xref: /linux/arch/s390/kernel/setup.c (revision 9abdb50cda0ffe33bbb2e40cbad97b32fb7ff892)
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-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>
52 
53 #include <asm/ipl.h>
54 #include <asm/facility.h>
55 #include <asm/smp.h>
56 #include <asm/mmu_context.h>
57 #include <asm/cpcmd.h>
58 #include <asm/lowcore.h>
59 #include <asm/nmi.h>
60 #include <asm/irq.h>
61 #include <asm/page.h>
62 #include <asm/ptrace.h>
63 #include <asm/sections.h>
64 #include <asm/ebcdic.h>
65 #include <asm/diag.h>
66 #include <asm/os_info.h>
67 #include <asm/sclp.h>
68 #include <asm/sysinfo.h>
69 #include <asm/numa.h>
70 #include <asm/alternative.h>
71 #include <asm/nospec-branch.h>
72 #include <asm/mem_detect.h>
73 #include "entry.h"
74 
75 /*
76  * Machine setup..
77  */
78 unsigned int console_mode = 0;
79 EXPORT_SYMBOL(console_mode);
80 
81 unsigned int console_devno = -1;
82 EXPORT_SYMBOL(console_devno);
83 
84 unsigned int console_irq = -1;
85 EXPORT_SYMBOL(console_irq);
86 
87 unsigned long elf_hwcap __read_mostly = 0;
88 char elf_platform[ELF_PLATFORM_SIZE];
89 
90 unsigned long int_hwcap = 0;
91 
92 int __bootdata(noexec_disabled);
93 int __bootdata(memory_end_set);
94 unsigned long __bootdata(memory_end);
95 unsigned long __bootdata(max_physmem_end);
96 struct mem_detect_info __bootdata(mem_detect);
97 
98 unsigned long VMALLOC_START;
99 EXPORT_SYMBOL(VMALLOC_START);
100 
101 unsigned long VMALLOC_END;
102 EXPORT_SYMBOL(VMALLOC_END);
103 
104 struct page *vmemmap;
105 EXPORT_SYMBOL(vmemmap);
106 
107 unsigned long MODULES_VADDR;
108 unsigned long MODULES_END;
109 
110 /* An array with a pointer to the lowcore of every CPU. */
111 struct lowcore *lowcore_ptr[NR_CPUS];
112 EXPORT_SYMBOL(lowcore_ptr);
113 
114 /*
115  * This is set up by the setup-routine at boot-time
116  * for S390 need to find out, what we have to setup
117  * using address 0x10400 ...
118  */
119 
120 #include <asm/setup.h>
121 
122 /*
123  * condev= and conmode= setup parameter.
124  */
125 
126 static int __init condev_setup(char *str)
127 {
128 	int vdev;
129 
130 	vdev = simple_strtoul(str, &str, 0);
131 	if (vdev >= 0 && vdev < 65536) {
132 		console_devno = vdev;
133 		console_irq = -1;
134 	}
135 	return 1;
136 }
137 
138 __setup("condev=", condev_setup);
139 
140 static void __init set_preferred_console(void)
141 {
142 	if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
143 		add_preferred_console("ttyS", 0, NULL);
144 	else if (CONSOLE_IS_3270)
145 		add_preferred_console("tty3270", 0, NULL);
146 	else if (CONSOLE_IS_VT220)
147 		add_preferred_console("ttyS", 1, NULL);
148 	else if (CONSOLE_IS_HVC)
149 		add_preferred_console("hvc", 0, NULL);
150 }
151 
152 static int __init conmode_setup(char *str)
153 {
154 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
155 	if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
156                 SET_CONSOLE_SCLP;
157 #endif
158 #if defined(CONFIG_TN3215_CONSOLE)
159 	if (strncmp(str, "3215", 5) == 0)
160 		SET_CONSOLE_3215;
161 #endif
162 #if defined(CONFIG_TN3270_CONSOLE)
163 	if (strncmp(str, "3270", 5) == 0)
164 		SET_CONSOLE_3270;
165 #endif
166 	set_preferred_console();
167         return 1;
168 }
169 
170 __setup("conmode=", conmode_setup);
171 
172 static void __init conmode_default(void)
173 {
174 	char query_buffer[1024];
175 	char *ptr;
176 
177         if (MACHINE_IS_VM) {
178 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
179 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
180 		ptr = strstr(query_buffer, "SUBCHANNEL =");
181 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
182 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
183 		ptr = strstr(query_buffer, "CONMODE");
184 		/*
185 		 * Set the conmode to 3215 so that the device recognition
186 		 * will set the cu_type of the console to 3215. If the
187 		 * conmode is 3270 and we don't set it back then both
188 		 * 3215 and the 3270 driver will try to access the console
189 		 * device (3215 as console and 3270 as normal tty).
