arch/s390/kernel/setup.c

   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)
 309                 __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
 310                                      VMALLOC_START, VMALLOC_END,
 311                                      THREADINFO_GFP,
 312                                      PAGE_KERNEL, 0, NUMA_NO_NODE,
 313                                      __builtin_return_address(0));
 314 #else
 315         return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 316 #endif
 317 }
 318
 319 void stack_free(unsigned long stack)
 320 {
 321 #ifdef CONFIG_VMAP_STACK
 322         vfree((void *) stack);
 323 #else
 324         free_pages(stack, THREAD_SIZE_ORDER);
 325 #endif
 326 }
 327
 328 int __init arch_early_irq_init(void)
 329 {
 330         unsigned long stack;
 331
 332         stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 333         if (!stack)
 334                 panic("Couldn't allocate async stack");
 335         S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
 336         return 0;
 337 }
 338
 339 static int __init async_stack_realloc(void)
 340 {
 341         unsigned long old, new;
 342
 343         old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
 344         new = stack_alloc();
 345         if (!new)
 346                 panic("Couldn't allocate async stack");
 347         S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
 348         free_pages(old, THREAD_SIZE_ORDER);
 349         return 0;
 350 }
 351 early_initcall(async_stack_realloc);
 352
 353 void __init arch_call_rest_init(void)
 354 {
 355         unsigned long stack;
 356
 357         stack = stack_alloc();
 358         if (!stack)
 359                 panic("Couldn't allocate kernel stack");
 360         current->stack = (void *) stack;
 361 #ifdef CONFIG_VMAP_STACK
 362         current->stack_vm_area = (void *) stack;
 363 #endif
 364         set_task_stack_end_magic(current);
 365         stack += STACK_INIT_OFFSET;
 366         S390_lowcore.kernel_stack = stack;
 367         CALL_ON_STACK_NORETURN(rest_init, stack);
 368 }
 369
 370 static void __init setup_lowcore_dat_off(void)
 371 {
 372         struct lowcore *lc;
 373
 374         /*
 375          * Setup lowcore for boot cpu
 376          */
 377         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 378         lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
 379         if (!lc)
 380                 panic("%s: Failed to allocate %zu bytes align=%zx\n",
 381                       __func__, sizeof(*lc), sizeof(*lc));
 382
 383         lc->restart_psw.mask = PSW_KERNEL_BITS;
 384         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 385         lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 386         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 387         lc->svc_new_psw.mask = PSW_KERNEL_BITS |
 388                 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 389         lc->svc_new_psw.addr = (unsigned long) system_call;
 390         lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 391         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 392         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 393         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 394         lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 395         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 396         lc->clock_comparator = clock_comparator_max;
 397         lc->nodat_stack = ((unsigned long) &init_thread_union)
 398                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 399         lc->current_task = (unsigned long)&init_task;
 400         lc->lpp = LPP_MAGIC;
 401         lc->machine_flags = S390_lowcore.machine_flags;
 402         lc->preempt_count = S390_lowcore.preempt_count;
 403         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 404         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 405                sizeof(lc->stfle_fac_list));
 406         memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
 407                sizeof(lc->alt_stfle_fac_list));
 408         nmi_alloc_boot_cpu(lc);
 409         vdso_alloc_boot_cpu(lc);
 410         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 411         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 412         lc->exit_timer = S390_lowcore.exit_timer;
 413         lc->user_timer = S390_lowcore.user_timer;
 414         lc->system_timer = S390_lowcore.system_timer;
 415         lc->steal_timer = S390_lowcore.steal_timer;
 416         lc->last_update_timer = S390_lowcore.last_update_timer;
 417         lc->last_update_clock = S390_lowcore.last_update_clock;
 418
 419         /*
 420          * Allocate the global restart stack which is the same for
 421          * all CPUs in cast *one* of them does a PSW restart.
 422          */
 423         restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 424         if (!restart_stack)
 425                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 426                       __func__, THREAD_SIZE, THREAD_SIZE);
 427         restart_stack += STACK_INIT_OFFSET;
 428
 429         /*
 430          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 431          * restart data to the absolute zero lowcore. This is necessary if
 432          * PSW restart is done on an offline CPU that has lowcore zero.
