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-map-ops.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 #include <linux/hugetlb.h>
  53 #include <linux/kmemleak.h>
  54
  55 #include <asm/archrandom.h>
  56 #include <asm/boot_data.h>
  57 #include <asm/ipl.h>
  58 #include <asm/facility.h>
  59 #include <asm/smp.h>
  60 #include <asm/mmu_context.h>
  61 #include <asm/cpcmd.h>
  62 #include <asm/abs_lowcore.h>
  63 #include <asm/nmi.h>
  64 #include <asm/irq.h>
  65 #include <asm/page.h>
  66 #include <asm/ptrace.h>
  67 #include <asm/sections.h>
  68 #include <asm/ebcdic.h>
  69 #include <asm/diag.h>
  70 #include <asm/os_info.h>
  71 #include <asm/sclp.h>
  72 #include <asm/stacktrace.h>
  73 #include <asm/sysinfo.h>
  74 #include <asm/numa.h>
  75 #include <asm/alternative.h>
  76 #include <asm/nospec-branch.h>
  77 #include <asm/physmem_info.h>
  78 #include <asm/maccess.h>
  79 #include <asm/uv.h>
  80 #include <asm/asm-offsets.h>
  81 #include "entry.h"
  82
  83 /*
  84  * Machine setup..
  85  */
  86 unsigned int console_mode = 0;
  87 EXPORT_SYMBOL(console_mode);
  88
  89 unsigned int console_devno = -1;
  90 EXPORT_SYMBOL(console_devno);
  91
  92 unsigned int console_irq = -1;
  93 EXPORT_SYMBOL(console_irq);
  94
  95 /*
  96  * Some code and data needs to stay below 2 GB, even when the kernel would be
  97  * relocated above 2 GB, because it has to use 31 bit addresses.
  98  * Such code and data is part of the .amode31 section.
  99  */
 100 unsigned long __amode31_ref __samode31 = (unsigned long)&_samode31;
 101 unsigned long __amode31_ref __eamode31 = (unsigned long)&_eamode31;
 102 unsigned long __amode31_ref __stext_amode31 = (unsigned long)&_stext_amode31;
 103 unsigned long __amode31_ref __etext_amode31 = (unsigned long)&_etext_amode31;
 104 struct exception_table_entry __amode31_ref *__start_amode31_ex_table = _start_amode31_ex_table;
 105 struct exception_table_entry __amode31_ref *__stop_amode31_ex_table = _stop_amode31_ex_table;
 106
 107 /*
 108  * Control registers CR2, CR5 and CR15 are initialized with addresses
 109  * of tables that must be placed below 2G which is handled by the AMODE31
 110  * sections.
 111  * Because the AMODE31 sections are relocated below 2G at startup,
 112  * the content of control registers CR2, CR5 and CR15 must be updated
 113  * with new addresses after the relocation. The initial initialization of
 114  * control registers occurs in head64.S and then gets updated again after AMODE31
 115  * relocation. We must access the relevant AMODE31 tables indirectly via
 116  * pointers placed in the .amode31.refs linker section. Those pointers get
 117  * updated automatically during AMODE31 relocation and always contain a valid
 118  * address within AMODE31 sections.
