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 || MACHINE_HAS_GS) { 343 unsigned long bits, size; 344 345 bits = MACHINE_HAS_GS ? 11 : 10; 346 size = 1UL << bits; 347 lc->mcesad = (__u64) memblock_virt_alloc(size, size); 348 if (MACHINE_HAS_GS) 349 lc->mcesad |= bits; 350 } 351 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0]; 352 lc->sync_enter_timer = S390_lowcore.sync_enter_timer; 353 lc->async_enter_timer = S390_lowcore.async_enter_timer; 354 lc->exit_timer = S390_lowcore.exit_timer; 355 lc->user_timer = S390_lowcore.user_timer; 356 lc->system_timer = S390_lowcore.system_timer; 357 lc->steal_timer = S390_lowcore.steal_timer; 358 lc->last_update_timer = S390_lowcore.last_update_timer; 359 lc->last_update_clock = S390_lowcore.last_update_clock; 360 361 restart_stack = memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE); 362 restart_stack += ASYNC_SIZE; 363 364 /* 365 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 366 * restart data to the absolute zero lowcore. This is necessary if 367 * PSW restart is done on an offline CPU that has lowcore zero. 368 */ 369 lc->restart_stack = (unsigned long) restart_stack; 370 lc->restart_fn = (unsigned long) do_restart; 371 lc->restart_data = 0; 372 lc->restart_source = -1UL; 373 374 /* Setup absolute zero lowcore */ 375 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack); 376 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn); 377 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data); 378 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source); 379 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw); 380 381 #ifdef CONFIG_SMP 382 lc->spinlock_lockval = arch_spin_lockval(0); 383 #endif 384 385 set_prefix((u32)(unsigned long) lc); 386 lowcore_ptr[0] = lc; 387 } 388 389 static struct resource code_resource = { 390 .name = "Kernel code", 391 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 392 }; 393 394 static struct resource data_resource = { 395 .name = "Kernel data", 396 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 397 }; 398 399 static struct resource bss_resource = { 400 .name = "Kernel bss", 401 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 402 }; 403 404 static struct resource __initdata *standard_resources[] = { 405 &code_resource, 406 &data_resource, 407 &bss_resource, 408 }; 409 410 static void __init setup_resources(void) 411 { 412 struct resource *res, *std_res, *sub_res; 413 struct memblock_region *reg; 414 int j; 415 416 code_resource.start = (unsigned long) &_text; 417 code_resource.end = (unsigned long) &_etext - 1; 418 data_resource.start = (unsigned long) &_etext; 419 data_resource.end = (unsigned long) &_edata - 1; 420 bss_resource.start = (unsigned long) &__bss_start; 421 bss_resource.end = (unsigned long) &__bss_stop - 1; 422 423 for_each_memblock(memory, reg) { 424 res = memblock_virt_alloc(sizeof(*res), 8); 425 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM; 426 427 res->name = "System RAM"; 428 res->start = reg->base; 429 res->end = reg->base + reg->size - 1; 430 request_resource(&iomem_resource, res); 431 432 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 433 std_res = standard_resources[j]; 434 if (std_res->start < res->start || 435 std_res->start > res->end) 436 continue; 437 if (std_res->end > res->end) { 438 sub_res = memblock_virt_alloc(sizeof(*sub_res), 8); 439 *sub_res = *std_res; 440 sub_res->end = res->end; 441 std_res->start = res->end + 1; 442 request_resource(res, sub_res); 443 } else { 444 request_resource(res, std_res); 445 } 446 } 447 } 448 #ifdef CONFIG_CRASH_DUMP 449 /* 450 * Re-add removed crash kernel memory as reserved memory. This makes 451 * sure it will be mapped with the identity mapping and struct pages 452 * will be created, so it can be resized later on. 453 * However add it later since the crash kernel resource should not be 454 * part of the System RAM resource. 