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-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>
54 #include <asm/boot_data.h>
56 #include <asm/facility.h>
58 #include <asm/mmu_context.h>
59 #include <asm/cpcmd.h>
60 #include <asm/lowcore.h>
64 #include <asm/ptrace.h>
65 #include <asm/sections.h>
66 #include <asm/ebcdic.h>
68 #include <asm/os_info.h>
70 #include <asm/stacktrace.h>
71 #include <asm/sysinfo.h>
73 #include <asm/alternative.h>
74 #include <asm/nospec-branch.h>
75 #include <asm/mem_detect.h>
77 #include <asm/asm-offsets.h>
83 unsigned int console_mode = 0;
84 EXPORT_SYMBOL(console_mode);
86 unsigned int console_devno = -1;
87 EXPORT_SYMBOL(console_devno);
89 unsigned int console_irq = -1;
90 EXPORT_SYMBOL(console_irq);
93 * Some code and data needs to stay below 2 GB, even when the kernel would be
94 * relocated above 2 GB, because it has to use 31 bit addresses.
95 * Such code and data is part of the .dma section.
97 unsigned long __dma_ref __sdma = __pa(&_sdma);
98 unsigned long __dma_ref __edma = __pa(&_edma);
99 unsigned long __dma_ref __stext_dma = __pa(&_stext_dma);
100 unsigned long __dma_ref __etext_dma = __pa(&_etext_dma);
101 struct exception_table_entry __dma_ref *__start_dma_ex_table = _start_dma_ex_table;
102 struct exception_table_entry __dma_ref *__stop_dma_ex_table = _stop_dma_ex_table;
105 * Control registers CR2, CR5 and CR15 are initialized with addresses
106 * of tables that must be placed below 2G which is handled by the DMA
108 * Because the DMA sections are relocated below 2G at startup,
109 * the content of control registers CR2, CR5 and CR15 must be updated
110 * with new addresses after the relocation. The initial initialization of
111 * control registers occurs in head64.S and then gets updated again after DMA
112 * relocation. We must access the relevant DMA tables indirectly via
113 * pointers placed in the .dma.refs linker section. Those pointers get
114 * updated automatically during DMA relocation and always contain a valid
115 * address within DMA sections.
118 static __dma_data u32 __ctl_duct_dma[16] __aligned(64);
120 static __dma_data u64 __ctl_aste_dma[8] __aligned(64) = {
121 [1] = 0xffffffffffffffff
124 static __dma_data u32 __ctl_duald_dma[32] __aligned(128) = {
135 static __dma_data u32 __ctl_linkage_stack_dma[8] __aligned(64) = {
140 static u64 __dma_ref *__ctl_aste = __ctl_aste_dma;
141 static u32 __dma_ref *__ctl_duald = __ctl_duald_dma;
142 static u32 __dma_ref *__ctl_linkage_stack = __ctl_linkage_stack_dma;
143 static u32 __dma_ref *__ctl_duct = __ctl_duct_dma;
145 int __bootdata(noexec_disabled);
146 unsigned long __bootdata(ident_map_size);
147 struct mem_detect_info __bootdata(mem_detect);
148 struct initrd_data __bootdata(initrd_data);
150 unsigned long __bootdata_preserved(__kaslr_offset);
151 unsigned int __bootdata_preserved(zlib_dfltcc_support);
152 EXPORT_SYMBOL(zlib_dfltcc_support);
153 u64 __bootdata_preserved(stfle_fac_list[16]);
154 EXPORT_SYMBOL(stfle_fac_list);
155 u64 __bootdata_preserved(alt_stfle_fac_list[16]);
156 struct oldmem_data __bootdata_preserved(oldmem_data);
158 unsigned long VMALLOC_START;
159 EXPORT_SYMBOL(VMALLOC_START);
161 unsigned long VMALLOC_END;
162 EXPORT_SYMBOL(VMALLOC_END);
164 struct page *vmemmap;
165 EXPORT_SYMBOL(vmemmap);
166 unsigned long vmemmap_size;
168 unsigned long MODULES_VADDR;
169 unsigned long MODULES_END;
171 /* An array with a pointer to the lowcore of every CPU. */
172 struct lowcore *lowcore_ptr[NR_CPUS];
173 EXPORT_SYMBOL(lowcore_ptr);
176 * The Write Back bit position in the physaddr is given by the SLPC PCI.
