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 #include <linux/kmemleak.h>
54 
55 #include <asm/boot_data.h>
56 #include <asm/ipl.h>
57 #include <asm/facility.h>
58 #include <asm/smp.h>
59 #include <asm/mmu_context.h>
60 #include <asm/cpcmd.h>
61 #include <asm/lowcore.h>
62 #include <asm/nmi.h>
63 #include <asm/irq.h>
64 #include <asm/page.h>
65 #include <asm/ptrace.h>
66 #include <asm/sections.h>
67 #include <asm/ebcdic.h>
68 #include <asm/diag.h>
69 #include <asm/os_info.h>
70 #include <asm/sclp.h>
71 #include <asm/stacktrace.h>
72 #include <asm/sysinfo.h>
73 #include <asm/numa.h>
74 #include <asm/alternative.h>
75 #include <asm/nospec-branch.h>
76 #include <asm/mem_detect.h>
77 #include <asm/uv.h>
78 #include <asm/asm-offsets.h>
79 #include "entry.h"
80 
81 /*
82  * Machine setup..
83  */
84 unsigned int console_mode = 0;
85 EXPORT_SYMBOL(console_mode);
86 
87 unsigned int console_devno = -1;
88 EXPORT_SYMBOL(console_devno);
89 
90 unsigned int console_irq = -1;
91 EXPORT_SYMBOL(console_irq);
92 
93 /*
94  * Some code and data needs to stay below 2 GB, even when the kernel would be
95  * relocated above 2 GB, because it has to use 31 bit addresses.
96  * Such code and data is part of the .amode31 section.
97  */
98 unsigned long __amode31_ref __samode31 = (unsigned long)&_samode31;
99 unsigned long __amode31_ref __eamode31 = (unsigned long)&_eamode31;
100 unsigned long __amode31_ref __stext_amode31 = (unsigned long)&_stext_amode31;
101 unsigned long __amode31_ref __etext_amode31 = (unsigned long)&_etext_amode31;
102 struct exception_table_entry __amode31_ref *__start_amode31_ex_table = _start_amode31_ex_table;
103 struct exception_table_entry __amode31_ref *__stop_amode31_ex_table = _stop_amode31_ex_table;
104 
105 /*
106  * Control registers CR2, CR5 and CR15 are initialized with addresses
107  * of tables that must be placed below 2G which is handled by the AMODE31
108  * sections.
109  * Because the AMODE31 sections are relocated below 2G at startup,
110  * the content of control registers CR2, CR5 and CR15 must be updated
111  * with new addresses after the relocation. The initial initialization of
112  * control registers occurs in head64.S and then gets updated again after AMODE31
113  * relocation. We must access the relevant AMODE31 tables indirectly via
114  * pointers placed in the .amode31.refs linker section. Those pointers get
115  * updated automatically during AMODE31 relocation and always contain a valid
116  * address within AMODE31 sections.
117  */
118 
119 static __amode31_data u32 __ctl_duct_amode31[16] __aligned(64);
120 
121 static __amode31_data u64 __ctl_aste_amode31[8] __aligned(64) = {
122 	[1] = 0xffffffffffffffff
123 };
124 
125 static __amode31_data u32 __ctl_duald_amode31[32] __aligned(128) = {
126 	0x80000000, 0, 0, 0,
127 	0x80000000, 0, 0, 0,
128 	0x80000000, 0, 0, 0,
129 	0x80000000, 0, 0, 0,
130 	0x80000000, 0, 0, 0,
131 	0x80000000, 0, 0, 0,
132 	0x80000000, 0, 0, 0,
133 	0x80000000, 0, 0, 0
134 };
135 
136 static __amode31_data u32 __ctl_linkage_stack_amode31[8] __aligned(64) = {
137 	0, 0, 0x89000000, 0,
138 	0, 0, 0x8a000000, 0
139 };
140 
141 static u64 __amode31_ref *__ctl_aste = __ctl_aste_amode31;
142 static u32 __amode31_ref *__ctl_duald = __ctl_duald_amode31;
143 static u32 __amode31_ref *__ctl_linkage_stack = __ctl_linkage_stack_amode31;
144 static u32 __amode31_ref *__ctl_duct = __ctl_duct_amode31;
145 
146 int __bootdata(noexec_disabled);
147 unsigned long __bootdata(ident_map_size);
148 struct mem_detect_info __bootdata(mem_detect);
149 struct initrd_data __bootdata(initrd_data);
150 
151 unsigned long __bootdata_preserved(__kaslr_offset);
152 unsigned long __bootdata(__amode31_base);
153 unsigned int __bootdata_preserved(zlib_dfltcc_support);
154 EXPORT_SYMBOL(zlib_dfltcc_support);
155 u64 __bootdata_preserved(stfle_fac_list[16]);
156 EXPORT_SYMBOL(stfle_fac_list);
157 u64 __bootdata_preserved(alt_stfle_fac_list[16]);
158 struct oldmem_data __bootdata_preserved(oldmem_data);
159 
160 unsigned long VMALLOC_START;
161 EXPORT_SYMBOL(VMALLOC_START);
162 
163 unsigned long VMALLOC_END;
164 EXPORT_SYMBOL(VMALLOC_END);
165 
166 struct page *vmemmap;
167 EXPORT_SYMBOL(vmemmap);
168 unsigned long vmemmap_size;
169 
170 unsigned long MODULES_VADDR;
171 unsigned long MODULES_END;
172 
173 /* An array with a pointer to the lowcore of every CPU. */
174 struct lowcore *lowcore_ptr[NR_CPUS];
175 EXPORT_SYMBOL(lowcore_ptr);
176 
177 DEFINE_STATIC_KEY_FALSE(cpu_has_bear);
178 
179 /*
180  * The Write Back bit position in the physaddr is given by the SLPC PCI.
