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/root_dev.h>
38 #include <linux/console.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/dma-contiguous.h>
41 #include <linux/device.h>
42 #include <linux/notifier.h>
43 #include <linux/pfn.h>
44 #include <linux/ctype.h>
45 #include <linux/reboot.h>
46 #include <linux/topology.h>
47 #include <linux/kexec.h>
48 #include <linux/crash_dump.h>
49 #include <linux/memory.h>
50 #include <linux/compat.h>
51 #include <linux/start_kernel.h>
53 #include <asm/boot_data.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/stacktrace.h>
70 #include <asm/sysinfo.h>
72 #include <asm/alternative.h>
73 #include <asm/nospec-branch.h>
74 #include <asm/mem_detect.h>
76 #include <asm/asm-offsets.h>
82 unsigned int console_mode = 0;
83 EXPORT_SYMBOL(console_mode);
85 unsigned int console_devno = -1;
86 EXPORT_SYMBOL(console_devno);
88 unsigned int console_irq = -1;
89 EXPORT_SYMBOL(console_irq);
91 unsigned long elf_hwcap __read_mostly = 0;
92 char elf_platform[ELF_PLATFORM_SIZE];
94 unsigned long int_hwcap = 0;
96 int __bootdata(noexec_disabled);
97 int __bootdata(memory_end_set);
98 unsigned long __bootdata(memory_end);
99 unsigned long __bootdata(vmalloc_size);
100 unsigned long __bootdata(max_physmem_end);
101 struct mem_detect_info __bootdata(mem_detect);
103 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
104 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
105 unsigned long __bootdata_preserved(__swsusp_reset_dma);
106 unsigned long __bootdata_preserved(__stext_dma);
107 unsigned long __bootdata_preserved(__etext_dma);
108 unsigned long __bootdata_preserved(__sdma);
109 unsigned long __bootdata_preserved(__edma);
110 unsigned long __bootdata_preserved(__kaslr_offset);
111 unsigned int __bootdata_preserved(zlib_dfltcc_support);
112 EXPORT_SYMBOL(zlib_dfltcc_support);
114 unsigned long VMALLOC_START;
115 EXPORT_SYMBOL(VMALLOC_START);
117 unsigned long VMALLOC_END;
118 EXPORT_SYMBOL(VMALLOC_END);
120 struct page *vmemmap;
121 EXPORT_SYMBOL(vmemmap);
122 unsigned long vmemmap_size;
124 unsigned long MODULES_VADDR;
125 unsigned long MODULES_END;
127 /* An array with a pointer to the lowcore of every CPU. */
128 struct lowcore *lowcore_ptr[NR_CPUS];
129 EXPORT_SYMBOL(lowcore_ptr);
132 * The Write Back bit position in the physaddr is given by the SLPC PCI.
133 * Leaving the mask zero always uses write through which is safe
135 unsigned long mio_wb_bit_mask __ro_after_init;
138 * This is set up by the setup-routine at boot-time
139 * for S390 need to find out, what we have to setup
140 * using address 0x10400 ...
143 #include <asm/setup.h>
146 * condev= and conmode= setup parameter.
149 static int __init condev_setup(char *str)
153 vdev = simple_strtoul(str, &str, 0);
154 if (vdev >= 0 && vdev < 65536) {
155 console_devno = vdev;
161 __setup("condev=", condev_setup);
163 static void __init set_preferred_console(void)
165 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
166 add_preferred_console("ttyS", 0, NULL);
167 else if (CONSOLE_IS_3270)
168 add_preferred_console("tty3270", 0, NULL);
169 else if (CONSOLE_IS_VT220)
170 add_preferred_console("ttyS", 1, NULL);
171 else if (CONSOLE_IS_HVC)
172 add_preferred_console("hvc", 0, NULL);
175 static int __init conmode_setup(char *str)
177 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
178 if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
181 #if defined(CONFIG_TN3215_CONSOLE)
182 if (!strcmp(str, "3215"))
185 #if defined(CONFIG_TN3270_CONSOLE)
186 if (!strcmp(str, "3270"))
189 set_preferred_console();
193 __setup("conmode=", conmode_setup);
195 static void __init conmode_default(void)
197 char query_buffer[1024];
201 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
202 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
203 ptr = strstr(query_buffer, "SUBCHANNEL =");
204 console_irq = simple_strtoul(ptr + 13, NULL, 16);
205 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
206 ptr = strstr(query_buffer, "CONMODE");
208 * Set the conmode to 3215 so that the device recognition
209 * will set the cu_type of the console to 3215. If the
210 * conmode is 3270 and we don't set it back then both
211 * 3215 and the 3270 driver will try to access the console
212 * device (3215 as console and 3270 as normal tty).
