[PATCH] io_uring: Fix reversed nonblock flag

[PATCH] io_uring: Fix reversed nonblock flag

[Pavel Begunkov (Silence)]

From: Pavel Begunkov <asml.silence@gmail.com>

io_queue_link_head() accepts @force_nonblock flag, but io_ring_submit()
passes something opposite.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 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 c934f91c51e9..ffe66512ca07 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -2761,7 +2761,7 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
 
 	if (link)
 		io_queue_link_head(ctx, link, &link->submit, shadow_req,
-					block_for_last);
+					force_nonblock);
 	if (statep)
 		io_submit_state_end(statep);
 
-- 
2.23.0

Re: [PATCH] io_uring: Fix reversed nonblock flag

[Pavel Begunkov]

[-- Attachment #1.1: Type: text/plain, Size: 940 bytes --]

I haven't followed the code path properly, but it looks strange to me.
Jens, could you take a look?


On 04/10/2019 11:25, Pavel Begunkov (Silence) wrote:
> From: Pavel Begunkov <asml.silence@gmail.com>
> 
> io_queue_link_head() accepts @force_nonblock flag, but io_ring_submit()
> passes something opposite.
> 
> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
> ---
>  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 c934f91c51e9..ffe66512ca07 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -2761,7 +2761,7 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
>  
>  	if (link)
>  		io_queue_link_head(ctx, link, &link->submit, shadow_req,
> -					block_for_last);
> +					force_nonblock);
>  	if (statep)
>  		io_submit_state_end(statep);
>  
> 

-- 
Yours sincerely,
Pavel Begunkov


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH] io_uring: Fix reversed nonblock flag

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 3882 bytes --]

Hi "Pavel,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on linus/master]
[cannot apply to v5.4-rc1 next-20191004]
[if your patch is applied to the wrong git tree, please drop us a note to help
improve the system. BTW, we also suggest to use '--base' option to specify the
base tree in git format-patch, please see https://stackoverflow.com/a/37406982]

url:    https://github.com/0day-ci/linux/commits/Pavel-Begunkov-Silence/io_uring-Fix-reversed-nonblock-flag/20191004-163432
config: m68k-allyesconfig (attached as .config)
compiler: m68k-linux-gcc (GCC) 7.4.0
reproduce:
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        GCC_VERSION=7.4.0 make.cross ARCH=m68k 

If you fix the issue, kindly add following tag
Reported-by: kbuild test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

   fs/io_uring.c: In function 'io_ring_submit':
>> fs/io_uring.c:2764:6: error: 'force_nonblock' undeclared (first use in this function); did you mean 'tcf_block'?
         force_nonblock);
         ^~~~~~~~~~~~~~
         tcf_block
   fs/io_uring.c:2764:6: note: each undeclared identifier is reported only once for each function it appears in

