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/abs_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/maccess.h>
78 #include <asm/uv.h>
79 #include <asm/asm-offsets.h>
80 #include "entry.h"
81
82 /*
83 * Machine setup..
84 */
85 unsigned int console_mode = 0;
86 EXPORT_SYMBOL(console_mode);
87
88 unsigned int console_devno = -1;
89 EXPORT_SYMBOL(console_devno);
90
91 unsigned int console_irq = -1;
92 EXPORT_SYMBOL(console_irq);
93
94 /*
95 * Some code and data needs to stay below 2 GB, even when the kernel would be
96 * relocated above 2 GB, because it has to use 31 bit addresses.
97 * Such code and data is part of the .amode31 section.
98 */
99 unsigned long __amode31_ref __samode31 = (unsigned long)&_samode31;
100 unsigned long __amode31_ref __eamode31 = (unsigned long)&_eamode31;
101 unsigned long __amode31_ref __stext_amode31 = (unsigned long)&_stext_amode31;
102 unsigned long __amode31_ref __etext_amode31 = (unsigned long)&_etext_amode31;
103 struct exception_table_entry __amode31_ref *__start_amode31_ex_table = _start_amode31_ex_table;
104 struct exception_table_entry __amode31_ref *__stop_amode31_ex_table = _stop_amode31_ex_table;
105
106 /*
107 * Control registers CR2, CR5 and CR15 are initialized with addresses
108 * of tables that must be placed below 2G which is handled by the AMODE31
109 * sections.
110 * Because the AMODE31 sections are relocated below 2G at startup,
111 * the content of control registers CR2, CR5 and CR15 must be updated
112 * with new addresses after the relocation. The initial initialization of
113 * control registers occurs in head64.S and then gets updated again after AMODE31
114 * relocation. We must access the relevant AMODE31 tables indirectly via
115 * pointers placed in the .amode31.refs linker section. Those pointers get
116 * updated automatically during AMODE31 relocation and always contain a valid
117 * address within AMODE31 sections.
118 */
119
120 static __amode31_data u32 __ctl_duct_amode31[16] __aligned(64);
121
122 static __amode31_data u64 __ctl_aste_amode31[8] __aligned(64) = {
123 [1] = 0xffffffffffffffff
124 };
125
126 static __amode31_data u32 __ctl_duald_amode31[32] __aligned(128) = {
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 0x80000000, 0, 0, 0
135 };
136
137 static __amode31_data u32 __ctl_linkage_stack_amode31[8] __aligned(64) = {
138 0, 0, 0x89000000, 0,
139 0, 0, 0x8a000000, 0
140 };
141
142 static u64 __amode31_ref *__ctl_aste = __ctl_aste_amode31;
143 static u32 __amode31_ref *__ctl_duald = __ctl_duald_amode31;
144 static u32 __amode31_ref *__ctl_linkage_stack = __ctl_linkage_stack_amode31;
145 static u32 __amode31_ref *__ctl_duct = __ctl_duct_amode31;
146
147 int __bootdata(noexec_disabled);
148 unsigned long __bootdata(ident_map_size);
149 struct mem_detect_info __bootdata(mem_detect);
150 struct initrd_data __bootdata(initrd_data);
151
152 unsigned long __bootdata_preserved(__kaslr_offset);
153 unsigned long __bootdata(__amode31_base);
154 unsigned int __bootdata_preserved(zlib_dfltcc_support);
155 EXPORT_SYMBOL(zlib_dfltcc_support);
156 u64 __bootdata_preserved(stfle_fac_list[16]);
157 EXPORT_SYMBOL(stfle_fac_list);
158 u64 __bootdata_preserved(alt_stfle_fac_list[16]);
159 struct oldmem_data __bootdata_preserved(oldmem_data);
160
161 unsigned long VMALLOC_START;
162 EXPORT_SYMBOL(VMALLOC_START);
163
164 unsigned long VMALLOC_END;
165 EXPORT_SYMBOL(VMALLOC_END);
166
167 struct page *vmemmap;
168 EXPORT_SYMBOL(vmemmap);
169 unsigned long vmemmap_size;
170
171 unsigned long MODULES_VADDR;
172 unsigned long MODULES_END;
173
174 /* An array with a pointer to the lowcore of every CPU. */
175 struct lowcore *lowcore_ptr[NR_CPUS];
176 EXPORT_SYMBOL(lowcore_ptr);
177
178 DEFINE_STATIC_KEY_FALSE(cpu_has_bear);
179
180 /*
181 * The Write Back bit position in the physaddr is given by the SLPC PCI.
