thermal: fix Mediatek thermal controller build
[linux/fpc-iii.git] / arch / x86 / platform / efi / efi_64.c
blob49e4dd4a1f58257630c8b341bf8b12922200cd2e
1 /*
2 * x86_64 specific EFI support functions
3 * Based on Extensible Firmware Interface Specification version 1.0
5 * Copyright (C) 2005-2008 Intel Co.
6 * Fenghua Yu <fenghua.yu@intel.com>
7 * Bibo Mao <bibo.mao@intel.com>
8 * Chandramouli Narayanan <mouli@linux.intel.com>
9 * Huang Ying <ying.huang@intel.com>
11 * Code to convert EFI to E820 map has been implemented in elilo bootloader
12 * based on a EFI patch by Edgar Hucek. Based on the E820 map, the page table
13 * is setup appropriately for EFI runtime code.
14 * - mouli 06/14/2007.
18 #define pr_fmt(fmt) "efi: " fmt
20 #include <linux/kernel.h>
21 #include <linux/init.h>
22 #include <linux/mm.h>
23 #include <linux/types.h>
24 #include <linux/spinlock.h>
25 #include <linux/bootmem.h>
26 #include <linux/ioport.h>
27 #include <linux/module.h>
28 #include <linux/efi.h>
29 #include <linux/uaccess.h>
30 #include <linux/io.h>
31 #include <linux/reboot.h>
32 #include <linux/slab.h>
34 #include <asm/setup.h>
35 #include <asm/page.h>
36 #include <asm/e820.h>
37 #include <asm/pgtable.h>
38 #include <asm/tlbflush.h>
39 #include <asm/proto.h>
40 #include <asm/efi.h>
41 #include <asm/cacheflush.h>
42 #include <asm/fixmap.h>
43 #include <asm/realmode.h>
44 #include <asm/time.h>
45 #include <asm/pgalloc.h>
48 * We allocate runtime services regions bottom-up, starting from -4G, i.e.
49 * 0xffff_ffff_0000_0000 and limit EFI VA mapping space to 64G.
51 static u64 efi_va = EFI_VA_START;
53 struct efi_scratch efi_scratch;
55 static void __init early_code_mapping_set_exec(int executable)
57 efi_memory_desc_t *md;
58 void *p;
60 if (!(__supported_pte_mask & _PAGE_NX))
61 return;
63 /* Make EFI service code area executable */
64 for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
65 md = p;
66 if (md->type == EFI_RUNTIME_SERVICES_CODE ||
67 md->type == EFI_BOOT_SERVICES_CODE)
68 efi_set_executable(md, executable);
72 pgd_t * __init efi_call_phys_prolog(void)
74 unsigned long vaddress;
75 pgd_t *save_pgd;
77 int pgd;
78 int n_pgds;
80 if (!efi_enabled(EFI_OLD_MEMMAP)) {
81 save_pgd = (pgd_t *)read_cr3();
82 write_cr3((unsigned long)efi_scratch.efi_pgt);
83 goto out;
86 early_code_mapping_set_exec(1);
88 n_pgds = DIV_ROUND_UP((max_pfn << PAGE_SHIFT), PGDIR_SIZE);
89 save_pgd = kmalloc(n_pgds * sizeof(pgd_t), GFP_KERNEL);
91 for (pgd = 0; pgd < n_pgds; pgd++) {
92 save_pgd[pgd] = *pgd_offset_k(pgd * PGDIR_SIZE);
93 vaddress = (unsigned long)__va(pgd * PGDIR_SIZE);
94 set_pgd(pgd_offset_k(pgd * PGDIR_SIZE), *pgd_offset_k(vaddress));
96 out:
97 __flush_tlb_all();
99 return save_pgd;
102 void __init efi_call_phys_epilog(pgd_t *save_pgd)
105 * After the lock is released, the original page table is restored.
