xref: /linux/fs/bcachefs/chardev.c (revision 8886640dade4ae2595fcdce511c8bcc716aa47d3)
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.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 __must_check
26 static int copy_to_user_errcode(void __user *to, const void *from, unsigned long n)
27 {
28 	return copy_to_user(to, from, n) ? -EFAULT : 0;
29 }
30 
31 /* returns with ref on ca->ref */
32 static struct bch_dev *bch2_device_lookup(struct bch_fs *c, u64 dev,
33 					  unsigned flags)
34 {
35 	struct bch_dev *ca;
36 
37 	if (flags & BCH_BY_INDEX) {
38 		if (dev >= c->sb.nr_devices)
39 			return ERR_PTR(-EINVAL);
40 
41 		rcu_read_lock();
42 		ca = rcu_dereference(c->devs[dev]);
43 		if (ca)
44 			percpu_ref_get(&ca->ref);
45 		rcu_read_unlock();
46 
47 		if (!ca)
48 			return ERR_PTR(-EINVAL);
49 	} else {
50 		char *path;
51 
52 		path = strndup_user((const char __user *)
53 				    (unsigned long) dev, PATH_MAX);
54 		if (IS_ERR(path))
55 			return ERR_CAST(path);
56 
57 		ca = bch2_dev_lookup(c, path);
58 		kfree(path);
59 	}
60 
61 	return ca;
62 }
63 
64 #if 0
65 static long bch2_ioctl_assemble(struct bch_ioctl_assemble __user *user_arg)
66 {
67 	struct bch_ioctl_assemble arg;
68 	struct bch_fs *c;
69 	u64 *user_devs = NULL;
70 	char **devs = NULL;
71 	unsigned i;
72 	int ret = -EFAULT;
73 
74 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
75 		return -EFAULT;
76 
77 	if (arg.flags || arg.pad)
78 		return -EINVAL;
79 
80 	user_devs = kmalloc_array(arg.nr_devs, sizeof(u64), GFP_KERNEL);
81 	if (!user_devs)
82 		return -ENOMEM;
83 
84 	devs = kcalloc(arg.nr_devs, sizeof(char *), GFP_KERNEL);
85 
86 	if (copy_from_user(user_devs, user_arg->devs,
87 			   sizeof(u64) * arg.nr_devs))
88 		goto err;
89 
90 	for (i = 0; i < arg.nr_devs; i++) {
91 		devs[i] = strndup_user((const char __user *)(unsigned long)
92 				       user_devs[i],
93 				       PATH_MAX);
94 		ret= PTR_ERR_OR_ZERO(devs[i]);
95 		if (ret)
96 			goto err;
97 	}
98 
99 	c = bch2_fs_open(devs, arg.nr_devs, bch2_opts_empty());
100 	ret = PTR_ERR_OR_ZERO(c);
101 	if (!ret)
102 		closure_put(&c->cl);
103 err:
104 	if (devs)
105 		for (i = 0; i < arg.nr_devs; i++)
106 			kfree(devs[i]);
107 	kfree(devs);
108 	return ret;
109 }
110 
111 static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg)
112 {
113 	struct bch_ioctl_incremental arg;
114 	const char *err;
115 	char *path;
116 
117 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
118 		return -EFAULT;
119 
120 	if (arg.flags || arg.pad)
121 		return -EINVAL;
122 
123 	path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
124 	ret = PTR_ERR_OR_ZERO(path);
125 	if (ret)
126 		return ret;
127 
128 	err = bch2_fs_open_incremental(path);
129 	kfree(path);
130 
131 	if (err) {
132 		pr_err("Could not register bcachefs devices: %s", err);
133 		return -EINVAL;
134 	}
135 
136 	return 0;
137 }
138 #endif
139 
140 struct fsck_thread {
141 	struct thread_with_stdio thr;
142 	struct bch_fs		*c;
143 	char			**devs;
144 	size_t			nr_devs;
145 	struct bch_opts		opts;
146 };
147 
148 static void bch2_fsck_thread_exit(struct thread_with_stdio *_thr)
149 {
