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 (arg.flags || arg.pad)
163 return bch2_fs_start(c);
166 static long bch2_ioctl_stop(struct bch_fs *c)
173 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
178 if (arg.flags || arg.pad)
181 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
185 ret = bch2_dev_add(c, path);
191 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
195 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
196 BCH_FORCE_IF_METADATA_LOST|
197 BCH_FORCE_IF_DEGRADED|
202 ca = bch2_device_lookup(c, arg.dev, arg.flags);
206 return bch2_dev_remove(c, ca, arg.flags);
209 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
214 if (arg.flags || arg.pad)
217 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
221 ret = bch2_dev_online(c, path);
226 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
231 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
232 BCH_FORCE_IF_METADATA_LOST|
233 BCH_FORCE_IF_DEGRADED|
238 ca = bch2_device_lookup(c, arg.dev, arg.flags);
242 ret = bch2_dev_offline(c, ca, arg.flags);
243 percpu_ref_put(&ca->ref);
247 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
248 struct bch_ioctl_disk_set_state arg)
253 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
254 BCH_FORCE_IF_METADATA_LOST|
255 BCH_FORCE_IF_DEGRADED|
257 arg.pad[0] || arg.pad[1] || arg.pad[2])
260 ca = bch2_device_lookup(c, arg.dev, arg.flags);
264 ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
266 percpu_ref_put(&ca->ref);
270 struct bch_data_ctx {
272 struct bch_ioctl_data arg;
273 struct bch_move_stats stats;
277 struct task_struct *thread;
280 static int bch2_data_thread(void *arg)
282 struct bch_data_ctx *ctx = arg;
284 ctx->ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
286 ctx->stats.data_type = U8_MAX;
290 static int bch2_data_job_release(struct inode *inode, struct file *file)
292 struct bch_data_ctx *ctx = file->private_data;
294 kthread_stop(ctx->thread);
295 put_task_struct(ctx->thread);
300 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
301 size_t len, loff_t *ppos)
303 struct bch_data_ctx *ctx = file->private_data;
304 struct bch_fs *c = ctx->c;
305 struct bch_ioctl_data_event e = {
306 .type = BCH_DATA_EVENT_PROGRESS,
307 .p.data_type = ctx->stats.data_type,
308 .p.btree_id = ctx->stats.btree_id,
309 .p.pos = ctx->stats.pos,
310 .p.sectors_done = atomic64_read(&ctx->stats.sectors_seen),
311 .p.sectors_total = bch2_fs_usage_read_short(c).used,
317 return copy_to_user(buf, &e, sizeof(e)) ?: sizeof(e);
320 static const struct file_operations bcachefs_data_ops = {
321 .release = bch2_data_job_release,
322 .read = bch2_data_job_read,
326 static long bch2_ioctl_data(struct bch_fs *c,
327 struct bch_ioctl_data arg)
329 struct bch_data_ctx *ctx = NULL;
330 struct file *file = NULL;
331 unsigned flags = O_RDONLY|O_CLOEXEC|O_NONBLOCK;
334 if (arg.op >= BCH_DATA_OP_NR || arg.flags)
337 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
344 ctx->thread = kthread_create(bch2_data_thread, ctx,
345 "bch-data/%s", c->name);
346 if (IS_ERR(ctx->thread)) {
347 ret = PTR_ERR(ctx->thread);
351 ret = get_unused_fd_flags(flags);
356 file = anon_inode_getfile("[bcachefs]", &bcachefs_data_ops, ctx, flags);
362 fd_install(fd, file);
364 get_task_struct(ctx->thread);
365 wake_up_process(ctx->thread);
371 if (!IS_ERR_OR_NULL(ctx->thread))
372 kthread_stop(ctx->thread);
377 static long bch2_ioctl_fs_usage(struct bch_fs *c,
