]> git.ipfire.org Git - thirdparty/kernel/linux.git/commitdiff
netkit: Allow for configuring needed_{head,tail}room
authorDaniel Borkmann <daniel@iogearbox.net>
Fri, 20 Dec 2024 23:46:56 +0000 (00:46 +0100)
committerDaniel Borkmann <daniel@iogearbox.net>
Mon, 6 Jan 2025 08:48:49 +0000 (09:48 +0100)
Allow the user to configure needed_{head,tail}room for both netkit
devices. The idea is similar to 163e529200af ("veth: implement
ndo_set_rx_headroom") with the difference that the two parameters
can be specified upon device creation. By default the current behavior
stays as is which is needed_{head,tail}room is 0.

In case of Cilium, for example, the netkit devices are not enslaved
into a bridge or openvswitch device (rather, BPF-based redirection
is used out of tcx), and as such these parameters are not propagated
into the Pod's netns via peer device.

Given Cilium can run in vxlan/geneve tunneling mode (needed_headroom)
and/or be used in combination with WireGuard (needed_{head,tail}room),
allow the Cilium CNI plugin to specify these two upon netkit device
creation.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://lore.kernel.org/bpf/20241220234658.490686-1-daniel@iogearbox.net
drivers/net/netkit.c
include/uapi/linux/if_link.h
tools/include/uapi/linux/if_link.h

index c1d881dc6409a4dae9bd99562a4f11e93cce3d68..1e1b00756be7d974623bf1062725dac589a73b84 100644 (file)
@@ -338,6 +338,7 @@ static int netkit_new_link(struct net *peer_net, struct net_device *dev,
        enum netkit_scrub scrub_peer = NETKIT_SCRUB_DEFAULT;
        enum netkit_mode mode = NETKIT_L3;
        unsigned char ifname_assign_type;
+       u16 headroom = 0, tailroom = 0;
        struct ifinfomsg *ifmp = NULL;
        struct net_device *peer;
        char ifname[IFNAMSIZ];
@@ -371,6 +372,10 @@ static int netkit_new_link(struct net *peer_net, struct net_device *dev,
                        if (err < 0)
                                return err;
                }
+               if (data[IFLA_NETKIT_HEADROOM])
+                       headroom = nla_get_u16(data[IFLA_NETKIT_HEADROOM]);
+               if (data[IFLA_NETKIT_TAILROOM])
+                       tailroom = nla_get_u16(data[IFLA_NETKIT_TAILROOM]);
        }
 
        if (ifmp && tbp[IFLA_IFNAME]) {
@@ -390,6 +395,14 @@ static int netkit_new_link(struct net *peer_net, struct net_device *dev,
                return PTR_ERR(peer);
 
        netif_inherit_tso_max(peer, dev);
+       if (headroom) {
+               peer->needed_headroom = headroom;
+               dev->needed_headroom = headroom;
+       }
+       if (tailroom) {
+               peer->needed_tailroom = tailroom;
+               dev->needed_tailroom = tailroom;
+       }
 
        if (mode == NETKIT_L2 && !(ifmp && tbp[IFLA_ADDRESS]))
                eth_hw_addr_random(peer);
@@ -401,6 +414,7 @@ static int netkit_new_link(struct net *peer_net, struct net_device *dev,
        nk->policy = policy_peer;
        nk->scrub = scrub_peer;
        nk->mode = mode;
+       nk->headroom = headroom;
        bpf_mprog_bundle_init(&nk->bundle);
 
        err = register_netdevice(peer);
@@ -426,6 +440,7 @@ static int netkit_new_link(struct net *peer_net, struct net_device *dev,
        nk->policy = policy_prim;
        nk->scrub = scrub_prim;
        nk->mode = mode;
+       nk->headroom = headroom;
        bpf_mprog_bundle_init(&nk->bundle);
 
        err = register_netdevice(dev);
@@ -850,7 +865,18 @@ static int netkit_change_link(struct net_device *dev, struct nlattr *tb[],
        struct net_device *peer = rtnl_dereference(nk->peer);
        enum netkit_action policy;
        struct nlattr *attr;
-       int err;
+       int err, i;
+       static const struct {
+               u32 attr;
+               char *name;
+       } fixed_params[] = {
+               { IFLA_NETKIT_MODE,       "operating mode" },
+               { IFLA_NETKIT_SCRUB,      "scrubbing" },
+               { IFLA_NETKIT_PEER_SCRUB, "peer scrubbing" },
+               { IFLA_NETKIT_PEER_INFO,  "peer info" },
+               { IFLA_NETKIT_HEADROOM,   "headroom" },
+               { IFLA_NETKIT_TAILROOM,   "tailroom" },
+       };
 
        if (!nk->primary) {
                NL_SET_ERR_MSG(extack,
@@ -858,28 +884,14 @@ static int netkit_change_link(struct net_device *dev, struct nlattr *tb[],
                return -EACCES;
        }
 
