Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/drivers/char/mem.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | * | |
d7d4d849 | 7 | * Added devfs support. |
1da177e4 | 8 | * Jan-11-1998, C. Scott Ananian <cananian@alumni.princeton.edu> |
af901ca1 | 9 | * Shared /dev/zero mmapping support, Feb 2000, Kanoj Sarcar <kanoj@sgi.com> |
1da177e4 LT |
10 | */ |
11 | ||
1da177e4 LT |
12 | #include <linux/mm.h> |
13 | #include <linux/miscdevice.h> | |
14 | #include <linux/slab.h> | |
15 | #include <linux/vmalloc.h> | |
16 | #include <linux/mman.h> | |
17 | #include <linux/random.h> | |
18 | #include <linux/init.h> | |
19 | #include <linux/raw.h> | |
20 | #include <linux/tty.h> | |
21 | #include <linux/capability.h> | |
1da177e4 LT |
22 | #include <linux/ptrace.h> |
23 | #include <linux/device.h> | |
50b1fdbd | 24 | #include <linux/highmem.h> |
1da177e4 | 25 | #include <linux/backing-dev.h> |
c01d5b30 | 26 | #include <linux/shmem_fs.h> |
d6b29d7c | 27 | #include <linux/splice.h> |
b8a3ad5b | 28 | #include <linux/pfn.h> |
66300e66 | 29 | #include <linux/export.h> |
e1612de9 | 30 | #include <linux/io.h> |
e2e40f2c | 31 | #include <linux/uio.h> |
1da177e4 | 32 | |
35b6c7e4 | 33 | #include <linux/uaccess.h> |
1da177e4 LT |
34 | |
35 | #ifdef CONFIG_IA64 | |
36 | # include <linux/efi.h> | |
37 | #endif | |
38 | ||
e1612de9 HM |
39 | #define DEVPORT_MINOR 4 |
40 | ||
f222318e WF |
41 | static inline unsigned long size_inside_page(unsigned long start, |
42 | unsigned long size) | |
43 | { | |
44 | unsigned long sz; | |
45 | ||
7fabaddd | 46 | sz = PAGE_SIZE - (start & (PAGE_SIZE - 1)); |
f222318e | 47 | |
7fabaddd | 48 | return min(sz, size); |
f222318e WF |
49 | } |
50 | ||
1da177e4 | 51 | #ifndef ARCH_HAS_VALID_PHYS_ADDR_RANGE |
7e6735c3 | 52 | static inline int valid_phys_addr_range(phys_addr_t addr, size_t count) |
1da177e4 | 53 | { |
cfaf346c | 54 | return addr + count <= __pa(high_memory); |
1da177e4 | 55 | } |
80851ef2 | 56 | |
06c67bef | 57 | static inline int valid_mmap_phys_addr_range(unsigned long pfn, size_t size) |
80851ef2 BH |
58 | { |
59 | return 1; | |
60 | } | |
1da177e4 LT |
61 | #endif |
62 | ||
d092633b | 63 | #ifdef CONFIG_STRICT_DEVMEM |
a4866aa8 KC |
64 | static inline int page_is_allowed(unsigned long pfn) |
65 | { | |
66 | return devmem_is_allowed(pfn); | |
67 | } | |
e2beb3ea | 68 | static inline int range_is_allowed(unsigned long pfn, unsigned long size) |
ae531c26 | 69 | { |
e2beb3ea VP |
70 | u64 from = ((u64)pfn) << PAGE_SHIFT; |
71 | u64 to = from + size; | |
72 | u64 cursor = from; | |
73 | ||
74 | while (cursor < to) { | |
39380b80 | 75 | if (!devmem_is_allowed(pfn)) |
ae531c26 | 76 | return 0; |
e2beb3ea VP |
77 | cursor += PAGE_SIZE; |
78 | pfn++; | |
ae531c26 AV |
79 | } |
80 | return 1; | |
81 | } | |
82 | #else | |
a4866aa8 KC |
83 | static inline int page_is_allowed(unsigned long pfn) |
84 | { | |
85 | return 1; | |
86 | } | |
e2beb3ea | 87 | static inline int range_is_allowed(unsigned long pfn, unsigned long size) |
ae531c26 AV |
88 | { |
89 | return 1; | |
90 | } | |
91 | #endif | |
92 | ||
4707a341 TR |
93 | #ifndef unxlate_dev_mem_ptr |
94 | #define unxlate_dev_mem_ptr unxlate_dev_mem_ptr | |
95 | void __weak unxlate_dev_mem_ptr(phys_addr_t phys, void *addr) | |
e045fb2a | 96 | { |
97 | } | |
4707a341 | 98 | #endif |
e045fb2a | 99 | |
1da177e4 | 100 | /* |
d7d4d849 AM |
101 | * This funcion reads the *physical* memory. The f_pos points directly to the |
102 | * memory location. | |
1da177e4 | 103 | */ |
d7d4d849 | 104 | static ssize_t read_mem(struct file *file, char __user *buf, |
1da177e4 LT |
105 | size_t count, loff_t *ppos) |
106 | { | |
7e6735c3 | 107 | phys_addr_t p = *ppos; |
1da177e4 | 108 | ssize_t read, sz; |
4707a341 | 109 | void *ptr; |
22ec1a2a KC |
110 | char *bounce; |
111 | int err; | |
1da177e4 | 112 | |
08d2d00b PT |
113 | if (p != *ppos) |
114 | return 0; | |
115 | ||
136939a2 | 116 | if (!valid_phys_addr_range(p, count)) |
1da177e4 LT |
117 | return -EFAULT; |
118 | read = 0; | |
119 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
