xref: /openbmc/linux/arch/s390/kernel/setup.c (revision 0da7e432)
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/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 
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 "entry.h"
73 
74 /*
75  * Machine setup..
76  */
77 unsigned int console_mode = 0;
78 EXPORT_SYMBOL(console_mode);
79 
80 unsigned int console_devno = -1;
81 EXPORT_SYMBOL(console_devno);
82 
83 unsigned int console_irq = -1;
84 EXPORT_SYMBOL(console_irq);
85 
86 unsigned long elf_hwcap __read_mostly = 0;
87 char elf_platform[ELF_PLATFORM_SIZE];
88 
89 unsigned long int_hwcap = 0;
90 
91 int __initdata memory_end_set;
92 unsigned long __initdata memory_end;
93 unsigned long __initdata max_physmem_end;
94 
95 unsigned long VMALLOC_START;
96 EXPORT_SYMBOL(VMALLOC_START);
97 
98 unsigned long VMALLOC_END;
99 EXPORT_SYMBOL(VMALLOC_END);
100 
101 struct page *vmemmap;
102 EXPORT_SYMBOL(vmemmap);
103 
104 unsigned long MODULES_VADDR;
105 unsigned long MODULES_END;
106 
107 /* An array with a pointer to the lowcore of every CPU. */
108 struct lowcore *lowcore_ptr[NR_CPUS];
109 EXPORT_SYMBOL(lowcore_ptr);
110 
111 /*
112  * This is set up by the setup-routine at boot-time
113  * for S390 need to find out, what we have to setup
114  * using address 0x10400 ...
115  */
116 
117 #include <asm/setup.h>
118 
119 /*
120  * condev= and conmode= setup parameter.
121  */
122 
123 static int __init condev_setup(char *str)
124 {
125 	int vdev;
126 
127 	vdev = simple_strtoul(str, &str, 0);
128 	if (vdev >= 0 && vdev < 65536) {
129 		console_devno = vdev;
130 		console_irq = -1;
131 	}
132 	return 1;
133 }
134 
135 __setup("condev=", condev_setup);
136 
137 static void __init set_preferred_console(void)
138 {
139 	if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
140 		add_preferred_console("ttyS", 0, NULL);
141 	else if (CONSOLE_IS_3270)
142 		add_preferred_console("tty3270", 0, NULL);
143 	else if (CONSOLE_IS_VT220)
144 		add_preferred_console("ttyS", 1, NULL);
145 	else if (CONSOLE_IS_HVC)
146 		add_preferred_console("hvc", 0, NULL);
147 }
148 
149 static int __init conmode_setup(char *str)
150 {
151 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
152 	if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
153                 SET_CONSOLE_SCLP;
154 #endif
155 #if defined(CONFIG_TN3215_CONSOLE)
156 	if (strncmp(str, "3215", 5) == 0)
157 		SET_CONSOLE_3215;
158 #endif
159 #if defined(CONFIG_TN3270_CONSOLE)
160 	if (strncmp(str, "3270", 5) == 0)
161 		SET_CONSOLE_3270;
162 #endif
163 	set_preferred_console();
164         return 1;
165 }
166 
167 __setup("conmode=", conmode_setup);
168 
169 static void __init conmode_default(void)
170 {
171 	char query_buffer[1024];
172 	char *ptr;
173 
174         if (MACHINE_IS_VM) {
175 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
176 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
177 		ptr = strstr(query_buffer, "SUBCHANNEL =");
178 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
179 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
180 		ptr = strstr(query_buffer, "CONMODE");
181 		/*
182 		 * Set the conmode to 3215 so that the device recognition
183 		 * will set the cu_type of the console to 3215. If the
184 		 * conmode is 3270 and we don't set it back then both
185 		 * 3215 and the 3270 driver will try to access the console
186 		 * device (3215 as console and 3270 as normal tty).
