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 VG |
23 | #include <linux/highmem.h> |
24 | #include <linux/crash_dump.h> | |
1da177e4 | 25 | #include <linux/backing-dev.h> |
315c215c | 26 | #include <linux/bootmem.h> |
d6b29d7c | 27 | #include <linux/splice.h> |
b8a3ad5b | 28 | #include <linux/pfn.h> |
1da177e4 LT |
29 | |
30 | #include <asm/uaccess.h> | |
31 | #include <asm/io.h> | |
32 | ||
33 | #ifdef CONFIG_IA64 | |
34 | # include <linux/efi.h> | |
35 | #endif | |
36 | ||
f222318e WF |
37 | static inline unsigned long size_inside_page(unsigned long start, |
38 | unsigned long size) | |
39 | { | |
40 | unsigned long sz; | |
41 | ||
7fabaddd | 42 | sz = PAGE_SIZE - (start & (PAGE_SIZE - 1)); |
f222318e | 43 | |
7fabaddd | 44 | return min(sz, size); |
f222318e WF |
45 | } |
46 | ||
1da177e4 | 47 | #ifndef ARCH_HAS_VALID_PHYS_ADDR_RANGE |
136939a2 | 48 | static inline int valid_phys_addr_range(unsigned long addr, size_t count) |
1da177e4 | 49 | { |
cfaf346c | 50 | return addr + count <= __pa(high_memory); |
1da177e4 | 51 | } |
80851ef2 | 52 | |
06c67bef | 53 | static inline int valid_mmap_phys_addr_range(unsigned long pfn, size_t size) |
80851ef2 BH |
54 | { |
55 | return 1; | |
56 | } | |
1da177e4 LT |
57 | #endif |
58 | ||
d092633b | 59 | #ifdef CONFIG_STRICT_DEVMEM |
e2beb3ea | 60 | static inline int range_is_allowed(unsigned long pfn, unsigned long size) |
ae531c26 | 61 | { |
e2beb3ea VP |
62 | u64 from = ((u64)pfn) << PAGE_SHIFT; |
63 | u64 to = from + size; | |
64 | u64 cursor = from; | |
65 | ||
66 | while (cursor < to) { | |
67 | if (!devmem_is_allowed(pfn)) { | |
68 | printk(KERN_INFO | |
69 | "Program %s tried to access /dev/mem between %Lx->%Lx.\n", | |
ae531c26 AV |
70 | current->comm, from, to); |
71 | return 0; | |
72 | } | |
e2beb3ea VP |
73 | cursor += PAGE_SIZE; |
74 | pfn++; | |
ae531c26 AV |
75 | } |
76 | return 1; | |
77 | } | |
78 | #else | |
e2beb3ea | 79 | static inline int range_is_allowed(unsigned long pfn, unsigned long size) |
ae531c26 AV |
80 | { |
81 | return 1; | |
82 | } | |
83 | #endif | |
84 | ||
d7d4d849 | 85 | void __weak unxlate_dev_mem_ptr(unsigned long phys, void *addr) |
e045fb2a | 86 | { |
87 | } | |
88 | ||
1da177e4 | 89 | /* |
d7d4d849 AM |
90 | * This funcion reads the *physical* memory. The f_pos points directly to the |
91 | * memory location. | |
1da177e4 | 92 | */ |
d7d4d849 | 93 | static ssize_t read_mem(struct file *file, char __user *buf, |
1da177e4 LT |
94 | size_t count, loff_t *ppos) |
95 | { | |
96 | unsigned long p = *ppos; | |
97 | ssize_t read, sz; | |
98 | char *ptr; | |
99 | ||
136939a2 | 100 | if (!valid_phys_addr_range(p, count)) |
1da177e4 LT |
101 | return -EFAULT; |
102 | read = 0; | |
103 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
104 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
105 | if (p < PAGE_SIZE) { | |
7fabaddd | 106 | sz = size_inside_page(p, count); |
1da177e4 LT |
107 | if (sz > 0) { |
108 | if (clear_user(buf, sz)) | |
109 | return -EFAULT; | |
d7d4d849 AM |
110 | buf += sz; |
111 | p += sz; | |
112 | count -= sz; | |
113 | read += sz; | |
1da177e4 LT |
114 | } |
115 | } | |
116 | #endif | |
117 | ||
118 | while (count > 0) { | |
fa29e97b WF |
119 | unsigned long remaining; |
120 | ||
f222318e | 121 | sz = size_inside_page(p, count); |
1da177e4 | 122 | |
e045fb2a | 123 | if (!range_is_allowed(p >> PAGE_SHIFT, count)) |
124 | return -EPERM; | |
125 | ||
1da177e4 | 126 | /* |
d7d4d849 AM |
127 | * On ia64 if a page has been mapped somewhere as uncached, then |
128 | * it must also be accessed uncached by the kernel or data | |
129 | * corruption may occur. | |
1da177e4 LT |
130 | */ |
131 | ptr = xlate_dev_mem_ptr(p); | |
e045fb2a | 132 | if (!ptr) |
133 | return -EFAULT; | |
1da177e4 | 134 | |
fa29e97b | 135 | remaining = copy_to_user(buf, ptr, sz); |
e045fb2a | 136 | unxlate_dev_mem_ptr(p, ptr); |
fa29e97b WF |