 433          */
 434         lc->restart_stack = (unsigned long) restart_stack;
 435         lc->restart_fn = (unsigned long) do_restart;
 436         lc->restart_data = 0;
 437         lc->restart_source = -1UL;
 438
 439         /* Setup absolute zero lowcore */
 440         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 441         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 442         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 443         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 444         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 445
 446         lc->spinlock_lockval = arch_spin_lockval(0);
 447         lc->spinlock_index = 0;
 448         arch_spin_lock_setup(0);
 449         lc->br_r1_trampoline = 0x07f1;  /* br %r1 */
 450         lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
 451         lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
 452
 453         set_prefix((u32)(unsigned long) lc);
 454         lowcore_ptr[0] = lc;
 455 }
 456
 457 static void __init setup_lowcore_dat_on(void)
 458 {
 459         __ctl_clear_bit(0, 28);
 460         S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
 461         S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
 462         S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
 463         S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
 464         __ctl_set_bit(0, 28);
 465 }
 466
 467 static struct resource code_resource = {
 468         .name  = "Kernel code",
 469         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 470 };
 471
 472 static struct resource data_resource = {
 473         .name = "Kernel data",
 474         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 475 };
 476
 477 static struct resource bss_resource = {
 478         .name = "Kernel bss",
 479         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 480 };
 481
 482 static struct resource __initdata *standard_resources[] = {
 483         &code_resource,
 484         &data_resource,
 485         &bss_resource,
 486 };
 487
 488 static void __init setup_resources(void)
 489 {
 490         struct resource *res, *std_res, *sub_res;
 491         struct memblock_region *reg;
 492         int j;
 493
 494         code_resource.start = (unsigned long) _text;
 495         code_resource.end = (unsigned long) _etext - 1;
 496         data_resource.start = (unsigned long) _etext;
 497         data_resource.end = (unsigned long) _edata - 1;
 498         bss_resource.start = (unsigned long) __bss_start;
 499         bss_resource.end = (unsigned long) __bss_stop - 1;
 500
 501         for_each_memblock(memory, reg) {
 502                 res = memblock_alloc(sizeof(*res), 8);
 503                 if (!res)
 504                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 505                               __func__, sizeof(*res), 8);
 506                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 507
 508                 res->name = "System RAM";
 509                 res->start = reg->base;
 510                 res->end = reg->base + reg->size - 1;
 511                 request_resource(&iomem_resource, res);
 512
 513                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 514                         std_res = standard_resources[j];
 515                         if (std_res->start < res->start ||
 516                             std_res->start > res->end)
 517                                 continue;
 518                         if (std_res->end > res->end) {
 519                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
 520                                 if (!sub_res)
 521                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 522                                               __func__, sizeof(*sub_res), 8);
 523                                 *sub_res = *std_res;
 524                                 sub_res->end = res->end;
 525                                 std_res->start = res->end + 1;
 526                                 request_resource(res, sub_res);
 527                         } else {
 528                                 request_resource(res, std_res);
 529                         }
 530                 }
 531         }
 532 #ifdef CONFIG_CRASH_DUMP
 533         /*
 534          * Re-add removed crash kernel memory as reserved memory. This makes
 535          * sure it will be mapped with the identity mapping and struct pages
 536          * will be created, so it can be resized later on.
 537          * However add it later since the crash kernel resource should not be
 538          * part of the System RAM resource.
