xref: /openbmc/linux/arch/s390/kernel/setup.c (revision 3d37ef41)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  S390 version
4  *    Copyright IBM Corp. 1999, 2012
5  *    Author(s): Hartmut Penner (hp@de.ibm.com),
6  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
7  *
8  *  Derived from "arch/i386/kernel/setup.c"
9  *    Copyright (C) 1995, Linus Torvalds
10  */
11 
12 /*
13  * This file handles the architecture-dependent parts of initialization
14  */
15 
16 #define KMSG_COMPONENT "setup"
17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
18 
19 #include <linux/errno.h>
20 #include <linux/export.h>
21 #include <linux/sched.h>
22 #include <linux/sched/task.h>
23 #include <linux/cpu.h>
24 #include <linux/kernel.h>
25 #include <linux/memblock.h>
26 #include <linux/mm.h>
27 #include <linux/stddef.h>
28 #include <linux/unistd.h>
29 #include <linux/ptrace.h>
30 #include <linux/random.h>
31 #include <linux/user.h>
32 #include <linux/tty.h>
33 #include <linux/ioport.h>
34 #include <linux/delay.h>
35 #include <linux/init.h>
36 #include <linux/initrd.h>
37 #include <linux/root_dev.h>
38 #include <linux/console.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/dma-map-ops.h>
41 #include <linux/device.h>
42 #include <linux/notifier.h>
43 #include <linux/pfn.h>
44 #include <linux/ctype.h>
45 #include <linux/reboot.h>
46 #include <linux/topology.h>
47 #include <linux/kexec.h>
48 #include <linux/crash_dump.h>
49 #include <linux/memory.h>
50 #include <linux/compat.h>
51 #include <linux/start_kernel.h>
52 #include <linux/hugetlb.h>
53 
54 #include <asm/boot_data.h>
55 #include <asm/ipl.h>
56 #include <asm/facility.h>
57 #include <asm/smp.h>
58 #include <asm/mmu_context.h>
59 #include <asm/cpcmd.h>
60 #include <asm/lowcore.h>
61 #include <asm/nmi.h>
62 #include <asm/irq.h>
63 #include <asm/page.h>
64 #include <asm/ptrace.h>
65 #include <asm/sections.h>
66 #include <asm/ebcdic.h>
67 #include <asm/diag.h>
68 #include <asm/os_info.h>
69 #include <asm/sclp.h>
70 #include <asm/stacktrace.h>
71 #include <asm/sysinfo.h>
72 #include <asm/numa.h>
73 #include <asm/alternative.h>
74 #include <asm/nospec-branch.h>
75 #include <asm/mem_detect.h>
76 #include <asm/uv.h>
77 #include <asm/asm-offsets.h>
78 #include "entry.h"
79 
80 /*
81  * Machine setup..
82  */
83 unsigned int console_mode = 0;
84 EXPORT_SYMBOL(console_mode);
85 
86 unsigned int console_devno = -1;
87 EXPORT_SYMBOL(console_devno);
88 
89 unsigned int console_irq = -1;
90 EXPORT_SYMBOL(console_irq);
91 
92 unsigned long elf_hwcap __read_mostly = 0;
93 char elf_platform[ELF_PLATFORM_SIZE];
94 
95 unsigned long int_hwcap = 0;
96 
97 int __bootdata(noexec_disabled);
98 unsigned long __bootdata(ident_map_size);
99 unsigned long __bootdata(vmalloc_size);
100 struct mem_detect_info __bootdata(mem_detect);
101 
102 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
103 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
104 unsigned long __bootdata_preserved(__stext_dma);
105 unsigned long __bootdata_preserved(__etext_dma);
106 unsigned long __bootdata_preserved(__sdma);
107 unsigned long __bootdata_preserved(__edma);
108 unsigned long __bootdata_preserved(__kaslr_offset);
109 unsigned int __bootdata_preserved(zlib_dfltcc_support);
110 EXPORT_SYMBOL(zlib_dfltcc_support);
111 
112 unsigned long VMALLOC_START;
113 EXPORT_SYMBOL(VMALLOC_START);
114 
115 unsigned long VMALLOC_END;
116 EXPORT_SYMBOL(VMALLOC_END);
117 
118 struct page *vmemmap;
119 EXPORT_SYMBOL(vmemmap);
120 unsigned long vmemmap_size;
121 
122 unsigned long MODULES_VADDR;
123 unsigned long MODULES_END;
124 
125 /* An array with a pointer to the lowcore of every CPU. */
126 struct lowcore *lowcore_ptr[NR_CPUS];
127 EXPORT_SYMBOL(lowcore_ptr);
128 
129 /*
130  * The Write Back bit position in the physaddr is given by the SLPC PCI.
131  * Leaving the mask zero always uses write through which is safe
132  */
133 unsigned long mio_wb_bit_mask __ro_after_init;
134 
135 /*
136  * This is set up by the setup-routine at boot-time
137  * for S390 need to find out, what we have to setup
138  * using address 0x10400 ...
139  */
140 
141 #include <asm/setup.h>
142 
143 /*
144  * condev= and conmode= setup parameter.