378 struct bch_ioctl_fs_usage __user *user_arg)
380 struct bch_ioctl_fs_usage *arg = NULL;
381 struct bch_replicas_usage *dst_e, *dst_end;
382 struct bch_fs_usage_online *src;
383 u32 replica_entries_bytes;
387 if (!test_bit(BCH_FS_STARTED, &c->flags))
390 if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
393 arg = kzalloc(sizeof(*arg) + replica_entries_bytes, GFP_KERNEL);
397 src = bch2_fs_usage_read(c);
403 arg->capacity = c->capacity;
404 arg->used = bch2_fs_sectors_used(c, src);
405 arg->online_reserved = src->online_reserved;
407 for (i = 0; i < BCH_REPLICAS_MAX; i++)
408 arg->persistent_reserved[i] = src->u.persistent_reserved[i];
410 dst_e = arg->replicas;
411 dst_end = (void *) arg->replicas + replica_entries_bytes;
413 for (i = 0; i < c->replicas.nr; i++) {
414 struct bch_replicas_entry *src_e =
415 cpu_replicas_entry(&c->replicas, i);
417 if (replicas_usage_next(dst_e) > dst_end) {
422 dst_e->sectors = src->u.replicas[i];
425 /* recheck after setting nr_devs: */
426 if (replicas_usage_next(dst_e) > dst_end) {
431 memcpy(dst_e->r.devs, src_e->devs, src_e->nr_devs);
433 dst_e = replicas_usage_next(dst_e);
436 arg->replica_entries_bytes = (void *) dst_e - (void *) arg->replicas;
438 percpu_up_read(&c->mark_lock);
442 ret = copy_to_user(user_arg, arg,
443 sizeof(*arg) + arg->replica_entries_bytes);
449 static long bch2_ioctl_dev_usage(struct bch_fs *c,
450 struct bch_ioctl_dev_usage __user *user_arg)
452 struct bch_ioctl_dev_usage arg;
453 struct bch_dev_usage src;
457 if (!test_bit(BCH_FS_STARTED, &c->flags))
460 if (copy_from_user(&arg, user_arg, sizeof(arg)))
463 if ((arg.flags & ~BCH_BY_INDEX) ||
469 ca = bch2_device_lookup(c, arg.dev, arg.flags);
473 src = bch2_dev_usage_read(ca);
475 arg.state = ca->mi.state;
476 arg.bucket_size = ca->mi.bucket_size;
477 arg.nr_buckets = ca->mi.nbuckets - ca->mi.first_bucket;
478 arg.available_buckets = arg.nr_buckets - src.buckets_unavailable;
479 arg.ec_buckets = src.buckets_ec;
482 for (i = 0; i < BCH_DATA_NR; i++) {
483 arg.buckets[i] = src.d[i].buckets;
484 arg.sectors[i] = src.d[i].sectors;
487 percpu_ref_put(&ca->ref);
489 return copy_to_user(user_arg, &arg, sizeof(arg));
492 static long bch2_ioctl_read_super(struct bch_fs *c,
493 struct bch_ioctl_read_super arg)
495 struct bch_dev *ca = NULL;
499 if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
503 mutex_lock(&c->sb_lock);
505 if (arg.flags & BCH_READ_DEV) {
506 ca = bch2_device_lookup(c, arg.dev, arg.flags);
518 if (vstruct_bytes(sb) > arg.size) {
523 ret = copy_to_user((void __user *)(unsigned long)arg.sb,
524 sb, vstruct_bytes(sb));
527 percpu_ref_put(&ca->ref);
528 mutex_unlock(&c->sb_lock);
532 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
533 struct bch_ioctl_disk_get_idx arg)
535 dev_t dev = huge_decode_dev(arg.dev);
539 for_each_online_member(ca, c, i)
540 if (ca->disk_sb.bdev->bd_dev == dev) {
541 percpu_ref_put(&ca->io_ref);
548 static long bch2_ioctl_disk_resize(struct bch_fs *c,
549 struct bch_ioctl_disk_resize arg)
554 if ((arg.flags & ~BCH_BY_INDEX) ||
558 ca = bch2_device_lookup(c, arg.dev, arg.flags);
562 ret = bch2_dev_resize(c, ca, arg.nbuckets);
564 percpu_ref_put(&ca->ref);
568 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
569 struct bch_ioctl_disk_resize_journal arg)
574 if ((arg.flags & ~BCH_BY_INDEX) ||
578 ca = bch2_device_lookup(c, arg.dev, arg.flags);
582 ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
