xref: /linux/fs/bcachefs/chardev.c (revision 566ab427f827b0256d3e8ce0235d088e6a9c28bd)
1 // SPDX-License-Identifier: GPL-2.0
2 #ifndef NO_BCACHEFS_CHARDEV
3 
4 #include "bcachefs.h"
5 #include "bcachefs_ioctl.h"
6 #include "buckets.h"
7 #include "chardev.h"
8 #include "disk_accounting.h"
9 #include "journal.h"
10 #include "move.h"
11 #include "recovery_passes.h"
12 #include "replicas.h"
13 #include "super.h"
14 #include "super-io.h"
15 #include "thread_with_file.h"
16 
17 #include <linux/cdev.h>
18 #include <linux/device.h>
19 #include <linux/fs.h>
20 #include <linux/ioctl.h>
21 #include <linux/major.h>
22 #include <linux/sched/task.h>
23 #include <linux/slab.h>
24 #include <linux/uaccess.h>
25 
26 /* returns with ref on ca->ref */
27 static struct bch_dev *bch2_device_lookup(struct bch_fs *c, u64 dev,
28 					  unsigned flags)
29 {
30 	struct bch_dev *ca;
31 
32 	if (flags & BCH_BY_INDEX) {
33 		if (dev >= c->sb.nr_devices)
34 			return ERR_PTR(-EINVAL);
35 
36 		ca = bch2_dev_tryget_noerror(c, dev);
37 		if (!ca)
38 			return ERR_PTR(-EINVAL);
39 	} else {
40 		char *path;
41 
42 		path = strndup_user((const char __user *)
43 				    (unsigned long) dev, PATH_MAX);
44 		if (IS_ERR(path))
45 			return ERR_CAST(path);
46 
47 		ca = bch2_dev_lookup(c, path);
48 		kfree(path);
49 	}
50 
51 	return ca;
52 }
53 
54 #if 0
55 static long bch2_ioctl_assemble(struct bch_ioctl_assemble __user *user_arg)
56 {
57 	struct bch_ioctl_assemble arg;
58 	struct bch_fs *c;
59 	u64 *user_devs = NULL;
60 	char **devs = NULL;
61 	unsigned i;
62 	int ret = -EFAULT;
63 
64 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
65 		return -EFAULT;
66 
67 	if (arg.flags || arg.pad)
68 		return -EINVAL;
69 
70 	user_devs = kmalloc_array(arg.nr_devs, sizeof(u64), GFP_KERNEL);
71 	if (!user_devs)
72 		return -ENOMEM;
73 
74 	devs = kcalloc(arg.nr_devs, sizeof(char *), GFP_KERNEL);
75 
76 	if (copy_from_user(user_devs, user_arg->devs,
77 			   sizeof(u64) * arg.nr_devs))
78 		goto err;
79 
80 	for (i = 0; i < arg.nr_devs; i++) {
81 		devs[i] = strndup_user((const char __user *)(unsigned long)
82 				       user_devs[i],
83 				       PATH_MAX);
84 		ret= PTR_ERR_OR_ZERO(devs[i]);
85 		if (ret)
86 			goto err;
87 	}
88 
89 	c = bch2_fs_open(devs, arg.nr_devs, bch2_opts_empty());
90 	ret = PTR_ERR_OR_ZERO(c);
91 	if (!ret)
92 		closure_put(&c->cl);
93 err:
94 	if (devs)
95 		for (i = 0; i < arg.nr_devs; i++)
96 			kfree(devs[i]);
97 	kfree(devs);
98 	return ret;
99 }
100 
101 static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg)
102 {
103 	struct bch_ioctl_incremental arg;
104 	const char *err;
105 	char *path;
106 
107 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
108 		return -EFAULT;
109 
110 	if (arg.flags || arg.pad)
111 		return -EINVAL;
112 
113 	path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
114 	ret = PTR_ERR_OR_ZERO(path);
115 	if (ret)
116 		return ret;
117 
118 	err = bch2_fs_open_incremental(path);
119 	kfree(path);
120 
121 	if (err) {
122 		pr_err("Could not register bcachefs devices: %s", err);
123 		return -EINVAL;
124 	}
125 
126 	return 0;
127 }
128 #endif
129 
130 struct fsck_thread {
131 	struct thread_with_stdio thr;
132 	struct bch_fs		*c;
133 	struct bch_opts		opts;
134 };
135 
136 static void bch2_fsck_thread_exit(struct thread_with_stdio *_thr)
137 {