145  */
146 
147 static int __init condev_setup(char *str)
148 {
149 	int vdev;
150 
151 	vdev = simple_strtoul(str, &str, 0);
152 	if (vdev >= 0 && vdev < 65536) {
153 		console_devno = vdev;
154 		console_irq = -1;
155 	}
156 	return 1;
157 }
158 
159 __setup("condev=", condev_setup);
160 
161 static void __init set_preferred_console(void)
162 {
163 	if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
164 		add_preferred_console("ttyS", 0, NULL);
165 	else if (CONSOLE_IS_3270)
166 		add_preferred_console("tty3270", 0, NULL);
167 	else if (CONSOLE_IS_VT220)
168 		add_preferred_console("ttyS", 1, NULL);
169 	else if (CONSOLE_IS_HVC)
170 		add_preferred_console("hvc", 0, NULL);
171 }
172 
173 static int __init conmode_setup(char *str)
174 {
175 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
176 	if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
177                 SET_CONSOLE_SCLP;
178 #endif
179 #if defined(CONFIG_TN3215_CONSOLE)
180 	if (!strcmp(str, "3215"))
181 		SET_CONSOLE_3215;
182 #endif
183 #if defined(CONFIG_TN3270_CONSOLE)
184 	if (!strcmp(str, "3270"))
185 		SET_CONSOLE_3270;
186 #endif
187 	set_preferred_console();
188         return 1;
189 }
190 
191 __setup("conmode=", conmode_setup);
192 
193 static void __init conmode_default(void)
194 {
195 	char query_buffer[1024];
196 	char *ptr;
197 
198         if (MACHINE_IS_VM) {
199 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
200 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
201 		ptr = strstr(query_buffer, "SUBCHANNEL =");
202 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
203 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
204 		ptr = strstr(query_buffer, "CONMODE");
205 		/*
206 		 * Set the conmode to 3215 so that the device recognition
207 		 * will set the cu_type of the console to 3215. If the
208 		 * conmode is 3270 and we don't set it back then both
209 		 * 3215 and the 3270 driver will try to access the console
210 		 * device (3215 as console and 3270 as normal tty).
211 		 */
212 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
213 		if (ptr == NULL) {
214 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
215 			SET_CONSOLE_SCLP;
216 #endif
217 			return;
218 		}
219 		if (str_has_prefix(ptr + 8, "3270")) {
220 #if defined(CONFIG_TN3270_CONSOLE)
221 			SET_CONSOLE_3270;
222 #elif defined(CONFIG_TN3215_CONSOLE)
223 			SET_CONSOLE_3215;
224 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
225 			SET_CONSOLE_SCLP;
226 #endif
227 		} else if (str_has_prefix(ptr + 8, "3215")) {
228 #if defined(CONFIG_TN3215_CONSOLE)
229 			SET_CONSOLE_3215;
230 #elif defined(CONFIG_TN3270_CONSOLE)
231 			SET_CONSOLE_3270;
232 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
233 			SET_CONSOLE_SCLP;
234 #endif
235 		}
236 	} else if (MACHINE_IS_KVM) {
237 		if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
238 			SET_CONSOLE_VT220;
239 		else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
240 			SET_CONSOLE_SCLP;
241 		else
242 			SET_CONSOLE_HVC;
243 	} else {
244 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
245 		SET_CONSOLE_SCLP;
246 #endif
247 	}
248 }
249 
250 #ifdef CONFIG_CRASH_DUMP
251 static void __init setup_zfcpdump(void)
252 {
253 	if (!is_ipl_type_dump())
254 		return;
255 	if (OLDMEM_BASE)
256 		return;
257 	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
258 	console_loglevel = 2;
259 }
260 #else
261 static inline void setup_zfcpdump(void) {}
262 #endif /* CONFIG_CRASH_DUMP */
263 
264  /*
265  * Reboot, halt and power_off stubs. They just call _machine_restart,
266  * _machine_halt or _machine_power_off.
267  */
268 
269 void machine_restart(char *command)
270 {
271 	if ((!in_interrupt() && !in_atomic()) || 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_restart(command);
278 }
279 
280 void machine_halt(void)
281 {
282 	if (!in_interrupt() || oops_in_progress)
283 		/*
284 		 * Only unblank the console if we are called in enabled
285 		 * context or a bust_spinlocks cleared the way for us.
286 		 */
287 		console_unblank();
288 	_machine_halt();
289 }
290 
291 void machine_power_off(void)
292 {
293 	if (!in_interrupt() || oops_in_progress)
294 		/*
295 		 * Only unblank the console if we are called in enabled
296 		 * context or a bust_spinlocks cleared the way for us.
297 		 */
298 		console_unblank();
299 	_machine_power_off();
300 }
301 
302 /*
303  * Dummy power off function.
