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/bootmem.h>
  38 #include <linux/root_dev.h>
  39 #include <linux/console.h>
  40 #include <linux/kernel_stat.h>
  41 #include <linux/dma-contiguous.h>
  42 #include <linux/device.h>
  43 #include <linux/notifier.h>
  44 #include <linux/pfn.h>
  45 #include <linux/ctype.h>
  46 #include <linux/reboot.h>
  47 #include <linux/topology.h>
  48 #include <linux/kexec.h>
  49 #include <linux/crash_dump.h>
  50 #include <linux/memory.h>
  51 #include <linux/compat.h>
  52
  53 #include <asm/ipl.h>
  54 #include <asm/facility.h>
  55 #include <asm/smp.h>
  56 #include <asm/mmu_context.h>
  57 #include <asm/cpcmd.h>
  58 #include <asm/lowcore.h>
  59 #include <asm/nmi.h>
  60 #include <asm/irq.h>
  61 #include <asm/page.h>
  62 #include <asm/ptrace.h>
  63 #include <asm/sections.h>
  64 #include <asm/ebcdic.h>
  65 #include <asm/diag.h>
  66 #include <asm/os_info.h>
  67 #include <asm/sclp.h>
  68 #include <asm/sysinfo.h>
  69 #include <asm/numa.h>
  70 #include <asm/alternative.h>
  71 #include <asm/nospec-branch.h>
  72 #include "entry.h"
  73
  74 /*
  75  * Machine setup..
  76  */
  77 unsigned int console_mode = 0;
  78 EXPORT_SYMBOL(console_mode);
  79
  80 unsigned int console_devno = -1;
  81 EXPORT_SYMBOL(console_devno);
  82
  83 unsigned int console_irq = -1;
  84 EXPORT_SYMBOL(console_irq);
  85
  86 unsigned long elf_hwcap __read_mostly = 0;
  87 char elf_platform[ELF_PLATFORM_SIZE];
  88
  89 unsigned long int_hwcap = 0;
  90
  91 int __initdata memory_end_set;
  92 unsigned long __initdata memory_end;
  93 unsigned long __initdata max_physmem_end;
  94
  95 unsigned long VMALLOC_START;
  96 EXPORT_SYMBOL(VMALLOC_START);
  97
  98 unsigned long VMALLOC_END;
  99 EXPORT_SYMBOL(VMALLOC_END);
 100
 101 struct page *vmemmap;
 102 EXPORT_SYMBOL(vmemmap);
 103
 104 unsigned long MODULES_VADDR;
 105 unsigned long MODULES_END;
 106
 107 /* An array with a pointer to the lowcore of every CPU. */
 108 struct lowcore *lowcore_ptr[NR_CPUS];
 109 EXPORT_SYMBOL(lowcore_ptr);
 110
 111 /*
 112  * This is set up by the setup-routine at boot-time
 113  * for S390 need to find out, what we have to setup
 114  * using address 0x10400 ...
 115  */
 116
 117 #include <asm/setup.h>
 118
 119 /*
 120  * condev= and conmode= setup parameter.
