net: allow dissecting/matching tunnel control flags
Subsystems (4)
v4 cover net: allow dissecting/matching tunnel control flags
Ilya says: "for correct matching on decapsulated packets, we should match on not only tunnel id and headers, but also on tunnel configuration flags like TUNNEL_NO_CSUM and TUNNEL_DONT_FRAGMENT. This is done to distinguish similar tunnels with slightly different configs. And it is important since tunnel configuration is flow based, i.e. can be different for every packet, even though the main tunnel port is the same."
- patch 1 extends the kernel's flow dissector to extract these flags from the packet's tunnel metadata.
- patch 2 extends TC flower to match on any combination of TUNNEL_NO_CSUM, TUNNEL_DONT_FRAGMENT, TUNNEL_OAM, TUNNEL_CRIT_OPT
v4:
- fix kernel-doc warning in flow_dissector.h (thanks Jakub)
v3:
- rebase on top of new uAPI bits and internals after commit 5832c4a77d69 ("ip_tunnel: convert __be16 tunnel flags to bitmaps"). Use of network byte order is no more needed, since these bits match on metadata: convert netlink attributes to be u32.
- also include TUNNEL_CRIT_OPT
v2:
- use NL_REQ_ATTR_CHECK() where possible (thanks Jamal)
- don't overwrite 'ret' in the error path of fl_set_key_flags()
Davide Caratti (2):
flow_dissector: add support for tunnel control flags net/sched: cls_flower: add support for matching tunnel control flags
include/net/flow_dissector.h | 9 ++++++ include/net/ip_tunnels.h | 12 ++++++++ include/uapi/linux/pkt_cls.h | 3 ++ net/core/flow_dissector.c | 16 ++++++++++- net/sched/cls_flower.c | 56 +++++++++++++++++++++++++++++++++++- 5 files changed, 94 insertions(+), 2 deletions(-)
Signature · 1 line
2.44.0
PATCH v4 1/2 flow_dissector: add support for tunnel control flags
Dissect [no]csum, [no]dontfrag, [no]oam, [no]crit flags from skb metadata. This is a prerequisite for matching these control flags using TC flower.
--- include/net/flow_dissector.h | 9 +++++++++ include/net/ip_tunnels.h | 12 ++++++++++++ net/core/flow_dissector.c | 16 +++++++++++++++- 3 files changed, 36 insertions(+), 1 deletion(-)
include/net/flow_dissector.h +9 -0
@@ -329,6 +329,14 @@ struct flow_dissector_key_cfm {#define FLOW_DIS_CFM_MDL_MASK GENMASK(7, 5)#define FLOW_DIS_CFM_MDL_MAX 7 +/** +* struct flow_dissector_key_enc_flags: tunnel metadata control flags +* @flags: tunnel control flags +*/ +struct flow_dissector_key_enc_flags { +u32 flags; +}; +enum flow_dissector_key_id {FLOW_DISSECTOR_KEY_CONTROL ,/* struct flow_dissector_key_control */ FLOW_DISSECTOR_KEY_BASIC ,/* struct flow_dissector_key_basic */
@@ -363,6 +371,7 @@ enum flow_dissector_key_id {FLOW_DISSECTOR_KEY_L2TPV3 ,/* struct flow_dissector_key_l2tpv3 */ FLOW_DISSECTOR_KEY_CFM ,/* struct flow_dissector_key_cfm */ FLOW_DISSECTOR_KEY_IPSEC ,/* struct flow_dissector_key_ipsec */ +FLOW_DISSECTOR_KEY_ENC_FLAGS ,/* struct flow_dissector_key_enc_flags */ FLOW_DISSECTOR_KEY_MAX, };
include/net/ip_tunnels.h +12 -0
@@ -247,6 +247,18 @@ static inline bool ip_tunnel_is_options_present(const unsigned long *flags)return ip_tunnel_flags_intersect (flags, present); } +static inline void ip_tunnel_set_encflags_present (unsigned long *flags) +{ +IP_TUNNEL_DECLARE_FLAGS (present) = { }; + +__set_bit (IP_TUNNEL_CSUM_BIT , present); +__set_bit (IP_TUNNEL_DONT_FRAGMENT_BIT , present); +__set_bit (IP_TUNNEL_OAM_BIT , present); +__set_bit (IP_TUNNEL_CRIT_OPT_BIT , present); + +ip_tunnel_flags_or (flags, flags, present); +} +static inline bool ip_tunnel_flags_is_be16_compat (const unsigned long *flags) {IP_TUNNEL_DECLARE_FLAGS (supp) = { };
