xref: /linux/fs/bcachefs/chardev.c (revision b4ba157044ea433a66126603ad7140e12dbc794b)
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 		ca = bch2_dev_tryget_noerror(c, dev);
36 		if (!ca)
37 			return ERR_PTR(-EINVAL);
38 	} else {
39 		char *path;
40 
41 		path = strndup_user((const char __user *)
42 				    (unsigned long) dev, PATH_MAX);
43 		if (IS_ERR(path))
44 			return ERR_CAST(path);
45 
46 		ca = bch2_dev_lookup(c, path);
47 		kfree(path);
48 	}
49 
50 	return ca;
51 }
52 
53 #if 0
54 static long bch2_ioctl_assemble(struct bch_ioctl_assemble __user *user_arg)
55 {
56 	struct bch_ioctl_assemble arg;
57 	struct bch_fs *c;
58 	u64 *user_devs = NULL;
59 	char **devs = NULL;
60 	unsigned i;
61 	int ret = -EFAULT;
62 
63 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
64 		return -EFAULT;
65 
66 	if (arg.flags || arg.pad)
67 		return -EINVAL;
68 
69 	user_devs = kmalloc_array(arg.nr_devs, sizeof(u64), GFP_KERNEL);
70 	if (!user_devs)
71 		return -ENOMEM;
72 
73 	devs = kcalloc(arg.nr_devs, sizeof(char *), GFP_KERNEL);
74 
75 	if (copy_from_user(user_devs, user_arg->devs,
76 			   sizeof(u64) * arg.nr_devs))
77 		goto err;
78 
79 	for (i = 0; i < arg.nr_devs; i++) {
80 		devs[i] = strndup_user((const char __user *)(unsigned long)
81 				       user_devs[i],
82 				       PATH_MAX);
83 		ret= PTR_ERR_OR_ZERO(devs[i]);
84 		if (ret)
85 			goto err;
86 	}
87 
88 	c = bch2_fs_open(devs, arg.nr_devs, bch2_opts_empty());
89 	ret = PTR_ERR_OR_ZERO(c);
90 	if (!ret)
91 		closure_put(&c->cl);
92 err:
93 	if (devs)
94 		for (i = 0; i < arg.nr_devs; i++)
95 			kfree(devs[i]);
96 	kfree(devs);
97 	return ret;
98 }
99 
100 static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg)
101 {
102 	struct bch_ioctl_incremental arg;
103 	const char *err;
104 	char *path;
105 
106 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
107 		return -EFAULT;
108 
109 	if (arg.flags || arg.pad)
110 		return -EINVAL;
111 
112 	path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
113 	ret = PTR_ERR_OR_ZERO(path);
114 	if (ret)
115 		return ret;
116 
117 	err = bch2_fs_open_incremental(path);
118 	kfree(path);
119 
120 	if (err) {
121 		pr_err("Could not register bcachefs devices: %s", err);
122 		return -EINVAL;
123 	}
124 
125 	return 0;
126 }
127 #endif
128 
129 struct fsck_thread {
130 	struct thread_with_stdio thr;
131 	struct bch_fs		*c;
132 	struct bch_opts		opts;
133 };
134 
135 static void bch2_fsck_thread_exit(struct thread_with_stdio *_thr)
136 {
137 	struct fsck_thread *thr = container_of(_thr, struct fsck_thread, thr);
138 	kfree(thr);
139 }
140 
141 static int bch2_fsck_offline_thread_fn(struct thread_with_stdio *stdio)
142 {
143 	struct fsck_thread *thr = container_of(stdio, struct fsck_thread, thr);
144 	struct bch_fs *c = thr->c;
145 
146 	int ret = PTR_ERR_OR_ZERO(c);
147 	if (ret)
148 		return ret;
149 
150 	ret = bch2_fs_start(thr->c);
151 	if (ret)
152 		goto err;
153 
154 	if (test_bit(BCH_FS_errors_fixed, &c->flags)) {
