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-contiguous.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
53 #include <asm/boot_data.h>
54 #include <asm/ipl.h>
55 #include <asm/facility.h>
56 #include <asm/smp.h>
57 #include <asm/mmu_context.h>
58 #include <asm/cpcmd.h>
59 #include <asm/lowcore.h>
60 #include <asm/nmi.h>
61 #include <asm/irq.h>
62 #include <asm/page.h>
63 #include <asm/ptrace.h>
64 #include <asm/sections.h>
65 #include <asm/ebcdic.h>
66 #include <asm/diag.h>
67 #include <asm/os_info.h>
68 #include <asm/sclp.h>
69 #include <asm/stacktrace.h>
70 #include <asm/sysinfo.h>
71 #include <asm/numa.h>
72 #include <asm/alternative.h>
73 #include <asm/nospec-branch.h>
74 #include <asm/mem_detect.h>
75 #include <asm/uv.h>
76 #include "entry.h"
77
78 /*
79 * Machine setup..
80 */
81 unsigned int console_mode = 0;
82 EXPORT_SYMBOL(console_mode);
83
84 unsigned int console_devno = -1;
85 EXPORT_SYMBOL(console_devno);
86
87 unsigned int console_irq = -1;
88 EXPORT_SYMBOL(console_irq);
89
90 unsigned long elf_hwcap __read_mostly = 0;
91 char elf_platform[ELF_PLATFORM_SIZE];
92
93 unsigned long int_hwcap = 0;
94
95 #ifdef CONFIG_PROTECTED_VIRTUALIZATION_GUEST
96 int __bootdata_preserved(prot_virt_guest);
97 #endif
98
99 int __bootdata(noexec_disabled);
100 int __bootdata(memory_end_set);
101 unsigned long __bootdata(memory_end);
102 unsigned long __bootdata(vmalloc_size);
103 unsigned long __bootdata(max_physmem_end);
104 struct mem_detect_info __bootdata(mem_detect);
105
106 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
107 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
108 unsigned long __bootdata_preserved(__swsusp_reset_dma);
109 unsigned long __bootdata_preserved(__stext_dma);
110 unsigned long __bootdata_preserved(__etext_dma);
111 unsigned long __bootdata_preserved(__sdma);
112 unsigned long __bootdata_preserved(__edma);
113 unsigned long __bootdata_preserved(__kaslr_offset);
114
115 unsigned long VMALLOC_START;
116 EXPORT_SYMBOL(VMALLOC_START);
117
118 unsigned long VMALLOC_END;
119 EXPORT_SYMBOL(VMALLOC_END);
120
121 struct page *vmemmap;
122 EXPORT_SYMBOL(vmemmap);
123
124 unsigned long MODULES_VADDR;
125 unsigned long MODULES_END;
126
127 /* An array with a pointer to the lowcore of every CPU. */
128 struct lowcore *lowcore_ptr[NR_CPUS];
129 EXPORT_SYMBOL(lowcore_ptr);
130
131 /*
132 * This is set up by the setup-routine at boot-time
133 * for S390 need to find out, what we have to setup
134 * using address 0x10400 ...
135 */
136
137 #include <asm/setup.h>
138
139 /*
140 * condev= and conmode= setup parameter.
141 */
142
condev_setup(char * str)143 static int __init condev_setup(char *str)
144 {
145 int vdev;
146
147 vdev = simple_strtoul(str, &str, 0);
148 if (vdev >= 0 && vdev < 65536) {
149 console_devno = vdev;
150 console_irq = -1;
151 }
152 return 1;
153 }
154
155 __setup("condev=", condev_setup);
156
set_preferred_console(void)157 static void __init set_preferred_console(void)
158 {
159 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
160 add_preferred_console("ttyS", 0, NULL);
161 else if (CONSOLE_IS_3270)
162 add_preferred_console("tty3270", 0, NULL);
163 else if (CONSOLE_IS_VT220)
164 add_preferred_console("ttyS", 1, NULL);
165 else if (CONSOLE_IS_HVC)
166 add_preferred_console("hvc", 0, NULL);
167 }
168
conmode_setup(char * str)169 static int __init conmode_setup(char *str)
170 {
171 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
172 if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
173 SET_CONSOLE_SCLP;
174 #endif
175 #if defined(CONFIG_TN3215_CONSOLE)
176 if (!strcmp(str, "3215"))
177 SET_CONSOLE_3215;
178 #endif
179 #if defined(CONFIG_TN3270_CONSOLE)
180 if (!strcmp(str, "3270"))
181 SET_CONSOLE_3270;
182 #endif
183 set_preferred_console();
184 return 1;
185 }
186
187 __setup("conmode=", conmode_setup);
188
conmode_default(void)189 static void __init conmode_default(void)
190 {
191 char query_buffer[1024];
192 char *ptr;
193
194 if (MACHINE_IS_VM) {
195 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
196 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
197 ptr = strstr(query_buffer, "SUBCHANNEL =");
198 console_irq = simple_strtoul(ptr + 13, NULL, 16);
199 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
200 ptr = strstr(query_buffer, "CONMODE");
201 /*
202 * Set the conmode to 3215 so that the device recognition
203 * will set the cu_type of the console to 3215. If the
204 * conmode is 3270 and we don't set it back then both
205 * 3215 and the 3270 driver will try to access the console
206 * device (3215 as console and 3270 as normal tty).
