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 /* 68 * User copy operations. 69 */ 70 struct uaccess_ops uaccess; 71 EXPORT_SYMBOL(uaccess); 72 73 /* 74 * Machine setup.. 75 */ 76 unsigned int console_mode = 0; 77 EXPORT_SYMBOL(console_mode); 78 79 unsigned int console_devno = -1; 80 EXPORT_SYMBOL(console_devno); 81 82 unsigned int console_irq = -1; 83 EXPORT_SYMBOL(console_irq); 84 85 unsigned long elf_hwcap = 0; 86 char elf_platform[ELF_PLATFORM_SIZE]; 87 88 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS]; 89 90 int __initdata memory_end_set; 91 unsigned long __initdata memory_end; 92 93 unsigned long VMALLOC_START; 94 EXPORT_SYMBOL(VMALLOC_START); 95 96 unsigned long VMALLOC_END; 97 EXPORT_SYMBOL(VMALLOC_END); 98 99 struct page *vmemmap; 100 EXPORT_SYMBOL(vmemmap); 101 102 #ifdef CONFIG_64BIT 103 unsigned long MODULES_VADDR; 104 unsigned long MODULES_END; 105 #endif 106 107 /* An array with a pointer to the lowcore of every CPU. */ 108 struct _lowcore *lowcore_ptr[NR_CPUS]; 109 EXPORT_SYMBOL(lowcore_ptr); 110 111 /* 112 * This is set up by the setup-routine at boot-time 113 * for S390 need to find out, what we have to setup 114 * using address 0x10400 ... 115 */ 116 117 #include <asm/setup.h> 118 119 /* 120 * condev= and conmode= setup parameter. 121 */ 122 123 static int __init condev_setup(char *str) 124 { 125 int vdev; 126 127 vdev = simple_strtoul(str, &str, 0); 128 if (vdev >= 0 && vdev < 65536) { 129 console_devno = vdev; 130 console_irq = -1; 131 } 132 return 1; 133 } 134 135 __setup("condev=", condev_setup); 136 137 static void __init set_preferred_console(void) 138 { 139 if (MACHINE_IS_KVM) { 140 if (sclp_has_vt220()) 141 add_preferred_console("ttyS", 1, NULL); 142 else if (sclp_has_linemode()) 143 add_preferred_console("ttyS", 0, NULL); 144 else 145 add_preferred_console("hvc", 0, NULL); 146 } else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP) 147 add_preferred_console("ttyS", 0, NULL); 148 else if (CONSOLE_IS_3270) 149 add_preferred_console("tty3270", 0, NULL); 150 } 151 152 static int __init conmode_setup(char *str) 153 { 154 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 155 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0) 156 SET_CONSOLE_SCLP; 157 #endif 158 #if defined(CONFIG_TN3215_CONSOLE) 159 if (strncmp(str, "3215", 5) == 0) 160 SET_CONSOLE_3215; 161 #endif 162 #if defined(CONFIG_TN3270_CONSOLE) 163 if (strncmp(str, "3270", 5) == 0) 164 SET_CONSOLE_3270; 165 #endif 166 set_preferred_console(); 167 return 1; 168 } 169 170 __setup("conmode=", conmode_setup); 171 172 static void __init conmode_default(void) 173 { 174 char query_buffer[1024]; 175 char *ptr; 176 177 if (MACHINE_IS_VM) { 178 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL); 179 console_devno = simple_strtoul(query_buffer + 5, NULL, 16); 180 ptr = strstr(query_buffer, "SUBCHANNEL ="); 181 console_irq = simple_strtoul(ptr + 13, NULL, 16); 182 cpcmd("QUERY TERM", query_buffer, 1024, NULL); 183 ptr = strstr(query_buffer, "CONMODE"); 184 /* 185 * Set the conmode to 3215 so that the device recognition 186 * will set the cu_type of the console to 3215. If the 187 * conmode is 3270 and we don't set it back then both 188 * 3215 and the 3270 driver will try to access the console 189 * device (3215 as console and 3270 as normal tty). 