 121  */
 122
 123 static int __init condev_setup(char *str)
 124 {
 125         int vdev;
 126
 127         vdev = simple_strtoul(str, &str, 0);
 128         if (vdev >= 0 && vdev < 65536) {
 129                 console_devno = vdev;
 130                 console_irq = -1;
 131         }
 132         return 1;
 133 }
 134
 135 __setup("condev=", condev_setup);
 136
 137 static void __init set_preferred_console(void)
 138 {
 139         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 140                 add_preferred_console("ttyS", 0, NULL);
 141         else if (CONSOLE_IS_3270)
 142                 add_preferred_console("tty3270", 0, NULL);
 143         else if (CONSOLE_IS_VT220)
 144                 add_preferred_console("ttyS", 1, NULL);
 145         else if (CONSOLE_IS_HVC)
 146                 add_preferred_console("hvc", 0, NULL);
 147 }
 148
 149 static int __init conmode_setup(char *str)
 150 {
 151 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 152         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 153                 SET_CONSOLE_SCLP;
 154 #endif
 155 #if defined(CONFIG_TN3215_CONSOLE)
 156         if (strncmp(str, "3215", 5) == 0)
 157                 SET_CONSOLE_3215;
 158 #endif
 159 #if defined(CONFIG_TN3270_CONSOLE)
 160         if (strncmp(str, "3270", 5) == 0)
 161                 SET_CONSOLE_3270;
 162 #endif
 163         set_preferred_console();
 164         return 1;
 165 }
 166
 167 __setup("conmode=", conmode_setup);
 168
 169 static void __init conmode_default(void)
 170 {
 171         char query_buffer[1024];
 172         char *ptr;
 173
 174         if (MACHINE_IS_VM) {
 175                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 176                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 177                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 178                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 179                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 180                 ptr = strstr(query_buffer, "CONMODE");
 181                 /*
 182                  * Set the conmode to 3215 so that the device recognition
 183                  * will set the cu_type of the console to 3215. If the
 184                  * conmode is 3270 and we don't set it back then both
 185                  * 3215 and the 3270 driver will try to access the console
 186                  * device (3215 as console and 3270 as normal tty).
 187                  */
 188                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 189                 if (ptr == NULL) {
 190 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 191                         SET_CONSOLE_SCLP;
 192 #endif
 193                         return;
 194                 }
 195                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 196 #if defined(CONFIG_TN3270_CONSOLE)
 197                         SET_CONSOLE_3270;
 198 #elif defined(CONFIG_TN3215_CONSOLE)
 199                         SET_CONSOLE_3215;
 200 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 201                         SET_CONSOLE_SCLP;
 202 #endif
 203                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 204 #if defined(CONFIG_TN3215_CONSOLE)
 205                         SET_CONSOLE_3215;
 206 #elif defined(CONFIG_TN3270_CONSOLE)
 207                         SET_CONSOLE_3270;
 208 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 209                         SET_CONSOLE_SCLP;
 210 #endif
 211                 }
 212         } else if (MACHINE_IS_KVM) {
 213                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
 214                         SET_CONSOLE_VT220;
 215                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
 216                         SET_CONSOLE_SCLP;
 217                 else
 218                         SET_CONSOLE_HVC;
 219         } else {
 220 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 221                 SET_CONSOLE_SCLP;
 222 #endif
 223         }
 224         if (IS_ENABLED(CONFIG_VT) && IS_ENABLED(CONFIG_DUMMY_CONSOLE))
 225                 conswitchp = &dummy_con;
 226 }
 227
 228 #ifdef CONFIG_CRASH_DUMP
 229 static void __init setup_zfcpdump(void)
 230 {
 231         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 232                 return;
 233         if (OLDMEM_BASE)
 234                 return;
 235         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 236         console_loglevel = 2;
 237 }
 238 #else
 239 static inline void setup_zfcpdump(void) {}
 240 #endif /* CONFIG_CRASH_DUMP */
 241
 242  /*
 243  * Reboot, halt and power_off stubs. They just call _machine_restart,
 244  * _machine_halt or _machine_power_off.
 245  */
 246
 247 void machine_restart(char *command)
 248 {
 249         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 250                 /*
 251                  * Only unblank the console if we are called in enabled
 252                  * context or a bust_spinlocks cleared the way for us.
 253                  */
 254                 console_unblank();
 255         _machine_restart(command);
 256 }
 257
 258 void machine_halt(void)
 259 {
 260         if (!in_interrupt() || oops_in_progress)
 261                 /*
 262                  * Only unblank the console if we are called in enabled
 263                  * context or a bust_spinlocks cleared the way for us.
 264                  */
 265                 console_unblank();
 266         _machine_halt();
 267 }
 268
 269 void machine_power_off(void)
 270 {
 271         if (!in_interrupt() || oops_in_progress)
 272                 /*
 273                  * Only unblank the console if we are called in enabled
 274                  * context or a bust_spinlocks cleared the way for us.
