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
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 <asm/asm-offsets.h>
77 #include "entry.h"
78
79 /*
80 * Machine setup..
81 */
82 unsigned int console_mode = 0;
83 EXPORT_SYMBOL(console_mode);
84
85 unsigned int console_devno = -1;
86 EXPORT_SYMBOL(console_devno);
87
88 unsigned int console_irq = -1;
89 EXPORT_SYMBOL(console_irq);
90
91 unsigned long elf_hwcap __read_mostly = 0;
92 char elf_platform[ELF_PLATFORM_SIZE];
93
94 unsigned long int_hwcap = 0;
95
96 int __bootdata(noexec_disabled);
97 int __bootdata(memory_end_set);
98 unsigned long __bootdata(memory_end);
99 unsigned long __bootdata(vmalloc_size);
100 unsigned long __bootdata(max_physmem_end);
101 struct mem_detect_info __bootdata(mem_detect);
102
103 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
104 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
105 unsigned long __bootdata_preserved(__stext_dma);
106 unsigned long __bootdata_preserved(__etext_dma);
107 unsigned long __bootdata_preserved(__sdma);
108 unsigned long __bootdata_preserved(__edma);
109 unsigned long __bootdata_preserved(__kaslr_offset);
110 unsigned int __bootdata_preserved(zlib_dfltcc_support);
111 EXPORT_SYMBOL(zlib_dfltcc_support);
112
113 unsigned long VMALLOC_START;
114 EXPORT_SYMBOL(VMALLOC_START);
115
116 unsigned long VMALLOC_END;
117 EXPORT_SYMBOL(VMALLOC_END);
118
119 struct page *vmemmap;
120 EXPORT_SYMBOL(vmemmap);
121 unsigned long vmemmap_size;
122
123 unsigned long MODULES_VADDR;
124 unsigned long MODULES_END;
125
126 /* An array with a pointer to the lowcore of every CPU. */
127 struct lowcore *lowcore_ptr[NR_CPUS];
128 EXPORT_SYMBOL(lowcore_ptr);
129
130 /*
131 * The Write Back bit position in the physaddr is given by the SLPC PCI.
132 * Leaving the mask zero always uses write through which is safe
133 */
134 unsigned long mio_wb_bit_mask __ro_after_init;
135
136 /*
137 * This is set up by the setup-routine at boot-time
138 * for S390 need to find out, what we have to setup
139 * using address 0x10400 ...
140 */
141
142 #include <asm/setup.h>
143
144 /*
145 * condev= and conmode= setup parameter.
146 */
147
condev_setup(char * str)148 static int __init condev_setup(char *str)
149 {
150 int vdev;
151
152 vdev = simple_strtoul(str, &str, 0);
153 if (vdev >= 0 && vdev < 65536) {
154 console_devno = vdev;
155 console_irq = -1;
156 }
157 return 1;
158 }
159
160 __setup("condev=", condev_setup);
161
set_preferred_console(void)162 static void __init set_preferred_console(void)
163 {
164 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
165 add_preferred_console("ttyS", 0, NULL);
166 else if (CONSOLE_IS_3270)
167 add_preferred_console("tty3270", 0, NULL);
168 else if (CONSOLE_IS_VT220)
169 add_preferred_console("ttysclp", 0, NULL);
170 else if (CONSOLE_IS_HVC)
171 add_preferred_console("hvc", 0, NULL);
172 }
173
conmode_setup(char * str)174 static int __init conmode_setup(char *str)
175 {
176 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
177 if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
178 SET_CONSOLE_SCLP;
179 #endif
180 #if defined(CONFIG_TN3215_CONSOLE)
181 if (!strcmp(str, "3215"))
182 SET_CONSOLE_3215;
183 #endif
184 #if defined(CONFIG_TN3270_CONSOLE)
185 if (!strcmp(str, "3270"))
186 SET_CONSOLE_3270;
187 #endif
188 set_preferred_console();
189 return 1;
190 }
191
192 __setup("conmode=", conmode_setup);
193
conmode_default(void)194 static void __init conmode_default(void)
195 {
196 char query_buffer[1024];
197 char *ptr;
198
199 if (MACHINE_IS_VM) {
200 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
201 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
202 ptr = strstr(query_buffer, "SUBCHANNEL =");
203 console_irq = simple_strtoul(ptr + 13, NULL, 16);
204 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
205 ptr = strstr(query_buffer, "CONMODE");
206 /*
207 * Set the conmode to 3215 so that the device recognition
208 * will set the cu_type of the console to 3215. If the
209 * conmode is 3270 and we don't set it back then both
210 * 3215 and the 3270 driver will try to access the console
211 * device (3215 as console and 3270 as normal tty).
