1 /*
2 * S390 version
3 * Copyright IBM Corp. 1999, 2012
4 * Author(s): Hartmut Penner (hp@de.ibm.com),
5 * Martin Schwidefsky (schwidefsky@de.ibm.com)
6 *
7 * Derived from "arch/i386/kernel/setup.c"
8 * Copyright (C) 1995, Linus Torvalds
9 */
10
11 /*
12 * This file handles the architecture-dependent parts of initialization
13 */
14
15 #define KMSG_COMPONENT "setup"
16 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
17
18 #include <linux/errno.h>
19 #include <linux/export.h>
20 #include <linux/sched.h>
21 #include <linux/sched/task.h>
22 #include <linux/cpu.h>
23 #include <linux/kernel.h>
24 #include <linux/memblock.h>
25 #include <linux/mm.h>
26 #include <linux/stddef.h>
27 #include <linux/unistd.h>
28 #include <linux/ptrace.h>
29 #include <linux/random.h>
30 #include <linux/user.h>
31 #include <linux/tty.h>
32 #include <linux/ioport.h>
33 #include <linux/delay.h>
34 #include <linux/init.h>
35 #include <linux/initrd.h>
36 #include <linux/bootmem.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
52 #include <asm/ipl.h>
53 #include <asm/facility.h>
54 #include <asm/smp.h>
55 #include <asm/mmu_context.h>
56 #include <asm/cpcmd.h>
57 #include <asm/lowcore.h>
58 #include <asm/irq.h>
59 #include <asm/page.h>
60 #include <asm/ptrace.h>
61 #include <asm/sections.h>
62 #include <asm/ebcdic.h>
63 #include <asm/kvm_virtio.h>
64 #include <asm/diag.h>
65 #include <asm/os_info.h>
66 #include <asm/sclp.h>
67 #include <asm/sysinfo.h>
68 #include <asm/numa.h>
69 #include <asm/alternative.h>
70 #include <asm/nospec-branch.h>
71 #include "entry.h"
72
73 /*
74 * Machine setup..
75 */
76 unsigned int console_mode = 0;
77 EXPORT_SYMBOL(console_mode);
78
79 unsigned int console_devno = -1;
80 EXPORT_SYMBOL(console_devno);
81
82 unsigned int console_irq = -1;
83 EXPORT_SYMBOL(console_irq);
84
85 unsigned long elf_hwcap __read_mostly = 0;
86 char elf_platform[ELF_PLATFORM_SIZE];
87
88 unsigned long int_hwcap = 0;
89
90 int __initdata memory_end_set;
91 unsigned long __initdata memory_end;
92 unsigned long __initdata max_physmem_end;
93
94 unsigned long VMALLOC_START;
95 EXPORT_SYMBOL(VMALLOC_START);
96
97 unsigned long VMALLOC_END;
98 EXPORT_SYMBOL(VMALLOC_END);
99
100 struct page *vmemmap;
101 EXPORT_SYMBOL(vmemmap);
102
103 unsigned long MODULES_VADDR;
104 unsigned long MODULES_END;
105
106 /* An array with a pointer to the lowcore of every CPU. */
107 struct lowcore *lowcore_ptr[NR_CPUS];
108 EXPORT_SYMBOL(lowcore_ptr);
109
110 /*
111 * This is set up by the setup-routine at boot-time
112 * for S390 need to find out, what we have to setup
113 * using address 0x10400 ...
114 */
115
116 #include <asm/setup.h>
117
118 /*
119 * condev= and conmode= setup parameter.
