2008-03-21 06:19:07

by Aneesh Kumar K.V

[permalink] [raw]
Subject: [RFC PATCH] ext4: Update i_diskzie during writepage for delayed allocation.

We need to update i_disksize after allocating blocks. writepages
already does this. Update writepage to also update i_disksize.

Signed-off-by: Aneesh Kumar K.V <[email protected]>
---
fs/ext4/inode.c | 40 +++++++++++++++++++++++++++++++++++++++-
1 files changed, 39 insertions(+), 1 deletions(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 79930df..50d700f 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1488,6 +1488,44 @@ out:

return ret;
}
+/* FIXME!! only support data=writeback mode */
+static int ext4_da_writepage(struct page *page,
+ struct writeback_control *wbc)
+{
+ struct inode *inode = page->mapping->host;
+ handle_t *handle = NULL;
+ int ret = 0;
+ int err;
+
+ if (ext4_journal_current_handle())
+ goto out_fail;
+
+ handle = ext4_journal_start(inode, ext4_writepage_trans_blocks(inode));
+ if (IS_ERR(handle)) {
+ ret = PTR_ERR(handle);
+ goto out_fail;
+ }
+
+ if (test_opt(inode->i_sb, NOBH) && ext4_should_writeback_data(inode))
+ ret = nobh_writepage(page, ext4_get_block, wbc);
+ else
+ ret = block_write_full_page(page, ext4_get_block, wbc);
+
+ if (!ret && inode->i_size > EXT4_I(inode)->i_disksize) {
+ EXT4_I(inode)->i_disksize = inode->i_size;
+ ext4_mark_inode_dirty(handle, inode);
+ }
+
+ err = ext4_journal_stop(handle);
+ if (!ret)
+ ret = err;
+ return ret;
+
+out_fail:
+ redirty_page_for_writepage(wbc, page);
+ unlock_page(page);
+ return ret;
+}

static int ext4_da_writepages(struct address_space *mapping,
struct writeback_control *wbc)
@@ -2015,7 +2053,7 @@ static const struct address_space_operations ext4_journalled_aops = {
static const struct address_space_operations ext4_da_aops = {
.readpage = ext4_readpage,
.readpages = ext4_readpages,
- .writepage = ext4_writeback_writepage,
+ .writepage = ext4_da_writepage,
.writepages = ext4_da_writepages,
.sync_page = block_sync_page,
.write_begin = ext4_da_write_begin,
--
1.5.5.rc0.16.g02b00.dirty



2008-03-21 06:19:11

by Aneesh Kumar K.V

[permalink] [raw]
Subject: [RFC PATCH] ext4: Return unwritten buffer head when trying to read from prealloc space.

ext4_ext_get_blocks returns number of blocks allocated with buffer head
unmapped for a read from prealloc space. This is needed so that delayed
allocation doesn't do block reservation for prealloc space since the blocks
are already resevred on disk. Fix ext4_ext_get_blocks to not return greater
than max_blocks and also mark the buffer head unwritten. Some code path tries
to read the block if buffer_head is not new and no uptodate. Marking the buffer
head unwritten avoid this reading.

Signed-off-by: Aneesh Kumar K.V <[email protected]>
---
fs/ext4/extents.c | 14 +++++++++++++-
1 files changed, 13 insertions(+), 1 deletions(-)

diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index d6ae40a..edd1bf2 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -2600,8 +2600,20 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
}
if (create == EXT4_CREATE_UNINITIALIZED_EXT)
goto out;
- if (!create)
+ if (!create) {
+ /*
+ * We have blocks reserved already. We
+ * return allocated blocks so that delalloc
+ * won't do block reservation for us. But
+ * the buffer head will be unmapped so that
+ * a read from the block return 0
+ */
+ if (allocated > max_blocks)
+ allocated = max_blocks;
+ /* mark the buffer unwritten */
+ __set_bit(BH_Unwritten, &bh_result->b_state);
goto out2;
+ }

ret = ext4_ext_convert_to_initialized(handle, inode,
path, iblock,
--
1.5.5.rc0.16.g02b00.dirty


2008-03-21 22:08:35

by Mingming Cao

[permalink] [raw]
Subject: Re: [RFC PATCH] ext4: Update i_diskzie during writepage for delayed allocation.