 539          */
 540         if (crashk_res.end) {
 541                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 542                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 543                 insert_resource(&iomem_resource, &crashk_res);
 544         }
 545 #endif
 546 }
 547
 548 static void __init setup_memory_end(void)
 549 {
 550         unsigned long vmax, tmp;
 551
 552         /* Choose kernel address space layout: 3 or 4 levels. */
 553         if (IS_ENABLED(CONFIG_KASAN)) {
 554                 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
 555                            ? _REGION1_SIZE
 556                            : _REGION2_SIZE;
 557         } else {
 558                 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
 559                 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
 560                 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
 561                         vmax = _REGION2_SIZE; /* 3-level kernel page table */
 562                 else
 563                         vmax = _REGION1_SIZE; /* 4-level kernel page table */
 564         }
 565
 566         if (is_prot_virt_host())
 567                 adjust_to_uv_max(&vmax);
 568
 569         /* module area is at the end of the kernel address space. */
 570         MODULES_END = vmax;
 571         MODULES_VADDR = MODULES_END - MODULES_LEN;
 572         VMALLOC_END = MODULES_VADDR;
 573         VMALLOC_START = VMALLOC_END - vmalloc_size;
 574
 575         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 576         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 577         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 578         tmp = SECTION_ALIGN_UP(tmp);
 579         tmp = VMALLOC_START - tmp * sizeof(struct page);
 580         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 581         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 582         vmemmap = (struct page *) tmp;
 583
 584         /* Take care that memory_end is set and <= vmemmap */
 585         memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
 586 #ifdef CONFIG_KASAN
 587         /* fit in kasan shadow memory region between 1:1 and vmemmap */
 588         memory_end = min(memory_end, KASAN_SHADOW_START);
 589         vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
 590 #endif
 591         max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 592         memblock_remove(memory_end, ULONG_MAX);
 593
 594         pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
 595 }
 596
 597 #ifdef CONFIG_CRASH_DUMP
 598
 599 /*
 600  * When kdump is enabled, we have to ensure that no memory from
 601  * the area [0 - crashkernel memory size] and
 602  * [crashk_res.start - crashk_res.end] is set offline.
 603  */
 604 static int kdump_mem_notifier(struct notifier_block *nb,
 605                               unsigned long action, void *data)
 606 {
 607         struct memory_notify *arg = data;
 608
 609         if (action != MEM_GOING_OFFLINE)
 610                 return NOTIFY_OK;
 611         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 612                 return NOTIFY_BAD;
 613         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 614                 return NOTIFY_OK;
 615         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 616                 return NOTIFY_OK;
 617         return NOTIFY_BAD;
 618 }
 619
 620 static struct notifier_block kdump_mem_nb = {
 621         .notifier_call = kdump_mem_notifier,
 622 };
 623
 624 #endif
 625
 626 /*
 627  * Make sure that the area behind memory_end is protected
 628  */
 629 static void reserve_memory_end(void)
 630 {
 631         if (memory_end_set)
 632                 memblock_reserve(memory_end, ULONG_MAX);
 633 }
 634
 635 /*
 636  * Make sure that oldmem, where the dump is stored, is protected
 637  */
 638 static void reserve_oldmem(void)
 639 {
 640 #ifdef CONFIG_CRASH_DUMP
 641         if (OLDMEM_BASE)
 642                 /* Forget all memory above the running kdump system */
 643                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 644 #endif
 645 }
 646
 647 /*
 648  * Make sure that oldmem, where the dump is stored, is protected
 649  */
 650 static void remove_oldmem(void)
 651 {
 652 #ifdef CONFIG_CRASH_DUMP
 653         if (OLDMEM_BASE)
 654                 /* Forget all memory above the running kdump system */
 655                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 656 #endif
 657 }
 658
 659 /*
 660  * Reserve memory for kdump kernel to be loaded with kexec
 661  */
 662 static void __init reserve_crashkernel(void)
 663 {
 664 #ifdef CONFIG_CRASH_DUMP
 665         unsigned long long crash_base, crash_size;
 666         phys_addr_t low, high;
 667         int rc;
 668
 669         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 670                                &crash_base);
 671
 672         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 673         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 674         if (rc || crash_size == 0)
 675                 return;