182 * Leaving the mask zero always uses write through which is safe
183 */
184 unsigned long mio_wb_bit_mask __ro_after_init;
185
186 /*
187 * This is set up by the setup-routine at boot-time
188 * for S390 need to find out, what we have to setup
189 * using address 0x10400 ...
190 */
191
192 #include <asm/setup.h>
193
194 /*
195 * condev= and conmode= setup parameter.
196 */
197
condev_setup(char * str)198 static int __init condev_setup(char *str)
199 {
200 int vdev;
201
202 vdev = simple_strtoul(str, &str, 0);
203 if (vdev >= 0 && vdev < 65536) {
204 console_devno = vdev;
205 console_irq = -1;
206 }
207 return 1;
208 }
209
210 __setup("condev=", condev_setup);
211
set_preferred_console(void)212 static void __init set_preferred_console(void)
213 {
214 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
215 add_preferred_console("ttyS", 0, NULL);
216 else if (CONSOLE_IS_3270)
217 add_preferred_console("tty3270", 0, NULL);
218 else if (CONSOLE_IS_VT220)
219 add_preferred_console("ttysclp", 0, NULL);
220 else if (CONSOLE_IS_HVC)
221 add_preferred_console("hvc", 0, NULL);
222 }
223
conmode_setup(char * str)224 static int __init conmode_setup(char *str)
225 {
226 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
227 if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
228 SET_CONSOLE_SCLP;
229 #endif
230 #if defined(CONFIG_TN3215_CONSOLE)
231 if (!strcmp(str, "3215"))
232 SET_CONSOLE_3215;
233 #endif
234 #if defined(CONFIG_TN3270_CONSOLE)
235 if (!strcmp(str, "3270"))
236 SET_CONSOLE_3270;
237 #endif
238 set_preferred_console();
239 return 1;
240 }
241
242 __setup("conmode=", conmode_setup);
243
conmode_default(void)244 static void __init conmode_default(void)
245 {
246 char query_buffer[1024];
247 char *ptr;
248
249 if (MACHINE_IS_VM) {
250 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
251 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
252 ptr = strstr(query_buffer, "SUBCHANNEL =");
253 console_irq = simple_strtoul(ptr + 13, NULL, 16);
254 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
255 ptr = strstr(query_buffer, "CONMODE");
256 /*
257 * Set the conmode to 3215 so that the device recognition
258 * will set the cu_type of the console to 3215. If the
259 * conmode is 3270 and we don't set it back then both
260 * 3215 and the 3270 driver will try to access the console
261 * device (3215 as console and 3270 as normal tty).
262 */
263 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
264 if (ptr == NULL) {
265 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
266 SET_CONSOLE_SCLP;
267 #endif
268 return;
269 }
270 if (str_has_prefix(ptr + 8, "3270")) {
271 #if defined(CONFIG_TN3270_CONSOLE)
272 SET_CONSOLE_3270;
273 #elif defined(CONFIG_TN3215_CONSOLE)
274 SET_CONSOLE_3215;
275 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
276 SET_CONSOLE_SCLP;
277 #endif
278 } else if (str_has_prefix(ptr + 8, "3215")) {
279 #if defined(CONFIG_TN3215_CONSOLE)
280 SET_CONSOLE_3215;
281 #elif defined(CONFIG_TN3270_CONSOLE)
282 SET_CONSOLE_3270;
283 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
284 SET_CONSOLE_SCLP;
285 #endif
286 }
287 } else if (MACHINE_IS_KVM) {
288 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
289 SET_CONSOLE_VT220;
290 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
291 SET_CONSOLE_SCLP;
292 else
293 SET_CONSOLE_HVC;
294 } else {
295 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
296 SET_CONSOLE_SCLP;
297 #endif
298 }
299 }
300
301 #ifdef CONFIG_CRASH_DUMP
setup_zfcpdump(void)302 static void __init setup_zfcpdump(void)
303 {
304 if (!is_ipl_type_dump())
305 return;
306 if (oldmem_data.start)
307 return;
308 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
309 console_loglevel = 2;
310 }
311 #else
setup_zfcpdump(void)312 static inline void setup_zfcpdump(void) {}
313 #endif /* CONFIG_CRASH_DUMP */
314
315 /*
316 * Reboot, halt and power_off stubs. They just call _machine_restart,
317 * _machine_halt or _machine_power_off.
