Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/drivers/char/mem.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * | |
6 | * Added devfs support. | |
7 | * Jan-11-1998, C. Scott Ananian <cananian@alumni.princeton.edu> | |
8 | * Shared /dev/zero mmaping support, Feb 2000, Kanoj Sarcar <kanoj@sgi.com> | |
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 | ||
1da177e4 LT |
37 | /* |
38 | * Architectures vary in how they handle caching for addresses | |
39 | * outside of main memory. | |
40 | * | |
41 | */ | |
42 | static inline int uncached_access(struct file *file, unsigned long addr) | |
43 | { | |
44 | #if defined(__i386__) | |
45 | /* | |
46 | * On the PPro and successors, the MTRRs are used to set | |
47 | * memory types for physical addresses outside main memory, | |
48 | * so blindly setting PCD or PWT on those pages is wrong. | |
49 | * For Pentiums and earlier, the surround logic should disable | |
50 | * caching for the high addresses through the KEN pin, but | |
51 | * we maintain the tradition of paranoia in this code. | |
52 | */ | |
53 | if (file->f_flags & O_SYNC) | |
54 | return 1; | |
55 | return !( test_bit(X86_FEATURE_MTRR, boot_cpu_data.x86_capability) || | |
56 | test_bit(X86_FEATURE_K6_MTRR, boot_cpu_data.x86_capability) || | |
57 | test_bit(X86_FEATURE_CYRIX_ARR, boot_cpu_data.x86_capability) || | |
58 | test_bit(X86_FEATURE_CENTAUR_MCR, boot_cpu_data.x86_capability) ) | |
59 | && addr >= __pa(high_memory); | |
60 | #elif defined(__x86_64__) | |
61 | /* | |
62 | * This is broken because it can generate memory type aliases, | |
63 | * which can cause cache corruptions | |
64 | * But it is only available for root and we have to be bug-to-bug | |
65 | * compatible with i386. | |
66 | */ | |
67 | if (file->f_flags & O_SYNC) | |
68 | return 1; | |
69 | /* same behaviour as i386. PAT always set to cached and MTRRs control the | |
70 | caching behaviour. | |
71 | Hopefully a full PAT implementation will fix that soon. */ | |
72 | return 0; | |
73 | #elif defined(CONFIG_IA64) | |
74 | /* | |
75 | * On ia64, we ignore O_SYNC because we cannot tolerate memory attribute aliases. | |
76 | */ | |
77 | return !(efi_mem_attributes(addr) & EFI_MEMORY_WB); | |
24e9d0b9 RB |
78 | #elif defined(CONFIG_MIPS) |
79 | { | |
80 | extern int __uncached_access(struct file *file, | |
81 | unsigned long addr); | |
82 | ||
83 | return __uncached_access(file, addr); | |
84 | } | |
1da177e4 LT |
85 | #else |
86 | /* | |
87 | * Accessing memory above the top the kernel knows about or through a file pointer | |
88 | * that was marked O_SYNC will be done non-cached. | |
89 | */ | |
90 | if (file->f_flags & O_SYNC) | |
91 | return 1; | |
92 | return addr >= __pa(high_memory); | |
93 | #endif | |
94 | } | |
95 | ||
96 | #ifndef ARCH_HAS_VALID_PHYS_ADDR_RANGE | |
136939a2 | 97 | static inline int valid_phys_addr_range(unsigned long addr, size_t count) |
1da177e4 | 98 | { |
136939a2 | 99 | if (addr + count > __pa(high_memory)) |
1da177e4 LT |
100 | return 0; |
101 | ||
1da177e4 LT |
102 | return 1; |
103 | } | |
80851ef2 | 104 | |
06c67bef | 105 | static inline int valid_mmap_phys_addr_range(unsigned long pfn, size_t size) |
80851ef2 BH |
106 | { |
107 | return 1; | |
108 | } | |
1da177e4 LT |
109 | #endif |
110 | ||
111 | /* | |
112 | * This funcion reads the *physical* memory. The f_pos points directly to the | |
113 | * memory location. | |
114 | */ | |
115 | static ssize_t read_mem(struct file * file, char __user * buf, | |
116 | size_t count, loff_t *ppos) | |
117 | { | |
118 | unsigned long p = *ppos; | |