187 		 */
188 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
189 		if (ptr == NULL) {
190 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
191 			SET_CONSOLE_SCLP;
192 #endif
193 			return;
194 		}
195 		if (strncmp(ptr + 8, "3270", 4) == 0) {
196 #if defined(CONFIG_TN3270_CONSOLE)
197 			SET_CONSOLE_3270;
198 #elif defined(CONFIG_TN3215_CONSOLE)
199 			SET_CONSOLE_3215;
200 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
201 			SET_CONSOLE_SCLP;
202 #endif
203 		} else if (strncmp(ptr + 8, "3215", 4) == 0) {
204 #if defined(CONFIG_TN3215_CONSOLE)
205 			SET_CONSOLE_3215;
206 #elif defined(CONFIG_TN3270_CONSOLE)
207 			SET_CONSOLE_3270;
208 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
209 			SET_CONSOLE_SCLP;
210 #endif
211 		}
212 	} else if (MACHINE_IS_KVM) {
213 		if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
214 			SET_CONSOLE_VT220;
215 		else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
216 			SET_CONSOLE_SCLP;
217 		else
218 			SET_CONSOLE_HVC;
219 	} else {
220 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
221 		SET_CONSOLE_SCLP;
222 #endif
223 	}
224 	if (IS_ENABLED(CONFIG_VT) && IS_ENABLED(CONFIG_DUMMY_CONSOLE))
225 		conswitchp = &dummy_con;
226 }
227 
228 #ifdef CONFIG_CRASH_DUMP
229 static void __init setup_zfcpdump(void)
230 {
231 	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
232 		return;
233 	if (OLDMEM_BASE)
234 		return;
235 	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
236 	console_loglevel = 2;
237 }
238 #else
239 static inline void setup_zfcpdump(void) {}
240 #endif /* CONFIG_CRASH_DUMP */
241 
242  /*
243  * Reboot, halt and power_off stubs. They just call _machine_restart,
244  * _machine_halt or _machine_power_off.
245  */
246 
247 void machine_restart(char *command)
248 {
249 	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
250 		/*
251 		 * Only unblank the console if we are called in enabled
252 		 * context or a bust_spinlocks cleared the way for us.
253 		 */
254 		console_unblank();
255 	_machine_restart(command);
256 }
257 
258 void machine_halt(void)
259 {
260 	if (!in_interrupt() || oops_in_progress)
261 		/*
262 		 * Only unblank the console if we are called in enabled
263 		 * context or a bust_spinlocks cleared the way for us.
264 		 */
265 		console_unblank();
266 	_machine_halt();
267 }
268 
269 void machine_power_off(void)
270 {
271 	if (!in_interrupt() || oops_in_progress)
272 		/*
273 		 * Only unblank the console if we are called in enabled
274 		 * context or a bust_spinlocks cleared the way for us.
275 		 */
276 		console_unblank();
277 	_machine_power_off();
278 }
279 
280 /*
281  * Dummy power off function.
282  */
283 void (*pm_power_off)(void) = machine_power_off;
284 EXPORT_SYMBOL_GPL(pm_power_off);
285 
286 static int __init early_parse_mem(char *p)
287 {
288 	memory_end = memparse(p, &p);
289 	memory_end &= PAGE_MASK;
290 	memory_end_set = 1;
291 	return 0;
292 }
293 early_param("mem", early_parse_mem);
294 
295 static int __init parse_vmalloc(char *arg)
296 {
297 	if (!arg)
298 		return -EINVAL;
299 	VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
300 	return 0;
301 }
302 early_param("vmalloc", parse_vmalloc);
303 
304 void *restart_stack __section(.data);
305 
306 static void __init setup_lowcore(void)
307 {
308 	struct lowcore *lc;
309 
310 	/*
311 	 * Setup lowcore for boot cpu
312 	 */
313 	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
314 	lc = memblock_virt_alloc_low(sizeof(*lc), sizeof(*lc));
315 	lc->restart_psw.mask = PSW_KERNEL_BITS;
316 	lc->restart_psw.addr = (unsigned long) restart_int_handler;
317 	lc->external_new_psw.mask = PSW_KERNEL_BITS |
318 		PSW_MASK_DAT | PSW_MASK_MCHECK;
319 	lc->external_new_psw.addr = (unsigned long) ext_int_handler;
320 	lc->svc_new_psw.mask = PSW_KERNEL_BITS |
321 		PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
322 	lc->svc_new_psw.addr = (unsigned long) system_call;
323 	lc->program_new_psw.mask = PSW_KERNEL_BITS |
324 		PSW_MASK_DAT | PSW_MASK_MCHECK;
325 	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
326 	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
327 	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
328 	lc->io_new_psw.mask = PSW_KERNEL_BITS |
329 		PSW_MASK_DAT | PSW_MASK_MCHECK;
330 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
331 	lc->clock_comparator = clock_comparator_max;
332 	lc->kernel_stack = ((unsigned long) &init_thread_union)
333 		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