214 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
216 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
221 if (str_has_prefix(ptr + 8, "3270")) {
222 #if defined(CONFIG_TN3270_CONSOLE)
224 #elif defined(CONFIG_TN3215_CONSOLE)
226 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
229 } else if (str_has_prefix(ptr + 8, "3215")) {
230 #if defined(CONFIG_TN3215_CONSOLE)
232 #elif defined(CONFIG_TN3270_CONSOLE)
234 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
238 } else if (MACHINE_IS_KVM) {
239 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
241 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
246 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
252 #ifdef CONFIG_CRASH_DUMP
253 static void __init setup_zfcpdump(void)
255 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
259 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
260 console_loglevel = 2;
263 static inline void setup_zfcpdump(void) {}
264 #endif /* CONFIG_CRASH_DUMP */
267 * Reboot, halt and power_off stubs. They just call _machine_restart,
268 * _machine_halt or _machine_power_off.
271 void machine_restart(char *command)
273 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
275 * Only unblank the console if we are called in enabled
276 * context or a bust_spinlocks cleared the way for us.
279 _machine_restart(command);
282 void machine_halt(void)
284 if (!in_interrupt() || oops_in_progress)
286 * Only unblank the console if we are called in enabled
287 * context or a bust_spinlocks cleared the way for us.
293 void machine_power_off(void)
295 if (!in_interrupt() || oops_in_progress)
297 * Only unblank the console if we are called in enabled
298 * context or a bust_spinlocks cleared the way for us.
301 _machine_power_off();
305 * Dummy power off function.
307 void (*pm_power_off)(void) = machine_power_off;
308 EXPORT_SYMBOL_GPL(pm_power_off);
312 unsigned long stack_alloc(void)
314 #ifdef CONFIG_VMAP_STACK
315 return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE,
316 THREADINFO_GFP, NUMA_NO_NODE,
317 __builtin_return_address(0));
319 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
323 void stack_free(unsigned long stack)
325 #ifdef CONFIG_VMAP_STACK
326 vfree((void *) stack);
328 free_pages(stack, THREAD_SIZE_ORDER);
332 int __init arch_early_irq_init(void)
336 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
338 panic("Couldn't allocate async stack");
339 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
343 static int __init async_stack_realloc(void)
345 unsigned long old, new;
347 old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
350 panic("Couldn't allocate async stack");
351 S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
352 free_pages(old, THREAD_SIZE_ORDER);
355 early_initcall(async_stack_realloc);
357 void __init arch_call_rest_init(void)
361 stack = stack_alloc();
363 panic("Couldn't allocate kernel stack");
364 current->stack = (void *) stack;
365 #ifdef CONFIG_VMAP_STACK
366 current->stack_vm_area = (void *) stack;
368 set_task_stack_end_magic(current);
369 stack += STACK_INIT_OFFSET;
370 S390_lowcore.kernel_stack = stack;
371 CALL_ON_STACK_NORETURN(rest_init, stack);
374 static void __init setup_lowcore_dat_off(void)
379 * Setup lowcore for boot cpu
381 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
382 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
384 panic("%s: Failed to allocate %zu bytes align=%zx\n",
385 __func__, sizeof(*lc), sizeof(*lc));
387 lc->restart_psw.mask = PSW_KERNEL_BITS;
388 lc->restart_psw.addr = (unsigned long) restart_int_handler;
389 lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
390 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
391 lc->svc_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
392 lc->svc_new_psw.addr = (unsigned long) system_call;
393 lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
394 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
395 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
396 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
397 lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
398 lc->io_new_psw.addr = (unsigned long) io_int_handler;