190 		 */
191 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
192 		if (ptr == NULL) {
193 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
194 			SET_CONSOLE_SCLP;
195 #endif
196 			return;
197 		}
198 		if (strncmp(ptr + 8, "3270", 4) == 0) {
199 #if defined(CONFIG_TN3270_CONSOLE)
200 			SET_CONSOLE_3270;
201 #elif defined(CONFIG_TN3215_CONSOLE)
202 			SET_CONSOLE_3215;
203 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
204 			SET_CONSOLE_SCLP;
205 #endif
206 		} else if (strncmp(ptr + 8, "3215", 4) == 0) {
207 #if defined(CONFIG_TN3215_CONSOLE)
208 			SET_CONSOLE_3215;
209 #elif defined(CONFIG_TN3270_CONSOLE)
210 			SET_CONSOLE_3270;
211 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
212 			SET_CONSOLE_SCLP;
213 #endif
214 		}
215 	} else if (MACHINE_IS_KVM) {
216 		if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
217 			SET_CONSOLE_VT220;
218 		else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
219 			SET_CONSOLE_SCLP;
220 		else
221 			SET_CONSOLE_HVC;
222 	} else {
223 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
224 		SET_CONSOLE_SCLP;
225 #endif
226 	}
227 	if (IS_ENABLED(CONFIG_VT) && IS_ENABLED(CONFIG_DUMMY_CONSOLE))
228 		conswitchp = &dummy_con;
229 }
230 
231 #ifdef CONFIG_CRASH_DUMP
232 static void __init setup_zfcpdump(void)
233 {
234 	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
235 		return;
236 	if (OLDMEM_BASE)
237 		return;
238 	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
239 	console_loglevel = 2;
240 }
241 #else
242 static inline void setup_zfcpdump(void) {}
243 #endif /* CONFIG_CRASH_DUMP */
244 
245  /*
246  * Reboot, halt and power_off stubs. They just call _machine_restart,
247  * _machine_halt or _machine_power_off.
248  */
249 
250 void machine_restart(char *command)
251 {
252 	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
253 		/*
254 		 * Only unblank the console if we are called in enabled
255 		 * context or a bust_spinlocks cleared the way for us.
256 		 */
257 		console_unblank();
258 	_machine_restart(command);
259 }
260 
261 void machine_halt(void)
262 {
263 	if (!in_interrupt() || oops_in_progress)
264 		/*
265 		 * Only unblank the console if we are called in enabled
266 		 * context or a bust_spinlocks cleared the way for us.
267 		 */
268 		console_unblank();
269 	_machine_halt();
270 }
271 
272 void machine_power_off(void)
273 {
274 	if (!in_interrupt() || oops_in_progress)
275 		/*
276 		 * Only unblank the console if we are called in enabled
277 		 * context or a bust_spinlocks cleared the way for us.
278 		 */
279 		console_unblank();
280 	_machine_power_off();
281 }
282 
283 /*
284  * Dummy power off function.
