xref: /openbmc/linux/arch/s390/kernel/setup.c (revision 4e1a33b1)
1 /*
2  *  S390 version
3  *    Copyright IBM Corp. 1999, 2012
4  *    Author(s): Hartmut Penner (hp@de.ibm.com),
5  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
6  *
7  *  Derived from "arch/i386/kernel/setup.c"
8  *    Copyright (C) 1995, Linus Torvalds
9  */
10 
11 /*
12  * This file handles the architecture-dependent parts of initialization
13  */
14 
15 #define KMSG_COMPONENT "setup"
16 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
17 
18 #include <linux/errno.h>
19 #include <linux/export.h>
20 #include <linux/sched.h>
21 #include <linux/kernel.h>
22 #include <linux/memblock.h>
23 #include <linux/mm.h>
24 #include <linux/stddef.h>
25 #include <linux/unistd.h>
26 #include <linux/ptrace.h>
27 #include <linux/random.h>
28 #include <linux/user.h>
29 #include <linux/tty.h>
30 #include <linux/ioport.h>
31 #include <linux/delay.h>
32 #include <linux/init.h>
33 #include <linux/initrd.h>
34 #include <linux/bootmem.h>
35 #include <linux/root_dev.h>
36 #include <linux/console.h>
37 #include <linux/kernel_stat.h>
38 #include <linux/dma-contiguous.h>
39 #include <linux/device.h>
40 #include <linux/notifier.h>
41 #include <linux/pfn.h>
42 #include <linux/ctype.h>
43 #include <linux/reboot.h>
44 #include <linux/topology.h>
45 #include <linux/kexec.h>
46 #include <linux/crash_dump.h>
47 #include <linux/memory.h>
48 #include <linux/compat.h>
49 
50 #include <asm/ipl.h>
51 #include <asm/facility.h>
52 #include <asm/smp.h>
53 #include <asm/mmu_context.h>
54 #include <asm/cpcmd.h>
55 #include <asm/lowcore.h>
56 #include <asm/irq.h>
57 #include <asm/page.h>
58 #include <asm/ptrace.h>
59 #include <asm/sections.h>
60 #include <asm/ebcdic.h>
61 #include <asm/kvm_virtio.h>
62 #include <asm/diag.h>
63 #include <asm/os_info.h>
64 #include <asm/sclp.h>
65 #include <asm/sysinfo.h>
66 #include <asm/numa.h>
67 #include "entry.h"
68 
69 /*
70  * Machine setup..
71  */
72 unsigned int console_mode = 0;
73 EXPORT_SYMBOL(console_mode);
74 
75 unsigned int console_devno = -1;
76 EXPORT_SYMBOL(console_devno);
77 
78 unsigned int console_irq = -1;
79 EXPORT_SYMBOL(console_irq);
80 
81 unsigned long elf_hwcap __read_mostly = 0;
82 char elf_platform[ELF_PLATFORM_SIZE];
83 
84 unsigned long int_hwcap = 0;
85 
86 int __initdata memory_end_set;
87 unsigned long __initdata memory_end;
88 unsigned long __initdata max_physmem_end;
89 
90 unsigned long VMALLOC_START;
91 EXPORT_SYMBOL(VMALLOC_START);
92 
93 unsigned long VMALLOC_END;
94 EXPORT_SYMBOL(VMALLOC_END);
95 
96 struct page *vmemmap;
97 EXPORT_SYMBOL(vmemmap);
98 
99 unsigned long MODULES_VADDR;
100 unsigned long MODULES_END;
101 
102 /* An array with a pointer to the lowcore of every CPU. */
103 struct lowcore *lowcore_ptr[NR_CPUS];
104 EXPORT_SYMBOL(lowcore_ptr);
105 
106 /*
107  * This is set up by the setup-routine at boot-time
108  * for S390 need to find out, what we have to setup
109  * using address 0x10400 ...
110  */
111 
112 #include <asm/setup.h>
113 
114 /*
115  * condev= and conmode= setup parameter.