 275                  */
 276                 console_unblank();
 277         _machine_power_off();
 278 }
 279
 280 /*
 281  * Dummy power off function.
 282  */
 283 void (*pm_power_off)(void) = machine_power_off;
 284 EXPORT_SYMBOL_GPL(pm_power_off);
 285
 286 static int __init early_parse_mem(char *p)
 287 {
 288         memory_end = memparse(p, &p);
 289         memory_end &= PAGE_MASK;
 290         memory_end_set = 1;
 291         return 0;
 292 }
 293 early_param("mem", early_parse_mem);
 294
 295 static int __init parse_vmalloc(char *arg)
 296 {
 297         if (!arg)
 298                 return -EINVAL;
 299         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
 300         return 0;
 301 }
 302 early_param("vmalloc", parse_vmalloc);
 303
 304 void *restart_stack __section(.data);
 305
 306 static void __init setup_lowcore_dat_off(void)
 307 {
 308         struct lowcore *lc;
 309
 310         /*
 311          * Setup lowcore for boot cpu
 312          */
 313         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 314         lc = memblock_virt_alloc_low(sizeof(*lc), sizeof(*lc));
 315         lc->restart_psw.mask = PSW_KERNEL_BITS;
 316         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 317         lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 318         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 319         lc->svc_new_psw.mask = PSW_KERNEL_BITS |
 320                 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 321         lc->svc_new_psw.addr = (unsigned long) system_call;
 322         lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 323         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 324         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 325         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 326         lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 327         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 328         lc->clock_comparator = clock_comparator_max;
 329         lc->kernel_stack = ((unsigned long) &init_thread_union)
 330                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 331         lc->async_stack = (unsigned long)
 332                 memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE)
 333                 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 334         lc->panic_stack = (unsigned long)
 335                 memblock_virt_alloc(PAGE_SIZE, PAGE_SIZE)
 336                 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 337         lc->current_task = (unsigned long)&init_task;
 338         lc->lpp = LPP_MAGIC;
 339         lc->machine_flags = S390_lowcore.machine_flags;
 340         lc->preempt_count = S390_lowcore.preempt_count;
 341         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 342         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 343                sizeof(lc->stfle_fac_list));
 344         memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
 345                sizeof(lc->alt_stfle_fac_list));
 346         nmi_alloc_boot_cpu(lc);
 347         vdso_alloc_boot_cpu(lc);
 348         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 349         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 350         lc->exit_timer = S390_lowcore.exit_timer;
 351         lc->user_timer = S390_lowcore.user_timer;
 352         lc->system_timer = S390_lowcore.system_timer;
 353         lc->steal_timer = S390_lowcore.steal_timer;
 354         lc->last_update_timer = S390_lowcore.last_update_timer;
 355         lc->last_update_clock = S390_lowcore.last_update_clock;
 356
 357         restart_stack = memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE);
 358         restart_stack += ASYNC_SIZE;
 359
 360         /*
 361          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 362          * restart data to the absolute zero lowcore. This is necessary if
 363          * PSW restart is done on an offline CPU that has lowcore zero.
