, " ");
 235         strcat(COMMAND_LINE, str);
 236         console_loglevel = 2;
 237 }
 238 #else
 239 static inline void setup_zfcpdump(unsigned int console_devno) {}
 240 #endif /* CONFIG_ZFCPDUMP */
 241
 242  /*
 243  * Reboot, halt and power_off stubs. They just call _machine_restart,
 244  * _machine_halt or _machine_power_off.
 245  */
 246
 247 void machine_restart(char *command)
 248 {
 249         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 250                 /*
 251                  * Only unblank the console if we are called in enabled
 252                  * context or a bust_spinlocks cleared the way for us.
 253                  */
 254                 console_unblank();
 255         _machine_restart(command);
 256 }
 257
 258 void machine_halt(void)
 259 {
 260         if (!in_interrupt() || oops_in_progress)
 261                 /*
 262                  * Only unblank the console if we are called in enabled
 263                  * context or a bust_spinlocks cleared the way for us.
 264                  */
 265                 console_unblank();
 266         _machine_halt();
 267 }
 268
 269 void machine_power_off(void)
 270 {
 271         if (!in_interrupt() || oops_in_progress)
 272                 /*
 273                  * Only unblank the console if we are called in enabled
 274                  * context or a bust_spinlocks cleared the way for us.
 275                  */
 276                 console_unblank();
 277         _machine_power_off();
 278 }
 279
 280 /*
 281  * Dummy power off function.
 282  */
 283 void (*pm_power_off)(void) = machine_power_off;
 284
 285 static int __init early_parse_mem(char *p)
 286 {
 287         memory_end = memparse(p, &p);
 288         return 0;
 289 }
 290 early_param("mem", early_parse_mem);
 291
 292 /*
 293  * "ipldelay=XXX[sm]" sets ipl delay in seconds or minutes
 294  */
 295 static int __init early_parse_ipldelay(char *p)
 296 {
 297         unsigned long delay = 0;
 298
 299         delay = simple_strtoul(p, &p, 0);
 300
 301         switch (*p) {
 302         case 's':
 303         case 'S':
 304                 delay *= 1000000;
 305                 break;
 306         case 'm':
 307         case 'M':
 308                 delay *= 60 * 1000000;
 309         }
 310
 311         /* now wait for the requested amount of time */
 312         udelay(delay);
 313
 314         return 0;
 315 }
 316 early_param("ipldelay", early_parse_ipldelay);
 317
 318 #ifdef CONFIG_S390_SWITCH_AMODE
 319 unsigned int switch_amode = 0;
 320 EXPORT_SYMBOL_GPL(switch_amode);
 321
 322 static void set_amode_and_uaccess(unsigned long user_amode,
 323                                   unsigned long user32_amode)
 324 {
 325         psw_user_bits = PSW_BASE_BITS | PSW_MASK_DAT | user_amode |
 326                         PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
 327                         PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
 328 #ifdef CONFIG_COMPAT
 329         psw_user32_bits = PSW_BASE32_BITS | PSW_MASK_DAT | user_amode |
 330                           PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
 331                           PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
 332         psw32_user_bits = PSW32_BASE_BITS | PSW32_MASK_DAT | user32_amode |
 333                           PSW32_MASK_IO | PSW32_MASK_EXT | PSW32_MASK_MCHECK |
 334                           PSW32_MASK_PSTATE;
 335 #endif
 336         psw_kernel_bits = PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
 337                           PSW_MASK_MCHECK | PSW_DEFAULT_KEY;
 338
 339         if (MACHINE_HAS_MVCOS) {
 340                 printk("mvcos available.\n");
 341                 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
 342         } else {
 343                 printk("mvcos not available.\n");
 344                 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
 345         }
 346 }
 347
 348 /*
 349  * Switch kernel/user addressing modes?
 350  */
 351 static int __init early_parse_switch_amode(char *p)
 352 {
 353         switch_amode = 1;
 354         return 0;
 355 }
 356 early_param("switch_amode", early_parse_switch_amode);
 357
 358 #else /* CONFIG_S390_SWITCH_AMODE */
 359 static inline void set_amode_and_uaccess(unsigned long user_amode,
 360                                          unsigned long user32_amode)
 361 {
 362 }
 363 #endif /* CONFIG_S390_SWITCH_AMODE */
 364
 365 #ifdef CONFIG_S390_EXEC_PROTECT
 366 unsigned int s390_noexec = 0;
 367 EXPORT_SYMBOL_GPL(s390_noexec);
 368
 369 /*
 370  * Enable execute protection?