181  * Leaving the mask zero always uses write through which is safe
182  */
183 unsigned long mio_wb_bit_mask __ro_after_init;
184 
185 /*
186  * This is set up by the setup-routine at boot-time
187  * for S390 need to find out, what we have to setup
188  * using address 0x10400 ...
189  */
190 
191 #include <asm/setup.h>
192 
193 /*
194  * condev= and conmode= setup parameter.
195  */
196 
condev_setup(char * str)197 static int __init condev_setup(char *str)
198 {
199 	int vdev;
200 
201 	vdev = simple_strtoul(str, &str, 0);
202 	if (vdev >= 0 && vdev < 65536) {
203 		console_devno = vdev;
204 		console_irq = -1;
205 	}
206 	return 1;
207 }
208 
209 __setup("condev=", condev_setup);
210 
set_preferred_console(void)211 static void __init set_preferred_console(void)
212 {
213 	if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
214 		add_preferred_console("ttyS", 0, NULL);
215 	else if (CONSOLE_IS_3270)
216 		add_preferred_console("tty3270", 0, NULL);
217 	else if (CONSOLE_IS_VT220)
218 		add_preferred_console("ttysclp", 0, NULL);
219 	else if (CONSOLE_IS_HVC)
220 		add_preferred_console("hvc", 0, NULL);
221 }
222 
conmode_setup(char * str)223 static int __init conmode_setup(char *str)
224 {
225 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
226 	if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
227                 SET_CONSOLE_SCLP;
228 #endif
229 #if defined(CONFIG_TN3215_CONSOLE)
230 	if (!strcmp(str, "3215"))
231 		SET_CONSOLE_3215;
232 #endif
233 #if defined(CONFIG_TN3270_CONSOLE)
234 	if (!strcmp(str, "3270"))
235 		SET_CONSOLE_3270;
236 #endif
237 	set_preferred_console();
238         return 1;
239 }
240 
241 __setup("conmode=", conmode_setup);
242 
conmode_default(void)243 static void __init conmode_default(void)
244 {
245 	char query_buffer[1024];
246 	char *ptr;
247 
248         if (MACHINE_IS_VM) {
249 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
250 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
251 		ptr = strstr(query_buffer, "SUBCHANNEL =");
252 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
253 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
254 		ptr = strstr(query_buffer, "CONMODE");
255 		/*
256 		 * Set the conmode to 3215 so that the device recognition
257 		 * will set the cu_type of the console to 3215. If the
258 		 * conmode is 3270 and we don't set it back then both
259 		 * 3215 and the 3270 driver will try to access the console
260 		 * device (3215 as console and 3270 as normal tty).
