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) + THREAD_SIZE; 381 lc->async_stack = (unsigned long) 382 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE; 383 lc->panic_stack = (unsigned long) 384 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE; 385 lc->current_task = (unsigned long) init_thread_union.thread_info.task; 386 lc->thread_info = (unsigned long) &init_thread_union; 387 lc->machine_flags = S390_lowcore.machine_flags; 388 lc->stfl_fac_list = S390_lowcore.stfl_fac_list; 389 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list, 390 MAX_FACILITY_BIT/8); 391 #ifndef CONFIG_64BIT 392 if (MACHINE_HAS_IEEE) { 393 lc->extended_save_area_addr = (__u32) 394 __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0); 395 /* enable extended save area */ 396 __ctl_set_bit(14, 29); 397 } 398 #else 399 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0]; 400 #endif 401 lc->sync_enter_timer = S390_lowcore.sync_enter_timer; 402 lc->async_enter_timer = S390_lowcore.async_enter_timer; 403 lc->exit_timer = S390_lowcore.exit_timer; 404 lc->user_timer = S390_lowcore.user_timer; 405 lc->system_timer = S390_lowcore.system_timer; 406 lc->steal_timer = S390_lowcore.steal_timer; 407 lc->last_update_timer = S390_lowcore.last_update_timer; 408 lc->last_update_clock = S390_lowcore.last_update_clock; 409 lc->ftrace_func = S390_lowcore.ftrace_func; 410 411 restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0); 412 restart_stack += ASYNC_SIZE; 413 414 /* 415 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 416 * restart data to the absolute zero lowcore. This is necesary if 417 * PSW restart is done on an offline CPU that has lowcore zero. 418 */ 419 lc->restart_stack = (unsigned long) restart_stack; 420 lc->restart_fn = (unsigned long) do_restart; 421 lc->restart_data = 0; 422 lc->restart_source = -1UL; 423 424 /* Setup absolute zero lowcore */ 425 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack); 426 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn); 427 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data); 428 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source); 429 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw); 430 431 set_prefix((u32)(unsigned long) lc); 432 lowcore_ptr[0] = lc; 433 } 434 435 static struct resource code_resource = { 436 .name = "Kernel code", 437 .flags = IORESOURCE_BUSY | IORESOURCE_MEM, 438 }; 439 440 static struct resource data_resource = { 441 .name = "Kernel data", 442 .flags = IORESOURCE_BUSY | IORESOURCE_MEM, 443 }; 444 445 static struct resource bss_resource = { 446 .name = "Kernel bss", 447 .flags = IORESOURCE_BUSY | IORESOURCE_MEM, 448 }; 449 450 static struct resource __initdata *standard_resources[] = { 451 &code_resource, 452 &data_resource, 453 &bss_resource, 454 }; 455 456 static void __init setup_resources(void) 457 { 458 struct resource *res, *std_res, *sub_res; 459 int i, j; 460 461 code_resource.start = (unsigned long) &_text; 462 code_resource.end = (unsigned long) &_etext - 1; 463 data_resource.start = (unsigned long) &_etext; 464 data_resource.end = (unsigned long) &_edata - 1; 465 bss_resource.start = (unsigned long) &__bss_start; 466 bss_resource.end = (unsigned long) &__bss_stop - 1; 467 468 for (i = 0; i < MEMORY_CHUNKS; i++) { 469 if (!memory_chunk[i].size) 470 continue; 471 if (memory_chunk[i].type == CHUNK_OLDMEM || 472 memory_chunk[i].type == CHUNK_CRASHK) 473 continue; 474 res = alloc_bootmem_low(sizeof(*res)); 475 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM; 476 switch (memory_chunk[i].type) { 477 case CHUNK_READ_WRITE: 478 case CHUNK_CRASHK: 479 res->name = "System RAM"; 480 break; 481 case