RE: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Chao Yu]

Hi Jaegeuk,

> -Original Message-
> From: Chao Yu [mailto:chao2...@samsung.com]
> Sent: Wednesday, December 23, 2015 5:47 PM
> To: 'Jaegeuk Kim'
> Cc: linux-fsde...@vger.kernel.org; linux-kernel@vger.kernel.org;
> linux-f2fs-de...@lists.sourceforge.net
> Subject: Re: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node 
> was changed
> 
> Hi Jaegeuk,
> 
> > -Original Message-
> > From: Jaegeuk Kim [mailto:jaeg...@kernel.org]
> > Sent: Wednesday, December 23, 2015 9:00 AM
> > To: linux-kernel@vger.kernel.org; linux-fsde...@vger.kernel.org;
> > linux-f2fs-de...@lists.sourceforge.net
> > Cc: Jaegeuk Kim
> > Subject: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node 
> > was changed
> >
> > If user tries to update or read data, we don't need to call f2fs_balance_fs
> > which triggers f2fs_gc, which increases unnecessary long latency.
> 
> One missing case is get_data_block_dio, how about also covering it based on
> following patch?
> 
> 
> >From 7175efac7473e7a04285055c69edfb7432f8ca4e Mon Sep 17 00:00:00 2001
> From: Chao Yu 
> Date: Wed, 23 Dec 2015 17:11:43 +0800
> Subject: [PATCH] f2fs: reduce covered region of sbi->cp_rwsem in
>  f2fs_map_blocks

Title of this commit log was not merged into dev-test branch correctly,
could you please reedit it? :)

Thanks,


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

RE: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Chao Yu]

Hi Jaegeuk,

> -Original Message-
> From: Chao Yu [mailto:chao2...@samsung.com]
> Sent: Wednesday, December 23, 2015 5:47 PM
> To: 'Jaegeuk Kim'
> Cc: linux-fsde...@vger.kernel.org; linux-kernel@vger.kernel.org;
> linux-f2fs-de...@lists.sourceforge.net
> Subject: Re: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node 
> was changed
> 
> Hi Jaegeuk,
> 
> > -Original Message-
> > From: Jaegeuk Kim [mailto:jaeg...@kernel.org]
> > Sent: Wednesday, December 23, 2015 9:00 AM
> > To: linux-kernel@vger.kernel.org; linux-fsde...@vger.kernel.org;
> > linux-f2fs-de...@lists.sourceforge.net
> > Cc: Jaegeuk Kim
> > Subject: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node 
> > was changed
> >
> > If user tries to update or read data, we don't need to call f2fs_balance_fs
> > which triggers f2fs_gc, which increases unnecessary long latency.
> 
> One missing case is get_data_block_dio, how about also covering it based on
> following patch?
> 
> 
> >From 7175efac7473e7a04285055c69edfb7432f8ca4e Mon Sep 17 00:00:00 2001
> From: Chao Yu <chao2...@samsung.com>
> Date: Wed, 23 Dec 2015 17:11:43 +0800
> Subject: [PATCH] f2fs: reduce covered region of sbi->cp_rwsem in
>  f2fs_map_blocks

Title of this commit log was not merged into dev-test branch correctly,
could you please reedit it? :)

Thanks,


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

RE: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Chao Yu]

> -Original Message-
> From: Jaegeuk Kim [mailto:jaeg...@kernel.org]
> Sent: Thursday, December 24, 2015 3:14 AM
> To: linux-kernel@vger.kernel.org; linux-fsde...@vger.kernel.org;
> linux-f2fs-de...@lists.sourceforge.net
> Subject: Re: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node 
> was changed
> 
> Change log v2:
>  - add dio case
> 
> >From c2d16a526371954671f9c8cff5f09f9d230f7993 Mon Sep 17 00:00:00 2001
> From: Jaegeuk Kim 
> Date: Tue, 22 Dec 2015 13:23:35 -0800
> Subject: [PATCH] f2fs: call f2fs_balance_fs only when node was changed
> 
> If user tries to update or read data, we don't need to call f2fs_balance_fs
> which triggers f2fs_gc, which increases unnecessary long latency.
> 
> Signed-off-by: Jaegeuk Kim 

Reviewed-by: Chao Yu 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Jaegeuk Kim]

Change log v2:
 - add dio case

>From c2d16a526371954671f9c8cff5f09f9d230f7993 Mon Sep 17 00:00:00 2001
From: Jaegeuk Kim 
Date: Tue, 22 Dec 2015 13:23:35 -0800
Subject: [PATCH] f2fs: call f2fs_balance_fs only when node was changed

If user tries to update or read data, we don't need to call f2fs_balance_fs
which triggers f2fs_gc, which increases unnecessary long latency.

