diff options
Diffstat (limited to 'usr/src/uts/common/inet/ip.h')
| -rw-r--r-- | usr/src/uts/common/inet/ip.h | 2395 |
1 files changed, 1163 insertions, 1232 deletions
diff --git a/usr/src/uts/common/inet/ip.h b/usr/src/uts/common/inet/ip.h index 5a7e05b210..88a14068bb 100644 --- a/usr/src/uts/common/inet/ip.h +++ b/usr/src/uts/common/inet/ip.h @@ -55,8 +55,6 @@ extern "C" { #include <sys/squeue.h> #include <net/route.h> #include <sys/systm.h> -#include <sys/multidata.h> -#include <sys/list.h> #include <net/radix.h> #include <sys/modhash.h> @@ -94,6 +92,7 @@ typedef uint32_t ipaddr_t; /* Number of bits in an address */ #define IP_ABITS 32 +#define IPV4_ABITS IP_ABITS #define IPV6_ABITS 128 #define IP_HOST_MASK (ipaddr_t)0xffffffffU @@ -101,14 +100,6 @@ typedef uint32_t ipaddr_t; #define IP_CSUM(mp, off, sum) (~ip_cksum(mp, off, sum) & 0xFFFF) #define IP_CSUM_PARTIAL(mp, off, sum) ip_cksum(mp, off, sum) #define IP_BCSUM_PARTIAL(bp, len, sum) bcksum(bp, len, sum) -#define IP_MD_CSUM(pd, off, sum) (~ip_md_cksum(pd, off, sum) & 0xffff) -#define IP_MD_CSUM_PARTIAL(pd, off, sum) ip_md_cksum(pd, off, sum) - -/* - * Flag to IP write side to indicate that the appln has sent in a pre-built - * IP header. Stored in ipha_ident (which is otherwise zero). - */ -#define IP_HDR_INCLUDED 0xFFFF #define ILL_FRAG_HASH_TBL_COUNT ((unsigned int)64) #define ILL_FRAG_HASH_TBL_SIZE (ILL_FRAG_HASH_TBL_COUNT * sizeof (ipfb_t)) @@ -137,17 +128,12 @@ typedef uint32_t ipaddr_t; #define UDPH_SIZE 8 -/* Leave room for ip_newroute to tack on the src and target addresses */ -#define OK_RESOLVER_MP(mp) \ - ((mp) && ((mp)->b_wptr - (mp)->b_rptr) >= (2 * IP_ADDR_LEN)) - /* * Constants and type definitions to support IP IOCTL commands */ #define IP_IOCTL (('i'<<8)|'p') #define IP_IOC_IRE_DELETE 4 #define IP_IOC_IRE_DELETE_NO_REPLY 5 -#define IP_IOC_IRE_ADVISE_NO_REPLY 6 #define IP_IOC_RTS_REQUEST 7 /* Common definitions used by IP IOCTL data structures */ @@ -157,31 +143,6 @@ typedef struct ipllcmd_s { uint_t ipllc_name_length; } ipllc_t; -/* IP IRE Change Command Structure. */ -typedef struct ipic_s { - ipllc_t ipic_ipllc; - uint_t ipic_ire_type; - uint_t ipic_max_frag; - uint_t ipic_addr_offset; - uint_t ipic_addr_length; - uint_t ipic_mask_offset; - uint_t ipic_mask_length; - uint_t ipic_src_addr_offset; - uint_t ipic_src_addr_length; - uint_t ipic_ll_hdr_offset; - uint_t ipic_ll_hdr_length; - uint_t ipic_gateway_addr_offset; - uint_t ipic_gateway_addr_length; - clock_t ipic_rtt; - uint32_t ipic_ssthresh; - clock_t ipic_rtt_sd; - uchar_t ipic_ire_marks; -} ipic_t; - -#define ipic_cmd ipic_ipllc.ipllc_cmd -#define ipic_ll_name_length ipic_ipllc.ipllc_name_length -#define ipic_ll_name_offset ipic_ipllc.ipllc_name_offset - /* IP IRE Delete Command Structure. */ typedef struct ipid_s { ipllc_t ipid_ipllc; @@ -257,16 +218,8 @@ typedef struct ipoptp_s #define Q_TO_ICMP(q) (Q_TO_CONN((q))->conn_icmp) #define Q_TO_RTS(q) (Q_TO_CONN((q))->conn_rts) -/* - * The following two macros are used by IP to get the appropriate - * wq and rq for a conn. If it is a TCP conn, then we need - * tcp_wq/tcp_rq else, conn_wq/conn_rq. IP can use conn_wq and conn_rq - * from a conn directly if it knows that the conn is not TCP. - */ -#define CONNP_TO_WQ(connp) \ - (IPCL_IS_TCP(connp) ? (connp)->conn_tcp->tcp_wq : (connp)->conn_wq) - -#define CONNP_TO_RQ(connp) RD(CONNP_TO_WQ(connp)) +#define CONNP_TO_WQ(connp) ((connp)->conn_wq) +#define CONNP_TO_RQ(connp) ((connp)->conn_rq) #define GRAB_CONN_LOCK(q) { \ if (q != NULL && CONN_Q(q)) \ @@ -278,9 +231,6 @@ typedef struct ipoptp_s mutex_exit(&(Q_TO_CONN(q))->conn_lock); \ } -/* "Congestion controlled" protocol */ -#define IP_FLOW_CONTROLLED_ULP(p) ((p) == IPPROTO_TCP || (p) == IPPROTO_SCTP) - /* * Complete the pending operation. Usually an ioctl. Can also * be a bind or option management request that got enqueued @@ -295,63 +245,13 @@ typedef struct ipoptp_s } /* - * Flags for the various ip_fanout_* routines. - */ -#define IP_FF_SEND_ICMP 0x01 /* Send an ICMP error */ -#define IP_FF_HDR_COMPLETE 0x02 /* Call ip_hdr_complete if error */ -#define IP_FF_CKSUM 0x04 /* Recompute ipha_cksum if error */ -#define IP_FF_RAWIP 0x08 /* Use rawip mib variable */ -#define IP_FF_SRC_QUENCH 0x10 /* OK to send ICMP_SOURCE_QUENCH */ -#define IP_FF_SYN_ADDIRE 0x20 /* Add IRE if TCP syn packet */ -#define IP_FF_IPINFO 0x80 /* Used for both V4 and V6 */ -#define IP_FF_SEND_SLLA 0x100 /* Send source link layer info ? */ -#define IPV6_REACHABILITY_CONFIRMATION 0x200 /* Flags for ip_xmit_v6 */ -#define IP_FF_NO_MCAST_LOOP 0x400 /* No multicasts for sending zone */ - -/* - * Following flags are used by IPQoS to determine if policy processing is - * required. - */ -#define IP6_NO_IPPOLICY 0x800 /* Don't do IPQoS processing */ -#define IP6_IN_LLMCAST 0x1000 /* Multicast */ - -#define IP_FF_LOOPBACK 0x2000 /* Loopback fanout */ -#define IP_FF_SCTP_CSUM_ERR 0x4000 /* sctp pkt has failed chksum */ - -#ifndef IRE_DB_TYPE -#define IRE_DB_TYPE M_SIG -#endif - -#ifndef IRE_DB_REQ_TYPE -#define IRE_DB_REQ_TYPE M_PCSIG -#endif - -#ifndef IRE_ARPRESOLVE_TYPE -#define IRE_ARPRESOLVE_TYPE M_EVENT -#endif - -/* * Values for squeue switch: */ - #define IP_SQUEUE_ENTER_NODRAIN 1 #define IP_SQUEUE_ENTER 2 -/* - * This is part of the interface between Transport provider and - * IP which can be used to set policy information. This is usually - * accompanied with O_T_BIND_REQ/T_BIND_REQ.ip_bind assumes that - * only IPSEC_POLICY_SET is there when it is found in the chain. - * The information contained is an struct ipsec_req_t. On success - * or failure, either the T_BIND_ACK or the T_ERROR_ACK is returned. - * IPSEC_POLICY_SET is never returned. - */ -#define IPSEC_POLICY_SET M_SETOPTS +#define IP_SQUEUE_FILL 3 -#define IRE_IS_LOCAL(ire) ((ire != NULL) && \ - ((ire)->ire_type & (IRE_LOCAL | IRE_LOOPBACK))) - -#define IRE_IS_TARGET(ire) ((ire != NULL) && \ - ((ire)->ire_type != IRE_BROADCAST)) +extern int ip_squeue_flag; /* IP Fragmentation Reassembly Header */ typedef struct ipf_s { @@ -387,71 +287,6 @@ typedef struct ipf_s { #define ipf_src V4_PART_OF_V6(ipf_v6src) #define ipf_dst V4_PART_OF_V6(ipf_v6dst) -typedef enum { - IB_PKT = 0x01, - OB_PKT = 0x02 -} ip_pkt_t; - -#define UPDATE_IB_PKT_COUNT(ire)\ - { \ - (ire)->ire_ib_pkt_count++; \ - if ((ire)->ire_ipif != NULL) { \ - /* \ - * forwarding packet \ - */ \ - if ((ire)->ire_type & (IRE_LOCAL|IRE_BROADCAST)) \ - atomic_add_32(&(ire)->ire_ipif->ipif_ib_pkt_count, 1);\ - else \ - atomic_add_32(&(ire)->ire_ipif->ipif_fo_pkt_count, 1);\ - } \ - } - -#define UPDATE_OB_PKT_COUNT(ire)\ - { \ - (ire)->ire_ob_pkt_count++;\ - if ((ire)->ire_ipif != NULL) { \ - atomic_add_32(&(ire)->ire_ipif->ipif_ob_pkt_count, 1); \ - } \ - } - -#define IP_RPUT_LOCAL(q, mp, ipha, ire, recv_ill) \ -{ \ - switch (ipha->ipha_protocol) { \ - case IPPROTO_UDP: \ - ip_udp_input(q, mp, ipha, ire, recv_ill); \ - break; \ - default: \ - ip_proto_input(q, mp, ipha, ire, recv_ill, 0); \ - break; \ - } \ -} - -/* - * NCE_EXPIRED is TRUE when we have a non-permanent nce that was - * found to be REACHABLE more than ip_ire_arp_interval ms ago. - * This macro is used to age existing nce_t entries. The - * nce's will get cleaned up in the following circumstances: - * - ip_ire_trash_reclaim will free nce's using ndp_cache_reclaim - * when memory is low, - * - ip_arp_news, when updates are received. - * - if the nce is NCE_EXPIRED(), it will deleted, so that a new - * arp request will need to be triggered from an ND_INITIAL nce. - * - * Note that the nce state transition follows the pattern: - * ND_INITIAL -> ND_INCOMPLETE -> ND_REACHABLE - * after which the nce is deleted when it has expired. - * - * nce_last is the timestamp that indicates when the nce_res_mp in the - * nce_t was last updated to a valid link-layer address. nce_last gets - * modified/updated : - * - when the nce is created - * - every time we get a sane arp response for the nce. - */ -#define NCE_EXPIRED(nce, ipst) (nce->nce_last > 0 && \ - ((nce->nce_flags & NCE_F_PERMANENT) == 0) && \ - ((TICK_TO_MSEC(lbolt64) - nce->nce_last) > \ - (ipst)->ips_ip_ire_arp_interval)) - #endif /* _KERNEL */ /* ICMP types */ @@ -560,7 +395,17 @@ typedef struct ipha_s { #define IPH_DF 0x4000 /* Don't fragment */ #define IPH_MF 0x2000 /* More fragments to come */ #define IPH_OFFSET 0x1FFF /* Where the offset lives */ -#define IPH_FRAG_HDR 0x8000 /* IPv6 don't fragment bit */ + +/* Byte-order specific values */ +#ifdef _BIG_ENDIAN +#define IPH_DF_HTONS 0x4000 /* Don't fragment */ +#define IPH_MF_HTONS 0x2000 /* More fragments to come */ +#define IPH_OFFSET_HTONS 0x1FFF /* Where the offset lives */ +#else +#define IPH_DF_HTONS 0x0040 /* Don't fragment */ +#define IPH_MF_HTONS 0x0020 /* More fragments to come */ +#define IPH_OFFSET_HTONS 0xFF1F /* Where the offset lives */ +#endif /* ECN code points for IPv4 TOS byte and IPv6 traffic class octet. */ #define IPH_ECN_NECT 0x0 /* Not ECN-Capable Transport */ @@ -571,10 +416,8 @@ typedef struct ipha_s { struct ill_s; typedef void ip_v6intfid_func_t(struct ill_s *, in6_addr_t *); -typedef boolean_t ip_v6mapinfo_func_t(uint_t, uint8_t *, uint8_t *, uint32_t *, - in6_addr_t *); -typedef boolean_t ip_v4mapinfo_func_t(uint_t, uint8_t *, uint8_t *, uint32_t *, - ipaddr_t *); +typedef void ip_v6mapinfo_func_t(struct ill_s *, uchar_t *, uchar_t *); +typedef void ip_v4mapinfo_func_t(struct ill_s *, uchar_t *, uchar_t *); /* IP Mac info structure */ typedef struct ip_m_s { @@ -582,8 +425,8 @@ typedef struct ip_m_s { int ip_m_type; /* From <net/if_types.h> */ t_uscalar_t ip_m_ipv4sap; t_uscalar_t ip_m_ipv6sap; - ip_v4mapinfo_func_t *ip_m_v4mapinfo; - ip_v6mapinfo_func_t *ip_m_v6mapinfo; + ip_v4mapinfo_func_t *ip_m_v4mapping; + ip_v6mapinfo_func_t *ip_m_v6mapping; ip_v6intfid_func_t *ip_m_v6intfid; ip_v6intfid_func_t *ip_m_v6destintfid; } ip_m_t; @@ -591,20 +434,14 @@ typedef struct ip_m_s { /* * The following functions attempt to reduce the link layer dependency * of the IP stack. The current set of link specific operations are: - * a. map from IPv4 class D (224.0/4) multicast address range to the link - * layer multicast address range. - * b. map from IPv6 multicast address range (ff00::/8) to the link - * layer multicast address range. - * c. derive the default IPv6 interface identifier from the interface. - * d. derive the default IPv6 destination interface identifier from + * a. map from IPv4 class D (224.0/4) multicast address range or the + * IPv6 multicast address range (ff00::/8) to the link layer multicast + * address. + * b. derive the default IPv6 interface identifier from the interface. + * c. derive the default IPv6 destination interface identifier from * the interface (point-to-point only). */ -#define MEDIA_V4MINFO(ip_m, plen, bphys, maddr, hwxp, v4ptr) \ - (((ip_m)->ip_m_v4mapinfo != NULL) && \ - (*(ip_m)->ip_m_v4mapinfo)(plen, bphys, maddr, hwxp, v4ptr)) -#define MEDIA_V6MINFO(ip_m, plen, bphys, maddr, hwxp, v6ptr) \ - (((ip_m)->ip_m_v6mapinfo != NULL) && \ - (*(ip_m)->ip_m_v6mapinfo)(plen, bphys, maddr, hwxp, v6ptr)) +extern void ip_mcast_mapping(struct ill_s *, uchar_t *, uchar_t *); /* ip_m_v6*intfid return void and are never NULL */ #define MEDIA_V6INTFID(ip_m, ill, v6ptr) (ip_m)->ip_m_v6intfid(ill, v6ptr) #define MEDIA_V6DESTINTFID(ip_m, ill, v6ptr) \ @@ -616,107 +453,38 @@ typedef struct ip_m_s { #define IRE_LOCAL 0x0004 /* Route entry for local address */ #define IRE_LOOPBACK 0x0008 /* Route entry for loopback address */ #define IRE_PREFIX 0x0010 /* Route entry for prefix routes */ +#ifndef _KERNEL +/* Keep so user-level still compiles */ #define IRE_CACHE 0x0020 /* Cached Route entry */ +#endif #define IRE_IF_NORESOLVER 0x0040 /* Route entry for local interface */ /* net without any address mapping. */ #define IRE_IF_RESOLVER 0x0080 /* Route entry for local interface */ /* net with resolver. */ #define IRE_HOST 0x0100 /* Host route entry */ +/* Keep so user-level still compiles */ #define IRE_HOST_REDIRECT 0x0200 /* only used for T_SVR4_OPTMGMT_REQ */ +#define IRE_IF_CLONE 0x0400 /* Per host clone of IRE_IF */ +#define IRE_MULTICAST 0x0800 /* Special - not in table */ +#define IRE_NOROUTE 0x1000 /* Special - not in table */ #define IRE_INTERFACE (IRE_IF_NORESOLVER | IRE_IF_RESOLVER) -#define IRE_OFFSUBNET (IRE_DEFAULT | IRE_PREFIX | IRE_HOST) -#define IRE_CACHETABLE (IRE_CACHE | IRE_BROADCAST | IRE_LOCAL | \ - IRE_LOOPBACK) -#define IRE_FORWARDTABLE (IRE_INTERFACE | IRE_OFFSUBNET) - -/* - * If an IRE is marked with IRE_MARK_CONDEMNED, the last walker of - * the bucket should delete this IRE from this bucket. - */ -#define IRE_MARK_CONDEMNED 0x0001 - -/* - * An IRE with IRE_MARK_PMTU has ire_max_frag set from an ICMP error. - */ -#define IRE_MARK_PMTU 0x0002 - -/* - * An IRE with IRE_MARK_TESTHIDDEN is used by in.mpathd for test traffic. It - * can only be looked up by requesting MATCH_IRE_MARK_TESTHIDDEN. - */ -#define IRE_MARK_TESTHIDDEN 0x0004 - -/* - * An IRE with IRE_MARK_NOADD is created in ip_newroute_ipif when the outgoing - * interface is specified by e.g. IP_PKTINFO. The IRE is not added to the IRE - * cache table. - */ -#define IRE_MARK_NOADD 0x0008 /* Mark not to add ire in cache */ - -/* - * IRE marked with IRE_MARK_TEMPORARY means that this IRE has been used - * either for forwarding a packet or has not been used for sending - * traffic on TCP connections terminated on this system. In both - * cases, this IRE is the first to go when IRE is being cleaned up. - */ -#define IRE_MARK_TEMPORARY 0x0010 - -/* - * IRE marked with IRE_MARK_USESRC_CHECK means that while adding an IRE with - * this mark, additional atomic checks need to be performed. For eg: by the - * time an IRE_CACHE is created, sent up to ARP and then comes back to IP; the - * usesrc grouping could have changed in which case we want to fail adding - * the IRE_CACHE entry - */ -#define IRE_MARK_USESRC_CHECK 0x0020 - -/* - * IRE_MARK_PRIVATE_ADDR is used for IP_NEXTHOP. When IP_NEXTHOP is set, the - * routing table lookup for the destination is bypassed and the packet is - * sent directly to the specified nexthop. The associated IRE_CACHE entries - * should be marked with IRE_MARK_PRIVATE_ADDR flag so that they don't show up - * in regular ire cache lookups. - */ -#define IRE_MARK_PRIVATE_ADDR 0x0040 +#define IRE_IF_ALL (IRE_IF_NORESOLVER | IRE_IF_RESOLVER | \ + IRE_IF_CLONE) +#define IRE_OFFSUBNET (IRE_DEFAULT | IRE_PREFIX | IRE_HOST) +#define IRE_OFFLINK IRE_OFFSUBNET /* - * When we send an ARP resolution query for the nexthop gateway's ire, - * we use esballoc to create the ire_t in the AR_ENTRY_QUERY mblk - * chain, and mark its ire_marks with IRE_MARK_UNCACHED. This flag - * indicates that information from ARP has not been transferred to a - * permanent IRE_CACHE entry. The flag is reset only when the - * information is successfully transferred to an ire_cache entry (in - * ire_add()). Attempting to free the AR_ENTRY_QUERY mblk chain prior - * to ire_add (e.g., from arp, or from ip`ip_wput_nondata) will - * require that the resources (incomplete ire_cache and/or nce) must - * be cleaned up. The free callback routine (ire_freemblk()) checks - * for IRE_MARK_UNCACHED to see if any resources that are pinned down - * will need to be cleaned up or not. + * Note that we view IRE_NOROUTE as ONLINK since we can "send" to them without + * going through a router; the result of sending will be an error/icmp error. */ - -#define IRE_MARK_UNCACHED 0x0080 - -/* - * The comment below (and for other netstack_t references) refers - * to the fact that we only do netstack_hold in particular cases, - * such as the references from open streams (ill_t and conn_t's - * pointers). Internally within IP we rely on IP's ability to cleanup e.g. - * ire_t's when an ill goes away. - */ -typedef struct ire_expire_arg_s { - int iea_flush_flag; - ip_stack_t *iea_ipst; /* Does not have a netstack_hold */ -} ire_expire_arg_t; - -/* Flags with ire_expire routine */ -#define FLUSH_ARP_TIME 0x0001 /* ARP info potentially stale timer */ -#define FLUSH_REDIRECT_TIME 0x0002 /* Redirects potentially stale */ -#define FLUSH_MTU_TIME 0x0004 /* Include path MTU per RFC 1191 */ +#define IRE_ONLINK (IRE_IF_ALL|IRE_LOCAL|IRE_LOOPBACK| \ + IRE_BROADCAST|IRE_MULTICAST|IRE_NOROUTE) /* Arguments to ire_flush_cache() */ #define IRE_FLUSH_DELETE 0 #define IRE_FLUSH_ADD 1 +#define IRE_FLUSH_GWCHANGE 2 /* * Open/close synchronization flags. @@ -724,31 +492,21 @@ typedef struct ire_expire_arg_s { * depends on the atomic 32 bit access to that field. */ #define CONN_CLOSING 0x01 /* ip_close waiting for ip_wsrv */ -#define CONN_IPSEC_LOAD_WAIT 0x02 /* waiting for load */ -#define CONN_CONDEMNED 0x04 /* conn is closing, no more refs */ -#define CONN_INCIPIENT 0x08 /* conn not yet visible, no refs */ -#define CONN_QUIESCED 0x10 /* conn is now quiescent */ - -/* Used to check connection state flags before caching the IRE */ -#define CONN_CACHE_IRE(connp) \ - (!