 119  */
 120
 121 static __amode31_data u32 __ctl_duct_amode31[16] __aligned(64);
 122
 123 static __amode31_data u64 __ctl_aste_amode31[8] __aligned(64) = {
 124         [1] = 0xffffffffffffffff
 125 };
 126
 127 static __amode31_data u32 __ctl_duald_amode31[32] __aligned(128) = {
 128         0x80000000, 0, 0, 0,
 129         0x80000000, 0, 0, 0,
 130         0x80000000, 0, 0, 0,
 131         0x80000000, 0, 0, 0,
 132         0x80000000, 0, 0, 0,
 133         0x80000000, 0, 0, 0,
 134         0x80000000, 0, 0, 0,
 135         0x80000000, 0, 0, 0
 136 };
 137
 138 static __amode31_data u32 __ctl_linkage_stack_amode31[8] __aligned(64) = {
 139         0, 0, 0x89000000, 0,
 140         0, 0, 0x8a000000, 0
 141 };
 142
 143 static u64 __amode31_ref *__ctl_aste = __ctl_aste_amode31;
 144 static u32 __amode31_ref *__ctl_duald = __ctl_duald_amode31;
 145 static u32 __amode31_ref *__ctl_linkage_stack = __ctl_linkage_stack_amode31;
 146 static u32 __amode31_ref *__ctl_duct = __ctl_duct_amode31;
 147
 148 int __bootdata(noexec_disabled);
 149 unsigned long __bootdata(ident_map_size);
 150 struct physmem_info __bootdata(physmem_info);
 151
 152 unsigned long __bootdata_preserved(__kaslr_offset);
 153 unsigned int __bootdata_preserved(zlib_dfltcc_support);
 154 EXPORT_SYMBOL(zlib_dfltcc_support);
 155 u64 __bootdata_preserved(stfle_fac_list[16]);
 156 EXPORT_SYMBOL(stfle_fac_list);
 157 u64 __bootdata_preserved(alt_stfle_fac_list[16]);
 158 struct oldmem_data __bootdata_preserved(oldmem_data);
 159
 160 unsigned long VMALLOC_START;
 161 EXPORT_SYMBOL(VMALLOC_START);
 162
 163 unsigned long VMALLOC_END;
 164 EXPORT_SYMBOL(VMALLOC_END);
 165
 166 struct page *vmemmap;
 167 EXPORT_SYMBOL(vmemmap);
 168 unsigned long vmemmap_size;
 169
 170 unsigned long MODULES_VADDR;
 171 unsigned long MODULES_END;
 172
 173 /* An array with a pointer to the lowcore of every CPU. */
 174 struct lowcore *lowcore_ptr[NR_CPUS];
 175 EXPORT_SYMBOL(lowcore_ptr);
 176
 177 DEFINE_STATIC_KEY_FALSE(cpu_has_bear);
 178
 179 /*
 180  * The Write Back bit position in the physaddr is given by the SLPC PCI.
 181  * Leaving the mask zero always uses write through which is safe
 182  */
 183 unsigned long mio_wb_bit_mask __ro_after_init;
 184
 185 /*
 186  * This is set up by the setup-routine at boot-time
 187  * for S390 need to find out, what we have to setup
 188  * using address 0x10400 ...
 189  */
 190
 191 #include <asm/setup.h>
 192
 193 /*
 194  * condev= and conmode= setup parameter.
 195  */
 196
 197 static int __init condev_setup(char *str)
 198 {
 199         int vdev;
 200
 201         vdev = simple_strtoul(str, &str, 0);
 202         if (vdev >= 0 && vdev < 65536) {
 203                 console_devno = vdev;
 204                 console_irq = -1;
 205         }
 206         return 1;
 207 }
 208
 209 __setup("condev=", condev_setup);
 210
 211 static void __init set_preferred_console(void)
 212 {
 213         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 214                 add_preferred_console("ttyS", 0, NULL);
 215         else if (CONSOLE_IS_3270)
 216                 add_preferred_console("tty3270", 0, NULL);
 217         else if (CONSOLE_IS_VT220)
 218                 add_preferred_console("ttysclp", 0, NULL);
 219         else if (CONSOLE_IS_HVC)
 220                 add_preferred_console("hvc", 0, NULL);
 221 }
 222
 223 static int __init conmode_setup(char *str)
 224 {
 225 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 226         if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
 227                 SET_CONSOLE_SCLP;
 228 #endif
 229 #if defined(CONFIG_TN3215_CONSOLE)
 230         if (!strcmp(str, "3215"))
 231                 SET_CONSOLE_3215;
 232 #endif
 233 #if defined(CONFIG_TN3270_CONSOLE)
 234         if (!strcmp(str, "3270"))
 235                 SET_CONSOLE_3270;
 236 #endif
 237         set_preferred_console();
 238         return 1;
 239 }
 240
 241 __setup("conmode=", conmode_setup);
 242
 243 static void __init conmode_default(void)
 244 {
 245         char query_buffer[1024];
 246         char *ptr;
 247
 248         if (MACHINE_IS_VM) {
 249                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 250                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 251                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 252                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 253                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 254                 ptr = strstr(query_buffer, "CONMODE");
 255                 /*
 256                  * Set the conmode to 3215 so that the device recognition
 257                  * will set the cu_type of the console to 3215. If the
 258                  * conmode is 3270 and we don't set it back then both
 259                  * 3215 and the 3270 driver will try to access the console
 260                  * device (3215 as console and 3270 as normal tty).