261 		 */
262 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
263 		if (ptr == NULL) {
264 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
265 			SET_CONSOLE_SCLP;
266 #endif
267 			return;
268 		}
269 		if (str_has_prefix(ptr + 8, "3270")) {
270 #if defined(CONFIG_TN3270_CONSOLE)
271 			SET_CONSOLE_3270;
272 #elif defined(CONFIG_TN3215_CONSOLE)
273 			SET_CONSOLE_3215;
274 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
275 			SET_CONSOLE_SCLP;
276 #endif
277 		} else if (str_has_prefix(ptr + 8, "3215")) {
278 #if defined(CONFIG_TN3215_CONSOLE)
279 			SET_CONSOLE_3215;
280 #elif defined(CONFIG_TN3270_CONSOLE)
281 			SET_CONSOLE_3270;
282 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
283 			SET_CONSOLE_SCLP;
284 #endif
285 		}
286 	} else if (MACHINE_IS_KVM) {
287 		if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
288 			SET_CONSOLE_VT220;
289 		else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
290 			SET_CONSOLE_SCLP;
291 		else
292 			SET_CONSOLE_HVC;
293 	} else {
294 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
295 		SET_CONSOLE_SCLP;
296 #endif
297 	}
298 }
299 
300 #ifdef CONFIG_CRASH_DUMP
setup_zfcpdump(void)301 static void __init setup_zfcpdump(void)
302 {
303 	if (!is_ipl_type_dump())
304 		return;
305 	if (oldmem_data.start)
306 		return;
307 	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
308 	console_loglevel = 2;
309 }
310 #else
setup_zfcpdump(void)311 static inline void setup_zfcpdump(void) {}
312 #endif /* CONFIG_CRASH_DUMP */
313 
314  /*
315  * Reboot, halt and power_off stubs. They just call _machine_restart,
316  * _machine_halt or _machine_power_off.
317  */
318 
machine_restart(char * command)319 void machine_restart(char *command)
320 {
321 	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
322 		/*
323 		 * Only unblank the console if we are called in enabled
324 		 * context or a bust_spinlocks cleared the way for us.
325 		 */
326 		console_unblank();
327 	_machine_restart(command);
328 }
329 
machine_halt(void)330 void machine_halt(void)
331 {
332 	if (!in_interrupt() || oops_in_progress)
333 		/*
334 		 * Only unblank the console if we are called in enabled
335 		 * context or a bust_spinlocks cleared the way for us.
336 		 */
337 		console_unblank();
338 	_machine_halt();
339 }
340 
machine_power_off(void)341 void machine_power_off(void)
342 {
343 	if (!in_interrupt() || oops_in_progress)
344 		/*
345 		 * Only unblank the console if we are called in enabled
346 		 * context or a bust_spinlocks cleared the way for us.
347 		 */
348 		console_unblank();
349 	_machine_power_off();
350 }
351 
352 /*
353  * Dummy power off function.
354  */
355 void (*pm_power_off)(void) = machine_power_off;
356 EXPORT_SYMBOL_GPL(pm_power_off);
357 
358 void *restart_stack;
359 
stack_alloc(void)360 unsigned long stack_alloc(void)
361 {
362 #ifdef CONFIG_VMAP_STACK
363 	void *ret;
364 
365 	ret = __vmalloc_node(THREAD_SIZE, THREAD_SIZE, THREADINFO_GFP,
366 			     NUMA_NO_NODE, __builtin_return_address(0));
367 	kmemleak_not_leak(ret);
368 	return (unsigned long)ret;
369 #else
370 	return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
371 #endif
372 }
373 
stack_free(unsigned long stack)374 void stack_free(unsigned long stack)
375 {
376 #ifdef CONFIG_VMAP_STACK
377 	vfree((void *) stack);
378 #else
379 	free_pages(stack, THREAD_SIZE_ORDER);
380 #endif
381 }
382 
arch_early_irq_init(void)383 int __init arch_early_irq_init(void)
384 {
385 	unsigned long stack;
386 
387 	stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
388 	if (!stack)
389 		panic("Couldn't allocate async stack");
390 	S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
391 	return 0;
392 }
393 
arch_call_rest_init(void)394 void __init arch_call_rest_init(void)
395 {
396 	unsigned long stack;
397 
398 	stack = stack_alloc();
399 	if (!stack)
400 		panic("Couldn't allocate kernel stack");
401 	current->stack = (void *) stack;
402 #ifdef CONFIG_VMAP_STACK
403 	current->stack_vm_area = (void *) stack;
404 #endif
405 	set_task_stack_end_magic(current);
406 	stack += STACK_INIT_OFFSET;
407 	S390_lowcore.kernel_stack = stack;
408 	call_on_stack_noreturn(rest_init, stack);
409 }
410 
setup_lowcore_dat_off(void)411 static void __init setup_lowcore_dat_off(void)
412 {
413 	unsigned long int_psw_mask = PSW_KERNEL_BITS;
414 	unsigned long mcck_stack;
415 	struct lowcore *lc;
416 
417 	if (IS_ENABLED(CONFIG_KASAN))
418 		int_psw_mask |= PSW_MASK_DAT;
419 
420 	/*
421 	 * Setup lowcore for boot cpu
422 	 */
423 	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
424 	lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
425 	if (!lc)
426 		panic("%s: Failed to allocate %zu bytes align=%zx\n",
427 		      __func__, sizeof(*lc), sizeof(*lc));
428 
429 	lc->restart_psw.mask = PSW_KERNEL_BITS;
430 	lc->restart_psw.addr = (unsigned long) restart_int_handler;
431 	lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
432 	lc->external_new_psw.addr = (unsigned long) ext_int_handler;
433 	lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
434 	lc->svc_new_psw.addr = (unsigned long) system_call;