177 * Leaving the mask zero always uses write through which is safe
179 unsigned long mio_wb_bit_mask __ro_after_init;
182 * This is set up by the setup-routine at boot-time
183 * for S390 need to find out, what we have to setup
184 * using address 0x10400 ...
187 #include <asm/setup.h>
190 * condev= and conmode= setup parameter.
193 static int __init condev_setup(char *str)
197 vdev = simple_strtoul(str, &str, 0);
198 if (vdev >= 0 && vdev < 65536) {
199 console_devno = vdev;
205 __setup("condev=", condev_setup);
207 static void __init set_preferred_console(void)
209 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
210 add_preferred_console("ttyS", 0, NULL);
211 else if (CONSOLE_IS_3270)
212 add_preferred_console("tty3270", 0, NULL);
213 else if (CONSOLE_IS_VT220)
214 add_preferred_console("ttysclp", 0, NULL);
215 else if (CONSOLE_IS_HVC)
216 add_preferred_console("hvc", 0, NULL);
219 static int __init conmode_setup(char *str)
221 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
222 if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
225 #if defined(CONFIG_TN3215_CONSOLE)
226 if (!strcmp(str, "3215"))
229 #if defined(CONFIG_TN3270_CONSOLE)
230 if (!strcmp(str, "3270"))
233 set_preferred_console();
237 __setup("conmode=", conmode_setup);
239 static void __init conmode_default(void)
241 char query_buffer[1024];
245 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
246 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
247 ptr = strstr(query_buffer, "SUBCHANNEL =");
248 console_irq = simple_strtoul(ptr + 13, NULL, 16);
249 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
250 ptr = strstr(query_buffer, "CONMODE");
252 * Set the conmode to 3215 so that the device recognition
253 * will set the cu_type of the console to 3215. If the
254 * conmode is 3270 and we don't set it back then both
255 * 3215 and the 3270 driver will try to access the console
256 * device (3215 as console and 3270 as normal tty).
258 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
260 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
265 if (str_has_prefix(ptr + 8, "3270")) {
266 #if defined(CONFIG_TN3270_CONSOLE)
268 #elif defined(CONFIG_TN3215_CONSOLE)
270 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
273 } else if (str_has_prefix(ptr + 8, "3215")) {
274 #if defined(CONFIG_TN3215_CONSOLE)
276 #elif defined(CONFIG_TN3270_CONSOLE)
278 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
282 } else if (MACHINE_IS_KVM) {
283 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
285 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
290 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
296 #ifdef CONFIG_CRASH_DUMP
297 static void __init setup_zfcpdump(void)
299 if (!is_ipl_type_dump())
301 if (oldmem_data.start)
303 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
304 console_loglevel = 2;
307 static inline void setup_zfcpdump(void) {}
308 #endif /* CONFIG_CRASH_DUMP */
311 * Reboot, halt and power_off stubs. They just call _machine_restart,
312 * _machine_halt or _machine_power_off.
315 void machine_restart(char *command)
317 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
319 * Only unblank the console if we are called in enabled
320 * context or a bust_spinlocks cleared the way for us.
323 _machine_restart(command);
326 void machine_halt(void)
328 if (!in_interrupt() || oops_in_progress)
330 * Only unblank the console if we are called in enabled
331 * context or a bust_spinlocks cleared the way for us.
337 void machine_power_off(void)
339 if (!in_interrupt() || oops_in_progress)
341 * Only unblank the console if we are called in enabled
342 * context or a bust_spinlocks cleared the way for us.
345 _machine_power_off();
349 * Dummy power off function.