584 percpu_ref_put(&ca->ref);
588 #define BCH_IOCTL(_name, _argtype) \
592 if (copy_from_user(&i, arg, sizeof(i))) \
594 return bch2_ioctl_##_name(c, i); \
597 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
599 /* ioctls that don't require admin cap: */
601 case BCH_IOCTL_QUERY_UUID:
602 return bch2_ioctl_query_uuid(c, arg);
603 case BCH_IOCTL_FS_USAGE:
604 return bch2_ioctl_fs_usage(c, arg);
605 case BCH_IOCTL_DEV_USAGE:
606 return bch2_ioctl_dev_usage(c, arg);
609 if (!capable(CAP_SYS_ADMIN))
614 case BCH_IOCTL_START:
615 BCH_IOCTL(start, struct bch_ioctl_start);
617 return bch2_ioctl_stop(c);
619 case BCH_IOCTL_READ_SUPER:
620 BCH_IOCTL(read_super, struct bch_ioctl_read_super);
621 case BCH_IOCTL_DISK_GET_IDX:
622 BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
625 if (!test_bit(BCH_FS_STARTED, &c->flags))
628 /* ioctls that do require admin cap: */
630 case BCH_IOCTL_DISK_ADD:
631 BCH_IOCTL(disk_add, struct bch_ioctl_disk);
632 case BCH_IOCTL_DISK_REMOVE:
633 BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
634 case BCH_IOCTL_DISK_ONLINE:
635 BCH_IOCTL(disk_online, struct bch_ioctl_disk);
636 case BCH_IOCTL_DISK_OFFLINE:
637 BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
638 case BCH_IOCTL_DISK_SET_STATE:
639 BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
641 BCH_IOCTL(data, struct bch_ioctl_data);
642 case BCH_IOCTL_DISK_RESIZE:
643 BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
644 case BCH_IOCTL_DISK_RESIZE_JOURNAL:
645 BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
652 static DEFINE_IDR(bch_chardev_minor);
654 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
656 unsigned minor = iminor(file_inode(filp));
657 struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
658 void __user *arg = (void __user *) v;
661 ? bch2_fs_ioctl(c, cmd, arg)
662 : bch2_global_ioctl(cmd, arg);
665 static const struct file_operations bch_chardev_fops = {
666 .owner = THIS_MODULE,
667 .unlocked_ioctl = bch2_chardev_ioctl,
668 .open = nonseekable_open,
671 static int bch_chardev_major;
672 static struct class *bch_chardev_class;
673 static struct device *bch_chardev;
675 void bch2_fs_chardev_exit(struct bch_fs *c)
677 if (!IS_ERR_OR_NULL(c->chardev))
678 device_unregister(c->chardev);
680 idr_remove(&bch_chardev_minor, c->minor);
683 int bch2_fs_chardev_init(struct bch_fs *c)
685 c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
689 c->chardev = device_create(bch_chardev_class, NULL,
690 MKDEV(bch_chardev_major, c->minor), c,
691 "bcachefs%u-ctl", c->minor);
692 if (IS_ERR(c->chardev))
693 return PTR_ERR(c->chardev);
698 void bch2_chardev_exit(void)
700 if (!IS_ERR_OR_NULL(bch_chardev_class))
701 device_destroy(bch_chardev_class,
702 MKDEV(bch_chardev_major, U8_MAX));
703 if (!IS_ERR_OR_NULL(bch_chardev_class))
704 class_destroy(bch_chardev_class);
705 if (bch_chardev_major > 0)
706 unregister_chrdev(bch_chardev_major, "bcachefs");
709 int __init bch2_chardev_init(void)
711 bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
712 if (bch_chardev_major < 0)
713 return bch_chardev_major;
715 bch_chardev_class = class_create(THIS_MODULE, "bcachefs");
716 if (IS_ERR(bch_chardev_class))
717 return PTR_ERR(bch_chardev_class);
719 bch_chardev = device_create(bch_chardev_class, NULL,
720 MKDEV(bch_chardev_major, U8_MAX),
721 NULL, "bcachefs-ctl");
722 if (IS_ERR(bch_chardev))
723 return PTR_ERR(bch_chardev);
728 #endif /* NO_BCACHEFS_CHARDEV */