Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1355584 > unrolled thread

[PATCH 0/8] some cleanup of inline flag checking

Started byShawn Lin <shawn.lin@rock-chips.com>
First post2016-03-11 04:40 +0100
Last post2016-03-11 08:30 +0100
Articles 7 — 3 participants

Back to article view | Back to linux.kernel

Contents

  [PATCH 0/8] some cleanup of inline flag checking Shawn Lin <shawn.lin@rock-chips.com> - 2016-03-11 04:40 +0100
    [PATCH 3/8] f2fs: check inline flag ahead for f2fs_read_inline_data Shawn Lin <shawn.lin@rock-chips.com> - 2016-03-11 04:40 +0100
    [PATCH 5/8] f2fs: f2fs: check inline flag ahead for f2fs_inline_data_fiemap Shawn Lin <shawn.lin@rock-chips.com> - 2016-03-11 04:40 +0100
    [PATCH 2/8] f2fs: remove checing inline data flag for f2fs_write_data_page Shawn Lin <shawn.lin@rock-chips.com> - 2016-03-11 04:40 +0100
    RE: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking Chao Yu <chao2.yu@samsung.com> - 2016-03-11 06:40 +0100
      Re: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking Shawn Lin <shawn.lin@kernel-upstream.org> - 2016-03-11 07:40 +0100
        RE: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking Chao Yu <chao2.yu@samsung.com> - 2016-03-11 08:30 +0100

#1355584 — [PATCH 0/8] some cleanup of inline flag checking

FromShawn Lin <shawn.lin@rock-chips.com>
Date2016-03-11 04:40 +0100
Subject[PATCH 0/8] some cleanup of inline flag checking
Message-ID<rbm5I-8y-5@gated-at.bofh.it>
This patchset is going to remove some redunant checking
of inline data flag and also going to avoid some unnecessary
cpu waste when doing inline stuff.

Note:
Sorry for sending previous four patches in separate, let
drop them and make them in this thread for better review.



Shawn Lin (8):
  f2fs: check inline flag ahead for f2fs_write_inline_data
  f2fs: remove checing inline data flag for f2fs_write_data_page
  f2fs: check inline flag ahead for f2fs_read_inline_data
  f2fs: remove redundant checking of inline data flag
  f2fs: f2fs: check inline flag ahead for f2fs_inline_data_fiemap
  f2fs: remove checing inline data flag for f2fs_fiemap
  f2fs: remove unnecessary inline checking for f2fs_convert_inline_inode
  f2fs: check inline flag ahead for get_dnode_of_data

 fs/f2fs/data.c   | 17 +++++++----------
 fs/f2fs/inline.c | 27 ++++++++++-----------------
 fs/f2fs/node.c   | 12 +++++-------
 3 files changed, 22 insertions(+), 34 deletions(-)

-- 
2.3.7

[toc] | [next] | [standalone]

#1355585 — [PATCH 3/8] f2fs: check inline flag ahead for f2fs_read_inline_data

FromShawn Lin <shawn.lin@rock-chips.com>
Date2016-03-11 04:40 +0100
Subject[PATCH 3/8] f2fs: check inline flag ahead for f2fs_read_inline_data
Message-ID<rbm5I-8y-23@gated-at.bofh.it>
In reply to#1355584
No matter inline data flag is set or not, get_node_page is
going work now. But actually we can avoid it by puting the
check of inline data flag in advance to save this cpu cycle.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
---

 fs/f2fs/inline.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/fs/f2fs/inline.c b/fs/f2fs/inline.c
index 0926eab..69a4806 100644
--- a/fs/f2fs/inline.c
+++ b/fs/f2fs/inline.c
@@ -81,17 +81,15 @@ int f2fs_read_inline_data(struct inode *inode, struct page *page)
 {
 	struct page *ipage;
 
+	if (!f2fs_has_inline_data(inode))
+		return -EAGAIN;
+
 	ipage = get_node_page(F2FS_I_SB(inode), inode->i_ino);
 	if (IS_ERR(ipage)) {
 		unlock_page(page);
 		return PTR_ERR(ipage);
 	}
 
-	if (!f2fs_has_inline_data(inode)) {
-		f2fs_put_page(ipage, 1);
-		return -EAGAIN;
-	}
-
 	if (page->index)
 		zero_user_segment(page, 0, PAGE_CACHE_SIZE);
 	else
-- 
2.3.7

