arch/s390/kernel/setup.c

   1 /*
   2  *  S390 version
   3  *    Copyright IBM Corp. 1999, 2012
   4  *    Author(s): Hartmut Penner (hp@de.ibm.com),
   5  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
   6  *
   7  *  Derived from "arch/i386/kernel/setup.c"
   8  *    Copyright (C) 1995, Linus Torvalds
   9  */
  10
  11 /*
  12  * This file handles the architecture-dependent parts of initialization
  13  */
  14
  15 #define KMSG_COMPONENT "setup"
  16 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  17
  18 #include <linux/errno.h>
  19 #include <linux/export.h>
  20 #include <linux/sched.h>
  21 #include <linux/kernel.h>
  22 #include <linux/memblock.h>
  23 #include <linux/mm.h>
  24 #include <linux/stddef.h>
  25 #include <linux/unistd.h>
  26 #include <linux/ptrace.h>
  27 #include <linux/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 #ifdef CONFIG_64BIT
 109 unsigned long MODULES_VADDR;
 110 unsigned long MODULES_END;
 111 #endif
 112
 113 /* An array with a pointer to the lowcore of every CPU. */
 114 struct _lowcore *lowcore_ptr[NR_CPUS];
 115 EXPORT_SYMBOL(lowcore_ptr);
 116
 117 /*
 118  * This is set up by the setup-routine at boot-time
 119  * for S390 need to find out, what we have to setup
 120  * using address 0x10400 ...
 121  */
 122
 123 #include <asm/setup.h>
 124
 125 /*
 126  * condev= and conmode= setup parameter.
 127  */
 128
 129 static int __init condev_setup(char *str)
 130 {
 131         int vdev;
 132
 133         vdev = simple_strtoul(str, &str, 0);
 134         if (vdev >= 0 && vdev < 65536) {
 135                 console_devno = vdev;
 136                 console_irq = -1;
 137         }
 138         return 1;
 139 }
 140
 141 __setup("condev=", condev_setup);
 142
 143 static void __init set_preferred_console(void)
 144 {
 145         if (MACHINE_IS_KVM) {
 146                 if (sclp_has_vt220())
 147                         add_preferred_console("ttyS", 1, NULL);
 148                 else if (sclp_has_linemode())
 149                         add_preferred_console("ttyS", 0, NULL);
 150                 else
 151                         add_preferred_console("hvc", 0, NULL);
 152         } else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 153                 add_preferred_console("ttyS", 0, NULL);
 154         else if (CONSOLE_IS_3270)
 155                 add_preferred_console("tty3270", 0, NULL);
 156 }
 157
 158 static int __init conmode_setup(char *str)
 159 {
 160 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 161         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 162                 SET_CONSOLE_SCLP;
 163 #endif
 164 #if defined(CONFIG_TN3215_CONSOLE)
 165         if (strncmp(str, "3215", 5) == 0)
 166                 SET_CONSOLE_3215;
 167 #endif
 168 #if defined(CONFIG_TN3270_CONSOLE)
 169         if (strncmp(str, "3270", 5) == 0)
 170                 SET_CONSOLE_3270;
 171 #endif
 172         set_preferred_console();
 173         return 1;
 174 }
 175
 176 __setup("conmode=", conmode_setup);
 177
 178 static void __init conmode_default(void)
 179 {
 180         char query_buffer[1024];
 181         char *ptr;
 182
 183         if (MACHINE_IS_VM) {
 184                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 185                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 186                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 187                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 188                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 189                 ptr = strstr(query_buffer, "CONMODE");
 190                 /*
 191                  * Set the conmode to 3215 so that the device recognition
 192                  * will set the cu_type of the console to 3215. If the
 193                  * conmode is 3270 and we don't set it back then both
 194                  * 3215 and the 3270 driver will try to access the console
 195                  * device (3215 as console and 3270 as normal tty).
 196                  */
 197                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 198                 if (ptr == NULL) {
 199 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 200                         SET_CONSOLE_SCLP;
 201 #endif
 202                         return;
 203                 }
 204                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 205 #if defined(CONFIG_TN3270_CONSOLE)
 206                         SET_CONSOLE_3270;
 207 #elif defined(CONFIG_TN3215_CONSOLE)
 208                         SET_CONSOLE_3215;
 209 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 210                         SET_CONSOLE_SCLP;
 211 #endif
 212                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 213 #if defined(CONFIG_TN3215_CONSOLE)
 214                         SET_CONSOLE_3215;
 215 #elif defined(CONFIG_TN3270_CONSOLE)
 216                         SET_CONSOLE_3270;
 217 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 218                         SET_CONSOLE_SCLP;
 219 #endif
 220                 }
 221         } else {
 222 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 223                 SET_CONSOLE_SCLP;
 224 #endif
 225         }
 226 }
 227
 228 #ifdef CONFIG_ZFCPDUMP
 229 static void __init setup_zfcpdump(unsigned int console_devno)
 230 {
 231         static char str[41];
 232
 233         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 234                 return;
 235         if (OLDMEM_BASE)
 236                 return;
 237         if (console_devno != -1)
 238                 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
 239                         ipl_info.data.fcp.dev_id.devno, console_devno);
 240         else
 241                 sprintf(str, " cio_ignore=all,!0.0.%04x",
 242                         ipl_info.data.fcp.dev_id.devno);
 243         strcat(boot_command_line, str);
 244         console_loglevel = 2;
 245 }
 246 #else
 247 static inline void setup_zfcpdump(unsigned int console_devno) {}
 248 #endif /* CONFIG_ZFCPDUMP */
 249
 250  /*
 251  * Reboot, halt and power_off stubs. They just call _machine_restart,
 252  * _machine_halt or _machine_power_off.
 253  */
 254
 255 void machine_restart(char *command)
 256 {
 257         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 258                 /*
 259                  * Only unblank the console if we are called in enabled
 260                  * context or a bust_spinlocks cleared the way for us.
 261                  */
 262                 console_unblank();
 263         _machine_restart(command);
 264 }
 265
 266 void machine_halt(void)
 267 {
 268         if (!in_interrupt() || oops_in_progress)
 269                 /*
 270                  * Only unblank the console if we are called in enabled
 271                  * context or a bust_spinlocks cleared the way for us.
 272                  */
 273                 console_unblank();
 274         _machine_halt();
 275 }
 276
 277 void machine_power_off(void)
 278 {
 279         if (!in_interrupt() || oops_in_progress)
 280                 /*
 281                  * Only unblank the console if we are called in enabled
 282                  * context or a bust_spinlocks cleared the way for us.
 283                  */
 284                 console_unblank();
 285         _machine_power_off();
 286 }
 287
 288 /*
 289  * Dummy power off function.
 290  */
 291 void (*pm_power_off)(void) = machine_power_off;
 292 EXPORT_SYMBOL_GPL(pm_power_off);
 293
 294 static int __init early_parse_mem(char *p)
 295 {
 296         memory_end = memparse(p, &p);
 297         memory_end_set = 1;
 298         return 0;
 299 }
 300 early_param("mem", early_parse_mem);
 301
 302 static int __init parse_vmalloc(char *arg)
 303 {
 304         if (!arg)
 305                 return -EINVAL;
 306         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
 307         return 0;
 308 }
 309 early_param("vmalloc", parse_vmalloc);
 310
 311 unsigned int s390_user_mode = PRIMARY_SPACE_MODE;
 312 EXPORT_SYMBOL_GPL(s390_user_mode);
 313
 314 static void __init set_user_mode_primary(void)
 315 {
 316         psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
 317         psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
 318 #ifdef CONFIG_COMPAT
 319         psw32_user_bits =
 320                 (psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
 321 #endif
 322         uaccess = MACHINE_HAS_MVCOS ? uaccess_mvcos_switch : uaccess_pt;
 323 }
 324
 325 static int __init early_parse_user_mode(char *p)
 326 {
 327         if (p && strcmp(p, "primary") == 0)
 328                 s390_user_mode = PRIMARY_SPACE_MODE;
 329         else if (!p || strcmp(p, "home") == 0)
 330                 s390_user_mode = HOME_SPACE_MODE;
 331         else
 332                 return 1;
 333         return 0;
 334 }
 335 early_param("user_mode", early_parse_user_mode);
 336
 337 static void __init setup_addressing_mode(void)
 338 {
 339         if (s390_user_mode != PRIMARY_SPACE_MODE)
 340                 return;
 341         set_user_mode_primary();
 342         if (MACHINE_HAS_MVCOS)
 343                 pr_info("Address spaces switched, mvcos available\n");
 344         else
 345                 pr_info("Address spaces switched, mvcos not available\n");
 346 }
 347
 348 void *restart_stack __attribute__((__section__(".data")));
 349
 350 static void __init setup_lowcore(void)
 351 {
 352         struct _lowcore *lc;
 353
 354         /*
 355          * Setup lowcore for boot cpu
 356          */
 357         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
 358         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 359         lc->restart_psw.mask = psw_kernel_bits;
 360         lc->restart_psw.addr =
 361                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
 362         lc->external_new_psw.mask = psw_kernel_bits |
 363                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 364         lc->external_new_psw.addr =
 365                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
 366         lc->svc_new_psw.mask = psw_kernel_bits |
 367                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 368         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
 369         lc->program_new_psw.mask = psw_kernel_bits |
 370                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 371         lc->program_new_psw.addr =
 372                 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
 373         lc->mcck_new_psw.mask = psw_kernel_bits;
 374         lc->mcck_new_psw.addr =
 375                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
 376         lc->io_new_psw.mask = psw_kernel_bits |
 377                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 378         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 379         lc->clock_comparator = -1ULL;
 380         lc->kernel_stack = ((unsigned long) &init_thread_union)
 381                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 382         lc->async_stack = (unsigned long)
 383                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
 384                 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 385         lc->panic_stack = (unsigned long)
 386                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
 387                 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 388         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 389         lc->thread_info = (unsigned long) &init_thread_union;
 390         lc->machine_flags = S390_lowcore.machine_flags;
 391         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 392         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 393                MAX_FACILITY_BIT/8);
 394 #ifndef CONFIG_64BIT
 395         if (MACHINE_HAS_IEEE) {
 396                 lc->extended_save_area_addr = (__u32)
 397                         __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
 398                 /* enable extended save area */
 399                 __ctl_set_bit(14, 29);
 400         }
 401 #else
 402         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
 403 #endif
 404         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 405         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 406         lc->exit_timer = S390_lowcore.exit_timer;
 407         lc->user_timer = S390_lowcore.user_timer;
 408         lc->system_timer = S390_lowcore.system_timer;
 409         lc->steal_timer = S390_lowcore.steal_timer;
 410         lc->last_update_timer = S390_lowcore.last_update_timer;
 411         lc->last_update_clock = S390_lowcore.last_update_clock;
 412         lc->ftrace_func = S390_lowcore.ftrace_func;
 413
 414         restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
 415         restart_stack += ASYNC_SIZE;
 416
 417         /*
 418          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 419          * restart data to the absolute zero lowcore. This is necesary if
 420          * PSW restart is done on an offline CPU that has lowcore zero.
 421          */
 422         lc->restart_stack = (unsigned long) restart_stack;
 423         lc->restart_fn = (unsigned long) do_restart;
 424         lc->restart_data = 0;
 425         lc->restart_source = -1UL;
 426
 427         /* Setup absolute zero lowcore */
 428         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 429         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 430         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 431         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 432         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 433
 434         set_prefix((u32)(unsigned long) lc);
 435         lowcore_ptr[0] = lc;
 436 }
 437
 438 static struct resource code_resource = {
 439         .name  = "Kernel code",
 440         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 441 };
 442
 443 static struct resource data_resource = {
 444         .name = "Kernel data",
 445         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 446 };
 447
 448 static struct resource bss_resource = {
 449         .name = "Kernel bss",
 450         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 451 };
 452
 453 static struct resource __initdata *standard_resources[] = {
 454         &code_resource,
 455         &data_resource,
 456         &bss_resource,
 457 };
 458
 459 static void __init setup_resources(void)
 460 {
 461         struct resource *res, *std_res, *sub_res;
 462         int i, j;
 463
 464         code_resource.start = (unsigned long) &_text;
 465         code_resource.end = (unsigned long) &_etext - 1;
 466         data_resource.start = (unsigned long) &_etext;
 467         data_resource.end = (unsigned long) &_edata - 1;
 468         bss_resource.start = (unsigned long) &__bss_start;
 469         bss_resource.end = (unsigned long) &__bss_stop - 1;
 470
 471         for (i = 0; i < MEMORY_CHUNKS; i++) {
 472                 if (!memory_chunk[i].size)
 473                         continue;
 474                 if (memory_chunk[i].type == CHUNK_OLDMEM ||
 475                     memory_chunk[i].type == CHUNK_CRASHK)
 476                         continue;
 477                 res = alloc_bootmem_low(sizeof(*res));
 478                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
 479                 switch (memory_chunk[i].type) {
 480                 case CHUNK_READ_WRITE:
 481                 case CHUNK_CRASHK:
 482                         res->name = "System RAM";
 483                         break;
 484                 case CHUNK_READ_ONLY:
 485                         res->name = "System ROM";
 486                         res->flags |= IORESOURCE_READONLY;
 487                         break;
 488                 default:
 489                         res->name = "reserved";
 490                 }
 491                 res->start = memory_chunk[i].addr;
 492                 res->end = res->start + memory_chunk[i].size - 1;
 493                 request_resource(&iomem_resource, res);
 494
 495                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 496                         std_res = standard_resources[j];
 497                         if (std_res->start < res->start ||
 498                             std_res->start > res->end)
 499                                 continue;
 500                         if (std_res->end > res->end) {
 501                                 sub_res = alloc_bootmem_low(sizeof(*sub_res));
 502                                 *sub_res = *std_res;
 503                                 sub_res->end = res->end;
 504                                 std_res->start = res->end + 1;
 505                                 request_resource(res, sub_res);
 506                         } else {
 507                                 request_resource(res, std_res);
 508                         }
 509                 }
 510         }
 511 }
 512
 513 unsigned long real_memory_size;
 514 EXPORT_SYMBOL_GPL(real_memory_size);
 515
 516 static void __init setup_memory_end(void)
 517 {
 518         unsigned long vmax, vmalloc_size, tmp;
 519         int i;
 520
 521
 522 #ifdef CONFIG_ZFCPDUMP
 523         if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) {
 524                 memory_end = ZFCPDUMP_HSA_SIZE;
 525                 memory_end_set = 1;
 526         }
 527 #endif
 528         real_memory_size = 0;
 529         memory_end &= PAGE_MASK;
 530
 531         /*
 532          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 533          * extra checks that HOLES_IN_ZONE would require.
 534          */
 535         for (i = 0; i < MEMORY_CHUNKS; i++) {
 536                 unsigned long start, end;
 537                 struct mem_chunk *chunk;
 538                 unsigned long align;
 539
 540                 chunk = &memory_chunk[i];
 541                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
 542                 start = (chunk->addr + align - 1) & ~(align - 1);
 543                 end = (chunk->addr + chunk->size) & ~(align - 1);
 544                 if (start >= end)
 545                         memset(chunk, 0, sizeof(*chunk));
 546                 else {
 547                         chunk->addr = start;
 548                         chunk->size = end - start;
 549                 }
 550                 real_memory_size = max(real_memory_size,
 551                                        chunk->addr + chunk->size);
 552         }
 553
 554         /* Choose kernel address space layout: 2, 3, or 4 levels. */
 555 #ifdef CONFIG_64BIT
 556         vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
 557         tmp = (memory_end ?: real_memory_size) / PAGE_SIZE;
 558         tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
 559         if (tmp <= (1UL << 42))
 560                 vmax = 1UL << 42;       /* 3-level kernel page table */
 561         else
 562                 vmax = 1UL << 53;       /* 4-level kernel page table */
 563         /* module area is at the end of the kernel address space. */
 564         MODULES_END = vmax;
 565         MODULES_VADDR = MODULES_END - MODULES_LEN;
 566         VMALLOC_END = MODULES_VADDR;
 567 #else
 568         vmalloc_size = VMALLOC_END ?: 96UL << 20;
 569         vmax = 1UL << 31;               /* 2-level kernel page table */
 570         /* vmalloc area is at the end of the kernel address space. */
 571         VMALLOC_END = vmax;
 572 #endif
 573         VMALLOC_START = vmax - vmalloc_size;
 574
 575         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 576         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 577         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 578         tmp = SECTION_ALIGN_UP(tmp);
 579         tmp = VMALLOC_START - tmp * sizeof(struct page);
 580         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 581         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 582         vmemmap = (struct page *) tmp;
 583
 584         /* Take care that memory_end is set and <= vmemmap */
 585         memory_end = min(memory_end ?: real_memory_size, tmp);
 586
 587         /* Fixup memory chunk array to fit into 0..memory_end */
 588         for (i = 0; i < MEMORY_CHUNKS; i++) {
 589                 struct mem_chunk *chunk = &memory_chunk[i];
 590
 591                 if (chunk->addr >= memory_end) {
 592                         memset(chunk, 0, sizeof(*chunk));
 593                         continue;
 594                 }
 595                 if (chunk->addr + chunk->size > memory_end)
 596                         chunk->size = memory_end - chunk->addr;
 597         }
 598 }
 599
 600 static void __init setup_vmcoreinfo(void)
 601 {
 602         mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
 603 }
 604
 605 #ifdef CONFIG_CRASH_DUMP
 606
 607 /*
 608  * Find suitable location for crashkernel memory
 609  */
 610 static unsigned long __init find_crash_base(unsigned long crash_size,
 611                                             char **msg)
 612 {
 613         unsigned long crash_base;
 614         struct mem_chunk *chunk;
 615         int i;
 616
 617         if (memory_chunk[0].size < crash_size) {
 618                 *msg = "first memory chunk must be at least crashkernel size";
 619                 return 0;
 620         }
 621         if (OLDMEM_BASE && crash_size == OLDMEM_SIZE)
 622                 return OLDMEM_BASE;
 623
 624         for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
 625                 chunk = &memory_chunk[i];
 626                 if (chunk->size == 0)
 627                         continue;
 628                 if (chunk->type != CHUNK_READ_WRITE)
 629                         continue;
 630                 if (chunk->size < crash_size)
 631                         continue;
 632                 crash_base = (chunk->addr + chunk->size) - crash_size;
 633                 if (crash_base < crash_size)
 634                         continue;
 635                 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
 636                         continue;
 637                 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
 638                         continue;
 639                 return crash_base;
 640         }
 641         *msg = "no suitable area found";
 642         return 0;
 643 }
 644
 645 /*
 646  * Check if crash_base and crash_size is valid
 647  */
 648 static int __init verify_crash_base(unsigned long crash_base,
 649                                     unsigned long crash_size,
 650                                     char **msg)
 651 {
 652         struct mem_chunk *chunk;
 653         int i;
 654
 655         /*
 656          * Because we do the swap to zero, we must have at least 'crash_size'
 657          * bytes free space before crash_base
 658          */
 659         if (crash_size > crash_base) {
 660                 *msg = "crashkernel offset must be greater than size";
 661                 return -EINVAL;
 662         }
 663
 664         /* First memory chunk must be at least crash_size */
 665         if (memory_chunk[0].size < crash_size) {
 666                 *msg = "first memory chunk must be at least crashkernel size";
 667                 return -EINVAL;
 668         }
 669         /* Check if we fit into the respective memory chunk */
 670         for (i = 0; i < MEMORY_CHUNKS; i++) {
 671                 chunk = &memory_chunk[i];
 672                 if (chunk->size == 0)
 673                         continue;
 674                 if (crash_base < chunk->addr)
 675                         continue;
 676                 if (crash_base >= chunk->addr + chunk->size)
 677                         continue;
 678                 /* we have found the memory chunk */
 679                 if (crash_base + crash_size > chunk->addr + chunk->size) {
 680                         *msg = "selected memory chunk is too small for "
 681                                 "crashkernel memory";
 682                         return -EINVAL;
 683                 }
 684                 return 0;
 685         }
 686         *msg = "invalid memory range specified";
 687         return -EINVAL;
 688 }
 689
 690 /*
 691  * Reserve kdump memory by creating a memory hole in the mem_chunk array
 692  */
 693 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
 694                                          int type)
 695 {
 696         create_mem_hole(memory_chunk, addr, size, type);
 697 }
 698
 699 /*
 700  * When kdump is enabled, we have to ensure that no memory from
 701  * the area [0 - crashkernel memory size] and
 702  * [crashk_res.start - crashk_res.end] is set offline.
