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> |
66300e66 | 29 | #include <linux/export.h> |
1da177e4 LT |
30 | |
31 | #include <asm/uaccess.h> | |
32 | #include <asm/io.h> | |
33 | ||
34 | #ifdef CONFIG_IA64 | |
35 | # include <linux/efi.h> | |
36 | #endif | |
37 | ||
f222318e WF |
38 | static inline unsigned long size_inside_page(unsigned long start, |
39 | unsigned long size) | |
40 | { | |
41 | unsigned long sz; | |
42 | ||
7fabaddd | 43 | sz = PAGE_SIZE - (start & (PAGE_SIZE - 1)); |
f222318e | 44 | |
7fabaddd | 45 | return min(sz, size); |
f222318e WF |
46 | } |
47 | ||
1da177e4 | 48 | #ifndef ARCH_HAS_VALID_PHYS_ADDR_RANGE |
136939a2 | 49 | static inline int valid_phys_addr_range(unsigned long addr, size_t count) |
1da177e4 | 50 | { |
cfaf346c | 51 | return addr + count <= __pa(high_memory); |
1da177e4 | 52 | } |
80851ef2 | 53 | |
06c67bef | 54 | static inline int valid_mmap_phys_addr_range(unsigned long pfn, size_t size) |
80851ef2 BH |
55 | { |
56 | return 1; | |
57 | } | |
1da177e4 LT |
58 | #endif |
59 | ||
d092633b | 60 | #ifdef CONFIG_STRICT_DEVMEM |
e2beb3ea | 61 | static inline int range_is_allowed(unsigned long pfn, unsigned long size) |
ae531c26 | 62 | { |
e2beb3ea VP |
63 | u64 from = ((u64)pfn) << PAGE_SHIFT; |
64 | u64 to = from + size; | |
65 | u64 cursor = from; | |
66 | ||
67 | while (cursor < to) { | |
68 | if (!devmem_is_allowed(pfn)) { | |
69 | printk(KERN_INFO | |
70 | "Program %s tried to access /dev/mem between %Lx->%Lx.\n", | |
ae531c26 AV |
71 | current->comm, from, to); |
72 | return 0; | |
73 | } | |
e2beb3ea VP |
74 | cursor += PAGE_SIZE; |
75 | pfn++; | |
ae531c26 AV |
76 | } |
77 | return 1; | |
78 | } | |
79 | #else | |
e2beb3ea | 80 | static inline int range_is_allowed(unsigned long pfn, unsigned long size) |
ae531c26 AV |
81 | { |
82 | return 1; | |
83 | } | |
84 | #endif | |
85 | ||
d7d4d849 | 86 | void __weak unxlate_dev_mem_ptr(unsigned long phys, void *addr) |
e045fb2a | 87 | { |
88 | } | |
89 | ||
1da177e4 | 90 | /* |
d7d4d849 AM |
91 | * This funcion reads the *physical* memory. The f_pos points directly to the |
92 | * memory location. | |
1da177e4 | 93 | */ |
d7d4d849 | 94 | static ssize_t read_mem(struct file *file, char __user *buf, |
1da177e4 LT |
95 | size_t count, loff_t *ppos) |
96 | { | |
97 | unsigned long p = *ppos; | |
98 | ssize_t read, sz; | |
99 | char *ptr; | |
100 | ||
136939a2 | 101 | if (!valid_phys_addr_range(p, count)) |
1da177e4 LT |
102 | return -EFAULT; |
103 | read = 0; | |
104 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
105 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
106 | if (p < PAGE_SIZE) { | |
7fabaddd | 107 | sz = size_inside_page(p, count); |
1da177e4 LT |
108 | if (sz > 0) { |
109 | if (clear_user(buf, sz)) | |
110 | return -EFAULT; | |
d7d4d849 AM |
111 | buf += sz; |
112 | p += sz; | |
113 | count -= sz; | |
114 | read += sz; | |
1da177e4 LT |
115 | } |
116 | } | |
117 | #endif | |
118 | ||
119 | while (count > 0) { | |
fa29e97b WF |
120 | unsigned long remaining; |
121 | ||
f222318e | 122 | sz = size_inside_page(p, count); |
1da177e4 | 123 | |
e045fb2a | 124 | if (!range_is_allowed(p >> PAGE_SHIFT, count)) |
125 | return -EPERM; | |
126 | ||
1da177e4 | 127 | /* |
d7d4d849 AM |
128 | * On ia64 if a page has been mapped somewhere as uncached, then |
129 | * it must also be accessed uncached by the kernel or data | |
130 | * corruption may occur. | |
1da177e4 LT |
131 | */ |
132 | ptr = xlate_dev_mem_ptr(p); | |
e045fb2a | 133 | if (!ptr) |
134 | return -EFAULT; | |
1da177e4 | 135 | |
fa29e97b | 136 | remaining = copy_to_user(buf, ptr, sz); |
e045fb2a | 137 | unxlate_dev_mem_ptr(p, ptr); |
fa29e97b WF |
138 | if (remaining) |
139 | return -EFAULT; | |
e045fb2a | 140 | |
1da177e4 LT |
141 | buf += sz; |
142 | p += sz; | |
143 | count -= sz; | |
