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 user_mode = HOME_SPACE_MODE; 306 EXPORT_SYMBOL_GPL(user_mode); 307 308 static int set_amode_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 316 317 if (MACHINE_HAS_MVCOS) { 318 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess)); 319 return 1; 320 } else { 321 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess)); 322 return 0; 323 } 324 } 325 326 /* 327 * Switch kernel/user addressing modes? 328 */ 329 static int __init early_parse_switch_amode(char *p) 330 { 331 user_mode = PRIMARY_SPACE_MODE; 332 return 0; 333 } 334 early_param("switch_amode", early_parse_switch_amode); 335 336 static int __init early_parse_user_mode(char *p) 337 { 338 if (p && strcmp(p, "primary") == 0) 339 user_mode = PRIMARY_SPACE_MODE; 340 else if (!p || strcmp(p, "home") == 0) 341 user_mode = HOME_SPACE_MODE; 342 else 343 return 1; 344 return 0; 345 } 346 early_param("user_mode", early_parse_user_mode); 347 348 static void setup_addressing_mode(void) 349 { 350 if (user_mode == PRIMARY_SPACE_MODE) { 351 if (set_amode_primary()) 352 pr_info("Address spaces switched, " 353 "mvcos available\n"); 354 else 355 pr_info("Address spaces switched, " 356 "mvcos not available\n"); 357 } 358 } 359 360 void *restart_stack __attribute__((__section__(".data"))); 361 362 static void __init setup_lowcore(void) 363 { 364 struct _lowcore *lc; 365 366 /* 367 * Setup lowcore for boot cpu 368 */ 369 BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096); 370 lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0); 371 lc->restart_psw.mask = psw_kernel_bits; 372 lc->restart_psw.addr = 373 PSW_ADDR_AMODE | (unsigned long) restart_int_handler; 374 lc->external_new_psw.mask = psw_kernel_bits | 375 PSW_MASK_DAT | PSW_MASK_MCHECK; 376 lc->external_new_psw.addr = 377 PSW_ADDR_AMODE | (unsigned long) ext_int_handler; 378 lc->svc_new_psw.mask = psw_kernel_bits | 379 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK; 380 lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call; 381 lc->program_new_psw.mask = psw_kernel_bits | 382 PSW_MASK_DAT | PSW_MASK_MCHECK; 383 lc->program_new_psw.addr = 384 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler; 385 lc->mcck_new_psw.mask = psw_kernel_bits; 386 lc->mcck_new_psw.addr = 387 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler; 388 lc->io_new_psw.mask = psw_kernel_bits | 389 PSW_MASK_DAT | PSW_MASK_MCHECK; 390 lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler; 391 lc->clock_comparator = -1ULL; 392 lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE; 393 lc->async_stack = (unsigned long) 394 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE; 395 lc->panic_stack = (unsigned long) 396 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE; 397 lc->current_task = (unsigned long) init_thread_union.thread_info.task; 398 lc->thread_info = (unsigned long) &init_thread_union; 399 lc->machine_flags = S390_lowcore.machine_flags; 400 lc->stfl_fac_list = S390_lowcore.stfl_fac_list; 401 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list, 402 MAX_FACILITY_BIT/8); 403 #ifndef CONFIG_64BIT 404 if (MACHINE_HAS_IEEE) { 405 lc->extended_save_area_addr = (__u32) 406 __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0); 407 /* enable extended save area */ 408 __ctl_set_bit(14, 29); 409 } 410 #else 411 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0]; 412 #endif 413 lc->sync_enter_timer = S390_lowcore.sync_enter_timer; 414 lc->async_enter_timer = S390_lowcore.async_enter_timer; 415 lc->exit_timer = S390_lowcore.exit_timer; 416 lc->user_timer = S390_lowcore.user_timer; 417 lc->system_timer = S390_lowcore.system_timer; 418 lc->steal_timer = S390_lowcore.steal_timer; 419 lc->last_update_timer = S390_lowcore.last_update_timer; 420 lc->last_update_clock = S390_lowcore.last_update_clock; 421 lc->ftrace_func = S390_lowcore.ftrace_func; 422 423 restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0); 424 restart_stack += ASYNC_SIZE; 425 426 /* 427 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 428 * restart data to the absolute zero lowcore. This is necesary if 429 * PSW restart is done on an offline CPU that has lowcore zero. 430 */ 431 lc->restart_stack = (unsigned long) restart_stack; 432 lc->restart_fn = (unsigned long) do_restart; 433 lc->restart_data = 0; 434 lc->restart_source = -1UL; 435 436 /* Setup absolute zero lowcore */ 437 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack); 438 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn); 439 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data); 440 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source); 441 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw); 442 443 set_prefix((u32)(unsigned long) lc); 444 lowcore_ptr[0] = lc; 445 } 446 447 static struct resource code_resource = { 448 .name = "Kernel code", 449 .flags = IORESOURCE_BUSY | IORESOURCE_MEM, 450 }; 451 452 static struct resource data_resource = { 453 .name = "Kernel data", 454 .flags = IORESOURCE_BUSY | IORESOURCE_MEM, 455 }; 456 457 static struct resource bss_resource = { 458 .name = "Kernel bss", 459 .flags = IORESOURCE_BUSY | IORESOURCE_MEM, 460 }; 461 462 static struct resource __initdata *standard_resources[] = { 463 &code_resource, 464 &data_resource, 465 &bss_resource, 466 }; 467 468 static void __init setup_resources(void) 469 { 470 struct resource *res, *std_res, *sub_res; 471 int i, j; 472 473 code_resource.start = (unsigned long) &_text; 474 code_resource.end = (unsigned long) &_etext - 1; 475 data_resource.start = (unsigned long) &_etext; 476 data_resource.end = (unsigned long) &_edata - 1; 477 bss_resource.start = (unsigned long) &__bss_start; 478 bss_resource.end = (unsigned long) &__bss_stop - 1; 479 480 for (i = 0; i < MEMORY_CHUNKS; i++) { 481 if (!memory_chunk[i].size) 482 continue; 483 if (memory_chunk[i].type == CHUNK_OLDMEM || 484 memory_chunk[i].type == CHUNK_CRASHK) 485 continue; 486 res = alloc_bootmem_low(sizeof(*res)); 487 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM; 488 switch (memory_chunk[i].type) { 489 case CHUNK_READ_WRITE: 490 case CHUNK_CRASHK: 491 res->name = "System RAM"; 492 break; 493 case CHUNK_READ_ONLY: 494 res->name = "System ROM"; 495 res->flags |= IORESOURCE_READONLY; 496 break; 497 default: 498 res->name = "reserved"; 499 } 500 res->start = memory_chunk[i].addr; 501 res->end = res->start + memory_chunk[i].size - 1; 502 request_resource(&iomem_resource, res); 503 504 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 505 std_res = standard_resources[j]; 506 if (std_res->start < res->start || 507 std_res->start > res->end) 508 continue; 509 if (std_res->end > res->end) { 510 sub_res = alloc_bootmem_low(sizeof(*sub_res)); 511 *sub_res = *std_res; 512 sub_res->end = res->end; 513 std_res->start = res->end + 1; 514 request_resource(res, sub_res); 515 } else { 516 request_resource(res, std_res); 