Linux-BTRFS Archive on lore.kernel.org
 help / Atom feed
From: Nikolay Borisov <nborisov@suse.com>
To: linux-btrfs@vger.kernel.org
Cc: jeffm@suse.com, Nikolay Borisov <nborisov@suse.com>
Subject: [PATCH 12/15] btrfs: Optimize unallocated chunks discard
Date: Wed, 30 Jan 2019 16:50:59 +0200
Message-ID: <20190130145102.4708-13-nborisov@suse.com> (raw)
In-Reply-To: <20190130145102.4708-1-nborisov@suse.com>

Currently unallocated chunks are always trimmed. For example
2 consecutive trims on large storage would trim freespace twice
irrespective of whether the space was actually allocated or not between
those trims.

Optimise this behavior by exploiting the newly introduced alloc_state
tree of btrfs_device. A new CHUNK_TRIMMED bit is used to mark
those unallocated chunks which have been trimmed and have not been
allocated afterwards. On chunk allocation the respective underlying devices'
physical space will have its CHUNK_TRIMMED flag cleared. This avoids
submitting discards for space which hasn't been changed since the last
time discard was issued.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
---
 fs/btrfs/extent-tree.c | 57 +++++++++++++++++++++++++++++++++++++++++-
 fs/btrfs/extent_io.h   |  8 +++++-
 fs/btrfs/extent_map.c  |  4 ++-
 3 files changed, 66 insertions(+), 3 deletions(-)

diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 610bed028511..f5005ef39f98 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -11127,6 +11127,54 @@ int btrfs_error_unpin_extent_range(struct btrfs_fs_info *fs_info,
 	return unpin_extent_range(fs_info, start, end, false);
 }
 
+static bool should_skip_trim(struct btrfs_device *device, u64 *start, u64 *len)
+{
+	u64 trimmed_start = 0, trimmed_end = 0;
+	u64 end = *start + *len - 1;
+
+	if (!find_first_extent_bit(&device->alloc_state, *start, &trimmed_start,
+				   &trimmed_end, CHUNK_TRIMMED, NULL)) {
+		u64 trimmed_len = trimmed_end - trimmed_start + 1;
+
+		if (*start < trimmed_start) {
+			if (in_range(end, trimmed_start, trimmed_len) ||
+			    end > trimmed_end) {
+				/*
+				 * start|------|end
+				 *      ts|--|trimmed_len
+				 *      OR
+				 * start|-----|end
+				 *      ts|-----|trimmed_len
+				 */
+				*len = trimmed_start - *start;
+				return false;
+			} else if (end < trimmed_start) {
+				/*
+				 * start|------|end
+				 *             ts|--|trimmed_len
+				 */
+				return false;
+			}
+		} else if (in_range(*start, trimmed_start, trimmed_len)) {
+			if (in_range(end, trimmed_start, trimmed_len)) {
+				/*
+				 * start|------|end
+				 *  ts|----------|trimmed_len
+				 */
+				return true;
+			} else {
+				/*
+				 * start|-----------|end
+				 *  ts|----------|trimmed_len
+				 */
+				*start = trimmed_end + 1;
+				*len = end - *start + 1;
+				return false;
+			}
+		}
+	}
+	return false;
+}
 /*
  * It used to be that old block groups would be left around forever.
  * Iterating over them would be enough to trim unused space.  Since we
@@ -11197,7 +11245,14 @@ static int btrfs_trim_free_extents(struct btrfs_device *device,
 		start = max(range->start, start);
 		len = min(range->len, len);
 
-		ret = btrfs_issue_discard(device->bdev, start, len, &bytes);
+		if (!should_skip_trim(device, &start, &len)) {
+			ret = btrfs_issue_discard(device->bdev, start, len,
+						  &bytes);
+			if (!ret)
+				set_extent_bits(&device->alloc_state, start,
+						start + bytes - 1,
+						CHUNK_TRIMMED);
+		}
 		mutex_unlock(&fs_info->chunk_mutex);
 
 		if (ret)
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index d4227e40c8ee..d238efd628cf 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -30,8 +30,14 @@
 #define EXTENT_CTLBITS		(EXTENT_DO_ACCOUNTING)
 
 
-/* Redefined bits above which are used only in the device allocation tree */
+/*
+ * Redefined bits above which are used only in the device allocation tree,
+ * shouldn't be using EXTENT_IOBITS(EXTENT_LOCKED/EXTENT_WRITEBACK) /
+ * EXTENT_BOUNDARY / EXTENT_CLEAR_META_RESV / EXTENT_CLEAR_DATA_RESV because
+ * they have special meaning to the bit manipulation functions
+ */
 #define CHUNK_ALLOCATED EXTENT_DIRTY
+#define CHUNK_TRIMMED   EXTENT_DEFRAG
 
