[PATCH net] selftests: net: udpgso_bench_tx: honor CPU zero in the affinity option

[PATCH net] selftests: net: udpgso_bench_tx: honor CPU zero in the affinity option

[Jingjing Xu]

The -C option accepts a CPU number, with -1 meaning that affinity was
not requested. Testing cfg_cpu > 0 ignores an explicit -C 0, so the
benchmark can run on other CPUs despite the requested pinning.

Accept zero as well, keeping the negative default unchanged.

Discovered by Codex-assisted code inspection. Validated before and after
the fix with loopback UDP traffic, inheriting CPUs 0 and 1 and reading
the process affinity with sched_getaffinity(). Before the fix, -C 0
retains both CPUs; after it, only CPU 0 is allowed. The default and -C 1
cases retain their expected masks. Built with GCC using -Wall -Werror.

No kernel instrumentation or fault injection was used.

Fixes: 3a687bef148d ("selftests: udp gso benchmark")
Signed-off-by: Jingjing Xu <xujingjing@cmss.chinamobile.com>
---
 tools/testing/selftests/net/udpgso_bench_tx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/testing/selftests/net/udpgso_bench_tx.c b/tools/testing/selftests/net/udpgso_bench_tx.c
index 86d80cce55b4..797c8d6800f0 100644
--- a/tools/testing/selftests/net/udpgso_bench_tx.c
+++ b/tools/testing/selftests/net/udpgso_bench_tx.c
@@ -642,7 +642,7 @@ int main(int argc, char **argv)
 
 	parse_opts(argc, argv);
 
-	if (cfg_cpu > 0)
+	if (cfg_cpu >= 0)
 		set_cpu(cfg_cpu);
 
 	for (i = 0; i < sizeof(buf[0]); i++)
base-commit: 30c71476d842b094651bd7712f6c8ee8716147d0
-- 
2.53.0

Re: [PATCH net] selftests: net: udpgso_bench_tx: honor CPU zero in the affinity option

[Hangbin Liu]

On Tue, Nov 10, 2026 at 03:39:20AM +0000, Jingjing Xu wrote:
> The -C option accepts a CPU number, with -1 meaning that affinity was
> not requested. Testing cfg_cpu > 0 ignores an explicit -C 0, so the
> benchmark can run on other CPUs despite the requested pinning.
> 
> Accept zero as well, keeping the negative default unchanged.
> 
> Discovered by Codex-assisted code inspection. Validated before and after
> the fix with loopback UDP traffic, inheriting CPUs 0 and 1 and reading
> the process affinity with sched_getaffinity(). Before the fix, -C 0
> retains both CPUs; after it, only CPU 0 is allowed. The default and -C 1
> cases retain their expected masks. Built with GCC using -Wall -Werror.
> 
> No kernel instrumentation or fault injection was used.
> 
> Fixes: 3a687bef148d ("selftests: udp gso benchmark")
> Signed-off-by: Jingjing Xu <xujingjing@cmss.chinamobile.com>
> ---
>  tools/testing/selftests/net/udpgso_bench_tx.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/testing/selftests/net/udpgso_bench_tx.c b/tools/testing/selftests/net/udpgso_bench_tx.c
> index 86d80cce55b4..797c8d6800f0 100644
> --- a/tools/testing/selftests/net/udpgso_bench_tx.c
> +++ b/tools/testing/selftests/net/udpgso_bench_tx.c
> @@ -642,7 +642,7 @@ int main(int argc, char **argv)
>  
>  	parse_opts(argc, argv);
>  
> -	if (cfg_cpu > 0)
> +	if (cfg_cpu >= 0)
>  		set_cpu(cfg_cpu);
>  
>  	for (i = 0; i < sizeof(buf[0]); i++)
> base-commit: 30c71476d842b094651bd7712f6c8ee8716147d0
> -- 
> 2.53.0
> 

Reviewed-by: Hangbin Liu <liuhangbin@kylinos.cn>

Re: [PATCH net] selftests: net: udpgso_bench_tx: honor CPU zero in the affinity option

[Xia Yan]

On 11/10/26 03:39, Jingjing Xu wrote:
> Testing cfg_cpu > 0 ignores an explicit -C 0, so the
> benchmark can run on other CPUs despite the requested pinning.

Looks right. cfg_cpu defaults to -1 and set_cpu() treats 0 like any
other CPU number, so ">= 0" honors -C 0 without changing the
default. 3a687bef148d added the option and the "> 0" test together,
so the Fixes tag is accurate.

Reviewed-by: Xia Yan <yanxia@cmss.chinamobile.com>

Re: [PATCH net] selftests: net: udpgso_bench_tx: honor CPU zero in the affinity option

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Tue, 10 Nov 2026 03:39:20 +0000 you wrote:
> The -C option accepts a CPU number, with -1 meaning that affinity was
> not requested. Testing cfg_cpu > 0 ignores an explicit -C 0, so the
> benchmark can run on other CPUs despite the requested pinning.
> 
> Accept zero as well, keeping the negative default unchanged.
> 
> Discovered by Codex-assisted code inspection. Validated before and after
> the fix with loopback UDP traffic, inheriting CPUs 0 and 1 and reading
> the process affinity with sched_getaffinity(). Before the fix, -C 0
> retains both CPUs; after it, only CPU 0 is allowed. The default and -C 1
> cases retain their expected masks. Built with GCC using -Wall -Werror.
> 
> [...]

Here is the summary with links:
  - [net] selftests: net: udpgso_bench_tx: honor CPU zero in the affinity option
    https://git.kernel.org/netdev/net-next/c/c2d5ef7e600b

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html