[PATCH] ext2: optimize ext2_xattr_get()

[PATCH] ext2: optimize ext2_xattr_get()

[Chengguang Xu]

Since xattr entry names are sorted, we don't have
to continue when current entry name is greater than
target.

Signed-off-by: Chengguang Xu <cgxu519@zoho.com.cn>
---
 fs/ext2/xattr.c | 16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)

diff --git a/fs/ext2/xattr.c b/fs/ext2/xattr.c
index d21dbf297b74..f1f857b83b45 100644
--- a/fs/ext2/xattr.c
+++ b/fs/ext2/xattr.c
@@ -178,7 +178,7 @@ ext2_xattr_get(struct inode *inode, int name_index, const char *name,
 	struct ext2_xattr_entry *entry;
 	size_t name_len, size;
 	char *end;
-	int error;
+	int error, not_found;
 	struct mb_cache *ea_block_cache = EA_BLOCK_CACHE(inode);
 
 	ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
@@ -220,10 +220,18 @@ ext2_xattr_get(struct inode *inode, int name_index, const char *name,
 			goto bad_block;
 		if (!ext2_xattr_entry_valid(entry, inode->i_sb->s_blocksize))
 			goto bad_block;
-		if (name_index == entry->e_name_index &&
-		    name_len == entry->e_name_len &&
-		    memcmp(name, entry->e_name, name_len) == 0)
+
+		not_found = name_index - entry->e_name_index;
+		if (!not_found)
+			not_found = name_len - entry->e_name_len;
+		if (!not_found)
+			not_found = memcmp(name, entry->e_name,
+						   name_len);
+		if (!not_found)
 			goto found;
+		if (not_found < 0)
+			break;
+
 		entry = next;
 	}
 	if (ext2_xattr_cache_insert(ea_block_cache, bh))
-- 
2.20.1

Re: [PATCH] ext2: optimize ext2_xattr_get()

[Jan Kara]

On Tue 21-05-19 16:21:39, Chengguang Xu wrote:
> Since xattr entry names are sorted, we don't have
> to continue when current entry name is greater than
> target.
> 
> Signed-off-by: Chengguang Xu <cgxu519@zoho.com.cn>

Thanks for the patch! If we are going to do these comparisons in multiple
places, then please create a helper function to do the comparison (so that
we have the same comparison in every place). Something like:

int ext2_xattr_cmp(int name_index, size_t name_len, const char *name,
		   struct ext2_xattr_entry *entry)

Thanks!

								Honza
> ---
>  fs/ext2/xattr.c | 16 ++++++++++++----
>  1 file changed, 12 insertions(+), 4 deletions(-)
> 
> diff --git a/fs/ext2/xattr.c b/fs/ext2/xattr.c
> index d21dbf297b74..f1f857b83b45 100644
> --- a/fs/ext2/xattr.c
> +++ b/fs/ext2/xattr.c
> @@ -178,7 +178,7 @@ ext2_xattr_get(struct inode *inode, int name_index, const char *name,
>  	struct ext2_xattr_entry *entry;
>  	size_t name_len, size;
>  	char *end;
> -	int error;
> +	int error, not_found;
>  	struct mb_cache *ea_block_cache = EA_BLOCK_CACHE(inode);
>  
>  	ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
> @@ -220,10 +220,18 @@ ext2_xattr_get(struct inode *inode, int name_index, const char *name,
>  			goto bad_block;
>  		if (!ext2_xattr_entry_valid(entry, inode->i_sb->s_blocksize))
>  			goto bad_block;
> -		if (name_index == entry->e_name_index &&
> -		    name_len == entry->e_name_len &&
> -		    memcmp(name, entry->e_name, name_len) == 0)
> +
> +		not_found = name_index - entry->e_name_index;
> +		if (!not_found)
> +			not_found = name_len - entry->e_name_len;
> +		if (!not_found)
> +			not_found = memcmp(name, entry->e_name,
> +						   name_len);
> +		if (!not_found)
>  			goto found;
> +		if (not_found < 0)
> +			break;
> +
>  		entry = next;
>  	}
>  	if (ext2_xattr_cache_insert(ea_block_cache, bh))
> -- 
> 2.20.1
> 
> 
> 
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: [PATCH] ext2: optimize ext2_xattr_get()