 371  */
 372 static int __init early_parse_noexec(char *p)
 373 {
 374         if (!strncmp(p, "off", 3))
 375                 return 0;
 376         switch_amode = 1;
 377         s390_noexec = 1;
 378         return 0;
 379 }
 380 early_param("noexec", early_parse_noexec);
 381 #endif /* CONFIG_S390_EXEC_PROTECT */
 382
 383 static void setup_addressing_mode(void)
 384 {
 385         if (s390_noexec) {
 386                 printk("S390 execute protection active, ");
 387                 set_amode_and_uaccess(PSW_ASC_SECONDARY, PSW32_ASC_SECONDARY);
 388         } else if (switch_amode) {
 389                 printk("S390 address spaces switched, ");
 390                 set_amode_and_uaccess(PSW_ASC_PRIMARY, PSW32_ASC_PRIMARY);
 391         }
 392 #ifdef CONFIG_TRACE_IRQFLAGS
 393         sysc_restore_trace_psw.mask = psw_kernel_bits & ~PSW_MASK_MCHECK;
 394         io_restore_trace_psw.mask = psw_kernel_bits & ~PSW_MASK_MCHECK;
 395 #endif
 396 }
 397
 398 static void __init
 399 setup_lowcore(void)
 400 {
 401         struct _lowcore *lc;
 402         int lc_pages;
 403
 404         /*
 405          * Setup lowcore for boot cpu
 406          */
 407         lc_pages = sizeof(void *) == 8 ? 2 : 1;
 408         lc = (struct _lowcore *)
 409                 __alloc_bootmem(lc_pages * PAGE_SIZE, lc_pages * PAGE_SIZE, 0);
 410         memset(lc, 0, lc_pages * PAGE_SIZE);
 411         lc->restart_psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
 412         lc->restart_psw.addr =
 413                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
 414         if (switch_amode)
 415                 lc->restart_psw.mask |= PSW_ASC_HOME;
 416         lc->external_new_psw.mask = psw_kernel_bits;
 417         lc->external_new_psw.addr =
 418                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
 419         lc->svc_new_psw.mask = psw_kernel_bits | PSW_MASK_IO | PSW_MASK_EXT;
 420         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
 421         lc->program_new_psw.mask = psw_kernel_bits;
 422         lc->program_new_psw.addr =
 423                 PSW_ADDR_AMODE | (unsigned long)pgm_check_handler;
 424         lc->mcck_new_psw.mask =
 425                 psw_kernel_bits & ~PSW_MASK_MCHECK & ~PSW_MASK_DAT;
 426         lc->mcck_new_psw.addr =
 427                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
 428         lc->io_new_psw.mask = psw_kernel_bits;
 429         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 430         lc->ipl_device = S390_lowcore.ipl_device;
 431         lc->clock_comparator = -1ULL;
 432         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
 433         lc->async_stack = (unsigned long)
 434                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
 435         lc->panic_stack = (unsigned long)
 436                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
 437         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 438         lc->thread_info = (unsigned long) &init_thread_union;
 439 #ifndef CONFIG_64BIT
 440         if (MACHINE_HAS_IEEE) {
 441                 lc->extended_save_area_addr = (__u32)
 442                         __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0);
 443                 /* enable extended save area */
 444                 __ctl_set_bit(14, 29);
 445         }
 446 #endif
 447         set_prefix((u32)(unsigned long) lc);
 448 }
 449
 450 static void __init
 451 setup_resources(void)
 452 {
 453         struct resource *res, *sub_res;
 454         int i;
 455
 456         code_resource.start = (unsigned long) &_text;
 457         code_resource.end = (unsigned long) &_etext - 1;
 458         data_resource.start = (unsigned long) &_etext;
 459         data_resource.end = (unsigned long) &_edata - 1;
 460
 461         for (i = 0; i < MEMORY_CHUNKS; i++) {
 462                 if (!memory_chunk[i].size)
 463                         continue;
 464                 res = alloc_bootmem_low(sizeof(struct resource));
 465                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
 466                 switch (memory_chunk[i].type) {
 467                 case CHUNK_READ_WRITE:
 468                         res->name = "System RAM";
 469                         break;
 470                 case CHUNK_READ_ONLY:
 471                         res->name = "System ROM";
 472                         res->flags |= IORESOURCE_READONLY;
 473                         break;
 474                 default:
 475                         res->name = "reserved";
 476                 }
 477                 res->start = memory_chunk[i].addr;
 478                 res->end = memory_chunk[i].addr +  memory_chunk[i].size - 1;