 /*
  * flags for bio submission. The high bits indicate the compression
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index 0820f6fcf3a6..9e8c0904f623 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -389,8 +389,10 @@ int add_extent_mapping(struct extent_map_tree *tree,
 		goto out;
 
 	setup_extent_mapping(tree, em, modified);
-	if (test_bit(EXTENT_FLAG_FS_MAPPING, &em->flags))
+	if (test_bit(EXTENT_FLAG_FS_MAPPING, &em->flags)) {
 		extent_map_device_set_bits(em, CHUNK_ALLOCATED);
+		extent_map_device_clear_bits(em, CHUNK_TRIMMED);
+	}
 out:
 	return ret;
 }
-- 
2.17.1


  parent reply index

Thread overview: 36+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-01-30 14:50 [PATCH 00/15] Improvements to fitrim Nikolay Borisov
2019-01-30 14:50 ` [PATCH 01/15] btrfs: Honour FITRIM range constraints during free space trim Nikolay Borisov
2019-01-31 15:21   ` David Sterba
2019-01-31 15:35     ` Nikolay Borisov
2019-01-31 15:48       ` David Sterba
2019-01-31 19:30       ` Jeff Mahoney
2019-01-31 21:45         ` Nikolay Borisov
2019-01-30 14:50 ` [PATCH 02/15] btrfs: Make WARN_ON in a canonical form Nikolay Borisov
2019-01-31 15:22   ` David Sterba
2019-02-04 13:12   ` Johannes Thumshirn
2019-01-30 14:50 ` [PATCH 03/15] btrfs: Remove EXTENT_FIRST_DELALLOC bit Nikolay Borisov
2019-01-31 15:23   ` David Sterba
2019-02-04 13:15   ` Johannes Thumshirn
2019-01-30 14:50 ` [PATCH 04/15] btrfs: combine device update operations during transaction commit Nikolay Borisov
2019-02-04 13:25   ` Johannes Thumshirn
2019-02-04 14:48     ` Nikolay Borisov
2019-02-05  9:21       ` Johannes Thumshirn
2019-01-30 14:50 ` [PATCH 05/15] btrfs: Handle pending/pinned chunks before blockgroup relocation during device shrink Nikolay Borisov
2019-02-04 13:29   ` Johannes Thumshirn
2019-01-30 14:50 ` [PATCH 06/15] btrfs: Rename and export clear_btree_io_tree Nikolay Borisov
2019-02-04 13:31   ` Johannes Thumshirn
2019-01-30 14:50 ` [PATCH 07/15] btrfs: Populate ->orig_block_len during read_one_chunk Nikolay Borisov
2019-01-30 14:50 ` [PATCH 08/15] btrfs: Introduce new bits for device allocation tree Nikolay Borisov
2019-01-30 14:50 ` [PATCH 09/15] btrfs: replace pending/pinned chunks lists with io tree Nikolay Borisov
2019-01-30 14:50 ` [PATCH 10/15] btrfs: Remove 'trans' argument from find_free_dev_extent(_start) Nikolay Borisov
2019-02-04 14:36   ` Johannes Thumshirn
2019-01-30 14:50 ` [PATCH 11/15] btrfs: Factor out in_range macro Nikolay Borisov
2019-02-04 13:57   ` Johannes Thumshirn
2019-01-30 14:50 ` Nikolay Borisov [this message]
2019-01-30 14:51 ` [PATCH 13/15] btrfs: Fix gross misnaming Nikolay Borisov
2019-01-30 14:51 ` [PATCH 14/15] btrfs: Implement find_first_clear_extent_bit Nikolay Borisov
2019-02-04 14:04   ` Johannes Thumshirn
2019-02-04 16:57   ` Nikolay Borisov
2019-01-30 14:51 ` [PATCH 15/15] btrfs: Switch btrfs_trim_free_extents to find_first_clear_extent_bit Nikolay Borisov
2019-01-31 15:38   ` [PATCH v2] " Nikolay Borisov
2019-01-31 15:41   ` [PATCH v3] " Nikolay Borisov

Reply instructions:

You may reply publically to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190130145102.4708-13-nborisov@suse.com \
    --to=nborisov@suse.com \
    --cc=jeffm@suse.com \
    --cc=linux-btrfs@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-BTRFS Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-btrfs/0 linux-btrfs/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-btrfs linux-btrfs/ https://lore.kernel.org/linux-btrfs \
		linux-btrfs@vger.kernel.org linux-btrfs@archiver.kernel.org
	public-inbox-index linux-btrfs


Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-btrfs


AGPL code for this site: git clone https://public-inbox.org/ public-inbox