517 } 518 } 519 } 520 } 521 522 unsigned long real_memory_size; 523 EXPORT_SYMBOL_GPL(real_memory_size); 524 525 static void __init setup_memory_end(void) 526 { 527 unsigned long vmax, vmalloc_size, tmp; 528 int i; 529 530 531 #ifdef CONFIG_ZFCPDUMP 532 if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) { 533 memory_end = ZFCPDUMP_HSA_SIZE; 534 memory_end_set = 1; 535 } 536 #endif 537 real_memory_size = 0; 538 memory_end &= PAGE_MASK; 539 540 /* 541 * Make sure all chunks are MAX_ORDER aligned so we don't need the 542 * extra checks that HOLES_IN_ZONE would require. 543 */ 544 for (i = 0; i < MEMORY_CHUNKS; i++) { 545 unsigned long start, end; 546 struct mem_chunk *chunk; 547 unsigned long align; 548 549 chunk = &memory_chunk[i]; 550 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1); 551 start = (chunk->addr + align - 1) & ~(align - 1); 552 end = (chunk->addr + chunk->size) & ~(align - 1); 553 if (start >= end) 554 memset(chunk, 0, sizeof(*chunk)); 555 else { 556 chunk->addr = start; 557 chunk->size = end - start; 558 } 559 real_memory_size = max(real_memory_size, 560 chunk->addr + chunk->size); 561 } 562 563 /* Choose kernel address space layout: 2, 3, or 4 levels. */ 564 #ifdef CONFIG_64BIT 565 vmalloc_size = VMALLOC_END ?: 128UL << 30; 566 tmp = (memory_end ?: real_memory_size) / PAGE_SIZE; 567 tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size; 568 if (tmp <= (1UL << 42)) 569 vmax = 1UL << 42; /* 3-level kernel page table */ 570 else 571 vmax = 1UL << 53; /* 4-level kernel page table */ 572 #else 573 vmalloc_size = VMALLOC_END ?: 96UL << 20; 574 vmax = 1UL << 31; /* 2-level kernel page table */ 575 #endif 576 /* vmalloc area is at the end of the kernel address space. */ 577 VMALLOC_END = vmax; 578 VMALLOC_START = vmax - vmalloc_size; 579 580 /* Split remaining virtual space between 1:1 mapping & vmemmap array */ 581 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page)); 582 tmp = VMALLOC_START - tmp * sizeof(struct page); 583 tmp &= ~((vmax >> 11) - 1); /* align to page table level */ 584 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS); 585 vmemmap = (struct page *) tmp; 586 587 /* Take care that memory_end is set and <= vmemmap */ 588 memory_end = min(memory_end ?: real_memory_size, tmp); 589 590 /* Fixup memory chunk array to fit into 0..memory_end */ 591 for (i = 0; i < MEMORY_CHUNKS; i++) { 592 struct mem_chunk *chunk = &memory_chunk[i]; 593 594 if (chunk->addr >= memory_end) { 595 memset(chunk, 0, sizeof(*chunk)); 596 continue; 597 } 598 if (chunk->addr + chunk->size > memory_end) 599 chunk->size = memory_end - chunk->addr; 600 } 601 } 602 603 static void __init setup_vmcoreinfo(void) 604 { 605 #ifdef CONFIG_KEXEC 606 mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note()); 607 #endif 608 } 609 610 #ifdef CONFIG_CRASH_DUMP 611 612 /* 613 * Find suitable location for crashkernel memory 614 */ 615 static unsigned long __init find_crash_base(unsigned long crash_size, 616 char **msg) 617 { 618 unsigned long crash_base; 619 struct mem_chunk *chunk; 620 int i; 621 622 if (memory_chunk[0].size < crash_size) { 623 *msg = "first memory chunk must be at least crashkernel size"; 624 return 0; 625 } 626 if (OLDMEM_BASE && crash_size == OLDMEM_SIZE) 627 return OLDMEM_BASE; 628 629 for (i = MEMORY_CHUNKS - 1; i >= 0; i--) { 630 chunk = &memory_chunk[i]; 631 if (chunk->size == 0) 632 continue; 633 if (chunk->type != CHUNK_READ_WRITE) 634 continue; 635 if (chunk->size < crash_size) 636 continue; 637 crash_base = (chunk->addr + chunk->size) - crash_size; 638 if (crash_base < crash_size) 639 continue; 640 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX) 641 continue; 642 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE) 643 continue; 644 return crash_base; 645 } 646 *msg = "no suitable area found"; 647 return 0; 648 } 649 650 /* 651 * Check if crash_base and crash_size is valid 652 */ 653 static int __init verify_crash_base(unsigned long crash_base, 654 unsigned long crash_size, 655 char **msg) 656 { 657 struct mem_chunk *chunk; 658 int i; 659 660 /* 661 * Because we do the swap to zero, we must have at least 'crash_size' 662 * bytes free space before crash_base 663 */ 664 if (crash_size > crash_base) { 665 *msg = "crashkernel offset must be greater than size"; 666 return -EINVAL; 667 } 668 669 /* First memory chunk must be at least crash_size */ 670 if (memory_chunk[0].size < crash_size) { 671 *msg = "first memory chunk must be at least crashkernel size"; 672 return -EINVAL; 673 } 674 /* Check if we fit into the respective memory chunk */ 675 for (i = 0; i < MEMORY_CHUNKS; i++) { 676 chunk = &memory_chunk[i]; 677 if (chunk->size == 0) 678 continue; 679 if (crash_base < chunk->addr) 680 continue; 681 if (crash_base >= chunk->addr + chunk->size) 682 continue; 683 /* we have found the memory chunk */ 684 if (crash_base + crash_size > chunk->addr + chunk->size) { 685 *msg = "selected memory chunk is too small for " 686 "crashkernel memory"; 687 return -EINVAL; 688 } 689 return 0; 690 } 691 *msg = "invalid memory range specified"; 692 return -EINVAL; 693 } 694 695 /* 696 * Reserve kdump memory by creating a memory hole in the mem_chunk array 697 */ 698 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size, 699 int type) 700 { 701 create_mem_hole(memory_chunk, addr, size, type); 702 } 703 704 /* 705 * When kdump is enabled, we have to ensure that no memory from 706 * the area [0 - crashkernel memory size] and 707 * [crashk_res.start - crashk_res.end] is set offline. 708 */ 709 static int kdump_mem_notifier(struct notifier_block *nb, 710 unsigned long action, void *data) 711 { 712 struct memory_notify *arg = data; 713 714 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 715 return NOTIFY_BAD; 716 if (arg->start_pfn > PFN_DOWN(crashk_res.end)) 717 return NOTIFY_OK; 718 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start)) 719 return NOTIFY_OK; 720 return NOTIFY_BAD; 721 } 722 723 static struct notifier_block kdump_mem_nb = { 724 .notifier_call = kdump_mem_notifier, 725 }; 726 727 #endif 728 729 /* 730 * Make sure that oldmem, where the dump is stored, is protected 731 */ 732 static void reserve_oldmem(void) 733 { 734 #ifdef CONFIG_CRASH_DUMP 735 if (!OLDMEM_BASE) 736 return; 737 738 reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM); 739 reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE, 740 CHUNK_OLDMEM); 741 if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size) 742 saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1; 743 else 744 saved_max_pfn = PFN_DOWN(real_memory_size) - 1; 745 #endif 746 } 747 748 /* 749 * Reserve memory for kdump kernel to be loaded with kexec 750 */ 751 static void __init reserve_crashkernel(void) 752 { 753 #ifdef CONFIG_CRASH_DUMP 754 unsigned long long crash_base, crash_size; 755 char *msg = NULL; 756 int rc; 757 758 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size, 