116  */
117 
118 static int __init condev_setup(char *str)
119 {
120 	int vdev;
121 
122 	vdev = simple_strtoul(str, &str, 0);
123 	if (vdev >= 0 && vdev < 65536) {
124 		console_devno = vdev;
125 		console_irq = -1;
126 	}
127 	return 1;
128 }
129 
130 __setup("condev=", condev_setup);
131 
132 static void __init set_preferred_console(void)
133 {
134 	if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
135 		add_preferred_console("ttyS", 0, NULL);
136 	else if (CONSOLE_IS_3270)
137 		add_preferred_console("tty3270", 0, NULL);
138 	else if (CONSOLE_IS_VT220)
139 		add_preferred_console("ttyS", 1, NULL);
140 	else if (CONSOLE_IS_HVC)
141 		add_preferred_console("hvc", 0, NULL);
142 }
143 
144 static int __init conmode_setup(char *str)
145 {
146 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
147 	if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
148                 SET_CONSOLE_SCLP;
149 #endif
150 #if defined(CONFIG_TN3215_CONSOLE)
151 	if (strncmp(str, "3215", 5) == 0)
152 		SET_CONSOLE_3215;
153 #endif
154 #if defined(CONFIG_TN3270_CONSOLE)
155 	if (strncmp(str, "3270", 5) == 0)
156 		SET_CONSOLE_3270;
157 #endif
158 	set_preferred_console();
159         return 1;
160 }
161 
162 __setup("conmode=", conmode_setup);
163 
164 static void __init conmode_default(void)
165 {
166 	char query_buffer[1024];
167 	char *ptr;
168 
169         if (MACHINE_IS_VM) {
170 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
171 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
172 		ptr = strstr(query_buffer, "SUBCHANNEL =");
173 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
174 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
175 		ptr = strstr(query_buffer, "CONMODE");
176 		/*
177 		 * Set the conmode to 3215 so that the device recognition
178 		 * will set the cu_type of the console to 3215. If the
179 		 * conmode is 3270 and we don't set it back then both
180 		 * 3215 and the 3270 driver will try to access the console
181 		 * device (3215 as console and 3270 as normal tty).
182 		 */
183 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
184 		if (ptr == NULL) {
185 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
186 			SET_CONSOLE_SCLP;
187 #endif
188 			return;
189 		}
190 		if (strncmp(ptr + 8, "3270", 4) == 0) {
191 #if defined(CONFIG_TN3270_CONSOLE)
192 			SET_CONSOLE_3270;
193 #elif defined(CONFIG_TN3215_CONSOLE)
194 			SET_CONSOLE_3215;
195 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
196 			SET_CONSOLE_SCLP;
197 #endif
198 		} else if (strncmp(ptr + 8, "3215", 4) == 0) {
199 #if defined(CONFIG_TN3215_CONSOLE)
200 			SET_CONSOLE_3215;
201 #elif defined(CONFIG_TN3270_CONSOLE)
202 			SET_CONSOLE_3270;
203 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
204 			SET_CONSOLE_SCLP;
205 #endif
206 		}
207 	} else if (MACHINE_IS_KVM) {
208 		if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
209 			SET_CONSOLE_VT220;
210 		else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
211 			SET_CONSOLE_SCLP;
212 		else
213 			SET_CONSOLE_HVC;
214 	} else {
215 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
216 		SET_CONSOLE_SCLP;
217 #endif
218 	}
219 }
220 
221 #ifdef CONFIG_CRASH_DUMP
222 static void __init setup_zfcpdump(void)
223 {
224 	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
225 		return;
226 	if (OLDMEM_BASE)
227 		return;
228 	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
229 	console_loglevel = 2;
230 }
231 #else
232 static inline void setup_zfcpdump(void) {}
233 #endif /* CONFIG_CRASH_DUMP */
234 
235  /*
236  * Reboot, halt and power_off stubs. They just call _machine_restart,
237  * _machine_halt or _machine_power_off.
