[PATCH 1/2] f2fs: fix a missing discard prefree segments

[PATCH 1/2] f2fs: fix a missing discard prefree segments

[Yunlei He]

If userspace issue a fstrim with a range not involve prefree segments,
it will reuse these segments without discard. This patch fix it.

Signed-off-by: Yunlei He <heyunlei@huawei.com>
---
 fs/f2fs/segment.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index d7d5727..5b4468f 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -899,6 +899,9 @@ void clear_prefree_segments(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 	unsigned int secno, start_segno;
 	bool force = (cpc->reason == CP_DISCARD);
 
+	if (!test_opt(sbi, DISCARD))
+		return;
+
 	blk_start_plug(&plug);
 
 	mutex_lock(&dirty_i->seglist_lock);
@@ -910,15 +913,15 @@ void clear_prefree_segments(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 			break;
 		end = find_next_zero_bit(prefree_map, MAIN_SEGS(sbi),
 								start + 1);
+		if (force && start * sbi->blocks_per_seg >= cpc->trim_start &&
+					end * sbi->blocks_per_seg <= cpc->trim_end)
+			continue;
 
 		for (i = start; i < end; i++)
 			clear_bit(i, prefree_map);
 
 		dirty_i->nr_dirty[PRE] -= end - start;
 
-		if (force || !test_opt(sbi, DISCARD))
-			continue;
-
 		if (!test_opt(sbi, LFS) || sbi->segs_per_sec == 1) {
 			f2fs_issue_discard(sbi, START_BLOCK(sbi, start),
 				(end - start) << sbi->log_blocks_per_seg);
-- 
2.10.1


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

[PATCH 2/2] f2fs: add a case of no need to read a page in write begin

[Yunlei He]

If the range we write cover the whole valid data
in the last page, we do not need to read it.

Signed-off-by: Yunlei He <heyunlei@huawei.com>
---
 fs/f2fs/data.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index b90fb01..d9f7339 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -1715,6 +1715,9 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping,
 	if (len == PAGE_SIZE || PageUptodate(page))
 		return 0;
 
+	if (!(pos % PAGE_SHIFT) && (pos + len) > i_size_read(inode))
+		return 0;
+
 	if (blkaddr == NEW_ADDR) {
 		zero_user_segment(page, 0, PAGE_SIZE);
 		SetPageUptodate(page);
-- 
2.10.1


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

Re: [PATCH 2/2] f2fs: add a case of no need to read a page in write begin

[Jaegeuk Kim]

Hi,

On 12/16, Yunlei He wrote:
> If the range we write cover the whole valid data
> in the last page, we do not need to read it.
> 
> Signed-off-by: Yunlei He <heyunlei@huawei.com>
> ---
>  fs/f2fs/data.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
> index b90fb01..d9f7339 100644
> --- a/fs/f2fs/data.c
> +++ b/fs/f2fs/data.c
> @@ -1715,6 +1715,9 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping,
>  	if (len == PAGE_SIZE || PageUptodate(page))
>  		return 0;
>  
> +	if (!(pos % PAGE_SHIFT) && (pos + len) > i_size_read(inode))

PAGE_SHIFT?

Do you mean like this?

	if (!(pos % PAGE_MASK) && (pos + len) >= i_size_read(inode))

And, it needs to change f2fs_write_end() together.

	/* Need to change comments */
	if (!PageUptodate(page)) {
		if (unlikely(copied != len))  <== like this?
			copied = 0;
		else
			...

Thanks,

> +		return 0;
> +
>  	if (blkaddr == NEW_ADDR) {
>  		zero_user_segment(page, 0, PAGE_SIZE);
>  		SetPageUptodate(page);
> -- 
> 2.10.1
> 
> 
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most 
> engaging tech sites, SlashDot.org! http://sdm.link/slashdot
> _______________________________________________
> Linux-f2fs-devel mailing list
> Linux-f2fs-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

------------------------------------------------------------------------------
Developer Access Program for Intel Xeon Phi Processors
Access to Intel Xeon Phi processor-based developer platforms.
With one year of Intel Parallel Studio XE.
Training and support from Colfax.
Order your platform today.http://sdm.link/intel

Re: [PATCH 1/2] f2fs: fix a missing discard prefree segments

[Jaegeuk Kim]

Hi Yunlei,

On 12/16, Yunlei He wrote:
> If userspace issue a fstrim with a range not involve prefree segments,
> it will reuse these segments without discard. This patch fix it.
> 
> Signed-off-by: Yunlei He <heyunlei@huawei.com>
> ---
>  fs/f2fs/segment.c | 9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> index d7d5727..5b4468f 100644
> --- a/fs/f2fs/segment.c
> +++ b/fs/f2fs/segment.c
> @@ -899,6 +899,9 @@ void clear_prefree_segments(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>  	unsigned int secno, start_segno;
>  	bool force = (cpc->reason == CP_DISCARD);
>  
> +	if (!test_opt(sbi, DISCARD))
> +		return;

We need to clear the prefree bits, even if discard option was not set.

> +
>  	blk_start_plug(&plug);
>  
>  	mutex_lock(&dirty_i->seglist_lock);
> @@ -910,15 +913,15 @@ void clear_prefree_segments(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>  			break;
>  		end = find_next_zero_bit(prefree_map, MAIN_SEGS(sbi),
>  								start + 1);
> +		if (force && start * sbi->blocks_per_seg >= cpc->trim_start &&
> +					end * sbi->blocks_per_seg <= cpc->trim_end)
> +			continue;

ditto.

>  
>  		for (i = start; i < end; i++)
>  			clear_bit(i, prefree_map);
>  
>  		dirty_i->nr_dirty[PRE] -= end - start;
>  
> -		if (force || !test_opt(sbi, DISCARD))
> -			continue;

Why not just adding that condition here?

Thanks,

> -
>  		if (!test_opt(sbi, LFS) || sbi->segs_per_sec == 1) {
>  			f2fs_issue_discard(sbi, START_BLOCK(sbi, start),
>  				(end - start) << sbi->log_blocks_per_seg);
> -- 
> 2.10.1
> 
> 
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most 
> engaging tech sites, SlashDot.org! http://sdm.link/slashdot
> _______________________________________________
> Linux-f2fs-devel mailing list
> Linux-f2fs-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

------------------------------------------------------------------------------
Developer Access Program for Intel Xeon Phi Processors
Access to Intel Xeon Phi processor-based developer platforms.
With one year of Intel Parallel Studio XE.
Training and support from Colfax.
Order your platform today.http://sdm.link/intel