144 | read += sz; | |
145 | } | |
146 | ||
147 | *ppos += read; | |
148 | return read; | |
149 | } | |
150 | ||
d7d4d849 | 151 | static ssize_t write_mem(struct file *file, const char __user *buf, |
1da177e4 LT |
152 | size_t count, loff_t *ppos) |
153 | { | |
154 | unsigned long p = *ppos; | |
155 | ssize_t written, sz; | |
156 | unsigned long copied; | |
157 | void *ptr; | |
158 | ||
136939a2 | 159 | if (!valid_phys_addr_range(p, count)) |
1da177e4 LT |
160 | return -EFAULT; |
161 | ||
162 | written = 0; | |
163 | ||
164 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
165 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
166 | if (p < PAGE_SIZE) { | |
7fabaddd | 167 | sz = size_inside_page(p, count); |
1da177e4 LT |
168 | /* Hmm. Do something? */ |
169 | buf += sz; | |
170 | p += sz; | |
171 | count -= sz; | |
172 | written += sz; | |
173 | } | |
174 | #endif | |
175 | ||
176 | while (count > 0) { | |
f222318e | 177 | sz = size_inside_page(p, count); |
1da177e4 | 178 | |
e045fb2a | 179 | if (!range_is_allowed(p >> PAGE_SHIFT, sz)) |
180 | return -EPERM; | |
181 | ||
1da177e4 | 182 | /* |
d7d4d849 AM |
183 | * On ia64 if a page has been mapped somewhere as uncached, then |
184 | * it must also be accessed uncached by the kernel or data | |
185 | * corruption may occur. | |
1da177e4 LT |
186 | */ |
187 | ptr = xlate_dev_mem_ptr(p); | |
e045fb2a | 188 | if (!ptr) { |
189 | if (written) | |
190 | break; | |
191 | return -EFAULT; | |
192 | } | |
1da177e4 LT |
193 | |
194 | copied = copy_from_user(ptr, buf, sz); | |
fa29e97b | 195 | unxlate_dev_mem_ptr(p, ptr); |
1da177e4 | 196 | if (copied) { |
c654d60e JB |
197 | written += sz - copied; |
198 | if (written) | |
199 | break; | |
1da177e4 LT |
200 | return -EFAULT; |
201 | } | |
e045fb2a | 202 | |
1da177e4 LT |
203 | buf += sz; |
204 | p += sz; | |
205 | count -= sz; | |
206 | written += sz; | |
207 | } | |
208 | ||
209 | *ppos += written; | |
210 | return written; | |
211 | } | |
212 | ||
d7d4d849 | 213 | int __weak phys_mem_access_prot_allowed(struct file *file, |
f0970c13 | 214 | unsigned long pfn, unsigned long size, pgprot_t *vma_prot) |
215 | { | |
216 | return 1; | |
217 | } | |
218 | ||
44ac8413 | 219 | #ifndef __HAVE_PHYS_MEM_ACCESS_PROT |
d7d4d849 AM |
220 | |
221 | /* | |
222 | * Architectures vary in how they handle caching for addresses | |
223 | * outside of main memory. | |
224 | * | |
225 | */ | |
ea56f411 | 226 | #ifdef pgprot_noncached |
d7d4d849 AM |
227 | static int uncached_access(struct file *file, unsigned long addr) |
228 | { | |
229 | #if defined(CONFIG_IA64) | |
230 | /* | |
231 | * On ia64, we ignore O_DSYNC because we cannot tolerate memory | |
232 | * attribute aliases. | |
233 | */ | |
234 | return !(efi_mem_attributes(addr) & EFI_MEMORY_WB); | |
235 | #elif defined(CONFIG_MIPS) | |
236 | { | |
237 | extern int __uncached_access(struct file *file, | |
238 | unsigned long addr); | |
239 | ||
240 | return __uncached_access(file, addr); | |
241 | } | |
242 | #else | |
243 | /* | |
244 | * Accessing memory above the top the kernel knows about or through a | |
245 | * file pointer | |
246 | * that was marked O_DSYNC will be done non-cached. | |
247 | */ | |
248 | if (file->f_flags & O_DSYNC) | |
249 | return 1; | |
250 | return addr >= __pa(high_memory); | |
251 | #endif | |
252 | } | |
ea56f411 | 253 | #endif |
d7d4d849 | 254 | |
44ac8413 BH |
255 | static pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, |
256 | unsigned long size, pgprot_t vma_prot) | |
257 | { | |
258 | #ifdef pgprot_noncached | |
259 | unsigned long offset = pfn << PAGE_SHIFT; | |
260 | ||
261 | if (uncached_access(file, offset)) | |
262 | return pgprot_noncached(vma_prot); | |
263 | #endif | |
264 | return vma_prot; | |
265 | } | |
266 | #endif | |
267 | ||
5da6185b DH |
