Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0-only */
2/*
3 * Based on arch/arm/include/asm/memory.h
4 *
5 * Copyright (C) 2000-2002 Russell King
6 * Copyright (C) 2012 ARM Ltd.
7 *
8 * Note: this file should not be included by non-asm/.h files
9 */
10#ifndef __ASM_MEMORY_H
11#define __ASM_MEMORY_H
12
13#include <linux/const.h>
14#include <linux/sizes.h>
15#include <asm/page-def.h>
16
17/*
18 * Size of the PCI I/O space. This must remain a power of two so that
19 * IO_SPACE_LIMIT acts as a mask for the low bits of I/O addresses.
20 */
21#define PCI_IO_SIZE SZ_16M
22
23/*
24 * VMEMMAP_SIZE - allows the whole linear region to be covered by
25 * a struct page array
26 *
27 * If we are configured with a 52-bit kernel VA then our VMEMMAP_SIZE
28 * needs to cover the memory region from the beginning of the 52-bit
29 * PAGE_OFFSET all the way to PAGE_END for 48-bit. This allows us to
30 * keep a constant PAGE_OFFSET and "fallback" to using the higher end
31 * of the VMEMMAP where 52-bit support is not available in hardware.
32 */
33#define VMEMMAP_SHIFT (PAGE_SHIFT - STRUCT_PAGE_MAX_SHIFT)
34#define VMEMMAP_SIZE ((_PAGE_END(VA_BITS_MIN) - PAGE_OFFSET) >> VMEMMAP_SHIFT)
35
36/*
37 * PAGE_OFFSET - the virtual address of the start of the linear map, at the
38 * start of the TTBR1 address space.
39 * PAGE_END - the end of the linear map, where all other kernel mappings begin.
40 * KIMAGE_VADDR - the virtual address of the start of the kernel image.
41 * VA_BITS - the maximum number of bits for virtual addresses.
42 */
43#define VA_BITS (CONFIG_ARM64_VA_BITS)
44#define _PAGE_OFFSET(va) (-(UL(1) << (va)))
45#define PAGE_OFFSET (_PAGE_OFFSET(VA_BITS))
46#define KIMAGE_VADDR (MODULES_END)
47#define MODULES_END (MODULES_VADDR + MODULES_VSIZE)
48#define MODULES_VADDR (_PAGE_END(VA_BITS_MIN))
49#define MODULES_VSIZE (SZ_2G)
50#define VMEMMAP_START (-(UL(1) << (VA_BITS - VMEMMAP_SHIFT)))
51#define VMEMMAP_END (VMEMMAP_START + VMEMMAP_SIZE)
52#define PCI_IO_END (VMEMMAP_START - SZ_8M)
53#define PCI_IO_START (PCI_IO_END - PCI_IO_SIZE)
54#define FIXADDR_TOP (VMEMMAP_START - SZ_32M)
55
56#if VA_BITS > 48
57#define VA_BITS_MIN (48)
58#else
59#define VA_BITS_MIN (VA_BITS)
60#endif
61
62#define _PAGE_END(va) (-(UL(1) << ((va) - 1)))
63
64#define KERNEL_START _text
65#define KERNEL_END _end
66
67/*
68 * Generic and Software Tag-Based KASAN modes require 1/8th and 1/16th of the
69 * kernel virtual address space for storing the shadow memory respectively.
70 *
71 * The mapping between a virtual memory address and its corresponding shadow
72 * memory address is defined based on the formula:
73 *
74 * shadow_addr = (addr >> KASAN_SHADOW_SCALE_SHIFT) + KASAN_SHADOW_OFFSET
75 *
76 * where KASAN_SHADOW_SCALE_SHIFT is the order of the number of bits that map
77 * to a single shadow byte and KASAN_SHADOW_OFFSET is a constant that offsets
78 * the mapping. Note that KASAN_SHADOW_OFFSET does not point to the start of
79 * the shadow memory region.