155 		bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: errors fixed\n", c->name);
156 		ret |= 1;
157 	}
158 	if (test_bit(BCH_FS_error, &c->flags)) {
159 		bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: still has errors\n", c->name);
160 		ret |= 4;
161 	}
162 err:
163 	bch2_fs_stop(c);
164 	return ret;
165 }
166 
167 static const struct thread_with_stdio_ops bch2_offline_fsck_ops = {
168 	.exit		= bch2_fsck_thread_exit,
169 	.fn		= bch2_fsck_offline_thread_fn,
170 };
171 
172 static long bch2_ioctl_fsck_offline(struct bch_ioctl_fsck_offline __user *user_arg)
173 {
174 	struct bch_ioctl_fsck_offline arg;
175 	struct fsck_thread *thr = NULL;
176 	darray_str(devs) = {};
177 	long ret = 0;
178 
179 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
180 		return -EFAULT;
181 
182 	if (arg.flags)
183 		return -EINVAL;
184 
185 	if (!capable(CAP_SYS_ADMIN))
186 		return -EPERM;
187 
188 	for (size_t i = 0; i < arg.nr_devs; i++) {
189 		u64 dev_u64;
190 		ret = copy_from_user_errcode(&dev_u64, &user_arg->devs[i], sizeof(u64));
191 		if (ret)
192 			goto err;
193 
194 		char *dev_str = strndup_user((char __user *)(unsigned long) dev_u64, PATH_MAX);
195 		ret = PTR_ERR_OR_ZERO(dev_str);
196 		if (ret)
197 			goto err;
198 
199 		ret = darray_push(&devs, dev_str);
200 		if (ret) {
201 			kfree(dev_str);
202 			goto err;
203 		}
204 	}
205 
206 	thr = kzalloc(sizeof(*thr), GFP_KERNEL);
207 	if (!thr) {
208 		ret = -ENOMEM;
209 		goto err;
210 	}
211 
212 	thr->opts = bch2_opts_empty();
213 
214 	if (arg.opts) {
215 		char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
216 
217 		ret =   PTR_ERR_OR_ZERO(optstr) ?:
218 			bch2_parse_mount_opts(NULL, &thr->opts, 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 
228 	/* We need request_key() to be called before we punt to kthread: */
229 	opt_set(thr->opts, nostart, true);
230 
231 	bch2_thread_with_stdio_init(&thr->thr, &bch2_offline_fsck_ops);
232 
233 	thr->c = bch2_fs_open(devs.data, arg.nr_devs, thr->opts);
234 
235 	if (!IS_ERR(thr->c) &&
236 	    thr->c->opts.errors == BCH_ON_ERROR_panic)
237 		thr->c->opts.errors = BCH_ON_ERROR_ro;
238 
239 	ret = __bch2_run_thread_with_stdio(&thr->thr);
240 out:
241 	darray_for_each(devs, i)
242 		kfree(*i);
243 	darray_exit(&devs);
244 	return ret;
245 err:
246 	if (thr)
247 		bch2_fsck_thread_exit(&thr->thr);
248 	pr_err("ret %s", bch2_err_str(ret));
249 	goto out;
250 }
251 
252 static long bch2_global_ioctl(unsigned cmd, void __user *arg)
253 {
254 	long ret;
255 
256 	switch (cmd) {
257 #if 0
258 	case BCH_IOCTL_ASSEMBLE:
259 		return bch2_ioctl_assemble(arg);
260 	case BCH_IOCTL_INCREMENTAL:
261 		return bch2_ioctl_incremental(arg);
262 #endif
263 	case BCH_IOCTL_FSCK_OFFLINE: {
264 		ret = bch2_ioctl_fsck_offline(arg);
265 		break;
266 	}
267 	default:
268 		ret = -ENOTTY;
269 		break;
270 	}
271 
272 	if (ret < 0)
273 		ret = bch2_err_class(ret);
274 	return ret;
275 }
276 
277 static long bch2_ioctl_query_uuid(struct bch_fs *c,
278 			struct bch_ioctl_query_uuid __user *user_arg)
279 {
