2008-12-14 18:26:49

by Coly Li

Subject: [PATCH] avoid unnecessary assignment in lookup_bdev()

In normal cases, lookup_bdev() should return successfully, therefore it's not a good idea to always
assign error values before checking truth/false conditions. This patch modifies code to only assign
error value in necessary false conditions.

If I missed something, please correct me.

Signed-off-by: Coly Li <[email protected]>
---
fs/block_dev.c | 15 +++++++++------
1 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/fs/block_dev.c b/fs/block_dev.c
index 99e0ae1..10b656f 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -1282,16 +1282,19 @@ struct block_device *lookup_bdev(const char *pathname)
return ERR_PTR(error);

inode = path.dentry->d_inode;
- error = -ENOTBLK;
- if (!S_ISBLK(inode->i_mode))
+ if (!S_ISBLK(inode->i_mode)) {
+ error = -ENOTBLK;
goto fail;
- error = -EACCES;
- if (path.mnt->mnt_flags & MNT_NODEV)
+ }
+ if (path.mnt->mnt_flags & MNT_NODEV) {
+ error = -EACCES;
goto fail;
- error = -ENOMEM;
+ }
bdev = bd_acquire(inode);
- if (!bdev)
+ if (!bdev) {
+ error = -ENOMEM;
goto fail;
+ }
out:
path_put(&path);
return bdev;
--
Coly Li
SuSE PRC Labs

2008-12-14 19:13:51

by Greg KH

Subject: Re: [PATCH] avoid unnecessary assignment in lookup_bdev()

On Mon, Dec 15, 2008 at 02:33:20AM +0800, Coly Li wrote:
> In normal cases, lookup_bdev() should return successfully, therefore it's not a good idea to always
> assign error values before checking truth/false conditions. This patch modifies code to only assign
> error value in necessary false conditions.
>
> If I missed something, please correct me.

I don't understand, is there some way that the current code could
incorrectly return an error value?

Is your change somehow faster? I don't think this is on any fast-path
code, right?

thanks,

greg k-h

2008-12-14 19:23:55

by Coly Li

Subject: Re: [PATCH] avoid unnecessary assignment in lookup_bdev()

Greg KH Wrote:
> On Mon, Dec 15, 2008 at 02:33:20AM +0800, Coly Li wrote:
>> In normal cases, lookup_bdev() should return successfully, therefore it's not a good idea to always
>> assign error values before checking truth/false conditions. This patch modifies code to only assign
>> error value in necessary false conditions.
>>
>> If I missed something, please correct me.
>
> I don't understand, is there some way that the current code could
> incorrectly return an error value?
>
> Is your change somehow faster? I don't think this is on any fast-path
> code, right?
For 1 or 2 lines to set error value before checking condition, that's a clean coding style. But for
more places in one function to set error value this way, I can't say that's perfect.

lookup_bdev() is just in my code review stack, the way to set error value just makes me
uncomfortable. Maybe it is because my threshold (>=3) is lower than others. This patch can not bring
up any *real* faster, but sending it out makes me feel better at least :)

--
Coly Li
SuSE PRC Labs