150 	struct fsck_thread *thr = container_of(_thr, struct fsck_thread, thr);
151 	if (thr->devs)
152 		for (size_t i = 0; i < thr->nr_devs; i++)
153 			kfree(thr->devs[i]);
154 	kfree(thr->devs);
155 	kfree(thr);
156 }
157 
158 static int bch2_fsck_offline_thread_fn(void *arg)
159 {
160 	struct fsck_thread *thr = container_of(arg, struct fsck_thread, thr);
161 	struct bch_fs *c = bch2_fs_open(thr->devs, thr->nr_devs, thr->opts);
162 
163 	thr->thr.thr.ret = PTR_ERR_OR_ZERO(c);
164 	if (!thr->thr.thr.ret)
165 		bch2_fs_stop(c);
166 
167 	thread_with_stdio_done(&thr->thr);
168 	return 0;
169 }
170 
171 static long bch2_ioctl_fsck_offline(struct bch_ioctl_fsck_offline __user *user_arg)
172 {
173 	struct bch_ioctl_fsck_offline arg;
174 	struct fsck_thread *thr = NULL;
175 	u64 *devs = NULL;
176 	long ret = 0;
177 
178 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
179 		return -EFAULT;
180 
181 	if (arg.flags)
182 		return -EINVAL;
183 
184 	if (!capable(CAP_SYS_ADMIN))
185 		return -EPERM;
186 
187 	if (!(devs = kcalloc(arg.nr_devs, sizeof(*devs), GFP_KERNEL)) ||
188 	    !(thr = kzalloc(sizeof(*thr), GFP_KERNEL)) ||
189 	    !(thr->devs = kcalloc(arg.nr_devs, sizeof(*thr->devs), GFP_KERNEL))) {
190 		ret = -ENOMEM;
191 		goto err;
192 	}
193 
194 	thr->opts = bch2_opts_empty();
195 	thr->nr_devs = arg.nr_devs;
196 
197 	if (copy_from_user(devs, &user_arg->devs[0],
198 			   array_size(sizeof(user_arg->devs[0]), arg.nr_devs))) {
199 		ret = -EINVAL;
200 		goto err;
201 	}
202 
203 	for (size_t i = 0; i < arg.nr_devs; i++) {
204 		thr->devs[i] = strndup_user((char __user *)(unsigned long) devs[i], PATH_MAX);
205 		ret = PTR_ERR_OR_ZERO(thr->devs[i]);
206 		if (ret)
207 			goto err;
208 	}
209 
210 	if (arg.opts) {
211 		char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
212 
213 		ret =   PTR_ERR_OR_ZERO(optstr) ?:
214 			bch2_parse_mount_opts(NULL, &thr->opts, optstr);
215 		kfree(optstr);
216 
217 		if (ret)
218 			goto err;
219 	}
220 
221 	opt_set(thr->opts, stdio, (u64)(unsigned long)&thr->thr.stdio);
222 
223 	ret = bch2_run_thread_with_stdio(&thr->thr,
224 			bch2_fsck_thread_exit,
225 			bch2_fsck_offline_thread_fn);
226 err:
227 	if (ret < 0) {
228 		if (thr)
229 			bch2_fsck_thread_exit(&thr->thr);
230 		pr_err("ret %s", bch2_err_str(ret));
231 	}
232 	kfree(devs);
233 	return ret;
234 }
235 
236 static long bch2_global_ioctl(unsigned cmd, void __user *arg)
237 {
238 	long ret;
239 
240 	switch (cmd) {
241 #if 0
242 	case BCH_IOCTL_ASSEMBLE:
243 		return bch2_ioctl_assemble(arg);
244 	case BCH_IOCTL_INCREMENTAL:
245 		return bch2_ioctl_incremental(arg);
246 #endif
247 	case BCH_IOCTL_FSCK_OFFLINE: {
248 		ret = bch2_ioctl_fsck_offline(arg);
249 		break;
250 	}
251 	default:
252 		ret = -ENOTTY;
253 		break;
254 	}
255 
256 	if (ret < 0)
257 		ret = bch2_err_class(ret);
258 	return ret;
259 }
260 
261 static long bch2_ioctl_query_uuid(struct bch_fs *c,
262 			struct bch_ioctl_query_uuid __user *user_arg)
263 {
264 	return copy_to_user_errcode(&user_arg->uuid, &c->sb.user_uuid,
265 				    sizeof(c->sb.user_uuid));
266 }
267 
268 #if 0
269 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
270 {
