2022-08-11 08:21:19

by Zhang chunchao

Subject: [PATCH] Modify the return value ret to EOPNOTSUPP when initialized to reduce repeated assignment of errno

Remove unnecessary initialization assignments.

Signed-off-by: Zhang chunchao <[email protected]>
---
io_uring/io_uring.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index b54218da075c..8c267af06401 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -3859,14 +3859,13 @@ SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
void __user *, arg, unsigned int, nr_args)
{
struct io_ring_ctx *ctx;
- long ret = -EBADF;
+ long ret = -EOPNOTSUPP;
struct fd f;

f = fdget(fd);
if (!f.file)
return -EBADF;

- ret = -EOPNOTSUPP;
if (!io_is_uring_fops(f.file))
goto out_fput;

--
2.18.2

2022-08-11 15:48:16

by Stefano Garzarella

Subject: Re: [PATCH] Modify the return value ret to EOPNOTSUPP when initialized to reduce repeated assignment of errno

On Thu, Aug 11, 2022 at 03:56:38PM +0800, Zhang chunchao wrote:
>Remove unnecessary initialization assignments.
>
>Signed-off-by: Zhang chunchao <[email protected]>
>---
> io_uring/io_uring.c | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)
>
>diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
>index b54218da075c..8c267af06401 100644
>--- a/io_uring/io_uring.c
>+++ b/io_uring/io_uring.c
>@@ -3859,14 +3859,13 @@ SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
> void __user *, arg, unsigned int, nr_args)
> {
> struct io_ring_ctx *ctx;
>- long ret = -EBADF;
>+ long ret = -EOPNOTSUPP;
> struct fd f;
>
> f = fdget(fd);
> if (!f.file)
> return -EBADF;
>
>- ret = -EOPNOTSUPP;
> if (!io_is_uring_fops(f.file))
> goto out_fput;
>

What about remove the initialization and assign it in the if branch?
I find it a bit easier to read.

I mean something like this:

--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -3859,16 +3859,17 @@ SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
void __user *, arg, unsigned int, nr_args)
{
struct io_ring_ctx *ctx;
- long ret = -EBADF;
+ long ret;
struct fd f;

f = fdget(fd);
if (!f.file)
return -EBADF;

- ret = -EOPNOTSUPP;
- if (!io_is_uring_fops(f.file))
+ if (!io_is_uring_fops(f.file)) {
+ ret = -EOPNOTSUPP;
goto out_fput;
+ }

ctx = f.file->private_data;


Otherwise remove the initialization, but leave the assignment as it is
now.

Thanks,
Stefano

2022-08-11 16:13:30

by Jens Axboe

Subject: Re: [PATCH] Modify the return value ret to EOPNOTSUPP when initialized to reduce repeated assignment of errno

On 8/11/22 9:02 AM, Stefano Garzarella wrote:
> On Thu, Aug 11, 2022 at 03:56:38PM +0800, Zhang chunchao wrote:
>> Remove unnecessary initialization assignments.
>>
>> Signed-off-by: Zhang chunchao <[email protected]>
>> ---
>> io_uring/io_uring.c | 3 +--
>> 1 file changed, 1 insertion(+), 2 deletions(-)
>>
>> diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
>> index b54218da075c..8c267af06401 100644
>> --- a/io_uring/io_uring.c
>> +++ b/io_uring/io_uring.c
>> @@ -3859,14 +3859,13 @@ SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
>> void __user *, arg, unsigned int, nr_args)
>> {
>> struct io_ring_ctx *ctx;
>> - long ret = -EBADF;
>> + long ret = -EOPNOTSUPP;
>> struct fd f;
>>
>> f = fdget(fd);
>> if (!f.file)
>> return -EBADF;
>>
>> - ret = -EOPNOTSUPP;
>> if (!io_is_uring_fops(f.file))
>> goto out_fput;
>>
>
> What about remove the initialization and assign it in the if branch?
> I find it a bit easier to read.
>
> I mean something like this:
>
> --- a/io_uring/io_uring.c
> +++ b/io_uring/io_uring.c
> @@ -3859,16 +3859,17 @@ SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
> void __user *, arg, unsigned int, nr_args)
> {
> struct io_ring_ctx *ctx;
> - long ret = -EBADF;
> + long ret;
> struct fd f;
>
> f = fdget(fd);
> if (!f.file)
> return -EBADF;
>
> - ret = -EOPNOTSUPP;
> - if (!io_is_uring_fops(f.file))
> + if (!io_is_uring_fops(f.file)) {
> + ret = -EOPNOTSUPP;
> goto out_fput;
> + }
>
> ctx = f.file->private_data;
>
>
> Otherwise remove the initialization, but leave the assignment as it is now.