 364          */
 365         lc->restart_stack = (unsigned long) restart_stack;
 366         lc->restart_fn = (unsigned long) do_restart;
 367         lc->restart_data = 0;
 368         lc->restart_source = -1UL;
 369
 370         /* Setup absolute zero lowcore */
 371         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 372         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 373         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 374         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 375         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 376
 377 #ifdef CONFIG_SMP
 378         lc->spinlock_lockval = arch_spin_lockval(0);
 379         lc->spinlock_index = 0;
 380         arch_spin_lock_setup(0);
 381 #endif
 382         lc->br_r1_trampoline = 0x07f1;  /* br %r1 */
 383
 384         set_prefix((u32)(unsigned long) lc);
 385         lowcore_ptr[0] = lc;
 386 }
 387
 388 static void __init setup_lowcore_dat_on(void)
 389 {
 390         __ctl_clear_bit(0, 28);
 391         S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
 392         S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
 393         S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
 394         S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
 395         __ctl_set_bit(0, 28);
 396 }
 397
 398 static struct resource code_resource = {
 399         .name  = "Kernel code",
 400         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 401 };
 402
 403 static struct resource data_resource = {
 404         .name = "Kernel data",
 405         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 406 };
 407
 408 static struct resource bss_resource = {
 409         .name = "Kernel bss",
 410         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 411 };
 412
 413 static struct resource __initdata *standard_resources[] = {
 414         &code_resource,
 415         &data_resource,
 416         &bss_resource,
 417 };
 418
 419 static void __init setup_resources(void)
 420 {
 421         struct resource *res, *std_res, *sub_res;
 422         struct memblock_region *reg;
 423         int j;
 424
 425         code_resource.start = (unsigned long) _text;
 426         code_resource.end = (unsigned long) _etext - 1;
 427         data_resource.start = (unsigned long) _etext;
 428         data_resource.end = (unsigned long) _edata - 1;
 429         bss_resource.start = (unsigned long) __bss_start;
 430         bss_resource.end = (unsigned long) __bss_stop - 1;
 431
 432         for_each_memblock(memory, reg) {
 433                 res = memblock_virt_alloc(sizeof(*res), 8);
 434                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 435
 436                 res->name = "System RAM";
 437                 res->start = reg->base;
 438                 res->end = reg->base + reg->size - 1;
 439                 request_resource(&iomem_resource, res);
 440
 441                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 442                         std_res = standard_resources[j];
 443                         if (std_res->start < res->start ||
 444                             std_res->start > res->end)
 445                                 continue;
 446                         if (std_res->end > res->end) {
 447                                 sub_res = memblock_virt_alloc(sizeof(*sub_res), 8);
 448                                 *sub_res = *std_res;
 449                                 sub_res->end = res->end;
 450                                 std_res->start = res->end + 1;
 451                                 request_resource(res, sub_res);
 452                         } else {
 453                                 request_resource(res, std_res);
 454                         }
 455                 }
 456         }
 457 #ifdef CONFIG_CRASH_DUMP
 458         /*
 459          * Re-add removed crash kernel memory as reserved memory. This makes
 460          * sure it will be mapped with the identity mapping and struct pages
 461          * will be created, so it can be resized later on.
 462          * However add it later since the crash kernel resource should not be
 463          * part of the System RAM resource.
 464          */
 465         if (crashk_res.end) {
 466                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 467                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 468                 insert_resource(&iomem_resource, &crashk_res);
 469         }
 470 #endif
 471 }
 472
 473 static void __init setup_memory_end(void)
 474 {
 475         unsigned long vmax, vmalloc_size, tmp;
 476
 477         /* Choose kernel address space layout: 2, 3, or 4 levels. */
 478         vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
 479         tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
 480         tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
 481         if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
 482                 vmax = _REGION2_SIZE; /* 3-level kernel page table */
 483         else
 484                 vmax = _REGION1_SIZE; /* 4-level kernel page table */
 485         /* module area is at the end of the kernel address space. */
 486         MODULES_END = vmax;
 487         MODULES_VADDR = MODULES_END - MODULES_LEN;
 488         VMALLOC_END = MODULES_VADDR;
 489         VMALLOC_START = vmax - vmalloc_size;
 490
 491         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 492         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 493         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 494         tmp = SECTION_ALIGN_UP(tmp);
 495         tmp = VMALLOC_START - tmp * sizeof(struct page);
 496         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 497         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 498         vmemmap = (struct page *) tmp;
 499
 500         /* Take care that memory_end is set and <= vmemmap */
 501         memory_end = min(memory_end ?: max_physmem_end, tmp);
 502         max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 503         memblock_remove(memory_end, ULONG_MAX);
 504
 505         pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
 506 }
 507
 508 #ifdef CONFIG_CRASH_DUMP
 509
 510 /*
 511  * When kdump is enabled, we have to ensure that no memory from
 512  * the area [0 - crashkernel memory size] and
 513  * [crashk_res.start - crashk_res.end] is set offline.