120 */
121
condev_setup(char * str)122 static int __init condev_setup(char *str)
123 {
124 int vdev;
125
126 vdev = simple_strtoul(str, &str, 0);
127 if (vdev >= 0 && vdev < 65536) {
128 console_devno = vdev;
129 console_irq = -1;
130 }
131 return 1;
132 }
133
134 __setup("condev=", condev_setup);
135
set_preferred_console(void)136 static void __init set_preferred_console(void)
137 {
138 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
139 add_preferred_console("ttyS", 0, NULL);
140 else if (CONSOLE_IS_3270)
141 add_preferred_console("tty3270", 0, NULL);
142 else if (CONSOLE_IS_VT220)
143 add_preferred_console("ttyS", 1, NULL);
144 else if (CONSOLE_IS_HVC)
145 add_preferred_console("hvc", 0, NULL);
146 }
147
conmode_setup(char * str)148 static int __init conmode_setup(char *str)
149 {
150 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
151 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
152 SET_CONSOLE_SCLP;
153 #endif
154 #if defined(CONFIG_TN3215_CONSOLE)
155 if (strncmp(str, "3215", 5) == 0)
156 SET_CONSOLE_3215;
157 #endif
158 #if defined(CONFIG_TN3270_CONSOLE)
159 if (strncmp(str, "3270", 5) == 0)
160 SET_CONSOLE_3270;
161 #endif
162 set_preferred_console();
163 return 1;
164 }
165
166 __setup("conmode=", conmode_setup);
167
conmode_default(void)168 static void __init conmode_default(void)
169 {
170 char query_buffer[1024];
171 char *ptr;
172
173 if (MACHINE_IS_VM) {
174 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
175 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
176 ptr = strstr(query_buffer, "SUBCHANNEL =");
177 console_irq = simple_strtoul(ptr + 13, NULL, 16);
178 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
179 ptr = strstr(query_buffer, "CONMODE");
180 /*
181 * Set the conmode to 3215 so that the device recognition
182 * will set the cu_type of the console to 3215. If the
183 * conmode is 3270 and we don't set it back then both
184 * 3215 and the 3270 driver will try to access the console
185 * device (3215 as console and 3270 as normal tty).
186 */
187 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
188 if (ptr == NULL) {
189 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
190 SET_CONSOLE_SCLP;
191 #endif
192 return;
193 }
194 if (strncmp(ptr + 8, "3270", 4) == 0) {
195 #if defined(CONFIG_TN3270_CONSOLE)
196 SET_CONSOLE_3270;
197 #elif defined(CONFIG_TN3215_CONSOLE)
198 SET_CONSOLE_3215;
199 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
200 SET_CONSOLE_SCLP;
201 #endif
202 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
203 #if defined(CONFIG_TN3215_CONSOLE)
204 SET_CONSOLE_3215;
205 #elif defined(CONFIG_TN3270_CONSOLE)
206 SET_CONSOLE_3270;
207 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
208 SET_CONSOLE_SCLP;
209 #endif
210 }
211 } else if (MACHINE_IS_KVM) {
212 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
213 SET_CONSOLE_VT220;
214 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
215 SET_CONSOLE_SCLP;
216 else
217 SET_CONSOLE_HVC;
218 } else {
219 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
220 SET_CONSOLE_SCLP;
221 #endif
222 }
223 }
224
225 #ifdef CONFIG_CRASH_DUMP
setup_zfcpdump(void)226 static void __init setup_zfcpdump(void)
227 {
228 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
229 return;
230 if (OLDMEM_BASE)
231 return;
232 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
233 console_loglevel = 2;
234 }
235 #else
setup_zfcpdump(void)236 static inline void setup_zfcpdump(void) {}
237 #endif /* CONFIG_CRASH_DUMP */
238
239 /*
240 * Reboot, halt and power_off stubs. They just call _machine_restart,
241 * _machine_halt or _machine_power_off.
242 */
243
machine_restart(char * command)244 void machine_restart(char *command)
245 {
246 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
247 /*
248 * Only unblank the console if we are called in enabled
249 * context or a bust_spinlocks cleared the way for us.
250 */
251 console_unblank();
252 _machine_restart(command);
253 }
254
machine_halt(void)255 void machine_halt(void)
256 {
257 if (!in_interrupt() || oops_in_progress)
258 /*
259 * Only unblank the console if we are called in enabled
260 * context or a bust_spinlocks cleared the way for us.
