Patchwork ext4: do not normalize block requests from fallocate.

login
register
mail settings
Submitter Vivek Haldar
Date May 13, 2011, 9:19 p.m.
Message ID <1305321545-17214-1-git-send-email-haldar@google.com>
Download mbox | patch
Permalink /patch/95516/
State Accepted
Headers show

Comments

Vivek Haldar - May 13, 2011, 9:19 p.m.
Currently, an fallocate request of size slightly larger than a power of
2 is turned into two block requests, each a power of 2, with the extra
blocks pre-allocated for future use. When an application calls
fallocate, it already has an idea about how large the file may grow so
there is usually little benefit to reserve extra blocks on the
preallocation list. This reduces disk fragmentation.

Tested: fsstress. Also verified manually that fallocat'ed files are
contiguously laid out with this change (whereas without it they begin at
power-of-2 boundaries, leaving blocks in between). CPU usage of
fallocate is not appreciably higher. In a tight fallocate loop, CPU
usage hovers between 5%-8% with this change, and 5%-7% without it.

Signed-off-by: Vivek Haldar <haldar@google.com>
---
 fs/ext4/ext4.h    |    2 ++
 fs/ext4/extents.c |    5 ++++-
 2 files changed, 6 insertions(+), 1 deletions(-)
Lukas Czerner - May 16, 2011, 9:39 a.m.
On Fri, 13 May 2011, Vivek Haldar wrote:

> Currently, an fallocate request of size slightly larger than a power of
> 2 is turned into two block requests, each a power of 2, with the extra
> blocks pre-allocated for future use. When an application calls
> fallocate, it already has an idea about how large the file may grow so
> there is usually little benefit to reserve extra blocks on the
> preallocation list. This reduces disk fragmentation.
> 
> Tested: fsstress. Also verified manually that fallocat'ed files are
> contiguously laid out with this change (whereas without it they begin at
> power-of-2 boundaries, leaving blocks in between). CPU usage of
> fallocate is not appreciably higher. In a tight fallocate loop, CPU
> usage hovers between 5%-8% with this change, and 5%-7% without it.

Hi Vivek,

the patch looks good, but I do not understand why you are introducing
new #define when you can simply use EXT4_MB_HINT_NOPREALLOC and then in
the ext4_ext_map_blocks() just do :

ar.flags |= (flags & EXT4_MB_HINT_NOPREALLOC)

and you do not need to introduce new condition. Am I missing something ?

Thanks!
-Lukas

> 
> Signed-off-by: Vivek Haldar <haldar@google.com>
> ---
>  fs/ext4/ext4.h    |    2 ++
>  fs/ext4/extents.c |    5 ++++-
>  2 files changed, 6 insertions(+), 1 deletions(-)
> 
> diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
> index 076c5d2..e606d34 100644
> --- a/fs/ext4/ext4.h
> +++ b/fs/ext4/ext4.h
> @@ -512,6 +512,8 @@ struct ext4_new_group_data {
>  	/* Convert extent to initialized after IO complete */
>  #define EXT4_GET_BLOCKS_IO_CONVERT_EXT		(EXT4_GET_BLOCKS_CONVERT|\
>  					 EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
> +	/* Don't normalize when fallocat'ing */
> +#define EXT4_GET_BLOCKS_NO_NORMALIZE		0x0020
>  
>  /*
>   * Flags used by ext4_free_blocks
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index e363f21..f02dd52 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -3304,6 +3304,8 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
>  	else
>  		/* disable in-core preallocation for non-regular files */
>  		ar.flags = 0;
> +	if (flags & EXT4_GET_BLOCKS_NO_NORMALIZE)
> +		ar.flags |= EXT4_MB_HINT_NOPREALLOC;
>  	newblock = ext4_mb_new_blocks(handle, &ar, &err);
>  	if (!newblock)
>  		goto out2;
> @@ -3549,7 +3551,8 @@ retry:
>  			break;
>  		}
>  		ret = ext4_map_blocks(handle, inode, &map,
> -				      EXT4_GET_BLOCKS_CREATE_UNINIT_EXT);
> +				      EXT4_GET_BLOCKS_CREATE_UNINIT_EXT |
> +				      EXT4_GET_BLOCKS_NO_NORMALIZE);
>  		if (ret <= 0) {
>  #ifdef EXT4FS_DEBUG
>  			WARN_ON(ret <= 0);
>
Vivek Haldar - May 16, 2011, 5:18 p.m.
EXT4_GET_BLOCKS_* and EXT4_MB_HINT_* are different flag spaces and
should not be mixed. ext4_ext_map_blocks should not be using
EXT4_MB_HINT_*, it only uses EXT4_GET_BLOCKS_*.