119 | ssize_t read, sz; | |
120 | char *ptr; | |
121 | ||
136939a2 | 122 | if (!valid_phys_addr_range(p, count)) |
1da177e4 LT |
123 | return -EFAULT; |
124 | read = 0; | |
125 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
126 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
127 | if (p < PAGE_SIZE) { | |
128 | sz = PAGE_SIZE - p; | |
129 | if (sz > count) | |
130 | sz = count; | |
131 | if (sz > 0) { | |
132 | if (clear_user(buf, sz)) | |
133 | return -EFAULT; | |
134 | buf += sz; | |
135 | p += sz; | |
136 | count -= sz; | |
137 | read += sz; | |
138 | } | |
139 | } | |
140 | #endif | |
141 | ||
142 | while (count > 0) { | |
143 | /* | |
144 | * Handle first page in case it's not aligned | |
145 | */ | |
146 | if (-p & (PAGE_SIZE - 1)) | |
147 | sz = -p & (PAGE_SIZE - 1); | |
148 | else | |
149 | sz = PAGE_SIZE; | |
150 | ||
151 | sz = min_t(unsigned long, sz, count); | |
152 | ||
153 | /* | |
154 | * On ia64 if a page has been mapped somewhere as | |
155 | * uncached, then it must also be accessed uncached | |
156 | * by the kernel or data corruption may occur | |
157 | */ | |
158 | ptr = xlate_dev_mem_ptr(p); | |
159 | ||
160 | if (copy_to_user(buf, ptr, sz)) | |
161 | return -EFAULT; | |
162 | buf += sz; | |
163 | p += sz; | |
164 | count -= sz; | |
165 | read += sz; | |
166 | } | |
167 | ||
168 | *ppos += read; | |
169 | return read; | |
170 | } | |
171 | ||
172 | static ssize_t write_mem(struct file * file, const char __user * buf, | |
173 | size_t count, loff_t *ppos) | |
174 | { | |
175 | unsigned long p = *ppos; | |
176 | ssize_t written, sz; | |
177 | unsigned long copied; | |
178 | void *ptr; | |
179 | ||
136939a2 | 180 | if (!valid_phys_addr_range(p, count)) |
1da177e4 LT |
181 | return -EFAULT; |
182 | ||
183 | written = 0; | |
184 | ||
185 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
186 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
187 | if (p < PAGE_SIZE) { | |
188 | unsigned long sz = PAGE_SIZE - p; | |
189 | if (sz > count) | |
190 | sz = count; | |
191 | /* Hmm. Do something? */ | |
192 | buf += sz; | |
193 | p += sz; | |
194 | count -= sz; | |
195 | written += sz; | |
196 | } | |
197 | #endif | |
198 | ||
199 | while (count > 0) { | |
200 | /* | |
201 | * Handle first page in case it's not aligned | |
202 | */ | |
203 | if (-p & (PAGE_SIZE - 1)) | |
204 | sz = -p & (PAGE_SIZE - 1); | |
205 | else | |
206 | sz = PAGE_SIZE; | |
207 | ||
208 | sz = min_t(unsigned long, sz, count); | |
209 | ||
210 | /* | |
211 | * On ia64 if a page has been mapped somewhere as | |
212 | * uncached, then it must also be accessed uncached | |
213 | * by the kernel or data corruption may occur | |
214 | */ | |
215 | ptr = xlate_dev_mem_ptr(p); | |
216 | ||
217 | copied = copy_from_user(ptr, buf, sz); | |
218 | if (copied) { | |
c654d60e JB |
219 | written += sz - copied; |
220 | if (written) | |
221 | break; | |
1da177e4 LT |
222 | return -EFAULT; |
223 | } | |
224 | buf += sz; | |
225 | p += sz; | |
226 | count -= sz; | |
227 | written += sz; | |
228 | } | |
229 | ||
230 | *ppos += written; | |
231 | return written; | |
232 | } | |
233 | ||
44ac8413 BH |
234 | #ifndef __HAVE_PHYS_MEM_ACCESS_PROT |
235 | static pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, | |
236 | unsigned long size, pgprot_t vma_prot) | |
237 | { | |
238 | #ifdef pgprot_noncached | |
239 | unsigned long offset = pfn << PAGE_SHIFT; | |
240 | ||
241 | if (uncached_access(file, offset)) | |
242 | return pgprot_noncached(vma_prot); | |
243 | #endif | |
244 | return vma_prot; | |
245 | } | |