190 */ 191 cpcmd("TERM CONMODE 3215", NULL, 0, NULL); 192 if (ptr == NULL) { 193 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 194 SET_CONSOLE_SCLP; 195 #endif 196 return; 197 } 198 if (strncmp(ptr + 8, "3270", 4) == 0) { 199 #if defined(CONFIG_TN3270_CONSOLE) 200 SET_CONSOLE_3270; 201 #elif defined(CONFIG_TN3215_CONSOLE) 202 SET_CONSOLE_3215; 203 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 204 SET_CONSOLE_SCLP; 205 #endif 206 } else if (strncmp(ptr + 8, "3215", 4) == 0) { 207 #if defined(CONFIG_TN3215_CONSOLE) 208 SET_CONSOLE_3215; 209 #elif defined(CONFIG_TN3270_CONSOLE) 210 SET_CONSOLE_3270; 211 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 212 SET_CONSOLE_SCLP; 213 #endif 214 } 215 } else { 216 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 217 SET_CONSOLE_SCLP; 218 #endif 219 } 220 } 221 222 #ifdef CONFIG_ZFCPDUMP 223 static void __init setup_zfcpdump(void) 224 { 225 if (ipl_info.type != IPL_TYPE_FCP_DUMP) 226 return; 227 if (OLDMEM_BASE) 228 return; 229 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev"); 230 console_loglevel = 2; 231 } 232 #else 233 static inline void setup_zfcpdump(void) {} 234 #endif /* CONFIG_ZFCPDUMP */ 235 236 /* 237 * Reboot, halt and power_off stubs. They just call _machine_restart, 238 * _machine_halt or _machine_power_off. 239 */ 240 241 void machine_restart(char *command) 242 { 243 if ((!in_interrupt() && !in_atomic()) || oops_in_progress) 244 /* 245 * Only unblank the console if we are called in enabled 246 * context or a bust_spinlocks cleared the way for us. 247 */ 248 console_unblank(); 249 _machine_restart(command); 250 } 251 252 void machine_halt(void) 253 { 254 if (!in_interrupt() || oops_in_progress) 255 /* 256 * Only unblank the console if we are called in enabled 257 * context or a bust_spinlocks cleared the way for us. 258 */ 259 console_unblank(); 260 _machine_halt(); 261 } 262 263 void machine_power_off(void) 264 { 265 if (!in_interrupt() || oops_in_progress) 266 /* 267 * Only unblank the console if we are called in enabled 268 * context or a bust_spinlocks cleared the way for us. 269 */ 270 console_unblank(); 271 _machine_power_off(); 272 } 273 274 /* 275 * Dummy power off function. 276 */ 277 void (*pm_power_off)(void) = machine_power_off; 278 EXPORT_SYMBOL_GPL(pm_power_off); 279 280 static int __init early_parse_mem(char *p) 281 { 282 memory_end = memparse(p, &p); 283 memory_end_set = 1; 284 return 0; 285 } 286 early_param("mem", early_parse_mem); 287 288 static int __init parse_vmalloc(char *arg) 289 { 290 if (!arg) 291 return -EINVAL; 292 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK; 293 return 0; 294 } 295 early_param("vmalloc", parse_vmalloc); 296 297 static int __init early_parse_user_mode(char *p) 298 { 299 if (!p || strcmp(p, "primary") == 0) 300 return 0; 301 return 1; 302 } 303 early_param("user_mode", early_parse_user_mode); 304 305 void *restart_stack __attribute__((__section__(".data"))); 306 307 static void __init setup_lowcore(void) 308 { 309 struct _lowcore *lc; 310 311 /* 312 * Setup lowcore for boot cpu 313 */ 314 BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096); 315 lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0); 316 lc->restart_psw.mask = PSW_KERNEL_BITS; 317 lc->restart_psw.addr = 318 PSW_ADDR_AMODE | (unsigned long) restart_int_handler; 319 lc->external_new_psw.mask = PSW_KERNEL_BITS | 320 PSW_MASK_DAT | PSW_MASK_MCHECK; 321 lc->external_new_psw.addr = 322 PSW_ADDR_AMODE | (unsigned long) ext_int_handler; 323 lc->svc_new_psw.mask = PSW_KERNEL_BITS | 324 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK; 325 lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call; 326 lc->program_new_psw.mask = PSW_KERNEL_BITS | 327 PSW_MASK_DAT | PSW_MASK_MCHECK; 328 lc->program_new_psw.addr = 329 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler; 330 lc->mcck_new_psw.mask = PSW_KERNEL_BITS; 331 lc->mcck_new_psw.addr = 332 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler; 333 