[toc] | [prev] | [next] | [standalone]

#1355587 — [PATCH 5/8] f2fs: f2fs: check inline flag ahead for f2fs_inline_data_fiemap

FromShawn Lin <shawn.lin@rock-chips.com>
Date2016-03-11 04:40 +0100
Subject[PATCH 5/8] f2fs: f2fs: check inline flag ahead for f2fs_inline_data_fiemap
Message-ID<rbm5J-8y-27@gated-at.bofh.it>
In reply to#1355584
No matter inline data flag is set or not, get_node_page is
going work now. But actually we can avoid it by puting the
check of inline data flag in advance to save this cpu cycle.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
---

 fs/f2fs/inline.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/fs/f2fs/inline.c b/fs/f2fs/inline.c
index 69a4806..394feee 100644
--- a/fs/f2fs/inline.c
+++ b/fs/f2fs/inline.c
@@ -566,15 +566,13 @@ int f2fs_inline_data_fiemap(struct inode *inode,
 	struct page *ipage;
 	int err = 0;
 
+	if (!f2fs_has_inline_data(inode))
+		return -EAGAIN;
+
 	ipage = get_node_page(F2FS_I_SB(inode), inode->i_ino);
 	if (IS_ERR(ipage))
 		return PTR_ERR(ipage);
 
-	if (!f2fs_has_inline_data(inode)) {
-		err = -EAGAIN;
-		goto out;
-	}
-
 	ilen = min_t(size_t, MAX_INLINE_DATA, i_size_read(inode));
 	if (start >= ilen)
 		goto out;
-- 
2.3.7

[toc] | [prev] | [next] | [standalone]

#1355588 — [PATCH 2/8] f2fs: remove checing inline data flag for f2fs_write_data_page

FromShawn Lin <shawn.lin@rock-chips.com>
Date2016-03-11 04:40 +0100
Subject[PATCH 2/8] f2fs: remove checing inline data flag for f2fs_write_data_page
Message-ID<rbm5J-8y-29@gated-at.bofh.it>
In reply to#1355584
Remove the f2fs_has_inline_data for f2fs_write_data_page,
and let f2fs_write_inline_data take over gatekeeper of
checking inline data flag.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
---

 fs/f2fs/data.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index e5c762b..38c834e 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -1192,8 +1192,7 @@ write:
 
 	err = -EAGAIN;
 	f2fs_lock_op(sbi);
-	if (f2fs_has_inline_data(inode))
-		err = f2fs_write_inline_data(inode, page);
+	err = f2fs_write_inline_data(inode, page);
 	if (err == -EAGAIN)
 		err = do_write_data_page(&fio);
 	f2fs_unlock_op(sbi);
-- 
2.3.7

[toc] | [prev] | [next] | [standalone]

#1355616 — RE: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking

FromChao Yu <chao2.yu@samsung.com>
Date2016-03-11 06:40 +0100
SubjectRE: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking
Message-ID<rbnXP-1nu-1@gated-at.bofh.it>
In reply to#1355584
Hi Shawn,

> -----Original Message-----
> From: Shawn Lin [mailto:shawn.lin@rock-chips.com]
> Sent: Friday, March 11, 2016 11:28 AM
> To: Jaegeuk Kim
> Cc: Shawn Lin; linux-kernel@vger.kernel.org; linux-f2fs-devel@lists.sourceforge.net
> Subject: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking
> 
> 
> This patchset is going to remove some redunant checking
> of inline data flag and also going to avoid some unnecessary
> cpu waste when doing inline stuff.

When we are accessing inline inode, inline inode conversion can happen
concurrently, we should check inline flag again under inode page's lock
to avoid accessing the wrong inline data which may have been converted.