((connp)->conn_state_flags & (CONN_CLOSING|CONN_CONDEMNED))) - -/* - * Parameter to ip_output giving the identity of the caller. - * IP_WSRV means the packet was enqueued in the STREAMS queue - * due to flow control and is now being reprocessed in the context of - * the STREAMS service procedure, consequent to flow control relief. - * IRE_SEND means the packet is being reprocessed consequent to an - * ire cache creation and addition and this may or may not be happening - * in the service procedure context. Anything other than the above 2 - * cases is identified as IP_WPUT. Most commonly this is the case of - * packets coming down from the application. +#define CONN_CONDEMNED 0x02 /* conn is closing, no more refs */ +#define CONN_INCIPIENT 0x04 /* conn not yet visible, no refs */ +#define CONN_QUIESCED 0x08 /* conn is now quiescent */ +#define CONN_UPDATE_ILL 0x10 /* conn_update_ill in progress */ + +/* + * Flags for dce_flags field. Specifies which information has been set. + * dce_ident is always present, but the other ones are identified by the flags. */ -#ifdef _KERNEL -#define IP_WSRV 1 /* Called from ip_wsrv */ -#define IP_WPUT 2 /* Called from ip_wput */ -#define IRE_SEND 3 /* Called from ire_send */ +#define DCEF_DEFAULT 0x0001 /* Default DCE - no pmtu or uinfo */ +#define DCEF_PMTU 0x0002 /* Different than interface MTU */ +#define DCEF_UINFO 0x0004 /* dce_uinfo set */ +#define DCEF_TOO_SMALL_PMTU 0x0008 /* Smaller than IPv4/IPv6 MIN */ +#ifdef _KERNEL /* * Extra structures need for per-src-addr filtering (IGMPv3/MLDv2) */ @@ -786,90 +544,80 @@ typedef struct mrec_s { } mrec_t; /* Group membership list per upper conn */ + /* - * XXX add ilg info for ifaddr/ifindex. - * XXX can we make ilg survive an ifconfig unplumb + plumb - * by setting the ipif/ill to NULL and recover that later? + * We record the multicast information from the socket option in + * ilg_ifaddr/ilg_ifindex. This allows rejoining the group in the case when + * the ifaddr (or ifindex) disappears and later reappears, potentially on + * a different ill. The IPv6 multicast socket options and ioctls all specify + * the interface using an ifindex. For IPv4 some socket options/ioctls use + * the interface address and others use the index. We record here the method + * that was actually used (and leave the other of ilg_ifaddr or ilg_ifindex) + * at zero so that we can rejoin the way the application intended. * - * ilg_ipif is used by IPv4 as multicast groups are joined using an interface - * address (ipif). - * ilg_ill is used by IPv6 as multicast groups are joined using an interface - * index (phyint->phyint_ifindex). - * ilg_ill is NULL for IPv4 and ilg_ipif is NULL for IPv6. + * We track the ill on which we will or already have joined an ilm using + * ilg_ill. When we have succeeded joining the ilm and have a refhold on it + * then we set ilg_ilm. Thus intentionally there is a window where ilg_ill is + * set and ilg_ilm is not set. This allows clearing ilg_ill as a signal that + * the ill is being unplumbed and the ilm should be discarded. * * ilg records the state of multicast memberships of a socket end point. * ilm records the state of multicast memberships with the driver and is * maintained per interface. * - * There is no direct link between a given ilg and ilm. If the - * application has joined a group G with ifindex I, we will have - * an ilg with ilg_v6group and ilg_ill. There will be a corresponding - * ilm with ilm_ill/ilm_v6addr recording the multicast membership. - * To delete the membership: - * - * a) Search for ilg matching on G and I with ilg_v6group - * and ilg_ill. Delete ilg_ill. - * b) Search the corresponding ilm matching on G and I with - * ilm_v6addr and ilm_ill. Delete ilm. - * - * For IPv4 the only difference is that we look using ipifs, not ills. + * The ilg state is protected by conn_ilg_lock. + * The ilg will not be freed until ilg_refcnt drops to zero. */ - -/* - * The ilg_t and ilm_t members are protected by ipsq. They can be changed only - * by a thread executing in the ipsq. In other words add/delete of a - * multicast group has to execute in the ipsq. - */ -#define ILG_DELETED 0x1 /* ilg_flags */ typedef struct ilg_s { + struct ilg_s *ilg_next; + struct ilg_s **ilg_ptpn; + struct conn_s *ilg_connp; /* Back pointer to get lock */ in6_addr_t ilg_v6group; - struct ipif_s *ilg_ipif; /* Logical interface we are member on */ - struct ill_s *ilg_ill; /* Used by IPv6 */ - uint_t ilg_flags; + ipaddr_t ilg_ifaddr; /* For some IPv4 cases */ + uint_t ilg_ifindex; /* IPv6 and some other IPv4 cases */ + struct ill_s *ilg_ill; /* Where ilm is joined. No refhold */ + struct ilm_s *ilg_ilm; /* With ilm_refhold */ + uint_t ilg_refcnt; mcast_record_t ilg_fmode; /* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */ slist_t *ilg_filter; + boolean_t ilg_condemned; /* Conceptually deleted */ } ilg_t; /* * Multicast address list entry for ill. - * ilm_ipif is used by IPv4 as multicast groups are joined using ipif. - * ilm_ill is used by IPv6 as multicast groups are joined using ill. - * ilm_ill is NULL for IPv4 and ilm_ipif is NULL for IPv6. + * ilm_ill is used by IPv4 and IPv6 + * + * The ilm state (and other multicast state on the ill) is protected by + * ill_mcast_lock. Operations that change state on both an ilg and ilm + * in addition use ill_mcast_serializer to ensure that we can't have + * interleaving between e.g., add and delete operations for the same conn_t, + * group, and ill. * * The comment below (and for other netstack_t references) refers * to the fact that we only do netstack_hold in particular cases, - * such as the references from open streams (ill_t and conn_t's + * such as the references from open endpoints (ill_t and conn_t's * pointers). Internally within IP we rely on IP's ability to cleanup e.g. * ire_t's when an ill goes away. */ -#define ILM_DELETED 0x1 /* ilm_flags */ typedef struct ilm_s { in6_addr_t ilm_v6addr; int ilm_refcnt; uint_t ilm_timer; /* IGMP/MLD query resp timer, in msec */ - struct ipif_s *ilm_ipif; /* Back pointer to ipif for IPv4 */ struct ilm_s *ilm_next; /* Linked list for each ill */ uint_t ilm_state; /* state of the membership */ - struct ill_s *ilm_ill; /* Back pointer to ill for IPv6 */ - uint_t ilm_flags; - boolean_t ilm_notify_driver; /* Need to notify the driver */ + struct ill_s *ilm_ill; /* Back pointer to ill - ill_ilm_cnt */ zoneid_t ilm_zoneid; int ilm_no_ilg_cnt; /* number of joins w/ no ilg */ mcast_record_t ilm_fmode; /* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */ slist_t *ilm_filter; /* source filter list */ slist_t *ilm_pendsrcs; /* relevant src addrs for pending req */ rtx_state_t ilm_rtx; /* SCR retransmission state */ + ipaddr_t ilm_ifaddr; /* For IPv4 netstat */ ip_stack_t *ilm_ipst; /* Does not have a netstack_hold */ } ilm_t; #define ilm_addr V4_PART_OF_V6(ilm_v6addr) -typedef struct ilm_walker { - struct ill_s *ilw_ill; /* associated ill */ - struct ill_s *ilw_ipmp_ill; /* associated ipmp ill (if any) */ - struct ill_s *ilw_walk_ill; /* current ill being walked */ -} ilm_walker_t; - /* * Soft reference to an IPsec SA. * @@ -898,40 +646,28 @@ typedef struct ipsa_ref_s * In the presence of IPsec policy, fully-bound conn's bind a connection * to more than just the 5-tuple, but also a specific IPsec action and * identity-pair. - * - * As an optimization, we also cache soft references to IPsec SA's - * here so that we can fast-path around most of the work needed for + * The identity pair is accessed from both the receive and transmit side + * hence it is maintained in the ipsec_latch_t structure. conn_latch and + * ixa_ipsec_latch points to it. + * The policy and actions are stored in conn_latch_in_policy and + * conn_latch_in_action for the inbound side, and in ixa_ipsec_policy and + * ixa_ipsec_action for the transmit side. + * + * As an optimization, we also cache soft references to IPsec SA's in + * ip_xmit_attr_t so that we can fast-path around most of the work needed for * outbound IPsec SA selection. - * - * Were it not for TCP's detached connections, this state would be - * in-line in conn_t; instead, this is in a separate structure so it - * can be handed off to TCP when a connection is detached. */ typedef struct ipsec_latch_s { kmutex_t ipl_lock; uint32_t ipl_refcnt; - uint64_t ipl_unique; - struct ipsec_policy_s *ipl_in_policy; /* latched policy (in) */ - struct ipsec_policy_s *ipl_out_policy; /* latched policy (out) */ - struct ipsec_action_s *ipl_in_action; /* latched action (in) */ - struct ipsec_action_s *ipl_out_action; /* latched action (out) */ - cred_t *ipl_local_id; struct ipsid_s *ipl_local_cid; struct ipsid_s *ipl_remote_cid; unsigned int - ipl_out_action_latched : 1, - ipl_in_action_latched : 1, - ipl_out_policy_latched : 1, - ipl_in_policy_latched : 1, - ipl_ids_latched : 1, - ipl_pad_to_bit_31 : 27; - - ipsa_ref_t ipl_ref[2]; /* 0: ESP, 1: AH */ - + ipl_pad_to_bit_31 : 31; } ipsec_latch_t; #define IPLATCH_REFHOLD(ipl) { \ @@ -939,97 +675,19 @@ typedef struct ipsec_latch_s ASSERT((ipl)->ipl_refcnt != 0); \ } -#define IPLATCH_REFRELE(ipl, ns) { \ +#define IPLATCH_REFRELE(ipl) { \ ASSERT((ipl)->ipl_refcnt != 0); \ membar_exit(); \ if (atomic_add_32_nv(&(ipl)->ipl_refcnt, -1) == 0) \ - iplatch_free(ipl, ns); \ + iplatch_free(ipl); \ } /* * peer identity structure. */ - typedef struct conn_s conn_t; /* - * The old IP client structure "ipc_t" is gone. All the data is stored in the - * connection structure "conn_t" now. The mapping of old and new fields looks - * like this: - * - * ipc_ulp conn_ulp - * ipc_rq conn_rq - * ipc_wq conn_wq - * - * ipc_laddr conn_src - * ipc_faddr conn_rem - * ipc_v6laddr conn_srcv6 - * ipc_v6faddr conn_remv6 - * - * ipc_lport conn_lport - * ipc_fport conn_fport - * ipc_ports conn_ports - * - * ipc_policy conn_policy - * ipc_latch conn_latch - * - * ipc_irc_lock conn_lock - * ipc_ire_cache conn_ire_cache - * - * ipc_state_flags conn_state_flags - * ipc_outgoing_ill conn_outgoing_ill - * - * ipc_dontroute conn_dontroute - * ipc_loopback conn_loopback - * ipc_broadcast conn_broadcast - * ipc_reuseaddr conn_reuseaddr - * - * ipc_multicast_loop conn_multicast_loop - * ipc_multi_router conn_multi_router - * ipc_draining conn_draining - * - * ipc_did_putbq conn_did_putbq - * ipc_unspec_src conn_unspec_src - * ipc_policy_cached conn_policy_cached - * - * ipc_in_enforce_policy conn_in_enforce_policy - * ipc_out_enforce_policy conn_out_enforce_policy - * ipc_af_isv6 conn_af_isv6 - * ipc_pkt_isv6 