xref: /openbmc/linux/arch/s390/kernel/setup.c (revision 78700c0a)
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/device.h>
39 #include <linux/notifier.h>
40 #include <linux/pfn.h>
41 #include <linux/ctype.h>
42 #include <linux/reboot.h>
43 #include <linux/topology.h>
44 #include <linux/kexec.h>
45 #include <linux/crash_dump.h>
46 #include <linux/memory.h>
47 #include <linux/compat.h>
48 
49 #include <asm/ipl.h>
50 #include <asm/facility.h>
51 #include <asm/smp.h>
52 #include <asm/mmu_context.h>
53 #include <asm/cpcmd.h>
54 #include <asm/lowcore.h>
55 #include <asm/irq.h>
56 #include <asm/page.h>
57 #include <asm/ptrace.h>
58 #include <asm/sections.h>
59 #include <asm/ebcdic.h>
60 #include <asm/kvm_virtio.h>
61 #include <asm/diag.h>
62 #include <asm/os_info.h>
63 #include <asm/sclp.h>
64 #include <asm/sysinfo.h>
65 #include <asm/numa.h>
66 #include "entry.h"
67 
68 /*
69  * Machine setup..
70  */
71 unsigned int console_mode = 0;
72 EXPORT_SYMBOL(console_mode);
73 
74 unsigned int console_devno = -1;
75 EXPORT_SYMBOL(console_devno);
76 
77 unsigned int console_irq = -1;
78 EXPORT_SYMBOL(console_irq);
79 
80 unsigned long elf_hwcap __read_mostly = 0;
81 char elf_platform[ELF_PLATFORM_SIZE];
82 
83 unsigned long int_hwcap = 0;
84 
85 int __initdata memory_end_set;
86 unsigned long __initdata memory_end;
87 unsigned long __initdata max_physmem_end;
88 
89 unsigned long VMALLOC_START;
90 EXPORT_SYMBOL(VMALLOC_START);
91 
92 unsigned long VMALLOC_END;
93 EXPORT_SYMBOL(VMALLOC_END);
94 
95 struct page *vmemmap;
96 EXPORT_SYMBOL(vmemmap);
97 
98 unsigned long MODULES_VADDR;
99 unsigned long MODULES_END;
100 
101 /* An array with a pointer to the lowcore of every CPU. */
102 struct lowcore *lowcore_ptr[NR_CPUS];
103 EXPORT_SYMBOL(lowcore_ptr);
104 
105 /*
106  * This is set up by the setup-routine at boot-time
107  * for S390 need to find out, what we have to setup
108  * using address 0x10400 ...
109  */
110 
111 #include <asm/setup.h>
112 
113 /*
114  * condev= and conmode= setup parameter.
115  */
116 
117 static int __init condev_setup(char *str)
118 {
119 	int vdev;
120 
121 	vdev = simple_strtoul(str, &str, 0);
122 	if (vdev >= 0 && vdev < 65536) {
123 		console_devno = vdev;
124 		console_irq = -1;
125 	}
126 	return 1;
127 }
128 
129 __setup("condev=", condev_setup);
130 
131 static void __init set_preferred_console(void)
132 {
133 	if (MACHINE_IS_KVM) {
134 		if (sclp.has_vt220)
135 			add_preferred_console("ttyS", 1, NULL);
136 		else if (sclp.has_linemode)
137 			add_preferred_console("ttyS", 0, NULL);
138 		else
139 			add_preferred_console("hvc", 0, NULL);
140 	} else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
141 		add_preferred_console("ttyS", 0, NULL);
142 	else if (CONSOLE_IS_3270)
143 		add_preferred_console("tty3270", 0, NULL);
144 }
145 
146 static int __init conmode_setup(char *str)
147 {
148 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
149 	if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
150                 SET_CONSOLE_SCLP;
151 #endif
152 #if defined(CONFIG_TN3215_CONSOLE)
153 	if (strncmp(str, "3215", 5) == 0)
154 		SET_CONSOLE_3215;
155 #endif
156 #if defined(CONFIG_TN3270_CONSOLE)
157 	if (strncmp(str, "3270", 5) == 0)
158 		SET_CONSOLE_3270;
159 #endif
160 	set_preferred_console();
161         return 1;
162 }
163 
164 __setup("conmode=", conmode_setup);
165 
166 static void __init conmode_default(void)
167 {
168 	char query_buffer[1024];
169 	char *ptr;
170 
171         if (MACHINE_IS_VM) {
172 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
173 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
174 		ptr = strstr(query_buffer, "SUBCHANNEL =");
175 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
176 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
177 		ptr = strstr(query_buffer, "CONMODE");
178 		/*
179 		 * Set the conmode to 3215 so that the device recognition
180 		 * will set the cu_type of the console to 3215. If the
181 		 * conmode is 3270 and we don't set it back then both
182 		 * 3215 and the 3270 driver will try to access the console
183 		 * device (3215 as console and 3270 as normal tty).