vim +2764 fs/io_uring.c

  2697	
  2698	static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
  2699				  bool block_for_last)
  2700	{
  2701		struct io_submit_state state, *statep = NULL;
  2702		struct io_kiocb *link = NULL;
  2703		struct io_kiocb *shadow_req = NULL;
  2704		bool prev_was_link = false;
  2705		int i, submit = 0;
  2706	
  2707		if (to_submit > IO_PLUG_THRESHOLD) {
  2708			io_submit_state_start(&state, ctx, to_submit);
  2709			statep = &state;
  2710		}
  2711	
  2712		for (i = 0; i < to_submit; i++) {
  2713			bool force_nonblock = true;
  2714			struct sqe_submit s;
  2715	
  2716			if (!io_get_sqring(ctx, &s))
  2717				break;
  2718	
  2719			/*
  2720			 * If previous wasn't linked and we have a linked command,
  2721			 * that's the end of the chain. Submit the previous link.
  2722			 */
  2723			if (!prev_was_link && link) {
  2724				io_queue_link_head(ctx, link, &link->submit, shadow_req,
  2725							force_nonblock);
  2726				link = NULL;
  2727				shadow_req = NULL;
  2728			}
  2729			prev_was_link = (s.sqe->flags & IOSQE_IO_LINK) != 0;
  2730	
  2731			if (link && (s.sqe->flags & IOSQE_IO_DRAIN)) {
  2732				if (!shadow_req) {
  2733					shadow_req = io_get_req(ctx, NULL);
  2734					if (unlikely(!shadow_req))
  2735						goto out;
  2736					shadow_req->flags |= (REQ_F_IO_DRAIN | REQ_F_SHADOW_DRAIN);
  2737					refcount_dec(&shadow_req->refs);
  2738				}
  2739				shadow_req->sequence = s.sequence;
  2740			}
  2741	
  2742	out:
  2743			s.has_user = true;
  2744			s.needs_lock = false;
  2745			s.needs_fixed_file = false;
  2746			submit++;
  2747	
  2748			/*
  2749			 * The caller will block for events after submit, submit the
  2750			 * last IO non-blocking. This is either the only IO it's
  2751			 * submitting, or it already submitted the previous ones. This
  2752			 * improves performance by avoiding an async punt that we don't
  2753			 * need to do.
  2754			 */
  2755			if (block_for_last && submit == to_submit)
  2756				force_nonblock = false;
  2757	
  2758			io_submit_sqe(ctx, &s, statep, &link, force_nonblock);
  2759		}
  2760		io_commit_sqring(ctx);
  2761	
  2762		if (link)
  2763			io_queue_link_head(ctx, link, &link->submit, shadow_req,
> 2764						force_nonblock);
  2765		if (statep)
  2766			io_submit_state_end(statep);
  2767	
  2768		return submit;
  2769	}
  2770	

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 51648 bytes --]

Re: [PATCH] io_uring: Fix reversed nonblock flag

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 3895 bytes --]

Hi "Pavel,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on linus/master]
[cannot apply to v5.4-rc1 next-20191002]
[if your patch is applied to the wrong git tree, please drop us a note to help
improve the system. BTW, we also suggest to use '--base' option to specify the
base tree in git format-patch, please see https://stackoverflow.com/a/37406982]

url:    https://github.com/0day-ci/linux/commits/Pavel-Begunkov-Silence/io_uring-Fix-reversed-nonblock-flag/20191004-163432
config: sparc-allyesconfig (attached as .config)
compiler: sparc64-linux-gcc (GCC) 7.4.0
reproduce:
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        GCC_VERSION=7.4.0 make.cross ARCH=sparc 

If you fix the issue, kindly add following tag
Reported-by: kbuild test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

   fs/io_uring.c: In function 'io_ring_submit':
>> fs/io_uring.c:2764:6: error: 'force_nonblock' undeclared (first use in this function); did you mean 'lock_pin_lock'?
         force_nonblock);
         ^~~~~~~~~~~~~~
         lock_pin_lock
   fs/io_uring.c:2764:6: note: each undeclared identifier is reported only once for each function it appears in

vim +2764 fs/io_uring.c

  2697	
  2698	static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
  2699				  bool block_for_last)
  2700	{
  2701		struct io_submit_state state, *statep = NULL;
  2702		struct io_kiocb *link = NULL;
  2703		struct io_kiocb *shadow_req = NULL;
  2704		bool prev_was_link = false;
  2705		int i, submit = 0;
  2706	
  2707		if (to_submit > IO_PLUG_THRESHOLD) {
  2708			io_submit_state_start(&state, ctx, to_submit);
  2709			statep = &state;
  2710		}
  2711	
  2712		for (i = 0; i < to_submit; i++) {
  2713			bool force_nonblock = true;
  2714			struct sqe_submit s;
  2715	
  2716			if (!io_get_sqring(ctx, &s))
  2717				break;
  2718	
  2719			/*
  2720			 * If previous wasn't linked and we have a linked command,
  2721			 * that's the end of the chain. Submit the previous link.
  2722			 */
  2723			if (!prev_was_link && link) {
  2724				io_queue_link_head(ctx, link, &link->submit, shadow_req,
  2725							force_nonblock);
  2726				link = NULL;
  2727				shadow_req = NULL;
  2728			}
  2729			prev_was_link = (s.sqe->flags & IOSQE_IO_LINK) != 0;
  2730	
  2731			if (link && (s.sqe->flags & IOSQE_IO_DRAIN)) {
  2732				if (!shadow_req) {
  2733					shadow_req = io_get_req(ctx, NULL);
  2734					if (unlikely(!shadow_req))
  2735						goto out;
  2736					shadow_req->flags |= (REQ_F_IO_DRAIN | REQ_F_SHADOW_DRAIN);
  2737					refcount_dec(&shadow_req->refs);
  2738				}
  2739				shadow_req->sequence = s.sequence;
  2740			}
  2741	
  2742	out:
  2743			s.has_user = true;
  2744			s.needs_lock = false;
  2745			s.needs_fixed_file = false;
  2746			submit++;
  2747	
  2748			/*
  2749			 * The caller will block for events after submit, submit the
  2750			 * last IO non-blocking. This is either the only IO it's
  2751			 * submitting, or it already submitted the previous ones. This
  2752			 * improves performance by avoiding an async punt that we don't
  2753			 * need to do.
  2754			 */
  2755			if (block_for_last && submit == to_submit)
  2756				force_nonblock = false;
  2757	
  2758			io_submit_sqe(ctx, &s, statep, &link, force_nonblock);
  2759		}
  2760		io_commit_sqring(ctx);
  2761	
  2762		if (link)
  2763			io_queue_link_head(ctx, link, &link->submit, shadow_req,
> 2764						force_nonblock);
  2765		if (statep)
  2766			io_submit_state_end(statep);
  2767	
  2768		return submit;
  2769	}
  2770	