280 	return copy_to_user_errcode(&user_arg->uuid, &c->sb.user_uuid,
281 				    sizeof(c->sb.user_uuid));
282 }
283 
284 #if 0
285 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
286 {
287 	if (!capable(CAP_SYS_ADMIN))
288 		return -EPERM;
289 
290 	if (arg.flags || arg.pad)
291 		return -EINVAL;
292 
293 	return bch2_fs_start(c);
294 }
295 
296 static long bch2_ioctl_stop(struct bch_fs *c)
297 {
298 	if (!capable(CAP_SYS_ADMIN))
299 		return -EPERM;
300 
301 	bch2_fs_stop(c);
302 	return 0;
303 }
304 #endif
305 
306 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
307 {
308 	char *path;
309 	int ret;
310 
311 	if (!capable(CAP_SYS_ADMIN))
312 		return -EPERM;
313 
314 	if (arg.flags || arg.pad)
315 		return -EINVAL;
316 
317 	path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
318 	ret = PTR_ERR_OR_ZERO(path);
319 	if (ret)
320 		return ret;
321 
322 	ret = bch2_dev_add(c, path);
323 	if (!IS_ERR(path))
324 		kfree(path);
325 
326 	return ret;
327 }
328 
329 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
330 {
331 	struct bch_dev *ca;
332 
333 	if (!capable(CAP_SYS_ADMIN))
334 		return -EPERM;
335 
336 	if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
337 			   BCH_FORCE_IF_METADATA_LOST|
338 			   BCH_FORCE_IF_DEGRADED|
339 			   BCH_BY_INDEX)) ||
340 	    arg.pad)
341 		return -EINVAL;
342 
343 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
344 	if (IS_ERR(ca))
345 		return PTR_ERR(ca);
346 
347 	return bch2_dev_remove(c, ca, arg.flags);
348 }
349 
350 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
351 {
352 	char *path;
353 	int ret;
354 
355 	if (!capable(CAP_SYS_ADMIN))
356 		return -EPERM;
357 
358 	if (arg.flags || arg.pad)
359 		return -EINVAL;
360 
361 	path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
362 	ret = PTR_ERR_OR_ZERO(path);
363 	if (ret)
364 		return ret;
365 
366 	ret = bch2_dev_online(c, path);
367 	kfree(path);
368 	return ret;
369 }
370 
371 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
372 {
373 	struct bch_dev *ca;
374 	int ret;
375 
376 	if (!capable(CAP_SYS_ADMIN))
377 		return -EPERM;
378 
379 	if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
380 			   BCH_FORCE_IF_METADATA_LOST|
381 			   BCH_FORCE_IF_DEGRADED|
382 			   BCH_BY_INDEX)) ||
383 	    arg.pad)
384 		return -EINVAL;
385 
386 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
387 	if (IS_ERR(ca))
388 		return PTR_ERR(ca);
389 
390 	ret = bch2_dev_offline(c, ca, arg.flags);
391 	bch2_dev_put(ca);
392 	return ret;
393 }
394 
395 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
396 			struct bch_ioctl_disk_set_state arg)
397 {
398 	struct bch_dev *ca;
399 	int ret;
400 
401 	if (!capable(CAP_SYS_ADMIN))
402 		return -EPERM;
403 
404 	if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
405 			   BCH_FORCE_IF_METADATA_LOST|
406 			   BCH_FORCE_IF_DEGRADED|
407 			   BCH_BY_INDEX)) ||
408 	    arg.pad[0] || arg.pad[1] || arg.pad[2] ||
409 	    arg.new_state >= BCH_MEMBER_STATE_NR)
410 		return -EINVAL;
411 
