1 // SPDX-License-Identifier: GPL-2.0
4 * Copyright IBM Corp. 1999, 2012
5 * Author(s): Hartmut Penner (hp@de.ibm.com),
6 * Martin Schwidefsky (schwidefsky@de.ibm.com)
8 * Derived from "arch/i386/kernel/setup.c"
9 * Copyright (C) 1995, Linus Torvalds
13 * This file handles the architecture-dependent parts of initialization
16 #define KMSG_COMPONENT "setup"
17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
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>
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 #include <linux/start_kernel.h>
55 #include <asm/facility.h>
57 #include <asm/mmu_context.h>
58 #include <asm/cpcmd.h>
59 #include <asm/lowcore.h>
63 #include <asm/ptrace.h>
64 #include <asm/sections.h>
65 #include <asm/ebcdic.h>
67 #include <asm/os_info.h>
69 #include <asm/sysinfo.h>
71 #include <asm/alternative.h>
72 #include <asm/nospec-branch.h>
73 #include <asm/mem_detect.h>
79 unsigned int console_mode = 0;
80 EXPORT_SYMBOL(console_mode);
82 unsigned int console_devno = -1;
83 EXPORT_SYMBOL(console_devno);
85 unsigned int console_irq = -1;
86 EXPORT_SYMBOL(console_irq);
88 unsigned long elf_hwcap __read_mostly = 0;
89 char elf_platform[ELF_PLATFORM_SIZE];
91 unsigned long int_hwcap = 0;
93 int __bootdata(noexec_disabled);
94 int __bootdata(memory_end_set);
95 unsigned long __bootdata(memory_end);
96 unsigned long __bootdata(max_physmem_end);
97 struct mem_detect_info __bootdata(mem_detect);
99 unsigned long VMALLOC_START;
100 EXPORT_SYMBOL(VMALLOC_START);
102 unsigned long VMALLOC_END;
103 EXPORT_SYMBOL(VMALLOC_END);
105 struct page *vmemmap;
106 EXPORT_SYMBOL(vmemmap);
108 unsigned long MODULES_VADDR;
109 unsigned long MODULES_END;
111 /* An array with a pointer to the lowcore of every CPU. */
112 struct lowcore *lowcore_ptr[NR_CPUS];
113 EXPORT_SYMBOL(lowcore_ptr);
116 * This is set up by the setup-routine at boot-time
117 * for S390 need to find out, what we have to setup
118 * using address 0x10400 ...
121 #include <asm/setup.h>
124 * condev= and conmode= setup parameter.
127 static int __init condev_setup(char *str)
131 vdev = simple_strtoul(str, &str, 0);
132 if (vdev >= 0 && vdev < 65536) {
133 console_devno = vdev;
139 __setup("condev=", condev_setup);
141 static void __init set_preferred_console(void)
143 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
144 add_preferred_console("ttyS", 0, NULL);
145 else if (CONSOLE_IS_3270)
146 add_preferred_console("tty3270", 0, NULL);
147 else if (CONSOLE_IS_VT220)
148 add_preferred_console("ttyS", 1, NULL);
149 else if (CONSOLE_IS_HVC)
150 add_preferred_console("hvc", 0, NULL);
153 static int __init conmode_setup(char *str)
155 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
156 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
159 #if defined(CONFIG_TN3215_CONSOLE)
160 if (strncmp(str, "3215", 5) == 0)
163 #if defined(CONFIG_TN3270_CONSOLE)
164 if (strncmp(str, "3270", 5) == 0)
167 set_preferred_console();
171 __setup("conmode=", conmode_setup);
173 static void __init conmode_default(void)
175 char query_buffer[1024];
179 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
180 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
181 ptr = strstr(query_buffer, "SUBCHANNEL =");
182 console_irq = simple_strtoul(ptr + 13, NULL, 16);
183 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
184 ptr = strstr(query_buffer, "CONMODE");
186 * Set the conmode to 3215 so that the device recognition
187 * will set the cu_type of the console to 3215. If the
188 * conmode is 3270 and we don't set it back then both
189 * 3215 and the 3270 driver will try to access the console
190 * device (3215 as console and 3270 as normal tty).
