xref: /linux/fs/bcachefs/chardev.c (revision 7255fcc80d4b525cc10cfaaf7f485830d4ed2000)
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 "journal.h"
9 #include "move.h"
10 #include "recovery_passes.h"
11 #include "replicas.h"
12 #include "super.h"
13 #include "super-io.h"
14 #include "thread_with_file.h"
15 
16 #include <linux/cdev.h>
17 #include <linux/device.h>
18 #include <linux/fs.h>
19 #include <linux/ioctl.h>
20 #include <linux/major.h>
21 #include <linux/sched/task.h>
22 #include <linux/slab.h>
23 #include <linux/uaccess.h>
24 
25 /* returns with ref on ca->ref */
26 static struct bch_dev *bch2_device_lookup(struct bch_fs *c, u64 dev,
27 					  unsigned flags)
28 {
29 	struct bch_dev *ca;
30 
31 	if (flags & BCH_BY_INDEX) {
32 		if (dev >= c->sb.nr_devices)
33 			return ERR_PTR(-EINVAL);
34 
35 		rcu_read_lock();
36 		ca = rcu_dereference(c->devs[dev]);
37 		if (ca)
38 			percpu_ref_get(&ca->ref);
39 		rcu_read_unlock();
40 
41 		if (!ca)
42 			return ERR_PTR(-EINVAL);
43 	} else {
44 		char *path;
45 
46 		path = strndup_user((const char __user *)
47 				    (unsigned long) dev, PATH_MAX);
48 		if (IS_ERR(path))
49 			return ERR_CAST(path);
50 
51 		ca = bch2_dev_lookup(c, path);
52 		kfree(path);
53 	}
54 
55 	return ca;
56 }
57 
58 #if 0
59 static long bch2_ioctl_assemble(struct bch_ioctl_assemble __user *user_arg)
60 {
61 	struct bch_ioctl_assemble arg;
62 	struct bch_fs *c;
63 	u64 *user_devs = NULL;
64 	char **devs = NULL;
65 	unsigned i;
66 	int ret = -EFAULT;
67 
68 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
69 		return -EFAULT;
70 
71 	if (arg.flags || arg.pad)
72 		return -EINVAL;
73 
74 	user_devs = kmalloc_array(arg.nr_devs, sizeof(u64), GFP_KERNEL);
75 	if (!user_devs)
76 		return -ENOMEM;
77 
78 	devs = kcalloc(arg.nr_devs, sizeof(char *), GFP_KERNEL);
79 
80 	if (copy_from_user(user_devs, user_arg->devs,
81 			   sizeof(u64) * arg.nr_devs))
82 		goto err;
83 
84 	for (i = 0; i < arg.nr_devs; i++) {
85 		devs[i] = strndup_user((const char __user *)(unsigned long)
86 				       user_devs[i],
87 				       PATH_MAX);
88 		ret= PTR_ERR_OR_ZERO(devs[i]);
89 		if (ret)
90 			goto err;
91 	}
92 
93 	c = bch2_fs_open(devs, arg.nr_devs, bch2_opts_empty());
94 	ret = PTR_ERR_OR_ZERO(c);
95 	if (!ret)
96 		closure_put(&c->cl);
97 err:
98 	if (devs)
99 		for (i = 0; i < arg.nr_devs; i++)
100 			kfree(devs[i]);
101 	kfree(devs);
102 	return ret;
103 }
104 
105 static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg)
106 {
107 	struct bch_ioctl_incremental arg;
108 	const char *err;
109 	char *path;
110 
111 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
112 		return -EFAULT;
113 
114 	if (arg.flags || arg.pad)
115 		return -EINVAL;
116 
117 	path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
118 	ret = PTR_ERR_OR_ZERO(path);
119 	if (ret)
120 		return ret;
121 
122 	err = bch2_fs_open_incremental(path);
123 	kfree(path);
124 
125 	if (err) {
126 		pr_err("Could not register bcachefs devices: %s", err);
127 		return -EINVAL;
128 	}
129 
130 	return 0;
131 }
132 #endif
133 
134 struct fsck_thread {
135 	struct thread_with_stdio thr;
136 	struct bch_fs		*c;
137 	struct bch_opts		opts;
138 };
139 