CHUNK_READ_ONLY: 482 res->name = "System ROM"; 483 res->flags |= IORESOURCE_READONLY; 484 break; 485 default: 486 res->name = "reserved"; 487 } 488 res->start = memory_chunk[i].addr; 489 res->end = res->start + memory_chunk[i].size - 1; 490 request_resource(&iomem_resource, res); 491 492 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 493 std_res = standard_resources[j]; 494 if (std_res->start < res->start || 495 std_res->start > res->end) 496 continue; 497 if (std_res->end > res->end) { 498 sub_res = alloc_bootmem_low(sizeof(*sub_res)); 499 *sub_res = *std_res; 500 sub_res->end = res->end; 501 std_res->start = res->end + 1; 502 request_resource(res, sub_res); 503 } else { 504 request_resource(res, std_res); 505 } 506 } 507 } 508 } 509 510 unsigned long real_memory_size; 511 EXPORT_SYMBOL_GPL(real_memory_size); 512 513 static void __init setup_memory_end(void) 514 { 515 unsigned long vmax, vmalloc_size, tmp; 516 int i; 517 518 519 #ifdef CONFIG_ZFCPDUMP 520 if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) { 521 memory_end = ZFCPDUMP_HSA_SIZE; 522 memory_end_set = 1; 523 } 524 #endif 525 real_memory_size = 0; 526 memory_end &= PAGE_MASK; 527 528 /* 529 * Make sure all chunks are MAX_ORDER aligned so we don't need the 530 * extra checks that HOLES_IN_ZONE would require. 531 */ 532 for (i = 0; i < MEMORY_CHUNKS; i++) { 533 unsigned long start, end; 534 struct mem_chunk *chunk; 535 unsigned long align; 536 537 chunk = &memory_chunk[i]; 538 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1); 539 start = (chunk->addr + align - 1) & ~(align - 1); 540 end = (chunk->addr + chunk->size) & ~(align - 1); 541 if (start >= end) 542 memset(chunk, 0, sizeof(*chunk)); 543 else { 544 chunk->addr = start; 545 chunk->size = end - start; 546 } 547 real_memory_size = max(real_memory_size, 548 chunk->addr + chunk->size); 549 } 550 551 /* Choose kernel address space layout: 2, 3, or 4 levels. */ 552 #ifdef CONFIG_64BIT 553 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN; 554 tmp = (memory_end ?: real_memory_size) / PAGE_SIZE; 555 tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size; 556 if (tmp <= (1UL << 42)) 557 vmax = 1UL << 42; /* 3-level kernel page table */ 558 else 559 vmax = 1UL << 53; /* 4-level kernel page table */ 560 /* module area is at the end of the kernel address space. */ 561 MODULES_END = vmax; 562 MODULES_VADDR = MODULES_END - MODULES_LEN; 563 VMALLOC_END = MODULES_VADDR; 564 #else 565 vmalloc_size = VMALLOC_END ?: 96UL << 20; 566 vmax = 1UL << 31; /* 2-level kernel page table */ 567 /* vmalloc area is at the end of the kernel address space. */ 568 VMALLOC_END = vmax; 569 #endif 570 VMALLOC_START = vmax - vmalloc_size; 571 572 /* Split remaining virtual space between 1:1 mapping & vmemmap array */ 573 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page)); 574 tmp = VMALLOC_START - tmp * sizeof(struct page); 575 tmp &= ~((vmax >> 11) - 1); /* align to page table level */ 576 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS); 577 vmemmap = (struct page *) tmp; 578 579 /* Take care that memory_end is set and <= vmemmap */ 580 memory_end = min(memory_end ?: real_memory_size, tmp); 581 582 /* Fixup memory chunk array to fit into 0..memory_end */ 583 for (i = 0; i < MEMORY_CHUNKS; i++) { 584 struct mem_chunk *chunk = &memory_chunk[i]; 585 586 if (chunk->addr >= memory_end) { 587 memset(chunk, 0, sizeof(*chunk)); 588 continue; 589 } 590 if (chunk->addr + chunk->size > memory_end) 591 chunk->size = memory_end - chunk->addr; 592 } 593 } 594 595 static void __init setup_vmcoreinfo(void) 596 { 597 mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note()); 598 } 599 600 #ifdef