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_RANGE (_PAGE_END(VA_BITS_MIN) - PAGE_OFFSET)
34 #define VMEMMAP_SIZE ((VMEMMAP_RANGE >> PAGE_SHIFT) * sizeof(struct page))
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 (VMEMMAP_END - VMEMMAP_SIZE)
51 #define VMEMMAP_END (-UL(SZ_1G))
52 #define PCI_IO_START (VMEMMAP_END + SZ_8M)
53 #define PCI_IO_END (PCI_IO_START + PCI_IO_SIZE)
54 #define FIXADDR_TOP (-UL(SZ_8M))
55
56 #if VA_BITS > 48
57 #ifdef CONFIG_ARM64_16K_PAGES
58 #define VA_BITS_MIN (47)
59 #else
60 #define VA_BITS_MIN (48)
61 #endif
62 #else
63 #define VA_BITS_MIN (VA_BITS)
64 #endif
65
66 #define _PAGE_END(va) (-(UL(1) << ((va) - 1)))
67
68 #define KERNEL_START _text
69 #define KERNEL_END _end
70
71 /*
72 * Generic and Software Tag-Based KASAN modes require 1/8th and 1/16th of the
73 * kernel virtual address space for storing the shadow memory respectively.
74 *
75 * The mapping between a virtual memory address and its corresponding shadow
76 * memory address is defined based on the formula:
77 *
78 * shadow_addr = (addr >> KASAN_SHADOW_SCALE_SHIFT) + KASAN_SHADOW_OFFSET
79 *
80 * where KASAN_SHADOW_SCALE_SHIFT is the order of the number of bits that map
81 * to a single shadow byte and KASAN_SHADOW_OFFSET is a constant that offsets
82 * the mapping. Note that KASAN_SHADOW_OFFSET does not point to the start of
83 * the shadow memory region.
84 *
85 * Based on this mapping, we define two constants:
86 *
87 * KASAN_SHADOW_START: the start of the shadow memory region;
88 * KASAN_SHADOW_END: the end of the shadow memory region.
89 *
90 * KASAN_SHADOW_END is defined first as the shadow address that corresponds to
91 * the upper bound of possible virtual kernel memory addresses UL(1) << 64
92 * according to the mapping formula.
93 *
94 * KASAN_SHADOW_START is defined second based on KASAN_SHADOW_END. The shadow
95 * memory start must map to the lowest possible kernel virtual memory address
96 * and thus it depends on the actual bitness of the address space.
97 *
98 * As KASAN inserts redzones between stack variables, this increases the stack
99 * memory usage significantly. Thus, we double the (minimum) stack size.
100 */
101 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
102 #define KASAN_SHADOW_OFFSET _AC(CONFIG_KASAN_SHADOW_OFFSET, UL)
103 #define KASAN_SHADOW_END ((UL(1) << (64 - KASAN_SHADOW_SCALE_SHIFT)) + KASAN_SHADOW_OFFSET)
104 #define _KASAN_SHADOW_START(va) (KASAN_SHADOW_END - (UL(1) << ((va) - KASAN_SHADOW_SCALE_SHIFT)))
105 #define KASAN_SHADOW_START _KASAN_SHADOW_START(vabits_actual)
106 #define PAGE_END KASAN_SHADOW_START
107 #define KASAN_THREAD_SHIFT 1
108 #else
109 #define KASAN_THREAD_SHIFT 0
110 #define PAGE_END (_PAGE_END(VA_BITS_MIN))
111 #endif /* CONFIG_KASAN */
112
113 #define DIRECT_MAP_PHYSMEM_END __pa(PAGE_END - 1)
114
115 #define MIN_THREAD_SHIFT (14 + KASAN_THREAD_SHIFT)
116
117 /*
118 * VMAP'd stacks are allocated at page granularity, so we must ensure that such
119 * stacks are a multiple of page size.
120 */
121 #if defined(CONFIG_VMAP_STACK) && (MIN_THREAD_SHIFT < PAGE_SHIFT)
122 #define THREAD_SHIFT PAGE_SHIFT
123 #else
124 #define THREAD_SHIFT MIN_THREAD_SHIFT
125 #endif
126
127 #if THREAD_SHIFT >= PAGE_SHIFT
128 #define THREAD_SIZE_ORDER (THREAD_SHIFT - PAGE_SHIFT)
129 #endif
130
131 #define THREAD_SIZE (UL(1) << THREAD_SHIFT)
132
133 /*
134 * By aligning VMAP'd stacks to 2 * THREAD_SIZE, we can detect overflow by
135 * checking sp & (1 << THREAD_SHIFT), which we can do cheaply in the entry
136 * assembly.
