arch/s390/kernel/setup.c

   1 /*
   2  *  S390 version
   3  *    Copyright IBM Corp. 1999, 2012
   4  *    Author(s): Hartmut Penner (hp@de.ibm.com),
   5  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
   6  *
   7  *  Derived from "arch/i386/kernel/setup.c"
   8  *    Copyright (C) 1995, Linus Torvalds
   9  */
  10
  11 /*
  12  * This file handles the architecture-dependent parts of initialization
  13  */
  14
  15 #define KMSG_COMPONENT "setup"
  16 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  17
  18 #include <linux/errno.h>
  19 #include <linux/module.h>
  20 #include <linux/sched.h>
  21 #include <linux/kernel.h>
  22 #include <linux/memblock.h>
  23 #include <linux/mm.h>
  24 #include <linux/stddef.h>
  25 #include <linux/unistd.h>
  26 #include <linux/ptrace.h>
  27 #include <linux/user.h>
  28 #include <linux/tty.h>
  29 #include <linux/ioport.h>
  30 #include <linux/delay.h>
  31 #include <linux/init.h>
  32 #include <linux/initrd.h>
  33 #include <linux/bootmem.h>
  34 #include <linux/root_dev.h>
  35 #include <linux/console.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 #include <linux/ftrace.h>
  44 #include <linux/kexec.h>
  45 #include <linux/crash_dump.h>
  46 #include <linux/memory.h>
  47 #include <linux/compat.h>
  48
  49 #include <asm/ipl.h>
  50 #include <asm/uaccess.h>
  51 #include <asm/facility.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/kvm_virtio.h>
  62 #include <asm/diag.h>
  63 #include <asm/os_info.h>
  64 #include <asm/sclp.h>
  65 #include "entry.h"
  66
  67 long psw_kernel_bits    = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_ASC_PRIMARY |
  68                           PSW_MASK_EA | PSW_MASK_BA;
  69 long psw_user_bits      = PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT |
  70                           PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK |
  71                           PSW_MASK_PSTATE | PSW_ASC_HOME;
  72
  73 /*
  74  * User copy operations.
  75  */
  76 struct uaccess_ops uaccess;
  77 EXPORT_SYMBOL(uaccess);
  78
  79 /*
  80  * Machine setup..
  81  */
  82 unsigned int console_mode = 0;
  83 EXPORT_SYMBOL(console_mode);
  84
  85 unsigned int console_devno = -1;
  86 EXPORT_SYMBOL(console_devno);
  87
  88 unsigned int console_irq = -1;
  89 EXPORT_SYMBOL(console_irq);
  90
  91 unsigned long elf_hwcap = 0;
  92 char elf_platform[ELF_PLATFORM_SIZE];
  93
  94 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
  95
  96 int __initdata memory_end_set;
  97 unsigned long __initdata memory_end;
  98
  99 unsigned long VMALLOC_START;
 100 EXPORT_SYMBOL(VMALLOC_START);
 101
 102 unsigned long VMALLOC_END;
 103 EXPORT_SYMBOL(VMALLOC_END);
 104
 105 struct page *vmemmap;
 106 EXPORT_SYMBOL(vmemmap);
 107
 108 /* An array with a pointer to the lowcore of every CPU. */
 109 struct _lowcore *lowcore_ptr[NR_CPUS];
 110 EXPORT_SYMBOL(lowcore_ptr);
 111
 112 /*
 113  * This is set up by the setup-routine at boot-time
 114  * for S390 need to find out, what we have to setup
 115  * using address 0x10400 ...
 116  */
 117
 118 #include <asm/setup.h>
 119
 120 /*
 121  * condev= and conmode= setup parameter.