138 	struct fsck_thread *thr = container_of(_thr, struct fsck_thread, thr);
139 	kfree(thr);
140 }
141 
142 static int bch2_fsck_offline_thread_fn(struct thread_with_stdio *stdio)
143 {
144 	struct fsck_thread *thr = container_of(stdio, struct fsck_thread, thr);
145 	struct bch_fs *c = thr->c;
146 
147 	int ret = PTR_ERR_OR_ZERO(c);
148 	if (ret)
149 		return ret;
150 
151 	ret = bch2_fs_start(thr->c);
152 	if (ret)
153 		goto err;
154 
155 	if (test_bit(BCH_FS_errors_fixed, &c->flags)) {
156 		bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: errors fixed\n", c->name);
157 		ret |= 1;
158 	}
159 	if (test_bit(BCH_FS_error, &c->flags)) {
160 		bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: still has errors\n", c->name);
161 		ret |= 4;
162 	}
163 err:
164 	bch2_fs_stop(c);
165 	return ret;
166 }
167 
168 static const struct thread_with_stdio_ops bch2_offline_fsck_ops = {
169 	.exit		= bch2_fsck_thread_exit,
170 	.fn		= bch2_fsck_offline_thread_fn,
171 };
172 
173 static long bch2_ioctl_fsck_offline(struct bch_ioctl_fsck_offline __user *user_arg)
174 {
175 	struct bch_ioctl_fsck_offline arg;
176 	struct fsck_thread *thr = NULL;
177 	darray_str(devs) = {};
178 	long ret = 0;
179 
180 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
181 		return -EFAULT;
182 
183 	if (arg.flags)
184 		return -EINVAL;
185 
186 	if (!capable(CAP_SYS_ADMIN))
187 		return -EPERM;
188 
189 	for (size_t i = 0; i < arg.nr_devs; i++) {
190 		u64 dev_u64;
191 		ret = copy_from_user_errcode(&dev_u64, &user_arg->devs[i], sizeof(u64));
192 		if (ret)
193 			goto err;
194 
195 		char *dev_str = strndup_user((char __user *)(unsigned long) dev_u64, PATH_MAX);
196 		ret = PTR_ERR_OR_ZERO(dev_str);
197 		if (ret)
198 			goto err;
199 
200 		ret = darray_push(&devs, dev_str);
201 		if (ret) {
202 			kfree(dev_str);
203 			goto err;
204 		}
205 	}
206 
207 	thr = kzalloc(sizeof(*thr), GFP_KERNEL);
208 	if (!thr) {
209 		ret = -ENOMEM;
210 		goto err;
211 	}
212 
213 	thr->opts = bch2_opts_empty();
214 
215 	if (arg.opts) {
216 		char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
217 		ret =   PTR_ERR_OR_ZERO(optstr) ?:
218 			bch2_parse_mount_opts(NULL, &thr->opts, NULL, optstr);
219 		if (!IS_ERR(optstr))
220 			kfree(optstr);
221 
222 		if (ret)
223 			goto err;
224 	}
225 
226 	opt_set(thr->opts, stdio, (u64)(unsigned long)&thr->thr.stdio);
227 	opt_set(thr->opts, read_only, 1);
228 
229 	/* We need request_key() to be called before we punt to kthread: */
230 	opt_set(thr->opts, nostart, true);
231 
232 	bch2_thread_with_stdio_init(&thr->thr, &bch2_offline_fsck_ops);
233 
234 	thr->c = bch2_fs_open(devs.data, arg.nr_devs, thr->opts);
235 
236 	if (!IS_ERR(thr->c) &&
237 	    thr->c->opts.errors == BCH_ON_ERROR_panic)
238 		thr->c->opts.errors = BCH_ON_ERROR_ro;
239 
240 	ret = __bch2_run_thread_with_stdio(&thr->thr);
241 out:
242 	darray_for_each(devs, i)
243 		kfree(*i);
244 	darray_exit(&devs);
245 	return ret;
246 err:
247 	if (thr)
248 		bch2_fsck_thread_exit(&thr->thr);
249 	pr_err("ret %s", bch2_err_str(ret));
250 	goto out;
251 }
252 
253 static long bch2_global_ioctl(unsigned cmd, void __user *arg)
254 {
255 	long ret;
256 
257 	switch (cmd) {
258 #if 0
259 	case BCH_IOCTL_ASSEMBLE:
260 		return bch2_ioctl_assemble(arg);
261 	case BCH_IOCTL_INCREMENTAL:
262 		return bch2_ioctl_incremental(arg);
263 #endif
264 	case BCH_IOCTL_FSCK_OFFLINE: {