268 | #ifndef CONFIG_MMU |
269 | static unsigned long get_unmapped_area_mem(struct file *file, | |
270 | unsigned long addr, | |
271 | unsigned long len, | |
272 | unsigned long pgoff, | |
273 | unsigned long flags) | |
274 | { | |
275 | if (!valid_mmap_phys_addr_range(pgoff, len)) | |
276 | return (unsigned long) -EINVAL; | |
8a93258c | 277 | return pgoff << PAGE_SHIFT; |
5da6185b DH |
278 | } |
279 | ||
280 | /* can't do an in-place private mapping if there's no MMU */ | |
281 | static inline int private_mapping_ok(struct vm_area_struct *vma) | |
282 | { | |
283 | return vma->vm_flags & VM_MAYSHARE; | |
284 | } | |
285 | #else | |
286 | #define get_unmapped_area_mem NULL | |
287 | ||
288 | static inline int private_mapping_ok(struct vm_area_struct *vma) | |
289 | { | |
290 | return 1; | |
291 | } | |
292 | #endif | |
293 | ||
f0f37e2f | 294 | static const struct vm_operations_struct mmap_mem_ops = { |
7ae8ed50 RR |
295 | #ifdef CONFIG_HAVE_IOREMAP_PROT |
296 | .access = generic_access_phys | |
297 | #endif | |
e7f260a2 | 298 | }; |
299 | ||
d7d4d849 | 300 | static int mmap_mem(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 301 | { |
80851ef2 BH |
302 | size_t size = vma->vm_end - vma->vm_start; |
303 | ||
06c67bef | 304 | if (!valid_mmap_phys_addr_range(vma->vm_pgoff, size)) |
80851ef2 BH |
305 | return -EINVAL; |
306 | ||
5da6185b DH |
307 | if (!private_mapping_ok(vma)) |
308 | return -ENOSYS; | |
309 | ||
e2beb3ea VP |
310 | if (!range_is_allowed(vma->vm_pgoff, size)) |
311 | return -EPERM; | |
312 | ||
f0970c13 | 313 | if (!phys_mem_access_prot_allowed(file, vma->vm_pgoff, size, |
314 | &vma->vm_page_prot)) | |
315 | return -EINVAL; | |
316 | ||
8b150478 | 317 | vma->vm_page_prot = phys_mem_access_prot(file, vma->vm_pgoff, |
80851ef2 | 318 | size, |
1da177e4 | 319 | vma->vm_page_prot); |
1da177e4 | 320 | |
e7f260a2 | 321 | vma->vm_ops = &mmap_mem_ops; |
322 | ||
1da177e4 LT |
323 | /* Remap-pfn-range will mark the range VM_IO and VM_RESERVED */ |
324 | if (remap_pfn_range(vma, | |
325 | vma->vm_start, | |
326 | vma->vm_pgoff, | |
80851ef2 | 327 | size, |
e7f260a2 | 328 | vma->vm_page_prot)) { |
1da177e4 | 329 | return -EAGAIN; |
e7f260a2 | 330 | } |
1da177e4 LT |
331 | return 0; |
332 | } | |
333 | ||
b781ecb6 | 334 | #ifdef CONFIG_DEVKMEM |
d7d4d849 | 335 | static int mmap_kmem(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 336 | { |
4bb82551 LT |
337 | unsigned long pfn; |
338 | ||
6d3154cc LT |
339 | /* Turn a kernel-virtual address into a physical page frame */ |
340 | pfn = __pa((u64)vma->vm_pgoff << PAGE_SHIFT) >> PAGE_SHIFT; | |
4bb82551 | 341 | |
1da177e4 | 342 | /* |
d7d4d849 AM |
343 | * RED-PEN: on some architectures there is more mapped memory than |
344 | * available in mem_map which pfn_valid checks for. Perhaps should add a | |
345 | * new macro here. | |
1da177e4 LT |
346 | * |
347 | * RED-PEN: vmalloc is not supported right now. | |
348 | */ | |
4bb82551 | 349 | if (!pfn_valid(pfn)) |
1da177e4 | 350 | return -EIO; |
4bb82551 LT |
351 | |
352 | vma->vm_pgoff = pfn; | |
1da177e4 LT |
353 | return mmap_mem(file, vma); |
354 | } | |
b781ecb6 | 355 | #endif |
1da177e4 | 356 | |
50b1fdbd VG |
357 | #ifdef CONFIG_CRASH_DUMP |
358 | /* | |
359 | * Read memory corresponding to the old kernel. | |
50b1fdbd | 360 | */ |
315c215c | 361 | static ssize_t read_oldmem(struct file *file, char __user *buf, |
50b1fdbd VG |
362 | size_t count, loff_t *ppos) |
363 | { | |
315c215c VG |
364 | unsigned long pfn, offset; |
365 | size_t read = 0, csize; | |
366 | int rc = 0; | |
50b1fdbd | 367 | |
72414d3f | 368 | while (count) { |
50b1fdbd | 369 | pfn = *ppos / PAGE_SIZE; |
315c215c VG |
370 | if (pfn > saved_max_pfn) |
371 | return read; | |
50b1fdbd | 372 | |
315c215c VG |
373 | offset = (unsigned long)(*ppos % PAGE_SIZE); |
