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/sched/task.h> 22 #include <linux/cpu.h> 23 #include <linux/kernel.h> 24 #include <linux/memblock.h> 25 #include <linux/mm.h> 26 #include <linux/stddef.h> 27 #include <linux/unistd.h> 28 #include <linux/ptrace.h> 29 #include <linux/random.h> 30 #include <linux/user.h> 31 #include <linux/tty.h> 32 #include <linux/ioport.h> 33 #include <linux/delay.h> 34 #include <linux/init.h> 35 #include <linux/initrd.h> 36 #include <linux/bootmem.h> 37 #include <linux/root_dev.h> 38 #include <linux/console.h> 39 #include <linux/kernel_stat.h> 40 #include <linux/dma-contiguous.h> 41 #include <linux/device.h> 42 #include <linux/notifier.h> 43 #include <linux/pfn.h> 44 #include <linux/ctype.h> 45 #include <linux/reboot.h> 46 #include <linux/topology.h> 47 #include <linux/kexec.h> 48 #include <linux/crash_dump.h> 49 #include <linux/memory.h> 50 #include <linux/compat.h> 51 52 #include <asm/ipl.h> 53 #include <asm/facility.h> 54 #include <asm/smp.h> 55 #include <asm/mmu_context.h> 56 #include <asm/cpcmd.h> 57 #include <asm/lowcore.h> 58 #include <asm/irq.h> 59 #include <asm/page.h> 60 #include <asm/ptrace.h> 61 #include <asm/sections.h> 62 #include <asm/ebcdic.h> 63 #include <asm/kvm_virtio.h> 64 #include <asm/diag.h> 65 #include <asm/os_info.h> 66 #include <asm/sclp.h> 67 #include <asm/sysinfo.h> 68 #include <asm/numa.h> 69 #include "entry.h" 70 71 /* 72 * Machine setup.. 73 */ 74 unsigned int console_mode = 0; 75 EXPORT_SYMBOL(console_mode); 76 77 unsigned int console_devno = -1; 78 EXPORT_SYMBOL(console_devno); 79 80 unsigned int console_irq = -1; 81 EXPORT_SYMBOL(console_irq); 82 83 unsigned long elf_hwcap __read_mostly = 0; 84 char elf_platform[ELF_PLATFORM_SIZE]; 85 86 unsigned long int_hwcap = 0; 87 88 int __initdata memory_end_set; 89 unsigned long __initdata memory_end; 90 unsigned long __initdata max_physmem_end; 91 92 unsigned long VMALLOC_START; 93 EXPORT_SYMBOL(VMALLOC_START); 94 95 unsigned long VMALLOC_END; 96 EXPORT_SYMBOL(VMALLOC_END); 97 98 struct page *vmemmap; 99 EXPORT_SYMBOL(vmemmap); 100 101 unsigned long MODULES_VADDR; 102 unsigned long MODULES_END; 103 104 /* An array with a pointer to the lowcore of every CPU. */ 105 struct lowcore *lowcore_ptr[NR_CPUS]; 106 EXPORT_SYMBOL(lowcore_ptr); 107 108 /* 109 * This is set up by the setup-routine at boot-time 110 * for S390 need to find out, what we have to setup 111 * using address 0x10400 ... 112 */ 113 114 #include <asm/setup.h> 115 116 /* 117 * condev= and conmode= setup parameter. 118 */ 119 120 static int __init condev_setup(char *str) 121 { 122 int vdev; 123 124 vdev = simple_strtoul(str, &str, 0); 125 if (vdev >= 0 && vdev < 65536) { 126 console_devno = vdev; 127 console_irq = -1; 128 } 129 return 1; 130 } 131 132 __setup("condev=", condev_setup); 133 134 static void __init set_preferred_console(void) 135 { 136 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP) 137 add_preferred_console("ttyS", 0, NULL); 138 else if (CONSOLE_IS_3270) 139 add_preferred_console("tty3270", 0, NULL); 140 else if (CONSOLE_IS_VT220) 141 add_preferred_console("ttyS", 1, NULL); 142 else if (CONSOLE_IS_HVC) 143 add_preferred_console("hvc", 0, NULL); 144 } 145 146 static int __init conmode_setup(char *str) 147 { 148 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 149 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0) 150 SET_CONSOLE_SCLP; 151 #endif 152 #if defined(CONFIG_TN3215_CONSOLE) 153 if (strncmp(str, "3215", 