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)
95 c = bch2_fs_open(devs, arg.nr_devs, bch2_opts_empty());
96 ret = PTR_ERR_OR_ZERO(c);
101 for (i = 0; i < arg.nr_devs; i++)
107 static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg)
109 struct bch_ioctl_incremental arg;
113 if (copy_from_user(&arg, user_arg, sizeof(arg)))
116 if (arg.flags || arg.pad)
119 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
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 return copy_to_user(&user_arg->uuid,
154 sizeof(c->sb.user_uuid));
158 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
160 if (!capable(CAP_SYS_ADMIN))
163 if (arg.flags || arg.pad)
166 return bch2_fs_start(c);
169 static long bch2_ioctl_stop(struct bch_fs *c)
171 if (!capable(CAP_SYS_ADMIN))
179 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
184 if (!capable(CAP_SYS_ADMIN))
187 if (arg.flags || arg.pad)
190 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
194 ret = bch2_dev_add(c, path);
200 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
204 if (!capable(CAP_SYS_ADMIN))
207 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
208 BCH_FORCE_IF_METADATA_LOST|
209 BCH_FORCE_IF_DEGRADED|
214 ca = bch2_device_lookup(c, arg.dev, arg.flags);
218 return bch2_dev_remove(c, ca, arg.flags);
221 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
226 if (!capable(CAP_SYS_ADMIN))
229 if (arg.flags || arg.pad)
232 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
236 ret = bch2_dev_online(c, path);
241 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
246 if (!capable(CAP_SYS_ADMIN))
249 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
250 BCH_FORCE_IF_METADATA_LOST|
251 BCH_FORCE_IF_DEGRADED|
256 ca = bch2_device_lookup(c, arg.dev, arg.flags);
260 ret = bch2_dev_offline(c, ca, arg.flags);
261 percpu_ref_put(&ca->ref);
265 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
266 struct bch_ioctl_disk_set_state arg)
271 if (!capable(CAP_SYS_ADMIN))
274 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
275 BCH_FORCE_IF_METADATA_LOST|
276 BCH_FORCE_IF_DEGRADED|
278 arg.pad[0] || arg.pad[1] || arg.pad[2] ||
279 arg.new_state >= BCH_MEMBER_STATE_NR)
282 ca = bch2_device_lookup(c, arg.dev, arg.flags);
286 ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
288 bch_err(c, "Error setting device state: %s", bch2_err_str(ret));
290 percpu_ref_put(&ca->ref);
294 struct bch_data_ctx {
296 struct bch_ioctl_data arg;
297 struct bch_move_stats stats;
301 struct task_struct *thread;
304 static int bch2_data_thread(void *arg)
306 struct bch_data_ctx *ctx = arg;
308 ctx->ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
310 ctx->stats.data_type = U8_MAX;
314 static int bch2_data_job_release(struct inode *inode, struct file *file)
316 struct bch_data_ctx *ctx = file->private_data;
318 kthread_stop(ctx->thread);
319 put_task_struct(ctx->thread);
324 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
325 size_t len, loff_t *ppos)
327 struct bch_data_ctx *ctx = file->private_data;
328 struct bch_fs *c = ctx->c;
329 struct bch_ioctl_data_event e = {
330 .type = BCH_DATA_EVENT_PROGRESS,
331 .p.data_type = ctx->stats.data_type,
332 .p.btree_id = ctx->stats.btree_id,
333 .p.pos = ctx->stats.pos,
334 .p.sectors_done = atomic64_read(&ctx->stats.sectors_seen),
335 .p.sectors_total = bch2_fs_usage_read_short(c).used,
341 return copy_to_user(buf, &e, sizeof(e)) ?: sizeof(e);
344 static const struct file_operations bcachefs_data_ops = {
345 .release = bch2_data_job_release,
346 .read = bch2_data_job_read,
350 static long bch2_ioctl_data(struct bch_fs *c,
351 struct bch_ioctl_data arg)
353 struct bch_data_ctx *ctx = NULL;
354 struct file *file = NULL;
355 unsigned flags = O_RDONLY|O_CLOEXEC|O_NONBLOCK;
358 if (!capable(CAP_SYS_ADMIN))