137 | if (remaining) |
138 | return -EFAULT; | |
e045fb2a | 139 | |
1da177e4 LT |
140 | buf += sz; |
141 | p += sz; | |
142 | count -= sz; | |
143 | read += sz; | |
144 | } | |
145 | ||
146 | *ppos += read; | |
147 | return read; | |
148 | } | |
149 | ||
d7d4d849 | 150 | static ssize_t write_mem(struct file *file, const char __user *buf, |
1da177e4 LT |
151 | size_t count, loff_t *ppos) |
152 | { | |
153 | unsigned long p = *ppos; | |
154 | ssize_t written, sz; | |
155 | unsigned long copied; | |
156 | void *ptr; | |
157 | ||
136939a2 | 158 | if (!valid_phys_addr_range(p, count)) |
1da177e4 LT |
159 | return -EFAULT; |
160 | ||
161 | written = 0; | |
162 | ||
163 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
164 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
165 | if (p < PAGE_SIZE) { | |
7fabaddd | 166 | sz = size_inside_page(p, count); |
1da177e4 LT |
167 | /* Hmm. Do something? */ |
168 | buf += sz; | |
169 | p += sz; | |
170 | count -= sz; | |
171 | written += sz; | |
172 | } | |
173 | #endif | |
174 | ||
175 | while (count > 0) { | |
f222318e | 176 | sz = size_inside_page(p, count); |
1da177e4 | 177 | |
e045fb2a | 178 | if (!range_is_allowed(p >> PAGE_SHIFT, sz)) |
179 | return -EPERM; | |
180 | ||
1da177e4 | 181 | /* |
d7d4d849 AM |
182 | * On ia64 if a page has been mapped somewhere as uncached, then |
183 | * it must also be accessed uncached by the kernel or data | |
184 | * corruption may occur. | |
1da177e4 LT |
185 | */ |
186 | ptr = xlate_dev_mem_ptr(p); | |
e045fb2a | 187 | if (!ptr) { |
188 | if (written) | |
189 | break; | |
190 | return -EFAULT; | |
191 | } | |
1da177e4 LT |
192 | |
193 | copied = copy_from_user(ptr, buf, sz); | |
fa29e97b | 194 | unxlate_dev_mem_ptr(p, ptr); |
1da177e4 | 195 | if (copied) { |
c654d60e JB |
196 | written += sz - copied; |
197 | if (written) | |
198 | break; | |
1da177e4 LT |
199 | return -EFAULT; |
200 | } | |
e045fb2a | 201 | |
1da177e4 LT |
202 | buf += sz; |
203 | p += sz; | |
204 | count -= sz; | |
205 | written += sz; | |
206 | } | |
207 | ||
208 | *ppos += written; | |
209 | return written; | |
210 | } | |
211 | ||
d7d4d849 | 212 | int __weak phys_mem_access_prot_allowed(struct file *file, |
f0970c13 | 213 | unsigned long pfn, unsigned long size, pgprot_t *vma_prot) |
214 | { | |
215 | return 1; | |
216 | } | |
217 | ||
44ac8413 | 218 | #ifndef __HAVE_PHYS_MEM_ACCESS_PROT |
d7d4d849 AM |
219 | |
220 | /* | |
221 | * Architectures vary in how they handle caching for addresses | |
222 | * outside of main memory. | |
223 | * | |
224 | */ | |
ea56f411 | 225 | #ifdef pgprot_noncached |
d7d4d849 AM |
226 | static int uncached_access(struct file *file, unsigned long addr) |
227 | { | |
228 | #if defined(CONFIG_IA64) | |
229 | /* | |
230 | * On ia64, we ignore O_DSYNC because we cannot tolerate memory | |
231 | * attribute aliases. | |
232 | */ | |
233 | return !(efi_mem_attributes(addr) & EFI_MEMORY_WB); | |
234 | #elif defined(CONFIG_MIPS) | |
235 | { | |
236 | extern int __uncached_access(struct file *file, | |
237 | unsigned long addr); | |
238 | ||
239 | return __uncached_access(file, addr); | |
240 | } | |
241 | #else | |
242 | /* | |
243 | * Accessing memory above the top the kernel knows about or through a | |
244 | * file pointer | |
245 | * that was marked O_DSYNC will be done non-cached. | |
246 | */ | |
247 | if (file->f_flags & O_DSYNC) | |
248 | return 1; | |
249 | return addr >= __pa(high_memory); | |
250 | #endif | |
251 | } | |
ea56f411 | 252 | #endif |
d7d4d849 | 253 | |
44ac8413 BH |
254 | static pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, |
255 | unsigned long size, pgprot_t vma_prot) | |
256 | { | |
257 | #ifdef pgprot_noncached | |
258 | unsigned long offset = pfn << PAGE_SHIFT; | |
259 | ||
260 | if (uncached_access(file, offset)) | |
261 | return pgprot_noncached(vma_prot); | |
262 | #endif | |
263 | return vma_prot; | |
264 | } | |