137 */
138 #ifdef CONFIG_VMAP_STACK
139 #define THREAD_ALIGN (2 * THREAD_SIZE)
140 #else
141 #define THREAD_ALIGN THREAD_SIZE
142 #endif
143
144 #define IRQ_STACK_SIZE THREAD_SIZE
145
146 #define OVERFLOW_STACK_SIZE SZ_4K
147
148 #define NVHE_STACK_SHIFT PAGE_SHIFT
149 #define NVHE_STACK_SIZE (UL(1) << NVHE_STACK_SHIFT)
150
151 /*
152 * With the minimum frame size of [x29, x30], exactly half the combined
153 * sizes of the hyp and overflow stacks is the maximum size needed to
154 * save the unwinded stacktrace; plus an additional entry to delimit the
155 * end.
156 */
157 #define NVHE_STACKTRACE_SIZE ((OVERFLOW_STACK_SIZE + NVHE_STACK_SIZE) / 2 + sizeof(long))
158
159 /*
160 * Alignment of kernel segments (e.g. .text, .data).
161 *
162 * 4 KB granule: 16 level 3 entries, with contiguous bit
163 * 16 KB granule: 4 level 3 entries, without contiguous bit
164 * 64 KB granule: 1 level 3 entry
165 */
166 #define SEGMENT_ALIGN SZ_64K
167
168 /*
169 * Memory types available.
170 *
171 * IMPORTANT: MT_NORMAL must be index 0 since vm_get_page_prot() may 'or' in
172 * the MT_NORMAL_TAGGED memory type for PROT_MTE mappings. Note
173 * that protection_map[] only contains MT_NORMAL attributes.
174 */
175 #define MT_NORMAL 0
176 #define MT_NORMAL_TAGGED 1
177 #define MT_NORMAL_NC 2
178 #define MT_DEVICE_nGnRnE 3
179 #define MT_DEVICE_nGnRE 4
180
181 /*
182 * Memory types for Stage-2 translation
183 */
184 #define MT_S2_NORMAL 0xf
185 #define MT_S2_NORMAL_NC 0x5
186 #define MT_S2_DEVICE_nGnRE 0x1
187
188 /*
189 * Memory types for Stage-2 translation when ID_AA64MMFR2_EL1.FWB is 0001
190 * Stage-2 enforces Normal-WB and Device-nGnRE
191 */
192 #define MT_S2_FWB_NORMAL 6
193 #define MT_S2_FWB_NORMAL_NC 5
194 #define MT_S2_FWB_DEVICE_nGnRE 1
195
196 #ifdef CONFIG_ARM64_4K_PAGES
197 #define IOREMAP_MAX_ORDER (PUD_SHIFT)
198 #else
199 #define IOREMAP_MAX_ORDER (PMD_SHIFT)
200 #endif
201
202 /*
203 * Open-coded (swapper_pg_dir - reserved_pg_dir) as this cannot be calculated
204 * until link time.
205 */
206 #define RESERVED_SWAPPER_OFFSET (PAGE_SIZE)
207
208 /*
209 * Open-coded (swapper_pg_dir - tramp_pg_dir) as this cannot be calculated
210 * until link time.
211 */
212 #define TRAMP_SWAPPER_OFFSET (2 * PAGE_SIZE)
213
214 #ifndef __ASSEMBLY__
215
216 #include <linux/bitops.h>
217 #include <linux/compiler.h>
218 #include <linux/mmdebug.h>
219 #include <linux/types.h>
220 #include <asm/boot.h>
221 #include <asm/bug.h>
222 #include <asm/sections.h>
223 #include <asm/sysreg.h>
224
read_tcr(void)225 static inline u64 __pure read_tcr(void)
226 {
227 u64 tcr;
228
229 // read_sysreg() uses asm volatile, so avoid it here
230 asm("mrs %0, tcr_el1" : "=r"(tcr));
231 return tcr;
232 }
233
234 #if VA_BITS > 48
235 // For reasons of #include hell, we can't use TCR_T1SZ_OFFSET/TCR_T1SZ_MASK here
236 #define vabits_actual (64 - ((read_tcr() >> 16) & 63))
237 #else
238 #define vabits_actual ((u64)VA_BITS)
239 #endif
240
241 extern s64 memstart_addr;
242 /* PHYS_OFFSET - the physical address of the start of memory. */
243 #define PHYS_OFFSET ({ VM_BUG_ON(memstart_addr & 1); memstart_addr; })
244
245 /* the offset between the kernel virtual and physical mappings */
246 extern u64 kimage_voffset;
247
kaslr_offset(void)248 static inline unsigned long kaslr_offset(void)
249 {
250 return (u64)&_text - KIMAGE_VADDR;
251 }
252
253 #ifdef CONFIG_RANDOMIZE_BASE
254 void kaslr_init(void);
kaslr_enabled(void)255 static inline bool kaslr_enabled(void)
256 {
257 extern bool __kaslr_is_enabled;
258 return __kaslr_is_enabled;
259 }
260 #else
kaslr_init(void)261 static inline void kaslr_init(void) { }
kaslr_enabled(void)262 static inline bool kaslr_enabled(void) { return false; }
263 #endif
264
265 /*
266 * Allow all memory at the discovery stage. We will clip it later.
267 */
268 #define MIN_MEMBLOCK_ADDR 0
269 #define MAX_MEMBLOCK_ADDR U64_MAX
270
271 /*
272 * PFNs are used to describe any physical page; this means
273 * PFN 0 == physical address 0.
274 *
275 * This is the PFN of the first RAM page in the kernel
276 * direct-mapped view. We assume this is the first page
277 * of RAM in the mem_map as well.
278 */
279 #define PHYS_PFN_OFFSET (PHYS_OFFSET >> PAGE_SHIFT)
280
281 /*
282 * When dealing with data aborts, watchpoints, or instruction traps we may end
283 * up with a tagged userland pointer. Clear the tag to get a sane pointer to
284 * pass on to access_ok(), for instance.
285 */
286 #define __untagged_addr(addr) \
287 ((__force __typeof__(addr))sign_extend64((__force u64)(addr), 55))
288
289 #define untagged_addr(addr) ({ \
290 u64 __addr = (__force u64)(addr); \
291 __addr &= __untagged_addr(__addr); \
292 (__force __typeof__(addr))__addr; \
293 })
294
295 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
296 #define __tag_shifted(tag) ((u64)(tag) << 56)
297 #define __tag_reset(addr) __untagged_addr(addr)
298 #define __tag_get(addr) (__u8)((u64)(addr) >> 56)
299 #else
300 #define __tag_shifted(tag) 0UL
301 #define __tag_reset(addr) (addr)
302 #define __tag_get(addr) 0
303 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
304
__tag_set(const void * addr,u8 tag)305 static inline const void *__tag_set(const void *addr, u8 tag)
306 {
307 u64 __addr = (u64)addr & ~__tag_shifted(0xff);
308 return (const void *)(__addr | __tag_shifted(tag));
309 }
310
311 #ifdef CONFIG_KASAN_HW_TAGS
312 #define arch_enable_tag_checks_sync() mte_enable_kernel_sync()
313 #define arch_enable_tag_checks_async() mte_enable_kernel_async()
314 #define arch_enable_tag_checks_asymm() mte_enable_kernel_asymm()
315 #define arch_suppress_tag_checks_start() mte_enable_tco()
316 #define arch_suppress_tag_checks_stop() mte_disable_tco()
317 #define arch_force_async_tag_fault() mte_check_tfsr_exit()
318 #define arch_get_random_tag() mte_get_random_tag()
319 #define arch_get_mem_tag(addr) mte_get_mem_tag(addr)
320 #define arch_set_mem_tag_range(addr, size, tag, init) \
321 mte_set_mem_tag_range((addr), (size), (tag), (init))
322 #endif /* CONFIG_KASAN_HW_TAGS */
323
324 /*
325 * Physical vs virtual RAM address space conversion. These are
326 * private definitions which should NOT be used outside memory.h
327 * files. Use virt_to_phys/phys_to_virt/__pa/__va instead.
328 */
329
330
331 /*
332 * Check whether an arbitrary address is within the linear map, which
333 * lives in the [PAGE_OFFSET, PAGE_END) interval at the bottom of the
334 * kernel's TTBR1 address range.
335 */
336 #define __is_lm_address(addr) (((u64)(addr) - PAGE_OFFSET) < (PAGE_END - PAGE_OFFSET))
337
338 #define __lm_to_phys(addr) (((addr) - PAGE_OFFSET) + PHYS_OFFSET)
339 #define __kimg_to_phys(addr) ((addr) - kimage_voffset)
340
341 #define __virt_to_phys_nodebug(x) ({ \
342 phys_addr_t __x = (phys_addr_t)(__tag_reset(x)); \
343 __is_lm_address(__x) ? __lm_to_phys(__x) : __kimg_to_phys(__x); \
344 })
345
346 #define __pa_symbol_nodebug(x) __kimg_to_phys((phys_addr_t)(x))
347
348 #ifdef CONFIG_DEBUG_VIRTUAL
349 extern phys_addr_t __virt_to_phys(unsigned long x);
350 extern phys_addr_t __phys_addr_symbol(unsigned long x);
351 #else
352 #define __virt_to_phys(x) __virt_to_phys_nodebug(x)
353 #define __phys_addr_symbol(x) __pa_symbol_nodebug(x)
354 #endif /* CONFIG_DEBUG_VIRTUAL */
355
356 #define __phys_to_virt(x) ((unsigned long)((x) - PHYS_OFFSET) | PAGE_OFFSET)
357 #define __phys_to_kimg(x) ((unsigned long)((x) + kimage_voffset))
358
359 /*
360 * Note: Drivers should NOT use these. They are the wrong
361 * translation for translating DMA addresses. Use the driver
362 * DMA support - see dma-mapping.h.