net/core/flow_dissector.c +15 -1
@@ -382,7 +382,9 @@ skb_flow_dissect_tunnel_info(const struct sk_buff *skb, !dissector_uses_key (flow_dissector,FLOW_DISSECTOR_KEY_ENC_IP ) && !dissector_uses_key (flow_dissector, - FLOW_DISSECTOR_KEY_ENC_OPTS)) +FLOW_DISSECTOR_KEY_ENC_OPTS ) && + !dissector_uses_key (flow_dissector, + FLOW_DISSECTOR_KEY_ENC_FLAGS)) return; info =skb_tunnel_info (skb);
@@ -475,6 +477,18 @@ skb_flow_dissect_tunnel_info(const struct sk_buff *skb, IP_TUNNEL_GENEVE_OPT_BIT); enc_opt->dst_opt_type = val < __IP_TUNNEL_FLAG_NUM ? val :0 ; } + +if (dissector_uses_key (flow_dissector,FLOW_DISSECTOR_KEY_ENC_FLAGS )) { +struct flow_dissector_key_enc_flags *enc_flags; +IP_TUNNEL_DECLARE_FLAGS (flags) = {}; + + enc_flags =skb_flow_dissector_target (flow_dissector, +FLOW_DISSECTOR_KEY_ENC_FLAGS , + target_container); +ip_tunnel_set_encflags_present (flags); +ip_tunnel_flags_and (flags, flags, info->key.tun_flags); + enc_flags->flags =bitmap_read (flags,IP_TUNNEL_CSUM_BIT ,32 ); + } }EXPORT_SYMBOL (skb_flow_dissect_tunnel_info);
Signature · 1 line
2.44.0
PATCH v4 2/2 net/sched: cls_flower: add support for matching tunnel control flags
In reply to Davide Carattiextend cls_flower to match TUNNEL_FLAGS_PRESENT bits in tunnel metadata.
--- include/uapi/linux/pkt_cls.h | 3 ++ net/sched/cls_flower.c | 56 +++++++++++++++++++++++++++++++++++- 2 files changed, 58 insertions(+), 1 deletion(-)
include/uapi/linux/pkt_cls.h +3 -0
@@ -554,6 +554,9 @@ enum {TCA_FLOWER_KEY_SPI ,/* be32 */ TCA_FLOWER_KEY_SPI_MASK ,/* be32 */ +TCA_FLOWER_KEY_ENC_FLAGS ,/* u32 */ +TCA_FLOWER_KEY_ENC_FLAGS_MASK ,/* u32 */ + __TCA_FLOWER_MAX, };
net/sched/cls_flower.c +55 -1
106 lines not on this page · raw message
Signature · 1 line
2.44.0
Re: [PATCH net-next v4 1/2] flow_dissector: add support for tunnel control flags
In reply to Davide CarattiQuoted 5 lines · Davide Caratti
Dissect [no]csum, [no]dontfrag, [no]oam, [no]crit flags from skb metadata. This is a prerequisite for matching these control flags using TC flower.
Suggested-by Ilya Maximets <[email protected]>Signed-off-by Davide Caratti <[email protected]>
Re: [PATCH net-next v4 2/2] net/sched: cls_flower: add support for matching tunnel control flags
In reply to Davide CarattiQuoted 5 lines · Davide Caratti
extend
cls_flowerto matchTUNNEL_FLAGS_PRESENTbits in tunnel metadata.Suggested-by Ilya Maximets <[email protected]>Acked-by Jamal Hadi Salim <[email protected]>Signed-off-by Davide Caratti <[email protected]>
This series was applied to netdev/net-next.git (main) by Paolo Abeni <[email protected]>:
Hello:
This series was applied to netdev/net-next.git (main) by Paolo Abeni <[email protected]>:
Quoted 8 lines · On Thu, 30 May 2024 19:08:33 +0200 you wrote:
Ilya says: "for correct matching on decapsulated packets, we should match on not only tunnel id and headers, but also on tunnel configuration flags like
TUNNEL_NO_CSUMandTUNNEL_DONT_FRAGMENT. This is done to distinguish similar tunnels with slightly different configs. And it is important since tunnel configuration is flow based, i.e. can be different for every packet, even though the main tunnel port is the same."[...]