140 static void bch2_fsck_thread_exit(struct thread_with_stdio *_thr)
141 {
142 	struct fsck_thread *thr = container_of(_thr, struct fsck_thread, thr);
143 	kfree(thr);
144 }
145 
146 static int bch2_fsck_offline_thread_fn(struct thread_with_stdio *stdio)
147 {
148 	struct fsck_thread *thr = container_of(stdio, struct fsck_thread, thr);
149 	struct bch_fs *c = thr->c;
150 
151 	int ret = PTR_ERR_OR_ZERO(c);
152 	if (ret)
153 		return ret;
154 
155 	ret = bch2_fs_start(thr->c);
156 	if (ret)
157 		goto err;
158 
159 	if (test_bit(BCH_FS_errors_fixed, &c->flags)) {
160 		bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: errors fixed\n", c->name);
161 		ret |= 1;
162 	}
163 	if (test_bit(BCH_FS_error, &c->flags)) {
164 		bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: still has errors\n", c->name);
165 		ret |= 4;
166 	}
167 err:
168 	bch2_fs_stop(c);
169 	return ret;
170 }
171 
172 static const struct thread_with_stdio_ops bch2_offline_fsck_ops = {
173 	.exit		= bch2_fsck_thread_exit,
174 	.fn		= bch2_fsck_offline_thread_fn,
175 };
176 
177 static long bch2_ioctl_fsck_offline(struct bch_ioctl_fsck_offline __user *user_arg)
178 {
179 	struct bch_ioctl_fsck_offline arg;
180 	struct fsck_thread *thr = NULL;
181 	darray_str(devs) = {};
182 	long ret = 0;
183 
184 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
185 		return -EFAULT;
186 
187 	if (arg.flags)
188 		return -EINVAL;
189 
190 	if (!capable(CAP_SYS_ADMIN))
191 		return -EPERM;
192 
193 	for (size_t i = 0; i < arg.nr_devs; i++) {
194 		u64 dev_u64;
195 		ret = copy_from_user_errcode(&dev_u64, &user_arg->devs[i], sizeof(u64));
196 		if (ret)
197 			goto err;
198 
199 		char *dev_str = strndup_user((char __user *)(unsigned long) dev_u64, PATH_MAX);
200 		ret = PTR_ERR_OR_ZERO(dev_str);
201 		if (ret)
202 			goto err;
203 
204 		ret = darray_push(&devs, dev_str);
205 		if (ret) {
206 			kfree(dev_str);
207 			goto err;
208 		}
209 	}
210 
211 	thr = kzalloc(sizeof(*thr), GFP_KERNEL);
212 	if (!thr) {
213 		ret = -ENOMEM;
214 		goto err;
215 	}
216 
217 	thr->opts = bch2_opts_empty();
218 
219 	if (arg.opts) {
220 		char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
221 
222 		ret =   PTR_ERR_OR_ZERO(optstr) ?:
223 			bch2_parse_mount_opts(NULL, &thr->opts, optstr);
224 		kfree(optstr);
225 
226 		if (ret)
227 			goto err;
228 	}
229 
230 	opt_set(thr->opts, stdio, (u64)(unsigned long)&thr->thr.stdio);
231 
232 	/* We need request_key() to be called before we punt to kthread: */
233 	opt_set(thr->opts, nostart, true);
234 
235 	thr->c = bch2_fs_open(devs.data, arg.nr_devs, thr->opts);
236 
237 	if (!IS_ERR(thr->c) &&
238 	    thr->c->opts.errors == BCH_ON_ERROR_panic)
239 		thr->c->opts.errors = BCH_ON_ERROR_ro;
240 
241 	ret = bch2_run_thread_with_stdio(&thr->thr, &bch2_offline_fsck_ops);
242 out:
243 	darray_for_each(devs, i)
244 		kfree(*i);
245 	darray_exit(&devs);
246 	return ret;
247 err:
248 	if (thr)
249 		bch2_fsck_thread_exit(&thr->thr);
250 	pr_err("ret %s", bch2_err_str(ret));
251 	goto out;
252 }
253 
254 static long bch2_global_ioctl(unsigned cmd, void __user *arg)
255 {
256 	long ret;
257 
258 	switch (cmd) {
259 #if 0
260 	case BCH_IOCTL_ASSEMBLE:
261 		return bch2_ioctl_assemble(arg);