lc->io_new_psw.mask = PSW_KERNEL_BITS | 334 PSW_MASK_DAT | PSW_MASK_MCHECK; 335 lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler; 336 lc->clock_comparator = -1ULL; 337 lc->kernel_stack = ((unsigned long) &init_thread_union) 338 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); 339 lc->async_stack = (unsigned long) 340 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) 341 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); 342 lc->panic_stack = (unsigned long) 343 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) 344 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); 345 lc->current_task = (unsigned long) init_thread_union.thread_info.task; 346 lc->thread_info = (unsigned long) &init_thread_union; 347 lc->machine_flags = S390_lowcore.machine_flags; 348 lc->stfl_fac_list = S390_lowcore.stfl_fac_list; 349 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list, 350 MAX_FACILITY_BIT/8); 351 #ifndef CONFIG_64BIT 352 if (MACHINE_HAS_IEEE) { 353 lc->extended_save_area_addr = (__u32) 354 __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0); 355 /* enable extended save area */ 356 __ctl_set_bit(14, 29); 357 } 358 #else 359 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0]; 360 #endif 361 lc->sync_enter_timer = S390_lowcore.sync_enter_timer; 362 lc->async_enter_timer = S390_lowcore.async_enter_timer; 363 lc->exit_timer = S390_lowcore.exit_timer; 364 lc->user_timer = S390_lowcore.user_timer; 365 lc->system_timer = S390_lowcore.system_timer; 366 lc->steal_timer = S390_lowcore.steal_timer; 367 lc->last_update_timer = S390_lowcore.last_update_timer; 368 lc->last_update_clock = S390_lowcore.last_update_clock; 369 lc->ftrace_func = S390_lowcore.ftrace_func; 370 371 restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0); 372 restart_stack += ASYNC_SIZE; 373 374 /* 375 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 376 * restart data to the absolute zero lowcore. This is necesary if 377 * PSW restart is done on an offline CPU that has lowcore zero. 378 */ 379 lc->restart_stack = (unsigned long) restart_stack; 380 lc->restart_fn = (unsigned long) do_restart; 381 lc->restart_data = 0; 382 lc->restart_source = -1UL; 383 384 /* Setup absolute zero lowcore */ 385 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack); 386 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn); 387 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data); 388 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source); 389 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw); 390 391 set_prefix((u32)(unsigned long) lc); 392 lowcore_ptr[0] = lc; 393 } 394 395 static struct resource code_resource = { 396 .name = "Kernel code", 397 .flags = IORESOURCE_BUSY | IORESOURCE_MEM, 398 }; 399 400 static struct resource data_resource = { 401 .name = "Kernel data", 402 .flags = IORESOURCE_BUSY | IORESOURCE_MEM, 403 }; 404 405 static struct resource bss_resource = { 406 .name = "Kernel bss", 407 .flags = IORESOURCE_BUSY | IORESOURCE_MEM, 408 }; 409 410 static struct resource __initdata *standard_resources[] = { 411 &code_resource, 412 &data_resource, 413 &bss_resource, 414 }; 415 416 static void __init setup_resources(void) 417 { 418 struct resource *res, *std_res, *sub_res; 419 int i, j; 420 421 code_resource.start = (unsigned long) &_text; 422 code_resource.end = (unsigned long) &_etext - 1; 423 data_resource.start = (unsigned long) &_etext; 424 data_resource.end = (unsigned long) &_edata - 1; 425 bss_resource.start = (unsigned long) &__bss_start; 426 bss_resource.end = (unsigned long) &__bss_stop - 1; 427 428 for (i = 0; i < MEMORY_CHUNKS; i++) { 429 if (!memory_chunk[i].size) 