Signed-off-by: Jaegeuk Kim 
---
 fs/f2fs/data.c   | 26 ++
 fs/f2fs/file.c   | 26 +-
 fs/f2fs/inline.c |  4 
 3 files changed, 35 insertions(+), 21 deletions(-)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 82ecaa30..958d826 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -509,7 +509,6 @@ static void __allocate_data_blocks(struct inode *inode, 
loff_t offset,
u64 end_offset;
 
while (len) {
-   f2fs_balance_fs(sbi);
f2fs_lock_op(sbi);
 
/* When reading holes, we need its node page */
@@ -542,6 +541,9 @@ static void __allocate_data_blocks(struct inode *inode, 
loff_t offset,
 
f2fs_put_dnode();
f2fs_unlock_op(sbi);
+
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
}
return;
 
@@ -551,6 +553,8 @@ sync_out:
f2fs_put_dnode();
 out:
f2fs_unlock_op(sbi);
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
return;
 }
 
@@ -649,6 +653,8 @@ get_next:
 
if (create) {
f2fs_unlock_op(sbi);
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
f2fs_lock_op(sbi);
}
 
@@ -706,8 +712,11 @@ sync_out:
 put_out:
f2fs_put_dnode();
 unlock_out:
-   if (create)
+   if (create) {
f2fs_unlock_op(sbi);
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
+   }
 out:
trace_f2fs_map_blocks(inode, map, err);
return err;
@@ -1415,8 +1424,6 @@ static int f2fs_write_begin(struct file *file, struct 
address_space *mapping,
 
trace_f2fs_write_begin(inode, pos, len, flags);
 
-   f2fs_balance_fs(sbi);
-
/*
 * We should check this at this moment to avoid deadlock on inode page
 * and #0 page. The locking rule for inline_data conversion should be:
@@ -1466,6 +1473,17 @@ put_next:
f2fs_put_dnode();
f2fs_unlock_op(sbi);
 
+   if (dn.node_changed && has_not_enough_free_secs(sbi, 0)) {
+   unlock_page(page);
+   f2fs_balance_fs(sbi);
+   lock_page(page);
+   if (page->mapping != mapping) {
+   /* The page got truncated from under us */
+   f2fs_put_page(page, 1);
+   goto repeat;
+   }
+   }
+
f2fs_wait_on_page_writeback(page, DATA);
 
/* wait for GCed encrypted page writeback */
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index f2effe1..888ce47 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -40,8 +40,6 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
struct dnode_of_data dn;
int err;
 
-   f2fs_balance_fs(sbi);
-
sb_start_pagefault(inode->i_sb);
 
f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
@@ -57,6 +55,9 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
f2fs_put_dnode();
f2fs_unlock_op(sbi);
 
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
+
file_update_time(vma->vm_file);
lock_page(page);
if (unlikely(page->mapping != inode->i_mapping ||
@@ -233,9 +234,6 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t 
end, int datasync)
goto out;
}
 go_write:
-   /* guarantee free sections for fsync */
-   f2fs_balance_fs(sbi);
-
/*
 * Both of fdatasync() and fsync() are able to be recovered from
 * sudden-power-off.
@@ -267,6 +265,8 @@ sync_nodes:
if (need_inode_block_update(sbi, ino)) {
mark_inode_dirty_sync(inode);
f2fs_write_inode(inode, NULL);
+
+   f2fs_balance_fs(sbi);
goto sync_nodes;
}
 
@@ -945,8 +945,6 @@ static int f2fs_collapse_range(struct inode *inode, loff_t 
offset, loff_t len)
if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
return -EINVAL;
 
-   f2fs_balance_fs(F2FS_I_SB(inode));
-
ret = f2fs_convert_inline_inode(inode);
if (ret)
return ret;
@@ -993,8 +991,6 @@ static int f2fs_zero_range(struct inode *inode, loff_t 
offset, loff_t len,
if (ret)
return ret;
 
