xref: /linux/arch/s390/kernel/setup.c (revision fd639726bf15fca8ee1a00dce8e0096d0ad9bd18)
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 "entry.h"
72 
73 /*
74  * Machine setup..
75  */
76 unsigned int console_mode = 0;
77 EXPORT_SYMBOL(console_mode);
78 
79 unsigned int console_devno = -1;
80 EXPORT_SYMBOL(console_devno);
81 
82 unsigned int console_irq = -1;
83 EXPORT_SYMBOL(console_irq);
84 
85 unsigned long elf_hwcap __read_mostly = 0;
86 char elf_platform[ELF_PLATFORM_SIZE];
87 
88 unsigned long int_hwcap = 0;
89 
90 int __initdata memory_end_set;
91 unsigned long __initdata memory_end;
92 unsigned long __initdata max_physmem_end;
93 
94 unsigned long VMALLOC_START;
95 EXPORT_SYMBOL(VMALLOC_START);
96 
97 unsigned long VMALLOC_END;
98 EXPORT_SYMBOL(VMALLOC_END);
99 
100 struct page *vmemmap;
101 EXPORT_SYMBOL(vmemmap);
102 
103 unsigned long MODULES_VADDR;
104 unsigned long MODULES_END;
105 
106 /* An array with a pointer to the lowcore of every CPU. */
107 struct lowcore *lowcore_ptr[NR_CPUS];
108 EXPORT_SYMBOL(lowcore_ptr);
109 
110 /*
111  * This is set up by the setup-routine at boot-time
112  * for S390 need to find out, what we have to setup
113  * using address 0x10400 ...
114  */
115 
116 #include <asm/setup.h>
117 
118 /*
119  * condev= and conmode= setup parameter.
120  */
121 
122 static int __init condev_setup(char *str)
123 {
124 	int vdev;
125 
126 	vdev = simple_strtoul(str, &str, 0);
127 	if (vdev >= 0 && vdev < 65536) {
128 		console_devno = vdev;
129 		console_irq = -1;
130 	}
131 	return 1;
132 }
133 
134 __setup("condev=", condev_setup);
135 
136 static void __init set_preferred_console(void)
137 {
138 	if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
139 		add_preferred_console("ttyS", 0, NULL);
140 	else if (CONSOLE_IS_3270)
141 		add_preferred_console("tty3270", 0, NULL);
142 	else if (CONSOLE_IS_VT220)
143 		add_preferred_console("ttyS", 1, NULL);
144 	else if (CONSOLE_IS_HVC)
145 		add_preferred_console("hvc", 0, NULL);
146 }
147 
148 static int __init conmode_setup(char *str)
149 {
150 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
151 	if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
152                 SET_CONSOLE_SCLP;
153 #endif
154 #if defined(CONFIG_TN3215_CONSOLE)
155 	if (strncmp(str, "3215", 5) == 0)
156 		SET_CONSOLE_3215;
157 #endif
158 #if defined(CONFIG_TN3270_CONSOLE)
159 	if (strncmp(str, "3270", 5) == 0)
160 		SET_CONSOLE_3270;
161 #endif
162 	set_preferred_console();
163         return 1;
164 }
165 
166 __setup("conmode=", conmode_setup);
167 
168 static void __init conmode_default(void)
169 {
170 	char query_buffer[1024];
171 	char *ptr;
172 
173         if (MACHINE_IS_VM) {
174 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
175 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
176 		ptr = strstr(query_buffer, "SUBCHANNEL =");
177 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
178 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
179 		ptr = strstr(query_buffer, "CONMODE");
180 		/*
181 		 * Set the conmode to 3215 so that the device recognition
182 		 * will set the cu_type of the console to 3215. If the
183 		 * conmode is 3270 and we don't set it back then both
184 		 * 3215 and the 3270 driver will try to access the console
185 		 * device (3215 as console and 3270 as normal tty).
