1 #ifndef NO_BCACHEFS_CHARDEV
5 #include "bcachefs_ioctl.h"
12 #include <linux/anon_inodes.h>
13 #include <linux/cdev.h>
14 #include <linux/device.h>
15 #include <linux/file.h>
17 #include <linux/ioctl.h>
18 #include <linux/kthread.h>
19 #include <linux/major.h>
20 #include <linux/sched/task.h>
21 #include <linux/slab.h>
22 #include <linux/uaccess.h>
24 /* returns with ref on ca->ref */
25 static struct bch_dev *bch2_device_lookup(struct bch_fs *c, u64 dev,
30 if (flags & BCH_BY_INDEX) {
31 if (dev >= c->sb.nr_devices)
32 return ERR_PTR(-EINVAL);
35 ca = rcu_dereference(c->devs[dev]);
37 percpu_ref_get(&ca->ref);
41 return ERR_PTR(-EINVAL);
45 path = strndup_user((const char __user *)
46 (unsigned long) dev, PATH_MAX);
48 return ERR_CAST(path);
50 ca = bch2_dev_lookup(c, path);
58 static long bch2_ioctl_assemble(struct bch_ioctl_assemble __user *user_arg)
60 struct bch_ioctl_assemble arg;
62 u64 *user_devs = NULL;
67 if (copy_from_user(&arg, user_arg, sizeof(arg)))
70 if (arg.flags || arg.pad)
73 user_devs = kmalloc_array(arg.nr_devs, sizeof(u64), GFP_KERNEL);
77 devs = kcalloc(arg.nr_devs, sizeof(char *), GFP_KERNEL);
79 if (copy_from_user(user_devs, user_arg->devs,
80 sizeof(u64) * arg.nr_devs))
83 for (i = 0; i < arg.nr_devs; i++) {
84 devs[i] = strndup_user((const char __user *)(unsigned long)
93 c = bch2_fs_open(devs, arg.nr_devs, bch2_opts_empty());
94 ret = PTR_ERR_OR_ZERO(c);
99 for (i = 0; i < arg.nr_devs; i++)
105 static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg)
107 struct bch_ioctl_incremental arg;
111 if (copy_from_user(&arg, user_arg, sizeof(arg)))
114 if (arg.flags || arg.pad)
117 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
121 err = bch2_fs_open_incremental(path);
125 pr_err("Could not register bcachefs devices: %s", err);
133 static long bch2_global_ioctl(unsigned cmd, void __user *arg)
137 case BCH_IOCTL_ASSEMBLE:
138 return bch2_ioctl_assemble(arg);
139 case BCH_IOCTL_INCREMENTAL:
140 return bch2_ioctl_incremental(arg);
147 static long bch2_ioctl_query_uuid(struct bch_fs *c,
148 struct bch_ioctl_query_uuid __user *user_arg)
150 return copy_to_user(&user_arg->uuid,
152 sizeof(c->sb.user_uuid));
156 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
158 if (arg.flags || arg.pad)
161 return bch2_fs_start(c) ? -EIO : 0;
164 static long bch2_ioctl_stop(struct bch_fs *c)
171 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
176 if (arg.flags || arg.pad)
179 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
183 ret = bch2_dev_add(c, path);
189 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
193 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
194 BCH_FORCE_IF_METADATA_LOST|
195 BCH_FORCE_IF_DEGRADED|
200 ca = bch2_device_lookup(c, arg.dev, arg.flags);
204 return bch2_dev_remove(c, ca, arg.flags);
207 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
212 if (arg.flags || arg.pad)
215 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
219 ret = bch2_dev_online(c, path);
224 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
229 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
230 BCH_FORCE_IF_METADATA_LOST|
231 BCH_FORCE_IF_DEGRADED|
236 ca = bch2_device_lookup(c, arg.dev, arg.flags);
240 ret = bch2_dev_offline(c, ca, arg.flags);
241 percpu_ref_put(&ca->ref);
245 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
246 struct bch_ioctl_disk_set_state arg)
251 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
252 BCH_FORCE_IF_METADATA_LOST|
253 BCH_FORCE_IF_DEGRADED|
255 arg.pad[0] || arg.pad[1] || arg.pad[2])
258 ca = bch2_device_lookup(c, arg.dev, arg.flags);
262 ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
264 percpu_ref_put(&ca->ref);
268 struct bch_data_ctx {
270 struct bch_ioctl_data arg;
271 struct bch_move_stats stats;