-   f2fs_balance_fs(sbi);
-
ret = f2fs_convert_inline_inode(inode);
if (ret)
return ret;
@@ -1104,12 +1100,12 @@ static int f2fs_insert_range(struct inode *inode, 
loff_t offset, 

Re: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Jaegeuk Kim]

Hi Chao,

On Wed, Dec 23, 2015 at 05:46:59PM +0800, Chao Yu wrote:
> Hi Jaegeuk,
> 
> > -Original Message-
> > From: Jaegeuk Kim [mailto:jaeg...@kernel.org]
> > Sent: Wednesday, December 23, 2015 9:00 AM
> > To: linux-kernel@vger.kernel.org; linux-fsde...@vger.kernel.org;
> > linux-f2fs-de...@lists.sourceforge.net
> > Cc: Jaegeuk Kim
> > Subject: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node 
> > was changed
> > 
> > If user tries to update or read data, we don't need to call f2fs_balance_fs
> > which triggers f2fs_gc, which increases unnecessary long latency.
> 
> One missing case is get_data_block_dio, how about also covering it based on
> following patch?

It makes sense.
I'll submit v2.

Thanks,

> 
> 
> >From 7175efac7473e7a04285055c69edfb7432f8ca4e Mon Sep 17 00:00:00 2001
> From: Chao Yu 
> Date: Wed, 23 Dec 2015 17:11:43 +0800
> Subject: [PATCH] f2fs: reduce covered region of sbi->cp_rwsem in
>  f2fs_map_blocks
> 
> Only cover sbi->cp_rwsem on one dnode page's allocation and modification
> instead of multiple's in f2fs_map_blocks, it can reduce the covered region
> of cp_rwsem, then we can avoid potential long time delay for concurrent
> checkpointer.
> 
> Signed-off-by: Chao Yu 
> ---
>  fs/f2fs/data.c | 9 +++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
> index 8f8f8b0..3c83b16 100644
> --- a/fs/f2fs/data.c
> +++ b/fs/f2fs/data.c
> @@ -594,7 +594,7 @@ int f2fs_map_blocks(struct inode *inode, struct 
> f2fs_map_blocks *map,
>   }
>  
>   if (create)
> - f2fs_lock_op(F2FS_I_SB(inode));
> + f2fs_lock_op(sbi);
>  
>   /* When reading holes, we need its node page */
>   set_new_dnode(, inode, NULL, NULL, 0);
> @@ -651,6 +651,11 @@ get_next:
>   allocated = false;
>   f2fs_put_dnode();
>  
> + if (create) {
> + f2fs_unlock_op(sbi);
> + f2fs_lock_op(sbi);
> + }
> +
>   set_new_dnode(, inode, NULL, NULL, 0);
>   err = get_dnode_of_data(, pgofs, mode);
>   if (err) {
> @@ -706,7 +711,7 @@ put_out:
>   f2fs_put_dnode();
>  unlock_out:
>   if (create)
> - f2fs_unlock_op(F2FS_I_SB(inode));
> + f2fs_unlock_op(sbi);
>  out:
>   trace_f2fs_map_blocks(inode, map, err);
>   return err;
> -- 
> 2.6.3
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

RE: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Chao Yu]

Hi Jaegeuk,

> -Original Message-
> From: Jaegeuk Kim [mailto:jaeg...@kernel.org]
> Sent: Wednesday, December 23, 2015 9:00 AM
> To: linux-kernel@vger.kernel.org; linux-fsde...@vger.kernel.org;
> linux-f2fs-de...@lists.sourceforge.net
> Cc: Jaegeuk Kim
> Subject: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was 
> changed
> 
> If user tries to update or read data, we don't need to call f2fs_balance_fs
> which triggers f2fs_gc, which increases unnecessary long latency.

One missing case is get_data_block_dio, how about also covering it based on
following patch?


>From 7175efac7473e7a04285055c69edfb7432f8ca4e Mon Sep 17 00:00:00 2001
From: Chao Yu 
Date: Wed, 23 Dec 2015 17:11:43 +0800
Subject: [PATCH] f2fs: reduce covered region of sbi->cp_rwsem in
 f2fs_map_blocks

Only cover sbi->cp_rwsem on one dnode page's allocation and modification
instead of multiple's in f2fs_map_blocks, it can reduce the covered region
of cp_rwsem, then we can avoid potential long time delay for concurrent
checkpointer.