265 		ret = bch2_ioctl_fsck_offline(arg);
266 		break;
267 	}
268 	default:
269 		ret = -ENOTTY;
270 		break;
271 	}
272 
273 	if (ret < 0)
274 		ret = bch2_err_class(ret);
275 	return ret;
276 }
277 
278 static long bch2_ioctl_query_uuid(struct bch_fs *c,
279 			struct bch_ioctl_query_uuid __user *user_arg)
280 {
281 	return copy_to_user_errcode(&user_arg->uuid, &c->sb.user_uuid,
282 				    sizeof(c->sb.user_uuid));
283 }
284 
285 #if 0
286 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
287 {
288 	if (!capable(CAP_SYS_ADMIN))
289 		return -EPERM;
290 
291 	if (arg.flags || arg.pad)
292 		return -EINVAL;
293 
294 	return bch2_fs_start(c);
295 }
296 
297 static long bch2_ioctl_stop(struct bch_fs *c)
298 {
299 	if (!capable(CAP_SYS_ADMIN))
300 		return -EPERM;
301 
302 	bch2_fs_stop(c);
303 	return 0;
304 }
305 #endif
306 
307 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
308 {
309 	char *path;
310 	int ret;
311 
312 	if (!capable(CAP_SYS_ADMIN))
313 		return -EPERM;
314 
315 	if (arg.flags || arg.pad)
316 		return -EINVAL;
317 
318 	path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
319 	ret = PTR_ERR_OR_ZERO(path);
320 	if (ret)
321 		return ret;
322 
323 	ret = bch2_dev_add(c, path);
324 	if (!IS_ERR(path))
325 		kfree(path);
326 
327 	return ret;
328 }
329 
330 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
331 {
332 	struct bch_dev *ca;
333 
334 	if (!capable(CAP_SYS_ADMIN))
335 		return -EPERM;
336 
337 	if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
338 			   BCH_FORCE_IF_METADATA_LOST|
339 			   BCH_FORCE_IF_DEGRADED|
340 			   BCH_BY_INDEX)) ||
341 	    arg.pad)
342 		return -EINVAL;
343 
344 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
345 	if (IS_ERR(ca))
346 		return PTR_ERR(ca);
347 
348 	return bch2_dev_remove(c, ca, arg.flags);
349 }
350 
351 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
352 {
353 	char *path;
354 	int ret;
355 
356 	if (!capable(CAP_SYS_ADMIN))
357 		return -EPERM;
358 
359 	if (arg.flags || arg.pad)
360 		return -EINVAL;
361 
362 	path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
363 	ret = PTR_ERR_OR_ZERO(path);
364 	if (ret)
365 		return ret;
366 
367 	ret = bch2_dev_online(c, path);
368 	kfree(path);
369 	return ret;
370 }
371 
372 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
373 {
374 	struct bch_dev *ca;
375 	int ret;
376 
377 	if (!capable(CAP_SYS_ADMIN))
378 		return -EPERM;
379 
380 	if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
381 			   BCH_FORCE_IF_METADATA_LOST|
382 			   BCH_FORCE_IF_DEGRADED|
383 			   BCH_BY_INDEX)) ||
384 	    arg.pad)
385 		return -EINVAL;
386 
387 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
388 	if (IS_ERR(ca))
389 		return PTR_ERR(ca);
390 
391 	ret = bch2_dev_offline(c, ca, arg.flags);
392 	bch2_dev_put(ca);
393 	return ret;
394 }
395 
396 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
397 			struct bch_ioctl_disk_set_state arg)
398 {
399 	struct bch_dev *ca;
400 	int ret;
401 
402 	if (!capable(CAP_SYS_ADMIN))
403 		return -EPERM;
404 
405 	if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
406 			   BCH_FORCE_IF_METADATA_LOST|
407 			   BCH_FORCE_IF_DEGRADED|
408 			   BCH_BY_INDEX)) ||
409 	    arg.pad[0] || arg.pad[1] || arg.pad[2] ||
410 	    arg.new_state >= BCH_MEMBER_STATE_NR)
411 		return -EINVAL;
412 