246 | #endif | |
247 | ||
5da6185b DH |
248 | #ifndef CONFIG_MMU |
249 | static unsigned long get_unmapped_area_mem(struct file *file, | |
250 | unsigned long addr, | |
251 | unsigned long len, | |
252 | unsigned long pgoff, | |
253 | unsigned long flags) | |
254 | { | |
255 | if (!valid_mmap_phys_addr_range(pgoff, len)) | |
256 | return (unsigned long) -EINVAL; | |
8a93258c | 257 | return pgoff << PAGE_SHIFT; |
5da6185b DH |
258 | } |
259 | ||
260 | /* can't do an in-place private mapping if there's no MMU */ | |
261 | static inline int private_mapping_ok(struct vm_area_struct *vma) | |
262 | { | |
263 | return vma->vm_flags & VM_MAYSHARE; | |
264 | } | |
265 | #else | |
266 | #define get_unmapped_area_mem NULL | |
267 | ||
268 | static inline int private_mapping_ok(struct vm_area_struct *vma) | |
269 | { | |
270 | return 1; | |
271 | } | |
272 | #endif | |
273 | ||
1da177e4 LT |
274 | static int mmap_mem(struct file * file, struct vm_area_struct * vma) |
275 | { | |
80851ef2 BH |
276 | size_t size = vma->vm_end - vma->vm_start; |
277 | ||
06c67bef | 278 | if (!valid_mmap_phys_addr_range(vma->vm_pgoff, size)) |
80851ef2 BH |
279 | return -EINVAL; |
280 | ||
5da6185b DH |
281 | if (!private_mapping_ok(vma)) |
282 | return -ENOSYS; | |
283 | ||
8b150478 | 284 | vma->vm_page_prot = phys_mem_access_prot(file, vma->vm_pgoff, |
80851ef2 | 285 | size, |
1da177e4 | 286 | vma->vm_page_prot); |
1da177e4 LT |
287 | |
288 | /* Remap-pfn-range will mark the range VM_IO and VM_RESERVED */ | |
289 | if (remap_pfn_range(vma, | |
290 | vma->vm_start, | |
291 | vma->vm_pgoff, | |
80851ef2 | 292 | size, |
1da177e4 LT |
293 | vma->vm_page_prot)) |
294 | return -EAGAIN; | |
295 | return 0; | |
296 | } | |
297 | ||
298 | static int mmap_kmem(struct file * file, struct vm_area_struct * vma) | |
299 | { | |
4bb82551 LT |
300 | unsigned long pfn; |
301 | ||
6d3154cc LT |
302 | /* Turn a kernel-virtual address into a physical page frame */ |
303 | pfn = __pa((u64)vma->vm_pgoff << PAGE_SHIFT) >> PAGE_SHIFT; | |
4bb82551 | 304 | |
1da177e4 LT |
305 | /* |
306 | * RED-PEN: on some architectures there is more mapped memory | |
307 | * than available in mem_map which pfn_valid checks | |
308 | * for. Perhaps should add a new macro here. | |
309 | * | |
310 | * RED-PEN: vmalloc is not supported right now. | |
311 | */ | |
4bb82551 | 312 | if (!pfn_valid(pfn)) |
1da177e4 | 313 | return -EIO; |
4bb82551 LT |
314 | |
315 | vma->vm_pgoff = pfn; | |
1da177e4 LT |
316 | return mmap_mem(file, vma); |
317 | } | |
318 | ||
50b1fdbd VG |
319 | #ifdef CONFIG_CRASH_DUMP |
320 | /* | |
321 | * Read memory corresponding to the old kernel. | |
50b1fdbd | 322 | */ |
315c215c | 323 | static ssize_t read_oldmem(struct file *file, char __user *buf, |
50b1fdbd VG |
324 | size_t count, loff_t *ppos) |
325 | { | |
315c215c VG |
326 | unsigned long pfn, offset; |
327 | size_t read = 0, csize; | |
328 | int rc = 0; | |
50b1fdbd | 329 | |
72414d3f | 330 | while (count) { |
50b1fdbd | 331 | pfn = *ppos / PAGE_SIZE; |
315c215c VG |
332 | if (pfn > saved_max_pfn) |
333 | return read; | |
50b1fdbd | 334 | |
315c215c VG |
335 | offset = (unsigned long)(*ppos % PAGE_SIZE); |
336 | if (count > PAGE_SIZE - offset) | |
337 | csize = PAGE_SIZE - offset; | |
338 | else | |
339 | csize = count; | |
50b1fdbd | 340 | |
315c215c VG |
341 | rc = copy_oldmem_page(pfn, buf, csize, offset, 1); |
342 | if (rc < 0) | |
343 | return rc; | |
50b1fdbd VG |
344 | buf += csize; |
345 | *ppos += csize; | |
346 | read += csize; | |
347 | count -= csize; | |
348 | } | |
50b1fdbd VG |