261 */
262 console_unblank();
263 _machine_halt();
264 }
265
machine_power_off(void)266 void machine_power_off(void)
267 {
268 if (!in_interrupt() || oops_in_progress)
269 /*
270 * Only unblank the console if we are called in enabled
271 * context or a bust_spinlocks cleared the way for us.
272 */
273 console_unblank();
274 _machine_power_off();
275 }
276
277 /*
278 * Dummy power off function.
279 */
280 void (*pm_power_off)(void) = machine_power_off;
281 EXPORT_SYMBOL_GPL(pm_power_off);
282
early_parse_mem(char * p)283 static int __init early_parse_mem(char *p)
284 {
285 memory_end = memparse(p, &p);
286 memory_end &= PAGE_MASK;
287 memory_end_set = 1;
288 return 0;
289 }
290 early_param("mem", early_parse_mem);
291
parse_vmalloc(char * arg)292 static int __init parse_vmalloc(char *arg)
293 {
294 if (!arg)
295 return -EINVAL;
296 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
297 return 0;
298 }
299 early_param("vmalloc", parse_vmalloc);
300
301 void *restart_stack __section(.data);
302
setup_lowcore_dat_off(void)303 static void __init setup_lowcore_dat_off(void)
304 {
305 struct lowcore *lc;
306
307 /*
308 * Setup lowcore for boot cpu
309 */
310 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
311 lc = memblock_virt_alloc_low(sizeof(*lc), sizeof(*lc));
312 lc->restart_psw.mask = PSW_KERNEL_BITS;
313 lc->restart_psw.addr = (unsigned long) restart_int_handler;
314 lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
315 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
316 lc->svc_new_psw.mask = PSW_KERNEL_BITS |
317 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
318 lc->svc_new_psw.addr = (unsigned long) system_call;
319 lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
320 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
321 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
322 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
323 lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
324 lc->io_new_psw.addr = (unsigned long) io_int_handler;
325 lc->clock_comparator = clock_comparator_max;
326 lc->kernel_stack = ((unsigned long) &init_thread_union)
327 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
328 lc->async_stack = (unsigned long)
329 memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE)
330 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
331 lc->panic_stack = (unsigned long)
332 memblock_virt_alloc(PAGE_SIZE, PAGE_SIZE)
333 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
334 lc->current_task = (unsigned long)&init_task;
335 lc->lpp = LPP_MAGIC;
336 lc->machine_flags = S390_lowcore.machine_flags;
337 lc->preempt_count = S390_lowcore.preempt_count;
338 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
339 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
340 sizeof(lc->stfle_fac_list));
341 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
342 sizeof(lc->alt_stfle_fac_list));
343 if (MACHINE_HAS_VX || MACHINE_HAS_GS) {
344 unsigned long bits, size;
345
346 bits = MACHINE_HAS_GS ? 11 : 10;
347 size = 1UL << bits;
348 lc->mcesad = (__u64) memblock_virt_alloc(size, size);
349 if (MACHINE_HAS_GS)
350 lc->mcesad |= bits;
351 }
352 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
353 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
354 lc->async_enter_timer = S390_lowcore.async_enter_timer;
355 lc->exit_timer = S390_lowcore.exit_timer;
356 lc->user_timer = S390_lowcore.user_timer;
357 lc->system_timer = S390_lowcore.system_timer;
358 lc->steal_timer = S390_lowcore.steal_timer;
359 lc->last_update_timer = S390_lowcore.last_update_timer;
360 lc->last_update_clock = S390_lowcore.last_update_clock;
361
362 restart_stack = memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE);
363 restart_stack += ASYNC_SIZE;
364
365 /*
366 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
367 * restart data to the absolute zero lowcore. This is necessary if
368 * PSW restart is done on an offline CPU that has lowcore zero.
