1 // SPDX-License-Identifier: GPL-2.0
2 #ifndef NO_BCACHEFS_CHARDEV
5 #include "bcachefs_ioctl.h"
14 #include <linux/anon_inodes.h>
15 #include <linux/cdev.h>
16 #include <linux/device.h>
17 #include <linux/file.h>
19 #include <linux/ioctl.h>
20 #include <linux/kthread.h>
21 #include <linux/major.h>
22 #include <linux/sched/task.h>
23 #include <linux/slab.h>
24 #include <linux/uaccess.h>
26 /* returns with ref on ca->ref */
27 static struct bch_dev *bch2_device_lookup(struct bch_fs *c, u64 dev,
32 if (flags & BCH_BY_INDEX) {
33 if (dev >= c->sb.nr_devices)
34 return ERR_PTR(-EINVAL);
37 ca = rcu_dereference(c->devs[dev]);
39 percpu_ref_get(&ca->ref);
43 return ERR_PTR(-EINVAL);
47 path = strndup_user((const char __user *)
48 (unsigned long) dev, PATH_MAX);
50 return ERR_CAST(path);
52 ca = bch2_dev_lookup(c, path);
60 static long bch2_ioctl_assemble(struct bch_ioctl_assemble __user *user_arg)
62 struct bch_ioctl_assemble arg;
64 u64 *user_devs = NULL;
69 if (copy_from_user(&arg, user_arg, sizeof(arg)))
72 if (arg.flags || arg.pad)
75 user_devs = kmalloc_array(arg.nr_devs, sizeof(u64), GFP_KERNEL);
79 devs = kcalloc(arg.nr_devs, sizeof(char *), GFP_KERNEL);
81 if (copy_from_user(user_devs, user_arg->devs,
82 sizeof(u64) * arg.nr_devs))
85 for (i = 0; i < arg.nr_devs; i++) {
86 devs[i] = strndup_user((const char __user *)(unsigned long)
89 ret= PTR_ERR_OR_ZERO(devs[i]);
94 c = bch2_fs_open(devs, arg.nr_devs, bch2_opts_empty());
95 ret = PTR_ERR_OR_ZERO(c);
100 for (i = 0; i < arg.nr_devs; i++)
106 static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg)
108 struct bch_ioctl_incremental arg;
112 if (copy_from_user(&arg, user_arg, sizeof(arg)))
115 if (arg.flags || arg.pad)
118 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
119 ret = PTR_ERR_OR_ZERO(path);
123 err = bch2_fs_open_incremental(path);
127 pr_err("Could not register bcachefs devices: %s", err);
135 static long bch2_global_ioctl(unsigned cmd, void __user *arg)
139 case BCH_IOCTL_ASSEMBLE:
140 return bch2_ioctl_assemble(arg);
141 case BCH_IOCTL_INCREMENTAL:
142 return bch2_ioctl_incremental(arg);
149 static long bch2_ioctl_query_uuid(struct bch_fs *c,
150 struct bch_ioctl_query_uuid __user *user_arg)
152 if (copy_to_user(&user_arg->uuid, &c->sb.user_uuid,
153 sizeof(c->sb.user_uuid)))
159 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
161 if (!capable(CAP_SYS_ADMIN))
164 if (arg.flags || arg.pad)
167 return bch2_fs_start(c);
170 static long bch2_ioctl_stop(struct bch_fs *c)
172 if (!capable(CAP_SYS_ADMIN))
180 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
185 if (!capable(CAP_SYS_ADMIN))
188 if (arg.flags || arg.pad)
191 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
192 ret = PTR_ERR_OR_ZERO(path);
196 ret = bch2_dev_add(c, path);
202 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
206 if (!capable(CAP_SYS_ADMIN))
209 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
210 BCH_FORCE_IF_METADATA_LOST|
211 BCH_FORCE_IF_DEGRADED|
216 ca = bch2_device_lookup(c, arg.dev, arg.flags);
220 return bch2_dev_remove(c, ca, arg.flags);
223 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
228 if (!capable(CAP_SYS_ADMIN))
231 if (arg.flags || arg.pad)
234 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
235 ret = PTR_ERR_OR_ZERO(path);