349 | return read; |
350 | } | |
351 | #endif | |
352 | ||
1da177e4 LT |
353 | extern long vread(char *buf, char *addr, unsigned long count); |
354 | extern long vwrite(char *buf, char *addr, unsigned long count); | |
355 | ||
356 | /* | |
357 | * This function reads the *virtual* memory as seen by the kernel. | |
358 | */ | |
359 | static ssize_t read_kmem(struct file *file, char __user *buf, | |
360 | size_t count, loff_t *ppos) | |
361 | { | |
362 | unsigned long p = *ppos; | |
363 | ssize_t low_count, read, sz; | |
364 | char * kbuf; /* k-addr because vread() takes vmlist_lock rwlock */ | |
365 | ||
366 | read = 0; | |
367 | if (p < (unsigned long) high_memory) { | |
368 | low_count = count; | |
369 | if (count > (unsigned long) high_memory - p) | |
370 | low_count = (unsigned long) high_memory - p; | |
371 | ||
372 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
373 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
374 | if (p < PAGE_SIZE && low_count > 0) { | |
375 | size_t tmp = PAGE_SIZE - p; | |
376 | if (tmp > low_count) tmp = low_count; | |
377 | if (clear_user(buf, tmp)) | |
378 | return -EFAULT; | |
379 | buf += tmp; | |
380 | p += tmp; | |
381 | read += tmp; | |
382 | low_count -= tmp; | |
383 | count -= tmp; | |
384 | } | |
385 | #endif | |
386 | while (low_count > 0) { | |
387 | /* | |
388 | * Handle first page in case it's not aligned | |
389 | */ | |
390 | if (-p & (PAGE_SIZE - 1)) | |
391 | sz = -p & (PAGE_SIZE - 1); | |
392 | else | |
393 | sz = PAGE_SIZE; | |
394 | ||
395 | sz = min_t(unsigned long, sz, low_count); | |
396 | ||
397 | /* | |
398 | * On ia64 if a page has been mapped somewhere as | |
399 | * uncached, then it must also be accessed uncached | |
400 | * by the kernel or data corruption may occur | |
401 | */ | |
402 | kbuf = xlate_dev_kmem_ptr((char *)p); | |
403 | ||
404 | if (copy_to_user(buf, kbuf, sz)) | |
405 | return -EFAULT; | |
406 | buf += sz; | |
407 | p += sz; | |
408 | read += sz; | |
409 | low_count -= sz; | |
410 | count -= sz; | |
411 | } | |
412 | } | |
413 | ||
414 | if (count > 0) { | |
415 | kbuf = (char *)__get_free_page(GFP_KERNEL); | |
416 | if (!kbuf) | |
417 | return -ENOMEM; | |
418 | while (count > 0) { | |
419 | int len = count; | |
420 | ||
421 | if (len > PAGE_SIZE) | |
422 | len = PAGE_SIZE; | |
423 | len = vread(kbuf, (char *)p, len); | |
424 | if (!len) | |
425 | break; | |
426 | if (copy_to_user(buf, kbuf, len)) { | |
427 | free_page((unsigned long)kbuf); | |
428 | return -EFAULT; | |
429 | } | |
430 | count -= len; | |
431 | buf += len; | |
432 | read += len; | |
433 | p += len; | |
434 | } | |
435 | free_page((unsigned long)kbuf); | |
436 | } | |
437 | *ppos = p; | |
438 | return read; | |
439 | } | |
440 | ||
441 | ||
442 | static inline ssize_t | |
443 | do_write_kmem(void *p, unsigned long realp, const char __user * buf, | |
444 | size_t count, loff_t *ppos) | |
445 | { | |
446 | ssize_t written, sz; | |
447 | unsigned long copied; | |
448 | ||
449 | written = 0; | |
450 | #ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED | |
451 | /* we don't have page 0 mapped on sparc and m68k.. */ | |
452 | if (realp < PAGE_SIZE) { | |
453 | unsigned long sz = PAGE_SIZE - realp; | |
454 | if (sz > count) | |
455 | sz = count; | |
456 | /* Hmm. Do something? */ | |
457 | buf += sz; | |
458 | p += sz; | |
459 | realp += sz; | |
460 | count -= sz; | |
461 | written += sz; | |
462 | } | |
463 | #endif | |
464 | ||
465 | while (count > 0) { | |
466 | char *ptr; | |
467 | /* | |
468 | * Handle first page in case it's not aligned | |