412 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
413 	if (IS_ERR(ca))
414 		return PTR_ERR(ca);
415 
416 	ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
417 	if (ret)
418 		bch_err(c, "Error setting device state: %s", bch2_err_str(ret));
419 
420 	bch2_dev_put(ca);
421 	return ret;
422 }
423 
424 struct bch_data_ctx {
425 	struct thread_with_file		thr;
426 
427 	struct bch_fs			*c;
428 	struct bch_ioctl_data		arg;
429 	struct bch_move_stats		stats;
430 };
431 
432 static int bch2_data_thread(void *arg)
433 {
434 	struct bch_data_ctx *ctx = container_of(arg, struct bch_data_ctx, thr);
435 
436 	ctx->thr.ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
437 	ctx->stats.data_type = U8_MAX;
438 	return 0;
439 }
440 
441 static int bch2_data_job_release(struct inode *inode, struct file *file)
442 {
443 	struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
444 
445 	bch2_thread_with_file_exit(&ctx->thr);
446 	kfree(ctx);
447 	return 0;
448 }
449 
450 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
451 				  size_t len, loff_t *ppos)
452 {
453 	struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
454 	struct bch_fs *c = ctx->c;
455 	struct bch_ioctl_data_event e = {
456 		.type			= BCH_DATA_EVENT_PROGRESS,
457 		.p.data_type		= ctx->stats.data_type,
458 		.p.btree_id		= ctx->stats.pos.btree,
459 		.p.pos			= ctx->stats.pos.pos,
460 		.p.sectors_done		= atomic64_read(&ctx->stats.sectors_seen),
461 		.p.sectors_total	= bch2_fs_usage_read_short(c).used,
462 	};
463 
464 	if (len < sizeof(e))
465 		return -EINVAL;
466 
467 	return copy_to_user_errcode(buf, &e, sizeof(e)) ?: sizeof(e);
468 }
469 
470 static const struct file_operations bcachefs_data_ops = {
471 	.release	= bch2_data_job_release,
472 	.read		= bch2_data_job_read,
473 	.llseek		= no_llseek,
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 = NULL;
507 	struct bch_replicas_usage *dst_e, *dst_end;
508 	struct bch_fs_usage_online *src;
509 	u32 replica_entries_bytes;
510 	unsigned i;
511 	int ret = 0;
512 
513 	if (!test_bit(BCH_FS_started, &c->flags))
514 		return -EINVAL;
515 
516 	if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
517 		return -EFAULT;
518 
519 	arg = kzalloc(size_add(sizeof(*arg), replica_entries_bytes), GFP_KERNEL);
520 	if (!arg)
521 		return -ENOMEM;
522 
523 	src = bch2_fs_usage_read(c);
524 	if (!src) {
525 		ret = -ENOMEM;
526 		goto err;
527 	}
528 
529 	arg->capacity		= c->capacity;
530 	arg->used		= bch2_fs_sectors_used(c, src);
531 	arg->online_reserved	= src->online_reserved;
532 
533 	for (i = 0; i < BCH_REPLICAS_MAX; i++)
534 		arg->persistent_reserved[i] = src->u.persistent_reserved[i];
535 
536 	dst_e	= arg->replicas;
537 	dst_end = (void *) arg->replicas + replica_entries_bytes;
538 
539 	for (i = 0; i < c->replicas.nr; i++) {
540 		struct bch_replicas_entry_v1 *src_e =
541 			cpu_replicas_entry(&c->replicas, i);
542 
543 		/* check that we have enough space for one replicas entry */
544 		if (dst_e + 1 > dst_end) {
545 			ret = -ERANGE;
546 			break;