413 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
414 	if (IS_ERR(ca))
415 		return PTR_ERR(ca);
416 
417 	ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
418 	if (ret)
419 		bch_err(c, "Error setting device state: %s", bch2_err_str(ret));
420 
421 	bch2_dev_put(ca);
422 	return ret;
423 }
424 
425 struct bch_data_ctx {
426 	struct thread_with_file		thr;
427 
428 	struct bch_fs			*c;
429 	struct bch_ioctl_data		arg;
430 	struct bch_move_stats		stats;
431 };
432 
433 static int bch2_data_thread(void *arg)
434 {
435 	struct bch_data_ctx *ctx = container_of(arg, struct bch_data_ctx, thr);
436 
437 	ctx->thr.ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
438 	ctx->stats.data_type = U8_MAX;
439 	return 0;
440 }
441 
442 static int bch2_data_job_release(struct inode *inode, struct file *file)
443 {
444 	struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
445 
446 	bch2_thread_with_file_exit(&ctx->thr);
447 	kfree(ctx);
448 	return 0;
449 }
450 
451 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
452 				  size_t len, loff_t *ppos)
453 {
454 	struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
455 	struct bch_fs *c = ctx->c;
456 	struct bch_ioctl_data_event e = {
457 		.type			= BCH_DATA_EVENT_PROGRESS,
458 		.p.data_type		= ctx->stats.data_type,
459 		.p.btree_id		= ctx->stats.pos.btree,
460 		.p.pos			= ctx->stats.pos.pos,
461 		.p.sectors_done		= atomic64_read(&ctx->stats.sectors_seen),
462 		.p.sectors_total	= bch2_fs_usage_read_short(c).used,
463 	};
464 
465 	if (len < sizeof(e))
466 		return -EINVAL;
467 
468 	return copy_to_user_errcode(buf, &e, sizeof(e)) ?: sizeof(e);
469 }
470 
471 static const struct file_operations bcachefs_data_ops = {
472 	.release	= bch2_data_job_release,
473 	.read		= bch2_data_job_read,
474 };
475 
476 static long bch2_ioctl_data(struct bch_fs *c,
477 			    struct bch_ioctl_data arg)
478 {
479 	struct bch_data_ctx *ctx;
480 	int ret;
481 
482 	if (!capable(CAP_SYS_ADMIN))
483 		return -EPERM;
484 
485 	if (arg.op >= BCH_DATA_OP_NR || arg.flags)
486 		return -EINVAL;
487 
488 	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
489 	if (!ctx)
490 		return -ENOMEM;
491 
492 	ctx->c = c;
493 	ctx->arg = arg;
494 
495 	ret = bch2_run_thread_with_file(&ctx->thr,
496 			&bcachefs_data_ops,
497 			bch2_data_thread);
498 	if (ret < 0)
499 		kfree(ctx);
500 	return ret;
501 }
502 
503 static long bch2_ioctl_fs_usage(struct bch_fs *c,
504 				struct bch_ioctl_fs_usage __user *user_arg)
505 {
506 	struct bch_ioctl_fs_usage arg = {};
507 	darray_char replicas = {};
508 	u32 replica_entries_bytes;
509 	int ret = 0;
510 
511 	if (!test_bit(BCH_FS_started, &c->flags))
512 		return -EINVAL;
513 
514 	if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
515 		return -EFAULT;
516 
517 	ret   = bch2_fs_replicas_usage_read(c, &replicas) ?:
518 		(replica_entries_bytes < replicas.nr ? -ERANGE : 0) ?:
519 		copy_to_user_errcode(&user_arg->replicas, replicas.data, replicas.nr);
520 	if (ret)
521 		goto err;
522 
523 	struct bch_fs_usage_short u = bch2_fs_usage_read_short(c);
524 	arg.capacity		= c->capacity;
525 	arg.used		= u.used;
526 	arg.online_reserved	= percpu_u64_get(c->online_reserved);
527 	arg.replica_entries_bytes = replicas.nr;
528 
529 	for (unsigned i = 0; i < BCH_REPLICAS_MAX; i++) {
530 		struct disk_accounting_pos k = {
