1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * S390 version 4 * Copyright IBM Corp. 1999, 2012 5 * Author(s): Hartmut Penner (hp@de.ibm.com), 6 * Martin Schwidefsky (schwidefsky@de.ibm.com) 7 * 8 * Derived from "arch/i386/kernel/setup.c" 9 * Copyright (C) 1995, Linus Torvalds 10 */ 11 12 /* 13 * This file handles the architecture-dependent parts of initialization 14 */ 15 16 #define KMSG_COMPONENT "setup" 17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt 18 19 #include <linux/errno.h> 20 #include <linux/export.h> 21 #include <linux/sched.h> 22 #include <linux/sched/task.h> 23 #include <linux/cpu.h> 24 #include <linux/kernel.h> 25 #include <linux/memblock.h> 26 #include <linux/mm.h> 27 #include <linux/stddef.h> 28 #include <linux/unistd.h> 29 #include <linux/ptrace.h> 30 #include <linux/random.h> 31 #include <linux/user.h> 32 #include <linux/tty.h> 33 #include <linux/ioport.h> 34 #include <linux/delay.h> 35 #include <linux/init.h> 36 #include <linux/initrd.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 #include <linux/start_kernel.h> 52 53 #include <asm/boot_data.h> 54 #include <asm/ipl.h> 55 #include <asm/facility.h> 56 #include <asm/smp.h> 57 #include <asm/mmu_context.h> 58 #include <asm/cpcmd.h> 59 #include <asm/lowcore.h> 60 #include <asm/nmi.h> 61 #include <asm/irq.h> 62 #include <asm/page.h> 63 #include <asm/ptrace.h> 64 #include <asm/sections.h> 65 #include <asm/ebcdic.h> 66 #include <asm/diag.h> 67 #include <asm/os_info.h> 68 #include <asm/sclp.h> 69 #include <asm/stacktrace.h> 70 #include <asm/sysinfo.h> 71 #include <asm/numa.h> 72 #include <asm/alternative.h> 73 #include <asm/nospec-branch.h> 74 #include <asm/mem_detect.h> 75 #include <asm/uv.h> 76 #include <asm/asm-offsets.h> 77 #include "entry.h" 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 __read_mostly = 0; 92 char elf_platform[ELF_PLATFORM_SIZE]; 93 94 unsigned long int_hwcap = 0; 95 96 int __bootdata(noexec_disabled); 97 int __bootdata(memory_end_set); 98 unsigned long __bootdata(memory_end); 99 unsigned long __bootdata(vmalloc_size); 100 unsigned long __bootdata(max_physmem_end); 101 struct mem_detect_info __bootdata(mem_detect); 102 103 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table); 104 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table); 105 unsigned long __bootdata_preserved(__swsusp_reset_dma); 106 unsigned long __bootdata_preserved(__stext_dma); 107 unsigned long __bootdata_preserved(__etext_dma); 108 unsigned long __bootdata_preserved(__sdma); 109 unsigned long __bootdata_preserved(__edma); 110 unsigned long __bootdata_preserved(__kaslr_offset); 111 unsigned int __bootdata_preserved(zlib_dfltcc_support); 112 EXPORT_SYMBOL(zlib_dfltcc_support); 113 114 unsigned long VMALLOC_START; 115 EXPORT_SYMBOL(VMALLOC_START); 116 117 unsigned long VMALLOC_END; 118 EXPORT_SYMBOL(VMALLOC_END); 119 120 struct page *vmemmap; 121 EXPORT_SYMBOL(vmemmap); 122 123 unsigned long MODULES_VADDR; 124 unsigned long MODULES_END; 125 126 /* An array with a pointer to the lowcore of every CPU. */ 127 struct lowcore *lowcore_ptr[NR_CPUS]; 128 EXPORT_SYMBOL(lowcore_ptr); 129 130 /* 131 * This is set up by the setup-routine at boot-time 132 * for S390 need to find out, what we have to setup 133 * using address 0x10400 ... 134 */ 135 136 #include <asm/setup.h> 137 138 /* 139 * condev= and conmode= setup parameter. 140 */ 141 142 static int __init condev_setup(char *str) 143 { 144 int vdev; 145 146 vdev = simple_strtoul(str, &str, 0); 147 if (vdev >= 0 && vdev < 65536) { 148 console_devno = vdev; 149 console_irq = -1; 150 } 151 return 1; 152 } 153 154 __setup("condev=", condev_setup); 155 156 static void __init set_preferred_console(void) 157 { 158 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP) 159 add_preferred_console("ttyS", 0, NULL); 160 else if (CONSOLE_IS_3270) 161 add_preferred_console("tty3270", 0, NULL); 162 else if (CONSOLE_IS_VT220) 163 add_preferred_console("ttyS", 1, NULL); 164 else if (CONSOLE_IS_HVC) 165 add_preferred_console("hvc", 0, NULL); 166 } 167 168 static int __init conmode_setup(char *str) 169 { 170 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 171 if (!strcmp(str, "hwc") || !strcmp(str, "sclp")) 172 SET_CONSOLE_SCLP; 173 #endif 174 #if defined(CONFIG_TN3215_CONSOLE) 175 if (!strcmp(str, "3215")) 176 