]> Git Repo - linux.git/blob - arch/s390/kernel/setup.c
s390/mm: add support to allocate gigantic hugepages using CMA
[linux.git] / arch / s390 / kernel / setup.c
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  S390 version
4  *    Copyright IBM Corp. 1999, 2012
5  *    Author(s): Hartmut Penner ([email protected]),
6  *               Martin Schwidefsky ([email protected])
7  *
8  *  Derived from "arch/i386/kernel/setup.c"
9  *    Copyright (C) 1995, Linus Torvalds
10  */
11
12 /*
13  * This file handles the architecture-dependent parts of initialization
14  */
15
16 #define KMSG_COMPONENT "setup"
17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
18
19 #include <linux/errno.h>
20 #include <linux/export.h>
21 #include <linux/sched.h>
22 #include <linux/sched/task.h>
23 #include <linux/cpu.h>
24 #include <linux/kernel.h>
25 #include <linux/memblock.h>
26 #include <linux/mm.h>
27 #include <linux/stddef.h>
28 #include <linux/unistd.h>
29 #include <linux/ptrace.h>
30 #include <linux/random.h>
31 #include <linux/user.h>
32 #include <linux/tty.h>
33 #include <linux/ioport.h>
34 #include <linux/delay.h>
35 #include <linux/init.h>
36 #include <linux/initrd.h>
37 #include <linux/root_dev.h>
38 #include <linux/console.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/dma-map-ops.h>
41 #include <linux/device.h>
42 #include <linux/notifier.h>
43 #include <linux/pfn.h>
44 #include <linux/ctype.h>
45 #include <linux/reboot.h>
46 #include <linux/topology.h>
47 #include <linux/kexec.h>
48 #include <linux/crash_dump.h>
49 #include <linux/memory.h>
50 #include <linux/compat.h>
51 #include <linux/start_kernel.h>
52 #include <linux/hugetlb.h>
53
54 #include <asm/boot_data.h>
55 #include <asm/ipl.h>
56 #include <asm/facility.h>
57 #include <asm/smp.h>
58 #include <asm/mmu_context.h>
59 #include <asm/cpcmd.h>
60 #include <asm/lowcore.h>
61 #include <asm/nmi.h>
62 #include <asm/irq.h>
63 #include <asm/page.h>
64 #include <asm/ptrace.h>
65 #include <asm/sections.h>
66 #include <asm/ebcdic.h>
67 #include <asm/diag.h>
68 #include <asm/os_info.h>
69 #include <asm/sclp.h>
70 #include <asm/stacktrace.h>
71 #include <asm/sysinfo.h>
72 #include <asm/numa.h>
73 #include <asm/alternative.h>
74 #include <asm/nospec-branch.h>
75 #include <asm/mem_detect.h>
76 #include <asm/uv.h>
77 #include <asm/asm-offsets.h>
78 #include "entry.h"
79
80 /*
81  * Machine setup..
82  */
83 unsigned int console_mode = 0;
84 EXPORT_SYMBOL(console_mode);
85
86 unsigned int console_devno = -1;
87 EXPORT_SYMBOL(console_devno);
88
89 unsigned int console_irq = -1;
90 EXPORT_SYMBOL(console_irq);
91
92 unsigned long elf_hwcap __read_mostly = 0;
93 char elf_platform[ELF_PLATFORM_SIZE];
94
95 unsigned long int_hwcap = 0;
96
97 int __bootdata(noexec_disabled);
98 unsigned long __bootdata(ident_map_size);
99 unsigned long __bootdata(vmalloc_size);
100 struct mem_detect_info __bootdata(mem_detect);
101
102 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
103 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
104 unsigned long __bootdata_preserved(__stext_dma);
105 unsigned long __bootdata_preserved(__etext_dma);
106 unsigned long __bootdata_preserved(__sdma);
107 unsigned long __bootdata_preserved(__edma);
108 unsigned long __bootdata_preserved(__kaslr_offset);
109 unsigned int __bootdata_preserved(zlib_dfltcc_support);
110 EXPORT_SYMBOL(zlib_dfltcc_support);
111
112 unsigned long VMALLOC_START;
113 EXPORT_SYMBOL(VMALLOC_START);
114
115 unsigned long VMALLOC_END;
116 EXPORT_SYMBOL(VMALLOC_END);
117
118 struct page *vmemmap;
119 EXPORT_SYMBOL(vmemmap);
120 unsigned long vmemmap_size;
121
122 unsigned long MODULES_VADDR;
123 unsigned long MODULES_END;
124
125 /* An array with a pointer to the lowcore of every CPU. */
126 struct lowcore *lowcore_ptr[NR_CPUS];
127 EXPORT_SYMBOL(lowcore_ptr);
128
129 /*
130  * The Write Back bit position in the physaddr is given by the SLPC PCI.
