There are totally 9 ndo_bridge_setlink handlers in the current kernel,
which are 1) bnxt_bridge_setlink, 2) be_ndo_bridge_setlink 3)
i40e_ndo_bridge_setlink 4) ice_bridge_setlink 5)
ixgbe_ndo_bridge_setlink 6) mlx5e_bridge_setlink 7)
nfp_net_bridge_setlink 8) qeth_l2_bridge_setlink 9) br_setlink.
By investigating the code, we find that 1-7 parse and use nlattr
IFLA_BRIDGE_MODE but 3 and 4 forget to do the nla_len check. This can
lead to an out-of-attribute read and allow a malformed nlattr (e.g.,
length 0) to be viewed as a 2 byte integer.
To avoid such issues, also for other ndo_bridge_setlink handlers in the
future. This patch adds the nla_len check in rtnl_bridge_setlink and
does an early error return if length mismatches. To make it works, the
break is removed from the parsing for IFLA_BRIDGE_FLAGS to make sure
this nla_for_each_nested iterates every attribute.
Fixes: b1edc14a3fbf ("ice: Implement ice_bridge_getlink and ice_bridge_setlink")
Fixes: 51616018dd1b ("i40e: Add support for getlink, setlink ndo ops")
Suggested-by: Jakub Kicinski <[email protected]>
Signed-off-by: Lin Ma <[email protected]>
---
V1 -> V2: removes the break in parsing for IFLA_BRIDGE_FLAGS suggested
by Hangbin Liu <[email protected]>
net/core/rtnetlink.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 3ad4e030846d..aef25aa5cf1d 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -5140,13 +5140,17 @@ static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
if (br_spec) {
nla_for_each_nested(attr, br_spec, rem) {
- if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
+ if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !have_flags) {
if (nla_len(attr) < sizeof(flags))
return -EINVAL;
have_flags = true;
flags = nla_get_u16(attr);
- break;
+ }
+
+ if (nla_type(attr) == IFLA_BRIDGE_MODE) {
+ if (nla_len(attr) < sizeof(u16))
+ return -EINVAL;
}
}
}
--
2.17.1
Hi Ma Lin,
Please add the target branch in your subject. e.g. [PATCHv2 net]
On Tue, Jul 25, 2023 at 01:57:06PM +0800, Lin Ma wrote:
> There are totally 9 ndo_bridge_setlink handlers in the current kernel,
> which are 1) bnxt_bridge_setlink, 2) be_ndo_bridge_setlink 3)
> i40e_ndo_bridge_setlink 4) ice_bridge_setlink 5)
> ixgbe_ndo_bridge_setlink 6) mlx5e_bridge_setlink 7)
> nfp_net_bridge_setlink 8) qeth_l2_bridge_setlink 9) br_setlink.
>
> By investigating the code, we find that 1-7 parse and use nlattr
> IFLA_BRIDGE_MODE but 3 and 4 forget to do the nla_len check. This can
> lead to an out-of-attribute read and allow a malformed nlattr (e.g.,
> length 0) to be viewed as a 2 byte integer.
>
> To avoid such issues, also for other ndo_bridge_setlink handlers in the
> future. This patch adds the nla_len check in rtnl_bridge_setlink and
> does an early error return if length mismatches. To make it works, the
> break is removed from the parsing for IFLA_BRIDGE_FLAGS to make sure
> this nla_for_each_nested iterates every attribute.
Since you have checked the length in rtnl_bridge_setlink(). Can we remove
the check in the driver handlers to avoid duplicate code? You can hold on
this update and see if others have different opinion.