318 */
319
machine_restart(char * command)320 void machine_restart(char *command)
321 {
322 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
323 /*
324 * Only unblank the console if we are called in enabled
325 * context or a bust_spinlocks cleared the way for us.
326 */
327 console_unblank();
328 _machine_restart(command);
329 }
330
machine_halt(void)331 void machine_halt(void)
332 {
333 if (!in_interrupt() || oops_in_progress)
334 /*
335 * Only unblank the console if we are called in enabled
336 * context or a bust_spinlocks cleared the way for us.
337 */
338 console_unblank();
339 _machine_halt();
340 }
341
machine_power_off(void)342 void machine_power_off(void)
343 {
344 if (!in_interrupt() || oops_in_progress)
345 /*
346 * Only unblank the console if we are called in enabled
347 * context or a bust_spinlocks cleared the way for us.
348 */
349 console_unblank();
350 _machine_power_off();
351 }
352
353 /*
354 * Dummy power off function.
355 */
356 void (*pm_power_off)(void) = machine_power_off;
357 EXPORT_SYMBOL_GPL(pm_power_off);
358
359 void *restart_stack;
360
stack_alloc(void)361 unsigned long stack_alloc(void)
362 {
363 #ifdef CONFIG_VMAP_STACK
364 void *ret;
365
366 ret = __vmalloc_node(THREAD_SIZE, THREAD_SIZE, THREADINFO_GFP,
367 NUMA_NO_NODE, __builtin_return_address(0));
368 kmemleak_not_leak(ret);
369 return (unsigned long)ret;
370 #else
371 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
372 #endif
373 }
374
stack_free(unsigned long stack)375 void stack_free(unsigned long stack)
376 {
377 #ifdef CONFIG_VMAP_STACK
378 vfree((void *) stack);
379 #else
380 free_pages(stack, THREAD_SIZE_ORDER);
381 #endif
382 }
383
arch_early_irq_init(void)384 int __init arch_early_irq_init(void)
385 {
386 unsigned long stack;
387
388 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
389 if (!stack)
390 panic("Couldn't allocate async stack");
391 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
392 return 0;
393 }
394
arch_call_rest_init(void)395 void __init arch_call_rest_init(void)
396 {
397 unsigned long stack;
398
399 smp_reinit_ipl_cpu();
400 stack = stack_alloc();
401 if (!stack)
402 panic("Couldn't allocate kernel stack");
403 current->stack = (void *) stack;
404 #ifdef CONFIG_VMAP_STACK
405 current->stack_vm_area = (void *) stack;
406 #endif
407 set_task_stack_end_magic(current);
408 stack += STACK_INIT_OFFSET;
409 S390_lowcore.kernel_stack = stack;
410 call_on_stack_noreturn(rest_init, stack);
411 }
412
setup_lowcore_dat_off(void)413 static void __init setup_lowcore_dat_off(void)
414 {
415 unsigned long int_psw_mask = PSW_KERNEL_BITS;
416 struct lowcore *abs_lc, *lc;
417 unsigned long mcck_stack;
418 unsigned long flags;
419
420 if (IS_ENABLED(CONFIG_KASAN))
421 int_psw_mask |= PSW_MASK_DAT;
422
423 /*
424 * Setup lowcore for boot cpu
425 */
426 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
427 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
428 if (!lc)
429 panic("%s: Failed to allocate %zu bytes align=%zx\n",
430 __func__, sizeof(*lc), sizeof(*lc));
431
432 lc->restart_psw.mask = PSW_KERNEL_BITS;
433 lc->restart_psw.addr = (unsigned long) restart_int_handler;
434 lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
435 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
436 lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
437 lc->svc_new_psw.addr = (unsigned long) system_call;
438 lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
439 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
440 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
441 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
442 lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
443 lc->io_new_psw.addr = (unsigned long) io_int_handler;
444 lc->clock_comparator = clock_comparator_max;
445 lc->nodat_stack = ((unsigned long) &init_thread_union)
446 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
447 lc->current_task = (unsigned long)&init_task;
448 lc->lpp = LPP_MAGIC;
449 lc->machine_flags = S390_lowcore.machine_flags;
450 lc->preempt_count = S390_lowcore.preempt_count;
451 nmi_alloc_mcesa_early(&lc->mcesad);
452 lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
453 lc->exit_timer = S390_lowcore.exit_timer;
454 lc->user_timer = S390_lowcore.user_timer;
455 lc->system_timer = S390_lowcore.system_timer;
456 lc->steal_timer = S390_lowcore.steal_timer;
457 lc->last_update_timer = S390_lowcore.last_update_timer;
458 lc->last_update_clock = S390_lowcore.last_update_clock;
459
460 /*
461 * Allocate the global restart stack which is the same for
462 * all CPUs in cast *one* of them does a PSW restart.
