]>
Commit | Line | Data |
---|---|---|
1 | #include <linux/capability.h> | |
2 | #include <linux/blkdev.h> | |
3 | #include <linux/export.h> | |
4 | #include <linux/gfp.h> | |
5 | #include <linux/blkpg.h> | |
6 | #include <linux/hdreg.h> | |
7 | #include <linux/backing-dev.h> | |
8 | #include <linux/fs.h> | |
9 | #include <linux/blktrace_api.h> | |
10 | #include <asm/uaccess.h> | |
11 | ||
12 | static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user *arg) | |
13 | { | |
14 | struct block_device *bdevp; | |
15 | struct gendisk *disk; | |
16 | struct hd_struct *part, *lpart; | |
17 | struct blkpg_ioctl_arg a; | |
18 | struct blkpg_partition p; | |
19 | struct disk_part_iter piter; | |
20 | long long start, length; | |
21 | int partno; | |
22 | ||
23 | if (!capable(CAP_SYS_ADMIN)) | |
24 | return -EACCES; | |
25 | if (copy_from_user(&a, arg, sizeof(struct blkpg_ioctl_arg))) | |
26 | return -EFAULT; | |
27 | if (copy_from_user(&p, a.data, sizeof(struct blkpg_partition))) | |
28 | return -EFAULT; | |
29 | disk = bdev->bd_disk; | |
30 | if (bdev != bdev->bd_contains) | |
31 | return -EINVAL; | |
32 | partno = p.pno; | |
33 | if (partno <= 0) | |
34 | return -EINVAL; | |
35 | switch (a.op) { | |
36 | case BLKPG_ADD_PARTITION: | |
37 | start = p.start >> 9; | |
38 | length = p.length >> 9; | |
39 | /* check for fit in a hd_struct */ | |
40 | if (sizeof(sector_t) == sizeof(long) && | |
41 | sizeof(long long) > sizeof(long)) { | |
42 | long pstart = start, plength = length; | |
43 | if (pstart != start || plength != length | |
44 | || pstart < 0 || plength < 0 || partno > 65535) | |
45 | return -EINVAL; | |
46 | } | |
47 | ||
48 | mutex_lock(&bdev->bd_mutex); | |
49 | ||
50 | /* overlap? */ | |
51 | disk_part_iter_init(&piter, disk, | |
52 | DISK_PITER_INCL_EMPTY); | |
53 | while ((part = disk_part_iter_next(&piter))) { | |
54 | if (!(start + length <= part->start_sect || | |
55 | start >= part->start_sect + part->nr_sects)) { | |
56 | disk_part_iter_exit(&piter); | |
57 | mutex_unlock(&bdev->bd_mutex); | |
58 | return -EBUSY; | |
59 | } | |
60 | } | |
61 | disk_part_iter_exit(&piter); | |
62 | ||
63 | /* all seems OK */ | |
64 | part = add_partition(disk, partno, start, length, | |
65 | ADDPART_FLAG_NONE, NULL); | |
66 | mutex_unlock(&bdev->bd_mutex); | |
67 | return PTR_ERR_OR_ZERO(part); | |
68 | case BLKPG_DEL_PARTITION: | |
69 | part = disk_get_part(disk, partno); | |
70 | if (!part) | |
71 | return -ENXIO; | |
72 | ||
73 | bdevp = bdget(part_devt(part)); | |
74 | disk_put_part(part); | |
75 | if (!bdevp) | |
76 | return -ENOMEM; | |
77 | ||
78 | mutex_lock(&bdevp->bd_mutex); | |
79 | if (bdevp->bd_openers) { | |
80 | mutex_unlock(&bdevp->bd_mutex); | |
81 | bdput(bdevp); | |
82 | return -EBUSY; | |
83 | } | |
84 | /* all seems OK */ | |
85 | fsync_bdev(bdevp); | |
86 | invalidate_bdev(bdevp); | |
87 | ||
88 | mutex_lock_nested(&bdev->bd_mutex, 1); | |
89 | delete_partition(disk, partno); | |
90 | mutex_unlock(&bdev->bd_mutex); | |
91 | mutex_unlock(&bdevp->bd_mutex); | |