192 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
194 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
199 if (strncmp(ptr + 8, "3270", 4) == 0) {
200 #if defined(CONFIG_TN3270_CONSOLE)
202 #elif defined(CONFIG_TN3215_CONSOLE)
204 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
207 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
208 #if defined(CONFIG_TN3215_CONSOLE)
210 #elif defined(CONFIG_TN3270_CONSOLE)
212 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
216 } else if (MACHINE_IS_KVM) {
217 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
219 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
224 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
228 if (IS_ENABLED(CONFIG_VT) && IS_ENABLED(CONFIG_DUMMY_CONSOLE))
229 conswitchp = &dummy_con;
232 #ifdef CONFIG_CRASH_DUMP
233 static void __init setup_zfcpdump(void)
235 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
239 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
240 console_loglevel = 2;
243 static inline void setup_zfcpdump(void) {}
244 #endif /* CONFIG_CRASH_DUMP */
247 * Reboot, halt and power_off stubs. They just call _machine_restart,
248 * _machine_halt or _machine_power_off.
251 void machine_restart(char *command)
253 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
255 * Only unblank the console if we are called in enabled
256 * context or a bust_spinlocks cleared the way for us.
259 _machine_restart(command);
262 void machine_halt(void)
264 if (!in_interrupt() || oops_in_progress)
266 * Only unblank the console if we are called in enabled
267 * context or a bust_spinlocks cleared the way for us.
273 void machine_power_off(void)
275 if (!in_interrupt() || oops_in_progress)
277 * Only unblank the console if we are called in enabled
278 * context or a bust_spinlocks cleared the way for us.
281 _machine_power_off();
285 * Dummy power off function.
287 void (*pm_power_off)(void) = machine_power_off;
288 EXPORT_SYMBOL_GPL(pm_power_off);
290 static int __init parse_vmalloc(char *arg)
294 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
297 early_param("vmalloc", parse_vmalloc);
299 void *restart_stack __section(.data);
301 unsigned long stack_alloc(void)
303 #ifdef CONFIG_VMAP_STACK
304 return (unsigned long)
305 __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
306 VMALLOC_START, VMALLOC_END,
308 PAGE_KERNEL, 0, NUMA_NO_NODE,
309 __builtin_return_address(0));
311 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
315 void stack_free(unsigned long stack)
317 #ifdef CONFIG_VMAP_STACK
318 vfree((void *) stack);
320 free_pages(stack, THREAD_SIZE_ORDER);
324 int __init arch_early_irq_init(void)
328 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
330 panic("Couldn't allocate async stack");
331 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
335 static int __init async_stack_realloc(void)
337 unsigned long old, new;
339 old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
342 panic("Couldn't allocate async stack");
343 S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
344 free_pages(old, THREAD_SIZE_ORDER);
347 early_initcall(async_stack_realloc);
349 void __init arch_call_rest_init(void)
351 struct stack_frame *frame;
354 stack = stack_alloc();
356 panic("Couldn't allocate kernel stack");
357 current->stack = (void *) stack;
358 #ifdef CONFIG_VMAP_STACK
359 current->stack_vm_area = (void *) stack;
361 set_task_stack_end_magic(current);
362 stack += STACK_INIT_OFFSET;
363 S390_lowcore.kernel_stack = stack;
364 frame = (struct stack_frame *) stack;
365 memset(frame, 0, sizeof(*frame));
366 /* Branch to rest_init on the new stack, never returns */
368 " la 15,0(%[_frame])\n"
370 : : [_frame] "a" (frame));
373 static void __init setup_lowcore(void)
378 * Setup lowcore for boot cpu
380 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
381 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
382 lc->restart_psw.mask = PSW_KERNEL_BITS;
383 lc->restart_psw.addr = (unsigned long) restart_int_handler;
384 lc->external_new_psw.mask = PSW_KERNEL_BITS |
385 PSW_MASK_DAT | 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_DAT | 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 |
391 PSW_MASK_DAT | PSW_MASK_MCHECK;
392 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
393 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
394 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
395 lc->io_new_psw.mask = PSW_KERNEL_BITS |
396 PSW_MASK_DAT | PSW_MASK_MCHECK;
397 lc->io_new_psw.addr = (unsigned long) io_int_handler;
398 lc->clock_comparator = clock_comparator_max;
399 lc->nodat_stack = ((unsigned long) &init_thread_union)
400 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
401 lc->current_task = (unsigned long)&init_task;
403 lc->machine_flags = S390_lowcore.machine_flags;
404 lc->preempt_count = S390_lowcore.preempt_count;
405 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
406 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
407 sizeof(lc->stfle_fac_list));
408 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
409 sizeof(lc->alt_stfle_fac_list));
410 nmi_alloc_boot_cpu(lc);
411 vdso_alloc_boot_cpu(lc);
412 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
413 lc->async_enter_timer = S390_lowcore.async_enter_timer;
414 lc->exit_timer = S390_lowcore.exit_timer;
415 lc->user_timer = S390_lowcore.user_timer;
416 lc->system_timer = S390_lowcore.system_timer;
417 lc->steal_timer = S390_lowcore.steal_timer;
418 lc->last_update_timer = S390_lowcore.last_update_timer;
419 lc->last_update_clock = S390_lowcore.last_update_clock;
422 * Allocate the global restart stack which is the same for
423 * all CPUs in cast *one* of them does a PSW restart.