Signed-off-by: Chao Yu 
---
 fs/f2fs/data.c | 9 +++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 8f8f8b0..3c83b16 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -594,7 +594,7 @@ int f2fs_map_blocks(struct inode *inode, struct 
f2fs_map_blocks *map,
}
 
if (create)
-   f2fs_lock_op(F2FS_I_SB(inode));
+   f2fs_lock_op(sbi);
 
/* When reading holes, we need its node page */
set_new_dnode(, inode, NULL, NULL, 0);
@@ -651,6 +651,11 @@ get_next:
allocated = false;
f2fs_put_dnode();
 
+   if (create) {
+   f2fs_unlock_op(sbi);
+   f2fs_lock_op(sbi);
+   }
+
set_new_dnode(, inode, NULL, NULL, 0);
err = get_dnode_of_data(, pgofs, mode);
if (err) {
@@ -706,7 +711,7 @@ put_out:
f2fs_put_dnode();
 unlock_out:
if (create)
-   f2fs_unlock_op(F2FS_I_SB(inode));
+   f2fs_unlock_op(sbi);
 out:
trace_f2fs_map_blocks(inode, map, err);
return err;
-- 
2.6.3


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

RE: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Chao Yu]

Hi Jaegeuk,

> -Original Message-
> From: Jaegeuk Kim [mailto:jaeg...@kernel.org]
> Sent: Wednesday, December 23, 2015 9:00 AM
> To: linux-kernel@vger.kernel.org; linux-fsde...@vger.kernel.org;
> linux-f2fs-de...@lists.sourceforge.net
> Cc: Jaegeuk Kim
> Subject: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was 
> changed
> 
> If user tries to update or read data, we don't need to call f2fs_balance_fs
> which triggers f2fs_gc, which increases unnecessary long latency.

One missing case is get_data_block_dio, how about also covering it based on
following patch?


>From 7175efac7473e7a04285055c69edfb7432f8ca4e Mon Sep 17 00:00:00 2001
From: Chao Yu <chao2...@samsung.com>
Date: Wed, 23 Dec 2015 17:11:43 +0800
Subject: [PATCH] f2fs: reduce covered region of sbi->cp_rwsem in
 f2fs_map_blocks

Only cover sbi->cp_rwsem on one dnode page's allocation and modification
instead of multiple's in f2fs_map_blocks, it can reduce the covered region
of cp_rwsem, then we can avoid potential long time delay for concurrent
checkpointer.

Signed-off-by: Chao Yu <chao2...@samsung.com>
---
 fs/f2fs/data.c | 9 +++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 8f8f8b0..3c83b16 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -594,7 +594,7 @@ int f2fs_map_blocks(struct inode *inode, struct 
f2fs_map_blocks *map,
}
 
if (create)
-   f2fs_lock_op(F2FS_I_SB(inode));
+   f2fs_lock_op(sbi);
 
/* When reading holes, we need its node page */
set_new_dnode(, inode, NULL, NULL, 0);
@@ -651,6 +651,11 @@ get_next:
allocated = false;
f2fs_put_dnode();
 
+   if (create) {
+   f2fs_unlock_op(sbi);
+   f2fs_lock_op(sbi);
+   }
+
set_new_dnode(, inode, NULL, NULL, 0);
err = get_dnode_of_data(, pgofs, mode);
if (err) {
@@ -706,7 +711,7 @@ put_out:
f2fs_put_dnode();
 unlock_out:
if (create)
-   f2fs_unlock_op(F2FS_I_SB(inode));
+   f2fs_unlock_op(sbi);
 out:
trace_f2fs_map_blocks(inode, map, err);
return err;
-- 
2.6.3


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Jaegeuk Kim]

Hi Chao,

On Wed, Dec 23, 2015 at 05:46:59PM +0800, Chao Yu wrote:
> Hi Jaegeuk,
> 
> > -Original Message-
> > From: Jaegeuk Kim [mailto:jaeg...@kernel.org]
> > Sent: Wednesday, December 23, 2015 9:00 AM
> > To: linux-kernel@vger.kernel.org; linux-fsde...@vger.kernel.org;
> > linux-f2fs-de...@lists.sourceforge.net
> > Cc: Jaegeuk Kim
> > Subject: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node 
> > was changed
> > 
> > If user tries to update or read data, we don't need to call f2fs_balance_fs
> > which triggers f2fs_gc, which increases unnecessary long latency.
> 
> One missing case is get_data_block_dio, how about also covering it based on
> following patch?