 703  */
 704 static int kdump_mem_notifier(struct notifier_block *nb,
 705                               unsigned long action, void *data)
 706 {
 707         struct memory_notify *arg = data;
 708
 709         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 710                 return NOTIFY_BAD;
 711         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 712                 return NOTIFY_OK;
 713         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 714                 return NOTIFY_OK;
 715         return NOTIFY_BAD;
 716 }
 717
 718 static struct notifier_block kdump_mem_nb = {
 719         .notifier_call = kdump_mem_notifier,
 720 };
 721
 722 #endif
 723
 724 /*
 725  * Make sure that oldmem, where the dump is stored, is protected
 726  */
 727 static void reserve_oldmem(void)
 728 {
 729 #ifdef CONFIG_CRASH_DUMP
 730         if (!OLDMEM_BASE)
 731                 return;
 732
 733         reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
 734         reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
 735                               CHUNK_OLDMEM);
 736         if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
 737                 saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
 738         else
 739                 saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
 740 #endif
 741 }
 742
 743 /*
 744  * Reserve memory for kdump kernel to be loaded with kexec
 745  */
 746 static void __init reserve_crashkernel(void)
 747 {
 748 #ifdef CONFIG_CRASH_DUMP
 749         unsigned long long crash_base, crash_size;
 750         char *msg = NULL;
 751         int rc;
 752
 753         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 754                                &crash_base);
 755         if (rc || crash_size == 0)
 756                 return;
 757         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 758         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 759         if (register_memory_notifier(&kdump_mem_nb))
 760                 return;
 761         if (!crash_base)
 762                 crash_base = find_crash_base(crash_size, &msg);
 763         if (!crash_base) {
 764                 pr_info("crashkernel reservation failed: %s\n", msg);
 765                 unregister_memory_notifier(&kdump_mem_nb);
 766                 return;
 767         }
 768         if (verify_crash_base(crash_base, crash_size, &msg)) {
 769                 pr_info("crashkernel reservation failed: %s\n", msg);
 770                 unregister_memory_notifier(&kdump_mem_nb);
 771                 return;
 772         }
 773         if (!OLDMEM_BASE && MACHINE_IS_VM)
 774                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 775         crashk_res.start = crash_base;
 776         crashk_res.end = crash_base + crash_size - 1;
 777         insert_resource(&iomem_resource, &crashk_res);
 778         reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
 779         pr_info("Reserving %lluMB of memory at %lluMB "
 780                 "for crashkernel (System RAM: %luMB)\n",
 781                 crash_size >> 20, crash_base >> 20, memory_end >> 20);
 782         os_info_crashkernel_add(crash_base, crash_size);
 783 #endif
 784 }
 785
 786 static void __init setup_memory(void)
 787 {
 788         unsigned long bootmap_size;
 789         unsigned long start_pfn, end_pfn;
 790         int i;
 791
 792         /*
 793          * partially used pages are not usable - thus
 794          * we are rounding upwards:
 795          */
 796         start_pfn = PFN_UP(__pa(&_end));
 797         end_pfn = max_pfn = PFN_DOWN(memory_end);
 798
 799 #ifdef CONFIG_BLK_DEV_INITRD
 800         /*
 801          * Move the initrd in case the bitmap of the bootmem allocater
 802          * would overwrite it.
 803          */
 804
 805         if (INITRD_START && INITRD_SIZE) {
 806                 unsigned long bmap_size;
 807                 unsigned long start;
 808
 809                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
 810                 bmap_size = PFN_PHYS(bmap_size);
 811
 812                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
 813                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
 814
 815 #ifdef CONFIG_CRASH_DUMP
 816                         if (OLDMEM_BASE) {
 817                                 /* Move initrd behind kdump oldmem */
 818                                 if (start + INITRD_SIZE > OLDMEM_BASE &&
 819                                     start < OLDMEM_BASE + OLDMEM_SIZE)
 820                                         start = OLDMEM_BASE + OLDMEM_SIZE;
 821                         }
 822 #endif
 823                         if (start + INITRD_SIZE > memory_end) {
 824                                 pr_err("initrd extends beyond end of "
 825                                        "memory (0x%08lx > 0x%08lx) "
 826                                        "disabling initrd\n",
 827                                        start + INITRD_SIZE, memory_end);
 828                                 INITRD_START = INITRD_SIZE = 0;
 829                         } else {
 830                                 pr_info("Moving initrd (0x%08lx -> "
 831                                         "0x%08lx, size: %ld)\n",
 832                                         INITRD_START, start, INITRD_SIZE);
 833                                 memmove((void *) start, (void *) INITRD_START,
 834                                         INITRD_SIZE);
 835                                 INITRD_START = start;
 836                         }
 837                 }
 838         }
 839 #endif
 840
 841         /*
 842          * Initialize the boot-time allocator
 843          */
 844         bootmap_size = init_bootmem(start_pfn, end_pfn);
 845
 846         /*
 847          * Register RAM areas with the bootmem allocator.