425 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
426 restart_stack += STACK_INIT_OFFSET;
429 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
430 * restart data to the absolute zero lowcore. This is necessary if
431 * PSW restart is done on an offline CPU that has lowcore zero.
433 lc->restart_stack = (unsigned long) restart_stack;
434 lc->restart_fn = (unsigned long) do_restart;
435 lc->restart_data = 0;
436 lc->restart_source = -1UL;
438 /* Setup absolute zero lowcore */
439 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
440 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
441 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
442 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
443 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
446 lc->spinlock_lockval = arch_spin_lockval(0);
447 lc->spinlock_index = 0;
448 arch_spin_lock_setup(0);
450 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
452 set_prefix((u32)(unsigned long) lc);
456 static struct resource code_resource = {
457 .name = "Kernel code",
458 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
461 static struct resource data_resource = {
462 .name = "Kernel data",
463 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
466 static struct resource bss_resource = {
467 .name = "Kernel bss",
468 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
471 static struct resource __initdata *standard_resources[] = {
477 static void __init setup_resources(void)
479 struct resource *res, *std_res, *sub_res;
480 struct memblock_region *reg;
483 code_resource.start = (unsigned long) _text;
484 code_resource.end = (unsigned long) _etext - 1;
485 data_resource.start = (unsigned long) _etext;
486 data_resource.end = (unsigned long) _edata - 1;
487 bss_resource.start = (unsigned long) __bss_start;
488 bss_resource.end = (unsigned long) __bss_stop - 1;
490 for_each_memblock(memory, reg) {
491 res = memblock_alloc(sizeof(*res), 8);
492 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
494 res->name = "System RAM";
495 res->start = reg->base;
496 res->end = reg->base + reg->size - 1;
497 request_resource(&iomem_resource, res);
499 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
500 std_res = standard_resources[j];
501 if (std_res->start < res->start ||
502 std_res->start > res->end)
504 if (std_res->end > res->end) {
505 sub_res = memblock_alloc(sizeof(*sub_res), 8);
507 sub_res->end = res->end;
508 std_res->start = res->end + 1;
509 request_resource(res, sub_res);
511 request_resource(res, std_res);
515 #ifdef CONFIG_CRASH_DUMP
517 * Re-add removed crash kernel memory as reserved memory. This makes
518 * sure it will be mapped with the identity mapping and struct pages
519 * will be created, so it can be resized later on.
520 * However add it later since the crash kernel resource should not be
521 * part of the System RAM resource.
523 if (crashk_res.end) {
524 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
525 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
526 insert_resource(&iomem_resource, &crashk_res);
531 static void __init setup_memory_end(void)
533 unsigned long vmax, vmalloc_size, tmp;
535 /* Choose kernel address space layout: 3 or 4 levels. */
536 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
537 if (IS_ENABLED(CONFIG_KASAN)) {
538 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
542 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
543 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
544 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
545 vmax = _REGION2_SIZE; /* 3-level kernel page table */
547 vmax = _REGION1_SIZE; /* 4-level kernel page table */
550 /* module area is at the end of the kernel address space. */
552 MODULES_VADDR = MODULES_END - MODULES_LEN;
553 VMALLOC_END = MODULES_VADDR;
554 VMALLOC_START = VMALLOC_END - vmalloc_size;
556 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
557 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
558 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
559 tmp = SECTION_ALIGN_UP(tmp);
560 tmp = VMALLOC_START - tmp * sizeof(struct page);
561 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
562 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
563 vmemmap = (struct page *) tmp;
565 /* Take care that memory_end is set and <= vmemmap */
566 memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
568 /* fit in kasan shadow memory region between 1:1 and vmemmap */
569 memory_end = min(memory_end, KASAN_SHADOW_START);
570 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
572 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
573 memblock_remove(memory_end, ULONG_MAX);
575 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
578 #ifdef CONFIG_CRASH_DUMP
581 * When kdump is enabled, we have to ensure that no memory from
582 * the area [0 - crashkernel memory size] and
583 * [crashk_res.start - crashk_res.end] is set offline.