On Fri, 2008-03-21 at 11:48 +0530, Aneesh Kumar K.V wrote:
> We need to update i_disksize after allocating blocks. writepages
> already does this. Update writepage to also update i_disksize.
>

I agree that we should update the i_disk size after the allocation, to
avoid possible exposing stale data in case of fs crash happened after
the metadata hit to disk but before data being written out.

> Signed-off-by: Aneesh Kumar K.V <[email protected]>
> ---
> fs/ext4/inode.c | 40 +++++++++++++++++++++++++++++++++++++++-
> 1 files changed, 39 insertions(+), 1 deletions(-)
>
> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index 79930df..50d700f 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -1488,6 +1488,44 @@ out:
>
> return ret;
> }
> +/* FIXME!! only support data=writeback mode */
> +static int ext4_da_writepage(struct page *page,
> + struct writeback_control *wbc)
> +{
> + struct inode *inode = page->mapping->host;
> + handle_t *handle = NULL;
> + int ret = 0;
> + int err;
> +
> + if (ext4_journal_current_handle())
> + goto out_fail;
> +
> + handle = ext4_journal_start(inode, ext4_writepage_trans_blocks(inode));
> + if (IS_ERR(handle)) {
> + ret = PTR_ERR(handle);
> + goto out_fail;
> + }
> +
> + if (test_opt(inode->i_sb, NOBH) && ext4_should_writeback_data(inode))
> + ret = nobh_writepage(page, ext4_get_block, wbc);
> + else
> + ret = block_write_full_page(page, ext4_get_block, wbc);
> +
> + if (!ret && inode->i_size > EXT4_I(inode)->i_disksize) {
> + EXT4_I(inode)->i_disksize = inode->i_size;
> + ext4_mark_inode_dirty(handle, inode);
> + }
> +
> + err = ext4_journal_stop(handle);
> + if (!ret)
> + ret = err;
> + return ret;
> +
> +out_fail:
> + redirty_page_for_writepage(wbc, page);
> + unlock_page(page);
> + return ret;
> +}
>
> static int ext4_da_writepages(struct address_space *mapping,
> struct writeback_control *wbc)
> @@ -2015,7 +2053,7 @@ static const struct address_space_operations ext4_journalled_aops = {
> static const struct address_space_operations ext4_da_aops = {
> .readpage = ext4_readpage,
> .readpages = ext4_readpages,
> - .writepage = ext4_writeback_writepage,
> + .writepage = ext4_da_writepage,
> .writepages = ext4_da_writepages,
> .sync_page = block_sync_page,
> .write_begin = ext4_da_write_begin,


2008-03-21 22:14:03

by Mingming Cao

[permalink] [raw]
Subject: Re: [RFC PATCH] ext4: Return unwritten buffer head when trying to read from prealloc space.

On Fri, 2008-03-21 at 11:48 +0530, Aneesh Kumar K.V wrote:
> ext4_ext_get_blocks returns number of blocks allocated with buffer head
> unmapped for a read from prealloc space. This is needed so that delayed
> allocation doesn't do block reservation for prealloc space since the blocks
> are already resevred on disk. Fix ext4_ext_get_blocks to not return greater
> than max_blocks and also mark the buffer head unwritten. Some code path tries
> to read the block if buffer_head is not new and no uptodate. Marking the buffer
> head unwritten avoid this reading.
>

Seems this patch fixes two bugs together, it would be nice to split to
two as they cause two different problems.

To fix the bh->b_size warning solof founds for delayed allocation, I
would say we need to fix it in ext4_get_blocks_handle() (for indirect
files) and ext4_ext_get_blocks()(for extent files) both, not allowing
returning b_size greater than what the caller asking for. Also, the
patch below seems only address the preallocation map case, we need to
fix in in general block lookup case.