363 */
364 #define virt_to_phys virt_to_phys
virt_to_phys(const volatile void * x)365 static inline phys_addr_t virt_to_phys(const volatile void *x)
366 {
367 return __virt_to_phys((unsigned long)(x));
368 }
369
370 #define phys_to_virt phys_to_virt
phys_to_virt(phys_addr_t x)371 static inline void *phys_to_virt(phys_addr_t x)
372 {
373 return (void *)(__phys_to_virt(x));
374 }
375
376 /* Needed already here for resolving __phys_to_pfn() in virt_to_pfn() */
377 #include <asm-generic/memory_model.h>
378
virt_to_pfn(const void * kaddr)379 static inline unsigned long virt_to_pfn(const void *kaddr)
380 {
381 return __phys_to_pfn(virt_to_phys(kaddr));
382 }
383
384 /*
385 * Drivers should NOT use these either.
386 */
387 #define __pa(x) __virt_to_phys((unsigned long)(x))
388 #define __pa_symbol(x) __phys_addr_symbol(RELOC_HIDE((unsigned long)(x), 0))
389 #define __pa_nodebug(x) __virt_to_phys_nodebug((unsigned long)(x))
390 #define __va(x) ((void *)__phys_to_virt((phys_addr_t)(x)))
391 #define pfn_to_kaddr(pfn) __va((pfn) << PAGE_SHIFT)
392 #define sym_to_pfn(x) __phys_to_pfn(__pa_symbol(x))
393
394 /*
395 * virt_to_page(x) convert a _valid_ virtual address to struct page *
396 * virt_addr_valid(x) indicates whether a virtual address is valid
397 */
398 #define ARCH_PFN_OFFSET ((unsigned long)PHYS_PFN_OFFSET)
399
400 #if defined(CONFIG_DEBUG_VIRTUAL)
401 #define page_to_virt(x) ({ \
402 __typeof__(x) __page = x; \
403 void *__addr = __va(page_to_phys(__page)); \
404 (void *)__tag_set((const void *)__addr, page_kasan_tag(__page));\
405 })
406 #define virt_to_page(x) pfn_to_page(virt_to_pfn(x))
407 #else
408 #define page_to_virt(x) ({ \
409 __typeof__(x) __page = x; \
410 u64 __idx = ((u64)__page - VMEMMAP_START) / sizeof(struct page);\
411 u64 __addr = PAGE_OFFSET + (__idx * PAGE_SIZE); \
412 (void *)__tag_set((const void *)__addr, page_kasan_tag(__page));\
413 })
414
415 #define virt_to_page(x) ({ \
416 u64 __idx = (__tag_reset((u64)x) - PAGE_OFFSET) / PAGE_SIZE; \
417 u64 __addr = VMEMMAP_START + (__idx * sizeof(struct page)); \
418 (struct page *)__addr; \
419 })
420 #endif /* CONFIG_DEBUG_VIRTUAL */
421
422 #define virt_addr_valid(addr) ({ \
423 __typeof__(addr) __addr = __tag_reset(addr); \
424 __is_lm_address(__addr) && pfn_is_map_memory(virt_to_pfn(__addr)); \
425 })
426
427 void dump_mem_limit(void);
428 #endif /* !ASSEMBLY */
429
430 /*
431 * Given that the GIC architecture permits ITS implementations that can only be
432 * configured with a LPI table address once, GICv3 systems with many CPUs may
433 * end up reserving a lot of different regions after a kexec for their LPI
434 * tables (one per CPU), as we are forced to reuse the same memory after kexec
435 * (and thus reserve it persistently with EFI beforehand)
436 */
437 #if defined(CONFIG_EFI) && defined(CONFIG_ARM_GIC_V3_ITS)
438 # define INIT_MEMBLOCK_RESERVED_REGIONS (INIT_MEMBLOCK_REGIONS + NR_CPUS + 1)
439 #endif
440
441 /*
442 * memory regions which marked with flag MEMBLOCK_NOMAP(for example, the memory
443 * of the EFI_UNUSABLE_MEMORY type) may divide a continuous memory block into
444 * multiple parts. As a result, the number of memory regions is large.
445 */
446 #ifdef CONFIG_EFI
447 #define INIT_MEMBLOCK_MEMORY_REGIONS (INIT_MEMBLOCK_REGIONS * 8)
448 #endif
449
450
451 #endif /* __ASM_MEMORY_H */
452