X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fflow.h;h=6e169d68ee8da442a5949b98ba092ee632168dc5;hb=485e2766d816a18e1a51ddf5e12e55f38aeedf89;hp=a19a9ae5e42b5356ec2a9cee4a5a9a1609059a80;hpb=1e37a2d75196c6e10ac078213ad94afb4f3e219a;p=sliver-openvswitch.git diff --git a/lib/flow.h b/lib/flow.h index a19a9ae5e..6e169d68e 100644 --- a/lib/flow.h +++ b/lib/flow.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2008, 2009, 2010 Nicira Networks. + * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -24,145 +24,276 @@ #include "openflow/nicira-ext.h" #include "openflow/openflow.h" #include "hash.h" -#include "openvswitch/datapath-protocol.h" #include "util.h" +struct dpif_flow_stats; struct ds; struct flow_wildcards; -struct ofp_match; +struct miniflow; +struct minimask; struct ofpbuf; -#define FLOW_N_REGS 3 +/* This sequence number should be incremented whenever anything involving flows + * or the wildcarding of flows changes. This will cause build assertion + * failures in places which likely need to be updated. */ +#define FLOW_WC_SEQ 20 + +#define FLOW_N_REGS 8 BUILD_ASSERT_DECL(FLOW_N_REGS <= NXM_NX_MAX_REGS); +/* Used for struct flow's dl_type member for frames that have no Ethernet + * type, that is, pure 802.2 frames. */ +#define FLOW_DL_TYPE_NONE 0x5ff + +/* Fragment bits, used for IPv4 and IPv6, always zero for non-IP flows. */ +#define FLOW_NW_FRAG_ANY (1 << 0) /* Set for any IP frag. */ +#define FLOW_NW_FRAG_LATER (1 << 1) /* Set for IP frag with nonzero offset. */ +#define FLOW_NW_FRAG_MASK (FLOW_NW_FRAG_ANY | FLOW_NW_FRAG_LATER) + +BUILD_ASSERT_DECL(FLOW_NW_FRAG_ANY == NX_IP_FRAG_ANY); +BUILD_ASSERT_DECL(FLOW_NW_FRAG_LATER == NX_IP_FRAG_LATER); + +#define FLOW_TNL_F_DONT_FRAGMENT (1 << 0) +#define FLOW_TNL_F_CSUM (1 << 1) +#define FLOW_TNL_F_KEY (1 << 2) + +const char *flow_tun_flag_to_string(uint32_t flags); + +struct flow_tnl { + ovs_be64 tun_id; + ovs_be32 ip_src; + ovs_be32 ip_dst; + uint16_t flags; + uint8_t ip_tos; + uint8_t ip_ttl; +}; + +/* +* A flow in the network. +* +* The meaning of 'in_port' is context-dependent. In most cases, it is a +* 16-bit OpenFlow 1.0 port number. In the software datapath interface (dpif) +* layer and its implementations (e.g. dpif-linux, dpif-netdev), it is instead +* a 32-bit datapath port number. +*/ struct flow { + struct flow_tnl tunnel; /* Encapsulating tunnel parameters. */ + ovs_be64 metadata; /* OpenFlow Metadata. */ + struct in6_addr ipv6_src; /* IPv6 source address. */ + struct in6_addr ipv6_dst; /* IPv6 destination address. */ + struct in6_addr nd_target; /* IPv6 neighbor discovery (ND) target. */ + uint32_t skb_priority; /* Packet priority for QoS. */ uint32_t regs[FLOW_N_REGS]; /* Registers. */ - ovs_be32 tun_id; /* Encapsulating tunnel ID. */ - ovs_be32 nw_src; /* IP source address. */ - ovs_be32 nw_dst; /* IP destination address. */ - uint16_t in_port; /* Input switch port. */ - ovs_be16 dl_vlan; /* Input VLAN. */ + ovs_be32 nw_src; /* IPv4 source address. */ + ovs_be32 nw_dst; /* IPv4 destination address. */ + ovs_be32 ipv6_label; /* IPv6 flow label. */ + uint32_t in_port; /* Input port. OpenFlow port number + unless in DPIF code, in which case it + is the datapath port number. */ + uint32_t skb_mark; /* Packet mark. */ + ovs_be32 mpls_lse; /* MPLS label stack entry. */ + uint16_t mpls_depth; /* Depth of MPLS stack. */ + ovs_be16 vlan_tci; /* If 802.1Q, TCI | VLAN_CFI; otherwise 0. */ ovs_be16 dl_type; /* Ethernet frame type. */ ovs_be16 tp_src; /* TCP/UDP source port. */ ovs_be16 tp_dst; /* TCP/UDP destination port. */ uint8_t dl_src[6]; /* Ethernet source address. */ uint8_t dl_dst[6]; /* Ethernet destination address. */ uint8_t nw_proto; /* IP protocol or low 8 bits of ARP opcode. */ - uint8_t dl_vlan_pcp; /* Input VLAN priority. */ - uint8_t