285  */
286 void (*pm_power_off)(void) = machine_power_off;
287 EXPORT_SYMBOL_GPL(pm_power_off);
288 
289 static int __init parse_vmalloc(char *arg)
290 {
291 	if (!arg)
292 		return -EINVAL;
293 	VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
294 	return 0;
295 }
296 early_param("vmalloc", parse_vmalloc);
297 
298 void *restart_stack __section(.data);
299 
300 unsigned long stack_alloc(void)
301 {
302 #ifdef CONFIG_VMAP_STACK
303 	return (unsigned long)
304 		__vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
305 				     VMALLOC_START, VMALLOC_END,
306 				     THREADINFO_GFP,
307 				     PAGE_KERNEL, 0, NUMA_NO_NODE,
308 				     __builtin_return_address(0));
309 #else
310 	return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
311 #endif
312 }
313 
314 void stack_free(unsigned long stack)
315 {
316 #ifdef CONFIG_VMAP_STACK
317 	vfree((void *) stack);
318 #else
319 	free_pages(stack, THREAD_SIZE_ORDER);
320 #endif
321 }
322 
323 int __init arch_early_irq_init(void)
324 {
325 	unsigned long stack;
326 
327 	stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
328 	if (!stack)
329 		panic("Couldn't allocate async stack");
330 	S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
331 	return 0;
332 }
333 
334 static int __init async_stack_realloc(void)
335 {
336 	unsigned long old, new;
337 
338 	old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
339 	new = stack_alloc();
340 	if (!new)
341 		panic("Couldn't allocate async stack");
342 	S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
343 	free_pages(old, THREAD_SIZE_ORDER);
344 	return 0;
345 }
346 early_initcall(async_stack_realloc);
347 
348 void __init arch_call_rest_init(void)
349 {
350 	struct stack_frame *frame;
351 	unsigned long stack;
352 
353 	stack = stack_alloc();
354 	if (!stack)
355 		panic("Couldn't allocate kernel stack");
356 	current->stack = (void *) stack;
357 #ifdef CONFIG_VMAP_STACK
358 	current->stack_vm_area = (void *) stack;
359 #endif
360 	set_task_stack_end_magic(current);
361 	stack += STACK_INIT_OFFSET;
362 	S390_lowcore.kernel_stack = stack;
363 	frame = (struct stack_frame *) stack;
364 	memset(frame, 0, sizeof(*frame));
365 	/* Branch to rest_init on the new stack, never returns */
366 	asm volatile(
367 		"	la	15,0(%[_frame])\n"
368 		"	jg	rest_init\n"
369 		: : [_frame] "a" (frame));
370 }
371 
372 static void __init setup_lowcore_dat_off(void)
373 {
374 	struct lowcore *lc;
375 
376 	/*
377 	 * Setup lowcore for boot cpu
378 	 */
379 	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
380 	lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
381 	lc->restart_psw.mask = PSW_KERNEL_BITS;
382 	lc->restart_psw.addr = (unsigned long) restart_int_handler;
383 	lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
384 	lc->external_new_psw.addr = (unsigned long) ext_int_handler;
385 	lc->svc_new_psw.mask = PSW_KERNEL_BITS |
386 		PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
387 	lc->svc_new_psw.addr = (unsigned long) system_call;
388 	lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
389 	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
390 	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
391 	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
392 	lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
393 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
394 	lc->clock_comparator = clock_comparator_max;
395 	lc->nodat_stack = ((unsigned long) &init_thread_union)
396 		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
397 	lc->current_task = (unsigned long)&init_task;
398 	lc->lpp = LPP_MAGIC;
399 	lc->machine_flags = S390_lowcore.machine_flags;
400 	lc->preempt_count = S390_lowcore.preempt_count;
401 	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
402 	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
403 	       sizeof(lc->stfle_fac_list));
404 	memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
405 	       sizeof(lc->alt_stfle_fac_list));
406 	nmi_alloc_boot_cpu(lc);
407 	vdso_alloc_boot_cpu(lc);
408 	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
409 	lc->async_enter_timer = S390_lowcore.async_enter_timer;
410 	lc->exit_timer = S390_lowcore.exit_timer;
411 	lc->user_timer = S390_lowcore.user_timer;
412 	lc->system_timer = S390_lowcore.system_timer;
413 	lc->steal_timer = S390_lowcore.steal_timer;
414 	lc->last_update_timer = S390_lowcore.last_update_timer;
415 	lc->last_update_clock = S390_lowcore.last_update_clock;
416 
417 	/*
418 	 * Allocate the global restart stack which is the same for
419 	 * all CPUs in cast *one* of them does a PSW restart.
420 	 */
421 	restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
422 	restart_stack += STACK_INIT_OFFSET;
423 
424 	/*
425 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
426 	 * restart data to the absolute zero lowcore. This is necessary if
427 	 * PSW restart is done on an offline CPU that has lowcore zero.