[cgxu519]

On Thu, 2019-05-23 at 16:46 +0200, Jan Kara wrote:
> On Tue 21-05-19 16:21:39, Chengguang Xu wrote:
> > Since xattr entry names are sorted, we don't have
> > to continue when current entry name is greater than
> > target.
> > 
> > Signed-off-by: Chengguang Xu <cgxu519@zoho.com.cn>
> 
> Thanks for the patch! If we are going to do these comparisons in multiple
> places, then please create a helper function to do the comparison (so that
> we have the same comparison in every place). Something like:
> 
> int ext2_xattr_cmp(int name_index, size_t name_len, const char *name,
> 		   struct ext2_xattr_entry *entry)
> 

Hi Jan,

Thanks for the review and advice. 

You are right we should introduce a helper to handle this part of work
and personally I think maybe implementing a helper to find target entry
will be more useful, do you think it makes sense?


Thanks,
Chengguang

Re: [PATCH] ext2: optimize ext2_xattr_get()

[Jan Kara]

On Fri 24-05-19 10:07:05, cgxu519@zoho.com.cn wrote:
> On Thu, 2019-05-23 at 16:46 +0200, Jan Kara wrote:
> > On Tue 21-05-19 16:21:39, Chengguang Xu wrote:
> > > Since xattr entry names are sorted, we don't have
> > > to continue when current entry name is greater than
> > > target.
> > > 
> > > Signed-off-by: Chengguang Xu <cgxu519@zoho.com.cn>
> > 
> > Thanks for the patch! If we are going to do these comparisons in multiple
> > places, then please create a helper function to do the comparison (so that
> > we have the same comparison in every place). Something like:
> > 
> > int ext2_xattr_cmp(int name_index, size_t name_len, const char *name,
> > 		   struct ext2_xattr_entry *entry)
> > 
> 
> Hi Jan,
> 
> Thanks for the review and advice. 
> 
> You are right we should introduce a helper to handle this part of work
> and personally I think maybe implementing a helper to find target entry
> will be more useful, do you think it makes sense?

It makes sense but ext2_xattr_set() also computes min_offs and last during
the search so using the search function in that case won't be a readbility
win I guess. So I'm not sure the search function pays off in the end.

								Honza

> 
> 
> Thanks,
> Chengguang
> 
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: [PATCH] ext2: optimize ext2_xattr_get()

[cgxu519]

On Fri, 2019-05-24 at 10:07 +0200, Jan Kara wrote:
> On Fri 24-05-19 10:07:05, cgxu519@zoho.com.cn wrote:
> > On Thu, 2019-05-23 at 16:46 +0200, Jan Kara wrote:
> > > On Tue 21-05-19 16:21:39, Chengguang Xu wrote:
> > > > Since xattr entry names are sorted, we don't have
> > > > to continue when current entry name is greater than
> > > > target.
> > > > 
> > > > Signed-off-by: Chengguang Xu <cgxu519@zoho.com.cn>
> > > 
> > > Thanks for the patch! If we are going to do these comparisons in multiple
> > > places, then please create a helper function to do the comparison (so that
> > > we have the same comparison in every place). Something like:
> > > 
> > > int ext2_xattr_cmp(int name_index, size_t name_len, const char *name,
> > > 		   struct ext2_xattr_entry *entry)
> > > 
> > 
> > Hi Jan,
> > 
> > Thanks for the review and advice. 
> > 
> > You are right we should introduce a helper to handle this part of work
> > and personally I think maybe implementing a helper to find target entry
> > will be more useful, do you think it makes sense?
> 
> It makes sense but ext2_xattr_set() also computes min_offs and last during
> the search so using the search function in that case won't be a readbility
> win I guess. So I'm not sure the search function pays off in the end.

Yes, I noticed that too, I plan to set min_offs pointer as function parameter
so that we can seperate different search modes based on it.

Thanks,
Chengguang