conn_pkt_isv6 - * - * ipc_ipv6_recvpktinfo conn_ipv6_recvpktinfo - * - * ipc_ipv6_recvhoplimit conn_ipv6_recvhoplimit - * ipc_ipv6_recvhopopts conn_ipv6_recvhopopts - * ipc_ipv6_recvdstopts conn_ipv6_recvdstopts - * - * ipc_ipv6_recvrthdr conn_ipv6_recvrthdr - * ipc_ipv6_recvrtdstopts conn_ipv6_recvrtdstopts - * ipc_fully_bound conn_fully_bound - * - * ipc_recvif conn_recvif - * - * ipc_recvslla conn_recvslla - * ipc_acking_unbind conn_acking_unbind - * ipc_pad_to_bit_31 conn_pad_to_bit_31 - * - * ipc_proto conn_proto - * ipc_incoming_ill conn_incoming_ill - * ipc_pending_ill conn_pending_ill - * ipc_unbind_mp conn_unbind_mp - * ipc_ilg conn_ilg - * ipc_ilg_allocated conn_ilg_allocated - * ipc_ilg_inuse conn_ilg_inuse - * ipc_ilg_walker_cnt conn_ilg_walker_cnt - * ipc_refcv conn_refcv - * ipc_multicast_ipif conn_multicast_ipif - * ipc_multicast_ill conn_multicast_ill - * ipc_drain_next conn_drain_next - * ipc_drain_prev conn_drain_prev - * ipc_idl conn_idl - */ - -/* * This is used to match an inbound/outbound datagram with policy. */ typedef struct ipsec_selector { @@ -1069,22 +727,6 @@ typedef struct ipsec_selector { #define IPSEC_POLICY_MAX 5 /* Always max + 1. */ /* - * Folowing macro is used whenever the code does not know whether there - * is a M_CTL present in the front and it needs to examine the actual mp - * i.e the IP header. As a M_CTL message could be in the front, this - * extracts the packet into mp and the M_CTL mp into first_mp. If M_CTL - * mp is not present, both first_mp and mp point to the same message. - */ -#define EXTRACT_PKT_MP(mp, first_mp, mctl_present) \ - (first_mp) = (mp); \ - if ((mp)->b_datap->db_type == M_CTL) { \ - (mp) = (mp)->b_cont; \ - (mctl_present) = B_TRUE; \ - } else { \ - (mctl_present) = B_FALSE; \ - } - -/* * Check with IPSEC inbound policy if * * 1) per-socket policy is present - indicated by conn_in_enforce_policy. @@ -1113,11 +755,6 @@ typedef struct ipsec_selector { /* * Information cached in IRE for upper layer protocol (ULP). - * - * Notice that ire_max_frag is not included in the iulp_t structure, which - * it may seem that it should. But ire_max_frag cannot really be cached. It - * is fixed for each interface. For MTU found by PMTUd, we may want to cache - * it. But currently, we do not do that. */ typedef struct iulp_s { boolean_t iulp_set; /* Is any metric set? */ @@ -1128,17 +765,21 @@ typedef struct iulp_s { uint32_t iulp_rpipe; /* Receive pipe size. */ uint32_t iulp_rtomax; /* Max round trip timeout. */ uint32_t iulp_sack; /* Use SACK option (TCP)? */ + uint32_t iulp_mtu; /* Setable with routing sockets */ + uint32_t iulp_tstamp_ok : 1, /* Use timestamp option (TCP)? */ iulp_wscale_ok : 1, /* Use window scale option (TCP)? */ iulp_ecn_ok : 1, /* Enable ECN (for TCP)? */ iulp_pmtud_ok : 1, /* Enable PMTUd? */ - iulp_not_used : 28; -} iulp_t; + /* These three are passed out by ip_set_destination */ + iulp_localnet: 1, /* IRE_ONLINK */ + iulp_loopback: 1, /* IRE_LOOPBACK */ + iulp_local: 1, /* IRE_LOCAL */ -/* Zero iulp_t. */ -extern const iulp_t ire_uinfo_null; + iulp_not_used : 25; +} iulp_t; /* * The conn drain list structure (idl_t). @@ -1173,7 +814,6 @@ struct idl_tx_list_s { struct idl_s { conn_t *idl_conn; /* Head of drain list */ kmutex_t idl_lock; /* Lock for this list */ - conn_t *idl_conn_draining; /* conn that is draining */ uint32_t idl_repeat : 1, /* Last conn must re-enable */ /* drain list again */ @@ -1182,36 +822,38 @@ struct idl_s { }; #define CONN_DRAIN_LIST_LOCK(connp) (&((connp)->conn_idl->idl_lock)) + /* * Interface route structure which holds the necessary information to recreate - * routes that are tied to an interface (namely where ire_ipif != NULL). + * routes that are tied to an interface i.e. have ire_ill set. + * * These routes which were initially created via a routing socket or via the * SIOCADDRT ioctl may be gateway routes (RTF_GATEWAY being set) or may be - * traditional interface routes. When an interface comes back up after being - * marked down, this information will be used to recreate the routes. These - * are part of an mblk_t chain that hangs off of the IPIF (ipif_saved_ire_mp). + * traditional interface routes. When an ill comes back up after being + * down, this information will be used to recreate the routes. These + * are part of an mblk_t chain that hangs off of the ILL (ill_saved_ire_mp). */ typedef struct ifrt_s { ushort_t ifrt_type; /* Type of IRE */ in6_addr_t ifrt_v6addr; /* Address IRE represents. */ - in6_addr_t ifrt_v6gateway_addr; /* Gateway if IRE_OFFSUBNET */ - in6_addr_t ifrt_v6src_addr; /* Src addr if RTF_SETSRC */ + in6_addr_t ifrt_v6gateway_addr; /* Gateway if IRE_OFFLINK */ + in6_addr_t ifrt_v6setsrc_addr; /* Src addr if RTF_SETSRC */ in6_addr_t ifrt_v6mask; /* Mask for matching IRE. */ uint32_t ifrt_flags; /* flags related to route */ - uint_t ifrt_max_frag; /* MTU (next hop or path). */ - iulp_t ifrt_iulp_info; /* Cached IRE ULP info. */ + iulp_t ifrt_metrics; /* Routing socket metrics */ + zoneid_t ifrt_zoneid; /* zoneid for route */ } ifrt_t; #define ifrt_addr V4_PART_OF_V6(ifrt_v6addr) #define ifrt_gateway_addr V4_PART_OF_V6(ifrt_v6gateway_addr) -#define ifrt_src_addr V4_PART_OF_V6(ifrt_v6src_addr) #define ifrt_mask V4_PART_OF_V6(ifrt_v6mask) +#define ifrt_setsrc_addr V4_PART_OF_V6(ifrt_v6setsrc_addr) /* Number of IP addresses that can be hosted on a physical interface */ #define MAX_ADDRS_PER_IF 8192 /* * Number of Source addresses to be considered for source address - * selection. Used by ipif_select_source[_v6]. + * selection. Used by ipif_select_source_v4/v6. */ #define MAX_IPIF_SELECT_SOURCE 50 @@ -1245,16 +887,13 @@ typedef struct th_hash_s { #define IPIF_CONDEMNED 0x1 /* The ipif is being removed */ #define IPIF_CHANGING 0x2 /* A critcal ipif field is changing */ #define IPIF_SET_LINKLOCAL 0x10 /* transient flag during bringup */ -#define IPIF_ZERO_SOURCE 0x20 /* transient flag during bringup */ /* IP interface structure, one per local address */ typedef struct ipif_s { struct ipif_s *ipif_next; struct ill_s *ipif_ill; /* Back pointer to our ill */ int ipif_id; /* Logical unit number */ - uint_t ipif_mtu; /* Starts at ipif_ill->ill_max_frag */ in6_addr_t ipif_v6lcl_addr; /* Local IP address for this if. */ - in6_addr_t ipif_v6src_addr; /* Source IP address for this if. */ in6_addr_t ipif_v6subnet; /* Subnet prefix for this if. */ in6_addr_t ipif_v6net_mask; /* Net mask for this interface. */ in6_addr_t ipif_v6brd_addr; /* Broadcast addr for this interface. */ @@ -1262,47 +901,29 @@ typedef struct ipif_s { uint64_t ipif_flags; /* Interface flags. */ uint_t ipif_metric; /* BSD if metric, for compatibility. */ uint_t ipif_ire_type; /* IRE_LOCAL or IRE_LOOPBACK */ - mblk_t *ipif_arp_del_mp; /* Allocated at time arp comes up, to */ - /* prevent awkward out of mem */ - /* condition later */ - mblk_t *ipif_saved_ire_mp; /* Allocated for each extra */ - /* IRE_IF_NORESOLVER/IRE_IF_RESOLVER */ - /* on this interface so that they */ - /* can survive ifconfig down. */ - kmutex_t ipif_saved_ire_lock; /* Protects ipif_saved_ire_mp */ - - mrec_t *ipif_igmp_rpt; /* List of group memberships which */ - /* will be reported on. Used when */ - /* handling an igmp timeout. */ /* - * The packet counts in the ipif contain the sum of the - * packet counts in dead IREs that were affiliated with - * this ipif. + * The packet count in the ipif contain the sum of the + * packet counts in dead IRE_LOCAL/LOOPBACK for this ipif. */ - uint_t ipif_fo_pkt_count; /* Forwarded thru our dead IREs */ uint_t ipif_ib_pkt_count; /* Inbound packets for our dead IREs */ - uint_t ipif_ob_pkt_count; /* Outbound packets to our dead IREs */ + /* Exclusive bit fields, protected by ipsq_t */ unsigned int - ipif_multicast_up : 1, /* ipif_multicast_up() successful */ ipif_was_up : 1, /* ipif was up before */ ipif_addr_ready : 1, /* DAD is done */ ipif_was_dup : 1, /* DAD had failed */ - - ipif_joined_allhosts : 1, /* allhosts joined */ ipif_added_nce : 1, /* nce added for local address */ - ipif_pad_to_31 : 26; + + ipif_pad_to_31 : 28; + + ilm_t *ipif_allhosts_ilm; /* For all-nodes join */ + ilm_t *ipif_solmulti_ilm; /* For IPv6 solicited multicast join */ uint_t ipif_seqid; /* unique index across all ills */ uint_t ipif_state_flags; /* See IPIF_* flag defs above */ uint_t ipif_refcnt; /* active consistent reader cnt */ - /* Number of ire's and ilm's referencing this ipif */ - uint_t ipif_ire_cnt; - uint_t ipif_ilm_cnt; - - uint_t ipif_saved_ire_cnt; zoneid_t ipif_zoneid; /* zone ID number */ timeout_id_t ipif_recovery_id; /* Timer for DAD recovery */ boolean_t ipif_trace_disable; /* True when alloc fails */ @@ -1313,40 +934,12 @@ typedef struct ipif_s { * part of a group will be pointed to, and an ill cannot disappear * while it's in a group. */ - struct ill_s *ipif_bound_ill; - struct ipif_s *ipif_bound_next; /* bound ipif chain */ - boolean_t ipif_bound; /* B_TRUE if we successfully bound */ -} ipif_t; + struct ill_s *ipif_bound_ill; + struct ipif_s *ipif_bound_next; /* bound ipif chain */ + boolean_t ipif_bound; /* B_TRUE if we successfully bound */ -/* - * IPIF_FREE_OK() means that