Here is the summary with links:
- [net-next,v4,1/2] flow_dissector: add support for tunnel control flags
https://git.kernel.org/netdev/net-next/c/668b6a2ef832 - [net-next,v4,2/2] net/sched: cls_flower: add support for matching tunnel control flags https://git.kernel.org/netdev/net-next/c/1d17568e74de
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html
Re: [PATCH net-next v4 2/2] net/sched: cls_flower: add support for matching tunnel control flags
In reply to Davide CarattiHi Davide and Ilva,
Quoted 89 lines · Davide Caratti
extend
cls_flowerto matchTUNNEL_FLAGS_PRESENTbits in tunnel metadata.Suggested-by Ilya Maximets <[email protected]>Acked-by Jamal Hadi Salim <[email protected]>Signed-off-by Davide Caratti <[email protected]>--- include/uapi/linux/pkt_cls.h | 3 ++ net/sched/cls_flower.c | 56 +++++++++++++++++++++++++++++++++++- 2 files changed, 58 insertions(+), 1 deletion(-)include/uapi/linux/pkt_cls.h +3
-0@@ -554,6 +554,9 @@ enum {TCA_FLOWER_KEY_SPI ,/* be32 */ TCA_FLOWER_KEY_SPI_MASK ,/* be32 */ +TCA_FLOWER_KEY_ENC_FLAGS ,/* u32 */ +TCA_FLOWER_KEY_ENC_FLAGS_MASK ,/* u32 */ + __TCA_FLOWER_MAX, };net/sched/cls_flower.c +27
-1@@ -41,6 +41,12 @@#define TCA_FLOWER_KEY_CT_FLAGS_MASK \ (TCA_FLOWER_KEY_CT_FLAGS_MAX - 1) +#define TUNNEL_FLAGS_PRESENT (\ + _BITUL(IP_TUNNEL_CSUM_BIT) | \ + _BITUL(IP_TUNNEL_DONT_FRAGMENT_BIT) | \ + _BITUL(IP_TUNNEL_OAM_BIT) | \ + _BITUL(IP_TUNNEL_CRIT_OPT_BIT)) +struct fl_flow_key {struct flow_dissector_key_meta meta;struct flow_dissector_key_control control;@@ -75,6 +81,7 @@ struct fl_flow_key {struct flow_dissector_key_l2tpv3 l2tpv3;struct flow_dissector_key_ipsec ipsec;struct flow_dissector_key_cfm cfm; +struct flow_dissector_key_enc_flags enc_flags; }__aligned(BITS_PER_LONG / 8) ;/* Ensure that we can do comparisons as longs. */ struct fl_flow_mask_range {@@ -732,6 +739,10 @@ static const struct nla_policy fl_policy[TCA_FLOWER_MAX + 1] = { [TCA_FLOWER_KEY_SPI_MASK] = { .type = NLA_U32 }, [TCA_FLOWER_L2_MISS] = NLA_POLICY_MAX(NLA_U8, 1), [TCA_FLOWER_KEY_CFM] = { .type = NLA_NESTED }, + [TCA_FLOWER_KEY_ENC_FLAGS ] =NLA_POLICY_MASK (NLA_U32 , +TUNNEL_FLAGS_PRESENT ), +` [TCA_FLOWER_KEY_ENC_FLAGS_MASK] = NLA_POLICY_MASK(NLA_U32, + TUNNEL_FLAGS_PRESENT), }; static const struct nla_policy@@ -1825,6 +1836,21 @@ static int fl_set_key_cfm(struct nlattr **tb,return 0 ; } +static int fl_set_key_enc_flags (struct nlattr **tb,u32 *flags_key, +u32 *flags_mask,struct netlink_ext_ack *extack) +{ +/* mask is mandatory for flags */ +if (NL_REQ_ATTR_CHECK (extack,NULL , tb,TCA_FLOWER_KEY_ENC_FLAGS_MASK )) { +NL_SET_ERR_MSG (extack,"missing enc_flags mask" ); +return -EINVAL ; + } + + *flags_key =nla_get_u32 (tb[TCA_FLOWER_KEY_ENC_FLAGS ]); + *flags_mask =nla_get_u32 (tb[TCA_FLOWER_KEY_ENC_FLAGS_MASK ]); + +return 0 ; +} +static int fl_set_key (struct net *net,struct nlattr **tb,struct fl_flow_key *key,struct fl_flow_key *mask,struct netlink_ext_ack *extack)@@ -2059,9 +2085,16 @@ static int fl_set_key(struct net *net, struct nlattr **tb,if (ret)return ret; -if (tb[TCA_FLOWER_KEY_FLAGS ]) +if (tb[TCA_FLOWER_KEY_FLAGS ]) { ret = fl_set_key_flags(tb, &key->control.flags,
net/sched/cls_flower.c static int fl_set_key(struct net *net, struct nlattr **tb, · line 2099
Quoted 10 lines
net/sched/cls_flower.c +7
-0&mask->control.flags, extack); +if (ret) +return ret; + } + +if (tb[TCA_FLOWER_KEY_ENC_FLAGS ]) + ret =fl_set_key_enc_flags (tb, &key->enc_flags.flags, + &mask->enc_flags.flags, extack);return ret;
Sorry, that I am late to the party, I have been away camping at Electromagnetic Field in the UK.