759 &crash_base); 760 if (rc || crash_size == 0) 761 return; 762 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 763 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 764 if (register_memory_notifier(&kdump_mem_nb)) 765 return; 766 if (!crash_base) 767 crash_base = find_crash_base(crash_size, &msg); 768 if (!crash_base) { 769 pr_info("crashkernel reservation failed: %s\n", msg); 770 unregister_memory_notifier(&kdump_mem_nb); 771 return; 772 } 773 if (verify_crash_base(crash_base, crash_size, &msg)) { 774 pr_info("crashkernel reservation failed: %s\n", msg); 775 unregister_memory_notifier(&kdump_mem_nb); 776 return; 777 } 778 if (!OLDMEM_BASE && MACHINE_IS_VM) 779 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 780 crashk_res.start = crash_base; 781 crashk_res.end = crash_base + crash_size - 1; 782 insert_resource(&iomem_resource, &crashk_res); 783 reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK); 784 pr_info("Reserving %lluMB of memory at %lluMB " 785 "for crashkernel (System RAM: %luMB)\n", 786 crash_size >> 20, crash_base >> 20, memory_end >> 20); 787 os_info_crashkernel_add(crash_base, crash_size); 788 #endif 789 } 790 791 static void __init setup_memory(void) 792 { 793 unsigned long bootmap_size; 794 unsigned long start_pfn, end_pfn; 795 int i; 796 797 /* 798 * partially used pages are not usable - thus 799 * we are rounding upwards: 800 */ 801 start_pfn = PFN_UP(__pa(&_end)); 802 end_pfn = max_pfn = PFN_DOWN(memory_end); 803 804 #ifdef CONFIG_BLK_DEV_INITRD 805 /* 806 * Move the initrd in case the bitmap of the bootmem allocater 807 * would overwrite it. 808 */ 809 810 if (INITRD_START && INITRD_SIZE) { 811 unsigned long bmap_size; 812 unsigned long start; 813 814 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1); 815 bmap_size = PFN_PHYS(bmap_size); 816 817 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) { 818 start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE; 819 820 #ifdef CONFIG_CRASH_DUMP 821 if (OLDMEM_BASE) { 822 /* Move initrd behind kdump oldmem */ 823 if (start + INITRD_SIZE > OLDMEM_BASE && 824 start < OLDMEM_BASE + OLDMEM_SIZE) 825 start = OLDMEM_BASE + OLDMEM_SIZE; 826 } 827 #endif 828 if (start + INITRD_SIZE > memory_end) { 829 pr_err("initrd extends beyond end of " 830 "memory (0x%08lx > 0x%08lx) " 831 "disabling initrd\n", 832 start + INITRD_SIZE, memory_end); 833 INITRD_START = INITRD_SIZE = 0; 834 } else { 835 pr_info("Moving initrd (0x%08lx -> " 836 "0x%08lx, size: %ld)\n", 837 INITRD_START, start, INITRD_SIZE); 838 memmove((void *) start, (void *) INITRD_START, 839 INITRD_SIZE); 840 INITRD_START = start; 841 } 842 } 843 } 844 #endif 845 846 /* 847 * Initialize the boot-time allocator 848 */ 849 bootmap_size = init_bootmem(start_pfn, end_pfn); 850 851 /* 852 * Register RAM areas with the bootmem allocator. 853 */ 854 855 for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) { 856 unsigned long start_chunk, end_chunk, pfn; 857 858 if (memory_chunk[i].type != CHUNK_READ_WRITE && 859 memory_chunk[i].type != CHUNK_CRASHK) 860 continue; 861 start_chunk = PFN_DOWN(memory_chunk[i].addr); 862 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size); 863 end_chunk = min(end_chunk, end_pfn); 864 if (start_chunk >= end_chunk) 865 continue; 866 memblock_add_node(PFN_PHYS(start_chunk), 867 PFN_PHYS(end_chunk - start_chunk), 0); 868 pfn = max(start_chunk, start_pfn); 869 for (; pfn < end_chunk; pfn++) 870 page_set_storage_key(PFN_PHYS(pfn), 871 PAGE_DEFAULT_KEY, 0); 872 } 873 874 psw_set_key(PAGE_DEFAULT_KEY); 875 876 free_bootmem_with_active_regions(0, max_pfn); 877 878 /* 879 * Reserve memory used for lowcore/command line/kernel image. 