there are no incoming references - * to the ipif. Incoming refs would prevent the ipif from being freed. - */ -#define IPIF_FREE_OK(ipif) \ - ((ipif)->ipif_ire_cnt == 0 && (ipif)->ipif_ilm_cnt == 0) -/* - * IPIF_DOWN_OK() determines whether the incoming pointer reference counts - * would permit the ipif to be considered quiescent. In order for - * an ipif or ill to be considered quiescent, the ire and nce references - * to that ipif/ill must be zero. - * - * We do not require the ilm references to go to zero for quiescence - * because the quiescence checks are done to ensure that - * outgoing packets do not use addresses from the ipif/ill after it - * has been marked down, and incoming packets to addresses on a - * queiscent interface are rejected. This implies that all the - * ire/nce's using that source address need to be deleted and future - * creation of any ires using that source address must be prevented. - * Similarly incoming unicast packets destined to the 'down' address - * will not be accepted once that ire is gone. However incoming - * multicast packets are not destined to the downed address. - * They are only related to the ill in question. Furthermore - * the current API behavior allows applications to join or leave - * multicast groups, i.e., IP_ADD_MEMBERSHIP / LEAVE_MEMBERSHIP, using a - * down address. Therefore the ilm references are not included in - * the _DOWN_OK macros. - */ -#define IPIF_DOWN_OK(ipif) ((ipif)->ipif_ire_cnt == 0) + struct ire_s *ipif_ire_local; /* Our IRE_LOCAL or LOOPBACK */ +} ipif_t; /* * The following table lists the protection levels of the various members @@ -1371,9 +964,7 @@ typedef struct ipif_s { * ill_g_lock ill_g_lock * ipif_ill ipsq + down ipif write once * ipif_id ipsq + down ipif write once - * ipif_mtu ipsq * ipif_v6lcl_addr ipsq + down ipif up ipif - * ipif_v6src_addr ipsq + down ipif up ipif * ipif_v6subnet ipsq + down ipif up ipif * ipif_v6net_mask ipsq + down ipif up ipif * @@ -1383,28 +974,30 @@ typedef struct ipif_s { * ipif_metric * ipif_ire_type ipsq + down ill up ill * - * ipif_arp_del_mp ipsq ipsq - * ipif_saved_ire_mp ipif_saved_ire_lock ipif_saved_ire_lock - * ipif_igmp_rpt ipsq ipsq - * - * ipif_fo_pkt_count Approx * ipif_ib_pkt_count Approx - * ipif_ob_pkt_count Approx * * bit fields ill_lock ill_lock * + * ipif_allhosts_ilm ipsq ipsq + * ipif_solmulti_ilm ipsq ipsq + * * ipif_seqid ipsq Write once * * ipif_state_flags ill_lock ill_lock * ipif_refcnt ill_lock ill_lock - * ipif_ire_cnt ill_lock ill_lock - * ipif_ilm_cnt ill_lock ill_lock - * ipif_saved_ire_cnt - * * ipif_bound_ill ipsq + ipmp_lock ipsq OR ipmp_lock * ipif_bound_next ipsq ipsq * ipif_bound ipsq ipsq + * + * ipif_ire_local ipsq + ips_ill_g_lock ipsq OR ips_ill_g_lock + */ + +/* + * Return values from ip_laddr_verify_{v4,v6} */ +typedef enum { IPVL_UNICAST_UP, IPVL_UNICAST_DOWN, IPVL_MCAST, IPVL_BCAST, + IPVL_BAD} ip_laddr_t; + #define IP_TR_HASH(tid) ((((uintptr_t)tid) >> 6) & (IP_TR_HASH_MAX - 1)) @@ -1422,18 +1015,12 @@ typedef struct ipif_s { /* IPv4 compatibility macros */ #define ipif_lcl_addr V4_PART_OF_V6(ipif_v6lcl_addr) -#define ipif_src_addr V4_PART_OF_V6(ipif_v6src_addr) #define ipif_subnet V4_PART_OF_V6(ipif_v6subnet) #define ipif_net_mask V4_PART_OF_V6(ipif_v6net_mask) #define ipif_brd_addr V4_PART_OF_V6(ipif_v6brd_addr) #define ipif_pp_dst_addr V4_PART_OF_V6(ipif_v6pp_dst_addr) /* Macros for easy backreferences to the ill. */ -#define ipif_wq ipif_ill->ill_wq -#define ipif_rq ipif_ill->ill_rq -#define ipif_net_type ipif_ill->ill_net_type -#define ipif_ipif_up_count ipif_ill->ill_ipif_up_count -#define ipif_type ipif_ill->ill_type #define ipif_isv6 ipif_ill->ill_isv6 #define SIOCLIFADDR_NDX 112 /* ndx of SIOCLIFADDR in the ndx ioctl table */ @@ -1524,7 +1111,7 @@ typedef struct ipxop_s { boolean_t ipx_current_done; /* is the current operation done? */ int ipx_current_ioctl; /* current ioctl, or 0 if no ioctl */ ipif_t *ipx_current_ipif; /* ipif for current op */ - ipif_t *ipx_pending_ipif; /* ipif for ipsq_pending_mp */ + ipif_t *ipx_pending_ipif; /* ipif for ipx_pending_mp */ mblk_t *ipx_pending_mp; /* current ioctl mp while waiting */ boolean_t ipx_forced; /* debugging aid */ #ifdef DEBUG @@ -1642,24 +1229,62 @@ typedef struct irb { krwlock_t irb_lock; /* Protect this bucket */ uint_t irb_refcnt; /* Protected by irb_lock */ uchar_t irb_marks; /* CONDEMNED ires in this bucket ? */ -#define IRB_MARK_CONDEMNED 0x0001 -#define IRB_MARK_FTABLE 0x0002 +#define IRB_MARK_CONDEMNED 0x0001 /* Contains some IRE_IS_CONDEMNED */ +#define IRB_MARK_DYNAMIC 0x0002 /* Dynamically allocated */ + /* Once IPv6 uses radix then IRB_MARK_DYNAMIC will be always be set */ uint_t irb_ire_cnt; /* Num of active IRE in this bucket */ - uint_t irb_tmp_ire_cnt; /* Num of temporary IRE */ - struct ire_s *irb_rr_origin; /* origin for round-robin */ int irb_nire; /* Num of ftable ire's that ref irb */ ip_stack_t *irb_ipst; /* Does not have a netstack_hold */ } irb_t; #define IRB2RT(irb) (rt_t *)((caddr_t)(irb) - offsetof(rt_t, rt_irb)) -/* The following are return values of ip_xmit_v4() */ -typedef enum { - SEND_PASSED = 0, /* sent packet out on wire */ - SEND_FAILED, /* sending of packet failed */ - LOOKUP_IN_PROGRESS, /* ire cache found, ARP resolution in progress */ - LLHDR_RESLV_FAILED /* macaddr resl of onlink dst or nexthop failed */ -} ipxmit_state_t; +/* Forward declarations */ +struct dce_s; +typedef struct dce_s dce_t; +struct ire_s; +typedef struct ire_s ire_t; +struct ncec_s; +typedef struct ncec_s ncec_t; +struct nce_s; +typedef struct nce_s nce_t; +struct ip_recv_attr_s; +typedef struct ip_recv_attr_s ip_recv_attr_t; +struct ip_xmit_attr_s; +typedef struct ip_xmit_attr_s ip_xmit_attr_t; + +struct tsol_ire_gw_secattr_s; +typedef struct tsol_ire_gw_secattr_s tsol_ire_gw_secattr_t; + +/* + * This is a structure for a one-element route cache that is passed + * by reference between ip_input and ill_inputfn. + */ +typedef struct { + ire_t *rtc_ire; + ipaddr_t rtc_ipaddr; + in6_addr_t rtc_ip6addr; +} rtc_t; + +/* + * Note: Temporarily use 64 bits, and will probably go back to 32 bits after + * more cleanup work is done. + */ +typedef uint64_t iaflags_t; + +/* The ill input function pointer type */ +typedef void (*pfillinput_t)(mblk_t *, void *, void *, ip_recv_attr_t *, + rtc_t *); + +/* The ire receive function pointer type */ +typedef void (*pfirerecv_t)(ire_t *, mblk_t *, void *, ip_recv_attr_t *); + +/* The ire send and postfrag function pointer types */ +typedef int (*pfiresend_t)(ire_t *, mblk_t *, void *, + ip_xmit_attr_t *, uint32_t *); +typedef int (*pfirepostfrag_t)(mblk_t *, nce_t *, iaflags_t, uint_t, uint32_t, + zoneid_t, zoneid_t, uintptr_t *); + #define IP_V4_G_HEAD 0 #define IP_V6_G_HEAD 1 @@ -1733,26 +1358,12 @@ typedef union ill_g_head_u { /* * Capabilities, possible flags for ill_capabilities. */ - -#define ILL_CAPAB_AH 0x01 /* IPsec AH acceleration */ -#define ILL_CAPAB_ESP 0x02 /* IPsec ESP acceleration */ -#define ILL_CAPAB_MDT 0x04 /* Multidata Transmit */ +#define ILL_CAPAB_LSO 0x04 /* Large Send Offload */ #define ILL_CAPAB_HCKSUM 0x08 /* Hardware checksumming */ #define ILL_CAPAB_ZEROCOPY 0x10 /* Zero-copy */ #define ILL_CAPAB_DLD 0x20 /* DLD capabilities */ #define ILL_CAPAB_DLD_POLL 0x40 /* Polling */ #define ILL_CAPAB_DLD_DIRECT 0x80 /* Direct function call */ -#define ILL_CAPAB_DLD_LSO 0x100 /* Large Segment Offload */ - -/* - * Per-ill Multidata Transmit capabilities. - */ -typedef struct ill_mdt_capab_s ill_mdt_capab_t; - -/* - * Per-ill IPsec capabilities. - */ -typedef struct ill_ipsec_capab_s ill_ipsec_capab_t; /* * Per-ill Hardware Checksumming capbilities. @@ -1775,15 +1386,18 @@ typedef struct ill_dld_capab_s ill_dld_capab_t; typedef struct ill_rx_ring ill_rx_ring_t; /* - * Per-ill Large Segment Offload capabilities. + * Per-ill Large Send Offload capabilities. */ typedef struct ill_lso_capab_s ill_lso_capab_t; /* The following are ill_state_flags */ #define ILL_LL_SUBNET_PENDING 0x01 /* Waiting for DL_INFO_ACK from drv */ #define ILL_CONDEMNED 0x02 /* No more new ref's to the ILL */ -#define ILL_CHANGING 0x04 /* ILL not globally visible */ -#define ILL_DL_UNBIND_IN_PROGRESS 0x08 /* UNBIND_REQ is sent */ +#define ILL_DL_UNBIND_IN_PROGRESS 0x04 /* UNBIND_REQ is sent */ +#define ILL_DOWN_IN_PROGRESS 0x08 /* ILL is going down - no new nce's */ +#define ILL_LL_BIND_PENDING 0x0020 /* XXX Reuse ILL_LL_SUBNET_PENDING ? */ +#define ILL_LL_UP 0x0040 +#define ILL_LL_DOWN 0x0080 /* Is this an ILL whose source address is used by other ILL's ? */ #define IS_USESRC_ILL(ill) \ @@ -1796,10 +1410,9 @@ typedef struct ill_lso_capab_s ill_lso_capab_t; ((ill)->ill_usesrc_grp_next != NULL)) /* Is this an virtual network interface (vni) ILL ? */ -#define IS_VNI(ill) \ - (((ill) != NULL) && \ +#define IS_VNI(ill) \ (((ill)->ill_phyint->phyint_flags & (PHYI_LOOPBACK|PHYI_VIRTUAL)) == \ - PHYI_VIRTUAL)) + PHYI_VIRTUAL) /* Is this a loopback ILL? */ #define IS_LOOPBACK(ill) \ @@ -1900,18 +1513,41 @@ typedef struct ipmp_grp_s { * ARP up-to-date as the active set of interfaces in the group changes. */ typedef struct ipmp_arpent_s { - mblk_t *ia_area_mp; /* AR_ENTRY_ADD pointer */ ipaddr_t ia_ipaddr; /* IP address for this entry */ boolean_t ia_proxyarp; /* proxy ARP entry? */ boolean_t ia_notified; /* ARP notified about this entry? */ list_node_t ia_node; /* next ARP entry in list */ + uint16_t ia_flags; /* nce_flags for the address */ + size_t ia_lladdr_len; + uchar_t *ia_lladdr; } ipmp_arpent_t; +struct arl_s; + +/* + * Per-ill capabilities. + */ +struct ill_hcksum_capab_s { + uint_t ill_hcksum_version; /* interface version */ + uint_t ill_hcksum_txflags; /* capabilities on transmit */ +}; + +struct ill_zerocopy_capab_s { + uint_t ill_zerocopy_version; /* interface version */ + uint_t ill_zerocopy_flags; /* capabilities */ +}; + +struct ill_lso_capab_s { + uint_t ill_lso_flags; /* capabilities */ + uint_t ill_lso_max; /* maximum size of payload */ +}; + /* * IP Lower level Structure. * Instance data structure in ip_open when there is a device below us. */ typedef struct ill_s { + pfillinput_t ill_inputfn; /* Fast input function selector */ ill_if_t *ill_ifptr; /* pointer to interface type */ queue_t *ill_rq; /* Read queue. */ queue_t *ill_wq; /* Write queue. */ @@ -1922,6 +1558,8 @@ typedef struct ill_s { uint_t ill_ipif_up_count; /* Number of IPIFs currently up. */ uint_t ill_max_frag; /* Max IDU from DLPI. */ + uint_t ill_current_frag; /* Current IDU from DLPI. */ + uint_t ill_mtu; /* User-specified MTU; SIOCSLIFMTU */ char *ill_name; /* Our name. */ uint_t ill_ipif_dup_count; /* Number of duplicate addresses. */ uint_t ill_name_length; /* Name length, incl. terminator. */ @@ -1941,8 +1579,9 @@ typedef struct ill_s { uint8_t *ill_frag_ptr; /* Reassembly state. */ timeout_id_t ill_frag_timer_id; /* timeout id for the frag timer */ ipfb_t *ill_frag_hash_tbl; /* Fragment hash list head. */ - ipif_t *ill_pending_ipif; /* IPIF waiting for DL operation. */ + krwlock_t ill_mcast_lock; /* Protects multicast state */ + kmutex_t ill_mcast_serializer; /* Serialize across ilg and ilm state */ ilm_t *ill_ilm; /* Multicast membership for ill */ uint_t ill_global_timer; /* for IGMPv3/MLDv2 general queries */ int ill_mcast_type; /* type of router which is querier */ @@ -1955,22 +1594,20 @@ typedef struct ill_s { uint8_t ill_mcast_rv; /* IGMPv3/MLDv2 robustness variable */ int ill_mcast_qi; /* IGMPv3/MLDv2 query interval var */ - mblk_t *ill_pending_mp; /* IOCTL/DLPI awaiting completion. */ /* * All non-NULL cells between 'ill_first_mp_to_free' and * 'ill_last_mp_to_free' are freed in ill_delete. */ #define ill_first_mp_to_free ill_bcast_mp mblk_t *ill_bcast_mp; /* DLPI header for broadcasts. */ - mblk_t *ill_resolver_mp; /* Resolver template. */ mblk_t *ill_unbind_mp; /* unbind mp from ill_dl_up() */ mblk_t *ill_promiscoff_mp; /* for ill_leave_allmulti() */ mblk_t *ill_dlpi_deferred; /* b_next chain of control messages */ - mblk_t *ill_ardeact_mp; /* deact mp from ipmp_ill_activate() */ mblk_t *ill_dest_addr_mp; /* mblk which holds ill_dest_addr */ mblk_t *ill_replumb_mp; /* replumb mp from ill_replumb() */ mblk_t *ill_phys_addr_mp; /* mblk which holds ill_phys_addr */ -#define ill_last_mp_to_free ill_phys_addr_mp + mblk_t *ill_mcast_deferred; /* b_next chain of IGMP/MLD packets */ +#define ill_last_mp_to_free ill_mcast_deferred cred_t *ill_credp; /* opener's credentials */ uint8_t *ill_phys_addr; /* ill_phys_addr_mp->b_rptr + off */ @@ -1986,37 +1623,33 @@ typedef struct ill_s { ill_dlpi_style_set : 1, ill_ifname_pending : 1, - ill_join_allmulti : 1, ill_logical_down : 1, ill_dl_up : 1, - ill_up_ipifs : 1, + ill_note_link : 1, /* supports link-up notification */ ill_capab_reneg : 1, /* capability renegotiation to be done */ ill_dld_capab_inprog : 1, /* direct dld capab call in prog */ - ill_need_recover_multicast : 1, - ill_pad_to_bit_31 : 19; + + ill_replumbing : 1, + ill_arl_dlpi_pending : 1, + + ill_pad_to_bit_31 : 18; /* Following bit fields protected by ill_lock */ uint_t ill_fragtimer_executing : 1, ill_fragtimer_needrestart : 1, - ill_ilm_cleanup_reqd : 1, - ill_arp_closing : 1, - - ill_arp_bringup_pending : 1, - ill_arp_extend : 1, /* ARP has DAD extensions */ ill_manual_token : 1, /* system won't override ill_token */ ill_manual_linklocal : 1, /* system won't auto-conf linklocal */ - ill_pad_bit_31 : 24; + ill_pad_bit_31 : 28; /* * Used in SIOCSIFMUXID and SIOCGIFMUXID for 'ifconfig unplumb'. */ - int ill_arp_muxid; /* muxid returned from plink for arp */ - int ill_ip_muxid; /* muxid returned from plink for ip */ + int ill_muxid; /* muxid returned from plink */ /* Used for IP frag reassembly throttling on a per ILL basis. */ uint_t ill_ipf_gen; /* Generation of next fragment queue */ @@ -2033,20 +1666,13 @@ typedef struct ill_s { uint_t ill_dlpi_capab_state; /* State of capability query, IDCS_* */ uint_t ill_capab_pending_cnt; uint64_t ill_capabilities; /* Enabled capabilities, ILL_CAPAB_* */ - ill_mdt_capab_t *ill_mdt_capab; /* Multidata Transmit capabilities */ - ill_ipsec_capab_t *ill_ipsec_capab_ah; /* IPsec AH capabilities */ - ill_ipsec_capab_t *ill_ipsec_capab_esp; /* IPsec ESP capabilities */ ill_hcksum_capab_t *ill_hcksum_capab; /* H/W cksumming capabilities */ ill_zerocopy_capab_t *ill_zerocopy_capab; /* Zero-copy capabilities */ ill_dld_capab_t *ill_dld_capab; /* DLD capabilities */ ill_lso_capab_t *ill_lso_capab; /* Large Segment Offload capabilities */ mblk_t *ill_capab_reset_mp; /* Preallocated mblk for capab reset */ - /* - * Fields for IPv6 - */ uint8_t ill_max_hops; /* Maximum hops for any logical interface */ - uint_t ill_max_mtu; /* Maximum MTU for any logical interface */ uint_t ill_user_mtu; /* User-specified MTU via SIOCSLIFLNKINFO */ uint32_t ill_reachable_time; /* Value for ND algorithm in msec */ uint32_t ill_reachable_retrans_time; /* Value for ND algorithm msec */ @@ -2057,20 +1683,6 @@ typedef struct ill_s { uint32_t ill_xmit_count; /* ndp max multicast xmits */ mib2_ipIfStatsEntry_t *ill_ip_mib; /* ver indep. interface mib */ mib2_ipv6IfIcmpEntry_t *ill_icmp6_mib; /* Per interface mib */ - /* - * Following two mblks are allocated common to all - * the ipifs when the first interface is coming up. - * It is sent up to arp when the last ipif is coming - * down. - */ - mblk_t *ill_arp_down_mp; - mblk_t *ill_arp_del_mapping_mp; - /* - * Used for implementing IFF_NOARP. As IFF_NOARP is used - * to turn off for all the logicals, it is here instead - * of the ipif. - */ - mblk_t *ill_arp_on_mp; phyint_t *ill_phyint; uint64_t ill_flags; @@ -2094,11 +1706,11 @@ typedef struct ill_s { */ uint_t ill_ifname_pending_err; avl_node_t ill_avl_byppa; /* avl node based on ppa */ - void *ill_fastpath_list; /* both ire and nce hang off this */ + list_t ill_nce; /* pointer to nce_s list */ uint_t ill_refcnt; /* active refcnt by threads */ uint_t ill_ire_cnt; /* ires associated with this ill */ kcondvar_t ill_cv; - uint_t ill_ilm_walker_cnt; /* snmp ilm walkers */ + uint_t ill_ncec_cnt; /* ncecs associated with this ill */ uint_t ill_nce_cnt; /* nces associated with this ill */ uint_t ill_waiters; /* threads waiting in ipsq_enter */ /* @@ -2119,6 +1731,17 @@ typedef struct ill_s { void *ill_flownotify_mh; /* Tx flow ctl, mac cb handle */ uint_t ill_ilm_cnt; /* ilms referencing this ill */ uint_t ill_ipallmulti_cnt; /* ip_join_allmulti() calls */ + ilm_t *ill_ipallmulti_ilm; + + mblk_t *ill_saved_ire_mp; /* Allocated for each extra IRE */ + /* with ire_ill set so they can */ + /* survive the ill going down and up. */ + kmutex_t ill_saved_ire_lock; /* Protects ill_saved_ire_mp, cnt */ + uint_t ill_saved_ire_cnt; /* # entries */ + struct arl_ill_common_s *ill_common; + ire_t *ill_ire_multicast; /* IRE_MULTICAST for ill */ + clock_t ill_defend_start; /* start of 1 hour period */ + uint_t ill_defend_count; /* # of announce/defends per ill */ /* * IPMP fields. */ @@ -2131,6 +1754,8 @@ typedef struct ill_s { uint_t ill_bound_cnt; /* # of data addresses bound to ill */ ipif_t *ill_bound_ipif; /* ipif chain bound to ill */ timeout_id_t ill_refresh_tid; /* ill refresh retry timeout id */ + + uint32_t ill_mrouter_cnt; /* mrouter allmulti joins */ } ill_t; /* @@ -2139,15 +1764,17 @@ typedef struct ill_s { */ #define ILL_FREE_OK(ill) \ ((ill)->ill_ire_cnt == 0 && (ill)->ill_ilm_cnt == 0 && \ - (ill)->ill_nce_cnt == 0) + (ill)->ill_ncec_cnt == 0 && (ill)->ill_nce_cnt == 0) /* - * An ipif/ill can be marked down only when the ire and nce references + * An ipif/ill can be marked down only when the ire and ncec references * to that ipif/ill goes to zero. ILL_DOWN_OK() is a necessary condition * quiescence checks. See comments above IPIF_DOWN_OK for details * on why ires and nces are selectively considered for this macro. */ -#define ILL_DOWN_OK(ill) (ill->ill_ire_cnt == 0 && ill->ill_nce_cnt == 0) +#define ILL_DOWN_OK(ill) \ + (ill->ill_ire_cnt == 0 && ill->ill_ncec_cnt == 0 && \ + ill->ill_nce_cnt == 0) /* * The following table lists the protection levels of the various members @@ -2162,7 +1789,8 @@ typedef struct ill_s { * ill_error ipsq None * ill_ipif ill_g_lock + ipsq ill_g_lock OR ipsq * ill_ipif_up_count ill_lock + ipsq ill_lock OR ipsq - * ill_max_frag ipsq Write once + * ill_max_frag ill_lock ill_lock + * ill_current_frag