It makes sense.
I'll submit v2.

Thanks,

> 
> 
> >From 7175efac7473e7a04285055c69edfb7432f8ca4e Mon Sep 17 00:00:00 2001
> From: Chao Yu <chao2...@samsung.com>
> Date: Wed, 23 Dec 2015 17:11:43 +0800
> Subject: [PATCH] f2fs: reduce covered region of sbi->cp_rwsem in
>  f2fs_map_blocks
> 
> Only cover sbi->cp_rwsem on one dnode page's allocation and modification
> instead of multiple's in f2fs_map_blocks, it can reduce the covered region
> of cp_rwsem, then we can avoid potential long time delay for concurrent
> checkpointer.
> 
> Signed-off-by: Chao Yu <chao2...@samsung.com>
> ---
>  fs/f2fs/data.c | 9 +++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
> index 8f8f8b0..3c83b16 100644
> --- a/fs/f2fs/data.c
> +++ b/fs/f2fs/data.c
> @@ -594,7 +594,7 @@ int f2fs_map_blocks(struct inode *inode, struct 
> f2fs_map_blocks *map,
>   }
>  
>   if (create)
> - f2fs_lock_op(F2FS_I_SB(inode));
> + f2fs_lock_op(sbi);
>  
>   /* When reading holes, we need its node page */
>   set_new_dnode(, inode, NULL, NULL, 0);
> @@ -651,6 +651,11 @@ get_next:
>   allocated = false;
>   f2fs_put_dnode();
>  
> + if (create) {
> + f2fs_unlock_op(sbi);
> + f2fs_lock_op(sbi);
> + }
> +
>   set_new_dnode(, inode, NULL, NULL, 0);
>   err = get_dnode_of_data(, pgofs, mode);
>   if (err) {
> @@ -706,7 +711,7 @@ put_out:
>   f2fs_put_dnode();
>  unlock_out:
>   if (create)
> - f2fs_unlock_op(F2FS_I_SB(inode));
> + f2fs_unlock_op(sbi);
>  out:
>   trace_f2fs_map_blocks(inode, map, err);
>   return err;
> -- 
> 2.6.3
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Jaegeuk Kim]

Change log v2:
 - add dio case

>From c2d16a526371954671f9c8cff5f09f9d230f7993 Mon Sep 17 00:00:00 2001
From: Jaegeuk Kim 
Date: Tue, 22 Dec 2015 13:23:35 -0800
Subject: [PATCH] f2fs: call f2fs_balance_fs only when node was changed

If user tries to update or read data, we don't need to call f2fs_balance_fs
which triggers f2fs_gc, which increases unnecessary long latency.

Signed-off-by: Jaegeuk Kim 
---
 fs/f2fs/data.c   | 26 ++
 fs/f2fs/file.c   | 26 +-
 fs/f2fs/inline.c |  4 
 3 files changed, 35 insertions(+), 21 deletions(-)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 82ecaa30..958d826 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -509,7 +509,6 @@ static void __allocate_data_blocks(struct inode *inode, 
loff_t offset,
u64 end_offset;
 
while (len) {
-   f2fs_balance_fs(sbi);
f2fs_lock_op(sbi);
 
/* When reading holes, we need its node page */
@@ -542,6 +541,9 @@ static void __allocate_data_blocks(struct inode *inode, 
loff_t offset,
 
f2fs_put_dnode();
f2fs_unlock_op(sbi);
+
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
}
return;
 
@@ -551,6 +553,8 @@ sync_out:
f2fs_put_dnode();
 out:
f2fs_unlock_op(sbi);
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
return;
 }
 
@@ -649,6 +653,8 @@ get_next:
 
if (create) {
f2fs_unlock_op(sbi);
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
f2fs_lock_op(sbi);
}
 
@@ -706,8 +712,11 @@ sync_out:
 put_out:
f2fs_put_dnode();
 unlock_out:
-   if (create)
+   if (create) {
f2fs_unlock_op(sbi);
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
+   }
 out:
trace_f2fs_map_blocks(inode, map, err);
return err;
@@ -1415,8 +1424,6 @@ static int f2fs_write_begin(struct file *file, struct 
address_space *mapping,
 
trace_f2fs_write_begin(inode, pos, len, flags);
 