-       if (data[IFLA_NETKIT_MODE]) {
-               NL_SET_ERR_MSG_ATTR(extack, data[IFLA_NETKIT_MODE],
-                                   "netkit link operating mode cannot be changed after device creation");
-               return -EACCES;
-       }
-
-       if (data[IFLA_NETKIT_SCRUB]) {
-               NL_SET_ERR_MSG_ATTR(extack, data[IFLA_NETKIT_SCRUB],
-                                   "netkit scrubbing cannot be changed after device creation");
-               return -EACCES;
-       }
-
-       if (data[IFLA_NETKIT_PEER_SCRUB]) {
-               NL_SET_ERR_MSG_ATTR(extack, data[IFLA_NETKIT_PEER_SCRUB],
-                                   "netkit scrubbing cannot be changed after device creation");
-               return -EACCES;
-       }
-
-       if (data[IFLA_NETKIT_PEER_INFO]) {
-               NL_SET_ERR_MSG_ATTR(extack, data[IFLA_NETKIT_PEER_INFO],
-                                   "netkit peer info cannot be changed after device creation");
-               return -EINVAL;
+       for (i = 0; i < ARRAY_SIZE(fixed_params); i++) {
+               attr = data[fixed_params[i].attr];
+               if (attr) {
+                       NL_SET_ERR_MSG_ATTR_FMT(extack, attr,
+                                               "netkit link %s cannot be changed after device creation",
+                                               fixed_params[i].name);
+                       return -EACCES;
+               }
        }
 
        if (data[IFLA_NETKIT_POLICY]) {
@@ -914,6 +926,8 @@ static size_t netkit_get_size(const struct net_device *dev)
               nla_total_size(sizeof(u32)) + /* IFLA_NETKIT_PEER_SCRUB */
               nla_total_size(sizeof(u32)) + /* IFLA_NETKIT_MODE */
               nla_total_size(sizeof(u8))  + /* IFLA_NETKIT_PRIMARY */
+              nla_total_size(sizeof(u16)) + /* IFLA_NETKIT_HEADROOM */
+              nla_total_size(sizeof(u16)) + /* IFLA_NETKIT_TAILROOM */
               0;
 }
 
@@ -930,6 +944,10 @@ static int netkit_fill_info(struct sk_buff *skb, const struct net_device *dev)
                return -EMSGSIZE;
        if (nla_put_u32(skb, IFLA_NETKIT_SCRUB, nk->scrub))
                return -EMSGSIZE;
+       if (nla_put_u16(skb, IFLA_NETKIT_HEADROOM, dev->needed_headroom))
+               return -EMSGSIZE;
+       if (nla_put_u16(skb, IFLA_NETKIT_TAILROOM, dev->needed_tailroom))
+               return -EMSGSIZE;
 
        if (peer) {
                nk = netkit_priv(peer);
@@ -947,6 +965,8 @@ static const struct nla_policy netkit_policy[IFLA_NETKIT_MAX + 1] = {
        [IFLA_NETKIT_MODE]              = NLA_POLICY_MAX(NLA_U32, NETKIT_L3),
        [IFLA_NETKIT_POLICY]            = { .type = NLA_U32 },
        [IFLA_NETKIT_PEER_POLICY]       = { .type = NLA_U32 },
+       [IFLA_NETKIT_HEADROOM]          = { .type = NLA_U16 },
+       [IFLA_NETKIT_TAILROOM]          = { .type = NLA_U16 },
        [IFLA_NETKIT_SCRUB]             = NLA_POLICY_MAX(NLA_U32, NETKIT_SCRUB_DEFAULT),
        [IFLA_NETKIT_PEER_SCRUB]        = NLA_POLICY_MAX(NLA_U32, NETKIT_SCRUB_DEFAULT),
        [IFLA_NETKIT_PRIMARY]           = { .type = NLA_REJECT,
index 2575e0cd9b482e78984ca6528dd7d4a24f03e290..2fa2c265dcbaf6502f4df814e2258a0cb0767aed 100644 (file)
@@ -1315,6 +1315,8 @@ enum {
        IFLA_NETKIT_MODE,
        IFLA_NETKIT_SCRUB,
        IFLA_NETKIT_PEER_SCRUB,
+       IFLA_NETKIT_HEADROOM,
+       IFLA_NETKIT_TAILROOM,
        __IFLA_NETKIT_MAX,
 };
 #define IFLA_NETKIT_MAX        (__IFLA_NETKIT_MAX - 1)
index 8516c1ccd57a7c7634a538fe3ac16c858f647420..7e46ca4cd31bb5cbeb7f386c3cd151e782aba5d9 100644 (file)
@@ -1315,6 +1315,8 @@ enum {
        IFLA_NETKIT_MODE,
        IFLA_NETKIT_SCRUB,
        IFLA_NETKIT_PEER_SCRUB,
+       IFLA_NETKIT_HEADROOM,
+       IFLA_NETKIT_TAILROOM,
        __IFLA_NETKIT_MAX,
 };
 #define IFLA_NETKIT_MAX        (__IFLA_NETKIT_MAX - 1)