120 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
121 | if (p < PAGE_SIZE) { | |
7fabaddd | 122 | sz = size_inside_page(p, count); |
1da177e4 LT |
123 | if (sz > 0) { |
124 | if (clear_user(buf, sz)) | |
125 | return -EFAULT; | |
d7d4d849 AM |
126 | buf += sz; |
127 | p += sz; | |
128 | count -= sz; | |
129 | read += sz; | |
1da177e4 LT |
130 | } |
131 | } | |
132 | #endif | |
133 | ||
22ec1a2a KC |
134 | bounce = kmalloc(PAGE_SIZE, GFP_KERNEL); |
135 | if (!bounce) | |
136 | return -ENOMEM; | |
137 | ||
1da177e4 | 138 | while (count > 0) { |
fa29e97b | 139 | unsigned long remaining; |
b5b38200 | 140 | int allowed, probe; |
fa29e97b | 141 | |
f222318e | 142 | sz = size_inside_page(p, count); |
1da177e4 | 143 | |
22ec1a2a | 144 | err = -EPERM; |
a4866aa8 KC |
145 | allowed = page_is_allowed(p >> PAGE_SHIFT); |
146 | if (!allowed) | |
22ec1a2a KC |
147 | goto failed; |
148 | ||
149 | err = -EFAULT; | |
a4866aa8 KC |
150 | if (allowed == 2) { |
151 | /* Show zeros for restricted memory. */ | |
152 | remaining = clear_user(buf, sz); | |
153 | } else { | |
154 | /* | |
155 | * On ia64 if a page has been mapped somewhere as | |
156 | * uncached, then it must also be accessed uncached | |
157 | * by the kernel or data corruption may occur. | |
158 | */ | |
159 | ptr = xlate_dev_mem_ptr(p); | |
160 | if (!ptr) | |
22ec1a2a | 161 | goto failed; |
a4866aa8 | 162 | |
b5b38200 | 163 | probe = probe_kernel_read(bounce, ptr, sz); |
a4866aa8 | 164 | unxlate_dev_mem_ptr(p, ptr); |
b5b38200 | 165 | if (probe) |
22ec1a2a KC |
166 | goto failed; |
167 | ||
168 | remaining = copy_to_user(buf, bounce, sz); | |
a4866aa8 | 169 | } |
1da177e4 | 170 | |
fa29e97b | 171 | if (remaining) |
22ec1a2a | 172 | goto failed; |
e045fb2a | 173 | |
1da177e4 LT |
174 | buf += sz; |
175 | p += sz; | |
176 | count -= sz; | |
177 | read += sz; | |
178 | } | |
22ec1a2a | 179 | kfree(bounce); |
1da177e4 LT |
180 | |
181 | *ppos += read; | |
182 | return read; | |
22ec1a2a KC |
183 | |
184 | failed: | |
185 | kfree(bounce); | |
186 | return err; | |
1da177e4 LT |
187 | } |
188 | ||
d7d4d849 | 189 | static ssize_t write_mem(struct file *file, const char __user *buf, |
1da177e4 LT |
190 | size_t count, loff_t *ppos) |
191 | { | |
7e6735c3 | 192 | phys_addr_t p = *ppos; |
1da177e4 LT |
193 | ssize_t written, sz; |
194 | unsigned long copied; | |
195 | void *ptr; | |
196 | ||
08d2d00b PT |
197 | if (p != *ppos) |
198 | return -EFBIG; | |
199 | ||
136939a2 | 200 | if (!valid_phys_addr_range(p, count)) |
1da177e4 LT |
201 | return -EFAULT; |
202 | ||
203 | written = 0; | |
204 | ||
205 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
206 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
207 | if (p < PAGE_SIZE) { | |
7fabaddd | 208 | sz = size_inside_page(p, count); |
1da177e4 LT |
209 | /* Hmm. Do something? */ |
210 | buf += sz; | |
211 | p += sz; | |
212 | count -= sz; | |
213 | written += sz; | |
214 | } | |
215 | #endif | |
216 | ||
217 | while (count > 0) { | |
a4866aa8 KC |
218 | int allowed; |
219 | ||
f222318e | 220 | sz = size_inside_page(p, count); |
1da177e4 | 221 | |
a4866aa8 KC |
222 | allowed = page_is_allowed(p >> PAGE_SHIFT); |
223 | if (!allowed) | |
e045fb2a | 224 | return -EPERM; |
225 | ||
a4866aa8 KC |
226 | /* Skip actual writing when a page is marked as restricted. */ |
227 | if (allowed == 1) { | |
228 | /* | |
229 | * On ia64 if a page has been mapped somewhere as | |
230 | * uncached, then it must also be accessed uncached | |
231 | * by the kernel or data corruption may occur. | |
232 | */ | |
233 | ptr = xlate_dev_mem_ptr(p); | |
234 | if (!ptr) { | |
235 | if (written) | |
236 | break; | |
237 | return -EFAULT; | |
238 | } | |
239 | ||
240 | copied = copy_from_user(ptr, buf, sz); | |
241 | unxlate_dev_mem_ptr(p, ptr); | |
242 | if (copied) { | |
243 | written += sz - copied; | |
244 | if (written) | |
245 | break; | |
246 | return -EFAULT; | |
247 | } | |
1da177e4 | 248 | } |
e045fb2a | 249 | |
1da177e4 LT |
250 | buf += sz; |
251 | p += sz; | |
252 | count -= sz; | |
253 | written += sz; | |
254 | } | |
255 | ||
256 | *ppos += written; | |
257 | return written; | |
258 | } | |
259 | ||
d7d4d849 | 260 | int __weak phys_mem_access_prot_allowed(struct file *file, |
f0970c13 | 261 | unsigned long pfn, unsigned long size, pgprot_t *vma_prot) |
262 | { | |
263 | return 1; | |
264 | } | |
265 | ||
44ac8413 | 266 | #ifndef __HAVE_PHYS_MEM_ACCESS_PROT |
d7d4d849 AM |
267 | |
268 | /* | |
269 | * Architectures vary in how they handle caching for addresses | |
270 | * outside of main memory. | |
271 | * | |
272 | */ | |
ea56f411 | 273 | #ifdef pgprot_noncached |
7e6735c3 | 274 | static int uncached_access(struct file *file, phys_addr_t addr) |
d7d4d849 AM |
275 | { |
276 | #if defined(CONFIG_IA64) | |
277 | /* | |
278 | * On ia64, we ignore O_DSYNC because we cannot tolerate memory | |
279 | * attribute aliases. | |
280 | */ | |
281 | return !(efi_mem_attributes(addr) & EFI_MEMORY_WB); | |
282 | #elif defined(CONFIG_MIPS) | |
283 | { | |
284 | extern int __uncached_access(struct file *file, | |
285 | unsigned long addr); | |
286 | ||
287 | return __uncached_access(file, addr); | |
288 | } | |
289 | #else | |
290 | /* | |
291 | * Accessing memory above the top the kernel knows about or through a | |
292 | * file pointer | |
293 | * that was marked O_DSYNC will be done non-cached. | |
294 | */ | |
295 | if (file->f_flags & O_DSYNC) | |
296 | return 1; | |
297 | return addr >= __pa(high_memory); | |
298 | #endif | |
299 | } | |
ea56f411 | 300 | #endif |
d7d4d849 | 301 | |
44ac8413 BH |
302 | static pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, |
303 | unsigned long size, pgprot_t vma_prot) | |
304 | { | |
305 | #ifdef pgprot_noncached | |
7e6735c3 | 306 | phys_addr_t offset = pfn << PAGE_SHIFT; |
44ac8413 BH |
307 | |
308 | if (uncached_access(file, offset)) | |
309 | return pgprot_noncached(vma_prot); | |
310 | #endif | |
311 | return vma_prot; | |
312 | } | |
313 | #endif | |
314 | ||
5da6185b DH |
315 | #ifndef CONFIG_MMU |
316 | static unsigned long get_unmapped_area_mem(struct file *file, | |
317 | unsigned long addr, | |
318 | unsigned long len, | |
319 | unsigned long pgoff, | |
320 | unsigned long flags) | |
321 | { | |
322 | if (!valid_mmap_phys_addr_range(pgoff, len)) | |
323 | return (unsigned long) -EINVAL; | |
8a93258c | 324 | return pgoff << PAGE_SHIFT; |
5da6185b DH |
325 | } |
326 | ||
b4caecd4 CH |
327 | /* permit direct mmap, for read, write or exec */ |
328 | static unsigned memory_mmap_capabilities(struct file *file) | |
329 | { | |
330 | return NOMMU_MAP_DIRECT | | |
331 | NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC; | |
332 | } | |
333 | ||
334 | static unsigned zero_mmap_capabilities(struct file *file) | |
335 | { | |
336 | return NOMMU_MAP_COPY; | |
337 | } | |
338 | ||
5da6185b DH |
339 | /* can't do an in-place private mapping if there's no MMU */ |
340 | static inline int private_mapping_ok(struct vm_area_struct *vma) | |
341 | { | |
342 | return vma->vm_flags & VM_MAYSHARE; | |
343 | } | |
344 | #else | |
5da6185b DH |
345 | |
346 | static inline int private_mapping_ok(struct vm_area_struct *vma) | |
347 | { | |
348 | return 1; | |
349 | } | |
350 | #endif | |
351 | ||
f0f37e2f | 352 | static const struct vm_operations_struct mmap_mem_ops = { |
7ae8ed50 RR |
353 | #ifdef CONFIG_HAVE_IOREMAP_PROT |
354 | .access = generic_access_phys | |
355 | #endif | |
e7f260a2 | 356 | }; |
357 | ||
d7d4d849 | 358 | static int mmap_mem(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 359 | { |
80851ef2 | 360 | size_t size = vma->vm_end - vma->vm_start; |
b299cde2 JW |
361 | phys_addr_t offset = (phys_addr_t)vma->vm_pgoff << PAGE_SHIFT; |
362 | ||
be62a320 CB |
363 | /* Does it even fit in phys_addr_t? */ |
364 | if (offset >> PAGE_SHIFT != vma->vm_pgoff) | |
365 | return -EINVAL; | |
366 | ||
b299cde2 | 367 | /* It's illegal to wrap around the end of the physical address space. */ |
32829da5 | 368 | if (offset + (phys_addr_t)size - 1 < offset) |
b299cde2 | 369 | return -EINVAL; |
80851ef2 | 370 | |
06c67bef | 371 | if (!valid_mmap_phys_addr_range(vma->vm_pgoff, size)) |
80851ef2 BH |
372 | return -EINVAL; |
373 | ||
5da6185b DH |
374 | if (!private_mapping_ok(vma)) |
375 | return -ENOSYS; | |
376 | ||
e2beb3ea VP |
377 | if (!range_is_allowed(vma->vm_pgoff, size)) |
378 | return -EPERM; | |
379 | ||
f0970c13 | 380 | if (!phys_mem_access_prot_allowed(file, vma->vm_pgoff, size, |
381 | &vma->vm_page_prot)) | |
382 | return -EINVAL; | |
383 | ||
8b150478 | 384 | vma->vm_page_prot = phys_mem_access_prot(file, vma->vm_pgoff, |
80851ef2 | 385 | size, |
1da177e4 | 386 | vma->vm_page_prot); |
1da177e4 | 387 | |
e7f260a2 | 388 | vma->vm_ops = &mmap_mem_ops; |
389 | ||
314e51b9 | 390 | /* Remap-pfn-range will mark the range VM_IO */ |
1da177e4 LT |
391 | if (remap_pfn_range(vma, |
392 | vma->vm_start, | |
393 | vma->vm_pgoff, | |
80851ef2 | 394 | size, |
e7f260a2 | 395 | vma->vm_page_prot)) { |
1da177e4 | 396 | return -EAGAIN; |
e7f260a2 | 397 | } |
1da177e4 LT |
398 | return 0; |
399 | } | |
400 | ||
d7d4d849 | 401 | static int mmap_kmem(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 402 | { |
4bb82551 LT |
403 | unsigned long pfn; |
404 | ||
6d3154cc LT |
405 | /* Turn a kernel-virtual address into a physical page frame */ |
406 | pfn = __pa((u64)vma->vm_pgoff << PAGE_SHIFT) >> PAGE_SHIFT; | |
4bb82551 | 407 | |
1da177e4 | 408 | /* |
d7d4d849 AM |
409 | * RED-PEN: on some architectures there is more mapped memory than |
410 | * available in mem_map which pfn_valid checks for. Perhaps should add a | |
411 | * new macro here. | |
1da177e4 LT |
412 | * |
413 | * RED-PEN: vmalloc is not supported right now. | |
414 | */ | |
4bb82551 | 415 | if (!pfn_valid(pfn)) |
1da177e4 | 416 | return -EIO; |
4bb82551 LT |
417 | |
418 | vma->vm_pgoff = pfn; | |
1da177e4 LT |
419 | return mmap_mem(file, vma); |
420 | } | |
421 | ||
1da177e4 LT |
422 | /* |
423 | * This function reads the *virtual* memory as seen by the kernel. | |
424 | */ | |
d7d4d849 | 425 | static ssize_t read_kmem(struct file *file, char __user *buf, |
1da177e4 LT |
426 | size_t count, loff_t *ppos) |
427 | { | |
428 | unsigned long p = *ppos; | |
429 | ssize_t low_count, read, sz; | |
890537b3 | 430 | char *kbuf; /* k-addr because vread() takes vmlist_lock rwlock */ |
325fda71 | 431 | int err = 0; |
1da177e4 LT |
432 | |
433 | read = 0; | |
434 | if (p < (unsigned long) high_memory) { | |
435 | low_count = count; | |
d7d4d849 AM |
436 | if (count > (unsigned long)high_memory - p) |
437 | low_count = (unsigned long)high_memory - p; | |
1da177e4 LT |
438 | |
439 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
440 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
441 | if (p < PAGE_SIZE && low_count > 0) { | |
7fabaddd WF |
442 | sz = size_inside_page(p, low_count); |
443 | if (clear_user(buf, sz)) | |
1da177e4 | 444 | return -EFAULT; |
7fabaddd WF |
445 | buf += sz; |
446 | p += sz; | |
447 | read += sz; | |
448 | low_count -= sz; | |
449 | count -= sz; | |
1da177e4 LT |
450 | } |
451 | #endif | |
452 | while (low_count > 0) { | |
f222318e | 453 | sz = size_inside_page(p, low_count); |
1da177e4 LT |
454 | |
455 | /* | |
456 | * On ia64 if a page has been mapped somewhere as | |
457 | * uncached, then it must also be accessed uncached | |
458 | * by the kernel or data corruption may occur | |
459 | */ | |
4707a341 | 460 | kbuf = xlate_dev_kmem_ptr((void *)p); |
488debb9 RM |
461 | if (!virt_addr_valid(kbuf)) |
462 | return -ENXIO; | |
1da177e4 LT |
463 | |
464 | if (copy_to_user(buf, kbuf, sz)) | |
465 | return -EFAULT; | |
466 | buf += sz; | |
467 | p += sz; | |
468 | read += sz; | |
469 | low_count -= sz; | |
470 | count -= sz; | |
471 | } | |
472 | } | |
473 | ||
474 | if (count > 0) { | |
475 | kbuf = (char *)__get_free_page(GFP_KERNEL); | |
476 | if (!kbuf) | |
477 | return -ENOMEM; | |
478 | while (count > 0) { | |
80ad89a0 | 479 | sz = size_inside_page(p, count); |
325fda71 KH |
480 | if (!is_vmalloc_or_module_addr((void *)p)) { |
481 | err = -ENXIO; | |
482 | break; | |
483 | } | |
80ad89a0 WF |
484 | sz = vread(kbuf, (char *)p, sz); |
485 | if (!sz) | |
1da177e4 | 486 | break; |
80ad89a0 | 487 | if (copy_to_user(buf, kbuf, sz)) { |
325fda71 KH |
488 | err = -EFAULT; |
489 | break; | |
1da177e4 | 490 | } |
80ad89a0 WF |
491 | count -= sz; |
492 | buf += sz; | |
493 | read += sz; | |
494 | p += sz; | |
1da177e4 LT |
495 | } |
496 | free_page((unsigned long)kbuf); | |
497 | } | |
325fda71 KH |
498 | *ppos = p; |
499 | return read ? read : err; | |
1da177e4 LT |
500 | } |
501 | ||
502 | ||
d7d4d849 AM |
503 | static ssize_t do_write_kmem(unsigned long p, const char __user *buf, |
504 | size_t count, loff_t *ppos) | |
1da177e4 LT |
505 | { |
506 | ssize_t written, sz; | |
507 | unsigned long copied; | |
508 | ||
509 | written = 0; | |
510 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
511 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
ee32398f WF |
512 | if (p < PAGE_SIZE) { |
513 | sz = size_inside_page(p, count); | |
1da177e4 LT |
514 | /* Hmm. Do something? */ |
515 | buf += sz; | |
516 | p += sz; | |
1da177e4 LT |
517 | count -= sz; |
518 | written += sz; | |
519 | } | |
520 | #endif | |
521 | ||
522 | while (count > 0) { | |
4707a341 | 523 | void *ptr; |
1da177e4 | 524 | |
ee32398f | 525 | sz = size_inside_page(p, count); |
1da177e4 LT |
526 | |
527 | /* | |
d7d4d849 AM |
528 | * On ia64 if a page has been mapped somewhere as uncached, then |
529 | * it must also be accessed uncached by the kernel or data | |
530 | * corruption may occur. | |
1da177e4 | 531 | */ |
4707a341 | 532 | ptr = xlate_dev_kmem_ptr((void *)p); |
488debb9 RM |
533 | if (!virt_addr_valid(ptr)) |
534 | return -ENXIO; | |
1da177e4 LT |
535 | |
536 | copied = copy_from_user(ptr, buf, sz); | |
537 | if (copied) { | |
c654d60e JB |
538 | written += sz - copied; |
539 | if (written) | |
540 | break; | |
1da177e4 LT |
541 | return -EFAULT; |
542 | } | |
543 | buf += sz; | |
544 | p += sz; | |
1da177e4 LT |
545 | count -= sz; |
546 | written += sz; | |
547 | } | |
548 | ||
549 | *ppos += written; | |
550 | return written; | |
551 | } | |
552 | ||
1da177e4 LT |
553 | /* |
554 | * This function writes to the *virtual* memory as seen by the kernel. | |
555 | */ | |
d7d4d849 | 556 | static ssize_t write_kmem(struct file *file, const char __user *buf, |
1da177e4 LT |
557 | size_t count, loff_t *ppos) |
558 | { | |
559 | unsigned long p = *ppos; | |
560 | ssize_t wrote = 0; | |
561 | ssize_t virtr = 0; | |
890537b3 | 562 | char *kbuf; /* k-addr because vwrite() takes vmlist_lock rwlock */ |
325fda71 | 563 | int err = 0; |
1da177e4 LT |
564 | |
565 | if (p < (unsigned long) high_memory) { | |
80ad89a0 WF |
566 | unsigned long to_write = min_t(unsigned long, count, |
567 | (unsigned long)high_memory - p); | |
ee32398f | 568 | wrote = do_write_kmem(p, buf, to_write, ppos); |
80ad89a0 WF |
569 | if (wrote != to_write) |
570 | return wrote; | |
1da177e4 LT |
571 | p += wrote; |
572 | buf += wrote; | |
573 | count -= wrote; | |
574 | } | |
575 | ||
576 | if (count > 0) { | |
577 | kbuf = (char *)__get_free_page(GFP_KERNEL); | |
578 | if (!kbuf) | |
579 | return wrote ? wrote : -ENOMEM; | |
580 | while (count > 0) { | |
80ad89a0 WF |
581 | unsigned long sz = size_inside_page(p, count); |
582 | unsigned long n; | |
1da177e4 | 583 | |
325fda71 KH |
584 | if (!is_vmalloc_or_module_addr((void *)p)) { |
585 | err = -ENXIO; | |
586 | break; | |
587 | } | |
80ad89a0 WF |
588 | n = copy_from_user(kbuf, buf, sz); |
589 | if (n) { | |
325fda71 KH |
590 | err = -EFAULT; |
591 | break; | |
1da177e4 | 592 | } |
c85e9a97 | 593 | vwrite(kbuf, (char *)p, sz); |
80ad89a0 WF |
594 | count -= sz; |
595 | buf += sz; | |
596 | virtr += sz; | |
597 | p += sz; | |
1da177e4 LT |
598 | } |
599 | free_page((unsigned long)kbuf); | |
600 | } | |
601 | ||
325fda71 KH |
602 | *ppos = p; |
603 | return virtr + wrote ? : err; | |
1da177e4 LT |
604 | } |
605 | ||
d7d4d849 | 606 | static ssize_t read_port(struct file *file, char __user *buf, |
1da177e4 LT |
607 | size_t count, loff_t *ppos) |
608 | { | |
609 | unsigned long i = *ppos; | |
610 | char __user *tmp = buf; | |
611 | ||
96d4f267 | 612 | if (!access_ok(buf, count)) |
d7d4d849 | 613 | return -EFAULT; |
1da177e4 | 614 | while (count-- > 0 && i < 65536) { |
d7d4d849 AM |
615 | if (__put_user(inb(i), tmp) < 0) |
616 | return -EFAULT; | |
1da177e4 LT |
617 | i++; |
618 | tmp++; | |
619 | } | |
620 | *ppos = i; | |
621 | return tmp-buf; | |
622 | } | |
623 | ||
d7d4d849 | 624 | static ssize_t write_port(struct file *file, const char __user *buf, |
1da177e4 LT |
625 | size_t count, loff_t *ppos) |
626 | { | |
627 | unsigned long i = *ppos; | |
890537b3 | 628 | const char __user *tmp = buf; |
1da177e4 | 629 | |
96d4f267 | 630 | if (!access_ok(buf, count)) |
1da177e4 LT |
631 | return -EFAULT; |
632 | while (count-- > 0 && i < 65536) { | |
633 | char c; | |
6a0061ba | 634 | |
c654d60e JB |
635 | if (__get_user(c, tmp)) { |
636 | if (tmp > buf) | |
637 | break; | |
d7d4d849 | 638 | return -EFAULT; |
c654d60e | 639 | } |
d7d4d849 | 640 | outb(c, i); |
1da177e4 LT |
641 | i++; |
642 | tmp++; | |
643 | } | |
644 | *ppos = i; | |
645 | return tmp-buf; | |
646 | } | |
1da177e4 | 647 | |
d7d4d849 | 648 | static ssize_t read_null(struct file *file, char __user *buf, |
1da177e4 LT |
649 | size_t count, loff_t *ppos) |
650 | { | |
651 | return 0; | |
652 | } | |
653 | ||
d7d4d849 | 654 | static ssize_t write_null(struct file *file, const char __user *buf, |
1da177e4 LT |
655 | size_t count, loff_t *ppos) |
656 | { | |
657 | return count; | |
658 | } | |
659 | ||
cd28e28d | 660 | static ssize_t read_iter_null(struct kiocb *iocb, struct iov_iter *to) |
162934de ZB |
661 | { |
662 | return 0; | |
663 | } | |
664 | ||
cd28e28d | 665 | static ssize_t write_iter_null(struct kiocb *iocb, struct iov_iter *from) |
162934de | 666 | { |
cd28e28d AV |
667 | size_t count = iov_iter_count(from); |
668 | iov_iter_advance(from, count); | |
669 | return count; | |
162934de ZB |
670 | } |
671 | ||
1ebd32fc JA |
672 | static int pipe_to_null(struct pipe_inode_info *info, struct pipe_buffer *buf, |
673 | struct splice_desc *sd) | |
674 | { | |
675 | return sd->len; | |
676 | } | |
677 | ||
d7d4d849 | 678 | static ssize_t splice_write_null(struct pipe_inode_info *pipe, struct file *out, |
1ebd32fc JA |
679 | loff_t *ppos, size_t len, unsigned int flags) |
680 | { | |
681 | return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_null); | |
682 | } | |
683 | ||
13ba33e8 | 684 | static ssize_t read_iter_zero(struct kiocb *iocb, struct iov_iter *iter) |
1da177e4 | 685 | { |
13ba33e8 | 686 | size_t written = 0; |
1da177e4 | 687 | |
13ba33e8 AV |
688 | while (iov_iter_count(iter)) { |
689 | size_t chunk = iov_iter_count(iter), n; | |
6a0061ba | 690 | |
557ed1fa NP |
691 | if (chunk > PAGE_SIZE) |
692 | chunk = PAGE_SIZE; /* Just for latency reasons */ | |
13ba33e8 AV |
693 | n = iov_iter_zero(chunk, iter); |
694 | if (!n && iov_iter_count(iter)) | |
695 | return written ? written : -EFAULT; | |
696 | written += n; | |
2b838687 LT |
697 | if (signal_pending(current)) |
698 | return written ? written : -ERESTARTSYS; | |
1da177e4 LT |
699 | cond_resched(); |
700 | } | |
13ba33e8 | 701 | return written; |
162934de ZB |
702 | } |
703 | ||
d7d4d849 | 704 | static int mmap_zero(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 705 | { |
557ed1fa | 706 | #ifndef CONFIG_MMU |
1da177e4 | 707 | return -ENOSYS; |
557ed1fa NP |
708 | #endif |
709 | if (vma->vm_flags & VM_SHARED) | |
710 | return shmem_zero_setup(vma); | |
bfd40eaf | 711 | vma_set_anonymous(vma); |
557ed1fa | 712 | return 0; |
1da177e4 | 713 | } |
1da177e4 | 714 | |
c01d5b30 HD |
715 | static unsigned long get_unmapped_area_zero(struct file *file, |
716 | unsigned long addr, unsigned long len, | |
717 | unsigned long pgoff, unsigned long flags) | |
718 | { | |
719 | #ifdef CONFIG_MMU | |
720 | if (flags & MAP_SHARED) { | |
721 | /* | |
722 | * mmap_zero() will call shmem_zero_setup() to create a file, | |
723 | * so use shmem's get_unmapped_area in case it can be huge; | |
724 | * and pass NULL for file as in mmap.c's get_unmapped_area(), | |
725 | * so as not to confuse shmem with our handle on "/dev/zero". | |
726 | */ | |
727 | return shmem_get_unmapped_area(NULL, addr, len, pgoff, flags); | |
728 | } | |
729 | ||
730 | /* Otherwise flags & MAP_PRIVATE: with no shmem object beneath it */ | |
731 | return current->mm->get_unmapped_area(file, addr, len, pgoff, flags); | |
732 | #else | |
733 | return -ENOSYS; | |
734 | #endif | |
735 | } | |
736 | ||
d7d4d849 | 737 | static ssize_t write_full(struct file *file, const char __user *buf, |
1da177e4 LT |
738 | size_t count, loff_t *ppos) |
739 | { | |
740 | return -ENOSPC; | |
741 | } | |
742 | ||
743 | /* | |
744 | * Special lseek() function for /dev/null and /dev/zero. Most notably, you | |
745 | * can fopen() both devices with "a" now. This was previously impossible. | |
746 | * -- SRB. | |
747 | */ | |
d7d4d849 | 748 | static loff_t null_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
749 | { |
750 | return file->f_pos = 0; | |
751 | } | |
752 | ||
753 | /* | |
754 | * The memory devices use the full 32/64 bits of the offset, and so we cannot | |
755 | * check against negative addresses: they are ok. The return value is weird, | |
756 | * though, in that case (0). | |
757 | * | |
758 | * also note that seeking relative to the "end of file" isn't supported: | |
759 | * it has no meaning, so it returns -EINVAL. | |
760 | */ | |
d7d4d849 | 761 | static loff_t memory_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
762 | { |
763 | loff_t ret; | |
764 | ||
5955102c | 765 | inode_lock(file_inode(file)); |
1da177e4 | 766 | switch (orig) { |
d7d4d849 AM |
767 | case SEEK_CUR: |
768 | offset += file->f_pos; | |
5faecb01 | 769 | /* fall through */ |
d7d4d849 AM |
770 | case SEEK_SET: |
771 | /* to avoid userland mistaking f_pos=-9 as -EBADF=-9 */ | |
ecb63a1b | 772 | if ((unsigned long long)offset >= -MAX_ERRNO) { |
d7d4d849 | 773 | ret = -EOVERFLOW; |
1da177e4 | 774 | break; |
d7d4d849 AM |
775 | } |
776 | file->f_pos = offset; | |
777 | ret = file->f_pos; | |
778 | force_successful_syscall_return(); | |
779 | break; | |
780 | default: | |
781 | ret = -EINVAL; | |
1da177e4 | 782 | } |
5955102c | 783 | inode_unlock(file_inode(file)); |
1da177e4 LT |
784 | return ret; |
785 | } | |
786 | ||
890537b3 | 787 | static int open_port(struct inode *inode, struct file *filp) |
1da177e4 LT |
788 | { |
789 | return capable(CAP_SYS_RAWIO) ? 0 : -EPERM; | |
790 | } | |
791 | ||
792 | #define zero_lseek null_lseek | |
793 | #define full_lseek null_lseek | |
794 | #define write_zero write_null | |
cd28e28d | 795 | #define write_iter_zero write_iter_null |
1da177e4 LT |
796 | #define open_mem open_port |
797 | #define open_kmem open_mem | |
798 | ||
73f0718e | 799 | static const struct file_operations __maybe_unused mem_fops = { |
1da177e4 LT |
800 | .llseek = memory_lseek, |
801 | .read = read_mem, | |
802 | .write = write_mem, | |
803 | .mmap = mmap_mem, | |
804 | .open = open_mem, | |
b4caecd4 | 805 | #ifndef CONFIG_MMU |
5da6185b | 806 | .get_unmapped_area = get_unmapped_area_mem, |
b4caecd4 CH |
807 | .mmap_capabilities = memory_mmap_capabilities, |
808 | #endif | |
1da177e4 LT |
809 | }; |
810 | ||
a8c91252 | 811 | static const struct file_operations __maybe_unused kmem_fops = { |
1da177e4 LT |
812 | .llseek = memory_lseek, |
813 | .read = read_kmem, | |
814 | .write = write_kmem, | |
815 | .mmap = mmap_kmem, | |
816 | .open = open_kmem, | |
b4caecd4 | 817 | #ifndef CONFIG_MMU |
5da6185b | 818 | .get_unmapped_area = get_unmapped_area_mem, |
b4caecd4 CH |
819 | .mmap_capabilities = memory_mmap_capabilities, |
820 | #endif | |
1da177e4 LT |
821 | }; |
822 | ||
62322d25 | 823 | static const struct file_operations null_fops = { |
1da177e4 LT |
824 | .llseek = null_lseek, |
825 | .read = read_null, | |
826 | .write = write_null, | |
cd28e28d AV |
827 | .read_iter = read_iter_null, |
828 | .write_iter = write_iter_null, | |
1ebd32fc | 829 | .splice_write = splice_write_null, |
1da177e4 LT |
830 | }; |
831 | ||
3a4bc2fb | 832 | static const struct file_operations __maybe_unused port_fops = { |
1da177e4 LT |
833 | .llseek = memory_lseek, |
834 | .read = read_port, | |
835 | .write = write_port, | |
836 | .open = open_port, | |
837 | }; | |
1da177e4 | 838 | |