5) == 0) 154 SET_CONSOLE_3215; 155 #endif 156 #if defined(CONFIG_TN3270_CONSOLE) 157 if (strncmp(str, "3270", 5) == 0) 158 SET_CONSOLE_3270; 159 #endif 160 set_preferred_console(); 161 return 1; 162 } 163 164 __setup("conmode=", conmode_setup); 165 166 static void __init conmode_default(void) 167 { 168 char query_buffer[1024]; 169 char *ptr; 170 171 if (MACHINE_IS_VM) { 172 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL); 173 console_devno = simple_strtoul(query_buffer + 5, NULL, 16); 174 ptr = strstr(query_buffer, "SUBCHANNEL ="); 175 console_irq = simple_strtoul(ptr + 13, NULL, 16); 176 cpcmd("QUERY TERM", query_buffer, 1024, NULL); 177 ptr = strstr(query_buffer, "CONMODE"); 178 /* 179 * Set the conmode to 3215 so that the device recognition 180 * will set the cu_type of the console to 3215. If the 181 * conmode is 3270 and we don't set it back then both 182 * 3215 and the 3270 driver will try to access the console 183 * device (3215 as console and 3270 as normal tty). 184 */ 185 cpcmd("TERM CONMODE 3215", NULL, 0, NULL); 186 if (ptr == NULL) { 187 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 188 SET_CONSOLE_SCLP; 189 #endif 190 return; 191 } 192 if (strncmp(ptr + 8, "3270", 4) == 0) { 193 #if defined(CONFIG_TN3270_CONSOLE) 194 SET_CONSOLE_3270; 195 #elif defined(CONFIG_TN3215_CONSOLE) 196 SET_CONSOLE_3215; 197 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 198 SET_CONSOLE_SCLP; 199 #endif 200 } else if (strncmp(ptr + 8, "3215", 4) == 0) { 201 #if defined(CONFIG_TN3215_CONSOLE) 202 SET_CONSOLE_3215; 203 #elif defined(CONFIG_TN3270_CONSOLE) 204 SET_CONSOLE_3270; 205 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 206 SET_CONSOLE_SCLP; 207 #endif 208 } 209 } else if (MACHINE_IS_KVM) { 210 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE)) 211 SET_CONSOLE_VT220; 212 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE)) 213 SET_CONSOLE_SCLP; 214 else 215 SET_CONSOLE_HVC; 216 } else { 217 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 218 SET_CONSOLE_SCLP; 219 #endif 220 } 221 } 222 223 #ifdef CONFIG_CRASH_DUMP 224 static void __init setup_zfcpdump(void) 225 { 226 if (ipl_info.type != IPL_TYPE_FCP_DUMP) 227 return; 228 if (OLDMEM_BASE) 229 return; 230 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev"); 231 console_loglevel = 2; 232 } 233 #else 234 static inline void setup_zfcpdump(void) {} 235 #endif /* CONFIG_CRASH_DUMP */ 236 237 /* 238 * Reboot, halt and power_off stubs. They just call _machine_restart, 239 * _machine_halt or _machine_power_off. 240 */ 241 242 void machine_restart(char *command) 243 { 244 if ((!in_interrupt() && !in_atomic()) || oops_in_progress) 245 /* 246 * Only unblank the console if we are called in enabled 247 * context or a bust_spinlocks cleared the way for us. 248 */ 249 console_unblank(); 250 _machine_restart(command); 251 } 252 253 void machine_halt(void) 254 { 255 if (!in_interrupt() || oops_in_progress) 256 /* 257 * Only unblank the console if we are called in enabled 258 * context or a bust_spinlocks cleared the way for us. 259 */ 260 console_unblank(); 261 _machine_halt(); 262 } 263 264 void machine_power_off(void) 265 { 266 if (!in_interrupt() || oops_in_progress) 267 /* 268 * Only unblank the console if we are called in enabled 269 * context or a bust_spinlocks cleared the way for us. 270 */ 271 console_unblank(); 272 _machine_power_off(); 273 } 274 275 /* 276 * Dummy power off function. 