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