351 void (*pm_power_off)(void) = machine_power_off;
352 EXPORT_SYMBOL_GPL(pm_power_off);
356 unsigned long stack_alloc(void)
358 #ifdef CONFIG_VMAP_STACK
359 return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE,
360 THREADINFO_GFP, NUMA_NO_NODE,
361 __builtin_return_address(0));
363 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
367 void stack_free(unsigned long stack)
369 #ifdef CONFIG_VMAP_STACK
370 vfree((void *) stack);
372 free_pages(stack, THREAD_SIZE_ORDER);
376 int __init arch_early_irq_init(void)
380 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
382 panic("Couldn't allocate async stack");
383 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
387 void __init arch_call_rest_init(void)
391 stack = stack_alloc();
393 panic("Couldn't allocate kernel stack");
394 current->stack = (void *) stack;
395 #ifdef CONFIG_VMAP_STACK
396 current->stack_vm_area = (void *) stack;
398 set_task_stack_end_magic(current);
399 stack += STACK_INIT_OFFSET;
400 S390_lowcore.kernel_stack = stack;
401 call_on_stack_noreturn(rest_init, stack);
404 static void __init setup_lowcore_dat_off(void)
406 unsigned long int_psw_mask = PSW_KERNEL_BITS;
407 unsigned long mcck_stack;
410 if (IS_ENABLED(CONFIG_KASAN))
411 int_psw_mask |= PSW_MASK_DAT;
414 * Setup lowcore for boot cpu
416 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
417 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
419 panic("%s: Failed to allocate %zu bytes align=%zx\n",
420 __func__, sizeof(*lc), sizeof(*lc));
422 lc->restart_psw.mask = PSW_KERNEL_BITS;
423 lc->restart_psw.addr = (unsigned long) restart_int_handler;
424 lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
425 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
426 lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
427 lc->svc_new_psw.addr = (unsigned long) system_call;
428 lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
429 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
430 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
431 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
432 lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
433 lc->io_new_psw.addr = (unsigned long) io_int_handler;
434 lc->clock_comparator = clock_comparator_max;
435 lc->nodat_stack = ((unsigned long) &init_thread_union)
436 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
437 lc->current_task = (unsigned long)&init_task;
439 lc->machine_flags = S390_lowcore.machine_flags;
440 lc->preempt_count = S390_lowcore.preempt_count;
441 nmi_alloc_boot_cpu(lc);
442 lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
443 lc->exit_timer = S390_lowcore.exit_timer;
444 lc->user_timer = S390_lowcore.user_timer;
445 lc->system_timer = S390_lowcore.system_timer;
446 lc->steal_timer = S390_lowcore.steal_timer;
447 lc->last_update_timer = S390_lowcore.last_update_timer;
448 lc->last_update_clock = S390_lowcore.last_update_clock;
451 * Allocate the global restart stack which is the same for
452 * all CPUs in cast *one* of them does a PSW restart.
454 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
456 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
457 __func__, THREAD_SIZE, THREAD_SIZE);
458 restart_stack += STACK_INIT_OFFSET;
461 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
462 * restart data to the absolute zero lowcore. This is necessary if
463 * PSW restart is done on an offline CPU that has lowcore zero.
465 lc->restart_stack = (unsigned long) restart_stack;
466 lc->restart_fn = (unsigned long) do_restart;
467 lc->restart_data = 0;
468 lc->restart_source = -1UL;
470 mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
472 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
473 __func__, THREAD_SIZE, THREAD_SIZE);
474 lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
476 /* Setup absolute zero lowcore */
477 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
478 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
479 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
480 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
481 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
483 lc->spinlock_lockval = arch_spin_lockval(0);
484 lc->spinlock_index = 0;
485 arch_spin_lock_setup(0);
486 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
487 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
488 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
489 lc->preempt_count = PREEMPT_DISABLED;
491 set_prefix((u32)(unsigned long) lc);
495 static void __init setup_lowcore_dat_on(void)
497 __ctl_clear_bit(0, 28);
498 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
499 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
500 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
501 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
502 __ctl_set_bit(0, 28);
505 static struct resource code_resource = {
506 .name = "Kernel code",
507 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
510 static struct resource data_resource = {
511 .name = "Kernel data",
512 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
515 static struct resource bss_resource = {
516 .name = "Kernel bss",
517 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
520 static struct resource __initdata *standard_resources[] = {
526 static void __init setup_resources(void)
528 struct resource *res, *std_res, *sub_res;
529 phys_addr_t start, end;
533 code_resource.start = (unsigned long) _text;
534 code_resource.end = (unsigned long) _etext - 1;
535 data_resource.start = (unsigned long) _etext;
536 data_resource.end = (unsigned long) _edata - 1;
537 bss_resource.start = (unsigned long) __bss_start;
538 bss_resource.end = (unsigned long) __bss_stop - 1;
540 for_each_mem_range(i, &start, &end) {
541 res = memblock_alloc(sizeof(*res), 8);
543 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
544 __func__, sizeof(*res), 8);
545 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
547 res->name = "System RAM";
550 * In memblock, end points to the first byte after the
551 * range while in resourses, end points to the last byte in
555 request_resource(&iomem_resource, res);
557 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
558 std_res = standard_resources[j];
559 if (std_res->start < res->start ||
560 std_res->start > res->end)
562 if (std_res->end > res->end) {
563 sub_res = memblock_alloc(sizeof(*sub_res), 8);
565 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
566 __func__, sizeof(*sub_res), 8);
568 sub_res->end = res->end;
569 std_res->start = res->end + 1;
570 request_resource(res, sub_res);
572 request_resource(res, std_res);
576 #ifdef CONFIG_CRASH_DUMP
578 * Re-add removed crash kernel memory as reserved memory. This makes
579 * sure it will be mapped with the identity mapping and struct pages
580 * will be created, so it can be resized later on.