 514  */
 515 static int kdump_mem_notifier(struct notifier_block *nb,
 516                               unsigned long action, void *data)
 517 {
 518         struct memory_notify *arg = data;
 519
 520         if (action != MEM_GOING_OFFLINE)
 521                 return NOTIFY_OK;
 522         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 523                 return NOTIFY_BAD;
 524         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 525                 return NOTIFY_OK;
 526         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 527                 return NOTIFY_OK;
 528         return NOTIFY_BAD;
 529 }
 530
 531 static struct notifier_block kdump_mem_nb = {
 532         .notifier_call = kdump_mem_notifier,
 533 };
 534
 535 #endif
 536
 537 /*
 538  * Make sure that the area behind memory_end is protected
 539  */
 540 static void reserve_memory_end(void)
 541 {
 542 #ifdef CONFIG_CRASH_DUMP
 543         if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
 544             !OLDMEM_BASE && sclp.hsa_size) {
 545                 memory_end = sclp.hsa_size;
 546                 memory_end &= PAGE_MASK;
 547                 memory_end_set = 1;
 548         }
 549 #endif
 550         if (!memory_end_set)
 551                 return;
 552         memblock_reserve(memory_end, ULONG_MAX);
 553 }
 554
 555 /*
 556  * Make sure that oldmem, where the dump is stored, is protected
 557  */
 558 static void reserve_oldmem(void)
 559 {
 560 #ifdef CONFIG_CRASH_DUMP
 561         if (OLDMEM_BASE)
 562                 /* Forget all memory above the running kdump system */
 563                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 564 #endif
 565 }
 566
 567 /*
 568  * Make sure that oldmem, where the dump is stored, is protected
 569  */
 570 static void remove_oldmem(void)
 571 {
 572 #ifdef CONFIG_CRASH_DUMP
 573         if (OLDMEM_BASE)
 574                 /* Forget all memory above the running kdump system */
 575                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 576 #endif
 577 }
 578
 579 /*
 580  * Reserve memory for kdump kernel to be loaded with kexec
 581  */
 582 static void __init reserve_crashkernel(void)
 583 {
 584 #ifdef CONFIG_CRASH_DUMP
 585         unsigned long long crash_base, crash_size;
 586         phys_addr_t low, high;
 587         int rc;
 588
 589         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 590                                &crash_base);
 591
 592         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 593         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 594         if (rc || crash_size == 0)
 595                 return;
 596
 597         if (memblock.memory.regions[0].size < crash_size) {
 598                 pr_info("crashkernel reservation failed: %s\n",
 599                         "first memory chunk must be at least crashkernel size");
 600                 return;
 601         }
 602
 603         low = crash_base ?: OLDMEM_BASE;
 604         high = low + crash_size;
 605         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 606                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 607                 crash_base = low;
 608         } else {
 609                 /* Find suitable area in free memory */
 610                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 611                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 612
 613                 if (crash_base && crash_base < low) {
 614                         pr_info("crashkernel reservation failed: %s\n",
 615                                 "crash_base too low");
 616                         return;
 617                 }
 618                 low = crash_base ?: low;
 619                 crash_base = memblock_find_in_range(low, high, crash_size,
 620                                                     KEXEC_CRASH_MEM_ALIGN);
 621         }
 622
 623         if (!crash_base) {
 624                 pr_info("crashkernel reservation failed: %s\n",
 625                         "no suitable area found");
 626                 return;
 627         }
 628
 629         if (register_memory_notifier(&kdump_mem_nb))
 630                 return;
 631
 632         if (!OLDMEM_BASE && MACHINE_IS_VM)
 633                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 634         crashk_res.start = crash_base;
 635         crashk_res.end = crash_base + crash_size - 1;
