Skip to contentSkip to conversation

f2fs: avoid very large discard command

PATCH In review Jaegeuk Kim 3 participants 7 messages Feb 23, 2017 – Feb 27, 2017

Subsystems (3)
lore
Share & tools
lore

Jaegeuk Kim

author maintainer
#0 lore raw

PATCH 1/4 f2fs: avoid very large discard command

This patch adds MAX_DISCARD_BLOCKS() to avoid issuing too much large single discard command.

---
 fs/f2fs/f2fs.h    | 3 ++-
 fs/f2fs/segment.c | 3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)
fs/f2fs/f2fs.h +2 -1
index d076b94530bc..5f3fe97df055 100644
@@ -133,7 +133,8 @@ enum {
 		(SM_I(sbi)->trim_sections * (sbi)->segs_per_sec)
 #define BATCHED_TRIM_BLOCKS(sbi)	\
 		(BATCHED_TRIM_SEGMENTS(sbi) << (sbi)->log_blocks_per_seg)
-
+#define MAX_DISCARD_BLOCKS(sbi)						\
+		((1 << (sbi)->log_blocks_per_seg) * (sbi)->segs_per_sec)
 #define DISCARD_ISSUE_RATE	8
 #define DEF_CP_INTERVAL			60	/* 60 secs */
 #define DEF_IDLE_INTERVAL		5	/* 5 secs */
fs/f2fs/segment.c +2 -1
index fe434cd872b4..567019940e9b 100644
@@ -886,7 +886,8 @@ static void __add_discard_entry(struct f2fs_sb_info *sbi,
 	if (!list_empty(head)) {
 		last = list_last_entry(head, struct discard_entry, list);
 		if (START_BLOCK(sbi, cpc->trim_start) + start ==
-						last->blkaddr + last->len) {
+				last->blkaddr + last->len &&
+				last->len < MAX_DISCARD_BLOCKS(sbi)) {
 			last->len += end - start;
 			goto done;
 		}
Signature · 1 line
2.11.0
author maintainer
#1 lore raw

PATCH 2/4 f2fs: much larger batched trim_fs job

We have a kernel thread to issue discard commands, so we can increase the number of batched discard sections. By default, now it becomes 4GB range.

---
 fs/f2fs/f2fs.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/f2fs/f2fs.h +1 -1
index 5f3fe97df055..8f3171f12fa5 100644
@@ -128,7 +128,7 @@ enum {
 	CP_DISCARD,
 };
 
-#define DEF_BATCHED_TRIM_SECTIONS	2
+#define DEF_BATCHED_TRIM_SECTIONS	2048
 #define BATCHED_TRIM_SEGMENTS(sbi)	\
 		(SM_I(sbi)->trim_sections * (sbi)->segs_per_sec)
 #define BATCHED_TRIM_BLOCKS(sbi)	\
Signature · 1 line
2.11.0
author maintainer
#2 lore raw

PATCH 3/4 f2fs: wait for discard completion after submission

In reply to Jaegeuk Kim

We don't need to wait for each discard commands when unmounting the image.

---
 fs/f2fs/segment.c | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)
fs/f2fs/segment.c +13 -3
index 567019940e9b..d76e49c3d1b4 100644
@@ -676,8 +676,12 @@ void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
 	struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
 	struct list_head *wait_list = &(dcc->discard_cmd_list);
 	struct discard_cmd *dc, *tmp;
+	struct blk_plug plug;
 
 	mutex_lock(&dcc->cmd_lock);
+
+	blk_start_plug(&plug);
+
 	list_for_each_entry_safe(dc, tmp, wait_list, list) {
 
 		if (blkaddr == NULL_ADDR) {
@@ -686,9 +690,6 @@ void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
 				submit_bio(dc->bio);
 				atomic_inc(&dcc->submit_discard);
 			}
-			wait_for_completion_io(&dc->wait);
-
-			__remove_discard_cmd(sbi, dc);
 			continue;
 		}
 