 122  */
 123
 124 static int __init condev_setup(char *str)
 125 {
 126         int vdev;
 127
 128         vdev = simple_strtoul(str, &str, 0);
 129         if (vdev >= 0 && vdev < 65536) {
 130                 console_devno = vdev;
 131                 console_irq = -1;
 132         }
 133         return 1;
 134 }
 135
 136 __setup("condev=", condev_setup);
 137
 138 static void __init set_preferred_console(void)
 139 {
 140         if (MACHINE_IS_KVM) {
 141                 if (sclp_has_vt220())
 142                         add_preferred_console("ttyS", 1, NULL);
 143                 else if (sclp_has_linemode())
 144                         add_preferred_console("ttyS", 0, NULL);
 145                 else
 146                         add_preferred_console("hvc", 0, NULL);
 147         } else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 148                 add_preferred_console("ttyS", 0, NULL);
 149         else if (CONSOLE_IS_3270)
 150                 add_preferred_console("tty3270", 0, NULL);
 151 }
 152
 153 static int __init conmode_setup(char *str)
 154 {
 155 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 156         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 157                 SET_CONSOLE_SCLP;
 158 #endif
 159 #if defined(CONFIG_TN3215_CONSOLE)
 160         if (strncmp(str, "3215", 5) == 0)
 161                 SET_CONSOLE_3215;
 162 #endif
 163 #if defined(CONFIG_TN3270_CONSOLE)
 164         if (strncmp(str, "3270", 5) == 0)
 165                 SET_CONSOLE_3270;
 166 #endif
 167         set_preferred_console();
 168         return 1;
 169 }
 170
 171 __setup("conmode=", conmode_setup);
 172
 173 static void __init conmode_default(void)
 174 {
 175         char query_buffer[1024];
 176         char *ptr;
 177
 178         if (MACHINE_IS_VM) {
 179                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 180                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 181                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 182                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 183                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 184                 ptr = strstr(query_buffer, "CONMODE");
 185                 /*
 186                  * Set the conmode to 3215 so that the device recognition
 187                  * will set the cu_type of the console to 3215. If the
 188                  * conmode is 3270 and we don't set it back then both
 189                  * 3215 and the 3270 driver will try to access the console
 190                  * device (3215 as console and 3270 as normal tty).
 191                  */
 192                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 193                 if (ptr == NULL) {
 194 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 195                         SET_CONSOLE_SCLP;
 196 #endif
 197                         return;
 198                 }
 199                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 200 #if defined(CONFIG_TN3270_CONSOLE)
 201                         SET_CONSOLE_3270;
 202 #elif defined(CONFIG_TN3215_CONSOLE)
 203                         SET_CONSOLE_3215;
 204 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 205                         SET_CONSOLE_SCLP;
 206 #endif
 207                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 208 #if defined(CONFIG_TN3215_CONSOLE)
 209                         SET_CONSOLE_3215;
 210 #elif defined(CONFIG_TN3270_CONSOLE)
 211                         SET_CONSOLE_3270;
 212 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 213                         SET_CONSOLE_SCLP;
 214 #endif
 215                 }
 216         } else {
 217 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 218                 SET_CONSOLE_SCLP;
 219 #endif
 220         }
 221 }
 222
 223 #ifdef CONFIG_ZFCPDUMP
 224 static void __init setup_zfcpdump(unsigned int console_devno)
 225 {
 226         static char str[41];
 227
 228         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 229                 return;
 230         if (OLDMEM_BASE)
 231                 return;
 232         if (console_devno != -1)
 233                 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
 234                         ipl_info.data.fcp.dev_id.devno, console_devno);
 235         else
 236                 sprintf(str, " cio_ignore=all,!0.0.%04x",
 237                         ipl_info.data.fcp.dev_id.devno);
 238         strcat(boot_command_line, str);
 239         console_loglevel = 2;
 240 }
 241 #else
 242 static inline void setup_zfcpdump(unsigned int console_devno) {}
 243 #endif /* CONFIG_ZFCPDUMP */
 244
 245  /*
 246  * Reboot, halt and power_off stubs. They just call _machine_restart,
 247  * _machine_halt or _machine_power_off.
 248  */
 249
 250 void machine_restart(char *command)
 251 {
 252         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 253                 /*
 254                  * Only unblank the console if we are called in enabled
 255                  * context or a bust_spinlocks cleared the way for us.
 256                  */
 257                 console_unblank();
 258         _machine_restart(command);
 259 }
 260
 261 void machine_halt(void)
 262 {
 263         if (!in_interrupt() || oops_in_progress)
 264                 /*
 265                  * Only unblank the console if we are called in enabled
 266                  * context or a bust_spinlocks cleared the way for us.
 267                  */
 268                 console_unblank();
 269         _machine_halt();
 270 }
 271
 272 void machine_power_off(void)
 273 {
 274         if (!in_interrupt() || oops_in_progress)
 275                 /*
 276                  * Only unblank the console if we are called in enabled
 277                  * context or a bust_spinlocks cleared the way for us.
 278                  */
 279                 console_unblank();
 280         _machine_power_off();
 281 }
 282
 283 /*
 284  * Dummy power off function.