 676
 677         if (memblock.memory.regions[0].size < crash_size) {
 678                 pr_info("crashkernel reservation failed: %s\n",
 679                         "first memory chunk must be at least crashkernel size");
 680                 return;
 681         }
 682
 683         low = crash_base ?: OLDMEM_BASE;
 684         high = low + crash_size;
 685         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 686                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 687                 crash_base = low;
 688         } else {
 689                 /* Find suitable area in free memory */
 690                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 691                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 692
 693                 if (crash_base && crash_base < low) {
 694                         pr_info("crashkernel reservation failed: %s\n",
 695                                 "crash_base too low");
 696                         return;
 697                 }
 698                 low = crash_base ?: low;
 699                 crash_base = memblock_find_in_range(low, high, crash_size,
 700                                                     KEXEC_CRASH_MEM_ALIGN);
 701         }
 702
 703         if (!crash_base) {
 704                 pr_info("crashkernel reservation failed: %s\n",
 705                         "no suitable area found");
 706                 return;
 707         }
 708
 709         if (register_memory_notifier(&kdump_mem_nb))
 710                 return;
 711
 712         if (!OLDMEM_BASE && MACHINE_IS_VM)
 713                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 714         crashk_res.start = crash_base;
 715         crashk_res.end = crash_base + crash_size - 1;
 716         memblock_remove(crash_base, crash_size);
 717         pr_info("Reserving %lluMB of memory at %lluMB "
 718                 "for crashkernel (System RAM: %luMB)\n",
 719                 crash_size >> 20, crash_base >> 20,
 720                 (unsigned long)memblock.memory.total_size >> 20);
 721         os_info_crashkernel_add(crash_base, crash_size);
 722 #endif
 723 }
 724
 725 /*
 726  * Reserve the initrd from being used by memblock
 727  */
 728 static void __init reserve_initrd(void)
 729 {
 730 #ifdef CONFIG_BLK_DEV_INITRD
 731         if (!INITRD_START || !INITRD_SIZE)
 732                 return;
 733         initrd_start = INITRD_START;
 734         initrd_end = initrd_start + INITRD_SIZE;
 735         memblock_reserve(INITRD_START, INITRD_SIZE);
 736 #endif
 737 }
 738
 739 /*
 740  * Reserve the memory area used to pass the certificate lists
 741  */
 742 static void __init reserve_certificate_list(void)
 743 {
 744         if (ipl_cert_list_addr)
 745                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
 746 }
 747
 748 static void __init reserve_mem_detect_info(void)
 749 {
 750         unsigned long start, size;
 751
 752         get_mem_detect_reserved(&start, &size);
 753         if (size)
 754                 memblock_reserve(start, size);
 755 }
 756
 757 static void __init free_mem_detect_info(void)
 758 {
 759         unsigned long start, size;
 760
 761         get_mem_detect_reserved(&start, &size);
 762         if (size)
 763                 memblock_free(start, size);
 764 }
 765
 766 static const char * __init get_mem_info_source(void)
 767 {
 768         switch (mem_detect.info_source) {
 769         case MEM_DETECT_SCLP_STOR_INFO:
 770                 return "sclp storage info";
 771         case MEM_DETECT_DIAG260:
 772                 return "diag260";
 773         case MEM_DETECT_SCLP_READ_INFO:
 774                 return "sclp read info";
 775         case MEM_DETECT_BIN_SEARCH:
 776                 return "binary search";
 777         }
 778         return "none";
 779 }
 780
 781 static void __init memblock_add_mem_detect_info(void)
 782 {
 783         unsigned long start, end;
 784         int i;
 785
 786         memblock_dbg("physmem info source: %s (%hhd)\n",
 787                      get_mem_info_source(), mem_detect.info_source);
 788         /* keep memblock lists close to the kernel */
 789         memblock_set_bottom_up(true);
 790         for_each_mem_detect_block(i, &start, &end) {
 791                 memblock_add(start, end - start);
 792                 memblock_physmem_add(start, end - start);
 793         }
 794         memblock_set_bottom_up(false);
 795         memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
 796         memblock_dump_all();
 797 }
 798
 799 /*
 800  * Check for initrd being in usable memory
 801  */
 802 static void __init check_initrd(void)
 803 {
 804 #ifdef CONFIG_BLK_DEV_INITRD
 805         if (INITRD_START && INITRD_SIZE &&
 806             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 807                 pr_err("The initial RAM disk does not fit into the memory\n");
 808                 memblock_free(INITRD_START, INITRD_SIZE);
 809                 initrd_start = initrd_end = 0;
 810         }
 811 #endif
 812 }
 813
 814 /*
 815  * Reserve memory used for lowcore/command line/kernel image.