212 */
213 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
214 if (ptr == NULL) {
215 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
216 SET_CONSOLE_SCLP;
217 #endif
218 return;
219 }
220 if (str_has_prefix(ptr + 8, "3270")) {
221 #if defined(CONFIG_TN3270_CONSOLE)
222 SET_CONSOLE_3270;
223 #elif defined(CONFIG_TN3215_CONSOLE)
224 SET_CONSOLE_3215;
225 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
226 SET_CONSOLE_SCLP;
227 #endif
228 } else if (str_has_prefix(ptr + 8, "3215")) {
229 #if defined(CONFIG_TN3215_CONSOLE)
230 SET_CONSOLE_3215;
231 #elif defined(CONFIG_TN3270_CONSOLE)
232 SET_CONSOLE_3270;
233 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
234 SET_CONSOLE_SCLP;
235 #endif
236 }
237 } else if (MACHINE_IS_KVM) {
238 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
239 SET_CONSOLE_VT220;
240 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
241 SET_CONSOLE_SCLP;
242 else
243 SET_CONSOLE_HVC;
244 } else {
245 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
246 SET_CONSOLE_SCLP;
247 #endif
248 }
249 }
250
251 #ifdef CONFIG_CRASH_DUMP
setup_zfcpdump(void)252 static void __init setup_zfcpdump(void)
253 {
254 if (!is_ipl_type_dump())
255 return;
256 if (OLDMEM_BASE)
257 return;
258 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
259 console_loglevel = 2;
260 }
261 #else
setup_zfcpdump(void)262 static inline void setup_zfcpdump(void) {}
263 #endif /* CONFIG_CRASH_DUMP */
264
265 /*
266 * Reboot, halt and power_off stubs. They just call _machine_restart,
267 * _machine_halt or _machine_power_off.
268 */
269
machine_restart(char * command)270 void machine_restart(char *command)
271 {
272 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
273 /*
274 * Only unblank the console if we are called in enabled
275 * context or a bust_spinlocks cleared the way for us.
276 */
277 console_unblank();
278 _machine_restart(command);
279 }
280
machine_halt(void)281 void machine_halt(void)
282 {
283 if (!in_interrupt() || oops_in_progress)
284 /*
285 * Only unblank the console if we are called in enabled
286 * context or a bust_spinlocks cleared the way for us.
287 */
288 console_unblank();
289 _machine_halt();
290 }
291
machine_power_off(void)292 void machine_power_off(void)
293 {
294 if (!in_interrupt() || oops_in_progress)
295 /*
296 * Only unblank the console if we are called in enabled
297 * context or a bust_spinlocks cleared the way for us.
298 */
299 console_unblank();
300 _machine_power_off();
301 }
302
303 /*
304 * Dummy power off function.
305 */
306 void (*pm_power_off)(void) = machine_power_off;
307 EXPORT_SYMBOL_GPL(pm_power_off);
308
309 void *restart_stack;
310
stack_alloc(void)311 unsigned long stack_alloc(void)
312 {
313 #ifdef CONFIG_VMAP_STACK
314 return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE,
315 THREADINFO_GFP, 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 unsigned long stack;
359
360 stack = stack_alloc();
361 if (!stack)
362 panic("Couldn't allocate kernel stack");
363 current->stack = (void *) stack;
364 #ifdef CONFIG_VMAP_STACK
365 current->stack_vm_area = (void *) stack;
366 #endif
367 set_task_stack_end_magic(current);
368 stack += STACK_INIT_OFFSET;
369 S390_lowcore.kernel_stack = stack;
370 CALL_ON_STACK_NORETURN(rest_init, stack);
371 }
372
setup_lowcore_dat_off(void)373 static void __init setup_lowcore_dat_off(void)
374 {
375 unsigned long int_psw_mask = PSW_KERNEL_BITS;
376 struct lowcore *lc;
377
378 if (IS_ENABLED(CONFIG_KASAN))
379 int_psw_mask |= PSW_MASK_DAT;
380
381 /*
382 * Setup lowcore for boot cpu
383 */
384 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
385 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
386 if (!lc)
387 panic("%s: Failed to allocate %zu bytes align=%zx\n",
388 __func__, sizeof(*lc), sizeof(*lc));
389
390 lc->restart_psw.mask = PSW_KERNEL_BITS;
391 lc->restart_psw.addr = (unsigned long) restart_int_handler;
392 lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
393 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
394 lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
395 lc->svc_new_psw.addr = (unsigned long) system_call;
396 lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
397 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
398 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
399 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
400 lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