80 *
81 * Based on this mapping, we define two constants:
82 *
83 * KASAN_SHADOW_START: the start of the shadow memory region;
84 * KASAN_SHADOW_END: the end of the shadow memory region.
85 *
86 * KASAN_SHADOW_END is defined first as the shadow address that corresponds to
87 * the upper bound of possible virtual kernel memory addresses UL(1) << 64
88 * according to the mapping formula.
89 *
90 * KASAN_SHADOW_START is defined second based on KASAN_SHADOW_END. The shadow
91 * memory start must map to the lowest possible kernel virtual memory address
92 * and thus it depends on the actual bitness of the address space.
93 *
94 * As KASAN inserts redzones between stack variables, this increases the stack
95 * memory usage significantly. Thus, we double the (minimum) stack size.
96 */
97#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
98#define KASAN_SHADOW_OFFSET _AC(CONFIG_KASAN_SHADOW_OFFSET, UL)
99#define KASAN_SHADOW_END ((UL(1) << (64 - KASAN_SHADOW_SCALE_SHIFT)) + KASAN_SHADOW_OFFSET)
100#define _KASAN_SHADOW_START(va) (KASAN_SHADOW_END - (UL(1) << ((va) - KASAN_SHADOW_SCALE_SHIFT)))
101#define KASAN_SHADOW_START _KASAN_SHADOW_START(vabits_actual)
102#define PAGE_END KASAN_SHADOW_START
103#define KASAN_THREAD_SHIFT 1
104#else
105#define KASAN_THREAD_SHIFT 0
106#define PAGE_END (_PAGE_END(VA_BITS_MIN))
107#endif /* CONFIG_KASAN */
108
109#define MIN_THREAD_SHIFT (14 + KASAN_THREAD_SHIFT)
110
111/*
112 * VMAP'd stacks are allocated at page granularity, so we must ensure that such
113 * stacks are a multiple of page size.
114 */
115#if defined(CONFIG_VMAP_STACK) && (MIN_THREAD_SHIFT < PAGE_SHIFT)
116#define THREAD_SHIFT PAGE_SHIFT
117#else
118#define THREAD_SHIFT MIN_THREAD_SHIFT
119#endif
120
121#if THREAD_SHIFT >= PAGE_SHIFT
122#define THREAD_SIZE_ORDER (THREAD_SHIFT - PAGE_SHIFT)
123#endif
124
125#define THREAD_SIZE (UL(1) << THREAD_SHIFT)
126
127/*
128 * By aligning VMAP'd stacks to 2 * THREAD_SIZE, we can detect overflow by
129 * checking sp & (1 << THREAD_SHIFT), which we can do cheaply in the entry
130 * assembly.
131 */
132#ifdef CONFIG_VMAP_STACK
133#define THREAD_ALIGN (2 * THREAD_SIZE)
134#else
135#define THREAD_ALIGN THREAD_SIZE
136#endif
137
138#define IRQ_STACK_SIZE THREAD_SIZE
139
140#define OVERFLOW_STACK_SIZE SZ_4K
141
142/*
143 * With the minimum frame size of [x29, x30], exactly half the combined
144 * sizes of the hyp and overflow stacks is the maximum size needed to
145 * save the unwinded stacktrace; plus an additional entry to delimit the
146 * end.
147 */
148#define NVHE_STACKTRACE_SIZE ((OVERFLOW_STACK_SIZE + PAGE_SIZE) / 2 + sizeof(long))
149
150/*
151 * Alignment of kernel segments (e.g. .text, .data).
152 *
153 * 4 KB granule: 16 level 3 entries, with contiguous bit
154 * 16 KB granule: 4 level 3 entries, without contiguous bit
155 * 64 KB granule: 1 level 3 entry
156 */
157#define SEGMENT_ALIGN SZ_64K
158
159/*
160 * Memory types available.
161 *
162 * IMPORTANT: MT_NORMAL must be index 0 since vm_get_page_prot() may 'or' in
163 * the MT_NORMAL_TAGGED memory type for PROT_MTE mappings. Note
164 * that protection_map[] only contains MT_NORMAL attributes.
