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])
281 ca = bch2_device_lookup(c, arg.dev, arg.flags);
285 ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
287 percpu_ref_put(&ca->ref);
291 struct bch_data_ctx {
293 struct bch_ioctl_data arg;
294 struct bch_move_stats stats;
298 struct task_struct *thread;
301 static int bch2_data_thread(void *arg)
303 struct bch_data_ctx *ctx = arg;
305 ctx->ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
307 ctx->stats.data_type = U8_MAX;
311 static int bch2_data_job_release(struct inode *inode, struct file *file)
313 struct bch_data_ctx *ctx = file->private_data;
315 kthread_stop(ctx->thread);
316 put_task_struct(ctx->thread);
321 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
322 size_t len, loff_t *ppos)
324 struct bch_data_ctx *ctx = file->private_data;
325 struct bch_fs *c = ctx->c;
326 struct bch_ioctl_data_event e = {
327 .type = BCH_DATA_EVENT_PROGRESS,
328 .p.data_type = ctx->stats.data_type,
329 .p.btree_id = ctx->stats.btree_id,
330 .p.pos = ctx->stats.pos,
331 .p.sectors_done = atomic64_read(&ctx->stats.sectors_seen),
332 .p.sectors_total = bch2_fs_usage_read_short(c).used,
338 return copy_to_user(buf, &e, sizeof(e)) ?: sizeof(e);
341 static const struct file_operations bcachefs_data_ops = {
342 .release = bch2_data_job_release,
343 .read = bch2_data_job_read,
347 static long bch2_ioctl_data(struct bch_fs *c,
348 struct bch_ioctl_data arg)
350 struct bch_data_ctx *ctx = NULL;
351 struct file *file = NULL;
352 unsigned flags = O_RDONLY|O_CLOEXEC|O_NONBLOCK;
355 if (!capable(CAP_SYS_ADMIN))
358 if (arg.op >= BCH_DATA_OP_NR || arg.flags)
361 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
368 ctx->thread = kthread_create(bch2_data_thread, ctx,
369 "bch-data/%s", c->name);
370 if (IS_ERR(ctx->thread)) {
371 ret = PTR_ERR(ctx->thread);
375 ret = get_unused_fd_flags(flags);
380 file = anon_inode_getfile("[bcachefs]", &bcachefs_data_ops, ctx, flags);
386 fd_install(fd, file);
388 get_task_struct(ctx->thread);
389 wake_up_process(ctx->thread);
395 if (!IS_ERR_OR_NULL(ctx->thread))
396 kthread_stop(ctx->thread);
401 static long bch2_ioctl_fs_usage(struct bch_fs *c,
402 struct bch_ioctl_fs_usage __user *user_arg)
404 struct bch_ioctl_fs_usage *arg = NULL;
405 struct bch_replicas_usage *dst_e, *dst_end;
406 struct bch_fs_usage_online *src;
407 u32 replica_entries_bytes;
411 if (!test_bit(BCH_FS_STARTED, &c->flags))
414 if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
417 arg = kzalloc(sizeof(*arg) + replica_entries_bytes, GFP_KERNEL);
421 src = bch2_fs_usage_read(c);
427 arg->capacity = c->capacity;
428 arg->used = bch2_fs_sectors_used(c, src);
429 arg->online_reserved = src->online_reserved;
431 for (i = 0; i < BCH_REPLICAS_MAX; i++)
432 arg->persistent_reserved[i] = src->u.persistent_reserved[i];
434 dst_e = arg->replicas;
435 dst_end = (void *) arg->replicas + replica_entries_bytes;
437 for (i = 0; i < c->replicas.nr; i++) {
438 struct bch_replicas_entry *src_e =
439 cpu_replicas_entry(&c->replicas, i);
441 /* check that we have enough space for one replicas entry */
442 if (dst_e + 1 > dst_end) {
447 dst_e->sectors = src->u.replicas[i];
450 /* recheck after setting nr_devs: */
451 if (replicas_usage_next(dst_e) > dst_end) {
456 memcpy(dst_e->r.devs, src_e->devs, src_e->nr_devs);
458 dst_e = replicas_usage_next(dst_e);
461 arg->replica_entries_bytes = (void *) dst_e - (void *) arg->replicas;
463 percpu_up_read(&c->mark_lock);
467 ret = copy_to_user(user_arg, arg,
468 sizeof(*arg) + arg->replica_entries_bytes);
474 static long bch2_ioctl_dev_usage(struct bch_fs *c,
475 struct bch_ioctl_dev_usage __user *user_arg)
477 struct bch_ioctl_dev_usage arg;
478 struct bch_dev_usage src;
482 if (!test_bit(BCH_FS_STARTED, &c->flags))
485 if (copy_from_user(&arg, user_arg, sizeof(arg)))
488 if ((arg.flags & ~BCH_BY_INDEX) ||
494 ca = bch2_device_lookup(c, arg.dev, arg.flags);
498 src = bch2_dev_usage_read(ca);
500 arg.state = ca->mi.state;
501 arg.bucket_size = ca->mi.bucket_size;
502 arg.nr_buckets = ca->mi.nbuckets - ca->mi.first_bucket;
503 arg.available_buckets = arg.nr_buckets - src.buckets_unavailable;
504 arg.ec_buckets = src.buckets_ec;
507 for (i = 0; i < BCH_DATA_NR; i++) {
508 arg.buckets[i] = src.d[i].buckets;
509 arg.sectors[i] = src.d[i].sectors;
512 percpu_ref_put(&ca->ref);
514 return copy_to_user(user_arg, &arg, sizeof(arg));
517 static long bch2_ioctl_read_super(struct bch_fs *c,
518 struct bch_ioctl_read_super arg)
520 struct bch_dev *ca = NULL;
524 if (!capable(CAP_SYS_ADMIN))