239 ret = bch2_dev_online(c, path);
244 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
249 if (!capable(CAP_SYS_ADMIN))
252 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
253 BCH_FORCE_IF_METADATA_LOST|
254 BCH_FORCE_IF_DEGRADED|
259 ca = bch2_device_lookup(c, arg.dev, arg.flags);
263 ret = bch2_dev_offline(c, ca, arg.flags);
264 percpu_ref_put(&ca->ref);
268 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
269 struct bch_ioctl_disk_set_state arg)
274 if (!capable(CAP_SYS_ADMIN))
277 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
278 BCH_FORCE_IF_METADATA_LOST|
279 BCH_FORCE_IF_DEGRADED|
281 arg.pad[0] || arg.pad[1] || arg.pad[2] ||
282 arg.new_state >= BCH_MEMBER_STATE_NR)
285 ca = bch2_device_lookup(c, arg.dev, arg.flags);
289 ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
291 bch_err(c, "Error setting device state: %s", bch2_err_str(ret));
293 percpu_ref_put(&ca->ref);
297 struct bch_data_ctx {
299 struct bch_ioctl_data arg;
300 struct bch_move_stats stats;
304 struct task_struct *thread;
307 static int bch2_data_thread(void *arg)
309 struct bch_data_ctx *ctx = arg;
311 ctx->ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
313 ctx->stats.data_type = U8_MAX;
317 static int bch2_data_job_release(struct inode *inode, struct file *file)
319 struct bch_data_ctx *ctx = file->private_data;
321 kthread_stop(ctx->thread);
322 put_task_struct(ctx->thread);
327 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
328 size_t len, loff_t *ppos)
330 struct bch_data_ctx *ctx = file->private_data;
331 struct bch_fs *c = ctx->c;
332 struct bch_ioctl_data_event e = {
333 .type = BCH_DATA_EVENT_PROGRESS,
334 .p.data_type = ctx->stats.data_type,
335 .p.btree_id = ctx->stats.pos.btree,
336 .p.pos = ctx->stats.pos.pos,
337 .p.sectors_done = atomic64_read(&ctx->stats.sectors_seen),
338 .p.sectors_total = bch2_fs_usage_read_short(c).used,
344 if (copy_to_user(buf, &e, sizeof(e)))
350 static const struct file_operations bcachefs_data_ops = {
351 .release = bch2_data_job_release,
352 .read = bch2_data_job_read,
356 static long bch2_ioctl_data(struct bch_fs *c,
357 struct bch_ioctl_data arg)
359 struct bch_data_ctx *ctx = NULL;
360 struct file *file = NULL;
361 unsigned flags = O_RDONLY|O_CLOEXEC|O_NONBLOCK;
364 if (!capable(CAP_SYS_ADMIN))
367 if (arg.op >= BCH_DATA_OP_NR || arg.flags)
370 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
377 ctx->thread = kthread_create(bch2_data_thread, ctx,
378 "bch-data/%s", c->name);
379 if (IS_ERR(ctx->thread)) {
380 ret = PTR_ERR(ctx->thread);
384 ret = get_unused_fd_flags(flags);
389 file = anon_inode_getfile("[bcachefs]", &bcachefs_data_ops, ctx, flags);
395 fd_install(fd, file);
397 get_task_struct(ctx->thread);
398 wake_up_process(ctx->thread);
404 if (!IS_ERR_OR_NULL(ctx->thread))
405 kthread_stop(ctx->thread);
410 static long bch2_ioctl_fs_usage(struct bch_fs *c,
411 struct bch_ioctl_fs_usage __user *user_arg)
413 struct bch_ioctl_fs_usage *arg = NULL;
414 struct bch_replicas_usage *dst_e, *dst_end;
415 struct bch_fs_usage_online *src;
416 u32 replica_entries_bytes;
420 if (!test_bit(BCH_FS_STARTED, &c->flags))
423 if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
426 arg = kzalloc(size_add(sizeof(*arg), replica_entries_bytes), GFP_KERNEL);
430 src = bch2_fs_usage_read(c);