62322d25 | 839 | static const struct file_operations zero_fops = { |
1da177e4 | 840 | .llseek = zero_lseek, |
1da177e4 | 841 | .write = write_zero, |
13ba33e8 | 842 | .read_iter = read_iter_zero, |
cd28e28d | 843 | .write_iter = write_iter_zero, |
1da177e4 | 844 | .mmap = mmap_zero, |
c01d5b30 | 845 | .get_unmapped_area = get_unmapped_area_zero, |
b4caecd4 CH |
846 | #ifndef CONFIG_MMU |
847 | .mmap_capabilities = zero_mmap_capabilities, | |
848 | #endif | |
1da177e4 LT |
849 | }; |
850 | ||
62322d25 | 851 | static const struct file_operations full_fops = { |
1da177e4 | 852 | .llseek = full_lseek, |
13ba33e8 | 853 | .read_iter = read_iter_zero, |
1da177e4 LT |
854 | .write = write_full, |
855 | }; | |
856 | ||
389e0cb9 KS |
857 | static const struct memdev { |
858 | const char *name; | |
2c9ede55 | 859 | umode_t mode; |
389e0cb9 | 860 | const struct file_operations *fops; |
b4caecd4 | 861 | fmode_t fmode; |
389e0cb9 | 862 | } devlist[] = { |
73f0718e | 863 | #ifdef CONFIG_DEVMEM |
b4caecd4 | 864 | [1] = { "mem", 0, &mem_fops, FMODE_UNSIGNED_OFFSET }, |
73f0718e | 865 | #endif |
b781ecb6 | 866 | #ifdef CONFIG_DEVKMEM |
b4caecd4 | 867 | [2] = { "kmem", 0, &kmem_fops, FMODE_UNSIGNED_OFFSET }, |
b781ecb6 | 868 | #endif |
b4caecd4 | 869 | [3] = { "null", 0666, &null_fops, 0 }, |
4f911d64 | 870 | #ifdef CONFIG_DEVPORT |
b4caecd4 | 871 | [4] = { "port", 0, &port_fops, 0 }, |
1da177e4 | 872 | #endif |
b4caecd4 CH |
873 | [5] = { "zero", 0666, &zero_fops, 0 }, |
874 | [7] = { "full", 0666, &full_fops, 0 }, | |
875 | [8] = { "random", 0666, &random_fops, 0 }, | |
876 | [9] = { "urandom", 0666, &urandom_fops, 0 }, | |
7f3a781d | 877 | #ifdef CONFIG_PRINTK |
b4caecd4 | 878 | [11] = { "kmsg", 0644, &kmsg_fops, 0 }, |
7f3a781d | 879 | #endif |
d6f47bef ASF |
880 | }; |
881 | ||
882 | static int memory_open(struct inode *inode, struct file *filp) | |
883 | { | |
389e0cb9 KS |
884 | int minor; |
885 | const struct memdev *dev; | |
d6f47bef | 886 | |
389e0cb9 KS |
887 | minor = iminor(inode); |
888 | if (minor >= ARRAY_SIZE(devlist)) | |
205153aa | 889 | return -ENXIO; |
d6f47bef | 890 | |
389e0cb9 KS |
891 | dev = &devlist[minor]; |
892 | if (!dev->fops) | |
205153aa | 893 | return -ENXIO; |
d6f47bef | 894 | |
389e0cb9 | 895 | filp->f_op = dev->fops; |
b4caecd4 | 896 | filp->f_mode |= dev->fmode; |
4a3956c7 | 897 | |
389e0cb9 | 898 | if (dev->fops->open) |
205153aa FW |
899 | return dev->fops->open(inode, filp); |
900 | ||
901 | return 0; | |
1da177e4 LT |
902 | } |
903 | ||
62322d25 | 904 | static const struct file_operations memory_fops = { |
d7d4d849 | 905 | .open = memory_open, |
6038f373 | 906 | .llseek = noop_llseek, |
1da177e4 LT |
907 | }; |
908 | ||
2c9ede55 | 909 | static char *mem_devnode(struct device *dev, umode_t *mode) |
e454cea2 KS |
910 | { |
911 | if (mode && devlist[MINOR(dev->devt)].mode) | |
912 | *mode = devlist[MINOR(dev->devt)].mode; | |
913 | return NULL; | |
914 | } | |
915 | ||
ca8eca68 | 916 | static struct class *mem_class; |
1da177e4 LT |
917 | |
918 | static int __init chr_dev_init(void) | |
919 | { | |
389e0cb9 | 920 | int minor; |
1da177e4 | 921 | |
d7d4d849 | 922 | if (register_chrdev(MEM_MAJOR, "mem", &memory_fops)) |
1da177e4 LT |
923 | printk("unable to get major %d for memory devs\n", MEM_MAJOR); |
924 | ||
ca8eca68 | 925 | mem_class = class_create(THIS_MODULE, "mem"); |
6e191f7b AB |
926 | if (IS_ERR(mem_class)) |
927 | return PTR_ERR(mem_class); | |
928 | ||
e454cea2 | 929 | mem_class->devnode = mem_devnode; |
389e0cb9 KS |
930 | for (minor = 1; minor < ARRAY_SIZE(devlist); minor++) { |
931 | if (!devlist[minor].name) | |
932 | continue; | |
e1612de9 HM |
933 | |
934 | /* | |
890537b3 | 935 | * Create /dev/port? |
e1612de9 HM |
936 | */ |
937 | if ((minor == DEVPORT_MINOR) && !arch_has_dev_port()) | |
938 | continue; | |
939 | ||
389e0cb9 KS |
940 | device_create(mem_class, NULL, MKDEV(MEM_MAJOR, minor), |
941 | NULL, devlist[minor].name); | |
942 | } | |
ebf644c4 | 943 | |
31d1d48e | 944 | return tty_init(); |
1da177e4 LT |
945 | } |
946 | ||
947 | fs_initcall(chr_dev_init); |