463 */
464 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
465 if (!restart_stack)
466 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
467 __func__, THREAD_SIZE, THREAD_SIZE);
468 restart_stack += STACK_INIT_OFFSET;
469
470 /*
471 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
472 * restart data to the absolute zero lowcore. This is necessary if
473 * PSW restart is done on an offline CPU that has lowcore zero.
474 */
475 lc->restart_stack = (unsigned long) restart_stack;
476 lc->restart_fn = (unsigned long) do_restart;
477 lc->restart_data = 0;
478 lc->restart_source = -1U;
479
480 abs_lc = get_abs_lowcore(&flags);
481 abs_lc->restart_stack = lc->restart_stack;
482 abs_lc->restart_fn = lc->restart_fn;
483 abs_lc->restart_data = lc->restart_data;
484 abs_lc->restart_source = lc->restart_source;
485 abs_lc->restart_psw = lc->restart_psw;
486 abs_lc->mcesad = lc->mcesad;
487 put_abs_lowcore(abs_lc, flags);
488
489 mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
490 if (!mcck_stack)
491 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
492 __func__, THREAD_SIZE, THREAD_SIZE);
493 lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
494
495 lc->spinlock_lockval = arch_spin_lockval(0);
496 lc->spinlock_index = 0;
497 arch_spin_lock_setup(0);
498 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
499 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
500 lc->preempt_count = PREEMPT_DISABLED;
501
502 set_prefix(__pa(lc));
503 lowcore_ptr[0] = lc;
504 }
505
setup_lowcore_dat_on(void)506 static void __init setup_lowcore_dat_on(void)
507 {
508 struct lowcore *abs_lc;
509 unsigned long flags;
510
511 __ctl_clear_bit(0, 28);
512 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
513 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
514 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
515 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
516 __ctl_set_bit(0, 28);
517 __ctl_store(S390_lowcore.cregs_save_area, 0, 15);
518 if (abs_lowcore_map(0, lowcore_ptr[0], true))
519 panic("Couldn't setup absolute lowcore");
520 abs_lowcore_mapped = true;
521 abs_lc = get_abs_lowcore(&flags);
522 abs_lc->restart_flags = RESTART_FLAG_CTLREGS;
523 abs_lc->program_new_psw = S390_lowcore.program_new_psw;
524 memcpy(abs_lc->cregs_save_area, S390_lowcore.cregs_save_area,
525 sizeof(abs_lc->cregs_save_area));
526 put_abs_lowcore(abs_lc, flags);
527 }
528
529 static struct resource code_resource = {
530 .name = "Kernel code",
531 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
532 };
533
534 static struct resource data_resource = {
535 .name = "Kernel data",
536 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
537 };
538
539 static struct resource bss_resource = {
540 .name = "Kernel bss",
541 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
542 };
543
544 static struct resource __initdata *standard_resources[] = {
545 &code_resource,
546 &data_resource,
547 &bss_resource,
548 };
549
setup_resources(void)550 static void __init setup_resources(void)
551 {
552 struct resource *res, *std_res, *sub_res;
553 phys_addr_t start, end;
554 int j;
555 u64 i;
556
557 code_resource.start = (unsigned long) _text;
558 code_resource.end = (unsigned long) _etext - 1;
559 data_resource.start = (unsigned long) _etext;
560 data_resource.end = (unsigned long) _edata - 1;
561 bss_resource.start = (unsigned long) __bss_start;
562 bss_resource.end = (unsigned long) __bss_stop - 1;
563
564 for_each_mem_range(i, &start, &end) {
565 res = memblock_alloc(sizeof(*res), 8);
566 if (!res)
567 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
568 __func__, sizeof(*res), 8);
569 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
570
571 res->name = "System RAM";
572 res->start = start;
573 /*
574 * In memblock, end points to the first byte after the
575 * range while in resourses, end points to the last byte in
576 * the range.