On Mon, May 16, 2011 at 2:39 AM, Lukas Czerner <lczerner@redhat.com> wrote:
> On Fri, 13 May 2011, Vivek Haldar wrote:
>
>> Currently, an fallocate request of size slightly larger than a power of
>> 2 is turned into two block requests, each a power of 2, with the extra
>> blocks pre-allocated for future use. When an application calls
>> fallocate, it already has an idea about how large the file may grow so
>> there is usually little benefit to reserve extra blocks on the
>> preallocation list. This reduces disk fragmentation.
>>
>> Tested: fsstress. Also verified manually that fallocat'ed files are
>> contiguously laid out with this change (whereas without it they begin at
>> power-of-2 boundaries, leaving blocks in between). CPU usage of
>> fallocate is not appreciably higher. In a tight fallocate loop, CPU
>> usage hovers between 5%-8% with this change, and 5%-7% without it.
>
> Hi Vivek,
>
> the patch looks good, but I do not understand why you are introducing
> new #define when you can simply use EXT4_MB_HINT_NOPREALLOC and then in
> the ext4_ext_map_blocks() just do :
>
> ar.flags |= (flags & EXT4_MB_HINT_NOPREALLOC)
>
> and you do not need to introduce new condition. Am I missing something ?
>
> Thanks!
> -Lukas
>
>>
>> Signed-off-by: Vivek Haldar <haldar@google.com>
>> ---
>>  fs/ext4/ext4.h    |    2 ++
>>  fs/ext4/extents.c |    5 ++++-
>>  2 files changed, 6 insertions(+), 1 deletions(-)
>>
>> diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
>> index 076c5d2..e606d34 100644
>> --- a/fs/ext4/ext4.h
>> +++ b/fs/ext4/ext4.h
>> @@ -512,6 +512,8 @@ struct ext4_new_group_data {
>>       /* Convert extent to initialized after IO complete */
>>  #define EXT4_GET_BLOCKS_IO_CONVERT_EXT               (EXT4_GET_BLOCKS_CONVERT|\
>>                                        EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
>> +     /* Don't normalize when fallocat'ing */
>> +#define EXT4_GET_BLOCKS_NO_NORMALIZE         0x0020
>>
>>  /*
>>   * Flags used by ext4_free_blocks
>> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
>> index e363f21..f02dd52 100644
>> --- a/fs/ext4/extents.c
>> +++ b/fs/ext4/extents.c
>> @@ -3304,6 +3304,8 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
>>       else
>>               /* disable in-core preallocation for non-regular files */
>>               ar.flags = 0;
>> +     if (flags & EXT4_GET_BLOCKS_NO_NORMALIZE)
>> +             ar.flags |= EXT4_MB_HINT_NOPREALLOC;
>>       newblock = ext4_mb_new_blocks(handle, &ar, &err);
>>       if (!newblock)
>>               goto out2;
>> @@ -3549,7 +3551,8 @@ retry:
>>                       break;
>>               }
>>               ret = ext4_map_blocks(handle, inode, &map,
>> -                                   EXT4_GET_BLOCKS_CREATE_UNINIT_EXT);
>> +                                   EXT4_GET_BLOCKS_CREATE_UNINIT_EXT |
>> +                                   EXT4_GET_BLOCKS_NO_NORMALIZE);
>>               if (ret <= 0) {
>>  #ifdef EXT4FS_DEBUG
>>                       WARN_ON(ret <= 0);
>>
>
> --
>
Lukas Czerner - May 17, 2011, 6:47 a.m.
On Mon, 16 May 2011, Vivek Haldar wrote:

> EXT4_GET_BLOCKS_* and EXT4_MB_HINT_* are different flag spaces and
> should not be mixed. ext4_ext_map_blocks should not be using
> EXT4_MB_HINT_*, it only uses EXT4_GET_BLOCKS_*.

Agh, yes you're right of course, somehow I just missed that
ext4_map_blocks() take separate flag argument and EXT4_GET_* flag might
use the same bit as EXT4_MB_HINT_NOPREALLOC. Sorry.