184 		 */
185 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
186 		if (ptr == NULL) {
187 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
188 			SET_CONSOLE_SCLP;
189 #endif
190 			return;
191 		}
192 		if (strncmp(ptr + 8, "3270", 4) == 0) {
193 #if defined(CONFIG_TN3270_CONSOLE)
194 			SET_CONSOLE_3270;
195 #elif defined(CONFIG_TN3215_CONSOLE)
196 			SET_CONSOLE_3215;
197 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
198 			SET_CONSOLE_SCLP;
199 #endif
200 		} else if (strncmp(ptr + 8, "3215", 4) == 0) {
201 #if defined(CONFIG_TN3215_CONSOLE)
202 			SET_CONSOLE_3215;
203 #elif defined(CONFIG_TN3270_CONSOLE)
204 			SET_CONSOLE_3270;
205 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
206 			SET_CONSOLE_SCLP;
207 #endif
208 		}
209 	} else {
210 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
211 		SET_CONSOLE_SCLP;
212 #endif
213 	}
214 }
215 
216 #ifdef CONFIG_CRASH_DUMP
217 static void __init setup_zfcpdump(void)
218 {
219 	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
220 		return;
221 	if (OLDMEM_BASE)
222 		return;
223 	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
224 	console_loglevel = 2;
225 }
226 #else
227 static inline void setup_zfcpdump(void) {}
228 #endif /* CONFIG_CRASH_DUMP */
229 
230  /*
231  * Reboot, halt and power_off stubs. They just call _machine_restart,
232  * _machine_halt or _machine_power_off.
233  */
234 
235 void machine_restart(char *command)
236 {
237 	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
238 		/*
239 		 * Only unblank the console if we are called in enabled
240 		 * context or a bust_spinlocks cleared the way for us.
241 		 */
242 		console_unblank();
243 	_machine_restart(command);
244 }
245 
246 void machine_halt(void)
247 {
248 	if (!in_interrupt() || oops_in_progress)
249 		/*
250 		 * Only unblank the console if we are called in enabled
251 		 * context or a bust_spinlocks cleared the way for us.
252 		 */
253 		console_unblank();
254 	_machine_halt();
255 }
256 
257 void machine_power_off(void)
258 {
259 	if (!in_interrupt() || oops_in_progress)
260 		/*
261 		 * Only unblank the console if we are called in enabled
262 		 * context or a bust_spinlocks cleared the way for us.
263 		 */
264 		console_unblank();
265 	_machine_power_off();
266 }
267 
268 /*
269  * Dummy power off function.