238  */
239 
240 void machine_restart(char *command)
241 {
242 	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
243 		/*
244 		 * Only unblank the console if we are called in enabled
245 		 * context or a bust_spinlocks cleared the way for us.
246 		 */
247 		console_unblank();
248 	_machine_restart(command);
249 }
250 
251 void machine_halt(void)
252 {
253 	if (!in_interrupt() || oops_in_progress)
254 		/*
255 		 * Only unblank the console if we are called in enabled
256 		 * context or a bust_spinlocks cleared the way for us.
257 		 */
258 		console_unblank();
259 	_machine_halt();
260 }
261 
262 void machine_power_off(void)
263 {
264 	if (!in_interrupt() || oops_in_progress)
265 		/*
266 		 * Only unblank the console if we are called in enabled
267 		 * context or a bust_spinlocks cleared the way for us.
268 		 */
269 		console_unblank();
270 	_machine_power_off();
271 }
272 
273 /*
274  * Dummy power off function.
275  */
276 void (*pm_power_off)(void) = machine_power_off;
277 EXPORT_SYMBOL_GPL(pm_power_off);
278 
279 static int __init early_parse_mem(char *p)
280 {
281 	memory_end = memparse(p, &p);
282 	memory_end &= PAGE_MASK;
283 	memory_end_set = 1;
284 	return 0;
285 }
286 early_param("mem", early_parse_mem);
287 
288 static int __init parse_vmalloc(char *arg)
289 {
290 	if (!arg)
291 		return -EINVAL;
292 	VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
293 	return 0;
294 }
295 early_param("vmalloc", parse_vmalloc);
296 
297 void *restart_stack __section(.data);
298 
299 static void __init setup_lowcore(void)
300 {
301 	struct lowcore *lc;
302 
303 	/*
304 	 * Setup lowcore for boot cpu
305 	 */
306 	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * 4096);
307 	lc = memblock_virt_alloc_low(sizeof(*lc), sizeof(*lc));
308 	lc->restart_psw.mask = PSW_KERNEL_BITS;
309 	lc->restart_psw.addr = (unsigned long) restart_int_handler;
310 	lc->external_new_psw.mask = PSW_KERNEL_BITS |
311 		PSW_MASK_DAT | PSW_MASK_MCHECK;
312 	lc->external_new_psw.addr = (unsigned long) ext_int_handler;
313 	lc->svc_new_psw.mask = PSW_KERNEL_BITS |
314 		PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
315 	lc->svc_new_psw.addr = (unsigned long) system_call;
316 	lc->program_new_psw.mask = PSW_KERNEL_BITS |
317 		PSW_MASK_DAT | PSW_MASK_MCHECK;
318 	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
319 	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
320 	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
321 	lc->io_new_psw.mask = PSW_KERNEL_BITS |
322 		PSW_MASK_DAT | PSW_MASK_MCHECK;
323 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
324 	lc->clock_comparator = -1ULL;
325 	lc->kernel_stack = ((unsigned long) &init_thread_union)
326 		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
327 	lc->async_stack = (unsigned long)
328 		memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE)
329 		+ ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
330 	lc->panic_stack = (unsigned long)
331 		memblock_virt_alloc(PAGE_SIZE, PAGE_SIZE)
332 		+ PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
333 	lc->current_task = (unsigned long)&init_task;
334 	lc->lpp = LPP_MAGIC;
335 	lc->machine_flags = S390_lowcore.machine_flags;
336 	lc->preempt_count = S390_lowcore.preempt_count;
337 	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
338 	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
339 	       MAX_FACILITY_BIT/8);
340 	if (MACHINE_HAS_VX)
341 		lc->vector_save_area_addr =
342 			(unsigned long) &lc->vector_save_area;
343 	lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
344 	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
345 	lc->async_enter_timer = S390_lowcore.async_enter_timer;
346 	lc->exit_timer = S390_lowcore.exit_timer;
347 	lc->user_timer = S390_lowcore.user_timer;
348 	lc->system_timer = S390_lowcore.system_timer;
349 	lc->steal_timer = S390_lowcore.steal_timer;
350 	lc->last_update_timer = S390_lowcore.last_update_timer;
351 	lc->last_update_clock = S390_lowcore.last_update_clock;
352 
353 	restart_stack = memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE);
354 	restart_stack += ASYNC_SIZE;
355 
356 	/*
357 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
358 	 * restart data to the absolute zero lowcore. This is necessary if
359 	 * PSW restart is done on an offline CPU that has lowcore zero.