430 continue; 431 res = alloc_bootmem_low(sizeof(*res)); 432 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM; 433 switch (memory_chunk[i].type) { 434 case CHUNK_READ_WRITE: 435 res->name = "System RAM"; 436 break; 437 case CHUNK_READ_ONLY: 438 res->name = "System ROM"; 439 res->flags |= IORESOURCE_READONLY; 440 break; 441 default: 442 res->name = "reserved"; 443 } 444 res->start = memory_chunk[i].addr; 445 res->end = res->start + memory_chunk[i].size - 1; 446 request_resource(&iomem_resource, res); 447 448 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 449 std_res = standard_resources[j]; 450 if (std_res->start < res->start || 451 std_res->start > res->end) 452 continue; 453 if (std_res->end > res->end) { 454 sub_res = alloc_bootmem_low(sizeof(*sub_res)); 455 *sub_res = *std_res; 456 sub_res->end = res->end; 457 std_res->start = res->end + 1; 458 request_resource(res, sub_res); 459 } else { 460 request_resource(res, std_res); 461 } 462 } 463 } 464 } 465 466 static void __init setup_memory_end(void) 467 { 468 unsigned long vmax, vmalloc_size, tmp; 469 unsigned long real_memory_size = 0; 470 int i; 471 472 473 #ifdef CONFIG_ZFCPDUMP 474 if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) { 475 memory_end = ZFCPDUMP_HSA_SIZE; 476 memory_end_set = 1; 477 } 478 #endif 479 memory_end &= PAGE_MASK; 480 481 /* 482 * Make sure all chunks are MAX_ORDER aligned so we don't need the 483 * extra checks that HOLES_IN_ZONE would require. 484 */ 485 for (i = 0; i < MEMORY_CHUNKS; i++) { 486 unsigned long start, end; 487 struct mem_chunk *chunk; 488 unsigned long align; 489 490 chunk = &memory_chunk[i]; 491 if (!chunk->size) 492 continue; 493 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1); 494 start = (chunk->addr + align - 1) & ~(align - 1); 495 end = (chunk->addr + chunk->size) & ~(align - 1); 496 if (start >= end) 497 memset(chunk, 0, sizeof(*chunk)); 498 else { 499 chunk->addr = start; 500 chunk->size = end - start; 501 } 502 real_memory_size = max(real_memory_size, 503 chunk->addr + chunk->size); 504 } 505 506 /* Choose kernel address space layout: 2, 3, or 4 levels. */ 507 #ifdef CONFIG_64BIT 508 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN; 509 tmp = (memory_end ?: real_memory_size) / PAGE_SIZE; 510 tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size; 511 if (tmp <= (1UL << 42)) 512 vmax = 1UL << 42; /* 3-level kernel page table */ 513 else 514 vmax = 1UL << 53; /* 4-level kernel page table */ 515 /* module area is at the end of the kernel address space. */ 516 MODULES_END = vmax; 517 MODULES_VADDR = MODULES_END - MODULES_LEN; 518 VMALLOC_END = MODULES_VADDR; 519 #else 520 vmalloc_size = VMALLOC_END ?: 96UL << 20; 521 vmax = 1UL << 31; /* 2-level kernel page table */ 522 /* vmalloc area is at the end of the kernel address space. */ 523 VMALLOC_END = vmax; 524 #endif 525 VMALLOC_START = vmax - vmalloc_size; 526 527 /* Split remaining virtual space between 1:1 mapping & vmemmap array */ 528 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page)); 529 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */ 530 tmp = SECTION_ALIGN_UP(tmp); 531 tmp = VMALLOC_START - tmp * sizeof(struct page); 532 tmp &= ~((vmax >> 11) - 1); /* align to page table level */ 533 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS); 534 vmemmap = (struct page *) tmp; 535 536 /* Take care that memory_end is set and <= vmemmap */ 537 memory_end = min(memory_end ?: real_memory_size, tmp); 538 539 /* Fixup memory chunk array to fit into 0..memory_end */ 540 for (i = 0; i < MEMORY_CHUNKS; i++) { 541 struct mem_chunk *chunk = &memory_chunk[i]; 542 543 if (!chunk->size) 544 