Thanks
Hangbin
>
> Fixes: b1edc14a3fbf ("ice: Implement ice_bridge_getlink and ice_bridge_setlink")
> Fixes: 51616018dd1b ("i40e: Add support for getlink, setlink ndo ops")
> Suggested-by: Jakub Kicinski <[email protected]>
> Signed-off-by: Lin Ma <[email protected]>
> ---
> V1 -> V2: removes the break in parsing for IFLA_BRIDGE_FLAGS suggested
> by Hangbin Liu <[email protected]>
>
> net/core/rtnetlink.c | 8 ++++++--
> 1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
> index 3ad4e030846d..aef25aa5cf1d 100644
> --- a/net/core/rtnetlink.c
> +++ b/net/core/rtnetlink.c
> @@ -5140,13 +5140,17 @@ static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
> br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
> if (br_spec) {
> nla_for_each_nested(attr, br_spec, rem) {
> - if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
> + if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !have_flags) {
> if (nla_len(attr) < sizeof(flags))
> return -EINVAL;
>
> have_flags = true;
> flags = nla_get_u16(attr);
> - break;
> + }
> +
> + if (nla_type(attr) == IFLA_BRIDGE_MODE) {
> + if (nla_len(attr) < sizeof(u16))
> + return -EINVAL;
> }
> }
> }
> --
> 2.17.1
>
On 7/25/23 08:57, Lin Ma wrote:
> There are totally 9 ndo_bridge_setlink handlers in the current kernel,
> which are 1) bnxt_bridge_setlink, 2) be_ndo_bridge_setlink 3)
> i40e_ndo_bridge_setlink 4) ice_bridge_setlink 5)
> ixgbe_ndo_bridge_setlink 6) mlx5e_bridge_setlink 7)
> nfp_net_bridge_setlink 8) qeth_l2_bridge_setlink 9) br_setlink.
>
> By investigating the code, we find that 1-7 parse and use nlattr
> IFLA_BRIDGE_MODE but 3 and 4 forget to do the nla_len check. This can
> lead to an out-of-attribute read and allow a malformed nlattr (e.g.,
> length 0) to be viewed as a 2 byte integer.
>
> To avoid such issues, also for other ndo_bridge_setlink handlers in the
> future. This patch adds the nla_len check in rtnl_bridge_setlink and
> does an early error return if length mismatches. To make it works, the
> break is removed from the parsing for IFLA_BRIDGE_FLAGS to make sure
> this nla_for_each_nested iterates every attribute.
>
> Fixes: b1edc14a3fbf ("ice: Implement ice_bridge_getlink and ice_bridge_setlink")
> Fixes: 51616018dd1b ("i40e: Add support for getlink, setlink ndo ops")
> Suggested-by: Jakub Kicinski <[email protected]>
> Signed-off-by: Lin Ma <[email protected]>
> ---
> V1 -> V2: removes the break in parsing for IFLA_BRIDGE_FLAGS suggested
> by Hangbin Liu <[email protected]>
>
> net/core/rtnetlink.c | 8 ++++++--
> 1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
> index 3ad4e030846d..aef25aa5cf1d 100644
> --- a/net/core/rtnetlink.c
> +++ b/net/core/rtnetlink.c
> @@ -5140,13 +5140,17 @@ static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
> br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
> if (br_spec) {
> nla_for_each_nested(attr, br_spec, rem) {
> - if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
> + if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !have_flags) {
> if (nla_len(attr) < sizeof(flags))
> return -EINVAL;
>
> have_flags = true;
> flags = nla_get_u16(attr);
> - break;
> + }
> +
> + if (nla_type(attr) == IFLA_BRIDGE_MODE) {
> + if (nla_len(attr) < sizeof(u16))
> + return -EINVAL;
> }
> }
> }
Patch looks good now, you should probably remove the extra checks done
by each driver that are now unnecessary (net-next material). As Hangbin
commented you should target this fix at -net, with that:
Acked-by: Nikolay Aleksandrov <[email protected]>
Hi Nikolay,
>
> Patch looks good now, you should probably remove the extra checks done
> by each driver that are now unnecessary (net-next material). As Hangbin
> commented you should target this fix at -net, with that:
>
> Acked-by: Nikolay Aleksandrov <[email protected]>
Cool, I agree with Hangbin that another patch which removes the redundant
checks in driver is needed.
But I have a simple question. I will send this patch to net one and another
to net-next one. How can I ensure the latter one depends on the former one?
Or should I send a patch series to net-next that contains the former one :)
(I currently choose the method 2 and please let me know if I do this wrong)
Regards
Lin
On Wed, 26 Jul 2023 15:49:02 +0800 (GMT+08:00) Lin Ma wrote:
> Cool, I agree with Hangbin that another patch which removes the redundant
> checks in driver is needed.
>
> But I have a simple question. I will send this patch to net one and another
> to net-next one. How can I ensure the latter one depends on the former one?
> Or should I send a patch series to net-next that contains the former one :)
> (I currently choose the method 2 and please let me know if I do this wrong)
You'll need to wait for the patch to propagate before posting.
Our trees merge each Thursday, so if you post on Friday the fix
should be in net-next.
Hi Jakub,
>
> You'll need to wait for the patch to propagate before posting.
> Our trees merge each Thursday, so if you post on Friday the fix
> should be in net-next.
Cool, I understand now. Thanks!
Regards
Lin