2008-06-18 17:52:22

by Aneesh Kumar K.V

[permalink] [raw]
Subject: [PATCH] ext4: Fix ext4_bmap to flush the data to the disk with delalloc

With delalloc we don't do block allocation in the write_begin/write_end.
So when using bmap we first need to flush data to the disk so that blocks
get allocated and then call generic_block_bmap.

Signed-off-by: Aneesh Kumar K.V <[email protected]>
---
fs/ext4/inode.c | 11 +++++++++++
1 files changed, 11 insertions(+), 0 deletions(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 7035621..cfeb869 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1833,6 +1833,17 @@ sector_t ext4_bmap(struct address_space *mapping, sector_t block)
journal_t *journal;
int err;

+ if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
+ test_opt(inode->i_sb, DELALLOC)) {
+ /*
+ * With delalloc we want to sync the file
+ * so that we can make sure we allocate
+ * blocks for file
+ */
+ filemap_fdatawrite(mapping);
+ filemap_fdatawait(mapping);
+ }
+
if (EXT4_I(inode)->i_state & EXT4_STATE_JDATA) {
/*
* This is a REALLY heavyweight approach, but the use of
--
1.5.6.rc2.15.g457bb.dirty



2008-06-18 18:23:41

by Eric Sandeen

[permalink] [raw]
Subject: Re: [PATCH] ext4: Fix ext4_bmap to flush the data to the disk with delalloc

Aneesh Kumar K.V wrote:
> With delalloc we don't do block allocation in the write_begin/write_end.
> So when using bmap we first need to flush data to the disk so that blocks
> get allocated and then call generic_block_bmap.
>
> Signed-off-by: Aneesh Kumar K.V <[email protected]>
> ---
> fs/ext4/inode.c | 11 +++++++++++
> 1 files changed, 11 insertions(+), 0 deletions(-)
>
> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index 7035621..cfeb869 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -1833,6 +1833,17 @@ sector_t ext4_bmap(struct address_space *mapping, sector_t block)
> journal_t *journal;
> int err;
>
> + if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
> + test_opt(inode->i_sb, DELALLOC)) {
> + /*
> + * With delalloc we want to sync the file
> + * so that we can make sure we allocate
> + * blocks for file
> + */
> + filemap_fdatawrite(mapping);
> + filemap_fdatawait(mapping);
> + }

This seems fine.

I wonder, does it make any sense at all to only do the flushing if the
block we wish to map is actually in a delalloc state at the moment?

-Eric

> +
> if (EXT4_I(inode)->i_state & EXT4_STATE_JDATA) {
> /*
> * This is a REALLY heavyweight approach, but the use of


2008-06-18 18:31:07

by Aneesh Kumar K.V

[permalink] [raw]
Subject: Re: [PATCH] ext4: Fix ext4_bmap to flush the data to the disk with delalloc

On Wed, Jun 18, 2008 at 01:03:32PM -0500, Eric Sandeen wrote:
> Aneesh Kumar K.V wrote:
> > With delalloc we don't do block allocation in the write_begin/write_end.
> > So when using bmap we first need to flush data to the disk so that blocks
> > get allocated and then call generic_block_bmap.
> >
> > Signed-off-by: Aneesh Kumar K.V <[email protected]>
> > ---
> > fs/ext4/inode.c | 11 +++++++++++
> > 1 files changed, 11 insertions(+), 0 deletions(-)
> >
> > diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> > index 7035621..cfeb869 100644
> > --- a/fs/ext4/inode.c
> > +++ b/fs/ext4/inode.c
> > @@ -1833,6 +1833,17 @@ sector_t ext4_bmap(struct address_space *mapping, sector_t block)
> > journal_t *journal;
> > int err;
> >
> > + if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
> > + test_opt(inode->i_sb, DELALLOC)) {
> > + /*
> > + * With delalloc we want to sync the file
> > + * so that we can make sure we allocate
> > + * blocks for file
> > + */
> > + filemap_fdatawrite(mapping);
> > + filemap_fdatawait(mapping);
> > + }
>
> This seems fine.
>
> I wonder, does it make any sense at all to only do the flushing if the
> block we wish to map is actually in a delalloc state at the moment?
>

I am not sure it is worth the complexity considering bmap is not a
frequently used API. Initially I was even thinking should i be doing
it only for delalloc. That test seems to be a simple one to put.