92 | bdput(bdevp); | |
93 | ||
94 | return 0; | |
95 | case BLKPG_RESIZE_PARTITION: | |
96 | start = p.start >> 9; | |
97 | /* new length of partition in bytes */ | |
98 | length = p.length >> 9; | |
99 | /* check for fit in a hd_struct */ | |
100 | if (sizeof(sector_t) == sizeof(long) && | |
101 | sizeof(long long) > sizeof(long)) { | |
102 | long pstart = start, plength = length; | |
103 | if (pstart != start || plength != length | |
104 | || pstart < 0 || plength < 0) | |
105 | return -EINVAL; | |
106 | } | |
107 | part = disk_get_part(disk, partno); | |
108 | if (!part) | |
109 | return -ENXIO; | |
110 | bdevp = bdget(part_devt(part)); | |
111 | if (!bdevp) { | |
112 | disk_put_part(part); | |
113 | return -ENOMEM; | |
114 | } | |
115 | mutex_lock(&bdevp->bd_mutex); | |
116 | mutex_lock_nested(&bdev->bd_mutex, 1); | |
117 | if (start != part->start_sect) { | |
118 | mutex_unlock(&bdevp->bd_mutex); | |
119 | mutex_unlock(&bdev->bd_mutex); | |
120 | bdput(bdevp); | |
121 | disk_put_part(part); | |
122 | return -EINVAL; | |
123 | } | |
124 | /* overlap? */ | |
125 | disk_part_iter_init(&piter, disk, | |
126 | DISK_PITER_INCL_EMPTY); | |
127 | while ((lpart = disk_part_iter_next(&piter))) { | |
128 | if (lpart->partno != partno && | |
129 | !(start + length <= lpart->start_sect || | |
130 | start >= lpart->start_sect + lpart->nr_sects) | |
131 | ) { | |
132 | disk_part_iter_exit(&piter); | |
133 | mutex_unlock(&bdevp->bd_mutex); | |
134 | mutex_unlock(&bdev->bd_mutex); | |
135 | bdput(bdevp); | |
136 | disk_put_part(part); | |
137 | return -EBUSY; | |
138 | } | |
139 | } | |
140 | disk_part_iter_exit(&piter); | |
141 | part_nr_sects_write(part, (sector_t)length); | |
142 | i_size_write(bdevp->bd_inode, p.length); | |
143 | mutex_unlock(&bdevp->bd_mutex); | |
144 | mutex_unlock(&bdev->bd_mutex); | |
145 | bdput(bdevp); | |
146 | disk_put_part(part); | |
147 | return 0; | |
148 | default: | |
149 | return -EINVAL; | |
150 | } | |
151 | } | |
152 | ||
153 | /* | |
154 | * This is an exported API for the block driver, and will not | |
155 | * acquire bd_mutex. This API should be used in case that | |
156 | * caller has held bd_mutex already. | |
157 | */ | |
158 | int __blkdev_reread_part(struct block_device *bdev) | |
159 | { | |
160 | struct gendisk *disk = bdev->bd_disk; | |
161 | ||
162 | if (!disk_part_scan_enabled(disk) || bdev != bdev->bd_contains) | |
163 | return -EINVAL; | |
164 | if (!capable(CAP_SYS_ADMIN)) | |
165 | return -EACCES; | |
166 | ||
167 | lockdep_assert_held(&bdev->bd_mutex); | |
168 | ||
169 | return rescan_partitions(disk, bdev); | |
170 | } | |
171 | EXPORT_SYMBOL(__blkdev_reread_part); | |
172 | ||
173 | /* | |
174 | * This is an exported API for the block driver, and will | |
175 | * try to acquire bd_mutex. If bd_mutex has been held already | |
176 | * in current context, please call __blkdev_reread_part(). | |
177 | */ | |
178 | int blkdev_reread_part(struct block_device *bdev) | |
179 | { | |
180 | int res; | |