SET_CONSOLE_3215; 177 #endif 178 #if defined(CONFIG_TN3270_CONSOLE) 179 if (!strcmp(str, "3270")) 180 SET_CONSOLE_3270; 181 #endif 182 set_preferred_console(); 183 return 1; 184 } 185 186 __setup("conmode=", conmode_setup); 187 188 static void __init conmode_default(void) 189 { 190 char query_buffer[1024]; 191 char *ptr; 192 193 if (MACHINE_IS_VM) { 194 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL); 195 console_devno = simple_strtoul(query_buffer + 5, NULL, 16); 196 ptr = strstr(query_buffer, "SUBCHANNEL ="); 197 console_irq = simple_strtoul(ptr + 13, NULL, 16); 198 cpcmd("QUERY TERM", query_buffer, 1024, NULL); 199 ptr = strstr(query_buffer, "CONMODE"); 200 /* 201 * Set the conmode to 3215 so that the device recognition 202 * will set the cu_type of the console to 3215. If the 203 * conmode is 3270 and we don't set it back then both 204 * 3215 and the 3270 driver will try to access the console 205 * device (3215 as console and 3270 as normal tty). 206 */ 207 cpcmd("TERM CONMODE 3215", NULL, 0, NULL); 208 if (ptr == NULL) { 209 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 210 SET_CONSOLE_SCLP; 211 #endif 212 return; 213 } 214 if (str_has_prefix(ptr + 8, "3270")) { 215 #if defined(CONFIG_TN3270_CONSOLE) 216 SET_CONSOLE_3270; 217 #elif defined(CONFIG_TN3215_CONSOLE) 218 SET_CONSOLE_3215; 219 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 220 SET_CONSOLE_SCLP; 221 #endif 222 } else if (str_has_prefix(ptr + 8, "3215")) { 223 #if defined(CONFIG_TN3215_CONSOLE) 224 SET_CONSOLE_3215; 225 #elif defined(CONFIG_TN3270_CONSOLE) 226 SET_CONSOLE_3270; 227 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 228 SET_CONSOLE_SCLP; 229 #endif 230 } 231 } else if (MACHINE_IS_KVM) { 232 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE)) 233 SET_CONSOLE_VT220; 234 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE)) 235 SET_CONSOLE_SCLP; 236 else 237 SET_CONSOLE_HVC; 238 } else { 239 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 240 SET_CONSOLE_SCLP; 241 #endif 242 } 243 } 244 245 #ifdef CONFIG_CRASH_DUMP 246 static void __init setup_zfcpdump(void) 247 { 248 if (ipl_info.type != IPL_TYPE_FCP_DUMP) 249 return; 250 if (OLDMEM_BASE) 251 return; 252 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev"); 253 console_loglevel = 2; 254 } 255 #else 256 static inline void setup_zfcpdump(void) {} 257 #endif /* CONFIG_CRASH_DUMP */ 258 259 /* 260 * Reboot, halt and power_off stubs. They just call _machine_restart, 261 * _machine_halt or _machine_power_off. 262 */ 263 264 void machine_restart(char *command) 265 { 266 if ((!in_interrupt() && !in_atomic()) || 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_restart(command); 273 } 274 275 void machine_halt(void) 276 { 277 if (!in_interrupt() || oops_in_progress) 278 /* 279 * Only unblank the console if we are called in enabled 280 * context or a bust_spinlocks cleared the way for us. 281 */ 282 console_unblank(); 283 _machine_halt(); 284 } 285 286 void machine_power_off(void) 287 { 288 if (!in_interrupt() || oops_in_progress) 289 /* 290 * Only unblank the console if we are called in enabled 291 * context or a bust_spinlocks cleared the way for us. 292 */ 293 console_unblank(); 294 _machine_power_off(); 295 } 296 297 /* 298 * Dummy power off function. 299 */ 300 void (*pm_power_off)(void) = machine_power_off; 301 EXPORT_SYMBOL_GPL(pm_power_off); 302 303 void *restart_stack __section(.data); 304 305 unsigned long stack_alloc(void) 306 { 307 #ifdef CONFIG_VMAP_STACK 308 return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE, 309 THREADINFO_GFP, NUMA_NO_NODE, 310 __builtin_return_address(0)); 311 #else 312 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 313 #endif 314 } 315 316 void stack_free(unsigned long stack) 317 { 318 #ifdef CONFIG_VMAP_STACK 319 vfree((void *) stack); 320 #else 321 free_pages(stack, THREAD_SIZE_ORDER); 322 #endif 323 } 324 325 int __init arch_early_irq_init(void) 326 { 327 unsigned long stack; 328 329 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 330 if (!stack) 331 panic("Couldn't allocate async stack"); 332 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET; 333 return 0; 334 } 335 336 static int __init async_stack_realloc(void) 337 { 338 unsigned long old, new; 339 340 old = S390_lowcore.async_stack - STACK_INIT_OFFSET; 