270  */
271 void (*pm_power_off)(void) = machine_power_off;
272 EXPORT_SYMBOL_GPL(pm_power_off);
273 
274 static int __init early_parse_mem(char *p)
275 {
276 	memory_end = memparse(p, &p);
277 	memory_end &= PAGE_MASK;
278 	memory_end_set = 1;
279 	return 0;
280 }
281 early_param("mem", early_parse_mem);
282 
283 static int __init parse_vmalloc(char *arg)
284 {
285 	if (!arg)
286 		return -EINVAL;
287 	VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
288 	return 0;
289 }
290 early_param("vmalloc", parse_vmalloc);
291 
292 void *restart_stack __attribute__((__section__(".data")));
293 
294 static void __init setup_lowcore(void)
295 {
296 	struct lowcore *lc;
297 
298 	/*
299 	 * Setup lowcore for boot cpu
300 	 */
301 	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * 4096);
302 	lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
303 	lc->restart_psw.mask = PSW_KERNEL_BITS;
304 	lc->restart_psw.addr = (unsigned long) restart_int_handler;
305 	lc->external_new_psw.mask = PSW_KERNEL_BITS |
306 		PSW_MASK_DAT | PSW_MASK_MCHECK;
307 	lc->external_new_psw.addr = (unsigned long) ext_int_handler;
308 	lc->svc_new_psw.mask = PSW_KERNEL_BITS |
309 		PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
310 	lc->svc_new_psw.addr = (unsigned long) system_call;
311 	lc->program_new_psw.mask = PSW_KERNEL_BITS |
312 		PSW_MASK_DAT | PSW_MASK_MCHECK;
313 	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
314 	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
315 	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
316 	lc->io_new_psw.mask = PSW_KERNEL_BITS |
317 		PSW_MASK_DAT | PSW_MASK_MCHECK;
318 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
319 	lc->clock_comparator = -1ULL;
320 	lc->kernel_stack = ((unsigned long) &init_thread_union)
321 		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
322 	lc->async_stack = (unsigned long)
323 		__alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
324 		+ ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
325 	lc->panic_stack = (unsigned long)
326 		__alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
327 		+ PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
328 	lc->current_task = (unsigned long) init_thread_union.thread_info.task;
329 	lc->thread_info = (unsigned long) &init_thread_union;
330 	lc->lpp = LPP_MAGIC;
331 	lc->machine_flags = S390_lowcore.machine_flags;
332 	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
333 	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
334 	       MAX_FACILITY_BIT/8);
335 	if (MACHINE_HAS_VX)
336 		lc->vector_save_area_addr =
337 			(unsigned long) &lc->vector_save_area;
338 	lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
339 	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
340 	lc->async_enter_timer = S390_lowcore.async_enter_timer;
341 	lc->exit_timer = S390_lowcore.exit_timer;
342 	lc->user_timer = S390_lowcore.user_timer;
343 	lc->system_timer = S390_lowcore.system_timer;
344 	lc->steal_timer = S390_lowcore.steal_timer;
345 	lc->last_update_timer = S390_lowcore.last_update_timer;
346 	lc->last_update_clock = S390_lowcore.last_update_clock;
347 
348 	restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
349 	restart_stack += ASYNC_SIZE;
350 
351 	/*
352 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
353 	 * restart data to the absolute zero lowcore. This is necessary if
354 	 * PSW restart is done on an offline CPU that has lowcore zero.
355 	 */
356 	lc->restart_stack = (unsigned long) restart_stack;
357 	lc->restart_fn = (unsigned long) do_restart;
358 	lc->restart_data = 0;
359 	lc->restart_source = -1UL;
360 
361 	/* Setup absolute zero lowcore */
362 	mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
363 	mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
364 	mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
365 	mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
366 	mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
367 
368 #ifdef CONFIG_SMP
369 	lc->spinlock_lockval = arch_spin_lockval(0);
370 #endif
371 
372 	set_prefix((u32)(unsigned long) lc);
373 	lowcore_ptr[0] = lc;
374 }
375 
376 static struct resource code_resource = {
377 	.name  = "Kernel code",
378 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
379 };
