[PATCH] Improve FL_KEEP_SIZE handling

[PATCH] Improve FL_KEEP_SIZE handling

[Davide Italiano]

This matches the logic of ext4. I think it's more
correct passing (offset + len) to inode_newsize_ok() 
rather than rounding up to block size.
The call can be skipped in some cases too. It works for
me but I'm new to this code so I might miss something.
Let me know what you think.

Davide Italiano (1):
  Btrfs: Improve FL_KEEP_SIZE handling in fallocate.

 fs/btrfs/file.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

-- 
2.3.4

[PATCH] Btrfs: Improve FL_KEEP_SIZE handling in fallocate.

[Davide Italiano]

- We call inode_size_ok() only if FL_KEEP_SIZE isn't specified.
- As an optimisation we can skip the call if (off + len)
  isn't greater than the current size of the file. This operation
  is called under the lock so the less work we do, the better.
- If we call inode_size_ok() pass to it the correct value rather
  than a more conservative estimation.

Signed-off-by: Davide Italiano <dccitaliano@gmail.com>
---
 fs/btrfs/file.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 30982bb..f649bfc 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -2586,9 +2586,13 @@ static long btrfs_fallocate(struct file *file, int mode,
 	}
 
 	mutex_lock(&inode->i_mutex);
-	ret = inode_newsize_ok(inode, alloc_end);
-	if (ret)
-		goto out;
+
+	if (!(mode & FALLOC_FL_KEEP_SIZE) &&
+	    offset + len > inode->i_size) {
+		ret = inode_newsize_ok(inode, offset + len);
+		if (ret)
+			goto out;
+	}
 
 	if (alloc_start > inode->i_size) {
 		ret = btrfs_cont_expand(inode, i_size_read(inode),
-- 
2.3.4

Re: [PATCH] Btrfs: Improve FL_KEEP_SIZE handling in fallocate.

[Davide Italiano]

On Mon, Apr 6, 2015 at 10:09 PM, Davide Italiano <dccitaliano@gmail.com> wrote:
> - We call inode_size_ok() only if FL_KEEP_SIZE isn't specified.
> - As an optimisation we can skip the call if (off + len)
>   isn't greater than the current size of the file. This operation
>   is called under the lock so the less work we do, the better.
> - If we call inode_size_ok() pass to it the correct value rather
>   than a more conservative estimation.
>
> Signed-off-by: Davide Italiano <dccitaliano@gmail.com>
> ---
>  fs/btrfs/file.c | 10 +++++++---
>  1 file changed, 7 insertions(+), 3 deletions(-)
>
> diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
> index 30982bb..f649bfc 100644
> --- a/fs/btrfs/file.c
> +++ b/fs/btrfs/file.c
> @@ -2586,9 +2586,13 @@ static long btrfs_fallocate(struct file *file, int mode,
>         }
>
>         mutex_lock(&inode->i_mutex);
> -       ret = inode_newsize_ok(inode, alloc_end);
> -       if (ret)
> -               goto out;
> +
> +       if (!(mode & FALLOC_FL_KEEP_SIZE) &&
> +           offset + len > inode->i_size) {
> +               ret = inode_newsize_ok(inode, offset + len);
> +               if (ret)
> +                       goto out;
> +       }
>
>         if (alloc_start > inode->i_size) {
>                 ret = btrfs_cont_expand(inode, i_size_read(inode),
> --
> 2.3.4
>

Any comment on this?

Re: [PATCH] Btrfs: Improve FL_KEEP_SIZE handling in fallocate.

[Davide Italiano]

On Mon, Apr 20, 2015 at 1:49 PM, Davide Italiano <dccitaliano@gmail.com> wrote:
> On Mon, Apr 6, 2015 at 10:09 PM, Davide Italiano <dccitaliano@gmail.com> wrote:
>> - We call inode_size_ok() only if FL_KEEP_SIZE isn't specified.
>> - As an optimisation we can skip the call if (off + len)
>>   isn't greater than the current size of the file. This operation
>>   is called under the lock so the less work we do, the better.
>> - If we call inode_size_ok() pass to it the correct value rather
>>   than a more conservative estimation.
>>
>> Signed-off-by: Davide Italiano <dccitaliano@gmail.com>
>> ---
>>  fs/btrfs/file.c | 10 +++++++---
>>  1 file changed, 7 insertions(+), 3 deletions(-)
>>
>> diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
>> index 30982bb..f649bfc 100644
>> --- a/fs/btrfs/file.c
>> +++ b/fs/btrfs/file.c
>> @@ -2586,9 +2586,13 @@ static long btrfs_fallocate(struct file *file, int mode,
>>         }
>>
>>         mutex_lock(&inode->i_mutex);
>> -       ret = inode_newsize_ok(inode, alloc_end);
>> -       if (ret)
>> -               goto out;
>> +
>> +       if (!(mode & FALLOC_FL_KEEP_SIZE) &&
>> +           offset + len > inode->i_size) {
>> +               ret = inode_newsize_ok(inode, offset + len);
>> +               if (ret)
>> +                       goto out;
>> +       }
>>
>>         if (alloc_start > inode->i_size) {
>>                 ret = btrfs_cont_expand(inode, i_size_read(inode),
>> --
>> 2.3.4
>>
>
> Any comment on this?

Very gentle ping after couple of months.

Re: [PATCH] Btrfs: Improve FL_KEEP_SIZE handling in fallocate.