341 new = stack_alloc(); 342 if (!new) 343 panic("Couldn't allocate async stack"); 344 S390_lowcore.async_stack = new + STACK_INIT_OFFSET; 345 free_pages(old, THREAD_SIZE_ORDER); 346 return 0; 347 } 348 early_initcall(async_stack_realloc); 349 350 void __init arch_call_rest_init(void) 351 { 352 unsigned long stack; 353 354 stack = stack_alloc(); 355 if (!stack) 356 panic("Couldn't allocate kernel stack"); 357 current->stack = (void *) stack; 358 #ifdef CONFIG_VMAP_STACK 359 current->stack_vm_area = (void *) stack; 360 #endif 361 set_task_stack_end_magic(current); 362 stack += STACK_INIT_OFFSET; 363 S390_lowcore.kernel_stack = stack; 364 CALL_ON_STACK_NORETURN(rest_init, stack); 365 } 366 367 static void __init setup_lowcore_dat_off(void) 368 { 369 struct lowcore *lc; 370 371 /* 372 * Setup lowcore for boot cpu 373 */ 374 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE); 375 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc)); 376 if (!lc) 377 panic("%s: Failed to allocate %zu bytes align=%zx\n", 378 __func__, sizeof(*lc), sizeof(*lc)); 379 380 lc->restart_psw.mask = PSW_KERNEL_BITS; 381 lc->restart_psw.addr = (unsigned long) restart_int_handler; 382 lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK; 383 lc->external_new_psw.addr = (unsigned long) ext_int_handler; 384 lc->svc_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK; 385 lc->svc_new_psw.addr = (unsigned long) system_call; 386 lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK; 387 lc->program_new_psw.addr = (unsigned long) pgm_check_handler; 388 lc->mcck_new_psw.mask = PSW_KERNEL_BITS; 389 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler; 390 lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK; 391 lc->io_new_psw.addr = (unsigned long) io_int_handler; 392 lc->clock_comparator = clock_comparator_max; 393 lc->nodat_stack = ((unsigned long) &init_thread_union) 394 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); 395 lc->current_task = (unsigned long)&init_task; 396 lc->lpp = LPP_MAGIC; 397 lc->machine_flags = S390_lowcore.machine_flags; 398 lc->preempt_count = S390_lowcore.preempt_count; 399 lc->stfl_fac_list = S390_lowcore.stfl_fac_list; 400 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list, 401 sizeof(lc->stfle_fac_list)); 402 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list, 403 sizeof(lc->alt_stfle_fac_list)); 404 nmi_alloc_boot_cpu(lc); 405 vdso_alloc_boot_cpu(lc); 406 lc->sync_enter_timer = S390_lowcore.sync_enter_timer; 407 lc->async_enter_timer = S390_lowcore.async_enter_timer; 408 lc->exit_timer = S390_lowcore.exit_timer; 409 lc->user_timer = S390_lowcore.user_timer; 410 lc->system_timer = S390_lowcore.system_timer; 411 lc->steal_timer = S390_lowcore.steal_timer; 412 lc->last_update_timer = S390_lowcore.last_update_timer; 413 lc->last_update_clock = S390_lowcore.last_update_clock; 414 415 /* 416 * Allocate the global restart stack which is the same for 417 * all CPUs in cast *one* of them does a PSW restart. 418 */ 419 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE); 420 if (!restart_stack) 421 panic("%s: Failed to allocate %lu bytes align=0x%lx\n", 422 __func__, THREAD_SIZE, THREAD_SIZE); 423 restart_stack += STACK_INIT_OFFSET; 424 425 /* 426 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 427 * restart data to the absolute zero lowcore. This is necessary if 428 * PSW restart is done on an offline CPU that has lowcore zero. 429 */ 430 lc->restart_stack = (unsigned long) restart_stack; 431 lc->restart_fn = (unsigned long) do_restart; 432 lc->restart_data = 0; 433 lc->restart_source = -1UL; 434 435 /* Setup absolute zero lowcore */ 436 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack); 437 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn); 438 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data); 439 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source); 440 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw); 441 442 lc->spinlock_lockval = arch_spin_lockval(0); 443 lc->spinlock_index = 0; 444 arch_spin_lock_setup(0); 445 lc->br_r1_trampoline = 0x07f1; /* br %r1 */ 446 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW); 447 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW); 448 449 set_prefix((u32)(unsigned long) lc); 450 lowcore_ptr[0] = lc; 451 } 452 