271 	if (!capable(CAP_SYS_ADMIN))
272 		return -EPERM;
273 
274 	if (arg.flags || arg.pad)
275 		return -EINVAL;
276 
277 	return bch2_fs_start(c);
278 }
279 
280 static long bch2_ioctl_stop(struct bch_fs *c)
281 {
282 	if (!capable(CAP_SYS_ADMIN))
283 		return -EPERM;
284 
285 	bch2_fs_stop(c);
286 	return 0;
287 }
288 #endif
289 
290 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
291 {
292 	char *path;
293 	int ret;
294 
295 	if (!capable(CAP_SYS_ADMIN))
296 		return -EPERM;
297 
298 	if (arg.flags || arg.pad)
299 		return -EINVAL;
300 
301 	path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
302 	ret = PTR_ERR_OR_ZERO(path);
303 	if (ret)
304 		return ret;
305 
306 	ret = bch2_dev_add(c, path);
307 	kfree(path);
308 
309 	return ret;
310 }
311 
312 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
313 {
314 	struct bch_dev *ca;
315 
316 	if (!capable(CAP_SYS_ADMIN))
317 		return -EPERM;
318 
319 	if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
320 			   BCH_FORCE_IF_METADATA_LOST|
321 			   BCH_FORCE_IF_DEGRADED|
322 			   BCH_BY_INDEX)) ||
323 	    arg.pad)
324 		return -EINVAL;
325 
326 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
327 	if (IS_ERR(ca))
328 		return PTR_ERR(ca);
329 
330 	return bch2_dev_remove(c, ca, arg.flags);
331 }
332 
333 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
334 {
335 	char *path;
336 	int ret;
337 
338 	if (!capable(CAP_SYS_ADMIN))
339 		return -EPERM;
340 
341 	if (arg.flags || arg.pad)
342 		return -EINVAL;
343 
344 	path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
345 	ret = PTR_ERR_OR_ZERO(path);
346 	if (ret)
347 		return ret;
348 
349 	ret = bch2_dev_online(c, path);
350 	kfree(path);
351 	return ret;
352 }
353 
354 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
355 {
356 	struct bch_dev *ca;
357 	int ret;
358 
359 	if (!capable(CAP_SYS_ADMIN))
360 		return -EPERM;
361 
362 	if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
363 			   BCH_FORCE_IF_METADATA_LOST|
364 			   BCH_FORCE_IF_DEGRADED|
365 			   BCH_BY_INDEX)) ||
366 	    arg.pad)
367 		return -EINVAL;
368 
369 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
370 	if (IS_ERR(ca))
371 		return PTR_ERR(ca);
372 
373 	ret = bch2_dev_offline(c, ca, arg.flags);
374 	percpu_ref_put(&ca->ref);
375 	return ret;
376 }
377 
378 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
379 			struct bch_ioctl_disk_set_state arg)
380 {
381 	struct bch_dev *ca;
382 	int ret;
383 
384 	if (!capable(CAP_SYS_ADMIN))
385 		return -EPERM;
386 
387 	if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
388 			   BCH_FORCE_IF_METADATA_LOST|
389 			   BCH_FORCE_IF_DEGRADED|
390 			   BCH_BY_INDEX)) ||
391 	    arg.pad[0] || arg.pad[1] || arg.pad[2] ||
392 	    arg.new_state >= BCH_MEMBER_STATE_NR)
393 		return -EINVAL;
394 
395 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
396 	if (IS_ERR(ca))
397 		return PTR_ERR(ca);
398 
399 	ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
400 	if (ret)
401 		bch_err(c, "Error setting device state: %s", bch2_err_str(ret));
402 
403 	percpu_ref_put(&ca->ref);
404 	return ret;
405 }
406 
407 struct bch_data_ctx {