531 			.type = BCH_DISK_ACCOUNTING_persistent_reserved,
532 			.persistent_reserved.nr_replicas = i,
533 		};
534 
535 		bch2_accounting_mem_read(c,
536 					 disk_accounting_pos_to_bpos(&k),
537 					 &arg.persistent_reserved[i], 1);
538 	}
539 
540 	ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
541 err:
542 	darray_exit(&replicas);
543 	return ret;
544 }
545 
546 static long bch2_ioctl_query_accounting(struct bch_fs *c,
547 			struct bch_ioctl_query_accounting __user *user_arg)
548 {
549 	struct bch_ioctl_query_accounting arg;
550 	darray_char accounting = {};
551 	int ret = 0;
552 
553 	if (!test_bit(BCH_FS_started, &c->flags))
554 		return -EINVAL;
555 
556 	ret   = copy_from_user_errcode(&arg, user_arg, sizeof(arg)) ?:
557 		bch2_fs_accounting_read(c, &accounting, arg.accounting_types_mask) ?:
558 		(arg.accounting_u64s * sizeof(u64) < accounting.nr ? -ERANGE : 0) ?:
559 		copy_to_user_errcode(&user_arg->accounting, accounting.data, accounting.nr);
560 	if (ret)
561 		goto err;
562 
563 	arg.capacity		= c->capacity;
564 	arg.used		= bch2_fs_usage_read_short(c).used;
565 	arg.online_reserved	= percpu_u64_get(c->online_reserved);
566 	arg.accounting_u64s	= accounting.nr / sizeof(u64);
567 
568 	ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
569 err:
570 	darray_exit(&accounting);
571 	return ret;
572 }
573 
574 /* obsolete, didn't allow for new data types: */
575 static long bch2_ioctl_dev_usage(struct bch_fs *c,
576 				 struct bch_ioctl_dev_usage __user *user_arg)
577 {
578 	struct bch_ioctl_dev_usage arg;
579 	struct bch_dev_usage src;
580 	struct bch_dev *ca;
581 	unsigned i;
582 
583 	if (!test_bit(BCH_FS_started, &c->flags))
584 		return -EINVAL;
585 
586 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
587 		return -EFAULT;
588 
589 	if ((arg.flags & ~BCH_BY_INDEX) ||
590 	    arg.pad[0] ||
591 	    arg.pad[1] ||
592 	    arg.pad[2])
593 		return -EINVAL;
594 
595 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
596 	if (IS_ERR(ca))
597 		return PTR_ERR(ca);
598 
599 	src = bch2_dev_usage_read(ca);
600 
601 	arg.state		= ca->mi.state;
602 	arg.bucket_size		= ca->mi.bucket_size;
603 	arg.nr_buckets		= ca->mi.nbuckets - ca->mi.first_bucket;
604 
605 	for (i = 0; i < ARRAY_SIZE(arg.d); i++) {
606 		arg.d[i].buckets	= src.d[i].buckets;
607 		arg.d[i].sectors	= src.d[i].sectors;
608 		arg.d[i].fragmented	= src.d[i].fragmented;
609 	}
610 
611 	bch2_dev_put(ca);
612 
613 	return copy_to_user_errcode(user_arg, &arg, sizeof(arg));
614 }
615 
616 static long bch2_ioctl_dev_usage_v2(struct bch_fs *c,
617 				 struct bch_ioctl_dev_usage_v2 __user *user_arg)
618 {
619 	struct bch_ioctl_dev_usage_v2 arg;
620 	struct bch_dev_usage src;
621 	struct bch_dev *ca;
622 	int ret = 0;
623 
624 	if (!test_bit(BCH_FS_started, &c->flags))
625 		return -EINVAL;
626 
627 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
628 		return -EFAULT;
629 
630 	if ((arg.flags & ~BCH_BY_INDEX) ||
631 	    arg.pad[0] ||
632 	    arg.pad[1] ||
633 	    arg.pad[2])
634 		return -EINVAL;
635 
636 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
637 	if (IS_ERR(ca))
638 		return PTR_ERR(ca);
639 
640 	src = bch2_dev_usage_read(ca);
641 
642 	arg.state		= ca->mi.state;
643 	arg.bucket_size		= ca->mi.bucket_size;
644 	arg.nr_data_types	= min(arg.nr_data_types, BCH_DATA_NR);
645 	arg.nr_buckets		= ca->mi.nbuckets - ca->mi.first_bucket;