186 		 */
187 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
188 		if (ptr == NULL) {
189 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
190 			SET_CONSOLE_SCLP;
191 #endif
192 			return;
193 		}
194 		if (strncmp(ptr + 8, "3270", 4) == 0) {
195 #if defined(CONFIG_TN3270_CONSOLE)
196 			SET_CONSOLE_3270;
197 #elif defined(CONFIG_TN3215_CONSOLE)
198 			SET_CONSOLE_3215;
199 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
200 			SET_CONSOLE_SCLP;
201 #endif
202 		} else if (strncmp(ptr + 8, "3215", 4) == 0) {
203 #if defined(CONFIG_TN3215_CONSOLE)
204 			SET_CONSOLE_3215;
205 #elif defined(CONFIG_TN3270_CONSOLE)
206 			SET_CONSOLE_3270;
207 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
208 			SET_CONSOLE_SCLP;
209 #endif
210 		}
211 	} else if (MACHINE_IS_KVM) {
212 		if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
213 			SET_CONSOLE_VT220;
214 		else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
215 			SET_CONSOLE_SCLP;
216 		else
217 			SET_CONSOLE_HVC;
218 	} else {
219 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
220 		SET_CONSOLE_SCLP;
221 #endif
222 	}
223 }
224 
225 #ifdef CONFIG_CRASH_DUMP
226 static void __init setup_zfcpdump(void)
227 {
228 	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
229 		return;
230 	if (OLDMEM_BASE)
231 		return;
232 	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
233 	console_loglevel = 2;
234 }
235 #else
236 static inline void setup_zfcpdump(void) {}
237 #endif /* CONFIG_CRASH_DUMP */
238 
239  /*
240  * Reboot, halt and power_off stubs. They just call _machine_restart,
241  * _machine_halt or _machine_power_off.
242  */
243 
244 void machine_restart(char *command)
245 {
246 	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
247 		/*
248 		 * Only unblank the console if we are called in enabled
249 		 * context or a bust_spinlocks cleared the way for us.
250 		 */
251 		console_unblank();
252 	_machine_restart(command);
253 }
254 
255 void machine_halt(void)
256 {
257 	if (!in_interrupt() || oops_in_progress)
258 		/*
259 		 * Only unblank the console if we are called in enabled
260 		 * context or a bust_spinlocks cleared the way for us.
261 		 */
262 		console_unblank();
263 	_machine_halt();
264 }
265 
266 void machine_power_off(void)
267 {
268 	if (!in_interrupt() || oops_in_progress)
269 		/*
270 		 * Only unblank the console if we are called in enabled
271 		 * context or a bust_spinlocks cleared the way for us.
272 		 */
273 		console_unblank();
274 	_machine_power_off();
275 }
276 
277 /*
278  * Dummy power off function.
279  */
280 void (*pm_power_off)(void) = machine_power_off;
281 EXPORT_SYMBOL_GPL(pm_power_off);
282 
283 static int __init early_parse_mem(char *p)
284 {
285 	memory_end = memparse(p, &p);
286 	memory_end &= PAGE_MASK;
287 	memory_end_set = 1;
288 	return 0;
289 }
290 early_param("mem", early_parse_mem);
291 
292 static int __init parse_vmalloc(char *arg)
293 {
294 	if (!arg)
295 		return -EINVAL;
296 	VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
297 	return 0;
298 }
299 early_param("vmalloc", parse_vmalloc);
300 
301 void *restart_stack __section(.data);
302 
303 static void __init setup_lowcore(void)
304 {
305 	struct lowcore *lc;
306 
307 	/*
308 	 * Setup lowcore for boot cpu
309 	 */
310 	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
311 	lc = memblock_virt_alloc_low(sizeof(*lc), sizeof(*lc));
312 	lc->restart_psw.mask = PSW_KERNEL_BITS;
313 	lc->restart_psw.addr = (unsigned long) restart_int_handler;
314 	lc->external_new_psw.mask = PSW_KERNEL_BITS |
315 		PSW_MASK_DAT | PSW_MASK_MCHECK;
316 	lc->external_new_psw.addr = (unsigned long) ext_int_handler;
317 	lc->svc_new_psw.mask = PSW_KERNEL_BITS |
318 		PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
319 	lc->svc_new_psw.addr = (unsigned long) system_call;
320 	lc->program_new_psw.mask = PSW_KERNEL_BITS |
321 		PSW_MASK_DAT | PSW_MASK_MCHECK;
322 	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
323 	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
324 	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
325 	lc->io_new_psw.mask = PSW_KERNEL_BITS |
326 		PSW_MASK_DAT | PSW_MASK_MCHECK;
327 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
328 	lc->clock_comparator = clock_comparator_max;
329 	lc->kernel_stack = ((unsigned long) &init_thread_union)