277 */ 278 void (*pm_power_off)(void) = machine_power_off; 279 EXPORT_SYMBOL_GPL(pm_power_off); 280 281 static int __init early_parse_mem(char *p) 282 { 283 memory_end = memparse(p, &p); 284 memory_end &= PAGE_MASK; 285 memory_end_set = 1; 286 return 0; 287 } 288 early_param("mem", early_parse_mem); 289 290 static int __init parse_vmalloc(char *arg) 291 { 292 if (!arg) 293 return -EINVAL; 294 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK; 295 return 0; 296 } 297 early_param("vmalloc", parse_vmalloc); 298 299 void *restart_stack __section(.data); 300 301 static void __init setup_lowcore(void) 302 { 303 struct lowcore *lc; 304 305 /* 306 * Setup lowcore for boot cpu 307 */ 308 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * 4096); 309 lc = memblock_virt_alloc_low(sizeof(*lc), sizeof(*lc)); 310 lc->restart_psw.mask = PSW_KERNEL_BITS; 311 lc->restart_psw.addr = (unsigned long) restart_int_handler; 312 lc->external_new_psw.mask = PSW_KERNEL_BITS | 313 PSW_MASK_DAT | PSW_MASK_MCHECK; 314 lc->external_new_psw.addr = (unsigned long) ext_int_handler; 315 lc->svc_new_psw.mask = PSW_KERNEL_BITS | 316 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK; 317 lc->svc_new_psw.addr = (unsigned long) system_call; 318 lc->program_new_psw.mask = PSW_KERNEL_BITS | 319 PSW_MASK_DAT | PSW_MASK_MCHECK; 320 lc->program_new_psw.addr = (unsigned long) pgm_check_handler; 321 lc->mcck_new_psw.mask = PSW_KERNEL_BITS; 322 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler; 323 lc->io_new_psw.mask = PSW_KERNEL_BITS | 324 PSW_MASK_DAT | PSW_MASK_MCHECK; 325 lc->io_new_psw.addr = (unsigned long) io_int_handler; 326 lc->clock_comparator = -1ULL; 327 lc->kernel_stack = ((unsigned long) &init_thread_union) 328 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); 329 lc->async_stack = (unsigned long) 330 memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE) 331 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); 332 lc->panic_stack = (unsigned long) 333 memblock_virt_alloc(PAGE_SIZE, PAGE_SIZE) 334 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); 335 lc->current_task = (unsigned long)&init_task; 336 lc->lpp = LPP_MAGIC; 337 lc->machine_flags = S390_lowcore.machine_flags; 338 lc->preempt_count = S390_lowcore.preempt_count; 339 lc->stfl_fac_list = S390_lowcore.stfl_fac_list; 340 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list, 341 MAX_FACILITY_BIT/8); 342 if (MACHINE_HAS_VX) 343 lc->vector_save_area_addr = 344 (unsigned long) &lc->vector_save_area; 345 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0]; 346 lc->sync_enter_timer = S390_lowcore.sync_enter_timer; 347 lc->async_enter_timer = S390_lowcore.async_enter_timer; 348 lc->exit_timer = S390_lowcore.exit_timer; 349 lc->user_timer = S390_lowcore.user_timer; 350 lc->system_timer = S390_lowcore.system_timer; 351 lc->steal_timer = S390_lowcore.steal_timer; 352 lc->last_update_timer = S390_lowcore.last_update_timer; 353 lc->last_update_clock = S390_lowcore.last_update_clock; 354 355 restart_stack = memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE); 356 restart_stack += ASYNC_SIZE; 357 358 /* 359 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 360 * restart data to the absolute zero lowcore. This is necessary if 361 * PSW restart is done on an offline CPU that has lowcore zero. 362 */ 363 lc->restart_stack = (unsigned long) restart_stack; 364 lc->restart_fn = (unsigned long) do_restart; 365 lc->restart_data = 0; 366 lc->restart_source = -1UL; 367 368 /* Setup