547 		}
548 
549 		dst_e->sectors		= src->u.replicas[i];
550 		dst_e->r		= *src_e;
551 
552 		/* recheck after setting nr_devs: */
553 		if (replicas_usage_next(dst_e) > dst_end) {
554 			ret = -ERANGE;
555 			break;
556 		}
557 
558 		memcpy(dst_e->r.devs, src_e->devs, src_e->nr_devs);
559 
560 		dst_e = replicas_usage_next(dst_e);
561 	}
562 
563 	arg->replica_entries_bytes = (void *) dst_e - (void *) arg->replicas;
564 
565 	percpu_up_read(&c->mark_lock);
566 	kfree(src);
567 
568 	if (ret)
569 		goto err;
570 
571 	ret = copy_to_user_errcode(user_arg, arg,
572 			sizeof(*arg) + arg->replica_entries_bytes);
573 err:
574 	kfree(arg);
575 	return ret;
576 }
577 
578 /* obsolete, didn't allow for new data types: */
579 static long bch2_ioctl_dev_usage(struct bch_fs *c,
580 				 struct bch_ioctl_dev_usage __user *user_arg)
581 {
582 	struct bch_ioctl_dev_usage arg;
583 	struct bch_dev_usage src;
584 	struct bch_dev *ca;
585 	unsigned i;
586 
587 	if (!test_bit(BCH_FS_started, &c->flags))
588 		return -EINVAL;
589 
590 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
591 		return -EFAULT;
592 
593 	if ((arg.flags & ~BCH_BY_INDEX) ||
594 	    arg.pad[0] ||
595 	    arg.pad[1] ||
596 	    arg.pad[2])
597 		return -EINVAL;
598 
599 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
600 	if (IS_ERR(ca))
601 		return PTR_ERR(ca);
602 
603 	src = bch2_dev_usage_read(ca);
604 
605 	arg.state		= ca->mi.state;
606 	arg.bucket_size		= ca->mi.bucket_size;
607 	arg.nr_buckets		= ca->mi.nbuckets - ca->mi.first_bucket;
608 
609 	for (i = 0; i < BCH_DATA_NR; i++) {
610 		arg.d[i].buckets	= src.d[i].buckets;
611 		arg.d[i].sectors	= src.d[i].sectors;
612 		arg.d[i].fragmented	= src.d[i].fragmented;
613 	}
614 
615 	bch2_dev_put(ca);
616 
617 	return copy_to_user_errcode(user_arg, &arg, sizeof(arg));
618 }
619 
620 static long bch2_ioctl_dev_usage_v2(struct bch_fs *c,
621 				 struct bch_ioctl_dev_usage_v2 __user *user_arg)
622 {
623 	struct bch_ioctl_dev_usage_v2 arg;
624 	struct bch_dev_usage src;
625 	struct bch_dev *ca;
626 	int ret = 0;
627 
628 	if (!test_bit(BCH_FS_started, &c->flags))
629 		return -EINVAL;
630 
631 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
632 		return -EFAULT;
633 
634 	if ((arg.flags & ~BCH_BY_INDEX) ||
635 	    arg.pad[0] ||
636 	    arg.pad[1] ||
637 	    arg.pad[2])
638 		return -EINVAL;
639 
640 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
641 	if (IS_ERR(ca))
642 		return PTR_ERR(ca);
643 
644 	src = bch2_dev_usage_read(ca);
645 
646 	arg.state		= ca->mi.state;
647 	arg.bucket_size		= ca->mi.bucket_size;
648 	arg.nr_data_types	= min(arg.nr_data_types, BCH_DATA_NR);
649 	arg.nr_buckets		= ca->mi.nbuckets - ca->mi.first_bucket;
650 
651 	ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
652 	if (ret)
653 		goto err;
654 
655 	for (unsigned i = 0; i < arg.nr_data_types; i++) {
656 		struct bch_ioctl_dev_usage_type t = {
657 			.buckets	= src.d[i].buckets,
658 			.sectors	= src.d[i].sectors,
659 			.fragmented	= src.d[i].fragmented,