-aneesh

2008-06-18 22:26:00

by Mingming Cao

[permalink] [raw]
Subject: Re: [PATCH] ext4: Fix ext4_bmap to flush the data to the disk with delalloc


On Wed, 2008-06-18 at 23:22 +0530, Aneesh Kumar K.V wrote:
> With delalloc we don't do block allocation in the write_begin/write_end.
> So when using bmap we first need to flush data to the disk so that blocks
> get allocated and then call generic_block_bmap.
>
> Signed-off-by: Aneesh Kumar K.V <[email protected]>
> ---
> fs/ext4/inode.c | 11 +++++++++++
> 1 files changed, 11 insertions(+), 0 deletions(-)
>
> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index 7035621..cfeb869 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -1833,6 +1833,17 @@ sector_t ext4_bmap(struct address_space *mapping, sector_t block)
> journal_t *journal;
> int err;
>
> + if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
> + test_opt(inode->i_sb, DELALLOC)) {
> + /*
> + * With delalloc we want to sync the file
> + * so that we can make sure we allocate
> + * blocks for file
> + */
> + filemap_fdatawrite(mapping);
> + filemap_fdatawait(mapping);
> + }
> +

I would use filemap_write_and_wait(), it is essentially the same as
above two functions, but with a little extra checking.

Mingming
> if (EXT4_I(inode)->i_state & EXT4_STATE_JDATA) {
> /*
> * This is a REALLY heavyweight approach, but the use of


2008-06-19 03:29:25

by Aneesh Kumar K.V

[permalink] [raw]
Subject: Re: [PATCH] ext4: Fix ext4_bmap to flush the data to the disk with delalloc

On Wed, Jun 18, 2008 at 03:26:25PM -0700, Mingming wrote:
>
> On Wed, 2008-06-18 at 23:22 +0530, Aneesh Kumar K.V wrote:
> > With delalloc we don't do block allocation in the write_begin/write_end.
> > So when using bmap we first need to flush data to the disk so that blocks
> > get allocated and then call generic_block_bmap.
> >
> > Signed-off-by: Aneesh Kumar K.V <[email protected]>
> > ---
> > fs/ext4/inode.c | 11 +++++++++++
> > 1 files changed, 11 insertions(+), 0 deletions(-)
> >
> > diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> > index 7035621..cfeb869 100644
> > --- a/fs/ext4/inode.c
> > +++ b/fs/ext4/inode.c
> > @@ -1833,6 +1833,17 @@ sector_t ext4_bmap(struct address_space *mapping, sector_t block)
> > journal_t *journal;
> > int err;
> >
> > + if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
> > + test_opt(inode->i_sb, DELALLOC)) {
> > + /*
> > + * With delalloc we want to sync the file
> > + * so that we can make sure we allocate
> > + * blocks for file
> > + */
> > + filemap_fdatawrite(mapping);
> > + filemap_fdatawait(mapping);
> > + }
> > +
>
> I would use filemap_write_and_wait(), it is essentially the same as
> above two functions, but with a little extra checking.
>

Updated patch below

ext4: Fix ext4_bmap to flush the data to the disk with delalloc

From: Aneesh Kumar K.V <[email protected]>

With delalloc we don't do block allocation in the write_begin/write_end.
So when using bmap we first need to flush data to the disk so that blocks
get allocated and then call generic_block_bmap.

Signed-off-by: Aneesh Kumar K.V <[email protected]>
---

fs/ext4/inode.c | 10 ++++++++++
1 files changed, 10 insertions(+), 0 deletions(-)


diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 7035621..f1bd194 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1833,6 +1833,16 @@ sector_t ext4_bmap(struct address_space *mapping, sector_t block)
journal_t *journal;
int err;

+ if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
+ test_opt(inode->i_sb, DELALLOC)) {
+ /*
+ * With delalloc we want to sync the file
+ * so that we can make sure we allocate
+ * blocks for file
+ */
+ filemap_write_and_wait(mapping);
+ }
+
if (EXT4_I(inode)->i_state & EXT4_STATE_JDATA) {
/*
* This is a REALLY heavyweight approach, but the use of