nw_tos; /* IP ToS (DSCP field, 6 bits). */ + uint8_t nw_tos; /* IP ToS (including DSCP and ECN). */ + uint8_t arp_sha[6]; /* ARP/ND source hardware address. */ + uint8_t arp_tha[6]; /* ARP/ND target hardware address. */ + uint8_t nw_ttl; /* IP TTL/Hop Limit. */ + uint8_t nw_frag; /* FLOW_FRAG_* flags. */ + uint8_t zeros[6]; }; +BUILD_ASSERT_DECL(sizeof(struct flow) % 4 == 0); + +#define FLOW_U32S (sizeof(struct flow) / 4) + +/* Remember to update FLOW_WC_SEQ when changing 'struct flow'. */ +BUILD_ASSERT_DECL(sizeof(struct flow) == sizeof(struct flow_tnl) + 160 && + FLOW_WC_SEQ == 20); + +/* Represents the metadata fields of struct flow. */ +struct flow_metadata { + ovs_be64 tun_id; /* Encapsulating tunnel ID. */ + ovs_be64 metadata; /* OpenFlow 1.1+ metadata field. */ + uint32_t regs[FLOW_N_REGS]; /* Registers. */ + uint16_t in_port; /* OpenFlow port or zero. */ +}; + +void flow_extract(struct ofpbuf *, uint32_t priority, uint32_t mark, + const struct flow_tnl *, uint16_t in_port, struct flow *); + +void flow_zero_wildcards(struct flow *, const struct flow_wildcards *); +void flow_get_metadata(const struct flow *, struct flow_metadata *); -/* Assert that there are FLOW_SIG_SIZE bytes of significant data in "struct - * flow", followed by FLOW_PAD_SIZE bytes of padding. */ -#define FLOW_SIG_SIZE (37 + FLOW_N_REGS * 4) -#define FLOW_PAD_SIZE 3 -BUILD_ASSERT_DECL(offsetof(struct flow, nw_tos) == FLOW_SIG_SIZE - 1); -BUILD_ASSERT_DECL(sizeof(((struct flow *)0)->nw_tos) == 1); -BUILD_ASSERT_DECL(sizeof(struct flow) == FLOW_SIG_SIZE + FLOW_PAD_SIZE); - -int flow_extract(struct ofpbuf *, ovs_be32 tun_id, uint16_t in_port, - struct flow *); -void flow_extract_stats(const struct flow *flow, struct ofpbuf *packet, - struct odp_flow_stats *stats); -void flow_to_match(const struct flow *, uint32_t wildcards, int flow_format, - struct ofp_match *); -void flow_from_match(const struct ofp_match *, int flow_format, - ovs_be64 cookie, struct flow *, struct flow_wildcards *); char *flow_to_string(const struct flow *); +void format_flags(struct ds *ds, const char *(*bit_to_string)(uint32_t), + uint32_t flags, char del); + void flow_format(struct ds *, const struct flow *); void flow_print(FILE *, const struct flow *); -static inline int flow_compare(const struct flow *, const struct flow *); +static inline int flow_compare_3way(const struct flow *, const struct flow *); static inline bool flow_equal(const struct flow *, const struct flow *); static inline size_t flow_hash(const struct flow *, uint32_t basis); +void flow_set_dl_vlan(struct flow *, ovs_be16 vid); +void flow_set_vlan_vid(struct flow *, ovs_be16 vid); +void flow_set_vlan_pcp(struct flow *, uint8_t pcp); + +void flow_set_mpls_label(struct flow *flow, ovs_be32 label); +void flow_set_mpls_ttl(struct flow *flow, uint8_t ttl); +void flow_set_mpls_tc(struct flow *flow, uint8_t tc); +void flow_set_mpls_bos(struct flow *flow, uint8_t stack); + +void flow_compose(struct ofpbuf *, const struct flow *); + static inline int -flow_compare(const struct flow *a, const struct flow *b) +flow_compare_3way(const struct flow *a, const struct flow *b) { - return memcmp(a, b, FLOW_SIG_SIZE); + return memcmp(a, b, sizeof *a); } static inline bool flow_equal(const struct flow *a, const struct flow *b) { - return !flow_compare(a, b); + return !flow_compare_3way(a, b); } static inline size_t flow_hash(const struct flow *flow, uint32_t basis) { - return hash_bytes(flow, FLOW_SIG_SIZE, basis); + return hash_words((const uint32_t *) flow, sizeof *flow / 4, basis); } -/* Open vSwitch internal-only wildcard bits. +uint32_t flow_hash_in_minimask(const struct flow *, const struct minimask *, + uint32_t basis); + +/* Wildcards for a flow. * - * These are used only internally to Open vSwitch, in the 