107 int pgd_idx;
108 int nr_pgds;
110 if (!efi_enabled(EFI_OLD_MEMMAP)) {
111 write_cr3((unsigned long)save_pgd);
112 __flush_tlb_all();
113 return;
116 nr_pgds = DIV_ROUND_UP((max_pfn << PAGE_SHIFT) , PGDIR_SIZE);
118 for (pgd_idx = 0; pgd_idx < nr_pgds; pgd_idx++)
119 set_pgd(pgd_offset_k(pgd_idx * PGDIR_SIZE), save_pgd[pgd_idx]);
121 kfree(save_pgd);
123 __flush_tlb_all();
124 early_code_mapping_set_exec(0);
127 static pgd_t *efi_pgd;
130 * We need our own copy of the higher levels of the page tables
131 * because we want to avoid inserting EFI region mappings (EFI_VA_END
132 * to EFI_VA_START) into the standard kernel page tables. Everything
133 * else can be shared, see efi_sync_low_kernel_mappings().
135 int __init efi_alloc_page_tables(void)
137 pgd_t *pgd;
138 pud_t *pud;
139 gfp_t gfp_mask;
141 if (efi_enabled(EFI_OLD_MEMMAP))
142 return 0;
144 gfp_mask = GFP_KERNEL | __GFP_NOTRACK | __GFP_REPEAT | __GFP_ZERO;
145 efi_pgd = (pgd_t *)__get_free_page(gfp_mask);
146 if (!efi_pgd)
147 return -ENOMEM;
149 pgd = efi_pgd + pgd_index(EFI_VA_END);
151 pud = pud_alloc_one(NULL, 0);
152 if (!pud) {
153 free_page((unsigned long)efi_pgd);
154 return -ENOMEM;
157 pgd_populate(NULL, pgd, pud);
159 return 0;
163 * Add low kernel mappings for passing arguments to EFI functions.
165 void efi_sync_low_kernel_mappings(void)
167 unsigned num_entries;
168 pgd_t *pgd_k, *pgd_efi;
169 pud_t *pud_k, *pud_efi;
171 if (efi_enabled(EFI_OLD_MEMMAP))
172 return;
175 * We can share all PGD entries apart from the one entry that
176 * covers the EFI runtime mapping space.
178 * Make sure the EFI runtime region mappings are guaranteed to
179 * only span a single PGD entry and that the entry also maps
180 * other important kernel regions.
182 BUILD_BUG_ON(pgd_index(EFI_VA_END) != pgd_index(MODULES_END));
183 BUILD_BUG_ON((EFI_VA_START & PGDIR_MASK) !=
184 (EFI_VA_END & PGDIR_MASK));
186 pgd_efi = efi_pgd + pgd_index(PAGE_OFFSET);
187 pgd_k = pgd_offset_k(PAGE_OFFSET);
189 num_entries = pgd_index(EFI_VA_END) - pgd_index(PAGE_OFFSET);
190 memcpy(pgd_efi, pgd_k, sizeof(pgd_t) * num_entries);
193 * We share all the PUD entries apart from those that map the
194 * EFI regions. Copy around them.
196 BUILD_BUG_ON((EFI_VA_START & ~PUD_MASK) != 0);
197 BUILD_BUG_ON((EFI_VA_END & ~PUD_MASK) != 0);
199 pgd_efi = efi_pgd + pgd_index(EFI_VA_END);
200 pud_efi = pud_offset(pgd_efi, 0);
202 pgd_k = pgd_offset_k(EFI_VA_END);
203 pud_k = pud_offset(pgd_k, 0);
205 num_entries = pud_index(EFI_VA_END);
206 memcpy(pud_efi, pud_k, sizeof(pud_t) * num_entries);
208 pud_efi = pud_offset(pgd_efi, EFI_VA_START);
209 pud_k = pud_offset(pgd_k, EFI_VA_START);
211 num_entries = PTRS_PER_PUD - pud_index(EFI_VA_START);
212 memcpy(pud_efi, pud_k, sizeof(pud_t) * num_entries);
215 int __init efi_setup_page_tables(unsigned long pa_memmap, unsigned num_pages)
217 unsigned long pfn, text;
218 efi_memory_desc_t *md;
219 struct page *page;
220 unsigned npages;
221 pgd_t *pgd;
223 if (efi_enabled(EFI_OLD_MEMMAP))
224 return 0;
226 efi_scratch.efi_pgt = (pgd_t *)__pa(efi_pgd);
227 pgd = efi_pgd;
230 * It can happen that the physical address of new_memmap lands in memory
231 * which is not mapped in the EFI page table. Therefore we need to go
232 * and ident-map those pages containing the map before calling
233 * phys_efi_set_virtual_address_map().