continue; 545 if (chunk->addr >= memory_end) { 546 memset(chunk, 0, sizeof(*chunk)); 547 continue; 548 } 549 if (chunk->addr + chunk->size > memory_end) 550 chunk->size = memory_end - chunk->addr; 551 } 552 } 553 554 static void __init setup_vmcoreinfo(void) 555 { 556 mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note()); 557 } 558 559 #ifdef CONFIG_CRASH_DUMP 560 561 /* 562 * Find suitable location for crashkernel memory 563 */ 564 static unsigned long __init find_crash_base(unsigned long crash_size, 565 char **msg) 566 { 567 unsigned long crash_base; 568 struct mem_chunk *chunk; 569 int i; 570 571 if (memory_chunk[0].size < crash_size) { 572 *msg = "first memory chunk must be at least crashkernel size"; 573 return 0; 574 } 575 if (OLDMEM_BASE && crash_size == OLDMEM_SIZE) 576 return OLDMEM_BASE; 577 578 for (i = MEMORY_CHUNKS - 1; i >= 0; i--) { 579 chunk = &memory_chunk[i]; 580 if (chunk->size == 0) 581 continue; 582 if (chunk->type != CHUNK_READ_WRITE) 583 continue; 584 if (chunk->size < crash_size) 585 continue; 586 crash_base = (chunk->addr + chunk->size) - crash_size; 587 if (crash_base < crash_size) 588 continue; 589 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX) 590 continue; 591 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE) 592 continue; 593 return crash_base; 594 } 595 *msg = "no suitable area found"; 596 return 0; 597 } 598 599 /* 600 * Check if crash_base and crash_size is valid 601 */ 602 static int __init verify_crash_base(unsigned long crash_base, 603 unsigned long crash_size, 604 char **msg) 605 { 606 struct mem_chunk *chunk; 607 int i; 608 609 /* 610 * Because we do the swap to zero, we must have at least 'crash_size' 611 * bytes free space before crash_base 612 */ 613 if (crash_size > crash_base) { 614 *msg = "crashkernel offset must be greater than size"; 615 return -EINVAL; 616 } 617 618 /* First memory chunk must be at least crash_size */ 619 if (memory_chunk[0].size < crash_size) { 620 *msg = "first memory chunk must be at least crashkernel size"; 621 return -EINVAL; 622 } 623 /* Check if we fit into the respective memory chunk */ 624 for (i = 0; i < MEMORY_CHUNKS; i++) { 625 chunk = &memory_chunk[i]; 626 if (chunk->size == 0) 627 continue; 628 if (crash_base < chunk->addr) 629 continue; 630 if (crash_base >= chunk->addr + chunk->size) 631 continue; 632 /* we have found the memory chunk */ 633 if (crash_base + crash_size > chunk->addr + chunk->size) { 634 *msg = "selected memory chunk is too small for " 635 "crashkernel memory"; 636 return -EINVAL; 637 } 638 return 0; 639 } 640 *msg = "invalid memory range specified"; 641 return -EINVAL; 642 } 643 644 /* 645 * When kdump is enabled, we have to ensure that no memory from 646 * the area [0 - crashkernel memory size] and 647 * [crashk_res.start - crashk_res.end] is set offline. 648 */ 649 static int kdump_mem_notifier(struct notifier_block *nb, 650 unsigned long action, void *data) 651 { 652 struct memory_notify *arg = data; 653 654 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 655 return NOTIFY_BAD; 656 if (arg->start_pfn > PFN_DOWN(crashk_res.end)) 657 return NOTIFY_OK; 658 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start)) 659 return NOTIFY_OK; 660 return NOTIFY_BAD; 661 } 662 663 static struct notifier_block kdump_mem_nb = { 664 .notifier_call = kdump_mem_notifier, 665 }; 666 667 #endif 668 669 /* 670 * Make sure that oldmem, where the dump is stored, is protected 671 */ 672 static void reserve_oldmem(void) 673 { 674 #ifdef CONFIG_CRASH_DUMP 675 unsigned long real_size = 0; 676 int i; 677 678 if (!OLDMEM_BASE) 679 return; 680 for (i = 0; i < MEMORY_CHUNKS; i++) { 681 