401 lc->io_new_psw.addr = (unsigned long) io_int_handler;
402 lc->clock_comparator = clock_comparator_max;
403 lc->nodat_stack = ((unsigned long) &init_thread_union)
404 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
405 lc->current_task = (unsigned long)&init_task;
406 lc->lpp = LPP_MAGIC;
407 lc->machine_flags = S390_lowcore.machine_flags;
408 lc->preempt_count = S390_lowcore.preempt_count;
409 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
410 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
411 sizeof(lc->stfle_fac_list));
412 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
413 sizeof(lc->alt_stfle_fac_list));
414 nmi_alloc_boot_cpu(lc);
415 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
416 lc->async_enter_timer = S390_lowcore.async_enter_timer;
417 lc->exit_timer = S390_lowcore.exit_timer;
418 lc->user_timer = S390_lowcore.user_timer;
419 lc->system_timer = S390_lowcore.system_timer;
420 lc->steal_timer = S390_lowcore.steal_timer;
421 lc->last_update_timer = S390_lowcore.last_update_timer;
422 lc->last_update_clock = S390_lowcore.last_update_clock;
423
424 /*
425 * Allocate the global restart stack which is the same for
426 * all CPUs in cast *one* of them does a PSW restart.
427 */
428 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
429 if (!restart_stack)
430 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
431 __func__, THREAD_SIZE, THREAD_SIZE);
432 restart_stack += STACK_INIT_OFFSET;
433
434 /*
435 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
436 * restart data to the absolute zero lowcore. This is necessary if
437 * PSW restart is done on an offline CPU that has lowcore zero.
438 */
439 lc->restart_stack = (unsigned long) restart_stack;
440 lc->restart_fn = (unsigned long) do_restart;
441 lc->restart_data = 0;
442 lc->restart_source = -1UL;
443
444 /* Setup absolute zero lowcore */
445 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
446 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
447 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
448 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
449 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
450
451 lc->spinlock_lockval = arch_spin_lockval(0);
452 lc->spinlock_index = 0;
453 arch_spin_lock_setup(0);
454 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
455 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
456 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
457 lc->preempt_count = PREEMPT_DISABLED;
458
459 set_prefix((u32)(unsigned long) lc);
460 lowcore_ptr[0] = lc;
461 }
462
setup_lowcore_dat_on(void)463 static void __init setup_lowcore_dat_on(void)
464 {
465 __ctl_clear_bit(0, 28);
466 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
467 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
468 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
469 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
470 __ctl_set_bit(0, 28);
471 }
472
473 static struct resource code_resource = {
474 .name = "Kernel code",
475 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
476 };
477
478 static struct resource data_resource = {
479 .name = "Kernel data",
480 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
481 };
482
483 static struct resource bss_resource = {
484 .name = "Kernel bss",
485 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
486 };
487
488 static struct resource __initdata *standard_resources[] = {
489 &code_resource,
490 &data_resource,
491 &bss_resource,
492 };
493
setup_resources(void)494 static void __init setup_resources(void)
495 {
496 struct resource *res, *std_res, *sub_res;
497 phys_addr_t start, end;
498 int j;
499 u64 i;
500
501 code_resource.start = (unsigned long) _text;
502 code_resource.end = (unsigned long) _etext - 1;
503 data_resource.start = (unsigned long) _etext;
504 data_resource.end = (unsigned long) _edata - 1;
505 bss_resource.start = (unsigned long) __bss_start;
506 bss_resource.end = (unsigned long) __bss_stop - 1;
507
508 for_each_mem_range(i, &start, &end) {
509 res = memblock_alloc(sizeof(*res), 8);
510 if (!res)
511 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
512 __func__, sizeof(*res), 8);
513 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
514
515 res->name = "System RAM";
516 res->start = start;
517 /*
518 * In memblock, end points to the first byte after the
519 * range while in resourses, end points to the last byte in
520 * the range.
