[PATCH][next] net: inet_sock.h: Avoid thousands of -Wflex-array-member-not-at-end warnings

[PATCH][next] net: inet_sock.h: Avoid thousands of -Wflex-array-member-not-at-end warnings

[Gustavo A. R. Silva]

Use the new TRAILING_OVERLAP() helper to fix 2600 of the following
warnings:

2600 ./include/net/inet_sock.h:65:33: warning: structure containing a flexible array member is not at the end of another structure [-Wflex-array-member-not-at-end]

This helper creates a union between a flexible-array member (FAM)
and a set of members that would otherwise follow it (in this case
`char data[40];) This overlays the trailing members (data) onto the FAM
(__data) while keeping the FAM and the start of MEMBERS aligned.

The static_assert() ensures this alignment remains, and it's
intentionally placed inmediately after `struct ip_options_data`
(no blank line in between).

Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
---

I think it's worth mentioning that the introduction of the new
TRAILING_OVERLAP() helper saves us from making changes like the
following, for this particular case:

	https://lore.kernel.org/linux-hardening/ZzK-n_C2yl8mW2Tz@kspp/

Thanks

 include/net/inet_sock.h | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/include/net/inet_sock.h b/include/net/inet_sock.h
index 1086256549fa..a974588803af 100644
--- a/include/net/inet_sock.h
+++ b/include/net/inet_sock.h
@@ -62,9 +62,12 @@ struct ip_options_rcu {
 };
 
 struct ip_options_data {
-	struct ip_options_rcu	opt;
-	char			data[40];
+	TRAILING_OVERLAP(struct ip_options_rcu, opt, opt.__data,
+			 char			data[40];
+	);
 };
+static_assert(offsetof(struct ip_options_data, opt.opt.__data) ==
+	      offsetof(struct ip_options_data, data));
 
 struct inet_request_sock {
 	struct request_sock	req;
-- 
2.43.0

Re: [PATCH][next] net: inet_sock.h: Avoid thousands of -Wflex-array-member-not-at-end warnings

[Simon Horman]

On Tue, Oct 21, 2025 at 12:43:30PM +0100, Gustavo A. R. Silva wrote:
> Use the new TRAILING_OVERLAP() helper to fix 2600 of the following
> warnings:
> 
> 2600 ./include/net/inet_sock.h:65:33: warning: structure containing a flexible array member is not at the end of another structure [-Wflex-array-member-not-at-end]
> 
> This helper creates a union between a flexible-array member (FAM)
> and a set of members that would otherwise follow it (in this case
> `char data[40];) This overlays the trailing members (data) onto the FAM
> (__data) while keeping the FAM and the start of MEMBERS aligned.
> 
> The static_assert() ensures this alignment remains, and it's
> intentionally placed inmediately after `struct ip_options_data`
> (no blank line in between).
> 
> Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
> ---
> 
> I think it's worth mentioning that the introduction of the new
> TRAILING_OVERLAP() helper saves us from making changes like the
> following, for this particular case:
> 
> 	https://lore.kernel.org/linux-hardening/ZzK-n_C2yl8mW2Tz@kspp/

Thanks,

I was able to reproduce a (small) subset of those warnings
and agree that this is a very nice way to address them.

Reviewed-by: Simon Horman <horms@kernel.org>

...

Re: [PATCH][next] net: inet_sock.h: Avoid thousands of -Wflex-array-member-not-at-end warnings

[Jakub Kicinski]

On Tue, 21 Oct 2025 12:43:30 +0100 Gustavo A. R. Silva wrote:
>  struct ip_options_data {
> -	struct ip_options_rcu	opt;
> -	char			data[40];
> +	TRAILING_OVERLAP(struct ip_options_rcu, opt, opt.__data,
> +			 char			data[40];
> +	);
>  };

Is there a way to reserve space for flexible length array on the stack
without resorting to any magic macros? This struct has total of 5 users.
-- 
pw-bot: cr

Re: [PATCH][next] net: inet_sock.h: Avoid thousands of -Wflex-array-member-not-at-end warnings

[Gustavo A. R. Silva]

On 10/24/25 01:25, Jakub Kicinski wrote:
> On Tue, 21 Oct 2025 12:43:30 +0100 Gustavo A. R. Silva wrote:
>>   struct ip_options_data {
>> -	struct ip_options_rcu	opt;
>> -	char			data[40];
>> +	TRAILING_OVERLAP(struct ip_options_rcu, opt, opt.__data,
>> +			 char			data[40];
>> +	);
>>   };
> 
> Is there a way to reserve space for flexible length array on the stack
> without resorting to any magic macros? This struct has total of 5 users.