453 static void __init setup_lowcore_dat_on(void) 454 { 455 __ctl_clear_bit(0, 28); 456 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT; 457 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT; 458 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT; 459 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT; 460 __ctl_set_bit(0, 28); 461 } 462 463 static struct resource code_resource = { 464 .name = "Kernel code", 465 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 466 }; 467 468 static struct resource data_resource = { 469 .name = "Kernel data", 470 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 471 }; 472 473 static struct resource bss_resource = { 474 .name = "Kernel bss", 475 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 476 }; 477 478 static struct resource __initdata *standard_resources[] = { 479 &code_resource, 480 &data_resource, 481 &bss_resource, 482 }; 483 484 static void __init setup_resources(void) 485 { 486 struct resource *res, *std_res, *sub_res; 487 struct memblock_region *reg; 488 int j; 489 490 code_resource.start = (unsigned long) _text; 491 code_resource.end = (unsigned long) _etext - 1; 492 data_resource.start = (unsigned long) _etext; 493 data_resource.end = (unsigned long) _edata - 1; 494 bss_resource.start = (unsigned long) __bss_start; 495 bss_resource.end = (unsigned long) __bss_stop - 1; 496 497 for_each_memblock(memory, reg) { 498 res = memblock_alloc(sizeof(*res), 8); 499 if (!res) 500 panic("%s: Failed to allocate %zu bytes align=0x%x\n", 501 __func__, sizeof(*res), 8); 502 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM; 503 504 res->name = "System RAM"; 505 res->start = reg->base; 506 res->end = reg->base + reg->size - 1; 507 request_resource(&iomem_resource, res); 508 509 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 510 std_res = standard_resources[j]; 511 if (std_res->start < res->start || 512 std_res->start > res->end) 513 continue; 514 if (std_res->end > res->end) { 515 sub_res = memblock_alloc(sizeof(*sub_res), 8); 516 if (!sub_res) 517 panic("%s: Failed to allocate %zu bytes align=0x%x\n", 518 __func__, sizeof(*sub_res), 8); 519 *sub_res = *std_res; 520 sub_res->end = res->end; 521 std_res->start = res->end + 1; 522 request_resource(res, sub_res); 523 } else { 524 request_resource(res, std_res); 525 } 526 } 527 } 528 #ifdef CONFIG_CRASH_DUMP 529 /* 530 * Re-add removed crash kernel memory as reserved memory. This makes 531 * sure it will be mapped with the identity mapping and struct pages 532 * will be created, so it can be resized later on. 533 * However add it later since the crash kernel resource should not be 534 * part of the System RAM resource. 535 */ 536 if (crashk_res.end) { 537 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0); 538 memblock_reserve(crashk_res.start, resource_size(&crashk_res)); 539 insert_resource(&iomem_resource, &crashk_res); 540 } 541 #endif 542 } 543 544 static void __init setup_memory_end(void) 545 { 546 unsigned long vmax, tmp; 547 548 /* Choose kernel address space layout: 3 or 4 levels. */ 549 if (IS_ENABLED(CONFIG_KASAN)) { 550 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING) 551 ? _REGION1_SIZE 552 : _REGION2_SIZE; 553 } else { 554 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE; 555 tmp = tmp * (sizeof(struct page) + PAGE_SIZE); 556 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE) 557 vmax = _REGION2_SIZE; /* 3-level kernel page table */ 558 else 559 vmax = _REGION1_SIZE; /* 4-level kernel page table */ 560 } 561 562 if (is_prot_virt_host()) 563 adjust_to_uv_max(&vmax); 564 565 /* module area is at the end of the kernel address space. */ 566 MODULES_END = vmax; 567 MODULES_VADDR = MODULES_END - MODULES_LEN; 568 VMALLOC_END = MODULES_VADDR; 569 VMALLOC_START = VMALLOC_END - vmalloc_size; 570 571 /* Split remaining virtual space between 1:1 mapping & vmemmap array */ 572 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page)); 573 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */ 574 tmp = SECTION_ALIGN_UP(tmp); 575 tmp = VMALLOC_START - tmp * sizeof(struct page); 576 tmp &= ~((vmax >> 11) - 1); /* align to page table level */ 577 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS); 578 vmemmap = (struct page *) tmp; 579 580 /* Take care that memory_end is set and <= vmemmap */ 581 memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap); 582 #ifdef CONFIG_KASAN 583 /* fit in kasan shadow memory region between 1:1 and vmemmap */ 584 memory_end = min(memory_end, KASAN_SHADOW_START); 585 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END); 586 #endif 587 max_pfn = max_low_pfn = PFN_DOWN(memory_end); 588 memblock_remove(memory_end, ULONG_MAX); 589 590 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20); 591 } 592 593 #ifdef CONFIG_CRASH_DUMP 594 595 /* 596 * When kdump is enabled, we have to ensure that no memory from the area 597 * [0 - crashkernel memory size] is set offline - it will be exchanged with 598 * the crashkernel memory region when kdump is triggered. The crashkernel 599 * memory region can never get offlined (pages are unmovable). 600 */ 601 static int kdump_mem_notifier(struct notifier_block *nb, 602 unsigned long action, void *data) 603 { 604 struct memory_notify *arg = data; 605 606 if (action != MEM_GOING_OFFLINE) 607 return NOTIFY_OK; 608 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 609 return NOTIFY_BAD; 610 return NOTIFY_OK; 611 } 612 613 static struct notifier_block kdump_mem_nb = { 614 .notifier_call = kdump_mem_notifier, 615 }; 616 617 #endif 618 619 /* 620 * Make sure that the area behind memory_end is protected 621 */ 622 static void reserve_memory_end(void) 623 { 624 if (memory_end_set) 625 memblock_reserve(memory_end, ULONG_MAX); 626 } 627 628 /* 629 * Make sure that oldmem, where the dump is stored, is protected 630 */ 631 static void reserve_oldmem(void) 632 { 633 #ifdef CONFIG_CRASH_DUMP 634 if (OLDMEM_BASE) 635 /* Forget all memory above the running kdump system */ 636 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 637 #endif 638 } 639 640 /* 641 * Make sure that oldmem, where the dump is stored, is protected 642 */ 643 static void remove_oldmem(void) 644 { 645 #ifdef CONFIG_CRASH_DUMP 646 if (OLDMEM_BASE) 647 /* Forget all memory above the running kdump system */ 648 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 649 #endif 650 } 651 652 /* 653 * Reserve memory for kdump kernel to be loaded with kexec 654 */ 655 static void __init reserve_crashkernel(void) 656 { 657 #ifdef CONFIG_CRASH_DUMP 658 unsigned long long crash_base, crash_size; 659 phys_addr_t low, high; 660 int rc; 661 662 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size, 663 &crash_base); 664 665 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 666 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 667 if (rc || crash_size == 0) 668 return; 669 670 if (memblock.memory.regions[0].size < crash_size) { 671 pr_info("crashkernel reservation failed: %s\n", 672 "first memory chunk must be at least crashkernel size"); 673 return; 674 } 675 676 low = crash_base ?: OLDMEM_BASE; 677 high = low + crash_size; 678 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) { 679 /* The crashkernel fits into OLDMEM, reuse OLDMEM */ 680 crash_base = low; 681 } else { 682 /* Find suitable area in free memory */ 683 low = max_t(unsigned long, crash_size, sclp.hsa_size); 684 high = crash_base ? crash_base + crash_size : ULONG_MAX; 685 686 if (crash_base && crash_base < low) { 687 pr_info("crashkernel reservation failed: %s\n", 688 "crash_base too low"); 689 return; 690 } 691 low = crash_base ?: low; 692 crash_base = memblock_find_in_range(low, high, crash_size, 693 KEXEC_CRASH_MEM_ALIGN); 694 } 695 696 if (!crash_base) { 697 pr_info("crashkernel reservation failed: %s\n", 698 "no suitable area found"); 699 return; 700 } 701 702 if (register_memory_notifier(&kdump_mem_nb)) 703 return; 704 705 if (!OLDMEM_BASE && MACHINE_IS_VM) 706 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 707 crashk_res.start = crash_base; 708 crashk_res.end = crash_base + crash_size - 1; 709 memblock_remove(crash_base, crash_size); 710 pr_info("Reserving %lluMB of memory at %lluMB " 711 "for crashkernel (System RAM: %luMB)\n", 712 crash_size >> 20, crash_base >> 20, 713 (unsigned long)memblock.memory.total_size >> 20); 714 os_info_crashkernel_add(crash_base, crash_size); 715 #endif 716 } 717 718 /* 719 * Reserve the initrd from being used by memblock 720 */ 721 static void __init reserve_initrd(void) 722 { 723 #ifdef CONFIG_BLK_DEV_INITRD 724 if (!INITRD_START || !INITRD_SIZE) 725 return; 726 