Thanks,

> 
> Note:
> Sorry for sending previous four patches in separate, let
> drop them and make them in this thread for better review.
> 
> 
> 
> Shawn Lin (8):
>   f2fs: check inline flag ahead for f2fs_write_inline_data
>   f2fs: remove checing inline data flag for f2fs_write_data_page
>   f2fs: check inline flag ahead for f2fs_read_inline_data
>   f2fs: remove redundant checking of inline data flag
>   f2fs: f2fs: check inline flag ahead for f2fs_inline_data_fiemap
>   f2fs: remove checing inline data flag for f2fs_fiemap
>   f2fs: remove unnecessary inline checking for f2fs_convert_inline_inode
>   f2fs: check inline flag ahead for get_dnode_of_data
> 
>  fs/f2fs/data.c   | 17 +++++++----------
>  fs/f2fs/inline.c | 27 ++++++++++-----------------
>  fs/f2fs/node.c   | 12 +++++-------
>  3 files changed, 22 insertions(+), 34 deletions(-)
> 
> --
> 2.3.7
> 
> 
> 
> ------------------------------------------------------------------------------
> Transform Data into Opportunity.
> Accelerate data analysis in your applications with
> Intel Data Analytics Acceleration Library.
> Click to learn more.
> http://pubads.g.doubleclick.net/gampad/clk?id=278785111&iu=/4140
> _______________________________________________
> Linux-f2fs-devel mailing list
> Linux-f2fs-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

[toc] | [prev] | [next] | [standalone]

#1355626 — Re: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking

FromShawn Lin <shawn.lin@kernel-upstream.org>
Date2016-03-11 07:40 +0100
SubjectRe: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking
Message-ID<rboTT-22L-1@gated-at.bofh.it>
In reply to#1355616
Hi Chao Yu,

On 2016/3/11 13:29, Chao Yu wrote:
> Hi Shawn,
>
>> -----Original Message-----
>> From: Shawn Lin [mailto:shawn.lin@rock-chips.com]
>> Sent: Friday, March 11, 2016 11:28 AM
>> To: Jaegeuk Kim
>> Cc: Shawn Lin; linux-kernel@vger.kernel.org; linux-f2fs-devel@lists.sourceforge.net
>> Subject: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking
>>
>>
>> This patchset is going to remove some redunant checking
>> of inline data flag and also going to avoid some unnecessary
>> cpu waste when doing inline stuff.
>
> When we are accessing inline inode, inline inode conversion can happen
> concurrently, we should check inline flag again under inode page's lock
> to avoid accessing the wrong inline data which may have been converted.
>

that sounds reasonable at first glance, and it more seems like that
mopst part of this patchset is just puting the checking in the right way.

If we need to check the inline inode under the protection of inode
page's lock, it means any callers who calling inline API stuff is
wasting time on checing the flag outside the API, right?

So we can just remove the redundant checking of the caller, but not
change the behaviour of checing inline flag under page's lock?

Thanks for catching it.

> Thanks,
>
>>
>> Note:
>> Sorry for sending previous four patches in separate, let
>> drop them and make them in this thread for better review.
>>
>>
>>
>> Shawn Lin (8):
>>    f2fs: check inline flag ahead for f2fs_write_inline_data
>>    f2fs: remove checing inline data flag for f2fs_write_data_page
>>    f2fs: check inline flag ahead for f2fs_read_inline_data
>>    f2fs: remove redundant checking of inline data flag
>>    f2fs: f2fs: check inline flag ahead for f2fs_inline_data_fiemap
>>    f2fs: remove checing inline data flag for f2fs_fiemap
>>    f2fs: remove unnecessary inline checking for f2fs_convert_inline_inode
>>    f2fs: check inline flag ahead for get_dnode_of_data
>>
>>   fs/f2fs/data.c   | 17 +++++++----------
>>   fs/f2fs/inline.c | 27 ++++++++++-----------------
>>   fs/f2fs/node.c   | 12 +++++-------
>>   3 files changed, 22 insertions(+), 34 deletions(-)
>>
>> --
>> 2.3.7
>>
>>
>>
>> ------------------------------------------------------------------------------
>> Transform Data into Opportunity.
>> Accelerate data analysis in your applications with
>> Intel Data Analytics Acceleration Library.
>> Click to learn more.
>> http://pubads.g.doubleclick.net/gampad/clk?id=278785111&iu=/4140
>> _______________________________________________
>> Linux-f2fs-devel mailing list
>> Linux-f2fs-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
>
>
> ------------------------------------------------------------------------------
> Transform Data into Opportunity.
> Accelerate data analysis in your applications with
> Intel Data Analytics Acceleration Library.
> Click to learn more.
> http://pubads.g.doubleclick.net/gampad/clk?id=278785111&iu=/4140
> _______________________________________________
> Linux-f2fs-devel mailing list
> Linux-f2fs-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
>

[toc] | [prev] | [next] | [standalone]

#1355637 — RE: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking

FromChao Yu <chao2.yu@samsung.com>
Date2016-03-11 08:30 +0100
SubjectRE: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking
Message-ID<rbpGi-2IW-5@gated-at.bofh.it>
In reply to#1355626
Hi Shawn,