 848          */
 849
 850         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
 851                 unsigned long start_chunk, end_chunk, pfn;
 852
 853                 if (memory_chunk[i].type != CHUNK_READ_WRITE &&
 854                     memory_chunk[i].type != CHUNK_CRASHK)
 855                         continue;
 856                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
 857                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
 858                 end_chunk = min(end_chunk, end_pfn);
 859                 if (start_chunk >= end_chunk)
 860                         continue;
 861                 memblock_add_node(PFN_PHYS(start_chunk),
 862                                   PFN_PHYS(end_chunk - start_chunk), 0);
 863                 pfn = max(start_chunk, start_pfn);
 864                 storage_key_init_range(PFN_PHYS(pfn), PFN_PHYS(end_chunk));
 865         }
 866
 867         psw_set_key(PAGE_DEFAULT_KEY);
 868
 869         free_bootmem_with_active_regions(0, max_pfn);
 870
 871         /*
 872          * Reserve memory used for lowcore/command line/kernel image.
 873          */
 874         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
 875         reserve_bootmem((unsigned long)_stext,
 876                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
 877                         BOOTMEM_DEFAULT);
 878         /*
 879          * Reserve the bootmem bitmap itself as well. We do this in two
 880          * steps (first step was init_bootmem()) because this catches
 881          * the (very unlikely) case of us accidentally initializing the
 882          * bootmem allocator with an invalid RAM area.
 883          */
 884         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
 885                         BOOTMEM_DEFAULT);
 886
 887 #ifdef CONFIG_CRASH_DUMP
 888         if (crashk_res.start)
 889                 reserve_bootmem(crashk_res.start,
 890                                 crashk_res.end - crashk_res.start + 1,
 891                                 BOOTMEM_DEFAULT);
 892         if (is_kdump_kernel())
 893                 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
 894                                 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
 895 #endif
 896 #ifdef CONFIG_BLK_DEV_INITRD
 897         if (INITRD_START && INITRD_SIZE) {
 898                 if (INITRD_START + INITRD_SIZE <= memory_end) {
 899                         reserve_bootmem(INITRD_START, INITRD_SIZE,
 900                                         BOOTMEM_DEFAULT);
 901                         initrd_start = INITRD_START;
 902                         initrd_end = initrd_start + INITRD_SIZE;
 903                 } else {
 904                         pr_err("initrd extends beyond end of "
 905                                "memory (0x%08lx > 0x%08lx) "
 906                                "disabling initrd\n",
 907                                initrd_start + INITRD_SIZE, memory_end);
 908                         initrd_start = initrd_end = 0;
 909                 }
 910         }
 911 #endif
 912 }
 913
 914 /*
 915  * Setup hardware capabilities.
 916  */
 917 static void __init setup_hwcaps(void)
 918 {
 919         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 920         struct cpuid cpu_id;
 921         int i;
 922
 923         /*
 924          * The store facility list bits numbers as found in the principles
 925          * of operation are numbered with bit 1UL<<31 as number 0 to
 926          * bit 1UL<<0 as number 31.
 927          *   Bit 0: instructions named N3, "backported" to esa-mode
 928          *   Bit 2: z/Architecture mode is active
 929          *   Bit 7: the store-facility-list-extended facility is installed
 930          *   Bit 17: the message-security assist is installed
 931          *   Bit 19: the long-displacement facility is installed
 932          *   Bit 21: the extended-immediate facility is installed
 933          *   Bit 22: extended-translation facility 3 is installed
 934          *   Bit 30: extended-translation facility 3 enhancement facility
 935          * These get translated to:
 936          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 937          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 938          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 939          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 940          */
 941         for (i = 0; i < 6; i++)
 942                 if (test_facility(stfl_bits[i]))
 943                         elf_hwcap |= 1UL << i;
 944
 945         if (test_facility(22) && test_facility(30))
 946                 elf_hwcap |= HWCAP_S390_ETF3EH;
 947
 948         /*
 949          * Check for additional facilities with store-facility-list-extended.
 950          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 951          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 952          * as stored by stfl, bits 32-xxx contain additional facilities.
 953          * How many facility words are stored depends on the number of
 954          * doublewords passed to the instruction. The additional facilities
 955          * are:
 956          *   Bit 42: decimal floating point facility is installed
 957          *   Bit 44: perform floating point operation facility is installed
 958          * translated to:
 959          *   HWCAP_S390_DFP bit 6 (42 && 44).
 960          */
 961         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 962                 elf_hwcap |= HWCAP_S390_DFP;
 963
 964         /*
 965          * Huge page support HWCAP_S390_HPAGE is bit 7.