 636         memblock_remove(crash_base, crash_size);
 637         pr_info("Reserving %lluMB of memory at %lluMB "
 638                 "for crashkernel (System RAM: %luMB)\n",
 639                 crash_size >> 20, crash_base >> 20,
 640                 (unsigned long)memblock.memory.total_size >> 20);
 641         os_info_crashkernel_add(crash_base, crash_size);
 642 #endif
 643 }
 644
 645 /*
 646  * Reserve the initrd from being used by memblock
 647  */
 648 static void __init reserve_initrd(void)
 649 {
 650 #ifdef CONFIG_BLK_DEV_INITRD
 651         if (!INITRD_START || !INITRD_SIZE)
 652                 return;
 653         initrd_start = INITRD_START;
 654         initrd_end = initrd_start + INITRD_SIZE;
 655         memblock_reserve(INITRD_START, INITRD_SIZE);
 656 #endif
 657 }
 658
 659 /*
 660  * Check for initrd being in usable memory
 661  */
 662 static void __init check_initrd(void)
 663 {
 664 #ifdef CONFIG_BLK_DEV_INITRD
 665         if (INITRD_START && INITRD_SIZE &&
 666             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 667                 pr_err("The initial RAM disk does not fit into the memory\n");
 668                 memblock_free(INITRD_START, INITRD_SIZE);
 669                 initrd_start = initrd_end = 0;
 670         }
 671 #endif
 672 }
 673
 674 /*
 675  * Reserve memory used for lowcore/command line/kernel image.
 676  */
 677 static void __init reserve_kernel(void)
 678 {
 679         unsigned long start_pfn = PFN_UP(__pa(_end));
 680
 681 #ifdef CONFIG_DMA_API_DEBUG
 682         /*
 683          * DMA_API_DEBUG code stumbles over addresses from the
 684          * range [PARMAREA_END, _stext]. Mark the memory as reserved
 685          * so it is not used for CONFIG_DMA_API_DEBUG=y.
 686          */
 687         memblock_reserve(0, PFN_PHYS(start_pfn));
 688 #else
 689         memblock_reserve(0, PARMAREA_END);
 690         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 691                          - (unsigned long)_stext);
 692 #endif
 693 }
 694
 695 static void __init setup_memory(void)
 696 {
 697         struct memblock_region *reg;
 698
 699         /*
 700          * Init storage key for present memory
 701          */
 702         for_each_memblock(memory, reg) {
 703                 storage_key_init_range(reg->base, reg->base + reg->size);
 704         }
 705         psw_set_key(PAGE_DEFAULT_KEY);
 706
 707         /* Only cosmetics */
 708         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 709 }
 710
 711 /*
 712  * Setup hardware capabilities.
 713  */
 714 static int __init setup_hwcaps(void)
 715 {
 716         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 717         struct cpuid cpu_id;
 718         int i;
 719
 720         /*
 721          * The store facility list bits numbers as found in the principles
 722          * of operation are numbered with bit 1UL<<31 as number 0 to
 723          * bit 1UL<<0 as number 31.
 724          *   Bit 0: instructions named N3, "backported" to esa-mode
 725          *   Bit 2: z/Architecture mode is active
 726          *   Bit 7: the store-facility-list-extended facility is installed
 727          *   Bit 17: the message-security assist is installed
 728          *   Bit 19: the long-displacement facility is installed
 729          *   Bit 21: the extended-immediate facility is installed
 730          *   Bit 22: extended-translation facility 3 is installed
 731          *   Bit 30: extended-translation facility 3 enhancement facility
 732          * These get translated to:
 733          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 734          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 735          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 736          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 737          */
 738         for (i = 0; i < 6; i++)
 739                 if (test_facility(stfl_bits[i]))
 740                         elf_hwcap |= 1UL << i;
 741
 742         if (test_facility(22) && test_facility(30))
 743                 elf_hwcap |= HWCAP_S390_ETF3EH;
 744
 745         /*
 746          * Check for additional facilities with store-facility-list-extended.
 747          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 748          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 749          * as stored by stfl, bits 32-xxx contain additional facilities.