struct mem_chunk *chunk = &memory_chunk[i]; 682 683 real_size = max(real_size, chunk->addr + chunk->size); 684 } 685 create_mem_hole(memory_chunk, OLDMEM_BASE, OLDMEM_SIZE); 686 create_mem_hole(memory_chunk, OLDMEM_SIZE, real_size - OLDMEM_SIZE); 687 #endif 688 } 689 690 /* 691 * Reserve memory for kdump kernel to be loaded with kexec 692 */ 693 static void __init reserve_crashkernel(void) 694 { 695 #ifdef CONFIG_CRASH_DUMP 696 unsigned long long crash_base, crash_size; 697 char *msg = NULL; 698 int rc; 699 700 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size, 701 &crash_base); 702 if (rc || crash_size == 0) 703 return; 704 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 705 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 706 if (register_memory_notifier(&kdump_mem_nb)) 707 return; 708 if (!crash_base) 709 crash_base = find_crash_base(crash_size, &msg); 710 if (!crash_base) { 711 pr_info("crashkernel reservation failed: %s\n", msg); 712 unregister_memory_notifier(&kdump_mem_nb); 713 return; 714 } 715 if (verify_crash_base(crash_base, crash_size, &msg)) { 716 pr_info("crashkernel reservation failed: %s\n", msg); 717 unregister_memory_notifier(&kdump_mem_nb); 718 return; 719 } 720 if (!OLDMEM_BASE && MACHINE_IS_VM) 721 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 722 crashk_res.start = crash_base; 723 crashk_res.end = crash_base + crash_size - 1; 724 insert_resource(&iomem_resource, &crashk_res); 725 create_mem_hole(memory_chunk, crash_base, crash_size); 726 pr_info("Reserving %lluMB of memory at %lluMB " 727 "for crashkernel (System RAM: %luMB)\n", 728 crash_size >> 20, crash_base >> 20, memory_end >> 20); 729 os_info_crashkernel_add(crash_base, crash_size); 730 #endif 731 } 732 733 static void __init setup_memory(void) 734 { 735 unsigned long bootmap_size; 736 unsigned long start_pfn, end_pfn; 737 int i; 738 739 /* 740 * partially used pages are not usable - thus 741 * we are rounding upwards: 742 */ 743 start_pfn = PFN_UP(__pa(&_end)); 744 end_pfn = max_pfn = PFN_DOWN(memory_end); 745 746 #ifdef CONFIG_BLK_DEV_INITRD 747 /* 748 * Move the initrd in case the bitmap of the bootmem allocater 749 * would overwrite it. 750 */ 751 752 if (INITRD_START && INITRD_SIZE) { 753 unsigned long bmap_size; 754 unsigned long start; 755 756 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1); 757 bmap_size = PFN_PHYS(bmap_size); 758 759 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) { 760 start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE; 761 762 #ifdef CONFIG_CRASH_DUMP 763 if (OLDMEM_BASE) { 764 /* Move initrd behind kdump oldmem */ 765 if (start + INITRD_SIZE > OLDMEM_BASE && 766 start < OLDMEM_BASE + OLDMEM_SIZE) 767 start = OLDMEM_BASE + OLDMEM_SIZE; 768 } 769 #endif 770 if (start + INITRD_SIZE > memory_end) { 771 pr_err("initrd extends beyond end of " 772 "memory (0x%08lx > 0x%08lx) " 773 "disabling initrd\n", 774 start + INITRD_SIZE, memory_end); 775 INITRD_START = INITRD_SIZE = 0; 776 } else { 777 pr_info("Moving initrd (0x%08lx -> " 778 "0x%08lx, size: %ld)\n", 779 INITRD_START, start, INITRD_SIZE); 780 memmove((void *) start, (void *) INITRD_START, 781 INITRD_SIZE); 782 INITRD_START = start; 783 } 784 } 785 } 786 #endif 787 788 /* 789 * Initialize the boot-time allocator 790 */ 791 bootmap_size = init_bootmem(start_pfn, end_pfn); 792 793 /* 794 * Register RAM areas with the bootmem allocator. 