585 static int kdump_mem_notifier(struct notifier_block *nb,
586 unsigned long action, void *data)
588 struct memory_notify *arg = data;
590 if (action != MEM_GOING_OFFLINE)
592 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
594 if (arg->start_pfn > PFN_DOWN(crashk_res.end))
596 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
601 static struct notifier_block kdump_mem_nb = {
602 .notifier_call = kdump_mem_notifier,
608 * Make sure that the area behind memory_end is protected
610 static void reserve_memory_end(void)
613 memblock_reserve(memory_end, ULONG_MAX);
617 * Make sure that oldmem, where the dump is stored, is protected
619 static void reserve_oldmem(void)
621 #ifdef CONFIG_CRASH_DUMP
623 /* Forget all memory above the running kdump system */
624 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
629 * Make sure that oldmem, where the dump is stored, is protected
631 static void remove_oldmem(void)
633 #ifdef CONFIG_CRASH_DUMP
635 /* Forget all memory above the running kdump system */
636 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
641 * Reserve memory for kdump kernel to be loaded with kexec
643 static void __init reserve_crashkernel(void)
645 #ifdef CONFIG_CRASH_DUMP
646 unsigned long long crash_base, crash_size;
647 phys_addr_t low, high;
650 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
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)
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");
664 low = crash_base ?: OLDMEM_BASE;
665 high = low + crash_size;
666 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
667 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
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;
674 if (crash_base && crash_base < low) {
675 pr_info("crashkernel reservation failed: %s\n",
676 "crash_base too low");
679 low = crash_base ?: low;
680 crash_base = memblock_find_in_range(low, high, crash_size,
681 KEXEC_CRASH_MEM_ALIGN);
685 pr_info("crashkernel reservation failed: %s\n",
686 "no suitable area found");
690 if (register_memory_notifier(&kdump_mem_nb))
693 if (!OLDMEM_BASE && MACHINE_IS_VM)
694 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
695 crashk_res.start = crash_base;
696 crashk_res.end = crash_base + crash_size - 1;
697 memblock_remove(crash_base, crash_size);
698 pr_info("Reserving %lluMB of memory at %lluMB "
699 "for crashkernel (System RAM: %luMB)\n",
700 crash_size >> 20, crash_base >> 20,
701 (unsigned long)memblock.memory.total_size >> 20);
702 os_info_crashkernel_add(crash_base, crash_size);
707 * Reserve the initrd from being used by memblock
709 static void __init reserve_initrd(void)
711 #ifdef CONFIG_BLK_DEV_INITRD
712 if (!INITRD_START || !INITRD_SIZE)
714 initrd_start = INITRD_START;
715 initrd_end = initrd_start + INITRD_SIZE;
716 memblock_reserve(INITRD_START, INITRD_SIZE);
720 static void __init reserve_mem_detect_info(void)
722 unsigned long start, size;
724 get_mem_detect_reserved(&start, &size);
726 memblock_reserve(start, size);
729 static void __init free_mem_detect_info(void)
731 unsigned long start, size;
733 get_mem_detect_reserved(&start, &size);
735 memblock_free(start, size);
738 static void __init memblock_physmem_add(phys_addr_t start, phys_addr_t size)
740 memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n",
741 start, start + size - 1);
742 memblock_add_range(&memblock.memory, start, size, 0, 0);
743 memblock_add_range(&memblock.physmem, start, size, 0, 0);
746 static const char * __init get_mem_info_source(void)
748 switch (mem_detect.info_source) {
749 case MEM_DETECT_SCLP_STOR_INFO:
750 return "sclp storage info";
751 case MEM_DETECT_DIAG260:
753 case MEM_DETECT_SCLP_READ_INFO:
754 return "sclp read info";
755 case MEM_DETECT_BIN_SEARCH:
756 return "binary search";
761 static void __init memblock_add_mem_detect_info(void)
763 unsigned long start, end;
766 memblock_dbg("physmem info source: %s (%hhd)\n",
767 get_mem_info_source(), mem_detect.info_source);
768 /* keep memblock lists close to the kernel */
769 memblock_set_bottom_up(true);
770 for_each_mem_detect_block(i, &start, &end)
771 memblock_physmem_add(start, end - start);
772 memblock_set_bottom_up(false);
777 * Check for initrd being in usable memory
779 static void __init check_initrd(void)
781 #ifdef CONFIG_BLK_DEV_INITRD
782 if (INITRD_START && INITRD_SIZE &&
783 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
784 pr_err("The initial RAM disk does not fit into the memory\n");
785 memblock_free(INITRD_START, INITRD_SIZE);
786 initrd_start = initrd_end = 0;
792 * Reserve memory used for lowcore/command line/kernel image.