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 59356 bytes --]

[PATCH v2] io_uring: Fix reversed nonblock flag

[Pavel Begunkov (Silence)]

From: Pavel Begunkov <asml.silence@gmail.com>

io_queue_link_head() accepts @force_nonblock flag, but io_ring_submit()
passes something opposite.

v2: fix build error by test robot: Rebase from custom tree
Reported-by: kbuild test robot <lkp@intel.com>

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index c934f91c51e9..c909ea2b84e9 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -2703,6 +2703,7 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
 	struct io_kiocb *shadow_req = NULL;
 	bool prev_was_link = false;
 	int i, submit = 0;
+	bool force_nonblock = true;
 
 	if (to_submit > IO_PLUG_THRESHOLD) {
 		io_submit_state_start(&state, ctx, to_submit);
@@ -2710,9 +2711,9 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
 	}
 
 	for (i = 0; i < to_submit; i++) {
-		bool force_nonblock = true;
 		struct sqe_submit s;
 
+		force_nonblock = true;
 		if (!io_get_sqring(ctx, &s))
 			break;
 
@@ -2761,7 +2762,7 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
 
 	if (link)
 		io_queue_link_head(ctx, link, &link->submit, shadow_req,
-					block_for_last);
+					force_nonblock);
 	if (statep)
 		io_submit_state_end(statep);
 
-- 
2.23.0

Re: [PATCH v2] io_uring: Fix reversed nonblock flag

[Jens Axboe]

On 10/4/19 4:07 AM, Pavel Begunkov (Silence) wrote:
> From: Pavel Begunkov <asml.silence@gmail.com>
> 
> io_queue_link_head() accepts @force_nonblock flag, but io_ring_submit()
> passes something opposite.
> 
> v2: fix build error by test robot: Rebase from custom tree
> Reported-by: kbuild test robot <lkp@intel.com>
> 
> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
> ---
>   fs/io_uring.c | 5 +++--
>   1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index c934f91c51e9..c909ea2b84e9 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -2703,6 +2703,7 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
>   	struct io_kiocb *shadow_req = NULL;
>   	bool prev_was_link = false;
>   	int i, submit = 0;
> +	bool force_nonblock = true;
>   
>   	if (to_submit > IO_PLUG_THRESHOLD) {
>   		io_submit_state_start(&state, ctx, to_submit);
> @@ -2710,9 +2711,9 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
>   	}
>   
>   	for (i = 0; i < to_submit; i++) {
> -		bool force_nonblock = true;
>   		struct sqe_submit s;
>   
> +		force_nonblock = true;
>   		if (!io_get_sqring(ctx, &s))
>   			break;
>   
> @@ -2761,7 +2762,7 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
>   
>   	if (link)
>   		io_queue_link_head(ctx, link, &link->submit, shadow_req,
> -					block_for_last);
> +					force_nonblock);
>   	if (statep)
>   		io_submit_state_end(statep);

Shouldn't this just be:

   		io_queue_link_head(ctx, link, &link->submit, shadow_req,
 					!block_for_last);

We're outside the loop, so by definition at the end of what we need to
do. We don't need to factor in the fiddling of force_nonblock here,
it'll be false at this point anyway. Only exception is error handling,
if the caller asked for more than what was in the ring. Not a big
deal...