@@ -699,6 +700,15 @@ void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
 				__remove_discard_cmd(sbi, dc);
 		}
 	}
+	blk_finish_plug(&plug);
+
+	/* this comes from f2fs_put_super */
+	if (blkaddr == NULL_ADDR) {
+		list_for_each_entry_safe(dc, tmp, wait_list, list) {
+			wait_for_completion_io(&dc->wait);
+			__remove_discard_cmd(sbi, dc);
+		}
+	}
 	mutex_unlock(&dcc->cmd_lock);
 }
 
Signature · 1 line
2.11.0
author maintainer
#3 lore raw

PATCH 4/4 f2fs: check discard alignment only for SEQWRITE zones

In reply to Jaegeuk Kim

For converntional zones, we don't need to align discard commands to exact zone size.

---
 fs/f2fs/segment.c | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
fs/f2fs/segment.c +12 -12
index d76e49c3d1b4..6fff1c58b3b1 100644
@@ -793,24 +793,13 @@ static int __f2fs_issue_discard_async(struct f2fs_sb_info *sbi,
 static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
 		struct block_device *bdev, block_t blkstart, block_t blklen)
 {
-	sector_t nr_sects = SECTOR_FROM_BLOCK(blklen);
-	sector_t sector;
+	sector_t sector, nr_sects;
 	int devi = 0;
 
 	if (sbi->s_ndevs) {
 		devi = f2fs_target_device_index(sbi, blkstart);
 		blkstart -= FDEV(devi).start_blk;
 	}
-	sector = SECTOR_FROM_BLOCK(blkstart);
-
-	if (sector & (bdev_zone_sectors(bdev) - 1) ||
-	    nr_sects != bdev_zone_sectors(bdev)) {
-		f2fs_msg(sbi->sb, KERN_INFO,
-			"(%d) %s: Unaligned discard attempted (block %x + %x)",
-			devi, sbi->s_ndevs ? FDEV(devi).path: "",
-			blkstart, blklen);
-		return -EIO;
-	}
 
 	/*
 	 * We need to know the type of the zone: for conventional zones,
@@ -825,6 +814,17 @@ static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
 		return __f2fs_issue_discard_async(sbi, bdev, blkstart, blklen);
 	case BLK_ZONE_TYPE_SEQWRITE_REQ:
 	case BLK_ZONE_TYPE_SEQWRITE_PREF:
+		sector = SECTOR_FROM_BLOCK(blkstart);
+		nr_sects = SECTOR_FROM_BLOCK(blklen);
+
+		if (sector & (bdev_zone_sectors(bdev) - 1) ||
+				nr_sects != bdev_zone_sectors(bdev)) {
+			f2fs_msg(sbi->sb, KERN_INFO,
+				"(%d) %s: Unaligned discard attempted (block %x + %x)",
+				devi, sbi->s_ndevs ? FDEV(devi).path: "",
+				blkstart, blklen);
+			return -EIO;
+		}
 		trace_f2fs_issue_reset_zone(bdev, blkstart);
 		return blkdev_reset_zones(bdev, sector,
 					  nr_sects, GFP_NOFS);
Signature · 6 lines
2.11.0


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
#4 lore raw
In reply to Jaegeuk Kim
Quoted 2 lines · Jaegeuk Kim

This patch adds MAX_DISCARD_BLOCKS() to avoid issuing too much large single discard command.

Needs an explanation in the code on why this number was chosen. In doubt I suspect it should be a quirk in the driver for the device, and not something decided by the fs.