360 	 */
361 	lc->restart_stack = (unsigned long) restart_stack;
362 	lc->restart_fn = (unsigned long) do_restart;
363 	lc->restart_data = 0;
364 	lc->restart_source = -1UL;
365 
366 	/* Setup absolute zero lowcore */
367 	mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
368 	mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
369 	mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
370 	mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
371 	mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
372 
373 #ifdef CONFIG_SMP
374 	lc->spinlock_lockval = arch_spin_lockval(0);
375 #endif
376 
377 	set_prefix((u32)(unsigned long) lc);
378 	lowcore_ptr[0] = lc;
379 }
380 
381 static struct resource code_resource = {
382 	.name  = "Kernel code",
383 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
384 };
385 
386 static struct resource data_resource = {
387 	.name = "Kernel data",
388 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
389 };
390 
391 static struct resource bss_resource = {
392 	.name = "Kernel bss",
393 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
394 };
395 
396 static struct resource __initdata *standard_resources[] = {
397 	&code_resource,
398 	&data_resource,
399 	&bss_resource,
400 };
401 
402 static void __init setup_resources(void)
403 {
404 	struct resource *res, *std_res, *sub_res;
405 	struct memblock_region *reg;
406 	int j;
407 
408 	code_resource.start = (unsigned long) &_text;
409 	code_resource.end = (unsigned long) &_etext - 1;
410 	data_resource.start = (unsigned long) &_etext;
411 	data_resource.end = (unsigned long) &_edata - 1;
412 	bss_resource.start = (unsigned long) &__bss_start;
413 	bss_resource.end = (unsigned long) &__bss_stop - 1;
414 
415 	for_each_memblock(memory, reg) {
416 		res = memblock_virt_alloc(sizeof(*res), 8);
417 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
418 
419 		res->name = "System RAM";
420 		res->start = reg->base;
421 		res->end = reg->base + reg->size - 1;
422 		request_resource(&iomem_resource, res);
423 
424 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
425 			std_res = standard_resources[j];
426 			if (std_res->start < res->start ||
427 			    std_res->start > res->end)
428 				continue;
429 			if (std_res->end > res->end) {
430 				sub_res = memblock_virt_alloc(sizeof(*sub_res), 8);
431 				*sub_res = *std_res;
432 				sub_res->end = res->end;
433 				std_res->start = res->end + 1;
434 				request_resource(res, sub_res);
435 			} else {
436 				request_resource(res, std_res);
437 			}
438 		}
439 	}
440 #ifdef CONFIG_CRASH_DUMP
441 	/*
442 	 * Re-add removed crash kernel memory as reserved memory. This makes
443 	 * sure it will be mapped with the identity mapping and struct pages
444 	 * will be created, so it can be resized later on.
445 	 * However add it later since the crash kernel resource should not be
446 	 * part of the System RAM resource.