646 
647 	ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
648 	if (ret)
649 		goto err;
650 
651 	for (unsigned i = 0; i < arg.nr_data_types; i++) {
652 		struct bch_ioctl_dev_usage_type t = {
653 			.buckets	= src.d[i].buckets,
654 			.sectors	= src.d[i].sectors,
655 			.fragmented	= src.d[i].fragmented,
656 		};
657 
658 		ret = copy_to_user_errcode(&user_arg->d[i], &t, sizeof(t));
659 		if (ret)
660 			goto err;
661 	}
662 err:
663 	bch2_dev_put(ca);
664 	return ret;
665 }
666 
667 static long bch2_ioctl_read_super(struct bch_fs *c,
668 				  struct bch_ioctl_read_super arg)
669 {
670 	struct bch_dev *ca = NULL;
671 	struct bch_sb *sb;
672 	int ret = 0;
673 
674 	if (!capable(CAP_SYS_ADMIN))
675 		return -EPERM;
676 
677 	if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
678 	    arg.pad)
679 		return -EINVAL;
680 
681 	mutex_lock(&c->sb_lock);
682 
683 	if (arg.flags & BCH_READ_DEV) {
684 		ca = bch2_device_lookup(c, arg.dev, arg.flags);
685 		ret = PTR_ERR_OR_ZERO(ca);
686 		if (ret)
687 			goto err_unlock;
688 
689 		sb = ca->disk_sb.sb;
690 	} else {
691 		sb = c->disk_sb.sb;
692 	}
693 
694 	if (vstruct_bytes(sb) > arg.size) {
695 		ret = -ERANGE;
696 		goto err;
697 	}
698 
699 	ret = copy_to_user_errcode((void __user *)(unsigned long)arg.sb, sb,
700 				   vstruct_bytes(sb));
701 err:
702 	bch2_dev_put(ca);
703 err_unlock:
704 	mutex_unlock(&c->sb_lock);
705 	return ret;
706 }
707 
708 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
709 				    struct bch_ioctl_disk_get_idx arg)
710 {
711 	dev_t dev = huge_decode_dev(arg.dev);
712 
713 	if (!capable(CAP_SYS_ADMIN))
714 		return -EPERM;
715 
716 	if (!dev)
717 		return -EINVAL;
718 
719 	for_each_online_member(c, ca)
720 		if (ca->dev == dev) {
721 			percpu_ref_put(&ca->io_ref);
722 			return ca->dev_idx;
723 		}
724 
725 	return -BCH_ERR_ENOENT_dev_idx_not_found;
726 }
727 
728 static long bch2_ioctl_disk_resize(struct bch_fs *c,
729 				   struct bch_ioctl_disk_resize arg)
730 {
731 	struct bch_dev *ca;
732 	int ret;
733 
734 	if (!capable(CAP_SYS_ADMIN))
735 		return -EPERM;
736 
737 	if ((arg.flags & ~BCH_BY_INDEX) ||
738 	    arg.pad)
739 		return -EINVAL;
740 
741 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
742 	if (IS_ERR(ca))
743 		return PTR_ERR(ca);
744 
745 	ret = bch2_dev_resize(c, ca, arg.nbuckets);
746 
747 	bch2_dev_put(ca);
748 	return ret;
749 }
750 
751 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
752 				   struct bch_ioctl_disk_resize_journal arg)
753 {
754 	struct bch_dev *ca;
755 	int ret;
756 
757 	if (!capable(CAP_SYS_ADMIN))
758 		return -EPERM;
759 
760 	if ((arg.flags & ~BCH_BY_INDEX) ||
761 	    arg.pad)
762 		return -EINVAL;
763 
764 	if (arg.nbuckets > U32_MAX)
765 		return -EINVAL;
766 
767 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
768 	if (IS_ERR(ca))
769 		return PTR_ERR(ca);
770 
771 	ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
772 
773 	bch2_dev_put(ca);
774 	return ret;
775 }
776 
777 static int bch2_fsck_online_thread_fn(struct thread_with_stdio *stdio)
778 {
779 	struct fsck_thread *thr = container_of(stdio, struct fsck_thread, thr);
780 	struct bch_fs *c = thr->c;
781 
782 	c->stdio_filter = current;
783 	c->stdio = &thr->thr.stdio;
784 
785 	/*
786 	 * XXX: can we figure out a way to do this without mucking with c->opts?