521 */
522 res->end = end - 1;
523 request_resource(&iomem_resource, res);
524
525 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
526 std_res = standard_resources[j];
527 if (std_res->start < res->start ||
528 std_res->start > res->end)
529 continue;
530 if (std_res->end > res->end) {
531 sub_res = memblock_alloc(sizeof(*sub_res), 8);
532 if (!sub_res)
533 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
534 __func__, sizeof(*sub_res), 8);
535 *sub_res = *std_res;
536 sub_res->end = res->end;
537 std_res->start = res->end + 1;
538 request_resource(res, sub_res);
539 } else {
540 request_resource(res, std_res);
541 }
542 }
543 }
544 #ifdef CONFIG_CRASH_DUMP
545 /*
546 * Re-add removed crash kernel memory as reserved memory. This makes
547 * sure it will be mapped with the identity mapping and struct pages
548 * will be created, so it can be resized later on.
549 * However add it later since the crash kernel resource should not be
550 * part of the System RAM resource.
551 */
552 if (crashk_res.end) {
553 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
554 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
555 insert_resource(&iomem_resource, &crashk_res);
556 }
557 #endif
558 }
559
setup_memory_end(void)560 static void __init setup_memory_end(void)
561 {
562 unsigned long vmax, tmp;
563
564 /* Choose kernel address space layout: 3 or 4 levels. */
565 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
566 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
567 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
568 vmax = _REGION2_SIZE; /* 3-level kernel page table */
569 else
570 vmax = _REGION1_SIZE; /* 4-level kernel page table */
571 if (is_prot_virt_host())
572 adjust_to_uv_max(&vmax);
573 #ifdef CONFIG_KASAN
574 vmax = kasan_vmax;
575 #endif
576 /* module area is at the end of the kernel address space. */
577 MODULES_END = vmax;
578 MODULES_VADDR = MODULES_END - MODULES_LEN;
579 VMALLOC_END = MODULES_VADDR;
580 VMALLOC_START = VMALLOC_END - vmalloc_size;
581
582 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
583 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
584 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
585 tmp = SECTION_ALIGN_UP(tmp);
586 tmp = VMALLOC_START - tmp * sizeof(struct page);
587 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
588 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
589 vmemmap = (struct page *) tmp;
590
591 /* Take care that memory_end is set and <= vmemmap */
592 memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
593 #ifdef CONFIG_KASAN
594 memory_end = min(memory_end, KASAN_SHADOW_START);
595 #endif
596 vmemmap_size = SECTION_ALIGN_UP(memory_end / PAGE_SIZE) * sizeof(struct page);
597 #ifdef CONFIG_KASAN
598 /* move vmemmap above kasan shadow only if stands in a way */
599 if (KASAN_SHADOW_END > (unsigned long)vmemmap &&
600 (unsigned long)vmemmap + vmemmap_size > KASAN_SHADOW_START)
601 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
602 #endif
603 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
604 memblock_remove(memory_end, ULONG_MAX);
605
606 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
607 }
608
609 #ifdef CONFIG_CRASH_DUMP
610
611 /*
612 * When kdump is enabled, we have to ensure that no memory from the area
613 * [0 - crashkernel memory size] is set offline - it will be exchanged with
614 * the crashkernel memory region when kdump is triggered. The crashkernel
615 * memory region can never get offlined (pages are unmovable).
