Skip to contentSkip to conversation

netfilter: conntrack: fix this -Wformat clang warning:

PATCH Superseded Justin Stitt 3 participants 8 messages Jun 6, 2022 – Jun 7, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 netfilter: conntrack: fix this -Wformat clang warning:

 | net/netfilter/nf_conntrack_standalone.c:63:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.tcp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:64:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.tcp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:69:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.udp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:70:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.udp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:75:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.dccp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:76:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.dccp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:80:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.sctp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:81:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.sctp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Variadic functions (printf-like) undergo default argument promotion. Documentation/core-api/printk-formats.rst specifically recommends using the promoted-to-type's format flag.

Also, as per C11 6.3.1.1:
(https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1548.pdf)
If an int can represent all values of the original type ..., the value is converted to an int; otherwise, it is converted to an unsigned int. These are called the integer promotions. Thus it makes sense to change %hu (as well as %u) to %d.

---
 net/netfilter/nf_conntrack_standalone.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
net/netfilter/nf_conntrack_standalone.c +5 -5
index 6ad7bbc90d38..afbec8a12c5e 100644
@@ -53,30 +53,30 @@ print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple,
 
 	switch (l4proto->l4proto) {
 	case IPPROTO_ICMP:
-		seq_printf(s, "type=%u code=%u id=%u ",
+		seq_printf(s, "type=%d code=%d id=%d ",
 			   tuple->dst.u.icmp.type,
 			   tuple->dst.u.icmp.code,
 			   ntohs(tuple->src.u.icmp.id));
 		break;
 	case IPPROTO_TCP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.tcp.port),
 			   ntohs(tuple->dst.u.tcp.port));
 		break;
 	case IPPROTO_UDPLITE:
 	case IPPROTO_UDP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.udp.port),
 			   ntohs(tuple->dst.u.udp.port));
 
 		break;
 	case IPPROTO_DCCP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.dccp.port),
 			   ntohs(tuple->dst.u.dccp.port));
 		break;
 	case IPPROTO_SCTP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.sctp.port),
 			   ntohs(tuple->dst.u.sctp.port));
 		break;
Signature · 1 line
2.30.2
Reviewed-by on v1 1/1
#1 lore raw

Hi Justin,

The subject might be a little more descriptive if if were something like:

netfilter: conntrack: Fix clang -Wformat in print_tuple()

rather than saying "this warning", as subjects should be standalone from the rest of the message. Minor nit but some maintainers have high standards for commit messages.

On Mon, Jun 06, 2022 at 02:28:19PM -0700, Justin Stitt wrote:

It would be helpful to note here that this warning appears when building arm64 with clang using -Wformat; the architecture is critical for reproducing this warning.

Quoted 37 lines
 | net/netfilter/nf_conntrack_standalone.c:63:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.tcp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:64:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.tcp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:69:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.udp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:70:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.udp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:75:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.dccp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:76:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.dccp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:80:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.sctp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:81:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.sctp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Variadic functions (printf-like) undergo default argument promotion. Documentation/core-api/printk-formats.rst specifically recommends using the promoted-to-type's format flag.

Quoted 5 lines

Also, as per C11 6.3.1.1:
(https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1548.pdf)
If an int can represent all values of the original type ..., the value is converted to an int; otherwise, it is converted to an unsigned int. These are called the integer promotions.

I think that it is probably worth noting that ntohs() has type 'int' in this case because arm64 does not define CONFIG_ARCH_USE_BUILTIN_BSWAP to use the version of __swab16() with a cast to __u16; rather, it uses the one with the conditional operator, which per 6.5.15.5 means the type of the result of the conditional operator is determined through these promotion rules as well.

I suspect that the conditional versions of __swab{16,32,64}() in include/uapi/linux/swab.h want a cast on the outside of the conditional operator with the specified type but I am not sure what other repurcussions that might have so it is probably safer to just fix these warnings up one by one.

Quoted 1 line

Thus it makes sense to change %hu (as well as %u) to %d.

Regardless of that, changing from '%hu' to '%d' is recommended so that part of the change is fine. I am not sure that changing '%u' to '%d' does anything other than add to the diff, as there was no warnings from those, but consistency is key.

Quoted 1 line
Quoted 48 lines
---
 net/netfilter/nf_conntrack_standalone.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
net/netfilter/nf_conntrack_standalone.c +5 -5
index 6ad7bbc90d38..afbec8a12c5e 100644
@@ -53,30 +53,30 @@ print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple,
 
 	switch (l4proto->l4proto) {
 	case IPPROTO_ICMP:
-		seq_printf(s, "type=%u code=%u id=%u ",
+		seq_printf(s, "type=%d code=%d id=%d ",
 			   tuple->dst.u.icmp.type,
 			   tuple->dst.u.icmp.code,
 			   ntohs(tuple->src.u.icmp.id));
 		break;
 	case IPPROTO_TCP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.tcp.port),
 			   ntohs(tuple->dst.u.tcp.port));
 		break;
 	case IPPROTO_UDPLITE:
 	case IPPROTO_UDP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.udp.port),
 			   ntohs(tuple->dst.u.udp.port));
 
 		break;
 	case IPPROTO_DCCP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.dccp.port),
 			   ntohs(tuple->dst.u.dccp.port));
 		break;
 	case IPPROTO_SCTP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.sctp.port),
 			   ntohs(tuple->dst.u.sctp.port));
 		break;
Signature · 1 line
2.30.2
#2 lore raw

PATCH v2 1/1 netfilter: conntrack: Fix clang -Wformat warning in print_tuple()

 | net/netfilter/nf_conntrack_standalone.c:63:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.tcp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:64:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.tcp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:69:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.udp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:70:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.udp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:75:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.dccp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:76:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.dccp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:80:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.sctp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:81:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.sctp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Variadic functions (printf-like) undergo default argument promotion. Documentation/core-api/printk-formats.rst specifically recommends using the promoted-to-type's format flag.

Also, as per C11 6.3.1.1:
(https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1548.pdf)
If an int can represent all values of the original type ..., the value is converted to an int; otherwise, it is converted to an unsigned int. These are called the integer promotions. Thus it makes sense to change %hu (as well as %u) to %d.

It should be noted that %u does not produce the same warning as %hu in this context. However, it should probably be changed as well for consistency.

---
Diff between v1 -> v2:
* update commit message and subject line

Note: The architecture (arm64) is critical for reproducing this warning.

 net/netfilter/nf_conntrack_standalone.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
net/netfilter/nf_conntrack_standalone.c +5 -5
index 6ad7bbc90d38..afbec8a12c5e 100644
@@ -53,30 +53,30 @@ print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple,
 
 	switch (l4proto->l4proto) {
 	case IPPROTO_ICMP:
-		seq_printf(s, "type=%u code=%u id=%u ",
+		seq_printf(s, "type=%d code=%d id=%d ",
 			   tuple->dst.u.icmp.type,
 			   tuple->dst.u.icmp.code,
 			   ntohs(tuple->src.u.icmp.id));
 		break;
 	case IPPROTO_TCP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.tcp.port),
 			   ntohs(tuple->dst.u.tcp.port));
 		break;
 	case IPPROTO_UDPLITE:
 	case IPPROTO_UDP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.udp.port),
 			   ntohs(tuple->dst.u.udp.port));
 
 		break;
 	case IPPROTO_DCCP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.dccp.port),
 			   ntohs(tuple->dst.u.dccp.port));
 		break;
 	case IPPROTO_SCTP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.sctp.port),
 			   ntohs(tuple->dst.u.sctp.port));
 		break;
Signature · 1 line
2.30.2
Reviewed-by on v2 1/1
#3 lore raw

Re: [PATCH v2] netfilter: conntrack: Fix clang -Wformat warning in print_tuple()

Quoted 45 lines · Justin Stitt
 | net/netfilter/nf_conntrack_standalone.c:63:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.tcp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:64:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.tcp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:69:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.udp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:70:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.udp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:75:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.dccp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:76:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.dccp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:80:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.sctp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:81:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.sctp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Variadic functions (printf-like) undergo default argument promotion. Documentation/core-api/printk-formats.rst specifically recommends using the promoted-to-type's format flag.

Also, as per C11 6.3.1.1:
(https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1548.pdf)
If an int can represent all values of the original type ..., the value is converted to an int; otherwise, it is converted to an unsigned int. These are called the integer promotions. Thus it makes sense to change %hu (as well as %u) to %d.

Quoted 2 lines

It should be noted that %u does not produce the same warning as %hu in this context. However, it should probably be changed as well for consistency.

Right, because they are unsigned char and the parameter is unnamed for variadic functions they are also default-argument-promoted to int. -Wformat won't warn on signedness.

Thanks for the patch!

Also, Nathan supplied his RB tag on v1; it's ok next time to include it on subsequent revisions of patches, so long as you don't change the patch too much between revisions.

Quoted 54 lines

---
Diff between v1 -> v2:
* update commit message and subject line

Note: The architecture (arm64) is critical for reproducing this warning.

 net/netfilter/nf_conntrack_standalone.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
net/netfilter/nf_conntrack_standalone.c +5 -6
index 6ad7bbc90d38..afbec8a12c5e 100644
@@ -53,30 +53,30 @@ print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple,
 
        switch (l4proto->l4proto) {
        case IPPROTO_ICMP:
-               seq_printf(s, "type=%u code=%u id=%u ",
+               seq_printf(s, "type=%d code=%d id=%d ",
                           tuple->dst.u.icmp.type,
                           tuple->dst.u.icmp.code,
                           ntohs(tuple->src.u.icmp.id));
                break;
        case IPPROTO_TCP:
-               seq_printf(s, "sport=%hu dport=%hu ",
+               seq_printf(s, "sport=%d dport=%d ",
                           ntohs(tuple->src.u.tcp.port),
                           ntohs(tuple->dst.u.tcp.port));
                break;
        case IPPROTO_UDPLITE:
        case IPPROTO_UDP:
-               seq_printf(s, "sport=%hu dport=%hu ",
+               seq_printf(s, "sport=%d dport=%d ",
                           ntohs(tuple->src.u.udp.port),
                           ntohs(tuple->dst.u.udp.port));
 
                break;
        case IPPROTO_DCCP:
-               seq_printf(s, "sport=%hu dport=%hu ",
+               seq_printf(s, "sport=%d dport=%d ",
                           ntohs(tuple->src.u.dccp.port),
                           ntohs(tuple->dst.u.dccp.port));
                break;
        case IPPROTO_SCTP:
-               seq_printf(s, "sport=%hu dport=%hu ",
+               seq_printf(s, "sport=%d dport=%d ",
                           ntohs(tuple->src.u.sctp.port),
                           ntohs(tuple->dst.u.sctp.port));
                break;
--

2.30.2

Signature · 2 lines
Thanks,
~Nick Desaulniers
#4 lore raw

Re: [PATCH v2] netfilter: conntrack: Fix clang -Wformat warning in print_tuple()

Also, please make sure to run scripts/get_maintainer.pl on your patch file and CC everyone it recommends, with the maintainers you want to pick up your patch in the To field.
https://lore.kernel.org/llvm/[email protected]/T/#u

$ ./scripts/get_maintainer.pl
0001-netfilter-conntrack-Fix-clang-Wformat-warning-in-pri.patch
Pablo Neira Ayuso <[email protected]> (maintainer:NETFILTER)
Jozsef Kadlecsik <[email protected]> (maintainer:NETFILTER)
Florian Westphal <[email protected]> (maintainer:NETFILTER)
"David S. Miller" <[email protected]> (maintainer:NETWORKING [GENERAL])
Eric Dumazet <[email protected]> (maintainer:NETWORKING [GENERAL])
Jakub Kicinski <[email protected]> (maintainer:NETWORKING [GENERAL])
Paolo Abeni <[email protected]> (maintainer:NETWORKING [GENERAL])
Nathan Chancellor <[email protected]> (supporter:CLANG/LLVM BUILD SUPPORT)
Nick Desaulniers <[email protected]> (supporter:CLANG/LLVM BUILD SUPPORT)
Tom Rix <[email protected]> (reviewer:CLANG/LLVM BUILD SUPPORT)
[email protected] (open list:NETFILTER)
[email protected] (open list:NETFILTER)
[email protected] (open list:NETWORKING [GENERAL])
[email protected] (open list)
[email protected] (open list:CLANG/LLVM BUILD SUPPORT
Quoted 122 lines · Nick Desaulniers
Quoted 45 lines · Justin Stitt
 | net/netfilter/nf_conntrack_standalone.c:63:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.tcp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:64:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.tcp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:69:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.udp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:70:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.udp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:75:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.dccp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:76:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.dccp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:80:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.sctp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:81:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.sctp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Variadic functions (printf-like) undergo default argument promotion. Documentation/core-api/printk-formats.rst specifically recommends using the promoted-to-type's format flag.

Also, as per C11 6.3.1.1:
(https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1548.pdf)
If an int can represent all values of the original type ..., the value is converted to an int; otherwise, it is converted to an unsigned int. These are called the integer promotions. Thus it makes sense to change %hu (as well as %u) to %d.

Quoted 2 lines

It should be noted that %u does not produce the same warning as %hu in this context. However, it should probably be changed as well for consistency.

Right, because they are unsigned char and the parameter is unnamed for variadic functions they are also default-argument-promoted to int. -Wformat won't warn on signedness.

Thanks for the patch!

Also, Nathan supplied his RB tag on v1; it's ok next time to include it on subsequent revisions of patches, so long as you don't change the patch too much between revisions.

Quoted 52 lines

---
Diff between v1 -> v2:
* update commit message and subject line

Note: The architecture (arm64) is critical for reproducing this warning.

 net/netfilter/nf_conntrack_standalone.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
net/netfilter/nf_conntrack_standalone.c +5 -6
index 6ad7bbc90d38..afbec8a12c5e 100644
@@ -53,30 +53,30 @@ print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple,
 
        switch (l4proto->l4proto) {
        case IPPROTO_ICMP:
-               seq_printf(s, "type=%u code=%u id=%u ",
+               seq_printf(s, "type=%d code=%d id=%d ",
                           tuple->dst.u.icmp.type,
                           tuple->dst.u.icmp.code,
                           ntohs(tuple->src.u.icmp.id));
                break;
        case IPPROTO_TCP:
-               seq_printf(s, "sport=%hu dport=%hu ",
+               seq_printf(s, "sport=%d dport=%d ",
                           ntohs(tuple->src.u.tcp.port),
                           ntohs(tuple->dst.u.tcp.port));
                break;
        case IPPROTO_UDPLITE:
        case IPPROTO_UDP:
-               seq_printf(s, "sport=%hu dport=%hu ",
+               seq_printf(s, "sport=%d dport=%d ",
                           ntohs(tuple->src.u.udp.port),
                           ntohs(tuple->dst.u.udp.port));
 
                break;
        case IPPROTO_DCCP:
-               seq_printf(s, "sport=%hu dport=%hu ",
+               seq_printf(s, "sport=%d dport=%d ",
                           ntohs(tuple->src.u.dccp.port),
                           ntohs(tuple->dst.u.dccp.port));
                break;
        case IPPROTO_SCTP:
-               seq_printf(s, "sport=%hu dport=%hu ",
+               seq_printf(s, "sport=%d dport=%d ",
                           ntohs(tuple->src.u.sctp.port),
                           ntohs(tuple->dst.u.sctp.port));
                break;
--
Quoted 2 lines

2.30.2

--
Thanks,
~Nick Desaulniers

Signature · 2 lines
Thanks,
~Nick Desaulniers
Reviewed-by on v2 1/1
#5 lore raw

Re: [PATCH v2] netfilter: conntrack: Fix clang -Wformat warning in print_tuple()

In reply to Justin Stitt
Quoted 44 lines · Justin Stitt
 | net/netfilter/nf_conntrack_standalone.c:63:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.tcp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:64:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.tcp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:69:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.udp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:70:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.udp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:75:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.dccp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:76:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.dccp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:80:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.sctp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:81:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.sctp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Variadic functions (printf-like) undergo default argument promotion. Documentation/core-api/printk-formats.rst specifically recommends using the promoted-to-type's format flag.

Also, as per C11 6.3.1.1:
(https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1548.pdf)
If an int can represent all values of the original type ..., the value is converted to an int; otherwise, it is converted to an unsigned int. These are called the integer promotions. Thus it makes sense to change %hu (as well as %u) to %d.

Quoted 4 lines

It should be noted that %u does not produce the same warning as %hu in this context. However, it should probably be changed as well for consistency.

Quoted 52 lines

---
Diff between v1 -> v2:
* update commit message and subject line

Note: The architecture (arm64) is critical for reproducing this warning.

 net/netfilter/nf_conntrack_standalone.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
net/netfilter/nf_conntrack_standalone.c +5 -5
index 6ad7bbc90d38..afbec8a12c5e 100644
@@ -53,30 +53,30 @@ print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple,
 
 	switch (l4proto->l4proto) {
 	case IPPROTO_ICMP:
-		seq_printf(s, "type=%u code=%u id=%u ",
+		seq_printf(s, "type=%d code=%d id=%d ",
 			   tuple->dst.u.icmp.type,
 			   tuple->dst.u.icmp.code,
 			   ntohs(tuple->src.u.icmp.id));
 		break;
 	case IPPROTO_TCP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.tcp.port),
 			   ntohs(tuple->dst.u.tcp.port));
 		break;
 	case IPPROTO_UDPLITE:
 	case IPPROTO_UDP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.udp.port),
 			   ntohs(tuple->dst.u.udp.port));
 
 		break;
 	case IPPROTO_DCCP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.dccp.port),
 			   ntohs(tuple->dst.u.dccp.port));
 		break;
 	case IPPROTO_SCTP:
-		seq_printf(s, "sport=%hu dport=%hu ",
+		seq_printf(s, "sport=%d dport=%d ",
 			   ntohs(tuple->src.u.sctp.port),
 			   ntohs(tuple->dst.u.sctp.port));
 		break;
Signature · 1 line
2.30.2
#6 lore raw
In reply to Nathan Chancellor
Quoted 68 lines · Nathan Chancellor

Hi Justin,

The subject might be a little more descriptive if if were something like:

netfilter: conntrack: Fix clang -Wformat in print_tuple()

rather than saying "this warning", as subjects should be standalone from the rest of the message. Minor nit but some maintainers have high standards for commit messages.

On Mon, Jun 06, 2022 at 02:28:19PM -0700, Justin Stitt wrote:

It would be helpful to note here that this warning appears when building arm64 with clang using -Wformat; the architecture is critical for reproducing this warning.

Quoted 37 lines
 | net/netfilter/nf_conntrack_standalone.c:63:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.tcp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:64:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.tcp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:69:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.udp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:70:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.udp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:75:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.dccp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:76:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.dccp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:80:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.sctp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:81:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.sctp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Variadic functions (printf-like) undergo default argument promotion. Documentation/core-api/printk-formats.rst specifically recommends using the promoted-to-type's format flag.

Quoted 5 lines

Also, as per C11 6.3.1.1:
(https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1548.pdf)
If an int can represent all values of the original type ..., the value is converted to an int; otherwise, it is converted to an unsigned int. These are called the integer promotions.

I think that it is probably worth noting that ntohs() has type 'int' in this case because arm64 does not define CONFIG_ARCH_USE_BUILTIN_BSWAP to use the version of __swab16() with a cast to __u16; rather, it uses the one with the conditional operator, which per 6.5.15.5 means the type of the result of the conditional operator is determined through these promotion rules as well.

Quoted 5 lines

I suspect that the conditional versions of __swab{16,32,64}() in include/uapi/linux/swab.h want a cast on the outside of the conditional operator with the specified type but I am not sure what other repurcussions that might have so it is probably safer to just fix these warnings up one by one.

Yes, there's a subtly where for ternary's the resulting type is the largest of all operands. When the LHS of the ? is larger than either results, the result is STILL the larger type.
https://godbolt.org/z/EGjW8sqzf

That said, regardless of that change, unnamed parameters of variadic functions will still undergo default argument promotion, so the __u16 would still be promoted to int. So we still want this patch, too.

Justin, can you send a separate patch adding the explicit cast to to __u16 for the ternary expression in __swab16 when __HAVE_BUILTIN_BSWAP16__ is not defined? include/uapi/linux/swab.h
Please include Nathan's suggested by tag; cc me and I'll provide review.

Quoted 7 lines
Quoted 2 lines

Thus it makes sense to change %hu (as well as %u) to %d.

Regardless of that, changing from '%hu' to '%d' is recommended so that part of the change is fine. I am not sure that changing '%u' to '%d' does anything other than add to the diff, as there was no warnings from those, but consistency is key.

These types in the case of icmp are uint8_t; they also are default-argument-promoted to int. All possible values of uint8_t are representable by int, and so %d and %u will make no difference. It is "no functional change" and simply being consistent, as you say.

Quoted 45 lines
Quoted 2 lines
Quoted 40 lines
---
 net/netfilter/nf_conntrack_standalone.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
net/netfilter/nf_conntrack_standalone.c +4 -5
index 6ad7bbc90d38..afbec8a12c5e 100644
@@ -53,30 +53,30 @@ print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple,
 
      switch (l4proto->l4proto) {
      case IPPROTO_ICMP:
-             seq_printf(s, "type=%u code=%u id=%u ",
+             seq_printf(s, "type=%d code=%d id=%d ",
                         tuple->dst.u.icmp.type,
                         tuple->dst.u.icmp.code,
                         ntohs(tuple->src.u.icmp.id));
              break;
      case IPPROTO_TCP:
-             seq_printf(s, "sport=%hu dport=%hu ",
+             seq_printf(s, "sport=%d dport=%d ",
                         ntohs(tuple->src.u.tcp.port),
                         ntohs(tuple->dst.u.tcp.port));
              break;
      case IPPROTO_UDPLITE:
      case IPPROTO_UDP:
-             seq_printf(s, "sport=%hu dport=%hu ",
+             seq_printf(s, "sport=%d dport=%d ",
                         ntohs(tuple->src.u.udp.port),
                         ntohs(tuple->dst.u.udp.port));
 
              break;
      case IPPROTO_DCCP:
-             seq_printf(s, "sport=%hu dport=%hu ",
+             seq_printf(s, "sport=%d dport=%d ",
                         ntohs(tuple->src.u.dccp.port),
                         ntohs(tuple->dst.u.dccp.port));
              break;
      case IPPROTO_SCTP:
-             seq_printf(s, "sport=%hu dport=%hu ",
Quoted 8 lines
Quoted 8 lines
net/netfilter/nf_conntrack_standalone.c +1 -1
+             seq_printf(s, "sport=%d dport=%d ",
                         ntohs(tuple->src.u.sctp.port),
                         ntohs(tuple->dst.u.sctp.port));
              break;
--

2.30.2

--
Thanks,
~Nick Desaulniers

#7 lore raw
Quoted 75 lines · Nick Desaulniers
Quoted 68 lines · Nathan Chancellor

Hi Justin,

The subject might be a little more descriptive if if were something like:

netfilter: conntrack: Fix clang -Wformat in print_tuple()

rather than saying "this warning", as subjects should be standalone from the rest of the message. Minor nit but some maintainers have high standards for commit messages.

On Mon, Jun 06, 2022 at 02:28:19PM -0700, Justin Stitt wrote:

It would be helpful to note here that this warning appears when building arm64 with clang using -Wformat; the architecture is critical for reproducing this warning.

Quoted 37 lines
 | net/netfilter/nf_conntrack_standalone.c:63:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.tcp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:64:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.tcp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:69:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.udp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:70:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.udp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:75:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.dccp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:76:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.dccp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:80:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->src.u.sctp.port),
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 | net/netfilter/nf_conntrack_standalone.c:81:7: warning: format specifies type
 | 'unsigned short' but the argument has type 'int' [-Wformat]
 |                            ntohs(tuple->dst.u.sctp.port));
 |                            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Variadic functions (printf-like) undergo default argument promotion. Documentation/core-api/printk-formats.rst specifically recommends using the promoted-to-type's format flag.

Quoted 5 lines

Also, as per C11 6.3.1.1:
(https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1548.pdf)
If an int can represent all values of the original type ..., the value is converted to an int; otherwise, it is converted to an unsigned int. These are called the integer promotions.

I think that it is probably worth noting that ntohs() has type 'int' in this case because arm64 does not define CONFIG_ARCH_USE_BUILTIN_BSWAP to use the version of __swab16() with a cast to __u16; rather, it uses the one with the conditional operator, which per 6.5.15.5 means the type of the result of the conditional operator is determined through these promotion rules as well.

Quoted 5 lines

I suspect that the conditional versions of __swab{16,32,64}() in include/uapi/linux/swab.h want a cast on the outside of the conditional operator with the specified type but I am not sure what other repurcussions that might have so it is probably safer to just fix these warnings up one by one.

Quoted 4 lines

Yes, there's a subtly where for ternary's the resulting type is the largest of all operands. When the LHS of the ? is larger than either results, the result is STILL the larger type.
https://godbolt.org/z/EGjW8sqzf

Ha, I used the same test locally to make sure I was understanding things correctly :)

Quoted 3 lines

That said, regardless of that change, unnamed parameters of variadic functions will still undergo default argument promotion, so the __u16 would still be promoted to int. So we still want this patch, too.

Ah right, I forgot about the default argument promotion rules. While the result of ntohs() would still be promoted to int (and we should use the wider specifier for that reason as you said), we wouldn't see a warning in the first place because the type of ntohs() would match the specifier in all cases, rather than !CONFIG_ARCH_USE_BUILTIN_BSWAP, which is the only reason I made that comment. I suspect adding that cast will clear up more warnings than just this one, depending on how many there are from ntohs() (and __swab16()).

Quoted 14 lines

Justin, can you send a separate patch adding the explicit cast to to __u16 for the ternary expression in __swab16 when __HAVE_BUILTIN_BSWAP16__ is not defined? include/uapi/linux/swab.h
Please include Nathan's suggested by tag; cc me and I'll provide review.

Quoted 7 lines
Quoted 2 lines

Thus it makes sense to change %hu (as well as %u) to %d.

Regardless of that, changing from '%hu' to '%d' is recommended so that part of the change is fine. I am not sure that changing '%u' to '%d' does anything other than add to the diff, as there was no warnings from those, but consistency is key.

Quoted 4 lines

These types in the case of icmp are uint8_t; they also are default-argument-promoted to int. All possible values of uint8_t are representable by int, and so %d and %u will make no difference. It is "no functional change" and simply being consistent, as you say.

Right, I should have made that more clear by adding ", so that part is fine as well", thanks for spelling it out!

Cheers,
Nathan