369 */
370 lc->restart_stack = (unsigned long) restart_stack;
371 lc->restart_fn = (unsigned long) do_restart;
372 lc->restart_data = 0;
373 lc->restart_source = -1UL;
374
375 /* Setup absolute zero lowcore */
376 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
377 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
378 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
379 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
380 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
381
382 #ifdef CONFIG_SMP
383 lc->spinlock_lockval = arch_spin_lockval(0);
384 #endif
385 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
386
387 set_prefix((u32)(unsigned long) lc);
388 lowcore_ptr[0] = lc;
389 }
390
setup_lowcore_dat_on(void)391 static void __init setup_lowcore_dat_on(void)
392 {
393 __ctl_clear_bit(0, 28);
394 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
395 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
396 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
397 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
398 __ctl_set_bit(0, 28);
399 }
400
401 static struct resource code_resource = {
402 .name = "Kernel code",
403 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
404 };
405
406 static struct resource data_resource = {
407 .name = "Kernel data",
408 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
409 };
410
411 static struct resource bss_resource = {
412 .name = "Kernel bss",
413 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
414 };
415
416 static struct resource __initdata *standard_resources[] = {
417 &code_resource,
418 &data_resource,
419 &bss_resource,
420 };
421
setup_resources(void)422 static void __init setup_resources(void)
423 {
424 struct resource *res, *std_res, *sub_res;
425 struct memblock_region *reg;
426 int j;
427
428 code_resource.start = (unsigned long) &_text;
429 code_resource.end = (unsigned long) &_etext - 1;
430 data_resource.start = (unsigned long) &_etext;
431 data_resource.end = (unsigned long) &_edata - 1;
432 bss_resource.start = (unsigned long) &__bss_start;
433 bss_resource.end = (unsigned long) &__bss_stop - 1;
434
435 for_each_memblock(memory, reg) {
436 res = memblock_virt_alloc(sizeof(*res), 8);
437 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
438
439 res->name = "System RAM";
440 res->start = reg->base;
441 res->end = reg->base + reg->size - 1;
442 request_resource(&iomem_resource, res);
443
444 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
445 std_res = standard_resources[j];
446 if (std_res->start < res->start ||
447 std_res->start > res->end)
448 continue;
449 if (std_res->end > res->end) {
450 sub_res = memblock_virt_alloc(sizeof(*sub_res), 8);
451 *sub_res = *std_res;
452 sub_res->end = res->end;
453 std_res->start = res->end + 1;
454 request_resource(res, sub_res);
455 } else {
456 request_resource(res, std_res);
457 }
458 }
459 }
460 #ifdef CONFIG_CRASH_DUMP
461 /*
462 * Re-add removed crash kernel memory as reserved memory. This makes
463 * sure it will be mapped with the identity mapping and struct pages
464 * will be created, so it can be resized later on.
465 * However add it later since the crash kernel resource should not be
466 * part of the System RAM resource.
467 */
468 if (crashk_res.end) {
469 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
470 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
471 insert_resource(&iomem_resource, &crashk_res);
472 }
473 #endif
474 }
475
setup_memory_end(void)476 static void __init setup_memory_end(void)
477 {
478 unsigned long vmax, vmalloc_size, tmp;
479
480 /* Choose kernel address space layout: 2, 3, or 4 levels. */
481 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
482 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
483 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
484 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
485 vmax = _REGION2_SIZE; /* 3-level kernel page table */
486 else
487 vmax = _REGION1_SIZE; /* 4-level kernel page table */
488 /* module area is at the end of the kernel address space. */
489 MODULES_END = vmax;
490 MODULES_VADDR = MODULES_END - MODULES_LEN;
491 VMALLOC_END = MODULES_VADDR;
492 VMALLOC_START = vmax - vmalloc_size;
493
494 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
495 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
496 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
497 tmp = SECTION_ALIGN_UP(tmp);
498 tmp = VMALLOC_START - tmp * sizeof(struct page);
499 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
500 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
501 vmemmap = (struct page *) tmp;
502
503 /* Take care that memory_end is set and <= vmemmap */
504 memory_end = min(memory_end ?: max_physmem_end, tmp);
505 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
506 memblock_remove(memory_end, ULONG_MAX);
507
508 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
509 }
510
511 #ifdef CONFIG_CRASH_DUMP
512
513 /*
514 * When kdump is enabled, we have to ensure that no memory from
515 * the area [0 - crashkernel memory size] and
516 * [crashk_res.start - crashk_res.end] is set offline.