469 | */ | |
470 | if (-realp & (PAGE_SIZE - 1)) | |
471 | sz = -realp & (PAGE_SIZE - 1); | |
472 | else | |
473 | sz = PAGE_SIZE; | |
474 | ||
475 | sz = min_t(unsigned long, sz, count); | |
476 | ||
477 | /* | |
478 | * On ia64 if a page has been mapped somewhere as | |
479 | * uncached, then it must also be accessed uncached | |
480 | * by the kernel or data corruption may occur | |
481 | */ | |
482 | ptr = xlate_dev_kmem_ptr(p); | |
483 | ||
484 | copied = copy_from_user(ptr, buf, sz); | |
485 | if (copied) { | |
c654d60e JB |
486 | written += sz - copied; |
487 | if (written) | |
488 | break; | |
1da177e4 LT |
489 | return -EFAULT; |
490 | } | |
491 | buf += sz; | |
492 | p += sz; | |
493 | realp += sz; | |
494 | count -= sz; | |
495 | written += sz; | |
496 | } | |
497 | ||
498 | *ppos += written; | |
499 | return written; | |
500 | } | |
501 | ||
502 | ||
503 | /* | |
504 | * This function writes to the *virtual* memory as seen by the kernel. | |
505 | */ | |
506 | static ssize_t write_kmem(struct file * file, const char __user * buf, | |
507 | size_t count, loff_t *ppos) | |
508 | { | |
509 | unsigned long p = *ppos; | |
510 | ssize_t wrote = 0; | |
511 | ssize_t virtr = 0; | |
512 | ssize_t written; | |
513 | char * kbuf; /* k-addr because vwrite() takes vmlist_lock rwlock */ | |
514 | ||
515 | if (p < (unsigned long) high_memory) { | |
516 | ||
517 | wrote = count; | |
518 | if (count > (unsigned long) high_memory - p) | |
519 | wrote = (unsigned long) high_memory - p; | |
520 | ||
521 | written = do_write_kmem((void*)p, p, buf, wrote, ppos); | |
522 | if (written != wrote) | |
523 | return written; | |
524 | wrote = written; | |
525 | p += wrote; | |
526 | buf += wrote; | |
527 | count -= wrote; | |
528 | } | |
529 | ||
530 | if (count > 0) { | |
531 | kbuf = (char *)__get_free_page(GFP_KERNEL); | |
532 | if (!kbuf) | |
533 | return wrote ? wrote : -ENOMEM; | |
534 | while (count > 0) { | |
535 | int len = count; | |
536 | ||
537 | if (len > PAGE_SIZE) | |
538 | len = PAGE_SIZE; | |
539 | if (len) { | |
540 | written = copy_from_user(kbuf, buf, len); | |
541 | if (written) { | |
c654d60e JB |
542 | if (wrote + virtr) |
543 | break; | |
1da177e4 | 544 | free_page((unsigned long)kbuf); |
c654d60e | 545 | return -EFAULT; |
1da177e4 LT |
546 | } |
547 | } | |
548 | len = vwrite(kbuf, (char *)p, len); | |
549 | count -= len; | |
550 | buf += len; | |
551 | virtr += len; | |
552 | p += len; | |
553 | } | |
554 | free_page((unsigned long)kbuf); | |
555 | } | |
556 | ||
557 | *ppos = p; | |
558 | return virtr + wrote; | |
559 | } | |
560 | ||
4f911d64 | 561 | #ifdef CONFIG_DEVPORT |
1da177e4 LT |
562 | static ssize_t read_port(struct file * file, char __user * buf, |
563 | size_t count, loff_t *ppos) | |
564 | { | |
565 | unsigned long i = *ppos; | |
566 | char __user *tmp = buf; | |
567 | ||
568 | if (!access_ok(VERIFY_WRITE, buf, count)) | |
569 | return -EFAULT; | |
570 | while (count-- > 0 && i < 65536) { | |
571 | if (__put_user(inb(i),tmp) < 0) | |
572 | return -EFAULT; | |
573 | i++; | |
574 | tmp++; | |
575 | } | |
576 | *ppos = i; | |
577 | return tmp-buf; | |
578 | } | |
579 | ||
580 | static ssize_t write_port(struct file * file, const char __user * buf, | |
581 | size_t count, loff_t *ppos) | |
582 | { | |
583 | unsigned long i = *ppos; | |
584 | const char __user * tmp = buf; | |
585 | ||
586 | if (!access_ok(VERIFY_READ,buf,count)) | |
587 | return -EFAULT; | |
588 | while (count-- > 0 && i < 65536) { | |
589 | char c; | |
c654d60e JB |
590 | if (__get_user(c, tmp)) { |
591 | if (tmp > buf) | |