447 	 */
448 	if (crashk_res.end) {
449 		memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
450 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
451 		insert_resource(&iomem_resource, &crashk_res);
452 	}
453 #endif
454 }
455 
456 static void __init setup_memory_end(void)
457 {
458 	unsigned long vmax, vmalloc_size, tmp;
459 
460 	/* Choose kernel address space layout: 2, 3, or 4 levels. */
461 	vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
462 	tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
463 	tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
464 	if (tmp + vmalloc_size + MODULES_LEN <= (1UL << 42))
465 		vmax = 1UL << 42;	/* 3-level kernel page table */
466 	else
467 		vmax = 1UL << 53;	/* 4-level kernel page table */
468 	/* module area is at the end of the kernel address space. */
469 	MODULES_END = vmax;
470 	MODULES_VADDR = MODULES_END - MODULES_LEN;
471 	VMALLOC_END = MODULES_VADDR;
472 	VMALLOC_START = vmax - vmalloc_size;
473 
474 	/* Split remaining virtual space between 1:1 mapping & vmemmap array */
475 	tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
476 	/* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
477 	tmp = SECTION_ALIGN_UP(tmp);
478 	tmp = VMALLOC_START - tmp * sizeof(struct page);
479 	tmp &= ~((vmax >> 11) - 1);	/* align to page table level */
480 	tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
481 	vmemmap = (struct page *) tmp;
482 
483 	/* Take care that memory_end is set and <= vmemmap */
484 	memory_end = min(memory_end ?: max_physmem_end, tmp);
485 	max_pfn = max_low_pfn = PFN_DOWN(memory_end);
486 	memblock_remove(memory_end, ULONG_MAX);
487 
488 	pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
489 }
490 
491 static void __init setup_vmcoreinfo(void)
492 {
493 	mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
494 }
495 
496 #ifdef CONFIG_CRASH_DUMP
497 
498 /*
499  * When kdump is enabled, we have to ensure that no memory from
500  * the area [0 - crashkernel memory size] and
501  * [crashk_res.start - crashk_res.end] is set offline.
502  */
503 static int kdump_mem_notifier(struct notifier_block *nb,
504 			      unsigned long action, void *data)
505 {
506 	struct memory_notify *arg = data;
507 
508 	if (action != MEM_GOING_OFFLINE)
509 		return NOTIFY_OK;
510 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
511 		return NOTIFY_BAD;
512 	if (arg->start_pfn > PFN_DOWN(crashk_res.end))
513 		return NOTIFY_OK;
514 	if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
515 		return NOTIFY_OK;
516 	return NOTIFY_BAD;
517 }
518 
519 static struct notifier_block kdump_mem_nb = {
520 	.notifier_call = kdump_mem_notifier,
521 };
522 
523 #endif
524 
525 /*
526  * Make sure that the area behind memory_end is protected
527  */
528 static void reserve_memory_end(void)
529 {
530 #ifdef CONFIG_CRASH_DUMP
531 	if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
532 	    !OLDMEM_BASE && sclp.hsa_size) {
533 		memory_end = sclp.hsa_size;
534 		memory_end &= PAGE_MASK;
535 		memory_end_set = 1;
536 	}
537 #endif
538 	if (!memory_end_set)
539 		return;
540 	memblock_reserve(memory_end, ULONG_MAX);
541 }
542 
543 /*
544  * Make sure that oldmem, where the dump is stored, is protected
545  */
546 static void reserve_oldmem(void)
547 {
548 #ifdef CONFIG_CRASH_DUMP
549 	if (OLDMEM_BASE)
550 		/* Forget all memory above the running kdump system */
551 		memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
552 #endif
553 }
554 
555 /*
556  * Make sure that oldmem, where the dump is stored, is protected
557  */
558 static void remove_oldmem(void)
559 {
560 #ifdef CONFIG_CRASH_DUMP
561 	if (OLDMEM_BASE)
562 		/* Forget all memory above the running kdump system */
563 		memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
564 #endif
565 }
566 
567 /*
568  * Reserve memory for kdump kernel to be loaded with kexec
569  */
570 static void __init reserve_crashkernel(void)
571 {
572 #ifdef CONFIG_CRASH_DUMP
573 	unsigned long long crash_base, crash_size;
574 	phys_addr_t low, high;
575 	int rc;
576 