absolute zero lowcore */ 369 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack); 370 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn); 371 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data); 372 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source); 373 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw); 374 375 #ifdef CONFIG_SMP 376 lc->spinlock_lockval = arch_spin_lockval(0); 377 #endif 378 379 set_prefix((u32)(unsigned long) lc); 380 lowcore_ptr[0] = lc; 381 } 382 383 static struct resource code_resource = { 384 .name = "Kernel code", 385 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 386 }; 387 388 static struct resource data_resource = { 389 .name = "Kernel data", 390 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 391 }; 392 393 static struct resource bss_resource = { 394 .name = "Kernel bss", 395 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 396 }; 397 398 static struct resource __initdata *standard_resources[] = { 399 &code_resource, 400 &data_resource, 401 &bss_resource, 402 }; 403 404 static void __init setup_resources(void) 405 { 406 struct resource *res, *std_res, *sub_res; 407 struct memblock_region *reg; 408 int j; 409 410 code_resource.start = (unsigned long) &_text; 411 code_resource.end = (unsigned long) &_etext - 1; 412 data_resource.start = (unsigned long) &_etext; 413 data_resource.end = (unsigned long) &_edata - 1; 414 bss_resource.start = (unsigned long) &__bss_start; 415 bss_resource.end = (unsigned long) &__bss_stop - 1; 416 417 for_each_memblock(memory, reg) { 418 res = memblock_virt_alloc(sizeof(*res), 8); 419 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM; 420 421 res->name = "System RAM"; 422 res->start = reg->base; 423 res->end = reg->base + reg->size - 1; 424 request_resource(&iomem_resource, res); 425 426 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 427 std_res = standard_resources[j]; 428 if (std_res->start < res->start || 429 std_res->start > res->end) 430 continue; 431 if (std_res->end > res->end) { 432 sub_res = memblock_virt_alloc(sizeof(*sub_res), 8); 433 *sub_res = *std_res; 434 sub_res->end = res->end; 435 std_res->start = res->end + 1; 436 request_resource(res, sub_res); 437 } else { 438 request_resource(res, std_res); 439 } 440 } 441 } 442 #ifdef CONFIG_CRASH_DUMP 443 /* 444 * Re-add removed crash kernel memory as reserved memory. This makes 445 * sure it will be mapped with the identity mapping and struct pages 446 * will be created, so it can be resized later on. 447 * However add it later since the crash kernel resource should not be 448 * part of the System RAM resource. 449 */ 450 if (crashk_res.end) { 451 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0); 452 memblock_reserve(crashk_res.start, resource_size(&crashk_res)); 453 insert_resource(&iomem_resource, &crashk_res); 454 } 455 #endif 456 } 457 458 static void __init setup_memory_end(void) 459 { 460 unsigned long vmax, vmalloc_size, tmp; 461 462 /* Choose kernel address space layout: 2, 3, or 4 levels. */ 463 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN; 464 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE; 465 tmp = tmp * (sizeof(struct page) + PAGE_SIZE); 466 if (tmp + vmalloc_size + MODULES_LEN <= (1UL << 42)) 467 vmax = 1UL << 42; /* 3-level kernel page table */ 468 else 469 vmax = 1UL << 53; /* 4-level kernel page table */ 470 /* module area is at the end of the kernel address space. */ 471 MODULES_END = vmax; 