616 */
kdump_mem_notifier(struct notifier_block * nb,unsigned long action,void * data)617 static int kdump_mem_notifier(struct notifier_block *nb,
618 unsigned long action, void *data)
619 {
620 struct memory_notify *arg = data;
621
622 if (action != MEM_GOING_OFFLINE)
623 return NOTIFY_OK;
624 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
625 return NOTIFY_BAD;
626 return NOTIFY_OK;
627 }
628
629 static struct notifier_block kdump_mem_nb = {
630 .notifier_call = kdump_mem_notifier,
631 };
632
633 #endif
634
635 /*
636 * Make sure that the area behind memory_end is protected
637 */
reserve_memory_end(void)638 static void __init reserve_memory_end(void)
639 {
640 if (memory_end_set)
641 memblock_reserve(memory_end, ULONG_MAX);
642 }
643
644 /*
645 * Make sure that oldmem, where the dump is stored, is protected
646 */
reserve_oldmem(void)647 static void __init reserve_oldmem(void)
648 {
649 #ifdef CONFIG_CRASH_DUMP
650 if (OLDMEM_BASE)
651 /* Forget all memory above the running kdump system */
652 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
653 #endif
654 }
655
656 /*
657 * Make sure that oldmem, where the dump is stored, is protected
658 */
remove_oldmem(void)659 static void __init remove_oldmem(void)
660 {
661 #ifdef CONFIG_CRASH_DUMP
662 if (OLDMEM_BASE)
663 /* Forget all memory above the running kdump system */
664 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
665 #endif
666 }
667
668 /*
669 * Reserve memory for kdump kernel to be loaded with kexec
670 */
reserve_crashkernel(void)671 static void __init reserve_crashkernel(void)
672 {
673 #ifdef CONFIG_CRASH_DUMP
674 unsigned long long crash_base, crash_size;
675 phys_addr_t low, high;
676 int rc;
677
678 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
679 &crash_base);
680
681 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
682 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
683 if (rc || crash_size == 0)
684 return;
685
686 if (memblock.memory.regions[0].size < crash_size) {
687 pr_info("crashkernel reservation failed: %s\n",
688 "first memory chunk must be at least crashkernel size");
689 return;
690 }
691
692 low = crash_base ?: OLDMEM_BASE;
693 high = low + crash_size;
694 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
695 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
696 crash_base = low;
697 } else {
698 /* Find suitable area in free memory */
699 low = max_t(unsigned long, crash_size, sclp.hsa_size);
700 high = crash_base ? crash_base + crash_size : ULONG_MAX;
701
702 if (crash_base && crash_base < low) {
703 pr_info("crashkernel reservation failed: %s\n",
704 "crash_base too low");
705 return;
706 }
707 low = crash_base ?: low;
708 crash_base = memblock_find_in_range(low, high, crash_size,
709 KEXEC_CRASH_MEM_ALIGN);
710 }
711
712 if (!crash_base) {
713 pr_info("crashkernel reservation failed: %s\n",
714 "no suitable area found");
715 return;
716 }
717
718 if (register_memory_notifier(&kdump_mem_nb))
719 return;
720
721 if (!OLDMEM_BASE && MACHINE_IS_VM)
722 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
723 crashk_res.start = crash_base;
724 crashk_res.end = crash_base + crash_size - 1;
725 memblock_remove(crash_base, crash_size);
726 pr_info("Reserving %lluMB of memory at %lluMB "
727 "for crashkernel (System RAM: %luMB)\n",
728 crash_size >> 20, crash_base >> 20,
729 (unsigned long)memblock.memory.total_size >> 20);
730 os_info_crashkernel_add(crash_base, crash_size);
731 #endif
732 }
733
734 /*
735 * Reserve the initrd from being used by memblock
736 */
reserve_initrd(void)737 static void __init reserve_initrd(void)
738 {
739 #ifdef CONFIG_BLK_DEV_INITRD
740 if (!INITRD_START || !INITRD_SIZE)
741 return;
742 initrd_start = INITRD_START;
743 initrd_end = initrd_start + INITRD_SIZE;
744 memblock_reserve(INITRD_START, INITRD_SIZE);
745 #endif
746 }
747
748 /*
749 * Reserve the memory area used to pass the certificate lists
750 */
reserve_certificate_list(void)751 static void __init reserve_certificate_list(void)
752 {
753 if (ipl_cert_list_addr)
754 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
755 }
756
reserve_mem_detect_info(void)757 static void __init reserve_mem_detect_info(void)
758 {
759 unsigned long start, size;
760
761 get_mem_detect_reserved(&start, &size);
762 if (size)
763 memblock_reserve(start, size);
764 }
765
free_mem_detect_info(void)766 static void __init free_mem_detect_info(void)
767 {
768 unsigned long start, size;
769
770 get_mem_detect_reserved(&start, &size);
771 if (size)
772 memblock_free(start, size);
773 }
774
get_mem_info_source(void)775 static const char * __init get_mem_info_source(void)
776 {
777 switch (mem_detect.info_source) {
778 case MEM_DETECT_SCLP_STOR_INFO:
779 return "sclp storage info";
780 case MEM_DETECT_DIAG260:
781 return "diag260";
782 case MEM_DETECT_SCLP_READ_INFO:
783 return "sclp read info";
784 case MEM_DETECT_BIN_SEARCH:
785 return "binary search";
786 }
787 return "none";
788 }
789
memblock_add_mem_detect_info(void)790 static void __init memblock_add_mem_detect_info(void)
791 {
792 unsigned long start, end;
793 int i;
794
795 pr_debug("physmem info source: %s (%hhd)\n",
796 get_mem_info_source(), mem_detect.info_source);
797 /* keep memblock lists close to the kernel */
798 memblock_set_bottom_up(true);
799 for_each_mem_detect_block(i, &start, &end) {
800 memblock_add(start, end - start);
801 memblock_physmem_add(start, end - start);
802 }
803 memblock_set_bottom_up(false);
804 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
805 memblock_dump_all();
806 }
807
808 /*
809 * Check for initrd being in usable memory
810 */
check_initrd(void)811 static void __init check_initrd(void)
812 {
813 #ifdef CONFIG_BLK_DEV_INITRD
814 if (INITRD_START && INITRD_SIZE &&
815 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
816 pr_err("The initial RAM disk does not fit into the memory\n");
817 memblock_free(INITRD_START, INITRD_SIZE);
818 initrd_start = initrd_end = 0;
819 }
820 #endif
821 }
822
823 /*
824 * Reserve memory used for lowcore/command line/kernel image.