'wildcards' member of - * struct flow_wildcards. They never appear in the wire protocol in this - * form. */ + * A 1-bit in each bit in 'masks' indicates that the corresponding bit of + * the flow is significant (must match). A 0-bit indicates that the + * corresponding bit of the flow is wildcarded (need not match). */ +struct flow_wildcards { + struct flow masks; +}; -/* Set to 1 if any bits in any of the reg_masks are wildcarded. This maintains - * the invariant that 'wildcards' is nonzero if and only if any bits are - * wildcarded. */ -#define FWW_REGS (1u << 31) +void flow_wildcards_init_catchall(struct flow_wildcards *); +void flow_wildcards_init_exact(struct flow_wildcards *); -/* Set to 1 if bit 0 (the multicast bit) of the flow's dl_dst is wildcarded. - * - * (We reinterpret OFPFW_DL_DST as excluding bit 0. Both OFPFW_DL_DST and - * FWW_ETH_MCAST have to be set to wildcard the entire Ethernet destination - * address.) */ -#define FWW_ETH_MCAST (1u << 30) +bool flow_wildcards_is_catchall(const struct flow_wildcards *); -/* Avoid collisions. */ -#define FWW_ALL (FWW_REGS | FWW_ETH_MCAST) -BUILD_ASSERT_DECL(!(FWW_ALL & OVSFW_ALL)); +void flow_wildcards_set_reg_mask(struct flow_wildcards *, + int idx, uint32_t mask); -/* Information on wildcards for a flow, as a supplement to "struct flow". +void flow_wildcards_combine(struct flow_wildcards *dst, + const struct flow_wildcards *src1, + const struct flow_wildcards *src2); +bool flow_wildcards_has_extra(const struct flow_wildcards *, + const struct flow_wildcards *); + +uint32_t flow_wildcards_hash(const struct flow_wildcards *, uint32_t basis); +bool flow_wildcards_equal(const struct flow_wildcards *, + const struct flow_wildcards *); +uint32_t flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis); + +uint32_t flow_hash_fields(const struct flow *, enum nx_hash_fields, + uint16_t basis); +const char *flow_hash_fields_to_str(enum nx_hash_fields); +bool flow_hash_fields_valid(enum nx_hash_fields); + +bool flow_equal_except(const struct flow *a, const struct flow *b, + const struct flow_wildcards *); + +/* Compressed flow. */ + +#define MINI_N_INLINE (sizeof(void *) == 4 ? 7 : 8) +#define MINI_N_MAPS DIV_ROUND_UP(FLOW_U32S, 32) + +/* A sparse representation of a "struct flow". + * + * A "struct flow" is fairly large and tends to be mostly zeros. Sparse + * representation has two advantages. First, it saves memory. Second, it + * saves time when the goal is to iterate over only the nonzero parts of the + * struct. * - * The flow_wildcards_*() functions below both depend on and maintain the - * following important invariants: + * The 'map' member holds one bit for each uint32_t in a "struct flow". Each + * 0-bit indicates that the corresponding uint32_t is zero, each 1-bit that it + * is nonzero. * - * 1. 'wildcards' is nonzero if and only if at least one bit or field is - * wildcarded. + * 'values' points to the start of an array that has one element for each 1-bit + * in 'map'. The least-numbered 1-bit is in values[0], the next 1-bit is in + * values[1], and so on. * - * 2. Bits in 'wildcards' not included in OVSFW_ALL or FWW_ALL are set to 0. - * (This is a corollary to invariant #1.) + * 'values' may point to a few different locations: * - * 3. The fields in 'wildcards' masked by OFPFW_NW_SRC_MASK and - * OFPFW_NW_DST_MASK have values between 0 and 32, inclusive. + * - If 'map' has MINI_N_INLINE or fewer 1-bits, it may point to + * 'inline_values'. One hopes that this is the common case. * - * 4. The fields masked by OFPFW_NW_SRC_MASK and OFPFW_NW_DST_MASK correspond - * correctly to the masks in 'nw_src_mask' and 'nw_dst_mask', respectively. + * - If 'map' has more than MINI_N_INLINE 1-bits, it may point to memory + * allocated with malloc(). * - * 5. FWW_REGS