374 | if (count > PAGE_SIZE - offset) | |
375 | csize = PAGE_SIZE - offset; | |
376 | else | |
377 | csize = count; | |
50b1fdbd | 378 | |
315c215c VG |
379 | rc = copy_oldmem_page(pfn, buf, csize, offset, 1); |
380 | if (rc < 0) | |
381 | return rc; | |
50b1fdbd VG |
382 | buf += csize; |
383 | *ppos += csize; | |
384 | read += csize; | |
385 | count -= csize; | |
386 | } | |
50b1fdbd VG |
387 | return read; |
388 | } | |
389 | #endif | |
1da177e4 | 390 | |
b781ecb6 | 391 | #ifdef CONFIG_DEVKMEM |
1da177e4 LT |
392 | /* |
393 | * This function reads the *virtual* memory as seen by the kernel. | |
394 | */ | |
d7d4d849 | 395 | static ssize_t read_kmem(struct file *file, char __user *buf, |
1da177e4 LT |
396 | size_t count, loff_t *ppos) |
397 | { | |
398 | unsigned long p = *ppos; | |
399 | ssize_t low_count, read, sz; | |
400 | char * kbuf; /* k-addr because vread() takes vmlist_lock rwlock */ | |
325fda71 | 401 | int err = 0; |
1da177e4 LT |
402 | |
403 | read = 0; | |
404 | if (p < (unsigned long) high_memory) { | |
405 | low_count = count; | |
d7d4d849 AM |
406 | if (count > (unsigned long)high_memory - p) |
407 | low_count = (unsigned long)high_memory - p; | |
1da177e4 LT |
408 | |
409 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
410 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
411 | if (p < PAGE_SIZE && low_count > 0) { | |
7fabaddd WF |
412 | sz = size_inside_page(p, low_count); |
413 | if (clear_user(buf, sz)) | |
1da177e4 | 414 | return -EFAULT; |
7fabaddd WF |
415 | buf += sz; |
416 | p += sz; | |
417 | read += sz; | |
418 | low_count -= sz; | |
419 | count -= sz; | |
1da177e4 LT |
420 | } |
421 | #endif | |
422 | while (low_count > 0) { | |
f222318e | 423 | sz = size_inside_page(p, low_count); |
1da177e4 LT |
424 | |
425 | /* | |
426 | * On ia64 if a page has been mapped somewhere as | |
427 | * uncached, then it must also be accessed uncached | |
428 | * by the kernel or data corruption may occur | |
429 | */ | |
430 | kbuf = xlate_dev_kmem_ptr((char *)p); | |
431 | ||
432 | if (copy_to_user(buf, kbuf, sz)) | |
433 | return -EFAULT; | |
434 | buf += sz; | |
435 | p += sz; | |
436 | read += sz; | |
437 | low_count -= sz; | |
438 | count -= sz; | |
439 | } | |
440 | } | |
441 | ||
442 | if (count > 0) { | |
443 | kbuf = (char *)__get_free_page(GFP_KERNEL); | |
444 | if (!kbuf) | |
445 | return -ENOMEM; | |
446 | while (count > 0) { | |
80ad89a0 | 447 | sz = size_inside_page(p, count); |
325fda71 KH |
448 | if (!is_vmalloc_or_module_addr((void *)p)) { |
449 | err = -ENXIO; | |
450 | break; | |
451 | } | |
80ad89a0 WF |
452 | sz = vread(kbuf, (char *)p, sz); |
453 | if (!sz) | |
1da177e4 | 454 | break; |
80ad89a0 | 455 | if (copy_to_user(buf, kbuf, sz)) { |
325fda71 KH |
456 | err = -EFAULT; |
457 | break; | |
1da177e4 | 458 | } |
80ad89a0 WF |
459 | count -= sz; |
460 | buf += sz; | |
461 | read += sz; | |
462 | p += sz; | |
1da177e4 LT |
463 | } |
464 | free_page((unsigned long)kbuf); | |
465 | } | |
325fda71 KH |
466 | *ppos = p; |
467 | return read ? read : err; | |
1da177e4 LT |
468 | } |
469 | ||
470 | ||
d7d4d849 AM |
471 | static ssize_t do_write_kmem(unsigned long p, const char __user *buf, |
472 | size_t count, loff_t *ppos) | |
1da177e4 LT |
473 | { |
474 | ssize_t written, sz; | |
475 | unsigned long copied; | |
476 | ||
477 | written = 0; | |
478 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
479 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
ee32398f WF |
480 | if (p < PAGE_SIZE) { |
481 | sz = size_inside_page(p, count); | |
1da177e4 LT |
482 | /* Hmm. Do something? */ |
483 | buf += sz; | |
484 | p += sz; | |
1da177e4 LT |
485 | count -= sz; |
486 | written += sz; | |
487 | } | |
488 | #endif | |
489 | ||
490 | while (count > 0) { | |
491 | char *ptr; | |
1da177e4 | 492 | |
