Add BLKDISCARD ioctl to allow userspace to discard sectors

We may well want mkfs tools to use this to mark the whole device as unwanted before they format it, for example. The ioctl takes a pair of uint64_ts, which are start offset and length in _bytes_. Although at the moment it might make sense for them both to be in 512-byte sectors, I don't want to limit the ABI to that. Signed-off-by: David Woodhouse <David.Woodhouse@intel.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Add BLKDISCARD ioctl to allow userspace to discard sectors
We may well want mkfs tools to use this to mark the whole device as unwanted before they format it, for example. The ioctl takes a pair of uint64_ts, which are start offset and length in _bytes_. Although at the moment it might make sense for them both to be in 512-byte sectors, I don't want to limit the ABI to that. Signed-off-by: David Woodhouse <David.Woodhouse@intel.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
David Woodhouse · Jens Axboe
1 parent 2ebca85abc
Showing 3 changed files with 78 additions and 0 deletions Side-by-side Diff
block/compat_ioctl.c
block/ioctl.c
include/linux/fs.h
@@ -788,6 +788,7 @@
 		return compat_hdio_getgeo(disk, bdev, compat_ptr(arg));
 	case BLKFLSBUF:
 	case BLKROSET:
+	case BLKDISCARD:
 	/*
 	 * the ones below are implemented in blkdev_locked_ioctl,
 	 * but we call blkdev_ioctl, which gets the lock for us
@@ -111,6 +111,69 @@
 	return res;
 }
  
+static void blk_ioc_discard_endio(struct bio *bio, int err)
+{
+	if (err) {
+		if (err == -EOPNOTSUPP)
+			set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
+		clear_bit(BIO_UPTODATE, &bio->bi_flags);
+	}
+	complete(bio->bi_private);
+}
+
+static int blk_ioctl_discard(struct block_device *bdev, uint64_t start,
+			     uint64_t len)
+{
+	struct request_queue *q = bdev_get_queue(bdev);
+	int ret = 0;
+
+	if (start & 511)
+		return -EINVAL;
+	if (len & 511)
+		return -EINVAL;
+	start >>= 9;
+	len >>= 9;
+
+	if (start + len > (bdev->bd_inode->i_size >> 9))
+		return -EINVAL;
+
+	if (!q->prepare_discard_fn)
+		return -EOPNOTSUPP;
+
+	while (len && !ret) {
+		DECLARE_COMPLETION_ONSTACK(wait);
+		struct bio *bio;
+
+		bio = bio_alloc(GFP_KERNEL, 0);
+		if (!bio)
+			return -ENOMEM;
+
+		bio->bi_end_io = blk_ioc_discard_endio;
+		bio->bi_bdev = bdev;
+		bio->bi_private = &wait;
+		bio->bi_sector = start;
+
+		if (len > q->max_hw_sectors) {
+			bio->bi_size = q->max_hw_sectors << 9;
+			len -= q->max_hw_sectors;
+			start += q->max_hw_sectors;
+		} else {
+			bio->bi_size = len << 9;
+			len = 0;
+		}
+		submit_bio(WRITE_DISCARD, bio);
+
+		wait_for_completion(&wait);
+
+		if (bio_flagged(bio, BIO_EOPNOTSUPP))
+			ret = -EOPNOTSUPP;
+		else if (!bio_flagged(bio, BIO_UPTODATE))
+			ret = -EIO;
+		bio_put(bio);
+	}
+	return ret;
+}
+
 static int put_ushort(unsigned long arg, unsigned short val)
 {
 	return put_user(val, (unsigned short __user *)arg);
@@ -258,6 +321,19 @@
 		set_device_ro(bdev, n);
 		unlock_kernel();
 		return 0;
+
+	case BLKDISCARD: {
+		uint64_t range[2];
+
+		if (!(file->f_mode & FMODE_WRITE))
+			return -EBADF;
+
+		if (copy_from_user(range, (void __user *)arg, sizeof(range)))
+			return -EFAULT;
+
+		return blk_ioctl_discard(bdev, range[0], range[1]);
+	}
+
 	case HDIO_GETGEO: {
 		struct hd_geometry geo;
  
@@ -223,6 +223,7 @@
 #define BLKTRACESTART _IO(0x12,116)
 #define BLKTRACESTOP _IO(0x12,117)
 #define BLKTRACETEARDOWN _IO(0x12,118)
+#define BLKDISCARD _IO(0x12,119)
  