 750          * How many facility words are stored depends on the number of
 751          * doublewords passed to the instruction. The additional facilities
 752          * are:
 753          *   Bit 42: decimal floating point facility is installed
 754          *   Bit 44: perform floating point operation facility is installed
 755          * translated to:
 756          *   HWCAP_S390_DFP bit 6 (42 && 44).
 757          */
 758         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 759                 elf_hwcap |= HWCAP_S390_DFP;
 760
 761         /*
 762          * Huge page support HWCAP_S390_HPAGE is bit 7.
 763          */
 764         if (MACHINE_HAS_EDAT1)
 765                 elf_hwcap |= HWCAP_S390_HPAGE;
 766
 767         /*
 768          * 64-bit register support for 31-bit processes
 769          * HWCAP_S390_HIGH_GPRS is bit 9.
 770          */
 771         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 772
 773         /*
 774          * Transactional execution support HWCAP_S390_TE is bit 10.
 775          */
 776         if (MACHINE_HAS_TE)
 777                 elf_hwcap |= HWCAP_S390_TE;
 778
 779         /*
 780          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 781          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 782          * instead of facility bit 129.
 783          */
 784         if (MACHINE_HAS_VX) {
 785                 elf_hwcap |= HWCAP_S390_VXRS;
 786                 if (test_facility(134))
 787                         elf_hwcap |= HWCAP_S390_VXRS_EXT;
 788                 if (test_facility(135))
 789                         elf_hwcap |= HWCAP_S390_VXRS_BCD;
 790         }
 791
 792         /*
 793          * Guarded storage support HWCAP_S390_GS is bit 12.
 794          */
 795         if (MACHINE_HAS_GS)
 796                 elf_hwcap |= HWCAP_S390_GS;
 797
 798         get_cpu_id(&cpu_id);
 799         add_device_randomness(&cpu_id, sizeof(cpu_id));
 800         switch (cpu_id.machine) {
 801         case 0x2064:
 802         case 0x2066:
 803         default:        /* Use "z900" as default for 64 bit kernels. */
 804                 strcpy(elf_platform, "z900");
 805                 break;
 806         case 0x2084:
 807         case 0x2086:
 808                 strcpy(elf_platform, "z990");
 809                 break;
 810         case 0x2094:
 811         case 0x2096:
 812                 strcpy(elf_platform, "z9-109");
 813                 break;
 814         case 0x2097:
 815         case 0x2098:
 816                 strcpy(elf_platform, "z10");
 817                 break;
 818         case 0x2817:
 819         case 0x2818:
 820                 strcpy(elf_platform, "z196");
 821                 break;
 822         case 0x2827:
 823         case 0x2828:
 824                 strcpy(elf_platform, "zEC12");
 825                 break;
 826         case 0x2964:
 827         case 0x2965:
 828                 strcpy(elf_platform, "z13");
 829                 break;
 830         case 0x3906:
 831         case 0x3907:
 832                 strcpy(elf_platform, "z14");
 833                 break;
 834         }
 835
 836         /*
 837          * Virtualization support HWCAP_INT_SIE is bit 0.
 838          */
 839         if (sclp.has_sief2)
 840                 int_hwcap |= HWCAP_INT_SIE;
 841
 842         return 0;
 843 }
 844 arch_initcall(setup_hwcaps);
 845
 846 /*
 847  * Add system information as device randomness
 848  */
 849 static void __init setup_randomness(void)
 850 {
 851         struct sysinfo_3_2_2 *vmms;
 852
 853         vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE);
 854         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 855                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
 856         memblock_free((unsigned long) vmms, PAGE_SIZE);
 857 }
 858
 859 /*
 860  * Find the correct size for the task_struct. This depends on
 861  * the size of the struct fpu at the end of the thread_struct
 862  * which is embedded in the task_struct.