207 */
208 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
209 if (ptr == NULL) {
210 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
211 SET_CONSOLE_SCLP;
212 #endif
213 return;
214 }
215 if (str_has_prefix(ptr + 8, "3270")) {
216 #if defined(CONFIG_TN3270_CONSOLE)
217 SET_CONSOLE_3270;
218 #elif defined(CONFIG_TN3215_CONSOLE)
219 SET_CONSOLE_3215;
220 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
221 SET_CONSOLE_SCLP;
222 #endif
223 } else if (str_has_prefix(ptr + 8, "3215")) {
224 #if defined(CONFIG_TN3215_CONSOLE)
225 SET_CONSOLE_3215;
226 #elif defined(CONFIG_TN3270_CONSOLE)
227 SET_CONSOLE_3270;
228 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
229 SET_CONSOLE_SCLP;
230 #endif
231 }
232 } else if (MACHINE_IS_KVM) {
233 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
234 SET_CONSOLE_VT220;
235 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
236 SET_CONSOLE_SCLP;
237 else
238 SET_CONSOLE_HVC;
239 } else {
240 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
241 SET_CONSOLE_SCLP;
242 #endif
243 }
244 if (IS_ENABLED(CONFIG_VT) && IS_ENABLED(CONFIG_DUMMY_CONSOLE))
245 conswitchp = &dummy_con;
246 }
247
248 #ifdef CONFIG_CRASH_DUMP
setup_zfcpdump(void)249 static void __init setup_zfcpdump(void)
250 {
251 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
252 return;
253 if (OLDMEM_BASE)
254 return;
255 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
256 console_loglevel = 2;
257 }
258 #else
setup_zfcpdump(void)259 static inline void setup_zfcpdump(void) {}
260 #endif /* CONFIG_CRASH_DUMP */
261
262 /*
263 * Reboot, halt and power_off stubs. They just call _machine_restart,
264 * _machine_halt or _machine_power_off.
265 */
266
machine_restart(char * command)267 void machine_restart(char *command)
268 {
269 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
270 /*
271 * Only unblank the console if we are called in enabled
272 * context or a bust_spinlocks cleared the way for us.
273 */
274 console_unblank();
275 _machine_restart(command);
276 }
277
machine_halt(void)278 void machine_halt(void)
279 {
280 if (!in_interrupt() || oops_in_progress)
281 /*
282 * Only unblank the console if we are called in enabled
283 * context or a bust_spinlocks cleared the way for us.
284 */
285 console_unblank();
286 _machine_halt();
287 }
288
machine_power_off(void)289 void machine_power_off(void)
290 {
291 if (!in_interrupt() || oops_in_progress)
292 /*
293 * Only unblank the console if we are called in enabled
294 * context or a bust_spinlocks cleared the way for us.
295 */
296 console_unblank();
297 _machine_power_off();
298 }
299
300 /*
301 * Dummy power off function.