CONFIG_CRASH_DUMP 601 602 /* 603 * Find suitable location for crashkernel memory 604 */ 605 static unsigned long __init find_crash_base(unsigned long crash_size, 606 char **msg) 607 { 608 unsigned long crash_base; 609 struct mem_chunk *chunk; 610 int i; 611 612 if (memory_chunk[0].size < crash_size) { 613 *msg = "first memory chunk must be at least crashkernel size"; 614 return 0; 615 } 616 if (OLDMEM_BASE && crash_size == OLDMEM_SIZE) 617 return OLDMEM_BASE; 618 619 for (i = MEMORY_CHUNKS - 1; i >= 0; i--) { 620 chunk = &memory_chunk[i]; 621 if (chunk->size == 0) 622 continue; 623 if (chunk->type != CHUNK_READ_WRITE) 624 continue; 625 if (chunk->size < crash_size) 626 continue; 627 crash_base = (chunk->addr + chunk->size) - crash_size; 628 if (crash_base < crash_size) 629 continue; 630 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX) 631 continue; 632 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE) 633 continue; 634 return crash_base; 635 } 636 *msg = "no suitable area found"; 637 return 0; 638 } 639 640 /* 641 * Check if crash_base and crash_size is valid 642 */ 643 static int __init verify_crash_base(unsigned long crash_base, 644 unsigned long crash_size, 645 char **msg) 646 { 647 struct mem_chunk *chunk; 648 int i; 649 650 /* 651 * Because we do the swap to zero, we must have at least 'crash_size' 652 * bytes free space before crash_base 653 */ 654 if (crash_size > crash_base) { 655 *msg = "crashkernel offset must be greater than size"; 656 return -EINVAL; 657 } 658 659 /* First memory chunk must be at least crash_size */ 660 if (memory_chunk[0].size < crash_size) { 661 *msg = "first memory chunk must be at least crashkernel size"; 662 return -EINVAL; 663 } 664 /* Check if we fit into the respective memory chunk */ 665 for (i = 0; i < MEMORY_CHUNKS; i++) { 666 chunk = &memory_chunk[i]; 667 if (chunk->size == 0) 668 continue; 669 if (crash_base < chunk->addr) 670 continue; 671 if (crash_base >= chunk->addr + chunk->size) 672 continue; 673 /* we have found the memory chunk */ 674 if (crash_base + crash_size > chunk->addr + chunk->size) { 675 *msg = "selected memory chunk is too small for " 676 "crashkernel memory"; 677 return -EINVAL; 678 } 679 return 0; 680 } 681 *msg = "invalid memory range specified"; 682 return -EINVAL; 683 } 684 685 /* 686 * Reserve kdump memory by creating a memory hole in the mem_chunk array 687 */ 688 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size, 689 int type) 690 { 691 create_mem_hole(memory_chunk, addr, size, type); 692 } 693 694 /* 695 * When kdump is enabled, we have to ensure that no memory from 696 * the area [0 - crashkernel memory size] and 697 * [crashk_res.start - crashk_res.end] is set offline. 698 */ 699 static int kdump_mem_notifier(struct notifier_block *nb, 700 unsigned long action, void *data) 701 { 702 struct memory_notify *arg = data; 703 704 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 705 return NOTIFY_BAD; 706 if (arg->start_pfn > PFN_DOWN(crashk_res.end)) 707 return NOTIFY_OK; 708 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start)) 709 return NOTIFY_OK; 710 return NOTIFY_BAD; 711 } 712 713 static struct notifier_block kdump_mem_nb = { 714 .notifier_call = kdump_mem_notifier, 715 }; 716 717 #endif 718 719 /* 720 * Make sure that oldmem, where the dump is stored, is protected 721 */ 722 static void reserve_oldmem(void) 723 { 724 #ifdef CONFIG_CRASH_DUMP 725 if (!OLDMEM_BASE) 726 return; 727 728 reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM); 729 reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE, 730 CHUNK_OLDMEM); 731 if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size) 732 saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1; 733 else 734 saved_max_pfn = PFN_DOWN(real_memory_size) - 1; 735 #endif 736 } 737 738 /* 739 * Reserve memory for kdump kernel to be loaded with kexec 740 */ 741 static void __init reserve_crashkernel(void) 742 { 743 #ifdef CONFIG_CRASH_DUMP 744 unsigned long long crash_base, crash_size; 745 char *msg = NULL; 746 int rc; 747 748 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size, 749 &crash_base); 750 if (rc || crash_size == 0) 751 return; 752 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 753 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 754 if (register_memory_notifier(&kdump_mem_nb)) 755 return; 756 if (!crash_base) 757 crash_base = find_crash_base(crash_size, &msg); 758 if (!crash_base) { 759 pr_info("crashkernel reservation failed: %s\n", msg); 760 unregister_memory_notifier(&kdump_mem_nb); 761 return; 762 } 763 if (verify_crash_base(crash_base, crash_size, &msg)) { 764 pr_info("crashkernel reservation failed: %s\n", msg); 765 unregister_memory_notifier(&kdump_mem_nb); 766 return; 767 } 768 if (!OLDMEM_BASE && MACHINE_IS_VM) 769 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 770 crashk_res.start = crash_base; 771 crashk_res.end = crash_base + crash_size - 1; 772 insert_resource(&iomem_resource, &crashk_res); 773 reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK); 774 pr_info("Reserving %lluMB of memory at %lluMB " 775 "for crashkernel (System RAM: %luMB)\n", 776 crash_size >> 20, crash_base >> 20, memory_end >> 20); 777 os_info_crashkernel_add(crash_base, crash_size); 778 #endif 779 } 780 781 static void __init setup_memory(void) 782 { 783 unsigned long bootmap_size; 784 unsigned long start_pfn, end_pfn; 785 int i; 786 787 /* 788 * partially used pages are not usable - thus 789 * we are rounding upwards: 790 */ 791 start_pfn = PFN_UP(__pa(&_end)); 792 end_pfn = max_pfn = PFN_DOWN(memory_end); 793 794 #ifdef CONFIG_BLK_DEV_INITRD 795 /* 796 * Move the initrd in case the bitmap of the bootmem allocater 797 * would overwrite it. 798 */ 799 800 if (INITRD_START && INITRD_SIZE) { 801 unsigned long bmap_size; 802 unsigned long start; 803 804 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1); 805 bmap_size = PFN_PHYS(bmap_size); 806 807 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) { 808 start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE; 809 810 #ifdef CONFIG_CRASH_DUMP 811 if (OLDMEM_BASE) { 812 /* Move initrd behind kdump oldmem */ 813 if (start + INITRD_SIZE > OLDMEM_BASE && 814 start < OLDMEM_BASE + OLDMEM_SIZE) 815 start = OLDMEM_BASE + OLDMEM_SIZE; 816 } 817 #endif 818 if (start + INITRD_SIZE > memory_end) { 819 pr_err("initrd extends beyond end of " 820 "memory (0x%08lx > 0x%08lx) " 821 "disabling initrd\n", 822 start + INITRD_SIZE, memory_end); 823 INITRD_START = INITRD_SIZE = 0; 824 } else { 825 pr_info("Moving initrd (0x%08lx -> " 826 "0x%08lx, size: %ld)\n", 827 INITRD_START, start, INITRD_SIZE); 828 memmove((void *) start, (void *) INITRD_START, 829 INITRD_SIZE); 830 INITRD_START = start; 831 } 832 } 833 } 834 #endif 835 836 /* 837 * Initialize the boot-time allocator 838 */ 839 bootmap_size = init_bootmem(start_pfn, end_pfn); 840 841 /* 842 * Register RAM areas with the bootmem allocator. 843 */ 844 845 for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) { 846 unsigned long start_chunk, end_chunk, pfn; 847 848 if (memory_chunk[i].type != CHUNK_READ_WRITE && 849 memory_chunk[i].type != CHUNK_CRASHK) 850 continue; 851 start_chunk = PFN_DOWN(memory_chunk[i].addr); 852 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size); 853 end_chunk = min(end_chunk, end_pfn); 854 if (start_chunk >= end_chunk) 855 continue; 856 memblock_add_node(PFN_PHYS(start_chunk), 857 PFN_PHYS(end_chunk - start_chunk), 0); 858 pfn = max(start_chunk, start_pfn); 859 storage_key_init_range(PFN_PHYS(pfn), PFN_PHYS(end_chunk)); 860 } 861 862 psw_set_key(PAGE_DEFAULT_KEY); 863 864 free_bootmem_with_active_regions(0, max_pfn); 865 866 /* 867 * Reserve memory used for lowcore/command line/kernel image. 