 261                  */
 262                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 263                 if (ptr == NULL) {
 264 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 265                         SET_CONSOLE_SCLP;
 266 #endif
 267                         return;
 268                 }
 269                 if (str_has_prefix(ptr + 8, "3270")) {
 270 #if defined(CONFIG_TN3270_CONSOLE)
 271                         SET_CONSOLE_3270;
 272 #elif defined(CONFIG_TN3215_CONSOLE)
 273                         SET_CONSOLE_3215;
 274 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 275                         SET_CONSOLE_SCLP;
 276 #endif
 277                 } else if (str_has_prefix(ptr + 8, "3215")) {
 278 #if defined(CONFIG_TN3215_CONSOLE)
 279                         SET_CONSOLE_3215;
 280 #elif defined(CONFIG_TN3270_CONSOLE)
 281                         SET_CONSOLE_3270;
 282 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 283                         SET_CONSOLE_SCLP;
 284 #endif
 285                 }
 286         } else if (MACHINE_IS_KVM) {
 287                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
 288                         SET_CONSOLE_VT220;
 289                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
 290                         SET_CONSOLE_SCLP;
 291                 else
 292                         SET_CONSOLE_HVC;
 293         } else {
 294 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 295                 SET_CONSOLE_SCLP;
 296 #endif
 297         }
 298 }
 299
 300 #ifdef CONFIG_CRASH_DUMP
 301 static void __init setup_zfcpdump(void)
 302 {
 303         if (!is_ipl_type_dump())
 304                 return;
 305         if (oldmem_data.start)
 306                 return;
 307         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 308         console_loglevel = 2;
 309 }
 310 #else
 311 static inline void setup_zfcpdump(void) {}
 312 #endif /* CONFIG_CRASH_DUMP */
 313
 314  /*
 315  * Reboot, halt and power_off stubs. They just call _machine_restart,
 316  * _machine_halt or _machine_power_off.
 317  */
 318
 319 void machine_restart(char *command)
 320 {
 321         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 322                 /*
 323                  * Only unblank the console if we are called in enabled
 324                  * context or a bust_spinlocks cleared the way for us.
 325                  */
 326                 console_unblank();
 327         _machine_restart(command);
 328 }
 329
 330 void machine_halt(void)
 331 {
 332         if (!in_interrupt() || oops_in_progress)
 333                 /*
 334                  * Only unblank the console if we are called in enabled
 335                  * context or a bust_spinlocks cleared the way for us.
 336                  */
 337                 console_unblank();
 338         _machine_halt();
 339 }
 340
 341 void machine_power_off(void)
 342 {
 343         if (!in_interrupt() || oops_in_progress)
 344                 /*
 345                  * Only unblank the console if we are called in enabled
 346                  * context or a bust_spinlocks cleared the way for us.
 347                  */
 348                 console_unblank();
 349         _machine_power_off();
 350 }
 351
 352 /*
 353  * Dummy power off function.
 354  */
 355 void (*pm_power_off)(void) = machine_power_off;
 356 EXPORT_SYMBOL_GPL(pm_power_off);
 357
 358 void *restart_stack;
 359
 360 unsigned long stack_alloc(void)
 361 {
 362 #ifdef CONFIG_VMAP_STACK
 363         void *ret;
 364
 365         ret = __vmalloc_node(THREAD_SIZE, THREAD_SIZE, THREADINFO_GFP,
 366                              NUMA_NO_NODE, __builtin_return_address(0));
 367         kmemleak_not_leak(ret);
 368         return (unsigned long)ret;
 369 #else
 370         return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 371 #endif
 372 }
 373
 374 void stack_free(unsigned long stack)
 375 {
 376 #ifdef CONFIG_VMAP_STACK
 377         vfree((void *) stack);
 378 #else
 379         free_pages(stack, THREAD_SIZE_ORDER);
 380 #endif
 381 }
 382
 383 int __init arch_early_irq_init(void)
 384 {
 385         unsigned long stack;
 386
 387         stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 388         if (!stack)
 389                 panic("Couldn't allocate async stack");
 390         S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
 391         return 0;
 392 }
 393
 394 void __init arch_call_rest_init(void)
 395 {
 396         unsigned long stack;
 397
 398         smp_reinit_ipl_cpu();
 399         stack = stack_alloc();
 400         if (!stack)
 401                 panic("Couldn't allocate kernel stack");
 402         current->stack = (void *) stack;