initrd_start = INITRD_START; 727 initrd_end = initrd_start + INITRD_SIZE; 728 memblock_reserve(INITRD_START, INITRD_SIZE); 729 #endif 730 } 731 732 /* 733 * Reserve the memory area used to pass the certificate lists 734 */ 735 static void __init reserve_certificate_list(void) 736 { 737 if (ipl_cert_list_addr) 738 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size); 739 } 740 741 static void __init reserve_mem_detect_info(void) 742 { 743 unsigned long start, size; 744 745 get_mem_detect_reserved(&start, &size); 746 if (size) 747 memblock_reserve(start, size); 748 } 749 750 static void __init free_mem_detect_info(void) 751 { 752 unsigned long start, size; 753 754 get_mem_detect_reserved(&start, &size); 755 if (size) 756 memblock_free(start, size); 757 } 758 759 static const char * __init get_mem_info_source(void) 760 { 761 switch (mem_detect.info_source) { 762 case MEM_DETECT_SCLP_STOR_INFO: 763 return "sclp storage info"; 764 case MEM_DETECT_DIAG260: 765 return "diag260"; 766 case MEM_DETECT_SCLP_READ_INFO: 767 return "sclp read info"; 768 case MEM_DETECT_BIN_SEARCH: 769 return "binary search"; 770 } 771 return "none"; 772 } 773 774 static void __init memblock_add_mem_detect_info(void) 775 { 776 unsigned long start, end; 777 int i; 778 779 memblock_dbg("physmem info source: %s (%hhd)\n", 780 get_mem_info_source(), mem_detect.info_source); 781 /* keep memblock lists close to the kernel */ 782 memblock_set_bottom_up(true); 783 for_each_mem_detect_block(i, &start, &end) { 784 memblock_add(start, end - start); 785 memblock_physmem_add(start, end - start); 786 } 787 memblock_set_bottom_up(false); 788 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0); 789 memblock_dump_all(); 790 } 791 792 /* 793 * Check for initrd being in usable memory 794 */ 795 static void __init check_initrd(void) 796 { 797 #ifdef CONFIG_BLK_DEV_INITRD 798 if (INITRD_START && INITRD_SIZE && 799 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) { 800 pr_err("The initial RAM disk does not fit into the memory\n"); 801 memblock_free(INITRD_START, INITRD_SIZE); 802 initrd_start = initrd_end = 0; 803 } 804 #endif 805 } 806 807 /* 808 * Reserve memory used for lowcore/command line/kernel image. 809 */ 810 static void __init reserve_kernel(void) 811 { 812 unsigned long start_pfn = PFN_UP(__pa(_end)); 813 814 memblock_reserve(0, HEAD_END); 815 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn) 816 - (unsigned long)_stext); 817 memblock_reserve(__sdma, __edma - __sdma); 818 } 819 820 static void __init setup_memory(void) 821 { 822 struct memblock_region *reg; 823 824 /* 825 * Init storage key for present memory 826 */ 827 for_each_memblock(memory, reg) { 828 storage_key_init_range(reg->base, reg->base + reg->size); 829 } 830 psw_set_key(PAGE_DEFAULT_KEY); 831 832 /* Only cosmetics */ 833 memblock_enforce_memory_limit(memblock_end_of_DRAM()); 834 } 835 836 /* 837 * Setup hardware capabilities. 838 */ 839 static int __init setup_hwcaps(void) 840 { 841 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 }; 842 struct cpuid cpu_id; 843 int i; 844 845 /* 846 * The store facility list bits numbers as found in the principles 847 * of operation are numbered with bit 1UL<<31 as number 0 to 848 * bit 1UL<<0 as number 31. 849 * Bit 0: instructions named N3, "backported" to esa-mode 850 * Bit 2: z/Architecture mode is active 851 * Bit 7: the store-facility-list-extended facility is installed 852 * Bit 17: the message-security assist is installed 853 * Bit 19: the long-displacement facility is installed 854 * Bit 21: the extended-immediate facility is installed 855 * Bit 22: extended-translation facility 3 is installed 856 * Bit 30: extended-translation facility 3 enhancement facility 857 * These get translated to: 858 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1, 859 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3, 860 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and 861 * HWCAP_S390_ETF3EH bit 8 (22 && 30). 862 */ 863 for (i = 0; i < 6; i++) 864 if (test_facility(stfl_bits[i])) 865 elf_hwcap |= 1UL << i; 866 867 if (test_facility(22) && test_facility(30)) 868 elf_hwcap |= HWCAP_S390_ETF3EH; 869 870 /* 871 * Check for additional facilities with store-facility-list-extended. 