436 arg->capacity = c->capacity;
437 arg->used = bch2_fs_sectors_used(c, src);
438 arg->online_reserved = src->online_reserved;
440 for (i = 0; i < BCH_REPLICAS_MAX; i++)
441 arg->persistent_reserved[i] = src->u.persistent_reserved[i];
443 dst_e = arg->replicas;
444 dst_end = (void *) arg->replicas + replica_entries_bytes;
446 for (i = 0; i < c->replicas.nr; i++) {
447 struct bch_replicas_entry *src_e =
448 cpu_replicas_entry(&c->replicas, i);
450 /* check that we have enough space for one replicas entry */
451 if (dst_e + 1 > dst_end) {
456 dst_e->sectors = src->u.replicas[i];
459 /* recheck after setting nr_devs: */
460 if (replicas_usage_next(dst_e) > dst_end) {
465 memcpy(dst_e->r.devs, src_e->devs, src_e->nr_devs);
467 dst_e = replicas_usage_next(dst_e);
470 arg->replica_entries_bytes = (void *) dst_e - (void *) arg->replicas;
472 percpu_up_read(&c->mark_lock);
477 if (copy_to_user(user_arg, arg,
478 sizeof(*arg) + arg->replica_entries_bytes))
485 static long bch2_ioctl_dev_usage(struct bch_fs *c,
486 struct bch_ioctl_dev_usage __user *user_arg)
488 struct bch_ioctl_dev_usage arg;
489 struct bch_dev_usage src;
493 if (!test_bit(BCH_FS_STARTED, &c->flags))
496 if (copy_from_user(&arg, user_arg, sizeof(arg)))
499 if ((arg.flags & ~BCH_BY_INDEX) ||
505 ca = bch2_device_lookup(c, arg.dev, arg.flags);
509 src = bch2_dev_usage_read(ca);
511 arg.state = ca->mi.state;
512 arg.bucket_size = ca->mi.bucket_size;
513 arg.nr_buckets = ca->mi.nbuckets - ca->mi.first_bucket;
514 arg.buckets_ec = src.buckets_ec;
516 for (i = 0; i < BCH_DATA_NR; i++) {
517 arg.d[i].buckets = src.d[i].buckets;
518 arg.d[i].sectors = src.d[i].sectors;
519 arg.d[i].fragmented = src.d[i].fragmented;
522 percpu_ref_put(&ca->ref);
524 if (copy_to_user(user_arg, &arg, sizeof(arg)))
530 static long bch2_ioctl_read_super(struct bch_fs *c,
531 struct bch_ioctl_read_super arg)
533 struct bch_dev *ca = NULL;
537 if (!capable(CAP_SYS_ADMIN))
540 if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
544 mutex_lock(&c->sb_lock);
546 if (arg.flags & BCH_READ_DEV) {
547 ca = bch2_device_lookup(c, arg.dev, arg.flags);
559 if (vstruct_bytes(sb) > arg.size) {
564 if (copy_to_user((void __user *)(unsigned long)arg.sb, sb,
568 if (!IS_ERR_OR_NULL(ca))
569 percpu_ref_put(&ca->ref);
570 mutex_unlock(&c->sb_lock);
574 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
575 struct bch_ioctl_disk_get_idx arg)
577 dev_t dev = huge_decode_dev(arg.dev);
581 if (!capable(CAP_SYS_ADMIN))
587 for_each_online_member(ca, c, i)
588 if (ca->dev == dev) {
589 percpu_ref_put(&ca->io_ref);
593 return -BCH_ERR_ENOENT_dev_idx_not_found;
596 static long bch2_ioctl_disk_resize(struct bch_fs *c,
597 struct bch_ioctl_disk_resize arg)
602 if (!capable(CAP_SYS_ADMIN))
605 if ((arg.flags & ~BCH_BY_INDEX) ||
609 ca = bch2_device_lookup(c, arg.dev, arg.flags);
613 ret = bch2_dev_resize(c, ca, arg.nbuckets);
615 percpu_ref_put(&ca->ref);
619 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
620 struct bch_ioctl_disk_resize_journal arg)
625 if (!capable(CAP_SYS_ADMIN))
628 if ((arg.flags & ~BCH_BY_INDEX) ||
632 if (arg.nbuckets > U32_MAX)
635 ca = bch2_device_lookup(c, arg.dev, arg.flags);
639 ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
641 percpu_ref_put(&ca->ref);
