1IBV_CREATE_FLOW(3) Libibverbs Programmer's Manual IBV_CREATE_FLOW(3)
2
6 ibv_create_flow, ibv_destroy_flow - create or destroy flow steering
7 rules
8
10 #include <infiniband/verbs.h>
11 struct ibv_flow *ibv_create_flow(struct ibv_qp *qp,
13 struct ibv_flow_attr *flow_attr);
14 int ibv_destroy_flow(struct ibv_flow *flow_id);
15
18 ibv_create_flow()
19 allows a user application QP qp to be attached into a specified flow
20 flow which is defined in <infiniband/verbs.h>
21 struct ibv_flow_attr {
23 uint32_t comp_mask; /* Future extendibility */
24 enum ibv_flow_attr_type type; /* Rule type - see below */
25 uint16_t size; /* Size of command */
26 uint16_t priority; /* Rule priority - see below */
27 uint8_t num_of_specs; /* Number of ibv_flow_spec_xxx */
28 uint8_t port; /* The uplink port number */
29 uint32_t flags; /* Extra flags for rule - see below */
30 /* Following are the optional layers according to user request
31 * struct ibv_flow_spec_xxx
32 * struct ibv_flow_spec_yyy
33 */
34 };
35 enum ibv_flow_attr_type {
37 IBV_FLOW_ATTR_NORMAL = 0x0, /* Steering according to rule specifications */
38 IBV_FLOW_ATTR_ALL_DEFAULT = 0x1, /* Default unicast and multicast rule - receive all Eth traffic which isn't steered to any QP */
39 IBV_FLOW_ATTR_MC_DEFAULT = 0x2, /* Default multicast rule - receive all Eth multicast traffic which isn't steered to any QP */
40 IBV_FLOW_ATTR_SNIFFER = 0x3, /* Sniffer rule - receive all port traffic */
41 };
42 enum ibv_flow_flags {
44 IBV_FLOW_ATTR_FLAGS_ALLOW_LOOP_BACK = 1 << 0, /* Apply the rules on packets that were sent from the attached QP through loopback */
45 IBV_FLOW_ATTR_FLAGS_DONT_TRAP = 1 << 1, /* Rule doesn't trap received packets, allowing them to match lower prioritized rules */
46 IBV_FLOW_ATTR_FLAGS_EGRESS = 1 << 2, /* Match sent packets against EGRESS rules and carry associated actions if required */
47 };
48 enum ibv_flow_spec_type {
50 IBV_FLOW_SPEC_ETH = 0x20, /* Flow specification of L2 header */
51 IBV_FLOW_SPEC_IPV4 = 0x30, /* Flow specification of IPv4 header */
52 IBV_FLOW_SPEC_IPV6 = 0x31, /* Flow specification of IPv6 header */
53 IBV_FLOW_SPEC_IPV4_EXT = 0x32, /* Extended flow specification of IPv4 */
54 IBV_FLOW_SPEC_ESP = 0x34, /* Flow specification of ESP (IPSec) header */
55 IBV_FLOW_SPEC_TCP = 0x40, /* Flow specification of TCP header */
56 IBV_FLOW_SPEC_UDP = 0x41, /* Flow specification of UDP header */
57 IBV_FLOW_SPEC_VXLAN_TUNNEL = 0x50, /* Flow specification of VXLAN header */
58 IBV_FLOW_SPEC_GRE = 0x51, /* Flow specification of GRE header */
59 IBV_FLOW_SPEC_MPLS = 0x60, /* Flow specification of MPLS header */
60 IBV_FLOW_SPEC_INNER = 0x100, /* Flag making L2/L3/L4 specifications to be applied on the inner header */
61 IBV_FLOW_SPEC_ACTION_TAG = 0x1000, /* Action tagging matched packet */
62 IBV_FLOW_SPEC_ACTION_DROP = 0x1001, /* Action dropping matched packet */
63 IBV_FLOW_SPEC_ACTION_HANDLE = 0x1002, /* Carry out an action created by ibv_create_flow_action_xxxx verb */
64 IBV_FLOW_SPEC_ACTION_COUNT = 0x1003, /* Action count matched packet with a ibv_counters handle */
65 };
66 Flow specification general structure:
68 struct ibv_flow_spec_xxx {
70 enum ibv_flow_spec_type type;
71 uint16_t size; /* Flow specification size = sizeof(struct ibv_flow_spec_xxx) */
72 struct ibv_flow_xxx_filter val;
73 struct ibv_flow_xxx_filter mask; /* Defines which bits from the filter value are applicable when looking for a match in the incoming packet */
74 };
75 Each spec struct holds the relevant network layer parameters for matching. To enforce the match, the user sets a mask for each parameter.
77 Packets coming from the wire are matched against the flow specification. If a match is found, the associated flow actions are executed on the packet.
78 In ingress flows, the QP parameter is treated as another action of scattering the packet to the respected QP.
79 If the bit is set in the mask, the corresponding bit in the value should be matched.
80 Note that most vendors support either full mask (all "1"s) or zero mask (all "0"s).
81 Network parameters in the relevant network structs should be given in network order (big endian).
82 Flow domains and priority
85 Flow steering defines the concept of domain and priority. Each domain
86 represents an application that can attach a flow. Domains are priori‐
87 tized. A higher priority domain will always supersede a lower priority
88 domain when their flow specifications overlap.