435 	lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
436 	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
437 	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
438 	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
439 	lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
440 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
441 	lc->clock_comparator = clock_comparator_max;
442 	lc->nodat_stack = ((unsigned long) &init_thread_union)
443 		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
444 	lc->current_task = (unsigned long)&init_task;
445 	lc->lpp = LPP_MAGIC;
446 	lc->machine_flags = S390_lowcore.machine_flags;
447 	lc->preempt_count = S390_lowcore.preempt_count;
448 	nmi_alloc_mcesa_early(&lc->mcesad);
449 	lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
450 	lc->exit_timer = S390_lowcore.exit_timer;
451 	lc->user_timer = S390_lowcore.user_timer;
452 	lc->system_timer = S390_lowcore.system_timer;
453 	lc->steal_timer = S390_lowcore.steal_timer;
454 	lc->last_update_timer = S390_lowcore.last_update_timer;
455 	lc->last_update_clock = S390_lowcore.last_update_clock;
456 
457 	/*
458 	 * Allocate the global restart stack which is the same for
459 	 * all CPUs in cast *one* of them does a PSW restart.
460 	 */
461 	restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
462 	if (!restart_stack)
463 		panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
464 		      __func__, THREAD_SIZE, THREAD_SIZE);
465 	restart_stack += STACK_INIT_OFFSET;
466 
467 	/*
468 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
469 	 * restart data to the absolute zero lowcore. This is necessary if
470 	 * PSW restart is done on an offline CPU that has lowcore zero.
471 	 */
472 	lc->restart_stack = (unsigned long) restart_stack;
473 	lc->restart_fn = (unsigned long) do_restart;
474 	lc->restart_data = 0;
475 	lc->restart_source = -1U;
476 
477 	mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
478 	if (!mcck_stack)
479 		panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
480 		      __func__, THREAD_SIZE, THREAD_SIZE);
481 	lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
482 
483 	/* Setup absolute zero lowcore */
484 	put_abs_lowcore(restart_stack, lc->restart_stack);
485 	put_abs_lowcore(restart_fn, lc->restart_fn);
486 	put_abs_lowcore(restart_data, lc->restart_data);
487 	put_abs_lowcore(restart_source, lc->restart_source);
488 	put_abs_lowcore(restart_psw, lc->restart_psw);
489 	put_abs_lowcore(mcesad, lc->mcesad);
490 
491 	lc->spinlock_lockval = arch_spin_lockval(0);
492 	lc->spinlock_index = 0;
493 	arch_spin_lock_setup(0);
494 	lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
495 	lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
496 	lc->preempt_count = PREEMPT_DISABLED;
497 
498 	set_prefix(__pa(lc));
499 	lowcore_ptr[0] = lc;
500 }
501 
setup_lowcore_dat_on(void)502 static void __init setup_lowcore_dat_on(void)
503 {
504 	struct lowcore *lc = lowcore_ptr[0];
505 	int cr;
506 
507 	__ctl_clear_bit(0, 28);
508 	S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
509 	S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
510 	S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
511 	S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
512 	__ctl_store(S390_lowcore.cregs_save_area, 0, 15);
513 	__ctl_set_bit(0, 28);
514 	put_abs_lowcore(restart_flags, RESTART_FLAG_CTLREGS);
515 	put_abs_lowcore(program_new_psw, lc->program_new_psw);
516 	for (cr = 0; cr < ARRAY_SIZE(lc->cregs_save_area); cr++)
517 		put_abs_lowcore(cregs_save_area[cr], lc->cregs_save_area[cr]);
518 }
519 
520 static struct resource code_resource = {
521 	.name  = "Kernel code",
522 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
523 };
524 
525 static struct resource data_resource = {
526 	.name = "Kernel data",
527 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
528 };
529 
530 static struct resource bss_resource = {
531 	.name = "Kernel bss",
532 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
533 };
534 
535 static struct resource __initdata *standard_resources[] = {
536 	&code_resource,
537 	&data_resource,
538 	&bss_resource,
539 };
540 
setup_resources(void)541 static void __init setup_resources(void)
542 {
543 	struct resource *res, *std_res, *sub_res;
544 	phys_addr_t start, end;
545 	int j;
546 	u64 i;
547 
548 	code_resource.start = (unsigned long) _text;
549 	code_resource.end = (unsigned long) _etext - 1;
550 	data_resource.start = (unsigned long) _etext;
551 	data_resource.end = (unsigned long) _edata - 1;
552 	bss_resource.start = (unsigned long) __bss_start;
553 	bss_resource.end = (unsigned long) __bss_stop - 1;
554 
555 	for_each_mem_range(i, &start, &end) {
556 		res = memblock_alloc(sizeof(*res), 8);
557 		if (!res)
558 			panic("%s: Failed to allocate %zu bytes align=0x%x\n",
559 			      __func__, sizeof(*res), 8);
560 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
561 
562 		res->name = "System RAM";
563 		res->start = start;
564 		/*
565 		 * In memblock, end points to the first byte after the
566 		 * range while in resourses, end points to the last byte in
567 		 * the range.