 479                 request_resource(&iomem_resource, res);
 480
 481                 if (code_resource.start >= res->start  &&
 482                         code_resource.start <= res->end &&
 483                         code_resource.end > res->end) {
 484                         sub_res = alloc_bootmem_low(sizeof(struct resource));
 485                         memcpy(sub_res, &code_resource,
 486                                 sizeof(struct resource));
 487                         sub_res->end = res->end;
 488                         code_resource.start = res->end + 1;
 489                         request_resource(res, sub_res);
 490                 }
 491
 492                 if (code_resource.start >= res->start &&
 493                         code_resource.start <= res->end &&
 494                         code_resource.end <= res->end)
 495                         request_resource(res, &code_resource);
 496
 497                 if (data_resource.start >= res->start &&
 498                         data_resource.start <= res->end &&
 499                         data_resource.end > res->end) {
 500                         sub_res = alloc_bootmem_low(sizeof(struct resource));
 501                         memcpy(sub_res, &data_resource,
 502                                 sizeof(struct resource));
 503                         sub_res->end = res->end;
 504                         data_resource.start = res->end + 1;
 505                         request_resource(res, sub_res);
 506                 }
 507
 508                 if (data_resource.start >= res->start &&
 509                         data_resource.start <= res->end &&
 510                         data_resource.end <= res->end)
 511                         request_resource(res, &data_resource);
 512         }
 513 }
 514
 515 unsigned long real_memory_size;
 516 EXPORT_SYMBOL_GPL(real_memory_size);
 517
 518 static void __init setup_memory_end(void)
 519 {
 520         unsigned long memory_size;
 521         unsigned long max_mem;
 522         int i;
 523
 524 #if defined(CONFIG_ZFCPDUMP) || defined(CONFIG_ZFCPDUMP_MODULE)
 525         if (ipl_info.type == IPL_TYPE_FCP_DUMP)
 526                 memory_end = ZFCPDUMP_HSA_SIZE;
 527 #endif
 528         memory_size = 0;
 529         memory_end &= PAGE_MASK;
 530
 531         max_mem = memory_end ? min(VMEM_MAX_PHYS, memory_end) : VMEM_MAX_PHYS;
 532         memory_end = min(max_mem, memory_end);
 533
 534         /*
 535          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 536          * extra checks that HOLES_IN_ZONE would require.
 537          */
 538         for (i = 0; i < MEMORY_CHUNKS; i++) {
 539                 unsigned long start, end;
 540                 struct mem_chunk *chunk;
 541                 unsigned long align;
 542
 543                 chunk = &memory_chunk[i];
 544                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
 545                 start = (chunk->addr + align - 1) & ~(align - 1);
 546                 end = (chunk->addr + chunk->size) & ~(align - 1);
 547                 if (start >= end)
 548                         memset(chunk, 0, sizeof(*chunk));
 549                 else {
 550                         chunk->addr = start;
 551                         chunk->size = end - start;
 552                 }
 553         }
 554
 555         for (i = 0; i < MEMORY_CHUNKS; i++) {
 556                 struct mem_chunk *chunk = &memory_chunk[i];
 557
 558                 real_memory_size = max(real_memory_size,
 559                                        chunk->addr + chunk->size);
 560                 if (chunk->addr >= max_mem) {
 561                         memset(chunk, 0, sizeof(*chunk));
 562                         continue;
 563                 }
 564                 if (chunk->addr + chunk->size > max_mem)
 565                         chunk->size = max_mem - chunk->addr;
 566                 memory_size = max(memory_size, chunk->addr + chunk->size);
 567         }
 568         if (!memory_end)
 569                 memory_end = memory_size;
 570 }
 571
 572 static void __init
 573 setup_memory(void)
 574 {
 575         unsigned long bootmap_size;
 576         unsigned long start_pfn, end_pfn;
 577         int i;
 578
 579         /*
 580          * partially used pages are not usable - thus
 581          * we are rounding upwards:
 582          */
 583         start_pfn = PFN_UP(__pa(&_end));
 584         end_pfn = max_pfn = PFN_DOWN(memory_end);
 585
 586 #ifdef CONFIG_BLK_DEV_INITRD
 587         /*
 588          * Move the initrd in case the bitmap of the bootmem allocater
 589          * would overwrite it.