181 | ||
182 | if (!mutex_trylock(&bdev->bd_mutex)) | |
183 | return -EBUSY; | |
184 | res = __blkdev_reread_part(bdev); | |
185 | mutex_unlock(&bdev->bd_mutex); | |
186 | ||
187 | return res; | |
188 | } | |
189 | EXPORT_SYMBOL(blkdev_reread_part); | |
190 | ||
191 | static int blk_ioctl_discard(struct block_device *bdev, uint64_t start, | |
192 | uint64_t len, int secure) | |
193 | { | |
194 | unsigned long flags = 0; | |
195 | ||
196 | if (start & 511) | |
197 | return -EINVAL; | |
198 | if (len & 511) | |
199 | return -EINVAL; | |
200 | start >>= 9; | |
201 | len >>= 9; | |
202 | ||
203 | if (start + len > (i_size_read(bdev->bd_inode) >> 9)) | |
204 | return -EINVAL; | |
205 | if (secure) | |
206 | flags |= BLKDEV_DISCARD_SECURE; | |
207 | return blkdev_issue_discard(bdev, start, len, GFP_KERNEL, flags); | |
208 | } | |
209 | ||
210 | static int blk_ioctl_zeroout(struct block_device *bdev, uint64_t start, | |
211 | uint64_t len) | |
212 | { | |
213 | if (start & 511) | |
214 | return -EINVAL; | |
215 | if (len & 511) | |
216 | return -EINVAL; | |
217 | start >>= 9; | |
218 | len >>= 9; | |
219 | ||
220 | if (start + len > (i_size_read(bdev->bd_inode) >> 9)) | |
221 | return -EINVAL; | |
222 | ||
223 | return blkdev_issue_zeroout(bdev, start, len, GFP_KERNEL, false); | |
224 | } | |
225 | ||
226 | static int put_ushort(unsigned long arg, unsigned short val) | |
227 | { | |
228 | return put_user(val, (unsigned short __user *)arg); | |
229 | } | |
230 | ||
231 | static int put_int(unsigned long arg, int val) | |
232 | { | |
233 | return put_user(val, (int __user *)arg); | |
234 | } | |
235 | ||
236 | static int put_uint(unsigned long arg, unsigned int val) | |
237 | { | |
238 | return put_user(val, (unsigned int __user *)arg); | |
239 | } | |
240 | ||
241 | static int put_long(unsigned long arg, long val) | |
242 | { | |
243 | return put_user(val, (long __user *)arg); | |
244 | } | |
245 | ||
246 | static int put_ulong(unsigned long arg, unsigned long val) | |
247 | { | |
248 | return put_user(val, (unsigned long __user *)arg); | |
249 | } | |
250 | ||
251 | static int put_u64(unsigned long arg, u64 val) | |
252 | { | |
253 | return put_user(val, (u64 __user *)arg); | |
254 | } | |
255 | ||
256 | int __blkdev_driver_ioctl(struct block_device *bdev, fmode_t mode, | |
257 | unsigned cmd, unsigned long arg) | |
258 | { | |
259 | struct gendisk *disk = bdev->bd_disk; | |
260 | ||
261 | if (disk->fops->ioctl) | |
262 | return disk->fops->ioctl(bdev, mode, cmd, arg); | |
263 | ||
264 | return -ENOTTY; | |
265 | } | |
266 | /* | |
267 | * For the record: _GPL here is only because somebody decided to slap it | |
268 | * on the previous export. Sheer idiocy, since it wasn't copyrightable | |
269 | * at all and could be open-coded without any exports by anybody who cares. | |
270 | */ | |
271 | EXPORT_SYMBOL_GPL(__blkdev_driver_ioctl); | |
272 | ||
273 | /* | |
274 | * Is it an unrecognized ioctl? The correct returns are either | |