Generally the kernel likes to do:

err = -EFOO;
if (something)
goto err_out;

rather than put it inside the if clause. I guess the rationale is it
makes it harder to forget to init the error value. I don't feel too
strongly, I'm fine with your patch too. Can you send it as a real patch?

--
Jens Axboe

2022-08-11 19:31:22

by Stefano Garzarella

Subject: Re: [PATCH] Modify the return value ret to EOPNOTSUPP when initialized to reduce repeated assignment of errno

On Thu, Aug 11, 2022 at 09:41:38AM -0600, Jens Axboe wrote:
>On 8/11/22 9:02 AM, Stefano Garzarella wrote:
>> On Thu, Aug 11, 2022 at 03:56:38PM +0800, Zhang chunchao wrote:
>>> Remove unnecessary initialization assignments.
>>>
>>> Signed-off-by: Zhang chunchao <[email protected]>
>>> ---
>>> io_uring/io_uring.c | 3 +--
>>> 1 file changed, 1 insertion(+), 2 deletions(-)
>>>
>>> diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
>>> index b54218da075c..8c267af06401 100644
>>> --- a/io_uring/io_uring.c
>>> +++ b/io_uring/io_uring.c
>>> @@ -3859,14 +3859,13 @@ SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
>>> void __user *, arg, unsigned int, nr_args)
>>> {
>>> struct io_ring_ctx *ctx;
>>> - long ret = -EBADF;
>>> + long ret = -EOPNOTSUPP;
>>> struct fd f;
>>>
>>> f = fdget(fd);
>>> if (!f.file)
>>> return -EBADF;
>>>
>>> - ret = -EOPNOTSUPP;
>>> if (!io_is_uring_fops(f.file))
>>> goto out_fput;
>>>
>>
>> What about remove the initialization and assign it in the if branch?
>> I find it a bit easier to read.
>>
>> I mean something like this:
>>
>> --- a/io_uring/io_uring.c
>> +++ b/io_uring/io_uring.c
>> @@ -3859,16 +3859,17 @@ SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
>> void __user *, arg, unsigned int, nr_args)
>> {
>> struct io_ring_ctx *ctx;
>> - long ret = -EBADF;
>> + long ret;
>> struct fd f;
>>
>> f = fdget(fd);
>> if (!f.file)
>> return -EBADF;
>>
>> - ret = -EOPNOTSUPP;
>> - if (!io_is_uring_fops(f.file))
>> + if (!io_is_uring_fops(f.file)) {
>> + ret = -EOPNOTSUPP;
>> goto out_fput;
>> + }
>>
>> ctx = f.file->private_data;
>>
>>
>> Otherwise remove the initialization, but leave the assignment as it is now.
>
>Generally the kernel likes to do:
>
>err = -EFOO;
>if (something)
> goto err_out;
>
>rather than put it inside the if clause. I guess the rationale is it
>makes it harder to forget to init the error value. I don't feel too

ah, thanks for pointing this out! Make sense to me, but I hope recent
compilers can spot that kind of issue :-)

>strongly, I'm fine with your patch too. Can you send it as a real patch?

@Zhang: if you want, feel free to change your patch following the
suggestions and send a new version, otherwise I can send mine of course.

Thanks,
Stefano