399 lc->clock_comparator = clock_comparator_max;
400 lc->nodat_stack = ((unsigned long) &init_thread_union)
401 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
402 lc->current_task = (unsigned long)&init_task;
404 lc->machine_flags = S390_lowcore.machine_flags;
405 lc->preempt_count = S390_lowcore.preempt_count;
406 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
407 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
408 sizeof(lc->stfle_fac_list));
409 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
410 sizeof(lc->alt_stfle_fac_list));
411 nmi_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);
427 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
428 __func__, THREAD_SIZE, THREAD_SIZE);
429 restart_stack += STACK_INIT_OFFSET;
432 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
433 * restart data to the absolute zero lowcore. This is necessary if
434 * PSW restart is done on an offline CPU that has lowcore zero.
436 lc->restart_stack = (unsigned long) restart_stack;
437 lc->restart_fn = (unsigned long) do_restart;
438 lc->restart_data = 0;
439 lc->restart_source = -1UL;
441 /* Setup absolute zero lowcore */
442 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
443 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
444 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
445 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
446 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
448 lc->spinlock_lockval = arch_spin_lockval(0);
449 lc->spinlock_index = 0;
450 arch_spin_lock_setup(0);
451 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
452 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
453 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
455 set_prefix((u32)(unsigned long) lc);
459 static void __init setup_lowcore_dat_on(void)
461 __ctl_clear_bit(0, 28);
462 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
463 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
464 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
465 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
466 __ctl_set_bit(0, 28);
469 static struct resource code_resource = {
470 .name = "Kernel code",
471 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
474 static struct resource data_resource = {
475 .name = "Kernel data",
476 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
479 static struct resource bss_resource = {
480 .name = "Kernel bss",
481 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
484 static struct resource __initdata *standard_resources[] = {
490 static void __init setup_resources(void)
492 struct resource *res, *std_res, *sub_res;
493 struct memblock_region *reg;
496 code_resource.start = (unsigned long) _text;
497 code_resource.end = (unsigned long) _etext - 1;
498 data_resource.start = (unsigned long) _etext;
499 data_resource.end = (unsigned long) _edata - 1;
500 bss_resource.start = (unsigned long) __bss_start;
501 bss_resource.end = (unsigned long) __bss_stop - 1;
503 for_each_memblock(memory, reg) {
504 res = memblock_alloc(sizeof(*res), 8);
506 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
507 __func__, sizeof(*res), 8);
508 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
510 res->name = "System RAM";
511 res->start = reg->base;
512 res->end = reg->base + reg->size - 1;
513 request_resource(&iomem_resource, res);
515 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
516 std_res = standard_resources[j];
517 if (std_res->start < res->start ||
518 std_res->start > res->end)
520 if (std_res->end > res->end) {
521 sub_res = memblock_alloc(sizeof(*sub_res), 8);
523 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
524 __func__, sizeof(*sub_res), 8);
526 sub_res->end = res->end;
527 std_res->start = res->end + 1;
528 request_resource(res, sub_res);
530 request_resource(res, std_res);
534 #ifdef CONFIG_CRASH_DUMP
536 * Re-add removed crash kernel memory as reserved memory. This makes
537 * sure it will be mapped with the identity mapping and struct pages
538 * will be created, so it can be resized later on.
539 * However add it later since the crash kernel resource should not be
540 * part of the System RAM resource.