592 | break; | |
1da177e4 | 593 | return -EFAULT; |
c654d60e | 594 | } |
1da177e4 LT |
595 | outb(c,i); |
596 | i++; | |
597 | tmp++; | |
598 | } | |
599 | *ppos = i; | |
600 | return tmp-buf; | |
601 | } | |
602 | #endif | |
603 | ||
604 | static ssize_t read_null(struct file * file, char __user * buf, | |
605 | size_t count, loff_t *ppos) | |
606 | { | |
607 | return 0; | |
608 | } | |
609 | ||
610 | static ssize_t write_null(struct file * file, const char __user * buf, | |
611 | size_t count, loff_t *ppos) | |
612 | { | |
613 | return count; | |
614 | } | |
615 | ||
1ebd32fc JA |
616 | static int pipe_to_null(struct pipe_inode_info *info, struct pipe_buffer *buf, |
617 | struct splice_desc *sd) | |
618 | { | |
619 | return sd->len; | |
620 | } | |
621 | ||
622 | static ssize_t splice_write_null(struct pipe_inode_info *pipe,struct file *out, | |
623 | loff_t *ppos, size_t len, unsigned int flags) | |
624 | { | |
625 | return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_null); | |
626 | } | |
627 | ||
1da177e4 LT |
628 | static ssize_t read_zero(struct file * file, char __user * buf, |
629 | size_t count, loff_t *ppos) | |
630 | { | |
557ed1fa | 631 | size_t written; |
1da177e4 LT |
632 | |
633 | if (!count) | |
634 | return 0; | |
635 | ||
636 | if (!access_ok(VERIFY_WRITE, buf, count)) | |
637 | return -EFAULT; | |
638 | ||
557ed1fa NP |
639 | written = 0; |
640 | while (count) { | |
641 | unsigned long unwritten; | |
642 | size_t chunk = count; | |
1da177e4 | 643 | |
557ed1fa NP |
644 | if (chunk > PAGE_SIZE) |
645 | chunk = PAGE_SIZE; /* Just for latency reasons */ | |
646 | unwritten = clear_user(buf, chunk); | |
647 | written += chunk - unwritten; | |
1da177e4 | 648 | if (unwritten) |
557ed1fa | 649 | break; |
1da177e4 | 650 | buf += chunk; |
557ed1fa | 651 | count -= chunk; |
1da177e4 LT |
652 | cond_resched(); |
653 | } | |
557ed1fa | 654 | return written ? written : -EFAULT; |
1da177e4 LT |
655 | } |
656 | ||
657 | static int mmap_zero(struct file * file, struct vm_area_struct * vma) | |
658 | { | |
557ed1fa | 659 | #ifndef CONFIG_MMU |
1da177e4 | 660 | return -ENOSYS; |
557ed1fa NP |
661 | #endif |
662 | if (vma->vm_flags & VM_SHARED) | |
663 | return shmem_zero_setup(vma); | |
664 | return 0; | |
1da177e4 | 665 | } |
1da177e4 LT |
666 | |
667 | static ssize_t write_full(struct file * file, const char __user * buf, | |
668 | size_t count, loff_t *ppos) | |
669 | { | |
670 | return -ENOSPC; | |
671 | } | |
672 | ||
673 | /* | |
674 | * Special lseek() function for /dev/null and /dev/zero. Most notably, you | |
675 | * can fopen() both devices with "a" now. This was previously impossible. | |
676 | * -- SRB. | |
677 | */ | |
678 | ||
679 | static loff_t null_lseek(struct file * file, loff_t offset, int orig) | |
680 | { | |
681 | return file->f_pos = 0; | |
682 | } | |
683 | ||
684 | /* | |
685 | * The memory devices use the full 32/64 bits of the offset, and so we cannot | |
686 | * check against negative addresses: they are ok. The return value is weird, | |
687 | * though, in that case (0). | |
688 | * | |
689 | * also note that seeking relative to the "end of file" isn't supported: | |
690 | * it has no meaning, so it returns -EINVAL. | |
691 | */ | |
692 | static loff_t memory_lseek(struct file * file, loff_t offset, int orig) | |
693 | { | |
694 | loff_t ret; | |
695 | ||
a7113a96 | 696 | mutex_lock(&file->f_path.dentry->d_inode->i_mutex); |
1da177e4 LT |
697 | switch (orig) { |
698 | case 0: | |
699 | file->f_pos = offset; | |
700 | ret = file->f_pos; | |