577 	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
578 			       &crash_base);
579 
580 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
581 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
582 	if (rc || crash_size == 0)
583 		return;
584 
585 	if (memblock.memory.regions[0].size < crash_size) {
586 		pr_info("crashkernel reservation failed: %s\n",
587 			"first memory chunk must be at least crashkernel size");
588 		return;
589 	}
590 
591 	low = crash_base ?: OLDMEM_BASE;
592 	high = low + crash_size;
593 	if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
594 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
595 		crash_base = low;
596 	} else {
597 		/* Find suitable area in free memory */
598 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
599 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
600 
601 		if (crash_base && crash_base < low) {
602 			pr_info("crashkernel reservation failed: %s\n",
603 				"crash_base too low");
604 			return;
605 		}
606 		low = crash_base ?: low;
607 		crash_base = memblock_find_in_range(low, high, crash_size,
608 						    KEXEC_CRASH_MEM_ALIGN);
609 	}
610 
611 	if (!crash_base) {
612 		pr_info("crashkernel reservation failed: %s\n",
613 			"no suitable area found");
614 		return;
615 	}
616 
617 	if (register_memory_notifier(&kdump_mem_nb))
618 		return;
619 
620 	if (!OLDMEM_BASE && MACHINE_IS_VM)
621 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
622 	crashk_res.start = crash_base;
623 	crashk_res.end = crash_base + crash_size - 1;
624 	memblock_remove(crash_base, crash_size);
625 	pr_info("Reserving %lluMB of memory at %lluMB "
626 		"for crashkernel (System RAM: %luMB)\n",
627 		crash_size >> 20, crash_base >> 20,
628 		(unsigned long)memblock.memory.total_size >> 20);
629 	os_info_crashkernel_add(crash_base, crash_size);
630 #endif
631 }
632 
633 /*
634  * Reserve the initrd from being used by memblock
635  */
636 static void __init reserve_initrd(void)
637 {
638 #ifdef CONFIG_BLK_DEV_INITRD
639 	if (!INITRD_START || !INITRD_SIZE)
640 		return;
641 	initrd_start = INITRD_START;
642 	initrd_end = initrd_start + INITRD_SIZE;
643 	memblock_reserve(INITRD_START, INITRD_SIZE);
644 #endif
645 }
646 
647 /*
648  * Check for initrd being in usable memory
649  */
650 static void __init check_initrd(void)
651 {
652 #ifdef CONFIG_BLK_DEV_INITRD
653 	if (INITRD_START && INITRD_SIZE &&
654 	    !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
655 		pr_err("The initial RAM disk does not fit into the memory\n");
656 		memblock_free(INITRD_START, INITRD_SIZE);
657 		initrd_start = initrd_end = 0;
658 	}
659 #endif
660 }
661 
662 /*
663  * Reserve memory used for lowcore/command line/kernel image.
664  */
665 static void __init reserve_kernel(void)
666 {
667 	unsigned long start_pfn = PFN_UP(__pa(&_end));
668 
669 #ifdef CONFIG_DMA_API_DEBUG
670 	/*
671 	 * DMA_API_DEBUG code stumbles over addresses from the
672 	 * range [_ehead, _stext]. Mark the memory as reserved
673 	 * so it is not used for CONFIG_DMA_API_DEBUG=y.
674 	 */
675 	memblock_reserve(0, PFN_PHYS(start_pfn));
676 #else
677 	memblock_reserve(0, (unsigned long)_ehead);
678 	memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
679 			 - (unsigned long)_stext);
680 #endif
681 }
682 
683 static void __init setup_memory(void)
684 {
685 	struct memblock_region *reg;
686 
687 	/*
688 	 * Init storage key for present memory
689 	 */
690 	for_each_memblock(memory, reg) {
691 		storage_key_init_range(reg->base, reg->base + reg->size);
692 	}
693 	psw_set_key(PAGE_DEFAULT_KEY);
694 
695 	/* Only cosmetics */
696 	memblock_enforce_memory_limit(memblock_end_of_DRAM());
697 }
698 
699 /*
700  * Setup hardware capabilities.
701  */
702 static int __init setup_hwcaps(void)
703 {
704 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
705 	struct cpuid cpu_id;
706 	int i;
707 
708 	/*
709 	 * The store facility list bits numbers as found in the principles
710 	 * of operation are numbered with bit 1UL<<31 as number 0 to
711 	 * bit 1UL<<0 as number 31.