795 */ 796 797 for (i = 0; i < MEMORY_CHUNKS; i++) { 798 unsigned long start_chunk, end_chunk, pfn; 799 800 if (!memory_chunk[i].size) 801 continue; 802 start_chunk = PFN_DOWN(memory_chunk[i].addr); 803 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size); 804 end_chunk = min(end_chunk, end_pfn); 805 if (start_chunk >= end_chunk) 806 continue; 807 memblock_add_node(PFN_PHYS(start_chunk), 808 PFN_PHYS(end_chunk - start_chunk), 0); 809 pfn = max(start_chunk, start_pfn); 810 storage_key_init_range(PFN_PHYS(pfn), PFN_PHYS(end_chunk)); 811 } 812 813 psw_set_key(PAGE_DEFAULT_KEY); 814 815 free_bootmem_with_active_regions(0, max_pfn); 816 817 /* 818 * Reserve memory used for lowcore/command line/kernel image. 819 */ 820 reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT); 821 reserve_bootmem((unsigned long)_stext, 822 PFN_PHYS(start_pfn) - (unsigned long)_stext, 823 BOOTMEM_DEFAULT); 824 /* 825 * Reserve the bootmem bitmap itself as well. We do this in two 826 * steps (first step was init_bootmem()) because this catches 827 * the (very unlikely) case of us accidentally initializing the 828 * bootmem allocator with an invalid RAM area. 829 */ 830 reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size, 831 BOOTMEM_DEFAULT); 832 833 #ifdef CONFIG_CRASH_DUMP 834 if (crashk_res.start) 835 reserve_bootmem(crashk_res.start, 836 crashk_res.end - crashk_res.start + 1, 837 BOOTMEM_DEFAULT); 838 if (is_kdump_kernel()) 839 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE, 840 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT); 841 #endif 842 #ifdef CONFIG_BLK_DEV_INITRD 843 if (INITRD_START && INITRD_SIZE) { 844 if (INITRD_START + INITRD_SIZE <= memory_end) { 845 reserve_bootmem(INITRD_START, INITRD_SIZE, 846 BOOTMEM_DEFAULT); 847 initrd_start = INITRD_START; 848 initrd_end = initrd_start + INITRD_SIZE; 849 } else { 850 pr_err("initrd extends beyond end of " 851 "memory (0x%08lx > 0x%08lx) " 852 "disabling initrd\n", 853 initrd_start + INITRD_SIZE, memory_end); 854 initrd_start = initrd_end = 0; 855 } 856 } 857 #endif 858 } 859 860 /* 861 * Setup hardware capabilities. 862 */ 863 static void __init setup_hwcaps(void) 864 { 865 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 }; 866 struct cpuid cpu_id; 867 int i; 868 869 /* 870 * The store facility list bits numbers as found in the principles 871 * of operation are numbered with bit 1UL<<31 as number 0 to 872 * bit 1UL<<0 as number 31. 873 * Bit 0: instructions named N3, "backported" to esa-mode 874 * Bit 2: z/Architecture mode is active 875 * Bit 7: the store-facility-list-extended facility is installed 876 * Bit 17: the message-security assist is installed 877 * Bit 19: the long-displacement facility is installed 878 * Bit 21: the extended-immediate facility is installed 879 * Bit 22: extended-translation facility 3 is installed 880 * Bit 30: extended-translation facility 3 enhancement facility 881 * These get translated to: 882 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1, 883 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3, 884 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and 885 * HWCAP_S390_ETF3EH bit 8 (22 && 30). 886 */ 887 for (i = 0; i < 6; i++) 888 if (test_facility(stfl_bits[i])) 889 elf_hwcap |= 1UL << i; 890 891 if (test_facility(22) && test_facility(30)) 892 elf_hwcap |= HWCAP_S390_ETF3EH; 893 894 /* 895 * Check for additional facilities with store-facility-list-extended. 896 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0 897 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information 898 * as stored by stfl, bits 32-xxx contain additional facilities. 899 * How many facility words are stored depends on the number of 900 * doublewords passed to the instruction. The additional facilities 901 * are: 902 * Bit 42: decimal floating point facility is installed 903 * Bit 44: perform floating point operation facility is installed 904 * translated to: 905 * HWCAP_S390_DFP bit 6 (42 && 44). 