262 	case BCH_IOCTL_INCREMENTAL:
263 		return bch2_ioctl_incremental(arg);
264 #endif
265 	case BCH_IOCTL_FSCK_OFFLINE: {
266 		ret = bch2_ioctl_fsck_offline(arg);
267 		break;
268 	}
269 	default:
270 		ret = -ENOTTY;
271 		break;
272 	}
273 
274 	if (ret < 0)
275 		ret = bch2_err_class(ret);
276 	return ret;
277 }
278 
279 static long bch2_ioctl_query_uuid(struct bch_fs *c,
280 			struct bch_ioctl_query_uuid __user *user_arg)
281 {
282 	return copy_to_user_errcode(&user_arg->uuid, &c->sb.user_uuid,
283 				    sizeof(c->sb.user_uuid));
284 }
285 
286 #if 0
287 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
288 {
289 	if (!capable(CAP_SYS_ADMIN))
290 		return -EPERM;
291 
292 	if (arg.flags || arg.pad)
293 		return -EINVAL;
294 
295 	return bch2_fs_start(c);
296 }
297 
298 static long bch2_ioctl_stop(struct bch_fs *c)
299 {
300 	if (!capable(CAP_SYS_ADMIN))
301 		return -EPERM;
302 
303 	bch2_fs_stop(c);
304 	return 0;
305 }
306 #endif
307 
308 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
309 {
310 	char *path;
311 	int ret;
312 
313 	if (!capable(CAP_SYS_ADMIN))
314 		return -EPERM;
315 
316 	if (arg.flags || arg.pad)
317 		return -EINVAL;
318 
319 	path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
320 	ret = PTR_ERR_OR_ZERO(path);
321 	if (ret)
322 		return ret;
323 
324 	ret = bch2_dev_add(c, 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 	percpu_ref_put(&ca->ref);
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 	percpu_ref_put(&ca->ref);
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 	.llseek		= no_llseek,
475 };
476 
477 static long bch2_ioctl_data(struct bch_fs *c,
478 			    struct bch_ioctl_data arg)
479 {
480 	struct bch_data_ctx *ctx;
481 	int ret;
482 
483 	if (!capable(CAP_SYS_ADMIN))
484 		return -EPERM;
485 
486 	if (arg.op >= BCH_DATA_OP_NR || arg.flags)
487 		return -EINVAL;
488 
489 	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
490 	if (!ctx)
491 		return -ENOMEM;
492 
493 	ctx->c = c;
494 	ctx->arg = arg;
495 
496 	ret = bch2_run_thread_with_file(&ctx->thr,
497 			&bcachefs_data_ops,
498 			bch2_data_thread);
499 	if (ret < 0)
500 		kfree(ctx);
501 	return ret;
502 }
503 
504 static long bch2_ioctl_fs_usage(struct bch_fs *c,
505 				struct bch_ioctl_fs_usage __user *user_arg)
506 {
507 	struct bch_ioctl_fs_usage *arg = NULL;
508 	struct bch_replicas_usage *dst_e, *dst_end;
509 	struct bch_fs_usage_online *src;
510 	u32 replica_entries_bytes;
511 	unsigned i;
512 	int ret = 0;
513 
514 	if (!test_bit(BCH_FS_started, &c->flags))
515 		return -EINVAL;
516 
517 	if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
518 		return -EFAULT;
519 
520 	arg = kzalloc(size_add(sizeof(*arg), replica_entries_bytes), GFP_KERNEL);
521 	if (!arg)
522 		return -ENOMEM;
523 
524 	src = bch2_fs_usage_read(c);
525 	if (!src) {
526 		ret = -ENOMEM;
527 		goto err;
528 	}
529 
530 	arg->capacity		= c->capacity;
531 	arg->used		= bch2_fs_sectors_used(c, src);
532 	arg->online_reserved	= src->online_reserved;
533 
534 	for (i = 0; i < BCH_REPLICAS_MAX; i++)
535 		arg->persistent_reserved[i] = src->u.persistent_reserved[i];
536 
537 	dst_e	= arg->replicas;
538 	dst_end = (void *) arg->replicas + replica_entries_bytes;