408 	struct thread_with_file		thr;
409 
410 	struct bch_fs			*c;
411 	struct bch_ioctl_data		arg;
412 	struct bch_move_stats		stats;
413 };
414 
415 static int bch2_data_thread(void *arg)
416 {
417 	struct bch_data_ctx *ctx = container_of(arg, struct bch_data_ctx, thr);
418 
419 	ctx->thr.ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
420 	ctx->stats.data_type = U8_MAX;
421 	return 0;
422 }
423 
424 static int bch2_data_job_release(struct inode *inode, struct file *file)
425 {
426 	struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
427 
428 	bch2_thread_with_file_exit(&ctx->thr);
429 	kfree(ctx);
430 	return 0;
431 }
432 
433 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
434 				  size_t len, loff_t *ppos)
435 {
436 	struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
437 	struct bch_fs *c = ctx->c;
438 	struct bch_ioctl_data_event e = {
439 		.type			= BCH_DATA_EVENT_PROGRESS,
440 		.p.data_type		= ctx->stats.data_type,
441 		.p.btree_id		= ctx->stats.pos.btree,
442 		.p.pos			= ctx->stats.pos.pos,
443 		.p.sectors_done		= atomic64_read(&ctx->stats.sectors_seen),
444 		.p.sectors_total	= bch2_fs_usage_read_short(c).used,
445 	};
446 
447 	if (len < sizeof(e))
448 		return -EINVAL;
449 
450 	return copy_to_user_errcode(buf, &e, sizeof(e)) ?: sizeof(e);
451 }
452 
453 static const struct file_operations bcachefs_data_ops = {
454 	.release	= bch2_data_job_release,
455 	.read		= bch2_data_job_read,
456 	.llseek		= no_llseek,
457 };
458 
459 static long bch2_ioctl_data(struct bch_fs *c,
460 			    struct bch_ioctl_data arg)
461 {
462 	struct bch_data_ctx *ctx;
463 	int ret;
464 
465 	if (!capable(CAP_SYS_ADMIN))
466 		return -EPERM;
467 
468 	if (arg.op >= BCH_DATA_OP_NR || arg.flags)
469 		return -EINVAL;
470 
471 	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
472 	if (!ctx)
473 		return -ENOMEM;
474 
475 	ctx->c = c;
476 	ctx->arg = arg;
477 
478 	ret = bch2_run_thread_with_file(&ctx->thr,
479 			&bcachefs_data_ops,
480 			bch2_data_thread);
481 	if (ret < 0)
482 		kfree(ctx);
483 	return ret;
484 }
485 
486 static long bch2_ioctl_fs_usage(struct bch_fs *c,
487 				struct bch_ioctl_fs_usage __user *user_arg)
488 {
489 	struct bch_ioctl_fs_usage *arg = NULL;
490 	struct bch_replicas_usage *dst_e, *dst_end;
491 	struct bch_fs_usage_online *src;
492 	u32 replica_entries_bytes;
493 	unsigned i;
494 	int ret = 0;
495 
496 	if (!test_bit(BCH_FS_started, &c->flags))
497 		return -EINVAL;
498 
499 	if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
500 		return -EFAULT;
501 
502 	arg = kzalloc(size_add(sizeof(*arg), replica_entries_bytes), GFP_KERNEL);
503 	if (!arg)
504 		return -ENOMEM;
505 
506 	src = bch2_fs_usage_read(c);
507 	if (!src) {
508 		ret = -ENOMEM;
509 		goto err;
510 	}
511 
512 	arg->capacity		= c->capacity;
513 	arg->used		= bch2_fs_sectors_used(c, src);
514 	arg->online_reserved	= src->online_reserved;
515 
516 	for (i = 0; i < BCH_REPLICAS_MAX; i++)
517 		arg->persistent_reserved[i] = src->u.persistent_reserved[i];
518 
519 	dst_e	= arg->replicas;
520 	dst_end = (void *) arg->replicas + replica_entries_bytes;
521 
522 	for (i = 0; i < c->replicas.nr; i++) {