275 | * ENOTTY (final) or ENOIOCTLCMD ("I don't know this one, try a | |
276 | * fallback"). ENOIOCTLCMD gets turned into ENOTTY by the ioctl | |
277 | * code before returning. | |
278 | * | |
279 | * Confused drivers sometimes return EINVAL, which is wrong. It | |
280 | * means "I understood the ioctl command, but the parameters to | |
281 | * it were wrong". | |
282 | * | |
283 | * We should aim to just fix the broken drivers, the EINVAL case | |
284 | * should go away. | |
285 | */ | |
286 | static inline int is_unrecognized_ioctl(int ret) | |
287 | { | |
288 | return ret == -EINVAL || | |
289 | ret == -ENOTTY || | |
290 | ret == -ENOIOCTLCMD; | |
291 | } | |
292 | ||
293 | /* | |
294 | * always keep this in sync with compat_blkdev_ioctl() | |
295 | */ | |
296 | int blkdev_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd, | |
297 | unsigned long arg) | |
298 | { | |
299 | struct gendisk *disk = bdev->bd_disk; | |
300 | struct backing_dev_info *bdi; | |
301 | loff_t size; | |
302 | int ret, n; | |
303 | unsigned int max_sectors; | |
304 | ||
305 | switch(cmd) { | |
306 | case BLKFLSBUF: | |
307 | if (!capable(CAP_SYS_ADMIN)) | |
308 | return -EACCES; | |
309 | ||
310 | ret = __blkdev_driver_ioctl(bdev, mode, cmd, arg); | |
311 | if (!is_unrecognized_ioctl(ret)) | |
312 | return ret; | |
313 | ||
314 | fsync_bdev(bdev); | |
315 | invalidate_bdev(bdev); | |
316 | return 0; | |
317 | ||
318 | case BLKROSET: | |
319 | ret = __blkdev_driver_ioctl(bdev, mode, cmd, arg); | |
320 | if (!is_unrecognized_ioctl(ret)) | |
321 | return ret; | |
322 | if (!capable(CAP_SYS_ADMIN)) | |
323 | return -EACCES; | |
324 | if (get_user(n, (int __user *)(arg))) | |
325 | return -EFAULT; | |
326 | set_device_ro(bdev, n); | |
327 | return 0; | |
328 | ||
329 | case BLKDISCARD: | |
330 | case BLKSECDISCARD: { | |
331 | uint64_t range[2]; | |
332 | ||
333 | if (!(mode & FMODE_WRITE)) | |
334 | return -EBADF; | |
335 | ||
336 | if (copy_from_user(range, (void __user *)arg, sizeof(range))) | |
337 | return -EFAULT; | |
338 | ||
339 | return blk_ioctl_discard(bdev, range[0], range[1], | |
340 | cmd == BLKSECDISCARD); | |
341 | } | |
342 | case BLKZEROOUT: { | |
343 | uint64_t range[2]; | |
344 | ||
345 | if (!(mode & FMODE_WRITE)) | |
346 | return -EBADF; | |
347 | ||
348 | if (copy_from_user(range, (void __user *)arg, sizeof(range))) | |
349 | return -EFAULT; | |
350 | ||
351 | return blk_ioctl_zeroout(bdev, range[0], range[1]); | |
352 | } | |
353 | ||
354 | case HDIO_GETGEO: { | |
355 | struct hd_geometry geo; | |
356 | ||
357 | if (!arg) | |
358 | return -EINVAL; | |
359 | if (!disk->fops->getgeo) | |
360 | return -ENOTTY; | |
361 | ||
362 | /* | |
363 | * We need to set the startsect first, the driver may | |
364 | * want to override it. | |
365 | */ | |
366 | memset(&geo, 0, sizeof(geo)); | |
367 | geo.start = get_start_sect(bdev); | |
368 | ret = disk->fops->getgeo(bdev, &geo); | |
369 | if (ret) | |
370 | return ret; | |
371 | if (copy_to_user((struct hd_geometry __user *)arg, &geo, | |
372 | sizeof(geo))) | |