304  */
305 void (*pm_power_off)(void) = machine_power_off;
306 EXPORT_SYMBOL_GPL(pm_power_off);
307 
308 void *restart_stack;
309 
310 unsigned long stack_alloc(void)
311 {
312 #ifdef CONFIG_VMAP_STACK
313 	return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE,
314 			THREADINFO_GFP, NUMA_NO_NODE,
315 			__builtin_return_address(0));
316 #else
317 	return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
318 #endif
319 }
320 
321 void stack_free(unsigned long stack)
322 {
323 #ifdef CONFIG_VMAP_STACK
324 	vfree((void *) stack);
325 #else
326 	free_pages(stack, THREAD_SIZE_ORDER);
327 #endif
328 }
329 
330 int __init arch_early_irq_init(void)
331 {
332 	unsigned long stack;
333 
334 	stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
335 	if (!stack)
336 		panic("Couldn't allocate async stack");
337 	S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
338 	return 0;
339 }
340 
341 static int __init stack_realloc(void)
342 {
343 	unsigned long old, new;
344 
345 	old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
346 	new = stack_alloc();
347 	if (!new)
348 		panic("Couldn't allocate async stack");
349 	WRITE_ONCE(S390_lowcore.async_stack, new + STACK_INIT_OFFSET);
350 	free_pages(old, THREAD_SIZE_ORDER);
351 
352 	old = S390_lowcore.mcck_stack - STACK_INIT_OFFSET;
353 	new = stack_alloc();
354 	if (!new)
355 		panic("Couldn't allocate machine check stack");
356 	WRITE_ONCE(S390_lowcore.mcck_stack, new + STACK_INIT_OFFSET);
357 	memblock_free_late(old, THREAD_SIZE);
358 	return 0;
359 }
360 early_initcall(stack_realloc);
361 
362 void __init arch_call_rest_init(void)
363 {
364 	unsigned long stack;
365 
366 	stack = stack_alloc();
367 	if (!stack)
368 		panic("Couldn't allocate kernel stack");
369 	current->stack = (void *) stack;
370 #ifdef CONFIG_VMAP_STACK
371 	current->stack_vm_area = (void *) stack;
372 #endif
373 	set_task_stack_end_magic(current);
374 	stack += STACK_INIT_OFFSET;
375 	S390_lowcore.kernel_stack = stack;
376 	CALL_ON_STACK_NORETURN(rest_init, stack);
377 }
378 
379 static void __init setup_lowcore_dat_off(void)
380 {
381 	unsigned long int_psw_mask = PSW_KERNEL_BITS;
382 	unsigned long mcck_stack;
383 	struct lowcore *lc;
384 
385 	if (IS_ENABLED(CONFIG_KASAN))
386 		int_psw_mask |= PSW_MASK_DAT;
387 
388 	/*
389 	 * Setup lowcore for boot cpu
390 	 */
391 	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
392 	lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
393 	if (!lc)
394 		panic("%s: Failed to allocate %zu bytes align=%zx\n",
395 		      __func__, sizeof(*lc), sizeof(*lc));
396 
397 	lc->restart_psw.mask = PSW_KERNEL_BITS;
398 	lc->restart_psw.addr = (unsigned long) restart_int_handler;
399 	lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
400 	lc->external_new_psw.addr = (unsigned long) ext_int_handler;
401 	lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
402 	lc->svc_new_psw.addr = (unsigned long) system_call;
403 	lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
404 	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
405 	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
406 	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
407 	lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
408 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
409 	lc->clock_comparator = clock_comparator_max;
410 	lc->nodat_stack = ((unsigned long) &init_thread_union)
411 		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
412 	lc->current_task = (unsigned long)&init_task;
413 	lc->lpp = LPP_MAGIC;
414 	lc->machine_flags = S390_lowcore.machine_flags;
415 	lc->preempt_count = S390_lowcore.preempt_count;
416 	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
417 	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
418 	       sizeof(lc->stfle_fac_list));
419 	memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
420 	       sizeof(lc->alt_stfle_fac_list));
421 	nmi_alloc_boot_cpu(lc);
422 	lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
423 	lc->exit_timer = S390_lowcore.exit_timer;
424 	lc->user_timer = S390_lowcore.user_timer;
425 	lc->system_timer = S390_lowcore.system_timer;
426 	lc->steal_timer = S390_lowcore.steal_timer;
427 	lc->last_update_timer = S390_lowcore.last_update_timer;
428 	lc->last_update_clock = S390_lowcore.last_update_clock;
429 
430 	/*
431 	 * Allocate the global restart stack which is the same for
432 	 * all CPUs in cast *one* of them does a PSW restart.
433 	 */
434 	restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
435 	if (!restart_stack)
436 		panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
437 		      __func__, THREAD_SIZE, THREAD_SIZE);
438 	restart_stack += STACK_INIT_OFFSET;
439 
440 	/*
441 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
442 	 * restart data to the absolute zero lowcore. This is necessary if
443 	 * PSW restart is done on an offline CPU that has lowcore zero.