568 		 */
569 		res->end = end - 1;
570 		request_resource(&iomem_resource, res);
571 
572 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
573 			std_res = standard_resources[j];
574 			if (std_res->start < res->start ||
575 			    std_res->start > res->end)
576 				continue;
577 			if (std_res->end > res->end) {
578 				sub_res = memblock_alloc(sizeof(*sub_res), 8);
579 				if (!sub_res)
580 					panic("%s: Failed to allocate %zu bytes align=0x%x\n",
581 					      __func__, sizeof(*sub_res), 8);
582 				*sub_res = *std_res;
583 				sub_res->end = res->end;
584 				std_res->start = res->end + 1;
585 				request_resource(res, sub_res);
586 			} else {
587 				request_resource(res, std_res);
588 			}
589 		}
590 	}
591 #ifdef CONFIG_CRASH_DUMP
592 	/*
593 	 * Re-add removed crash kernel memory as reserved memory. This makes
594 	 * sure it will be mapped with the identity mapping and struct pages
595 	 * will be created, so it can be resized later on.
596 	 * However add it later since the crash kernel resource should not be
597 	 * part of the System RAM resource.
598 	 */
599 	if (crashk_res.end) {
600 		memblock_add_node(crashk_res.start, resource_size(&crashk_res),
601 				  0, MEMBLOCK_NONE);
602 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
603 		insert_resource(&iomem_resource, &crashk_res);
604 	}
605 #endif
606 }
607 
setup_memory_end(void)608 static void __init setup_memory_end(void)
609 {
610 	memblock_remove(ident_map_size, PHYS_ADDR_MAX - ident_map_size);
611 	max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
612 	pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
613 }
614 
615 #ifdef CONFIG_CRASH_DUMP
616 
617 /*
618  * When kdump is enabled, we have to ensure that no memory from the area
619  * [0 - crashkernel memory size] is set offline - it will be exchanged with
620  * the crashkernel memory region when kdump is triggered. The crashkernel
621  * memory region can never get offlined (pages are unmovable).
622  */
kdump_mem_notifier(struct notifier_block * nb,unsigned long action,void * data)623 static int kdump_mem_notifier(struct notifier_block *nb,
624 			      unsigned long action, void *data)
625 {
626 	struct memory_notify *arg = data;
627 
628 	if (action != MEM_GOING_OFFLINE)
629 		return NOTIFY_OK;
630 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
631 		return NOTIFY_BAD;
632 	return NOTIFY_OK;
633 }
634 
635 static struct notifier_block kdump_mem_nb = {
636 	.notifier_call = kdump_mem_notifier,
637 };
638 
639 #endif
640 
641 /*
642  * Reserve memory for kdump kernel to be loaded with kexec
643  */
reserve_crashkernel(void)644 static void __init reserve_crashkernel(void)
645 {
646 #ifdef CONFIG_CRASH_DUMP
647 	unsigned long long crash_base, crash_size;
648 	phys_addr_t low, high;
649 	int rc;
650 
651 	rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
652 			       &crash_base);
653 
654 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
655 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
656 	if (rc || crash_size == 0)
657 		return;
658 
659 	if (memblock.memory.regions[0].size < crash_size) {
660 		pr_info("crashkernel reservation failed: %s\n",
661 			"first memory chunk must be at least crashkernel size");
662 		return;
663 	}
664 
665 	low = crash_base ?: oldmem_data.start;
666 	high = low + crash_size;
667 	if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
668 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
669 		crash_base = low;
670 	} else {
671 		/* Find suitable area in free memory */
672 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
673 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
674 
675 		if (crash_base && crash_base < low) {
676 			pr_info("crashkernel reservation failed: %s\n",