380 
381 static struct resource data_resource = {
382 	.name = "Kernel data",
383 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
384 };
385 
386 static struct resource bss_resource = {
387 	.name = "Kernel bss",
388 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
389 };
390 
391 static struct resource __initdata *standard_resources[] = {
392 	&code_resource,
393 	&data_resource,
394 	&bss_resource,
395 };
396 
397 static void __init setup_resources(void)
398 {
399 	struct resource *res, *std_res, *sub_res;
400 	struct memblock_region *reg;
401 	int j;
402 
403 	code_resource.start = (unsigned long) &_text;
404 	code_resource.end = (unsigned long) &_etext - 1;
405 	data_resource.start = (unsigned long) &_etext;
406 	data_resource.end = (unsigned long) &_edata - 1;
407 	bss_resource.start = (unsigned long) &__bss_start;
408 	bss_resource.end = (unsigned long) &__bss_stop - 1;
409 
410 	for_each_memblock(memory, reg) {
411 		res = alloc_bootmem_low(sizeof(*res));
412 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
413 
414 		res->name = "System RAM";
415 		res->start = reg->base;
416 		res->end = reg->base + reg->size - 1;
417 		request_resource(&iomem_resource, res);
418 
419 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
420 			std_res = standard_resources[j];
421 			if (std_res->start < res->start ||
422 			    std_res->start > res->end)
423 				continue;
424 			if (std_res->end > res->end) {
425 				sub_res = alloc_bootmem_low(sizeof(*sub_res));
426 				*sub_res = *std_res;
427 				sub_res->end = res->end;
428 				std_res->start = res->end + 1;
429 				request_resource(res, sub_res);
430 			} else {
431 				request_resource(res, std_res);
432 			}
433 		}
434 	}
435 }
436 
437 static void __init setup_memory_end(void)
438 {
439 	unsigned long vmax, vmalloc_size, tmp;
440 
441 	/* Choose kernel address space layout: 2, 3, or 4 levels. */
442 	vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
443 	tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
444 	tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
445 	if (tmp + vmalloc_size + MODULES_LEN <= (1UL << 42))
446 		vmax = 1UL << 42;	/* 3-level kernel page table */
447 	else
448 		vmax = 1UL << 53;	/* 4-level kernel page table */
449 	/* module area is at the end of the kernel address space. */
450 	MODULES_END = vmax;
451 	MODULES_VADDR = MODULES_END - MODULES_LEN;
452 	VMALLOC_END = MODULES_VADDR;
453 	VMALLOC_START = vmax - vmalloc_size;
454 
455 	/* Split remaining virtual space between 1:1 mapping & vmemmap array */
456 	tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
457 	/* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
458 	tmp = SECTION_ALIGN_UP(tmp);
459 	tmp = VMALLOC_START - tmp * sizeof(struct page);
460 	tmp &= ~((vmax >> 11) - 1);	/* align to page table level */
461 	tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
462 	vmemmap = (struct page *) tmp;
463 
464 	/* Take care that memory_end is set and <= vmemmap */
465 	memory_end = min(memory_end ?: max_physmem_end, tmp);
466 	max_pfn = max_low_pfn = PFN_DOWN(memory_end);
467 	memblock_remove(memory_end, ULONG_MAX);
468 
469 	pr_notice("Max memory size: %luMB\n", memory_end >> 20);
470 }
471 
472 static void __init setup_vmcoreinfo(void)
473 {
474 	mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
475 }
476 
477 #ifdef CONFIG_CRASH_DUMP
478 
479 /*
480  * When kdump is enabled, we have to ensure that no memory from
481  * the area [0 - crashkernel memory size] and
482  * [crashk_res.start - crashk_res.end] is set offline.
483  */
484 static int kdump_mem_notifier(struct notifier_block *nb,
485 			      unsigned long action, void *data)
486 {
487 	struct memory_notify *arg = data;
488 
489 	if (action != MEM_GOING_OFFLINE)
490 		return NOTIFY_OK;
491 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
492 		return NOTIFY_BAD;
493 	if (arg->start_pfn > PFN_DOWN(crashk_res.end))
494 		return NOTIFY_OK;
495 	if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
496 		return NOTIFY_OK;
497 	return NOTIFY_BAD;
498 }
499 
500 static struct notifier_block kdump_mem_nb = {
501 	.notifier_call = kdump_mem_notifier,
502 };
503 
504 #endif
505 
506 /*
507  * Make sure that the area behind memory_end is protected
508  */
509 static void reserve_memory_end(void)
