Re: [PATCH] GFS2: minor readability improvement in gfs2_mount()

[Alexey Khoroshilov <khoroshilov@ispras.ru>]

Hi Steve,

On 06/06/2013 01:24 PM, Steven Whitehouse wrote:
> Hi,
>
> On Thu, 2013-06-06 at 01:29 +0400, Alexey Khoroshilov wrote:
>> The patch makes usage pattern of IS_ERR-PTR_ERR more typical.
>>
>> Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
>> ---
>>  fs/gfs2/ops_fstype.c | 5 +++--
>>  1 file changed, 3 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
>> index 0262c19..0d30e8e 100644
>> --- a/fs/gfs2/ops_fstype.c
>> +++ b/fs/gfs2/ops_fstype.c
>> @@ -1313,9 +1313,10 @@ static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
>>  	}
>>  	s = sget(fs_type, test_gfs2_super, set_gfs2_super, flags, bdev);
>>  	mutex_unlock(&bdev->bd_fsfreeze_mutex);
>> -	error = PTR_ERR(s);
>> -	if (IS_ERR(s))
>> +	if (IS_ERR(s)) {
>> +		error = PTR_ERR(s);
>>  		goto error_bdev;
>> +	}
> I'm not sure what you are getting at here. Why add a jump into the fast
> path?
My experiments with gcc 4.7.3 on x86-64 do not show any difference in
code generated with patch and without it.
    227a:    e8 00 00 00 00           callq  227f <gfs2_mount+0xb1>
            227b: R_X86_64_PC32    sget-0x4
    227f:    4c 89 f7                 mov    %r14,%rdi
    2282:    49 89 c4                 mov    %rax,%r12
    2285:    e8 00 00 00 00           callq  228a <gfs2_mount+0xbc>
            2286: R_X86_64_PC32    mutex_unlock-0x4
    228a:    49 81 fc 00 f0 ff ff     cmp    $0xfffffffffffff000,%r12
    2291:    0f 87 2a 01 00 00        ja     23c1 <gfs2_mount+0x1f3>
    2297:    49 83 7c 24 60 00        cmpq   $0x0,0x60(%r12)
    229d:    74 0b                    je     22aa <gfs2_mount+0xdc>
    229f:    44 89 ee                 mov    %r13d,%esi
    22a2:    48 89 df                 mov    %rbx,%rdi
    22a5:    e8 00 00 00 00           callq  22aa <gfs2_mount+0xdc>
            22a6: R_X86_64_PC32    blkdev_put-0x4

It looks like 'error = PTR_ERR(s);' assignment is just ignored because s and error share the same register %r12.

So the only point of the patch is to improve readability.

--
Best regards,
Alexey