 966          */
 967         if (MACHINE_HAS_HPAGE)
 968                 elf_hwcap |= HWCAP_S390_HPAGE;
 969
 970 #if defined(CONFIG_64BIT)
 971         /*
 972          * 64-bit register support for 31-bit processes
 973          * HWCAP_S390_HIGH_GPRS is bit 9.
 974          */
 975         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 976
 977         /*
 978          * Transactional execution support HWCAP_S390_TE is bit 10.
 979          */
 980         if (test_facility(50) && test_facility(73))
 981                 elf_hwcap |= HWCAP_S390_TE;
 982 #endif
 983
 984         get_cpu_id(&cpu_id);
 985         switch (cpu_id.machine) {
 986         case 0x9672:
 987 #if !defined(CONFIG_64BIT)
 988         default:        /* Use "g5" as default for 31 bit kernels. */
 989 #endif
 990                 strcpy(elf_platform, "g5");
 991                 break;
 992         case 0x2064:
 993         case 0x2066:
 994 #if defined(CONFIG_64BIT)
 995         default:        /* Use "z900" as default for 64 bit kernels. */
 996 #endif
 997                 strcpy(elf_platform, "z900");
 998                 break;
 999         case 0x2084:
1000         case 0x2086:
1001                 strcpy(elf_platform, "z990");
1002                 break;
1003         case 0x2094:
1004         case 0x2096:
1005                 strcpy(elf_platform, "z9-109");
1006                 break;
1007         case 0x2097:
1008         case 0x2098:
1009                 strcpy(elf_platform, "z10");
1010                 break;
1011         case 0x2817:
1012         case 0x2818:
1013                 strcpy(elf_platform, "z196");
1014                 break;
1015         case 0x2827:
1016                 strcpy(elf_platform, "zEC12");
1017                 break;
1018         }
1019 }
1020
1021 /*
1022  * Setup function called from init/main.c just after the banner
1023  * was printed.
1024  */
1025
1026 void __init setup_arch(char **cmdline_p)
1027 {
1028         /*
1029          * print what head.S has found out about the machine
1030          */
1031 #ifndef CONFIG_64BIT
1032         if (MACHINE_IS_VM)
1033                 pr_info("Linux is running as a z/VM "
1034                         "guest operating system in 31-bit mode\n");
1035         else if (MACHINE_IS_LPAR)
1036                 pr_info("Linux is running natively in 31-bit mode\n");
1037         if (MACHINE_HAS_IEEE)
1038                 pr_info("The hardware system has IEEE compatible "
1039                         "floating point units\n");
1040         else
1041                 pr_info("The hardware system has no IEEE compatible "
1042                         "floating point units\n");
1043 #else /* CONFIG_64BIT */
1044         if (MACHINE_IS_VM)
1045                 pr_info("Linux is running as a z/VM "
1046                         "guest operating system in 64-bit mode\n");
1047         else if (MACHINE_IS_KVM)
1048                 pr_info("Linux is running under KVM in 64-bit mode\n");
1049         else if (MACHINE_IS_LPAR)
1050                 pr_info("Linux is running natively in 64-bit mode\n");
1051 #endif /* CONFIG_64BIT */
1052
1053         /* Have one command line that is parsed and saved in /proc/cmdline */
1054         /* boot_command_line has been already set up in early.c */
1055         *cmdline_p = boot_command_line;
1056
1057         ROOT_DEV = Root_RAM0;
1058
1059         init_mm.start_code = PAGE_OFFSET;
1060         init_mm.end_code = (unsigned long) &_etext;
1061         init_mm.end_data = (unsigned long) &_edata;
1062         init_mm.brk = (unsigned long) &_end;
1063
1064         if (MACHINE_HAS_MVCOS)
1065                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1066         else
1067                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1068
1069         parse_early_param();
1070
1071         os_info_init();
1072         setup_ipl();
1073         setup_memory_end();
1074         setup_addressing_mode();
1075         reserve_oldmem();
1076         reserve_crashkernel();
1077         setup_memory();
1078         setup_resources();
1079         setup_vmcoreinfo();
1080         setup_lowcore();
1081
1082         cpu_init();
1083         s390_init_cpu_topology();
1084
1085         /*
1086          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1087          */
1088         setup_hwcaps();
1089
1090         /*
1091          * Create kernel page tables and switch to virtual addressing.
1092          */
1093         paging_init();
1094
1095         /* Setup default console */
1096         conmode_default();
1097         set_preferred_console();
1098
1099         /* Setup zfcpdump support */
1100         setup_zfcpdump(console_devno);
1101 }