645 #define BCH_IOCTL(_name, _argtype) \
649 if (copy_from_user(&i, arg, sizeof(i))) \
651 ret = bch2_ioctl_##_name(c, i); \
655 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
660 case BCH_IOCTL_QUERY_UUID:
661 return bch2_ioctl_query_uuid(c, arg);
662 case BCH_IOCTL_FS_USAGE:
663 return bch2_ioctl_fs_usage(c, arg);
664 case BCH_IOCTL_DEV_USAGE:
665 return bch2_ioctl_dev_usage(c, arg);
667 case BCH_IOCTL_START:
668 BCH_IOCTL(start, struct bch_ioctl_start);
670 return bch2_ioctl_stop(c);
672 case BCH_IOCTL_READ_SUPER:
673 BCH_IOCTL(read_super, struct bch_ioctl_read_super);
674 case BCH_IOCTL_DISK_GET_IDX:
675 BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
678 if (!test_bit(BCH_FS_STARTED, &c->flags))
682 case BCH_IOCTL_DISK_ADD:
683 BCH_IOCTL(disk_add, struct bch_ioctl_disk);
684 case BCH_IOCTL_DISK_REMOVE:
685 BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
686 case BCH_IOCTL_DISK_ONLINE:
687 BCH_IOCTL(disk_online, struct bch_ioctl_disk);
688 case BCH_IOCTL_DISK_OFFLINE:
689 BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
690 case BCH_IOCTL_DISK_SET_STATE:
691 BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
693 BCH_IOCTL(data, struct bch_ioctl_data);
694 case BCH_IOCTL_DISK_RESIZE:
695 BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
696 case BCH_IOCTL_DISK_RESIZE_JOURNAL:
697 BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
704 ret = bch2_err_class(ret);
708 static DEFINE_IDR(bch_chardev_minor);
710 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
712 unsigned minor = iminor(file_inode(filp));
713 struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
714 void __user *arg = (void __user *) v;
717 ? bch2_fs_ioctl(c, cmd, arg)
718 : bch2_global_ioctl(cmd, arg);
721 static const struct file_operations bch_chardev_fops = {
722 .owner = THIS_MODULE,
723 .unlocked_ioctl = bch2_chardev_ioctl,
724 .open = nonseekable_open,
727 static int bch_chardev_major;
728 static struct class *bch_chardev_class;
729 static struct device *bch_chardev;
731 void bch2_fs_chardev_exit(struct bch_fs *c)
733 if (!IS_ERR_OR_NULL(c->chardev))
734 device_unregister(c->chardev);
736 idr_remove(&bch_chardev_minor, c->minor);
739 int bch2_fs_chardev_init(struct bch_fs *c)
741 c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
745 c->chardev = device_create(bch_chardev_class, NULL,
746 MKDEV(bch_chardev_major, c->minor), c,
747 "bcachefs%u-ctl", c->minor);
748 if (IS_ERR(c->chardev))
749 return PTR_ERR(c->chardev);
754 void bch2_chardev_exit(void)
756 if (!IS_ERR_OR_NULL(bch_chardev_class))
757 device_destroy(bch_chardev_class,
758 MKDEV(bch_chardev_major, U8_MAX));
759 if (!IS_ERR_OR_NULL(bch_chardev_class))
760 class_destroy(bch_chardev_class);
761 if (bch_chardev_major > 0)
762 unregister_chrdev(bch_chardev_major, "bcachefs");
765 int __init bch2_chardev_init(void)
767 bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
768 if (bch_chardev_major < 0)
769 return bch_chardev_major;
771 bch_chardev_class = class_create("bcachefs");
772 if (IS_ERR(bch_chardev_class))
773 return PTR_ERR(bch_chardev_class);
775 bch_chardev = device_create(bch_chardev_class, NULL,
776 MKDEV(bch_chardev_major, U8_MAX),
777 NULL, "bcachefs-ctl");
778 if (IS_ERR(bch_chardev))
779 return PTR_ERR(bch_chardev);
784 #endif /* NO_BCACHEFS_CHARDEV */