542 if (crashk_res.end) {
543 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
544 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
545 insert_resource(&iomem_resource, &crashk_res);
550 static void __init setup_memory_end(void)
552 unsigned long vmax, tmp;
554 /* Choose kernel address space layout: 3 or 4 levels. */
555 if (IS_ENABLED(CONFIG_KASAN)) {
556 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
560 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
561 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
562 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
563 vmax = _REGION2_SIZE; /* 3-level kernel page table */
565 vmax = _REGION1_SIZE; /* 4-level kernel page table */
568 if (is_prot_virt_host())
569 adjust_to_uv_max(&vmax);
571 /* module area is at the end of the kernel address space. */
573 MODULES_VADDR = MODULES_END - MODULES_LEN;
574 VMALLOC_END = MODULES_VADDR;
575 VMALLOC_START = VMALLOC_END - vmalloc_size;
577 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
578 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
579 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
580 tmp = SECTION_ALIGN_UP(tmp);
581 tmp = VMALLOC_START - tmp * sizeof(struct page);
582 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
583 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
584 vmemmap = (struct page *) tmp;
586 /* Take care that memory_end is set and <= vmemmap */
587 memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
589 memory_end = min(memory_end, KASAN_SHADOW_START);
591 vmemmap_size = SECTION_ALIGN_UP(memory_end / PAGE_SIZE) * sizeof(struct page);
593 /* move vmemmap above kasan shadow only if stands in a way */
594 if (KASAN_SHADOW_END > (unsigned long)vmemmap &&
595 (unsigned long)vmemmap + vmemmap_size > KASAN_SHADOW_START)
596 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
598 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
599 memblock_remove(memory_end, ULONG_MAX);
601 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
604 #ifdef CONFIG_CRASH_DUMP
607 * When kdump is enabled, we have to ensure that no memory from the area
608 * [0 - crashkernel memory size] is set offline - it will be exchanged with
609 * the crashkernel memory region when kdump is triggered. The crashkernel
610 * memory region can never get offlined (pages are unmovable).
612 static int kdump_mem_notifier(struct notifier_block *nb,
613 unsigned long action, void *data)
615 struct memory_notify *arg = data;
617 if (action != MEM_GOING_OFFLINE)
619 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
624 static struct notifier_block kdump_mem_nb = {
625 .notifier_call = kdump_mem_notifier,
631 * Make sure that the area behind memory_end is protected
633 static void reserve_memory_end(void)
636 memblock_reserve(memory_end, ULONG_MAX);
640 * Make sure that oldmem, where the dump is stored, is protected
642 static void reserve_oldmem(void)
644 #ifdef CONFIG_CRASH_DUMP
646 /* Forget all memory above the running kdump system */
647 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
652 * Make sure that oldmem, where the dump is stored, is protected
654 static void remove_oldmem(void)
656 #ifdef CONFIG_CRASH_DUMP
658 /* Forget all memory above the running kdump system */
659 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
664 * Reserve memory for kdump kernel to be loaded with kexec
666 static void __init reserve_crashkernel(void)
668 #ifdef CONFIG_CRASH_DUMP
669 unsigned long long crash_base, crash_size;
670 phys_addr_t low, high;
673 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
676 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
677 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
678 if (rc || crash_size == 0)
681 if (memblock.memory.regions[0].size < crash_size) {
682 pr_info("crashkernel reservation failed: %s\n",
683 "first memory chunk must be at least crashkernel size");
687 low = crash_base ?: OLDMEM_BASE;
688 high = low + crash_size;
689 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