131  * Leaving the mask zero always uses write through which is safe
132  */
133 unsigned long mio_wb_bit_mask __ro_after_init;
134
135 /*
136  * This is set up by the setup-routine at boot-time
137  * for S390 need to find out, what we have to setup
138  * using address 0x10400 ...
139  */
140
141 #include <asm/setup.h>
142
143 /*
144  * condev= and conmode= setup parameter.
145  */
146
147 static int __init condev_setup(char *str)
148 {
149         int vdev;
150
151         vdev = simple_strtoul(str, &str, 0);
152         if (vdev >= 0 && vdev < 65536) {
153                 console_devno = vdev;
154                 console_irq = -1;
155         }
156         return 1;
157 }
158
159 __setup("condev=", condev_setup);
160
161 static void __init set_preferred_console(void)
162 {
163         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
164                 add_preferred_console("ttyS", 0, NULL);
165         else if (CONSOLE_IS_3270)
166                 add_preferred_console("tty3270", 0, NULL);
167         else if (CONSOLE_IS_VT220)
168                 add_preferred_console("ttyS", 1, NULL);
169         else if (CONSOLE_IS_HVC)
170                 add_preferred_console("hvc", 0, NULL);
171 }
172
173 static int __init conmode_setup(char *str)
174 {
175 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
176         if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
177                 SET_CONSOLE_SCLP;
178 #endif
179 #if defined(CONFIG_TN3215_CONSOLE)
180         if (!strcmp(str, "3215"))
181                 SET_CONSOLE_3215;
182 #endif
183 #if defined(CONFIG_TN3270_CONSOLE)
184         if (!strcmp(str, "3270"))
185                 SET_CONSOLE_3270;
186 #endif
187         set_preferred_console();
188         return 1;
189 }
190
191 __setup("conmode=", conmode_setup);
192
193 static void __init conmode_default(void)
194 {
195         char query_buffer[1024];
196         char *ptr;
197
198         if (MACHINE_IS_VM) {
199                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
200                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
201                 ptr = strstr(query_buffer, "SUBCHANNEL =");
202                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
203                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
204                 ptr = strstr(query_buffer, "CONMODE");
205                 /*
206                  * Set the conmode to 3215 so that the device recognition 
207                  * will set the cu_type of the console to 3215. If the
208                  * conmode is 3270 and we don't set it back then both
209                  * 3215 and the 3270 driver will try to access the console
210                  * device (3215 as console and 3270 as normal tty).
211                  */
212                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
213                 if (ptr == NULL) {
214 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
215                         SET_CONSOLE_SCLP;
216 #endif
217                         return;
218                 }
219                 if (str_has_prefix(ptr + 8, "3270")) {
220 #if defined(CONFIG_TN3270_CONSOLE)
221                         SET_CONSOLE_3270;
222 #elif defined(CONFIG_TN3215_CONSOLE)
223                         SET_CONSOLE_3215;
224 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
225                         SET_CONSOLE_SCLP;
226 #endif
227                 } else if (str_has_prefix(ptr + 8, "3215")) {
228 #if defined(CONFIG_TN3215_CONSOLE)
229                         SET_CONSOLE_3215;
230 #elif defined(CONFIG_TN3270_CONSOLE)
231                         SET_CONSOLE_3270;
232 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
233                         SET_CONSOLE_SCLP;
234 #endif
235                 }
236         } else if (MACHINE_IS_KVM) {
237                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
238                         SET_CONSOLE_VT220;
239                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
240                         SET_CONSOLE_SCLP;
241                 else
242                         SET_CONSOLE_HVC;
243         } else {
244 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
245                 SET_CONSOLE_SCLP;
246 #endif
247         }
248 }
249
250 #ifdef CONFIG_CRASH_DUMP
251 static void __init setup_zfcpdump(void)
252 {
253         if (!is_ipl_type_dump())
254                 return;
255         if (OLDMEM_BASE)
256                 return;
257         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
258         console_loglevel = 2;
259 }
260 #else
261 static inline void setup_zfcpdump(void) {}
262 #endif /* CONFIG_CRASH_DUMP */
263
264  /*
265  * Reboot, halt and power_off stubs. They just call _machine_restart,
266  * _machine_halt or _machine_power_off. 