539 
540 	for (i = 0; i < c->replicas.nr; i++) {
541 		struct bch_replicas_entry_v1 *src_e =
542 			cpu_replicas_entry(&c->replicas, i);
543 
544 		/* check that we have enough space for one replicas entry */
545 		if (dst_e + 1 > dst_end) {
546 			ret = -ERANGE;
547 			break;
548 		}
549 
550 		dst_e->sectors		= src->u.replicas[i];
551 		dst_e->r		= *src_e;
552 
553 		/* recheck after setting nr_devs: */
554 		if (replicas_usage_next(dst_e) > dst_end) {
555 			ret = -ERANGE;
556 			break;
557 		}
558 
559 		memcpy(dst_e->r.devs, src_e->devs, src_e->nr_devs);
560 
561 		dst_e = replicas_usage_next(dst_e);
562 	}
563 
564 	arg->replica_entries_bytes = (void *) dst_e - (void *) arg->replicas;
565 
566 	percpu_up_read(&c->mark_lock);
567 	kfree(src);
568 
569 	if (ret)
570 		goto err;
571 
572 	ret = copy_to_user_errcode(user_arg, arg,
573 			sizeof(*arg) + arg->replica_entries_bytes);
574 err:
575 	kfree(arg);
576 	return ret;
577 }
578 
579 /* obsolete, didn't allow for new data types: */
580 static long bch2_ioctl_dev_usage(struct bch_fs *c,
581 				 struct bch_ioctl_dev_usage __user *user_arg)
582 {
583 	struct bch_ioctl_dev_usage arg;
584 	struct bch_dev_usage src;
585 	struct bch_dev *ca;
586 	unsigned i;
587 
588 	if (!test_bit(BCH_FS_started, &c->flags))
589 		return -EINVAL;
590 
591 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
592 		return -EFAULT;
593 
594 	if ((arg.flags & ~BCH_BY_INDEX) ||
595 	    arg.pad[0] ||
596 	    arg.pad[1] ||
597 	    arg.pad[2])
598 		return -EINVAL;
599 
600 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
601 	if (IS_ERR(ca))
602 		return PTR_ERR(ca);
603 
604 	src = bch2_dev_usage_read(ca);
605 
606 	arg.state		= ca->mi.state;
607 	arg.bucket_size		= ca->mi.bucket_size;
608 	arg.nr_buckets		= ca->mi.nbuckets - ca->mi.first_bucket;
609 
610 	for (i = 0; i < BCH_DATA_NR; i++) {
611 		arg.d[i].buckets	= src.d[i].buckets;
612 		arg.d[i].sectors	= src.d[i].sectors;
613 		arg.d[i].fragmented	= src.d[i].fragmented;
614 	}
615 
616 	percpu_ref_put(&ca->ref);
617 
618 	return copy_to_user_errcode(user_arg, &arg, sizeof(arg));
619 }
620 
621 static long bch2_ioctl_dev_usage_v2(struct bch_fs *c,
622 				 struct bch_ioctl_dev_usage_v2 __user *user_arg)
623 {
624 	struct bch_ioctl_dev_usage_v2 arg;
625 	struct bch_dev_usage src;
626 	struct bch_dev *ca;
627 	int ret = 0;
628 
629 	if (!test_bit(BCH_FS_started, &c->flags))
630 		return -EINVAL;
631 
632 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
633 		return -EFAULT;
634 
635 	if ((arg.flags & ~BCH_BY_INDEX) ||
636 	    arg.pad[0] ||
637 	    arg.pad[1] ||
638 	    arg.pad[2])
639 		return -EINVAL;
640 
641 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
642 	if (IS_ERR(ca))
643 		return PTR_ERR(ca);
644 
645 	src = bch2_dev_usage_read(ca);
646 
647 	arg.state		= ca->mi.state;
648 	arg.bucket_size		= ca->mi.bucket_size;
649 	arg.nr_data_types	= min(arg.nr_data_types, BCH_DATA_NR);
650 	arg.nr_buckets		= ca->mi.nbuckets - ca->mi.first_bucket;
651 
652 	ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
653 	if (ret)
654 		goto err;
655 
656 	for (unsigned i = 0; i < arg.nr_data_types; i++) {
657 		struct bch_ioctl_dev_usage_type t = {