428 	 */
429 	lc->restart_stack = (unsigned long) restart_stack;
430 	lc->restart_fn = (unsigned long) do_restart;
431 	lc->restart_data = 0;
432 	lc->restart_source = -1UL;
433 
434 	/* Setup absolute zero lowcore */
435 	mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
436 	mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
437 	mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
438 	mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
439 	mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
440 
441 #ifdef CONFIG_SMP
442 	lc->spinlock_lockval = arch_spin_lockval(0);
443 	lc->spinlock_index = 0;
444 	arch_spin_lock_setup(0);
445 #endif
446 	lc->br_r1_trampoline = 0x07f1;	/* br %r1 */
447 
448 	set_prefix((u32)(unsigned long) lc);
449 	lowcore_ptr[0] = lc;
450 }
451 
452 static void __init setup_lowcore_dat_on(void)
453 {
454 	__ctl_clear_bit(0, 28);
455 	S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
456 	S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
457 	S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
458 	S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
459 	__ctl_set_bit(0, 28);
460 }
461 
462 static struct resource code_resource = {
463 	.name  = "Kernel code",
464 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
465 };
466 
467 static struct resource data_resource = {
468 	.name = "Kernel data",
469 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
470 };
471 
472 static struct resource bss_resource = {
473 	.name = "Kernel bss",
474 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
475 };
476 
477 static struct resource __initdata *standard_resources[] = {
478 	&code_resource,
479 	&data_resource,
480 	&bss_resource,
481 };
482 
483 static void __init setup_resources(void)
484 {
485 	struct resource *res, *std_res, *sub_res;
486 	struct memblock_region *reg;
487 	int j;
488 
489 	code_resource.start = (unsigned long) _text;
490 	code_resource.end = (unsigned long) _etext - 1;
491 	data_resource.start = (unsigned long) _etext;
492 	data_resource.end = (unsigned long) _edata - 1;
493 	bss_resource.start = (unsigned long) __bss_start;
494 	bss_resource.end = (unsigned long) __bss_stop - 1;
495 
496 	for_each_memblock(memory, reg) {
497 		res = memblock_alloc(sizeof(*res), 8);
498 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
499 
500 		res->name = "System RAM";
501 		res->start = reg->base;
502 		res->end = reg->base + reg->size - 1;
503 		request_resource(&iomem_resource, res);
504 
505 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
506 			std_res = standard_resources[j];
507 			if (std_res->start < res->start ||
508 			    std_res->start > res->end)
509 				continue;
510 			if (std_res->end > res->end) {
511 				sub_res = memblock_alloc(sizeof(*sub_res), 8);
512 				*sub_res = *std_res;
513 				sub_res->end = res->end;
514 				std_res->start = res->end + 1;
515 				request_resource(res, sub_res);
516 			} else {
517 				request_resource(res, std_res);
518 			}
519 		}
520 	}
521 #ifdef CONFIG_CRASH_DUMP
522 	/*
523 	 * Re-add removed crash kernel memory as reserved memory. This makes
524 	 * sure it will be mapped with the identity mapping and struct pages
525 	 * will be created, so it can be resized later on.
526 	 * However add it later since the crash kernel resource should not be
527 	 * part of the System RAM resource.
528 	 */
529 	if (crashk_res.end) {
530 		memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
531 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
532 		insert_resource(&iomem_resource, &crashk_res);
533 	}
534 #endif
535 }
536 
537 static void __init setup_memory_end(void)
538 {
539 	unsigned long vmax, vmalloc_size, tmp;
540 
541 	/* Choose kernel address space layout: 3 or 4 levels. */
542 	vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
543 	if (IS_ENABLED(CONFIG_KASAN)) {
544 		vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
545 			   ? _REGION1_SIZE
546 			   : _REGION2_SIZE;
547 	} else {
548 		tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
549 		tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
550 		if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
551 			vmax = _REGION2_SIZE; /* 3-level kernel page table */
552 		else
553 			vmax = _REGION1_SIZE; /* 4-level kernel page table */
554 	}
555 
556 	/* module area is at the end of the kernel address space. */
557 	MODULES_END = vmax;
558 	MODULES_VADDR = MODULES_END - MODULES_LEN;
559 	VMALLOC_END = MODULES_VADDR;
560 	VMALLOC_START = VMALLOC_END - vmalloc_size;
561 
562 	/* Split remaining virtual space between 1:1 mapping & vmemmap array */
563 	tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
564 	/* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
565 	tmp = SECTION_ALIGN_UP(tmp);
566 	tmp = VMALLOC_START - tmp * sizeof(struct page);
567 	tmp &= ~((vmax >> 11) - 1);	/* align to page table level */
568 	tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
569 	vmemmap = (struct page *) tmp;
570 
571 	/* Take care that memory_end is set and <= vmemmap */
572 	memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
573 #ifdef CONFIG_KASAN
574 	/* fit in kasan shadow memory region between 1:1 and vmemmap */
575 	memory_end = min(memory_end, KASAN_SHADOW_START);
576 	vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
577 #endif
578 	max_pfn = max_low_pfn = PFN_DOWN(memory_end);
579 	memblock_remove(memory_end, ULONG_MAX);
580 
581 	pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
582 }
583 
584 #ifdef CONFIG_CRASH_DUMP
585 
586 /*
587  * When kdump is enabled, we have to ensure that no memory from
588  * the area [0 - crashkernel memory size] and
589  * [crashk_res.start - crashk_res.end] is set offline.