 285  */
 286 void (*pm_power_off)(void) = machine_power_off;
 287
 288 static int __init early_parse_mem(char *p)
 289 {
 290         memory_end = memparse(p, &p);
 291         memory_end_set = 1;
 292         return 0;
 293 }
 294 early_param("mem", early_parse_mem);
 295
 296 static int __init parse_vmalloc(char *arg)
 297 {
 298         if (!arg)
 299                 return -EINVAL;
 300         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
 301         return 0;
 302 }
 303 early_param("vmalloc", parse_vmalloc);
 304
 305 unsigned int s390_user_mode = PRIMARY_SPACE_MODE;
 306 EXPORT_SYMBOL_GPL(s390_user_mode);
 307
 308 static void __init set_user_mode_primary(void)
 309 {
 310         psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
 311         psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
 312 #ifdef CONFIG_COMPAT
 313         psw32_user_bits =
 314                 (psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
 315 #endif
 316         uaccess = MACHINE_HAS_MVCOS ? uaccess_mvcos_switch : uaccess_pt;
 317 }
 318
 319 static int __init early_parse_user_mode(char *p)
 320 {
 321         if (p && strcmp(p, "primary") == 0)
 322                 s390_user_mode = PRIMARY_SPACE_MODE;
 323         else if (!p || strcmp(p, "home") == 0)
 324                 s390_user_mode = HOME_SPACE_MODE;
 325         else
 326                 return 1;
 327         return 0;
 328 }
 329 early_param("user_mode", early_parse_user_mode);
 330
 331 static void __init setup_addressing_mode(void)
 332 {
 333         if (s390_user_mode != PRIMARY_SPACE_MODE)
 334                 return;
 335         set_user_mode_primary();
 336         if (MACHINE_HAS_MVCOS)
 337                 pr_info("Address spaces switched, mvcos available\n");
 338         else
 339                 pr_info("Address spaces switched, mvcos not available\n");
 340 }
 341
 342 void *restart_stack __attribute__((__section__(".data")));
 343
 344 static void __init setup_lowcore(void)
 345 {
 346         struct _lowcore *lc;
 347
 348         /*
 349          * Setup lowcore for boot cpu
 350          */
 351         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
 352         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 353         lc->restart_psw.mask = psw_kernel_bits;
 354         lc->restart_psw.addr =
 355                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
 356         lc->external_new_psw.mask = psw_kernel_bits |
 357                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 358         lc->external_new_psw.addr =
 359                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
 360         lc->svc_new_psw.mask = psw_kernel_bits |
 361                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 362         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
 363         lc->program_new_psw.mask = psw_kernel_bits |
 364                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 365         lc->program_new_psw.addr =
 366                 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
 367         lc->mcck_new_psw.mask = psw_kernel_bits;
 368         lc->mcck_new_psw.addr =
 369                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
 370         lc->io_new_psw.mask = psw_kernel_bits |
 371                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 372         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 373         lc->clock_comparator = -1ULL;
 374         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
 375         lc->async_stack = (unsigned long)
 376                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
 377         lc->panic_stack = (unsigned long)
 378                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
 379         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 380         lc->thread_info = (unsigned long) &init_thread_union;
 381         lc->machine_flags = S390_lowcore.machine_flags;
 382         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 383         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 384                MAX_FACILITY_BIT/8);
 385 #ifndef CONFIG_64BIT
 386         if (MACHINE_HAS_IEEE) {
 387                 lc->extended_save_area_addr = (__u32)
 388                         __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
 389                 /* enable extended save area */
 390                 __ctl_set_bit(14, 29);
 391         }
 392 #else
 393         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
 394 #endif
 395         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 396         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 397         lc->exit_timer = S390_lowcore.exit_timer;
 398         lc->user_timer = S390_lowcore.user_timer;
 399         lc->system_timer = S390_lowcore.system_timer;
 400         lc->steal_timer = S390_lowcore.steal_timer;
 401         lc->last_update_timer = S390_lowcore.last_update_timer;
 402         lc->last_update_clock = S390_lowcore.last_update_clock;
 403         lc->ftrace_func = S390_lowcore.ftrace_func;
 404
 405         restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
 406         restart_stack += ASYNC_SIZE;
 407
 408         /*
 409          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 410          * restart data to the absolute zero lowcore. This is necesary if
 411          * PSW restart is done on an offline CPU that has lowcore zero.