334 	lc->async_stack = (unsigned long)
335 		memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE)
336 		+ ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
337 	lc->panic_stack = (unsigned long)
338 		memblock_virt_alloc(PAGE_SIZE, PAGE_SIZE)
339 		+ PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
340 	lc->current_task = (unsigned long)&init_task;
341 	lc->lpp = LPP_MAGIC;
342 	lc->machine_flags = S390_lowcore.machine_flags;
343 	lc->preempt_count = S390_lowcore.preempt_count;
344 	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
345 	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
346 	       sizeof(lc->stfle_fac_list));
347 	memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
348 	       sizeof(lc->alt_stfle_fac_list));
349 	nmi_alloc_boot_cpu(lc);
350 	vdso_alloc_boot_cpu(lc);
351 	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
352 	lc->async_enter_timer = S390_lowcore.async_enter_timer;
353 	lc->exit_timer = S390_lowcore.exit_timer;
354 	lc->user_timer = S390_lowcore.user_timer;
355 	lc->system_timer = S390_lowcore.system_timer;
356 	lc->steal_timer = S390_lowcore.steal_timer;
357 	lc->last_update_timer = S390_lowcore.last_update_timer;
358 	lc->last_update_clock = S390_lowcore.last_update_clock;
359 
360 	restart_stack = memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE);
361 	restart_stack += ASYNC_SIZE;
362 
363 	/*
364 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
365 	 * restart data to the absolute zero lowcore. This is necessary if
366 	 * PSW restart is done on an offline CPU that has lowcore zero.
367 	 */
368 	lc->restart_stack = (unsigned long) restart_stack;
369 	lc->restart_fn = (unsigned long) do_restart;
370 	lc->restart_data = 0;
371 	lc->restart_source = -1UL;
372 
373 	/* Setup absolute zero lowcore */
374 	mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
375 	mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
376 	mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
377 	mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
378 	mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
379 
380 #ifdef CONFIG_SMP
381 	lc->spinlock_lockval = arch_spin_lockval(0);
382 	lc->spinlock_index = 0;
383 	arch_spin_lock_setup(0);
384 #endif
385 	lc->br_r1_trampoline = 0x07f1;	/* br %r1 */
386 
387 	set_prefix((u32)(unsigned long) lc);
388 	lowcore_ptr[0] = lc;
389 }
390 
391 static struct resource code_resource = {
392 	.name  = "Kernel code",
393 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
394 };
395 
396 static struct resource data_resource = {
397 	.name = "Kernel data",
398 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
399 };
400 
401 static struct resource bss_resource = {
402 	.name = "Kernel bss",
403 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
404 };
405 
406 static struct resource __initdata *standard_resources[] = {
407 	&code_resource,
408 	&data_resource,
409 	&bss_resource,
410 };
411 
412 static void __init setup_resources(void)
413 {
414 	struct resource *res, *std_res, *sub_res;
415 	struct memblock_region *reg;
416 	int j;
417 
418 	code_resource.start = (unsigned long) _text;
419 	code_resource.end = (unsigned long) _etext - 1;
420 	data_resource.start = (unsigned long) _etext;
421 	data_resource.end = (unsigned long) _edata - 1;
422 	bss_resource.start = (unsigned long) __bss_start;
423 	bss_resource.end = (unsigned long) __bss_stop - 1;
424 
425 	for_each_memblock(memory, reg) {
426 		res = memblock_virt_alloc(sizeof(*res), 8);
427 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
428 
429 		res->name = "System RAM";
430 		res->start = reg->base;
431 		res->end = reg->base + reg->size - 1;
432 		request_resource(&iomem_resource, res);
433 
434 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
435 			std_res = standard_resources[j];
436 			if (std_res->start < res->start ||
437 			    std_res->start > res->end)
438 				continue;
439 			if (std_res->end > res->end) {
440 				sub_res = memblock_virt_alloc(sizeof(*sub_res), 8);
441 				*sub_res = *std_res;
442 				sub_res->end = res->end;
443 				std_res->start = res->end + 1;
444 				request_resource(res, sub_res);
445 			} else {
446 				request_resource(res, std_res);
447 			}
448 		}
449 	}
450 #ifdef CONFIG_CRASH_DUMP
451 	/*
452 	 * Re-add removed crash kernel memory as reserved memory. This makes
453 	 * sure it will be mapped with the identity mapping and struct pages
454 	 * will be created, so it can be resized later on.