712 	 *   Bit 0: instructions named N3, "backported" to esa-mode
713 	 *   Bit 2: z/Architecture mode is active
714 	 *   Bit 7: the store-facility-list-extended facility is installed
715 	 *   Bit 17: the message-security assist is installed
716 	 *   Bit 19: the long-displacement facility is installed
717 	 *   Bit 21: the extended-immediate facility is installed
718 	 *   Bit 22: extended-translation facility 3 is installed
719 	 *   Bit 30: extended-translation facility 3 enhancement facility
720 	 * These get translated to:
721 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
722 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
723 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
724 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
725 	 */
726 	for (i = 0; i < 6; i++)
727 		if (test_facility(stfl_bits[i]))
728 			elf_hwcap |= 1UL << i;
729 
730 	if (test_facility(22) && test_facility(30))
731 		elf_hwcap |= HWCAP_S390_ETF3EH;
732 
733 	/*
734 	 * Check for additional facilities with store-facility-list-extended.
735 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
736 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
737 	 * as stored by stfl, bits 32-xxx contain additional facilities.
738 	 * How many facility words are stored depends on the number of
739 	 * doublewords passed to the instruction. The additional facilities
740 	 * are:
741 	 *   Bit 42: decimal floating point facility is installed
742 	 *   Bit 44: perform floating point operation facility is installed
743 	 * translated to:
744 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
745 	 */
746 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
747 		elf_hwcap |= HWCAP_S390_DFP;
748 
749 	/*
750 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
751 	 */
752 	if (MACHINE_HAS_EDAT1)
753 		elf_hwcap |= HWCAP_S390_HPAGE;
754 
755 	/*
756 	 * 64-bit register support for 31-bit processes
757 	 * HWCAP_S390_HIGH_GPRS is bit 9.
758 	 */
759 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
760 
761 	/*
762 	 * Transactional execution support HWCAP_S390_TE is bit 10.
763 	 */
764 	if (test_facility(50) && test_facility(73))
765 		elf_hwcap |= HWCAP_S390_TE;
766 
767 	/*
768 	 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
769 	 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
770 	 * instead of facility bit 129.
771 	 */
772 	if (MACHINE_HAS_VX) {
773 		elf_hwcap |= HWCAP_S390_VXRS;
774 		if (test_facility(134))
775 			elf_hwcap |= HWCAP_S390_VXRS_EXT;
776 		if (test_facility(135))
777 			elf_hwcap |= HWCAP_S390_VXRS_BCD;
778 	}
779 
780 	get_cpu_id(&cpu_id);
781 	add_device_randomness(&cpu_id, sizeof(cpu_id));
782 	switch (cpu_id.machine) {
783 	case 0x2064:
784 	case 0x2066:
785 	default:	/* Use "z900" as default for 64 bit kernels. */
786 		strcpy(elf_platform, "z900");
787 		break;
788 	case 0x2084:
789 	case 0x2086:
790 		strcpy(elf_platform, "z990");
791 		break;
792 	case 0x2094:
793 	case 0x2096:
794 		strcpy(elf_platform, "z9-109");
795 		break;
796 	case 0x2097:
797 	case 0x2098:
798 		strcpy(elf_platform, "z10");
799 		break;
800 	case 0x2817:
801 	case 0x2818:
802 		strcpy(elf_platform, "z196");
803 		break;
804 	case 0x2827:
805 	case 0x2828:
806 		strcpy(elf_platform, "zEC12");
807 		break;
808 	case 0x2964:
809 	case 0x2965:
810 		strcpy(elf_platform, "z13");
811 		break;
812 	}
813 
814 	/*
815 	 * Virtualization support HWCAP_INT_SIE is bit 0.