 403 #ifdef CONFIG_VMAP_STACK
 404         current->stack_vm_area = (void *) stack;
 405 #endif
 406         set_task_stack_end_magic(current);
 407         stack += STACK_INIT_OFFSET;
 408         S390_lowcore.kernel_stack = stack;
 409         call_on_stack_noreturn(rest_init, stack);
 410 }
 411
 412 static void __init setup_lowcore(void)
 413 {
 414         struct lowcore *lc, *abs_lc;
 415         unsigned long mcck_stack;
 416
 417         /*
 418          * Setup lowcore for boot cpu
 419          */
 420         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 421         lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
 422         if (!lc)
 423                 panic("%s: Failed to allocate %zu bytes align=%zx\n",
 424                       __func__, sizeof(*lc), sizeof(*lc));
 425
 426         lc->restart_psw.mask = PSW_KERNEL_BITS & ~PSW_MASK_DAT;
 427         lc->restart_psw.addr = __pa(restart_int_handler);
 428         lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 429         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 430         lc->svc_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 431         lc->svc_new_psw.addr = (unsigned long) system_call;
 432         lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 433         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 434         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 435         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 436         lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 437         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 438         lc->clock_comparator = clock_comparator_max;
 439         lc->nodat_stack = ((unsigned long) &init_thread_union)
 440                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 441         lc->current_task = (unsigned long)&init_task;
 442         lc->lpp = LPP_MAGIC;
 443         lc->machine_flags = S390_lowcore.machine_flags;
 444         lc->preempt_count = S390_lowcore.preempt_count;
 445         nmi_alloc_mcesa_early(&lc->mcesad);
 446         lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
 447         lc->exit_timer = S390_lowcore.exit_timer;
 448         lc->user_timer = S390_lowcore.user_timer;
 449         lc->system_timer = S390_lowcore.system_timer;
 450         lc->steal_timer = S390_lowcore.steal_timer;
 451         lc->last_update_timer = S390_lowcore.last_update_timer;
 452         lc->last_update_clock = S390_lowcore.last_update_clock;
 453
 454         /*
 455          * Allocate the global restart stack which is the same for
 456          * all CPUs in cast *one* of them does a PSW restart.
 457          */
 458         restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 459         if (!restart_stack)
 460                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 461                       __func__, THREAD_SIZE, THREAD_SIZE);
 462         restart_stack += STACK_INIT_OFFSET;
 463
 464         /*
 465          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 466          * restart data to the absolute zero lowcore. This is necessary if
 467          * PSW restart is done on an offline CPU that has lowcore zero.
 468          */
 469         lc->restart_stack = (unsigned long) restart_stack;
 470         lc->restart_fn = (unsigned long) do_restart;
 471         lc->restart_data = 0;
 472         lc->restart_source = -1U;
 473         __ctl_store(lc->cregs_save_area, 0, 15);
 474
 475         mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 476         if (!mcck_stack)
 477                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 478                       __func__, THREAD_SIZE, THREAD_SIZE);
 479         lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
 480
 481         lc->spinlock_lockval = arch_spin_lockval(0);
 482         lc->spinlock_index = 0;
 483         arch_spin_lock_setup(0);
 484         lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
 485         lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
 486         lc->preempt_count = PREEMPT_DISABLED;
 487         lc->kernel_asce = S390_lowcore.kernel_asce;
 488         lc->user_asce = S390_lowcore.user_asce;
 489
 490         abs_lc = get_abs_lowcore();
 491         abs_lc->restart_stack = lc->restart_stack;
 492         abs_lc->restart_fn = lc->restart_fn;
 493         abs_lc->restart_data = lc->restart_data;
 494         abs_lc->restart_source = lc->restart_source;
 495         abs_lc->restart_psw = lc->restart_psw;
 496         abs_lc->restart_flags = RESTART_FLAG_CTLREGS;
 497         memcpy(abs_lc->cregs_save_area, lc->cregs_save_area, sizeof(abs_lc->cregs_save_area));