302 */
303 void (*pm_power_off)(void) = machine_power_off;
304 EXPORT_SYMBOL_GPL(pm_power_off);
305
306 void *restart_stack __section(.data);
307
stack_alloc(void)308 unsigned long stack_alloc(void)
309 {
310 #ifdef CONFIG_VMAP_STACK
311 return (unsigned long)
312 __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
313 VMALLOC_START, VMALLOC_END,
314 THREADINFO_GFP,
315 PAGE_KERNEL, 0, NUMA_NO_NODE,
316 __builtin_return_address(0));
317 #else
318 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
319 #endif
320 }
321
stack_free(unsigned long stack)322 void stack_free(unsigned long stack)
323 {
324 #ifdef CONFIG_VMAP_STACK
325 vfree((void *) stack);
326 #else
327 free_pages(stack, THREAD_SIZE_ORDER);
328 #endif
329 }
330
arch_early_irq_init(void)331 int __init arch_early_irq_init(void)
332 {
333 unsigned long stack;
334
335 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
336 if (!stack)
337 panic("Couldn't allocate async stack");
338 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
339 return 0;
340 }
341
async_stack_realloc(void)342 static int __init async_stack_realloc(void)
343 {
344 unsigned long old, new;
345
346 old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
347 new = stack_alloc();
348 if (!new)
349 panic("Couldn't allocate async stack");
350 S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
351 free_pages(old, THREAD_SIZE_ORDER);
352 return 0;
353 }
354 early_initcall(async_stack_realloc);
355
arch_call_rest_init(void)356 void __init arch_call_rest_init(void)
357 {
358 struct stack_frame *frame;
359 unsigned long stack;
360
361 stack = stack_alloc();
362 if (!stack)
363 panic("Couldn't allocate kernel stack");
364 current->stack = (void *) stack;
365 #ifdef CONFIG_VMAP_STACK
366 current->stack_vm_area = (void *) stack;
367 #endif
368 set_task_stack_end_magic(current);
369 stack += STACK_INIT_OFFSET;
370 S390_lowcore.kernel_stack = stack;
371 frame = (struct stack_frame *) stack;
372 memset(frame, 0, sizeof(*frame));
373 /* Branch to rest_init on the new stack, never returns */
374 asm volatile(
375 " la 15,0(%[_frame])\n"
376 " jg rest_init\n"
377 : : [_frame] "a" (frame));
378 }
379
setup_lowcore_dat_off(void)380 static void __init setup_lowcore_dat_off(void)
381 {
382 struct lowcore *lc;
383
384 /*
385 * Setup lowcore for boot cpu
386 */
387 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
388 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
389 if (!lc)
390 panic("%s: Failed to allocate %zu bytes align=%zx\n",
391 __func__, sizeof(*lc), sizeof(*lc));
392
393 lc->restart_psw.mask = PSW_KERNEL_BITS;
394 lc->restart_psw.addr = (unsigned long) restart_int_handler;
395 lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
396 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
397 lc->svc_new_psw.mask = PSW_KERNEL_BITS |
398 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
399 lc->svc_new_psw.addr = (unsigned long) system_call;
400 lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
401 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
402 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
403 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
404 lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
405 lc->io_new_psw.addr = (unsigned long) io_int_handler;
406 lc->clock_comparator = clock_comparator_max;
407 lc->nodat_stack = ((unsigned long) &init_thread_union)
408 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
409 lc->current_task = (unsigned long)&init_task;
410 lc->lpp = LPP_MAGIC;
411 lc->machine_flags = S390_lowcore.machine_flags;
412 lc->preempt_count = S390_lowcore.preempt_count;
413 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
414 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
415 sizeof(lc->stfle_fac_list));
416 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
417 sizeof(lc->alt_stfle_fac_list));
418 nmi_alloc_boot_cpu(lc);
419 vdso_alloc_boot_cpu(lc);
420 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
421 lc->async_enter_timer = S390_lowcore.async_enter_timer;
422 lc->exit_timer = S390_lowcore.exit_timer;
423 lc->user_timer = S390_lowcore.user_timer;
424 lc->system_timer = S390_lowcore.system_timer;
425 lc->steal_timer = S390_lowcore.steal_timer;
426 lc->last_update_timer = S390_lowcore.last_update_timer;
427 lc->last_update_clock = S390_lowcore.last_update_clock;
428
429 /*
430 * Allocate the global restart stack which is the same for
431 * all CPUs in cast *one* of them does a PSW restart.
432 */
433 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
434 if (!restart_stack)
435 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
436 __func__, THREAD_SIZE, THREAD_SIZE);
437 restart_stack += STACK_INIT_OFFSET;
438
439 /*
440 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
441 * restart data to the absolute zero lowcore. This is necessary if
442 * PSW restart is done on an offline CPU that has lowcore zero.