373 | return -EFAULT; | |
374 | return 0; | |
375 | } | |
376 | case BLKRAGET: | |
377 | case BLKFRAGET: | |
378 | if (!arg) | |
379 | return -EINVAL; | |
380 | bdi = blk_get_backing_dev_info(bdev); | |
381 | return put_long(arg, (bdi->ra_pages * PAGE_CACHE_SIZE) / 512); | |
382 | case BLKROGET: | |
383 | return put_int(arg, bdev_read_only(bdev) != 0); | |
384 | case BLKBSZGET: /* get block device soft block size (cf. BLKSSZGET) */ | |
385 | return put_int(arg, block_size(bdev)); | |
386 | case BLKSSZGET: /* get block device logical block size */ | |
387 | return put_int(arg, bdev_logical_block_size(bdev)); | |
388 | case BLKPBSZGET: /* get block device physical block size */ | |
389 | return put_uint(arg, bdev_physical_block_size(bdev)); | |
390 | case BLKIOMIN: | |
391 | return put_uint(arg, bdev_io_min(bdev)); | |
392 | case BLKIOOPT: | |
393 | return put_uint(arg, bdev_io_opt(bdev)); | |
394 | case BLKALIGNOFF: | |
395 | return put_int(arg, bdev_alignment_offset(bdev)); | |
396 | case BLKDISCARDZEROES: | |
397 | return put_uint(arg, bdev_discard_zeroes_data(bdev)); | |
398 | case BLKSECTGET: | |
399 | max_sectors = min_t(unsigned int, USHRT_MAX, | |
400 | queue_max_sectors(bdev_get_queue(bdev))); | |
401 | return put_ushort(arg, max_sectors); | |
402 | case BLKROTATIONAL: | |
403 | return put_ushort(arg, !blk_queue_nonrot(bdev_get_queue(bdev))); | |
404 | case BLKRASET: | |
405 | case BLKFRASET: | |
406 | if(!capable(CAP_SYS_ADMIN)) | |
407 | return -EACCES; | |
408 | bdi = blk_get_backing_dev_info(bdev); | |
409 | bdi->ra_pages = (arg * 512) / PAGE_CACHE_SIZE; | |
410 | return 0; | |
411 | case BLKBSZSET: | |
412 | /* set the logical block size */ | |
413 | if (!capable(CAP_SYS_ADMIN)) | |
414 | return -EACCES; | |
415 | if (!arg) | |
416 | return -EINVAL; | |
417 | if (get_user(n, (int __user *) arg)) | |
418 | return -EFAULT; | |
419 | if (!(mode & FMODE_EXCL)) { | |
420 | bdgrab(bdev); | |
421 | if (blkdev_get(bdev, mode | FMODE_EXCL, &bdev) < 0) | |
422 | return -EBUSY; | |
423 | } | |
424 | ret = set_blocksize(bdev, n); | |
425 | if (!(mode & FMODE_EXCL)) | |
426 | blkdev_put(bdev, mode | FMODE_EXCL); | |
427 | return ret; | |
428 | case BLKPG: | |
429 | ret = blkpg_ioctl(bdev, (struct blkpg_ioctl_arg __user *) arg); | |
430 | break; | |
431 | case BLKRRPART: | |
432 | ret = blkdev_reread_part(bdev); | |
433 | break; | |
434 | case BLKGETSIZE: | |
435 | size = i_size_read(bdev->bd_inode); | |
436 | if ((size >> 9) > ~0UL) | |
437 | return -EFBIG; | |
438 | return put_ulong(arg, size >> 9); | |
439 | case BLKGETSIZE64: | |
440 | return put_u64(arg, i_size_read(bdev->bd_inode)); | |
441 | case BLKTRACESTART: | |
442 | case BLKTRACESTOP: | |
443 | case BLKTRACESETUP: | |
444 | case BLKTRACETEARDOWN: | |
445 | ret = blk_trace_ioctl(bdev, cmd, (char __user *) arg); | |
446 | break; | |
447 | default: | |
448 | ret = __blkdev_driver_ioctl(bdev, mode, cmd, arg); | |
449 | } | |
450 | return ret; | |
451 | } | |
452 | EXPORT_SYMBOL_GPL(blkdev_ioctl); |