ee32398f | 493 | sz = size_inside_page(p, count); |
1da177e4 LT |
494 | |
495 | /* | |
d7d4d849 AM |
496 | * On ia64 if a page has been mapped somewhere as uncached, then |
497 | * it must also be accessed uncached by the kernel or data | |
498 | * corruption may occur. | |
1da177e4 | 499 | */ |
ee32398f | 500 | ptr = xlate_dev_kmem_ptr((char *)p); |
1da177e4 LT |
501 | |
502 | copied = copy_from_user(ptr, buf, sz); | |
503 | if (copied) { | |
c654d60e JB |
504 | written += sz - copied; |
505 | if (written) | |
506 | break; | |
1da177e4 LT |
507 | return -EFAULT; |
508 | } | |
509 | buf += sz; | |
510 | p += sz; | |
1da177e4 LT |
511 | count -= sz; |
512 | written += sz; | |
513 | } | |
514 | ||
515 | *ppos += written; | |
516 | return written; | |
517 | } | |
518 | ||
1da177e4 LT |
519 | /* |
520 | * This function writes to the *virtual* memory as seen by the kernel. | |
521 | */ | |
d7d4d849 | 522 | static ssize_t write_kmem(struct file *file, const char __user *buf, |
1da177e4 LT |
523 | size_t count, loff_t *ppos) |
524 | { | |
525 | unsigned long p = *ppos; | |
526 | ssize_t wrote = 0; | |
527 | ssize_t virtr = 0; | |
1da177e4 | 528 | char * kbuf; /* k-addr because vwrite() takes vmlist_lock rwlock */ |
325fda71 | 529 | int err = 0; |
1da177e4 LT |
530 | |
531 | if (p < (unsigned long) high_memory) { | |
80ad89a0 WF |
532 | unsigned long to_write = min_t(unsigned long, count, |
533 | (unsigned long)high_memory - p); | |
ee32398f | 534 | wrote = do_write_kmem(p, buf, to_write, ppos); |
80ad89a0 WF |
535 | if (wrote != to_write) |
536 | return wrote; | |
1da177e4 LT |
537 | p += wrote; |
538 | buf += wrote; | |
539 | count -= wrote; | |
540 | } | |
541 | ||
542 | if (count > 0) { | |
543 | kbuf = (char *)__get_free_page(GFP_KERNEL); | |
544 | if (!kbuf) | |
545 | return wrote ? wrote : -ENOMEM; | |
546 | while (count > 0) { | |
80ad89a0 WF |
547 | unsigned long sz = size_inside_page(p, count); |
548 | unsigned long n; | |
1da177e4 | 549 | |
325fda71 KH |
550 | if (!is_vmalloc_or_module_addr((void *)p)) { |
551 | err = -ENXIO; | |
552 | break; | |
553 | } | |
80ad89a0 WF |
554 | n = copy_from_user(kbuf, buf, sz); |
555 | if (n) { | |
325fda71 KH |
556 | err = -EFAULT; |
557 | break; | |
1da177e4 | 558 | } |
c85e9a97 | 559 | vwrite(kbuf, (char *)p, sz); |
80ad89a0 WF |
560 | count -= sz; |
561 | buf += sz; | |
562 | virtr += sz; | |
563 | p += sz; | |
1da177e4 LT |
564 | } |
565 | free_page((unsigned long)kbuf); | |
566 | } | |
567 | ||
325fda71 KH |
568 | *ppos = p; |
569 | return virtr + wrote ? : err; | |
1da177e4 | 570 | } |
b781ecb6 | 571 | #endif |
1da177e4 | 572 | |
4f911d64 | 573 | #ifdef CONFIG_DEVPORT |
d7d4d849 | 574 | static ssize_t read_port(struct file *file, char __user *buf, |
1da177e4 LT |
575 | size_t count, loff_t *ppos) |
576 | { | |
577 | unsigned long i = *ppos; | |
578 | char __user *tmp = buf; | |
579 | ||
580 | if (!access_ok(VERIFY_WRITE, buf, count)) | |
d7d4d849 | 581 | return -EFAULT; |
1da177e4 | 582 | while (count-- > 0 && i < 65536) { |
d7d4d849 AM |
583 | if (__put_user(inb(i), tmp) < 0) |
584 | return -EFAULT; | |
1da177e4 LT |
585 | i++; |
586 | tmp++; | |
587 | } | |
588 | *ppos = i; | |
589 | return tmp-buf; | |
590 | } | |
591 | ||
d7d4d849 | 592 | static ssize_t write_port(struct file *file, const char __user *buf, |
1da177e4 LT |
593 | size_t count, loff_t *ppos) |
594 | { | |
595 | unsigned long i = *ppos; | |
596 | const char __user * tmp = buf; | |
597 | ||
d7d4d849 | 598 | if (!access_ok(VERIFY_READ, buf, count)) |
1da177e4 LT |
599 | return -EFAULT; |
600 | while (count-- > 0 && i < 65536) { | |
601 | char c; | |
c654d60e JB |
602 | if (__get_user(c, tmp)) { |
603 | if (tmp > buf) | |
604 | break; | |
d7d4d849 | 605 | return -EFAULT; |
c654d60e | 606 | } |
d7d4d849 | 607 | outb(c, i); |
1da177e4 LT |