510 {
511 #ifdef CONFIG_CRASH_DUMP
512 	if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
513 	    !OLDMEM_BASE && sclp.hsa_size) {
514 		memory_end = sclp.hsa_size;
515 		memory_end &= PAGE_MASK;
516 		memory_end_set = 1;
517 	}
518 #endif
519 	if (!memory_end_set)
520 		return;
521 	memblock_reserve(memory_end, ULONG_MAX);
522 }
523 
524 /*
525  * Make sure that oldmem, where the dump is stored, is protected
526  */
527 static void reserve_oldmem(void)
528 {
529 #ifdef CONFIG_CRASH_DUMP
530 	if (OLDMEM_BASE)
531 		/* Forget all memory above the running kdump system */
532 		memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
533 #endif
534 }
535 
536 /*
537  * Make sure that oldmem, where the dump is stored, is protected
538  */
539 static void remove_oldmem(void)
540 {
541 #ifdef CONFIG_CRASH_DUMP
542 	if (OLDMEM_BASE)
543 		/* Forget all memory above the running kdump system */
544 		memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
545 #endif
546 }
547 
548 /*
549  * Reserve memory for kdump kernel to be loaded with kexec
550  */
551 static void __init reserve_crashkernel(void)
552 {
553 #ifdef CONFIG_CRASH_DUMP
554 	unsigned long long crash_base, crash_size;
555 	phys_addr_t low, high;
556 	int rc;
557 
558 	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
559 			       &crash_base);
560 
561 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
562 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
563 	if (rc || crash_size == 0)
564 		return;
565 
566 	if (memblock.memory.regions[0].size < crash_size) {
567 		pr_info("crashkernel reservation failed: %s\n",
568 			"first memory chunk must be at least crashkernel size");
569 		return;
570 	}
571 
572 	low = crash_base ?: OLDMEM_BASE;
573 	high = low + crash_size;
574 	if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
575 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
576 		crash_base = low;
577 	} else {
578 		/* Find suitable area in free memory */
579 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
580 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
581 
582 		if (crash_base && crash_base < low) {
583 			pr_info("crashkernel reservation failed: %s\n",
584 				"crash_base too low");
585 			return;
586 		}
587 		low = crash_base ?: low;
588 		crash_base = memblock_find_in_range(low, high, crash_size,
589 						    KEXEC_CRASH_MEM_ALIGN);
590 	}
591 
592 	if (!crash_base) {
593 		pr_info("crashkernel reservation failed: %s\n",
594 			"no suitable area found");
595 		return;
596 	}
597 
598 	if (register_memory_notifier(&kdump_mem_nb))
599 		return;
600 
601 	if (!OLDMEM_BASE && MACHINE_IS_VM)
602 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
603 	crashk_res.start = crash_base;
604 	crashk_res.end = crash_base + crash_size - 1;
605 	insert_resource(&iomem_resource, &crashk_res);
606 	memblock_remove(crash_base, crash_size);
607 	pr_info("Reserving %lluMB of memory at %lluMB "
608 		"for crashkernel (System RAM: %luMB)\n",
609 		crash_size >> 20, crash_base >> 20,
610 		(unsigned long)memblock.memory.total_size >> 20);
611 	os_info_crashkernel_add(crash_base, crash_size);
612 #endif
613 }
614 
615 /*
616  * Reserve the initrd from being used by memblock
617  */
618 static void __init reserve_initrd(void)
619 {
620 #ifdef CONFIG_BLK_DEV_INITRD
621 	initrd_start = INITRD_START;
622 	initrd_end = initrd_start + INITRD_SIZE;
623 	memblock_reserve(INITRD_START, INITRD_SIZE);
624 #endif
625 }
626 
627 /*
628  * Check for initrd being in usable memory
629  */
630 static void __init check_initrd(void)
631 {
632 #ifdef CONFIG_BLK_DEV_INITRD
633 	if (INITRD_START && INITRD_SIZE &&
634 	    !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
635 		pr_err("initrd does not fit memory.\n");
636 		memblock_free(INITRD_START, INITRD_SIZE);
637 		initrd_start = initrd_end = 0;
638 	}
639 #endif
640 }
641 
642 /*
643  * Reserve memory used for lowcore/command line/kernel image.
644  */
645 static void __init reserve_kernel(void)
646 {
647 	unsigned long start_pfn = PFN_UP(__pa(&_end));
648 
649 #ifdef CONFIG_DMA_API_DEBUG
650 	/*
651 	 * DMA_API_DEBUG code stumbles over addresses from the
652 	 * range [_ehead, _stext]. Mark the memory as reserved
653 	 * so it is not used for CONFIG_DMA_API_DEBUG=y.