 #define BMAP_IOCTL 1		/* obsolete - kept for compatibility */
 #define FIBMAP	   _IO(0x00,1)	/* bmap access */
...	...	@@ -788,6 +788,7 @@
788	788	return compat_hdio_getgeo(disk, bdev, compat_ptr(arg));
789	789	case BLKFLSBUF:
790	790	case BLKROSET:
	791	+ case BLKDISCARD:
791	792	/*
792	793	* the ones below are implemented in blkdev_locked_ioctl,
793	794	* but we call blkdev_ioctl, which gets the lock for us
...	...	@@ -111,6 +111,69 @@
111	111	return res;
112	112	}
113	113
	114	+static void blk_ioc_discard_endio(struct bio *bio, int err)
	115	+{
	116	+ if (err) {
	117	+ if (err == -EOPNOTSUPP)
	118	+ set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
	119	+ clear_bit(BIO_UPTODATE, &bio->bi_flags);
	120	+ }
	121	+ complete(bio->bi_private);
	122	+}
	123	+
	124	+static int blk_ioctl_discard(struct block_device *bdev, uint64_t start,
	125	+ uint64_t len)
	126	+{
	127	+ struct request_queue *q = bdev_get_queue(bdev);
	128	+ int ret = 0;
	129	+
	130	+ if (start & 511)
	131	+ return -EINVAL;
	132	+ if (len & 511)
	133	+ return -EINVAL;
	134	+ start >>= 9;
	135	+ len >>= 9;
	136	+
	137	+ if (start + len > (bdev->bd_inode->i_size >> 9))
	138	+ return -EINVAL;
	139	+
	140	+ if (!q->prepare_discard_fn)
	141	+ return -EOPNOTSUPP;
	142	+
	143	+ while (len && !ret) {
	144	+ DECLARE_COMPLETION_ONSTACK(wait);
	145	+ struct bio *bio;
	146	+
	147	+ bio = bio_alloc(GFP_KERNEL, 0);
	148	+ if (!bio)
	149	+ return -ENOMEM;
	150	+
	151	+ bio->bi_end_io = blk_ioc_discard_endio;
	152	+ bio->bi_bdev = bdev;
	153	+ bio->bi_private = &wait;
	154	+ bio->bi_sector = start;
	155	+
	156	+ if (len > q->max_hw_sectors) {
	157	+ bio->bi_size = q->max_hw_sectors << 9;
	158	+ len -= q->max_hw_sectors;
	159	+ start += q->max_hw_sectors;
	160	+ } else {
	161	+ bio->bi_size = len << 9;
	162	+ len = 0;
	163	+ }
	164	+ submit_bio(WRITE_DISCARD, bio);
	165	+
	166	+ wait_for_completion(&wait);
	167	+
	168	+ if (bio_flagged(bio, BIO_EOPNOTSUPP))
	169	+ ret = -EOPNOTSUPP;
	170	+ else if (!bio_flagged(bio, BIO_UPTODATE))
	171	+ ret = -EIO;
	172	+ bio_put(bio);
	173	+ }
	174	+ return ret;
	175	+}
	176	+
114	177	static int put_ushort(unsigned long arg, unsigned short val)
115	178	{
116	179	return put_user(val, (unsigned short __user *)arg);
...	...	@@ -258,6 +321,19 @@
258	321	set_device_ro(bdev, n);
259	322	unlock_kernel();
260	323	return 0;
	324	+
	325	+ case BLKDISCARD: {
	326	+ uint64_t range[2];
	327	+
	328	+ if (!(file->f_mode & FMODE_WRITE))
	329	+ return -EBADF;
	330	+
	331	+ if (copy_from_user(range, (void __user *)arg, sizeof(range)))
	332	+ return -EFAULT;
	333	+
	334	+ return blk_ioctl_discard(bdev, range[0], range[1]);
	335	+ }
	336	+
261	337	case HDIO_GETGEO: {
262	338	struct hd_geometry geo;
263	339
...	...	@@ -223,6 +223,7 @@
223	223	#define BLKTRACESTART _IO(0x12,116)
224	224	#define BLKTRACESTOP _IO(0x12,117)
225	225	#define BLKTRACETEARDOWN _IO(0x12,118)
	226	+#define BLKDISCARD _IO(0x12,119)
226	227
227	228	#define BMAP_IOCTL 1 /* obsolete - kept for compatibility */
228	229	#define FIBMAP _IO(0x00,1) /* bmap access */