455 	 * However add it later since the crash kernel resource should not be
456 	 * part of the System RAM resource.
457 	 */
458 	if (crashk_res.end) {
459 		memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
460 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
461 		insert_resource(&iomem_resource, &crashk_res);
462 	}
463 #endif
464 }
465 
466 static void __init setup_memory_end(void)
467 {
468 	unsigned long vmax, vmalloc_size, tmp;
469 
470 	/* Choose kernel address space layout: 2, 3, or 4 levels. */
471 	vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
472 	tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
473 	tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
474 	if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
475 		vmax = _REGION2_SIZE; /* 3-level kernel page table */
476 	else
477 		vmax = _REGION1_SIZE; /* 4-level kernel page table */
478 	/* module area is at the end of the kernel address space. */
479 	MODULES_END = vmax;
480 	MODULES_VADDR = MODULES_END - MODULES_LEN;
481 	VMALLOC_END = MODULES_VADDR;
482 	VMALLOC_START = vmax - vmalloc_size;
483 
484 	/* Split remaining virtual space between 1:1 mapping & vmemmap array */
485 	tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
486 	/* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
487 	tmp = SECTION_ALIGN_UP(tmp);
488 	tmp = VMALLOC_START - tmp * sizeof(struct page);
489 	tmp &= ~((vmax >> 11) - 1);	/* align to page table level */
490 	tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
491 	vmemmap = (struct page *) tmp;
492 
493 	/* Take care that memory_end is set and <= vmemmap */
494 	memory_end = min(memory_end ?: max_physmem_end, tmp);
495 	max_pfn = max_low_pfn = PFN_DOWN(memory_end);
496 	memblock_remove(memory_end, ULONG_MAX);
497 
498 	pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
499 }
500 
501 #ifdef CONFIG_CRASH_DUMP
502 
503 /*
504  * When kdump is enabled, we have to ensure that no memory from
505  * the area [0 - crashkernel memory size] and
506  * [crashk_res.start - crashk_res.end] is set offline.
507  */
508 static int kdump_mem_notifier(struct notifier_block *nb,
509 			      unsigned long action, void *data)
510 {
511 	struct memory_notify *arg = data;
512 
513 	if (action != MEM_GOING_OFFLINE)
514 		return NOTIFY_OK;
515 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
516 		return NOTIFY_BAD;
517 	if (arg->start_pfn > PFN_DOWN(crashk_res.end))
518 		return NOTIFY_OK;
519 	if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
520 		return NOTIFY_OK;
521 	return NOTIFY_BAD;
522 }
523 
524 static struct notifier_block kdump_mem_nb = {
525 	.notifier_call = kdump_mem_notifier,
526 };
527 
528 #endif
529 
530 /*
531  * Make sure that the area behind memory_end is protected
532  */
533 static void reserve_memory_end(void)
534 {
535 #ifdef CONFIG_CRASH_DUMP
536 	if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
537 	    !OLDMEM_BASE && sclp.hsa_size) {
538 		memory_end = sclp.hsa_size;
539 		memory_end &= PAGE_MASK;
540 		memory_end_set = 1;
541 	}
542 #endif
543 	if (!memory_end_set)
544 		return;
545 	memblock_reserve(memory_end, ULONG_MAX);
546 }
547 
548 /*
549  * Make sure that oldmem, where the dump is stored, is protected
550  */
551 static void reserve_oldmem(void)
552 {
553 #ifdef CONFIG_CRASH_DUMP
554 	if (OLDMEM_BASE)
555 		/* Forget all memory above the running kdump system */
556 		memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
557 #endif
558 }
559 
560 /*
561  * Make sure that oldmem, where the dump is stored, is protected
562  */
563 static void remove_oldmem(void)
564 {
565 #ifdef CONFIG_CRASH_DUMP
566 	if (OLDMEM_BASE)
567 		/* Forget all memory above the running kdump system */
568 		memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
569 #endif
570 }
571 
572 /*
573  * Reserve memory for kdump kernel to be loaded with kexec
574  */