235 pfn = pa_memmap >> PAGE_SHIFT;
236 if (kernel_map_pages_in_pgd(pgd, pfn, pa_memmap, num_pages, _PAGE_NX | _PAGE_RW)) {
237 pr_err("Error ident-mapping new memmap (0x%lx)!\n", pa_memmap);
238 return 1;
241 efi_scratch.use_pgd = true;
244 * When making calls to the firmware everything needs to be 1:1
245 * mapped and addressable with 32-bit pointers. Map the kernel
246 * text and allocate a new stack because we can't rely on the
247 * stack pointer being < 4GB.
249 if (!IS_ENABLED(CONFIG_EFI_MIXED))
250 return 0;
253 * Map all of RAM so that we can access arguments in the 1:1
254 * mapping when making EFI runtime calls.
256 for_each_efi_memory_desc(&memmap, md) {
257 if (md->type != EFI_CONVENTIONAL_MEMORY &&
258 md->type != EFI_LOADER_DATA &&
259 md->type != EFI_LOADER_CODE)
260 continue;
262 pfn = md->phys_addr >> PAGE_SHIFT;
263 npages = md->num_pages;
265 if (kernel_map_pages_in_pgd(pgd, pfn, md->phys_addr, npages, _PAGE_RW)) {
266 pr_err("Failed to map 1:1 memory\n");
267 return 1;
271 page = alloc_page(GFP_KERNEL|__GFP_DMA32);
272 if (!page)
273 panic("Unable to allocate EFI runtime stack < 4GB\n");
275 efi_scratch.phys_stack = virt_to_phys(page_address(page));
276 efi_scratch.phys_stack += PAGE_SIZE; /* stack grows down */
278 npages = (_etext - _text) >> PAGE_SHIFT;
279 text = __pa(_text);
280 pfn = text >> PAGE_SHIFT;
282 if (kernel_map_pages_in_pgd(pgd, pfn, text, npages, _PAGE_RW)) {
283 pr_err("Failed to map kernel text 1:1\n");
284 return 1;
287 return 0;
290 void __init efi_cleanup_page_tables(unsigned long pa_memmap, unsigned num_pages)
292 kernel_unmap_pages_in_pgd(efi_pgd, pa_memmap, num_pages);
295 static void __init __map_region(efi_memory_desc_t *md, u64 va)
297 unsigned long flags = _PAGE_RW;
298 unsigned long pfn;
299 pgd_t *pgd = efi_pgd;
301 if (!(md->attribute & EFI_MEMORY_WB))
302 flags |= _PAGE_PCD;
304 pfn = md->phys_addr >> PAGE_SHIFT;
305 if (kernel_map_pages_in_pgd(pgd, pfn, va, md->num_pages, flags))
306 pr_warn("Error mapping PA 0x%llx -> VA 0x%llx!\n",
307 md->phys_addr, va);
310 void __init efi_map_region(efi_memory_desc_t *md)
312 unsigned long size = md->num_pages << PAGE_SHIFT;
313 u64 pa = md->phys_addr;
315 if (efi_enabled(EFI_OLD_MEMMAP))
316 return old_map_region(md);
319 * Make sure the 1:1 mappings are present as a catch-all for b0rked
320 * firmware which doesn't update all internal pointers after switching
321 * to virtual mode and would otherwise crap on us.
323 __map_region(md, md->phys_addr);
326 * Enforce the 1:1 mapping as the default virtual address when
327 * booting in EFI mixed mode, because even though we may be
328 * running a 64-bit kernel, the firmware may only be 32-bit.