444 	 */
445 	lc->restart_stack = (unsigned long) restart_stack;
446 	lc->restart_fn = (unsigned long) do_restart;
447 	lc->restart_data = 0;
448 	lc->restart_source = -1UL;
449 
450 	mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
451 	if (!mcck_stack)
452 		panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
453 		      __func__, THREAD_SIZE, THREAD_SIZE);
454 	lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
455 
456 	/* Setup absolute zero lowcore */
457 	mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
458 	mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
459 	mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
460 	mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
461 	mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
462 
463 	lc->spinlock_lockval = arch_spin_lockval(0);
464 	lc->spinlock_index = 0;
465 	arch_spin_lock_setup(0);
466 	lc->br_r1_trampoline = 0x07f1;	/* br %r1 */
467 	lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
468 	lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
469 
470 	set_prefix((u32)(unsigned long) lc);
471 	lowcore_ptr[0] = lc;
472 }
473 
474 static void __init setup_lowcore_dat_on(void)
475 {
476 	__ctl_clear_bit(0, 28);
477 	S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
478 	S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
479 	S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
480 	S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
481 	__ctl_set_bit(0, 28);
482 }
483 
484 static struct resource code_resource = {
485 	.name  = "Kernel code",
486 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
487 };
488 
489 static struct resource data_resource = {
490 	.name = "Kernel data",
491 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
492 };
493 
494 static struct resource bss_resource = {
495 	.name = "Kernel bss",
496 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
497 };
498 
499 static struct resource __initdata *standard_resources[] = {
500 	&code_resource,
501 	&data_resource,
502 	&bss_resource,
503 };
504 
505 static void __init setup_resources(void)
506 {
507 	struct resource *res, *std_res, *sub_res;
508 	phys_addr_t start, end;
509 	int j;
510 	u64 i;
511 
512 	code_resource.start = (unsigned long) _text;
513 	code_resource.end = (unsigned long) _etext - 1;
514 	data_resource.start = (unsigned long) _etext;
515 	data_resource.end = (unsigned long) _edata - 1;
516 	bss_resource.start = (unsigned long) __bss_start;
517 	bss_resource.end = (unsigned long) __bss_stop - 1;
518 
519 	for_each_mem_range(i, &start, &end) {
520 		res = memblock_alloc(sizeof(*res), 8);
521 		if (!res)
522 			panic("%s: Failed to allocate %zu bytes align=0x%x\n",
523 			      __func__, sizeof(*res), 8);
524 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
525 
526 		res->name = "System RAM";
527 		res->start = start;
528 		/*
529 		 * In memblock, end points to the first byte after the
530 		 * range while in resourses, end points to the last byte in
531 		 * the range.
532 		 */
533 		res->end = end - 1;
534 		request_resource(&iomem_resource, res);
535 
536 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
537 			std_res = standard_resources[j];
538 			if (std_res->start < res->start ||
539 			    std_res->start > res->end)
540 				continue;
541 			if (std_res->end > res->end) {
542 				sub_res = memblock_alloc(sizeof(*sub_res), 8);
543 				if (!sub_res)
544 					panic("%s: Failed to allocate %zu bytes align=0x%x\n",
545 					      __func__, sizeof(*sub_res), 8);
546 				*sub_res = *std_res;
547 				sub_res->end = res->end;
548 				std_res->start = res->end + 1;
549 				request_resource(res, sub_res);
550 			} else {
551 				request_resource(res, std_res);
552 			}
553 		}
554 	}
555 #ifdef CONFIG_CRASH_DUMP
556 	/*
557 	 * Re-add removed crash kernel memory as reserved memory. This makes
558 	 * sure it will be mapped with the identity mapping and struct pages
559 	 * will be created, so it can be resized later on.
560 	 * However add it later since the crash kernel resource should not be
561 	 * part of the System RAM resource.
562 	 */
563 	if (crashk_res.end) {
564 		memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
565 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
566 		insert_resource(&iomem_resource, &crashk_res);
567 	}
568 #endif
569 }
570 
571 static void __init setup_ident_map_size(void)
572 {
573 	unsigned long vmax, tmp;
574 
575 	/* Choose kernel address space layout: 3 or 4 levels. */
576 	tmp = ident_map_size / PAGE_SIZE;
577 	tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
578 	if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
579 		vmax = _REGION2_SIZE; /* 3-level kernel page table */
580 	else
581 		vmax = _REGION1_SIZE; /* 4-level kernel page table */
582 	/* module area is at the end of the kernel address space. */
583 	MODULES_END = vmax;
584 	if (is_prot_virt_host())
585 		adjust_to_uv_max(&MODULES_END);
586 #ifdef CONFIG_KASAN
587 	vmax = _REGION1_SIZE;
588 	MODULES_END = kasan_vmax;
589 #endif
590 	MODULES_VADDR = MODULES_END - MODULES_LEN;
591 	VMALLOC_END = MODULES_VADDR;
592 	VMALLOC_START = VMALLOC_END - vmalloc_size;
593 
594 	/* Split remaining virtual space between 1:1 mapping & vmemmap array */
595 	tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
596 	/* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
597 	tmp = SECTION_ALIGN_UP(tmp);
598 	tmp = VMALLOC_START - tmp * sizeof(struct page);
599 	tmp &= ~((vmax >> 11) - 1);	/* align to page table level */
600 	tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
601 	vmemmap = (struct page *) tmp;
602 
603 	/* Take care that ident_map_size <= vmemmap */
604 	ident_map_size = min(ident_map_size, (unsigned long)vmemmap);
605 #ifdef CONFIG_KASAN
606 	ident_map_size = min(ident_map_size, KASAN_SHADOW_START);
607 #endif
608 	vmemmap_size = SECTION_ALIGN_UP(ident_map_size / PAGE_SIZE) * sizeof(struct page);
609 #ifdef CONFIG_KASAN
610 	/* move vmemmap above kasan shadow only if stands in a way */
611 	if (KASAN_SHADOW_END > (unsigned long)vmemmap &&
612 	    (unsigned long)vmemmap + vmemmap_size > KASAN_SHADOW_START)
613 		vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
614 #endif
615 	max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
616 	memblock_remove(ident_map_size, ULONG_MAX);
617 
618 	pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
619 }
620 
621 #ifdef CONFIG_CRASH_DUMP
622 
623 /*
624  * When kdump is enabled, we have to ensure that no memory from the area
625  * [0 - crashkernel memory size] is set offline - it will be exchanged with
626  * the crashkernel memory region when kdump is triggered. The crashkernel
627  * memory region can never get offlined (pages are unmovable).