[Liu Bo]

On Mon, Apr 06, 2015 at 10:09:15PM -0700, Davide Italiano wrote:
> - We call inode_size_ok() only if FL_KEEP_SIZE isn't specified.
> - As an optimisation we can skip the call if (off + len)
>   isn't greater than the current size of the file. This operation
>   is called under the lock so the less work we do, the better.
> - If we call inode_size_ok() pass to it the correct value rather
>   than a more conservative estimation.

Looks good.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>

> 
> Signed-off-by: Davide Italiano <dccitaliano@gmail.com>
> ---
>  fs/btrfs/file.c | 10 +++++++---
>  1 file changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
> index 30982bb..f649bfc 100644
> --- a/fs/btrfs/file.c
> +++ b/fs/btrfs/file.c
> @@ -2586,9 +2586,13 @@ static long btrfs_fallocate(struct file *file, int mode,
>  	}
>  
>  	mutex_lock(&inode->i_mutex);
> -	ret = inode_newsize_ok(inode, alloc_end);
> -	if (ret)
> -		goto out;
> +
> +	if (!(mode & FALLOC_FL_KEEP_SIZE) &&
> +	    offset + len > inode->i_size) {
> +		ret = inode_newsize_ok(inode, offset + len);
> +		if (ret)
> +			goto out;
> +	}
>  
>  	if (alloc_start > inode->i_size) {
>  		ret = btrfs_cont_expand(inode, i_size_read(inode),
> -- 
> 2.3.4
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] Btrfs: Improve FL_KEEP_SIZE handling in fallocate.

[David Sterba]

On Mon, Apr 06, 2015 at 10:09:15PM -0700, Davide Italiano wrote:
> - We call inode_size_ok() only if FL_KEEP_SIZE isn't specified.
> - As an optimisation we can skip the call if (off + len)
>   isn't greater than the current size of the file. This operation
>   is called under the lock so the less work we do, the better.
> - If we call inode_size_ok() pass to it the correct value rather
>   than a more conservative estimation.
> 
> Signed-off-by: Davide Italiano <dccitaliano@gmail.com>

Reviewed-by: David Sterba <dsterba@suse.cz>

Re: [PATCH] Btrfs: Improve FL_KEEP_SIZE handling in fallocate.

[Davide Italiano]

On Fri, Jun 26, 2015 at 7:08 AM, David Sterba <dsterba@suse.cz> wrote:
> On Mon, Apr 06, 2015 at 10:09:15PM -0700, Davide Italiano wrote:
>> - We call inode_size_ok() only if FL_KEEP_SIZE isn't specified.
>> - As an optimisation we can skip the call if (off + len)
>>   isn't greater than the current size of the file. This operation
>>   is called under the lock so the less work we do, the better.
>> - If we call inode_size_ok() pass to it the correct value rather
>>   than a more conservative estimation.
>>
>> Signed-off-by: Davide Italiano <dccitaliano@gmail.com>
>
> Reviewed-by: David Sterba <dsterba@suse.cz>

Hi Chris, this has been around for a while and it's been reviewed by
multiple people. Any chances you can pull in your branch?

Thanks,

--
Davide

Re: [PATCH] Btrfs: Improve FL_KEEP_SIZE handling in fallocate.

[Davide Italiano]

On Wed, Jul 22, 2015 at 10:45 AM, Davide Italiano <dccitaliano@gmail.com> wrote:
> On Fri, Jun 26, 2015 at 7:08 AM, David Sterba <dsterba@suse.cz> wrote:
>> On Mon, Apr 06, 2015 at 10:09:15PM -0700, Davide Italiano wrote:
>>> - We call inode_size_ok() only if FL_KEEP_SIZE isn't specified.
>>> - As an optimisation we can skip the call if (off + len)
>>>   isn't greater than the current size of the file. This operation
>>>   is called under the lock so the less work we do, the better.
>>> - If we call inode_size_ok() pass to it the correct value rather
>>>   than a more conservative estimation.
>>>
>>> Signed-off-by: Davide Italiano <dccitaliano@gmail.com>
>>
>> Reviewed-by: David Sterba <dsterba@suse.cz>
>
> Hi Chris, this has been around for a while and it's been reviewed by
> multiple people. Any chances you can pull in your branch?
>
> Thanks,
>
> --
> Davide

Any chance to get this in?

Re: [PATCH] Btrfs: Improve FL_KEEP_SIZE handling in fallocate.

[David Sterba]

On Wed, Oct 21, 2015 at 10:16:46AM -0700, Davide Italiano wrote:
> On Wed, Jul 22, 2015 at 10:45 AM, Davide Italiano <dccitaliano@gmail.com> wrote:
> > On Fri, Jun 26, 2015 at 7:08 AM, David Sterba <dsterba@suse.cz> wrote:
> >> On Mon, Apr 06, 2015 at 10:09:15PM -0700, Davide Italiano wrote:
> >>> - We call inode_size_ok() only if FL_KEEP_SIZE isn't specified.
> >>> - As an optimisation we can skip the call if (off + len)
> >>>   isn't greater than the current size of the file. This operation
> >>>   is called under the lock so the less work we do, the better.
> >>> - If we call inode_size_ok() pass to it the correct value rather
> >>>   than a more conservative estimation.
> >>>
> >>> Signed-off-by: Davide Italiano <dccitaliano@gmail.com>
> >>
> >> Reviewed-by: David Sterba <dsterba@suse.cz>
> >
> > Hi Chris, this has been around for a while and it's been reviewed by
> > multiple people. Any chances you can pull in your branch?

I'm adding this patch to my for-next, sorry for late response.