658 			.buckets	= src.d[i].buckets,
659 			.sectors	= src.d[i].sectors,
660 			.fragmented	= src.d[i].fragmented,
661 		};
662 
663 		ret = copy_to_user_errcode(&user_arg->d[i], &t, sizeof(t));
664 		if (ret)
665 			goto err;
666 	}
667 err:
668 	percpu_ref_put(&ca->ref);
669 	return ret;
670 }
671 
672 static long bch2_ioctl_read_super(struct bch_fs *c,
673 				  struct bch_ioctl_read_super arg)
674 {
675 	struct bch_dev *ca = NULL;
676 	struct bch_sb *sb;
677 	int ret = 0;
678 
679 	if (!capable(CAP_SYS_ADMIN))
680 		return -EPERM;
681 
682 	if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
683 	    arg.pad)
684 		return -EINVAL;
685 
686 	mutex_lock(&c->sb_lock);
687 
688 	if (arg.flags & BCH_READ_DEV) {
689 		ca = bch2_device_lookup(c, arg.dev, arg.flags);
690 
691 		if (IS_ERR(ca)) {
692 			ret = PTR_ERR(ca);
693 			goto err;
694 		}
695 
696 		sb = ca->disk_sb.sb;
697 	} else {
698 		sb = c->disk_sb.sb;
699 	}
700 
701 	if (vstruct_bytes(sb) > arg.size) {
702 		ret = -ERANGE;
703 		goto err;
704 	}
705 
706 	ret = copy_to_user_errcode((void __user *)(unsigned long)arg.sb, sb,
707 				   vstruct_bytes(sb));
708 err:
709 	if (!IS_ERR_OR_NULL(ca))
710 		percpu_ref_put(&ca->ref);
711 	mutex_unlock(&c->sb_lock);
712 	return ret;
713 }
714 
715 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
716 				    struct bch_ioctl_disk_get_idx arg)
717 {
718 	dev_t dev = huge_decode_dev(arg.dev);
719 
720 	if (!capable(CAP_SYS_ADMIN))
721 		return -EPERM;
722 
723 	if (!dev)
724 		return -EINVAL;
725 
726 	for_each_online_member(c, ca)
727 		if (ca->dev == dev) {
728 			percpu_ref_put(&ca->io_ref);
729 			return ca->dev_idx;
730 		}
731 
732 	return -BCH_ERR_ENOENT_dev_idx_not_found;
733 }
734 
735 static long bch2_ioctl_disk_resize(struct bch_fs *c,
736 				   struct bch_ioctl_disk_resize arg)
737 {
738 	struct bch_dev *ca;
739 	int ret;
740 
741 	if (!capable(CAP_SYS_ADMIN))
742 		return -EPERM;
743 
744 	if ((arg.flags & ~BCH_BY_INDEX) ||
745 	    arg.pad)
746 		return -EINVAL;
747 
748 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
749 	if (IS_ERR(ca))
750 		return PTR_ERR(ca);
751 
752 	ret = bch2_dev_resize(c, ca, arg.nbuckets);
753 
754 	percpu_ref_put(&ca->ref);
755 	return ret;
756 }
757 
758 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
759 				   struct bch_ioctl_disk_resize_journal arg)
760 {
761 	struct bch_dev *ca;
762 	int ret;
763 
764 	if (!capable(CAP_SYS_ADMIN))
765 		return -EPERM;
766 
767 	if ((arg.flags & ~BCH_BY_INDEX) ||
768 	    arg.pad)
769 		return -EINVAL;
770 
771 	if (arg.nbuckets > U32_MAX)
772 		return -EINVAL;
773 
774 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
775 	if (IS_ERR(ca))
776 		return PTR_ERR(ca);
777 
778 	ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
779 
780 	percpu_ref_put(&ca->ref);
781 	return ret;
782 }
783 
784 static int bch2_fsck_online_thread_fn(struct thread_with_stdio *stdio)
785 {
786 	struct fsck_thread *thr = container_of(stdio, struct fsck_thread, thr);
787 	struct bch_fs *c = thr->c;
788 
789 	c->stdio_filter = current;
790 	c->stdio = &thr->thr.stdio;
791 
792 	/*
793 	 * XXX: can we figure out a way to do this without mucking with c->opts?