628  */
629 static int kdump_mem_notifier(struct notifier_block *nb,
630 			      unsigned long action, void *data)
631 {
632 	struct memory_notify *arg = data;
633 
634 	if (action != MEM_GOING_OFFLINE)
635 		return NOTIFY_OK;
636 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
637 		return NOTIFY_BAD;
638 	return NOTIFY_OK;
639 }
640 
641 static struct notifier_block kdump_mem_nb = {
642 	.notifier_call = kdump_mem_notifier,
643 };
644 
645 #endif
646 
647 /*
648  * Make sure that the area above identity mapping is protected
649  */
650 static void __init reserve_above_ident_map(void)
651 {
652 	memblock_reserve(ident_map_size, ULONG_MAX);
653 }
654 
655 /*
656  * Make sure that oldmem, where the dump is stored, is protected
657  */
658 static void __init reserve_oldmem(void)
659 {
660 #ifdef CONFIG_CRASH_DUMP
661 	if (OLDMEM_BASE)
662 		/* Forget all memory above the running kdump system */
663 		memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
664 #endif
665 }
666 
667 /*
668  * Make sure that oldmem, where the dump is stored, is protected
669  */
670 static void __init remove_oldmem(void)
671 {
672 #ifdef CONFIG_CRASH_DUMP
673 	if (OLDMEM_BASE)
674 		/* Forget all memory above the running kdump system */
675 		memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
676 #endif
677 }
678 
679 /*
680  * Reserve memory for kdump kernel to be loaded with kexec
681  */
682 static void __init reserve_crashkernel(void)
683 {
684 #ifdef CONFIG_CRASH_DUMP
685 	unsigned long long crash_base, crash_size;
686 	phys_addr_t low, high;
687 	int rc;
688 
689 	rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
690 			       &crash_base);
691 
692 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
693 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
694 	if (rc || crash_size == 0)
695 		return;
696 
697 	if (memblock.memory.regions[0].size < crash_size) {
698 		pr_info("crashkernel reservation failed: %s\n",
699 			"first memory chunk must be at least crashkernel size");
700 		return;
701 	}
702 
703 	low = crash_base ?: OLDMEM_BASE;
704 	high = low + crash_size;
705 	if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
706 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
707 		crash_base = low;
708 	} else {
709 		/* Find suitable area in free memory */
710 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
711 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
712 
713 		if (crash_base && crash_base < low) {
714 			pr_info("crashkernel reservation failed: %s\n",
715 				"crash_base too low");
716 			return;
717 		}
718 		low = crash_base ?: low;
719 		crash_base = memblock_find_in_range(low, high, crash_size,
720 						    KEXEC_CRASH_MEM_ALIGN);
721 	}
722 
723 	if (!crash_base) {
724 		pr_info("crashkernel reservation failed: %s\n",
725 			"no suitable area found");
726 		return;
727 	}
728 
729 	if (register_memory_notifier(&kdump_mem_nb))
730 		return;
731 
732 	if (!OLDMEM_BASE && MACHINE_IS_VM)
733 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
734 	crashk_res.start = crash_base;
735 	crashk_res.end = crash_base + crash_size - 1;
736 	memblock_remove(crash_base, crash_size);
737 	pr_info("Reserving %lluMB of memory at %lluMB "
738 		"for crashkernel (System RAM: %luMB)\n",
739 		crash_size >> 20, crash_base >> 20,
740 		(unsigned long)memblock.memory.total_size >> 20);
741 	os_info_crashkernel_add(crash_base, crash_size);
742 #endif
743 }
744 
745 /*
746  * Reserve the initrd from being used by memblock
747  */
748 static void __init reserve_initrd(void)
749 {
750 #ifdef CONFIG_BLK_DEV_INITRD
751 	if (!INITRD_START || !INITRD_SIZE)
752 		return;
753 	initrd_start = INITRD_START;
754 	initrd_end = initrd_start + INITRD_SIZE;
755 	memblock_reserve(INITRD_START, INITRD_SIZE);
756 #endif
757 }
758 
759 /*
760  * Reserve the memory area used to pass the certificate lists
761  */
762 static void __init reserve_certificate_list(void)
763 {
764 	if (ipl_cert_list_addr)
765 		memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
766 }
767 
768 static void __init reserve_mem_detect_info(void)
769 {