275 struct task_struct *thread;
278 static int bch2_data_thread(void *arg)
280 struct bch_data_ctx *ctx = arg;
282 ctx->ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
284 ctx->stats.data_type = U8_MAX;
288 static int bch2_data_job_release(struct inode *inode, struct file *file)
290 struct bch_data_ctx *ctx = file->private_data;
292 kthread_stop(ctx->thread);
293 put_task_struct(ctx->thread);
298 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
299 size_t len, loff_t *ppos)
301 struct bch_data_ctx *ctx = file->private_data;
302 struct bch_fs *c = ctx->c;
303 struct bch_ioctl_data_event e = {
304 .type = BCH_DATA_EVENT_PROGRESS,
305 .p.data_type = ctx->stats.data_type,
306 .p.btree_id = ctx->stats.iter.btree_id,
307 .p.pos = ctx->stats.iter.pos,
308 .p.sectors_done = atomic64_read(&ctx->stats.sectors_seen),
309 .p.sectors_total = bch2_fs_sectors_used(c, bch2_fs_usage_read(c)),
315 return copy_to_user(buf, &e, sizeof(e)) ?: sizeof(e);
318 static const struct file_operations bcachefs_data_ops = {
319 .release = bch2_data_job_release,
320 .read = bch2_data_job_read,
324 static long bch2_ioctl_data(struct bch_fs *c,
325 struct bch_ioctl_data arg)
327 struct bch_data_ctx *ctx = NULL;
328 struct file *file = NULL;
329 unsigned flags = O_RDONLY|O_CLOEXEC|O_NONBLOCK;
332 if (arg.op >= BCH_DATA_OP_NR || arg.flags)
335 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
342 ctx->thread = kthread_create(bch2_data_thread, ctx, "[bcachefs]");
343 if (IS_ERR(ctx->thread)) {
344 ret = PTR_ERR(ctx->thread);
348 ret = get_unused_fd_flags(flags);
353 file = anon_inode_getfile("[bcachefs]", &bcachefs_data_ops, ctx, flags);
359 fd_install(fd, file);
361 get_task_struct(ctx->thread);
362 wake_up_process(ctx->thread);
368 if (!IS_ERR_OR_NULL(ctx->thread))
369 kthread_stop(ctx->thread);
374 static long bch2_ioctl_usage(struct bch_fs *c,
375 struct bch_ioctl_usage __user *user_arg)
377 struct bch_ioctl_usage arg;
382 if (!test_bit(BCH_FS_STARTED, &c->flags))
385 if (copy_from_user(&arg, user_arg, sizeof(arg)))
388 for (i = 0; i < arg.nr_devices; i++) {
389 struct bch_ioctl_dev_usage dst = { .alive = 0 };
391 ret = copy_to_user(&user_arg->devs[i], &dst, sizeof(dst));
397 struct bch_fs_usage src = bch2_fs_usage_read(c);
398 struct bch_ioctl_fs_usage dst = {
399 .capacity = c->capacity,
400 .used = bch2_fs_sectors_used(c, src),
401 .online_reserved = src.online_reserved,
404 for (i = 0; i < BCH_REPLICAS_MAX; i++) {
405 dst.persistent_reserved[i] =
406 src.s[i].persistent_reserved;
408 for (j = 0; j < S_ALLOC_NR; j++)
409 dst.sectors[s_alloc_to_data_type(j)][i] =
413 ret = copy_to_user(&user_arg->fs, &dst, sizeof(dst));
418 for_each_member_device(ca, c, i) {
419 struct bch_dev_usage src = bch2_dev_usage_read(c, ca);
420 struct bch_ioctl_dev_usage dst = {
422 .state = ca->mi.state,
423 .bucket_size = ca->mi.bucket_size,
424 .nr_buckets = ca->mi.nbuckets - ca->mi.first_bucket,
427 if (ca->dev_idx >= arg.nr_devices) {
428 percpu_ref_put(&ca->ref);
432 if (percpu_ref_tryget(&ca->io_ref)) {
433 dst.dev = huge_encode_dev(ca->disk_sb.bdev->bd_dev);
434 percpu_ref_put(&ca->io_ref);
437 for (j = 0; j < BCH_DATA_NR; j++) {
438 dst.buckets[j] = src.buckets[j];
439 dst.sectors[j] = src.sectors[j];
442 ret = copy_to_user(&user_arg->devs[i], &dst, sizeof(dst));
450 static long bch2_ioctl_read_super(struct bch_fs *c,
451 struct bch_ioctl_read_super arg)
453 struct bch_dev *ca = NULL;
457 if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
461 mutex_lock(&c->sb_lock);
463 if (arg.flags & BCH_READ_DEV) {
464 ca = bch2_device_lookup(c, arg.dev, arg.flags);
476 if (vstruct_bytes(sb) > arg.size) {
481 ret = copy_to_user((void __user *)(unsigned long)arg.sb,
482 sb, vstruct_bytes(sb));
485 percpu_ref_put(&ca->ref);