Not that I know of. That's the reason why we had to implement macros like
TRAILING_OVERLAP(), DEFINE_FLEX(), DEFINE_RAW_FLEX().

Regarding these three macros, the simplest and least intrusive one to use is
actually TRAILING_OVERLAP(), when the flex-array member is not annotated with
the counted_by attribute (otherwise, DEFINE_FLEX() would be preferred).

Of course, the most straightforward alternative is to use fixed-size arrays
if flex arrays are not actually needed.

Thanks
-Gustavo

Re: [PATCH][next] net: inet_sock.h: Avoid thousands of -Wflex-array-member-not-at-end warnings

[Jakub Kicinski]

On Fri, 24 Oct 2025 12:24:09 +0100 Gustavo A. R. Silva wrote:
> On 10/24/25 01:25, Jakub Kicinski wrote:
> > On Tue, 21 Oct 2025 12:43:30 +0100 Gustavo A. R. Silva wrote:  
> >>   struct ip_options_data {
> >> -	struct ip_options_rcu	opt;
> >> -	char			data[40];
> >> +	TRAILING_OVERLAP(struct ip_options_rcu, opt, opt.__data,
> >> +			 char			data[40];
> >> +	);
> >>   };  
> > 
> > Is there a way to reserve space for flexible length array on the stack
> > without resorting to any magic macros? This struct has total of 5 users.  
> 
> Not that I know of. That's the reason why we had to implement macros like
> TRAILING_OVERLAP(), DEFINE_FLEX(), DEFINE_RAW_FLEX().
> 
> Regarding these three macros, the simplest and least intrusive one to use is
> actually TRAILING_OVERLAP(), when the flex-array member is not annotated with
> the counted_by attribute (otherwise, DEFINE_FLEX() would be preferred).
> 
> Of course, the most straightforward alternative is to use fixed-size arrays
> if flex arrays are not actually needed.

Honestly, I'm tired of the endless, nasty macros for no clear benefit.
This patch is not happening.

Re: [PATCH][next] net: inet_sock.h: Avoid thousands of -Wflex-array-member-not-at-end warnings

[Gustavo A. R. Silva]

On 10/25/25 08:23, Jakub Kicinski wrote:
> On Fri, 24 Oct 2025 12:24:09 +0100 Gustavo A. R. Silva wrote:
>> On 10/24/25 01:25, Jakub Kicinski wrote:
>>> On Tue, 21 Oct 2025 12:43:30 +0100 Gustavo A. R. Silva wrote:
>>>>    struct ip_options_data {
>>>> -	struct ip_options_rcu	opt;
>>>> -	char			data[40];
>>>> +	TRAILING_OVERLAP(struct ip_options_rcu, opt, opt.__data,
>>>> +			 char			data[40];
>>>> +	);
>>>>    };
>>>
>>> Is there a way to reserve space for flexible length array on the stack
>>> without resorting to any magic macros? This struct has total of 5 users.
>>
>> Not that I know of. That's the reason why we had to implement macros like
>> TRAILING_OVERLAP(), DEFINE_FLEX(), DEFINE_RAW_FLEX().
>>
>> Regarding these three macros, the simplest and least intrusive one to use is
>> actually TRAILING_OVERLAP(), when the flex-array member is not annotated with
>> the counted_by attribute (otherwise, DEFINE_FLEX() would be preferred).
>>
>> Of course, the most straightforward alternative is to use fixed-size arrays
>> if flex arrays are not actually needed.
> 
> Honestly, I'm tired of the endless, nasty macros for no clear benefit.
> This patch is not happening.

Not sure what you mean by "nasty" and "no clear benefit", but here is an
alternative. See it as RFC:

https://git.kernel.org/pub/scm/linux/kernel/git/gustavoars/linux.git/commit/?h=testing/wfamnae-next20251117&id=e0547082214e61b1db0f5068da0daa3d11f992a5

[PATCH RFC][next] ipv4/inet_sock.h: Avoid thousands of -Wflex-array-member-not-at-end warnings

Use DEFINE_RAW_FLEX() to avoid thousands of -Wflex-array-member-not-at-end
warnings.