Why not use the already existing key->enc_control.flags with dissector key FLOW_DISSECTOR_KEY_ENC_CONTROL for storing the flags?
Currently key->enc_control.flags are unused.
I haven't fixed the drivers to validate that field yet, so currently only sfc does so.
Look at include/uapi/linux/pkt_cls.h for netlink flags:
enum {
TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT = (1 << 0),
TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST = (1 << 1),
};and at include/net/flow_dissector.h for the dissector flags:
#define FLOW_DIS_IS_FRAGMENT BIT(0) #define FLOW_DIS_FIRST_FRAG BIT(1) #define FLOW_DIS_ENCAPSULATION BIT(2)
I would like to keep FLOW_DIS_ENCAPSULATION as the last flag, in order to keep parity with the netlink flags, since the dissector flags field is exposed to user space when some flags are not supported.
I realize that since this series is now merged, then fixing this up will have to go in another series, are you up for that?
Signature · 4 lines
Best regards Asbjørn Sloth Tønnesen Network Engineer Fiberby - AS42541
Re: [PATCH net-next v4 2/2] net/sched: cls_flower: add support for matching tunnel control flags
hello Asbjørn, thanks for looking at this!
Quoted 1 line · Asbjørn Sloth Tønnesen
Hi Davide and Ilva,
[...]
Quoted 2 lines
Sorry, that I am late to the party, I have been away camping at Electromagnetic Field in the UK.
guess you had fun :)
Quoted 4 lines
Why not use the already existing
key->enc_control.flagswith dissector keyFLOW_DISSECTOR_KEY_ENC_CONTROLfor storing the flags?Currently
key->enc_control.flagsare unused.
I looked at the struct, it currently stores information about IPv4 / IPv6 fragmentation of the inner/outer packet plus FLOW_DIS_ENCAPSULATION - that IIRC is true in case the packet has an inner IP header (key->enc_control.addr_type selects the address family). these 3 bits are usually set when FLOW_DISSECTOR_KEY_CONTROL is requested.
Quoted 2 lines
I haven't fixed the drivers to validate that field yet, so currently only sfc does so.
I see:
[1] https://elixir.bootlin.com/linux/v6.10-rc2/source/drivers/net/ethernet/sfc/tc.c#L276
(for FLOW_DISSECTOR_KEY_CONTROL)
[2] https://elixir.bootlin.com/linux/v6.10-rc2/source/drivers/net/ethernet/sfc/tc.c#L390
(for FLOW_DISSECTOR_KEY_ENC_CONTROL)
Quoted 9 lines
Look at
include/uapi/linux/pkt_cls.hfor netlink flags:enum { TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT = (1 << 0), TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST = (1 << 1), };and at
include/net/flow_dissector.hfor the dissector flags:
Quoted 7 lines
#define FLOW_DIS_IS_FRAGMENT BIT(0) #define FLOW_DIS_FIRST_FRAG BIT(1) #define FLOW_DIS_ENCAPSULATION BIT(2)I would like to keep
FLOW_DIS_ENCAPSULATIONas the last flag, in order to keep parity with the netlink flags, since the dissector flags field is exposed to user space when some flags are not supported.
to fully understand this, I need some more info on how you think IP_TUNNEL_<blah> bit should be written on the dissected data, since FLOW_DIS_IS_FRAGMENT has the same value as BIT(IP_TUNNEL_CSUM_BIT). So, in a way or another parity with netlink can't be guaranteed.
The main reason why I added a new dissector bit is: avoiding breaking functionality of drivers, because I don't really know what the hardware does. If we want to use FLOW_DISSECTOR_KEY_ENC_CONTROL, then all drivers supporting the offload of this bit need to return -EOPNOTSUPP if key->enc_control.flags contains IP_TUNNEL_<blah> bits (like done in [2] for FLOW_DIS_*FRAG*).
Moreover, if a driver supports offload of one of these bits (e.g. TUNNEL_OAM), we need to understand if it's possible for hardware to match when 'addr_type' is not specified, e.g. a geneve packet carrying a non-IP packet.
Quoted 2 lines
I realize that since this series is now merged, then fixing this up will have to go in another series, are you up for that?
I'm not against: if there's something to improve, this is the right moment _ later it will be more difficult; but we should clarify:
- how / if we want to handle overlap of 'flags'. I guess, pass the current tunnel flags to the arguments of skb_flow_dissect_set_enc_addr_type(), and write "something" in ctrl->flags.
- what to do with with flower netlink API. If these bits are in the same struct, we don't need TCA_FLOWER_KEY_ENC_FLAGS anymore? should we extend TCA_FLOWER_KEY_FLAGS instead?
thanks,
Signature · 1 line
davide