-   f2fs_balance_fs(sbi);
-
/*
 * We should check this at this moment to avoid deadlock on inode page
 * and #0 page. The locking rule for inline_data conversion should be:
@@ -1466,6 +1473,17 @@ put_next:
f2fs_put_dnode();
f2fs_unlock_op(sbi);
 
+   if (dn.node_changed && has_not_enough_free_secs(sbi, 0)) {
+   unlock_page(page);
+   f2fs_balance_fs(sbi);
+   lock_page(page);
+   if (page->mapping != mapping) {
+   /* The page got truncated from under us */
+   f2fs_put_page(page, 1);
+   goto repeat;
+   }
+   }
+
f2fs_wait_on_page_writeback(page, DATA);
 
/* wait for GCed encrypted page writeback */
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index f2effe1..888ce47 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -40,8 +40,6 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
struct dnode_of_data dn;
int err;
 
-   f2fs_balance_fs(sbi);
-
sb_start_pagefault(inode->i_sb);
 
f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
@@ -57,6 +55,9 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
f2fs_put_dnode();
f2fs_unlock_op(sbi);
 
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
+
file_update_time(vma->vm_file);
lock_page(page);
if (unlikely(page->mapping != inode->i_mapping ||
@@ -233,9 +234,6 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t 
end, int datasync)
goto out;
}
 go_write:
-   /* guarantee free sections for fsync */
-   f2fs_balance_fs(sbi);
-
/*
 * Both of fdatasync() and fsync() are able to be recovered from
 * sudden-power-off.
@@ -267,6 +265,8 @@ sync_nodes:
if (need_inode_block_update(sbi, ino)) {
mark_inode_dirty_sync(inode);
f2fs_write_inode(inode, NULL);
+
+   f2fs_balance_fs(sbi);
goto sync_nodes;
}
 
@@ -945,8 +945,6 @@ static int f2fs_collapse_range(struct inode *inode, loff_t 
offset, loff_t len)
if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
return -EINVAL;
 
-   f2fs_balance_fs(F2FS_I_SB(inode));
-
ret = f2fs_convert_inline_inode(inode);
if (ret)
return ret;
@@ -993,8 +991,6 @@ static int f2fs_zero_range(struct inode *inode, loff_t 
offset, loff_t len,
if (ret)
return ret;
 
-   f2fs_balance_fs(sbi);
-
ret = f2fs_convert_inline_inode(inode);
if (ret)
return ret;
@@ -1104,12 +1100,12 @@ static int 

RE: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Chao Yu]

> -Original Message-
> From: Jaegeuk Kim [mailto:jaeg...@kernel.org]
> Sent: Thursday, December 24, 2015 3:14 AM
> To: linux-kernel@vger.kernel.org; linux-fsde...@vger.kernel.org;
> linux-f2fs-de...@lists.sourceforge.net
> Subject: Re: [f2fs-dev] [PATCH 4/4] f2fs: call f2fs_balance_fs only when node 
> was changed
> 
> Change log v2:
>  - add dio case
> 
> >From c2d16a526371954671f9c8cff5f09f9d230f7993 Mon Sep 17 00:00:00 2001
> From: Jaegeuk Kim <jaeg...@kernel.org>
> Date: Tue, 22 Dec 2015 13:23:35 -0800
> Subject: [PATCH] f2fs: call f2fs_balance_fs only when node was changed
> 
> If user tries to update or read data, we don't need to call f2fs_balance_fs
> which triggers f2fs_gc, which increases unnecessary long latency.
> 
> Signed-off-by: Jaegeuk Kim <jaeg...@kernel.org>

Reviewed-by: Chao Yu <chao2...@samsung.com>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Jaegeuk Kim]

If user tries to update or read data, we don't need to call f2fs_balance_fs
which triggers f2fs_gc, which increases unnecessary long latency.