443 */
444 lc->restart_stack = (unsigned long) restart_stack;
445 lc->restart_fn = (unsigned long) do_restart;
446 lc->restart_data = 0;
447 lc->restart_source = -1UL;
448
449 /* Setup absolute zero lowcore */
450 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
451 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
452 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
453 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
454 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
455
456 lc->spinlock_lockval = arch_spin_lockval(0);
457 lc->spinlock_index = 0;
458 arch_spin_lock_setup(0);
459 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
460
461 set_prefix((u32)(unsigned long) lc);
462 lowcore_ptr[0] = lc;
463 }
464
setup_lowcore_dat_on(void)465 static void __init setup_lowcore_dat_on(void)
466 {
467 __ctl_clear_bit(0, 28);
468 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
469 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
470 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
471 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
472 __ctl_set_bit(0, 28);
473 }
474
475 static struct resource code_resource = {
476 .name = "Kernel code",
477 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
478 };
479
480 static struct resource data_resource = {
481 .name = "Kernel data",
482 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
483 };
484
485 static struct resource bss_resource = {
486 .name = "Kernel bss",
487 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
488 };
489
490 static struct resource __initdata *standard_resources[] = {
491 &code_resource,
492 &data_resource,
493 &bss_resource,
494 };
495
setup_resources(void)496 static void __init setup_resources(void)
497 {
498 struct resource *res, *std_res, *sub_res;
499 struct memblock_region *reg;
500 int j;
501
502 code_resource.start = (unsigned long) _text;
503 code_resource.end = (unsigned long) _etext - 1;
504 data_resource.start = (unsigned long) _etext;
505 data_resource.end = (unsigned long) _edata - 1;
506 bss_resource.start = (unsigned long) __bss_start;
507 bss_resource.end = (unsigned long) __bss_stop - 1;
508
509 for_each_memblock(memory, reg) {
510 res = memblock_alloc(sizeof(*res), 8);
511 if (!res)
512 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
513 __func__, sizeof(*res), 8);
514 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
515
516 res->name = "System RAM";
517 res->start = reg->base;
518 res->end = reg->base + reg->size - 1;
519 request_resource(&iomem_resource, res);
520
521 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
522 std_res = standard_resources[j];
523 if (std_res->start < res->start ||
524 std_res->start > res->end)
525 continue;
526 if (std_res->end > res->end) {
527 sub_res = memblock_alloc(sizeof(*sub_res), 8);
528 if (!sub_res)
529 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
530 __func__, sizeof(*sub_res), 8);
531 *sub_res = *std_res;
532 sub_res->end = res->end;
533 std_res->start = res->end + 1;
534 request_resource(res, sub_res);
535 } else {
536 request_resource(res, std_res);
537 }
538 }
539 }
540 #ifdef CONFIG_CRASH_DUMP
541 /*
542 * Re-add removed crash kernel memory as reserved memory. This makes
543 * sure it will be mapped with the identity mapping and struct pages
544 * will be created, so it can be resized later on.
545 * However add it later since the crash kernel resource should not be
546 * part of the System RAM resource.
547 */
548 if (crashk_res.end) {
549 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
550 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
551 insert_resource(&iomem_resource, &crashk_res);
552 }
553 #endif
554 }
555
setup_memory_end(void)556 static void __init setup_memory_end(void)
557 {
558 unsigned long vmax, tmp;
559
560 /* Choose kernel address space layout: 3 or 4 levels. */
561 if (IS_ENABLED(CONFIG_KASAN)) {
562 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
563 ? _REGION1_SIZE
564 : _REGION2_SIZE;
565 } else {
566 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
567 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
568 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
569 vmax = _REGION2_SIZE; /* 3-level kernel page table */
570 else
571 vmax = _REGION1_SIZE; /* 4-level kernel page table */
572 }
573
574 /* module area is at the end of the kernel address space. */
575 MODULES_END = vmax;
576 MODULES_VADDR = MODULES_END - MODULES_LEN;
577 VMALLOC_END = MODULES_VADDR;
578 VMALLOC_START = VMALLOC_END - vmalloc_size;
579
580 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
581 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
582 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
583 tmp = SECTION_ALIGN_UP(tmp);
584 tmp = VMALLOC_START - tmp * sizeof(struct page);
585 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
586 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
587 vmemmap = (struct page *) tmp;
588
589 /* Take care that memory_end is set and <= vmemmap */
590 memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
591 #ifdef CONFIG_KASAN
592 /* fit in kasan shadow memory region between 1:1 and vmemmap */
593 memory_end = min(memory_end, KASAN_SHADOW_START);
594 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
595 #endif
596 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
597 memblock_remove(memory_end, ULONG_MAX);
598
599 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
600 }
601
602 #ifdef CONFIG_CRASH_DUMP
603
604 /*
605 * When kdump is enabled, we have to ensure that no memory from
606 * the area [0 - crashkernel memory size] and
607 * [crashk_res.start - crashk_res.end] is set offline.