581 * However add it later since the crash kernel resource should not be
582 * part of the System RAM resource.
584 if (crashk_res.end) {
585 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
586 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
587 insert_resource(&iomem_resource, &crashk_res);
592 static void __init setup_memory_end(void)
594 memblock_remove(ident_map_size, ULONG_MAX);
595 max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
596 pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
599 #ifdef CONFIG_CRASH_DUMP
602 * When kdump is enabled, we have to ensure that no memory from the area
603 * [0 - crashkernel memory size] is set offline - it will be exchanged with
604 * the crashkernel memory region when kdump is triggered. The crashkernel
605 * memory region can never get offlined (pages are unmovable).
607 static int kdump_mem_notifier(struct notifier_block *nb,
608 unsigned long action, void *data)
610 struct memory_notify *arg = data;
612 if (action != MEM_GOING_OFFLINE)
614 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
619 static struct notifier_block kdump_mem_nb = {
620 .notifier_call = kdump_mem_notifier,
626 * Make sure that the area above identity mapping is protected
628 static void __init reserve_above_ident_map(void)
630 memblock_reserve(ident_map_size, ULONG_MAX);
634 * Reserve memory for kdump kernel to be loaded with kexec
636 static void __init reserve_crashkernel(void)
638 #ifdef CONFIG_CRASH_DUMP
639 unsigned long long crash_base, crash_size;
640 phys_addr_t low, high;
643 rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
646 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
647 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
648 if (rc || crash_size == 0)
651 if (memblock.memory.regions[0].size < crash_size) {
652 pr_info("crashkernel reservation failed: %s\n",
653 "first memory chunk must be at least crashkernel size");
657 low = crash_base ?: oldmem_data.start;
658 high = low + crash_size;
659 if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
660 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
663 /* Find suitable area in free memory */
664 low = max_t(unsigned long, crash_size, sclp.hsa_size);
665 high = crash_base ? crash_base + crash_size : ULONG_MAX;
667 if (crash_base && crash_base < low) {
668 pr_info("crashkernel reservation failed: %s\n",
669 "crash_base too low");
672 low = crash_base ?: low;
673 crash_base = memblock_find_in_range(low, high, crash_size,
674 KEXEC_CRASH_MEM_ALIGN);
678 pr_info("crashkernel reservation failed: %s\n",
679 "no suitable area found");
683 if (register_memory_notifier(&kdump_mem_nb))
686 if (!oldmem_data.start && MACHINE_IS_VM)
687 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
688 crashk_res.start = crash_base;
689 crashk_res.end = crash_base + crash_size - 1;
690 memblock_remove(crash_base, crash_size);
691 pr_info("Reserving %lluMB of memory at %lluMB "
692 "for crashkernel (System RAM: %luMB)\n",
693 crash_size >> 20, crash_base >> 20,
694 (unsigned long)memblock.memory.total_size >> 20);
695 os_info_crashkernel_add(crash_base, crash_size);
700 * Reserve the initrd from being used by memblock
702 static void __init reserve_initrd(void)
704 #ifdef CONFIG_BLK_DEV_INITRD
705 if (!initrd_data.start || !initrd_data.size)
707 initrd_start = initrd_data.start;
708 initrd_end = initrd_start + initrd_data.size;
709 memblock_reserve(initrd_data.start, initrd_data.size);
714 * Reserve the memory area used to pass the certificate lists
716 static void __init reserve_certificate_list(void)
718 if (ipl_cert_list_addr)
719 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
722 static void __init reserve_mem_detect_info(void)
724 unsigned long start, size;
726 get_mem_detect_reserved(&start, &size);
728 memblock_reserve(start, size);
731 static void __init free_mem_detect_info(void)