 590          */
 591
 592         if (INITRD_START && INITRD_SIZE) {
 593                 unsigned long bmap_size;
 594                 unsigned long start;
 595
 596                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
 597                 bmap_size = PFN_PHYS(bmap_size);
 598
 599                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
 600                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
 601
 602                         if (start + INITRD_SIZE > memory_end) {
 603                                 printk("initrd extends beyond end of memory "
 604                                        "(0x%08lx > 0x%08lx)\n"
 605                                        "disabling initrd\n",
 606                                        start + INITRD_SIZE, memory_end);
 607                                 INITRD_START = INITRD_SIZE = 0;
 608                         } else {
 609                                 printk("Moving initrd (0x%08lx -> 0x%08lx, "
 610                                        "size: %ld)\n",
 611                                        INITRD_START, start, INITRD_SIZE);
 612                                 memmove((void *) start, (void *) INITRD_START,
 613                                         INITRD_SIZE);
 614                                 INITRD_START = start;
 615                         }
 616                 }
 617         }
 618 #endif
 619
 620         /*
 621          * Initialize the boot-time allocator
 622          */
 623         bootmap_size = init_bootmem(start_pfn, end_pfn);
 624
 625         /*
 626          * Register RAM areas with the bootmem allocator.
 627          */
 628
 629         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
 630                 unsigned long start_chunk, end_chunk, pfn;
 631
 632                 if (memory_chunk[i].type != CHUNK_READ_WRITE)
 633                         continue;
 634                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
 635                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size) - 1;
 636                 end_chunk = min(end_chunk, end_pfn);
 637                 if (start_chunk >= end_chunk)
 638                         continue;
 639                 add_active_range(0, start_chunk, end_chunk);
 640                 pfn = max(start_chunk, start_pfn);
 641                 for (; pfn <= end_chunk; pfn++)
 642                         page_set_storage_key(PFN_PHYS(pfn), PAGE_DEFAULT_KEY);
 643         }
 644
 645         psw_set_key(PAGE_DEFAULT_KEY);
 646
 647         free_bootmem_with_active_regions(0, max_pfn);
 648
 649         /*
 650          * Reserve memory used for lowcore/command line/kernel image.
 651          */
 652         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
 653         reserve_bootmem((unsigned long)_stext,
 654                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
 655                         BOOTMEM_DEFAULT);
 656         /*
 657          * Reserve the bootmem bitmap itself as well. We do this in two
 658          * steps (first step was init_bootmem()) because this catches
 659          * the (very unlikely) case of us accidentally initializing the
 660          * bootmem allocator with an invalid RAM area.
 661          */
 662         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
 663                         BOOTMEM_DEFAULT);
 664
 665 #ifdef CONFIG_BLK_DEV_INITRD
 666         if (INITRD_START && INITRD_SIZE) {
 667                 if (INITRD_START + INITRD_SIZE <= memory_end) {
 668                         reserve_bootmem(INITRD_START, INITRD_SIZE,
 669                                         BOOTMEM_DEFAULT);
 670                         initrd_start = INITRD_START;
 671                         initrd_end = initrd_start + INITRD_SIZE;
 672                 } else {
 673                         printk("initrd extends beyond end of memory "
 674                                "(0x%08lx > 0x%08lx)\ndisabling initrd\n",
 675                                initrd_start + INITRD_SIZE, memory_end);
 676                         initrd_start = initrd_end = 0;
 677                 }
 678         }
 679 #endif
 680 }
 681
 682 static __init unsigned int stfl(void)
 683 {
 684         asm volatile(
 685                 "       .insn   s,0xb2b10000,0(0)\n" /* stfl */
 686                 "0:\n"
 687                 EX_TABLE(0b,0b));
 688         return S390_lowcore.stfl_fac_list;
 689 }
 690
 691 static int __init __stfle(unsigned long long *list, int doublewords)
 692 {
 693         typedef struct { unsigned long long _[doublewords]; } addrtype;
 694         register unsigned long __nr asm("0") = doublewords - 1;
 695
 696         asm volatile(".insn s,0xb2b00000,%0" /* stfle */
 697                      : "=m" (*(addrtype *) list), "+d" (__nr) : : "cc");
 698         return __nr + 1;
 699 }
 700
 701 int __init stfle(unsigned long long *list, int doublewords)
 702 {
 703         if (!(stfl() & (1UL << 24)))
 704                 return -EOPNOTSUPP;
 705         return __stfle(list, doublewords);
 706 }
 707
 708 /*
 709  * Setup hardware capabilities.