517 */
kdump_mem_notifier(struct notifier_block * nb,unsigned long action,void * data)518 static int kdump_mem_notifier(struct notifier_block *nb,
519 unsigned long action, void *data)
520 {
521 struct memory_notify *arg = data;
522
523 if (action != MEM_GOING_OFFLINE)
524 return NOTIFY_OK;
525 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
526 return NOTIFY_BAD;
527 if (arg->start_pfn > PFN_DOWN(crashk_res.end))
528 return NOTIFY_OK;
529 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
530 return NOTIFY_OK;
531 return NOTIFY_BAD;
532 }
533
534 static struct notifier_block kdump_mem_nb = {
535 .notifier_call = kdump_mem_notifier,
536 };
537
538 #endif
539
540 /*
541 * Make sure that the area behind memory_end is protected
542 */
reserve_memory_end(void)543 static void reserve_memory_end(void)
544 {
545 #ifdef CONFIG_CRASH_DUMP
546 if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
547 !OLDMEM_BASE && sclp.hsa_size) {
548 memory_end = sclp.hsa_size;
549 memory_end &= PAGE_MASK;
550 memory_end_set = 1;
551 }
552 #endif
553 if (!memory_end_set)
554 return;
555 memblock_reserve(memory_end, ULONG_MAX);
556 }
557
558 /*
559 * Make sure that oldmem, where the dump is stored, is protected
560 */
reserve_oldmem(void)561 static void reserve_oldmem(void)
562 {
563 #ifdef CONFIG_CRASH_DUMP
564 if (OLDMEM_BASE)
565 /* Forget all memory above the running kdump system */
566 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
567 #endif
568 }
569
570 /*
571 * Make sure that oldmem, where the dump is stored, is protected
572 */
remove_oldmem(void)573 static void remove_oldmem(void)
574 {
575 #ifdef CONFIG_CRASH_DUMP
576 if (OLDMEM_BASE)
577 /* Forget all memory above the running kdump system */
578 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
579 #endif
580 }
581
582 /*
583 * Reserve memory for kdump kernel to be loaded with kexec
584 */
reserve_crashkernel(void)585 static void __init reserve_crashkernel(void)
586 {
587 #ifdef CONFIG_CRASH_DUMP
588 unsigned long long crash_base, crash_size;
589 phys_addr_t low, high;
590 int rc;
591
592 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
593 &crash_base);
594
595 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
596 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
597 if (rc || crash_size == 0)
598 return;
599
600 if (memblock.memory.regions[0].size < crash_size) {
601 pr_info("crashkernel reservation failed: %s\n",
602 "first memory chunk must be at least crashkernel size");
603 return;
604 }
605
606 low = crash_base ?: OLDMEM_BASE;
607 high = low + crash_size;
608 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
609 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
610 crash_base = low;
611 } else {
612 /* Find suitable area in free memory */
613 low = max_t(unsigned long, crash_size, sclp.hsa_size);
614 high = crash_base ? crash_base + crash_size : ULONG_MAX;
615
616 if (crash_base && crash_base < low) {
617 pr_info("crashkernel reservation failed: %s\n",
618 "crash_base too low");
619 return;
620 }
621 low = crash_base ?: low;
622 crash_base = memblock_find_in_range(low, high, crash_size,
623 KEXEC_CRASH_MEM_ALIGN);
624 }
625
626 if (!crash_base) {
627 pr_info("crashkernel reservation failed: %s\n",
628 "no suitable area found");
629 return;
630 }
631
632 if (register_memory_notifier(&kdump_mem_nb))
633 return;
634
635 if (!OLDMEM_BASE && MACHINE_IS_VM)
636 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
637 crashk_res.start = crash_base;
638 crashk_res.end = crash_base + crash_size - 1;
