]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/drivers/char/mem.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * | |
d7d4d849 | 6 | * Added devfs support. |
1da177e4 | 7 | * Jan-11-1998, C. Scott Ananian <cananian@alumni.princeton.edu> |
af901ca1 | 8 | * Shared /dev/zero mmapping support, Feb 2000, Kanoj Sarcar <kanoj@sgi.com> |
1da177e4 LT |
9 | */ |
10 | ||
1da177e4 LT |
11 | #include <linux/mm.h> |
12 | #include <linux/miscdevice.h> | |
13 | #include <linux/slab.h> | |
14 | #include <linux/vmalloc.h> | |
15 | #include <linux/mman.h> | |
16 | #include <linux/random.h> | |
17 | #include <linux/init.h> | |
18 | #include <linux/raw.h> | |
19 | #include <linux/tty.h> | |
20 | #include <linux/capability.h> | |
1da177e4 LT |
21 | #include <linux/ptrace.h> |
22 | #include <linux/device.h> | |
50b1fdbd | 23 | #include <linux/highmem.h> |
1da177e4 | 24 | #include <linux/backing-dev.h> |
c01d5b30 | 25 | #include <linux/shmem_fs.h> |
d6b29d7c | 26 | #include <linux/splice.h> |
b8a3ad5b | 27 | #include <linux/pfn.h> |
66300e66 | 28 | #include <linux/export.h> |
e1612de9 | 29 | #include <linux/io.h> |
e2e40f2c | 30 | #include <linux/uio.h> |
e6540ba2 | 31 | #include <linux/module.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 |
3de39707 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 | |
3de39707 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; |
1da177e4 | 110 | |
08d2d00b PT |
111 | if (p != *ppos) |
112 | return 0; | |
113 | ||
136939a2 | 114 | if (!valid_phys_addr_range(p, count)) |
1da177e4 LT |
115 | return -EFAULT; |
116 | read = 0; | |
117 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
118 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
119 | if (p < PAGE_SIZE) { | |
7fabaddd | 120 | sz = size_inside_page(p, count); |
1da177e4 LT |
121 | if (sz > 0) { |
122 | if (clear_user(buf, sz)) | |
123 | return -EFAULT; | |
d7d4d849 AM |
124 | buf += sz; |
125 | p += sz; | |
126 | count -= sz; | |
127 | read += sz; | |
1da177e4 LT |
128 | } |
129 | } | |
130 | #endif | |
131 | ||
132 | while (count > 0) { | |
fa29e97b | 133 | unsigned long remaining; |
3de39707 | 134 | int allowed; |
fa29e97b | 135 | |
f222318e | 136 | sz = size_inside_page(p, count); |
1da177e4 | 137 | |
3de39707 KC |
138 | allowed = page_is_allowed(p >> PAGE_SHIFT); |
139 | if (!allowed) | |
e045fb2a | 140 | return -EPERM; |
3de39707 KC |
141 | if (allowed == 2) { |
142 | /* Show zeros for restricted memory. */ | |
143 | remaining = clear_user(buf, sz); | |
144 | } else { | |
145 | /* | |
146 | * On ia64 if a page has been mapped somewhere as | |
147 | * uncached, then it must also be accessed uncached | |
148 | * by the kernel or data corruption may occur. | |
149 | */ | |
150 | ptr = xlate_dev_mem_ptr(p); | |
151 | if (!ptr) | |
152 | return -EFAULT; | |
e045fb2a | 153 | |
3de39707 KC |
154 | remaining = copy_to_user(buf, ptr, sz); |
155 | ||
156 | unxlate_dev_mem_ptr(p, ptr); | |
157 | } | |
1da177e4 | 158 | |
fa29e97b WF |
159 | if (remaining) |
160 | return -EFAULT; | |
e045fb2a | 161 | |
1da177e4 LT |
162 | buf += sz; |
163 | p += sz; | |
164 | count -= sz; | |
165 | read += sz; | |
166 | } | |
167 | ||
168 | *ppos += read; | |
169 | return read; | |
170 | } | |
171 | ||
d7d4d849 | 172 | static ssize_t write_mem(struct file *file, const char __user *buf, |
1da177e4 LT |
173 | size_t count, loff_t *ppos) |
174 | { | |
7e6735c3 | 175 | phys_addr_t p = *ppos; |
1da177e4 LT |
176 | ssize_t written, sz; |
177 | unsigned long copied; | |
178 | void *ptr; | |
179 | ||
08d2d00b PT |
180 | if (p != *ppos) |
181 | return -EFBIG; | |
182 | ||
99f0760c MG |
183 | if (secure_modules()) |
184 | return -EPERM; | |
185 | ||