590  */
591 static int kdump_mem_notifier(struct notifier_block *nb,
592 			      unsigned long action, void *data)
593 {
594 	struct memory_notify *arg = data;
595 
596 	if (action != MEM_GOING_OFFLINE)
597 		return NOTIFY_OK;
598 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
599 		return NOTIFY_BAD;
600 	if (arg->start_pfn > PFN_DOWN(crashk_res.end))
601 		return NOTIFY_OK;
602 	if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
603 		return NOTIFY_OK;
604 	return NOTIFY_BAD;
605 }
606 
607 static struct notifier_block kdump_mem_nb = {
608 	.notifier_call = kdump_mem_notifier,
609 };
610 
611 #endif
612 
613 /*
614  * Make sure that the area behind memory_end is protected
615  */
616 static void reserve_memory_end(void)
617 {
618 	if (memory_end_set)
619 		memblock_reserve(memory_end, ULONG_MAX);
620 }
621 
622 /*
623  * Make sure that oldmem, where the dump is stored, is protected
624  */
625 static void reserve_oldmem(void)
626 {
627 #ifdef CONFIG_CRASH_DUMP
628 	if (OLDMEM_BASE)
629 		/* Forget all memory above the running kdump system */
630 		memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
631 #endif
632 }
633 
634 /*
635  * Make sure that oldmem, where the dump is stored, is protected
636  */
637 static void remove_oldmem(void)
638 {
639 #ifdef CONFIG_CRASH_DUMP
640 	if (OLDMEM_BASE)
641 		/* Forget all memory above the running kdump system */
642 		memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
643 #endif
644 }
645 
646 /*
647  * Reserve memory for kdump kernel to be loaded with kexec
648  */
649 static void __init reserve_crashkernel(void)
650 {
651 #ifdef CONFIG_CRASH_DUMP
652 	unsigned long long crash_base, crash_size;
653 	phys_addr_t low, high;
654 	int rc;
655 
656 	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
657 			       &crash_base);
658 
659 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
660 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
661 	if (rc || crash_size == 0)
662 		return;
663 
664 	if (memblock.memory.regions[0].size < crash_size) {
665 		pr_info("crashkernel reservation failed: %s\n",
666 			"first memory chunk must be at least crashkernel size");
667 		return;
668 	}
669 
670 	low = crash_base ?: OLDMEM_BASE;
671 	high = low + crash_size;
672 	if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
673 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
674 		crash_base = low;
675 	} else {
676 		/* Find suitable area in free memory */
677 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
678 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
679 
680 		if (crash_base && crash_base < low) {
681 			pr_info("crashkernel reservation failed: %s\n",
682 				"crash_base too low");
683 			return;
684 		}
685 		low = crash_base ?: low;
686 		crash_base = memblock_find_in_range(low, high, crash_size,
687 						    KEXEC_CRASH_MEM_ALIGN);
688 	}
689 
690 	if (!crash_base) {
691 		pr_info("crashkernel reservation failed: %s\n",
692 			"no suitable area found");
693 		return;
694 	}
695 
696 	if (register_memory_notifier(&kdump_mem_nb))
697 		return;
698 
699 	if (!OLDMEM_BASE && MACHINE_IS_VM)
700 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
701 	crashk_res.start = crash_base;
702 	crashk_res.end = crash_base + crash_size - 1;
703 	memblock_remove(crash_base, crash_size);
704 	pr_info("Reserving %lluMB of memory at %lluMB "
705 		"for crashkernel (System RAM: %luMB)\n",
706 		crash_size >> 20, crash_base >> 20,
707 		(unsigned long)memblock.memory.total_size >> 20);
708 	os_info_crashkernel_add(crash_base, crash_size);
709 #endif
710 }
711 
712 /*
713  * Reserve the initrd from being used by memblock
714  */
715 static void __init reserve_initrd(void)
716 {
717 #ifdef CONFIG_BLK_DEV_INITRD
718 	if (!INITRD_START || !INITRD_SIZE)
719 		return;
720 	initrd_start = INITRD_START;
721 	initrd_end = initrd_start + INITRD_SIZE;
722 	memblock_reserve(INITRD_START, INITRD_SIZE);
723 #endif
724 }
725 
726 static void __init reserve_mem_detect_info(void)
727 {
728 	unsigned long start, size;
729 