523 		struct bch_replicas_entry_v1 *src_e =
524 			cpu_replicas_entry(&c->replicas, i);
525 
526 		/* check that we have enough space for one replicas entry */
527 		if (dst_e + 1 > dst_end) {
528 			ret = -ERANGE;
529 			break;
530 		}
531 
532 		dst_e->sectors		= src->u.replicas[i];
533 		dst_e->r		= *src_e;
534 
535 		/* recheck after setting nr_devs: */
536 		if (replicas_usage_next(dst_e) > dst_end) {
537 			ret = -ERANGE;
538 			break;
539 		}
540 
541 		memcpy(dst_e->r.devs, src_e->devs, src_e->nr_devs);
542 
543 		dst_e = replicas_usage_next(dst_e);
544 	}
545 
546 	arg->replica_entries_bytes = (void *) dst_e - (void *) arg->replicas;
547 
548 	percpu_up_read(&c->mark_lock);
549 	kfree(src);
550 
551 	if (ret)
552 		goto err;
553 
554 	ret = copy_to_user_errcode(user_arg, arg,
555 			sizeof(*arg) + arg->replica_entries_bytes);
556 err:
557 	kfree(arg);
558 	return ret;
559 }
560 
561 /* obsolete, didn't allow for new data types: */
562 static long bch2_ioctl_dev_usage(struct bch_fs *c,
563 				 struct bch_ioctl_dev_usage __user *user_arg)
564 {
565 	struct bch_ioctl_dev_usage arg;
566 	struct bch_dev_usage src;
567 	struct bch_dev *ca;
568 	unsigned i;
569 
570 	if (!test_bit(BCH_FS_started, &c->flags))
571 		return -EINVAL;
572 
573 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
574 		return -EFAULT;
575 
576 	if ((arg.flags & ~BCH_BY_INDEX) ||
577 	    arg.pad[0] ||
578 	    arg.pad[1] ||
579 	    arg.pad[2])
580 		return -EINVAL;
581 
582 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
583 	if (IS_ERR(ca))
584 		return PTR_ERR(ca);
585 
586 	src = bch2_dev_usage_read(ca);
587 
588 	arg.state		= ca->mi.state;
589 	arg.bucket_size		= ca->mi.bucket_size;
590 	arg.nr_buckets		= ca->mi.nbuckets - ca->mi.first_bucket;
591 
592 	for (i = 0; i < BCH_DATA_NR; i++) {
593 		arg.d[i].buckets	= src.d[i].buckets;
594 		arg.d[i].sectors	= src.d[i].sectors;
595 		arg.d[i].fragmented	= src.d[i].fragmented;
596 	}
597 
598 	percpu_ref_put(&ca->ref);
599 
600 	return copy_to_user_errcode(user_arg, &arg, sizeof(arg));
601 }
602 
603 static long bch2_ioctl_dev_usage_v2(struct bch_fs *c,
604 				 struct bch_ioctl_dev_usage_v2 __user *user_arg)
605 {
606 	struct bch_ioctl_dev_usage_v2 arg;
607 	struct bch_dev_usage src;
608 	struct bch_dev *ca;
609 	int ret = 0;
610 
611 	if (!test_bit(BCH_FS_started, &c->flags))
612 		return -EINVAL;
613 
614 	if (copy_from_user(&arg, user_arg, sizeof(arg)))
615 		return -EFAULT;
616 
617 	if ((arg.flags & ~BCH_BY_INDEX) ||
618 	    arg.pad[0] ||
619 	    arg.pad[1] ||
620 	    arg.pad[2])
621 		return -EINVAL;
622 
623 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
624 	if (IS_ERR(ca))
625 		return PTR_ERR(ca);
626 
627 	src = bch2_dev_usage_read(ca);
628 
629 	arg.state		= ca->mi.state;
630 	arg.bucket_size		= ca->mi.bucket_size;
631 	arg.nr_data_types	= min(arg.nr_data_types, BCH_DATA_NR);
632 	arg.nr_buckets		= ca->mi.nbuckets - ca->mi.first_bucket;
633 
634 	ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
635 	if (ret)
636 		goto err;
637 
638 	for (unsigned i = 0; i < arg.nr_data_types; i++) {
639 		struct bch_ioctl_dev_usage_type t = {