770 	unsigned long start, size;
771 
772 	get_mem_detect_reserved(&start, &size);
773 	if (size)
774 		memblock_reserve(start, size);
775 }
776 
777 static void __init free_mem_detect_info(void)
778 {
779 	unsigned long start, size;
780 
781 	get_mem_detect_reserved(&start, &size);
782 	if (size)
783 		memblock_free(start, size);
784 }
785 
786 static const char * __init get_mem_info_source(void)
787 {
788 	switch (mem_detect.info_source) {
789 	case MEM_DETECT_SCLP_STOR_INFO:
790 		return "sclp storage info";
791 	case MEM_DETECT_DIAG260:
792 		return "diag260";
793 	case MEM_DETECT_SCLP_READ_INFO:
794 		return "sclp read info";
795 	case MEM_DETECT_BIN_SEARCH:
796 		return "binary search";
797 	}
798 	return "none";
799 }
800 
801 static void __init memblock_add_mem_detect_info(void)
802 {
803 	unsigned long start, end;
804 	int i;
805 
806 	pr_debug("physmem info source: %s (%hhd)\n",
807 		 get_mem_info_source(), mem_detect.info_source);
808 	/* keep memblock lists close to the kernel */
809 	memblock_set_bottom_up(true);
810 	for_each_mem_detect_block(i, &start, &end) {
811 		memblock_add(start, end - start);
812 		memblock_physmem_add(start, end - start);
813 	}
814 	memblock_set_bottom_up(false);
815 	memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
816 	memblock_dump_all();
817 }
818 
819 /*
820  * Check for initrd being in usable memory
821  */
822 static void __init check_initrd(void)
823 {
824 #ifdef CONFIG_BLK_DEV_INITRD
825 	if (INITRD_START && INITRD_SIZE &&
826 	    !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
827 		pr_err("The initial RAM disk does not fit into the memory\n");
828 		memblock_free(INITRD_START, INITRD_SIZE);
829 		initrd_start = initrd_end = 0;
830 	}
831 #endif
832 }
833 
834 /*
835  * Reserve memory used for lowcore/command line/kernel image.
836  */
837 static void __init reserve_kernel(void)
838 {
839 	unsigned long start_pfn = PFN_UP(__pa(_end));
840 
841 	memblock_reserve(0, HEAD_END);
842 	memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
843 			 - (unsigned long)_stext);
844 	memblock_reserve(__sdma, __edma - __sdma);
845 }
846 
847 static void __init setup_memory(void)
848 {
849 	phys_addr_t start, end;
850 	u64 i;
851 
852 	/*
853 	 * Init storage key for present memory
854 	 */
855 	for_each_mem_range(i, &start, &end)
856 		storage_key_init_range(start, end);
857 
858 	psw_set_key(PAGE_DEFAULT_KEY);
859 
860 	/* Only cosmetics */
861 	memblock_enforce_memory_limit(memblock_end_of_DRAM());
862 }
863 
864 /*
865  * Setup hardware capabilities.
866  */
867 static int __init setup_hwcaps(void)
868 {
869 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
870 	struct cpuid cpu_id;
871 	int i;
872 
873 	/*
874 	 * The store facility list bits numbers as found in the principles
875 	 * of operation are numbered with bit 1UL<<31 as number 0 to
876 	 * bit 1UL<<0 as number 31.
877 	 *   Bit 0: instructions named N3, "backported" to esa-mode
878 	 *   Bit 2: z/Architecture mode is active
879 	 *   Bit 7: the store-facility-list-extended facility is installed
880 	 *   Bit 17: the message-security assist is installed
881 	 *   Bit 19: the long-displacement facility is installed
882 	 *   Bit 21: the extended-immediate facility is installed
883 	 *   Bit 22: extended-translation facility 3 is installed
884 	 *   Bit 30: extended-translation facility 3 enhancement facility
885 	 * These get translated to:
886 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
887 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
888 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
889 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
890 	 */
891 	for (i = 0; i < 6; i++)
892 		if (test_facility(stfl_bits[i]))
893 			elf_hwcap |= 1UL << i;
894 
895 	if (test_facility(22) && test_facility(30))
896 		elf_hwcap |= HWCAP_S390_ETF3EH;
897 
898 	/*
899 	 * Check for additional facilities with store-facility-list-extended.
900 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
901 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
902 	 * as stored by stfl, bits 32-xxx contain additional facilities.