361 if (arg.op >= BCH_DATA_OP_NR || arg.flags)
364 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
371 ctx->thread = kthread_create(bch2_data_thread, ctx,
372 "bch-data/%s", c->name);
373 if (IS_ERR(ctx->thread)) {
374 ret = PTR_ERR(ctx->thread);
378 ret = get_unused_fd_flags(flags);
383 file = anon_inode_getfile("[bcachefs]", &bcachefs_data_ops, ctx, flags);
389 fd_install(fd, file);
391 get_task_struct(ctx->thread);
392 wake_up_process(ctx->thread);
398 if (!IS_ERR_OR_NULL(ctx->thread))
399 kthread_stop(ctx->thread);
404 static long bch2_ioctl_fs_usage(struct bch_fs *c,
405 struct bch_ioctl_fs_usage __user *user_arg)
407 struct bch_ioctl_fs_usage *arg = NULL;
408 struct bch_replicas_usage *dst_e, *dst_end;
409 struct bch_fs_usage_online *src;
410 u32 replica_entries_bytes;
414 if (!test_bit(BCH_FS_STARTED, &c->flags))
417 if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
420 arg = kzalloc(sizeof(*arg) + replica_entries_bytes, GFP_KERNEL);
424 src = bch2_fs_usage_read(c);
430 arg->capacity = c->capacity;
431 arg->used = bch2_fs_sectors_used(c, src);
432 arg->online_reserved = src->online_reserved;
434 for (i = 0; i < BCH_REPLICAS_MAX; i++)
435 arg->persistent_reserved[i] = src->u.persistent_reserved[i];
437 dst_e = arg->replicas;
438 dst_end = (void *) arg->replicas + replica_entries_bytes;
440 for (i = 0; i < c->replicas.nr; i++) {
441 struct bch_replicas_entry *src_e =
442 cpu_replicas_entry(&c->replicas, i);
444 /* check that we have enough space for one replicas entry */
445 if (dst_e + 1 > dst_end) {
450 dst_e->sectors = src->u.replicas[i];
453 /* recheck after setting nr_devs: */
454 if (replicas_usage_next(dst_e) > dst_end) {
459 memcpy(dst_e->r.devs, src_e->devs, src_e->nr_devs);
461 dst_e = replicas_usage_next(dst_e);
464 arg->replica_entries_bytes = (void *) dst_e - (void *) arg->replicas;
466 percpu_up_read(&c->mark_lock);
470 ret = copy_to_user(user_arg, arg,
471 sizeof(*arg) + arg->replica_entries_bytes);
477 static long bch2_ioctl_dev_usage(struct bch_fs *c,
478 struct bch_ioctl_dev_usage __user *user_arg)
480 struct bch_ioctl_dev_usage arg;
481 struct bch_dev_usage src;
485 if (!test_bit(BCH_FS_STARTED, &c->flags))
488 if (copy_from_user(&arg, user_arg, sizeof(arg)))
491 if ((arg.flags & ~BCH_BY_INDEX) ||
497 ca = bch2_device_lookup(c, arg.dev, arg.flags);
501 src = bch2_dev_usage_read(ca);
503 arg.state = ca->mi.state;
504 arg.bucket_size = ca->mi.bucket_size;
505 arg.nr_buckets = ca->mi.nbuckets - ca->mi.first_bucket;
506 arg.buckets_ec = src.buckets_ec;
508 for (i = 0; i < BCH_DATA_NR; i++) {
509 arg.d[i].buckets = src.d[i].buckets;
510 arg.d[i].sectors = src.d[i].sectors;
511 arg.d[i].fragmented = src.d[i].fragmented;
514 percpu_ref_put(&ca->ref);
516 return copy_to_user(user_arg, &arg, sizeof(arg));
519 static long bch2_ioctl_read_super(struct bch_fs *c,
520 struct bch_ioctl_read_super arg)
522 struct bch_dev *ca = NULL;
526 if (!capable(CAP_SYS_ADMIN))
529 if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
533 mutex_lock(&c->sb_lock);
535 if (arg.flags & BCH_READ_DEV) {
536 ca = bch2_device_lookup(c, arg.dev, arg.flags);
548 if (vstruct_bytes(sb) > arg.size) {
553 ret = copy_to_user((void __user *)(unsigned long)arg.sb,
554 sb, vstruct_bytes(sb));
556 if (!IS_ERR_OR_NULL(ca))
557 percpu_ref_put(&ca->ref);
558 mutex_unlock(&c->sb_lock);
562 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
563 struct bch_ioctl_disk_get_idx arg)
565 dev_t dev = huge_decode_dev(arg.dev);
569 if (!capable(CAP_SYS_ADMIN))
575 for_each_online_member(ca, c, i)
576 if (ca->dev == dev) {
577 percpu_ref_put(&ca->io_ref);
581 return -BCH_ERR_ENOENT_dev_idx_not_found;
584 static long bch2_ioctl_disk_resize(struct bch_fs *c,
585 struct bch_ioctl_disk_resize arg)
590 if (!capable(CAP_SYS_ADMIN))
