ubifs: Move update of cnt in get_znodes_to_commit

[PATCH] ubifs: Move update of cnt in get_znodes_to_commit

Posted by Waqar Hameed 1 month, 1 week ago

There is no need to update `cnt` for the first dirty node just before
the loop, and then again update it at the end of each iteration for the
next dirty nodes. Just update `cnt` at the beginning of each iteration
instead. This way, the first iteration will count the first dirty node
and any subsequent iterations will count the next dirty nodes.

Signed-off-by: Waqar Hameed <waqar.hameed@axis.com>
---
 fs/ubifs/tnc_commit.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/fs/ubifs/tnc_commit.c b/fs/ubifs/tnc_commit.c
index a55e04822d16..d42f7829389c 100644
--- a/fs/ubifs/tnc_commit.c
+++ b/fs/ubifs/tnc_commit.c
@@ -650,8 +650,8 @@ static int get_znodes_to_commit(struct ubifs_info *c)
 		dbg_cmt("no znodes to commit");
 		return 0;
 	}
-	cnt += 1;
 	while (1) {
+		cnt += 1;
 		ubifs_assert(c, !ubifs_zn_cow(znode));
 		__set_bit(COW_ZNODE, &znode->flags);
 		znode->alt = 0;
@@ -664,7 +664,6 @@ static int get_znodes_to_commit(struct ubifs_info *c)
 		znode->ciip = znode->iip;
 		znode->cnext = cnext;
 		znode = cnext;
-		cnt += 1;
 	}
 	dbg_cmt("committing %d znodes", cnt);
 	ubifs_assert(c, cnt == atomic_long_read(&c->dirty_zn_cnt));
-- 
2.39.2

Re: [PATCH] ubifs: Move update of cnt in get_znodes_to_commit

Posted by Zhihao Cheng 1 month, 1 week ago

在 2024/10/18 2:41, Waqar Hameed 写道:
> There is no need to update `cnt` for the first dirty node just before
> the loop, and then again update it at the end of each iteration for the
> next dirty nodes. Just update `cnt` at the beginning of each iteration
> instead. This way, the first iteration will count the first dirty node
> and any subsequent iterations will count the next dirty nodes.

Well, from my own view, I prefer the orignal style because it looks more 
readable.
   c->cnext = find_first_dirty(c->zroot.znode);
   znode = c->enext = c->cnext;
   cnt += 1;  // We get the first one.

   while (1) {
     cnext = find_next_dirty(znode);
     znode = cnext;
     cnt += 1; // We get another one.
   }

After applying this patch, the intention of 'cnt' updating is not so 
obviously. However, it does reduce the duplicated codes. I will add an 
acked-by to let Richard determine whether or not apply this patch.

Acked-by: Zhihao Cheng <chengzhihao1@huawei.com>
> 
> Signed-off-by: Waqar Hameed <waqar.hameed@axis.com>
> ---
>   fs/ubifs/tnc_commit.c | 3 +--
>   1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/fs/ubifs/tnc_commit.c b/fs/ubifs/tnc_commit.c
> index a55e04822d16..d42f7829389c 100644
> --- a/fs/ubifs/tnc_commit.c
> +++ b/fs/ubifs/tnc_commit.c
> @@ -650,8 +650,8 @@ static int get_znodes_to_commit(struct ubifs_info *c)
>   		dbg_cmt("no znodes to commit");
>   		return 0;
>   	}
> -	cnt += 1;
>   	while (1) {
> +		cnt += 1;
>   		ubifs_assert(c, !ubifs_zn_cow(znode));
>   		__set_bit(COW_ZNODE, &znode->flags);
>   		znode->alt = 0;
> @@ -664,7 +664,6 @@ static int get_znodes_to_commit(struct ubifs_info *c)
>   		znode->ciip = znode->iip;
>   		znode->cnext = cnext;
>   		znode = cnext;
> -		cnt += 1;
>   	}
>   	dbg_cmt("committing %d znodes", cnt);
>   	ubifs_assert(c, cnt == atomic_long_read(&c->dirty_zn_cnt));
>

Re: [PATCH] ubifs: Move update of cnt in get_znodes_to_commit

Posted by Waqar Hameed 2 weeks, 5 days ago

On Fri, Oct 18, 2024 at 09:24 +0800 Zhihao Cheng <chengzhihao1@huawei.com> wrote:

> 在 2024/10/18 2:41, Waqar Hameed 写道:
>> There is no need to update `cnt` for the first dirty node just before
>> the loop, and then again update it at the end of each iteration for the
>> next dirty nodes. Just update `cnt` at the beginning of each iteration
>> instead. This way, the first iteration will count the first dirty node
>> and any subsequent iterations will count the next dirty nodes.
>
> Well, from my own view, I prefer the orignal style because it looks more
> readable.
>   c->cnext = find_first_dirty(c->zroot.znode);
>   znode = c->enext = c->cnext;
>   cnt += 1;  // We get the first one.
>
>   while (1) {
>     cnext = find_next_dirty(znode);
>     znode = cnext;
>     cnt += 1; // We get another one.
>   }
>
> After applying this patch, the intention of 'cnt' updating is not so obviously.
> However, it does reduce the duplicated codes. I will add an acked-by to let
> Richard determine whether or not apply this patch.
>
> Acked-by: Zhihao Cheng <chengzhihao1@huawei.com>

[...]

As you say, there could be a subjective argument here (for me it was the
other way around :) ), and that the objective argument is that it
reduces the code. 

I'm fine with either! Richard, you decide then.