660 		};
661 
662 		ret = copy_to_user_errcode(&user_arg->d[i], &t, sizeof(t));
663 		if (ret)
664 			goto err;
665 	}
666 err:
667 	bch2_dev_put(ca);
668 	return ret;
669 }
670 
671 static long bch2_ioctl_read_super(struct bch_fs *c,
672 				  struct bch_ioctl_read_super arg)
673 {
674 	struct bch_dev *ca = NULL;
675 	struct bch_sb *sb;
676 	int ret = 0;
677 
678 	if (!capable(CAP_SYS_ADMIN))
679 		return -EPERM;
680 
681 	if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
682 	    arg.pad)
683 		return -EINVAL;
684 
685 	mutex_lock(&c->sb_lock);
686 
687 	if (arg.flags & BCH_READ_DEV) {
688 		ca = bch2_device_lookup(c, arg.dev, arg.flags);
689 		ret = PTR_ERR_OR_ZERO(ca);
690 		if (ret)
691 			goto err_unlock;
692 
693 		sb = ca->disk_sb.sb;
694 	} else {
695 		sb = c->disk_sb.sb;
696 	}
697 
698 	if (vstruct_bytes(sb) > arg.size) {
699 		ret = -ERANGE;
700 		goto err;
701 	}
702 
703 	ret = copy_to_user_errcode((void __user *)(unsigned long)arg.sb, sb,
704 				   vstruct_bytes(sb));
705 err:
706 	bch2_dev_put(ca);
707 err_unlock:
708 	mutex_unlock(&c->sb_lock);
709 	return ret;
710 }
711 
712 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
713 				    struct bch_ioctl_disk_get_idx arg)
714 {
715 	dev_t dev = huge_decode_dev(arg.dev);
716 
717 	if (!capable(CAP_SYS_ADMIN))
718 		return -EPERM;
719 
720 	if (!dev)
721 		return -EINVAL;
722 
723 	for_each_online_member(c, ca)
724 		if (ca->dev == dev) {
725 			percpu_ref_put(&ca->io_ref);
726 			return ca->dev_idx;
727 		}
728 
729 	return -BCH_ERR_ENOENT_dev_idx_not_found;
730 }
731 
732 static long bch2_ioctl_disk_resize(struct bch_fs *c,
733 				   struct bch_ioctl_disk_resize arg)
734 {
735 	struct bch_dev *ca;
736 	int ret;
737 
738 	if (!capable(CAP_SYS_ADMIN))
739 		return -EPERM;
740 
741 	if ((arg.flags & ~BCH_BY_INDEX) ||
742 	    arg.pad)
743 		return -EINVAL;
744 
745 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
746 	if (IS_ERR(ca))
747 		return PTR_ERR(ca);
748 
749 	ret = bch2_dev_resize(c, ca, arg.nbuckets);
750 
751 	bch2_dev_put(ca);
752 	return ret;
753 }
754 
755 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
756 				   struct bch_ioctl_disk_resize_journal arg)
757 {
758 	struct bch_dev *ca;
759 	int ret;
760 
761 	if (!capable(CAP_SYS_ADMIN))
762 		return -EPERM;
763 
764 	if ((arg.flags & ~BCH_BY_INDEX) ||
765 	    arg.pad)
766 		return -EINVAL;
767 
768 	if (arg.nbuckets > U32_MAX)
769 		return -EINVAL;
770 
771 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
772 	if (IS_ERR(ca))
773 		return PTR_ERR(ca);
774 
775 	ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
776 
777 	bch2_dev_put(ca);
778 	return ret;
779 }
780 
781 static int bch2_fsck_online_thread_fn(struct thread_with_stdio *stdio)
782 {
783 	struct fsck_thread *thr = container_of(stdio, struct fsck_thread, thr);
784 	struct bch_fs *c = thr->c;
785 
786 	c->stdio_filter = current;
787 	c->stdio = &thr->thr.stdio;
788 
789 	/*
790 	 * XXX: can we figure out a way to do this without mucking with c->opts?