825 */
reserve_kernel(void)826 static void __init reserve_kernel(void)
827 {
828 unsigned long start_pfn = PFN_UP(__pa(_end));
829
830 memblock_reserve(0, HEAD_END);
831 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
832 - (unsigned long)_stext);
833 memblock_reserve(__sdma, __edma - __sdma);
834 }
835
setup_memory(void)836 static void __init setup_memory(void)
837 {
838 phys_addr_t start, end;
839 u64 i;
840
841 /*
842 * Init storage key for present memory
843 */
844 for_each_mem_range(i, &start, &end)
845 storage_key_init_range(start, end);
846
847 psw_set_key(PAGE_DEFAULT_KEY);
848 }
849
850 /*
851 * Setup hardware capabilities.
852 */
setup_hwcaps(void)853 static int __init setup_hwcaps(void)
854 {
855 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
856 struct cpuid cpu_id;
857 int i;
858
859 /*
860 * The store facility list bits numbers as found in the principles
861 * of operation are numbered with bit 1UL<<31 as number 0 to
862 * bit 1UL<<0 as number 31.
863 * Bit 0: instructions named N3, "backported" to esa-mode
864 * Bit 2: z/Architecture mode is active
865 * Bit 7: the store-facility-list-extended facility is installed
866 * Bit 17: the message-security assist is installed
867 * Bit 19: the long-displacement facility is installed
868 * Bit 21: the extended-immediate facility is installed
869 * Bit 22: extended-translation facility 3 is installed
870 * Bit 30: extended-translation facility 3 enhancement facility
871 * These get translated to:
872 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
873 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
874 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
875 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
876 */
877 for (i = 0; i < 6; i++)
878 if (test_facility(stfl_bits[i]))
879 elf_hwcap |= 1UL << i;
880
881 if (test_facility(22) && test_facility(30))
882 elf_hwcap |= HWCAP_S390_ETF3EH;
883
884 /*
885 * Check for additional facilities with store-facility-list-extended.
886 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
887 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
888 * as stored by stfl, bits 32-xxx contain additional facilities.
889 * How many facility words are stored depends on the number of
890 * doublewords passed to the instruction. The additional facilities
891 * are:
892 * Bit 42: decimal floating point facility is installed
893 * Bit 44: perform floating point operation facility is installed
894 * translated to:
895 * HWCAP_S390_DFP bit 6 (42 && 44).
896 */
897 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
898 elf_hwcap |= HWCAP_S390_DFP;
899
900 /*
901 * Huge page support HWCAP_S390_HPAGE is bit 7.
902 */
903 if (MACHINE_HAS_EDAT1)
904 elf_hwcap |= HWCAP_S390_HPAGE;
905
906 /*
907 * 64-bit register support for 31-bit processes
908 * HWCAP_S390_HIGH_GPRS is bit 9.
909 */
910 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
911
912 /*
913 * Transactional execution support HWCAP_S390_TE is bit 10.
914 */
915 if (MACHINE_HAS_TE)
916 elf_hwcap |= HWCAP_S390_TE;
917
918 /*
919 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
920 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
921 * instead of facility bit 129.