 816  */
 817 static void __init reserve_kernel(void)
 818 {
 819         unsigned long start_pfn = PFN_UP(__pa(_end));
 820
 821         memblock_reserve(0, HEAD_END);
 822         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 823                          - (unsigned long)_stext);
 824         memblock_reserve(__sdma, __edma - __sdma);
 825 }
 826
 827 static void __init setup_memory(void)
 828 {
 829         struct memblock_region *reg;
 830
 831         /*
 832          * Init storage key for present memory
 833          */
 834         for_each_memblock(memory, reg) {
 835                 storage_key_init_range(reg->base, reg->base + reg->size);
 836         }
 837         psw_set_key(PAGE_DEFAULT_KEY);
 838
 839         /* Only cosmetics */
 840         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 841 }
 842
 843 /*
 844  * Setup hardware capabilities.
 845  */
 846 static int __init setup_hwcaps(void)
 847 {
 848         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 849         struct cpuid cpu_id;
 850         int i;
 851
 852         /*
 853          * The store facility list bits numbers as found in the principles
 854          * of operation are numbered with bit 1UL<<31 as number 0 to
 855          * bit 1UL<<0 as number 31.
 856          *   Bit 0: instructions named N3, "backported" to esa-mode
 857          *   Bit 2: z/Architecture mode is active
 858          *   Bit 7: the store-facility-list-extended facility is installed
 859          *   Bit 17: the message-security assist is installed
 860          *   Bit 19: the long-displacement facility is installed
 861          *   Bit 21: the extended-immediate facility is installed
 862          *   Bit 22: extended-translation facility 3 is installed
 863          *   Bit 30: extended-translation facility 3 enhancement facility
 864          * These get translated to:
 865          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 866          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 867          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 868          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 869          */
 870         for (i = 0; i < 6; i++)
 871                 if (test_facility(stfl_bits[i]))
 872                         elf_hwcap |= 1UL << i;
 873
 874         if (test_facility(22) && test_facility(30))
 875                 elf_hwcap |= HWCAP_S390_ETF3EH;
 876
 877         /*
 878          * Check for additional facilities with store-facility-list-extended.
 879          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 880          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 881          * as stored by stfl, bits 32-xxx contain additional facilities.
 882          * How many facility words are stored depends on the number of
 883          * doublewords passed to the instruction. The additional facilities
 884          * are:
 885          *   Bit 42: decimal floating point facility is installed
 886          *   Bit 44: perform floating point operation facility is installed
 887          * translated to:
 888          *   HWCAP_S390_DFP bit 6 (42 && 44).
 889          */
 890         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 891                 elf_hwcap |= HWCAP_S390_DFP;
 892
 893         /*
 894          * Huge page support HWCAP_S390_HPAGE is bit 7.
 895          */
 896         if (MACHINE_HAS_EDAT1)
 897                 elf_hwcap |= HWCAP_S390_HPAGE;
 898
 899         /*
 900          * 64-bit register support for 31-bit processes
 901          * HWCAP_S390_HIGH_GPRS is bit 9.
 902          */
 903         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 904
 905         /*
 906          * Transactional execution support HWCAP_S390_TE is bit 10.
 907          */
 908         if (MACHINE_HAS_TE)
 909                 elf_hwcap |= HWCAP_S390_TE;
 910
 911         /*
 912          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 913          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 914          * instead of facility bit 129.
 915          */
 916         if (MACHINE_HAS_VX) {
 917                 elf_hwcap |= HWCAP_S390_VXRS;
 918                 if (test_facility(134))
 919                         elf_hwcap |= HWCAP_S390_VXRS_EXT;
 920                 if (test_facility(135))
 921                         elf_hwcap |= HWCAP_S390_VXRS_BCD;
 922                 if (test_facility(148))
 923                         elf_hwcap |= HWCAP_S390_VXRS_EXT2;
 924                 if (test_facility(152))
 925                         elf_hwcap |= HWCAP_S390_VXRS_PDE;
 926         }
 927         if (test_facility(150))
 928                 elf_hwcap |= HWCAP_S390_SORT;
 929         if (test_facility(151))
 930                 elf_hwcap |= HWCAP_S390_DFLT;
 931
 932         /*
 933          * Guarded storage support HWCAP_S390_GS is bit 12.