608 */
kdump_mem_notifier(struct notifier_block * nb,unsigned long action,void * data)609 static int kdump_mem_notifier(struct notifier_block *nb,
610 unsigned long action, void *data)
611 {
612 struct memory_notify *arg = data;
613
614 if (action != MEM_GOING_OFFLINE)
615 return NOTIFY_OK;
616 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
617 return NOTIFY_BAD;
618 if (arg->start_pfn > PFN_DOWN(crashk_res.end))
619 return NOTIFY_OK;
620 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
621 return NOTIFY_OK;
622 return NOTIFY_BAD;
623 }
624
625 static struct notifier_block kdump_mem_nb = {
626 .notifier_call = kdump_mem_notifier,
627 };
628
629 #endif
630
631 /*
632 * Make sure that the area behind memory_end is protected
633 */
reserve_memory_end(void)634 static void reserve_memory_end(void)
635 {
636 if (memory_end_set)
637 memblock_reserve(memory_end, ULONG_MAX);
638 }
639
640 /*
641 * Make sure that oldmem, where the dump is stored, is protected
642 */
reserve_oldmem(void)643 static void reserve_oldmem(void)
644 {
645 #ifdef CONFIG_CRASH_DUMP
646 if (OLDMEM_BASE)
647 /* Forget all memory above the running kdump system */
648 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
649 #endif
650 }
651
652 /*
653 * Make sure that oldmem, where the dump is stored, is protected
654 */
remove_oldmem(void)655 static void remove_oldmem(void)
656 {
657 #ifdef CONFIG_CRASH_DUMP
658 if (OLDMEM_BASE)
659 /* Forget all memory above the running kdump system */
660 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
661 #endif
662 }
663
664 /*
665 * Reserve memory for kdump kernel to be loaded with kexec
666 */
reserve_crashkernel(void)667 static void __init reserve_crashkernel(void)
668 {
669 #ifdef CONFIG_CRASH_DUMP
670 unsigned long long crash_base, crash_size;
671 phys_addr_t low, high;
672 int rc;
673
674 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
675 &crash_base);
676
677 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
678 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
679 if (rc || crash_size == 0)
680 return;
681
682 if (memblock.memory.regions[0].size < crash_size) {
683 pr_info("crashkernel reservation failed: %s\n",
684 "first memory chunk must be at least crashkernel size");
685 return;
686 }
687
688 low = crash_base ?: OLDMEM_BASE;
689 high = low + crash_size;
690 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
691 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
692 crash_base = low;
693 } else {
694 /* Find suitable area in free memory */
695 low = max_t(unsigned long, crash_size, sclp.hsa_size);
696 high = crash_base ? crash_base + crash_size : ULONG_MAX;
697
698 if (crash_base && crash_base < low) {
699 pr_info("crashkernel reservation failed: %s\n",
700 "crash_base too low");
701 return;
702 }
703 low = crash_base ?: low;
704 crash_base = memblock_find_in_range(low, high, crash_size,
705 KEXEC_CRASH_MEM_ALIGN);
706 }
707
708 if (!crash_base) {
709 pr_info("crashkernel reservation failed: %s\n",
710 "no suitable area found");
711 return;
712 }
713
714 if (register_memory_notifier(&kdump_mem_nb))
715 return;
716
717 if (!OLDMEM_BASE && MACHINE_IS_VM)
718 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
719 crashk_res.start = crash_base;
720 crashk_res.end = crash_base + crash_size - 1;
721 memblock_remove(crash_base, crash_size);
722 pr_info("Reserving %lluMB of memory at %lluMB "
723 "for crashkernel (System RAM: %luMB)\n",
724 crash_size >> 20, crash_base >> 20,
725 (unsigned long)memblock.memory.total_size >> 20);
726 os_info_crashkernel_add(crash_base, crash_size);
727 #endif
728 }
729
730 /*
731 * Reserve the initrd from being used by memblock
732 */
reserve_initrd(void)733 static void __init reserve_initrd(void)
734 {
735 #ifdef CONFIG_BLK_DEV_INITRD
736 if (!INITRD_START || !INITRD_SIZE)
737 return;
738 initrd_start = INITRD_START;
739 initrd_end = initrd_start + INITRD_SIZE;
740 memblock_reserve(INITRD_START, INITRD_SIZE);
741 #endif
742 }
743
744 /*
745 * Reserve the memory area used to pass the certificate lists
746 */
reserve_certificate_list(void)747 static void __init reserve_certificate_list(void)
748 {
749 if (ipl_cert_list_addr)
750 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
751 }
752
reserve_mem_detect_info(void)753 static void __init reserve_mem_detect_info(void)
754 {
755 unsigned long start, size;
756
757 get_mem_detect_reserved(&start, &size);
758 if (size)
759 memblock_reserve(start, size);
760 }
761
free_mem_detect_info(void)762 static void __init free_mem_detect_info(void)
763 {
764 unsigned long start, size;
765
766 get_mem_detect_reserved(&start, &size);
767 if (size)
768 memblock_free(start, size);
769 }
770
memblock_physmem_add(phys_addr_t start,phys_addr_t size)771 static void __init memblock_physmem_add(phys_addr_t start, phys_addr_t size)