 498         abs_lc->program_new_psw = lc->program_new_psw;
 499         abs_lc->mcesad = lc->mcesad;
 500         put_abs_lowcore(abs_lc);
 501
 502         set_prefix(__pa(lc));
 503         lowcore_ptr[0] = lc;
 504         if (abs_lowcore_map(0, lowcore_ptr[0], false))
 505                 panic("Couldn't setup absolute lowcore");
 506 }
 507
 508 static struct resource code_resource = {
 509         .name  = "Kernel code",
 510         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 511 };
 512
 513 static struct resource data_resource = {
 514         .name = "Kernel data",
 515         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 516 };
 517
 518 static struct resource bss_resource = {
 519         .name = "Kernel bss",
 520         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 521 };
 522
 523 static struct resource __initdata *standard_resources[] = {
 524         &code_resource,
 525         &data_resource,
 526         &bss_resource,
 527 };
 528
 529 static void __init setup_resources(void)
 530 {
 531         struct resource *res, *std_res, *sub_res;
 532         phys_addr_t start, end;
 533         int j;
 534         u64 i;
 535
 536         code_resource.start = (unsigned long) _text;
 537         code_resource.end = (unsigned long) _etext - 1;
 538         data_resource.start = (unsigned long) _etext;
 539         data_resource.end = (unsigned long) _edata - 1;
 540         bss_resource.start = (unsigned long) __bss_start;
 541         bss_resource.end = (unsigned long) __bss_stop - 1;
 542
 543         for_each_mem_range(i, &start, &end) {
 544                 res = memblock_alloc(sizeof(*res), 8);
 545                 if (!res)
 546                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 547                               __func__, sizeof(*res), 8);
 548                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 549
 550                 res->name = "System RAM";
 551                 res->start = start;
 552                 /*
 553                  * In memblock, end points to the first byte after the
 554                  * range while in resourses, end points to the last byte in
 555                  * the range.
 556                  */
 557                 res->end = end - 1;
 558                 request_resource(&iomem_resource, res);
 559
 560                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 561                         std_res = standard_resources[j];
 562                         if (std_res->start < res->start ||
 563                             std_res->start > res->end)
 564                                 continue;
 565                         if (std_res->end > res->end) {
 566                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
 567                                 if (!sub_res)
 568                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 569                                               __func__, sizeof(*sub_res), 8);
 570                                 *sub_res = *std_res;
 571                                 sub_res->end = res->end;
 572                                 std_res->start = res->end + 1;
 573                                 request_resource(res, sub_res);
 574                         } else {
 575                                 request_resource(res, std_res);
 576                         }
 577                 }
 578         }
 579 #ifdef CONFIG_CRASH_DUMP
 580         /*
 581          * Re-add removed crash kernel memory as reserved memory. This makes
 582          * sure it will be mapped with the identity mapping and struct pages
 583          * will be created, so it can be resized later on.
 584          * However add it later since the crash kernel resource should not be
 585          * part of the System RAM resource.
 586          */
 587         if (crashk_res.end) {
 588                 memblock_add_node(crashk_res.start, resource_size(&crashk_res),
 589                                   0, MEMBLOCK_NONE);
 590                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 591                 insert_resource(&iomem_resource, &crashk_res);
 592         }
 593 #endif
 594 }
 595
 596 static void __init setup_memory_end(void)
 597 {
 598         max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
 599         pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
 600 }
 601
 602 #ifdef CONFIG_CRASH_DUMP
 603
 604 /*
 605  * When kdump is enabled, we have to ensure that no memory from the area
 606  * [0 - crashkernel memory size] is set offline - it will be exchanged with
 607  * the crashkernel memory region when kdump is triggered. The crashkernel
 608  * memory region can never get offlined (pages are unmovable).