> -----Original Message-----
> From: Shawn Lin [mailto:shawn.lin@kernel-upstream.org]
> Sent: Friday, March 11, 2016 2:34 PM
> To: Chao Yu; 'Shawn Lin'; 'Jaegeuk Kim'
> Cc: shawn.lin@kernel-upstream.org; linux-kernel@vger.kernel.org;
> linux-f2fs-devel@lists.sourceforge.net
> Subject: Re: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking
> 
> Hi Chao Yu,
> 
> On 2016/3/11 13:29, Chao Yu wrote:
> > Hi Shawn,
> >
> >> -----Original Message-----
> >> From: Shawn Lin [mailto:shawn.lin@rock-chips.com]
> >> Sent: Friday, March 11, 2016 11:28 AM
> >> To: Jaegeuk Kim
> >> Cc: Shawn Lin; linux-kernel@vger.kernel.org; linux-f2fs-devel@lists.sourceforge.net
> >> Subject: [f2fs-dev] [PATCH 0/8] some cleanup of inline flag checking
> >>
> >>
> >> This patchset is going to remove some redunant checking
> >> of inline data flag and also going to avoid some unnecessary
> >> cpu waste when doing inline stuff.
> >
> > When we are accessing inline inode, inline inode conversion can happen
> > concurrently, we should check inline flag again under inode page's lock
> > to avoid accessing the wrong inline data which may have been converted.
> >
> 
> that sounds reasonable at first glance, and it more seems like that
> mopst part of this patchset is just puting the checking in the right way.
> 
> If we need to check the inline inode under the protection of inode
> page's lock, it means any callers who calling inline API stuff is
> wasting time on checing the flag outside the API, right?

As you know, inline conversion was designed as one-way operation, which means
inline inode can only be converted to normal inode, but can not be converted
in the opposite way. So here, with original design, it is OK to handle inode
as regular one if we detect that it is a non-inline inode, since it won't be
converted to inline one, otherwise, we should take inode page's lock and check
the flag again.

Thanks,

> 
> So we can just remove the redundant checking of the caller, but not
> change the behaviour of checing inline flag under page's lock?
> 
> Thanks for catching it.
> 
> > Thanks,
> >
> >>
> >> Note:
> >> Sorry for sending previous four patches in separate, let
> >> drop them and make them in this thread for better review.
> >>
> >>
> >>
> >> Shawn Lin (8):
> >>    f2fs: check inline flag ahead for f2fs_write_inline_data
> >>    f2fs: remove checing inline data flag for f2fs_write_data_page
> >>    f2fs: check inline flag ahead for f2fs_read_inline_data
> >>    f2fs: remove redundant checking of inline data flag
> >>    f2fs: f2fs: check inline flag ahead for f2fs_inline_data_fiemap
> >>    f2fs: remove checing inline data flag for f2fs_fiemap
> >>    f2fs: remove unnecessary inline checking for f2fs_convert_inline_inode
> >>    f2fs: check inline flag ahead for get_dnode_of_data
> >>
> >>   fs/f2fs/data.c   | 17 +++++++----------
> >>   fs/f2fs/inline.c | 27 ++++++++++-----------------
> >>   fs/f2fs/node.c   | 12 +++++-------
> >>   3 files changed, 22 insertions(+), 34 deletions(-)
> >>
> >> --
> >> 2.3.7
> >>
> >>
> >>
> >> ------------------------------------------------------------------------------
> >> Transform Data into Opportunity.
> >> Accelerate data analysis in your applications with
> >> Intel Data Analytics Acceleration Library.
> >> Click to learn more.
> >> http://pubads.g.doubleclick.net/gampad/clk?id=278785111&iu=/4140
> >> _______________________________________________
> >> Linux-f2fs-devel mailing list
> >> Linux-f2fs-devel@lists.sourceforge.net
> >> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
> >
> >
> > ------------------------------------------------------------------------------
> > Transform Data into Opportunity.
> > Accelerate data analysis in your applications with
> > Intel Data Analytics Acceleration Library.
> > Click to learn more.
> > http://pubads.g.doubleclick.net/gampad/clk?id=278785111&iu=/4140
> > _______________________________________________
> > Linux-f2fs-devel mailing list
> > Linux-f2fs-devel@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
> >

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web