701 | force_successful_syscall_return(); | |
702 | break; | |
703 | case 1: | |
704 | file->f_pos += offset; | |
705 | ret = file->f_pos; | |
706 | force_successful_syscall_return(); | |
707 | break; | |
708 | default: | |
709 | ret = -EINVAL; | |
710 | } | |
a7113a96 | 711 | mutex_unlock(&file->f_path.dentry->d_inode->i_mutex); |
1da177e4 LT |
712 | return ret; |
713 | } | |
714 | ||
715 | static int open_port(struct inode * inode, struct file * filp) | |
716 | { | |
717 | return capable(CAP_SYS_RAWIO) ? 0 : -EPERM; | |
718 | } | |
719 | ||
720 | #define zero_lseek null_lseek | |
721 | #define full_lseek null_lseek | |
722 | #define write_zero write_null | |
723 | #define read_full read_zero | |
724 | #define open_mem open_port | |
725 | #define open_kmem open_mem | |
50b1fdbd | 726 | #define open_oldmem open_mem |
1da177e4 | 727 | |
62322d25 | 728 | static const struct file_operations mem_fops = { |
1da177e4 LT |
729 | .llseek = memory_lseek, |
730 | .read = read_mem, | |
731 | .write = write_mem, | |
732 | .mmap = mmap_mem, | |
733 | .open = open_mem, | |
5da6185b | 734 | .get_unmapped_area = get_unmapped_area_mem, |
1da177e4 LT |
735 | }; |
736 | ||
62322d25 | 737 | static const struct file_operations kmem_fops = { |
1da177e4 LT |
738 | .llseek = memory_lseek, |
739 | .read = read_kmem, | |
740 | .write = write_kmem, | |
741 | .mmap = mmap_kmem, | |
742 | .open = open_kmem, | |
5da6185b | 743 | .get_unmapped_area = get_unmapped_area_mem, |
1da177e4 LT |
744 | }; |
745 | ||
62322d25 | 746 | static const struct file_operations null_fops = { |
1da177e4 LT |
747 | .llseek = null_lseek, |
748 | .read = read_null, | |
749 | .write = write_null, | |
1ebd32fc | 750 | .splice_write = splice_write_null, |
1da177e4 LT |
751 | }; |
752 | ||
4f911d64 | 753 | #ifdef CONFIG_DEVPORT |
62322d25 | 754 | static const struct file_operations port_fops = { |
1da177e4 LT |
755 | .llseek = memory_lseek, |
756 | .read = read_port, | |
757 | .write = write_port, | |
758 | .open = open_port, | |
759 | }; | |
760 | #endif | |
761 | ||
62322d25 | 762 | static const struct file_operations zero_fops = { |
1da177e4 LT |
763 | .llseek = zero_lseek, |
764 | .read = read_zero, | |
765 | .write = write_zero, | |
766 | .mmap = mmap_zero, | |
767 | }; | |
768 | ||
5da6185b DH |
769 | /* |
770 | * capabilities for /dev/zero | |
771 | * - permits private mappings, "copies" are taken of the source of zeros | |
772 | */ | |
1da177e4 LT |
773 | static struct backing_dev_info zero_bdi = { |
774 | .capabilities = BDI_CAP_MAP_COPY, | |
775 | }; | |
776 | ||
62322d25 | 777 | static const struct file_operations full_fops = { |
1da177e4 LT |
778 | .llseek = full_lseek, |
779 | .read = read_full, | |
780 | .write = write_full, | |
781 | }; | |
782 | ||
50b1fdbd | 783 | #ifdef CONFIG_CRASH_DUMP |
62322d25 | 784 | static const struct file_operations oldmem_fops = { |
50b1fdbd VG |
785 | .read = read_oldmem, |
786 | .open = open_oldmem, | |
787 | }; | |
788 | #endif | |
789 | ||
1da177e4 LT |
790 | static ssize_t kmsg_write(struct file * file, const char __user * buf, |
791 | size_t count, loff_t *ppos) | |
792 | { | |
793 | char *tmp; | |
cd140a5c | 794 | ssize_t ret; |
1da177e4 LT |
795 | |
796 | tmp = kmalloc(count + 1, GFP_KERNEL); | |
797 | if (tmp == NULL) | |
798 | return -ENOMEM; | |
799 | ret = -EFAULT; | |
800 | if (!copy_from_user(tmp, buf, count)) { | |
801 | tmp[count] = 0; | |
802 | ret = printk("%s", tmp); | |
cd140a5c GC |
803 | if (ret > count) |
804 | /* printk can add a prefix */ | |
805 | ret = count; | |
1da177e4 LT |
806 | } |
807 | kfree(tmp); | |