 710  */
 711 static void __init setup_hwcaps(void)
 712 {
 713         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 714         struct cpuinfo_S390 *cpuinfo = &S390_lowcore.cpu_data;
 715         unsigned long long facility_list_extended;
 716         unsigned int facility_list;
 717         int i;
 718
 719         facility_list = stfl();
 720         /*
 721          * The store facility list bits numbers as found in the principles
 722          * of operation are numbered with bit 1UL<<31 as number 0 to
 723          * bit 1UL<<0 as number 31.
 724          *   Bit 0: instructions named N3, "backported" to esa-mode
 725          *   Bit 2: z/Architecture mode is active
 726          *   Bit 7: the store-facility-list-extended facility is installed
 727          *   Bit 17: the message-security assist is installed
 728          *   Bit 19: the long-displacement facility is installed
 729          *   Bit 21: the extended-immediate facility is installed
 730          * These get translated to:
 731          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 732          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 733          *   HWCAP_S390_LDISP bit 4, and HWCAP_S390_EIMM bit 5.
 734          */
 735         for (i = 0; i < 6; i++)
 736                 if (facility_list & (1UL << (31 - stfl_bits[i])))
 737                         elf_hwcap |= 1UL << i;
 738
 739         /*
 740          * Check for additional facilities with store-facility-list-extended.
 741          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 742          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 743          * as stored by stfl, bits 32-xxx contain additional facilities.
 744          * How many facility words are stored depends on the number of
 745          * doublewords passed to the instruction. The additional facilites
 746          * are:
 747          *   Bit 43: decimal floating point facility is installed
 748          * translated to:
 749          *   HWCAP_S390_DFP bit 6.
 750          */
 751         if ((elf_hwcap & (1UL << 2)) &&
 752             __stfle(&facility_list_extended, 1) > 0) {
 753                 if (facility_list_extended & (1ULL << (64 - 43)))
 754                         elf_hwcap |= 1UL << 6;
 755         }
 756
 757         switch (cpuinfo->cpu_id.machine) {
 758         case 0x9672:
 759 #if !defined(CONFIG_64BIT)
 760         default:        /* Use "g5" as default for 31 bit kernels. */
 761 #endif
 762                 strcpy(elf_platform, "g5");
 763                 break;
 764         case 0x2064:
 765         case 0x2066:
 766 #if defined(CONFIG_64BIT)
 767         default:        /* Use "z900" as default for 64 bit kernels. */
 768 #endif
 769                 strcpy(elf_platform, "z900");
 770                 break;
 771         case 0x2084:
 772         case 0x2086:
 773                 strcpy(elf_platform, "z990");
 774                 break;
 775         case 0x2094:
 776                 strcpy(elf_platform, "z9-109");
 777                 break;
 778         }
 779 }
 780
 781 /*
 782  * Setup function called from init/main.c just after the banner
 783  * was printed.
 784  */
 785
 786 void __init
 787 setup_arch(char **cmdline_p)
 788 {
 789         /*
 790          * print what head.S has found out about the machine
 791          */
 792 #ifndef CONFIG_64BIT
 793         printk((MACHINE_IS_VM) ?
 794                "We are running under VM (31 bit mode)\n" :
 795                "We are running native (31 bit mode)\n");
 796         printk((MACHINE_HAS_IEEE) ?
 797                "This machine has an IEEE fpu\n" :
 798                "This machine has no IEEE fpu\n");
 799 #else /* CONFIG_64BIT */
 800         printk((MACHINE_IS_VM) ?