730 	get_mem_detect_reserved(&start, &size);
731 	if (size)
732 		memblock_reserve(start, size);
733 }
734 
735 static void __init free_mem_detect_info(void)
736 {
737 	unsigned long start, size;
738 
739 	get_mem_detect_reserved(&start, &size);
740 	if (size)
741 		memblock_free(start, size);
742 }
743 
744 static void __init memblock_physmem_add(phys_addr_t start, phys_addr_t size)
745 {
746 	memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n",
747 		     start, start + size - 1);
748 	memblock_add_range(&memblock.memory, start, size, 0, 0);
749 	memblock_add_range(&memblock.physmem, start, size, 0, 0);
750 }
751 
752 static const char * __init get_mem_info_source(void)
753 {
754 	switch (mem_detect.info_source) {
755 	case MEM_DETECT_SCLP_STOR_INFO:
756 		return "sclp storage info";
757 	case MEM_DETECT_DIAG260:
758 		return "diag260";
759 	case MEM_DETECT_SCLP_READ_INFO:
760 		return "sclp read info";
761 	case MEM_DETECT_BIN_SEARCH:
762 		return "binary search";
763 	}
764 	return "none";
765 }
766 
767 static void __init memblock_add_mem_detect_info(void)
768 {
769 	unsigned long start, end;
770 	int i;
771 
772 	memblock_dbg("physmem info source: %s (%hhd)\n",
773 		     get_mem_info_source(), mem_detect.info_source);
774 	/* keep memblock lists close to the kernel */
775 	memblock_set_bottom_up(true);
776 	for_each_mem_detect_block(i, &start, &end)
777 		memblock_physmem_add(start, end - start);
778 	memblock_set_bottom_up(false);
779 	memblock_dump_all();
780 }
781 
782 /*
783  * Check for initrd being in usable memory
784  */
785 static void __init check_initrd(void)
786 {
787 #ifdef CONFIG_BLK_DEV_INITRD
788 	if (INITRD_START && INITRD_SIZE &&
789 	    !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
790 		pr_err("The initial RAM disk does not fit into the memory\n");
791 		memblock_free(INITRD_START, INITRD_SIZE);
792 		initrd_start = initrd_end = 0;
793 	}
794 #endif
795 }
796 
797 /*
798  * Reserve memory used for lowcore/command line/kernel image.
799  */
800 static void __init reserve_kernel(void)
801 {
802 	unsigned long start_pfn = PFN_UP(__pa(_end));
803 
804 	memblock_reserve(0, PARMAREA_END);
805 	memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
806 			 - (unsigned long)_stext);
807 }
808 
809 static void __init setup_memory(void)
810 {
811 	struct memblock_region *reg;
812 
813 	/*
814 	 * Init storage key for present memory
815 	 */
816 	for_each_memblock(memory, reg) {
817 		storage_key_init_range(reg->base, reg->base + reg->size);
818 	}
819 	psw_set_key(PAGE_DEFAULT_KEY);
820 
821 	/* Only cosmetics */
822 	memblock_enforce_memory_limit(memblock_end_of_DRAM());
823 }
824 
825 /*
826  * Setup hardware capabilities.
827  */
828 static int __init setup_hwcaps(void)
829 {
830 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
831 	struct cpuid cpu_id;
832 	int i;
833 
834 	/*
835 	 * The store facility list bits numbers as found in the principles
836 	 * of operation are numbered with bit 1UL<<31 as number 0 to
837 	 * bit 1UL<<0 as number 31.
838 	 *   Bit 0: instructions named N3, "backported" to esa-mode
839 	 *   Bit 2: z/Architecture mode is active
840 	 *   Bit 7: the store-facility-list-extended facility is installed
841 	 *   Bit 17: the message-security assist is installed
842 	 *   Bit 19: the long-displacement facility is installed
843 	 *   Bit 21: the extended-immediate facility is installed
844 	 *   Bit 22: extended-translation facility 3 is installed
845 	 *   Bit 30: extended-translation facility 3 enhancement facility
846 	 * These get translated to:
847 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
848 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
849 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
850 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
851 	 */
852 	for (i = 0; i < 6; i++)
853 		if (test_facility(stfl_bits[i]))
854 			elf_hwcap |= 1UL << i;
855 
856 	if (test_facility(22) && test_facility(30))
857 		elf_hwcap |= HWCAP_S390_ETF3EH;
858 
859 	/*
860 	 * Check for additional facilities with store-facility-list-extended.