690 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
693 /* Find suitable area in free memory */
694 low = max_t(unsigned long, crash_size, sclp.hsa_size);
695 high = crash_base ? crash_base + crash_size : ULONG_MAX;
697 if (crash_base && crash_base < low) {
698 pr_info("crashkernel reservation failed: %s\n",
699 "crash_base too low");
702 low = crash_base ?: low;
703 crash_base = memblock_find_in_range(low, high, crash_size,
704 KEXEC_CRASH_MEM_ALIGN);
708 pr_info("crashkernel reservation failed: %s\n",
709 "no suitable area found");
713 if (register_memory_notifier(&kdump_mem_nb))
716 if (!OLDMEM_BASE && MACHINE_IS_VM)
717 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
718 crashk_res.start = crash_base;
719 crashk_res.end = crash_base + crash_size - 1;
720 memblock_remove(crash_base, crash_size);
721 pr_info("Reserving %lluMB of memory at %lluMB "
722 "for crashkernel (System RAM: %luMB)\n",
723 crash_size >> 20, crash_base >> 20,
724 (unsigned long)memblock.memory.total_size >> 20);
725 os_info_crashkernel_add(crash_base, crash_size);
730 * Reserve the initrd from being used by memblock
732 static void __init reserve_initrd(void)
734 #ifdef CONFIG_BLK_DEV_INITRD
735 if (!INITRD_START || !INITRD_SIZE)
737 initrd_start = INITRD_START;
738 initrd_end = initrd_start + INITRD_SIZE;
739 memblock_reserve(INITRD_START, INITRD_SIZE);
744 * Reserve the memory area used to pass the certificate lists
746 static void __init reserve_certificate_list(void)
748 if (ipl_cert_list_addr)
749 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
752 static void __init reserve_mem_detect_info(void)
754 unsigned long start, size;
756 get_mem_detect_reserved(&start, &size);
758 memblock_reserve(start, size);
761 static void __init free_mem_detect_info(void)
763 unsigned long start, size;
765 get_mem_detect_reserved(&start, &size);
767 memblock_free(start, size);
770 static const char * __init get_mem_info_source(void)
772 switch (mem_detect.info_source) {
773 case MEM_DETECT_SCLP_STOR_INFO:
774 return "sclp storage info";
775 case MEM_DETECT_DIAG260:
777 case MEM_DETECT_SCLP_READ_INFO:
778 return "sclp read info";
779 case MEM_DETECT_BIN_SEARCH:
780 return "binary search";
785 static void __init memblock_add_mem_detect_info(void)
787 unsigned long start, end;
790 memblock_dbg("physmem info source: %s (%hhd)\n",
791 get_mem_info_source(), mem_detect.info_source);
792 /* keep memblock lists close to the kernel */
793 memblock_set_bottom_up(true);
794 for_each_mem_detect_block(i, &start, &end) {
795 memblock_add(start, end - start);
796 memblock_physmem_add(start, end - start);
798 memblock_set_bottom_up(false);
799 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
804 * Check for initrd being in usable memory
806 static void __init check_initrd(void)
808 #ifdef CONFIG_BLK_DEV_INITRD
809 if (INITRD_START && INITRD_SIZE &&
810 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
811 pr_err("The initial RAM disk does not fit into the memory\n");
812 memblock_free(INITRD_START, INITRD_SIZE);
813 initrd_start = initrd_end = 0;
819 * Reserve memory used for lowcore/command line/kernel image.
821 static void __init reserve_kernel(void)
823 unsigned long start_pfn = PFN_UP(__pa(_end));
825 memblock_reserve(0, HEAD_END);
826 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
827 - (unsigned long)_stext);
828 memblock_reserve(__sdma, __edma - __sdma);
831 static void __init setup_memory(void)
833 struct memblock_region *reg;
836 * Init storage key for present memory
838 for_each_memblock(memory, reg) {
839 storage_key_init_range(reg->base, reg->base + reg->size);
841 psw_set_key(PAGE_DEFAULT_KEY);
844 memblock_enforce_memory_limit(memblock_end_of_DRAM());
848 * Setup hardware capabilities.
850 static int __init setup_hwcaps(void)
852 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
857 * The store facility list bits numbers as found in the principles
858 * of operation are numbered with bit 1UL<<31 as number 0 to
859 * bit 1UL<<0 as number 31.