 801                "We are running under VM (64 bit mode)\n" :
 802                "We are running native (64 bit mode)\n");
 803 #endif /* CONFIG_64BIT */
 804
 805         /* Save unparsed command line copy for /proc/cmdline */
 806         strlcpy(boot_command_line, COMMAND_LINE, COMMAND_LINE_SIZE);
 807
 808         *cmdline_p = COMMAND_LINE;
 809         *(*cmdline_p + COMMAND_LINE_SIZE - 1) = '\0';
 810
 811         ROOT_DEV = Root_RAM0;
 812
 813         init_mm.start_code = PAGE_OFFSET;
 814         init_mm.end_code = (unsigned long) &_etext;
 815         init_mm.end_data = (unsigned long) &_edata;
 816         init_mm.brk = (unsigned long) &_end;
 817
 818         if (MACHINE_HAS_MVCOS)
 819                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
 820         else
 821                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
 822
 823         parse_early_param();
 824
 825         setup_ipl();
 826         setup_memory_end();
 827         setup_addressing_mode();
 828         setup_memory();
 829         setup_resources();
 830         setup_lowcore();
 831
 832         cpu_init();
 833         __cpu_logical_map[0] = S390_lowcore.cpu_data.cpu_addr;
 834         s390_init_cpu_topology();
 835
 836         /*
 837          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
 838          */
 839         setup_hwcaps();
 840
 841         /*
 842          * Create kernel page tables and switch to virtual addressing.
 843          */
 844         paging_init();
 845
 846         /* Setup default console */
 847         conmode_default();
 848
 849         /* Setup zfcpdump support */
 850         setup_zfcpdump(console_devno);
 851 }
 852
 853 void __cpuinit print_cpu_info(struct cpuinfo_S390 *cpuinfo)
 854 {
 855    printk(KERN_INFO "cpu %d "
 856 #ifdef CONFIG_SMP
 857            "phys_idx=%d "
 858 #endif
 859            "vers=%02X ident=%06X machine=%04X unused=%04X\n",
 860            cpuinfo->cpu_nr,
 861 #ifdef CONFIG_SMP
 862            cpuinfo->cpu_addr,
 863 #endif
 864            cpuinfo->cpu_id.version,
 865            cpuinfo->cpu_id.ident,
 866            cpuinfo->cpu_id.machine,
 867            cpuinfo->cpu_id.unused);
 868 }
 869
 870 /*
 871  * show_cpuinfo - Get information on one CPU for use by procfs.
 872  */
 873
 874 static int show_cpuinfo(struct seq_file *m, void *v)
 875 {
 876         static const char *hwcap_str[7] = {
 877                 "esan3", "zarch", "stfle", "msa", "ldisp", "eimm", "dfp"
 878         };
 879         struct cpuinfo_S390 *cpuinfo;
 880         unsigned long n = (unsigned long) v - 1;
 881         int i;
 882
 883         s390_adjust_jiffies();
 884         preempt_disable();
 885         if (!n) {
 886                 seq_printf(m, "vendor_id       : IBM/S390\n"
 887                                "# processors    : %i\n"
 888                                "bogomips per cpu: %lu.%02lu\n",
 889                                num_online_cpus(), loops_per_jiffy/(500000/HZ),
 890                                (loops_per_jiffy/(5000/HZ))%100);
 891                 seq_puts(m, "features\t: ");
 892                 for (i = 0; i < 7; i++)
 893                         if (hwcap_str[i] && (elf_hwcap & (1UL << i)))
 894                                 seq_printf(m, "%s ", hwcap_str[i]);
 895                 seq_puts(m, "\n");
 896         }
 897
 898         if (cpu_online(n)) {
 899 #ifdef CONFIG_SMP
 900                 if (smp_processor_id() == n)
 901                         cpuinfo = &S390_lowcore.cpu_data;
 902                 else
 903                         cpuinfo = &lowcore_ptr[n]->cpu_data;
 904 #else
 905                 cpuinfo = &S390_lowcore.cpu_data;
 906 #endif
 907                 seq_printf(m, "processor %li: "
 908                                "version = %02X,  "
 909                                "identification = %06X,  "
 910                                "machine = %04X\n",
 911                                n, cpuinfo->cpu_id.version,
 912                                cpuinfo->cpu_id.ident,
 913                                cpuinfo->cpu_id.machine);
 914         }
 915         preempt_enable();
 916         return 0;
 917 }
 918
 919 static void *c_start(struct seq_file *m, loff_t *pos)
 920 {
 921         return *pos < NR_CPUS ? (void *)((unsigned long) *pos + 1) : NULL;
 922 }
 923 static void *c_next(struct seq_file *m, void *v, loff_t *pos)
 924 {
 925         ++*pos;
 926         return c_start(m, pos);
 927 }
 928 static void c_stop(struct seq_file *m, void *v)
 929 {
 930 }
 931 const struct seq_operations cpuinfo_op = {
 932         .start  = c_start,
 933         .next   = c_next,
 934         .stop   = c_stop,
 935         .show   = show_cpuinfo,
 936 };
 937