 609  */
 610 static int kdump_mem_notifier(struct notifier_block *nb,
 611                               unsigned long action, void *data)
 612 {
 613         struct memory_notify *arg = data;
 614
 615         if (action != MEM_GOING_OFFLINE)
 616                 return NOTIFY_OK;
 617         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 618                 return NOTIFY_BAD;
 619         return NOTIFY_OK;
 620 }
 621
 622 static struct notifier_block kdump_mem_nb = {
 623         .notifier_call = kdump_mem_notifier,
 624 };
 625
 626 #endif
 627
 628 /*
 629  * Reserve page tables created by decompressor
 630  */
 631 static void __init reserve_pgtables(void)
 632 {
 633         unsigned long start, end;
 634         struct reserved_range *range;
 635
 636         for_each_physmem_reserved_type_range(RR_VMEM, range, &start, &end)
 637                 memblock_reserve(start, end - start);
 638 }
 639
 640 /*
 641  * Reserve memory for kdump kernel to be loaded with kexec
 642  */
 643 static void __init reserve_crashkernel(void)
 644 {
 645 #ifdef CONFIG_CRASH_DUMP
 646         unsigned long long crash_base, crash_size;
 647         phys_addr_t low, high;
 648         int rc;
 649
 650         rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
 651                                &crash_base);
 652
 653         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 654         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 655         if (rc || crash_size == 0)
 656                 return;
 657
 658         if (memblock.memory.regions[0].size < crash_size) {
 659                 pr_info("crashkernel reservation failed: %s\n",
 660                         "first memory chunk must be at least crashkernel size");
 661                 return;
 662         }
 663
 664         low = crash_base ?: oldmem_data.start;
 665         high = low + crash_size;
 666         if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
 667                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 668                 crash_base = low;
 669         } else {
 670                 /* Find suitable area in free memory */
 671                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 672                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 673
 674                 if (crash_base && crash_base < low) {
 675                         pr_info("crashkernel reservation failed: %s\n",
 676                                 "crash_base too low");
 677                         return;
 678                 }
 679                 low = crash_base ?: low;
 680                 crash_base = memblock_phys_alloc_range(crash_size,
 681                                                        KEXEC_CRASH_MEM_ALIGN,
 682                                                        low, high);
 683         }
 684
 685         if (!crash_base) {
 686                 pr_info("crashkernel reservation failed: %s\n",
 687                         "no suitable area found");
 688                 return;
 689         }
 690
 691         if (register_memory_notifier(&kdump_mem_nb)) {
 692                 memblock_phys_free(crash_base, crash_size);
 693                 return;
 694         }
 695
 696         if (!oldmem_data.start && MACHINE_IS_VM)
 697                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 698         crashk_res.start = crash_base;
 699         crashk_res.end = crash_base + crash_size - 1;
 700         memblock_remove(crash_base, crash_size);
 701         pr_info("Reserving %lluMB of memory at %lluMB "
 702                 "for crashkernel (System RAM: %luMB)\n",
 703                 crash_size >> 20, crash_base >> 20,
 704                 (unsigned long)memblock.memory.total_size >> 20);
 705         os_info_crashkernel_add(crash_base, crash_size);
 706 #endif
 707 }
 708
 709 /*
 710  * Reserve the initrd from being used by memblock
 711  */
 712 static void __init reserve_initrd(void)
 713 {
 714         unsigned long addr, size;
 715
 716         if (!IS_ENABLED(CONFIG_BLK_DEV_INITRD) || !get_physmem_reserved(RR_INITRD, &addr, &size))
 717                 return;
 718         initrd_start = (unsigned long)__va(addr);
 719         initrd_end = initrd_start + size;
 720         memblock_reserve(addr, size);
 721 }
 722
 723 /*
 724  * Reserve the memory area used to pass the certificate lists
 725  */
 726 static void __init reserve_certificate_list(void)
 727 {
 728         if (ipl_cert_list_addr)
 729                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
 730 }
 731
 732 static void __init reserve_physmem_info(void)
 733 {
 734         unsigned long addr, size;
 735
 736         if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size))
 737                 memblock_reserve(addr, size);
 738 }
 739
 740 static void __init free_physmem_info(void)
 741 {
 742         unsigned long addr, size;
 743
 744         if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size))
 745                 memblock_phys_free(addr, size);
 746 }
 747
 748 static void __init memblock_add_physmem_info(void)
 749 {
 750         unsigned long start, end;
 751         int i;
 752
 753         pr_debug("physmem info source: %s (%hhd)\n",
 754                  get_physmem_info_source(), physmem_info.info_source);
 755         /* keep memblock lists close to the kernel */
 756         memblock_set_bottom_up(true);
 757         for_each_physmem_usable_range(i, &start, &end)
 758                 memblock_add(start, end - start);
 759         for_each_physmem_online_range(i, &start, &end)
 760                 memblock_physmem_add(start, end - start);
 761         memblock_set_bottom_up(false);
 762         memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
 763 }
 764
 765 /*
 766  * Reserve memory used for lowcore/command line/kernel image.