330 if (!efi_is_native () && IS_ENABLED(CONFIG_EFI_MIXED)) {
331 md->virt_addr = md->phys_addr;
332 return;
335 efi_va -= size;
337 /* Is PA 2M-aligned? */
338 if (!(pa & (PMD_SIZE - 1))) {
339 efi_va &= PMD_MASK;
340 } else {
341 u64 pa_offset = pa & (PMD_SIZE - 1);
342 u64 prev_va = efi_va;
344 /* get us the same offset within this 2M page */
345 efi_va = (efi_va & PMD_MASK) + pa_offset;
347 if (efi_va > prev_va)
348 efi_va -= PMD_SIZE;
351 if (efi_va < EFI_VA_END) {
352 pr_warn(FW_WARN "VA address range overflow!\n");
353 return;
356 /* Do the VA map */
357 __map_region(md, efi_va);
358 md->virt_addr = efi_va;
362 * kexec kernel will use efi_map_region_fixed to map efi runtime memory ranges.
363 * md->virt_addr is the original virtual address which had been mapped in kexec
364 * 1st kernel.
366 void __init efi_map_region_fixed(efi_memory_desc_t *md)
368 __map_region(md, md->virt_addr);
371 void __iomem *__init efi_ioremap(unsigned long phys_addr, unsigned long size,
372 u32 type, u64 attribute)
374 unsigned long last_map_pfn;
376 if (type == EFI_MEMORY_MAPPED_IO)
377 return ioremap(phys_addr, size);
379 last_map_pfn = init_memory_mapping(phys_addr, phys_addr + size);
380 if ((last_map_pfn << PAGE_SHIFT) < phys_addr + size) {
381 unsigned long top = last_map_pfn << PAGE_SHIFT;
382 efi_ioremap(top, size - (top - phys_addr), type, attribute);
385 if (!(attribute & EFI_MEMORY_WB))
386 efi_memory_uc((u64)(unsigned long)__va(phys_addr), size);
388 return (void __iomem *)__va(phys_addr);
391 void __init parse_efi_setup(u64 phys_addr, u32 data_len)
393 efi_setup = phys_addr + sizeof(struct setup_data);
396 void __init efi_runtime_update_mappings(void)
398 unsigned long pfn;
399 pgd_t *pgd = efi_pgd;
400 efi_memory_desc_t *md;
401 void *p;
403 if (efi_enabled(EFI_OLD_MEMMAP)) {
404 if (__supported_pte_mask & _PAGE_NX)
405 runtime_code_page_mkexec();
406 return;
409 if (!efi_enabled(EFI_NX_PE_DATA))
410 return;
412 for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
413 unsigned long pf = 0;
414 md = p;
416 if (!(md->attribute & EFI_MEMORY_RUNTIME))
417 continue;
419 if (!(md->attribute & EFI_MEMORY_WB))
420 pf |= _PAGE_PCD;
422 if ((md->attribute & EFI_MEMORY_XP) ||
423 (md->type == EFI_RUNTIME_SERVICES_DATA))
424 pf |= _PAGE_NX;
426 if (!(md->attribute & EFI_MEMORY_RO) &&
427 (md->type != EFI_RUNTIME_SERVICES_CODE))
428 pf |= _PAGE_RW;
430 /* Update the 1:1 mapping */
431 pfn = md->phys_addr >> PAGE_SHIFT;
432 if (kernel_map_pages_in_pgd(pgd, pfn, md->phys_addr, md->num_pages, pf))
433 pr_warn("Error mapping PA 0x%llx -> VA 0x%llx!\n",
434 md->phys_addr, md->virt_addr);
436 if (kernel_map_pages_in_pgd(pgd, pfn, md->virt_addr, md->num_pages, pf))
437 pr_warn("Error mapping PA 0x%llx -> VA 0x%llx!\n",
438 md->phys_addr, md->virt_addr);
442 void __init efi_dump_pagetable(void)
444 #ifdef CONFIG_EFI_PGT_DUMP
445 ptdump_walk_pgd_level(NULL, efi_pgd);
446 #endif
449 #ifdef CONFIG_EFI_MIXED
450 extern efi_status_t efi64_thunk(u32, ...);
452 #define runtime_service32(func) \
453 ({ \
454 u32 table = (u32)(unsigned long)efi.systab; \
455 u32 *rt, *___f; \
457 rt = (u32 *)(table + offsetof(efi_system_table_32_t, runtime)); \
458 ___f = (u32 *)(*rt + offsetof(efi_runtime_services_32_t, func)); \
459 *___f; \
463 * Switch to the EFI page tables early so that we can access the 1:1
464 * runtime services mappings which are not mapped in any other page
465 * tables. This function must be called before runtime_service32().