903 	 * How many facility words are stored depends on the number of
904 	 * doublewords passed to the instruction. The additional facilities
905 	 * are:
906 	 *   Bit 42: decimal floating point facility is installed
907 	 *   Bit 44: perform floating point operation facility is installed
908 	 * translated to:
909 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
910 	 */
911 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
912 		elf_hwcap |= HWCAP_S390_DFP;
913 
914 	/*
915 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
916 	 */
917 	if (MACHINE_HAS_EDAT1)
918 		elf_hwcap |= HWCAP_S390_HPAGE;
919 
920 	/*
921 	 * 64-bit register support for 31-bit processes
922 	 * HWCAP_S390_HIGH_GPRS is bit 9.
923 	 */
924 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
925 
926 	/*
927 	 * Transactional execution support HWCAP_S390_TE is bit 10.
928 	 */
929 	if (MACHINE_HAS_TE)
930 		elf_hwcap |= HWCAP_S390_TE;
931 
932 	/*
933 	 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
934 	 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
935 	 * instead of facility bit 129.
936 	 */
937 	if (MACHINE_HAS_VX) {
938 		elf_hwcap |= HWCAP_S390_VXRS;
939 		if (test_facility(134))
940 			elf_hwcap |= HWCAP_S390_VXRS_BCD;
941 		if (test_facility(135))
942 			elf_hwcap |= HWCAP_S390_VXRS_EXT;
943 		if (test_facility(148))
944 			elf_hwcap |= HWCAP_S390_VXRS_EXT2;
945 		if (test_facility(152))
946 			elf_hwcap |= HWCAP_S390_VXRS_PDE;
947 	}
948 	if (test_facility(150))
949 		elf_hwcap |= HWCAP_S390_SORT;
950 	if (test_facility(151))
951 		elf_hwcap |= HWCAP_S390_DFLT;
952 
953 	/*
954 	 * Guarded storage support HWCAP_S390_GS is bit 12.
955 	 */
956 	if (MACHINE_HAS_GS)
957 		elf_hwcap |= HWCAP_S390_GS;
958 
959 	get_cpu_id(&cpu_id);
960 	add_device_randomness(&cpu_id, sizeof(cpu_id));
961 	switch (cpu_id.machine) {
962 	case 0x2064:
963 	case 0x2066:
964 	default:	/* Use "z900" as default for 64 bit kernels. */
965 		strcpy(elf_platform, "z900");
966 		break;
967 	case 0x2084:
968 	case 0x2086:
969 		strcpy(elf_platform, "z990");
970 		break;
971 	case 0x2094:
972 	case 0x2096:
973 		strcpy(elf_platform, "z9-109");
974 		break;
975 	case 0x2097:
976 	case 0x2098:
977 		strcpy(elf_platform, "z10");
978 		break;
979 	case 0x2817:
980 	case 0x2818:
981 		strcpy(elf_platform, "z196");
982 		break;
983 	case 0x2827:
984 	case 0x2828:
985 		strcpy(elf_platform, "zEC12");
986 		break;
987 	case 0x2964:
988 	case 0x2965:
989 		strcpy(elf_platform, "z13");
990 		break;
991 	case 0x3906:
992 	case 0x3907:
993 		strcpy(elf_platform, "z14");
994 		break;
995 	case 0x8561:
996 	case 0x8562:
997 		strcpy(elf_platform, "z15");
998 		break;
999 	}
1000 
1001 	/*
1002 	 * Virtualization support HWCAP_INT_SIE is bit 0.
1003 	 */
1004 	if (sclp.has_sief2)
1005 		int_hwcap |= HWCAP_INT_SIE;
1006 
1007 	return 0;
1008 }
1009 arch_initcall(setup_hwcaps);
1010 
1011 /*
1012  * Add system information as device randomness
1013  */
1014 static void __init setup_randomness(void)
1015 {
1016 	struct sysinfo_3_2_2 *vmms;
1017 
1018 	vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
1019 							    PAGE_SIZE);
1020 	if (!vmms)
1021 		panic("Failed to allocate memory for sysinfo structure\n");
1022 
1023 	if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1024 		add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1025 	memblock_free((unsigned long) vmms, PAGE_SIZE);
1026 }
1027 
1028 /*
1029  * Find the correct size for the task_struct. This depends on
1030  * the size of the struct fpu at the end of the thread_struct
1031  * which is embedded in the task_struct.
1032  */
1033 static void __init setup_task_size(void)
1034 {
1035 	int task_size = sizeof(struct task_struct);
1036 
1037 	if (!MACHINE_HAS_VX) {
1038 		task_size -= sizeof(__vector128) * __NUM_VXRS;
1039 		task_size += sizeof(freg_t) * __NUM_FPRS;
1040 	}
1041 	arch_task_struct_size = task_size;
1042 }
1043 
1044 /*
1045  * Issue diagnose 318 to set the control program name and
1046  * version codes.