165 */
166#define MT_NORMAL 0
167#define MT_NORMAL_TAGGED 1
168#define MT_NORMAL_NC 2
169#define MT_DEVICE_nGnRnE 3
170#define MT_DEVICE_nGnRE 4
171
172/*
173 * Memory types for Stage-2 translation
174 */
175#define MT_S2_NORMAL 0xf
176#define MT_S2_DEVICE_nGnRE 0x1
177
178/*
179 * Memory types for Stage-2 translation when ID_AA64MMFR2_EL1.FWB is 0001
180 * Stage-2 enforces Normal-WB and Device-nGnRE
181 */
182#define MT_S2_FWB_NORMAL 6
183#define MT_S2_FWB_DEVICE_nGnRE 1
184
185#ifdef CONFIG_ARM64_4K_PAGES
186#define IOREMAP_MAX_ORDER (PUD_SHIFT)
187#else
188#define IOREMAP_MAX_ORDER (PMD_SHIFT)
189#endif
190
191/*
192 * Open-coded (swapper_pg_dir - reserved_pg_dir) as this cannot be calculated
193 * until link time.
194 */
195#define RESERVED_SWAPPER_OFFSET (PAGE_SIZE)
196
197/*
198 * Open-coded (swapper_pg_dir - tramp_pg_dir) as this cannot be calculated
199 * until link time.
200 */
201#define TRAMP_SWAPPER_OFFSET (2 * PAGE_SIZE)
202
203#ifndef __ASSEMBLY__
204
205#include <linux/bitops.h>
206#include <linux/compiler.h>
207#include <linux/mmdebug.h>
208#include <linux/types.h>
209#include <asm/boot.h>
210#include <asm/bug.h>
211#include <asm/sections.h>
212
213#if VA_BITS > 48
214extern u64 vabits_actual;
215#else
216#define vabits_actual ((u64)VA_BITS)
217#endif
218
219extern s64 memstart_addr;
220/* PHYS_OFFSET - the physical address of the start of memory. */
221#define PHYS_OFFSET ({ VM_BUG_ON(memstart_addr & 1); memstart_addr; })
222
223/* the offset between the kernel virtual and physical mappings */
224extern u64 kimage_voffset;
225
226static inline unsigned long kaslr_offset(void)
227{
228 return (u64)&_text - KIMAGE_VADDR;
229}
230
231#ifdef CONFIG_RANDOMIZE_BASE
232void kaslr_init(void);
233static inline bool kaslr_enabled(void)
234{
235 extern bool __kaslr_is_enabled;
236 return __kaslr_is_enabled;
237}
238#else
239static inline void kaslr_init(void) { }
240static inline bool kaslr_enabled(void) { return false; }
241#endif
242
243/*
244 * Allow all memory at the discovery stage. We will clip it later.
245 */
246#define MIN_MEMBLOCK_ADDR 0
247#define MAX_MEMBLOCK_ADDR U64_MAX
248
249/*
250 * PFNs are used to describe any physical page; this means
251 * PFN 0 == physical address 0.
252 *
253 * This is the PFN of the first RAM page in the kernel
254 * direct-mapped view. We assume this is the first page
255 * of RAM in the mem_map as well.
256 */
257#define PHYS_PFN_OFFSET (PHYS_OFFSET >> PAGE_SHIFT)
258
259/*
260 * When dealing with data aborts, watchpoints, or instruction traps we may end
261 * up with a tagged userland pointer. Clear the tag to get a sane pointer to
262 * pass on to access_ok(), for instance.