467 * Also, disable interrupts because the IDT points to 64-bit handlers,
468 * which aren't going to function correctly when we switch to 32-bit.
470 #define efi_thunk(f, ...) \
471 ({ \
472 efi_status_t __s; \
473 unsigned long flags; \
474 u32 func; \
476 efi_sync_low_kernel_mappings(); \
477 local_irq_save(flags); \
479 efi_scratch.prev_cr3 = read_cr3(); \
480 write_cr3((unsigned long)efi_scratch.efi_pgt); \
481 __flush_tlb_all(); \
483 func = runtime_service32(f); \
484 __s = efi64_thunk(func, __VA_ARGS__); \
486 write_cr3(efi_scratch.prev_cr3); \
487 __flush_tlb_all(); \
488 local_irq_restore(flags); \
490 __s; \
493 efi_status_t efi_thunk_set_virtual_address_map(
494 void *phys_set_virtual_address_map,
495 unsigned long memory_map_size,
496 unsigned long descriptor_size,
497 u32 descriptor_version,
498 efi_memory_desc_t *virtual_map)
500 efi_status_t status;
501 unsigned long flags;
502 u32 func;
504 efi_sync_low_kernel_mappings();
505 local_irq_save(flags);
507 efi_scratch.prev_cr3 = read_cr3();
508 write_cr3((unsigned long)efi_scratch.efi_pgt);
509 __flush_tlb_all();
511 func = (u32)(unsigned long)phys_set_virtual_address_map;
512 status = efi64_thunk(func, memory_map_size, descriptor_size,
513 descriptor_version, virtual_map);
515 write_cr3(efi_scratch.prev_cr3);
516 __flush_tlb_all();
517 local_irq_restore(flags);
519 return status;
522 static efi_status_t efi_thunk_get_time(efi_time_t *tm, efi_time_cap_t *tc)
524 efi_status_t status;
525 u32 phys_tm, phys_tc;
527 spin_lock(&rtc_lock);
529 phys_tm = virt_to_phys(tm);
530 phys_tc = virt_to_phys(tc);
532 status = efi_thunk(get_time, phys_tm, phys_tc);
534 spin_unlock(&rtc_lock);
536 return status;
539 static efi_status_t efi_thunk_set_time(efi_time_t *tm)
541 efi_status_t status;
542 u32 phys_tm;
544 spin_lock(&rtc_lock);
546 phys_tm = virt_to_phys(tm);
548 status = efi_thunk(set_time, phys_tm);
550 spin_unlock(&rtc_lock);
552 return status;
555 static efi_status_t
556 efi_thunk_get_wakeup_time(efi_bool_t *enabled, efi_bool_t *pending,
557 efi_time_t *tm)
559 efi_status_t status;
560 u32 phys_enabled, phys_pending, phys_tm;
562 spin_lock(&rtc_lock);
564 phys_enabled = virt_to_phys(enabled);
565 phys_pending = virt_to_phys(pending);
566 phys_tm = virt_to_phys(tm);
568 status = efi_thunk(get_wakeup_time, phys_enabled,
569 phys_pending, phys_tm);
571 spin_unlock(&rtc_lock);
573 return status;
576 static efi_status_t
577 efi_thunk_set_wakeup_time(efi_bool_t enabled, efi_time_t *tm)
579 efi_status_t status;
580 u32 phys_tm;
582 spin_lock(&rtc_lock);
584 phys_tm = virt_to_phys(tm);
586 status = efi_thunk(set_wakeup_time, enabled, phys_tm);
588 spin_unlock(&rtc_lock);
590 return status;
594 static efi_status_t
595 efi_thunk_get_variable(efi_char16_t *name, efi_guid_t *vendor,
596 u32 *attr, unsigned long *data_size, void *data)
598 efi_status_t status;
599 u32 phys_name, phys_vendor, phys_attr;
600 u32 phys_data_size, phys_data;
602 phys_data_size = virt_to_phys(data_size);
603 phys_vendor = virt_to_phys(vendor);
604 phys_name = virt_to_phys(name);
605 phys_attr = virt_to_phys(attr);
606 phys_data = virt_to_phys(data);
608 status = efi_thunk(get_variable, phys_name, phys_vendor,
609 phys_attr, phys_data_size, phys_data);
611 return status;
614 static efi_status_t
