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);
57 static long bch2_ioctl_assemble(struct bch_ioctl_assemble __user *user_arg)
59 struct bch_ioctl_assemble arg;
61 u64 *user_devs = NULL;
66 if (copy_from_user(&arg, user_arg, sizeof(arg)))
69 if (arg.flags || arg.pad)
72 user_devs = kmalloc_array(arg.nr_devs, sizeof(u64), GFP_KERNEL);
76 devs = kcalloc(arg.nr_devs, sizeof(char *), GFP_KERNEL);
78 if (copy_from_user(user_devs, user_arg->devs,
79 sizeof(u64) * arg.nr_devs))
82 for (i = 0; i < arg.nr_devs; i++) {
83 devs[i] = strndup_user((const char __user *)(unsigned long)
92 c = bch2_fs_open(devs, arg.nr_devs, bch2_opts_empty());
93 ret = PTR_ERR_OR_ZERO(c);
98 for (i = 0; i < arg.nr_devs; i++)
104 static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg)
106 struct bch_ioctl_incremental arg;
110 if (copy_from_user(&arg, user_arg, sizeof(arg)))
113 if (arg.flags || arg.pad)
116 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
120 err = bch2_fs_open_incremental(path);
124 pr_err("Could not register bcachefs devices: %s", err);
131 static long bch2_global_ioctl(unsigned cmd, void __user *arg)
134 case BCH_IOCTL_ASSEMBLE:
135 return bch2_ioctl_assemble(arg);
136 case BCH_IOCTL_INCREMENTAL:
137 return bch2_ioctl_incremental(arg);
143 static long bch2_ioctl_query_uuid(struct bch_fs *c,
144 struct bch_ioctl_query_uuid __user *user_arg)
146 return copy_to_user(&user_arg->uuid,
148 sizeof(c->sb.user_uuid));
151 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
153 if (arg.flags || arg.pad)
156 return bch2_fs_start(c) ? -EIO : 0;
159 static long bch2_ioctl_stop(struct bch_fs *c)
165 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
170 if (arg.flags || arg.pad)
173 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
177 ret = bch2_dev_add(c, path);
183 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
187 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
188 BCH_FORCE_IF_METADATA_LOST|
189 BCH_FORCE_IF_DEGRADED|
194 ca = bch2_device_lookup(c, arg.dev, arg.flags);
198 return bch2_dev_remove(c, ca, arg.flags);
201 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
206 if (arg.flags || arg.pad)
209 path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
213 ret = bch2_dev_online(c, path);
218 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
223 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
224 BCH_FORCE_IF_METADATA_LOST|
225 BCH_FORCE_IF_DEGRADED|
230 ca = bch2_device_lookup(c, arg.dev, arg.flags);
234 ret = bch2_dev_offline(c, ca, arg.flags);
235 percpu_ref_put(&ca->ref);
239 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
240 struct bch_ioctl_disk_set_state arg)
245 if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
246 BCH_FORCE_IF_METADATA_LOST|
247 BCH_FORCE_IF_DEGRADED|
249 arg.pad[0] || arg.pad[1] || arg.pad[2])
252 ca = bch2_device_lookup(c, arg.dev, arg.flags);
256 ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
258 percpu_ref_put(&ca->ref);
262 struct bch_data_ctx {
264 struct bch_ioctl_data arg;
265 struct bch_move_stats stats;
269 struct task_struct *thread;
272 static int bch2_data_thread(void *arg)
274 struct bch_data_ctx *ctx = arg;
276 ctx->ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
278 ctx->stats.data_type = U8_MAX;
282 static int bch2_data_job_release(struct inode *inode, struct file *file)
284 struct bch_data_ctx *ctx = file->private_data;
286 kthread_stop(ctx->thread);
287 put_task_struct(ctx->thread);
292 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