639 memblock_remove(crash_base, crash_size);
640 pr_info("Reserving %lluMB of memory at %lluMB "
641 "for crashkernel (System RAM: %luMB)\n",
642 crash_size >> 20, crash_base >> 20,
643 (unsigned long)memblock.memory.total_size >> 20);
644 os_info_crashkernel_add(crash_base, crash_size);
645 #endif
646 }
647
648 /*
649 * Reserve the initrd from being used by memblock
650 */
reserve_initrd(void)651 static void __init reserve_initrd(void)
652 {
653 #ifdef CONFIG_BLK_DEV_INITRD
654 if (!INITRD_START || !INITRD_SIZE)
655 return;
656 initrd_start = INITRD_START;
657 initrd_end = initrd_start + INITRD_SIZE;
658 memblock_reserve(INITRD_START, INITRD_SIZE);
659 #endif
660 }
661
662 /*
663 * Check for initrd being in usable memory
664 */
check_initrd(void)665 static void __init check_initrd(void)
666 {
667 #ifdef CONFIG_BLK_DEV_INITRD
668 if (INITRD_START && INITRD_SIZE &&
669 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
670 pr_err("The initial RAM disk does not fit into the memory\n");
671 memblock_free(INITRD_START, INITRD_SIZE);
672 initrd_start = initrd_end = 0;
673 }
674 #endif
675 }
676
677 /*
678 * Reserve memory used for lowcore/command line/kernel image.
679 */
reserve_kernel(void)680 static void __init reserve_kernel(void)
681 {
682 unsigned long start_pfn = PFN_UP(__pa(&_end));
683
684 #ifdef CONFIG_DMA_API_DEBUG
685 /*
686 * DMA_API_DEBUG code stumbles over addresses from the
687 * range [_ehead, _stext]. Mark the memory as reserved
688 * so it is not used for CONFIG_DMA_API_DEBUG=y.
689 */
690 memblock_reserve(0, PFN_PHYS(start_pfn));
691 #else
692 memblock_reserve(0, (unsigned long)_ehead);
693 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
694 - (unsigned long)_stext);
695 #endif
696 }
697
setup_memory(void)698 static void __init setup_memory(void)
699 {
700 struct memblock_region *reg;
701
702 /*
703 * Init storage key for present memory
704 */
705 for_each_memblock(memory, reg) {
706 storage_key_init_range(reg->base, reg->base + reg->size);
707 }
708 psw_set_key(PAGE_DEFAULT_KEY);
709
710 /* Only cosmetics */
711 memblock_enforce_memory_limit(memblock_end_of_DRAM());
712 }
713
714 /*
715 * Setup hardware capabilities.
716 */
setup_hwcaps(void)717 static int __init setup_hwcaps(void)
718 {
719 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
720 struct cpuid cpu_id;
721 int i;
722
723 /*
724 * The store facility list bits numbers as found in the principles
725 * of operation are numbered with bit 1UL<<31 as number 0 to
726 * bit 1UL<<0 as number 31.
727 * Bit 0: instructions named N3, "backported" to esa-mode
728 * Bit 2: z/Architecture mode is active
729 * Bit 7: the store-facility-list-extended facility is installed
730 * Bit 17: the message-security assist is installed
731 * Bit 19: the long-displacement facility is installed
732 * Bit 21: the extended-immediate facility is installed
733 * Bit 22: extended-translation facility 3 is installed
734 * Bit 30: extended-translation facility 3 enhancement facility
735 * These get translated to:
736 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
737 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
738 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
739 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
740 */
741 for (i = 0; i < 6; i++)
742 if (test_facility(stfl_bits[i]))
743 elf_hwcap |= 1UL << i;
744
745 if (test_facility(22) && test_facility(30))
746 elf_hwcap |= HWCAP_S390_ETF3EH;
747
748 /*
749 * Check for additional facilities with store-facility-list-extended.