791 	 */
792 	unsigned old_fix_errors = c->opts.fix_errors;
793 	if (opt_defined(thr->opts, fix_errors))
794 		c->opts.fix_errors = thr->opts.fix_errors;
795 	else
796 		c->opts.fix_errors = FSCK_FIX_ask;
797 
798 	c->opts.fsck = true;
799 	set_bit(BCH_FS_fsck_running, &c->flags);
800 
801 	c->curr_recovery_pass = BCH_RECOVERY_PASS_check_alloc_info;
802 	int ret = bch2_run_online_recovery_passes(c);
803 
804 	clear_bit(BCH_FS_fsck_running, &c->flags);
805 	bch_err_fn(c, ret);
806 
807 	c->stdio = NULL;
808 	c->stdio_filter = NULL;
809 	c->opts.fix_errors = old_fix_errors;
810 
811 	up(&c->online_fsck_mutex);
812 	bch2_ro_ref_put(c);
813 	return ret;
814 }
815 
816 static const struct thread_with_stdio_ops bch2_online_fsck_ops = {
817 	.exit		= bch2_fsck_thread_exit,
818 	.fn		= bch2_fsck_online_thread_fn,
819 };
820 
821 static long bch2_ioctl_fsck_online(struct bch_fs *c,
822 				   struct bch_ioctl_fsck_online arg)
823 {
824 	struct fsck_thread *thr = NULL;
825 	long ret = 0;
826 
827 	if (arg.flags)
828 		return -EINVAL;
829 
830 	if (!capable(CAP_SYS_ADMIN))
831 		return -EPERM;
832 
833 	if (!bch2_ro_ref_tryget(c))
834 		return -EROFS;
835 
836 	if (down_trylock(&c->online_fsck_mutex)) {
837 		bch2_ro_ref_put(c);
838 		return -EAGAIN;
839 	}
840 
841 	thr = kzalloc(sizeof(*thr), GFP_KERNEL);
842 	if (!thr) {
843 		ret = -ENOMEM;
844 		goto err;
845 	}
846 
847 	thr->c = c;
848 	thr->opts = bch2_opts_empty();
849 
850 	if (arg.opts) {
851 		char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
852 
853 		ret =   PTR_ERR_OR_ZERO(optstr) ?:
854 			bch2_parse_mount_opts(c, &thr->opts, optstr);
855 		if (!IS_ERR(optstr))
856 			kfree(optstr);
857 
858 		if (ret)
859 			goto err;
860 	}
861 
862 	ret = bch2_run_thread_with_stdio(&thr->thr, &bch2_online_fsck_ops);
863 err:
864 	if (ret < 0) {
865 		bch_err_fn(c, ret);
866 		if (thr)
867 			bch2_fsck_thread_exit(&thr->thr);
868 		up(&c->online_fsck_mutex);
869 		bch2_ro_ref_put(c);
870 	}
871 	return ret;
872 }
873 
874 #define BCH_IOCTL(_name, _argtype)					\
875 do {									\
876 	_argtype i;							\
877 									\
878 	if (copy_from_user(&i, arg, sizeof(i)))				\
879 		return -EFAULT;						\
880 	ret = bch2_ioctl_##_name(c, i);					\
881 	goto out;							\
882 } while (0)
883 
884 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
885 {
886 	long ret;
887 
888 	switch (cmd) {
889 	case BCH_IOCTL_QUERY_UUID:
890 		return bch2_ioctl_query_uuid(c, arg);
891 	case BCH_IOCTL_FS_USAGE:
892 		return bch2_ioctl_fs_usage(c, arg);
893 	case BCH_IOCTL_DEV_USAGE:
894 		return bch2_ioctl_dev_usage(c, arg);
895 	case BCH_IOCTL_DEV_USAGE_V2:
896 		return bch2_ioctl_dev_usage_v2(c, arg);
897 #if 0
898 	case BCH_IOCTL_START:
899 		BCH_IOCTL(start, struct bch_ioctl_start);
900 	case BCH_IOCTL_STOP:
901 		return bch2_ioctl_stop(c);
902 #endif
903 	case BCH_IOCTL_READ_SUPER:
904 		BCH_IOCTL(read_super, struct bch_ioctl_read_super);
905 	case BCH_IOCTL_DISK_GET_IDX:
906 		BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
907 	}
908 
909 	if (!test_bit(BCH_FS_started, &c->flags))
910 		return -EINVAL;
911 
912 	switch (cmd) {
913 	case BCH_IOCTL_DISK_ADD:
914 		BCH_IOCTL(disk_add, struct bch_ioctl_disk);
915 	case BCH_IOCTL_DISK_REMOVE:
916 		BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
917 	case BCH_IOCTL_DISK_ONLINE:
918 		BCH_IOCTL(disk_online, struct bch_ioctl_disk);
919 	case BCH_IOCTL_DISK_OFFLINE:
920 		BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
921 	case BCH_IOCTL_DISK_SET_STATE:
922 		BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
923 	case BCH_IOCTL_DATA:
924 		BCH_IOCTL(data, struct bch_ioctl_data);
925 	case BCH_IOCTL_DISK_RESIZE:
926 		BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
927 	case BCH_IOCTL_DISK_RESIZE_JOURNAL:
928 		BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
929 	case BCH_IOCTL_FSCK_ONLINE:
930 		BCH_IOCTL(fsck_online, struct bch_ioctl_fsck_online);
931 	default:
932 		return -ENOTTY;
933 	}
934 out:
935 	if (ret < 0)
936 		ret = bch2_err_class(ret);
937 	return ret;
938 }
939 
940 static DEFINE_IDR(bch_chardev_minor);
941 
942 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
943 {
944 	unsigned minor = iminor(file_inode(filp));
945 	struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
946 	void __user *arg = (void __user *) v;
947 
948 	return c
949 		? bch2_fs_ioctl(c, cmd, arg)
950 		: bch2_global_ioctl(cmd, arg);
951 }
952 
953 static const struct file_operations bch_chardev_fops = {
954 	.owner		= THIS_MODULE,
955 	.unlocked_ioctl = bch2_chardev_ioctl,
956 	.open		= nonseekable_open,
957 };
958 
959 static int bch_chardev_major;
960 static const struct class bch_chardev_class = {
961 	.name = "bcachefs",
962 };
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 	device_destroy(&bch_chardev_class, MKDEV(bch_chardev_major, U8_MAX));
991 	class_unregister(&bch_chardev_class);
992 	if (bch_chardev_major > 0)
993 		unregister_chrdev(bch_chardev_major, "bcachefs");
994 }
995 
996 int __init bch2_chardev_init(void)
997 {
998 	int ret;
999 
1000 	bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
1001 	if (bch_chardev_major < 0)
1002 		return bch_chardev_major;
1003 
1004 	ret = class_register(&bch_chardev_class);
1005 	if (ret)
1006 		goto major_out;
1007 
1008 	bch_chardev = device_create(&bch_chardev_class, NULL,
1009 				    MKDEV(bch_chardev_major, U8_MAX),
1010 				    NULL, "bcachefs-ctl");
1011 	if (IS_ERR(bch_chardev)) {
1012 		ret = PTR_ERR(bch_chardev);
1013 		goto class_out;
1014 	}
1015 
1016 	return 0;
1017 
1018 class_out:
1019 	class_unregister(&bch_chardev_class);
1020 major_out:
1021 	unregister_chrdev(bch_chardev_major, "bcachefs-ctl");
1022 	return ret;
1023 }
1024 
1025 #endif /* NO_BCACHEFS_CHARDEV */
1026