ill_lock ill_lock * * ill_name ill_g_lock + ipsq Write once * ill_name_length ill_g_lock + ipsq Write once @@ -2179,23 +1807,22 @@ typedef struct ill_s { * * ill_frag_timer_id ill_lock ill_lock * ill_frag_hash_tbl ipsq up ill - * ill_ilm ipsq + ill_lock ill_lock - * ill_mcast_type ill_lock ill_lock - * ill_mcast_v1_time ill_lock ill_lock - * ill_mcast_v2_time ill_lock ill_lock - * ill_mcast_v1_tset ill_lock ill_lock - * ill_mcast_v2_tset ill_lock ill_lock - * ill_mcast_rv ill_lock ill_lock - * ill_mcast_qi ill_lock ill_lock - * ill_pending_mp ill_lock ill_lock - * - * ill_bcast_mp ipsq ipsq - * ill_resolver_mp ipsq only when ill is up + * ill_ilm ill_mcast_lock(WRITER) ill_mcast_lock(READER) + * ill_global_timer ill_mcast_lock(WRITER) ill_mcast_lock(READER) + * ill_mcast_type ill_mcast_lock(WRITER) ill_mcast_lock(READER) + * ill_mcast_v1_time ill_mcast_lock(WRITER) ill_mcast_lock(READER) + * ill_mcast_v2_time ill_mcast_lock(WRITER) ill_mcast_lock(READER) + * ill_mcast_v1_tset ill_mcast_lock(WRITER) ill_mcast_lock(READER) + * ill_mcast_v2_tset ill_mcast_lock(WRITER) ill_mcast_lock(READER) + * ill_mcast_rv ill_mcast_lock(WRITER) ill_mcast_lock(READER) + * ill_mcast_qi ill_mcast_lock(WRITER) ill_mcast_lock(READER) + * * ill_down_mp ipsq ipsq * ill_dlpi_deferred ill_lock ill_lock * ill_dlpi_pending ipsq + ill_lock ipsq or ill_lock or * absence of ipsq writer. * ill_phys_addr_mp ipsq + down ill only when ill is up + * ill_mcast_deferred ill_lock ill_lock * ill_phys_addr ipsq + down ill only when ill is up * ill_dest_addr_mp ipsq + down ill only when ill is up * ill_dest_addr ipsq + down ill only when ill is up @@ -2204,8 +1831,7 @@ typedef struct ill_s { * exclusive bit flags ipsq_t ipsq_t * shared bit flags ill_lock ill_lock * - * ill_arp_muxid ipsq Not atomic - * ill_ip_muxid ipsq Not atomic + * ill_muxid ipsq Not atomic * * ill_ipf_gen Not atomic * ill_frag_count atomics atomics @@ -2215,7 +1841,7 @@ typedef struct ill_s { * ill_dlpi_capab_state ipsq ipsq * ill_max_hops ipsq Not atomic * - * ill_max_mtu + * ill_mtu ill_lock None * * ill_user_mtu ipsq + ill_lock ill_lock * ill_reachable_time ipsq + ill_lock ill_lock @@ -2230,9 +1856,6 @@ typedef struct ill_s { * ill_xmit_count ipsq + down ill write once * ill_ip6_mib ipsq + down ill only when ill is up * ill_icmp6_mib ipsq + down ill only when ill is up - * ill_arp_down_mp ipsq ipsq - * ill_arp_del_mapping_mp ipsq ipsq - * ill_arp_on_mp ipsq ipsq * * ill_phyint ipsq, ill_g_lock, ill_lock Any of them * ill_flags ill_lock ill_lock @@ -2247,7 +1870,7 @@ typedef struct ill_s { * ill_refcnt ill_lock ill_lock * ill_ire_cnt ill_lock ill_lock * ill_cv ill_lock ill_lock - * ill_ilm_walker_cnt ill_lock ill_lock + * ill_ncec_cnt ill_lock ill_lock * ill_nce_cnt ill_lock ill_lock * ill_ilm_cnt ill_lock ill_lock * ill_src_ipif ill_g_lock ill_g_lock @@ -2256,8 +1879,12 @@ typedef struct ill_s { * ill_dhcpinit atomics atomics * ill_flownotify_mh write once write once * ill_capab_pending_cnt ipsq ipsq - * - * ill_bound_cnt ipsq ipsq + * ill_ipallmulti_cnt ill_lock ill_lock + * ill_ipallmulti_ilm ill_lock ill_lock + * ill_saved_ire_mp ill_saved_ire_lock ill_saved_ire_lock + * ill_saved_ire_cnt ill_saved_ire_lock ill_saved_ire_lock + * ill_arl ??? ??? + * ill_ire_multicast ipsq + quiescent none * ill_bound_ipif ipsq ipsq * ill_actnode ipsq + ipmp_lock ipsq OR ipmp_lock * ill_grpnode ipsq + ill_g_lock ipsq OR ill_g_lock @@ -2267,6 +1894,7 @@ typedef struct ill_s { * ill_refresh_tid ill_lock ill_lock * ill_grp (for IPMP ill) write once write once * ill_grp (for underlying ill) ipsq + ill_g_lock ipsq OR ill_g_lock + * ill_mrouter_cnt atomics atomics * * NOTE: It's OK to make heuristic decisions on an underlying interface * by using IS_UNDER_IPMP() or comparing ill_grp's raw pointer value. @@ -2311,7 +1939,6 @@ enum { IF_CMD = 1, LIF_CMD, ARP_CMD, XARP_CMD, MSFILT_CMD, MISC_CMD }; #define IPI_GET_CMD 0x8 /* branch to mi_copyout on success */ /* unused 0x10 */ #define IPI_NULL_BCONT 0x20 /* ioctl has not data and hence no b_cont */ -#define IPI_PASS_DOWN 0x40 /* pass this ioctl down when a module only */ extern ip_ioctl_cmd_t ip_ndx_ioctl_table[]; extern ip_ioctl_cmd_t ip_misc_ioctl_table[]; @@ -2362,6 +1989,430 @@ typedef struct ipndp_s { char *ip_ndp_name; } ipndp_t; +/* IXA Notification types */ +typedef enum { + IXAN_LSO, /* LSO capability change */ + IXAN_PMTU, /* PMTU change */ + IXAN_ZCOPY /* ZEROCOPY capability change */ +} ixa_notify_type_t; + +typedef uint_t ixa_notify_arg_t; + +typedef void (*ixa_notify_t)(void *, ip_xmit_attr_t *ixa, ixa_notify_type_t, + ixa_notify_arg_t); + +/* + * Attribute flags that are common to the transmit and receive attributes + */ +#define IAF_IS_IPV4 0x80000000 /* ipsec_*_v4 */ +#define IAF_TRUSTED_ICMP 0x40000000 /* ipsec_*_icmp_loopback */ +#define IAF_NO_LOOP_ZONEID_SET 0x20000000 /* Zone that shouldn't have */ + /* a copy */ +#define IAF_LOOPBACK_COPY 0x10000000 /* For multi and broadcast */ + +#define IAF_MASK 0xf0000000 /* Flags that are common */ + +/* + * Transmit side attributes used between the transport protocols and IP as + * well as inside IP. It is also used to cache information in the conn_t i.e. + * replaces conn_ire and the IPsec caching in the conn_t. + */ +struct ip_xmit_attr_s { + iaflags_t ixa_flags; /* IXAF_*. See below */ + + uint32_t ixa_free_flags; /* IXA_FREE_*. See below */ + uint32_t ixa_refcnt; /* Using atomics */ + + /* + * Always initialized independently of ixa_flags settings. + * Used by ip_xmit so we keep them up front for cache locality. + */ + uint32_t ixa_xmit_hint; /* For ECMP and GLD TX ring fanout */ + uint_t ixa_pktlen; /* Always set. For frag and stats */ + zoneid_t ixa_zoneid; /* Assumed always set */ + + /* Always set for conn_ip_output(); might be stale */ + /* + * Since TCP keeps the conn_t around past the process going away + * we need to use the "notr" (e.g, ire_refhold_notr) for ixa_ire, + * ixa_nce, and ixa_dce. + */ + ire_t *ixa_ire; /* Forwarding table entry */ + uint_t ixa_ire_generation; + nce_t *ixa_nce; /* Neighbor cache entry */ + dce_t *ixa_dce; /* Destination cache entry */ + uint_t ixa_dce_generation; + uint_t ixa_src_generation; /* If IXAF_VERIFY_SOURCE */ + + uint32_t ixa_src_preferences; /* prefs for src addr select */ + uint32_t ixa_pmtu; /* IXAF_VERIFY_PMTU */ + + /* Set by ULP if IXAF_VERIFY_PMTU; otherwise set by IP */ + uint32_t ixa_fragsize; + + int8_t ixa_use_min_mtu; /* IXAF_USE_MIN_MTU values */ + + pfirepostfrag_t ixa_postfragfn; /* Set internally in IP */ + + in6_addr_t ixa_nexthop_v6; /* IXAF_NEXTHOP_SET */ +#define ixa_nexthop_v4 V4_PART_OF_V6(ixa_nexthop_v6) + + zoneid_t ixa_no_loop_zoneid; /* IXAF_NO_LOOP_ZONEID_SET */ + + uint_t ixa_scopeid; /* For IPv6 link-locals */ + + uint_t ixa_broadcast_ttl; /* IXAF_BROACAST_TTL_SET */ + + uint_t ixa_multicast_ttl; /* Assumed set for multicast */ + uint_t ixa_multicast_ifindex; /* Assumed set for multicast */ + ipaddr_t ixa_multicast_ifaddr; /* Assumed set for multicast */ + + int ixa_raw_cksum_offset; /* If IXAF_SET_RAW_CKSUM */ + + uint32_t ixa_ident; /* For IPv6 fragment header */ + + /* + * Cached LSO information. + */ + ill_lso_capab_t ixa_lso_capab; /* Valid when IXAF_LSO_CAPAB */ + + uint64_t ixa_ipsec_policy_gen; /* Generation from iph_gen */ + /* + * The following IPsec fields are only initialized when + * IXAF_IPSEC_SECURE is set. Otherwise they contain garbage. + */ + ipsec_latch_t *ixa_ipsec_latch; /* Just the ids */ + struct ipsa_s *ixa_ipsec_ah_sa; /* Hard reference SA for AH */ + struct ipsa_s *ixa_ipsec_esp_sa; /* Hard reference SA for ESP */ + struct ipsec_policy_s *ixa_ipsec_policy; /* why are we here? */ + struct ipsec_action_s *ixa_ipsec_action; /* For reflected packets */ + ipsa_ref_t ixa_ipsec_ref[2]; /* Soft reference to SA */ + /* 0: ESP, 1: AH */ + + /* + * The selectors here are potentially different than the SPD rule's + * selectors, and we need to have both available for IKEv2. + * + * NOTE: "Source" and "Dest" are w.r.t. outbound datagrams. Ports can + * be zero, and the protocol number is needed to make the ports + * significant. + */ + uint16_t ixa_ipsec_src_port; /* Source port number of d-gram. */ + uint16_t ixa_ipsec_dst_port; /* Destination port number of d-gram. */ + uint8_t ixa_ipsec_icmp_type; /* ICMP type of d-gram */ + uint8_t ixa_ipsec_icmp_code; /* ICMP code of d-gram */ + + sa_family_t ixa_ipsec_inaf; /* Inner address family */ +#define IXA_MAX_ADDRLEN 4 /* Max addr len. (in 32-bit words) */ + uint32_t ixa_ipsec_insrc[IXA_MAX_ADDRLEN]; /* Inner src address */ + uint32_t ixa_ipsec_indst[IXA_MAX_ADDRLEN]; /* Inner dest address */ + uint8_t ixa_ipsec_insrcpfx; /* Inner source prefix */ + uint8_t ixa_ipsec_indstpfx; /* Inner destination prefix */ + + uint8_t ixa_ipsec_proto; /* IP protocol number for d-gram. */ + + /* Always initialized independently of ixa_flags settings */ + uint_t ixa_ifindex; /* Assumed always set */ + uint16_t ixa_ip_hdr_length; /* Points to ULP header */ + uint8_t ixa_protocol; /* Protocol number for ULP cksum */ + ts_label_t *ixa_tsl; /* Always set. NULL if not TX */ + ip_stack_t *ixa_ipst; /* Always set */ + uint32_t ixa_extra_ident; /* Set