575 static void __init reserve_crashkernel(void)
576 {
577 #ifdef CONFIG_CRASH_DUMP
578 	unsigned long long crash_base, crash_size;
579 	phys_addr_t low, high;
580 	int rc;
581 
582 	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
583 			       &crash_base);
584 
585 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
586 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
587 	if (rc || crash_size == 0)
588 		return;
589 
590 	if (memblock.memory.regions[0].size < crash_size) {
591 		pr_info("crashkernel reservation failed: %s\n",
592 			"first memory chunk must be at least crashkernel size");
593 		return;
594 	}
595 
596 	low = crash_base ?: OLDMEM_BASE;
597 	high = low + crash_size;
598 	if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
599 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
600 		crash_base = low;
601 	} else {
602 		/* Find suitable area in free memory */
603 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
604 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
605 
606 		if (crash_base && crash_base < low) {
607 			pr_info("crashkernel reservation failed: %s\n",
608 				"crash_base too low");
609 			return;
610 		}
611 		low = crash_base ?: low;
612 		crash_base = memblock_find_in_range(low, high, crash_size,
613 						    KEXEC_CRASH_MEM_ALIGN);
614 	}
615 
616 	if (!crash_base) {
617 		pr_info("crashkernel reservation failed: %s\n",
618 			"no suitable area found");
619 		return;
620 	}
621 
622 	if (register_memory_notifier(&kdump_mem_nb))
623 		return;
624 
625 	if (!OLDMEM_BASE && MACHINE_IS_VM)
626 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
627 	crashk_res.start = crash_base;
628 	crashk_res.end = crash_base + crash_size - 1;
629 	memblock_remove(crash_base, crash_size);
630 	pr_info("Reserving %lluMB of memory at %lluMB "
631 		"for crashkernel (System RAM: %luMB)\n",
632 		crash_size >> 20, crash_base >> 20,
633 		(unsigned long)memblock.memory.total_size >> 20);
634 	os_info_crashkernel_add(crash_base, crash_size);
635 #endif
636 }
637 
638 /*
639  * Reserve the initrd from being used by memblock
640  */
641 static void __init reserve_initrd(void)
642 {
643 #ifdef CONFIG_BLK_DEV_INITRD
644 	if (!INITRD_START || !INITRD_SIZE)
645 		return;
646 	initrd_start = INITRD_START;
647 	initrd_end = initrd_start + INITRD_SIZE;
648 	memblock_reserve(INITRD_START, INITRD_SIZE);
649 #endif
650 }
651 
652 /*
653  * Check for initrd being in usable memory
654  */
655 static void __init check_initrd(void)
656 {
657 #ifdef CONFIG_BLK_DEV_INITRD
658 	if (INITRD_START && INITRD_SIZE &&
659 	    !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
660 		pr_err("The initial RAM disk does not fit into the memory\n");
661 		memblock_free(INITRD_START, INITRD_SIZE);
662 		initrd_start = initrd_end = 0;
663 	}
664 #endif
665 }
666 
667 /*
668  * Reserve memory used for lowcore/command line/kernel image.
669  */
670 static void __init reserve_kernel(void)
671 {
672 	unsigned long start_pfn = PFN_UP(__pa(_end));
673 
674 #ifdef CONFIG_DMA_API_DEBUG
675 	/*
676 	 * DMA_API_DEBUG code stumbles over addresses from the
677 	 * range [_ehead, _stext]. Mark the memory as reserved
678 	 * so it is not used for CONFIG_DMA_API_DEBUG=y.
679 	 */
680 	memblock_reserve(0, PFN_PHYS(start_pfn));
681 #else
682 	memblock_reserve(0, (unsigned long)_ehead);
683 	memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
684 			 - (unsigned long)_stext);
685 #endif
686 }
687 
688 static void __init setup_memory(void)
689 {
690 	struct memblock_region *reg;
691 
692 	/*
693 	 * Init storage key for present memory
694 	 */
695 	for_each_memblock(memory, reg) {
696 		storage_key_init_range(reg->base, reg->base + reg->size);
697 	}
698 	psw_set_key(PAGE_DEFAULT_KEY);
699 
700 	/* Only cosmetics */
701 	memblock_enforce_memory_limit(memblock_end_of_DRAM());
702 }
703 
704 /*
705  * Setup hardware capabilities.