472 MODULES_VADDR = MODULES_END - MODULES_LEN; 473 VMALLOC_END = MODULES_VADDR; 474 VMALLOC_START = vmax - vmalloc_size; 475 476 /* Split remaining virtual space between 1:1 mapping & vmemmap array */ 477 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page)); 478 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */ 479 tmp = SECTION_ALIGN_UP(tmp); 480 tmp = VMALLOC_START - tmp * sizeof(struct page); 481 tmp &= ~((vmax >> 11) - 1); /* align to page table level */ 482 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS); 483 vmemmap = (struct page *) tmp; 484 485 /* Take care that memory_end is set and <= vmemmap */ 486 memory_end = min(memory_end ?: max_physmem_end, tmp); 487 max_pfn = max_low_pfn = PFN_DOWN(memory_end); 488 memblock_remove(memory_end, ULONG_MAX); 489 490 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20); 491 } 492 493 static void __init setup_vmcoreinfo(void) 494 { 495 mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note()); 496 } 497 498 #ifdef CONFIG_CRASH_DUMP 499 500 /* 501 * When kdump is enabled, we have to ensure that no memory from 502 * the area [0 - crashkernel memory size] and 503 * [crashk_res.start - crashk_res.end] is set offline. 504 */ 505 static int kdump_mem_notifier(struct notifier_block *nb, 506 unsigned long action, void *data) 507 { 508 struct memory_notify *arg = data; 509 510 if (action != MEM_GOING_OFFLINE) 511 return NOTIFY_OK; 512 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 513 return NOTIFY_BAD; 514 if (arg->start_pfn > PFN_DOWN(crashk_res.end)) 515 return NOTIFY_OK; 516 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start)) 517 return NOTIFY_OK; 518 return NOTIFY_BAD; 519 } 520 521 static struct notifier_block kdump_mem_nb = { 522 .notifier_call = kdump_mem_notifier, 523 }; 524 525 #endif 526 527 /* 528 * Make sure that the area behind memory_end is protected 529 */ 530 static void reserve_memory_end(void) 531 { 532 #ifdef CONFIG_CRASH_DUMP 533 if (ipl_info.type == IPL_TYPE_FCP_DUMP && 534 !OLDMEM_BASE && sclp.hsa_size) { 535 memory_end = sclp.hsa_size; 536 memory_end &= PAGE_MASK; 537 memory_end_set = 1; 538 } 539 #endif 540 if (!memory_end_set) 541 return; 542 memblock_reserve(memory_end, ULONG_MAX); 543 } 544 545 /* 546 * Make sure that oldmem, where the dump is stored, is protected 547 */ 548 static void reserve_oldmem(void) 549 { 550 #ifdef CONFIG_CRASH_DUMP 551 if (OLDMEM_BASE) 552 /* Forget all memory above the running kdump system */ 553 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 554 #endif 555 } 556 557 /* 558 * Make sure that oldmem, where the dump is stored, is protected 559 */ 560 static void remove_oldmem(void) 561 { 562 #ifdef CONFIG_CRASH_DUMP 563 if (OLDMEM_BASE) 564 /* Forget all memory above the running kdump system */ 565 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 566 #endif 567 } 568 569 /* 570 * Reserve memory for kdump kernel to be loaded with kexec 571 */ 572 static void __init reserve_crashkernel(void) 573 { 574 #ifdef CONFIG_CRASH_DUMP 575 unsigned long long crash_base, crash_size; 576 phys_addr_t low, high; 577 int rc; 578 579 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size, 580 &crash_base); 581 582 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 583 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 584 if (rc || crash_size == 0) 585 return; 586 587 if (memblock.memory.regions[0].size < crash_size) { 588 pr_info("crashkernel reservation failed: %s\n", 