872 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0 873 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information 874 * as stored by stfl, bits 32-xxx contain additional facilities. 875 * How many facility words are stored depends on the number of 876 * doublewords passed to the instruction. The additional facilities 877 * are: 878 * Bit 42: decimal floating point facility is installed 879 * Bit 44: perform floating point operation facility is installed 880 * translated to: 881 * HWCAP_S390_DFP bit 6 (42 && 44). 882 */ 883 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44)) 884 elf_hwcap |= HWCAP_S390_DFP; 885 886 /* 887 * Huge page support HWCAP_S390_HPAGE is bit 7. 888 */ 889 if (MACHINE_HAS_EDAT1) 890 elf_hwcap |= HWCAP_S390_HPAGE; 891 892 /* 893 * 64-bit register support for 31-bit processes 894 * HWCAP_S390_HIGH_GPRS is bit 9. 895 */ 896 elf_hwcap |= HWCAP_S390_HIGH_GPRS; 897 898 /* 899 * Transactional execution support HWCAP_S390_TE is bit 10. 900 */ 901 if (MACHINE_HAS_TE) 902 elf_hwcap |= HWCAP_S390_TE; 903 904 /* 905 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension 906 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX 907 * instead of facility bit 129. 908 */ 909 if (MACHINE_HAS_VX) { 910 elf_hwcap |= HWCAP_S390_VXRS; 911 if (test_facility(134)) 912 elf_hwcap |= HWCAP_S390_VXRS_EXT; 913 if (test_facility(135)) 914 elf_hwcap |= HWCAP_S390_VXRS_BCD; 915 if (test_facility(148)) 916 elf_hwcap |= HWCAP_S390_VXRS_EXT2; 917 if (test_facility(152)) 918 elf_hwcap |= HWCAP_S390_VXRS_PDE; 919 } 920 if (test_facility(150)) 921 elf_hwcap |= HWCAP_S390_SORT; 922 if (test_facility(151)) 923 elf_hwcap |= HWCAP_S390_DFLT; 924 925 /* 926 * Guarded storage support HWCAP_S390_GS is bit 12. 927 */ 928 if (MACHINE_HAS_GS) 929 elf_hwcap |= HWCAP_S390_GS; 930 931 get_cpu_id(&cpu_id); 932 add_device_randomness(&cpu_id, sizeof(cpu_id)); 933 switch (cpu_id.machine) { 934 case 0x2064: 935 case 0x2066: 936 default: /* Use "z900" as default for 64 bit kernels. */ 937 strcpy(elf_platform, "z900"); 938 break; 939 case 0x2084: 940 case 0x2086: 941 strcpy(elf_platform, "z990"); 942 break; 943 case 0x2094: 944 case 0x2096: 945 strcpy(elf_platform, "z9-109"); 946 break; 947 case 0x2097: 948 case 0x2098: 949 strcpy(elf_platform, "z10"); 950 break; 951 case 0x2817: 952 case 0x2818: 953 strcpy(elf_platform, "z196"); 954 break; 955 case 0x2827: 956 case 0x2828: 957 strcpy(elf_platform, "zEC12"); 958 break; 959 case 0x2964: 960 case 0x2965: 961 strcpy(elf_platform, "z13"); 962 break; 963 case 0x3906: 964 case 0x3907: 965 strcpy(elf_platform, "z14"); 966 break; 967 case 0x8561: 968 case 0x8562: 969 strcpy(elf_platform, "z15"); 970 break; 971 } 972 973 /* 974 * Virtualization support HWCAP_INT_SIE is bit 0. 975 */ 976 if (sclp.has_sief2) 977 int_hwcap |= HWCAP_INT_SIE; 978 979 return 0; 980 } 981 arch_initcall(setup_hwcaps); 982 983 /* 984 * Add system information as device randomness 985 */ 986 static void __init setup_randomness(void) 987 { 988 struct sysinfo_3_2_2 *vmms; 989 990 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE, 991 PAGE_SIZE); 992 if (!vmms) 993 panic("Failed to allocate memory for sysinfo structure\n"); 994 995 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count) 996 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count); 997 memblock_free((unsigned long) vmms, PAGE_SIZE); 998 } 999 1000 /* 1001 * Find the correct size for the task_struct. This depends on 1002 * the size of the struct fpu at the end of the thread_struct 1003 * which is embedded in the task_struct. 1004 */ 1005 static void __init setup_task_size(void) 1006 { 1007 int task_size = sizeof(struct task_struct); 1008 1009 if (!MACHINE_HAS_VX) { 1010 task_size -= sizeof(__vector128) * __NUM_VXRS; 1011 task_size += sizeof(freg_t) * __NUM_FPRS; 1012 } 1013 arch_task_struct_size = task_size; 1014 } 1015 1016 /* 1017 * Issue diagnose 318 to set the control program name and 1018 * version codes. 