Thanks!
-Lukas

> 
> On Mon, May 16, 2011 at 2:39 AM, Lukas Czerner <lczerner@redhat.com> wrote:
> > On Fri, 13 May 2011, Vivek Haldar wrote:
> >
> >> Currently, an fallocate request of size slightly larger than a power of
> >> 2 is turned into two block requests, each a power of 2, with the extra
> >> blocks pre-allocated for future use. When an application calls
> >> fallocate, it already has an idea about how large the file may grow so
> >> there is usually little benefit to reserve extra blocks on the
> >> preallocation list. This reduces disk fragmentation.
> >>
> >> Tested: fsstress. Also verified manually that fallocat'ed files are
> >> contiguously laid out with this change (whereas without it they begin at
> >> power-of-2 boundaries, leaving blocks in between). CPU usage of
> >> fallocate is not appreciably higher. In a tight fallocate loop, CPU
> >> usage hovers between 5%-8% with this change, and 5%-7% without it.
> >
> > Hi Vivek,
> >
> > the patch looks good, but I do not understand why you are introducing
> > new #define when you can simply use EXT4_MB_HINT_NOPREALLOC and then in
> > the ext4_ext_map_blocks() just do :
> >
> > ar.flags |= (flags & EXT4_MB_HINT_NOPREALLOC)
> >
> > and you do not need to introduce new condition. Am I missing something ?
> >
> > Thanks!
> > -Lukas
> >
> >>
> >> Signed-off-by: Vivek Haldar <haldar@google.com>
> >> ---
> >>  fs/ext4/ext4.h    |    2 ++
> >>  fs/ext4/extents.c |    5 ++++-
> >>  2 files changed, 6 insertions(+), 1 deletions(-)
> >>
> >> diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
> >> index 076c5d2..e606d34 100644
> >> --- a/fs/ext4/ext4.h
> >> +++ b/fs/ext4/ext4.h
> >> @@ -512,6 +512,8 @@ struct ext4_new_group_data {
> >>       /* Convert extent to initialized after IO complete */
> >>  #define EXT4_GET_BLOCKS_IO_CONVERT_EXT               (EXT4_GET_BLOCKS_CONVERT|\
> >>                                        EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
> >> +     /* Don't normalize when fallocat'ing */
> >> +#define EXT4_GET_BLOCKS_NO_NORMALIZE         0x0020
> >>
> >>  /*
> >>   * Flags used by ext4_free_blocks
> >> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> >> index e363f21..f02dd52 100644
> >> --- a/fs/ext4/extents.c
> >> +++ b/fs/ext4/extents.c
> >> @@ -3304,6 +3304,8 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
> >>       else
> >>               /* disable in-core preallocation for non-regular files */
> >>               ar.flags = 0;
> >> +     if (flags & EXT4_GET_BLOCKS_NO_NORMALIZE)
> >> +             ar.flags |= EXT4_MB_HINT_NOPREALLOC;
> >>       newblock = ext4_mb_new_blocks(handle, &ar, &err);
> >>       if (!newblock)
> >>               goto out2;
> >> @@ -3549,7 +3551,8 @@ retry:
> >>                       break;
> >>               }
> >>               ret = ext4_map_blocks(handle, inode, &map,
> >> -                                   EXT4_GET_BLOCKS_CREATE_UNINIT_EXT);
> >> +                                   EXT4_GET_BLOCKS_CREATE_UNINIT_EXT |
> >> +                                   EXT4_GET_BLOCKS_NO_NORMALIZE);
> >>               if (ret <= 0) {
> >>  #ifdef EXT4FS_DEBUG
> >>                       WARN_ON(ret <= 0);
> >>
> >
> > --
> >
> 
> 
> 
> 

--
Theodore Ts'o - May 23, 2011, 1:11 a.m.
On Fri, May 13, 2011 at 02:19:05PM -0700, Vivek Haldar wrote:
> Currently, an fallocate request of size slightly larger than a power of
> 2 is turned into two block requests, each a power of 2, with the extra
> blocks pre-allocated for future use. When an application calls
> fallocate, it already has an idea about how large the file may grow so
> there is usually little benefit to reserve extra blocks on the
> preallocation list. This reduces disk fragmentation.
> 
> Tested: fsstress. Also verified manually that fallocat'ed files are
> contiguously laid out with this change (whereas without it they begin at
> power-of-2 boundaries, leaving blocks in between). CPU usage of
> fallocate is not appreciably higher. In a tight fallocate loop, CPU
> usage hovers between 5%-8% with this change, and 5%-7% without it.
> 
> Signed-off-by: Vivek Haldar <haldar@google.com>

Applied, with a minor change to avoid a bit assignment conflict with
the punch patches.

Also, I changed the commit message to note that using a file aging
simulator which filled the file system to 70%, the percentage of free
extents greater than 8MB (as measured using e2freefrag) increased from
38.8% without this commit, to 69.4% with this commit.

      	      	   	      	    	      - Ted
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 076c5d2..e606d34 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -512,6 +512,8 @@  struct ext4_new_group_data {
 	/* Convert extent to initialized after IO complete */
 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT		(EXT4_GET_BLOCKS_CONVERT|\
 					 EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
+	/* Don't normalize when fallocat'ing */
+#define EXT4_GET_BLOCKS_NO_NORMALIZE		0x0020
 
 /*
  * Flags used by ext4_free_blocks
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index e363f21..f02dd52 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -3304,6 +3304,8 @@  int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
 	else
 		/* disable in-core preallocation for non-regular files */
 		ar.flags = 0;
+	if (flags & EXT4_GET_BLOCKS_NO_NORMALIZE)
+		ar.flags |= EXT4_MB_HINT_NOPREALLOC;
 	newblock = ext4_mb_new_blocks(handle, &ar, &err);
 	if (!newblock)
 		goto out2;
@@ -3549,7 +3551,8 @@  retry:
 			break;
 		}
 		ret = ext4_map_blocks(handle, inode, &map,
-				      EXT4_GET_BLOCKS_CREATE_UNINIT_EXT);
+				      EXT4_GET_BLOCKS_CREATE_UNINIT_EXT |
+				      EXT4_GET_BLOCKS_NO_NORMALIZE);
 		if (ret <= 0) {
 #ifdef EXT4FS_DEBUG
 			WARN_ON(ret <= 0);