89 IB verbs have the higher priority domain.
90 In addition to the domain, there is priority within each of the
91 domains. A lower priority numeric value (higher priority) takes prece‐
92 dence over matching rules with higher numeric priority value (lower
93 priority). It is important to note that the priority value of a flow
94 spec is used not only to establish the precedence of conflicting flow
95 matches but also as a way to abstract the order on which flow specs are
96 tested for matches. Flows with higher priorities will be tested before
97 flows with lower priorities.
98 Rules definition ordering
101 An application can provide the ibv_flow_spec_xxx rules in an un-ordered
102 scheme. In this case, each spec should be well defined and match a spe‐
103 cific network header layer. In some cases, when certain flow spec
104 types are present in the spec list, it is required to provide the list
105 in an ordered manner so that the position of that flow spec type in the
106 protocol stack is strictly defined. When the certain spec type, which
107 requires the ordering, resides in the inner network protocol stack (in
108 tunnel protocols) the ordering should be applied to the inner network
109 specs and should be combined with the inner spec indication using the
110 IBV_FLOW_SPEC_INNER flag. For example: An MPLS spec which attempts to
111 match an MPLS tag in the inner network should have the
112 IBV_FLOW_SPEC_INNER flag set and so do the rest of the inner network
113 specs. On top of that, all the inner network specs should be provided
114 in an ordered manner. This is essential to represent many of the
115 encapsulation tunnel protocols.
116 The flow spec types which require this sort of ordering are:
118 1. IBV_FLOW_SPEC_MPLS -
119 Since MPLS header can appear at several locations in the protocol stack
120 and can also be encapsulated on top of different layers, it is required
121 to place this spec according to its exact location in the protocol
122 stack.
123 ibv_destroy_flow()
125 destroys the flow flow_id.
126
128 ibv_create_flow() returns a pointer to the flow, or NULL if the request
129 fails. In case of an error, errno is updated.
130 ibv_destroy_flow() returns 0 on success, or the value of errno on fail‐
132 ure (which indicates the failure reason).
133
135 EINVAL
136 ibv_create_flow() flow specification, QP or priority are invalid
137 ibv_destroy_flow() flow_id is invalid
139 ENOMEM
141 Couldn't create/destroy flow, not enough memory
142 ENXIO
144 Device managed flow steering isn't currently supported
145 EPERM
147 No permissions to add the flow steering rule
148
150 1. These verbs are available only for devices supporting
151 IBV_DEVICE_MANAGED_FLOW_STEERING and only for QPs of Transport Ser‐
152 vice Type IBV_QPT_UD or IBV_QPT_RAW_PACKET
153 2. User must memset the spec struct with zeros before using it.
154 3. ether_type field in ibv_flow_eth_filter is the ethertype following
155 the last VLAN tag of the packet.
156 4. Only rule type IBV_FLOW_ATTR_NORMAL supports
157 IBV_FLOW_ATTR_FLAGS_DONT_TRAP flag.
158 5. No specifications are needed for IBV_FLOW_ATTR_SNIFFER rule type.
159 6. When IBV_FLOW_ATTR_FLAGS_EGRESS flag is set, the qp parameter is
160 used only as a mean to get the device.
161
163 Below flow_attr defines a rule in priority 0 to match a destination mac
164 address and a source ipv4 address. For that, L2 and L3 specs are used.
165 If there is a hit on this rule, means the received packet has destina‐
166 tion mac: 66:11:22:33:44:55 and source ip: 0x0B86C806, the packet is
167 steered to its attached qp.
168 struct raw_eth_flow_attr {
170 struct ibv_flow_attr attr;
171 struct ibv_flow_spec_eth spec_eth;
172 struct ibv_flow_spec_ipv4 spec_ipv4;
173 } __attribute__((packed));
174 struct raw_eth_flow_attr flow_attr = {
176 .attr = {
177 .comp_mask = 0,
178 .type = IBV_FLOW_ATTR_NORMAL,
179 .size = sizeof(flow_attr),
180 .priority = 0,
181 .num_of_specs = 2,
182 .port = 1,
183 .flags = 0,
184 },
185 .spec_eth = {
186 .type = IBV_FLOW_SPEC_ETH,
187 .size = sizeof(struct ibv_flow_spec_eth),
188 .val = {
189 .dst_mac = {0x66, 0x11, 0x22, 0x33, 0x44, 0x55},
190 .src_mac = { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
191 .ether_type = 0,
192 .vlan_tag = 0,
193 },
194 .mask = {
195 .dst_mac = { 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF},
196 .src_mac = { 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF},
197 .ether_type = 0,
198 .vlan_tag = 0,
199 }
200 },
201 .spec_ipv4 = {
202 .type = IBV_FLOW_SPEC_IPV4,
203 .size = sizeof(struct ibv_flow_spec_ipv4),
204 .val = {
205 .src_ip = 0x0B86C806,
206 .dst_ip = 0,
207 },
208 .mask = {
209 .src_ip = 0xFFFFFFFF,
210 .dst_ip = 0,
211 }
212 }
213 };
214
217 Hadar Hen Zion <hadarh@mellanox.com>
218 Matan Barak <matanb@mellanox.com>
220 Yishai Hadas <yishaih@mellanox.com>
222 Maor Gottlieb <maorg@mellanox.com>
224libibverbs 2016-03-15 IBV_CREATE_FLOW(3)