861 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
862 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
863 	 * as stored by stfl, bits 32-xxx contain additional facilities.
864 	 * How many facility words are stored depends on the number of
865 	 * doublewords passed to the instruction. The additional facilities
866 	 * are:
867 	 *   Bit 42: decimal floating point facility is installed
868 	 *   Bit 44: perform floating point operation facility is installed
869 	 * translated to:
870 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
871 	 */
872 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
873 		elf_hwcap |= HWCAP_S390_DFP;
874 
875 	/*
876 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
877 	 */
878 	if (MACHINE_HAS_EDAT1)
879 		elf_hwcap |= HWCAP_S390_HPAGE;
880 
881 	/*
882 	 * 64-bit register support for 31-bit processes
883 	 * HWCAP_S390_HIGH_GPRS is bit 9.
884 	 */
885 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
886 
887 	/*
888 	 * Transactional execution support HWCAP_S390_TE is bit 10.
889 	 */
890 	if (MACHINE_HAS_TE)
891 		elf_hwcap |= HWCAP_S390_TE;
892 
893 	/*
894 	 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
895 	 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
896 	 * instead of facility bit 129.
897 	 */
898 	if (MACHINE_HAS_VX) {
899 		elf_hwcap |= HWCAP_S390_VXRS;
900 		if (test_facility(134))
901 			elf_hwcap |= HWCAP_S390_VXRS_EXT;
902 		if (test_facility(135))
903 			elf_hwcap |= HWCAP_S390_VXRS_BCD;
904 	}
905 
906 	/*
907 	 * Guarded storage support HWCAP_S390_GS is bit 12.
908 	 */
909 	if (MACHINE_HAS_GS)
910 		elf_hwcap |= HWCAP_S390_GS;
911 
912 	get_cpu_id(&cpu_id);
913 	add_device_randomness(&cpu_id, sizeof(cpu_id));
914 	switch (cpu_id.machine) {
915 	case 0x2064:
916 	case 0x2066:
917 	default:	/* Use "z900" as default for 64 bit kernels. */
918 		strcpy(elf_platform, "z900");
919 		break;
920 	case 0x2084:
921 	case 0x2086:
922 		strcpy(elf_platform, "z990");
923 		break;
924 	case 0x2094:
925 	case 0x2096:
926 		strcpy(elf_platform, "z9-109");
927 		break;
928 	case 0x2097:
929 	case 0x2098:
930 		strcpy(elf_platform, "z10");
931 		break;
932 	case 0x2817:
933 	case 0x2818:
934 		strcpy(elf_platform, "z196");
935 		break;
936 	case 0x2827:
937 	case 0x2828:
938 		strcpy(elf_platform, "zEC12");
939 		break;
940 	case 0x2964:
941 	case 0x2965:
942 		strcpy(elf_platform, "z13");
943 		break;
944 	case 0x3906:
945 	case 0x3907:
946 		strcpy(elf_platform, "z14");
947 		break;
948 	}
949 
950 	/*
951 	 * Virtualization support HWCAP_INT_SIE is bit 0.
952 	 */
953 	if (sclp.has_sief2)
954 		int_hwcap |= HWCAP_INT_SIE;
955 
956 	return 0;
957 }
958 arch_initcall(setup_hwcaps);
959 
960 /*
961  * Add system information as device randomness
962  */
963 static void __init setup_randomness(void)
964 {
965 	struct sysinfo_3_2_2 *vmms;
966 
967 	vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
968 							    PAGE_SIZE);
969 	if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
970 		add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
971 	memblock_free((unsigned long) vmms, PAGE_SIZE);
972 }
973 
974 /*
975  * Find the correct size for the task_struct. This depends on
976  * the size of the struct fpu at the end of the thread_struct
977  * which is embedded in the task_struct.
978  */
979 static void __init setup_task_size(void)
980 {
981 	int task_size = sizeof(struct task_struct);
982 
983 	if (!MACHINE_HAS_VX) {
984 		task_size -= sizeof(__vector128) * __NUM_VXRS;
985 		task_size += sizeof(freg_t) * __NUM_FPRS;
986 	}
987 	arch_task_struct_size = task_size;
988 }
989 
990 /*
991  * Issue diagnose 318 to set the control program name and
992  * version codes.