733 unsigned long start, size;
735 get_mem_detect_reserved(&start, &size);
737 memblock_free(start, size);
740 static const char * __init get_mem_info_source(void)
742 switch (mem_detect.info_source) {
743 case MEM_DETECT_SCLP_STOR_INFO:
744 return "sclp storage info";
745 case MEM_DETECT_DIAG260:
747 case MEM_DETECT_SCLP_READ_INFO:
748 return "sclp read info";
749 case MEM_DETECT_BIN_SEARCH:
750 return "binary search";
755 static void __init memblock_add_mem_detect_info(void)
757 unsigned long start, end;
760 pr_debug("physmem info source: %s (%hhd)\n",
761 get_mem_info_source(), mem_detect.info_source);
762 /* keep memblock lists close to the kernel */
763 memblock_set_bottom_up(true);
764 for_each_mem_detect_block(i, &start, &end) {
765 memblock_add(start, end - start);
766 memblock_physmem_add(start, end - start);
768 memblock_set_bottom_up(false);
769 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
774 * Check for initrd being in usable memory
776 static void __init check_initrd(void)
778 #ifdef CONFIG_BLK_DEV_INITRD
779 if (initrd_data.start && initrd_data.size &&
780 !memblock_is_region_memory(initrd_data.start, initrd_data.size)) {
781 pr_err("The initial RAM disk does not fit into the memory\n");
782 memblock_free(initrd_data.start, initrd_data.size);
783 initrd_start = initrd_end = 0;
789 * Reserve memory used for lowcore/command line/kernel image.
791 static void __init reserve_kernel(void)
793 unsigned long start_pfn = PFN_UP(__pa(_end));
795 memblock_reserve(0, STARTUP_NORMAL_OFFSET);
796 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
797 - (unsigned long)_stext);
800 static void __init setup_memory(void)
802 phys_addr_t start, end;
806 * Init storage key for present memory
808 for_each_mem_range(i, &start, &end)
809 storage_key_init_range(start, end);
811 psw_set_key(PAGE_DEFAULT_KEY);
814 memblock_enforce_memory_limit(memblock_end_of_DRAM());
817 static void __init relocate_dma_section(void)
819 unsigned long dma_addr, dma_size;
823 /* Allocate a new DMA capable memory region */
824 dma_size = __edma - __sdma;
825 pr_info("Relocating DMA section of size 0x%08lx\n", dma_size);
826 dma_addr = (unsigned long)memblock_alloc_low(dma_size, PAGE_SIZE);
828 panic("Failed to allocate memory for DMA section\n");
829 dma_offset = dma_addr - __sdma;
831 /* Move original DMA section to the new one */
832 memmove((void *)dma_addr, (void *)__sdma, dma_size);
833 /* Zero out the old DMA section to catch invalid accesses within it */
834 memset((void *)__sdma, 0, dma_size);
836 /* Update all DMA region references */
837 for (ptr = _start_dma_refs; ptr != _end_dma_refs; ptr++)
841 /* This must be called after DMA relocation */
842 static void __init setup_cr(void)
848 __ctl_duct[1] = (unsigned long)__ctl_aste;
849 __ctl_duct[2] = (unsigned long)__ctl_aste;
850 __ctl_duct[4] = (unsigned long)__ctl_duald;
852 /* Update control registers CR2, CR5 and CR15 */
853 __ctl_store(cr2.val, 2, 2);
854 __ctl_store(cr5.val, 5, 5);
855 __ctl_store(cr15.val, 15, 15);
856 cr2.ducto = (unsigned long)__ctl_duct >> 6;
857 cr5.pasteo = (unsigned long)__ctl_duct >> 6;
858 cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
859 __ctl_load(cr2.val, 2, 2);
860 __ctl_load(cr5.val, 5, 5);
861 __ctl_load(cr15.val, 15, 15);
865 * Add system information as device randomness
867 static void __init setup_randomness(void)
869 struct sysinfo_3_2_2 *vmms;
871 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
874 panic("Failed to allocate memory for sysinfo structure\n");
876 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
877 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
878 memblock_free((unsigned long) vmms, PAGE_SIZE);
882 * Find the correct size for the task_struct. This depends on
883 * the size of the struct fpu at the end of the thread_struct
884 * which is embedded in the task_struct.
