[PATCH net-next] selftests: tls: use ASSERT_GE in test_mutliproc

[PATCH net-next] selftests: tls: use ASSERT_GE in test_mutliproc

[Geliang Tang]

From: Geliang Tang <tanggeliang@kylinos.cn>

In test_mutliproc(), when send() or recv() returns an error (e.g.,
-1), the test continues to execute the remaining code and fails
repeatedly due to using EXPECT_GE.

For example, if a TLS connection is broken and recv() returns -1,
EXPECT_GE(res, 0) records a failure but does not stop the test.
The test then proceeds with left -= res (where res = -1), causing
left to increase unexpectedly, and the loop continues indefinitely.

This results in a massive number of identical failure messages:

 # tls.c:1686:mutliproc_sendpage_writers:Expected res (-1) >= 0 (0)
 # tls.c:1686:mutliproc_sendpage_writers:Expected res (-1) >= 0 (0)
 ... (hundreds of identical failures)

Fix this by replacing EXPECT_GE with ASSERT_GE. When send() or recv()
fails, ASSERT_GE immediately aborts the current test, preventing
the subsequent undefined behavior and endless failure messages.

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 tools/testing/selftests/net/tls.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/net/tls.c b/tools/testing/selftests/net/tls.c
index 30a236b8e9f7..9b9a3cb2700d 100644
--- a/tools/testing/selftests/net/tls.c
+++ b/tools/testing/selftests/net/tls.c
@@ -1549,7 +1549,7 @@ test_mutliproc(struct __test_metadata *_metadata, struct _test_data_tls *self,
 			res = recv(self->cfd, rb,
 				   left > sizeof(rb) ? sizeof(rb) : left, 0);
 
-			EXPECT_GE(res, 0);
+			ASSERT_GE(res, 0);
 			left -= res;
 		}
 	} else {
@@ -1566,7 +1566,7 @@ test_mutliproc(struct __test_metadata *_metadata, struct _test_data_tls *self,
 				res = send(self->fd, buf,
 					   left > file_sz ? file_sz : left, 0);
 
-			EXPECT_GE(res, 0);
+			ASSERT_GE(res, 0);
 			left -= res;
 		}
 	}
-- 
2.53.0

Re: [PATCH net-next] selftests: tls: use ASSERT_GE in test_mutliproc

[Jiayuan Chen]

On 5/21/26 5:11 PM, Geliang Tang wrote:
> From: Geliang Tang <tanggeliang@kylinos.cn>
>
> In test_mutliproc(), when send() or recv() returns an error (e.g.,
> -1), the test continues to execute the remaining code and fails
> repeatedly due to using EXPECT_GE.
>
> For example, if a TLS connection is broken and recv() returns -1,
> EXPECT_GE(res, 0) records a failure but does not stop the test.
> The test then proceeds with left -= res (where res = -1), causing
> left to increase unexpectedly, and the loop continues indefinitely.
>
> This results in a massive number of identical failure messages:
>
>   # tls.c:1686:mutliproc_sendpage_writers:Expected res (-1) >= 0 (0)
>   # tls.c:1686:mutliproc_sendpage_writers:Expected res (-1) >= 0 (0)
>   ... (hundreds of identical failures)

I think it’s worth backporting, so a Fixes tag is necessary.


> Fix this by replacing EXPECT_GE with ASSERT_GE. When send() or recv()
> fails, ASSERT_GE immediately aborts the current test, preventing
> the subsequent undefined behavior and endless failure messages.
>
> Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
> ---
>   tools/testing/selftests/net/tls.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/tools/testing/selftests/net/tls.c b/tools/testing/selftests/net/tls.c
> index 30a236b8e9f7..9b9a3cb2700d 100644
> --- a/tools/testing/selftests/net/tls.c
> +++ b/tools/testing/selftests/net/tls.c
> @@ -1549,7 +1549,7 @@ test_mutliproc(struct __test_metadata *_metadata, struct _test_data_tls *self,
>   			res = recv(self->cfd, rb,
>   				   left > sizeof(rb) ? sizeof(rb) : left, 0);
>   
> -			EXPECT_GE(res, 0);
> +			ASSERT_GE(res, 0);
>   			left -= res;
>   		}
>   	} else {
> @@ -1566,7 +1566,7 @@ test_mutliproc(struct __test_metadata *_metadata, struct _test_data_tls *self,
>   				res = send(self->fd, buf,
>   					   left > file_sz ? file_sz : left, 0);
>   
> -			EXPECT_GE(res, 0);
> +			ASSERT_GE(res, 0);
>   			left -= res;
>   		}
>   	}

Re: [PATCH net-next] selftests: tls: use ASSERT_GE in test_mutliproc

[Jakub Kicinski]

On Thu, 21 May 2026 18:57:32 +0800 Jiayuan Chen wrote:
> > This results in a massive number of identical failure messages:
> >
> >   # tls.c:1686:mutliproc_sendpage_writers:Expected res (-1) >= 0 (0)
> >   # tls.c:1686:mutliproc_sendpage_writers:Expected res (-1) >= 0 (0)
> >   ... (hundreds of identical failures)  
> 
> I think it’s worth backporting, so a Fixes tag is necessary.

Why? Is this bothering someone? Are you guys hitting this in real life?
I don't recall seeing such failure in netdev CIs.

Re: [PATCH net-next] selftests: tls: use ASSERT_GE in test_mutliproc

[Jiayuan Chen]

On 5/21/26 10:07 PM, Jakub Kicinski wrote:
> On Thu, 21 May 2026 18:57:32 +0800 Jiayuan Chen wrote:
>>> This results in a massive number of identical failure messages:
>>>
>>>    # tls.c:1686:mutliproc_sendpage_writers:Expected res (-1) >= 0 (0)
>>>    # tls.c:1686:mutliproc_sendpage_writers:Expected res (-1) >= 0 (0)
>>>    ... (hundreds of identical failures)
>> I think it’s worth backporting, so a Fixes tag is necessary.
> Why? Is this bothering someone? Are you guys hitting this in real life?
> I don't recall seeing such failure in netdev CIs.


Upon second thought, if this issue occur during the development of a new 
feature,
then a Fixes tag is not necessary.