[bpf-next] bpf: Make sure internal and UAPI bpf_redirect flags don't overlap

Commit Message

Toke Høiland-Jørgensen Sept. 20, 2024, 12:56 p.m. UTC

The bpf_redirect_info is shared between the SKB and XDP redirect paths,
and the two paths use the same numeric flag values in the ri->flags
field (specifically, BPF_F_BROADCAST == BPF_F_NEXTHOP). This means that
if skb bpf_redirect_neigh() is used with a non-NULL params argument and,
subsequently, an XDP redirect is performed using the same
bpf_redirect_info struct, the XDP path will get confused and end up
crashing, which syzbot managed to trigger.

With the stack-allocated bpf_redirect_info, the structure is no longer
shared between the SKB and XDP paths, so the crash doesn't happen
anymore. However, different code paths using identically-numbered flag
values in the same struct field still seems like a bit of a mess, so
this patch cleans that up by moving the flag definitions together and
redefining the three flags in BPF_F_REDIRECT_INTERNAL to not overlap
with the flags used for XDP. It also adds a BUILD_BUG_ON() check to make
sure the overlap is not re-introduced by mistake.

Fixes: e624d4ed4aa8 ("xdp: Extend xdp_redirect_map with broadcast support")
Reported-by: [e-mail]
Closes: https://syzkaller.appspot.com/bug?extid=cca39e6e84a367a7e6f6
Signed-off-by: Toke Høiland-Jørgensen <[e-mail]>
---
 include/uapi/linux/bpf.h | 14 ++++++--------
 net/core/filter.c        |  8 +++++---
 2 files changed, 11 insertions(+), 11 deletions(-)

Comments

Daniel Borkmann Oct. 1, 2024, 7:49 p.m. UTC | #1

On 9/20/24 2:56 PM, Toke Høiland-Jørgensen wrote:
> The bpf_redirect_info is shared between the SKB and XDP redirect paths,
> and the two paths use the same numeric flag values in the ri->flags
> field (specifically, BPF_F_BROADCAST == BPF_F_NEXTHOP). This means that
> if skb bpf_redirect_neigh() is used with a non-NULL params argument and,
> subsequently, an XDP redirect is performed using the same
> bpf_redirect_info struct, the XDP path will get confused and end up
> crashing, which syzbot managed to trigger.
>
> With the stack-allocated bpf_redirect_info, the structure is no longer
> shared between the SKB and XDP paths, so the crash doesn't happen
> anymore. However, different code paths using identically-numbered flag
> values in the same struct field still seems like a bit of a mess, so
> this patch cleans that up by moving the flag definitions together and
> redefining the three flags in BPF_F_REDIRECT_INTERNAL to not overlap
> with the flags used for XDP. It also adds a BUILD_BUG_ON() check to make
> sure the overlap is not re-introduced by mistake.
>
> Fixes: e624d4ed4aa8 ("xdp: Extend xdp_redirect_map with broadcast support")
> Reported-by: [e-mail]
> Closes: https://syzkaller.appspot.com/bug?extid=cca39e6e84a367a7e6f6
> Signed-off-by: Toke Høiland-Jørgensen <[e-mail]>
> ---
>   include/uapi/linux/bpf.h | 14 ++++++--------
>   net/core/filter.c        |  8 +++++---
>   2 files changed, 11 insertions(+), 11 deletions(-)
Lgtm, applied, thanks! I also added a tools header sync.I took this into 
bpf tree, so that stable can pick it up.
> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> index c3a5728db115..0c6154272ab3 100644
> --- a/include/uapi/linux/bpf.h
> +++ b/include/uapi/linux/bpf.h
> @@ -6047,11 +6047,6 @@ enum {
>   	BPF_F_MARK_ENFORCE		= (1ULL << 6),
>   };
>   
> -/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
> -enum {
> -	BPF_F_INGRESS			= (1ULL << 0),
> -};
> -
>   /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
>   enum {
>   	BPF_F_TUNINFO_IPV6		= (1ULL << 0),
> @@ -6198,11 +6193,14 @@ enum {
>   	BPF_F_BPRM_SECUREEXEC	= (1ULL << 0),
>   };
>   
> -/* Flags for bpf_redirect_map helper */
> +/* Flags for bpf_redirect and bpf_redirect_map helpers */
>   enum {
> -	BPF_F_BROADCAST		= (1ULL << 3),
> -	BPF_F_EXCLUDE_INGRESS	= (1ULL << 4),
> +	BPF_F_INGRESS		= (1ULL << 0), /* used for skb path */
> +	BPF_F_BROADCAST		= (1ULL << 3), /* used for XDP path */
> +	BPF_F_EXCLUDE_INGRESS	= (1ULL << 4), /* used for XDP path */
>   };
> +#define BPF_F_REDIRECT_ALL_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS)
> +
>   
Also fixed up extra newline.

Thanks,
Daniel

[e-mail] Oct. 1, 2024, 7:50 p.m. UTC | #2

Hello:

This patch was applied to bpf/bpf.git (master)
by Daniel Borkmann <[e-mail]>:

On Fri, 20 Sep 2024 14:56:24 +0200 you wrote:
> The bpf_redirect_info is shared between the SKB and XDP redirect paths,
> and the two paths use the same numeric flag values in the ri->flags
> field (specifically, BPF_F_BROADCAST == BPF_F_NEXTHOP). This means that
> if skb bpf_redirect_neigh() is used with a non-NULL params argument and,
> subsequently, an XDP redirect is performed using the same
> bpf_redirect_info struct, the XDP path will get confused and end up
> crashing, which syzbot managed to trigger.
> 
> [...]

Here is the summary with links:
  - [bpf-next] bpf: Make sure internal and UAPI bpf_redirect flags don't overlap
    https://git.kernel.org/bpf/bpf/c/09d88791c7cd

You are awesome, thank you!

Toke Høiland-Jørgensen Oct. 2, 2024, 8:35 a.m. UTC | #3

Daniel Borkmann <[e-mail]> writes:

> On 9/20/24 2:56 PM, Toke Høiland-Jørgensen wrote:
>> The bpf_redirect_info is shared between the SKB and XDP redirect paths,
>> and the two paths use the same numeric flag values in the ri->flags
>> field (specifically, BPF_F_BROADCAST == BPF_F_NEXTHOP). This means that
>> if skb bpf_redirect_neigh() is used with a non-NULL params argument and,
>> subsequently, an XDP redirect is performed using the same
>> bpf_redirect_info struct, the XDP path will get confused and end up
>> crashing, which syzbot managed to trigger.
>>
>> With the stack-allocated bpf_redirect_info, the structure is no longer
>> shared between the SKB and XDP paths, so the crash doesn't happen
>> anymore. However, different code paths using identically-numbered flag
>> values in the same struct field still seems like a bit of a mess, so
>> this patch cleans that up by moving the flag definitions together and
>> redefining the three flags in BPF_F_REDIRECT_INTERNAL to not overlap
>> with the flags used for XDP. It also adds a BUILD_BUG_ON() check to make
>> sure the overlap is not re-introduced by mistake.
>>
>> Fixes: e624d4ed4aa8 ("xdp: Extend xdp_redirect_map with broadcast support")
>> Reported-by: [e-mail]
>> Closes: https://syzkaller.appspot.com/bug?extid=cca39e6e84a367a7e6f6
>> Signed-off-by: Toke Høiland-Jørgensen <[e-mail]>
>> ---
>>   include/uapi/linux/bpf.h | 14 ++++++--------
>>   net/core/filter.c        |  8 +++++---
>>   2 files changed, 11 insertions(+), 11 deletions(-)
> Lgtm, applied, thanks! I also added a tools header sync.I took this into 
> bpf tree, so that stable can pick it up.

Great! Thanks for the fixups :)

-Toke

Patch

diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index c3a5728db115..0c6154272ab3 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -6047,11 +6047,6 @@  enum {
 	BPF_F_MARK_ENFORCE		= (1ULL << 6),
 };
 
-/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
-enum {
-	BPF_F_INGRESS			= (1ULL << 0),
-};
-
 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
 enum {
 	BPF_F_TUNINFO_IPV6		= (1ULL << 0),
@@ -6198,11 +6193,14 @@  enum {
 	BPF_F_BPRM_SECUREEXEC	= (1ULL << 0),
 };
 
-/* Flags for bpf_redirect_map helper */
+/* Flags for bpf_redirect and bpf_redirect_map helpers */
 enum {
-	BPF_F_BROADCAST		= (1ULL << 3),
-	BPF_F_EXCLUDE_INGRESS	= (1ULL << 4),
+	BPF_F_INGRESS		= (1ULL << 0), /* used for skb path */
+	BPF_F_BROADCAST		= (1ULL << 3), /* used for XDP path */
+	BPF_F_EXCLUDE_INGRESS	= (1ULL << 4), /* used for XDP path */
 };
+#define BPF_F_REDIRECT_ALL_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS)
+
 
 #define __bpf_md_ptr(type, name)	\
 union {					\
diff --git a/net/core/filter.c b/net/core/filter.c
index e4a4454df5f9..db99f2b38e06 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -2437,9 +2437,9 @@  static int __bpf_redirect_neigh(struct sk_buff *skb, struct net_device *dev,
 
 /* Internal, non-exposed redirect flags. */
 enum {
-	BPF_F_NEIGH	= (1ULL << 1),
-	BPF_F_PEER	= (1ULL << 2),
-	BPF_F_NEXTHOP	= (1ULL << 3),
+	BPF_F_NEIGH	= (1ULL << 16),
+	BPF_F_PEER	= (1ULL << 17),
+	BPF_F_NEXTHOP	= (1ULL << 18),
 #define BPF_F_REDIRECT_INTERNAL	(BPF_F_NEIGH | BPF_F_PEER | BPF_F_NEXTHOP)
 };
 
@@ -2449,6 +2449,8 @@  BPF_CALL_3(bpf_clone_redirect, struct sk_buff *, skb, u32, ifindex, u64, flags)
 	struct sk_buff *clone;
 	int ret;
 
+	BUILD_BUG_ON(BPF_F_REDIRECT_INTERNAL & BPF_F_REDIRECT_ALL_FLAGS);
+
 	if (unlikely(flags & (~(BPF_F_INGRESS) | BPF_F_REDIRECT_INTERNAL)))
 		return -EINVAL;