Quoted 38 lines
---
 fs/f2fs/f2fs.h    | 3 ++-
 fs/f2fs/segment.c | 3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)
fs/f2fs/f2fs.h +2 -1
index d076b94530bc..5f3fe97df055 100644
@@ -133,7 +133,8 @@ enum {
 		(SM_I(sbi)->trim_sections * (sbi)->segs_per_sec)
 #define BATCHED_TRIM_BLOCKS(sbi)	\
 		(BATCHED_TRIM_SEGMENTS(sbi) << (sbi)->log_blocks_per_seg)
-
+#define MAX_DISCARD_BLOCKS(sbi)						\
+		((1 << (sbi)->log_blocks_per_seg) * (sbi)->segs_per_sec)
 #define DISCARD_ISSUE_RATE	8
 #define DEF_CP_INTERVAL			60	/* 60 secs */
 #define DEF_IDLE_INTERVAL		5	/* 5 secs */
fs/f2fs/segment.c +2 -1
index fe434cd872b4..567019940e9b 100644
@@ -886,7 +886,8 @@ static void __add_discard_entry(struct f2fs_sb_info *sbi,
 	if (!list_empty(head)) {
 		last = list_last_entry(head, struct discard_entry, list);
 		if (START_BLOCK(sbi, cpc->trim_start) + start ==
-						last->blkaddr + last->len) {
+				last->blkaddr + last->len &&
+				last->len < MAX_DISCARD_BLOCKS(sbi)) {
 			last->len += end - start;
 			goto done;
 		}
Signature · 1 line
2.11.0

---end quoted text---

author maintainer
#5 lore raw
Quoted 5 lines · Christoph Hellwig
Quoted 2 lines · Jaegeuk Kim

This patch adds MAX_DISCARD_BLOCKS() to avoid issuing too much large single discard command.

Needs an explanation in the code on why this number was chosen.

No need to add a trivial comment, since it's the section size which we can easily translate it.

Quoted 2 lines

In doubt I suspect it should be a quirk in the driver for the device, and not something decided by the fs.

The block allocator checks all the issued bios whether any of them contains the newly allocated address or not. If one large discard command was issued, it needs to wait for its completion, even if we want to allocate part of its address space.

In addition, HM-SMR is required to issue discard/reset in a unit of section size.

Thanks,

Quoted 32 lines
Quoted 32 lines
---
 fs/f2fs/f2fs.h    | 3 ++-
 fs/f2fs/segment.c | 3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)
fs/f2fs/f2fs.h +2 -1
index d076b94530bc..5f3fe97df055 100644
@@ -133,7 +133,8 @@ enum {
 		(SM_I(sbi)->trim_sections * (sbi)->segs_per_sec)
 #define BATCHED_TRIM_BLOCKS(sbi)	\
 		(BATCHED_TRIM_SEGMENTS(sbi) << (sbi)->log_blocks_per_seg)
-
+#define MAX_DISCARD_BLOCKS(sbi)						\
+		((1 << (sbi)->log_blocks_per_seg) * (sbi)->segs_per_sec)
 #define DISCARD_ISSUE_RATE	8
 #define DEF_CP_INTERVAL			60	/* 60 secs */
 #define DEF_IDLE_INTERVAL		5	/* 5 secs */
fs/f2fs/segment.c +1 -1
index fe434cd872b4..567019940e9b 100644
@@ -886,7 +886,8 @@ static void __add_discard_entry(struct f2fs_sb_info *sbi,
 	if (!list_empty(head)) {
 		last = list_last_entry(head, struct discard_entry, list);
 		if (START_BLOCK(sbi, cpc->trim_start) + start ==
-						last->blkaddr + last->len) {
+				last->blkaddr + last->len &&
Quoted 8 lines
Quoted 7 lines
fs/f2fs/segment.c +1 -0
+				last->len < MAX_DISCARD_BLOCKS(sbi)) {
 			last->len += end - start;
 			goto done;
 		}
Signature · 1 line
2.11.0

---end quoted text---

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot

Reviewed-by on v1 1/4
#6 lore raw
In reply to Jaegeuk Kim
Quoted 4 lines · Jaegeuk Kim

This patch adds MAX_DISCARD_BLOCKS() to avoid issuing too much large single discard command.

This patch set looks good to me, please add reviewed-by into all patches. :)

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot