[PATCH net-next v4 0/2] Flexible array for ip tunnel options

[PATCH net-next v4 0/2] Flexible array for ip tunnel options

[Gal Pressman]

Remove the hidden assumption that options are allocated at the end of
the struct, and teach the compiler about them using a flexible array.

First patch is converting hard-coded 'info + 1' to use ip_tunnel_info()
helper.
Second patch adds the 'options' flexible array and changes the helper to
use it.

Changelog -
v3=>v4: https://lore.kernel.org/netdev/20250217202503.265318-1-gal@nvidia.com/
* Remove the casts in first patch (Jakub).

v2->v3: https://lore.kernel.org/netdev/20250212140953.107533-1-gal@nvidia.com/
* Add a precursory patch to convert hard-coded user of options.
* Keep ip_tunnel_info() macro (Alexander).
* Use __aligned_largest (Alexander).

v1->v2: https://lore.kernel.org/netdev/20250209101853.15828-1-gal@nvidia.com/
* Remove change in struct layout, align 'options' field explicitly (Ilya, Kees, Jakub).
* Change allocation I missed in v1 in metadata_dst_alloc_percpu().

Thanks,
Gal

Gal Pressman (2):
  ip_tunnel: Use ip_tunnel_info() helper instead of 'info + 1'
  net: Add options as a flexible array to struct ip_tunnel_info

 include/net/dst_metadata.h | 7 ++-----
 include/net/ip_tunnels.h   | 7 ++++---
 net/core/dst.c             | 6 ++++--
 net/sched/act_tunnel_key.c | 8 ++++----
 4 files changed, 14 insertions(+), 14 deletions(-)

-- 
2.40.1

[PATCH net-next v4 1/2] ip_tunnel: Use ip_tunnel_info() helper instead of 'info + 1'

[Gal Pressman]

Tunnel options should not be accessed directly, use the ip_tunnel_info()
accessor instead.

Signed-off-by: Gal Pressman <gal@nvidia.com>
---
 include/net/ip_tunnels.h   | 2 +-
 net/sched/act_tunnel_key.c | 8 ++++----
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/include/net/ip_tunnels.h b/include/net/ip_tunnels.h
index 1aa31bdb2b31..7b54cea5de27 100644
--- a/include/net/ip_tunnels.h
+++ b/include/net/ip_tunnels.h
@@ -650,7 +650,7 @@ static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
 static inline void ip_tunnel_info_opts_get(void *to,
 					   const struct ip_tunnel_info *info)
 {
-	memcpy(to, info + 1, info->options_len);
+	memcpy(to, ip_tunnel_info_opts(info), info->options_len);
 }
 
 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
diff --git a/net/sched/act_tunnel_key.c b/net/sched/act_tunnel_key.c
index af7c99845948..ae5dea7c48a8 100644
--- a/net/sched/act_tunnel_key.c
+++ b/net/sched/act_tunnel_key.c
@@ -571,8 +571,8 @@ static void tunnel_key_release(struct tc_action *a)
 static int tunnel_key_geneve_opts_dump(struct sk_buff *skb,
 				       const struct ip_tunnel_info *info)
 {
+	const u8 *src = ip_tunnel_info_opts(info);
 	int len = info->options_len;
-	u8 *src = (u8 *)(info + 1);
 	struct nlattr *start;
 
 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_GENEVE);
@@ -580,7 +580,7 @@ static int tunnel_key_geneve_opts_dump(struct sk_buff *skb,
 		return -EMSGSIZE;
 
 	while (len > 0) {
-		struct geneve_opt *opt = (struct geneve_opt *)src;
+		const struct geneve_opt *opt = (const struct geneve_opt *)src;
 
 		if (nla_put_be16(skb, TCA_TUNNEL_KEY_ENC_OPT_GENEVE_CLASS,
 				 opt->opt_class) ||
@@ -603,7 +603,7 @@ static int tunnel_key_geneve_opts_dump(struct sk_buff *skb,
 static int tunnel_key_vxlan_opts_dump(struct sk_buff *skb,
 				      const struct ip_tunnel_info *info)
 {
-	struct vxlan_metadata *md = (struct vxlan_metadata *)(info + 1);
+	const struct vxlan_metadata *md = ip_tunnel_info_opts(info);
 	struct nlattr *start;
 