265 | #endif | |
266 | ||
5da6185b DH |
267 | #ifndef CONFIG_MMU |
268 | static unsigned long get_unmapped_area_mem(struct file *file, | |
269 | unsigned long addr, | |
270 | unsigned long len, | |
271 | unsigned long pgoff, | |
272 | unsigned long flags) | |
273 | { | |
274 | if (!valid_mmap_phys_addr_range(pgoff, len)) | |
275 | return (unsigned long) -EINVAL; | |
8a93258c | 276 | return pgoff << PAGE_SHIFT; |
5da6185b DH |
277 | } |
278 | ||
279 | /* can't do an in-place private mapping if there's no MMU */ | |
280 | static inline int private_mapping_ok(struct vm_area_struct *vma) | |
281 | { | |
282 | return vma->vm_flags & VM_MAYSHARE; | |
283 | } | |
284 | #else | |
285 | #define get_unmapped_area_mem NULL | |
286 | ||
287 | static inline int private_mapping_ok(struct vm_area_struct *vma) | |
288 | { | |
289 | return 1; | |
290 | } | |
291 | #endif | |
292 | ||
f0f37e2f | 293 | static const struct vm_operations_struct mmap_mem_ops = { |
7ae8ed50 RR |
294 | #ifdef CONFIG_HAVE_IOREMAP_PROT |
295 | .access = generic_access_phys | |
296 | #endif | |
e7f260a2 | 297 | }; |
298 | ||
d7d4d849 | 299 | static int mmap_mem(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 300 | { |
80851ef2 BH |
301 | size_t size = vma->vm_end - vma->vm_start; |
302 | ||
06c67bef | 303 | if (!valid_mmap_phys_addr_range(vma->vm_pgoff, size)) |
80851ef2 BH |
304 | return -EINVAL; |
305 | ||
5da6185b DH |
306 | if (!private_mapping_ok(vma)) |
307 | return -ENOSYS; | |
308 | ||
e2beb3ea VP |
309 | if (!range_is_allowed(vma->vm_pgoff, size)) |
310 | return -EPERM; | |
311 | ||
f0970c13 | 312 | if (!phys_mem_access_prot_allowed(file, vma->vm_pgoff, size, |
313 | &vma->vm_page_prot)) | |
314 | return -EINVAL; | |
315 | ||
8b150478 | 316 | vma->vm_page_prot = phys_mem_access_prot(file, vma->vm_pgoff, |
80851ef2 | 317 | size, |
1da177e4 | 318 | vma->vm_page_prot); |
1da177e4 | 319 | |
e7f260a2 | 320 | vma->vm_ops = &mmap_mem_ops; |
321 | ||
1da177e4 LT |
322 | /* Remap-pfn-range will mark the range VM_IO and VM_RESERVED */ |
323 | if (remap_pfn_range(vma, | |
324 | vma->vm_start, | |
325 | vma->vm_pgoff, | |
80851ef2 | 326 | size, |
e7f260a2 | 327 | vma->vm_page_prot)) { |
1da177e4 | 328 | return -EAGAIN; |
e7f260a2 | 329 | } |
1da177e4 LT |
330 | return 0; |
331 | } | |
332 | ||
b781ecb6 | 333 | #ifdef CONFIG_DEVKMEM |
d7d4d849 | 334 | static int mmap_kmem(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 335 | { |
4bb82551 LT |
336 | unsigned long pfn; |
337 | ||
6d3154cc LT |
338 | /* Turn a kernel-virtual address into a physical page frame */ |
339 | pfn = __pa((u64)vma->vm_pgoff << PAGE_SHIFT) >> PAGE_SHIFT; | |
4bb82551 | 340 | |
1da177e4 | 341 | /* |
d7d4d849 AM |
342 | * RED-PEN: on some architectures there is more mapped memory than |
343 | * available in mem_map which pfn_valid checks for. Perhaps should add a | |
344 | * new macro here. | |
1da177e4 LT |
345 | * |
346 | * RED-PEN: vmalloc is not supported right now. | |
347 | */ | |
4bb82551 | 348 | if (!pfn_valid(pfn)) |
1da177e4 | 349 | return -EIO; |
4bb82551 LT |
350 | |
351 | vma->vm_pgoff = pfn; | |
1da177e4 LT |
352 | return mmap_mem(file, vma); |
353 | } | |
b781ecb6 | 354 | #endif |
1da177e4 | 355 | |
50b1fdbd VG |
356 | #ifdef CONFIG_CRASH_DUMP |
357 | /* | |
358 | * Read memory corresponding to the old kernel. | |
50b1fdbd | 359 | */ |
315c215c | 360 | static ssize_t read_oldmem(struct file *file, char __user *buf, |
50b1fdbd VG |
361 | size_t count, loff_t *ppos) |
362 | { | |
315c215c VG |
363 | unsigned long pfn, offset; |
364 | size_t read = 0, csize; | |
365 | int rc = 0; | |
50b1fdbd | 366 | |
72414d3f | 367 | while (count) { |
50b1fdbd | 368 | pfn = *ppos / PAGE_SIZE; |
315c215c VG |
369 | if (pfn > saved_max_pfn) |
370 | return read; | |
50b1fdbd | 371 | |
315c215c VG |
372 | offset = (unsigned long)(*ppos % PAGE_SIZE); |
373 | if (count > PAGE_SIZE - offset) | |
374 | csize = PAGE_SIZE - offset; | |
375 | else | |
376 | csize = count; | |
50b1fdbd | 377 | |
315c215c VG |
378 | rc = copy_oldmem_page(pfn, buf, csize, offset, 1); |
379 | if (rc < 0) | |
380 | return rc; | |
50b1fdbd VG |
381 | buf += csize; |
382 | *ppos += csize; | |
383 | read += csize; | |
384 | count -= csize; | |
385 | } | |
50b1fdbd VG |
386 | return read; |
387 | } | |
388 | #endif | |
1da177e4 | 389 | |
b781ecb6 | 390 | #ifdef CONFIG_DEVKMEM |
1da177e4 LT |
391 | /* |
392 | * This function reads the *virtual* memory as seen by the kernel. | |
393 | */ | |
d7d4d849 | 394 | static ssize_t read_kmem(struct file *file, char __user *buf, |
1da177e4 LT |
395 | size_t count, loff_t *ppos) |
396 | { | |
397 | unsigned long p = *ppos; | |
398 | ssize_t low_count, read, sz; | |
399 | char * kbuf; /* k-addr because vread() takes vmlist_lock rwlock */ | |
325fda71 | 400 | int err = 0; |
1da177e4 LT |
401 | |
402 | read = 0; | |
403 | if (p < (unsigned long) high_memory) { | |
404 | low_count = count; | |
d7d4d849 AM |
405 | if (count > (unsigned long)high_memory - p) |
406 | low_count = (unsigned long)high_memory - p; | |
1da177e4 LT |
407 | |
408 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
409 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
410 | if (p < PAGE_SIZE && low_count > 0) { | |
7fabaddd WF |
411 | sz = size_inside_page(p, low_count); |
412 | if (clear_user(buf, sz)) | |
1da177e4 | 413 | return -EFAULT; |
7fabaddd WF |
414 | buf += sz; |
415 | p += sz; | |
416 | read += sz; | |
417 | low_count -= sz; | |
418 | count -= sz; | |
1da177e4 LT |
419 | } |
420 | #endif | |
421 | while (low_count > 0) { | |
f222318e | 422 | sz = size_inside_page(p, low_count); |
1da177e4 LT |
423 | |
424 | /* | |
425 | * On ia64 if a page has been mapped somewhere as | |
426 | * uncached, then it must also be accessed uncached | |
427 | * by the kernel or data corruption may occur | |
428 | */ | |
429 | kbuf = xlate_dev_kmem_ptr((char *)p); | |
430 | ||
431 | if (copy_to_user(buf, kbuf, sz)) | |
432 | return -EFAULT; | |
433 | buf += sz; | |
434 | p += sz; | |
435 | read += sz; | |
436 | low_count -= sz; | |
437 | count -= sz; | |
438 | } | |
439 | } | |
440 | ||
441 | if (count > 0) { | |
442 | kbuf = (char *)__get_free_page(GFP_KERNEL); | |
443 | if (!kbuf) | |
444 | return -ENOMEM; | |
445 | while (count > 0) { | |
80ad89a0 | 446 | sz = size_inside_page(p, count); |
325fda71 KH |
447 | if (!is_vmalloc_or_module_addr((void *)p)) { |
448 | err = -ENXIO; | |
449 | break; | |
450 | } | |
80ad89a0 WF |
451 | sz = vread(kbuf, (char *)p, sz); |
452 | if (!sz) | |
1da177e4 | 453 | break; |
80ad89a0 | 454 | if (copy_to_user(buf, kbuf, sz)) { |
325fda71 KH |
455 | err = -EFAULT; |
456 | break; | |
1da177e4 | 457 | } |
80ad89a0 WF |
458 | count -= sz; |
459 | buf += sz; | |
460 | read += sz; | |
461 | p += sz; | |
1da177e4 LT |
462 | } |
463 | free_page((unsigned long)kbuf); | |
464 | } | |
325fda71 KH |
465 | *ppos = p; |
466 | return read ? read : err; | |
1da177e4 LT |
467 | } |
468 | ||
469 | ||
d7d4d849 AM |
470 | static ssize_t do_write_kmem(unsigned long p, const char __user *buf, |
471 | size_t count, loff_t *ppos) | |
1da177e4 LT |
472 | { |
473 | ssize_t written, sz; | |
474 | unsigned long copied; | |
475 | ||
476 | written = 0; | |
477 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
478 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
ee32398f WF |
479 | if (p < PAGE_SIZE) { |
480 | sz = size_inside_page(p, count); | |
1da177e4 LT |
481 | /* Hmm. Do something? */ |
482 | buf += sz; | |
483 | p += sz; | |
1da177e4 LT |
484 | count -= sz; |
485 | written += sz; | |
486 | } | |
487 | #endif | |
488 | ||
489 | while (count > 0) { | |
490 | char *ptr; | |
1da177e4 | 491 | |
ee32398f | 492 | sz = size_inside_page(p, count); |
1da177e4 LT |
493 | |
494 | /* | |
d7d4d849 AM |
495 | * On ia64 if a page has been mapped somewhere as uncached, then |
496 | * it must also be accessed uncached by the kernel or data | |
497 | * corruption may occur. | |
1da177e4 | 498 | */ |
ee32398f | 499 | ptr = xlate_dev_kmem_ptr((char *)p); |
1da177e4 LT |
500 | |
501 | copied = copy_from_user(ptr, buf, sz); | |
502 | if (copied) { | |
c654d60e JB |
503 | written += sz - copied; |
504 | if (written) | |
505 | break; | |
1da177e4 LT |
506 | return -EFAULT; |
507 | } | |
508 | buf += sz; | |
509 | p += sz; | |
1da177e4 LT |
510 | count -= sz; |
511 | written += sz; | |
512 | } | |
513 | ||
514 | *ppos += written; | |
515 | return written; | |
516 | } | |
517 | ||
1da177e4 LT |
518 | /* |
519 | * This function writes to the *virtual* memory as seen by the kernel. | |
520 | */ | |
d7d4d849 | 521 | static ssize_t write_kmem(struct file *file, const char __user *buf, |
1da177e4 LT |
522 | size_t count, loff_t *ppos) |
523 | { | |
524 | unsigned long p = *ppos; | |
525 | ssize_t wrote = 0; | |
526 | ssize_t virtr = 0; | |
1da177e4 | 527 | char * kbuf; /* k-addr because vwrite() takes vmlist_lock rwlock */ |
325fda71 | 528 | int err = 0; |
1da177e4 LT |
529 | |
530 | if (p < (unsigned long) high_memory) { | |
80ad89a0 WF |
531 | unsigned long to_write = min_t(unsigned long, count, |
532 | (unsigned long)high_memory - p); | |
ee32398f | 533 | wrote = do_write_kmem(p, buf, to_write, ppos); |
80ad89a0 WF |
534 | if (wrote != to_write) |
535 | return wrote; | |
1da177e4 LT |
536 | p += wrote; |
537 | buf += wrote; | |
538 | count -= wrote; | |
539 | } | |
540 | ||
541 | if (count > 0) { | |
542 | kbuf = (char *)__get_free_page(GFP_KERNEL); | |
543 | if (!kbuf) | |
544 | return wrote ? wrote : -ENOMEM; | |
545 | while (count > 0) { | |
80ad89a0 WF |
546 | unsigned long sz = size_inside_page(p, count); |
547 | unsigned long n; | |
1da177e4 | 548 | |
325fda71 KH |
549 | if (!is_vmalloc_or_module_addr((void *)p)) { |
550 | err = -ENXIO; | |
551 | break; | |
552 | } | |
80ad89a0 WF |
553 | n = copy_from_user(kbuf, buf, sz); |
554 | if (n) { | |
325fda71 KH |
555 | err = -EFAULT; |
556 | break; | |
1da177e4 | 557 | } |
c85e9a97 | 558 | vwrite(kbuf, (char *)p, sz); |
80ad89a0 WF |
559 | count -= sz; |
560 | buf += sz; | |
561 | virtr += sz; | |
562 | p += sz; | |
1da177e4 LT |
563 | } |
564 | free_page((unsigned long)kbuf); | |
565 | } | |
566 | ||
325fda71 KH |
567 | *ppos = p; |
568 | return virtr + wrote ? : err; | |
1da177e4 | 569 | } |
b781ecb6 | 570 | #endif |
1da177e4 | 571 | |
4f911d64 | 572 | #ifdef CONFIG_DEVPORT |
d7d4d849 | 573 | static ssize_t read_port(struct file *file, char __user *buf, |
1da177e4 LT |
574 | size_t count, loff_t *ppos) |
575 | { | |
576 | unsigned long i = *ppos; | |
577 | char __user *tmp = buf; | |
578 | ||
579 | if (!access_ok(VERIFY_WRITE, buf, count)) | |
d7d4d849 | 580 | return -EFAULT; |
1da177e4 | 581 | while (count-- > 0 && i < 65536) { |
d7d4d849 AM |
582 | if (__put_user(inb(i), tmp) < 0) |
583 | return -EFAULT; | |
1da177e4 LT |
584 | i++; |
585 | tmp++; | |
586 | } | |
587 | *ppos = i; | |
588 | return tmp-buf; | |
589 | } | |
590 | ||
d7d4d849 | 591 | static ssize_t write_port(struct file *file, const char __user *buf, |
1da177e4 LT |
592 | size_t count, loff_t *ppos) |
593 | { | |
594 | unsigned long i = *ppos; | |
595 | const char __user * tmp = buf; | |
596 | ||
d7d4d849 | 597 | if (!access_ok(VERIFY_READ, buf, count)) |
1da177e4 LT |
598 | return -EFAULT; |
599 | while (count-- > 0 && i < 65536) { | |
600 | char c; | |
c654d60e JB |
601 | if (__get_user(c, tmp)) { |
602 | if (tmp > buf) | |
603 | break; | |
d7d4d849 | 604 | return -EFAULT; |
c654d60e | 605 | } |
d7d4d849 | 606 | outb(c, i); |
1da177e4 LT |
607 | i++; |
608 | tmp++; | |
609 | } | |
610 | *ppos = i; | |
611 | return tmp-buf; | |
612 | } | |
613 | #endif | |
614 | ||
d7d4d849 | 615 | static ssize_t read_null(struct file *file, char __user *buf, |
1da177e4 LT |
616 | size_t count, loff_t *ppos) |
617 | { | |
618 | return 0; | |
619 | } | |
620 | ||
d7d4d849 | 621 | static ssize_t write_null(struct file *file, const char __user *buf, |
1da177e4 LT |
622 | size_t count, loff_t *ppos) |
623 | { | |
624 | return count; | |
625 | } | |
626 | ||
1ebd32fc JA |
627 | static int pipe_to_null(struct pipe_inode_info *info, struct pipe_buffer *buf, |
628 | struct splice_desc *sd) | |
629 | { | |
630 | return sd->len; | |
631 | } | |
632 | ||
d7d4d849 | 633 | static ssize_t splice_write_null(struct pipe_inode_info *pipe, struct file *out, |
1ebd32fc JA |
634 | loff_t *ppos, size_t len, unsigned int flags) |
635 | { | |
636 | return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_null); | |
637 | } | |
638 | ||
d7d4d849 | 639 | static ssize_t read_zero(struct file *file, char __user *buf, |
1da177e4 LT |
640 | size_t count, loff_t *ppos) |
641 | { | |
557ed1fa | 642 | size_t written; |
1da177e4 LT |
643 | |
644 | if (!count) | |
645 | return 0; | |
646 | ||
647 | if (!access_ok(VERIFY_WRITE, buf, count)) | |
648 | return -EFAULT; | |
649 | ||
557ed1fa NP |
650 | written = 0; |
651 | while (count) { | |
652 | unsigned long unwritten; | |
653 | size_t chunk = count; | |
1da177e4 | 654 | |
557ed1fa NP |
655 | if (chunk > PAGE_SIZE) |
656 | chunk = PAGE_SIZE; /* Just for latency reasons */ | |
bb521c5d | 657 | unwritten = __clear_user(buf, chunk); |
557ed1fa | 658 | written += chunk - unwritten; |
1da177e4 | 659 | if (unwritten) |
557ed1fa | 660 | break; |
2b838687 LT |
661 | if (signal_pending(current)) |
662 | return written ? written : -ERESTARTSYS; | |
1da177e4 | 663 | buf += chunk; |
557ed1fa | 664 | count -= chunk; |
1da177e4 LT |
665 | cond_resched(); |
666 | } | |
557ed1fa | 667 | return written ? written : -EFAULT; |
1da177e4 LT |
668 | } |
669 | ||
d7d4d849 | 670 | static int mmap_zero(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 671 | { |
557ed1fa | 672 | #ifndef CONFIG_MMU |
1da177e4 | 673 | return -ENOSYS; |
557ed1fa NP |
674 | #endif |
675 | if (vma->vm_flags & VM_SHARED) | |
676 | return shmem_zero_setup(vma); | |
677 | return 0; | |
1da177e4 | 678 | } |
1da177e4 | 679 | |
d7d4d849 | 680 | static ssize_t write_full(struct file *file, const char __user *buf, |
1da177e4 LT |
681 | size_t count, loff_t *ppos) |
682 | { | |
683 | return -ENOSPC; | |
684 | } | |
685 | ||
686 | /* | |
687 | * Special lseek() function for /dev/null and /dev/zero. Most notably, you | |
688 | * can fopen() both devices with "a" now. This was previously impossible. | |
689 | * -- SRB. | |
690 | */ | |
d7d4d849 | 691 | static loff_t null_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
692 | { |
693 | return file->f_pos = 0; | |
694 | } | |
695 | ||
696 | /* | |
697 | * The memory devices use the full 32/64 bits of the offset, and so we cannot | |
698 | * check against negative addresses: they are ok. The return value is weird, | |
699 | * though, in that case (0). | |
700 | * | |
701 | * also note that seeking relative to the "end of file" isn't supported: | |
702 | * it has no meaning, so it returns -EINVAL. | |
703 | */ | |
d7d4d849 | 704 | static loff_t memory_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
705 | { |
706 | loff_t ret; | |
707 | ||
a7113a96 | 708 | mutex_lock(&file->f_path.dentry->d_inode->i_mutex); |
1da177e4 | 709 | switch (orig) { |
d7d4d849 AM |
710 | case SEEK_CUR: |
711 | offset += file->f_pos; | |
d7d4d849 AM |
712 | case SEEK_SET: |
713 | /* to avoid userland mistaking f_pos=-9 as -EBADF=-9 */ | |
714 | if ((unsigned long long)offset >= ~0xFFFULL) { | |
715 | ret = -EOVERFLOW; | |
1da177e4 | 716 | break; |
d7d4d849 AM |
717 | } |
718 | file->f_pos = offset; | |
719 | ret = file->f_pos; | |
720 | force_successful_syscall_return(); | |
721 | break; | |
722 | default: | |
723 | ret = -EINVAL; | |
1da177e4 | 724 | } |
a7113a96 | 725 | mutex_unlock(&file->f_path.dentry->d_inode->i_mutex); |
1da177e4 LT |
726 | return ret; |
727 | } | |
728 | ||
729 | static int open_port(struct inode * inode, struct file * filp) | |
730 | { | |
731 | return capable(CAP_SYS_RAWIO) ? 0 : -EPERM; | |
732 | } | |
733 | ||
734 | #define zero_lseek null_lseek | |
735 | #define full_lseek null_lseek | |
736 | #define write_zero write_null | |
737 | #define read_full read_zero | |
738 | #define open_mem open_port | |
739 | #define open_kmem open_mem | |
50b1fdbd | 740 | #define open_oldmem open_mem |
1da177e4 | 741 | |
62322d25 | 742 | static const struct file_operations mem_fops = { |
1da177e4 LT |
743 | .llseek = memory_lseek, |
744 | .read = read_mem, | |
745 | .write = write_mem, | |
746 | .mmap = mmap_mem, | |
747 | .open = open_mem, | |
5da6185b | 748 | .get_unmapped_area = get_unmapped_area_mem, |
1da177e4 LT |
749 | }; |
750 | ||
b781ecb6 | 751 | #ifdef CONFIG_DEVKMEM |
62322d25 | 752 | static const struct file_operations kmem_fops = { |
1da177e4 LT |
753 | .llseek = memory_lseek, |
754 | .read = read_kmem, | |
755 | .write = write_kmem, | |
756 | .mmap = mmap_kmem, | |
757 | .open = open_kmem, | |
5da6185b | 758 | .get_unmapped_area = get_unmapped_area_mem, |
1da177e4 | 759 | }; |
b781ecb6 | 760 | #endif |
1da177e4 | 761 | |
62322d25 | 762 | static const struct file_operations null_fops = { |
1da177e4 LT |
763 | .llseek = null_lseek, |
764 | .read = read_null, | |
765 | .write = write_null, | |
1ebd32fc | 766 | .splice_write = splice_write_null, |
1da177e4 LT |
767 | }; |
768 | ||
4f911d64 | 769 | #ifdef CONFIG_DEVPORT |
62322d25 | 770 | static const struct file_operations port_fops = { |
1da177e4 LT |
771 | .llseek = memory_lseek, |
772 | .read = read_port, | |
773 | .write = write_port, | |
774 | .open = open_port, | |
775 | }; | |
776 | #endif | |
777 | ||
62322d25 | 778 | static const struct file_operations zero_fops = { |
1da177e4 LT |
779 | .llseek = zero_lseek, |
780 | .read = read_zero, | |
781 | .write = write_zero, | |
782 | .mmap = mmap_zero, | |
783 | }; | |
784 | ||
5da6185b DH |
785 | /* |
786 | * capabilities for /dev/zero | |
787 | * - permits private mappings, "copies" are taken of the source of zeros | |
371d217e | 788 | * - no writeback happens |
5da6185b | 789 | */ |
1da177e4 | 790 | static struct backing_dev_info zero_bdi = { |
d993831f | 791 | .name = "char/mem", |
371d217e | 792 | .capabilities = BDI_CAP_MAP_COPY | BDI_CAP_NO_ACCT_AND_WRITEBACK, |
1da177e4 LT |
793 | }; |
794 | ||
62322d25 | 795 | static const struct file_operations full_fops = { |
1da177e4 LT |
796 | .llseek = full_lseek, |
797 | .read = read_full, | |
798 | .write = write_full, | |
799 | }; | |
800 | ||
50b1fdbd | 801 | #ifdef CONFIG_CRASH_DUMP |
62322d25 | 802 | static const struct file_operations oldmem_fops = { |
50b1fdbd VG |
803 | .read = read_oldmem, |
804 | .open = open_oldmem, | |
6038f373 | 805 | .llseek = default_llseek, |
50b1fdbd VG |
806 | }; |
807 | #endif | |
808 | ||
7e5b58bc KS |
809 | static ssize_t kmsg_writev(struct kiocb *iocb, const struct iovec *iv, |
810 | unsigned long count, loff_t pos) | |
1da177e4 | 811 | { |
7e5b58bc | 812 | char *line, *p; |
70a5f521 | 813 | int i; |
7e5b58bc | 814 | ssize_t ret = -EFAULT; |
70a5f521 | 815 | size_t len = iov_length(iv, count); |
1da177e4 | 816 | |
70a5f521 | 817 | line = kmalloc(len + 1, GFP_KERNEL); |
7e5b58bc | 818 | if (line == NULL) |
1da177e4 | 819 | return -ENOMEM; |
7e5b58bc KS |
820 | |
821 | /* | |
822 | * copy all vectors into a single string, to ensure we do | |
823 | * not interleave our log line with other printk calls | |
824 | */ | |
70a5f521 | 825 | p = line; |
7e5b58bc KS |
826 | for (i = 0; i < count; i++) { |
827 | if (copy_from_user(p, iv[i].iov_base, iv[i].iov_len)) | |
828 | goto out; | |
829 | p += iv[i].iov_len; | |
1da177e4 | 830 | } |
7e5b58bc KS |
831 | p[0] = '\0'; |
832 | ||
833 | ret = printk("%s", line); | |