293 size_t len, loff_t *ppos)
295 struct bch_data_ctx *ctx = file->private_data;
296 struct bch_fs *c = ctx->c;
297 struct bch_ioctl_data_progress p = {
298 .data_type = ctx->stats.data_type,
299 .btree_id = ctx->stats.iter.btree_id,
300 .pos = ctx->stats.iter.pos,
301 .sectors_done = atomic64_read(&ctx->stats.sectors_seen),
302 .sectors_total = bch2_fs_sectors_used(c, bch2_fs_usage_read(c)),
305 if (len != sizeof(p))
308 return copy_to_user(buf, &p, sizeof(p)) ?: sizeof(p);
311 static const struct file_operations bcachefs_data_ops = {
312 .release = bch2_data_job_release,
313 .read = bch2_data_job_read,
317 static long bch2_ioctl_data(struct bch_fs *c,
318 struct bch_ioctl_data arg)
320 struct bch_data_ctx *ctx = NULL;
321 struct file *file = NULL;
322 unsigned flags = O_RDONLY|O_CLOEXEC|O_NONBLOCK;
325 if (arg.op >= BCH_DATA_OP_NR || arg.flags)
328 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
335 ctx->thread = kthread_create(bch2_data_thread, ctx, "[bcachefs]");
336 if (IS_ERR(ctx->thread)) {
337 ret = PTR_ERR(ctx->thread);
341 ret = get_unused_fd_flags(flags);
346 file = anon_inode_getfile("[bcachefs]", &bcachefs_data_ops, ctx, flags);
352 fd_install(fd, file);
354 get_task_struct(ctx->thread);
355 wake_up_process(ctx->thread);
361 if (!IS_ERR_OR_NULL(ctx->thread))
362 kthread_stop(ctx->thread);
367 static long bch2_ioctl_usage(struct bch_fs *c,
368 struct bch_ioctl_usage __user *user_arg)
370 struct bch_ioctl_usage arg;
375 if (copy_from_user(&arg, user_arg, sizeof(arg)))
378 for (i = 0; i < arg.nr_devices; i++) {
379 struct bch_ioctl_dev_usage dst = { .alive = 0 };
381 ret = copy_to_user(&user_arg->devs[i], &dst, sizeof(dst));
387 struct bch_fs_usage src = bch2_fs_usage_read(c);
388 struct bch_ioctl_fs_usage dst = {
389 .capacity = c->capacity,
390 .used = bch2_fs_sectors_used(c, src),
391 .online_reserved = src.online_reserved,
394 for (i = 0; i < BCH_REPLICAS_MAX; i++) {
395 dst.persistent_reserved[i] =
396 src.s[i].persistent_reserved;
398 for (j = 0; j < S_ALLOC_NR; j++)
399 dst.sectors[s_alloc_to_data_type(j)][i] =
403 ret = copy_to_user(&user_arg->fs, &dst, sizeof(dst));
408 for_each_member_device(ca, c, i) {
409 struct bch_dev_usage src = bch2_dev_usage_read(c, ca);
410 struct bch_ioctl_dev_usage dst = {
412 .state = ca->mi.state,
413 .bucket_size = ca->mi.bucket_size,
414 .nr_buckets = ca->mi.nbuckets - ca->mi.first_bucket,
417 if (ca->dev_idx >= arg.nr_devices) {
418 percpu_ref_put(&ca->ref);
422 if (percpu_ref_tryget(&ca->io_ref)) {
423 dst.dev = huge_encode_dev(ca->disk_sb.bdev->bd_dev);
424 percpu_ref_put(&ca->io_ref);
427 for (j = 0; j < BCH_DATA_NR; j++) {
428 dst.buckets[j] = src.buckets[j];
429 dst.sectors[j] = src.sectors[j];
432 ret = copy_to_user(&user_arg->devs[i], &dst, sizeof(dst));
440 static long bch2_ioctl_read_super(struct bch_fs *c,
441 struct bch_ioctl_read_super arg)
443 struct bch_dev *ca = NULL;
447 if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
451 mutex_lock(&c->sb_lock);
453 if (arg.flags & BCH_READ_DEV) {
454 ca = bch2_device_lookup(c, arg.dev, arg.flags);
466 if (vstruct_bytes(sb) > arg.size) {
471 ret = copy_to_user((void __user *)(unsigned long)arg.sb,
472 sb, vstruct_bytes(sb));
475 percpu_ref_put(&ca->ref);
476 mutex_unlock(&c->sb_lock);
480 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
481 struct bch_ioctl_disk_get_idx arg)
483 dev_t dev = huge_decode_dev(arg.dev);
487 for_each_online_member(ca, c, i)
488 if (ca->disk_sb.bdev->bd_dev == dev) {