455 */ 456 if (crashk_res.end) { 457 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0); 458 memblock_reserve(crashk_res.start, resource_size(&crashk_res)); 459 insert_resource(&iomem_resource, &crashk_res); 460 } 461 #endif 462 } 463 464 static void __init setup_memory_end(void) 465 { 466 unsigned long vmax, vmalloc_size, tmp; 467 468 /* Choose kernel address space layout: 2, 3, or 4 levels. */ 469 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN; 470 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE; 471 tmp = tmp * (sizeof(struct page) + PAGE_SIZE); 472 if (tmp + vmalloc_size + MODULES_LEN <= (1UL << 42)) 473 vmax = 1UL << 42; /* 3-level kernel page table */ 474 else 475 vmax = 1UL << 53; /* 4-level kernel page table */ 476 /* module area is at the end of the kernel address space. */ 477 MODULES_END = vmax; 478 MODULES_VADDR = MODULES_END - MODULES_LEN; 479 VMALLOC_END = MODULES_VADDR; 480 VMALLOC_START = vmax - vmalloc_size; 481 482 /* Split remaining virtual space between 1:1 mapping & vmemmap array */ 483 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page)); 484 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */ 485 tmp = SECTION_ALIGN_UP(tmp); 486 tmp = VMALLOC_START - tmp * sizeof(struct page); 487 tmp &= ~((vmax >> 11) - 1); /* align to page table level */ 488 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS); 489 vmemmap = (struct page *) tmp; 490 491 /* Take care that memory_end is set and <= vmemmap */ 492 memory_end = min(memory_end ?: max_physmem_end, tmp); 493 max_pfn = max_low_pfn = PFN_DOWN(memory_end); 494 memblock_remove(memory_end, ULONG_MAX); 495 496 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20); 497 } 498 499 static void __init setup_vmcoreinfo(void) 500 { 501 mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note()); 502 } 503 504 #ifdef CONFIG_CRASH_DUMP 505 506 /* 507 * When kdump is enabled, we have to ensure that no memory from 508 * the area [0 - crashkernel memory size] and 509 * [crashk_res.start - crashk_res.end] is set offline. 510 */ 511 static int kdump_mem_notifier(struct notifier_block *nb, 512 unsigned long action, void *data) 513 { 514 struct memory_notify *arg = data; 515 516 if (action != MEM_GOING_OFFLINE) 517 return NOTIFY_OK; 518 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 519 return NOTIFY_BAD; 520 if (arg->start_pfn > PFN_DOWN(crashk_res.end)) 521 return NOTIFY_OK; 522 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start)) 523 return NOTIFY_OK; 524 return NOTIFY_BAD; 525 } 526 527 static struct notifier_block kdump_mem_nb = { 528 .notifier_call = kdump_mem_notifier, 529 }; 530 531 #endif 532 533 /* 534 * Make sure that the area behind memory_end is protected 535 */ 536 static void reserve_memory_end(void) 537 { 538 #ifdef CONFIG_CRASH_DUMP 539 if (ipl_info.type == IPL_TYPE_FCP_DUMP && 540 !OLDMEM_BASE && sclp.hsa_size) { 541 memory_end = sclp.hsa_size; 542 memory_end &= PAGE_MASK; 543 memory_end_set = 1; 544 } 545 #endif 546 if (!memory_end_set) 547 return; 548 memblock_reserve(memory_end, ULONG_MAX); 549 } 550 551 /* 552 * Make sure that oldmem, where the dump is stored, is protected 553 */ 554 static void reserve_oldmem(void) 555 { 556 #ifdef CONFIG_CRASH_DUMP 557 if (OLDMEM_BASE) 558 /* Forget all memory above the running kdump system */ 559 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 560 #endif 561 } 562 563 /* 564 * Make sure that oldmem, where the dump is stored, is protected 565 */ 566 static void remove_oldmem(void) 567 { 568 #ifdef CONFIG_CRASH_DUMP 569 if (OLDMEM_BASE) 570 /* Forget all memory above