834 | /* printk can add a prefix */ | |
835 | if (ret > len) | |
836 | ret = len; | |
837 | out: | |
838 | kfree(line); | |
1da177e4 LT |
839 | return ret; |
840 | } | |
841 | ||
62322d25 | 842 | static const struct file_operations kmsg_fops = { |
7e5b58bc | 843 | .aio_write = kmsg_writev, |
6038f373 | 844 | .llseek = noop_llseek, |
1da177e4 LT |
845 | }; |
846 | ||
389e0cb9 KS |
847 | static const struct memdev { |
848 | const char *name; | |
e454cea2 | 849 | mode_t mode; |
389e0cb9 KS |
850 | const struct file_operations *fops; |
851 | struct backing_dev_info *dev_info; | |
852 | } devlist[] = { | |
e454cea2 | 853 | [1] = { "mem", 0, &mem_fops, &directly_mappable_cdev_bdi }, |
b781ecb6 | 854 | #ifdef CONFIG_DEVKMEM |
e454cea2 | 855 | [2] = { "kmem", 0, &kmem_fops, &directly_mappable_cdev_bdi }, |
b781ecb6 | 856 | #endif |
e454cea2 | 857 | [3] = { "null", 0666, &null_fops, NULL }, |
4f911d64 | 858 | #ifdef CONFIG_DEVPORT |
e454cea2 | 859 | [4] = { "port", 0, &port_fops, NULL }, |
1da177e4 | 860 | #endif |
e454cea2 KS |
861 | [5] = { "zero", 0666, &zero_fops, &zero_bdi }, |
862 | [7] = { "full", 0666, &full_fops, NULL }, | |
863 | [8] = { "random", 0666, &random_fops, NULL }, | |
864 | [9] = { "urandom", 0666, &urandom_fops, NULL }, | |
865 | [11] = { "kmsg", 0, &kmsg_fops, NULL }, | |
50b1fdbd | 866 | #ifdef CONFIG_CRASH_DUMP |
e454cea2 | 867 | [12] = { "oldmem", 0, &oldmem_fops, NULL }, |
50b1fdbd | 868 | #endif |
d6f47bef ASF |
869 | }; |
870 | ||
871 | static int memory_open(struct inode *inode, struct file *filp) | |
872 | { | |
389e0cb9 KS |
873 | int minor; |
874 | const struct memdev *dev; | |
d6f47bef | 875 | |
389e0cb9 KS |
876 | minor = iminor(inode); |
877 | if (minor >= ARRAY_SIZE(devlist)) | |
205153aa | 878 | return -ENXIO; |
d6f47bef | 879 | |
389e0cb9 KS |
880 | dev = &devlist[minor]; |
881 | if (!dev->fops) | |
205153aa | 882 | return -ENXIO; |
d6f47bef | 883 | |
389e0cb9 KS |
884 | filp->f_op = dev->fops; |
885 | if (dev->dev_info) | |
886 | filp->f_mapping->backing_dev_info = dev->dev_info; | |
d6f47bef | 887 | |
4a3956c7 KH |
888 | /* Is /dev/mem or /dev/kmem ? */ |
889 | if (dev->dev_info == &directly_mappable_cdev_bdi) | |
890 | filp->f_mode |= FMODE_UNSIGNED_OFFSET; | |
891 | ||
389e0cb9 | 892 | if (dev->fops->open) |
205153aa FW |
893 | return dev->fops->open(inode, filp); |
894 | ||
895 | return 0; | |
1da177e4 LT |
896 | } |
897 | ||
62322d25 | 898 | static const struct file_operations memory_fops = { |
d7d4d849 | 899 | .open = memory_open, |
6038f373 | 900 | .llseek = noop_llseek, |
1da177e4 LT |
901 | }; |
902 | ||
e454cea2 KS |
903 | static char *mem_devnode(struct device *dev, mode_t *mode) |
904 | { | |
905 | if (mode && devlist[MINOR(dev->devt)].mode) | |
906 | *mode = devlist[MINOR(dev->devt)].mode; | |
907 | return NULL; | |
908 | } | |
909 | ||
ca8eca68 | 910 | static struct class *mem_class; |
1da177e4 LT |
911 | |
912 | static int __init chr_dev_init(void) | |
913 | { | |
389e0cb9 | 914 | int minor; |
e0bf68dd PZ |
915 | int err; |
916 | ||
917 | err = bdi_init(&zero_bdi); | |
918 | if (err) | |
919 | return err; | |
1da177e4 | 920 | |
d7d4d849 | 921 | if (register_chrdev(MEM_MAJOR, "mem", &memory_fops)) |
1da177e4 LT |
922 | printk("unable to get major %d for memory devs\n", MEM_MAJOR); |
923 | ||
ca8eca68 | 924 | mem_class = class_create(THIS_MODULE, "mem"); |
6e191f7b AB |
925 | if (IS_ERR(mem_class)) |
926 | return PTR_ERR(mem_class); | |
927 | ||
e454cea2 | 928 | mem_class->devnode = mem_devnode; |
389e0cb9 KS |
929 | for (minor = 1; minor < ARRAY_SIZE(devlist); minor++) { |
930 | if (!devlist[minor].name) | |
931 | continue; | |
932 | device_create(mem_class, NULL, MKDEV(MEM_MAJOR, minor), | |
933 | NULL, devlist[minor].name); | |
934 | } | |
ebf644c4 | 935 | |
31d1d48e | 936 | return tty_init(); |
1da177e4 LT |
937 | } |
938 | ||
939 | fs_initcall(chr_dev_init); |