787 	 */
788 	unsigned old_fix_errors = c->opts.fix_errors;
789 	if (opt_defined(thr->opts, fix_errors))
790 		c->opts.fix_errors = thr->opts.fix_errors;
791 	else
792 		c->opts.fix_errors = FSCK_FIX_ask;
793 
794 	c->opts.fsck = true;
795 	set_bit(BCH_FS_fsck_running, &c->flags);
796 
797 	c->curr_recovery_pass = BCH_RECOVERY_PASS_check_alloc_info;
798 	int ret = bch2_run_online_recovery_passes(c);
799 
800 	clear_bit(BCH_FS_fsck_running, &c->flags);
801 	bch_err_fn(c, ret);
802 
803 	c->stdio = NULL;
804 	c->stdio_filter = NULL;
805 	c->opts.fix_errors = old_fix_errors;
806 
807 	up(&c->online_fsck_mutex);
808 	bch2_ro_ref_put(c);
809 	return ret;
810 }
811 
812 static const struct thread_with_stdio_ops bch2_online_fsck_ops = {
813 	.exit		= bch2_fsck_thread_exit,
814 	.fn		= bch2_fsck_online_thread_fn,
815 };
816 
817 static long bch2_ioctl_fsck_online(struct bch_fs *c,
818 				   struct bch_ioctl_fsck_online arg)
819 {
820 	struct fsck_thread *thr = NULL;
821 	long ret = 0;
822 
823 	if (arg.flags)
824 		return -EINVAL;
825 
826 	if (!capable(CAP_SYS_ADMIN))
827 		return -EPERM;
828 
829 	if (!bch2_ro_ref_tryget(c))
830 		return -EROFS;
831 
832 	if (down_trylock(&c->online_fsck_mutex)) {
833 		bch2_ro_ref_put(c);
834 		return -EAGAIN;
835 	}
836 
837 	thr = kzalloc(sizeof(*thr), GFP_KERNEL);
838 	if (!thr) {
839 		ret = -ENOMEM;
840 		goto err;
841 	}
842 
843 	thr->c = c;
844 	thr->opts = bch2_opts_empty();
845 
846 	if (arg.opts) {
847 		char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
848 
849 		ret =   PTR_ERR_OR_ZERO(optstr) ?:
850 			bch2_parse_mount_opts(c, &thr->opts, NULL, optstr);
851 		if (!IS_ERR(optstr))
852 			kfree(optstr);
853 
854 		if (ret)
855 			goto err;
856 	}
857 
858 	ret = bch2_run_thread_with_stdio(&thr->thr, &bch2_online_fsck_ops);
859 err:
860 	if (ret < 0) {
861 		bch_err_fn(c, ret);
862 		if (thr)
863 			bch2_fsck_thread_exit(&thr->thr);
864 		up(&c->online_fsck_mutex);
865 		bch2_ro_ref_put(c);
866 	}
867 	return ret;
868 }
869 
870 #define BCH_IOCTL(_name, _argtype)					\
871 do {									\
872 	_argtype i;							\
873 									\
874 	if (copy_from_user(&i, arg, sizeof(i)))				\
875 		return -EFAULT;						\
876 	ret = bch2_ioctl_##_name(c, i);					\
877 	goto out;							\
878 } while (0)
879 
880 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
881 {
882 	long ret;
883 
884 	switch (cmd) {
885 	case BCH_IOCTL_QUERY_UUID:
886 		return bch2_ioctl_query_uuid(c, arg);
887 	case BCH_IOCTL_FS_USAGE:
888 		return bch2_ioctl_fs_usage(c, arg);
889 	case BCH_IOCTL_DEV_USAGE:
890 		return bch2_ioctl_dev_usage(c, arg);
891 	case BCH_IOCTL_DEV_USAGE_V2:
892 		return bch2_ioctl_dev_usage_v2(c, arg);
893 #if 0
894 	case BCH_IOCTL_START:
895 		BCH_IOCTL(start, struct bch_ioctl_start);
896 	case BCH_IOCTL_STOP:
897 		return bch2_ioctl_stop(c);
898 #endif
899 	case BCH_IOCTL_READ_SUPER:
900 		BCH_IOCTL(read_super, struct bch_ioctl_read_super);
901 	case BCH_IOCTL_DISK_GET_IDX:
902 		BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
903 	}
904 
905 	if (!test_bit(BCH_FS_started, &c->flags))
906 		return -EINVAL;
907 
908 	switch (cmd) {
909 	case BCH_IOCTL_DISK_ADD:
910 		BCH_IOCTL(disk_add, struct bch_ioctl_disk);
911 	case BCH_IOCTL_DISK_REMOVE:
912 		BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
913 	case BCH_IOCTL_DISK_ONLINE:
914 		BCH_IOCTL(disk_online, struct bch_ioctl_disk);
915 	case BCH_IOCTL_DISK_OFFLINE:
916 		BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
917 	case BCH_IOCTL_DISK_SET_STATE:
918 		BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
919 	case BCH_IOCTL_DATA:
920 		BCH_IOCTL(data, struct bch_ioctl_data);
921 	case BCH_IOCTL_DISK_RESIZE:
922 		BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
923 	case BCH_IOCTL_DISK_RESIZE_JOURNAL:
924 		BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
925 	case BCH_IOCTL_FSCK_ONLINE:
926 		BCH_IOCTL(fsck_online, struct bch_ioctl_fsck_online);
927 	case BCH_IOCTL_QUERY_ACCOUNTING:
928 		return bch2_ioctl_query_accounting(c, arg);
929 	default:
930 		return -ENOTTY;
931 	}
932 out:
933 	if (ret < 0)
934 		ret = bch2_err_class(ret);
935 	return ret;
936 }
937 
938 static DEFINE_IDR(bch_chardev_minor);
939 
940 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
941 {
942 	unsigned minor = iminor(file_inode(filp));
943 	struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
944 	void __user *arg = (void __user *) v;
945 
946 	return c
947 		? bch2_fs_ioctl(c, cmd, arg)
948 		: bch2_global_ioctl(cmd, arg);
949 }
950 
951 static const struct file_operations bch_chardev_fops = {
952 	.owner		= THIS_MODULE,
953 	.unlocked_ioctl = bch2_chardev_ioctl,
954 	.open		= nonseekable_open,
955 };
956 
957 static int bch_chardev_major;
958 static const struct class bch_chardev_class = {
959 	.name = "bcachefs",
960 };
961 static struct device *bch_chardev;
962 
963 void bch2_fs_chardev_exit(struct bch_fs *c)
964 {
965 	if (!IS_ERR_OR_NULL(c->chardev))
966 		device_unregister(c->chardev);
967 	if (c->minor >= 0)
968 		idr_remove(&bch_chardev_minor, c->minor);
969 }
970 
971 int bch2_fs_chardev_init(struct bch_fs *c)
972 {
973 	c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
974 	if (c->minor < 0)
975 		return c->minor;
976 
977 	c->chardev = device_create(&bch_chardev_class, NULL,
978 				   MKDEV(bch_chardev_major, c->minor), c,
979 				   "bcachefs%u-ctl", c->minor);
980 	if (IS_ERR(c->chardev))
981 		return PTR_ERR(c->chardev);
982 
983 	return 0;
984 }
985 
986 void bch2_chardev_exit(void)
987 {
988 	device_destroy(&bch_chardev_class, MKDEV(bch_chardev_major, U8_MAX));
989 	class_unregister(&bch_chardev_class);
990 	if (bch_chardev_major > 0)
991 		unregister_chrdev(bch_chardev_major, "bcachefs");
992 }
993 
994 int __init bch2_chardev_init(void)
995 {
996 	int ret;
997 
998 	bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
999 	if (bch_chardev_major < 0)
1000 		return bch_chardev_major;
1001 
1002 	ret = class_register(&bch_chardev_class);
1003 	if (ret)
1004 		goto major_out;
1005 
1006 	bch_chardev = device_create(&bch_chardev_class, NULL,
1007 				    MKDEV(bch_chardev_major, U8_MAX),
1008 				    NULL, "bcachefs-ctl");
1009 	if (IS_ERR(bch_chardev)) {
1010 		ret = PTR_ERR(bch_chardev);
1011 		goto class_out;
1012 	}
1013 
1014 	return 0;
1015 
1016 class_out:
1017 	class_unregister(&bch_chardev_class);
1018 major_out:
1019 	unregister_chrdev(bch_chardev_major, "bcachefs-ctl");
1020 	return ret;
1021 }
1022 
1023 #endif /* NO_BCACHEFS_CHARDEV */
1024