608 | i++; |
609 | tmp++; | |
610 | } | |
611 | *ppos = i; | |
612 | return tmp-buf; | |
613 | } | |
614 | #endif | |
615 | ||
d7d4d849 | 616 | static ssize_t read_null(struct file *file, char __user *buf, |
1da177e4 LT |
617 | size_t count, loff_t *ppos) |
618 | { | |
619 | return 0; | |
620 | } | |
621 | ||
d7d4d849 | 622 | static ssize_t write_null(struct file *file, const char __user *buf, |
1da177e4 LT |
623 | size_t count, loff_t *ppos) |
624 | { | |
625 | return count; | |
626 | } | |
627 | ||
1ebd32fc JA |
628 | static int pipe_to_null(struct pipe_inode_info *info, struct pipe_buffer *buf, |
629 | struct splice_desc *sd) | |
630 | { | |
631 | return sd->len; | |
632 | } | |
633 | ||
d7d4d849 | 634 | static ssize_t splice_write_null(struct pipe_inode_info *pipe, struct file *out, |
1ebd32fc JA |
635 | loff_t *ppos, size_t len, unsigned int flags) |
636 | { | |
637 | return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_null); | |
638 | } | |
639 | ||
d7d4d849 | 640 | static ssize_t read_zero(struct file *file, char __user *buf, |
1da177e4 LT |
641 | size_t count, loff_t *ppos) |
642 | { | |
557ed1fa | 643 | size_t written; |
1da177e4 LT |
644 | |
645 | if (!count) | |
646 | return 0; | |
647 | ||
648 | if (!access_ok(VERIFY_WRITE, buf, count)) | |
649 | return -EFAULT; | |
650 | ||
557ed1fa NP |
651 | written = 0; |
652 | while (count) { | |
653 | unsigned long unwritten; | |
654 | size_t chunk = count; | |
1da177e4 | 655 | |
557ed1fa NP |
656 | if (chunk > PAGE_SIZE) |
657 | chunk = PAGE_SIZE; /* Just for latency reasons */ | |
bb521c5d | 658 | unwritten = __clear_user(buf, chunk); |
557ed1fa | 659 | written += chunk - unwritten; |
1da177e4 | 660 | if (unwritten) |
557ed1fa | 661 | break; |
2b838687 LT |
662 | if (signal_pending(current)) |
663 | return written ? written : -ERESTARTSYS; | |
1da177e4 | 664 | buf += chunk; |
557ed1fa | 665 | count -= chunk; |
1da177e4 LT |
666 | cond_resched(); |
667 | } | |
557ed1fa | 668 | return written ? written : -EFAULT; |
1da177e4 LT |
669 | } |
670 | ||
d7d4d849 | 671 | static int mmap_zero(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 672 | { |
557ed1fa | 673 | #ifndef CONFIG_MMU |
1da177e4 | 674 | return -ENOSYS; |
557ed1fa NP |
675 | #endif |
676 | if (vma->vm_flags & VM_SHARED) | |
677 | return shmem_zero_setup(vma); | |
678 | return 0; | |
1da177e4 | 679 | } |
1da177e4 | 680 | |
d7d4d849 | 681 | static ssize_t write_full(struct file *file, const char __user *buf, |
1da177e4 LT |
682 | size_t count, loff_t *ppos) |
683 | { | |
684 | return -ENOSPC; | |
685 | } | |
686 | ||
687 | /* | |
688 | * Special lseek() function for /dev/null and /dev/zero. Most notably, you | |
689 | * can fopen() both devices with "a" now. This was previously impossible. | |
690 | * -- SRB. | |
691 | */ | |
d7d4d849 | 692 | static loff_t null_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
693 | { |
694 | return file->f_pos = 0; | |
695 | } | |
696 | ||
697 | /* | |
698 | * The memory devices use the full 32/64 bits of the offset, and so we cannot | |
699 | * check against negative addresses: they are ok. The return value is weird, | |
700 | * though, in that case (0). | |
701 | * | |
702 | * also note that seeking relative to the "end of file" isn't supported: | |
703 | * it has no meaning, so it returns -EINVAL. | |
704 | */ | |
d7d4d849 | 705 | static loff_t memory_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
706 | { |
707 | loff_t ret; | |
708 | ||
a7113a96 | 709 | mutex_lock(&file->f_path.dentry->d_inode->i_mutex); |
1da177e4 | 710 | switch (orig) { |
d7d4d849 AM |
711 | case SEEK_CUR: |
712 | offset += file->f_pos; | |
d7d4d849 AM |
713 | case SEEK_SET: |
714 | /* to avoid userland mistaking f_pos=-9 as -EBADF=-9 */ | |
715 | if ((unsigned long long)offset >= ~0xFFFULL) { | |