922 */
923 if (MACHINE_HAS_VX) {
924 elf_hwcap |= HWCAP_S390_VXRS;
925 if (test_facility(134))
926 elf_hwcap |= HWCAP_S390_VXRS_BCD;
927 if (test_facility(135))
928 elf_hwcap |= HWCAP_S390_VXRS_EXT;
929 if (test_facility(148))
930 elf_hwcap |= HWCAP_S390_VXRS_EXT2;
931 if (test_facility(152))
932 elf_hwcap |= HWCAP_S390_VXRS_PDE;
933 }
934 if (test_facility(150))
935 elf_hwcap |= HWCAP_S390_SORT;
936 if (test_facility(151))
937 elf_hwcap |= HWCAP_S390_DFLT;
938
939 /*
940 * Guarded storage support HWCAP_S390_GS is bit 12.
941 */
942 if (MACHINE_HAS_GS)
943 elf_hwcap |= HWCAP_S390_GS;
944
945 get_cpu_id(&cpu_id);
946 add_device_randomness(&cpu_id, sizeof(cpu_id));
947 switch (cpu_id.machine) {
948 case 0x2064:
949 case 0x2066:
950 default: /* Use "z900" as default for 64 bit kernels. */
951 strcpy(elf_platform, "z900");
952 break;
953 case 0x2084:
954 case 0x2086:
955 strcpy(elf_platform, "z990");
956 break;
957 case 0x2094:
958 case 0x2096:
959 strcpy(elf_platform, "z9-109");
960 break;
961 case 0x2097:
962 case 0x2098:
963 strcpy(elf_platform, "z10");
964 break;
965 case 0x2817:
966 case 0x2818:
967 strcpy(elf_platform, "z196");
968 break;
969 case 0x2827:
970 case 0x2828:
971 strcpy(elf_platform, "zEC12");
972 break;
973 case 0x2964:
974 case 0x2965:
975 strcpy(elf_platform, "z13");
976 break;
977 case 0x3906:
978 case 0x3907:
979 strcpy(elf_platform, "z14");
980 break;
981 case 0x8561:
982 case 0x8562:
983 strcpy(elf_platform, "z15");
984 break;
985 }
986
987 /*
988 * Virtualization support HWCAP_INT_SIE is bit 0.
989 */
990 if (sclp.has_sief2)
991 int_hwcap |= HWCAP_INT_SIE;
992
993 return 0;
994 }
995 arch_initcall(setup_hwcaps);
996
997 /*
998 * Add system information as device randomness
999 */
setup_randomness(void)1000 static void __init setup_randomness(void)
1001 {
1002 struct sysinfo_3_2_2 *vmms;
1003
1004 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
1005 PAGE_SIZE);
1006 if (!vmms)
1007 panic("Failed to allocate memory for sysinfo structure\n");
1008
1009 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1010 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1011 memblock_free((unsigned long) vmms, PAGE_SIZE);
1012
1013 #ifdef CONFIG_ARCH_RANDOM
1014 if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
1015 static_branch_enable(&s390_arch_random_available);
1016 #endif
1017 }
1018
1019 /*
1020 * Find the correct size for the task_struct. This depends on
1021 * the size of the struct fpu at the end of the thread_struct
1022 * which is embedded in the task_struct.
1023 */
setup_task_size(void)1024 static void __init setup_task_size(void)
1025 {
1026 int task_size = sizeof(struct task_struct);
1027
1028 if (!MACHINE_HAS_VX) {
1029 task_size -= sizeof(__vector128) * __NUM_VXRS;
1030 task_size += sizeof(freg_t) * __NUM_FPRS;
1031 }
1032 arch_task_struct_size = task_size;
1033 }
1034
1035 /*
1036 * Issue diagnose 318 to set the control program name and
1037 * version codes.
1038 */
setup_control_program_code(void)1039 static void __init setup_control_program_code(void)
1040 {
1041 union diag318_info diag318_info = {
1042 .cpnc = CPNC_LINUX,
1043 .cpvc = 0,
1044 };
1045
1046 if (!sclp.has_diag318)
1047 return;
1048
1049 diag_stat_inc(DIAG_STAT_X318);
1050 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1051 }
1052
1053 /*
1054 * Print the component list from the IPL report
1055 */
log_component_list(void)1056 static void __init log_component_list(void)
1057 {
1058 struct ipl_rb_component_entry *ptr, *end;
1059 char *str;
1060
1061 if (!early_ipl_comp_list_addr)
1062 return;
1063 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1064 pr_info("Linux is running with Secure-IPL enabled\n");
1065 else
1066 pr_info("Linux is running with Secure-IPL disabled\n");
1067 ptr = (void *) early_ipl_comp_list_addr;
1068 end = (void *) ptr + early_ipl_comp_list_size;
1069 pr_info("The IPL report contains the following components:\n");
1070 while (ptr < end) {
1071 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1072 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1073 str = "signed, verified";
1074 else
1075 str = "signed, verification failed";
1076 } else {
1077 str = "not signed";
1078 }
1079 pr_info("%016llx - %016llx (%s)\n",
1080 ptr->addr, ptr->addr + ptr->len, str);
1081 ptr++;
1082 }
1083 }
1084
1085 /*
1086 * Setup function called from init/main.c just after the banner
1087 * was printed.