 934          */
 935         if (MACHINE_HAS_GS)
 936                 elf_hwcap |= HWCAP_S390_GS;
 937
 938         get_cpu_id(&cpu_id);
 939         add_device_randomness(&cpu_id, sizeof(cpu_id));
 940         switch (cpu_id.machine) {
 941         case 0x2064:
 942         case 0x2066:
 943         default:        /* Use "z900" as default for 64 bit kernels. */
 944                 strcpy(elf_platform, "z900");
 945                 break;
 946         case 0x2084:
 947         case 0x2086:
 948                 strcpy(elf_platform, "z990");
 949                 break;
 950         case 0x2094:
 951         case 0x2096:
 952                 strcpy(elf_platform, "z9-109");
 953                 break;
 954         case 0x2097:
 955         case 0x2098:
 956                 strcpy(elf_platform, "z10");
 957                 break;
 958         case 0x2817:
 959         case 0x2818:
 960                 strcpy(elf_platform, "z196");
 961                 break;
 962         case 0x2827:
 963         case 0x2828:
 964                 strcpy(elf_platform, "zEC12");
 965                 break;
 966         case 0x2964:
 967         case 0x2965:
 968                 strcpy(elf_platform, "z13");
 969                 break;
 970         case 0x3906:
 971         case 0x3907:
 972                 strcpy(elf_platform, "z14");
 973                 break;
 974         case 0x8561:
 975         case 0x8562:
 976                 strcpy(elf_platform, "z15");
 977                 break;
 978         }
 979
 980         /*
 981          * Virtualization support HWCAP_INT_SIE is bit 0.
 982          */
 983         if (sclp.has_sief2)
 984                 int_hwcap |= HWCAP_INT_SIE;
 985
 986         return 0;
 987 }
 988 arch_initcall(setup_hwcaps);
 989
 990 /*
 991  * Add system information as device randomness
 992  */
 993 static void __init setup_randomness(void)
 994 {
 995         struct sysinfo_3_2_2 *vmms;
 996
 997         vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
 998                                                             PAGE_SIZE);
 999         if (!vmms)
1000                 panic("Failed to allocate memory for sysinfo structure\n");
1001
1002         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1003                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1004         memblock_free((unsigned long) vmms, PAGE_SIZE);
1005 }
1006
1007 /*
1008  * Find the correct size for the task_struct. This depends on
1009  * the size of the struct fpu at the end of the thread_struct
1010  * which is embedded in the task_struct.
1011  */
1012 static void __init setup_task_size(void)
1013 {
1014         int task_size = sizeof(struct task_struct);
1015
1016         if (!MACHINE_HAS_VX) {
1017                 task_size -= sizeof(__vector128) * __NUM_VXRS;
1018                 task_size += sizeof(freg_t) * __NUM_FPRS;
1019         }
1020         arch_task_struct_size = task_size;
1021 }
1022
1023 /*
1024  * Issue diagnose 318 to set the control program name and
1025  * version codes.
1026  */
1027 static void __init setup_control_program_code(void)
1028 {
1029         union diag318_info diag318_info = {
1030                 .cpnc = CPNC_LINUX,
1031                 .cpvc_linux = 0,
1032                 .cpvc_distro = {0},
1033         };
1034
1035         if (!sclp.has_diag318)
1036                 return;
1037
1038         diag_stat_inc(DIAG_STAT_X318);
1039         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1040 }
1041
1042 /*
1043  * Print the component list from the IPL report
1044  */
1045 static void __init log_component_list(void)
1046 {
1047         struct ipl_rb_component_entry *ptr, *end;
1048         char *str;
1049
1050         if (!early_ipl_comp_list_addr)
1051                 return;
1052         if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1053                 pr_info("Linux is running with Secure-IPL enabled\n");
1054         else
1055                 pr_info("Linux is running with Secure-IPL disabled\n");
1056         ptr = (void *) early_ipl_comp_list_addr;
1057         end = (void *) ptr + early_ipl_comp_list_size;
1058         pr_info("The IPL report contains the following components:\n");
1059         while (ptr < end) {
1060                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1061                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1062                                 str = "signed, verified";
1063                         else
1064                                 str = "signed, verification failed";
1065                 } else {
1066                         str = "not signed";
1067                 }
1068                 pr_info("%016llx - %016llx (%s)\n",
1069                         ptr->addr, ptr->addr + ptr->len, str);
1070                 ptr++;
1071         }
1072 }
1073
1074 /*
1075  * Setup function called from init/main.c just after the banner
1076  * was printed.