is set to 1 in 'wildcards' if and only if at least one bit in - * 'reg_masks[]' is nonzero. (This allows wildcarded 'reg_masks[]' to - * satisfy invariant #1.) + * - The caller could provide storage on the stack for situations where + * that makes sense. So far that's only proved useful for + * minimask_combine(), but the principle works elsewhere. * - * 6. If FWW_REGS is set to 0 in 'wildcards', then the values of all of the - * other members can be correctly predicted based on 'wildcards' alone. + * The implementation maintains and depends on the invariant that every element + * in 'values' is nonzero; that is, wherever a 1-bit appears in 'map', the + * corresponding element of 'values' must be nonzero. */ -struct flow_wildcards { - uint32_t wildcards; /* OFPFW_* | OVSFW_* | FWW_*. */ - uint32_t reg_masks[FLOW_N_REGS]; /* 1-bit in each significant regs bit. */ - ovs_be32 nw_src_mask; /* 1-bit in each significant nw_src bit. */ - ovs_be32 nw_dst_mask; /* 1-bit in each significant nw_dst bit. */ +struct miniflow { + uint32_t *values; + uint32_t inline_values[MINI_N_INLINE]; + uint32_t map[MINI_N_MAPS]; }; -ovs_be32 flow_nw_bits_to_mask(uint32_t wildcards, int shift); -void flow_wildcards_init(struct flow_wildcards *, uint32_t wildcards); -void flow_wildcards_init_exact(struct flow_wildcards *); +void miniflow_init(struct miniflow *, const struct flow *); +void miniflow_clone(struct miniflow *, const struct miniflow *); +void miniflow_destroy(struct miniflow *); -bool flow_wildcards_set_nw_src_mask(struct flow_wildcards *, ovs_be32); -bool flow_wildcards_set_nw_dst_mask(struct flow_wildcards *, ovs_be32); -void flow_wildcards_set_reg_mask(struct flow_wildcards *, - int idx, uint32_t mask); +void miniflow_expand(const struct miniflow *, struct flow *); -void flow_wildcards_combine(struct flow_wildcards *dst, - const struct flow_wildcards *src1, - const struct flow_wildcards *src2); -bool flow_wildcards_has_extra(const struct flow_wildcards *, - const struct flow_wildcards *); +uint32_t miniflow_get(const struct miniflow *, unsigned int u32_ofs); +uint16_t miniflow_get_vid(const struct miniflow *); -uint32_t flow_wildcards_hash(const struct flow_wildcards *); -bool flow_wildcards_equal(const struct flow_wildcards *, - const struct flow_wildcards *); +bool miniflow_equal(const struct miniflow *a, const struct miniflow *b); +bool miniflow_equal_in_minimask(const struct miniflow *a, + const struct miniflow *b, + const struct minimask *); +bool miniflow_equal_flow_in_minimask(const struct miniflow *a, + const struct flow *b, + const struct minimask *); +uint32_t miniflow_hash(const struct miniflow *, uint32_t basis); +uint32_t miniflow_hash_in_minimask(const struct miniflow *, + const struct minimask *, uint32_t basis); + +/* Compressed flow wildcards. */ + +/* A sparse representation of a "struct flow_wildcards". + * + * See the large comment on struct miniflow for details. */ +struct minimask { + struct miniflow masks; +}; + +void minimask_init(struct minimask *, const struct flow_wildcards *); +void minimask_clone(struct minimask *, const struct minimask *); +void minimask_combine(struct minimask *dst, + const struct minimask *a, const struct minimask *b, + uint32_t storage[FLOW_U32S]); +void minimask_destroy(struct minimask *); + +void minimask_expand(const struct minimask *, struct flow_wildcards *); + +uint32_t minimask_get(const struct minimask *, unsigned int u32_ofs); +uint16_t minimask_get_vid_mask(const struct minimask *); + +bool minimask_equal(const struct minimask *a, const struct minimask *b); +uint32_t minimask_hash(const struct minimask *, uint32_t basis); + +bool minimask_has_extra(const struct minimask *, const struct minimask *); +bool minimask_is_catchall(const struct minimask *); #endif /* flow.h */