772 {
773 memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n",
774 start, start + size - 1);
775 memblock_add_range(&memblock.memory, start, size, 0, 0);
776 memblock_add_range(&memblock.physmem, start, size, 0, 0);
777 }
778
get_mem_info_source(void)779 static const char * __init get_mem_info_source(void)
780 {
781 switch (mem_detect.info_source) {
782 case MEM_DETECT_SCLP_STOR_INFO:
783 return "sclp storage info";
784 case MEM_DETECT_DIAG260:
785 return "diag260";
786 case MEM_DETECT_SCLP_READ_INFO:
787 return "sclp read info";
788 case MEM_DETECT_BIN_SEARCH:
789 return "binary search";
790 }
791 return "none";
792 }
793
memblock_add_mem_detect_info(void)794 static void __init memblock_add_mem_detect_info(void)
795 {
796 unsigned long start, end;
797 int i;
798
799 memblock_dbg("physmem info source: %s (%hhd)\n",
800 get_mem_info_source(), mem_detect.info_source);
801 /* keep memblock lists close to the kernel */
802 memblock_set_bottom_up(true);
803 for_each_mem_detect_block(i, &start, &end)
804 memblock_physmem_add(start, end - start);
805 memblock_set_bottom_up(false);
806 memblock_dump_all();
807 }
808
809 /*
810 * Check for initrd being in usable memory
811 */
check_initrd(void)812 static void __init check_initrd(void)
813 {
814 #ifdef CONFIG_BLK_DEV_INITRD
815 if (INITRD_START && INITRD_SIZE &&
816 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
817 pr_err("The initial RAM disk does not fit into the memory\n");
818 memblock_free(INITRD_START, INITRD_SIZE);
819 initrd_start = initrd_end = 0;
820 }
821 #endif
822 }
823
824 /*
825 * Reserve memory used for lowcore/command line/kernel image.
826 */
reserve_kernel(void)827 static void __init reserve_kernel(void)
828 {
829 unsigned long start_pfn = PFN_UP(__pa(_end));
830
831 memblock_reserve(0, HEAD_END);
832 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
833 - (unsigned long)_stext);
834 memblock_reserve(__sdma, __edma - __sdma);
835 }
836
setup_memory(void)837 static void __init setup_memory(void)
838 {
839 struct memblock_region *reg;
840
841 /*
842 * Init storage key for present memory
843 */
844 for_each_memblock(memory, reg) {
845 storage_key_init_range(reg->base, reg->base + reg->size);
846 }
847 psw_set_key(PAGE_DEFAULT_KEY);
848
849 /* Only cosmetics */
850 memblock_enforce_memory_limit(memblock_end_of_DRAM());
851 }
852
853 /*
854 * Setup hardware capabilities.
855 */
setup_hwcaps(void)856 static int __init setup_hwcaps(void)
857 {
858 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
859 struct cpuid cpu_id;
860 int i;
861
862 /*
863 * The store facility list bits numbers as found in the principles
864 * of operation are numbered with bit 1UL<<31 as number 0 to
865 * bit 1UL<<0 as number 31.
866 * Bit 0: instructions named N3, "backported" to esa-mode
867 * Bit 2: z/Architecture mode is active
868 * Bit 7: the store-facility-list-extended facility is installed
869 * Bit 17: the message-security assist is installed
870 * Bit 19: the long-displacement facility is installed
871 * Bit 21: the extended-immediate facility is installed
872 * Bit 22: extended-translation facility 3 is installed
873 * Bit 30: extended-translation facility 3 enhancement facility
874 * These get translated to:
875 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
876 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
877 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
878 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
879 */
880 for (i = 0; i < 6; i++)
881 if (test_facility(stfl_bits[i]))
882 elf_hwcap |= 1UL << i;
883
884 if (test_facility(22) && test_facility(30))
885 elf_hwcap |= HWCAP_S390_ETF3EH;
886
887 /*
888 * Check for additional facilities with store-facility-list-extended.
889 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
890 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
891 * as stored by stfl, bits 32-xxx contain additional facilities.
892 * How many facility words are stored depends on the number of
893 * doublewords passed to the instruction. The additional facilities
894 * are:
895 * Bit 42: decimal floating point facility is installed
896 * Bit 44: perform floating point operation facility is installed
897 * translated to:
898 * HWCAP_S390_DFP bit 6 (42 && 44).
899 */
900 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
901 elf_hwcap |= HWCAP_S390_DFP;
902
903 /*
904 * Huge page support HWCAP_S390_HPAGE is bit 7.
905 */
906 if (MACHINE_HAS_EDAT1)
907 elf_hwcap |= HWCAP_S390_HPAGE;
908
909 /*
910 * 64-bit register support for 31-bit processes
911 * HWCAP_S390_HIGH_GPRS is bit 9.
912 */
913 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
914
915 /*
916 * Transactional execution support HWCAP_S390_TE is bit 10.
