ext4: remove unnecessary gotos in ext4_xattr_set_entry

In the "out" label we only iput old/new_ea_inode-s, in all these places
these variables are always NULL so there is no point in goto to "out".

Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
---
 fs/ext4/xattr.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

On May 31, 2019, at 6:10 AM, Pavel Tikhomirov <ptikhomirov@virtuozzo.com> wrote:
> 
> In the "out" label we only iput old/new_ea_inode-s, in all these places
> these variables are always NULL so there is no point in goto to "out".
> 
> Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>

I'm not a fan of changes like this, since it adds potential complexity/bugs
if the error handling path is changed in the future.  That is one of the major
benefits of the "goto out_*" model of error handling is that you only need to
add one new label to the end of the function when some new state is added that
needs to be cleaned up, compared to having to check each individual error to
see if something needs to be cleaned up.

Cheers, Andreas

> ---
> fs/ext4/xattr.c | 9 +++------
> 1 file changed, 3 insertions(+), 6 deletions(-)
> 
> diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c
> index 491f9ee4040e..ac2ddd4446b3 100644
> --- a/fs/ext4/xattr.c
> +++ b/fs/ext4/xattr.c
> @@ -1601,8 +1601,7 @@ static int ext4_xattr_set_entry(struct ext4_xattr_info *i,
> 		next = EXT4_XATTR_NEXT(last);
> 		if ((void *)next >= s->end) {
> 			EXT4_ERROR_INODE(inode, "corrupted xattr entries");
> -			ret = -EFSCORRUPTED;
> -			goto out;
> +			return -EFSCORRUPTED;
> 		}
> 		if (!last->e_value_inum && last->e_value_size) {
> 			size_t offs = le16_to_cpu(last->e_value_offs);
> @@ -1620,8 +1619,7 @@ static int ext4_xattr_set_entry(struct ext4_xattr_info *i,
> 			free += EXT4_XATTR_LEN(name_len) + old_size;
> 
> 		if (free < EXT4_XATTR_LEN(name_len) + new_size) {
> -			ret = -ENOSPC;
> -			goto out;
> +			return -ENOSPC;
> 		}
> 
> 		/*
> @@ -1634,8 +1632,7 @@ static int ext4_xattr_set_entry(struct ext4_xattr_info *i,
> 		    new_size && is_block &&
> 		    (min_offs + old_size - new_size) <
> 					EXT4_XATTR_BLOCK_RESERVE(inode)) {
> -			ret = -ENOSPC;
> -			goto out;
> +			return -ENOSPC;
> 		}
> 	}
> 
> --
> 2.20.1
> 


Cheers, Andreas

On Fri, May 31, 2019 at 03:46:54PM -0600, Andreas Dilger wrote:
> On May 31, 2019, at 6:10 AM, Pavel Tikhomirov <ptikhomirov@virtuozzo.com> wrote:
> > 
> > In the "out" label we only iput old/new_ea_inode-s, in all these places
> > these variables are always NULL so there is no point in goto to "out".
> > 
> > Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
> 
> I'm not a fan of changes like this, since it adds potential complexity/bugs
> if the error handling path is changed in the future.  That is one of the major
> benefits of the "goto out_*" model of error handling is that you only need to
> add one new label to the end of the function when some new state is added that
> needs to be cleaned up, compared to having to check each individual error to
> see if something needs to be cleaned up.

I'm not a fan either, for the reasons Andreas stated; if you ever move
code around, it's much more hazardous because you now have to check if
what had previously been a "return ret" now has to change into "goto
outl".  In some case, it's really obvious, if the code is at the very
beginning of the function, but when you're 35 lines down, well over
the size of many of an editor window, it's no longer quite so obvious
whether or not "goto out" is necessary.

						- Ted