Signed-off-by: Jaegeuk Kim 
---
 fs/f2fs/data.c   | 19 ---
 fs/f2fs/file.c   | 26 +-
 fs/f2fs/inline.c |  4 
 3 files changed, 29 insertions(+), 20 deletions(-)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index a7a9a05..8f8f8b0 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -509,7 +509,6 @@ static void __allocate_data_blocks(struct inode *inode, 
loff_t offset,
u64 end_offset;
 
while (len) {
-   f2fs_balance_fs(sbi);
f2fs_lock_op(sbi);
 
/* When reading holes, we need its node page */
@@ -542,6 +541,9 @@ static void __allocate_data_blocks(struct inode *inode, 
loff_t offset,
 
f2fs_put_dnode();
f2fs_unlock_op(sbi);
+
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
}
return;
 
@@ -551,6 +553,8 @@ sync_out:
f2fs_put_dnode();
 out:
f2fs_unlock_op(sbi);
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
return;
 }
 
@@ -1410,8 +1414,6 @@ static int f2fs_write_begin(struct file *file, struct 
address_space *mapping,
 
trace_f2fs_write_begin(inode, pos, len, flags);
 
-   f2fs_balance_fs(sbi);
-
/*
 * We should check this at this moment to avoid deadlock on inode page
 * and #0 page. The locking rule for inline_data conversion should be:
@@ -1461,6 +1463,17 @@ put_next:
f2fs_put_dnode();
f2fs_unlock_op(sbi);
 
+   if (dn.node_changed && has_not_enough_free_secs(sbi, 0)) {
+   unlock_page(page);
+   f2fs_balance_fs(sbi);
+   lock_page(page);
+   if (page->mapping != mapping) {
+   /* The page got truncated from under us */
+   f2fs_put_page(page, 1);
+   goto repeat;
+   }
+   }
+
f2fs_wait_on_page_writeback(page, DATA);
 
/* wait for GCed encrypted page writeback */
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 10ed357..dbc08bb 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -40,8 +40,6 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
struct dnode_of_data dn;
int err;
 
-   f2fs_balance_fs(sbi);
-
sb_start_pagefault(inode->i_sb);
 
f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
@@ -57,6 +55,9 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
f2fs_put_dnode();
f2fs_unlock_op(sbi);
 
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
+
file_update_time(vma->vm_file);
lock_page(page);
if (unlikely(page->mapping != inode->i_mapping ||
@@ -233,9 +234,6 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t 
end, int datasync)
goto out;
}
 go_write:
-   /* guarantee free sections for fsync */
-   f2fs_balance_fs(sbi);
-
/*
 * Both of fdatasync() and fsync() are able to be recovered from
 * sudden-power-off.
@@ -267,6 +265,8 @@ sync_nodes:
if (need_inode_block_update(sbi, ino)) {
mark_inode_dirty_sync(inode);
f2fs_write_inode(inode, NULL);
+
+   f2fs_balance_fs(sbi);
goto sync_nodes;
}
 
@@ -946,8 +946,6 @@ static int f2fs_collapse_range(struct inode *inode, loff_t 
offset, loff_t len)
if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
return -EINVAL;
 
-   f2fs_balance_fs(F2FS_I_SB(inode));
-
ret = f2fs_convert_inline_inode(inode);
if (ret)
return ret;
@@ -994,8 +992,6 @@ static int f2fs_zero_range(struct inode *inode, loff_t 
offset, loff_t len,
if (ret)
return ret;
 
-   f2fs_balance_fs(sbi);
-
ret = f2fs_convert_inline_inode(inode);
if (ret)
return ret;
@@ -1105,12 +1101,12 @@ static int f2fs_insert_range(struct inode *inode, 
loff_t offset, loff_t len)
if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
return -EINVAL;
 
-   f2fs_balance_fs(sbi);
-
ret = f2fs_convert_inline_inode(inode);
if (ret)
return ret;
 
+   f2fs_balance_fs(sbi);
+
ret = truncate_blocks(inode, i_size_read(inode), true);
if (ret)
return ret;
@@ -1153,8 +1149,6 @@ static int expand_inode_data(struct inode *inode, loff_t 
offset,
loff_t off_start, off_end;
int ret = 0;
 
-   f2fs_balance_fs(sbi);
-
ret = inode_newsize_ok(inode, (len + offset));
if (ret)
return ret;
@@ -1163,6 +1157,8 @@ static int expand_inode_data(struct inode *inode, loff_t 
offset,
if (ret)
return ret;
 
+   f2fs_balance_fs(sbi);
+

[PATCH 4/4] f2fs: call f2fs_balance_fs only when node was changed

[Jaegeuk Kim]

If user tries to update or read data, we don't need to call f2fs_balance_fs
which triggers f2fs_gc, which increases unnecessary long latency.