 863  */
 864 static void __init setup_task_size(void)
 865 {
 866         int task_size = sizeof(struct task_struct);
 867
 868         if (!MACHINE_HAS_VX) {
 869                 task_size -= sizeof(__vector128) * __NUM_VXRS;
 870                 task_size += sizeof(freg_t) * __NUM_FPRS;
 871         }
 872         arch_task_struct_size = task_size;
 873 }
 874
 875 /*
 876  * Setup function called from init/main.c just after the banner
 877  * was printed.
 878  */
 879
 880 void __init setup_arch(char **cmdline_p)
 881 {
 882         /*
 883          * print what head.S has found out about the machine
 884          */
 885         if (MACHINE_IS_VM)
 886                 pr_info("Linux is running as a z/VM "
 887                         "guest operating system in 64-bit mode\n");
 888         else if (MACHINE_IS_KVM)
 889                 pr_info("Linux is running under KVM in 64-bit mode\n");
 890         else if (MACHINE_IS_LPAR)
 891                 pr_info("Linux is running natively in 64-bit mode\n");
 892         else
 893                 pr_info("Linux is running as a guest in 64-bit mode\n");
 894
 895         /* Have one command line that is parsed and saved in /proc/cmdline */
 896         /* boot_command_line has been already set up in early.c */
 897         *cmdline_p = boot_command_line;
 898
 899         ROOT_DEV = Root_RAM0;
 900
 901         /* Is init_mm really needed? */
 902         init_mm.start_code = PAGE_OFFSET;
 903         init_mm.end_code = (unsigned long) _etext;
 904         init_mm.end_data = (unsigned long) _edata;
 905         init_mm.brk = (unsigned long) _end;
 906
 907         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
 908                 nospec_auto_detect();
 909
 910         parse_early_param();
 911 #ifdef CONFIG_CRASH_DUMP
 912         /* Deactivate elfcorehdr= kernel parameter */
 913         elfcorehdr_addr = ELFCORE_ADDR_MAX;
 914 #endif
 915
 916         os_info_init();
 917         setup_ipl();
 918         setup_task_size();
 919
 920         /* Do some memory reservations *before* memory is added to memblock */
 921         reserve_memory_end();
 922         reserve_oldmem();
 923         reserve_kernel();
 924         reserve_initrd();
 925         memblock_allow_resize();
 926
 927         /* Get information about *all* installed memory */
 928         detect_memory_memblock();
 929
 930         remove_oldmem();
 931
 932         /*
 933          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 934          * extra checks that HOLES_IN_ZONE would require.
 935          *
 936          * Is this still required?
 937          */
 938         memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
 939
 940         setup_memory_end();
 941         setup_memory();
 942         dma_contiguous_reserve(memory_end);
 943         vmcp_cma_reserve();
 944
 945         check_initrd();
 946         reserve_crashkernel();
 947 #ifdef CONFIG_CRASH_DUMP
 948         /*
 949          * Be aware that smp_save_dump_cpus() triggers a system reset.
 950          * Therefore CPU and device initialization should be done afterwards.
 951          */
 952         smp_save_dump_cpus();
 953 #endif
 954
 955         setup_resources();
 956         setup_lowcore_dat_off();
 957         smp_fill_possible_mask();
 958         cpu_detect_mhz_feature();
 959         cpu_init();
 960         numa_setup();
 961         smp_detect_cpus();
 962         topology_init_early();
 963
 964         /*
 965          * Create kernel page tables and switch to virtual addressing.
 966          */
 967         paging_init();
 968
 969         /*
 970          * After paging_init created the kernel page table, the new PSWs
 971          * in lowcore can now run with DAT enabled.
 972          */
 973         setup_lowcore_dat_on();
 974
 975         /* Setup default console */
 976         conmode_default();
 977         set_preferred_console();
 978
 979         apply_alternative_instructions();
 980         if (IS_ENABLED(CONFIG_EXPOLINE))
 981                 nospec_init_branches();
 982
 983         /* Setup zfcpdump support */
 984         setup_zfcpdump();
 985
 986         /* Add system specific data to the random pool */
 987         setup_randomness();
 988 }