486 mutex_unlock(&c->sb_lock);
490 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
491 struct bch_ioctl_disk_get_idx arg)
493 dev_t dev = huge_decode_dev(arg.dev);
497 for_each_online_member(ca, c, i)
498 if (ca->disk_sb.bdev->bd_dev == dev) {
499 percpu_ref_put(&ca->io_ref);
506 static long bch2_ioctl_disk_resize(struct bch_fs *c,
507 struct bch_ioctl_disk_resize arg)
512 if ((arg.flags & ~BCH_BY_INDEX) ||
516 ca = bch2_device_lookup(c, arg.dev, arg.flags);
520 ret = bch2_dev_resize(c, ca, arg.nbuckets);
522 percpu_ref_put(&ca->ref);
526 #define BCH_IOCTL(_name, _argtype) \
530 if (copy_from_user(&i, arg, sizeof(i))) \
532 return bch2_ioctl_##_name(c, i); \
535 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
537 /* ioctls that don't require admin cap: */
539 case BCH_IOCTL_QUERY_UUID:
540 return bch2_ioctl_query_uuid(c, arg);
541 case BCH_IOCTL_USAGE:
542 return bch2_ioctl_usage(c, arg);
545 if (!capable(CAP_SYS_ADMIN))
550 case BCH_IOCTL_START:
551 BCH_IOCTL(start, struct bch_ioctl_start);
553 return bch2_ioctl_stop(c);
555 case BCH_IOCTL_READ_SUPER:
556 BCH_IOCTL(read_super, struct bch_ioctl_read_super);
557 case BCH_IOCTL_DISK_GET_IDX:
558 BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
561 if (!test_bit(BCH_FS_STARTED, &c->flags))
564 /* ioctls that do require admin cap: */
566 case BCH_IOCTL_DISK_ADD:
567 BCH_IOCTL(disk_add, struct bch_ioctl_disk);
568 case BCH_IOCTL_DISK_REMOVE:
569 BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
570 case BCH_IOCTL_DISK_ONLINE:
571 BCH_IOCTL(disk_online, struct bch_ioctl_disk);
572 case BCH_IOCTL_DISK_OFFLINE:
573 BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
574 case BCH_IOCTL_DISK_SET_STATE:
575 BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
577 BCH_IOCTL(data, struct bch_ioctl_data);
578 case BCH_IOCTL_DISK_RESIZE:
579 BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
586 static DEFINE_IDR(bch_chardev_minor);
588 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
590 unsigned minor = iminor(file_inode(filp));
591 struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
592 void __user *arg = (void __user *) v;
595 ? bch2_fs_ioctl(c, cmd, arg)
596 : bch2_global_ioctl(cmd, arg);
599 static const struct file_operations bch_chardev_fops = {
600 .owner = THIS_MODULE,
601 .unlocked_ioctl = bch2_chardev_ioctl,
602 .open = nonseekable_open,
605 static int bch_chardev_major;
606 static struct class *bch_chardev_class;
607 static struct device *bch_chardev;
609 void bch2_fs_chardev_exit(struct bch_fs *c)
611 if (!IS_ERR_OR_NULL(c->chardev))
612 device_unregister(c->chardev);
614 idr_remove(&bch_chardev_minor, c->minor);
617 int bch2_fs_chardev_init(struct bch_fs *c)
619 c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
623 c->chardev = device_create(bch_chardev_class, NULL,
624 MKDEV(bch_chardev_major, c->minor), c,
625 "bcachefs%u-ctl", c->minor);
626 if (IS_ERR(c->chardev))
627 return PTR_ERR(c->chardev);
632 void bch2_chardev_exit(void)
634 if (!IS_ERR_OR_NULL(bch_chardev_class))
635 device_destroy(bch_chardev_class,
636 MKDEV(bch_chardev_major, U8_MAX));
637 if (!IS_ERR_OR_NULL(bch_chardev_class))
638 class_destroy(bch_chardev_class);
639 if (bch_chardev_major > 0)
640 unregister_chrdev(bch_chardev_major, "bcachefs");
643 int __init bch2_chardev_init(void)
645 bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
646 if (bch_chardev_major < 0)
647 return bch_chardev_major;
649 bch_chardev_class = class_create(THIS_MODULE, "bcachefs");
650 if (IS_ERR(bch_chardev_class))
651 return PTR_ERR(bch_chardev_class);
653 bch_chardev = device_create(bch_chardev_class, NULL,
654 MKDEV(bch_chardev_major, U8_MAX),
655 NULL, "bcachefs-ctl");
656 if (IS_ERR(bch_chardev))
657 return PTR_ERR(bch_chardev);
662 #endif /* NO_BCACHEFS_CHARDEV */