716 | ret = -EOVERFLOW; | |
1da177e4 | 717 | break; |
d7d4d849 AM |
718 | } |
719 | file->f_pos = offset; | |
720 | ret = file->f_pos; | |
721 | force_successful_syscall_return(); | |
722 | break; | |
723 | default: | |
724 | ret = -EINVAL; | |
1da177e4 | 725 | } |
a7113a96 | 726 | mutex_unlock(&file->f_path.dentry->d_inode->i_mutex); |
1da177e4 LT |
727 | return ret; |
728 | } | |
729 | ||
730 | static int open_port(struct inode * inode, struct file * filp) | |
731 | { | |
732 | return capable(CAP_SYS_RAWIO) ? 0 : -EPERM; | |
733 | } | |
734 | ||
735 | #define zero_lseek null_lseek | |
736 | #define full_lseek null_lseek | |
737 | #define write_zero write_null | |
738 | #define read_full read_zero | |
739 | #define open_mem open_port | |
740 | #define open_kmem open_mem | |
50b1fdbd | 741 | #define open_oldmem open_mem |
1da177e4 | 742 | |
62322d25 | 743 | static const struct file_operations mem_fops = { |
1da177e4 LT |
744 | .llseek = memory_lseek, |
745 | .read = read_mem, | |
746 | .write = write_mem, | |
747 | .mmap = mmap_mem, | |
748 | .open = open_mem, | |
5da6185b | 749 | .get_unmapped_area = get_unmapped_area_mem, |
1da177e4 LT |
750 | }; |
751 | ||
b781ecb6 | 752 | #ifdef CONFIG_DEVKMEM |
62322d25 | 753 | static const struct file_operations kmem_fops = { |
1da177e4 LT |
754 | .llseek = memory_lseek, |
755 | .read = read_kmem, | |
756 | .write = write_kmem, | |
757 | .mmap = mmap_kmem, | |
758 | .open = open_kmem, | |
5da6185b | 759 | .get_unmapped_area = get_unmapped_area_mem, |
1da177e4 | 760 | }; |
b781ecb6 | 761 | #endif |
1da177e4 | 762 | |
62322d25 | 763 | static const struct file_operations null_fops = { |
1da177e4 LT |
764 | .llseek = null_lseek, |
765 | .read = read_null, | |
766 | .write = write_null, | |
1ebd32fc | 767 | .splice_write = splice_write_null, |
1da177e4 LT |
768 | }; |
769 | ||
4f911d64 | 770 | #ifdef CONFIG_DEVPORT |
62322d25 | 771 | static const struct file_operations port_fops = { |
1da177e4 LT |
772 | .llseek = memory_lseek, |
773 | .read = read_port, | |
774 | .write = write_port, | |
775 | .open = open_port, | |
776 | }; | |
777 | #endif | |
778 | ||
62322d25 | 779 | static const struct file_operations zero_fops = { |
1da177e4 LT |
780 | .llseek = zero_lseek, |
781 | .read = read_zero, | |
782 | .write = write_zero, | |
783 | .mmap = mmap_zero, | |
784 | }; | |
785 | ||
5da6185b DH |
786 | /* |
787 | * capabilities for /dev/zero | |
788 | * - permits private mappings, "copies" are taken of the source of zeros | |
371d217e | 789 | * - no writeback happens |
5da6185b | 790 | */ |
1da177e4 | 791 | static struct backing_dev_info zero_bdi = { |
d993831f | 792 | .name = "char/mem", |
371d217e | 793 | .capabilities = BDI_CAP_MAP_COPY | BDI_CAP_NO_ACCT_AND_WRITEBACK, |
1da177e4 LT |
794 | }; |
795 | ||
62322d25 | 796 | static const struct file_operations full_fops = { |
1da177e4 LT |
797 | .llseek = full_lseek, |
798 | .read = read_full, | |
799 | .write = write_full, | |
800 | }; | |
801 | ||
50b1fdbd | 802 | #ifdef CONFIG_CRASH_DUMP |
62322d25 | 803 | static const struct file_operations oldmem_fops = { |
50b1fdbd VG |
804 | .read = read_oldmem, |
805 | .open = open_oldmem, | |
6038f373 | 806 | .llseek = default_llseek, |
50b1fdbd VG |
807 | }; |
808 | #endif | |
809 | ||
7e5b58bc KS |
810 | static ssize_t kmsg_writev(struct kiocb *iocb, const struct iovec *iv, |
811 | unsigned long count, loff_t pos) | |
1da177e4 | 812 | { |
7e5b58bc | 813 | char *line, *p; |
70a5f521 | 814 | int i; |
7e5b58bc | 815 | ssize_t ret = -EFAULT; |
70a5f521 | 816 | size_t len = iov_length(iv, count); |
1da177e4 | 817 | |
70a5f521 | 818 | line = kmalloc(len + 1, GFP_KERNEL); |
7e5b58bc | 819 | if (line == NULL) |
1da177e4 | 820 | return -ENOMEM; |
7e5b58bc KS |
821 | |
822 | /* | |
823 | * copy all vectors into a single string, to ensure we do | |