330 		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
331 	lc->async_stack = (unsigned long)
332 		memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE)
333 		+ ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
334 	lc->panic_stack = (unsigned long)
335 		memblock_virt_alloc(PAGE_SIZE, PAGE_SIZE)
336 		+ PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
337 	lc->current_task = (unsigned long)&init_task;
338 	lc->lpp = LPP_MAGIC;
339 	lc->machine_flags = S390_lowcore.machine_flags;
340 	lc->preempt_count = S390_lowcore.preempt_count;
341 	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
342 	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
343 	       MAX_FACILITY_BIT/8);
344 	nmi_alloc_boot_cpu(lc);
345 	vdso_alloc_boot_cpu(lc);
346 	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
347 	lc->async_enter_timer = S390_lowcore.async_enter_timer;
348 	lc->exit_timer = S390_lowcore.exit_timer;
349 	lc->user_timer = S390_lowcore.user_timer;
350 	lc->system_timer = S390_lowcore.system_timer;
351 	lc->steal_timer = S390_lowcore.steal_timer;
352 	lc->last_update_timer = S390_lowcore.last_update_timer;
353 	lc->last_update_clock = S390_lowcore.last_update_clock;
354 
355 	restart_stack = memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE);
356 	restart_stack += ASYNC_SIZE;
357 
358 	/*
359 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
360 	 * restart data to the absolute zero lowcore. This is necessary if
361 	 * PSW restart is done on an offline CPU that has lowcore zero.
362 	 */
363 	lc->restart_stack = (unsigned long) restart_stack;
364 	lc->restart_fn = (unsigned long) do_restart;
365 	lc->restart_data = 0;
366 	lc->restart_source = -1UL;
367 
368 	/* Setup absolute zero lowcore */
369 	mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
370 	mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
371 	mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
372 	mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
373 	mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
374 
375 #ifdef CONFIG_SMP
376 	lc->spinlock_lockval = arch_spin_lockval(0);
377 	lc->spinlock_index = 0;
378 	arch_spin_lock_setup(0);
379 #endif
380 
381 	set_prefix((u32)(unsigned long) lc);
382 	lowcore_ptr[0] = lc;
383 }
384 
385 static struct resource code_resource = {
386 	.name  = "Kernel code",
387 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
388 };
389 
390 static struct resource data_resource = {
391 	.name = "Kernel data",
392 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
393 };
394 
395 static struct resource bss_resource = {
396 	.name = "Kernel bss",
397 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
398 };
399 
400 static struct resource __initdata *standard_resources[] = {
401 	&code_resource,
402 	&data_resource,
403 	&bss_resource,
404 };
405 
406 static void __init setup_resources(void)
407 {
408 	struct resource *res, *std_res, *sub_res;
409 	struct memblock_region *reg;
410 	int j;
411 
412 	code_resource.start = (unsigned long) &_text;
413 	code_resource.end = (unsigned long) &_etext - 1;
414 	data_resource.start = (unsigned long) &_etext;
415 	data_resource.end = (unsigned long) &_edata - 1;
416 	bss_resource.start = (unsigned long) &__bss_start;
417 	bss_resource.end = (unsigned long) &__bss_stop - 1;
418 
419 	for_each_memblock(memory, reg) {
420 		res = memblock_virt_alloc(sizeof(*res), 8);
421 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
422 
423 		res->name = "System RAM";
424 		res->start = reg->base;
425 		res->end = reg->base + reg->size - 1;
426 		request_resource(&iomem_resource, res);
427 
428 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
429 			std_res = standard_resources[j];
430 			if (std_res->start < res->start ||
431 			    std_res->start > res->end)
432 				continue;
433 			if (std_res->end > res->end) {
434 				sub_res = memblock_virt_alloc(sizeof(*sub_res), 8);
435 				*sub_res = *std_res;
436 				sub_res->end = res->end;
437 				std_res->start = res->end + 1;
438 				request_resource(res, sub_res);
439 			} else {
440 				request_resource(res, std_res);
441 			}
442 		}
443 	}
444 #ifdef CONFIG_CRASH_DUMP
445 	/*
446 	 * Re-add removed crash kernel memory as reserved memory. This makes
447 	 * sure it will be mapped with the identity mapping and struct pages
448 	 * will be created, so it can be resized later on.