654 	 */
655 	memblock_reserve(0, PFN_PHYS(start_pfn));
656 #else
657 	memblock_reserve(0, (unsigned long)_ehead);
658 	memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
659 			 - (unsigned long)_stext);
660 #endif
661 }
662 
663 static void __init setup_memory(void)
664 {
665 	struct memblock_region *reg;
666 
667 	/*
668 	 * Init storage key for present memory
669 	 */
670 	for_each_memblock(memory, reg) {
671 		storage_key_init_range(reg->base, reg->base + reg->size);
672 	}
673 	psw_set_key(PAGE_DEFAULT_KEY);
674 
675 	/* Only cosmetics */
676 	memblock_enforce_memory_limit(memblock_end_of_DRAM());
677 }
678 
679 /*
680  * Setup hardware capabilities.
681  */
682 static int __init setup_hwcaps(void)
683 {
684 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
685 	struct cpuid cpu_id;
686 	int i;
687 
688 	/*
689 	 * The store facility list bits numbers as found in the principles
690 	 * of operation are numbered with bit 1UL<<31 as number 0 to
691 	 * bit 1UL<<0 as number 31.
692 	 *   Bit 0: instructions named N3, "backported" to esa-mode
693 	 *   Bit 2: z/Architecture mode is active
694 	 *   Bit 7: the store-facility-list-extended facility is installed
695 	 *   Bit 17: the message-security assist is installed
696 	 *   Bit 19: the long-displacement facility is installed
697 	 *   Bit 21: the extended-immediate facility is installed
698 	 *   Bit 22: extended-translation facility 3 is installed
699 	 *   Bit 30: extended-translation facility 3 enhancement facility
700 	 * These get translated to:
701 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
702 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
703 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
704 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
705 	 */
706 	for (i = 0; i < 6; i++)
707 		if (test_facility(stfl_bits[i]))
708 			elf_hwcap |= 1UL << i;
709 
710 	if (test_facility(22) && test_facility(30))
711 		elf_hwcap |= HWCAP_S390_ETF3EH;
712 
713 	/*
714 	 * Check for additional facilities with store-facility-list-extended.
715 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
716 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
717 	 * as stored by stfl, bits 32-xxx contain additional facilities.
718 	 * How many facility words are stored depends on the number of
719 	 * doublewords passed to the instruction. The additional facilities
720 	 * are:
721 	 *   Bit 42: decimal floating point facility is installed
722 	 *   Bit 44: perform floating point operation facility is installed
723 	 * translated to:
724 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
725 	 */
726 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
727 		elf_hwcap |= HWCAP_S390_DFP;
728 
729 	/*
730 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
731 	 */
732 	if (MACHINE_HAS_HPAGE)
733 		elf_hwcap |= HWCAP_S390_HPAGE;
734 
735 	/*
736 	 * 64-bit register support for 31-bit processes
737 	 * HWCAP_S390_HIGH_GPRS is bit 9.
738 	 */
739 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
740 
741 	/*
742 	 * Transactional execution support HWCAP_S390_TE is bit 10.
743 	 */
744 	if (test_facility(50) && test_facility(73))
745 		elf_hwcap |= HWCAP_S390_TE;
746 
747 	/*
748 	 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
749 	 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
750 	 * instead of facility bit 129.
751 	 */
752 	if (MACHINE_HAS_VX)
753 		elf_hwcap |= HWCAP_S390_VXRS;
754 	get_cpu_id(&cpu_id);
755 	add_device_randomness(&cpu_id, sizeof(cpu_id));
756 	switch (cpu_id.machine) {
757 	case 0x2064:
758 	case 0x2066:
759 	default:	/* Use "z900" as default for 64 bit kernels. */
760 		strcpy(elf_platform, "z900");
761 		break;
762 	case 0x2084:
763 	case 0x2086:
764 		strcpy(elf_platform, "z990");
765 		break;
766 	case 0x2094:
767 	case 0x2096:
768 		strcpy(elf_platform, "z9-109");
769 		break;
770 	case 0x2097:
771 	case 0x2098:
772 		strcpy(elf_platform, "z10");
773 		break;
774 	case 0x2817:
775 	case 0x2818:
776 		strcpy(elf_platform, "z196");
777 		break;
778 	case 0x2827:
779 	case 0x2828:
780 		strcpy(elf_platform, "zEC12");
781 		break;
782 	case 0x2964:
783 	case 0x2965:
784 		strcpy(elf_platform, "z13");
785 		break;
786 	}
787 
788 	/*
789 	 * Virtualization support HWCAP_INT_SIE is bit 0.