 767  */
 768 static void __init reserve_kernel(void)
 769 {
 770         memblock_reserve(0, STARTUP_NORMAL_OFFSET);
 771         memblock_reserve(OLDMEM_BASE, sizeof(unsigned long));
 772         memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long));
 773         memblock_reserve(physmem_info.reserved[RR_AMODE31].start, __eamode31 - __samode31);
 774         memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP);
 775         memblock_reserve(__pa(_stext), _end - _stext);
 776 }
 777
 778 static void __init setup_memory(void)
 779 {
 780         phys_addr_t start, end;
 781         u64 i;
 782
 783         /*
 784          * Init storage key for present memory
 785          */
 786         for_each_mem_range(i, &start, &end)
 787                 storage_key_init_range(start, end);
 788
 789         psw_set_key(PAGE_DEFAULT_KEY);
 790 }
 791
 792 static void __init relocate_amode31_section(void)
 793 {
 794         unsigned long amode31_size = __eamode31 - __samode31;
 795         long amode31_offset = physmem_info.reserved[RR_AMODE31].start - __samode31;
 796         long *ptr;
 797
 798         pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size);
 799
 800         /* Move original AMODE31 section to the new one */
 801         memmove((void *)physmem_info.reserved[RR_AMODE31].start, (void *)__samode31, amode31_size);
 802         /* Zero out the old AMODE31 section to catch invalid accesses within it */
 803         memset((void *)__samode31, 0, amode31_size);
 804
 805         /* Update all AMODE31 region references */
 806         for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++)
 807                 *ptr += amode31_offset;
 808 }
 809
 810 /* This must be called after AMODE31 relocation */
 811 static void __init setup_cr(void)
 812 {
 813         union ctlreg2 cr2;
 814         union ctlreg5 cr5;
 815         union ctlreg15 cr15;
 816
 817         __ctl_duct[1] = (unsigned long)__ctl_aste;
 818         __ctl_duct[2] = (unsigned long)__ctl_aste;
 819         __ctl_duct[4] = (unsigned long)__ctl_duald;
 820
 821         /* Update control registers CR2, CR5 and CR15 */
 822         __ctl_store(cr2.val, 2, 2);
 823         __ctl_store(cr5.val, 5, 5);
 824         __ctl_store(cr15.val, 15, 15);
 825         cr2.ducto = (unsigned long)__ctl_duct >> 6;
 826         cr5.pasteo = (unsigned long)__ctl_duct >> 6;
 827         cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
 828         __ctl_load(cr2.val, 2, 2);
 829         __ctl_load(cr5.val, 5, 5);
 830         __ctl_load(cr15.val, 15, 15);
 831 }
 832
 833 /*
 834  * Add system information as device randomness
 835  */
 836 static void __init setup_randomness(void)
 837 {
 838         struct sysinfo_3_2_2 *vmms;
 839
 840         vmms = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
 841         if (!vmms)
 842                 panic("Failed to allocate memory for sysinfo structure\n");
 843         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 844                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
 845         memblock_free(vmms, PAGE_SIZE);
 846
 847         if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
 848                 static_branch_enable(&s390_arch_random_available);
 849 }
 850
 851 /*
 852  * Find the correct size for the task_struct. This depends on
 853  * the size of the struct fpu at the end of the thread_struct
 854  * which is embedded in the task_struct.
 855  */
 856 static void __init setup_task_size(void)
 857 {
 858         int task_size = sizeof(struct task_struct);
 859
 860         if (!MACHINE_HAS_VX) {
 861                 task_size -= sizeof(__vector128) * __NUM_VXRS;
 862                 task_size += sizeof(freg_t) * __NUM_FPRS;
 863         }
 864         arch_task_struct_size = task_size;
 865 }
 866
 867 /*
 868  * Issue diagnose 318 to set the control program name and
 869  * version codes.
 870  */
 871 static void __init setup_control_program_code(void)
 872 {
 873         union diag318_info diag318_info = {
 874                 .cpnc = CPNC_LINUX,
 875                 .cpvc = 0,
 876         };
 877
 878         if (!sclp.has_diag318)
 879                 return;
 880
 881         diag_stat_inc(DIAG_STAT_X318);
 882         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
 883 }
 884
 885 /*
 886  * Print the component list from the IPL report
 887  */
 888 static void __init log_component_list(void)
 889 {
 890         struct ipl_rb_component_entry *ptr, *end;
 891         char *str;
 892
 893         if (!early_ipl_comp_list_addr)
 894                 return;
 895         if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
 896                 pr_info("Linux is running with Secure-IPL enabled\n");
 897         else
 898                 pr_info("Linux is running with Secure-IPL disabled\n");
 899         ptr = (void *) early_ipl_comp_list_addr;
 900         end = (void *) ptr + early_ipl_comp_list_size;
 901         pr_info("The IPL report contains the following components:\n");
 902         while (ptr < end) {
 903                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
 904                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
 905                                 str = "signed, verified";
 906                         else
 907                                 str = "signed, verification failed";
 908                 } else {
 909                         str = "not signed";
 910                 }
 911                 pr_info("%016llx - %016llx (%s)\n",
 912                         ptr->addr, ptr->addr + ptr->len, str);
 913                 ptr++;
 914         }
 915 }
 916
 917 /*
 918  * Setup function called from init/main.c just after the banner
 919  * was printed.