677 				"crash_base too low");
678 			return;
679 		}
680 		low = crash_base ?: low;
681 		crash_base = memblock_phys_alloc_range(crash_size,
682 						       KEXEC_CRASH_MEM_ALIGN,
683 						       low, high);
684 	}
685 
686 	if (!crash_base) {
687 		pr_info("crashkernel reservation failed: %s\n",
688 			"no suitable area found");
689 		return;
690 	}
691 
692 	if (register_memory_notifier(&kdump_mem_nb)) {
693 		memblock_phys_free(crash_base, crash_size);
694 		return;
695 	}
696 
697 	if (!oldmem_data.start && MACHINE_IS_VM)
698 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
699 	crashk_res.start = crash_base;
700 	crashk_res.end = crash_base + crash_size - 1;
701 	memblock_remove(crash_base, crash_size);
702 	pr_info("Reserving %lluMB of memory at %lluMB "
703 		"for crashkernel (System RAM: %luMB)\n",
704 		crash_size >> 20, crash_base >> 20,
705 		(unsigned long)memblock.memory.total_size >> 20);
706 	os_info_crashkernel_add(crash_base, crash_size);
707 #endif
708 }
709 
710 /*
711  * Reserve the initrd from being used by memblock
712  */
reserve_initrd(void)713 static void __init reserve_initrd(void)
714 {
715 #ifdef CONFIG_BLK_DEV_INITRD
716 	if (!initrd_data.start || !initrd_data.size)
717 		return;
718 	initrd_start = (unsigned long)__va(initrd_data.start);
719 	initrd_end = initrd_start + initrd_data.size;
720 	memblock_reserve(initrd_data.start, initrd_data.size);
721 #endif
722 }
723 
724 /*
725  * Reserve the memory area used to pass the certificate lists
726  */
reserve_certificate_list(void)727 static void __init reserve_certificate_list(void)
728 {
729 	if (ipl_cert_list_addr)
730 		memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
731 }
732 
reserve_mem_detect_info(void)733 static void __init reserve_mem_detect_info(void)
734 {
735 	unsigned long start, size;
736 
737 	get_mem_detect_reserved(&start, &size);
738 	if (size)
739 		memblock_reserve(start, size);
740 }
741 
free_mem_detect_info(void)742 static void __init free_mem_detect_info(void)
743 {
744 	unsigned long start, size;
745 
746 	get_mem_detect_reserved(&start, &size);
747 	if (size)
748 		memblock_phys_free(start, size);
749 }
750 
get_mem_info_source(void)751 static const char * __init get_mem_info_source(void)
752 {
753 	switch (mem_detect.info_source) {
754 	case MEM_DETECT_SCLP_STOR_INFO:
755 		return "sclp storage info";
756 	case MEM_DETECT_DIAG260:
757 		return "diag260";
758 	case MEM_DETECT_SCLP_READ_INFO:
759 		return "sclp read info";
760 	case MEM_DETECT_BIN_SEARCH:
761 		return "binary search";
762 	}
763 	return "none";
764 }
765 
memblock_add_mem_detect_info(void)766 static void __init memblock_add_mem_detect_info(void)
767 {
768 	unsigned long start, end;
769 	int i;
770 
771 	pr_debug("physmem info source: %s (%hhd)\n",
772 		 get_mem_info_source(), mem_detect.info_source);
773 	/* keep memblock lists close to the kernel */
774 	memblock_set_bottom_up(true);
775 	for_each_mem_detect_block(i, &start, &end) {
776 		memblock_add(start, end - start);
777 		memblock_physmem_add(start, end - start);
778 	}
779 	memblock_set_bottom_up(false);
780 	memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
781 }
782 
783 /*
784  * Check for initrd being in usable memory
785  */
check_initrd(void)786 static void __init check_initrd(void)
787 {
788 #ifdef CONFIG_BLK_DEV_INITRD
789 	if (initrd_data.start && initrd_data.size &&
790 	    !memblock_is_region_memory(initrd_data.start, initrd_data.size)) {
791 		pr_err("The initial RAM disk does not fit into the memory\n");
792 		memblock_phys_free(initrd_data.start, initrd_data.size);
793 		initrd_start = initrd_end = 0;
794 	}
795 #endif
796 }
797 
798 /*
799  * Reserve memory used for lowcore/command line/kernel image.
800  */
reserve_kernel(void)801 static void __init reserve_kernel(void)
802 {
803 	memblock_reserve(0, STARTUP_NORMAL_OFFSET);
804 	memblock_reserve(OLDMEM_BASE, sizeof(unsigned long));
805 	memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long));
806 	memblock_reserve(__amode31_base, __eamode31 - __samode31);
807 	memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP);
808 	memblock_reserve(__pa(_stext), _end - _stext);
809 }
810 
setup_memory(void)811 static void __init setup_memory(void)
812 {
813 	phys_addr_t start, end;
814 	u64 i;
815 
816 	/*
817 	 * Init storage key for present memory
818 	 */
819 	for_each_mem_range(i, &start, &end)
820 		storage_key_init_range(start, end);
821 
822 	psw_set_key(PAGE_DEFAULT_KEY);
823 }
824 
relocate_amode31_section(void)825 static void __init relocate_amode31_section(void)
826 {
827 	unsigned long amode31_size = __eamode31 - __samode31;
828 	long amode31_offset = __amode31_base - __samode31;
829 	long *ptr;
830 
831 	pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size);
832 
833 	/* Move original AMODE31 section to the new one */
834 	memmove((void *)__amode31_base, (void *)__samode31, amode31_size);
835 	/* Zero out the old AMODE31 section to catch invalid accesses within it */
836 	memset((void *)__samode31, 0, amode31_size);
837 
838 	/* Update all AMODE31 region references */
839 	for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++)
840 		*ptr += amode31_offset;
841 }
842 
843 /* This must be called after AMODE31 relocation */
setup_cr(void)844 static void __init setup_cr(void)
845 {
846 	union ctlreg2 cr2;
847 	union ctlreg5 cr5;
848 	union ctlreg15 cr15;
849 
850 	__ctl_duct[1] = (unsigned long)__ctl_aste;
851 	__ctl_duct[2] = (unsigned long)__ctl_aste;
852 	__ctl_duct[4] = (unsigned long)__ctl_duald;
853 
854 	/* Update control registers CR2, CR5 and CR15 */
855 	__ctl_store(cr2.val, 2, 2);
856 	__ctl_store(cr5.val, 5, 5);
857 	__ctl_store(cr15.val, 15, 15);
858 	cr2.ducto = (unsigned long)__ctl_duct >> 6;
859 	cr5.pasteo = (unsigned long)__ctl_duct >> 6;
860 	cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
861 	__ctl_load(cr2.val, 2, 2);
862 	__ctl_load(cr5.val, 5, 5);
863 	__ctl_load(cr15.val, 15, 15);
864 }
865 
866 /*
867  * Add system information as device randomness
868  */
setup_randomness(void)869 static void __init setup_randomness(void)
870 {
871 	struct sysinfo_3_2_2 *vmms;
872 
873 	vmms = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
874 	if (!vmms)
875 		panic("Failed to allocate memory for sysinfo structure\n");
876 	if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
877 		add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
878 	memblock_free(vmms, PAGE_SIZE);
879 
880 #ifdef CONFIG_ARCH_RANDOM
881 	if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
882 		static_branch_enable(&s390_arch_random_available);
883 #endif
884 }
885 
886 /*
887  * Find the correct size for the task_struct. This depends on
888  * the size of the struct fpu at the end of the thread_struct
889  * which is embedded in the task_struct.
890  */
setup_task_size(void)891 static void __init setup_task_size(void)
892 {
893 	int task_size = sizeof(struct task_struct);
894 
895 	if (!MACHINE_HAS_VX) {
896 		task_size -= sizeof(__vector128) * __NUM_VXRS;
897 		task_size += sizeof(freg_t) * __NUM_FPRS;
898 	}
899 	arch_task_struct_size = task_size;
900 }
901 
902 /*
903  * Issue diagnose 318 to set the control program name and
904  * version codes.
905  */
setup_control_program_code(void)906 static void __init setup_control_program_code(void)
907 {
908 	union diag318_info diag318_info = {
909 		.cpnc = CPNC_LINUX,
910 		.cpvc = 0,
911 	};
912 
913 	if (!sclp.has_diag318)
914 		return;
915 
916 	diag_stat_inc(DIAG_STAT_X318);
917 	asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
918 }
919 
920 /*
921  * Print the component list from the IPL report
922  */
log_component_list(void)923 static void __init log_component_list(void)
924 {
925 	struct ipl_rb_component_entry *ptr, *end;
926 	char *str;
927 
928 	if (!early_ipl_comp_list_addr)
929 		return;
930 	if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
931 		pr_info("Linux is running with Secure-IPL enabled\n");
932 	else
933 		pr_info("Linux is running with Secure-IPL disabled\n");
934 	ptr = (void *) early_ipl_comp_list_addr;
935 	end = (void *) ptr + early_ipl_comp_list_size;
936 	pr_info("The IPL report contains the following components:\n");
937 	while (ptr < end) {
938 		if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
939 			if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
940 				str = "signed, verified";
941 			else
942 				str = "signed, verification failed";
943 		} else {
944 			str = "not signed";
945 		}
946 		pr_info("%016llx - %016llx (%s)\n",
947 			ptr->addr, ptr->addr + ptr->len, str);
948 		ptr++;
949 	}
950 }
951 
952 /*
953  * Setup function called from init/main.c just after the banner
954  * was printed.