880 */ 881 reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT); 882 reserve_bootmem((unsigned long)_stext, 883 PFN_PHYS(start_pfn) - (unsigned long)_stext, 884 BOOTMEM_DEFAULT); 885 /* 886 * Reserve the bootmem bitmap itself as well. We do this in two 887 * steps (first step was init_bootmem()) because this catches 888 * the (very unlikely) case of us accidentally initializing the 889 * bootmem allocator with an invalid RAM area. 890 */ 891 reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size, 892 BOOTMEM_DEFAULT); 893 894 #ifdef CONFIG_CRASH_DUMP 895 if (crashk_res.start) 896 reserve_bootmem(crashk_res.start, 897 crashk_res.end - crashk_res.start + 1, 898 BOOTMEM_DEFAULT); 899 if (is_kdump_kernel()) 900 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE, 901 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT); 902 #endif 903 #ifdef CONFIG_BLK_DEV_INITRD 904 if (INITRD_START && INITRD_SIZE) { 905 if (INITRD_START + INITRD_SIZE <= memory_end) { 906 reserve_bootmem(INITRD_START, INITRD_SIZE, 907 BOOTMEM_DEFAULT); 908 initrd_start = INITRD_START; 909 initrd_end = initrd_start + INITRD_SIZE; 910 } else { 911 pr_err("initrd extends beyond end of " 912 "memory (0x%08lx > 0x%08lx) " 913 "disabling initrd\n", 914 initrd_start + INITRD_SIZE, memory_end); 915 initrd_start = initrd_end = 0; 916 } 917 } 918 #endif 919 } 920 921 /* 922 * Setup hardware capabilities. 923 */ 924 static void __init setup_hwcaps(void) 925 { 926 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 }; 927 struct cpuid cpu_id; 928 int i; 929 930 /* 931 * The store facility list bits numbers as found in the principles 932 * of operation are numbered with bit 1UL<<31 as number 0 to 933 * bit 1UL<<0 as number 31. 934 * Bit 0: instructions named N3, "backported" to esa-mode 935 * Bit 2: z/Architecture mode is active 936 * Bit 7: the store-facility-list-extended facility is installed 937 * Bit 17: the message-security assist is installed 938 * Bit 19: the long-displacement facility is installed 939 * Bit 21: the extended-immediate facility is installed 940 * Bit 22: extended-translation facility 3 is installed 941 * Bit 30: extended-translation facility 3 enhancement facility 942 * These get translated to: 943 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1, 944 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3, 945 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and 946 * HWCAP_S390_ETF3EH bit 8 (22 && 30). 947 */ 948 for (i = 0; i < 6; i++) 949 if (test_facility(stfl_bits[i])) 950 elf_hwcap |= 1UL << i; 951 952 if (test_facility(22) && test_facility(30)) 953 elf_hwcap |= HWCAP_S390_ETF3EH; 954 955 /* 956 * Check for additional facilities with store-facility-list-extended. 957 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0 958 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information 959 * as stored by stfl, bits 32-xxx contain additional facilities. 960 * How many facility words are stored depends on the number of 961 * doublewords passed to the instruction. The additional facilities 962 * are: 963 * Bit 42: decimal floating point facility is installed 964 * Bit 44: perform floating point operation facility is installed 965 * translated to: 966 * HWCAP_S390_DFP bit 6 (42 && 44). 