589 "first memory chunk must be at least crashkernel size"); 590 return; 591 } 592 593 low = crash_base ?: OLDMEM_BASE; 594 high = low + crash_size; 595 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) { 596 /* The crashkernel fits into OLDMEM, reuse OLDMEM */ 597 crash_base = low; 598 } else { 599 /* Find suitable area in free memory */ 600 low = max_t(unsigned long, crash_size, sclp.hsa_size); 601 high = crash_base ? crash_base + crash_size : ULONG_MAX; 602 603 if (crash_base && crash_base < low) { 604 pr_info("crashkernel reservation failed: %s\n", 605 "crash_base too low"); 606 return; 607 } 608 low = crash_base ?: low; 609 crash_base = memblock_find_in_range(low, high, crash_size, 610 KEXEC_CRASH_MEM_ALIGN); 611 } 612 613 if (!crash_base) { 614 pr_info("crashkernel reservation failed: %s\n", 615 "no suitable area found"); 616 return; 617 } 618 619 if (register_memory_notifier(&kdump_mem_nb)) 620 return; 621 622 if (!OLDMEM_BASE && MACHINE_IS_VM) 623 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 624 crashk_res.start = crash_base; 625 crashk_res.end = crash_base + crash_size - 1; 626 memblock_remove(crash_base, crash_size); 627 pr_info("Reserving %lluMB of memory at %lluMB " 628 "for crashkernel (System RAM: %luMB)\n", 629 crash_size >> 20, crash_base >> 20, 630 (unsigned long)memblock.memory.total_size >> 20); 631 os_info_crashkernel_add(crash_base, crash_size); 632 #endif 633 } 634 635 /* 636 * Reserve the initrd from being used by memblock 637 */ 638 static void __init reserve_initrd(void) 639 { 640 #ifdef CONFIG_BLK_DEV_INITRD 641 if (!INITRD_START || !INITRD_SIZE) 642 return; 643 initrd_start = INITRD_START; 644 initrd_end = initrd_start + INITRD_SIZE; 645 memblock_reserve(INITRD_START, INITRD_SIZE); 646 #endif 647 } 648 649 /* 650 * Check for initrd being in usable memory 651 */ 652 static void __init check_initrd(void) 653 { 654 #ifdef CONFIG_BLK_DEV_INITRD 655 if (INITRD_START && INITRD_SIZE && 656 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) { 657 pr_err("The initial RAM disk does not fit into the memory\n"); 658 memblock_free(INITRD_START, INITRD_SIZE); 659 initrd_start = initrd_end = 0; 660 } 661 #endif 662 } 663 664 /* 665 * Reserve memory used for lowcore/command line/kernel image. 666 */ 667 static void __init reserve_kernel(void) 668 { 669 unsigned long start_pfn = PFN_UP(__pa(&_end)); 670 671 #ifdef CONFIG_DMA_API_DEBUG 672 /* 673 * DMA_API_DEBUG code stumbles over addresses from the 674 * range [_ehead, _stext]. Mark the memory as reserved 675 * so it is not used for CONFIG_DMA_API_DEBUG=y. 676 */ 677 memblock_reserve(0, PFN_PHYS(start_pfn)); 678 #else 679 memblock_reserve(0, (unsigned long)_ehead); 680 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn) 681 - (unsigned long)_stext); 682 #endif 683 } 684 685 static void __init setup_memory(void) 686 { 687 struct memblock_region *reg; 688 689 /* 690 * Init storage key for present memory 691 */ 692 for_each_memblock(memory, reg) { 693 storage_key_init_range(reg->base, reg->base + reg->size); 694 } 695 psw_set_key(PAGE_DEFAULT_KEY); 696 697 /* Only cosmetics */ 698 memblock_enforce_memory_limit(memblock_end_of_DRAM()); 699 } 700 701 /* 702 * Setup hardware capabilities. 703 */ 704 static int __init setup_hwcaps(void) 705 { 706 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 }; 707 struct cpuid cpu_id; 708 int i; 709 710 /* 711 * The store facility list bits numbers as found in the principles 712 * of operation are numbered with bit 1UL<<31 as number 0 to 713 * bit 1UL<<0 as number 31. 