 412          */
 413         lc->restart_stack = (unsigned long) restart_stack;
 414         lc->restart_fn = (unsigned long) do_restart;
 415         lc->restart_data = 0;
 416         lc->restart_source = -1UL;
 417
 418         /* Setup absolute zero lowcore */
 419         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 420         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 421         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 422         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 423         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 424
 425         set_prefix((u32)(unsigned long) lc);
 426         lowcore_ptr[0] = lc;
 427 }
 428
 429 static struct resource code_resource = {
 430         .name  = "Kernel code",
 431         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 432 };
 433
 434 static struct resource data_resource = {
 435         .name = "Kernel data",
 436         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 437 };
 438
 439 static struct resource bss_resource = {
 440         .name = "Kernel bss",
 441         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 442 };
 443
 444 static struct resource __initdata *standard_resources[] = {
 445         &code_resource,
 446         &data_resource,
 447         &bss_resource,
 448 };
 449
 450 static void __init setup_resources(void)
 451 {
 452         struct resource *res, *std_res, *sub_res;
 453         int i, j;
 454
 455         code_resource.start = (unsigned long) &_text;
 456         code_resource.end = (unsigned long) &_etext - 1;
 457         data_resource.start = (unsigned long) &_etext;
 458         data_resource.end = (unsigned long) &_edata - 1;
 459         bss_resource.start = (unsigned long) &__bss_start;
 460         bss_resource.end = (unsigned long) &__bss_stop - 1;
 461
 462         for (i = 0; i < MEMORY_CHUNKS; i++) {
 463                 if (!memory_chunk[i].size)
 464                         continue;
 465                 if (memory_chunk[i].type == CHUNK_OLDMEM ||
 466                     memory_chunk[i].type == CHUNK_CRASHK)
 467                         continue;
 468                 res = alloc_bootmem_low(sizeof(*res));
 469                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
 470                 switch (memory_chunk[i].type) {
 471                 case CHUNK_READ_WRITE:
 472                 case CHUNK_CRASHK:
 473                         res->name = "System RAM";
 474                         break;
 475                 case CHUNK_READ_ONLY:
 476                         res->name = "System ROM";
 477                         res->flags |= IORESOURCE_READONLY;
 478                         break;
 479                 default:
 480                         res->name = "reserved";
 481                 }
 482                 res->start = memory_chunk[i].addr;
 483                 res->end = res->start + memory_chunk[i].size - 1;
 484                 request_resource(&iomem_resource, res);
 485
 486                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 487                         std_res = standard_resources[j];
 488                         if (std_res->start < res->start ||
 489                             std_res->start > res->end)
 490                                 continue;
 491                         if (std_res->end > res->end) {
 492                                 sub_res = alloc_bootmem_low(sizeof(*sub_res));
 493                                 *sub_res = *std_res;
 494                                 sub_res->end = res->end;
 495                                 std_res->start = res->end + 1;
 496                                 request_resource(res, sub_res);
 497                         } else {
 498                                 request_resource(res, std_res);
 499                         }
 500                 }
 501         }
 502 }
 503
 504 unsigned long real_memory_size;
 505 EXPORT_SYMBOL_GPL(real_memory_size);
 506
 507 static void __init setup_memory_end(void)
 508 {
 509         unsigned long vmax, vmalloc_size, tmp;
 510         int i;
 511
 512
 513 #ifdef CONFIG_ZFCPDUMP
 514         if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) {
 515                 memory_end = ZFCPDUMP_HSA_SIZE;
 516                 memory_end_set = 1;
 517         }
 518 #endif
 519         real_memory_size = 0;
 520         memory_end &= PAGE_MASK;
 521
 522         /*
 523          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 524          * extra checks that HOLES_IN_ZONE would require.
 525          */
 526         for (i = 0; i < MEMORY_CHUNKS; i++) {
 527                 unsigned long start, end;
 528                 struct mem_chunk *chunk;
 529                 unsigned long align;
 530
 531                 chunk = &memory_chunk[i];
 532                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
 533                 start = (chunk->addr + align - 1) & ~(align - 1);
 534                 end = (chunk->addr + chunk->size) & ~(align - 1);
 535                 if (start >= end)
 536                         memset(chunk, 0, sizeof(*chunk));
 537                 else {
 538                         chunk->addr = start;
 539                         chunk->size = end - start;
 540                 }
 541                 real_memory_size = max(real_memory_size,
 542                                        chunk->addr + chunk->size);
 543         }
 544
 545         /* Choose kernel address space layout: 2, 3, or 4 levels. */
 546 #ifdef CONFIG_64BIT
 547         vmalloc_size = VMALLOC_END ?: 128UL << 30;
 548         tmp = (memory_end ?: real_memory_size) / PAGE_SIZE;