967 */ 968 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44)) 969 elf_hwcap |= HWCAP_S390_DFP; 970 971 /* 972 * Huge page support HWCAP_S390_HPAGE is bit 7. 973 */ 974 if (MACHINE_HAS_HPAGE) 975 elf_hwcap |= HWCAP_S390_HPAGE; 976 977 /* 978 * 64-bit register support for 31-bit processes 979 * HWCAP_S390_HIGH_GPRS is bit 9. 980 */ 981 elf_hwcap |= HWCAP_S390_HIGH_GPRS; 982 983 get_cpu_id(&cpu_id); 984 switch (cpu_id.machine) { 985 case 0x9672: 986 #if !defined(CONFIG_64BIT) 987 default: /* Use "g5" as default for 31 bit kernels. */ 988 #endif 989 strcpy(elf_platform, "g5"); 990 break; 991 case 0x2064: 992 case 0x2066: 993 #if defined(CONFIG_64BIT) 994 default: /* Use "z900" as default for 64 bit kernels. */ 995 #endif 996 strcpy(elf_platform, "z900"); 997 break; 998 case 0x2084: 999 case 0x2086: 1000 strcpy(elf_platform, "z990"); 1001 break; 1002 case 0x2094: 1003 case 0x2096: 1004 strcpy(elf_platform, "z9-109"); 1005 break; 1006 case 0x2097: 1007 case 0x2098: 1008 strcpy(elf_platform, "z10"); 1009 break; 1010 case 0x2817: 1011 case 0x2818: 1012 strcpy(elf_platform, "z196"); 1013 break; 1014 } 1015 } 1016 1017 /* 1018 * Setup function called from init/main.c just after the banner 1019 * was printed. 1020 */ 1021 1022 void __init setup_arch(char **cmdline_p) 1023 { 1024 /* 1025 * print what head.S has found out about the machine 1026 */ 1027 #ifndef CONFIG_64BIT 1028 if (MACHINE_IS_VM) 1029 pr_info("Linux is running as a z/VM " 1030 "guest operating system in 31-bit mode\n"); 1031 else if (MACHINE_IS_LPAR) 1032 pr_info("Linux is running natively in 31-bit mode\n"); 1033 if (MACHINE_HAS_IEEE) 1034 pr_info("The hardware system has IEEE compatible " 1035 "floating point units\n"); 1036 else 1037 pr_info("The hardware system has no IEEE compatible " 1038 "floating point units\n"); 1039 #else /* CONFIG_64BIT */ 1040 if (MACHINE_IS_VM) 1041 pr_info("Linux is running as a z/VM " 1042 "guest operating system in 64-bit mode\n"); 1043 else if (MACHINE_IS_KVM) 1044 pr_info("Linux is running under KVM in 64-bit mode\n"); 1045 else if (MACHINE_IS_LPAR) 1046 pr_info("Linux is running natively in 64-bit mode\n"); 1047 #endif /* CONFIG_64BIT */ 1048 1049 /* Have one command line that is parsed and saved in /proc/cmdline */ 1050 /* boot_command_line has been already set up in early.c */ 1051 *cmdline_p = boot_command_line; 1052 1053 ROOT_DEV = Root_RAM0; 1054 1055 init_mm.start_code = PAGE_OFFSET; 1056 init_mm.end_code = (unsigned long) &_etext; 1057 init_mm.end_data = (unsigned long) &_edata; 1058 init_mm.brk = (unsigned long) &_end; 1059 1060 if (MACHINE_HAS_MVCOS) 1061 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess)); 1062 else 1063 memcpy(&uaccess, &uaccess_std, sizeof(uaccess)); 1064 1065 parse_early_param(); 1066 1067 os_info_init(); 1068 setup_ipl(); 1069 setup_memory_end(); 1070 setup_addressing_mode(); 1071 reserve_oldmem(); 1072 reserve_crashkernel(); 1073 setup_memory(); 1074 setup_resources(); 1075 setup_vmcoreinfo(); 1076 setup_lowcore(); 1077 1078 cpu_init(); 1079 s390_init_cpu_topology(); 1080 1081 /* 1082 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM). 1083 */ 1084 setup_hwcaps(); 1085 1086 /* 1087 * Create kernel page tables and switch to virtual addressing. 1088 */ 1089 paging_init(); 1090 1091 /* Setup default console */ 1092 conmode_default(); 1093 set_preferred_console(); 1094 1095 /* Setup zfcpdump support */ 1096 setup_zfcpdump(console_devno); 1097 } 1098