750 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
751 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
752 * as stored by stfl, bits 32-xxx contain additional facilities.
753 * How many facility words are stored depends on the number of
754 * doublewords passed to the instruction. The additional facilities
755 * are:
756 * Bit 42: decimal floating point facility is installed
757 * Bit 44: perform floating point operation facility is installed
758 * translated to:
759 * HWCAP_S390_DFP bit 6 (42 && 44).
760 */
761 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
762 elf_hwcap |= HWCAP_S390_DFP;
763
764 /*
765 * Huge page support HWCAP_S390_HPAGE is bit 7.
766 */
767 if (MACHINE_HAS_EDAT1)
768 elf_hwcap |= HWCAP_S390_HPAGE;
769
770 /*
771 * 64-bit register support for 31-bit processes
772 * HWCAP_S390_HIGH_GPRS is bit 9.
773 */
774 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
775
776 /*
777 * Transactional execution support HWCAP_S390_TE is bit 10.
778 */
779 if (test_facility(50) && test_facility(73))
780 elf_hwcap |= HWCAP_S390_TE;
781
782 /*
783 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
784 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
785 * instead of facility bit 129.
786 */
787 if (MACHINE_HAS_VX) {
788 elf_hwcap |= HWCAP_S390_VXRS;
789 if (test_facility(134))
790 elf_hwcap |= HWCAP_S390_VXRS_EXT;
791 if (test_facility(135))
792 elf_hwcap |= HWCAP_S390_VXRS_BCD;
793 }
794
795 /*
796 * Guarded storage support HWCAP_S390_GS is bit 12.
797 */
798 if (MACHINE_HAS_GS)
799 elf_hwcap |= HWCAP_S390_GS;
800
801 get_cpu_id(&cpu_id);
802 add_device_randomness(&cpu_id, sizeof(cpu_id));
803 switch (cpu_id.machine) {
804 case 0x2064:
805 case 0x2066:
806 default: /* Use "z900" as default for 64 bit kernels. */
807 strcpy(elf_platform, "z900");
808 break;
809 case 0x2084:
810 case 0x2086:
811 strcpy(elf_platform, "z990");
812 break;
813 case 0x2094:
814 case 0x2096:
815 strcpy(elf_platform, "z9-109");
816 break;
817 case 0x2097:
818 case 0x2098:
819 strcpy(elf_platform, "z10");
820 break;
821 case 0x2817:
822 case 0x2818:
823 strcpy(elf_platform, "z196");
824 break;
825 case 0x2827:
826 case 0x2828:
827 strcpy(elf_platform, "zEC12");
828 break;
829 case 0x2964:
830 case 0x2965:
831 strcpy(elf_platform, "z13");
832 break;
833 case 0x3906:
834 strcpy(elf_platform, "z14");
835 break;
836 }
837
838 /*
839 * Virtualization support HWCAP_INT_SIE is bit 0.
840 */
841 if (sclp.has_sief2)
842 int_hwcap |= HWCAP_INT_SIE;
843
844 return 0;
845 }
846 arch_initcall(setup_hwcaps);
847
848 /*
849 * Add system information as device randomness
850 */
setup_randomness(void)851 static void __init setup_randomness(void)
852 {
853 struct sysinfo_3_2_2 *vmms;
854
855 vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE);
856 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
857 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
858 memblock_free((unsigned long) vmms, PAGE_SIZE);
859 }
860
861 /*
862 * Find the correct size for the task_struct. This depends on
863 * the size of the struct fpu at the end of the thread_struct
864 * which is embedded in the task_struct.
865 */
setup_task_size(void)866 static void __init setup_task_size(void)
867 {
868 int task_size = sizeof(struct task_struct);
869
870 if (!MACHINE_HAS_VX) {
871 task_size -= sizeof(__vector128) * __NUM_VXRS;
872 task_size += sizeof(freg_t) * __NUM_FPRS;
873 }
874 arch_task_struct_size = task_size;
875 }
876
877 /*
878 * Setup function called from init/main.c just after the banner
879 * was printed.
880 */
881
setup_arch(char ** cmdline_p)882 void __init setup_arch(char **cmdline_p)
883 {
884 /*
885 * print what head.S has found out about the machine
886 */
887 if (MACHINE_IS_VM)
888 pr_info("Linux is running as a z/VM "
889 "guest operating system in 64-bit mode\n");
890 else if (MACHINE_IS_KVM)
891 pr_info("Linux is running under KVM in 64-bit mode\n");
892 else if (MACHINE_IS_LPAR)
893 pr_info("Linux is running natively in 64-bit mode\n");
894 else
895 pr_info("Linux is running as a guest in 64-bit mode\n");
896
897 /* Have one command line that is parsed and saved in /proc/cmdline */
898 /* boot_command_line has been already set up in early.c */
899 *cmdline_p = boot_command_line;
900
901 ROOT_DEV = Root_RAM0;
902
903 /* Is init_mm really needed? */
904 init_mm.start_code = PAGE_OFFSET;
905 init_mm.end_code = (unsigned long) &_etext;
906 init_mm.end_data = (unsigned long) &_edata;
907 init_mm.brk = (unsigned long) &_end;
908
909 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
910 nospec_auto_detect();
911
912 parse_early_param();
913 #ifdef CONFIG_CRASH_DUMP
914 /* Deactivate elfcorehdr= kernel parameter */
915 elfcorehdr_addr = ELFCORE_ADDR_MAX;
916 #endif
917
918 os_info_init();
919 setup_ipl();
920 setup_task_size();
921
922 /* Do some memory reservations *before* memory is added to memblock */
923 reserve_memory_end();
924 reserve_oldmem();
925 reserve_kernel();
926 reserve_initrd();
927 memblock_allow_resize();
928
929 /* Get information about *all* installed memory */
930 detect_memory_memblock();
931
932 remove_oldmem();
933
934 /*
935 * Make sure all chunks are MAX_ORDER aligned so we don't need the
936 * extra checks that HOLES_IN_ZONE would require.
937 *
938 * Is this still required?
939 */
940 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
941
942 setup_memory_end();
943 setup_memory();
944 dma_contiguous_reserve(memory_end);
945 vmcp_cma_reserve();
946
947 check_initrd();
948 reserve_crashkernel();
949 #ifdef CONFIG_CRASH_DUMP
950 /*
951 * Be aware that smp_save_dump_cpus() triggers a system reset.
952 * Therefore CPU and device initialization should be done afterwards.
953 */
954 smp_save_dump_cpus();
955 #endif
956
957 setup_resources();
958 setup_lowcore_dat_off();
959 smp_fill_possible_mask();
960 cpu_detect_mhz_feature();
961 cpu_init();
962 numa_setup();
963 smp_detect_cpus();
964 topology_init_early();
965
966 /*
967 * Create kernel page tables and switch to virtual addressing.
968 */
969 paging_init();
970
971 /*
972 * After paging_init created the kernel page table, the new PSWs
973 * in lowcore can now run with DAT enabled.
974 */
975 setup_lowcore_dat_on();
976
977 /* Setup default console */
978 conmode_default();
979 set_preferred_console();
980
981 apply_alternative_instructions();
982 if (IS_ENABLED(CONFIG_EXPOLINE))
983 nospec_init_branches();
984
985 /* Setup zfcpdump support */
986 setup_zfcpdump();
987
988 /* Add system specific data to the random pool */
989 setup_randomness();
990 }
991