527 if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
531 mutex_lock(&c->sb_lock);
533 if (arg.flags & BCH_READ_DEV) {
534 ca = bch2_device_lookup(c, arg.dev, arg.flags);
546 if (vstruct_bytes(sb) > arg.size) {
551 ret = copy_to_user((void __user *)(unsigned long)arg.sb,
552 sb, vstruct_bytes(sb));
554 if (!IS_ERR_OR_NULL(ca))
555 percpu_ref_put(&ca->ref);
556 mutex_unlock(&c->sb_lock);
560 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
561 struct bch_ioctl_disk_get_idx arg)
563 dev_t dev = huge_decode_dev(arg.dev);
567 if (!capable(CAP_SYS_ADMIN))
570 for_each_online_member(ca, c, i)
571 if (ca->disk_sb.bdev->bd_dev == dev) {
572 percpu_ref_put(&ca->io_ref);
579 static long bch2_ioctl_disk_resize(struct bch_fs *c,
580 struct bch_ioctl_disk_resize arg)
585 if (!capable(CAP_SYS_ADMIN))
588 if ((arg.flags & ~BCH_BY_INDEX) ||
592 ca = bch2_device_lookup(c, arg.dev, arg.flags);
596 ret = bch2_dev_resize(c, ca, arg.nbuckets);
598 percpu_ref_put(&ca->ref);
602 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
603 struct bch_ioctl_disk_resize_journal arg)
608 if (!capable(CAP_SYS_ADMIN))
611 if ((arg.flags & ~BCH_BY_INDEX) ||
615 ca = bch2_device_lookup(c, arg.dev, arg.flags);
619 ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
621 percpu_ref_put(&ca->ref);
625 #define BCH_IOCTL(_name, _argtype) \
629 if (copy_from_user(&i, arg, sizeof(i))) \
631 return bch2_ioctl_##_name(c, i); \
634 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
637 case BCH_IOCTL_QUERY_UUID:
638 return bch2_ioctl_query_uuid(c, arg);
639 case BCH_IOCTL_FS_USAGE:
640 return bch2_ioctl_fs_usage(c, arg);
641 case BCH_IOCTL_DEV_USAGE:
642 return bch2_ioctl_dev_usage(c, arg);
644 case BCH_IOCTL_START:
645 BCH_IOCTL(start, struct bch_ioctl_start);
647 return bch2_ioctl_stop(c);
649 case BCH_IOCTL_READ_SUPER:
650 BCH_IOCTL(read_super, struct bch_ioctl_read_super);
651 case BCH_IOCTL_DISK_GET_IDX:
652 BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
655 if (!test_bit(BCH_FS_STARTED, &c->flags))
659 case BCH_IOCTL_DISK_ADD:
660 BCH_IOCTL(disk_add, struct bch_ioctl_disk);
661 case BCH_IOCTL_DISK_REMOVE:
662 BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
663 case BCH_IOCTL_DISK_ONLINE:
664 BCH_IOCTL(disk_online, struct bch_ioctl_disk);
665 case BCH_IOCTL_DISK_OFFLINE:
666 BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
667 case BCH_IOCTL_DISK_SET_STATE:
668 BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
670 BCH_IOCTL(data, struct bch_ioctl_data);
671 case BCH_IOCTL_DISK_RESIZE:
672 BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
673 case BCH_IOCTL_DISK_RESIZE_JOURNAL:
674 BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
681 static DEFINE_IDR(bch_chardev_minor);
683 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
685 unsigned minor = iminor(file_inode(filp));
686 struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
687 void __user *arg = (void __user *) v;
690 ? bch2_fs_ioctl(c, cmd, arg)
691 : bch2_global_ioctl(cmd, arg);
694 static const struct file_operations bch_chardev_fops = {
695 .owner = THIS_MODULE,
696 .unlocked_ioctl = bch2_chardev_ioctl,
697 .open = nonseekable_open,
700 static int bch_chardev_major;
701 static struct class *bch_chardev_class;
702 static struct device *bch_chardev;
704 void bch2_fs_chardev_exit(struct bch_fs *c)
706 if (!IS_ERR_OR_NULL(c->chardev))
707 device_unregister(c->chardev);
709 idr_remove(&bch_chardev_minor, c->minor);
712 int bch2_fs_chardev_init(struct bch_fs *c)
714 c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
718 c->chardev = device_create(bch_chardev_class, NULL,
719 MKDEV(bch_chardev_major, c->minor), c,
720 "bcachefs%u-ctl", c->minor);
721 if (IS_ERR(c->chardev))
722 return PTR_ERR(c->chardev);
727 void bch2_chardev_exit(void)
729 if (!IS_ERR_OR_NULL(bch_chardev_class))
730 device_destroy(bch_chardev_class,
731 MKDEV(bch_chardev_major, U8_MAX));
732 if (!IS_ERR_OR_NULL(bch_chardev_class))
733 class_destroy(bch_chardev_class);
734 if (bch_chardev_major > 0)
735 unregister_chrdev(bch_chardev_major, "bcachefs");
738 int __init bch2_chardev_init(void)
740 bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
741 if (bch_chardev_major < 0)
742 return bch_chardev_major;
744 bch_chardev_class = class_create(THIS_MODULE, "bcachefs");
745 if (IS_ERR(bch_chardev_class))
746 return PTR_ERR(bch_chardev_class);
748 bch_chardev = device_create(bch_chardev_class, NULL,
749 MKDEV(bch_chardev_major, U8_MAX),
750 NULL, "bcachefs-ctl");
751 if (IS_ERR(bch_chardev))
752 return PTR_ERR(bch_chardev);
757 #endif /* NO_BCACHEFS_CHARDEV */