1077  */
1078
1079 void __init setup_arch(char **cmdline_p)
1080 {
1081         /*
1082          * print what head.S has found out about the machine
1083          */
1084         if (MACHINE_IS_VM)
1085                 pr_info("Linux is running as a z/VM "
1086                         "guest operating system in 64-bit mode\n");
1087         else if (MACHINE_IS_KVM)
1088                 pr_info("Linux is running under KVM in 64-bit mode\n");
1089         else if (MACHINE_IS_LPAR)
1090                 pr_info("Linux is running natively in 64-bit mode\n");
1091         else
1092                 pr_info("Linux is running as a guest in 64-bit mode\n");
1093
1094         log_component_list();
1095
1096         /* Have one command line that is parsed and saved in /proc/cmdline */
1097         /* boot_command_line has been already set up in early.c */
1098         *cmdline_p = boot_command_line;
1099
1100         ROOT_DEV = Root_RAM0;
1101
1102         init_mm.start_code = (unsigned long) _text;
1103         init_mm.end_code = (unsigned long) _etext;
1104         init_mm.end_data = (unsigned long) _edata;
1105         init_mm.brk = (unsigned long) _end;
1106
1107         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1108                 nospec_auto_detect();
1109
1110         parse_early_param();
1111 #ifdef CONFIG_CRASH_DUMP
1112         /* Deactivate elfcorehdr= kernel parameter */
1113         elfcorehdr_addr = ELFCORE_ADDR_MAX;
1114 #endif
1115
1116         os_info_init();
1117         setup_ipl();
1118         setup_task_size();
1119         setup_control_program_code();
1120
1121         /* Do some memory reservations *before* memory is added to memblock */
1122         reserve_memory_end();
1123         reserve_oldmem();
1124         reserve_kernel();
1125         reserve_initrd();
1126         reserve_certificate_list();
1127         reserve_mem_detect_info();
1128         memblock_allow_resize();
1129
1130         /* Get information about *all* installed memory */
1131         memblock_add_mem_detect_info();
1132
1133         free_mem_detect_info();
1134         remove_oldmem();
1135
1136         /*
1137          * Make sure all chunks are MAX_ORDER aligned so we don't need the
1138          * extra checks that HOLES_IN_ZONE would require.
1139          *
1140          * Is this still required?
1141          */
1142         memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
1143
1144         if (is_prot_virt_host())
1145                 setup_uv();
1146         setup_memory_end();
1147         setup_memory();
1148         dma_contiguous_reserve(memory_end);
1149         vmcp_cma_reserve();
1150
1151         check_initrd();
1152         reserve_crashkernel();
1153 #ifdef CONFIG_CRASH_DUMP
1154         /*
1155          * Be aware that smp_save_dump_cpus() triggers a system reset.
1156          * Therefore CPU and device initialization should be done afterwards.
1157          */
1158         smp_save_dump_cpus();
1159 #endif
1160
1161         setup_resources();
1162         setup_lowcore_dat_off();
1163         smp_fill_possible_mask();
1164         cpu_detect_mhz_feature();
1165         cpu_init();
1166         numa_setup();
1167         smp_detect_cpus();
1168         topology_init_early();
1169
1170         /*
1171          * Create kernel page tables and switch to virtual addressing.
1172          */
1173         paging_init();
1174
1175         /*
1176          * After paging_init created the kernel page table, the new PSWs
1177          * in lowcore can now run with DAT enabled.
1178          */
1179         setup_lowcore_dat_on();
1180
1181         /* Setup default console */
1182         conmode_default();
1183         set_preferred_console();
1184
1185         apply_alternative_instructions();
1186         if (IS_ENABLED(CONFIG_EXPOLINE))
1187                 nospec_init_branches();
1188
1189         /* Setup zfcpdump support */
1190         setup_zfcpdump();
1191
1192         /* Add system specific data to the random pool */
1193         setup_randomness();
1194 }