886 static void __init setup_task_size(void)
888 int task_size = sizeof(struct task_struct);
890 if (!MACHINE_HAS_VX) {
891 task_size -= sizeof(__vector128) * __NUM_VXRS;
892 task_size += sizeof(freg_t) * __NUM_FPRS;
894 arch_task_struct_size = task_size;
898 * Issue diagnose 318 to set the control program name and
901 static void __init setup_control_program_code(void)
903 union diag318_info diag318_info = {
908 if (!sclp.has_diag318)
911 diag_stat_inc(DIAG_STAT_X318);
912 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
916 * Print the component list from the IPL report
918 static void __init log_component_list(void)
920 struct ipl_rb_component_entry *ptr, *end;
923 if (!early_ipl_comp_list_addr)
925 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
926 pr_info("Linux is running with Secure-IPL enabled\n");
928 pr_info("Linux is running with Secure-IPL disabled\n");
929 ptr = (void *) early_ipl_comp_list_addr;
930 end = (void *) ptr + early_ipl_comp_list_size;
931 pr_info("The IPL report contains the following components:\n");
933 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
934 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
935 str = "signed, verified";
937 str = "signed, verification failed";
941 pr_info("%016llx - %016llx (%s)\n",
942 ptr->addr, ptr->addr + ptr->len, str);
948 * Setup function called from init/main.c just after the banner
952 void __init setup_arch(char **cmdline_p)
955 * print what head.S has found out about the machine
958 pr_info("Linux is running as a z/VM "
959 "guest operating system in 64-bit mode\n");
960 else if (MACHINE_IS_KVM)
961 pr_info("Linux is running under KVM in 64-bit mode\n");
962 else if (MACHINE_IS_LPAR)
963 pr_info("Linux is running natively in 64-bit mode\n");
965 pr_info("Linux is running as a guest in 64-bit mode\n");
967 log_component_list();
969 /* Have one command line that is parsed and saved in /proc/cmdline */
970 /* boot_command_line has been already set up in early.c */
971 *cmdline_p = boot_command_line;
973 ROOT_DEV = Root_RAM0;
975 setup_initial_init_mm(_text, _etext, _edata, _end);
977 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
978 nospec_auto_detect();
982 #ifdef CONFIG_CRASH_DUMP
983 /* Deactivate elfcorehdr= kernel parameter */
984 elfcorehdr_addr = ELFCORE_ADDR_MAX;
990 setup_control_program_code();
992 /* Do some memory reservations *before* memory is added to memblock */
993 reserve_above_ident_map();
996 reserve_certificate_list();
997 reserve_mem_detect_info();
998 memblock_allow_resize();
1000 /* Get information about *all* installed memory */
1001 memblock_add_mem_detect_info();
1003 free_mem_detect_info();
1005 relocate_dma_section();
1011 dma_contiguous_reserve(ident_map_size);
1013 if (MACHINE_HAS_EDAT2)
1014 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1017 reserve_crashkernel();
1018 #ifdef CONFIG_CRASH_DUMP
1020 * Be aware that smp_save_dump_cpus() triggers a system reset.
1021 * Therefore CPU and device initialization should be done afterwards.
1023 smp_save_dump_cpus();
1027 setup_lowcore_dat_off();
1028 smp_fill_possible_mask();
1029 cpu_detect_mhz_feature();
1033 topology_init_early();
1036 * Create kernel page tables and switch to virtual addressing.
1041 * After paging_init created the kernel page table, the new PSWs
1042 * in lowcore can now run with DAT enabled.
1044 setup_lowcore_dat_on();
1046 /* Setup default console */
1048 set_preferred_console();
1050 apply_alternative_instructions();
1051 if (IS_ENABLED(CONFIG_EXPOLINE))
1052 nospec_init_branches();
1054 /* Setup zfcp/nvme dump support */
1057 /* Add system specific data to the random pool */