Remove fixed-size array char data[40]; from struct ip_options_data, so
that flexible-array member struct ip_options_rcu::opt.__data[] ends last
in this (and other) structure(s).

Compensate for this by using the DEFINE_RAW_FLEX() helper to declare each
on-stack struct instance that contains struct ip_options_data as a member.

Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
---
  include/net/inet_sock.h |  1 -
  net/ipv4/icmp.c         | 88 +++++++++++++++++++++--------------------
  net/ipv4/ip_output.c    | 12 +++---
  net/ipv4/ping.c         |  6 +--
  net/ipv4/raw.c          |  6 +--
  net/ipv4/udp.c          |  6 +--
  6 files changed, 60 insertions(+), 59 deletions(-)

diff --git a/include/net/inet_sock.h b/include/net/inet_sock.h
index ac1c75975908..3b5da5e54673 100644
--- a/include/net/inet_sock.h
+++ b/include/net/inet_sock.h
@@ -63,7 +63,6 @@ struct ip_options_rcu {

  struct ip_options_data {
  	struct ip_options_rcu	opt;
-	char			data[40];
  };

  struct inet_request_sock {
diff --git a/net/ipv4/icmp.c b/net/ipv4/icmp.c
index 4abbec2f47ef..744d4e91cb9c 100644
--- a/net/ipv4/icmp.c
+++ b/net/ipv4/icmp.c
@@ -353,9 +353,11 @@ void icmp_out_count(struct net *net, unsigned char type)
  static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
  			  struct sk_buff *skb)
  {
-	struct icmp_bxm *icmp_param = from;
+	DEFINE_RAW_FLEX(struct icmp_bxm, icmp_param, replyopts.opt.opt.__data, 40);
  	__wsum csum;

+	icmp_param = from;
+
  	csum = skb_copy_and_csum_bits(icmp_param->skb,
  				      icmp_param->offset + offset,
  				      to, len);
@@ -775,9 +777,9 @@ icmp_ext_append(struct net *net, struct sk_buff *skb_in, struct icmphdr *icmph,
  void __icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info,
  		 const struct inet_skb_parm *parm)
  {
+	DEFINE_RAW_FLEX(struct icmp_bxm, icmp_param, replyopts.opt.opt.__data, 40);
  	struct iphdr *iph;
  	int room;
-	struct icmp_bxm icmp_param;
  	struct rtable *rt = skb_rtable(skb_in);
  	bool apply_ratelimit = false;
  	struct sk_buff *ext_skb;
@@ -906,7 +908,7 @@ void __icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info,
  					   iph->tos;
  	mark = IP4_REPLY_MARK(net, skb_in->mark);

-	if (__ip_options_echo(net, &icmp_param.replyopts.opt.opt, skb_in,
+	if (__ip_options_echo(net, &icmp_param->replyopts.opt.opt, skb_in,
  			      &parm->opt))
  		goto out_unlock;

@@ -915,21 +917,21 @@ void __icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info,
  	 *	Prepare data for ICMP header.
  	 */

-	icmp_param.data.icmph.type	 = type;
-	icmp_param.data.icmph.code	 = code;
-	icmp_param.data.icmph.un.gateway = info;
-	icmp_param.data.icmph.checksum	 = 0;
-	icmp_param.skb	  = skb_in;
-	icmp_param.offset = skb_network_offset(skb_in);
+	icmp_param->data.icmph.type	 = type;
+	icmp_param->data.icmph.code	 = code;
+	icmp_param->data.icmph.un.gateway = info;
+	icmp_param->data.icmph.checksum	 = 0;
+	icmp_param->skb	  = skb_in;
+	icmp_param->offset = skb_network_offset(skb_in);
  	ipcm_init(&ipc);
  	ipc.tos = tos;
  	ipc.addr = iph->saddr;
-	ipc.opt = &icmp_param.replyopts.opt;
+	ipc.opt = &icmp_param->replyopts.opt;
  	ipc.sockc.mark = mark;

  	rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr,
  			       inet_dsfield_to_dscp(tos), mark, type, code,
-			       &icmp_param);
+			       icmp_param);
  	if (IS_ERR(rt))
  		goto out_unlock;

@@ -942,7 +944,7 @@ void __icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info,
  	room = dst_mtu(&rt->dst);
  	if (room > 576)
  		room = 576;
-	room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen;
+	room -= sizeof(struct iphdr) + icmp_param->replyopts.opt.opt.optlen;
  	room -= sizeof(struct icmphdr);
  	/* Guard against tiny mtu. We need to include at least one
  	 * IP network header for this message to make any sense.
@@ -950,15 +952,15 @@ void __icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info,
  	if (room <= (int)sizeof(struct iphdr))
  		goto ende;

-	ext_skb = icmp_ext_append(net, skb_in, &icmp_param.data.icmph, room,
+	ext_skb = icmp_ext_append(net, skb_in, &icmp_param->data.icmph, room,
  				  parm->iif);
  	if (ext_skb)
-		icmp_param.skb = ext_skb;
+		icmp_param->skb = ext_skb;

-	icmp_param.data_len = icmp_param.skb->len - icmp_param.offset;
-	if (icmp_param.data_len > room)
-		icmp_param.data_len = room;
-	icmp_param.head_len = sizeof(struct icmphdr);
+	icmp_param->data_len = icmp_param->skb->len - icmp_param->offset;
+	if (icmp_param->data_len > room)
+		icmp_param->data_len = room;
+	icmp_param->head_len = sizeof(struct icmphdr);

  	/* if we don't have a source address at this point, fall back to the
  	 * dummy address instead of sending out a packet with a source address
@@ -969,7 +971,7 @@ void __icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info,

  	trace_icmp_send(skb_in, type, code);

-	icmp_push_reply(sk, &icmp_param, &fl4, &ipc, &rt);
+	icmp_push_reply(sk, icmp_param, &fl4, &ipc, &rt);

  	if (ext_skb)
  		consume_skb(ext_skb);
@@ -1206,7 +1208,7 @@ static enum skb_drop_reason icmp_redirect(struct sk_buff *skb)

  static enum skb_drop_reason icmp_echo(struct sk_buff *skb)
  {
-	struct icmp_bxm icmp_param;
+	DEFINE_RAW_FLEX(struct icmp_bxm, icmp_param, replyopts.opt.opt.__data, 40);
  	struct net *net;

  	net = skb_dst_dev_net_rcu(skb);
@@ -1214,18 +1216,18 @@ static enum skb_drop_reason icmp_echo(struct sk_buff *skb)
  	if (READ_ONCE(net->ipv4.sysctl_icmp_echo_ignore_all))
  		return SKB_NOT_DROPPED_YET;

-	icmp_param.data.icmph	   = *icmp_hdr(skb);
-	icmp_param.skb		   = skb;
-	icmp_param.offset	   = 0;
-	icmp_param.data_len	   = skb->len;
-	icmp_param.head_len	   = sizeof(struct icmphdr);
+	icmp_param->data.icmph	   = *icmp_hdr(skb);
+	icmp_param->skb		   = skb;
+	icmp_param->offset	   = 0;
+	icmp_param->data_len	   = skb->len;
+	icmp_param->head_len	   = sizeof(struct icmphdr);

-	if (icmp_param.data.icmph.type == ICMP_ECHO)
-		icmp_param.data.icmph.type = ICMP_ECHOREPLY;
-	else if (!icmp_build_probe(skb, &icmp_param.data.icmph))
+	if (icmp_param->data.icmph.type == ICMP_ECHO)
+		icmp_param->data.icmph.type = ICMP_ECHOREPLY;
+	else if (!icmp_build_probe(skb, &icmp_param->data.icmph))
  		return SKB_NOT_DROPPED_YET;

-	icmp_reply(&icmp_param, skb);
+	icmp_reply(icmp_param, skb);
  	return SKB_NOT_DROPPED_YET;
  }

@@ -1353,7 +1355,7 @@ EXPORT_SYMBOL_GPL(icmp_build_probe);
   */
  static enum skb_drop_reason icmp_timestamp(struct sk_buff *skb)
  {
-	struct icmp_bxm icmp_param;
+	DEFINE_RAW_FLEX(struct icmp_bxm, icmp_param, replyopts.opt.opt.__data, 40);
  	/*
  	 *	Too short.
  	 */
@@ -1363,19 +1365,19 @@ static enum skb_drop_reason icmp_timestamp(struct sk_buff *skb)
  	/*
  	 *	Fill in the current time as ms since midnight UT:
  	 */
-	icmp_param.data.times[1] = inet_current_timestamp();
-	icmp_param.data.times[2] = icmp_param.data.times[1];
-
-	BUG_ON(skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4));
-
-	icmp_param.data.icmph	   = *icmp_hdr(skb);
-	icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
-	icmp_param.data.icmph.code = 0;
-	icmp_param.skb		   = skb;
-	icmp_param.offset	   = 0;
-	icmp_param.data_len	   = 0;
-	icmp_param.head_len	   = sizeof(struct icmphdr) + 12;
-	icmp_reply(&icmp_param, skb);
+	icmp_param->data.times[1] = inet_current_timestamp();
+	icmp_param->data.times[2] = icmp_param->data.times[1];
+
+	BUG_ON(skb_copy_bits(skb, 0, &icmp_param->data.times[0], 4));
+
+	icmp_param->data.icmph	   = *icmp_hdr(skb);
+	icmp_param->data.icmph.type = ICMP_TIMESTAMPREPLY;
+	icmp_param->data.icmph.code = 0;
+	icmp_param->skb		   = skb;
+	icmp_param->offset	   = 0;
+	icmp_param->data_len	   = 0;
+	icmp_param->head_len	   = sizeof(struct icmphdr) + 12;
+	icmp_reply(icmp_param, skb);
  	return SKB_NOT_DROPPED_YET;

  out_err:
diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index ff11d3a85a36..e0b20226b0b7 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -1606,7 +1606,7 @@ void ip_send_unicast_reply(struct sock *sk, const struct sock *orig_sk,
  			   const struct ip_reply_arg *arg,
  			   unsigned int len, u64 transmit_time, u32 txhash)
  {
-	struct ip_options_data replyopts;
+	DEFINE_RAW_FLEX(struct ip_options_data, replyopts, opt.opt.__data, 40);
  	struct ipcm_cookie ipc;
  	struct flowi4 fl4;
  	struct rtable *rt = skb_rtable(skb);
@@ -1615,18 +1615,18 @@ void ip_send_unicast_reply(struct sock *sk, const struct sock *orig_sk,
  	int err;
  	int oif;

-	if (__ip_options_echo(net, &replyopts.opt.opt, skb, sopt))
+	if (__ip_options_echo(net, &replyopts->opt.opt, skb, sopt))
  		return;

  	ipcm_init(&ipc);
  	ipc.addr = daddr;
  	ipc.sockc.transmit_time = transmit_time;

-	if (replyopts.opt.opt.optlen) {
-		ipc.opt = &replyopts.opt;
+	if (replyopts->opt.opt.optlen) {
+		ipc.opt = &replyopts->opt;

-		if (replyopts.opt.opt.srr)
-			daddr = replyopts.opt.opt.faddr;
+		if (replyopts->opt.opt.srr)
+			daddr = replyopts->opt.opt.faddr;
  	}

  	oif = arg->bound_dev_if;
diff --git a/net/ipv4/ping.c b/net/ipv4/ping.c
index ad56588107cc..ebc7d24a4a68 100644
--- a/net/ipv4/ping.c
+++ b/net/ipv4/ping.c
@@ -690,6 +690,7 @@ EXPORT_IPV6_MOD_GPL(ping_common_sendmsg);

  static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
  {
+	DEFINE_RAW_FLEX(struct ip_options_data, opt_copy, opt.opt.__data, 40);
  	struct net *net = sock_net(sk);
  	struct flowi4 fl4;
  	struct inet_sock *inet = inet_sk(sk);
@@ -697,7 +698,6 @@ static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
  	struct icmphdr user_icmph;
  	struct pingfakehdr pfh;
  	struct rtable *rt = NULL;
-	struct ip_options_data opt_copy;
  	int free = 0;
  	__be32 saddr, daddr, faddr;
  	u8 scope;
@@ -746,9 +746,9 @@ static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
  		rcu_read_lock();
  		inet_opt = rcu_dereference(inet->inet_opt);
  		if (inet_opt) {
-			memcpy(&opt_copy, inet_opt,
+			memcpy(opt_copy, inet_opt,
  			       sizeof(*inet_opt) + inet_opt->opt.optlen);
-			ipc.opt = &opt_copy.opt;
+			ipc.opt = &opt_copy->opt;
  		}
  		rcu_read_unlock();
  	}
diff --git a/net/ipv4/raw.c b/net/ipv4/raw.c
index 5998c4cc6f47..802ebde0246d 100644
--- a/net/ipv4/raw.c
+++ b/net/ipv4/raw.c
@@ -481,6 +481,7 @@ static int raw_getfrag(void *from, char *to, int offset, int len, int odd,

  static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
  {
+	DEFINE_RAW_FLEX(struct ip_options_data, opt_copy, opt.opt.__data, 40);
  	struct inet_sock *inet = inet_sk(sk);
  	struct net *net = sock_net(sk);
  	struct ipcm_cookie ipc;
@@ -491,7 +492,6 @@ static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
  	__be32 daddr;
  	__be32 saddr;
  	int uc_index, err;
-	struct ip_options_data opt_copy;
  	struct raw_frag_vec rfv;
  	int hdrincl;

@@ -561,9 +561,9 @@ static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
  		rcu_read_lock();
  		inet_opt = rcu_dereference(inet->inet_opt);
  		if (inet_opt) {
-			memcpy(&opt_copy, inet_opt,
+			memcpy(opt_copy, inet_opt,
  			       sizeof(*inet_opt) + inet_opt->opt.optlen);
-			ipc.opt = &opt_copy.opt;
+			ipc.opt = &opt_copy->opt;
  		}
  		rcu_read_unlock();
  	}
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index ffe074cb5865..fa2c4eeed55c 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -1269,6 +1269,7 @@ EXPORT_IPV6_MOD_GPL(udp_cmsg_send);

  int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
  {
+	DEFINE_RAW_FLEX(struct ip_options_data, opt_copy, opt.opt.__data, 40);
  	struct inet_sock *inet = inet_sk(sk);
  	struct udp_sock *up = udp_sk(sk);
  	DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
@@ -1286,7 +1287,6 @@ int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
  	int corkreq = udp_test_bit(CORK, sk) || msg->msg_flags & MSG_MORE;
  	int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
  	struct sk_buff *skb;
-	struct ip_options_data opt_copy;
  	int uc_index;

  	if (len > 0xFFFF)
@@ -1368,9 +1368,9 @@ int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
  		rcu_read_lock();
  		inet_opt = rcu_dereference(inet->inet_opt);
  		if (inet_opt) {
-			memcpy(&opt_copy, inet_opt,
+			memcpy(opt_copy, inet_opt,
  			       sizeof(*inet_opt) + inet_opt->opt.optlen);
-			ipc.opt = &opt_copy.opt;
+			ipc.opt = &opt_copy->opt;
  		}
  		rcu_read_unlock();
  	}
-- 
2.43.0


Thanks
-Gustavo

Re: [PATCH][next] net: inet_sock.h: Avoid thousands of -Wflex-array-member-not-at-end warnings

[Jakub Kicinski]

On Tue, 18 Nov 2025 13:36:41 +0900 Gustavo A. R. Silva wrote:
> Not sure what you mean by "nasty" and "no clear benefit", but here is an
> alternative. See it as RFC:

Meaning of the words per dictionary definition.

> https://git.kernel.org/pub/scm/linux/kernel/git/gustavoars/linux.git/commit/?h=testing/wfamnae-next20251117&id=e0547082214e61b1db0f5068da0daa3d11f992a5
> 
> [PATCH RFC][next] ipv4/inet_sock.h: Avoid thousands of -Wflex-array-member-not-at-end warnings
> 
> Use DEFINE_RAW_FLEX() to avoid thousands of -Wflex-array-member-not-at-end
> warnings.
> 
> Remove fixed-size array char data[40]; from struct ip_options_data, so
> that flexible-array member struct ip_options_rcu::opt.__data[] ends last
> in this (and other) structure(s).
> 
> Compensate for this by using the DEFINE_RAW_FLEX() helper to declare each
> on-stack struct instance that contains struct ip_options_data as a member.

This is much better, thanks!

Re: [PATCH][next] net: inet_sock.h: Avoid thousands of -Wflex-array-member-not-at-end warnings

[Kees Cook]

On Tue, Nov 18, 2025 at 01:36:41PM +0900, Gustavo A. R. Silva wrote:
>  struct ip_options_data {
>  	struct ip_options_rcu	opt;
> -	char			data[40];
>  };

If this just contains struct ip_options_rcu now, can ip_options_data
just be removed entirely?

-Kees

-- 
Kees Cook