593 if ((arg.flags & ~BCH_BY_INDEX) ||
597 ca = bch2_device_lookup(c, arg.dev, arg.flags);
601 ret = bch2_dev_resize(c, ca, arg.nbuckets);
603 percpu_ref_put(&ca->ref);
607 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
608 struct bch_ioctl_disk_resize_journal arg)
613 if (!capable(CAP_SYS_ADMIN))
616 if ((arg.flags & ~BCH_BY_INDEX) ||
620 ca = bch2_device_lookup(c, arg.dev, arg.flags);
624 ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
626 percpu_ref_put(&ca->ref);
630 #define BCH_IOCTL(_name, _argtype) \
634 if (copy_from_user(&i, arg, sizeof(i))) \
636 ret = bch2_ioctl_##_name(c, i); \
640 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
645 case BCH_IOCTL_QUERY_UUID:
646 return bch2_ioctl_query_uuid(c, arg);
647 case BCH_IOCTL_FS_USAGE:
648 return bch2_ioctl_fs_usage(c, arg);
649 case BCH_IOCTL_DEV_USAGE:
650 return bch2_ioctl_dev_usage(c, arg);
652 case BCH_IOCTL_START:
653 BCH_IOCTL(start, struct bch_ioctl_start);
655 return bch2_ioctl_stop(c);
657 case BCH_IOCTL_READ_SUPER:
658 BCH_IOCTL(read_super, struct bch_ioctl_read_super);
659 case BCH_IOCTL_DISK_GET_IDX:
660 BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
663 if (!test_bit(BCH_FS_STARTED, &c->flags))
667 case BCH_IOCTL_DISK_ADD:
668 BCH_IOCTL(disk_add, struct bch_ioctl_disk);
669 case BCH_IOCTL_DISK_REMOVE:
670 BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
671 case BCH_IOCTL_DISK_ONLINE:
672 BCH_IOCTL(disk_online, struct bch_ioctl_disk);
673 case BCH_IOCTL_DISK_OFFLINE:
674 BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
675 case BCH_IOCTL_DISK_SET_STATE:
676 BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
678 BCH_IOCTL(data, struct bch_ioctl_data);
679 case BCH_IOCTL_DISK_RESIZE:
680 BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
681 case BCH_IOCTL_DISK_RESIZE_JOURNAL:
682 BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
689 ret = bch2_err_class(ret);
693 static DEFINE_IDR(bch_chardev_minor);
695 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
697 unsigned minor = iminor(file_inode(filp));
698 struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
699 void __user *arg = (void __user *) v;
702 ? bch2_fs_ioctl(c, cmd, arg)
703 : bch2_global_ioctl(cmd, arg);
706 static const struct file_operations bch_chardev_fops = {
707 .owner = THIS_MODULE,
708 .unlocked_ioctl = bch2_chardev_ioctl,
709 .open = nonseekable_open,
712 static int bch_chardev_major;
713 static struct class *bch_chardev_class;
714 static struct device *bch_chardev;
716 void bch2_fs_chardev_exit(struct bch_fs *c)
718 if (!IS_ERR_OR_NULL(c->chardev))
719 device_unregister(c->chardev);
721 idr_remove(&bch_chardev_minor, c->minor);
724 int bch2_fs_chardev_init(struct bch_fs *c)
726 c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
730 c->chardev = device_create(bch_chardev_class, NULL,
731 MKDEV(bch_chardev_major, c->minor), c,
732 "bcachefs%u-ctl", c->minor);
733 if (IS_ERR(c->chardev))
734 return PTR_ERR(c->chardev);
739 void bch2_chardev_exit(void)
741 if (!IS_ERR_OR_NULL(bch_chardev_class))
742 device_destroy(bch_chardev_class,
743 MKDEV(bch_chardev_major, U8_MAX));
744 if (!IS_ERR_OR_NULL(bch_chardev_class))
745 class_destroy(bch_chardev_class);
746 if (bch_chardev_major > 0)
747 unregister_chrdev(bch_chardev_major, "bcachefs");
750 int __init bch2_chardev_init(void)
752 bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
753 if (bch_chardev_major < 0)
754 return bch_chardev_major;
756 bch_chardev_class = class_create(THIS_MODULE, "bcachefs");
757 if (IS_ERR(bch_chardev_class))
758 return PTR_ERR(bch_chardev_class);
760 bch_chardev = device_create(bch_chardev_class, NULL,
761 MKDEV(bch_chardev_major, U8_MAX),
762 NULL, "bcachefs-ctl");
763 if (IS_ERR(bch_chardev))
764 return PTR_ERR(bch_chardev);
769 #endif /* NO_BCACHEFS_CHARDEV */