1047  */
1048 static void __init setup_control_program_code(void)
1049 {
1050 	union diag318_info diag318_info = {
1051 		.cpnc = CPNC_LINUX,
1052 		.cpvc = 0,
1053 	};
1054 
1055 	if (!sclp.has_diag318)
1056 		return;
1057 
1058 	diag_stat_inc(DIAG_STAT_X318);
1059 	asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1060 }
1061 
1062 /*
1063  * Print the component list from the IPL report
1064  */
1065 static void __init log_component_list(void)
1066 {
1067 	struct ipl_rb_component_entry *ptr, *end;
1068 	char *str;
1069 
1070 	if (!early_ipl_comp_list_addr)
1071 		return;
1072 	if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1073 		pr_info("Linux is running with Secure-IPL enabled\n");
1074 	else
1075 		pr_info("Linux is running with Secure-IPL disabled\n");
1076 	ptr = (void *) early_ipl_comp_list_addr;
1077 	end = (void *) ptr + early_ipl_comp_list_size;
1078 	pr_info("The IPL report contains the following components:\n");
1079 	while (ptr < end) {
1080 		if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1081 			if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1082 				str = "signed, verified";
1083 			else
1084 				str = "signed, verification failed";
1085 		} else {
1086 			str = "not signed";
1087 		}
1088 		pr_info("%016llx - %016llx (%s)\n",
1089 			ptr->addr, ptr->addr + ptr->len, str);
1090 		ptr++;
1091 	}
1092 }
1093 
1094 /*
1095  * Setup function called from init/main.c just after the banner
1096  * was printed.
1097  */
1098 
1099 void __init setup_arch(char **cmdline_p)
1100 {
1101         /*
1102          * print what head.S has found out about the machine
1103          */
1104 	if (MACHINE_IS_VM)
1105 		pr_info("Linux is running as a z/VM "
1106 			"guest operating system in 64-bit mode\n");
1107 	else if (MACHINE_IS_KVM)
1108 		pr_info("Linux is running under KVM in 64-bit mode\n");
1109 	else if (MACHINE_IS_LPAR)
1110 		pr_info("Linux is running natively in 64-bit mode\n");
1111 	else
1112 		pr_info("Linux is running as a guest in 64-bit mode\n");
1113 
1114 	log_component_list();
1115 
1116 	/* Have one command line that is parsed and saved in /proc/cmdline */
1117 	/* boot_command_line has been already set up in early.c */
1118 	*cmdline_p = boot_command_line;
1119 
1120         ROOT_DEV = Root_RAM0;
1121 
1122 	init_mm.start_code = (unsigned long) _text;
1123 	init_mm.end_code = (unsigned long) _etext;
1124 	init_mm.end_data = (unsigned long) _edata;
1125 	init_mm.brk = (unsigned long) _end;
1126 
1127 	if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1128 		nospec_auto_detect();
1129 
1130 	jump_label_init();
1131 	parse_early_param();
1132 #ifdef CONFIG_CRASH_DUMP
1133 	/* Deactivate elfcorehdr= kernel parameter */
1134 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
1135 #endif
1136 
1137 	os_info_init();
1138 	setup_ipl();
1139 	setup_task_size();
1140 	setup_control_program_code();
1141 
1142 	/* Do some memory reservations *before* memory is added to memblock */
1143 	reserve_above_ident_map();
1144 	reserve_oldmem();
1145 	reserve_kernel();
1146 	reserve_initrd();
1147 	reserve_certificate_list();
1148 	reserve_mem_detect_info();
1149 	memblock_allow_resize();
1150 
1151 	/* Get information about *all* installed memory */
1152 	memblock_add_mem_detect_info();
1153 
1154 	free_mem_detect_info();
1155 	remove_oldmem();
1156 
1157 	setup_uv();
1158 	setup_ident_map_size();
1159 	setup_memory();
1160 	dma_contiguous_reserve(ident_map_size);
1161 	vmcp_cma_reserve();
1162 	if (MACHINE_HAS_EDAT2)
1163 		hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1164 
1165 	check_initrd();
1166 	reserve_crashkernel();
1167 #ifdef CONFIG_CRASH_DUMP
1168 	/*
1169 	 * Be aware that smp_save_dump_cpus() triggers a system reset.
1170 	 * Therefore CPU and device initialization should be done afterwards.
1171 	 */
1172 	smp_save_dump_cpus();
1173 #endif
1174 
1175 	setup_resources();
1176 	setup_lowcore_dat_off();
1177 	smp_fill_possible_mask();
1178 	cpu_detect_mhz_feature();
1179         cpu_init();
1180 	numa_setup();
1181 	smp_detect_cpus();
1182 	topology_init_early();
1183 
1184 	/*
1185 	 * Create kernel page tables and switch to virtual addressing.
1186 	 */
1187         paging_init();
1188 
1189 	/*
1190 	 * After paging_init created the kernel page table, the new PSWs
1191 	 * in lowcore can now run with DAT enabled.
1192 	 */
1193 	setup_lowcore_dat_on();
1194 
1195         /* Setup default console */
1196 	conmode_default();
1197 	set_preferred_console();
1198 
1199 	apply_alternative_instructions();
1200 	if (IS_ENABLED(CONFIG_EXPOLINE))
1201 		nospec_init_branches();
1202 
1203 	/* Setup zfcp/nvme dump support */
1204 	setup_zfcpdump();
1205 
1206 	/* Add system specific data to the random pool */
1207 	setup_randomness();
1208 }
1209