714 * Bit 0: instructions named N3, "backported" to esa-mode 715 * Bit 2: z/Architecture mode is active 716 * Bit 7: the store-facility-list-extended facility is installed 717 * Bit 17: the message-security assist is installed 718 * Bit 19: the long-displacement facility is installed 719 * Bit 21: the extended-immediate facility is installed 720 * Bit 22: extended-translation facility 3 is installed 721 * Bit 30: extended-translation facility 3 enhancement facility 722 * These get translated to: 723 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1, 724 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3, 725 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and 726 * HWCAP_S390_ETF3EH bit 8 (22 && 30). 727 */ 728 for (i = 0; i < 6; i++) 729 if (test_facility(stfl_bits[i])) 730 elf_hwcap |= 1UL << i; 731 732 if (test_facility(22) && test_facility(30)) 733 elf_hwcap |= HWCAP_S390_ETF3EH; 734 735 /* 736 * Check for additional facilities with store-facility-list-extended. 737 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0 738 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information 739 * as stored by stfl, bits 32-xxx contain additional facilities. 740 * How many facility words are stored depends on the number of 741 * doublewords passed to the instruction. The additional facilities 742 * are: 743 * Bit 42: decimal floating point facility is installed 744 * Bit 44: perform floating point operation facility is installed 745 * translated to: 746 * HWCAP_S390_DFP bit 6 (42 && 44). 747 */ 748 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44)) 749 elf_hwcap |= HWCAP_S390_DFP; 750 751 /* 752 * Huge page support HWCAP_S390_HPAGE is bit 7. 753 */ 754 if (MACHINE_HAS_EDAT1) 755 elf_hwcap |= HWCAP_S390_HPAGE; 756 757 /* 758 * 64-bit register support for 31-bit processes 759 * HWCAP_S390_HIGH_GPRS is bit 9. 760 */ 761 elf_hwcap |= HWCAP_S390_HIGH_GPRS; 762 763 /* 764 * Transactional execution support HWCAP_S390_TE is bit 10. 765 */ 766 if (test_facility(50) && test_facility(73)) 767 elf_hwcap |= HWCAP_S390_TE; 768 769 /* 770 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension 771 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX 772 * instead of facility bit 129. 773 */ 774 if (MACHINE_HAS_VX) { 775 elf_hwcap |= HWCAP_S390_VXRS; 776 if (test_facility(134)) 777 elf_hwcap |= HWCAP_S390_VXRS_EXT; 778 if (test_facility(135)) 779 elf_hwcap |= HWCAP_S390_VXRS_BCD; 780 } 781 782 get_cpu_id(&cpu_id); 783 add_device_randomness(&cpu_id, sizeof(cpu_id)); 784 switch (cpu_id.machine) { 785 case 0x2064: 786 case 0x2066: 787 default: /* Use "z900" as default for 64 bit kernels. */ 788 strcpy(elf_platform, "z900"); 789 break; 790 case 0x2084: 791 case 0x2086: 792 strcpy(elf_platform, "z990"); 793 break; 794 case 0x2094: 795 case 0x2096: 796 strcpy(elf_platform, "z9-109"); 797 break; 798 case 0x2097: 799 case 0x2098: 800 strcpy(elf_platform, "z10"); 801 break; 802 case 0x2817: 803 case 0x2818: 804 strcpy(elf_platform, "z196"); 805 break; 806 case 0x2827: 807 case 0x2828: 808 strcpy(elf_platform, "zEC12"); 809 break; 810 case 0x2964: 811 case 0x2965: 812 strcpy(elf_platform, "z13"); 813 break; 814 } 815 816 /* 817 * Virtualization support HWCAP_INT_SIE is bit 0. 