263 */
264#define __untagged_addr(addr) \
265 ((__force __typeof__(addr))sign_extend64((__force u64)(addr), 55))
266
267#define untagged_addr(addr) ({ \
268 u64 __addr = (__force u64)(addr); \
269 __addr &= __untagged_addr(__addr); \
270 (__force __typeof__(addr))__addr; \
271})
272
273#if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
274#define __tag_shifted(tag) ((u64)(tag) << 56)
275#define __tag_reset(addr) __untagged_addr(addr)
276#define __tag_get(addr) (__u8)((u64)(addr) >> 56)
277#else
278#define __tag_shifted(tag) 0UL
279#define __tag_reset(addr) (addr)
280#define __tag_get(addr) 0
281#endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
282
283static inline const void *__tag_set(const void *addr, u8 tag)
284{
285 u64 __addr = (u64)addr & ~__tag_shifted(0xff);
286 return (const void *)(__addr | __tag_shifted(tag));
287}
288
289#ifdef CONFIG_KASAN_HW_TAGS
290#define arch_enable_tag_checks_sync() mte_enable_kernel_sync()
291#define arch_enable_tag_checks_async() mte_enable_kernel_async()
292#define arch_enable_tag_checks_asymm() mte_enable_kernel_asymm()
293#define arch_suppress_tag_checks_start() mte_enable_tco()
294#define arch_suppress_tag_checks_stop() mte_disable_tco()
295#define arch_force_async_tag_fault() mte_check_tfsr_exit()
296#define arch_get_random_tag() mte_get_random_tag()
297#define arch_get_mem_tag(addr) mte_get_mem_tag(addr)
298#define arch_set_mem_tag_range(addr, size, tag, init) \
299 mte_set_mem_tag_range((addr), (size), (tag), (init))
300#endif /* CONFIG_KASAN_HW_TAGS */
301
302/*
303 * Physical vs virtual RAM address space conversion. These are
304 * private definitions which should NOT be used outside memory.h
305 * files. Use virt_to_phys/phys_to_virt/__pa/__va instead.
306 */
307
308
309/*
310 * Check whether an arbitrary address is within the linear map, which
311 * lives in the [PAGE_OFFSET, PAGE_END) interval at the bottom of the
312 * kernel's TTBR1 address range.
313 */
314#define __is_lm_address(addr) (((u64)(addr) - PAGE_OFFSET) < (PAGE_END - PAGE_OFFSET))
315
316#define __lm_to_phys(addr) (((addr) - PAGE_OFFSET) + PHYS_OFFSET)
317#define __kimg_to_phys(addr) ((addr) - kimage_voffset)
318
319#define __virt_to_phys_nodebug(x) ({ \
320 phys_addr_t __x = (phys_addr_t)(__tag_reset(x)); \
321 __is_lm_address(__x) ? __lm_to_phys(__x) : __kimg_to_phys(__x); \
322})
323
324#define __pa_symbol_nodebug(x) __kimg_to_phys((phys_addr_t)(x))
325
326#ifdef CONFIG_DEBUG_VIRTUAL
327extern phys_addr_t __virt_to_phys(unsigned long x);
328extern phys_addr_t __phys_addr_symbol(unsigned long x);
329#else
330#define __virt_to_phys(x) __virt_to_phys_nodebug(x)
331#define __phys_addr_symbol(x) __pa_symbol_nodebug(x)
332#endif /* CONFIG_DEBUG_VIRTUAL */
333
334#define __phys_to_virt(x) ((unsigned long)((x) - PHYS_OFFSET) | PAGE_OFFSET)
335#define __phys_to_kimg(x) ((unsigned long)((x) + kimage_voffset))
336
337/*
338 * Convert a page to/from a physical address
339 */
340#define page_to_phys(page) (__pfn_to_phys(page_to_pfn(page)))
341#define phys_to_page(phys) (pfn_to_page(__phys_to_pfn(phys)))
342
343/*
344 * Note: Drivers should NOT use these. They are the wrong
345 * translation for translating DMA addresses. Use the driver
346 * DMA support - see dma-mapping.h.