640 			.buckets	= src.d[i].buckets,
641 			.sectors	= src.d[i].sectors,
642 			.fragmented	= src.d[i].fragmented,
643 		};
644 
645 		ret = copy_to_user_errcode(&user_arg->d[i], &t, sizeof(t));
646 		if (ret)
647 			goto err;
648 	}
649 err:
650 	percpu_ref_put(&ca->ref);
651 	return ret;
652 }
653 
654 static long bch2_ioctl_read_super(struct bch_fs *c,
655 				  struct bch_ioctl_read_super arg)
656 {
657 	struct bch_dev *ca = NULL;
658 	struct bch_sb *sb;
659 	int ret = 0;
660 
661 	if (!capable(CAP_SYS_ADMIN))
662 		return -EPERM;
663 
664 	if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
665 	    arg.pad)
666 		return -EINVAL;
667 
668 	mutex_lock(&c->sb_lock);
669 
670 	if (arg.flags & BCH_READ_DEV) {
671 		ca = bch2_device_lookup(c, arg.dev, arg.flags);
672 
673 		if (IS_ERR(ca)) {
674 			ret = PTR_ERR(ca);
675 			goto err;
676 		}
677 
678 		sb = ca->disk_sb.sb;
679 	} else {
680 		sb = c->disk_sb.sb;
681 	}
682 
683 	if (vstruct_bytes(sb) > arg.size) {
684 		ret = -ERANGE;
685 		goto err;
686 	}
687 
688 	ret = copy_to_user_errcode((void __user *)(unsigned long)arg.sb, sb,
689 				   vstruct_bytes(sb));
690 err:
691 	if (!IS_ERR_OR_NULL(ca))
692 		percpu_ref_put(&ca->ref);
693 	mutex_unlock(&c->sb_lock);
694 	return ret;
695 }
696 
697 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
698 				    struct bch_ioctl_disk_get_idx arg)
699 {
700 	dev_t dev = huge_decode_dev(arg.dev);
701 
702 	if (!capable(CAP_SYS_ADMIN))
703 		return -EPERM;
704 
705 	if (!dev)
706 		return -EINVAL;
707 
708 	for_each_online_member(c, ca)
709 		if (ca->dev == dev) {
710 			percpu_ref_put(&ca->io_ref);
711 			return ca->dev_idx;
712 		}
713 
714 	return -BCH_ERR_ENOENT_dev_idx_not_found;
715 }
716 
717 static long bch2_ioctl_disk_resize(struct bch_fs *c,
718 				   struct bch_ioctl_disk_resize arg)
719 {
720 	struct bch_dev *ca;
721 	int ret;
722 
723 	if (!capable(CAP_SYS_ADMIN))
724 		return -EPERM;
725 
726 	if ((arg.flags & ~BCH_BY_INDEX) ||
727 	    arg.pad)
728 		return -EINVAL;
729 
730 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
731 	if (IS_ERR(ca))
732 		return PTR_ERR(ca);
733 
734 	ret = bch2_dev_resize(c, ca, arg.nbuckets);
735 
736 	percpu_ref_put(&ca->ref);
737 	return ret;
738 }
739 
740 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
741 				   struct bch_ioctl_disk_resize_journal arg)
742 {
743 	struct bch_dev *ca;
744 	int ret;
745 
746 	if (!capable(CAP_SYS_ADMIN))
747 		return -EPERM;
748 
749 	if ((arg.flags & ~BCH_BY_INDEX) ||
750 	    arg.pad)
751 		return -EINVAL;
752 
753 	if (arg.nbuckets > U32_MAX)
754 		return -EINVAL;
755 
756 	ca = bch2_device_lookup(c, arg.dev, arg.flags);
757 	if (IS_ERR(ca))
758 		return PTR_ERR(ca);
759 
760 	ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
761 
762 	percpu_ref_put(&ca->ref);
763 	return ret;
764 }
765 
766 static int bch2_fsck_online_thread_fn(void *arg)
767 {
768 	struct fsck_thread *thr = container_of(arg, struct fsck_thread, thr);
769 	struct bch_fs *c = thr->c;
770 
771 	c->stdio_filter = current;
772 	c->stdio = &thr->thr.stdio;
773 
774 	/*
775 	 * XXX: can we figure out a way to do this without mucking with c->opts?