816 	 */
817 	if (sclp.has_sief2)
818 		int_hwcap |= HWCAP_INT_SIE;
819 
820 	return 0;
821 }
822 arch_initcall(setup_hwcaps);
823 
824 /*
825  * Add system information as device randomness
826  */
827 static void __init setup_randomness(void)
828 {
829 	struct sysinfo_3_2_2 *vmms;
830 
831 	vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE);
832 	if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
833 		add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
834 	memblock_free((unsigned long) vmms, PAGE_SIZE);
835 }
836 
837 /*
838  * Find the correct size for the task_struct. This depends on
839  * the size of the struct fpu at the end of the thread_struct
840  * which is embedded in the task_struct.
841  */
842 static void __init setup_task_size(void)
843 {
844 	int task_size = sizeof(struct task_struct);
845 
846 	if (!MACHINE_HAS_VX) {
847 		task_size -= sizeof(__vector128) * __NUM_VXRS;
848 		task_size += sizeof(freg_t) * __NUM_FPRS;
849 	}
850 	arch_task_struct_size = task_size;
851 }
852 
853 /*
854  * Setup function called from init/main.c just after the banner
855  * was printed.
856  */
857 
858 void __init setup_arch(char **cmdline_p)
859 {
860         /*
861          * print what head.S has found out about the machine
862          */
863 	if (MACHINE_IS_VM)
864 		pr_info("Linux is running as a z/VM "
865 			"guest operating system in 64-bit mode\n");
866 	else if (MACHINE_IS_KVM)
867 		pr_info("Linux is running under KVM in 64-bit mode\n");
868 	else if (MACHINE_IS_LPAR)
869 		pr_info("Linux is running natively in 64-bit mode\n");
870 
871 	/* Have one command line that is parsed and saved in /proc/cmdline */
872 	/* boot_command_line has been already set up in early.c */
873 	*cmdline_p = boot_command_line;
874 
875         ROOT_DEV = Root_RAM0;
876 
877 	/* Is init_mm really needed? */
878 	init_mm.start_code = PAGE_OFFSET;
879 	init_mm.end_code = (unsigned long) &_etext;
880 	init_mm.end_data = (unsigned long) &_edata;
881 	init_mm.brk = (unsigned long) &_end;
882 
883 	parse_early_param();
884 #ifdef CONFIG_CRASH_DUMP
885 	/* Deactivate elfcorehdr= kernel parameter */
886 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
887 #endif
888 
889 	os_info_init();
890 	setup_ipl();
891 	setup_task_size();
892 
893 	/* Do some memory reservations *before* memory is added to memblock */
894 	reserve_memory_end();
895 	reserve_oldmem();
896 	reserve_kernel();
897 	reserve_initrd();
898 	memblock_allow_resize();
899 
900 	/* Get information about *all* installed memory */
901 	detect_memory_memblock();
902 
903 	remove_oldmem();
904 
905 	/*
906 	 * Make sure all chunks are MAX_ORDER aligned so we don't need the
907 	 * extra checks that HOLES_IN_ZONE would require.
908 	 *
909 	 * Is this still required?
910 	 */
911 	memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
912 
913 	setup_memory_end();
914 	setup_memory();
915 	dma_contiguous_reserve(memory_end);
916 
917 	check_initrd();
918 	reserve_crashkernel();
919 #ifdef CONFIG_CRASH_DUMP
920 	/*
921 	 * Be aware that smp_save_dump_cpus() triggers a system reset.
922 	 * Therefore CPU and device initialization should be done afterwards.
923 	 */
924 	smp_save_dump_cpus();
925 #endif
926 
927 	setup_resources();
928 	setup_vmcoreinfo();
929 	setup_lowcore();
930 	smp_fill_possible_mask();
931 	cpu_detect_mhz_feature();
932         cpu_init();
933 	numa_setup();
934 	smp_detect_cpus();
935 	topology_init_early();
936 
937 	/*
938 	 * Create kernel page tables and switch to virtual addressing.
939 	 */
940         paging_init();
941 
942         /* Setup default console */
943 	conmode_default();
944 	set_preferred_console();
945 
946 	/* Setup zfcpdump support */
947 	setup_zfcpdump();
948 
949 	/* Add system specific data to the random pool */
950 	setup_randomness();
951 }
952