906 */ 907 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44)) 908 elf_hwcap |= HWCAP_S390_DFP; 909 910 /* 911 * Huge page support HWCAP_S390_HPAGE is bit 7. 912 */ 913 if (MACHINE_HAS_HPAGE) 914 elf_hwcap |= HWCAP_S390_HPAGE; 915 916 #if defined(CONFIG_64BIT) 917 /* 918 * 64-bit register support for 31-bit processes 919 * HWCAP_S390_HIGH_GPRS is bit 9. 920 */ 921 elf_hwcap |= HWCAP_S390_HIGH_GPRS; 922 923 /* 924 * Transactional execution support HWCAP_S390_TE is bit 10. 925 */ 926 if (test_facility(50) && test_facility(73)) 927 elf_hwcap |= HWCAP_S390_TE; 928 #endif 929 930 get_cpu_id(&cpu_id); 931 switch (cpu_id.machine) { 932 case 0x9672: 933 #if !defined(CONFIG_64BIT) 934 default: /* Use "g5" as default for 31 bit kernels. */ 935 #endif 936 strcpy(elf_platform, "g5"); 937 break; 938 case 0x2064: 939 case 0x2066: 940 #if defined(CONFIG_64BIT) 941 default: /* Use "z900" as default for 64 bit kernels. */ 942 #endif 943 strcpy(elf_platform, "z900"); 944 break; 945 case 0x2084: 946 case 0x2086: 947 strcpy(elf_platform, "z990"); 948 break; 949 case 0x2094: 950 case 0x2096: 951 strcpy(elf_platform, "z9-109"); 952 break; 953 case 0x2097: 954 case 0x2098: 955 strcpy(elf_platform, "z10"); 956 break; 957 case 0x2817: 958 case 0x2818: 959 strcpy(elf_platform, "z196"); 960 break; 961 case 0x2827: 962 case 0x2828: 963 strcpy(elf_platform, "zEC12"); 964 break; 965 } 966 } 967 968 /* 969 * Setup function called from init/main.c just after the banner 970 * was printed. 971 */ 972 973 void __init setup_arch(char **cmdline_p) 974 { 975 /* 976 * print what head.S has found out about the machine 977 */ 978 #ifndef CONFIG_64BIT 979 if (MACHINE_IS_VM) 980 pr_info("Linux is running as a z/VM " 981 "guest operating system in 31-bit mode\n"); 982 else if (MACHINE_IS_LPAR) 983 pr_info("Linux is running natively in 31-bit mode\n"); 984 if (MACHINE_HAS_IEEE) 985 pr_info("The hardware system has IEEE compatible " 986 "floating point units\n"); 987 else 988 pr_info("The hardware system has no IEEE compatible " 989 "floating point units\n"); 990 #else /* CONFIG_64BIT */ 991 if (MACHINE_IS_VM) 992 pr_info("Linux is running as a z/VM " 993 "guest operating system in 64-bit mode\n"); 994 else if (MACHINE_IS_KVM) 995 pr_info("Linux is running under KVM in 64-bit mode\n"); 996 else if (MACHINE_IS_LPAR) 997 pr_info("Linux is running natively in 64-bit mode\n"); 998 #endif /* CONFIG_64BIT */ 999 1000 /* Have one command line that is parsed and saved in /proc/cmdline */ 1001 /* boot_command_line has been already set up in early.c */ 1002 *cmdline_p = boot_command_line; 1003 1004 ROOT_DEV = Root_RAM0; 1005 1006 init_mm.start_code = PAGE_OFFSET; 1007 init_mm.end_code = (unsigned long) &_etext; 1008 init_mm.end_data = (unsigned long) &_edata; 1009 init_mm.brk = (unsigned long) &_end; 1010 1011 uaccess = MACHINE_HAS_MVCOS ? uaccess_mvcos : uaccess_pt; 1012 1013 parse_early_param(); 1014 detect_memory_layout(memory_chunk, memory_end); 1015 os_info_init(); 1016 setup_ipl(); 1017 reserve_oldmem(); 1018 setup_memory_end(); 1019 reserve_crashkernel(); 1020 setup_memory(); 1021 setup_resources(); 1022 setup_vmcoreinfo(); 1023 setup_lowcore(); 1024 1025 cpu_init(); 1026 s390_init_cpu_topology(); 1027 1028 /* 1029 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM). 1030 */ 1031 setup_hwcaps(); 1032 1033 /* 1034 * Create kernel page tables and switch to virtual addressing. 1035 */ 1036 paging_init(); 1037 1038 /* Setup default console */ 1039 conmode_default(); 1040 set_preferred_console(); 1041 1042 /* Setup zfcpdump support */ 1043 setup_zfcpdump(); 1044 } 1045