1019 */ 1020 static void __init setup_control_program_code(void) 1021 { 1022 union diag318_info diag318_info = { 1023 .cpnc = CPNC_LINUX, 1024 .cpvc_linux = 0, 1025 .cpvc_distro = {0}, 1026 }; 1027 1028 if (!sclp.has_diag318) 1029 return; 1030 1031 diag_stat_inc(DIAG_STAT_X318); 1032 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val)); 1033 } 1034 1035 /* 1036 * Print the component list from the IPL report 1037 */ 1038 static void __init log_component_list(void) 1039 { 1040 struct ipl_rb_component_entry *ptr, *end; 1041 char *str; 1042 1043 if (!early_ipl_comp_list_addr) 1044 return; 1045 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL) 1046 pr_info("Linux is running with Secure-IPL enabled\n"); 1047 else 1048 pr_info("Linux is running with Secure-IPL disabled\n"); 1049 ptr = (void *) early_ipl_comp_list_addr; 1050 end = (void *) ptr + early_ipl_comp_list_size; 1051 pr_info("The IPL report contains the following components:\n"); 1052 while (ptr < end) { 1053 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) { 1054 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED) 1055 str = "signed, verified"; 1056 else 1057 str = "signed, verification failed"; 1058 } else { 1059 str = "not signed"; 1060 } 1061 pr_info("%016llx - %016llx (%s)\n", 1062 ptr->addr, ptr->addr + ptr->len, str); 1063 ptr++; 1064 } 1065 } 1066 1067 /* 1068 * Setup function called from init/main.c just after the banner 1069 * was printed. 1070 */ 1071 1072 void __init setup_arch(char **cmdline_p) 1073 { 1074 /* 1075 * print what head.S has found out about the machine 1076 */ 1077 if (MACHINE_IS_VM) 1078 pr_info("Linux is running as a z/VM " 1079 "guest operating system in 64-bit mode\n"); 1080 else if (MACHINE_IS_KVM) 1081 pr_info("Linux is running under KVM in 64-bit mode\n"); 1082 else if (MACHINE_IS_LPAR) 1083 pr_info("Linux is running natively in 64-bit mode\n"); 1084 else 1085 pr_info("Linux is running as a guest in 64-bit mode\n"); 1086 1087 log_component_list(); 1088 1089 /* Have one command line that is parsed and saved in /proc/cmdline */ 1090 /* boot_command_line has been already set up in early.c */ 1091 *cmdline_p = boot_command_line; 1092 1093 ROOT_DEV = Root_RAM0; 1094 1095 init_mm.start_code = (unsigned long) _text; 1096 init_mm.end_code = (unsigned long) _etext; 1097 init_mm.end_data = (unsigned long) _edata; 1098 init_mm.brk = (unsigned long) _end; 1099 1100 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO)) 1101 nospec_auto_detect(); 1102 1103 jump_label_init(); 1104 parse_early_param(); 1105 #ifdef CONFIG_CRASH_DUMP 1106 /* Deactivate elfcorehdr= kernel parameter */ 1107 elfcorehdr_addr = ELFCORE_ADDR_MAX; 1108 #endif 1109 1110 os_info_init(); 1111 setup_ipl(); 1112 setup_task_size(); 1113 setup_control_program_code(); 1114 1115 /* Do some memory reservations *before* memory is added to memblock */ 1116 reserve_memory_end(); 1117 reserve_oldmem(); 1118 reserve_kernel(); 1119 reserve_initrd(); 1120 reserve_certificate_list(); 1121 reserve_mem_detect_info(); 1122 memblock_allow_resize(); 1123 1124 /* Get information about *all* installed memory */ 1125 memblock_add_mem_detect_info(); 1126 1127 free_mem_detect_info(); 1128 remove_oldmem(); 1129 1130 /* 1131 * Make sure all chunks are MAX_ORDER aligned so we don't need the 1132 * extra checks that HOLES_IN_ZONE would require. 1133 * 1134 * Is this still required? 1135 */ 1136 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT)); 1137 1138 if (is_prot_virt_host()) 1139 setup_uv(); 1140 setup_memory_end(); 1141 setup_memory(); 1142 dma_contiguous_reserve(memory_end); 1143 vmcp_cma_reserve(); 1144 1145 check_initrd(); 1146 reserve_crashkernel(); 1147 #ifdef CONFIG_CRASH_DUMP 1148 /* 1149 * Be aware that smp_save_dump_cpus() triggers a system reset. 1150 * Therefore CPU and device initialization should be done afterwards. 1151 */ 1152 smp_save_dump_cpus(); 1153 #endif 1154 1155 setup_resources(); 1156 setup_lowcore_dat_off(); 1157 smp_fill_possible_mask(); 1158 cpu_detect_mhz_feature(); 1159 cpu_init(); 1160 numa_setup(); 1161 smp_detect_cpus(); 1162 topology_init_early(); 1163 1164 /* 1165 * Create kernel page tables and switch to virtual addressing. 1166 */ 1167 paging_init(); 1168 1169 /* 1170 * After paging_init created the kernel page table, the new PSWs 1171 * in lowcore can now run with DAT enabled. 1172 */ 1173 setup_lowcore_dat_on(); 1174 1175 /* Setup default console */ 1176 conmode_default(); 1177 set_preferred_console(); 1178 1179 apply_alternative_instructions(); 1180 if (IS_ENABLED(CONFIG_EXPOLINE)) 1181 nospec_init_branches(); 1182 1183 /* Setup zfcpdump support */ 1184 setup_zfcpdump(); 1185 1186 /* Add system specific data to the random pool */ 1187 setup_randomness(); 1188 } 1189