577 */
578 res->end = end - 1;
579 request_resource(&iomem_resource, res);
580
581 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
582 std_res = standard_resources[j];
583 if (std_res->start < res->start ||
584 std_res->start > res->end)
585 continue;
586 if (std_res->end > res->end) {
587 sub_res = memblock_alloc(sizeof(*sub_res), 8);
588 if (!sub_res)
589 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
590 __func__, sizeof(*sub_res), 8);
591 *sub_res = *std_res;
592 sub_res->end = res->end;
593 std_res->start = res->end + 1;
594 request_resource(res, sub_res);
595 } else {
596 request_resource(res, std_res);
597 }
598 }
599 }
600 #ifdef CONFIG_CRASH_DUMP
601 /*
602 * Re-add removed crash kernel memory as reserved memory. This makes
603 * sure it will be mapped with the identity mapping and struct pages
604 * will be created, so it can be resized later on.
605 * However add it later since the crash kernel resource should not be
606 * part of the System RAM resource.
607 */
608 if (crashk_res.end) {
609 memblock_add_node(crashk_res.start, resource_size(&crashk_res),
610 0, MEMBLOCK_NONE);
611 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
612 insert_resource(&iomem_resource, &crashk_res);
613 }
614 #endif
615 }
616
setup_memory_end(void)617 static void __init setup_memory_end(void)
618 {
619 memblock_remove(ident_map_size, PHYS_ADDR_MAX - ident_map_size);
620 max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
621 pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
622 }
623
624 #ifdef CONFIG_CRASH_DUMP
625
626 /*
627 * When kdump is enabled, we have to ensure that no memory from the area
628 * [0 - crashkernel memory size] is set offline - it will be exchanged with
629 * the crashkernel memory region when kdump is triggered. The crashkernel
630 * memory region can never get offlined (pages are unmovable).
631 */
kdump_mem_notifier(struct notifier_block * nb,unsigned long action,void * data)632 static int kdump_mem_notifier(struct notifier_block *nb,
633 unsigned long action, void *data)
634 {
635 struct memory_notify *arg = data;
636
637 if (action != MEM_GOING_OFFLINE)
638 return NOTIFY_OK;
639 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
640 return NOTIFY_BAD;
641 return NOTIFY_OK;
642 }
643
644 static struct notifier_block kdump_mem_nb = {
645 .notifier_call = kdump_mem_notifier,
646 };
647
648 #endif
649
650 /*
651 * Reserve memory for kdump kernel to be loaded with kexec
652 */
reserve_crashkernel(void)653 static void __init reserve_crashkernel(void)
654 {
655 #ifdef CONFIG_CRASH_DUMP
656 unsigned long long crash_base, crash_size;
657 phys_addr_t low, high;
658 int rc;
659
660 rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
661 &crash_base);
662
663 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
664 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
665 if (rc || crash_size == 0)
666 return;
667
668 if (memblock.memory.regions[0].size < crash_size) {
669 pr_info("crashkernel reservation failed: %s\n",
670 "first memory chunk must be at least crashkernel size");
671 return;
672 }
673
674 low = crash_base ?: oldmem_data.start;
675 high = low + crash_size;
676 if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
677 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
678 crash_base = low;
679 } else {
680 /* Find suitable area in free memory */
681 low = max_t(unsigned long, crash_size, sclp.hsa_size);
682 high = crash_base ? crash_base + crash_size : ULONG_MAX;
683
684 if (crash_base && crash_base < low) {
685 pr_info("crashkernel reservation failed: %s\n",
