[PATCH net-next v2] selftests: mptcp: Mark xerror __noreturn

[PATCH net-next v2] selftests: mptcp: Mark xerror __noreturn

[Ankit Khushwaha]

Compiler reports potential uses of uninitialized variables in
mptcp_connect.c when xerror() is called from failure paths.

mptcp_connect.c:1262:11: warning: variable 'raw_addr' is used
      uninitialized whenever 'if' condition is false
      [-Wsometimes-uninitialized]

xerror() terminates execution by calling exit(), but it is not visible
to the compiler & assumes control flow may continue past the call.

Annotate xerror() with __noreturn so the compiler can correctly reason
about control flow and avoid false-positive uninitialized variable
warnings.

Signed-off-by: Ankit Khushwaha <ankitkhushwaha.linux@gmail.com>
---
changelog:
v2: 
- annotate 'xerror()' with __noreturn
- remove defining 'raw_addr' to NULL

v1: https://lore.kernel.org/all/20251126163046.58615-1-ankitkhushwaha.linux@gmail.com/
---
 tools/testing/selftests/net/mptcp/Makefile        | 4 ++++
 tools/testing/selftests/net/mptcp/mptcp_connect.c | 3 ++-
 tools/testing/selftests/net/mptcp/mptcp_inq.c     | 3 ++-
 tools/testing/selftests/net/mptcp/mptcp_sockopt.c | 3 ++-
 4 files changed, 10 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/net/mptcp/Makefile b/tools/testing/selftests/net/mptcp/Makefile
index 15d144a25d82..4c94c01b893a 100644
--- a/tools/testing/selftests/net/mptcp/Makefile
+++ b/tools/testing/selftests/net/mptcp/Makefile
@@ -35,3 +35,7 @@ TEST_INCLUDES := ../lib.sh $(wildcard ../lib/sh/*.sh)
 EXTRA_CLEAN := *.pcap

 include ../../lib.mk
+
+$(OUTPUT)/mptcp_connect: CFLAGS += -I$(top_srcdir)/tools/include
+$(OUTPUT)/mptcp_sockopt: CFLAGS += -I$(top_srcdir)/tools/include
+$(OUTPUT)/mptcp_inq: CFLAGS += -I$(top_srcdir)/tools/include
diff --git a/tools/testing/selftests/net/mptcp/mptcp_connect.c b/tools/testing/selftests/net/mptcp/mptcp_connect.c
index 404a77bf366a..10f6f99cfd4e 100644
--- a/tools/testing/selftests/net/mptcp/mptcp_connect.c
+++ b/tools/testing/selftests/net/mptcp/mptcp_connect.c
@@ -33,6 +33,7 @@
 #include <linux/tcp.h>
 #include <linux/time_types.h>
 #include <linux/sockios.h>
+#include <linux/compiler.h>

 extern int optind;

@@ -140,7 +141,7 @@ static void die_usage(void)
 	exit(1);
 }

-static void xerror(const char *fmt, ...)
+static void __noreturn xerror(const char *fmt, ...)
 {
 	va_list ap;

diff --git a/tools/testing/selftests/net/mptcp/mptcp_inq.c b/tools/testing/selftests/net/mptcp/mptcp_inq.c
index 8e8f6441ad8b..d4a729814662 100644
--- a/tools/testing/selftests/net/mptcp/mptcp_inq.c
+++ b/tools/testing/selftests/net/mptcp/mptcp_inq.c
@@ -28,6 +28,7 @@

 #include <linux/tcp.h>
 #include <linux/sockios.h>
+#include <linux/compiler.h>

 #ifndef IPPROTO_MPTCP
 #define IPPROTO_MPTCP 262
@@ -52,7 +53,7 @@ static void die_usage(int r)
 	exit(r);
 }

-static void xerror(const char *fmt, ...)
+static void __noreturn xerror(const char *fmt, ...)
 {
 	va_list ap;

diff --git a/tools/testing/selftests/net/mptcp/mptcp_sockopt.c b/tools/testing/selftests/net/mptcp/mptcp_sockopt.c
index 286164f7246e..15cea5e919b5 100644
--- a/tools/testing/selftests/net/mptcp/mptcp_sockopt.c
+++ b/tools/testing/selftests/net/mptcp/mptcp_sockopt.c
@@ -25,6 +25,7 @@
 #include <netinet/in.h>

 #include <linux/tcp.h>
+#include <linux/compiler.h>

 static int pf = AF_INET;

@@ -139,7 +140,7 @@ static void die_usage(int r)
 	exit(r);
 }

-static void xerror(const char *fmt, ...)
+static void __noreturn xerror(const char *fmt, ...)
 {
 	va_list ap;

--
2.52.0

Re: [PATCH net-next v2] selftests: mptcp: Mark xerror __noreturn

[Matthieu Baerts]

Hi Ankit,

On 29/11/2025 05:38, Ankit Khushwaha wrote:
> Compiler reports potential uses of uninitialized variables in
> mptcp_connect.c when xerror() is called from failure paths.
> 
> mptcp_connect.c:1262:11: warning: variable 'raw_addr' is used
>       uninitialized whenever 'if' condition is false
>       [-Wsometimes-uninitialized]
> 
> xerror() terminates execution by calling exit(), but it is not visible
> to the compiler & assumes control flow may continue past the call.
> 
> Annotate xerror() with __noreturn so the compiler can correctly reason
> about control flow and avoid false-positive uninitialized variable
> warnings.
> 
> Signed-off-by: Ankit Khushwaha <ankitkhushwaha.linux@gmail.com>
> ---
> changelog:
> v2: 
> - annotate 'xerror()' with __noreturn
> - remove defining 'raw_addr' to NULL

Thank you for the new version!