776 	 */
777 	unsigned old_fix_errors = c->opts.fix_errors;
778 	if (opt_defined(thr->opts, fix_errors))
779 		c->opts.fix_errors = thr->opts.fix_errors;
780 	else
781 		c->opts.fix_errors = FSCK_FIX_ask;
782 
783 	c->opts.fsck = true;
784 	set_bit(BCH_FS_fsck_running, &c->flags);
785 
786 	c->curr_recovery_pass = BCH_RECOVERY_PASS_check_alloc_info;
787 	int ret = bch2_run_online_recovery_passes(c);
788 
789 	clear_bit(BCH_FS_fsck_running, &c->flags);
790 	bch_err_fn(c, ret);
791 
792 	c->stdio = NULL;
793 	c->stdio_filter = NULL;
794 	c->opts.fix_errors = old_fix_errors;
795 
796 	thread_with_stdio_done(&thr->thr);
797 
798 	up(&c->online_fsck_mutex);
799 	bch2_ro_ref_put(c);
800 	return 0;
801 }
802 
803 static long bch2_ioctl_fsck_online(struct bch_fs *c,
804 				   struct bch_ioctl_fsck_online arg)
805 {
806 	struct fsck_thread *thr = NULL;
807 	long ret = 0;
808 
809 	if (arg.flags)
810 		return -EINVAL;
811 
812 	if (!capable(CAP_SYS_ADMIN))
813 		return -EPERM;
814 
815 	if (!bch2_ro_ref_tryget(c))
816 		return -EROFS;
817 
818 	if (down_trylock(&c->online_fsck_mutex)) {
819 		bch2_ro_ref_put(c);
820 		return -EAGAIN;
821 	}
822 
823 	thr = kzalloc(sizeof(*thr), GFP_KERNEL);
824 	if (!thr) {
825 		ret = -ENOMEM;
826 		goto err;
827 	}
828 
829 	thr->c = c;
830 	thr->opts = bch2_opts_empty();
831 
832 	if (arg.opts) {
833 		char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
834 
835 		ret =   PTR_ERR_OR_ZERO(optstr) ?:
836 			bch2_parse_mount_opts(c, &thr->opts, optstr);
837 		kfree(optstr);
838 
839 		if (ret)
840 			goto err;
841 	}
842 
843 	ret = bch2_run_thread_with_stdio(&thr->thr,
844 			bch2_fsck_thread_exit,
845 			bch2_fsck_online_thread_fn);
846 err:
847 	if (ret < 0) {
848 		bch_err_fn(c, ret);
849 		if (thr)
850 			bch2_fsck_thread_exit(&thr->thr);
851 		up(&c->online_fsck_mutex);
852 		bch2_ro_ref_put(c);
853 	}
854 	return ret;
855 }
856 
857 #define BCH_IOCTL(_name, _argtype)					\
858 do {									\
859 	_argtype i;							\
860 									\
861 	if (copy_from_user(&i, arg, sizeof(i)))				\
862 		return -EFAULT;						\
863 	ret = bch2_ioctl_##_name(c, i);					\
864 	goto out;							\
865 } while (0)
866 
867 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
868 {
869 	long ret;
870 
871 	switch (cmd) {
872 	case BCH_IOCTL_QUERY_UUID:
873 		return bch2_ioctl_query_uuid(c, arg);
874 	case BCH_IOCTL_FS_USAGE:
875 		return bch2_ioctl_fs_usage(c, arg);
876 	case BCH_IOCTL_DEV_USAGE:
877 		return bch2_ioctl_dev_usage(c, arg);
878 	case BCH_IOCTL_DEV_USAGE_V2:
879 		return bch2_ioctl_dev_usage_v2(c, arg);
880 #if 0
881 	case BCH_IOCTL_START:
882 		BCH_IOCTL(start, struct bch_ioctl_start);
883 	case BCH_IOCTL_STOP:
884 		return bch2_ioctl_stop(c);
885 #endif
886 	case BCH_IOCTL_READ_SUPER:
887 		BCH_IOCTL(read_super, struct bch_ioctl_read_super);
888 	case BCH_IOCTL_DISK_GET_IDX:
889 		BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
890 	}
891 
892 	if (!test_bit(BCH_FS_started, &c->flags))
893 		return -EINVAL;
894 
895 	switch (cmd) {
896 	case BCH_IOCTL_DISK_ADD:
897 		BCH_IOCTL(disk_add, struct bch_ioctl_disk);