489 percpu_ref_put(&ca->io_ref);
496 static long bch2_ioctl_disk_resize(struct bch_fs *c,
497 struct bch_ioctl_disk_resize arg)
502 if ((arg.flags & ~BCH_BY_INDEX) ||
506 ca = bch2_device_lookup(c, arg.dev, arg.flags);
510 ret = bch2_dev_resize(c, ca, arg.nbuckets);
512 percpu_ref_put(&ca->ref);
516 #define BCH_IOCTL(_name, _argtype) \
520 if (copy_from_user(&i, arg, sizeof(i))) \
522 return bch2_ioctl_##_name(c, i); \
525 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
527 /* ioctls that don't require admin cap: */
529 case BCH_IOCTL_QUERY_UUID:
530 return bch2_ioctl_query_uuid(c, arg);
531 case BCH_IOCTL_USAGE:
532 return bch2_ioctl_usage(c, arg);
535 if (!capable(CAP_SYS_ADMIN))
538 /* ioctls that do require admin cap: */
540 case BCH_IOCTL_START:
541 BCH_IOCTL(start, struct bch_ioctl_start);
543 return bch2_ioctl_stop(c);
545 case BCH_IOCTL_DISK_ADD:
546 BCH_IOCTL(disk_add, struct bch_ioctl_disk);
547 case BCH_IOCTL_DISK_REMOVE:
548 BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
549 case BCH_IOCTL_DISK_ONLINE:
550 BCH_IOCTL(disk_online, struct bch_ioctl_disk);
551 case BCH_IOCTL_DISK_OFFLINE:
552 BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
553 case BCH_IOCTL_DISK_SET_STATE:
554 BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
556 BCH_IOCTL(data, struct bch_ioctl_data);
557 case BCH_IOCTL_READ_SUPER:
558 BCH_IOCTL(read_super, struct bch_ioctl_read_super);
559 case BCH_IOCTL_DISK_GET_IDX:
560 BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
561 case BCH_IOCTL_DISK_RESIZE:
562 BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
569 static DEFINE_IDR(bch_chardev_minor);
571 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
573 unsigned minor = iminor(file_inode(filp));
574 struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
575 void __user *arg = (void __user *) v;
578 ? bch2_fs_ioctl(c, cmd, arg)
579 : bch2_global_ioctl(cmd, arg);
582 static const struct file_operations bch_chardev_fops = {
583 .owner = THIS_MODULE,
584 .unlocked_ioctl = bch2_chardev_ioctl,
585 .open = nonseekable_open,
588 static int bch_chardev_major;
589 static struct class *bch_chardev_class;
590 static struct device *bch_chardev;
592 void bch2_fs_chardev_exit(struct bch_fs *c)
594 if (!IS_ERR_OR_NULL(c->chardev))
595 device_unregister(c->chardev);
597 idr_remove(&bch_chardev_minor, c->minor);
600 int bch2_fs_chardev_init(struct bch_fs *c)
602 c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
606 c->chardev = device_create(bch_chardev_class, NULL,
607 MKDEV(bch_chardev_major, c->minor), c,
608 "bcachefs%u-ctl", c->minor);
609 if (IS_ERR(c->chardev))
610 return PTR_ERR(c->chardev);
615 void bch2_chardev_exit(void)
617 if (!IS_ERR_OR_NULL(bch_chardev_class))
618 device_destroy(bch_chardev_class,
619 MKDEV(bch_chardev_major, U8_MAX));
620 if (!IS_ERR_OR_NULL(bch_chardev_class))
621 class_destroy(bch_chardev_class);
622 if (bch_chardev_major > 0)
623 unregister_chrdev(bch_chardev_major, "bcachefs");
626 int __init bch2_chardev_init(void)
628 bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
629 if (bch_chardev_major < 0)
630 return bch_chardev_major;
632 bch_chardev_class = class_create(THIS_MODULE, "bcachefs");
633 if (IS_ERR(bch_chardev_class))
634 return PTR_ERR(bch_chardev_class);
636 bch_chardev = device_create(bch_chardev_class, NULL,
637 MKDEV(bch_chardev_major, U8_MAX),
638 NULL, "bcachefs-ctl");
639 if (IS_ERR(bch_chardev))
640 return PTR_ERR(bch_chardev);
645 #endif /* NO_BCACHEFS_CHARDEV */