2018-11-17 19:17:17

by Chengguang Xu

Subject: [PATCH] ext2: compare old and new mode before setting update_mode flag

If new mode is the same as old mode we don't have to reset
inode mode in the rest of the code, so compare old and new
mode before setting update_mode flag.

Signed-off-by: Chengguang Xu <[email protected]>
---
fs/ext2/acl.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/fs/ext2/acl.c b/fs/ext2/acl.c
index cf4c77f8dd08..f4dd728393c8 100644
--- a/fs/ext2/acl.c
+++ b/fs/ext2/acl.c
@@ -226,7 +226,8 @@ ext2_set_acl(struct inode *inode, struct posix_acl *acl, int type)
error = posix_acl_update_mode(inode, &mode, &acl);
if (error)
return error;
- update_mode = 1;
+ if (mode != inode->i_mode)
+ update_mode = 1;
}
error = __ext2_set_acl(inode, acl, type);
if (!error && update_mode) {
--
2.17.2

2018-11-22 01:04:31

by Jan Kara

Subject: Re: [PATCH] ext2: compare old and new mode before setting update_mode flag

On Wed 21-11-18 18:55:28, cgxu519 wrote:
> On 11/19/18 6:04 PM, Jan Kara wrote:
> > On Sat 17-11-18 17:01:00, Chengguang Xu wrote:
> > > If new mode is the same as old mode we don't have to reset
> > > inode mode in the rest of the code, so compare old and new
> > > mode before setting update_mode flag.
> > >
> > > Signed-off-by: Chengguang Xu <[email protected]>
> > I don't think this is quite correct. E.g. I would think that i_ctime should
> > be updated even if the effective mode resulting from acl did not change.
>
> I think? __ext2_set_acl() will probably update i_ctime in this case, am I
> missing something?

Yeah, you're right. But I'd still prefer to keep the code simple as is.
There's no great reason to optimize this since saved time is going to be
minimal and the code is not performance critical.

Honza


> > diff --git a/fs/ext2/acl.c b/fs/ext2/acl.c
> > index cf4c77f8dd08..f4dd728393c8 100644
> > --- a/fs/ext2/acl.c
> > +++ b/fs/ext2/acl.c
> > @@ -226,7 +226,8 @@ ext2_set_acl(struct inode *inode, struct posix_acl *acl, int type)
> > error = posix_acl_update_mode(inode, &mode, &acl);
> > if (error)
> > return error;
> > - update_mode = 1;
> > + if (mode != inode->i_mode)
> > + update_mode = 1;
> > }
> > error = __ext2_set_acl(inode, acl, type);
> > if (!error && update_mode) {
> > --
> > 2.17.2
> >
> >
--
Jan Kara <[email protected]>
SUSE Labs, CR

2018-11-19 20:27:34

by Jan Kara

Subject: Re: [PATCH] ext2: compare old and new mode before setting update_mode flag

On Sat 17-11-18 17:01:00, Chengguang Xu wrote:
> If new mode is the same as old mode we don't have to reset
> inode mode in the rest of the code, so compare old and new
> mode before setting update_mode flag.
>
> Signed-off-by: Chengguang Xu <[email protected]>

I don't think this is quite correct. E.g. I would think that i_ctime should
be updated even if the effective mode resulting from acl did not change.

Honza

> ---
> fs/ext2/acl.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/fs/ext2/acl.c b/fs/ext2/acl.c
> index cf4c77f8dd08..f4dd728393c8 100644
> --- a/fs/ext2/acl.c
> +++ b/fs/ext2/acl.c
> @@ -226,7 +226,8 @@ ext2_set_acl(struct inode *inode, struct posix_acl *acl, int type)
> error = posix_acl_update_mode(inode, &mode, &acl);
> if (error)
> return error;
> - update_mode = 1;
> + if (mode != inode->i_mode)
> + update_mode = 1;
> }
> error = __ext2_set_acl(inode, acl, type);
> if (!error && update_mode) {
> --
> 2.17.2
>
>
--
Jan Kara <[email protected]>
SUSE Labs, CR

2018-11-21 21:29:41

by Chengguang Xu

Subject: Re: [PATCH] ext2: compare old and new mode before setting update_mode flag

On 11/19/18 6:04 PM, Jan Kara wrote:
> On Sat 17-11-18 17:01:00, Chengguang Xu wrote:
>> If new mode is the same as old mode we don't have to reset
>> inode mode in the rest of the code, so compare old and new
>> mode before setting update_mode flag.
>>
>> Signed-off-by: Chengguang Xu <[email protected]>
> I don't think this is quite correct. E.g. I would think that i_ctime should
> be updated even if the effective mode resulting from acl did not change.

I think  __ext2_set_acl() will probably update i_ctime in this case, am
I missing something?

Thanks,


> ---
> fs/ext2/acl.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/fs/ext2/acl.c b/fs/ext2/acl.c
> index cf4c77f8dd08..f4dd728393c8 100644
> --- a/fs/ext2/acl.c
> +++ b/fs/ext2/acl.c
> @@ -226,7 +226,8 @@ ext2_set_acl(struct inode *inode, struct posix_acl *acl, int type)
> error = posix_acl_update_mode(inode, &mode, &acl);
> if (error)
> return error;
> - update_mode = 1;
> + if (mode != inode->i_mode)
> + update_mode = 1;
> }
> error = __ext2_set_acl(inode, acl, type);
> if (!error && update_mode) {
> --
> 2.17.2
>
>