917 */
918 if (MACHINE_HAS_TE)
919 elf_hwcap |= HWCAP_S390_TE;
920
921 /*
922 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
923 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
924 * instead of facility bit 129.
925 */
926 if (MACHINE_HAS_VX) {
927 elf_hwcap |= HWCAP_S390_VXRS;
928 if (test_facility(134))
929 elf_hwcap |= HWCAP_S390_VXRS_EXT;
930 if (test_facility(135))
931 elf_hwcap |= HWCAP_S390_VXRS_BCD;
932 if (test_facility(148))
933 elf_hwcap |= HWCAP_S390_VXRS_EXT2;
934 if (test_facility(152))
935 elf_hwcap |= HWCAP_S390_VXRS_PDE;
936 }
937 if (test_facility(150))
938 elf_hwcap |= HWCAP_S390_SORT;
939 if (test_facility(151))
940 elf_hwcap |= HWCAP_S390_DFLT;
941
942 /*
943 * Guarded storage support HWCAP_S390_GS is bit 12.
944 */
945 if (MACHINE_HAS_GS)
946 elf_hwcap |= HWCAP_S390_GS;
947
948 get_cpu_id(&cpu_id);
949 add_device_randomness(&cpu_id, sizeof(cpu_id));
950 switch (cpu_id.machine) {
951 case 0x2064:
952 case 0x2066:
953 default: /* Use "z900" as default for 64 bit kernels. */
954 strcpy(elf_platform, "z900");
955 break;
956 case 0x2084:
957 case 0x2086:
958 strcpy(elf_platform, "z990");
959 break;
960 case 0x2094:
961 case 0x2096:
962 strcpy(elf_platform, "z9-109");
963 break;
964 case 0x2097:
965 case 0x2098:
966 strcpy(elf_platform, "z10");
967 break;
968 case 0x2817:
969 case 0x2818:
970 strcpy(elf_platform, "z196");
971 break;
972 case 0x2827:
973 case 0x2828:
974 strcpy(elf_platform, "zEC12");
975 break;
976 case 0x2964:
977 case 0x2965:
978 strcpy(elf_platform, "z13");
979 break;
980 case 0x3906:
981 case 0x3907:
982 strcpy(elf_platform, "z14");
983 break;
984 case 0x8561:
985 case 0x8562:
986 strcpy(elf_platform, "z15");
987 break;
988 }
989
990 /*
991 * Virtualization support HWCAP_INT_SIE is bit 0.
992 */
993 if (sclp.has_sief2)
994 int_hwcap |= HWCAP_INT_SIE;
995
996 return 0;
997 }
998 arch_initcall(setup_hwcaps);
999
1000 /*
1001 * Add system information as device randomness
1002 */
setup_randomness(void)1003 static void __init setup_randomness(void)
1004 {
1005 struct sysinfo_3_2_2 *vmms;
1006
1007 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
1008 PAGE_SIZE);
1009 if (!vmms)
1010 panic("Failed to allocate memory for sysinfo structure\n");
1011
1012 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1013 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1014 memblock_free((unsigned long) vmms, PAGE_SIZE);
1015 }
1016
1017 /*
1018 * Find the correct size for the task_struct. This depends on
1019 * the size of the struct fpu at the end of the thread_struct
1020 * which is embedded in the task_struct.
1021 */
setup_task_size(void)1022 static void __init setup_task_size(void)
1023 {
1024 int task_size = sizeof(struct task_struct);
1025
1026 if (!MACHINE_HAS_VX) {
1027 task_size -= sizeof(__vector128) * __NUM_VXRS;
1028 task_size += sizeof(freg_t) * __NUM_FPRS;
1029 }
1030 arch_task_struct_size = task_size;
1031 }
1032
1033 /*
1034 * Issue diagnose 318 to set the control program name and
1035 * version codes.