868 */ 869 reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT); 870 reserve_bootmem((unsigned long)_stext, 871 PFN_PHYS(start_pfn) - (unsigned long)_stext, 872 BOOTMEM_DEFAULT); 873 /* 874 * Reserve the bootmem bitmap itself as well. We do this in two 875 * steps (first step was init_bootmem()) because this catches 876 * the (very unlikely) case of us accidentally initializing the 877 * bootmem allocator with an invalid RAM area. 878 */ 879 reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size, 880 BOOTMEM_DEFAULT); 881 882 #ifdef CONFIG_CRASH_DUMP 883 if (crashk_res.start) 884 reserve_bootmem(crashk_res.start, 885 crashk_res.end - crashk_res.start + 1, 886 BOOTMEM_DEFAULT); 887 if (is_kdump_kernel()) 888 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE, 889 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT); 890 #endif 891 #ifdef CONFIG_BLK_DEV_INITRD 892 if (INITRD_START && INITRD_SIZE) { 893 if (INITRD_START + INITRD_SIZE <= memory_end) { 894 reserve_bootmem(INITRD_START, INITRD_SIZE, 895 BOOTMEM_DEFAULT); 896 initrd_start = INITRD_START; 897 initrd_end = initrd_start + INITRD_SIZE; 898 } else { 899 pr_err("initrd extends beyond end of " 900 "memory (0x%08lx > 0x%08lx) " 901 "disabling initrd\n", 902 initrd_start + INITRD_SIZE, memory_end); 903 initrd_start = initrd_end = 0; 904 } 905 } 906 #endif 907 } 908 909 /* 910 * Setup hardware capabilities. 911 */ 912 static void __init setup_hwcaps(void) 913 { 914 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 }; 915 struct cpuid cpu_id; 916 int i; 917 918 /* 919 * The store facility list bits numbers as found in the principles 920 * of operation are numbered with bit 1UL<<31 as number 0 to 921 * bit 1UL<<0 as number 31. 922 * Bit 0: instructions named N3, "backported" to esa-mode 923 * Bit 2: z/Architecture mode is active 924 * Bit 7: the store-facility-list-extended facility is installed 925 * Bit 17: the message-security assist is installed 926 * Bit 19: the long-displacement facility is installed 927 * Bit 21: the extended-immediate facility is installed 928 * Bit 22: extended-translation facility 3 is installed 929 * Bit 30: extended-translation facility 3 enhancement facility 930 * These get translated to: 931 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1, 932 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3, 933 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and 934 * HWCAP_S390_ETF3EH bit 8 (22 && 30). 935 */ 936 for (i = 0; i < 6; i++) 937 if (test_facility(stfl_bits[i])) 938 elf_hwcap |= 1UL << i; 939 940 if (test_facility(22) && test_facility(30)) 941 elf_hwcap |= HWCAP_S390_ETF3EH; 942 943 /* 944 * Check for additional facilities with store-facility-list-extended. 945 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0 946 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information 947 * as stored by stfl, bits 32-xxx contain additional facilities. 948 * How many facility words are stored depends on the number of 949 * doublewords passed to the instruction. The additional facilities 950 * are: 951 * Bit 42: decimal floating point facility is installed 952 * Bit 44: perform floating point operation facility is installed 953 * translated to: 954 * HWCAP_S390_DFP bit 6 (42 && 44). 