824 | * not interleave our log line with other printk calls | |
825 | */ | |
70a5f521 | 826 | p = line; |
7e5b58bc KS |
827 | for (i = 0; i < count; i++) { |
828 | if (copy_from_user(p, iv[i].iov_base, iv[i].iov_len)) | |
829 | goto out; | |
830 | p += iv[i].iov_len; | |
1da177e4 | 831 | } |
7e5b58bc KS |
832 | p[0] = '\0'; |
833 | ||
834 | ret = printk("%s", line); | |
835 | /* printk can add a prefix */ | |
836 | if (ret > len) | |
837 | ret = len; | |
838 | out: | |
839 | kfree(line); | |
1da177e4 LT |
840 | return ret; |
841 | } | |
842 | ||
62322d25 | 843 | static const struct file_operations kmsg_fops = { |
7e5b58bc | 844 | .aio_write = kmsg_writev, |
6038f373 | 845 | .llseek = noop_llseek, |
1da177e4 LT |
846 | }; |
847 | ||
389e0cb9 KS |
848 | static const struct memdev { |
849 | const char *name; | |
e454cea2 | 850 | mode_t mode; |
389e0cb9 KS |
851 | const struct file_operations *fops; |
852 | struct backing_dev_info *dev_info; | |
853 | } devlist[] = { | |
e454cea2 | 854 | [1] = { "mem", 0, &mem_fops, &directly_mappable_cdev_bdi }, |
b781ecb6 | 855 | #ifdef CONFIG_DEVKMEM |
e454cea2 | 856 | [2] = { "kmem", 0, &kmem_fops, &directly_mappable_cdev_bdi }, |
b781ecb6 | 857 | #endif |
e454cea2 | 858 | [3] = { "null", 0666, &null_fops, NULL }, |
4f911d64 | 859 | #ifdef CONFIG_DEVPORT |
e454cea2 | 860 | [4] = { "port", 0, &port_fops, NULL }, |
1da177e4 | 861 | #endif |
e454cea2 KS |
862 | [5] = { "zero", 0666, &zero_fops, &zero_bdi }, |
863 | [7] = { "full", 0666, &full_fops, NULL }, | |
864 | [8] = { "random", 0666, &random_fops, NULL }, | |
865 | [9] = { "urandom", 0666, &urandom_fops, NULL }, | |
866 | [11] = { "kmsg", 0, &kmsg_fops, NULL }, | |
50b1fdbd | 867 | #ifdef CONFIG_CRASH_DUMP |
e454cea2 | 868 | [12] = { "oldmem", 0, &oldmem_fops, NULL }, |
50b1fdbd | 869 | #endif |
d6f47bef ASF |
870 | }; |
871 | ||
872 | static int memory_open(struct inode *inode, struct file *filp) | |
873 | { | |
389e0cb9 KS |
874 | int minor; |
875 | const struct memdev *dev; | |
d6f47bef | 876 | |
389e0cb9 KS |
877 | minor = iminor(inode); |
878 | if (minor >= ARRAY_SIZE(devlist)) | |
205153aa | 879 | return -ENXIO; |
d6f47bef | 880 | |
389e0cb9 KS |
881 | dev = &devlist[minor]; |
882 | if (!dev->fops) | |
205153aa | 883 | return -ENXIO; |
d6f47bef | 884 | |
389e0cb9 KS |
885 | filp->f_op = dev->fops; |
886 | if (dev->dev_info) | |
887 | filp->f_mapping->backing_dev_info = dev->dev_info; | |
d6f47bef | 888 | |
4a3956c7 KH |
889 | /* Is /dev/mem or /dev/kmem ? */ |
890 | if (dev->dev_info == &directly_mappable_cdev_bdi) | |
891 | filp->f_mode |= FMODE_UNSIGNED_OFFSET; | |
892 | ||
389e0cb9 | 893 | if (dev->fops->open) |
205153aa FW |
894 | return dev->fops->open(inode, filp); |
895 | ||
896 | return 0; | |
1da177e4 LT |
897 | } |
898 | ||
62322d25 | 899 | static const struct file_operations memory_fops = { |
d7d4d849 | 900 | .open = memory_open, |
6038f373 | 901 | .llseek = noop_llseek, |
1da177e4 LT |
902 | }; |
903 | ||
e454cea2 KS |
904 | static char *mem_devnode(struct device *dev, mode_t *mode) |
905 | { | |
906 | if (mode && devlist[MINOR(dev->devt)].mode) | |
907 | *mode = devlist[MINOR(dev->devt)].mode; | |
908 | return NULL; | |
909 | } | |
910 | ||
ca8eca68 | 911 | static struct class *mem_class; |
1da177e4 LT |
912 | |
913 | static int __init chr_dev_init(void) | |
914 | { | |
389e0cb9 | 915 | int minor; |
e0bf68dd PZ |
916 | int err; |
917 | ||
918 | err = bdi_init(&zero_bdi); | |
919 | if (err) | |
920 | return err; | |
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; | |
933 | device_create(mem_class, NULL, MKDEV(MEM_MAJOR, minor), | |
934 | NULL, devlist[minor].name); | |
935 | } | |
ebf644c4 | 936 | |
31d1d48e | 937 | return tty_init(); |
1da177e4 LT |
938 | } |
939 | ||
940 | fs_initcall(chr_dev_init); |