615 efi_thunk_set_variable(efi_char16_t *name, efi_guid_t *vendor,
616 u32 attr, unsigned long data_size, void *data)
618 u32 phys_name, phys_vendor, phys_data;
619 efi_status_t status;
621 phys_name = virt_to_phys(name);
622 phys_vendor = virt_to_phys(vendor);
623 phys_data = virt_to_phys(data);
625 /* If data_size is > sizeof(u32) we've got problems */
626 status = efi_thunk(set_variable, phys_name, phys_vendor,
627 attr, data_size, phys_data);
629 return status;
632 static efi_status_t
633 efi_thunk_get_next_variable(unsigned long *name_size,
634 efi_char16_t *name,
635 efi_guid_t *vendor)
637 efi_status_t status;
638 u32 phys_name_size, phys_name, phys_vendor;
640 phys_name_size = virt_to_phys(name_size);
641 phys_vendor = virt_to_phys(vendor);
642 phys_name = virt_to_phys(name);
644 status = efi_thunk(get_next_variable, phys_name_size,
645 phys_name, phys_vendor);
647 return status;
650 static efi_status_t
651 efi_thunk_get_next_high_mono_count(u32 *count)
653 efi_status_t status;
654 u32 phys_count;
656 phys_count = virt_to_phys(count);
657 status = efi_thunk(get_next_high_mono_count, phys_count);
659 return status;
662 static void
663 efi_thunk_reset_system(int reset_type, efi_status_t status,
664 unsigned long data_size, efi_char16_t *data)
666 u32 phys_data;
668 phys_data = virt_to_phys(data);
670 efi_thunk(reset_system, reset_type, status, data_size, phys_data);
673 static efi_status_t
674 efi_thunk_update_capsule(efi_capsule_header_t **capsules,
675 unsigned long count, unsigned long sg_list)
678 * To properly support this function we would need to repackage
679 * 'capsules' because the firmware doesn't understand 64-bit
680 * pointers.
682 return EFI_UNSUPPORTED;
685 static efi_status_t
686 efi_thunk_query_variable_info(u32 attr, u64 *storage_space,
687 u64 *remaining_space,
688 u64 *max_variable_size)
690 efi_status_t status;
691 u32 phys_storage, phys_remaining, phys_max;
693 if (efi.runtime_version < EFI_2_00_SYSTEM_TABLE_REVISION)
694 return EFI_UNSUPPORTED;
696 phys_storage = virt_to_phys(storage_space);
697 phys_remaining = virt_to_phys(remaining_space);
698 phys_max = virt_to_phys(max_variable_size);
700 status = efi_thunk(query_variable_info, attr, phys_storage,
701 phys_remaining, phys_max);
703 return status;
706 static efi_status_t
707 efi_thunk_query_capsule_caps(efi_capsule_header_t **capsules,
708 unsigned long count, u64 *max_size,
709 int *reset_type)
712 * To properly support this function we would need to repackage
713 * 'capsules' because the firmware doesn't understand 64-bit
714 * pointers.
716 return EFI_UNSUPPORTED;
719 void efi_thunk_runtime_setup(void)
721 efi.get_time = efi_thunk_get_time;
722 efi.set_time = efi_thunk_set_time;
723 efi.get_wakeup_time = efi_thunk_get_wakeup_time;
724 efi.set_wakeup_time = efi_thunk_set_wakeup_time;
725 efi.get_variable = efi_thunk_get_variable;
726 efi.get_next_variable = efi_thunk_get_next_variable;
727 efi.set_variable = efi_thunk_set_variable;
728 efi.get_next_high_mono_count = efi_thunk_get_next_high_mono_count;
729 efi.reset_system = efi_thunk_reset_system;
730 efi.query_variable_info = efi_thunk_query_variable_info;
731 efi.update_capsule = efi_thunk_update_capsule;
732 efi.query_capsule_caps = efi_thunk_query_capsule_caps;
734 #endif /* CONFIG_EFI_MIXED */