993  */
994 static void __init setup_control_program_code(void)
995 {
996 	union diag318_info diag318_info = {
997 		.cpnc = CPNC_LINUX,
998 		.cpvc_linux = 0,
999 		.cpvc_distro = {0},
1000 	};
1001 
1002 	if (!sclp.has_diag318)
1003 		return;
1004 
1005 	diag_stat_inc(DIAG_STAT_X318);
1006 	asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1007 }
1008 
1009 /*
1010  * Setup function called from init/main.c just after the banner
1011  * was printed.
1012  */
1013 
1014 void __init setup_arch(char **cmdline_p)
1015 {
1016         /*
1017          * print what head.S has found out about the machine
1018          */
1019 	if (MACHINE_IS_VM)
1020 		pr_info("Linux is running as a z/VM "
1021 			"guest operating system in 64-bit mode\n");
1022 	else if (MACHINE_IS_KVM)
1023 		pr_info("Linux is running under KVM in 64-bit mode\n");
1024 	else if (MACHINE_IS_LPAR)
1025 		pr_info("Linux is running natively in 64-bit mode\n");
1026 	else
1027 		pr_info("Linux is running as a guest in 64-bit mode\n");
1028 
1029 	/* Have one command line that is parsed and saved in /proc/cmdline */
1030 	/* boot_command_line has been already set up in early.c */
1031 	*cmdline_p = boot_command_line;
1032 
1033         ROOT_DEV = Root_RAM0;
1034 
1035 	/* Is init_mm really needed? */
1036 	init_mm.start_code = PAGE_OFFSET;
1037 	init_mm.end_code = (unsigned long) _etext;
1038 	init_mm.end_data = (unsigned long) _edata;
1039 	init_mm.brk = (unsigned long) _end;
1040 
1041 	if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1042 		nospec_auto_detect();
1043 
1044 	parse_early_param();
1045 #ifdef CONFIG_CRASH_DUMP
1046 	/* Deactivate elfcorehdr= kernel parameter */
1047 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
1048 #endif
1049 
1050 	os_info_init();
1051 	setup_ipl();
1052 	setup_task_size();
1053 	setup_control_program_code();
1054 
1055 	/* Do some memory reservations *before* memory is added to memblock */
1056 	reserve_memory_end();
1057 	reserve_oldmem();
1058 	reserve_kernel();
1059 	reserve_initrd();
1060 	reserve_mem_detect_info();
1061 	memblock_allow_resize();
1062 
1063 	/* Get information about *all* installed memory */
1064 	memblock_add_mem_detect_info();
1065 
1066 	free_mem_detect_info();
1067 	remove_oldmem();
1068 
1069 	/*
1070 	 * Make sure all chunks are MAX_ORDER aligned so we don't need the
1071 	 * extra checks that HOLES_IN_ZONE would require.
1072 	 *
1073 	 * Is this still required?
1074 	 */
1075 	memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
1076 
1077 	setup_memory_end();
1078 	setup_memory();
1079 	dma_contiguous_reserve(memory_end);
1080 	vmcp_cma_reserve();
1081 
1082 	check_initrd();
1083 	reserve_crashkernel();
1084 #ifdef CONFIG_CRASH_DUMP
1085 	/*
1086 	 * Be aware that smp_save_dump_cpus() triggers a system reset.
1087 	 * Therefore CPU and device initialization should be done afterwards.
1088 	 */
1089 	smp_save_dump_cpus();
1090 #endif
1091 
1092 	setup_resources();
1093 	setup_lowcore_dat_off();
1094 	smp_fill_possible_mask();
1095 	cpu_detect_mhz_feature();
1096         cpu_init();
1097 	numa_setup();
1098 	smp_detect_cpus();
1099 	topology_init_early();
1100 
1101 	/*
1102 	 * Create kernel page tables and switch to virtual addressing.
1103 	 */
1104         paging_init();
1105 
1106 	/*
1107 	 * After paging_init created the kernel page table, the new PSWs
1108 	 * in lowcore can now run with DAT enabled.
1109 	 */
1110 	setup_lowcore_dat_on();
1111 
1112         /* Setup default console */
1113 	conmode_default();
1114 	set_preferred_console();
1115 
1116 	apply_alternative_instructions();
1117 	if (IS_ENABLED(CONFIG_EXPOLINE))
1118 		nospec_init_branches();
1119 
1120 	/* Setup zfcpdump support */
1121 	setup_zfcpdump();
1122 
1123 	/* Add system specific data to the random pool */
1124 	setup_randomness();
1125 }
1126