955  */
956 
setup_arch(char ** cmdline_p)957 void __init setup_arch(char **cmdline_p)
958 {
959         /*
960          * print what head.S has found out about the machine
961          */
962 	if (MACHINE_IS_VM)
963 		pr_info("Linux is running as a z/VM "
964 			"guest operating system in 64-bit mode\n");
965 	else if (MACHINE_IS_KVM)
966 		pr_info("Linux is running under KVM in 64-bit mode\n");
967 	else if (MACHINE_IS_LPAR)
968 		pr_info("Linux is running natively in 64-bit mode\n");
969 	else
970 		pr_info("Linux is running as a guest in 64-bit mode\n");
971 
972 	log_component_list();
973 
974 	/* Have one command line that is parsed and saved in /proc/cmdline */
975 	/* boot_command_line has been already set up in early.c */
976 	*cmdline_p = boot_command_line;
977 
978         ROOT_DEV = Root_RAM0;
979 
980 	setup_initial_init_mm(_text, _etext, _edata, _end);
981 
982 	if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
983 		nospec_auto_detect();
984 
985 	jump_label_init();
986 	parse_early_param();
987 #ifdef CONFIG_CRASH_DUMP
988 	/* Deactivate elfcorehdr= kernel parameter */
989 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
990 #endif
991 
992 	os_info_init();
993 	setup_ipl();
994 	setup_task_size();
995 	setup_control_program_code();
996 
997 	/* Do some memory reservations *before* memory is added to memblock */
998 	reserve_kernel();
999 	reserve_initrd();
1000 	reserve_certificate_list();
1001 	reserve_mem_detect_info();
1002 	memblock_set_current_limit(ident_map_size);
1003 	memblock_allow_resize();
1004 
1005 	/* Get information about *all* installed memory */
1006 	memblock_add_mem_detect_info();
1007 
1008 	free_mem_detect_info();
1009 	setup_memory_end();
1010 	memblock_dump_all();
1011 	setup_memory();
1012 
1013 	relocate_amode31_section();
1014 	setup_cr();
1015 	setup_uv();
1016 	dma_contiguous_reserve(ident_map_size);
1017 	vmcp_cma_reserve();
1018 	if (MACHINE_HAS_EDAT2)
1019 		hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1020 
1021 	check_initrd();
1022 	reserve_crashkernel();
1023 #ifdef CONFIG_CRASH_DUMP
1024 	/*
1025 	 * Be aware that smp_save_dump_cpus() triggers a system reset.
1026 	 * Therefore CPU and device initialization should be done afterwards.
1027 	 */
1028 	smp_save_dump_cpus();
1029 #endif
1030 
1031 	setup_resources();
1032 	setup_lowcore_dat_off();
1033 	smp_fill_possible_mask();
1034 	cpu_detect_mhz_feature();
1035         cpu_init();
1036 	numa_setup();
1037 	smp_detect_cpus();
1038 	topology_init_early();
1039 
1040 	if (test_facility(193))
1041 		static_branch_enable(&cpu_has_bear);
1042 
1043 	/*
1044 	 * Create kernel page tables and switch to virtual addressing.
1045 	 */
1046         paging_init();
1047 
1048 	/*
1049 	 * After paging_init created the kernel page table, the new PSWs
1050 	 * in lowcore can now run with DAT enabled.
1051 	 */
1052 	setup_lowcore_dat_on();
1053 
1054         /* Setup default console */
1055 	conmode_default();
1056 	set_preferred_console();
1057 
1058 	apply_alternative_instructions();
1059 	if (IS_ENABLED(CONFIG_EXPOLINE))
1060 		nospec_init_branches();
1061 
1062 	/* Setup zfcp/nvme dump support */
1063 	setup_zfcpdump();
1064 
1065 	/* Add system specific data to the random pool */
1066 	setup_randomness();
1067 }
1068