267  */
268
269 void machine_restart(char *command)
270 {
271         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
272                 /*
273                  * Only unblank the console if we are called in enabled
274                  * context or a bust_spinlocks cleared the way for us.
275                  */
276                 console_unblank();
277         _machine_restart(command);
278 }
279
280 void machine_halt(void)
281 {
282         if (!in_interrupt() || oops_in_progress)
283                 /*
284                  * Only unblank the console if we are called in enabled
285                  * context or a bust_spinlocks cleared the way for us.
286                  */
287                 console_unblank();
288         _machine_halt();
289 }
290
291 void machine_power_off(void)
292 {
293         if (!in_interrupt() || oops_in_progress)
294                 /*
295                  * Only unblank the console if we are called in enabled
296                  * context or a bust_spinlocks cleared the way for us.
297                  */
298                 console_unblank();
299         _machine_power_off();
300 }
301
302 /*
303  * Dummy power off function.
304  */
305 void (*pm_power_off)(void) = machine_power_off;
306 EXPORT_SYMBOL_GPL(pm_power_off);
307
308 void *restart_stack;
309
310 unsigned long stack_alloc(void)
311 {
312 #ifdef CONFIG_VMAP_STACK
313         return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE,
314                         THREADINFO_GFP, NUMA_NO_NODE,
315                         __builtin_return_address(0));
316 #else
317         return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
318 #endif
319 }
320
321 void stack_free(unsigned long stack)
322 {
323 #ifdef CONFIG_VMAP_STACK
324         vfree((void *) stack);
325 #else
326         free_pages(stack, THREAD_SIZE_ORDER);
327 #endif
328 }
329
330 int __init arch_early_irq_init(void)
331 {
332         unsigned long stack;
333
334         stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
335         if (!stack)
336                 panic("Couldn't allocate async stack");
337         S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
338         udelay_enable();
339         return 0;
340 }
341
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
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
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
458         set_prefix((u32)(unsigned long) lc);
459         lowcore_ptr[0] = lc;
460 }
461
462 static void __init setup_lowcore_dat_on(void)
463 {
464         __ctl_clear_bit(0, 28);
465         S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
466         S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
467         S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
468         S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
469         __ctl_set_bit(0, 28);
470 }
471
472 static struct resource code_resource = {
473         .name  = "Kernel code",
474         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
475 };
476
477 static struct resource data_resource = {
478         .name = "Kernel data",
479         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
480 };
481
482 static struct resource bss_resource = {
483         .name = "Kernel bss",
484         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
485 };
486
487 static struct resource __initdata *standard_resources[] = {
488         &code_resource,
489         &data_resource,
490         &bss_resource,
491 };
492
493 static void __init setup_resources(void)
494 {
495         struct resource *res, *std_res, *sub_res;
496         phys_addr_t start, end;
497         int j;
498         u64 i;
499
500         code_resource.start = (unsigned long) _text;
501         code_resource.end = (unsigned long) _etext - 1;
502         data_resource.start = (unsigned long) _etext;
503         data_resource.end = (unsigned long) _edata - 1;
504         bss_resource.start = (unsigned long) __bss_start;
505         bss_resource.end = (unsigned long) __bss_stop - 1;
506
507         for_each_mem_range(i, &start, &end) {
508                 res = memblock_alloc(sizeof(*res), 8);
509                 if (!res)
510                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
511                               __func__, sizeof(*res), 8);
512                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
513
514                 res->name = "System RAM";
515                 res->start = start;
516                 /*
517                  * In memblock, end points to the first byte after the
518                  * range while in resourses, end points to the last byte in
519                  * the range.