860 * Bit 0: instructions named N3, "backported" to esa-mode
861 * Bit 2: z/Architecture mode is active
862 * Bit 7: the store-facility-list-extended facility is installed
863 * Bit 17: the message-security assist is installed
864 * Bit 19: the long-displacement facility is installed
865 * Bit 21: the extended-immediate facility is installed
866 * Bit 22: extended-translation facility 3 is installed
867 * Bit 30: extended-translation facility 3 enhancement facility
868 * These get translated to:
869 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
870 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
871 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
872 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
874 for (i = 0; i < 6; i++)
875 if (test_facility(stfl_bits[i]))
876 elf_hwcap |= 1UL << i;
878 if (test_facility(22) && test_facility(30))
879 elf_hwcap |= HWCAP_S390_ETF3EH;
882 * Check for additional facilities with store-facility-list-extended.
883 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
884 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
885 * as stored by stfl, bits 32-xxx contain additional facilities.
886 * How many facility words are stored depends on the number of
887 * doublewords passed to the instruction. The additional facilities
889 * Bit 42: decimal floating point facility is installed
890 * Bit 44: perform floating point operation facility is installed
892 * HWCAP_S390_DFP bit 6 (42 && 44).
894 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
895 elf_hwcap |= HWCAP_S390_DFP;
898 * Huge page support HWCAP_S390_HPAGE is bit 7.
900 if (MACHINE_HAS_EDAT1)
901 elf_hwcap |= HWCAP_S390_HPAGE;
904 * 64-bit register support for 31-bit processes
905 * HWCAP_S390_HIGH_GPRS is bit 9.
907 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
910 * Transactional execution support HWCAP_S390_TE is bit 10.
913 elf_hwcap |= HWCAP_S390_TE;
916 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
917 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
918 * instead of facility bit 129.
920 if (MACHINE_HAS_VX) {
921 elf_hwcap |= HWCAP_S390_VXRS;
922 if (test_facility(134))
923 elf_hwcap |= HWCAP_S390_VXRS_EXT;
924 if (test_facility(135))
925 elf_hwcap |= HWCAP_S390_VXRS_BCD;
926 if (test_facility(148))
927 elf_hwcap |= HWCAP_S390_VXRS_EXT2;
928 if (test_facility(152))
929 elf_hwcap |= HWCAP_S390_VXRS_PDE;
931 if (test_facility(150))
932 elf_hwcap |= HWCAP_S390_SORT;
933 if (test_facility(151))
934 elf_hwcap |= HWCAP_S390_DFLT;
937 * Guarded storage support HWCAP_S390_GS is bit 12.
940 elf_hwcap |= HWCAP_S390_GS;
943 add_device_randomness(&cpu_id, sizeof(cpu_id));
944 switch (cpu_id.machine) {
947 default: /* Use "z900" as default for 64 bit kernels. */
948 strcpy(elf_platform, "z900");
952 strcpy(elf_platform, "z990");
956 strcpy(elf_platform, "z9-109");
960 strcpy(elf_platform, "z10");
964 strcpy(elf_platform, "z196");
968 strcpy(elf_platform, "zEC12");
972 strcpy(elf_platform, "z13");
976 strcpy(elf_platform, "z14");
980 strcpy(elf_platform, "z15");
985 * Virtualization support HWCAP_INT_SIE is bit 0.
988 int_hwcap |= HWCAP_INT_SIE;
992 arch_initcall(setup_hwcaps);
995 * Add system information as device randomness
997 static void __init setup_randomness(void)
999 struct sysinfo_3_2_2 *vmms;
1001 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
1004 panic("Failed to allocate memory for sysinfo structure\n");
1006 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1007 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1008 memblock_free((unsigned long) vmms, PAGE_SIZE);
1012 * Find the correct size for the task_struct. This depends on
1013 * the size of the struct fpu at the end of the thread_struct
1014 * which is embedded in the task_struct.
