2020-05-04 15:32:43

by Gustavo A. R. Silva

Subject: [PATCH][next] io_uring: Remove logically dead code in io_splice

In case force_nonblock happens to be true, the function returns
at:

2779 if (force_nonblock)
2780 return -EAGAIN;

before reaching this line of code. So, the null check on force_nonblock
at 2785, is never actually being executed.

Addresses-Coverity-ID: 1492838 ("Logically dead code")
Fixes: 2fb3e82284fc ("io_uring: punt splice async because of inode mutex")
Signed-off-by: Gustavo A. R. Silva <[email protected]>
---
fs/io_uring.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index e5dfbbd2aa34..4b1efb062f7f 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -2782,7 +2782,7 @@ static int io_splice(struct io_kiocb *req, bool force_nonblock)
poff_in = (sp->off_in == -1) ? NULL : &sp->off_in;
poff_out = (sp->off_out == -1) ? NULL : &sp->off_out;
ret = do_splice(in, poff_in, out, poff_out, sp->len, flags);
- if (force_nonblock && ret == -EAGAIN)
+ if (ret == -EAGAIN)
return -EAGAIN;

io_put_file(req, in, (sp->flags & SPLICE_F_FD_IN_FIXED));
--
2.26.0

2020-05-04 15:35:12

by Pavel Begunkov

Subject: Re: [PATCH][next] io_uring: Remove logically dead code in io_splice

On 04/05/2020 18:19, Gustavo A. R. Silva wrote:
> In case force_nonblock happens to be true, the function returns
> at:
>
> 2779 if (force_nonblock)
> 2780 return -EAGAIN;
>
> before reaching this line of code. So, the null check on force_nonblock
> at 2785, is never actually being executed.
>
> Addresses-Coverity-ID: 1492838 ("Logically dead code")
> Fixes: 2fb3e82284fc ("io_uring: punt splice async because of inode mutex")

It's not a bug, so 'Fixes' tag is IMHO misleading. Anyway,

Acked-by: Pavel Begunkov <[email protected]>

> Signed-off-by: Gustavo A. R. Silva <[email protected]>
> ---
> fs/io_uring.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index e5dfbbd2aa34..4b1efb062f7f 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -2782,7 +2782,7 @@ static int io_splice(struct io_kiocb *req, bool force_nonblock)
> poff_in = (sp->off_in == -1) ? NULL : &sp->off_in;
> poff_out = (sp->off_out == -1) ? NULL : &sp->off_out;
> ret = do_splice(in, poff_in, out, poff_out, sp->len, flags);
> - if (force_nonblock && ret == -EAGAIN)
> + if (ret == -EAGAIN)
> return -EAGAIN;
>
> io_put_file(req, in, (sp->flags & SPLICE_F_FD_IN_FIXED));
>

--
Pavel Begunkov

2020-05-04 15:36:04

by Jens Axboe

Subject: Re: [PATCH][next] io_uring: Remove logically dead code in io_splice

On 5/4/20 9:19 AM, Gustavo A. R. Silva wrote:
> In case force_nonblock happens to be true, the function returns
> at:
>
> 2779 if (force_nonblock)
> 2780 return -EAGAIN;
>
> before reaching this line of code. So, the null check on force_nonblock
> at 2785, is never actually being executed.
>
> Addresses-Coverity-ID: 1492838 ("Logically dead code")
> Fixes: 2fb3e82284fc ("io_uring: punt splice async because of inode mutex")
> Signed-off-by: Gustavo A. R. Silva <[email protected]>
> ---
> fs/io_uring.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index e5dfbbd2aa34..4b1efb062f7f 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -2782,7 +2782,7 @@ static int io_splice(struct io_kiocb *req, bool force_nonblock)
> poff_in = (sp->off_in == -1) ? NULL : &sp->off_in;
> poff_out = (sp->off_out == -1) ? NULL : &sp->off_out;
> ret = do_splice(in, poff_in, out, poff_out, sp->len, flags);
> - if (force_nonblock && ret == -EAGAIN)
> + if (ret == -EAGAIN)
> return -EAGAIN;

This isn't right, it should just remove the two lines completely. But
also see:

https://lore.kernel.org/io-uring/[email protected]/T/#u

--
Jens Axboe

2020-05-04 15:56:12

by Pavel Begunkov

Subject: Re: [PATCH][next] io_uring: Remove logically dead code in io_splice

On 04/05/2020 18:25, Jens Axboe wrote:
> On 5/4/20 9:19 AM, Gustavo A. R. Silva wrote:
>> In case force_nonblock happens to be true, the function returns
>> at:
>>
>> 2779 if (force_nonblock)
>> 2780 return -EAGAIN;
>>
>> before reaching this line of code. So, the null check on force_nonblock
>> at 2785, is never actually being executed.
>>
>> Addresses-Coverity-ID: 1492838 ("Logically dead code")
>> Fixes: 2fb3e82284fc ("io_uring: punt splice async because of inode mutex")
>> Signed-off-by: Gustavo A. R. Silva <[email protected]>
>> ---
>> fs/io_uring.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index e5dfbbd2aa34..4b1efb062f7f 100644
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -2782,7 +2782,7 @@ static int io_splice(struct io_kiocb *req, bool force_nonblock)
>> poff_in = (sp->off_in == -1) ? NULL : &sp->off_in;
>> poff_out = (sp->off_out == -1) ? NULL : &sp->off_out;
>> ret = do_splice(in, poff_in, out, poff_out, sp->len, flags);
>> - if (force_nonblock && ret == -EAGAIN)
>> + if (ret == -EAGAIN)
>> return -EAGAIN;
>
> This isn't right, it should just remove the two lines completely. But
> also see:

Oh, right, it will ignore O_NONBLOCK and be resubmitted, as going through
io_wq_submit_work(). I need to be more attentive.


>
> https://lore.kernel.org/io-uring/[email protected]/T/#u
>

--
Pavel Begunkov

2020-05-04 16:21:24

by Gustavo A. R. Silva

Subject: Re: [PATCH][next] io_uring: Remove logically dead code in io_splice

On 5/4/20 10:25, Jens Axboe wrote:
>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index e5dfbbd2aa34..4b1efb062f7f 100644
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -2782,7 +2782,7 @@ static int io_splice(struct io_kiocb *req, bool force_nonblock)
>> poff_in = (sp->off_in == -1) ? NULL : &sp->off_in;
>> poff_out = (sp->off_out == -1) ? NULL : &sp->off_out;
>> ret = do_splice(in, poff_in, out, poff_out, sp->len, flags);
>> - if (force_nonblock && ret == -EAGAIN)
>> + if (ret == -EAGAIN)
>> return -EAGAIN;
>
> This isn't right, it should just remove the two lines completely. But
> also see:
>
> https://lore.kernel.org/io-uring/[email protected]/T/#u
>

Oh, I see now. Thanks for the feedback.

--
Gustavo