136939a2 | 186 | if (!valid_phys_addr_range(p, count)) |
1da177e4 LT |
187 | return -EFAULT; |
188 | ||
189 | written = 0; | |
190 | ||
191 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
192 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
193 | if (p < PAGE_SIZE) { | |
7fabaddd | 194 | sz = size_inside_page(p, count); |
1da177e4 LT |
195 | /* Hmm. Do something? */ |
196 | buf += sz; | |
197 | p += sz; | |
198 | count -= sz; | |
199 | written += sz; | |
200 | } | |
201 | #endif | |
202 | ||
203 | while (count > 0) { | |
3de39707 KC |
204 | int allowed; |
205 | ||
f222318e | 206 | sz = size_inside_page(p, count); |
1da177e4 | 207 | |
3de39707 KC |
208 | allowed = page_is_allowed(p >> PAGE_SHIFT); |
209 | if (!allowed) | |
e045fb2a | 210 | return -EPERM; |
211 | ||
3de39707 KC |
212 | /* Skip actual writing when a page is marked as restricted. */ |
213 | if (allowed == 1) { | |
214 | /* | |
215 | * On ia64 if a page has been mapped somewhere as | |
216 | * uncached, then it must also be accessed uncached | |
217 | * by the kernel or data corruption may occur. | |
218 | */ | |
219 | ptr = xlate_dev_mem_ptr(p); | |
220 | if (!ptr) { | |
221 | if (written) | |
222 | break; | |
223 | return -EFAULT; | |
224 | } | |
225 | ||
226 | copied = copy_from_user(ptr, buf, sz); | |
227 | unxlate_dev_mem_ptr(p, ptr); | |
228 | if (copied) { | |
229 | written += sz - copied; | |
230 | if (written) | |
231 | break; | |
232 | return -EFAULT; | |
233 | } | |
1da177e4 | 234 | } |
e045fb2a | 235 | |
1da177e4 LT |
236 | buf += sz; |
237 | p += sz; | |
238 | count -= sz; | |
239 | written += sz; | |
240 | } | |
241 | ||
242 | *ppos += written; | |
243 | return written; | |
244 | } | |
245 | ||
d7d4d849 | 246 | int __weak phys_mem_access_prot_allowed(struct file *file, |
f0970c13 | 247 | unsigned long pfn, unsigned long size, pgprot_t *vma_prot) |
248 | { | |
249 | return 1; | |
250 | } | |
251 | ||
44ac8413 | 252 | #ifndef __HAVE_PHYS_MEM_ACCESS_PROT |
d7d4d849 AM |
253 | |
254 | /* | |
255 | * Architectures vary in how they handle caching for addresses | |
256 | * outside of main memory. | |
257 | * | |
258 | */ | |
ea56f411 | 259 | #ifdef pgprot_noncached |
7e6735c3 | 260 | static int uncached_access(struct file *file, phys_addr_t addr) |
d7d4d849 AM |
261 | { |
262 | #if defined(CONFIG_IA64) | |
263 | /* | |
264 | * On ia64, we ignore O_DSYNC because we cannot tolerate memory | |
265 | * attribute aliases. | |
266 | */ | |
267 | return !(efi_mem_attributes(addr) & EFI_MEMORY_WB); | |
268 | #elif defined(CONFIG_MIPS) | |
269 | { | |
270 | extern int __uncached_access(struct file *file, | |
271 | unsigned long addr); | |
272 | ||
273 | return __uncached_access(file, addr); | |
274 | } | |
275 | #else | |
276 | /* | |
277 | * Accessing memory above the top the kernel knows about or through a | |
278 | * file pointer | |
279 | * that was marked O_DSYNC will be done non-cached. | |
280 | */ | |
281 | if (file->f_flags & O_DSYNC) | |
282 | return 1; | |
283 | return addr >= __pa(high_memory); | |
284 | #endif | |
285 | } | |
ea56f411 | 286 | #endif |
d7d4d849 | 287 | |
44ac8413 BH |
288 | static pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, |
289 | unsigned long size, pgprot_t vma_prot) | |
290 | { | |
291 | #ifdef pgprot_noncached | |
7e6735c3 | 292 | phys_addr_t offset = pfn << PAGE_SHIFT; |
44ac8413 BH |
293 | |
294 | if (uncached_access(file, offset)) | |
295 | return pgprot_noncached(vma_prot); | |
296 | #endif | |
297 | return vma_prot; | |
298 | } | |
299 | #endif | |
300 | ||
5da6185b DH |
301 | #ifndef CONFIG_MMU |
302 | static unsigned long get_unmapped_area_mem(struct file *file, | |
303 | unsigned long addr, | |