the running kdump system */ 571 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 572 #endif 573 } 574 575 /* 576 * Reserve memory for kdump kernel to be loaded with kexec 577 */ 578 static void __init reserve_crashkernel(void) 579 { 580 #ifdef CONFIG_CRASH_DUMP 581 unsigned long long crash_base, crash_size; 582 phys_addr_t low, high; 583 int rc; 584 585 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size, 586 &crash_base); 587 588 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 589 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 590 if (rc || crash_size == 0) 591 return; 592 593 if (memblock.memory.regions[0].size < crash_size) { 594 pr_info("crashkernel reservation failed: %s\n", 595 "first memory chunk must be at least crashkernel size"); 596 return; 597 } 598 599 low = crash_base ?: OLDMEM_BASE; 600 high = low + crash_size; 601 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) { 602 /* The crashkernel fits into OLDMEM, reuse OLDMEM */ 603 crash_base = low; 604 } else { 605 /* Find suitable area in free memory */ 606 low = max_t(unsigned long, crash_size, sclp.hsa_size); 607 high = crash_base ? crash_base + crash_size : ULONG_MAX; 608 609 if (crash_base && crash_base < low) { 610 pr_info("crashkernel reservation failed: %s\n", 611 "crash_base too low"); 612 return; 613 } 614 low = crash_base ?: low; 615 crash_base = memblock_find_in_range(low, high, crash_size, 616 KEXEC_CRASH_MEM_ALIGN); 617 } 618 619 if (!crash_base) { 620 pr_info("crashkernel reservation failed: %s\n", 621 "no suitable area found"); 622 return; 623 } 624 625 if (register_memory_notifier(&kdump_mem_nb)) 626 return; 627 628 if (!OLDMEM_BASE && MACHINE_IS_VM) 629 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 630 crashk_res.start = crash_base; 631 crashk_res.end = crash_base + crash_size - 1; 632 memblock_remove(crash_base, crash_size); 633 pr_info("Reserving %lluMB of memory at %lluMB " 634 "for crashkernel (System RAM: %luMB)\n", 635 crash_size >> 20, crash_base >> 20, 636 (unsigned long)memblock.memory.total_size >> 20); 637 os_info_crashkernel_add(crash_base, crash_size); 638 #endif 639 } 640 641 /* 642 * Reserve the initrd from being used by memblock 643 */ 644 static void __init reserve_initrd(void) 645 { 646 #ifdef CONFIG_BLK_DEV_INITRD 647 if (!INITRD_START || !INITRD_SIZE) 648 return; 649 initrd_start = INITRD_START; 650 initrd_end = initrd_start + INITRD_SIZE; 651 memblock_reserve(INITRD_START, INITRD_SIZE); 652 #endif 653 } 654 655 /* 656 * Check for initrd being in usable memory 657 */ 658 static void __init check_initrd(void) 659 { 660 #ifdef CONFIG_BLK_DEV_INITRD 661 if (INITRD_START && INITRD_SIZE && 662 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) { 663 pr_err("The initial RAM disk does not fit into the memory\n"); 664 memblock_free(INITRD_START, INITRD_SIZE); 665 initrd_start = initrd_end = 0; 666 } 667 #endif 668 } 669 670 /* 671 * Reserve memory used for lowcore/command line/kernel image. 672 */ 673 static void __init reserve_kernel(void) 674 { 675 unsigned long start_pfn = PFN_UP(__pa(&_end)); 676 677 #ifdef CONFIG_DMA_API_DEBUG 678 /* 679 * DMA_API_DEBUG code stumbles over addresses from the 680 * range [_ehead, _stext]. Mark the memory as reserved 681 * so it is not used for CONFIG_DMA_API_DEBUG=y. 682 */ 683 memblock_reserve(0, PFN_PHYS(start_pfn)); 684 #else 685 memblock_reserve(0, (unsigned long)_ehead); 686 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn) 687 - (unsigned long)_stext); 688 #endif 689 } 690 691 static void __init setup_memory(void) 692 { 693 struct memblock_region *reg; 694 695 /* 696 * Init storage key for present memory 697 */ 698 for_each_memblock(memory, reg) { 699 storage_key_init_range(reg->base, reg->base + reg->size); 700 } 701 psw_set_key(PAGE_DEFAULT_KEY); 702 703 /* Only cosmetics */ 704 memblock_enforce_memory_limit(memblock_end_of_DRAM()); 705 } 706 707 /* 708 * Setup hardware capabilities. 