 920  */
 921
 922 void __init setup_arch(char **cmdline_p)
 923 {
 924         /*
 925          * print what head.S has found out about the machine
 926          */
 927         if (MACHINE_IS_VM)
 928                 pr_info("Linux is running as a z/VM "
 929                         "guest operating system in 64-bit mode\n");
 930         else if (MACHINE_IS_KVM)
 931                 pr_info("Linux is running under KVM in 64-bit mode\n");
 932         else if (MACHINE_IS_LPAR)
 933                 pr_info("Linux is running natively in 64-bit mode\n");
 934         else
 935                 pr_info("Linux is running as a guest in 64-bit mode\n");
 936
 937         log_component_list();
 938
 939         /* Have one command line that is parsed and saved in /proc/cmdline */
 940         /* boot_command_line has been already set up in early.c */
 941         *cmdline_p = boot_command_line;
 942
 943         ROOT_DEV = Root_RAM0;
 944
 945         setup_initial_init_mm(_text, _etext, _edata, _end);
 946
 947         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
 948                 nospec_auto_detect();
 949
 950         jump_label_init();
 951         parse_early_param();
 952 #ifdef CONFIG_CRASH_DUMP
 953         /* Deactivate elfcorehdr= kernel parameter */
 954         elfcorehdr_addr = ELFCORE_ADDR_MAX;
 955 #endif
 956
 957         os_info_init();
 958         setup_ipl();
 959         setup_task_size();
 960         setup_control_program_code();
 961
 962         /* Do some memory reservations *before* memory is added to memblock */
 963         reserve_pgtables();
 964         reserve_kernel();
 965         reserve_initrd();
 966         reserve_certificate_list();
 967         reserve_physmem_info();
 968         memblock_set_current_limit(ident_map_size);
 969         memblock_allow_resize();
 970
 971         /* Get information about *all* installed memory */
 972         memblock_add_physmem_info();
 973
 974         free_physmem_info();
 975         setup_memory_end();
 976         memblock_dump_all();
 977         setup_memory();
 978
 979         relocate_amode31_section();
 980         setup_cr();
 981         setup_uv();
 982         dma_contiguous_reserve(ident_map_size);
 983         vmcp_cma_reserve();
 984         if (MACHINE_HAS_EDAT2)
 985                 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
 986
 987         reserve_crashkernel();
 988 #ifdef CONFIG_CRASH_DUMP
 989         /*
 990          * Be aware that smp_save_dump_secondary_cpus() triggers a system reset.
 991          * Therefore CPU and device initialization should be done afterwards.
 992          */
 993         smp_save_dump_secondary_cpus();
 994 #endif
 995
 996         setup_resources();
 997         setup_lowcore();
 998         smp_fill_possible_mask();
 999         cpu_detect_mhz_feature();
1000         cpu_init();
1001         numa_setup();
1002         smp_detect_cpus();
1003         topology_init_early();
1004
1005         if (test_facility(193))
1006                 static_branch_enable(&cpu_has_bear);
1007
1008         /*
1009          * Create kernel page tables.
1010          */
1011         paging_init();
1012
1013         /*
1014          * After paging_init created the kernel page table, the new PSWs
1015          * in lowcore can now run with DAT enabled.
1016          */
1017 #ifdef CONFIG_CRASH_DUMP
1018         smp_save_dump_ipl_cpu();
1019 #endif
1020
1021         /* Setup default console */
1022         conmode_default();
1023         set_preferred_console();
1024
1025         apply_alternative_instructions();
1026         if (IS_ENABLED(CONFIG_EXPOLINE))
1027                 nospec_init_branches();
1028
1029         /* Setup zfcp/nvme dump support */
1030         setup_zfcpdump();
1031
1032         /* Add system specific data to the random pool */
1033         setup_randomness();
1034 }