794 static void __init reserve_kernel(void)
796 unsigned long start_pfn = PFN_UP(__pa(_end));
798 #ifdef CONFIG_DMA_API_DEBUG
800 * DMA_API_DEBUG code stumbles over addresses from the
801 * range [PARMAREA_END, _stext]. Mark the memory as reserved
802 * so it is not used for CONFIG_DMA_API_DEBUG=y.
804 memblock_reserve(0, PFN_PHYS(start_pfn));
806 memblock_reserve(0, PARMAREA_END);
807 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
808 - (unsigned long)_stext);
812 static void __init setup_memory(void)
814 struct memblock_region *reg;
817 * Init storage key for present memory
819 for_each_memblock(memory, reg) {
820 storage_key_init_range(reg->base, reg->base + reg->size);
822 psw_set_key(PAGE_DEFAULT_KEY);
825 memblock_enforce_memory_limit(memblock_end_of_DRAM());
829 * Setup hardware capabilities.
831 static int __init setup_hwcaps(void)
833 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
838 * The store facility list bits numbers as found in the principles
839 * of operation are numbered with bit 1UL<<31 as number 0 to
840 * bit 1UL<<0 as number 31.
841 * Bit 0: instructions named N3, "backported" to esa-mode
842 * Bit 2: z/Architecture mode is active
843 * Bit 7: the store-facility-list-extended facility is installed
844 * Bit 17: the message-security assist is installed
845 * Bit 19: the long-displacement facility is installed
846 * Bit 21: the extended-immediate facility is installed
847 * Bit 22: extended-translation facility 3 is installed
848 * Bit 30: extended-translation facility 3 enhancement facility
849 * These get translated to:
850 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
851 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
852 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
853 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
855 for (i = 0; i < 6; i++)
856 if (test_facility(stfl_bits[i]))
857 elf_hwcap |= 1UL << i;
859 if (test_facility(22) && test_facility(30))
860 elf_hwcap |= HWCAP_S390_ETF3EH;
863 * Check for additional facilities with store-facility-list-extended.
864 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
865 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
866 * as stored by stfl, bits 32-xxx contain additional facilities.
867 * How many facility words are stored depends on the number of
868 * doublewords passed to the instruction. The additional facilities
870 * Bit 42: decimal floating point facility is installed
871 * Bit 44: perform floating point operation facility is installed
873 * HWCAP_S390_DFP bit 6 (42 && 44).
875 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
876 elf_hwcap |= HWCAP_S390_DFP;
879 * Huge page support HWCAP_S390_HPAGE is bit 7.
881 if (MACHINE_HAS_EDAT1)
882 elf_hwcap |= HWCAP_S390_HPAGE;
885 * 64-bit register support for 31-bit processes
886 * HWCAP_S390_HIGH_GPRS is bit 9.
888 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
891 * Transactional execution support HWCAP_S390_TE is bit 10.
894 elf_hwcap |= HWCAP_S390_TE;
897 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
898 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
899 * instead of facility bit 129.