709 */ 710 static int __init setup_hwcaps(void) 711 { 712 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 }; 713 struct cpuid cpu_id; 714 int i; 715 716 /* 717 * The store facility list bits numbers as found in the principles 718 * of operation are numbered with bit 1UL<<31 as number 0 to 719 * bit 1UL<<0 as number 31. 720 * Bit 0: instructions named N3, "backported" to esa-mode 721 * Bit 2: z/Architecture mode is active 722 * Bit 7: the store-facility-list-extended facility is installed 723 * Bit 17: the message-security assist is installed 724 * Bit 19: the long-displacement facility is installed 725 * Bit 21: the extended-immediate facility is installed 726 * Bit 22: extended-translation facility 3 is installed 727 * Bit 30: extended-translation facility 3 enhancement facility 728 * These get translated to: 729 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1, 730 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3, 731 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and 732 * HWCAP_S390_ETF3EH bit 8 (22 && 30). 733 */ 734 for (i = 0; i < 6; i++) 735 if (test_facility(stfl_bits[i])) 736 elf_hwcap |= 1UL << i; 737 738 if (test_facility(22) && test_facility(30)) 739 elf_hwcap |= HWCAP_S390_ETF3EH; 740 741 /* 742 * Check for additional facilities with store-facility-list-extended. 743 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0 744 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information 745 * as stored by stfl, bits 32-xxx contain additional facilities. 746 * How many facility words are stored depends on the number of 747 * doublewords passed to the instruction. The additional facilities 748 * are: 749 * Bit 42: decimal floating point facility is installed 750 * Bit 44: perform floating point operation facility is installed 751 * translated to: 752 * HWCAP_S390_DFP bit 6 (42 && 44). 753 */ 754 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44)) 755 elf_hwcap |= HWCAP_S390_DFP; 756 757 /* 758 * Huge page support HWCAP_S390_HPAGE is bit 7. 759 */ 760 if (MACHINE_HAS_EDAT1) 761 elf_hwcap |= HWCAP_S390_HPAGE; 762 763 /* 764 * 64-bit register support for 31-bit processes 765 * HWCAP_S390_HIGH_GPRS is bit 9. 766 */ 767 elf_hwcap |= HWCAP_S390_HIGH_GPRS; 768 769 /* 770 * Transactional execution support HWCAP_S390_TE is bit 10. 771 */ 772 if (test_facility(50) && test_facility(73)) 773 elf_hwcap |= HWCAP_S390_TE; 774 775 /* 776 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension 777 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX 778 * instead of facility bit 129. 779 */ 780 if (MACHINE_HAS_VX) { 781 elf_hwcap |= HWCAP_S390_VXRS; 782 if (test_facility(134)) 783 elf_hwcap |= HWCAP_S390_VXRS_EXT; 784 if (test_facility(135)) 785 elf_hwcap |= HWCAP_S390_VXRS_BCD; 786 } 787 788 /* 789 * Guarded storage support HWCAP_S390_GS is bit 12. 790 */ 791 if (MACHINE_HAS_GS) 792 elf_hwcap |= HWCAP_S390_GS; 793 794 get_cpu_id(&cpu_id); 795 add_device_randomness(&cpu_id, sizeof(cpu_id)); 796 switch (cpu_id.machine) { 797 case 0x2064: 798 case 0x2066: 799 default: /* Use "z900" as default for 64 bit kernels. */ 800 strcpy(elf_platform, "z900"); 801 break; 802 case 0x2084: 803 case 0x2086: 804 strcpy(elf_platform, "z990"); 805 break; 806 case 0x2094: 807 case 0x2096: 808 strcpy(elf_platform, "z9-109"); 809 break; 810 case 0x2097: 811 case 0x2098: 812 strcpy(elf_platform, "z10"); 813 break; 814 case 0x2817: 815 case 0x2818: 816 strcpy(elf_platform, "z196"); 817 break; 818 case 0x2827: 819 case 0x2828: 820 strcpy(elf_platform, "zEC12"); 821 break; 822 case 0x2964: 823 case 0x2965: 824 strcpy(elf_platform, "z13"); 825 break; 826 } 827 828 /* 829 * Virtualization support HWCAP_INT_SIE is bit 0. 