Note: this patch can target 'net' instead of 'net-next'.

> ---
>  tools/testing/selftests/net/mptcp/Makefile        | 4 ++++
>  tools/testing/selftests/net/mptcp/mptcp_connect.c | 3 ++-
>  tools/testing/selftests/net/mptcp/mptcp_inq.c     | 3 ++-
>  tools/testing/selftests/net/mptcp/mptcp_sockopt.c | 3 ++-

Good idea to fix the other tools too!

>  4 files changed, 10 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/testing/selftests/net/mptcp/Makefile b/tools/testing/selftests/net/mptcp/Makefile
> index 15d144a25d82..4c94c01b893a 100644
> --- a/tools/testing/selftests/net/mptcp/Makefile
> +++ b/tools/testing/selftests/net/mptcp/Makefile
> @@ -35,3 +35,7 @@ TEST_INCLUDES := ../lib.sh $(wildcard ../lib/sh/*.sh)
>  EXTRA_CLEAN := *.pcap
> 
>  include ../../lib.mk
> +
> +$(OUTPUT)/mptcp_connect: CFLAGS += -I$(top_srcdir)/tools/include
> +$(OUTPUT)/mptcp_sockopt: CFLAGS += -I$(top_srcdir)/tools/include
> +$(OUTPUT)/mptcp_inq: CFLAGS += -I$(top_srcdir)/tools/include

Small detail: I think you can simply append the "main" CFLAGS at the top
of the file instead of adding specific rules per tool.

Note: because the CFLAGS variable is already long, please split it like
it is done in tools/testing/selftests/net/Makefile.

While at it, do you mind adding __noreturn to die_perror() in
mptcp_diag.c mptcp_inq.c mptcp_sockopt.c as well please?

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.

Re: [PATCH net-next v2] selftests: mptcp: Mark xerror __noreturn

[Jakub Kicinski]

On Sat, 29 Nov 2025 19:13:08 +0100 Matthieu Baerts wrote:
> > +$(OUTPUT)/mptcp_connect: CFLAGS += -I$(top_srcdir)/tools/include
> > +$(OUTPUT)/mptcp_sockopt: CFLAGS += -I$(top_srcdir)/tools/include
> > +$(OUTPUT)/mptcp_inq: CFLAGS += -I$(top_srcdir)/tools/include  

I believe this is being added via MM or kselftest tree at level of the
main ksft makefile. Since this is a false positive, maybe let's defer
fixing the issue until after the merge window? When the -I.. flag
will be implicitly in place?

Re: [PATCH net-next v2] selftests: mptcp: Mark xerror __noreturn

[Matthieu Baerts]

Hi Jakub, Ankit,

On 30/11/2025 02:41, Jakub Kicinski wrote:
> On Sat, 29 Nov 2025 19:13:08 +0100 Matthieu Baerts wrote:
>>> +$(OUTPUT)/mptcp_connect: CFLAGS += -I$(top_srcdir)/tools/include
>>> +$(OUTPUT)/mptcp_sockopt: CFLAGS += -I$(top_srcdir)/tools/include
>>> +$(OUTPUT)/mptcp_inq: CFLAGS += -I$(top_srcdir)/tools/include  
> 
> I believe this is being added via MM or kselftest tree at level of the
> main ksft makefile. Since this is a false positive, maybe let's defer
> fixing the issue until after the merge window? When the -I.. flag
> will be implicitly in place?

@Jakub: Thank you for sharing this, I didn't know about that. That makes
sense to wait.

@Ankit: do you mind sending a v3 without the modifications of the
Makefile, and supporting die_perror() in a bit more than 2 weeks or
later, please?

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.

Re: [PATCH net-next v2] selftests: mptcp: Mark xerror __noreturn

[Ankit Khushwaha]

Hi Matthieu, Jakub,

On Sun, Nov 30, 2025 at 06:09:18PM +0100, Matthieu Baerts wrote:
> Hi Jakub, Ankit,
> 
> On 30/11/2025 02:41, Jakub Kicinski wrote:
> > On Sat, 29 Nov 2025 19:13:08 +0100 Matthieu Baerts wrote:
> >>> +$(OUTPUT)/mptcp_connect: CFLAGS += -I$(top_srcdir)/tools/include
> >>> +$(OUTPUT)/mptcp_sockopt: CFLAGS += -I$(top_srcdir)/tools/include
> >>> +$(OUTPUT)/mptcp_inq: CFLAGS += -I$(top_srcdir)/tools/include  
> > 
> > I believe this is being added via MM or kselftest tree at level of the
> > main ksft makefile. Since this is a false positive, maybe let's defer
> > fixing the issue until after the merge window? When the -I.. flag
> > will be implicitly in place?
> 
> @Ankit: do you mind sending a v3 without the modifications of the
> Makefile, and supporting die_perror() in a bit more than 2 weeks or
> later, please?

I will surely send v3 patch after merge window with the requested
changes.

Thanks,
-- Ankit