901 if (MACHINE_HAS_VX) {
902 elf_hwcap |= HWCAP_S390_VXRS;
903 if (test_facility(134))
904 elf_hwcap |= HWCAP_S390_VXRS_EXT;
905 if (test_facility(135))
906 elf_hwcap |= HWCAP_S390_VXRS_BCD;
910 * Guarded storage support HWCAP_S390_GS is bit 12.
913 elf_hwcap |= HWCAP_S390_GS;
916 add_device_randomness(&cpu_id, sizeof(cpu_id));
917 switch (cpu_id.machine) {
920 default: /* Use "z900" as default for 64 bit kernels. */
921 strcpy(elf_platform, "z900");
925 strcpy(elf_platform, "z990");
929 strcpy(elf_platform, "z9-109");
933 strcpy(elf_platform, "z10");
937 strcpy(elf_platform, "z196");
941 strcpy(elf_platform, "zEC12");
945 strcpy(elf_platform, "z13");
949 strcpy(elf_platform, "z14");
954 * Virtualization support HWCAP_INT_SIE is bit 0.
957 int_hwcap |= HWCAP_INT_SIE;
961 arch_initcall(setup_hwcaps);
964 * Add system information as device randomness
966 static void __init setup_randomness(void)
968 struct sysinfo_3_2_2 *vmms;
970 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
972 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
973 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
974 memblock_free((unsigned long) vmms, PAGE_SIZE);
978 * Find the correct size for the task_struct. This depends on
979 * the size of the struct fpu at the end of the thread_struct
980 * which is embedded in the task_struct.
982 static void __init setup_task_size(void)
984 int task_size = sizeof(struct task_struct);
986 if (!MACHINE_HAS_VX) {
987 task_size -= sizeof(__vector128) * __NUM_VXRS;
988 task_size += sizeof(freg_t) * __NUM_FPRS;
990 arch_task_struct_size = task_size;
994 * Setup function called from init/main.c just after the banner
998 void __init setup_arch(char **cmdline_p)
1001 * print what head.S has found out about the machine
1004 pr_info("Linux is running as a z/VM "
1005 "guest operating system in 64-bit mode\n");
1006 else if (MACHINE_IS_KVM)
1007 pr_info("Linux is running under KVM in 64-bit mode\n");
1008 else if (MACHINE_IS_LPAR)
1009 pr_info("Linux is running natively in 64-bit mode\n");
1011 /* Have one command line that is parsed and saved in /proc/cmdline */
1012 /* boot_command_line has been already set up in early.c */
1013 *cmdline_p = boot_command_line;
1015 ROOT_DEV = Root_RAM0;
1017 /* Is init_mm really needed? */
1018 init_mm.start_code = PAGE_OFFSET;
1019 init_mm.end_code = (unsigned long) _etext;
1020 init_mm.end_data = (unsigned long) _edata;
1021 init_mm.brk = (unsigned long) _end;
1023 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1024 nospec_auto_detect();
1026 parse_early_param();
1027 #ifdef CONFIG_CRASH_DUMP
1028 /* Deactivate elfcorehdr= kernel parameter */
1029 elfcorehdr_addr = ELFCORE_ADDR_MAX;
1036 /* Do some memory reservations *before* memory is added to memblock */
1037 reserve_memory_end();
1041 reserve_mem_detect_info();
1042 memblock_allow_resize();
1044 /* Get information about *all* installed memory */
1045 memblock_add_mem_detect_info();
1047 free_mem_detect_info();
1051 * Make sure all chunks are MAX_ORDER aligned so we don't need the
1052 * extra checks that HOLES_IN_ZONE would require.
1054 * Is this still required?
1056 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
1060 dma_contiguous_reserve(memory_end);
1064 reserve_crashkernel();
1065 #ifdef CONFIG_CRASH_DUMP
1067 * Be aware that smp_save_dump_cpus() triggers a system reset.
1068 * Therefore CPU and device initialization should be done afterwards.
1070 smp_save_dump_cpus();
1075 smp_fill_possible_mask();
1076 cpu_detect_mhz_feature();
1080 topology_init_early();
1083 * Create kernel page tables and switch to virtual addressing.
1087 /* Setup default console */
1089 set_preferred_console();
1091 apply_alternative_instructions();
1092 if (IS_ENABLED(CONFIG_EXPOLINE))
1093 nospec_init_branches();
1095 /* Setup zfcpdump support */
1098 /* Add system specific data to the random pool */