520                  */
521                 res->end = end - 1;
522                 request_resource(&iomem_resource, res);
523
524                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
525                         std_res = standard_resources[j];
526                         if (std_res->start < res->start ||
527                             std_res->start > res->end)
528                                 continue;
529                         if (std_res->end > res->end) {
530                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
531                                 if (!sub_res)
532                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
533                                               __func__, sizeof(*sub_res), 8);
534                                 *sub_res = *std_res;
535                                 sub_res->end = res->end;
536                                 std_res->start = res->end + 1;
537                                 request_resource(res, sub_res);
538                         } else {
539                                 request_resource(res, std_res);
540                         }
541                 }
542         }
543 #ifdef CONFIG_CRASH_DUMP
544         /*
545          * Re-add removed crash kernel memory as reserved memory. This makes
546          * sure it will be mapped with the identity mapping and struct pages
547          * will be created, so it can be resized later on.
548          * However add it later since the crash kernel resource should not be
549          * part of the System RAM resource.
550          */
551         if (crashk_res.end) {
552                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
553                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
554                 insert_resource(&iomem_resource, &crashk_res);
555         }
556 #endif
557 }
558
559 static void __init setup_ident_map_size(void)
560 {
561         unsigned long vmax, tmp;
562
563         /* Choose kernel address space layout: 3 or 4 levels. */
564         tmp = ident_map_size / PAGE_SIZE;
565         tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
566         if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
567                 vmax = _REGION2_SIZE; /* 3-level kernel page table */
568         else
569                 vmax = _REGION1_SIZE; /* 4-level kernel page table */
570         /* module area is at the end of the kernel address space. */
571         MODULES_END = vmax;
572         if (is_prot_virt_host())
573                 adjust_to_uv_max(&MODULES_END);
574 #ifdef CONFIG_KASAN
575         vmax = _REGION1_SIZE;
576         MODULES_END = kasan_vmax;
577 #endif
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 ident_map_size <= vmemmap */
592         ident_map_size = min(ident_map_size, (unsigned long)vmemmap);
593 #ifdef CONFIG_KASAN
594         ident_map_size = min(ident_map_size, KASAN_SHADOW_START);
595 #endif
596         vmemmap_size = SECTION_ALIGN_UP(ident_map_size / 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(ident_map_size);
604         memblock_remove(ident_map_size, ULONG_MAX);
605
606         pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 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  */
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 above identity mapping is protected
637  */
638 static void __init reserve_above_ident_map(void)
639 {
640         memblock_reserve(ident_map_size, ULONG_MAX);
641 }
642
643 /*
644  * Make sure that oldmem, where the dump is stored, is protected
645  */
646 static void __init reserve_oldmem(void)
647 {
648 #ifdef CONFIG_CRASH_DUMP
649         if (OLDMEM_BASE)
650                 /* Forget all memory above the running kdump system */
651                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
652 #endif
653 }
654
655 /*
656  * Make sure that oldmem, where the dump is stored, is protected
657  */
658 static void __init remove_oldmem(void)
659 {
660 #ifdef CONFIG_CRASH_DUMP
661         if (OLDMEM_BASE)
662                 /* Forget all memory above the running kdump system */
663                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
664 #endif
665 }
666
667 /*
668  * Reserve memory for kdump kernel to be loaded with kexec
669  */
670 static void __init reserve_crashkernel(void)
671 {
672 #ifdef CONFIG_CRASH_DUMP
673         unsigned long long crash_base, crash_size;
674         phys_addr_t low, high;
675         int rc;
676
677         rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
678                                &crash_base);
679
680         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
681         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
682         if (rc || crash_size == 0)
683                 return;
684
685         if (memblock.memory.regions[0].size < crash_size) {
686                 pr_info("crashkernel reservation failed: %s\n",
687                         "first memory chunk must be at least crashkernel size");
688                 return;
689         }
690
691         low = crash_base ?: OLDMEM_BASE;
692         high = low + crash_size;
693         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
694                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
695                 crash_base = low;
696         } else {
697                 /* Find suitable area in free memory */
698                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
699                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
700
701                 if (crash_base && crash_base < low) {
702                         pr_info("crashkernel reservation failed: %s\n",
703                                 "crash_base too low");
704                         return;
705                 }
706                 low = crash_base ?: low;
707                 crash_base = memblock_find_in_range(low, high, crash_size,
708                                                     KEXEC_CRASH_MEM_ALIGN);
709         }
710
711         if (!crash_base) {
712                 pr_info("crashkernel reservation failed: %s\n",
713                         "no suitable area found");
714                 return;
715         }
716
717         if (register_memory_notifier(&kdump_mem_nb))
718                 return;
719
720         if (!OLDMEM_BASE && MACHINE_IS_VM)
721                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
722         crashk_res.start = crash_base;
723         crashk_res.end = crash_base + crash_size - 1;
724         memblock_remove(crash_base, crash_size);
725         pr_info("Reserving %lluMB of memory at %lluMB "
726                 "for crashkernel (System RAM: %luMB)\n",
727                 crash_size >> 20, crash_base >> 20,
728                 (unsigned long)memblock.memory.total_size >> 20);
729         os_info_crashkernel_add(crash_base, crash_size);
730 #endif
731 }
732
733 /*
734  * Reserve the initrd from being used by memblock
735  */
736 static void __init reserve_initrd(void)
737 {
738 #ifdef CONFIG_BLK_DEV_INITRD
739         if (!INITRD_START || !INITRD_SIZE)
740                 return;
741         initrd_start = INITRD_START;
742         initrd_end = initrd_start + INITRD_SIZE;
743         memblock_reserve(INITRD_START, INITRD_SIZE);
744 #endif
745 }
746
747 /*
748  * Reserve the memory area used to pass the certificate lists
749  */
750 static void __init reserve_certificate_list(void)
751 {
752         if (ipl_cert_list_addr)
753                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
754 }
755
756 static void __init reserve_mem_detect_info(void)
757 {
758         unsigned long start, size;
759
760         get_mem_detect_reserved(&start, &size);
761         if (size)
762                 memblock_reserve(start, size);
763 }
764
765 static void __init free_mem_detect_info(void)
766 {
767         unsigned long start, size;
768
769         get_mem_detect_reserved(&start, &size);
770         if (size)
771                 memblock_free(start, size);
772 }
773
774 static const char * __init get_mem_info_source(void)
775 {
776         switch (mem_detect.info_source) {
777         case MEM_DETECT_SCLP_STOR_INFO:
778                 return "sclp storage info";
779         case MEM_DETECT_DIAG260:
780                 return "diag260";
781         case MEM_DETECT_SCLP_READ_INFO:
782                 return "sclp read info";
783         case MEM_DETECT_BIN_SEARCH:
784                 return "binary search";
785         }
786         return "none";
787 }
788
789 static void __init memblock_add_mem_detect_info(void)
790 {
791         unsigned long start, end;
792         int i;
793
794         pr_debug("physmem info source: %s (%hhd)\n",
795                  get_mem_info_source(), mem_detect.info_source);
796         /* keep memblock lists close to the kernel */
797         memblock_set_bottom_up(true);
798         for_each_mem_detect_block(i, &start, &end) {
799                 memblock_add(start, end - start);
800                 memblock_physmem_add(start, end - start);
801         }
802         memblock_set_bottom_up(false);
803         memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
804         memblock_dump_all();
805 }
806
807 /*
808  * Check for initrd being in usable memory
809  */
810 static void __init check_initrd(void)
811 {
812 #ifdef CONFIG_BLK_DEV_INITRD
813         if (INITRD_START && INITRD_SIZE &&
814             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
815                 pr_err("The initial RAM disk does not fit into the memory\n");
816                 memblock_free(INITRD_START, INITRD_SIZE);
817                 initrd_start = initrd_end = 0;
818         }
819 #endif
820 }
821
822 /*
823  * Reserve memory used for lowcore/command line/kernel image.
824  */
825 static void __init reserve_kernel(void)
826 {
827         unsigned long start_pfn = PFN_UP(__pa(_end));
828
829         memblock_reserve(0, HEAD_END);
830         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
831                          - (unsigned long)_stext);
832         memblock_reserve(__sdma, __edma - __sdma);
833 }
834
835 static void __init setup_memory(void)
836 {
837         phys_addr_t start, end;
838         u64 i;
839
840         /*
841          * Init storage key for present memory
842          */
843         for_each_mem_range(i, &start, &end)
844                 storage_key_init_range(start, end);
845
846         psw_set_key(PAGE_DEFAULT_KEY);
847
848         /* Only cosmetics */
849         memblock_enforce_memory_limit(memblock_end_of_DRAM());
850 }
851
852 /*
853  * Setup hardware capabilities.
854  */
855 static int __init setup_hwcaps(void)
856 {
857         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
858         struct cpuid cpu_id;
859         int i;
860
861         /*
862          * The store facility list bits numbers as found in the principles
863          * of operation are numbered with bit 1UL<<31 as number 0 to
864          * bit 1UL<<0 as number 31.
865          *   Bit 0: instructions named N3, "backported" to esa-mode
866          *   Bit 2: z/Architecture mode is active
867          *   Bit 7: the store-facility-list-extended facility is installed
868          *   Bit 17: the message-security assist is installed
869          *   Bit 19: the long-displacement facility is installed
870          *   Bit 21: the extended-immediate facility is installed
871          *   Bit 22: extended-translation facility 3 is installed
872          *   Bit 30: extended-translation facility 3 enhancement facility
873          * These get translated to:
874          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
875          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
876          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
877          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
878          */
879         for (i = 0; i < 6; i++)
880                 if (test_facility(stfl_bits[i]))
881                         elf_hwcap |= 1UL << i;
882
883         if (test_facility(22) && test_facility(30))
884                 elf_hwcap |= HWCAP_S390_ETF3EH;
885
886         /*
887          * Check for additional facilities with store-facility-list-extended.
888          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
889          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
890          * as stored by stfl, bits 32-xxx contain additional facilities.
891          * How many facility words are stored depends on the number of
892          * doublewords passed to the instruction. The additional facilities
893          * are:
894          *   Bit 42: decimal floating point facility is installed
895          *   Bit 44: perform floating point operation facility is installed
896          * translated to:
897          *   HWCAP_S390_DFP bit 6 (42 && 44).
898          */
899         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
900                 elf_hwcap |= HWCAP_S390_DFP;
901
902         /*
903          * Huge page support HWCAP_S390_HPAGE is bit 7.
904          */
905         if (MACHINE_HAS_EDAT1)
906                 elf_hwcap |= HWCAP_S390_HPAGE;
907
908         /*
909          * 64-bit register support for 31-bit processes
910          * HWCAP_S390_HIGH_GPRS is bit 9.
911          */
912         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
913
914         /*
915          * Transactional execution support HWCAP_S390_TE is bit 10.
916          */
917         if (MACHINE_HAS_TE)
918                 elf_hwcap |= HWCAP_S390_TE;
919
920         /*
921          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
922          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
923          * instead of facility bit 129.
924          */
925         if (MACHINE_HAS_VX) {
926                 elf_hwcap |= HWCAP_S390_VXRS;
927                 if (test_facility(134))
928                         elf_hwcap |= HWCAP_S390_VXRS_EXT;
929                 if (test_facility(135))
930                         elf_hwcap |= HWCAP_S390_VXRS_BCD;
931                 if (test_facility(148))
932                         elf_hwcap |= HWCAP_S390_VXRS_EXT2;
933                 if (test_facility(152))
934                         elf_hwcap |= HWCAP_S390_VXRS_PDE;
935         }
936         if (test_facility(150))
937                 elf_hwcap |= HWCAP_S390_SORT;
938         if (test_facility(151))
939                 elf_hwcap |= HWCAP_S390_DFLT;
940
941         /*
942          * Guarded storage support HWCAP_S390_GS is bit 12.