706  */
707 static int __init setup_hwcaps(void)
708 {
709 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
710 	struct cpuid cpu_id;
711 	int i;
712 
713 	/*
714 	 * The store facility list bits numbers as found in the principles
715 	 * of operation are numbered with bit 1UL<<31 as number 0 to
716 	 * bit 1UL<<0 as number 31.
717 	 *   Bit 0: instructions named N3, "backported" to esa-mode
718 	 *   Bit 2: z/Architecture mode is active
719 	 *   Bit 7: the store-facility-list-extended facility is installed
720 	 *   Bit 17: the message-security assist is installed
721 	 *   Bit 19: the long-displacement facility is installed
722 	 *   Bit 21: the extended-immediate facility is installed
723 	 *   Bit 22: extended-translation facility 3 is installed
724 	 *   Bit 30: extended-translation facility 3 enhancement facility
725 	 * These get translated to:
726 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
727 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
728 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
729 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
730 	 */
731 	for (i = 0; i < 6; i++)
732 		if (test_facility(stfl_bits[i]))
733 			elf_hwcap |= 1UL << i;
734 
735 	if (test_facility(22) && test_facility(30))
736 		elf_hwcap |= HWCAP_S390_ETF3EH;
737 
738 	/*
739 	 * Check for additional facilities with store-facility-list-extended.
740 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
741 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
742 	 * as stored by stfl, bits 32-xxx contain additional facilities.
743 	 * How many facility words are stored depends on the number of
744 	 * doublewords passed to the instruction. The additional facilities
745 	 * are:
746 	 *   Bit 42: decimal floating point facility is installed
747 	 *   Bit 44: perform floating point operation facility is installed
748 	 * translated to:
749 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
750 	 */
751 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
752 		elf_hwcap |= HWCAP_S390_DFP;
753 
754 	/*
755 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
756 	 */
757 	if (MACHINE_HAS_EDAT1)
758 		elf_hwcap |= HWCAP_S390_HPAGE;
759 
760 	/*
761 	 * 64-bit register support for 31-bit processes
762 	 * HWCAP_S390_HIGH_GPRS is bit 9.
763 	 */
764 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
765 
766 	/*
767 	 * Transactional execution support HWCAP_S390_TE is bit 10.
768 	 */
769 	if (MACHINE_HAS_TE)
770 		elf_hwcap |= HWCAP_S390_TE;
771 
772 	/*
773 	 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
774 	 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
775 	 * instead of facility bit 129.
776 	 */
777 	if (MACHINE_HAS_VX) {
778 		elf_hwcap |= HWCAP_S390_VXRS;
779 		if (test_facility(134))
780 			elf_hwcap |= HWCAP_S390_VXRS_EXT;
781 		if (test_facility(135))
782 			elf_hwcap |= HWCAP_S390_VXRS_BCD;
783 	}
784 
785 	/*
786 	 * Guarded storage support HWCAP_S390_GS is bit 12.
787 	 */
788 	if (MACHINE_HAS_GS)
789 		elf_hwcap |= HWCAP_S390_GS;
790 
791 	get_cpu_id(&cpu_id);
792 	add_device_randomness(&cpu_id, sizeof(cpu_id));
793 	switch (cpu_id.machine) {
794 	case 0x2064:
795 	case 0x2066:
796 	default:	/* Use "z900" as default for 64 bit kernels. */
797 		strcpy(elf_platform, "z900");
798 		break;
799 	case 0x2084:
800 	case 0x2086:
801 		strcpy(elf_platform, "z990");
802 		break;
803 	case 0x2094:
804 	case 0x2096:
805 		strcpy(elf_platform, "z9-109");
806 		break;
807 	case 0x2097:
808 	case 0x2098:
809 		strcpy(elf_platform, "z10");
810 		break;
811 	case 0x2817:
812 	case 0x2818:
813 		strcpy(elf_platform, "z196");
814 		break;
815 	case 0x2827:
816 	case 0x2828:
817 		strcpy(elf_platform, "zEC12");
818 		break;
819 	case 0x2964:
820 	case 0x2965:
821 		strcpy(elf_platform, "z13");
822 		break;
823 	case 0x3906:
824 	case 0x3907:
825 		strcpy(elf_platform, "z14");
826 		break;
827 	}
828 
829 	/*
830 	 * Virtualization support HWCAP_INT_SIE is bit 0.
831 	 */
832 	if (sclp.has_sief2)
833 		int_hwcap |= HWCAP_INT_SIE;
834 
835 	return 0;
836 }
837 arch_initcall(setup_hwcaps);
838 
839 /*
840  * Add system information as device randomness
841  */
842 static void __init setup_randomness(void)
843 {
844 	struct sysinfo_3_2_2 *vmms;
845 
846 	vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE);
847 	if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
848 		add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
849 	memblock_free((unsigned long) vmms, PAGE_SIZE);
850 }
851 
852 /*
853  * Find the correct size for the task_struct. This depends on
854  * the size of the struct fpu at the end of the thread_struct
855  * which is embedded in the task_struct.
856  */
857 static void __init setup_task_size(void)
858 {
859 	int task_size = sizeof(struct task_struct);
860 
861 	if (!MACHINE_HAS_VX) {
862 		task_size -= sizeof(__vector128) * __NUM_VXRS;
863 		task_size += sizeof(freg_t) * __NUM_FPRS;
864 	}
865 	arch_task_struct_size = task_size;
866 }
867 
868 /*
869  * Setup function called from init/main.c just after the banner
870  * was printed.
871  */
872 
873 void __init setup_arch(char **cmdline_p)
874 {
875         /*
876          * print what head.S has found out about the machine
877          */
878 	if (MACHINE_IS_VM)
879 		pr_info("Linux is running as a z/VM "
880 			"guest operating system in 64-bit mode\n");
881 	else if (MACHINE_IS_KVM)
882 		pr_info("Linux is running under KVM in 64-bit mode\n");
883 	else if (MACHINE_IS_LPAR)
884 		pr_info("Linux is running natively in 64-bit mode\n");
885 
886 	/* Have one command line that is parsed and saved in /proc/cmdline */
887 	/* boot_command_line has been already set up in early.c */
888 	*cmdline_p = boot_command_line;
889 
890         ROOT_DEV = Root_RAM0;
891 
892 	/* Is init_mm really needed? */
893 	init_mm.start_code = PAGE_OFFSET;
894 	init_mm.end_code = (unsigned long) _etext;
895 	init_mm.end_data = (unsigned long) _edata;
896 	init_mm.brk = (unsigned long) _end;
897 
898 	if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
899 		nospec_auto_detect();
900 
901 	parse_early_param();
902 #ifdef CONFIG_CRASH_DUMP
903 	/* Deactivate elfcorehdr= kernel parameter */
904 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
905 #endif
906 
907 	os_info_init();
908 	setup_ipl();
909 	setup_task_size();
910 
911 	/* Do some memory reservations *before* memory is added to memblock */
912 	reserve_memory_end();
913 	reserve_oldmem();
914 	reserve_kernel();
915 	reserve_initrd();
916 	memblock_allow_resize();
917 
918 	/* Get information about *all* installed memory */
919 	detect_memory_memblock();
920 
921 	remove_oldmem();
922 
923 	/*
924 	 * Make sure all chunks are MAX_ORDER aligned so we don't need the
925 	 * extra checks that HOLES_IN_ZONE would require.
926 	 *
927 	 * Is this still required?
928 	 */
929 	memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
930 
931 	setup_memory_end();
932 	setup_memory();
933 	dma_contiguous_reserve(memory_end);
934 	vmcp_cma_reserve();
935 
936 	check_initrd();
937 	reserve_crashkernel();
938 #ifdef CONFIG_CRASH_DUMP
939 	/*
940 	 * Be aware that smp_save_dump_cpus() triggers a system reset.
941 	 * Therefore CPU and device initialization should be done afterwards.
942 	 */
943 	smp_save_dump_cpus();
944 #endif
945 
946 	setup_resources();
947 	setup_lowcore();
948 	smp_fill_possible_mask();
949 	cpu_detect_mhz_feature();
950         cpu_init();
951 	numa_setup();
952 	smp_detect_cpus();
953 	topology_init_early();
954 
955 	/*
956 	 * Create kernel page tables and switch to virtual addressing.
957 	 */
958         paging_init();
959 
960         /* Setup default console */
961 	conmode_default();
962 	set_preferred_console();
963 
964 	apply_alternative_instructions();
965 	if (IS_ENABLED(CONFIG_EXPOLINE))
966 		nospec_init_branches();
967 
968 	/* Setup zfcpdump support */
969 	setup_zfcpdump();
970 
971 	/* Add system specific data to the random pool */
972 	setup_randomness();
973 }
974