1088 */
1089
setup_arch(char ** cmdline_p)1090 void __init setup_arch(char **cmdline_p)
1091 {
1092 /*
1093 * print what head.S has found out about the machine
1094 */
1095 if (MACHINE_IS_VM)
1096 pr_info("Linux is running as a z/VM "
1097 "guest operating system in 64-bit mode\n");
1098 else if (MACHINE_IS_KVM)
1099 pr_info("Linux is running under KVM in 64-bit mode\n");
1100 else if (MACHINE_IS_LPAR)
1101 pr_info("Linux is running natively in 64-bit mode\n");
1102 else
1103 pr_info("Linux is running as a guest in 64-bit mode\n");
1104
1105 log_component_list();
1106
1107 /* Have one command line that is parsed and saved in /proc/cmdline */
1108 /* boot_command_line has been already set up in early.c */
1109 *cmdline_p = boot_command_line;
1110
1111 ROOT_DEV = Root_RAM0;
1112
1113 init_mm.start_code = (unsigned long) _text;
1114 init_mm.end_code = (unsigned long) _etext;
1115 init_mm.end_data = (unsigned long) _edata;
1116 init_mm.brk = (unsigned long) _end;
1117
1118 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1119 nospec_auto_detect();
1120
1121 jump_label_init();
1122 parse_early_param();
1123 #ifdef CONFIG_CRASH_DUMP
1124 /* Deactivate elfcorehdr= kernel parameter */
1125 elfcorehdr_addr = ELFCORE_ADDR_MAX;
1126 #endif
1127
1128 os_info_init();
1129 setup_ipl();
1130 setup_task_size();
1131 setup_control_program_code();
1132
1133 /* Do some memory reservations *before* memory is added to memblock */
1134 reserve_memory_end();
1135 reserve_oldmem();
1136 reserve_kernel();
1137 reserve_initrd();
1138 reserve_certificate_list();
1139 reserve_mem_detect_info();
1140 memblock_allow_resize();
1141
1142 /* Get information about *all* installed memory */
1143 memblock_add_mem_detect_info();
1144
1145 free_mem_detect_info();
1146 remove_oldmem();
1147
1148 setup_uv();
1149 setup_memory_end();
1150 setup_memory();
1151 dma_contiguous_reserve(memory_end);
1152 vmcp_cma_reserve();
1153
1154 check_initrd();
1155 reserve_crashkernel();
1156 #ifdef CONFIG_CRASH_DUMP
1157 /*
1158 * Be aware that smp_save_dump_cpus() triggers a system reset.
1159 * Therefore CPU and device initialization should be done afterwards.
1160 */
1161 smp_save_dump_cpus();
1162 #endif
1163
1164 setup_resources();
1165 setup_lowcore_dat_off();
1166 smp_fill_possible_mask();
1167 cpu_detect_mhz_feature();
1168 cpu_init();
1169 numa_setup();
1170 smp_detect_cpus();
1171 topology_init_early();
1172
1173 /*
1174 * Create kernel page tables and switch to virtual addressing.
1175 */
1176 paging_init();
1177
1178 /*
1179 * After paging_init created the kernel page table, the new PSWs
1180 * in lowcore can now run with DAT enabled.
1181 */
1182 setup_lowcore_dat_on();
1183
1184 /* Setup default console */
1185 conmode_default();
1186 set_preferred_console();
1187
1188 apply_alternative_instructions();
1189 if (IS_ENABLED(CONFIG_EXPOLINE))
1190 nospec_init_branches();
1191
1192 /* Setup zfcp/nvme dump support */
1193 setup_zfcpdump();
1194
1195 /* Add system specific data to the random pool */
1196 setup_randomness();
1197 }
1198