943          */
944         if (MACHINE_HAS_GS)
945                 elf_hwcap |= HWCAP_S390_GS;
946
947         get_cpu_id(&cpu_id);
948         add_device_randomness(&cpu_id, sizeof(cpu_id));
949         switch (cpu_id.machine) {
950         case 0x2064:
951         case 0x2066:
952         default:        /* Use "z900" as default for 64 bit kernels. */
953                 strcpy(elf_platform, "z900");
954                 break;
955         case 0x2084:
956         case 0x2086:
957                 strcpy(elf_platform, "z990");
958                 break;
959         case 0x2094:
960         case 0x2096:
961                 strcpy(elf_platform, "z9-109");
962                 break;
963         case 0x2097:
964         case 0x2098:
965                 strcpy(elf_platform, "z10");
966                 break;
967         case 0x2817:
968         case 0x2818:
969                 strcpy(elf_platform, "z196");
970                 break;
971         case 0x2827:
972         case 0x2828:
973                 strcpy(elf_platform, "zEC12");
974                 break;
975         case 0x2964:
976         case 0x2965:
977                 strcpy(elf_platform, "z13");
978                 break;
979         case 0x3906:
980         case 0x3907:
981                 strcpy(elf_platform, "z14");
982                 break;
983         case 0x8561:
984         case 0x8562:
985                 strcpy(elf_platform, "z15");
986                 break;
987         }
988
989         /*
990          * Virtualization support HWCAP_INT_SIE is bit 0.
991          */
992         if (sclp.has_sief2)
993                 int_hwcap |= HWCAP_INT_SIE;
994
995         return 0;
996 }
997 arch_initcall(setup_hwcaps);
998
999 /*
1000  * Add system information as device randomness
1001  */
1002 static void __init setup_randomness(void)
1003 {
1004         struct sysinfo_3_2_2 *vmms;
1005
1006         vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
1007                                                             PAGE_SIZE);
1008         if (!vmms)
1009                 panic("Failed to allocate memory for sysinfo structure\n");
1010
1011         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1012                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1013         memblock_free((unsigned long) vmms, PAGE_SIZE);
1014 }
1015
1016 /*
1017  * Find the correct size for the task_struct. This depends on
1018  * the size of the struct fpu at the end of the thread_struct
1019  * which is embedded in the task_struct.
1020  */
1021 static void __init setup_task_size(void)
1022 {
1023         int task_size = sizeof(struct task_struct);
1024
1025         if (!MACHINE_HAS_VX) {
1026                 task_size -= sizeof(__vector128) * __NUM_VXRS;
1027                 task_size += sizeof(freg_t) * __NUM_FPRS;
1028         }
1029         arch_task_struct_size = task_size;
1030 }
1031
1032 /*
1033  * Issue diagnose 318 to set the control program name and
1034  * version codes.
1035  */
1036 static void __init setup_control_program_code(void)
1037 {
1038         union diag318_info diag318_info = {
1039                 .cpnc = CPNC_LINUX,
1040                 .cpvc = 0,
1041         };
1042
1043         if (!sclp.has_diag318)
1044                 return;
1045
1046         diag_stat_inc(DIAG_STAT_X318);
1047         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1048 }
1049
1050 /*
1051  * Print the component list from the IPL report
1052  */
1053 static void __init log_component_list(void)
1054 {
1055         struct ipl_rb_component_entry *ptr, *end;
1056         char *str;
1057
1058         if (!early_ipl_comp_list_addr)
1059                 return;
1060         if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1061                 pr_info("Linux is running with Secure-IPL enabled\n");
1062         else
1063                 pr_info("Linux is running with Secure-IPL disabled\n");
1064         ptr = (void *) early_ipl_comp_list_addr;
1065         end = (void *) ptr + early_ipl_comp_list_size;
1066         pr_info("The IPL report contains the following components:\n");
1067         while (ptr < end) {
1068                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1069                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1070                                 str = "signed, verified";
1071                         else
1072                                 str = "signed, verification failed";
1073                 } else {
1074                         str = "not signed";
1075                 }
1076                 pr_info("%016llx - %016llx (%s)\n",
1077                         ptr->addr, ptr->addr + ptr->len, str);
1078                 ptr++;
1079         }
1080 }
1081
1082 /*
1083  * Setup function called from init/main.c just after the banner
1084  * was printed.