 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_VXLAN);
@@ -622,7 +622,7 @@ static int tunnel_key_vxlan_opts_dump(struct sk_buff *skb,
 static int tunnel_key_erspan_opts_dump(struct sk_buff *skb,
 				       const struct ip_tunnel_info *info)
 {
-	struct erspan_metadata *md = (struct erspan_metadata *)(info + 1);
+	const struct erspan_metadata *md = ip_tunnel_info_opts(info);
 	struct nlattr *start;
 
 	start = nla_nest_start_noflag(skb, TCA_TUNNEL_KEY_ENC_OPTS_ERSPAN);
-- 
2.40.1

Re: [PATCH net-next v4 1/2] ip_tunnel: Use ip_tunnel_info() helper instead of 'info + 1'

[Kees Cook]

On Wed, Feb 19, 2025 at 04:32:55PM +0200, Gal Pressman wrote:
> Tunnel options should not be accessed directly, use the ip_tunnel_info()
> accessor instead.
> 
> Signed-off-by: Gal Pressman <gal@nvidia.com>

Reviewed-by: Kees Cook <kees@kernel.org>

-- 
Kees Cook

[PATCH net-next v4 2/2] net: Add options as a flexible array to struct ip_tunnel_info

[Gal Pressman]

Remove the hidden assumption that options are allocated at the end of
the struct, and teach the compiler about them using a flexible array.

With this, we can revert the unsafe_memcpy() call we have in
tun_dst_unclone() [1], and resolve the false field-spanning write
warning caused by the memcpy() in ip_tunnel_info_opts_set().

The layout of struct ip_tunnel_info remains the same with this patch.
Before this patch, there was an implicit padding at the end of the
struct, options would be written at 'info + 1' which is after the
padding.
This will remain the same as this patch explicitly aligns 'options'.
The alignment is needed as the options are later casted to different
structs, and might result in unaligned memory access.

Pahole output before this patch:
struct ip_tunnel_info {
    struct ip_tunnel_key       key;                  /*     0    64 */

    /* XXX last struct has 1 byte of padding */

    /* --- cacheline 1 boundary (64 bytes) --- */
    struct ip_tunnel_encap     encap;                /*    64     8 */
    struct dst_cache           dst_cache;            /*    72    16 */
    u8                         options_len;          /*    88     1 */
    u8                         mode;                 /*    89     1 */

    /* size: 96, cachelines: 2, members: 5 */
    /* padding: 6 */
    /* paddings: 1, sum paddings: 1 */
    /* last cacheline: 32 bytes */
};

Pahole output after this patch:
struct ip_tunnel_info {
    struct ip_tunnel_key       key;                  /*     0    64 */

    /* XXX last struct has 1 byte of padding */

    /* --- cacheline 1 boundary (64 bytes) --- */
    struct ip_tunnel_encap     encap;                /*    64     8 */
    struct dst_cache           dst_cache;            /*    72    16 */
    u8                         options_len;          /*    88     1 */
    u8                         mode;                 /*    89     1 */

    /* XXX 6 bytes hole, try to pack */

    u8                         options[] __attribute__((__aligned__(16))); /*    96     0 */

    /* size: 96, cachelines: 2, members: 6 */
    /* sum members: 90, holes: 1, sum holes: 6 */
    /* paddings: 1, sum paddings: 1 */
    /* forced alignments: 1, forced holes: 1, sum forced holes: 6 */
    /* last cacheline: 32 bytes */
} __attribute__((__aligned__(16)));

[1] Commit 13cfd6a6d7ac ("net: Silence false field-spanning write warning in metadata_dst memcpy")

Link: https://lore.kernel.org/all/53D1D353-B8F6-4ADC-8F29-8C48A7C9C6F1@kernel.org/
Suggested-by: Kees Cook <kees@kernel.org>
Reviewed-by: Cosmin Ratiu <cratiu@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Gal Pressman <gal@nvidia.com>
---
 include/net/dst_metadata.h | 7 ++-----
 include/net/ip_tunnels.h   | 5 +++--
 net/core/dst.c             | 6 ++++--
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/include/net/dst_metadata.h b/include/net/dst_metadata.h
index 84c15402931c..4160731dcb6e 100644
--- a/include/net/dst_metadata.h
+++ b/include/net/dst_metadata.h
@@ -163,11 +163,8 @@ static inline struct metadata_dst *tun_dst_unclone(struct sk_buff *skb)
 	if (!new_md)
 		return ERR_PTR(-ENOMEM);
 
-	unsafe_memcpy(&new_md->u.tun_info, &md_dst->u.tun_info,
-		      sizeof(struct ip_tunnel_info) + md_size,
-		      /* metadata_dst_alloc() reserves room (md_size bytes) for
-		       * options right after the ip_tunnel_info struct.
-		       */);
+	memcpy(&new_md->u.tun_info, &md_dst->u.tun_info,
+	       sizeof(struct ip_tunnel_info) + md_size);
 #ifdef CONFIG_DST_CACHE
 	/* Unclone the dst cache if there is one */
 	if (new_md->u.tun_info.dst_cache.cache) {
diff --git a/include/net/ip_tunnels.h b/include/net/ip_tunnels.h
index 7b54cea5de27..e041e4865373 100644
--- a/include/net/ip_tunnels.h
+++ b/include/net/ip_tunnels.h
@@ -95,8 +95,8 @@ struct ip_tunnel_encap {
 
 #define ip_tunnel_info_opts(info)				\
 	_Generic(info,						\
-		 const struct ip_tunnel_info * : ((const void *)((info) + 1)),\
-		 struct ip_tunnel_info * : ((void *)((info) + 1))\
+		 const struct ip_tunnel_info * : ((const void *)(info)->options),\
+		 struct ip_tunnel_info * : ((void *)(info)->options)\
 	)
 
 struct ip_tunnel_info {
@@ -107,6 +107,7 @@ struct ip_tunnel_info {
 #endif
 	u8			options_len;
 	u8			mode;
+	u8			options[] __aligned_largest __counted_by(options_len);
 };
 
 /* 6rd prefix/relay information */
diff --git a/net/core/dst.c b/net/core/dst.c
index 9552a90d4772..c99b95cf9cbb 100644
--- a/net/core/dst.c
+++ b/net/core/dst.c
@@ -286,7 +286,8 @@ struct metadata_dst *metadata_dst_alloc(u8 optslen, enum metadata_type type,
 {
 	struct metadata_dst *md_dst;
 
-	md_dst = kmalloc(sizeof(*md_dst) + optslen, flags);
+	md_dst = kmalloc(struct_size(md_dst, u.tun_info.options, optslen),
+			 flags);
 	if (!md_dst)
 		return NULL;
 
@@ -314,7 +315,8 @@ metadata_dst_alloc_percpu(u8 optslen, enum metadata_type type, gfp_t flags)
 	int cpu;
 	struct metadata_dst __percpu *md_dst;
 
-	md_dst = __alloc_percpu_gfp(sizeof(struct metadata_dst) + optslen,
+	md_dst = __alloc_percpu_gfp(struct_size(md_dst, u.tun_info.options,
+						optslen),
 				    __alignof__(struct metadata_dst), flags);
 	if (!md_dst)
 		return NULL;
-- 
2.40.1

Re: [PATCH net-next v4 2/2] net: Add options as a flexible array to struct ip_tunnel_info

[Kees Cook]

On Wed, Feb 19, 2025 at 04:32:56PM +0200, Gal Pressman wrote:
> Remove the hidden assumption that options are allocated at the end of
> the struct, and teach the compiler about them using a flexible array.
> 
> With this, we can revert the unsafe_memcpy() call we have in
> tun_dst_unclone() [1], and resolve the false field-spanning write
> warning caused by the memcpy() in ip_tunnel_info_opts_set().
> 
> The layout of struct ip_tunnel_info remains the same with this patch.
> Before this patch, there was an implicit padding at the end of the
> struct, options would be written at 'info + 1' which is after the
> padding.
> This will remain the same as this patch explicitly aligns 'options'.
> The alignment is needed as the options are later casted to different
> structs, and might result in unaligned memory access.
> 
> Pahole output before this patch:
> struct ip_tunnel_info {
>     struct ip_tunnel_key       key;                  /*     0    64 */
> 
>     /* XXX last struct has 1 byte of padding */
> 
>     /* --- cacheline 1 boundary (64 bytes) --- */
>     struct ip_tunnel_encap     encap;                /*    64     8 */
>     struct dst_cache           dst_cache;            /*    72    16 */
>     u8                         options_len;          /*    88     1 */
>     u8                         mode;                 /*    89     1 */
> 
>     /* size: 96, cachelines: 2, members: 5 */
>     /* padding: 6 */
>     /* paddings: 1, sum paddings: 1 */
>     /* last cacheline: 32 bytes */
> };
> 
> Pahole output after this patch:
> struct ip_tunnel_info {
>     struct ip_tunnel_key       key;                  /*     0    64 */
> 
>     /* XXX last struct has 1 byte of padding */
> 
>     /* --- cacheline 1 boundary (64 bytes) --- */
>     struct ip_tunnel_encap     encap;                /*    64     8 */
>     struct dst_cache           dst_cache;            /*    72    16 */
>     u8                         options_len;          /*    88     1 */
>     u8                         mode;                 /*    89     1 */
> 
>     /* XXX 6 bytes hole, try to pack */
> 
>     u8                         options[] __attribute__((__aligned__(16))); /*    96     0 */
> 
>     /* size: 96, cachelines: 2, members: 6 */
>     /* sum members: 90, holes: 1, sum holes: 6 */
>     /* paddings: 1, sum paddings: 1 */
>     /* forced alignments: 1, forced holes: 1, sum forced holes: 6 */
>     /* last cacheline: 32 bytes */
> } __attribute__((__aligned__(16)));
> 
> [1] Commit 13cfd6a6d7ac ("net: Silence false field-spanning write warning in metadata_dst memcpy")
> 
> Link: https://lore.kernel.org/all/53D1D353-B8F6-4ADC-8F29-8C48A7C9C6F1@kernel.org/
> Suggested-by: Kees Cook <kees@kernel.org>
> Reviewed-by: Cosmin Ratiu <cratiu@nvidia.com>
> Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
> Signed-off-by: Gal Pressman <gal@nvidia.com>

Thanks for these updates and the pahole output. :)

Reviewed-by: Kees Cook <kees@kernel.org>

-- 
Kees Cook

Re: [PATCH net-next v4 0/2] Flexible array for ip tunnel options

[patchwork-bot+netdevbpf]

Hello:

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

On Wed, 19 Feb 2025 16:32:54 +0200 you wrote:
> Remove the hidden assumption that options are allocated at the end of
> the struct, and teach the compiler about them using a flexible array.
> 
> First patch is converting hard-coded 'info + 1' to use ip_tunnel_info()
> helper.
> Second patch adds the 'options' flexible array and changes the helper to
> use it.
> 
> [...]

Here is the summary with links:
  - [net-next,v4,1/2] ip_tunnel: Use ip_tunnel_info() helper instead of 'info + 1'
    https://git.kernel.org/netdev/net-next/c/ba3fa6e8c1eb
  - [net-next,v4,2/2] net: Add options as a flexible array to struct ip_tunnel_info
    https://git.kernel.org/netdev/net-next/c/bb5e62f2d547

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