304 | unsigned long len, | |
305 | unsigned long pgoff, | |
306 | unsigned long flags) | |
307 | { | |
308 | if (!valid_mmap_phys_addr_range(pgoff, len)) | |
309 | return (unsigned long) -EINVAL; | |
8a93258c | 310 | return pgoff << PAGE_SHIFT; |
5da6185b DH |
311 | } |
312 | ||
b4caecd4 CH |
313 | /* permit direct mmap, for read, write or exec */ |
314 | static unsigned memory_mmap_capabilities(struct file *file) | |
315 | { | |
316 | return NOMMU_MAP_DIRECT | | |
317 | NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC; | |
318 | } | |
319 | ||
320 | static unsigned zero_mmap_capabilities(struct file *file) | |
321 | { | |
322 | return NOMMU_MAP_COPY; | |
323 | } | |
324 | ||
5da6185b DH |
325 | /* can't do an in-place private mapping if there's no MMU */ |
326 | static inline int private_mapping_ok(struct vm_area_struct *vma) | |
327 | { | |
328 | return vma->vm_flags & VM_MAYSHARE; | |
329 | } | |
330 | #else | |
5da6185b DH |
331 | |
332 | static inline int private_mapping_ok(struct vm_area_struct *vma) | |
333 | { | |
334 | return 1; | |
335 | } | |
336 | #endif | |
337 | ||
f0f37e2f | 338 | static const struct vm_operations_struct mmap_mem_ops = { |
7ae8ed50 RR |
339 | #ifdef CONFIG_HAVE_IOREMAP_PROT |
340 | .access = generic_access_phys | |
341 | #endif | |
e7f260a2 | 342 | }; |
343 | ||
d7d4d849 | 344 | static int mmap_mem(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 345 | { |
80851ef2 BH |
346 | size_t size = vma->vm_end - vma->vm_start; |
347 | ||
06c67bef | 348 | if (!valid_mmap_phys_addr_range(vma->vm_pgoff, size)) |
80851ef2 BH |
349 | return -EINVAL; |
350 | ||
5da6185b DH |
351 | if (!private_mapping_ok(vma)) |
352 | return -ENOSYS; | |
353 | ||
e2beb3ea VP |
354 | if (!range_is_allowed(vma->vm_pgoff, size)) |
355 | return -EPERM; | |
356 | ||
f0970c13 | 357 | if (!phys_mem_access_prot_allowed(file, vma->vm_pgoff, size, |
358 | &vma->vm_page_prot)) | |
359 | return -EINVAL; | |
360 | ||
8b150478 | 361 | vma->vm_page_prot = phys_mem_access_prot(file, vma->vm_pgoff, |
80851ef2 | 362 | size, |
1da177e4 | 363 | vma->vm_page_prot); |
1da177e4 | 364 | |
e7f260a2 | 365 | vma->vm_ops = &mmap_mem_ops; |
366 | ||
314e51b9 | 367 | /* Remap-pfn-range will mark the range VM_IO */ |
1da177e4 LT |
368 | if (remap_pfn_range(vma, |
369 | vma->vm_start, | |
370 | vma->vm_pgoff, | |
80851ef2 | 371 | size, |
e7f260a2 | 372 | vma->vm_page_prot)) { |
1da177e4 | 373 | return -EAGAIN; |
e7f260a2 | 374 | } |
1da177e4 LT |
375 | return 0; |
376 | } | |
377 | ||
d7d4d849 | 378 | static int mmap_kmem(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 379 | { |
4bb82551 LT |
380 | unsigned long pfn; |
381 | ||
6d3154cc LT |
382 | /* Turn a kernel-virtual address into a physical page frame */ |
383 | pfn = __pa((u64)vma->vm_pgoff << PAGE_SHIFT) >> PAGE_SHIFT; | |
4bb82551 | 384 | |
1da177e4 | 385 | /* |
d7d4d849 AM |
386 | * RED-PEN: on some architectures there is more mapped memory than |
387 | * available in mem_map which pfn_valid checks for. Perhaps should add a | |
388 | * new macro here. | |
1da177e4 LT |
389 | * |
390 | * RED-PEN: vmalloc is not supported right now. | |
391 | */ | |
4bb82551 | 392 | if (!pfn_valid(pfn)) |
1da177e4 | 393 | return -EIO; |
4bb82551 LT |
394 | |
395 | vma->vm_pgoff = pfn; | |
1da177e4 LT |
396 | return mmap_mem(file, vma); |
397 | } | |
398 | ||
1da177e4 LT |
399 | /* |
400 | * This function reads the *virtual* memory as seen by the kernel. | |
401 | */ | |
d7d4d849 | 402 | static ssize_t read_kmem(struct file *file, char __user *buf, |
1da177e4 LT |
403 | size_t count, loff_t *ppos) |
404 | { | |
405 | unsigned long p = *ppos; | |
406 | ssize_t low_count, read, sz; | |
890537b3 | 407 | char *kbuf; /* k-addr because vread() takes vmlist_lock rwlock */ |
325fda71 | 408 | int err = 0; |
1da177e4 LT |
409 | |
410 | read = 0; | |
411 | if (p < (unsigned long) high_memory) { | |
412 | low_count = count; | |
d7d4d849 AM |
413 | if (count > (unsigned long)high_memory - p) |
414 | low_count = (unsigned long)high_memory - p; | |
1da177e4 LT |
415 | |
416 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
417 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
418 | if (p < PAGE_SIZE && low_count > 0) { | |
7fabaddd WF |
419 | sz = size_inside_page(p, low_count); |
420 | if (clear_user(buf, sz)) | |
1da177e4 | 421 | return -EFAULT; |
7fabaddd WF |
422 | buf += sz; |
423 | p += sz; | |
424 | read += sz; | |
425 | low_count -= sz; | |
426 | count -= sz; | |
1da177e4 LT |
427 | } |
428 | #endif | |
429 | while (low_count > 0) { | |
f222318e | 430 | sz = size_inside_page(p, low_count); |
1da177e4 LT |
431 | |
432 | /* | |
433 | * On ia64 if a page has been mapped somewhere as | |
434 | * uncached, then it must also be accessed uncached | |
435 | * by the kernel or data corruption may occur | |
436 | */ | |
4707a341 | 437 | kbuf = xlate_dev_kmem_ptr((void *)p); |
488debb9 RM |
438 | if (!virt_addr_valid(kbuf)) |
439 | return -ENXIO; | |
1da177e4 LT |
440 | |
441 | if (copy_to_user(buf, kbuf, sz)) | |
442 | return -EFAULT; | |
443 | buf += sz; | |
444 | p += sz; | |
445 | read += sz; | |
446 | low_count -= sz; | |
447 | count -= sz; | |
448 | } | |
449 | } | |
450 | ||
451 | if (count > 0) { | |
452 | kbuf = (char *)__get_free_page(GFP_KERNEL); | |
453 | if (!kbuf) | |
454 | return -ENOMEM; | |
455 | while (count > 0) { | |
80ad89a0 | 456 | sz = size_inside_page(p, count); |
325fda71 KH |
457 | if (!is_vmalloc_or_module_addr((void *)p)) { |
458 | err = -ENXIO; | |
459 | break; | |
460 | } | |
80ad89a0 WF |
461 | sz = vread(kbuf, (char *)p, sz); |
462 | if (!sz) | |
1da177e4 | 463 | break; |
80ad89a0 | 464 | if (copy_to_user(buf, kbuf, sz)) { |
325fda71 KH |
465 | err = -EFAULT; |
466 | break; | |
1da177e4 | 467 | } |
80ad89a0 WF |
468 | count -= sz; |
469 | buf += sz; | |
470 | read += sz; | |
471 | p += sz; | |
1da177e4 LT |
472 | } |
473 | free_page((unsigned long)kbuf); | |
474 | } | |
325fda71 KH |
475 | *ppos = p; |
476 | return read ? read : err; | |
1da177e4 LT |
477 | } |
478 | ||
479 | ||
d7d4d849 AM |
480 | static ssize_t do_write_kmem(unsigned long p, const char __user *buf, |
481 | size_t count, loff_t *ppos) | |
1da177e4 LT |
482 | { |
483 | ssize_t written, sz; | |
484 | unsigned long copied; | |
485 | ||
486 | written = 0; | |
487 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
488 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
ee32398f WF |
489 | if (p < PAGE_SIZE) { |
490 | sz = size_inside_page(p, count); | |
1da177e4 LT |
491 | /* Hmm. Do something? */ |
492 | buf += sz; | |
493 | p += sz; | |
1da177e4 LT |
494 | count -= sz; |
495 | written += sz; | |
496 | } | |
497 | #endif | |
498 | ||
499 | while (count > 0) { | |
4707a341 | 500 | void *ptr; |
1da177e4 | 501 | |
ee32398f | 502 | sz = size_inside_page(p, count); |
1da177e4 LT |
503 | |
504 | /* | |
d7d4d849 AM |
505 | * On ia64 if a page has been mapped somewhere as uncached, then |
506 | * it must also be accessed uncached by the kernel or data | |
507 | * corruption may occur. | |
1da177e4 | 508 | */ |
4707a341 | 509 | ptr = xlate_dev_kmem_ptr((void *)p); |
488debb9 RM |
510 | if (!virt_addr_valid(ptr)) |
511 | return -ENXIO; | |
1da177e4 LT |
512 | |
513 | copied = copy_from_user(ptr, buf, sz); | |
514 | if (copied) { | |
c654d60e JB |
515 | written += sz - copied; |
516 | if (written) | |
517 | break; | |
1da177e4 LT |
518 | return -EFAULT; |
519 | } | |
520 | buf += sz; | |
521 | p += sz; | |
1da177e4 LT |
522 | count -= sz; |
523 | written += sz; | |
524 | } | |
525 | ||
526 | *ppos += written; | |
527 | return written; | |
528 | } | |
529 | ||
1da177e4 LT |
530 | /* |
531 | * This function writes to the *virtual* memory as seen by the kernel. | |
532 | */ | |
d7d4d849 | 533 | static ssize_t write_kmem(struct file *file, const char __user *buf, |
1da177e4 LT |
534 | size_t count, loff_t *ppos) |
535 | { | |
536 | unsigned long p = *ppos; | |
537 | ssize_t wrote = 0; | |
538 | ssize_t virtr = 0; | |
890537b3 | 539 | char *kbuf; /* k-addr because vwrite() takes vmlist_lock rwlock */ |
325fda71 | 540 | int err = 0; |
1da177e4 | 541 | |
99f0760c MG |
542 | if (!pfn_valid(PFN_DOWN(p))) |
543 | return -EIO; | |
544 | if (secure_modules()) | |
545 | return -EPERM; | |
546 | ||
1da177e4 | 547 | if (p < (unsigned long) high_memory) { |
80ad89a0 WF |
548 | unsigned long to_write = min_t(unsigned long, count, |
549 | (unsigned long)high_memory - p); | |
ee32398f | 550 | wrote = do_write_kmem(p, buf, to_write, ppos); |
80ad89a0 WF |
551 | if (wrote != to_write) |
552 | return wrote; | |
1da177e4 LT |
553 | p += wrote; |
554 | buf += wrote; | |
555 | count -= wrote; | |
556 | } | |
557 | ||
558 | if (count > 0) { | |
559 | kbuf = (char *)__get_free_page(GFP_KERNEL); | |
560 | if (!kbuf) | |
561 | return wrote ? wrote : -ENOMEM; | |
562 | while (count > 0) { | |
80ad89a0 WF |
563 | unsigned long sz = size_inside_page(p, count); |
564 | unsigned long n; | |
1da177e4 | 565 | |
325fda71 KH |
566 | if (!is_vmalloc_or_module_addr((void *)p)) { |
567 | err = -ENXIO; | |
568 | break; | |
569 | } | |
80ad89a0 WF |
570 | n = copy_from_user(kbuf, buf, sz); |
571 | if (n) { | |
325fda71 KH |
572 | err = -EFAULT; |
573 | break; | |
1da177e4 | 574 | } |
c85e9a97 | 575 | vwrite(kbuf, (char *)p, sz); |
80ad89a0 WF |
576 | count -= sz; |
577 | buf += sz; | |
578 | virtr += sz; | |
579 | p += sz; | |
1da177e4 LT |
580 | } |
581 | free_page((unsigned long)kbuf); | |
582 | } | |
583 | ||
325fda71 KH |
584 | *ppos = p; |
585 | return virtr + wrote ? : err; | |
1da177e4 LT |
586 | } |
587 | ||
d7d4d849 | 588 | static ssize_t read_port(struct file *file, char __user *buf, |
1da177e4 LT |
589 | size_t count, loff_t *ppos) |
590 | { | |
591 | unsigned long i = *ppos; | |
592 | char __user *tmp = buf; | |
593 | ||
594 | if (!access_ok(VERIFY_WRITE, buf, count)) | |
d7d4d849 | 595 | return -EFAULT; |
1da177e4 | 596 | while (count-- > 0 && i < 65536) { |
d7d4d849 AM |
597 | if (__put_user(inb(i), tmp) < 0) |
598 | return -EFAULT; | |
1da177e4 LT |
599 | i++; |
600 | tmp++; | |
601 | } | |
602 | *ppos = i; | |
603 | return tmp-buf; | |
604 | } | |
605 | ||
d7d4d849 | 606 | static ssize_t write_port(struct file *file, const char __user *buf, |
1da177e4 LT |
607 | size_t count, loff_t *ppos) |
608 | { | |
609 | unsigned long i = *ppos; | |
890537b3 | 610 | const char __user *tmp = buf; |
1da177e4 | 611 | |
e6540ba2 MG |
612 | if (secure_modules()) |
613 | return -EPERM; | |
614 | ||
d7d4d849 | 615 | if (!access_ok(VERIFY_READ, buf, count)) |
1da177e4 LT |
616 | return -EFAULT; |
617 | while (count-- > 0 && i < 65536) { | |
618 | char c; | |
6a0061ba | 619 | |
c654d60e JB |
620 | if (__get_user(c, tmp)) { |
621 | if (tmp > buf) | |
622 | break; | |
d7d4d849 | 623 | return -EFAULT; |
c654d60e | 624 | } |
d7d4d849 | 625 | outb(c, i); |
1da177e4 LT |
626 | i++; |
627 | tmp++; | |
628 | } | |
629 | *ppos = i; | |
630 | return tmp-buf; | |
631 | } | |
1da177e4 | 632 | |
d7d4d849 | 633 | static ssize_t read_null(struct file *file, char __user *buf, |
1da177e4 LT |
634 | size_t count, loff_t *ppos) |
635 | { | |
636 | return 0; | |
637 | } | |
638 | ||
d7d4d849 | 639 | static ssize_t write_null(struct file *file, const char __user *buf, |
1da177e4 LT |
640 | size_t count, loff_t *ppos) |
641 | { | |
642 | return count; | |
643 | } | |
644 | ||
cd28e28d | 645 | static ssize_t read_iter_null(struct kiocb *iocb, struct iov_iter *to) |
162934de ZB |
646 | { |
647 | return 0; | |
648 | } | |
649 | ||
cd28e28d | 650 | static ssize_t write_iter_null(struct kiocb *iocb, struct iov_iter *from) |
162934de | 651 | { |
cd28e28d AV |
652 | size_t count = iov_iter_count(from); |
653 | iov_iter_advance(from, count); | |
654 | return count; | |
162934de ZB |
655 | } |
656 | ||
1ebd32fc JA |
657 | static int pipe_to_null(struct pipe_inode_info *info, struct pipe_buffer *buf, |
658 | struct splice_desc *sd) | |
659 | { | |
660 | return sd->len; | |
661 | } | |
662 | ||
d7d4d849 | 663 | static ssize_t splice_write_null(struct pipe_inode_info *pipe, struct file *out, |
1ebd32fc JA |
664 | loff_t *ppos, size_t len, unsigned int flags) |
665 | { | |
666 | return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_null); | |
667 | } | |
668 | ||
13ba33e8 | 669 | static ssize_t read_iter_zero(struct kiocb *iocb, struct iov_iter *iter) |
1da177e4 | 670 | { |
13ba33e8 | 671 | size_t written = 0; |
1da177e4 | 672 | |
13ba33e8 AV |
673 | while (iov_iter_count(iter)) { |
674 | size_t chunk = iov_iter_count(iter), n; | |
6a0061ba | 675 | |
557ed1fa NP |
676 | if (chunk > PAGE_SIZE) |
677 | chunk = PAGE_SIZE; /* Just for latency reasons */ | |
13ba33e8 AV |
678 | n = iov_iter_zero(chunk, iter); |
679 | if (!n && iov_iter_count(iter)) | |
680 | return written ? written : -EFAULT; | |
681 | written += n; | |
2b838687 LT |
682 | if (signal_pending(current)) |
683 | return written ? written : -ERESTARTSYS; | |
1da177e4 LT |
684 | cond_resched(); |
685 | } | |
13ba33e8 | 686 | return written; |
162934de ZB |
687 | } |
688 | ||
d7d4d849 | 689 | static int mmap_zero(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 690 | { |
557ed1fa | 691 | #ifndef CONFIG_MMU |
1da177e4 | 692 | return -ENOSYS; |
557ed1fa NP |
693 | #endif |
694 | if (vma->vm_flags & VM_SHARED) | |
695 | return shmem_zero_setup(vma); | |
696 | return 0; | |
1da177e4 | 697 | } |
1da177e4 | 698 | |
c01d5b30 HD |
699 | static unsigned long get_unmapped_area_zero(struct file *file, |
700 | unsigned long addr, unsigned long len, | |
701 | unsigned long pgoff, unsigned long flags) | |
702 | { | |
703 | #ifdef CONFIG_MMU | |
704 | if (flags & MAP_SHARED) { | |
705 | /* | |
706 | * mmap_zero() will call shmem_zero_setup() to create a file, | |
707 | * so use shmem's get_unmapped_area in case it can be huge; | |
708 | * and pass NULL for file as in mmap.c's get_unmapped_area(), | |
709 | * so as not to confuse shmem with our handle on "/dev/zero". | |
710 | */ | |
711 | return shmem_get_unmapped_area(NULL, addr, len, pgoff, flags); | |
712 | } | |
713 | ||
714 | /* Otherwise flags & MAP_PRIVATE: with no shmem object beneath it */ | |
715 | return current->mm->get_unmapped_area(file, addr, len, pgoff, flags); | |
716 | #else | |
717 | return -ENOSYS; | |
718 | #endif | |
719 | } | |
720 | ||
d7d4d849 | 721 | static ssize_t write_full(struct file *file, const char __user *buf, |
1da177e4 LT |
722 | size_t count, loff_t *ppos) |
723 | { | |
724 | return -ENOSPC; | |
725 | } | |
726 | ||
727 | /* | |
728 | * Special lseek() function for /dev/null and /dev/zero. Most notably, you | |
729 | * can fopen() both devices with "a" now. This was previously impossible. | |
730 | * -- SRB. | |
731 | */ | |
d7d4d849 | 732 | static loff_t null_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
733 | { |
734 | return file->f_pos = 0; | |
735 | } | |
736 | ||
737 | /* | |
738 | * The memory devices use the full 32/64 bits of the offset, and so we cannot | |
739 | * check against negative addresses: they are ok. The return value is weird, | |
740 | * though, in that case (0). | |
741 | * | |
742 | * also note that seeking relative to the "end of file" isn't supported: | |
743 | * it has no meaning, so it returns -EINVAL. | |
744 | */ | |
d7d4d849 | 745 | static loff_t memory_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
746 | { |
747 | loff_t ret; | |
748 | ||
5955102c | 749 | inode_lock(file_inode(file)); |
1da177e4 | 750 | switch (orig) { |
d7d4d849 AM |
751 | case SEEK_CUR: |
752 | offset += file->f_pos; | |
d7d4d849 AM |
753 | case SEEK_SET: |
754 | /* to avoid userland mistaking f_pos=-9 as -EBADF=-9 */ | |
ecb63a1b | 755 | if ((unsigned long long)offset >= -MAX_ERRNO) { |
d7d4d849 | 756 | ret = -EOVERFLOW; |
1da177e4 | 757 | break; |
d7d4d849 AM |
758 | } |
759 | file->f_pos = offset; | |
760 | ret = file->f_pos; | |
761 | force_successful_syscall_return(); | |
762 | break; | |
763 | default: | |
764 | ret = -EINVAL; | |
1da177e4 | 765 | } |
5955102c | 766 | inode_unlock(file_inode(file)); |
1da177e4 LT |
767 | return ret; |
768 | } | |
769 | ||
890537b3 | 770 | static int open_port(struct inode *inode, struct file *filp) |
1da177e4 LT |
771 | { |
772 | return capable(CAP_SYS_RAWIO) ? 0 : -EPERM; | |
773 | } | |
774 | ||
775 | #define zero_lseek null_lseek | |
776 | #define full_lseek null_lseek | |
777 | #define write_zero write_null | |
cd28e28d | 778 | #define write_iter_zero write_iter_null |
1da177e4 LT |
779 | #define open_mem open_port |
780 | #define open_kmem open_mem | |
781 | ||
73f0718e | 782 | static const struct file_operations __maybe_unused mem_fops = { |
1da177e4 LT |
783 | .llseek = memory_lseek, |
784 | .read = read_mem, | |
785 | .write = write_mem, | |
786 | .mmap = mmap_mem, | |
787 | .open = open_mem, | |
b4caecd4 | 788 | #ifndef CONFIG_MMU |
5da6185b | 789 | .get_unmapped_area = get_unmapped_area_mem, |
b4caecd4 CH |
790 | .mmap_capabilities = memory_mmap_capabilities, |
791 | #endif | |
1da177e4 LT |
792 | }; |
793 | ||
a8c91252 | 794 | static const struct file_operations __maybe_unused kmem_fops = { |
1da177e4 LT |
795 | .llseek = memory_lseek, |
796 | .read = read_kmem, | |
797 | .write = write_kmem, | |
798 | .mmap = mmap_kmem, | |
799 | .open = open_kmem, | |
b4caecd4 | 800 | #ifndef CONFIG_MMU |
5da6185b | 801 | .get_unmapped_area = get_unmapped_area_mem, |
b4caecd4 CH |
802 | .mmap_capabilities = memory_mmap_capabilities, |
803 | #endif | |
1da177e4 LT |
804 | }; |
805 | ||
62322d25 | 806 | static const struct file_operations null_fops = { |
1da177e4 LT |
807 | .llseek = null_lseek, |
808 | .read = read_null, | |
809 | .write = write_null, | |
cd28e28d AV |
810 | .read_iter = read_iter_null, |
811 | .write_iter = write_iter_null, | |
1ebd32fc | 812 | .splice_write = splice_write_null, |
1da177e4 LT |
813 | }; |
814 | ||
3a4bc2fb | 815 | static const struct file_operations __maybe_unused port_fops = { |
1da177e4 LT |
816 | .llseek = memory_lseek, |
817 | .read = read_port, | |
818 | .write = write_port, | |
819 | .open = open_port, | |
820 | }; | |
1da177e4 | 821 | |
62322d25 | 822 | static const struct file_operations zero_fops = { |
1da177e4 | 823 | .llseek = zero_lseek, |
1da177e4 | 824 | .write = write_zero, |
13ba33e8 | 825 | .read_iter = read_iter_zero, |
cd28e28d | 826 | .write_iter = write_iter_zero, |
1da177e4 | 827 | .mmap = mmap_zero, |
c01d5b30 | 828 | .get_unmapped_area = get_unmapped_area_zero, |
b4caecd4 CH |
829 | #ifndef CONFIG_MMU |