808 | return ret; | |
809 | } | |
810 | ||
62322d25 | 811 | static const struct file_operations kmsg_fops = { |
1da177e4 LT |
812 | .write = kmsg_write, |
813 | }; | |
814 | ||
815 | static int memory_open(struct inode * inode, struct file * filp) | |
816 | { | |
817 | switch (iminor(inode)) { | |
818 | case 1: | |
819 | filp->f_op = &mem_fops; | |
5da6185b DH |
820 | filp->f_mapping->backing_dev_info = |
821 | &directly_mappable_cdev_bdi; | |
1da177e4 LT |
822 | break; |
823 | case 2: | |
824 | filp->f_op = &kmem_fops; | |
5da6185b DH |
825 | filp->f_mapping->backing_dev_info = |
826 | &directly_mappable_cdev_bdi; | |
1da177e4 LT |
827 | break; |
828 | case 3: | |
829 | filp->f_op = &null_fops; | |
830 | break; | |
4f911d64 | 831 | #ifdef CONFIG_DEVPORT |
1da177e4 LT |
832 | case 4: |
833 | filp->f_op = &port_fops; | |
834 | break; | |
835 | #endif | |
836 | case 5: | |
837 | filp->f_mapping->backing_dev_info = &zero_bdi; | |
838 | filp->f_op = &zero_fops; | |
839 | break; | |
840 | case 7: | |
841 | filp->f_op = &full_fops; | |
842 | break; | |
843 | case 8: | |
844 | filp->f_op = &random_fops; | |
845 | break; | |
846 | case 9: | |
847 | filp->f_op = &urandom_fops; | |
848 | break; | |
849 | case 11: | |
850 | filp->f_op = &kmsg_fops; | |
851 | break; | |
50b1fdbd VG |
852 | #ifdef CONFIG_CRASH_DUMP |
853 | case 12: | |
854 | filp->f_op = &oldmem_fops; | |
855 | break; | |
856 | #endif | |
1da177e4 LT |
857 | default: |
858 | return -ENXIO; | |
859 | } | |
860 | if (filp->f_op && filp->f_op->open) | |
861 | return filp->f_op->open(inode,filp); | |
862 | return 0; | |
863 | } | |
864 | ||
62322d25 | 865 | static const struct file_operations memory_fops = { |
1da177e4 LT |
866 | .open = memory_open, /* just a selector for the real open */ |
867 | }; | |
868 | ||
869 | static const struct { | |
870 | unsigned int minor; | |
871 | char *name; | |
872 | umode_t mode; | |
99ac48f5 | 873 | const struct file_operations *fops; |
1da177e4 LT |
874 | } devlist[] = { /* list of minor devices */ |
875 | {1, "mem", S_IRUSR | S_IWUSR | S_IRGRP, &mem_fops}, | |
876 | {2, "kmem", S_IRUSR | S_IWUSR | S_IRGRP, &kmem_fops}, | |
877 | {3, "null", S_IRUGO | S_IWUGO, &null_fops}, | |
4f911d64 | 878 | #ifdef CONFIG_DEVPORT |
1da177e4 LT |
879 | {4, "port", S_IRUSR | S_IWUSR | S_IRGRP, &port_fops}, |
880 | #endif | |
881 | {5, "zero", S_IRUGO | S_IWUGO, &zero_fops}, | |
882 | {7, "full", S_IRUGO | S_IWUGO, &full_fops}, | |
883 | {8, "random", S_IRUGO | S_IWUSR, &random_fops}, | |
884 | {9, "urandom", S_IRUGO | S_IWUSR, &urandom_fops}, | |
885 | {11,"kmsg", S_IRUGO | S_IWUSR, &kmsg_fops}, | |
50b1fdbd VG |
886 | #ifdef CONFIG_CRASH_DUMP |
887 | {12,"oldmem", S_IRUSR | S_IWUSR | S_IRGRP, &oldmem_fops}, | |
888 | #endif | |
1da177e4 LT |
889 | }; |
890 | ||
ca8eca68 | 891 | static struct class *mem_class; |
1da177e4 LT |
892 | |
893 | static int __init chr_dev_init(void) | |
894 | { | |
895 | int i; | |
e0bf68dd PZ |
896 | int err; |
897 | ||
898 | err = bdi_init(&zero_bdi); | |
899 | if (err) | |
900 | return err; | |
1da177e4 LT |
901 | |
902 | if (register_chrdev(MEM_MAJOR,"mem",&memory_fops)) | |
903 | printk("unable to get major %d for memory devs\n", MEM_MAJOR); | |
904 | ||
ca8eca68 | 905 | mem_class = class_create(THIS_MODULE, "mem"); |
7c69ef79 | 906 | for (i = 0; i < ARRAY_SIZE(devlist); i++) |
ebf644c4 GKH |
907 | device_create(mem_class, NULL, |
908 | MKDEV(MEM_MAJOR, devlist[i].minor), | |
909 | devlist[i].name); | |
910 | ||
1da177e4 LT |
911 | return 0; |
912 | } | |
913 | ||
914 | fs_initcall(chr_dev_init); |