686 "crash_base too low");
687 return;
688 }
689 low = crash_base ?: low;
690 crash_base = memblock_phys_alloc_range(crash_size,
691 KEXEC_CRASH_MEM_ALIGN,
692 low, high);
693 }
694
695 if (!crash_base) {
696 pr_info("crashkernel reservation failed: %s\n",
697 "no suitable area found");
698 return;
699 }
700
701 if (register_memory_notifier(&kdump_mem_nb)) {
702 memblock_phys_free(crash_base, crash_size);
703 return;
704 }
705
706 if (!oldmem_data.start && MACHINE_IS_VM)
707 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
708 crashk_res.start = crash_base;
709 crashk_res.end = crash_base + crash_size - 1;
710 memblock_remove(crash_base, crash_size);
711 pr_info("Reserving %lluMB of memory at %lluMB "
712 "for crashkernel (System RAM: %luMB)\n",
713 crash_size >> 20, crash_base >> 20,
714 (unsigned long)memblock.memory.total_size >> 20);
715 os_info_crashkernel_add(crash_base, crash_size);
716 #endif
717 }
718
719 /*
720 * Reserve the initrd from being used by memblock
721 */
reserve_initrd(void)722 static void __init reserve_initrd(void)
723 {
724 #ifdef CONFIG_BLK_DEV_INITRD
725 if (!initrd_data.start || !initrd_data.size)
726 return;
727 initrd_start = (unsigned long)__va(initrd_data.start);
728 initrd_end = initrd_start + initrd_data.size;
729 memblock_reserve(initrd_data.start, initrd_data.size);
730 #endif
731 }
732
733 /*
734 * Reserve the memory area used to pass the certificate lists
735 */
reserve_certificate_list(void)736 static void __init reserve_certificate_list(void)
737 {
738 if (ipl_cert_list_addr)
739 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
740 }
741
reserve_mem_detect_info(void)742 static void __init reserve_mem_detect_info(void)
743 {
744 unsigned long start, size;
745
746 get_mem_detect_reserved(&start, &size);
747 if (size)
748 memblock_reserve(start, size);
749 }
750
free_mem_detect_info(void)751 static void __init free_mem_detect_info(void)
752 {
753 unsigned long start, size;
754
755 get_mem_detect_reserved(&start, &size);
756 if (size)
757 memblock_phys_free(start, size);
758 }
759
get_mem_info_source(void)760 static const char * __init get_mem_info_source(void)
761 {
762 switch (mem_detect.info_source) {
763 case MEM_DETECT_SCLP_STOR_INFO:
764 return "sclp storage info";
765 case MEM_DETECT_DIAG260:
766 return "diag260";
767 case MEM_DETECT_SCLP_READ_INFO:
768 return "sclp read info";
769 case MEM_DETECT_BIN_SEARCH:
770 return "binary search";
771 }
772 return "none";
773 }
774
memblock_add_mem_detect_info(void)775 static void __init memblock_add_mem_detect_info(void)
776 {
777 unsigned long start, end;
778 int i;
779
780 pr_debug("physmem info source: %s (%hhd)\n",
781 get_mem_info_source(), mem_detect.info_source);
782 /* keep memblock lists close to the kernel */
783 memblock_set_bottom_up(true);
784 for_each_mem_detect_block(i, &start, &end) {
785 memblock_add(start, end - start);
786 memblock_physmem_add(start, end - start);
787 }
788 memblock_set_bottom_up(false);
789 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
790 }
791
792 /*
793 * Check for initrd being in usable memory
794 */
check_initrd(void)795 static void __init check_initrd(void)
796 {
797 #ifdef CONFIG_BLK_DEV_INITRD
798 if (initrd_data.start && initrd_data.size &&
799 !memblock_is_region_memory(initrd_data.start, initrd_data.size)) {
800 pr_err("The initial RAM disk does not fit into the memory\n");
801 memblock_phys_free(initrd_data.start, initrd_data.size);
802 initrd_start = initrd_end = 0;
803 }
804 #endif
805 }
806
807 /*
808 * Reserve memory used for lowcore/command line/kernel image.
809 */
reserve_kernel(void)810 static void __init reserve_kernel(void)
811 {
812 memblock_reserve(0, STARTUP_NORMAL_OFFSET);
813 memblock_reserve(OLDMEM_BASE, sizeof(unsigned long));
814 memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long));
815 memblock_reserve(__amode31_base, __eamode31 - __samode31);
816 memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP);
817 memblock_reserve(__pa(_stext), _end - _stext);
818 }
819
setup_memory(void)820 static void __init setup_memory(void)
821 {
822 phys_addr_t start, end;
823 u64 i;
824
825 /*
826 * Init storage key for present memory
827 */
828 for_each_mem_range(i, &start, &end)
829 storage_key_init_range(start, end);
830
831 psw_set_key(PAGE_DEFAULT_KEY);
832 }
833
relocate_amode31_section(void)834 static void __init relocate_amode31_section(void)
835 {
836 unsigned long amode31_size = __eamode31 - __samode31;
837 long amode31_offset = __amode31_base - __samode31;
838 long *ptr;
839
840 pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size);
841
842 /* Move original AMODE31 section to the new one */
843 memmove((void *)__amode31_base, (void *)__samode31, amode31_size);
844 /* Zero out the old AMODE31 section to catch invalid accesses within it */
845 memset((void *)__samode31, 0, amode31_size);
846
847 /* Update all AMODE31 region references */
848 for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++)
849 *ptr += amode31_offset;
850 }
851
852 /* This must be called after AMODE31 relocation */
setup_cr(void)853 static void __init setup_cr(void)
854 {
855 union ctlreg2 cr2;
856 union ctlreg5 cr5;
857 union ctlreg15 cr15;
858
859 __ctl_duct[1] = (unsigned long)__ctl_aste;
860 __ctl_duct[2] = (unsigned long)__ctl_aste;
861 __ctl_duct[4] = (unsigned long)__ctl_duald;
862
863 /* Update control registers CR2, CR5 and CR15 */
864 __ctl_store(cr2.val, 2, 2);
865 __ctl_store(cr5.val, 5, 5);
866 __ctl_store(cr15.val, 15, 15);
867 cr2.ducto = (unsigned long)__ctl_duct >> 6;
868 cr5.pasteo = (unsigned long)__ctl_duct >> 6;
869 cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
870 __ctl_load(cr2.val, 2, 2);
871 __ctl_load(cr5.val, 5, 5);
872 __ctl_load(cr15.val, 15, 15);
873 }
874
875 /*
876 * Add system information as device randomness
877 */
setup_randomness(void)878 static void __init setup_randomness(void)
879 {
880 struct sysinfo_3_2_2 *vmms;
881
882 vmms = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
883 if (!vmms)
884 panic("Failed to allocate memory for sysinfo structure\n");
885 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
886 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
887 memblock_free(vmms, PAGE_SIZE);
888
889 if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
890 static_branch_enable(&s390_arch_random_available);
891 }
892
893 /*
894 * Find the correct size for the task_struct. This depends on
895 * the size of the struct fpu at the end of the thread_struct
896 * which is embedded in the task_struct.
897 */
setup_task_size(void)898 static void __init setup_task_size(void)
899 {
900 int task_size = sizeof(struct task_struct);
901
902 if (!MACHINE_HAS_VX) {
903 task_size -= sizeof(__vector128) * __NUM_VXRS;
904 task_size += sizeof(freg_t) * __NUM_FPRS;
905 }
906 arch_task_struct_size = task_size;
907 }
908
909 /*
910 * Issue diagnose 318 to set the control program name and
911 * version codes.
912 */
setup_control_program_code(void)913 static void __init setup_control_program_code(void)
914 {
915 union diag318_info diag318_info = {
916 .cpnc = CPNC_LINUX,
917 .cpvc = 0,
918 };
919
920 if (!sclp.has_diag318)
921 return;
922
923 diag_stat_inc(DIAG_STAT_X318);
924 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
925 }
926
927 /*
928 * Print the component list from the IPL report
929 */
log_component_list(void)930 static void __init log_component_list(void)
931 {
932 struct ipl_rb_component_entry *ptr, *end;
933 char *str;
934
935 if (!early_ipl_comp_list_addr)
936 return;
937 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
938 pr_info("Linux is running with Secure-IPL enabled\n");
939 else
940 pr_info("Linux is running with Secure-IPL disabled\n");
941 ptr = (void *) early_ipl_comp_list_addr;
942 end = (void *) ptr + early_ipl_comp_list_size;
943 pr_info("The IPL report contains the following components:\n");
944 while (ptr < end) {
945 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
946 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
947 str = "signed, verified";
948 else
949 str = "signed, verification failed";
950 } else {
951 str = "not signed";
952 }
953 pr_info("%016llx - %016llx (%s)\n",
954 ptr->addr, ptr->addr + ptr->len, str);
955 ptr++;
956 }
957 }
958
959 /*
960 * Setup function called from init/main.c just after the banner
961 * was printed.
962 */
963
setup_arch(char ** cmdline_p)964 void __init setup_arch(char **cmdline_p)
965 {
966 /*
967 * print what head.S has found out about the machine
968 */
969 if (MACHINE_IS_VM)
970 pr_info("Linux is running as a z/VM "
971 "guest operating system in 64-bit mode\n");
972 else if (MACHINE_IS_KVM)
973 pr_info("Linux is running under KVM in 64-bit mode\n");
974 else if (MACHINE_IS_LPAR)
975 pr_info("Linux is running natively in 64-bit mode\n");
976 else
977 pr_info("Linux is running as a guest in 64-bit mode\n");
978
979 log_component_list();
980
981 /* Have one command line that is parsed and saved in /proc/cmdline */
982 /* boot_command_line has been already set up in early.c */
983 *cmdline_p = boot_command_line;
984
985 ROOT_DEV = Root_RAM0;
986
987 setup_initial_init_mm(_text, _etext, _edata, _end);
988
989 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
990 nospec_auto_detect();
991
992 jump_label_init();
993 parse_early_param();
994 #ifdef CONFIG_CRASH_DUMP
995 /* Deactivate elfcorehdr= kernel parameter */
996 elfcorehdr_addr = ELFCORE_ADDR_MAX;
997 #endif
998
999 os_info_init();
1000 setup_ipl();
1001 setup_task_size();
1002 setup_control_program_code();
1003
1004 /* Do some memory reservations *before* memory is added to memblock */
1005 reserve_kernel();
1006 reserve_initrd();
1007 reserve_certificate_list();
1008 reserve_mem_detect_info();
1009 memblock_set_current_limit(ident_map_size);
1010 memblock_allow_resize();
1011
1012 /* Get information about *all* installed memory */
1013 memblock_add_mem_detect_info();
1014
1015 free_mem_detect_info();
1016 setup_memory_end();
1017 memblock_dump_all();
1018 setup_memory();
1019
1020 relocate_amode31_section();
1021 setup_cr();
1022 setup_uv();
1023 dma_contiguous_reserve(ident_map_size);
1024 vmcp_cma_reserve();
1025 if (MACHINE_HAS_EDAT2)
1026 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1027
1028 check_initrd();
1029 reserve_crashkernel();
1030 #ifdef CONFIG_CRASH_DUMP
1031 /*
1032 * Be aware that smp_save_dump_secondary_cpus() triggers a system reset.
1033 * Therefore CPU and device initialization should be done afterwards.
1034 */
1035 smp_save_dump_secondary_cpus();
1036 #endif
1037
1038 setup_resources();
1039 setup_lowcore_dat_off();
1040 smp_fill_possible_mask();
1041 cpu_detect_mhz_feature();
1042 cpu_init();
1043 numa_setup();
1044 smp_detect_cpus();
1045 topology_init_early();
1046
1047 if (test_facility(193))
1048 static_branch_enable(&cpu_has_bear);
1049
1050 /*
1051 * Create kernel page tables and switch to virtual addressing.
1052 */
1053 paging_init();
1054 memcpy_real_init();
1055 /*
1056 * After paging_init created the kernel page table, the new PSWs
1057 * in lowcore can now run with DAT enabled.
1058 */
1059 setup_lowcore_dat_on();
1060 #ifdef CONFIG_CRASH_DUMP
1061 smp_save_dump_ipl_cpu();
1062 #endif
1063
1064 /* Setup default console */
1065 conmode_default();
1066 set_preferred_console();
1067
1068 apply_alternative_instructions();
1069 if (IS_ENABLED(CONFIG_EXPOLINE))
1070 nospec_init_branches();
1071
1072 /* Setup zfcp/nvme dump support */
1073 setup_zfcpdump();
1074
1075 /* Add system specific data to the random pool */
1076 setup_randomness();
1077 }
1078