818 */ 819 if (sclp.has_sief2) 820 int_hwcap |= HWCAP_INT_SIE; 821 822 return 0; 823 } 824 arch_initcall(setup_hwcaps); 825 826 /* 827 * Add system information as device randomness 828 */ 829 static void __init setup_randomness(void) 830 { 831 struct sysinfo_3_2_2 *vmms; 832 833 vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE); 834 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count) 835 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count); 836 memblock_free((unsigned long) vmms, PAGE_SIZE); 837 } 838 839 /* 840 * Find the correct size for the task_struct. This depends on 841 * the size of the struct fpu at the end of the thread_struct 842 * which is embedded in the task_struct. 843 */ 844 static void __init setup_task_size(void) 845 { 846 int task_size = sizeof(struct task_struct); 847 848 if (!MACHINE_HAS_VX) { 849 task_size -= sizeof(__vector128) * __NUM_VXRS; 850 task_size += sizeof(freg_t) * __NUM_FPRS; 851 } 852 arch_task_struct_size = task_size; 853 } 854 855 /* 856 * Setup function called from init/main.c just after the banner 857 * was printed. 858 */ 859 860 void __init setup_arch(char **cmdline_p) 861 { 862 /* 863 * print what head.S has found out about the machine 864 */ 865 if (MACHINE_IS_VM) 866 pr_info("Linux is running as a z/VM " 867 "guest operating system in 64-bit mode\n"); 868 else if (MACHINE_IS_KVM) 869 pr_info("Linux is running under KVM in 64-bit mode\n"); 870 else if (MACHINE_IS_LPAR) 871 pr_info("Linux is running natively in 64-bit mode\n"); 872 873 /* Have one command line that is parsed and saved in /proc/cmdline */ 874 /* boot_command_line has been already set up in early.c */ 875 *cmdline_p = boot_command_line; 876 877 ROOT_DEV = Root_RAM0; 878 879 /* Is init_mm really needed? */ 880 init_mm.start_code = PAGE_OFFSET; 881 init_mm.end_code = (unsigned long) &_etext; 882 init_mm.end_data = (unsigned long) &_edata; 883 init_mm.brk = (unsigned long) &_end; 884 885 parse_early_param(); 886 #ifdef CONFIG_CRASH_DUMP 887 /* Deactivate elfcorehdr= kernel parameter */ 888 elfcorehdr_addr = ELFCORE_ADDR_MAX; 889 #endif 890 891 os_info_init(); 892 setup_ipl(); 893 setup_task_size(); 894 895 /* Do some memory reservations *before* memory is added to memblock */ 896 reserve_memory_end(); 897 reserve_oldmem(); 898 reserve_kernel(); 899 reserve_initrd(); 900 memblock_allow_resize(); 901 902 /* Get information about *all* installed memory */ 903 detect_memory_memblock(); 904 905 remove_oldmem(); 906 907 /* 908 * Make sure all chunks are MAX_ORDER aligned so we don't need the 909 * extra checks that HOLES_IN_ZONE would require. 910 * 911 * Is this still required? 912 */ 913 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT)); 914 915 setup_memory_end(); 916 setup_memory(); 917 dma_contiguous_reserve(memory_end); 918 919 check_initrd(); 920 reserve_crashkernel(); 921 #ifdef CONFIG_CRASH_DUMP 922 /* 923 * Be aware that smp_save_dump_cpus() triggers a system reset. 924 * Therefore CPU and device initialization should be done afterwards. 925 */ 926 smp_save_dump_cpus(); 927 #endif 928 929 setup_resources(); 930 setup_vmcoreinfo(); 931 setup_lowcore(); 932 smp_fill_possible_mask(); 933 cpu_detect_mhz_feature(); 934 cpu_init(); 935 numa_setup(); 936 smp_detect_cpus(); 937 topology_init_early(); 938 939 /* 940 * Create kernel page tables and switch to virtual addressing. 941 */ 942 paging_init(); 943 944 /* Setup default console */ 945 conmode_default(); 946 set_preferred_console(); 947 948 /* Setup zfcpdump support */ 949 setup_zfcpdump(); 950 951 /* Add system specific data to the random pool */ 952 setup_randomness(); 953 } 954