> Signed-off-by: Aneesh Kumar K.V <[email protected]>
> ---
> fs/ext4/extents.c | 14 +++++++++++++-
> 1 files changed, 13 insertions(+), 1 deletions(-)
>
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index d6ae40a..edd1bf2 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -2600,8 +2600,20 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
> }
> if (create == EXT4_CREATE_UNINITIALIZED_EXT)
> goto out;
> - if (!create)
> + if (!create) {
> + /*
> + * We have blocks reserved already. We
> + * return allocated blocks so that delalloc
> + * won't do block reservation for us. But
> + * the buffer head will be unmapped so that
> + * a read from the block return 0
> + */
> + if (allocated > max_blocks)
> + allocated = max_blocks;
> + /* mark the buffer unwritten */
> + __set_bit(BH_Unwritten, &bh_result->b_state);
> goto out2;
> + }
>
> ret = ext4_ext_convert_to_initialized(handle, inode,
> path, iblock,


2008-03-24 08:11:45

by Aneesh Kumar K.V

[permalink] [raw]
Subject: Re: [RFC PATCH] ext4: Return unwritten buffer head when trying to read from prealloc space.

On Fri, Mar 21, 2008 at 03:13:58PM -0700, Mingming Cao wrote:
> On Fri, 2008-03-21 at 11:48 +0530, Aneesh Kumar K.V wrote:
> > ext4_ext_get_blocks returns number of blocks allocated with buffer head
> > unmapped for a read from prealloc space. This is needed so that delayed
> > allocation doesn't do block reservation for prealloc space since the blocks
> > are already resevred on disk. Fix ext4_ext_get_blocks to not return greater
> > than max_blocks and also mark the buffer head unwritten. Some code path tries
> > to read the block if buffer_head is not new and no uptodate. Marking the buffer
> > head unwritten avoid this reading.
> >
>
> Seems this patch fixes two bugs together, it would be nice to split to
> two as they cause two different problems.

Split patches here.

http://www.radian.org/~kvaneesh/ext4/patches-to-patchqueue/

>
> To fix the bh->b_size warning solof founds for delayed allocation, I
> would say we need to fix it in ext4_get_blocks_handle() (for indirect
> files) and ext4_ext_get_blocks()(for extent files) both, not allowing
> returning b_size greater than what the caller asking for. Also, the
> patch below seems only address the preallocation map case, we need to
> fix in in general block lookup case.
>

ext4_ext_get_blocks already make sure we doesn't return > max_blocks.
for normal allocation.

2694 out:
2695 if (allocated > max_blocks)
2696 allocated = max_blocks;


-aneesh

2008-03-24 16:04:29

by Mingming Cao

[permalink] [raw]
Subject: Re: [RFC PATCH] ext4: Return unwritten buffer head when trying to read from prealloc space.

On Mon, 2008-03-24 at 13:41 +0530, Aneesh Kumar K.V wrote:
> On Fri, Mar 21, 2008 at 03:13:58PM -0700, Mingming Cao wrote:
> > On Fri, 2008-03-21 at 11:48 +0530, Aneesh Kumar K.V wrote:
> > > ext4_ext_get_blocks returns number of blocks allocated with buffer head
> > > unmapped for a read from prealloc space. This is needed so that delayed
> > > allocation doesn't do block reservation for prealloc space since the blocks
> > > are already resevred on disk. Fix ext4_ext_get_blocks to not return greater
> > > than max_blocks and also mark the buffer head unwritten. Some code path tries
> > > to read the block if buffer_head is not new and no uptodate. Marking the buffer
> > > head unwritten avoid this reading.
> > >
> >
> > Seems this patch fixes two bugs together, it would be nice to split to
> > two as they cause two different problems.
>
> Split patches here.
>
> http://www.radian.org/~kvaneesh/ext4/patches-to-patchqueue/
>
I will pick them up in patch queue.
> >
> > To fix the bh->b_size warning solof founds for delayed allocation, I
> > would say we need to fix it in ext4_get_blocks_handle() (for indirect
> > files) and ext4_ext_get_blocks()(for extent files) both, not allowing
> > returning b_size greater than what the caller asking for. Also, the
> > patch below seems only address the preallocation map case, we need to
> > fix in in general block lookup case.
> >
>
> ext4_ext_get_blocks already make sure we doesn't return > max_blocks.
> for normal allocation.
>
> 2694 out:
> 2695 if (allocated > max_blocks)
> 2696 allocated = max_blocks;
>

Oh I see, I missed that.

Regards,
Mingming
>
> -aneesh