 549         tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
 550         if (tmp <= (1UL << 42))
 551                 vmax = 1UL << 42;       /* 3-level kernel page table */
 552         else
 553                 vmax = 1UL << 53;       /* 4-level kernel page table */
 554 #else
 555         vmalloc_size = VMALLOC_END ?: 96UL << 20;
 556         vmax = 1UL << 31;               /* 2-level kernel page table */
 557 #endif
 558         /* vmalloc area is at the end of the kernel address space. */
 559         VMALLOC_END = vmax;
 560         VMALLOC_START = vmax - vmalloc_size;
 561
 562         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 563         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 564         tmp = VMALLOC_START - tmp * sizeof(struct page);
 565         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 566         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 567         vmemmap = (struct page *) tmp;
 568
 569         /* Take care that memory_end is set and <= vmemmap */
 570         memory_end = min(memory_end ?: real_memory_size, tmp);
 571
 572         /* Fixup memory chunk array to fit into 0..memory_end */
 573         for (i = 0; i < MEMORY_CHUNKS; i++) {
 574                 struct mem_chunk *chunk = &memory_chunk[i];
 575
 576                 if (chunk->addr >= memory_end) {
 577                         memset(chunk, 0, sizeof(*chunk));
 578                         continue;
 579                 }
 580                 if (chunk->addr + chunk->size > memory_end)
 581                         chunk->size = memory_end - chunk->addr;
 582         }
 583 }
 584
 585 static void __init setup_vmcoreinfo(void)
 586 {
 587         mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
 588 }
 589
 590 #ifdef CONFIG_CRASH_DUMP
 591
 592 /*
 593  * Find suitable location for crashkernel memory
 594  */
 595 static unsigned long __init find_crash_base(unsigned long crash_size,
 596                                             char **msg)
 597 {
 598         unsigned long crash_base;
 599         struct mem_chunk *chunk;
 600         int i;
 601
 602         if (memory_chunk[0].size < crash_size) {
 603                 *msg = "first memory chunk must be at least crashkernel size";
 604                 return 0;
 605         }
 606         if (OLDMEM_BASE && crash_size == OLDMEM_SIZE)
 607                 return OLDMEM_BASE;
 608
 609         for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
 610                 chunk = &memory_chunk[i];
 611                 if (chunk->size == 0)
 612                         continue;
 613                 if (chunk->type != CHUNK_READ_WRITE)
 614                         continue;
 615                 if (chunk->size < crash_size)
 616                         continue;
 617                 crash_base = (chunk->addr + chunk->size) - crash_size;
 618                 if (crash_base < crash_size)
 619                         continue;
 620                 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
 621                         continue;
 622                 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
 623                         continue;
 624                 return crash_base;
 625         }
 626         *msg = "no suitable area found";
 627         return 0;
 628 }
 629
 630 /*
 631  * Check if crash_base and crash_size is valid
 632  */
 633 static int __init verify_crash_base(unsigned long crash_base,
 634                                     unsigned long crash_size,
 635                                     char **msg)
 636 {
 637         struct mem_chunk *chunk;
 638         int i;
 639
 640         /*
 641          * Because we do the swap to zero, we must have at least 'crash_size'
 642          * bytes free space before crash_base
 643          */
 644         if (crash_size > crash_base) {
 645                 *msg = "crashkernel offset must be greater than size";
 646                 return -EINVAL;
 647         }
 648
 649         /* First memory chunk must be at least crash_size */
 650         if (memory_chunk[0].size < crash_size) {
 651                 *msg = "first memory chunk must be at least crashkernel size";
 652                 return -EINVAL;
 653         }
 654         /* Check if we fit into the respective memory chunk */
 655         for (i = 0; i < MEMORY_CHUNKS; i++) {
 656                 chunk = &memory_chunk[i];
 657                 if (chunk->size == 0)
 658                         continue;
 659                 if (crash_base < chunk->addr)
 660                         continue;
 661                 if (crash_base >= chunk->addr + chunk->size)
 662                         continue;
 663                 /* we have found the memory chunk */
 664                 if (crash_base + crash_size > chunk->addr + chunk->size) {
 665                         *msg = "selected memory chunk is too small for "
 666                                 "crashkernel memory";
 667                         return -EINVAL;
 668                 }
 669                 return 0;
 670         }
 671         *msg = "invalid memory range specified";
 672         return -EINVAL;
 673 }
 674
 675 /*
 676  * Reserve kdump memory by creating a memory hole in the mem_chunk array
 677  */
 678 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
 679                                          int type)
 680 {
 681         create_mem_hole(memory_chunk, addr, size, type);
 682 }
 683
 684 /*
 685  * When kdump is enabled, we have to ensure that no memory from
 686  * the area [0 - crashkernel memory size] and
 687  * [crashk_res.start - crashk_res.end] is set offline.
 688  */
 689 static int kdump_mem_notifier(struct notifier_block *nb,
 690                               unsigned long action, void *data)
 691 {
 692         struct memory_notify *arg = data;
 693
 694         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 695                 return NOTIFY_BAD;
 696         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 697                 return NOTIFY_OK;
 698         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 699                 return NOTIFY_OK;
 700         return NOTIFY_BAD;
 701 }
 702
 703 static struct notifier_block kdump_mem_nb = {
 704         .notifier_call = kdump_mem_notifier,
 705 };
 706
 707 #endif
 708
 709 /*
 710  * Make sure that oldmem, where the dump is stored, is protected
 711  */
 712 static void reserve_oldmem(void)
 713 {
 714 #ifdef CONFIG_CRASH_DUMP
 715         if (!OLDMEM_BASE)
 716                 return;
 717
 718         reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
 719         reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
 720                               CHUNK_OLDMEM);
 721         if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
 722                 saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
 723         else
 724                 saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
 725 #endif
 726 }
 727
 728 /*
 729  * Reserve memory for kdump kernel to be loaded with kexec
 730  */
 731 static void __init reserve_crashkernel(void)
 732 {
 733 #ifdef CONFIG_CRASH_DUMP
 734         unsigned long long crash_base, crash_size;
 735         char *msg = NULL;
 736         int rc;
 737
 738         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 739                                &crash_base);
 740         if (rc || crash_size == 0)
 741                 return;
 742         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 743         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 744         if (register_memory_notifier(&kdump_mem_nb))
 745                 return;
 746         if (!crash_base)
 747                 crash_base = find_crash_base(crash_size, &msg);
 748         if (!crash_base) {
 749                 pr_info("crashkernel reservation failed: %s\n", msg);
 750                 unregister_memory_notifier(&kdump_mem_nb);
 751                 return;
 752         }
 753         if (verify_crash_base(crash_base, crash_size, &msg)) {
 754                 pr_info("crashkernel reservation failed: %s\n", msg);
 755                 unregister_memory_notifier(&kdump_mem_nb);
 756                 return;
 757         }
 758         if (!OLDMEM_BASE && MACHINE_IS_VM)
 759                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 760         crashk_res.start = crash_base;
 761         crashk_res.end = crash_base + crash_size - 1;
 762         insert_resource(&iomem_resource, &crashk_res);
 763         reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
 764         pr_info("Reserving %lluMB of memory at %lluMB "
 765                 "for crashkernel (System RAM: %luMB)\n",
 766                 crash_size >> 20, crash_base >> 20, memory_end >> 20);
 767         os_info_crashkernel_add(crash_base, crash_size);
 768 #endif
 769 }
 770
 771 static void __init setup_memory(void)
 772 {
 773         unsigned long bootmap_size;
 774         unsigned long start_pfn, end_pfn;
 775         int i;
 776
 777         /*
 778          * partially used pages are not usable - thus
 779          * we are rounding upwards:
 780          */
 781         start_pfn = PFN_UP(__pa(&_end));
 782         end_pfn = max_pfn = PFN_DOWN(memory_end);
 783
 784 #ifdef CONFIG_BLK_DEV_INITRD
 785         /*
 786          * Move the initrd in case the bitmap of the bootmem allocater
 787          * would overwrite it.
 788          */
 789
 790         if (INITRD_START && INITRD_SIZE) {
 791                 unsigned long bmap_size;
 792                 unsigned long start;
 793
 794                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
 795                 bmap_size = PFN_PHYS(bmap_size);
 796
 797                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
 798                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
 799
 800 #ifdef CONFIG_CRASH_DUMP
 801                         if (OLDMEM_BASE) {
 802                                 /* Move initrd behind kdump oldmem */
 803                                 if (start + INITRD_SIZE > OLDMEM_BASE &&
 804                                     start < OLDMEM_BASE + OLDMEM_SIZE)
 805                                         start = OLDMEM_BASE + OLDMEM_SIZE;
 806                         }
 807 #endif
 808                         if (start + INITRD_SIZE > memory_end) {
 809                                 pr_err("initrd extends beyond end of "
 810                                        "memory (0x%08lx > 0x%08lx) "
 811                                        "disabling initrd\n",
 812                                        start + INITRD_SIZE, memory_end);
 813                                 INITRD_START = INITRD_SIZE = 0;
 814                         } else {
 815                                 pr_info("Moving initrd (0x%08lx -> "
 816                                         "0x%08lx, size: %ld)\n",
 817                                         INITRD_START, start, INITRD_SIZE);
 818                                 memmove((void *) start, (void *) INITRD_START,
 819                                         INITRD_SIZE);
 820                                 INITRD_START = start;
 821                         }
 822                 }
 823         }
 824 #endif
 825
 826         /*
 827          * Initialize the boot-time allocator
 828          */
 829         bootmap_size = init_bootmem(start_pfn, end_pfn);
 830
 831         /*
 832          * Register RAM areas with the bootmem allocator.
 833          */
 834
 835         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
 836                 unsigned long start_chunk, end_chunk, pfn;
 837
 838                 if (memory_chunk[i].type != CHUNK_READ_WRITE &&
 839                     memory_chunk[i].type != CHUNK_CRASHK)
 840                         continue;
 841                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
 842                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
 843                 end_chunk = min(end_chunk, end_pfn);
 844                 if (start_chunk >= end_chunk)
 845                         continue;
 846                 memblock_add_node(PFN_PHYS(start_chunk),
 847                                   PFN_PHYS(end_chunk - start_chunk), 0);
 848                 pfn = max(start_chunk, start_pfn);
 849                 for (; pfn < end_chunk; pfn++)
 850                         page_set_storage_key(PFN_PHYS(pfn),
 851                                              PAGE_DEFAULT_KEY, 0);
 852         }
 853
 854         psw_set_key(PAGE_DEFAULT_KEY);
 855
 856         free_bootmem_with_active_regions(0, max_pfn);
 857
 858         /*
 859          * Reserve memory used for lowcore/command line/kernel image.
 860          */
 861         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
 862         reserve_bootmem((unsigned long)_stext,
 863                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
 864                         BOOTMEM_DEFAULT);
 865         /*
 866          * Reserve the bootmem bitmap itself as well. We do this in two
 867          * steps (first step was init_bootmem()) because this catches
 868          * the (very unlikely) case of us accidentally initializing the
 869          * bootmem allocator with an invalid RAM area.
 870          */
 871         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
 872                         BOOTMEM_DEFAULT);
 873
 874 #ifdef CONFIG_CRASH_DUMP
 875         if (crashk_res.start)
 876                 reserve_bootmem(crashk_res.start,
 877                                 crashk_res.end - crashk_res.start + 1,
 878                                 BOOTMEM_DEFAULT);
 879         if (is_kdump_kernel())
 880                 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
 881                                 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
 882 #endif
 883 #ifdef CONFIG_BLK_DEV_INITRD
 884         if (INITRD_START && INITRD_SIZE) {
 885                 if (INITRD_START + INITRD_SIZE <= memory_end) {
 886                         reserve_bootmem(INITRD_START, INITRD_SIZE,
 887                                         BOOTMEM_DEFAULT);
 888                         initrd_start = INITRD_START;
 889                         initrd_end = initrd_start + INITRD_SIZE;
 890                 } else {
 891                         pr_err("initrd extends beyond end of "
 892                                "memory (0x%08lx > 0x%08lx) "
 893                                "disabling initrd\n",
 894                                initrd_start + INITRD_SIZE, memory_end);
 895                         initrd_start = initrd_end = 0;
 896                 }
 897         }
 898 #endif
 899 }
 900
 901 /*
 902  * Setup hardware capabilities.
 903  */
 904 static void __init setup_hwcaps(void)
 905 {
 906         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 907         struct cpuid cpu_id;
 908         int i;
 909
 910         /*
 911          * The store facility list bits numbers as found in the principles
 912          * of operation are numbered with bit 1UL<<31 as number 0 to
 913          * bit 1UL<<0 as number 31.
 914          *   Bit 0: instructions named N3, "backported" to esa-mode
 915          *   Bit 2: z/Architecture mode is active
 916          *   Bit 7: the store-facility-list-extended facility is installed
 917          *   Bit 17: the message-security assist is installed
 918          *   Bit 19: the long-displacement facility is installed
 919          *   Bit 21: the extended-immediate facility is installed
 920          *   Bit 22: extended-translation facility 3 is installed
 921          *   Bit 30: extended-translation facility 3 enhancement facility
 922          * These get translated to:
 923          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 924          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 925          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 926          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 927          */
 928         for (i = 0; i < 6; i++)
 929                 if (test_facility(stfl_bits[i]))
 930                         elf_hwcap |= 1UL << i;
 931
 932         if (test_facility(22) && test_facility(30))
 933                 elf_hwcap |= HWCAP_S390_ETF3EH;
 934
 935         /*
 936          * Check for additional facilities with store-facility-list-extended.
 937          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 938          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 939          * as stored by stfl, bits 32-xxx contain additional facilities.
 940          * How many facility words are stored depends on the number of
 941          * doublewords passed to the instruction. The additional facilities
 942          * are:
 943          *   Bit 42: decimal floating point facility is installed
 944          *   Bit 44: perform floating point operation facility is installed
 945          * translated to:
 946          *   HWCAP_S390_DFP bit 6 (42 && 44).
 947          */
 948         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 949                 elf_hwcap |= HWCAP_S390_DFP;
 950
 951         /*
 952          * Huge page support HWCAP_S390_HPAGE is bit 7.
 953          */
 954         if (MACHINE_HAS_HPAGE)
 955                 elf_hwcap |= HWCAP_S390_HPAGE;
 956
 957 #if defined(CONFIG_64BIT)
 958         /*
 959          * 64-bit register support for 31-bit processes
 960          * HWCAP_S390_HIGH_GPRS is bit 9.
 961          */
 962         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 963
 964         /*
 965          * Transactional execution support HWCAP_S390_TE is bit 10.
 966          */
 967         if (test_facility(50) && test_facility(73))
 968                 elf_hwcap |= HWCAP_S390_TE;
 969 #endif
 970
 971         get_cpu_id(&cpu_id);
 972         switch (cpu_id.machine) {
 973         case 0x9672:
 974 #if !defined(CONFIG_64BIT)
 975         default:        /* Use "g5" as default for 31 bit kernels. */
 976 #endif
 977                 strcpy(elf_platform, "g5");
 978                 break;
 979         case 0x2064:
 980         case 0x2066:
 981 #if defined(CONFIG_64BIT)
 982         default:        /* Use "z900" as default for 64 bit kernels. */
 983 #endif
 984                 strcpy(elf_platform, "z900");
 985                 break;
 986         case 0x2084:
 987         case 0x2086:
 988                 strcpy(elf_platform, "z990");
 989                 break;
 990         case 0x2094:
 991         case 0x2096:
 992                 strcpy(elf_platform, "z9-109");
 993                 break;
 994         case 0x2097:
 995         case 0x2098:
 996                 strcpy(elf_platform, "z10");
 997                 break;
 998         case 0x2817:
 999         case 0x2818:
1000                 strcpy(elf_platform, "z196");
1001                 break;
1002         }
1003 }
1004
1005 /*
1006  * Setup function called from init/main.c just after the banner
1007  * was printed.
1008  */
1009
1010 void __init setup_arch(char **cmdline_p)
1011 {
1012         /*
1013          * print what head.S has found out about the machine
1014          */
1015 #ifndef CONFIG_64BIT
1016         if (MACHINE_IS_VM)
1017                 pr_info("Linux is running as a z/VM "
1018                         "guest operating system in 31-bit mode\n");
1019         else if (MACHINE_IS_LPAR)
1020                 pr_info("Linux is running natively in 31-bit mode\n");
1021         if (MACHINE_HAS_IEEE)
1022                 pr_info("The hardware system has IEEE compatible "
1023                         "floating point units\n");
1024         else
1025                 pr_info("The hardware system has no IEEE compatible "
1026                         "floating point units\n");
1027 #else /* CONFIG_64BIT */
1028         if (MACHINE_IS_VM)
1029                 pr_info("Linux is running as a z/VM "
1030                         "guest operating system in 64-bit mode\n");
1031         else if (MACHINE_IS_KVM)
1032                 pr_info("Linux is running under KVM in 64-bit mode\n");
1033         else if (MACHINE_IS_LPAR)
1034                 pr_info("Linux is running natively in 64-bit mode\n");
1035 #endif /* CONFIG_64BIT */
1036
1037         /* Have one command line that is parsed and saved in /proc/cmdline */
1038         /* boot_command_line has been already set up in early.c */
1039         *cmdline_p = boot_command_line;
1040
1041         ROOT_DEV = Root_RAM0;
1042
1043         init_mm.start_code = PAGE_OFFSET;
1044         init_mm.end_code = (unsigned long) &_etext;
1045         init_mm.end_data = (unsigned long) &_edata;
1046         init_mm.brk = (unsigned long) &_end;
1047
1048         if (MACHINE_HAS_MVCOS)
1049                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1050         else
1051                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1052
1053         parse_early_param();
1054
1055         os_info_init();
1056         setup_ipl();
1057         setup_memory_end();
1058         setup_addressing_mode();
1059         reserve_oldmem();
1060         reserve_crashkernel();
1061         setup_memory();
1062         setup_resources();
1063         setup_vmcoreinfo();
1064         setup_lowcore();
1065
1066         cpu_init();
1067         s390_init_cpu_topology();
1068
1069         /*
1070          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1071          */
1072         setup_hwcaps();
1073
1074         /*
1075          * Create kernel page tables and switch to virtual addressing.
1076          */
1077         paging_init();
1078
1079         /* Setup default console */
1080         conmode_default();
1081         set_preferred_console();
1082
1083         /* Setup zfcpdump support */
1084         setup_zfcpdump(console_devno);
1085 }