898 	case BCH_IOCTL_DISK_REMOVE:
899 		BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
900 	case BCH_IOCTL_DISK_ONLINE:
901 		BCH_IOCTL(disk_online, struct bch_ioctl_disk);
902 	case BCH_IOCTL_DISK_OFFLINE:
903 		BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
904 	case BCH_IOCTL_DISK_SET_STATE:
905 		BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
906 	case BCH_IOCTL_DATA:
907 		BCH_IOCTL(data, struct bch_ioctl_data);
908 	case BCH_IOCTL_DISK_RESIZE:
909 		BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
910 	case BCH_IOCTL_DISK_RESIZE_JOURNAL:
911 		BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
912 	case BCH_IOCTL_FSCK_ONLINE:
913 		BCH_IOCTL(fsck_online, struct bch_ioctl_fsck_online);
914 	default:
915 		return -ENOTTY;
916 	}
917 out:
918 	if (ret < 0)
919 		ret = bch2_err_class(ret);
920 	return ret;
921 }
922 
923 static DEFINE_IDR(bch_chardev_minor);
924 
925 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
926 {
927 	unsigned minor = iminor(file_inode(filp));
928 	struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
929 	void __user *arg = (void __user *) v;
930 
931 	return c
932 		? bch2_fs_ioctl(c, cmd, arg)
933 		: bch2_global_ioctl(cmd, arg);
934 }
935 
936 static const struct file_operations bch_chardev_fops = {
937 	.owner		= THIS_MODULE,
938 	.unlocked_ioctl = bch2_chardev_ioctl,
939 	.open		= nonseekable_open,
940 };
941 
942 static int bch_chardev_major;
943 static struct class *bch_chardev_class;
944 static struct device *bch_chardev;
945 
946 void bch2_fs_chardev_exit(struct bch_fs *c)
947 {
948 	if (!IS_ERR_OR_NULL(c->chardev))
949 		device_unregister(c->chardev);
950 	if (c->minor >= 0)
951 		idr_remove(&bch_chardev_minor, c->minor);
952 }
953 
954 int bch2_fs_chardev_init(struct bch_fs *c)
955 {
956 	c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
957 	if (c->minor < 0)
958 		return c->minor;
959 
960 	c->chardev = device_create(bch_chardev_class, NULL,
961 				   MKDEV(bch_chardev_major, c->minor), c,
962 				   "bcachefs%u-ctl", c->minor);
963 	if (IS_ERR(c->chardev))
964 		return PTR_ERR(c->chardev);
965 
966 	return 0;
967 }
968 
969 void bch2_chardev_exit(void)
970 {
971 	if (!IS_ERR_OR_NULL(bch_chardev_class))
972 		device_destroy(bch_chardev_class,
973 			       MKDEV(bch_chardev_major, U8_MAX));
974 	if (!IS_ERR_OR_NULL(bch_chardev_class))
975 		class_destroy(bch_chardev_class);
976 	if (bch_chardev_major > 0)
977 		unregister_chrdev(bch_chardev_major, "bcachefs");
978 }
979 
980 int __init bch2_chardev_init(void)
981 {
982 	bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
983 	if (bch_chardev_major < 0)
984 		return bch_chardev_major;
985 
986 	bch_chardev_class = class_create("bcachefs");
987 	if (IS_ERR(bch_chardev_class))
988 		return PTR_ERR(bch_chardev_class);
989 
990 	bch_chardev = device_create(bch_chardev_class, NULL,
991 				    MKDEV(bch_chardev_major, U8_MAX),
992 				    NULL, "bcachefs-ctl");
993 	if (IS_ERR(bch_chardev))
994 		return PTR_ERR(bch_chardev);
995 
996 	return 0;
997 }
998 
999 #endif /* NO_BCACHEFS_CHARDEV */
1000