449 	 * However add it later since the crash kernel resource should not be
450 	 * part of the System RAM resource.
451 	 */
452 	if (crashk_res.end) {
453 		memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
454 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
455 		insert_resource(&iomem_resource, &crashk_res);
456 	}
457 #endif
458 }
459 
460 static void __init setup_memory_end(void)
461 {
462 	unsigned long vmax, vmalloc_size, tmp;
463 
464 	/* Choose kernel address space layout: 2, 3, or 4 levels. */
465 	vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
466 	tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
467 	tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
468 	if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
469 		vmax = _REGION2_SIZE; /* 3-level kernel page table */
470 	else
471 		vmax = _REGION1_SIZE; /* 4-level kernel page table */
472 	/* module area is at the end of the kernel address space. */
473 	MODULES_END = vmax;
474 	MODULES_VADDR = MODULES_END - MODULES_LEN;
475 	VMALLOC_END = MODULES_VADDR;
476 	VMALLOC_START = vmax - vmalloc_size;
477 
478 	/* Split remaining virtual space between 1:1 mapping & vmemmap array */
479 	tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
480 	/* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
481 	tmp = SECTION_ALIGN_UP(tmp);
482 	tmp = VMALLOC_START - tmp * sizeof(struct page);
483 	tmp &= ~((vmax >> 11) - 1);	/* align to page table level */
484 	tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
485 	vmemmap = (struct page *) tmp;
486 
487 	/* Take care that memory_end is set and <= vmemmap */
488 	memory_end = min(memory_end ?: max_physmem_end, tmp);
489 	max_pfn = max_low_pfn = PFN_DOWN(memory_end);
490 	memblock_remove(memory_end, ULONG_MAX);
491 
492 	pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
493 }
494 
495 #ifdef CONFIG_CRASH_DUMP
496 
497 /*
498  * When kdump is enabled, we have to ensure that no memory from
499  * the area [0 - crashkernel memory size] and
500  * [crashk_res.start - crashk_res.end] is set offline.
501  */
502 static int kdump_mem_notifier(struct notifier_block *nb,
503 			      unsigned long action, void *data)
504 {
505 	struct memory_notify *arg = data;
506 
507 	if (action != MEM_GOING_OFFLINE)
508 		return NOTIFY_OK;
509 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
510 		return NOTIFY_BAD;
511 	if (arg->start_pfn > PFN_DOWN(crashk_res.end))
512 		return NOTIFY_OK;
513 	if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
514 		return NOTIFY_OK;
515 	return NOTIFY_BAD;
516 }
517 
518 static struct notifier_block kdump_mem_nb = {
519 	.notifier_call = kdump_mem_notifier,
520 };
521 
522 #endif
523 
524 /*
525  * Make sure that the area behind memory_end is protected
526  */
527 static void reserve_memory_end(void)
528 {
529 #ifdef CONFIG_CRASH_DUMP
530 	if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
531 	    !OLDMEM_BASE && sclp.hsa_size) {
532 		memory_end = sclp.hsa_size;
533 		memory_end &= PAGE_MASK;
534 		memory_end_set = 1;
535 	}
536 #endif
537 	if (!memory_end_set)
538 		return;
539 	memblock_reserve(memory_end, ULONG_MAX);
540 }
541 
542 /*
543  * Make sure that oldmem, where the dump is stored, is protected
544  */
545 static void reserve_oldmem(void)
546 {
547 #ifdef CONFIG_CRASH_DUMP
548 	if (OLDMEM_BASE)
549 		/* Forget all memory above the running kdump system */
550 		memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
551 #endif
552 }
553 
554 /*
555  * Make sure that oldmem, where the dump is stored, is protected
556  */
557 static void remove_oldmem(void)
558 {
559 #ifdef CONFIG_CRASH_DUMP
560 	if (OLDMEM_BASE)
561 		/* Forget all memory above the running kdump system */
562 		memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
563 #endif
564 }
565 
566 /*
567  * Reserve memory for kdump kernel to be loaded with kexec
568  */