790 	 */
791 	if (sclp.has_sief2)
792 		int_hwcap |= HWCAP_INT_SIE;
793 
794 	return 0;
795 }
796 arch_initcall(setup_hwcaps);
797 
798 /*
799  * Add system information as device randomness
800  */
801 static void __init setup_randomness(void)
802 {
803 	struct sysinfo_3_2_2 *vmms;
804 
805 	vmms = (struct sysinfo_3_2_2 *) alloc_page(GFP_KERNEL);
806 	if (vmms && stsi(vmms, 3, 2, 2) == 0 && vmms->count)
807 		add_device_randomness(&vmms, vmms->count);
808 	free_page((unsigned long) vmms);
809 }
810 
811 /*
812  * Find the correct size for the task_struct. This depends on
813  * the size of the struct fpu at the end of the thread_struct
814  * which is embedded in the task_struct.
815  */
816 static void __init setup_task_size(void)
817 {
818 	int task_size = sizeof(struct task_struct);
819 
820 	if (!MACHINE_HAS_VX) {
821 		task_size -= sizeof(__vector128) * __NUM_VXRS;
822 		task_size += sizeof(freg_t) * __NUM_FPRS;
823 	}
824 	arch_task_struct_size = task_size;
825 }
826 
827 /*
828  * Setup function called from init/main.c just after the banner
829  * was printed.
830  */
831 
832 void __init setup_arch(char **cmdline_p)
833 {
834         /*
835          * print what head.S has found out about the machine
836          */
837 	if (MACHINE_IS_VM)
838 		pr_info("Linux is running as a z/VM "
839 			"guest operating system in 64-bit mode\n");
840 	else if (MACHINE_IS_KVM)
841 		pr_info("Linux is running under KVM in 64-bit mode\n");
842 	else if (MACHINE_IS_LPAR)
843 		pr_info("Linux is running natively in 64-bit mode\n");
844 
845 	/* Have one command line that is parsed and saved in /proc/cmdline */
846 	/* boot_command_line has been already set up in early.c */
847 	*cmdline_p = boot_command_line;
848 
849         ROOT_DEV = Root_RAM0;
850 
851 	/* Is init_mm really needed? */
852 	init_mm.start_code = PAGE_OFFSET;
853 	init_mm.end_code = (unsigned long) &_etext;
854 	init_mm.end_data = (unsigned long) &_edata;
855 	init_mm.brk = (unsigned long) &_end;
856 
857 	parse_early_param();
858 #ifdef CONFIG_CRASH_DUMP
859 	/* Deactivate elfcorehdr= kernel parameter */
860 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
861 #endif
862 
863 	os_info_init();
864 	setup_ipl();
865 	setup_task_size();
866 
867 	/* Do some memory reservations *before* memory is added to memblock */
868 	reserve_memory_end();
869 	reserve_oldmem();
870 	reserve_kernel();
871 	reserve_initrd();
872 	memblock_allow_resize();
873 
874 	/* Get information about *all* installed memory */
875 	detect_memory_memblock();
876 
877 	remove_oldmem();
878 
879 	/*
880 	 * Make sure all chunks are MAX_ORDER aligned so we don't need the
881 	 * extra checks that HOLES_IN_ZONE would require.
882 	 *
883 	 * Is this still required?
884 	 */
885 	memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
886 
887 	setup_memory_end();
888 	setup_memory();
889 
890 	check_initrd();
891 	reserve_crashkernel();
892 #ifdef CONFIG_CRASH_DUMP
893 	/*
894 	 * Be aware that smp_save_dump_cpus() triggers a system reset.
895 	 * Therefore CPU and device initialization should be done afterwards.
896 	 */
897 	smp_save_dump_cpus();
898 #endif
899 
900 	setup_resources();
901 	setup_vmcoreinfo();
902 	setup_lowcore();
903 	smp_fill_possible_mask();
904         cpu_init();
905 	numa_setup();
906 
907 	/*
908 	 * Create kernel page tables and switch to virtual addressing.
909 	 */
910         paging_init();
911 
912         /* Setup default console */
913 	conmode_default();
914 	set_preferred_console();
915 
916 	/* Setup zfcpdump support */
917 	setup_zfcpdump();
918 
919 	/* Add system specific data to the random pool */
920 	setup_randomness();
921 }
922