830 */ 831 if (sclp.has_sief2) 832 int_hwcap |= HWCAP_INT_SIE; 833 834 return 0; 835 } 836 arch_initcall(setup_hwcaps); 837 838 /* 839 * Add system information as device randomness 840 */ 841 static void __init setup_randomness(void) 842 { 843 struct sysinfo_3_2_2 *vmms; 844 845 vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE); 846 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count) 847 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count); 848 memblock_free((unsigned long) vmms, PAGE_SIZE); 849 } 850 851 /* 852 * Find the correct size for the task_struct. This depends on 853 * the size of the struct fpu at the end of the thread_struct 854 * which is embedded in the task_struct. 855 */ 856 static void __init setup_task_size(void) 857 { 858 int task_size = sizeof(struct task_struct); 859 860 if (!MACHINE_HAS_VX) { 861 task_size -= sizeof(__vector128) * __NUM_VXRS; 862 task_size += sizeof(freg_t) * __NUM_FPRS; 863 } 864 arch_task_struct_size = task_size; 865 } 866 867 /* 868 * Setup function called from init/main.c just after the banner 869 * was printed. 870 */ 871 872 void __init setup_arch(char **cmdline_p) 873 { 874 /* 875 * print what head.S has found out about the machine 876 */ 877 if (MACHINE_IS_VM) 878 pr_info("Linux is running as a z/VM " 879 "guest operating system in 64-bit mode\n"); 880 else if (MACHINE_IS_KVM) 881 pr_info("Linux is running under KVM in 64-bit mode\n"); 882 else if (MACHINE_IS_LPAR) 883 pr_info("Linux is running natively in 64-bit mode\n"); 884 885 /* Have one command line that is parsed and saved in /proc/cmdline */ 886 /* boot_command_line has been already set up in early.c */ 887 *cmdline_p = boot_command_line; 888 889 ROOT_DEV = Root_RAM0; 890 891 /* Is init_mm really needed? */ 892 init_mm.start_code = PAGE_OFFSET; 893 init_mm.end_code = (unsigned long) &_etext; 894 init_mm.end_data = (unsigned long) &_edata; 895 init_mm.brk = (unsigned long) &_end; 896 897 parse_early_param(); 898 #ifdef CONFIG_CRASH_DUMP 899 /* Deactivate elfcorehdr= kernel parameter */ 900 elfcorehdr_addr = ELFCORE_ADDR_MAX; 901 #endif 902 903 os_info_init(); 904 setup_ipl(); 905 setup_task_size(); 906 907 /* Do some memory reservations *before* memory is added to memblock */ 908 reserve_memory_end(); 909 reserve_oldmem(); 910 reserve_kernel(); 911 reserve_initrd(); 912 memblock_allow_resize(); 913 914 /* Get information about *all* installed memory */ 915 detect_memory_memblock(); 916 917 remove_oldmem(); 918 919 /* 920 * Make sure all chunks are MAX_ORDER aligned so we don't need the 921 * extra checks that HOLES_IN_ZONE would require. 922 * 923 * Is this still required? 924 */ 925 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT)); 926 927 setup_memory_end(); 928 setup_memory(); 929 dma_contiguous_reserve(memory_end); 930 931 check_initrd(); 932 reserve_crashkernel(); 933 #ifdef CONFIG_CRASH_DUMP 934 /* 935 * Be aware that smp_save_dump_cpus() triggers a system reset. 936 * Therefore CPU and device initialization should be done afterwards. 937 */ 938 smp_save_dump_cpus(); 939 #endif 940 941 setup_resources(); 942 setup_vmcoreinfo(); 943 setup_lowcore(); 944 smp_fill_possible_mask(); 945 cpu_detect_mhz_feature(); 946 cpu_init(); 947 numa_setup(); 948 smp_detect_cpus(); 949 topology_init_early(); 950 951 /* 952 * Create kernel page tables and switch to virtual addressing. 953 */ 954 paging_init(); 955 956 /* Setup default console */ 957 conmode_default(); 958 set_preferred_console(); 959 960 /* Setup zfcpdump support */ 961 setup_zfcpdump(); 962 963 /* Add system specific data to the random pool */ 964 setup_randomness(); 965 } 966