1016 static void __init setup_task_size(void)
1018 int task_size = sizeof(struct task_struct);
1020 if (!MACHINE_HAS_VX) {
1021 task_size -= sizeof(__vector128) * __NUM_VXRS;
1022 task_size += sizeof(freg_t) * __NUM_FPRS;
1024 arch_task_struct_size = task_size;
1028 * Issue diagnose 318 to set the control program name and
1031 static void __init setup_control_program_code(void)
1033 union diag318_info diag318_info = {
1038 if (!sclp.has_diag318)
1041 diag_stat_inc(DIAG_STAT_X318);
1042 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1046 * Print the component list from the IPL report
1048 static void __init log_component_list(void)
1050 struct ipl_rb_component_entry *ptr, *end;
1053 if (!early_ipl_comp_list_addr)
1055 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1056 pr_info("Linux is running with Secure-IPL enabled\n");
1058 pr_info("Linux is running with Secure-IPL disabled\n");
1059 ptr = (void *) early_ipl_comp_list_addr;
1060 end = (void *) ptr + early_ipl_comp_list_size;
1061 pr_info("The IPL report contains the following components:\n");
1063 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1064 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1065 str = "signed, verified";
1067 str = "signed, verification failed";
1071 pr_info("%016llx - %016llx (%s)\n",
1072 ptr->addr, ptr->addr + ptr->len, str);
1078 * Setup function called from init/main.c just after the banner
1082 void __init setup_arch(char **cmdline_p)
1085 * print what head.S has found out about the machine
1088 pr_info("Linux is running as a z/VM "
1089 "guest operating system in 64-bit mode\n");
1090 else if (MACHINE_IS_KVM)
1091 pr_info("Linux is running under KVM in 64-bit mode\n");
1092 else if (MACHINE_IS_LPAR)
1093 pr_info("Linux is running natively in 64-bit mode\n");
1095 pr_info("Linux is running as a guest in 64-bit mode\n");
1097 log_component_list();
1099 /* Have one command line that is parsed and saved in /proc/cmdline */
1100 /* boot_command_line has been already set up in early.c */
1101 *cmdline_p = boot_command_line;
1103 ROOT_DEV = Root_RAM0;
1105 init_mm.start_code = (unsigned long) _text;
1106 init_mm.end_code = (unsigned long) _etext;
1107 init_mm.end_data = (unsigned long) _edata;
1108 init_mm.brk = (unsigned long) _end;
1110 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1111 nospec_auto_detect();
1114 parse_early_param();
1115 #ifdef CONFIG_CRASH_DUMP
1116 /* Deactivate elfcorehdr= kernel parameter */
1117 elfcorehdr_addr = ELFCORE_ADDR_MAX;
1123 setup_control_program_code();
1125 /* Do some memory reservations *before* memory is added to memblock */
1126 reserve_memory_end();
1130 reserve_certificate_list();
1131 reserve_mem_detect_info();
1132 memblock_allow_resize();
1134 /* Get information about *all* installed memory */
1135 memblock_add_mem_detect_info();
1137 free_mem_detect_info();
1140 if (is_prot_virt_host())
1144 dma_contiguous_reserve(memory_end);
1148 reserve_crashkernel();
1149 #ifdef CONFIG_CRASH_DUMP
1151 * Be aware that smp_save_dump_cpus() triggers a system reset.
1152 * Therefore CPU and device initialization should be done afterwards.
1154 smp_save_dump_cpus();
1158 setup_lowcore_dat_off();
1159 smp_fill_possible_mask();
1160 cpu_detect_mhz_feature();
1164 topology_init_early();
1167 * Create kernel page tables and switch to virtual addressing.
1172 * After paging_init created the kernel page table, the new PSWs
1173 * in lowcore can now run with DAT enabled.
1175 setup_lowcore_dat_on();
1177 /* Setup default console */
1179 set_preferred_console();
1181 apply_alternative_instructions();
1182 if (IS_ENABLED(CONFIG_EXPOLINE))
1183 nospec_init_branches();
1185 /* Setup zfcpdump support */
1188 /* Add system specific data to the random pool */