830 | .mmap_capabilities = zero_mmap_capabilities, | |
831 | #endif | |
1da177e4 LT |
832 | }; |
833 | ||
62322d25 | 834 | static const struct file_operations full_fops = { |
1da177e4 | 835 | .llseek = full_lseek, |
13ba33e8 | 836 | .read_iter = read_iter_zero, |
1da177e4 LT |
837 | .write = write_full, |
838 | }; | |
839 | ||
389e0cb9 KS |
840 | static const struct memdev { |
841 | const char *name; | |
2c9ede55 | 842 | umode_t mode; |
389e0cb9 | 843 | const struct file_operations *fops; |
b4caecd4 | 844 | fmode_t fmode; |
389e0cb9 | 845 | } devlist[] = { |
73f0718e | 846 | #ifdef CONFIG_DEVMEM |
b4caecd4 | 847 | [1] = { "mem", 0, &mem_fops, FMODE_UNSIGNED_OFFSET }, |
73f0718e | 848 | #endif |
b781ecb6 | 849 | #ifdef CONFIG_DEVKMEM |
b4caecd4 | 850 | [2] = { "kmem", 0, &kmem_fops, FMODE_UNSIGNED_OFFSET }, |
b781ecb6 | 851 | #endif |
b4caecd4 | 852 | [3] = { "null", 0666, &null_fops, 0 }, |
4f911d64 | 853 | #ifdef CONFIG_DEVPORT |
b4caecd4 | 854 | [4] = { "port", 0, &port_fops, 0 }, |
1da177e4 | 855 | #endif |
b4caecd4 CH |
856 | [5] = { "zero", 0666, &zero_fops, 0 }, |
857 | [7] = { "full", 0666, &full_fops, 0 }, | |
858 | [8] = { "random", 0666, &random_fops, 0 }, | |
859 | [9] = { "urandom", 0666, &urandom_fops, 0 }, | |
7f3a781d | 860 | #ifdef CONFIG_PRINTK |
b4caecd4 | 861 | [11] = { "kmsg", 0644, &kmsg_fops, 0 }, |
7f3a781d | 862 | #endif |
d6f47bef ASF |
863 | }; |
864 | ||
865 | static int memory_open(struct inode *inode, struct file *filp) | |
866 | { | |
389e0cb9 KS |
867 | int minor; |
868 | const struct memdev *dev; | |
d6f47bef | 869 | |
389e0cb9 KS |
870 | minor = iminor(inode); |
871 | if (minor >= ARRAY_SIZE(devlist)) | |
205153aa | 872 | return -ENXIO; |
d6f47bef | 873 | |
389e0cb9 KS |
874 | dev = &devlist[minor]; |
875 | if (!dev->fops) | |
205153aa | 876 | return -ENXIO; |
d6f47bef | 877 | |
389e0cb9 | 878 | filp->f_op = dev->fops; |
b4caecd4 | 879 | filp->f_mode |= dev->fmode; |
4a3956c7 | 880 | |
389e0cb9 | 881 | if (dev->fops->open) |
205153aa FW |
882 | return dev->fops->open(inode, filp); |
883 | ||
884 | return 0; | |
1da177e4 LT |
885 | } |
886 | ||
62322d25 | 887 | static const struct file_operations memory_fops = { |
d7d4d849 | 888 | .open = memory_open, |
6038f373 | 889 | .llseek = noop_llseek, |
1da177e4 LT |
890 | }; |
891 | ||
2c9ede55 | 892 | static char *mem_devnode(struct device *dev, umode_t *mode) |
e454cea2 KS |
893 | { |
894 | if (mode && devlist[MINOR(dev->devt)].mode) | |
895 | *mode = devlist[MINOR(dev->devt)].mode; | |
896 | return NULL; | |
897 | } | |
898 | ||
ca8eca68 | 899 | static struct class *mem_class; |
1da177e4 LT |
900 | |
901 | static int __init chr_dev_init(void) | |
902 | { | |
389e0cb9 | 903 | int minor; |
1da177e4 | 904 | |
d7d4d849 | 905 | if (register_chrdev(MEM_MAJOR, "mem", &memory_fops)) |
1da177e4 LT |
906 | printk("unable to get major %d for memory devs\n", MEM_MAJOR); |
907 | ||
ca8eca68 | 908 | mem_class = class_create(THIS_MODULE, "mem"); |
6e191f7b AB |
909 | if (IS_ERR(mem_class)) |
910 | return PTR_ERR(mem_class); | |
911 | ||
e454cea2 | 912 | mem_class->devnode = mem_devnode; |
389e0cb9 KS |
913 | for (minor = 1; minor < ARRAY_SIZE(devlist); minor++) { |
914 | if (!devlist[minor].name) | |
915 | continue; | |
e1612de9 HM |
916 | |
917 | /* | |
890537b3 | 918 | * Create /dev/port? |
e1612de9 HM |
919 | */ |
920 | if ((minor == DEVPORT_MINOR) && !arch_has_dev_port()) | |
921 | continue; | |
922 | ||
389e0cb9 KS |
923 | device_create(mem_class, NULL, MKDEV(MEM_MAJOR, minor), |
924 | NULL, devlist[minor].name); | |
925 | } | |
ebf644c4 | 926 | |
31d1d48e | 927 | return tty_init(); |
1da177e4 LT |
928 | } |
929 | ||
930 | fs_initcall(chr_dev_init); |