1036 */
setup_control_program_code(void)1037 static void __init setup_control_program_code(void)
1038 {
1039 union diag318_info diag318_info = {
1040 .cpnc = CPNC_LINUX,
1041 .cpvc_linux = 0,
1042 .cpvc_distro = {0},
1043 };
1044
1045 if (!sclp.has_diag318)
1046 return;
1047
1048 diag_stat_inc(DIAG_STAT_X318);
1049 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1050 }
1051
1052 /*
1053 * Print the component list from the IPL report
1054 */
log_component_list(void)1055 static void __init log_component_list(void)
1056 {
1057 struct ipl_rb_component_entry *ptr, *end;
1058 char *str;
1059
1060 if (!early_ipl_comp_list_addr)
1061 return;
1062 if (ipl_block.hdr.flags & IPL_PL_FLAG_IPLSR)
1063 pr_info("Linux is running with Secure-IPL enabled\n");
1064 else
1065 pr_info("Linux is running with Secure-IPL disabled\n");
1066 ptr = (void *) early_ipl_comp_list_addr;
1067 end = (void *) ptr + early_ipl_comp_list_size;
1068 pr_info("The IPL report contains the following components:\n");
1069 while (ptr < end) {
1070 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1071 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1072 str = "signed, verified";
1073 else
1074 str = "signed, verification failed";
1075 } else {
1076 str = "not signed";
1077 }
1078 pr_info("%016llx - %016llx (%s)\n",
1079 ptr->addr, ptr->addr + ptr->len, str);
1080 ptr++;
1081 }
1082 }
1083
1084 /*
1085 * Setup function called from init/main.c just after the banner
1086 * was printed.
1087 */
1088
setup_arch(char ** cmdline_p)1089 void __init setup_arch(char **cmdline_p)
1090 {
1091 /*
1092 * print what head.S has found out about the machine
1093 */
1094 if (MACHINE_IS_VM)
1095 pr_info("Linux is running as a z/VM "
1096 "guest operating system in 64-bit mode\n");
1097 else if (MACHINE_IS_KVM)
1098 pr_info("Linux is running under KVM in 64-bit mode\n");
1099 else if (MACHINE_IS_LPAR)
1100 pr_info("Linux is running natively in 64-bit mode\n");
1101 else
1102 pr_info("Linux is running as a guest in 64-bit mode\n");
1103
1104 log_component_list();
1105
1106 /* Have one command line that is parsed and saved in /proc/cmdline */
1107 /* boot_command_line has been already set up in early.c */
1108 *cmdline_p = boot_command_line;
1109
1110 ROOT_DEV = Root_RAM0;
1111
1112 init_mm.start_code = (unsigned long) _text;
1113 init_mm.end_code = (unsigned long) _etext;
1114 init_mm.end_data = (unsigned long) _edata;
1115 init_mm.brk = (unsigned long) _end;
1116
1117 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1118 nospec_auto_detect();
1119
1120 parse_early_param();
1121 #ifdef CONFIG_CRASH_DUMP
1122 /* Deactivate elfcorehdr= kernel parameter */
1123 elfcorehdr_addr = ELFCORE_ADDR_MAX;
1124 #endif
1125
1126 os_info_init();
1127 setup_ipl();
1128 setup_task_size();
1129 setup_control_program_code();
1130
1131 /* Do some memory reservations *before* memory is added to memblock */
1132 reserve_memory_end();
1133 reserve_oldmem();
1134 reserve_kernel();
1135 reserve_initrd();
1136 reserve_certificate_list();
1137 reserve_mem_detect_info();
1138 memblock_allow_resize();
1139
1140 /* Get information about *all* installed memory */
1141 memblock_add_mem_detect_info();
1142
1143 free_mem_detect_info();
1144 remove_oldmem();
1145
1146 /*
1147 * Make sure all chunks are MAX_ORDER aligned so we don't need the
1148 * extra checks that HOLES_IN_ZONE would require.
1149 *
1150 * Is this still required?
1151 */
1152 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
1153
1154 setup_memory_end();
1155 setup_memory();
1156 dma_contiguous_reserve(memory_end);
1157 vmcp_cma_reserve();
1158
1159 check_initrd();
1160 reserve_crashkernel();
1161 #ifdef CONFIG_CRASH_DUMP
1162 /*
1163 * Be aware that smp_save_dump_cpus() triggers a system reset.
1164 * Therefore CPU and device initialization should be done afterwards.
1165 */
1166 smp_save_dump_cpus();
1167 #endif
1168
1169 setup_resources();
1170 setup_lowcore_dat_off();
1171 smp_fill_possible_mask();
1172 cpu_detect_mhz_feature();
1173 cpu_init();
1174 numa_setup();
1175 smp_detect_cpus();
1176 topology_init_early();
1177
1178 /*
1179 * Create kernel page tables and switch to virtual addressing.
1180 */
1181 paging_init();
1182
1183 /*
1184 * After paging_init created the kernel page table, the new PSWs
1185 * in lowcore can now run with DAT enabled.
1186 */
1187 setup_lowcore_dat_on();
1188
1189 /* Setup default console */
1190 conmode_default();
1191 set_preferred_console();
1192
1193 apply_alternative_instructions();
1194 if (IS_ENABLED(CONFIG_EXPOLINE))
1195 nospec_init_branches();
1196
1197 /* Setup zfcpdump support */
1198 setup_zfcpdump();
1199
1200 /* Add system specific data to the random pool */
1201 setup_randomness();
1202 }
1203