955 */ 956 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44)) 957 elf_hwcap |= HWCAP_S390_DFP; 958 959 /* 960 * Huge page support HWCAP_S390_HPAGE is bit 7. 961 */ 962 if (MACHINE_HAS_HPAGE) 963 elf_hwcap |= HWCAP_S390_HPAGE; 964 965 #if defined(CONFIG_64BIT) 966 /* 967 * 64-bit register support for 31-bit processes 968 * HWCAP_S390_HIGH_GPRS is bit 9. 969 */ 970 elf_hwcap |= HWCAP_S390_HIGH_GPRS; 971 972 /* 973 * Transactional execution support HWCAP_S390_TE is bit 10. 974 */ 975 if (test_facility(50) && test_facility(73)) 976 elf_hwcap |= HWCAP_S390_TE; 977 #endif 978 979 get_cpu_id(&cpu_id); 980 switch (cpu_id.machine) { 981 case 0x9672: 982 #if !defined(CONFIG_64BIT) 983 default: /* Use "g5" as default for 31 bit kernels. */ 984 #endif 985 strcpy(elf_platform, "g5"); 986 break; 987 case 0x2064: 988 case 0x2066: 989 #if defined(CONFIG_64BIT) 990 default: /* Use "z900" as default for 64 bit kernels. */ 991 #endif 992 strcpy(elf_platform, "z900"); 993 break; 994 case 0x2084: 995 case 0x2086: 996 strcpy(elf_platform, "z990"); 997 break; 998 case 0x2094: 999 case 0x2096: 1000 strcpy(elf_platform, "z9-109"); 1001 break; 1002 case 0x2097: 1003 case 0x2098: 1004 strcpy(elf_platform, "z10"); 1005 break; 1006 case 0x2817: 1007 case 0x2818: 1008 strcpy(elf_platform, "z196"); 1009 break; 1010 case 0x2827: 1011 strcpy(elf_platform, "zEC12"); 1012 break; 1013 } 1014 } 1015 1016 /* 1017 * Setup function called from init/main.c just after the banner 1018 * was printed. 1019 */ 1020 1021 void __init setup_arch(char **cmdline_p) 1022 { 1023 /* 1024 * print what head.S has found out about the machine 1025 */ 1026 #ifndef CONFIG_64BIT 1027 if (MACHINE_IS_VM) 1028 pr_info("Linux is running as a z/VM " 1029 "guest operating system in 31-bit mode\n"); 1030 else if (MACHINE_IS_LPAR) 1031 pr_info("Linux is running natively in 31-bit mode\n"); 1032 if (MACHINE_HAS_IEEE) 1033 pr_info("The hardware system has IEEE compatible " 1034 "floating point units\n"); 1035 else 1036 pr_info("The hardware system has no IEEE compatible " 1037 "floating point units\n"); 1038 #else /* CONFIG_64BIT */ 1039 if (MACHINE_IS_VM) 1040 pr_info("Linux is running as a z/VM " 1041 "guest operating system in 64-bit mode\n"); 1042 else if (MACHINE_IS_KVM) 1043 pr_info("Linux is running under KVM in 64-bit mode\n"); 1044 else if (MACHINE_IS_LPAR) 1045 pr_info("Linux is running natively in 64-bit mode\n"); 1046 #endif /* CONFIG_64BIT */ 1047 1048 /* Have one command line that is parsed and saved in /proc/cmdline */ 1049 /* boot_command_line has been already set up in early.c */ 1050 *cmdline_p = boot_command_line; 1051 1052 ROOT_DEV = Root_RAM0; 1053 1054 init_mm.start_code = PAGE_OFFSET; 1055 init_mm.end_code = (unsigned long) &_etext; 1056 init_mm.end_data = (unsigned long) &_edata; 1057 init_mm.brk = (unsigned long) &_end; 1058 1059 if (MACHINE_HAS_MVCOS) 1060 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess)); 1061 else 1062 memcpy(&uaccess, &uaccess_std, sizeof(uaccess)); 1063 1064 parse_early_param(); 1065 1066 os_info_init(); 1067 setup_ipl(); 1068 setup_memory_end(); 1069 setup_addressing_mode(); 1070 reserve_oldmem(); 1071 reserve_crashkernel(); 1072 setup_memory(); 1073 setup_resources(); 1074 setup_vmcoreinfo(); 1075 setup_lowcore(); 1076 1077 cpu_init(); 1078 s390_init_cpu_topology(); 1079 1080 /* 1081 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM). 1082 */ 1083 setup_hwcaps(); 1084 1085 /* 1086 * Create kernel page tables and switch to virtual addressing. 1087 */ 1088 paging_init(); 1089 1090 /* Setup default console */ 1091 conmode_default(); 1092 set_preferred_console(); 1093 1094 /* Setup zfcpdump support */ 1095 setup_zfcpdump(console_devno); 1096 } 1097