794 	 */
795 	unsigned old_fix_errors = c->opts.fix_errors;
796 	if (opt_defined(thr->opts, fix_errors))
797 		c->opts.fix_errors = thr->opts.fix_errors;
798 	else
799 		c->opts.fix_errors = FSCK_FIX_ask;
800 
801 	c->opts.fsck = true;
802 	set_bit(BCH_FS_fsck_running, &c->flags);
803 
804 	c->curr_recovery_pass = BCH_RECOVERY_PASS_check_alloc_info;
805 	int ret = bch2_run_online_recovery_passes(c);
806 
807 	clear_bit(BCH_FS_fsck_running, &c->flags);
808 	bch_err_fn(c, ret);
809 
810 	c->stdio = NULL;
811 	c->stdio_filter = NULL;
812 	c->opts.fix_errors = old_fix_errors;
813 
814 	up(&c->online_fsck_mutex);
815 	bch2_ro_ref_put(c);
816 	return ret;
817 }
818 
819 static const struct thread_with_stdio_ops bch2_online_fsck_ops = {
820 	.exit		= bch2_fsck_thread_exit,
821 	.fn		= bch2_fsck_online_thread_fn,
822 };
823 
824 static long bch2_ioctl_fsck_online(struct bch_fs *c,
825 				   struct bch_ioctl_fsck_online arg)
826 {
827 	struct fsck_thread *thr = NULL;
828 	long ret = 0;
829 
830 	if (arg.flags)
831 		return -EINVAL;
832 
833 	if (!capable(CAP_SYS_ADMIN))
834 		return -EPERM;
835 
836 	if (!bch2_ro_ref_tryget(c))
837 		return -EROFS;
838 
839 	if (down_trylock(&c->online_fsck_mutex)) {
840 		bch2_ro_ref_put(c);
841 		return -EAGAIN;
842 	}
843 
844 	thr = kzalloc(sizeof(*thr), GFP_KERNEL);
845 	if (!thr) {
846 		ret = -ENOMEM;
847 		goto err;
848 	}
849 
850 	thr->c = c;
851 	thr->opts = bch2_opts_empty();
852 
853 	if (arg.opts) {
854 		char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
855 
856 		ret =   PTR_ERR_OR_ZERO(optstr) ?:
857 			bch2_parse_mount_opts(c, &thr->opts, optstr);
858 		kfree(optstr);
859 
860 		if (ret)
861 			goto err;
862 	}
863 
864 	ret = bch2_run_thread_with_stdio(&thr->thr, &bch2_online_fsck_ops);
865 err:
866 	if (ret < 0) {
867 		bch_err_fn(c, ret);
868 		if (thr)
869 			bch2_fsck_thread_exit(&thr->thr);
870 		up(&c->online_fsck_mutex);
871 		bch2_ro_ref_put(c);
872 	}
873 	return ret;
874 }
875 
876 #define BCH_IOCTL(_name, _argtype)					\
877 do {									\
878 	_argtype i;							\
879 									\
880 	if (copy_from_user(&i, arg, sizeof(i)))				\
881 		return -EFAULT;						\
882 	ret = bch2_ioctl_##_name(c, i);					\
883 	goto out;							\
884 } while (0)
885 
886 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
887 {
888 	long ret;
889 
890 	switch (cmd) {
891 	case BCH_IOCTL_QUERY_UUID:
892 		return bch2_ioctl_query_uuid(c, arg);
893 	case BCH_IOCTL_FS_USAGE:
894 		return bch2_ioctl_fs_usage(c, arg);
895 	case BCH_IOCTL_DEV_USAGE:
896 		return bch2_ioctl_dev_usage(c, arg);
897 	case BCH_IOCTL_DEV_USAGE_V2:
898 		return bch2_ioctl_dev_usage_v2(c, arg);
899 #if 0
900 	case BCH_IOCTL_START:
901 		BCH_IOCTL(start, struct bch_ioctl_start);
902 	case BCH_IOCTL_STOP:
903 		return bch2_ioctl_stop(c);
904 #endif
905 	case BCH_IOCTL_READ_SUPER:
906 		BCH_IOCTL(read_super, struct bch_ioctl_read_super);
907 	case BCH_IOCTL_DISK_GET_IDX:
908 		BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
909 	}
910 
911 	if (!test_bit(BCH_FS_started, &c->flags))
912 		return -EINVAL;
913 
914 	switch (cmd) {
915 	case BCH_IOCTL_DISK_ADD:
916 		BCH_IOCTL(disk_add, struct bch_ioctl_disk);
917 	case BCH_IOCTL_DISK_REMOVE:
918 		BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
919 	case BCH_IOCTL_DISK_ONLINE:
920 		BCH_IOCTL(disk_online, struct bch_ioctl_disk);
921 	case BCH_IOCTL_DISK_OFFLINE:
922 		BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
923 	case BCH_IOCTL_DISK_SET_STATE:
924 		BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
925 	case BCH_IOCTL_DATA:
926 		BCH_IOCTL(data, struct bch_ioctl_data);
927 	case BCH_IOCTL_DISK_RESIZE:
928 		BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
929 	case BCH_IOCTL_DISK_RESIZE_JOURNAL:
930 		BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
931 	case BCH_IOCTL_FSCK_ONLINE:
932 		BCH_IOCTL(fsck_online, struct bch_ioctl_fsck_online);
933 	default:
934 		return -ENOTTY;
935 	}
936 out:
937 	if (ret < 0)
938 		ret = bch2_err_class(ret);
939 	return ret;
940 }
941 
942 static DEFINE_IDR(bch_chardev_minor);
943 
944 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
945 {
946 	unsigned minor = iminor(file_inode(filp));
947 	struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
948 	void __user *arg = (void __user *) v;
949 
950 	return c
951 		? bch2_fs_ioctl(c, cmd, arg)
952 		: bch2_global_ioctl(cmd, arg);
953 }
954 
955 static const struct file_operations bch_chardev_fops = {
956 	.owner		= THIS_MODULE,
957 	.unlocked_ioctl = bch2_chardev_ioctl,
958 	.open		= nonseekable_open,
959 };
960 
961 static int bch_chardev_major;
962 static struct class *bch_chardev_class;
963 static struct device *bch_chardev;
964 
965 void bch2_fs_chardev_exit(struct bch_fs *c)
966 {
967 	if (!IS_ERR_OR_NULL(c->chardev))
968 		device_unregister(c->chardev);
969 	if (c->minor >= 0)
970 		idr_remove(&bch_chardev_minor, c->minor);
971 }
972 
973 int bch2_fs_chardev_init(struct bch_fs *c)
974 {
975 	c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
976 	if (c->minor < 0)
977 		return c->minor;
978 
979 	c->chardev = device_create(bch_chardev_class, NULL,
980 				   MKDEV(bch_chardev_major, c->minor), c,
981 				   "bcachefs%u-ctl", c->minor);
982 	if (IS_ERR(c->chardev))
983 		return PTR_ERR(c->chardev);
984 
985 	return 0;
986 }
987 
988 void bch2_chardev_exit(void)
989 {
990 	if (!IS_ERR_OR_NULL(bch_chardev_class))
991 		device_destroy(bch_chardev_class,
992 			       MKDEV(bch_chardev_major, U8_MAX));
993 	if (!IS_ERR_OR_NULL(bch_chardev_class))
994 		class_destroy(bch_chardev_class);
995 	if (bch_chardev_major > 0)
996 		unregister_chrdev(bch_chardev_major, "bcachefs");
997 }
998 
999 int __init bch2_chardev_init(void)
1000 {
1001 	bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
1002 	if (bch_chardev_major < 0)
1003 		return bch_chardev_major;
1004 
1005 	bch_chardev_class = class_create("bcachefs");
1006 	if (IS_ERR(bch_chardev_class))
1007 		return PTR_ERR(bch_chardev_class);
1008 
1009 	bch_chardev = device_create(bch_chardev_class, NULL,
1010 				    MKDEV(bch_chardev_major, U8_MAX),
1011 				    NULL, "bcachefs-ctl");
1012 	if (IS_ERR(bch_chardev))
1013 		return PTR_ERR(bch_chardev);
1014 
1015 	return 0;
1016 }
1017 
1018 #endif /* NO_BCACHEFS_CHARDEV */
1019