Signed-off-by: Jaegeuk Kim 
---
 fs/f2fs/data.c   | 19 ---
 fs/f2fs/file.c   | 26 +-
 fs/f2fs/inline.c |  4 
 3 files changed, 29 insertions(+), 20 deletions(-)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index a7a9a05..8f8f8b0 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -509,7 +509,6 @@ static void __allocate_data_blocks(struct inode *inode, 
loff_t offset,
u64 end_offset;
 
while (len) {
-   f2fs_balance_fs(sbi);
f2fs_lock_op(sbi);
 
/* When reading holes, we need its node page */
@@ -542,6 +541,9 @@ static void __allocate_data_blocks(struct inode *inode, 
loff_t offset,
 
f2fs_put_dnode();
f2fs_unlock_op(sbi);
+
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
}
return;
 
@@ -551,6 +553,8 @@ sync_out:
f2fs_put_dnode();
 out:
f2fs_unlock_op(sbi);
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
return;
 }
 
@@ -1410,8 +1414,6 @@ static int f2fs_write_begin(struct file *file, struct 
address_space *mapping,
 
trace_f2fs_write_begin(inode, pos, len, flags);
 
-   f2fs_balance_fs(sbi);
-
/*
 * We should check this at this moment to avoid deadlock on inode page
 * and #0 page. The locking rule for inline_data conversion should be:
@@ -1461,6 +1463,17 @@ put_next:
f2fs_put_dnode();
f2fs_unlock_op(sbi);
 
+   if (dn.node_changed && has_not_enough_free_secs(sbi, 0)) {
+   unlock_page(page);
+   f2fs_balance_fs(sbi);
+   lock_page(page);
+   if (page->mapping != mapping) {
+   /* The page got truncated from under us */
+   f2fs_put_page(page, 1);
+   goto repeat;
+   }
+   }
+
f2fs_wait_on_page_writeback(page, DATA);
 
/* wait for GCed encrypted page writeback */
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 10ed357..dbc08bb 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -40,8 +40,6 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
struct dnode_of_data dn;
int err;
 
-   f2fs_balance_fs(sbi);
-
sb_start_pagefault(inode->i_sb);
 
f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
@@ -57,6 +55,9 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
f2fs_put_dnode();
f2fs_unlock_op(sbi);
 
+   if (dn.node_changed)
+   f2fs_balance_fs(sbi);
+
file_update_time(vma->vm_file);
lock_page(page);
if (unlikely(page->mapping != inode->i_mapping ||
@@ -233,9 +234,6 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t 
end, int datasync)
goto out;
}
 go_write:
-   /* guarantee free sections for fsync */
-   f2fs_balance_fs(sbi);
-
/*
 * Both of fdatasync() and fsync() are able to be recovered from
 * sudden-power-off.
@@ -267,6 +265,8 @@ sync_nodes:
if (need_inode_block_update(sbi, ino)) {
mark_inode_dirty_sync(inode);
f2fs_write_inode(inode, NULL);
+
+   f2fs_balance_fs(sbi);
goto sync_nodes;
}
 
@@ -946,8 +946,6 @@ static int f2fs_collapse_range(struct inode *inode, loff_t 
offset, loff_t len)
if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
return -EINVAL;
 
-   f2fs_balance_fs(F2FS_I_SB(inode));
-
ret = f2fs_convert_inline_inode(inode);
if (ret)
return ret;
@@ -994,8 +992,6 @@ static int f2fs_zero_range(struct inode *inode, loff_t 
offset, loff_t len,
if (ret)
return ret;
 
-   f2fs_balance_fs(sbi);
-
ret = f2fs_convert_inline_inode(inode);
if (ret)
return ret;
@@ -1105,12 +1101,12 @@ static int f2fs_insert_range(struct inode *inode, 
loff_t offset, loff_t len)
if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
return -EINVAL;
 
-   f2fs_balance_fs(sbi);
-
ret = f2fs_convert_inline_inode(inode);
if (ret)
return ret;
 
+   f2fs_balance_fs(sbi);
+
ret = truncate_blocks(inode, i_size_read(inode), true);
if (ret)
return ret;
@@ -1153,8 +1149,6 @@ static int expand_inode_data(struct inode *inode, loff_t 
offset,
loff_t off_start, off_end;
int ret = 0;
 
-   f2fs_balance_fs(sbi);
-
ret = inode_newsize_ok(inode, (len + offset));
if (ret)
return ret;
@@ -1163,6 +1157,8 @@ static int expand_inode_data(struct inode *inode, loff_t 
offset,
if (ret)
return ret;
 
+