569 static void __init reserve_crashkernel(void)
570 {
571 #ifdef CONFIG_CRASH_DUMP
572 	unsigned long long crash_base, crash_size;
573 	phys_addr_t low, high;
574 	int rc;
575 
576 	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
577 			       &crash_base);
578 
579 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
580 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
581 	if (rc || crash_size == 0)
582 		return;
583 
584 	if (memblock.memory.regions[0].size < crash_size) {
585 		pr_info("crashkernel reservation failed: %s\n",
586 			"first memory chunk must be at least crashkernel size");
587 		return;
588 	}
589 
590 	low = crash_base ?: OLDMEM_BASE;
591 	high = low + crash_size;
592 	if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
593 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
594 		crash_base = low;
595 	} else {
596 		/* Find suitable area in free memory */
597 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
598 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
599 
600 		if (crash_base && crash_base < low) {
601 			pr_info("crashkernel reservation failed: %s\n",
602 				"crash_base too low");
603 			return;
604 		}
605 		low = crash_base ?: low;
606 		crash_base = memblock_find_in_range(low, high, crash_size,
607 						    KEXEC_CRASH_MEM_ALIGN);
608 	}
609 
610 	if (!crash_base) {
611 		pr_info("crashkernel reservation failed: %s\n",
612 			"no suitable area found");
613 		return;
614 	}
615 
616 	if (register_memory_notifier(&kdump_mem_nb))
617 		return;
618 
619 	if (!OLDMEM_BASE && MACHINE_IS_VM)
620 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
621 	crashk_res.start = crash_base;
622 	crashk_res.end = crash_base + crash_size - 1;
623 	memblock_remove(crash_base, crash_size);
624 	pr_info("Reserving %lluMB of memory at %lluMB "
625 		"for crashkernel (System RAM: %luMB)\n",
626 		crash_size >> 20, crash_base >> 20,
627 		(unsigned long)memblock.memory.total_size >> 20);
628 	os_info_crashkernel_add(crash_base, crash_size);
629 #endif
630 }
631 
632 /*
633  * Reserve the initrd from being used by memblock
634  */
635 static void __init reserve_initrd(void)
636 {
637 #ifdef CONFIG_BLK_DEV_INITRD
638 	if (!INITRD_START || !INITRD_SIZE)
639 		return;
640 	initrd_start = INITRD_START;
641 	initrd_end = initrd_start + INITRD_SIZE;
642 	memblock_reserve(INITRD_START, INITRD_SIZE);
643 #endif
644 }
645 
646 /*
647  * Check for initrd being in usable memory
648  */
649 static void __init check_initrd(void)
650 {
651 #ifdef CONFIG_BLK_DEV_INITRD
652 	if (INITRD_START && INITRD_SIZE &&
653 	    !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
654 		pr_err("The initial RAM disk does not fit into the memory\n");
655 		memblock_free(INITRD_START, INITRD_SIZE);
656 		initrd_start = initrd_end = 0;
657 	}
658 #endif
659 }
660 
661 /*
662  * Reserve memory used for lowcore/command line/kernel image.
663  */
664 static void __init reserve_kernel(void)
665 {
666 	unsigned long start_pfn = PFN_UP(__pa(&_end));
667 
668 #ifdef CONFIG_DMA_API_DEBUG
669 	/*
670 	 * DMA_API_DEBUG code stumbles over addresses from the
671 	 * range [_ehead, _stext]. Mark the memory as reserved
672 	 * so it is not used for CONFIG_DMA_API_DEBUG=y.
673 	 */
674 	memblock_reserve(0, PFN_PHYS(start_pfn));
675 #else
676 	memblock_reserve(0, (unsigned long)_ehead);
677 	memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
678 			 - (unsigned long)_stext);
679 #endif
680 }
681 
682 static void __init setup_memory(void)
683 {
684 	struct memblock_region *reg;
685 
686 	/*
687 	 * Init storage key for present memory
688 	 */
689 	for_each_memblock(memory, reg) {
690 		storage_key_init_range(reg->base, reg->base + reg->size);
691 	}
692 	psw_set_key(PAGE_DEFAULT_KEY);
693 
694 	/* Only cosmetics */
695 	memblock_enforce_memory_limit(memblock_end_of_DRAM());
696 }
697 
698 /*
699  * Setup hardware capabilities.
700  */
701 static int __init setup_hwcaps(void)
702 {
703 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
704 	struct cpuid cpu_id;
705 	int i;
706 
707 	/*
708 	 * The store facility list bits numbers as found in the principles
709 	 * of operation are numbered with bit 1UL<<31 as number 0 to
710 	 * bit 1UL<<0 as number 31.
711 	 *   Bit 0: instructions named N3, "backported" to esa-mode
712 	 *   Bit 2: z/Architecture mode is active
713 	 *   Bit 7: the store-facility-list-extended facility is installed
714 	 *   Bit 17: the message-security assist is installed
715 	 *   Bit 19: the long-displacement facility is installed
716 	 *   Bit 21: the extended-immediate facility is installed
717 	 *   Bit 22: extended-translation facility 3 is installed
718 	 *   Bit 30: extended-translation facility 3 enhancement facility
719 	 * These get translated to:
720 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
721 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
722 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
723 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
724 	 */
725 	for (i = 0; i < 6; i++)
726 		if (test_facility(stfl_bits[i]))
727 			elf_hwcap |= 1UL << i;
728 
729 	if (test_facility(22) && test_facility(30))
730 		elf_hwcap |= HWCAP_S390_ETF3EH;
731 
732 	/*
733 	 * Check for additional facilities with store-facility-list-extended.
734 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
735 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
736 	 * as stored by stfl, bits 32-xxx contain additional facilities.
737 	 * How many facility words are stored depends on the number of
738 	 * doublewords passed to the instruction. The additional facilities
739 	 * are:
740 	 *   Bit 42: decimal floating point facility is installed
741 	 *   Bit 44: perform floating point operation facility is installed
742 	 * translated to:
743 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
744 	 */
745 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
746 		elf_hwcap |= HWCAP_S390_DFP;
747 
748 	/*
749 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
750 	 */
751 	if (MACHINE_HAS_EDAT1)
752 		elf_hwcap |= HWCAP_S390_HPAGE;
753 
754 	/*
755 	 * 64-bit register support for 31-bit processes
756 	 * HWCAP_S390_HIGH_GPRS is bit 9.
757 	 */
758 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
759 
760 	/*
761 	 * Transactional execution support HWCAP_S390_TE is bit 10.
762 	 */
763 	if (MACHINE_HAS_TE)
764 		elf_hwcap |= HWCAP_S390_TE;
765 
766 	/*
767 	 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
768 	 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
769 	 * instead of facility bit 129.
770 	 */
771 	if (MACHINE_HAS_VX) {
772 		elf_hwcap |= HWCAP_S390_VXRS;
773 		if (test_facility(134))
774 			elf_hwcap |= HWCAP_S390_VXRS_EXT;
775 		if (test_facility(135))
776 			elf_hwcap |= HWCAP_S390_VXRS_BCD;
777 	}
778 
779 	/*
780 	 * Guarded storage support HWCAP_S390_GS is bit 12.
781 	 */
782 	if (MACHINE_HAS_GS)
783 		elf_hwcap |= HWCAP_S390_GS;
784 
785 	get_cpu_id(&cpu_id);
786 	add_device_randomness(&cpu_id, sizeof(cpu_id));
787 	switch (cpu_id.machine) {
788 	case 0x2064:
789 	case 0x2066:
790 	default:	/* Use "z900" as default for 64 bit kernels. */
791 		strcpy(elf_platform, "z900");
792 		break;
793 	case 0x2084:
794 	case 0x2086:
795 		strcpy(elf_platform, "z990");
796 		break;
797 	case 0x2094:
798 	case 0x2096:
799 		strcpy(elf_platform, "z9-109");
800 		break;
801 	case 0x2097:
802 	case 0x2098:
803 		strcpy(elf_platform, "z10");
804 		break;
805 	case 0x2817:
806 	case 0x2818:
807 		strcpy(elf_platform, "z196");
808 		break;
809 	case 0x2827:
810 	case 0x2828:
811 		strcpy(elf_platform, "zEC12");
812 		break;
813 	case 0x2964:
814 	case 0x2965:
815 		strcpy(elf_platform, "z13");
816 		break;
817 	case 0x3906:
818 		strcpy(elf_platform, "z14");
819 		break;
820 	}
821 
822 	/*
823 	 * Virtualization support HWCAP_INT_SIE is bit 0.
824 	 */
825 	if (sclp.has_sief2)
826 		int_hwcap |= HWCAP_INT_SIE;
827 
828 	return 0;
829 }
830 arch_initcall(setup_hwcaps);
831 
832 /*
833  * Add system information as device randomness
834  */
835 static void __init setup_randomness(void)
836 {
837 	struct sysinfo_3_2_2 *vmms;
838 
839 	vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE);
840 	if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
841 		add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
842 	memblock_free((unsigned long) vmms, PAGE_SIZE);
843 }
844 
845 /*
846  * Find the correct size for the task_struct. This depends on
847  * the size of the struct fpu at the end of the thread_struct
848  * which is embedded in the task_struct.
849  */
850 static void __init setup_task_size(void)
851 {
852 	int task_size = sizeof(struct task_struct);
853 
854 	if (!MACHINE_HAS_VX) {
855 		task_size -= sizeof(__vector128) * __NUM_VXRS;
856 		task_size += sizeof(freg_t) * __NUM_FPRS;
857 	}
858 	arch_task_struct_size = task_size;
859 }
860 
861 /*
862  * Setup function called from init/main.c just after the banner
863  * was printed.
864  */
865 
866 void __init setup_arch(char **cmdline_p)
867 {
868         /*
869          * print what head.S has found out about the machine
870          */
871 	if (MACHINE_IS_VM)
872 		pr_info("Linux is running as a z/VM "
873 			"guest operating system in 64-bit mode\n");
874 	else if (MACHINE_IS_KVM)
875 		pr_info("Linux is running under KVM in 64-bit mode\n");
876 	else if (MACHINE_IS_LPAR)
877 		pr_info("Linux is running natively in 64-bit mode\n");
878 
879 	/* Have one command line that is parsed and saved in /proc/cmdline */
880 	/* boot_command_line has been already set up in early.c */
881 	*cmdline_p = boot_command_line;
882 
883         ROOT_DEV = Root_RAM0;
884 
885 	/* Is init_mm really needed? */
886 	init_mm.start_code = PAGE_OFFSET;
887 	init_mm.end_code = (unsigned long) &_etext;
888 	init_mm.end_data = (unsigned long) &_edata;
889 	init_mm.brk = (unsigned long) &_end;
890 
891 	parse_early_param();
892 #ifdef CONFIG_CRASH_DUMP
893 	/* Deactivate elfcorehdr= kernel parameter */
894 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
895 #endif
896 
897 	os_info_init();
898 	setup_ipl();
899 	setup_task_size();
900 
901 	/* Do some memory reservations *before* memory is added to memblock */
902 	reserve_memory_end();
903 	reserve_oldmem();
904 	reserve_kernel();
905 	reserve_initrd();
906 	memblock_allow_resize();
907 
908 	/* Get information about *all* installed memory */
909 	detect_memory_memblock();
910 
911 	remove_oldmem();
912 
913 	/*
914 	 * Make sure all chunks are MAX_ORDER aligned so we don't need the
915 	 * extra checks that HOLES_IN_ZONE would require.
916 	 *
917 	 * Is this still required?
918 	 */
919 	memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
920 
921 	setup_memory_end();
922 	setup_memory();
923 	dma_contiguous_reserve(memory_end);
924 	vmcp_cma_reserve();
925 
926 	check_initrd();
927 	reserve_crashkernel();
928 #ifdef CONFIG_CRASH_DUMP
929 	/*
930 	 * Be aware that smp_save_dump_cpus() triggers a system reset.
931 	 * Therefore CPU and device initialization should be done afterwards.
932 	 */
933 	smp_save_dump_cpus();
934 #endif
935 
936 	setup_resources();
937 	setup_lowcore();
938 	smp_fill_possible_mask();
939 	cpu_detect_mhz_feature();
940         cpu_init();
941 	numa_setup();
942 	smp_detect_cpus();
943 	topology_init_early();
944 
945 	/*
946 	 * Create kernel page tables and switch to virtual addressing.
947 	 */
948         paging_init();
949 
950         /* Setup default console */
951 	conmode_default();
952 	set_preferred_console();
953 
954 	apply_alternative_instructions();
955 
956 	/* Setup zfcpdump support */
957 	setup_zfcpdump();
958 
959 	/* Add system specific data to the random pool */
960 	setup_randomness();
961 }
962