-- 
Jens Axboe

Re: [PATCH v2] io_uring: Fix reversed nonblock flag

[Pavel Begunkov]

[-- Attachment #1.1: Type: text/plain, Size: 2197 bytes --]

On 04/10/2019 16:05, Jens Axboe wrote:
> On 10/4/19 4:07 AM, Pavel Begunkov (Silence) wrote:
>> From: Pavel Begunkov <asml.silence@gmail.com>
>>
>> io_queue_link_head() accepts @force_nonblock flag, but io_ring_submit()
>> passes something opposite.
>>
>> v2: fix build error by test robot: Rebase from custom tree
>> Reported-by: kbuild test robot <lkp@intel.com>
>>
>> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
>> ---
>>   fs/io_uring.c | 5 +++--
>>   1 file changed, 3 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index c934f91c51e9..c909ea2b84e9 100644
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -2703,6 +2703,7 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
>>   	struct io_kiocb *shadow_req = NULL;
>>   	bool prev_was_link = false;
>>   	int i, submit = 0;
>> +	bool force_nonblock = true;
>>   
>>   	if (to_submit > IO_PLUG_THRESHOLD) {
>>   		io_submit_state_start(&state, ctx, to_submit);
>> @@ -2710,9 +2711,9 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
>>   	}
>>   
>>   	for (i = 0; i < to_submit; i++) {
>> -		bool force_nonblock = true;
>>   		struct sqe_submit s;
>>   
>> +		force_nonblock = true;
>>   		if (!io_get_sqring(ctx, &s))
>>   			break;
>>   
>> @@ -2761,7 +2762,7 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
>>   
>>   	if (link)
>>   		io_queue_link_head(ctx, link, &link->submit, shadow_req,
>> -					block_for_last);
>> +					force_nonblock);
>>   	if (statep)
>>   		io_submit_state_end(statep);
> 
> Shouldn't this just be:
> 
>    		io_queue_link_head(ctx, link, &link->submit, shadow_req,
>  					!block_for_last);
> 
> We're outside the loop, so by definition at the end of what we need to
> do. We don't need to factor in the fiddling of force_nonblock here,
> it'll be false at this point anyway. Only exception is error handling,
> if the caller asked for more than what was in the ring. Not a big
> deal...

Thanks for explaining this, I'll resend.
Played safe because of breaks in the loop.

-- 
Yours sincerely,
Pavel Begunkov


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[PATCH v3] io_uring: Fix reversed nonblock flag

[Pavel Begunkov (Silence)]

From: Pavel Begunkov <asml.silence@gmail.com>

io_queue_link_head() accepts @force_nonblock flag, but io_ring_submit()
passes something opposite.

v2: build error from test robot: Rebase to block-tree
v3: simplify with Jens suggestion

Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 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 c934f91c51e9..b58c3d8594d8 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -2761,7 +2761,7 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit,
 
 	if (link)
 		io_queue_link_head(ctx, link, &link->submit, shadow_req,
-					block_for_last);
+					!block_for_last);
 	if (statep)
 		io_submit_state_end(statep);
 
-- 
2.23.0

Re: [PATCH v3] io_uring: Fix reversed nonblock flag

[Jens Axboe]

On 10/4/19 8:01 AM, Pavel Begunkov (Silence) wrote:
> From: Pavel Begunkov <asml.silence@gmail.com>
> 
> io_queue_link_head() accepts @force_nonblock flag, but io_ring_submit()
> passes something opposite.
> 
> v2: build error from test robot: Rebase to block-tree
> v3: simplify with Jens suggestion

Thanks, looks good to me. Two minor notes that I fixed up:

1) The revision history should go below the --- line
2) I've added a Fixes tag

-- 
Jens Axboe