347 */
348#define virt_to_phys virt_to_phys
349static inline phys_addr_t virt_to_phys(const volatile void *x)
350{
351 return __virt_to_phys((unsigned long)(x));
352}
353
354#define phys_to_virt phys_to_virt
355static inline void *phys_to_virt(phys_addr_t x)
356{
357 return (void *)(__phys_to_virt(x));
358}
359
360/* Needed already here for resolving __phys_to_pfn() in virt_to_pfn() */
361#include <asm-generic/memory_model.h>
362
363static inline unsigned long virt_to_pfn(const void *kaddr)
364{
365 return __phys_to_pfn(virt_to_phys(kaddr));
366}
367
368/*
369 * Drivers should NOT use these either.
370 */
371#define __pa(x) __virt_to_phys((unsigned long)(x))
372#define __pa_symbol(x) __phys_addr_symbol(RELOC_HIDE((unsigned long)(x), 0))
373#define __pa_nodebug(x) __virt_to_phys_nodebug((unsigned long)(x))
374#define __va(x) ((void *)__phys_to_virt((phys_addr_t)(x)))
375#define pfn_to_kaddr(pfn) __va((pfn) << PAGE_SHIFT)
376#define sym_to_pfn(x) __phys_to_pfn(__pa_symbol(x))
377
378/*
379 * virt_to_page(x) convert a _valid_ virtual address to struct page *
380 * virt_addr_valid(x) indicates whether a virtual address is valid
381 */
382#define ARCH_PFN_OFFSET ((unsigned long)PHYS_PFN_OFFSET)
383
384#if defined(CONFIG_DEBUG_VIRTUAL)
385#define page_to_virt(x) ({ \
386 __typeof__(x) __page = x; \
387 void *__addr = __va(page_to_phys(__page)); \
388 (void *)__tag_set((const void *)__addr, page_kasan_tag(__page));\
389})
390#define virt_to_page(x) pfn_to_page(virt_to_pfn(x))
391#else
392#define page_to_virt(x) ({ \
393 __typeof__(x) __page = x; \
394 u64 __idx = ((u64)__page - VMEMMAP_START) / sizeof(struct page);\
395 u64 __addr = PAGE_OFFSET + (__idx * PAGE_SIZE); \
396 (void *)__tag_set((const void *)__addr, page_kasan_tag(__page));\
397})
398
399#define virt_to_page(x) ({ \
400 u64 __idx = (__tag_reset((u64)x) - PAGE_OFFSET) / PAGE_SIZE; \
401 u64 __addr = VMEMMAP_START + (__idx * sizeof(struct page)); \
402 (struct page *)__addr; \
403})
404#endif /* CONFIG_DEBUG_VIRTUAL */
405
406#define virt_addr_valid(addr) ({ \
407 __typeof__(addr) __addr = __tag_reset(addr); \
408 __is_lm_address(__addr) && pfn_is_map_memory(virt_to_pfn(__addr)); \
409})
410
411void dump_mem_limit(void);
412#endif /* !ASSEMBLY */
413
414/*
415 * Given that the GIC architecture permits ITS implementations that can only be
416 * configured with a LPI table address once, GICv3 systems with many CPUs may
417 * end up reserving a lot of different regions after a kexec for their LPI
418 * tables (one per CPU), as we are forced to reuse the same memory after kexec
419 * (and thus reserve it persistently with EFI beforehand)
420 */
421#if defined(CONFIG_EFI) && defined(CONFIG_ARM_GIC_V3_ITS)
422# define INIT_MEMBLOCK_RESERVED_REGIONS (INIT_MEMBLOCK_REGIONS + NR_CPUS + 1)
423#endif
424
425/*
426 * memory regions which marked with flag MEMBLOCK_NOMAP(for example, the memory
427 * of the EFI_UNUSABLE_MEMORY type) may divide a continuous memory block into
428 * multiple parts. As a result, the number of memory regions is large.
429 */
430#ifdef CONFIG_EFI
431#define INIT_MEMBLOCK_MEMORY_REGIONS (INIT_MEMBLOCK_REGIONS * 8)
432#endif
433
434
435#endif /* __ASM_MEMORY_H */