1085  */
1086
1087 void __init setup_arch(char **cmdline_p)
1088 {
1089         /*
1090          * print what head.S has found out about the machine
1091          */
1092         if (MACHINE_IS_VM)
1093                 pr_info("Linux is running as a z/VM "
1094                         "guest operating system in 64-bit mode\n");
1095         else if (MACHINE_IS_KVM)
1096                 pr_info("Linux is running under KVM in 64-bit mode\n");
1097         else if (MACHINE_IS_LPAR)
1098                 pr_info("Linux is running natively in 64-bit mode\n");
1099         else
1100                 pr_info("Linux is running as a guest in 64-bit mode\n");
1101
1102         log_component_list();
1103
1104         /* Have one command line that is parsed and saved in /proc/cmdline */
1105         /* boot_command_line has been already set up in early.c */
1106         *cmdline_p = boot_command_line;
1107
1108         ROOT_DEV = Root_RAM0;
1109
1110         init_mm.start_code = (unsigned long) _text;
1111         init_mm.end_code = (unsigned long) _etext;
1112         init_mm.end_data = (unsigned long) _edata;
1113         init_mm.brk = (unsigned long) _end;
1114
1115         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1116                 nospec_auto_detect();
1117
1118         jump_label_init();
1119         parse_early_param();
1120 #ifdef CONFIG_CRASH_DUMP
1121         /* Deactivate elfcorehdr= kernel parameter */
1122         elfcorehdr_addr = ELFCORE_ADDR_MAX;
1123 #endif
1124
1125         os_info_init();
1126         setup_ipl();
1127         setup_task_size();
1128         setup_control_program_code();
1129
1130         /* Do some memory reservations *before* memory is added to memblock */
1131         reserve_above_ident_map();
1132         reserve_oldmem();
1133         reserve_kernel();
1134         reserve_initrd();
1135         reserve_certificate_list();
1136         reserve_mem_detect_info();
1137         memblock_allow_resize();
1138
1139         /* Get information about *all* installed memory */
1140         memblock_add_mem_detect_info();
1141
1142         free_mem_detect_info();
1143         remove_oldmem();
1144
1145         setup_uv();
1146         setup_ident_map_size();
1147         setup_memory();
1148         dma_contiguous_reserve(ident_map_size);
1149         vmcp_cma_reserve();
1150         if (MACHINE_HAS_EDAT2)
1151                 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1152
1153         check_initrd();
1154         reserve_crashkernel();
1155 #ifdef CONFIG_CRASH_DUMP
1156         /*
1157          * Be aware that smp_save_dump_cpus() triggers a system reset.
1158          * Therefore CPU and device initialization should be done afterwards.
1159          */
1160         smp_save_dump_cpus();
1161 #endif
1162
1163         setup_resources();
1164         setup_lowcore_dat_off();
1165         smp_fill_possible_mask();
1166         cpu_detect_mhz_feature();
1167         cpu_init();
1168         numa_setup();
1169         smp_detect_cpus();
1170         topology_init_early();
1171
1172         /*
1173          * Create kernel page tables and switch to virtual addressing.
1174          */
1175         paging_init();
1176
1177         /*
1178          * After paging_init created the kernel page table, the new PSWs
1179          * in lowcore can now run with DAT enabled.
1180          */
1181         setup_lowcore_dat_on();
1182
1183         /* Setup default console */
1184         conmode_default();
1185         set_preferred_console();
1186
1187         apply_alternative_instructions();
1188         if (IS_ENABLED(CONFIG_EXPOLINE))
1189                 nospec_init_branches();
1190
1191         /* Setup zfcp/nvme dump support */
1192         setup_zfcpdump();
1193
1194         /* Add system specific data to the random pool */
1195         setup_randomness();
1196 }
This page took 0.101994 seconds and 4 git commands to generate.