if LSO */ + cred_t *ixa_cred; /* For getpeerucred */ + pid_t ixa_cpid; /* For getpeerucred */ + +#ifdef DEBUG + kthread_t *ixa_curthread; /* For serialization assert */ +#endif + squeue_t *ixa_sqp; /* Set from conn_sqp as a hint */ + uintptr_t ixa_cookie; /* cookie to use for tx flow control */ + + /* + * Must be set by ULP if any of IXAF_VERIFY_LSO, IXAF_VERIFY_PMTU, + * or IXAF_VERIFY_ZCOPY is set. + */ + ixa_notify_t ixa_notify; /* Registered upcall notify function */ + void *ixa_notify_cookie; /* ULP cookie for ixa_notify */ +}; + +/* + * Flags to indicate which transmit attributes are set. + * Split into "xxx_SET" ones which indicate that the "xxx" field it set, and + * single flags. + */ +#define IXAF_REACH_CONF 0x00000001 /* Reachability confirmation */ +#define IXAF_BROADCAST_TTL_SET 0x00000002 /* ixa_broadcast_ttl valid */ +#define IXAF_SET_SOURCE 0x00000004 /* Replace if broadcast */ +#define IXAF_USE_MIN_MTU 0x00000008 /* IPV6_USE_MIN_MTU */ + +#define IXAF_DONTFRAG 0x00000010 /* IP*_DONTFRAG */ +#define IXAF_VERIFY_PMTU 0x00000020 /* ixa_pmtu/ixa_fragsize set */ +#define IXAF_PMTU_DISCOVERY 0x00000040 /* Create/use PMTU state */ +#define IXAF_MULTICAST_LOOP 0x00000080 /* IP_MULTICAST_LOOP */ + +#define IXAF_IPSEC_SECURE 0x00000100 /* Need IPsec processing */ +#define IXAF_UCRED_TSL 0x00000200 /* ixa_tsl from SCM_UCRED */ +#define IXAF_DONTROUTE 0x00000400 /* SO_DONTROUTE */ +#define IXAF_NO_IPSEC 0x00000800 /* Ignore policy */ + +#define IXAF_PMTU_TOO_SMALL 0x00001000 /* PMTU too small */ +#define IXAF_SET_ULP_CKSUM 0x00002000 /* Calculate ULP checksum */ +#define IXAF_VERIFY_SOURCE 0x00004000 /* Check that source is ok */ +#define IXAF_NEXTHOP_SET 0x00008000 /* ixa_nexthop set */ + +#define IXAF_PMTU_IPV4_DF 0x00010000 /* Set IPv4 DF */ +#define IXAF_NO_DEV_FLOW_CTL 0x00020000 /* Protocol needs no flow ctl */ +#define IXAF_NO_TTL_CHANGE 0x00040000 /* Internal to IP */ +#define IXAF_IPV6_ADD_FRAGHDR 0x00080000 /* Add fragment header */ + +#define IXAF_IPSEC_TUNNEL 0x00100000 /* Tunnel mode */ +#define IXAF_NO_PFHOOK 0x00200000 /* Skip xmit pfhook */ +#define IXAF_NO_TRACE 0x00400000 /* When back from ARP/ND */ +#define IXAF_SCOPEID_SET 0x00800000 /* ixa_scopeid set */ + +#define IXAF_MULTIRT_MULTICAST 0x01000000 /* MULTIRT for multicast */ +#define IXAF_NO_HW_CKSUM 0x02000000 /* Force software cksum */ +#define IXAF_SET_RAW_CKSUM 0x04000000 /* Use ixa_raw_cksum_offset */ +#define IXAF_IPSEC_GLOBAL_POLICY 0x08000000 /* Policy came from global */ + +/* Note the following uses bits 0x10000000 through 0x80000000 */ +#define IXAF_IS_IPV4 IAF_IS_IPV4 +#define IXAF_TRUSTED_ICMP IAF_TRUSTED_ICMP +#define IXAF_NO_LOOP_ZONEID_SET IAF_NO_LOOP_ZONEID_SET +#define IXAF_LOOPBACK_COPY IAF_LOOPBACK_COPY + +/* Note: use the upper 32 bits */ +#define IXAF_VERIFY_LSO 0x100000000 /* Check LSO capability */ +#define IXAF_LSO_CAPAB 0x200000000 /* Capable of LSO */ +#define IXAF_VERIFY_ZCOPY 0x400000000 /* Check Zero Copy capability */ +#define IXAF_ZCOPY_CAPAB 0x800000000 /* Capable of ZEROCOPY */ + +/* + * The normal flags for sending packets e.g., icmp errors + */ +#define IXAF_BASIC_SIMPLE_V4 (IXAF_SET_ULP_CKSUM | IXAF_IS_IPV4) +#define IXAF_BASIC_SIMPLE_V6 (IXAF_SET_ULP_CKSUM) + +/* + * Normally these fields do not have a hold. But in some cases they do, for + * instance when we've gone through ip_*_attr_to/from_mblk. + * We use ixa_free_flags to indicate that they have a hold and need to be + * released on cleanup. + */ +#define IXA_FREE_CRED 0x00000001 /* ixa_cred needs to be rele */ +#define IXA_FREE_TSL 0x00000002 /* ixa_tsl needs to be rele */ + +/* + * Simplistic way to set the ixa_xmit_hint for locally generated traffic + * and forwarded traffic. The shift amount are based on the size of the + * structs to discard the low order bits which don't have much if any variation + * (coloring in kmem_cache_alloc might provide some variation). + * + * Basing the locally generated hint on the address of the conn_t means that + * the packets from the same socket/connection do not get reordered. + * Basing the hint for forwarded traffic on the ill_ring_t means that + * packets from the same NIC+ring are likely to use the same outbound ring + * hence we get low contention on the ring in the transmitting driver. + */ +#define CONN_TO_XMIT_HINT(connp) ((uint32_t)(((uintptr_t)connp) >> 11)) +#define ILL_RING_TO_XMIT_HINT(ring) ((uint32_t)(((uintptr_t)ring) >> 7)) + +/* + * IP set Destination Flags used by function ip_set_destination, + * ip_attr_connect, and conn_connect. + */ +#define IPDF_ALLOW_MCBC 0x1 /* Allow multi/broadcast */ +#define IPDF_VERIFY_DST 0x2 /* Verify destination addr */ +#define IPDF_SELECT_SRC 0x4 /* Select source address */ +#define IPDF_LSO 0x8 /* Try LSO */ +#define IPDF_IPSEC 0x10 /* Set IPsec policy */ +#define IPDF_ZONE_IS_GLOBAL 0x20 /* From conn_zone_is_global */ +#define IPDF_ZCOPY 0x40 /* Try ZEROCOPY */ +#define IPDF_UNIQUE_DCE 0x80 /* Get a per-destination DCE */ + +/* + * Receive side attributes used between the transport protocols and IP as + * well as inside IP. + */ +struct ip_recv_attr_s { + iaflags_t ira_flags; /* See below */ + + uint32_t ira_free_flags; /* IRA_FREE_*. See below */ + + /* + * This is a hint for TCP SYN packets. + * Always initialized independently of ira_flags settings + */ + squeue_t *ira_sqp; + ill_rx_ring_t *ira_ring; /* Internal to IP */ + + /* For ip_accept_tcp when IRAF_TARGET_SQP is set */ + squeue_t *ira_target_sqp; + mblk_t *ira_target_sqp_mp; + + /* Always initialized independently of ira_flags settings */ + uint32_t ira_xmit_hint; /* For ECMP and GLD TX ring fanout */ + zoneid_t ira_zoneid; /* ALL_ZONES unless local delivery */ + uint_t ira_pktlen; /* Always set. For frag and stats */ + uint16_t ira_ip_hdr_length; /* Points to ULP header */ + uint8_t ira_protocol; /* Protocol number for ULP cksum */ + uint_t ira_rifindex; /* Received ifindex */ + uint_t ira_ruifindex; /* Received upper ifindex */ + ts_label_t *ira_tsl; /* Always set. NULL if not TX */ + /* + * ira_rill and ira_ill is set inside IP, but not when conn_recv is + * called; ULPs should use ira_ruifindex instead. + */ + ill_t *ira_rill; /* ill where packet came */ + ill_t *ira_ill; /* ill where IP address hosted */ + cred_t *ira_cred; /* For getpeerucred */ + pid_t ira_cpid; /* For getpeerucred */ + + /* Used when IRAF_VERIFIED_SRC is set; this source was ok */ + ipaddr_t ira_verified_src; + + /* + * The following IPsec fields are only initialized when + * IRAF_IPSEC_SECURE is set. Otherwise they contain garbage. + */ + struct ipsec_action_s *ira_ipsec_action; /* how we made it in.. */ + struct ipsa_s *ira_ipsec_ah_sa; /* SA for AH */ + struct ipsa_s *ira_ipsec_esp_sa; /* SA for ESP */ + + ipaddr_t ira_mroute_tunnel; /* IRAF_MROUTE_TUNNEL_SET */ + + zoneid_t ira_no_loop_zoneid; /* IRAF_NO_LOOP_ZONEID_SET */ + + uint32_t ira_esp_udp_ports; /* IRAF_ESP_UDP_PORTS */ + + /* + * For IP_RECVSLLA and ip_ndp_conflict/find_solicitation. + * Same size as max for sockaddr_dl + */ +#define IRA_L2SRC_SIZE 244 + uint8_t ira_l2src[IRA_L2SRC_SIZE]; /* If IRAF_L2SRC_SET */ + + /* + * Local handle that we use to do lazy setting of ira_l2src. + * We defer setting l2src until needed but we do before any + * ip_input pullupmsg or copymsg. + */ + struct mac_header_info_s *ira_mhip; /* Could be NULL */ +}; + +/* + * Flags to indicate which receive attributes are set. + */ +#define IRAF_SYSTEM_LABELED 0x00000001 /* is_system_labeled() */ +#define IRAF_IPV4_OPTIONS 0x00000002 /* Performance */ +#define IRAF_MULTICAST 0x00000004 /* Was multicast at L3 */ +#define IRAF_BROADCAST 0x00000008 /* Was broadcast at L3 */ +#define IRAF_MULTIBROADCAST (IRAF_MULTICAST|IRAF_BROADCAST) + +#define IRAF_LOOPBACK 0x00000010 /* Looped back by IP */ +#define IRAF_VERIFY_IP_CKSUM 0x00000020 /* Need to verify IP */ +#define IRAF_VERIFY_ULP_CKSUM 0x00000040 /* Need to verify TCP,UDP,etc */ +#define IRAF_SCTP_CSUM_ERR 0x00000080 /* sctp pkt has failed chksum */ + +#define IRAF_IPSEC_SECURE 0x00000100 /* Passed AH and/or ESP */ +#define IRAF_DHCP_UNICAST 0x00000200 +#define IRAF_IPSEC_DECAPS 0x00000400 /* Was packet decapsulated */ + /* from a matching inner packet? */ +#define IRAF_TARGET_SQP 0x00000800 /* ira_target_sqp is set */ +#define IRAF_VERIFIED_SRC 0x00001000 /* ira_verified_src set */ +#define IRAF_RSVP 0x00002000 /* RSVP packet for rsvpd */ +#define IRAF_MROUTE_TUNNEL_SET 0x00004000 /* From ip_mroute_decap */ +#define IRAF_PIM_REGISTER 0x00008000 /* From register_mforward */ + +#define IRAF_TX_MAC_EXEMPTABLE 0x00010000 /* Allow MAC_EXEMPT readdown */ +#define IRAF_TX_SHARED_ADDR 0x00020000 /* Arrived on ALL_ZONES addr */ +#define IRAF_ESP_UDP_PORTS 0x00040000 /* NAT-traversal packet */ +#define IRAF_NO_HW_CKSUM 0x00080000 /* Force software cksum */ + +#define IRAF_ICMP_ERROR 0x00100000 /* Send to conn_recvicmp */ +#define IRAF_ROUTER_ALERT 0x00200000 /* IPv6 router alert */ +#define IRAF_L2SRC_SET 0x00400000 /* ira_l2src has been set */ +#define IRAF_L2SRC_LOOPBACK 0x00800000 /* Came from us */ + +#define IRAF_L2DST_MULTICAST 0x01000000 /* Multicast at L2 */ +#define IRAF_L2DST_BROADCAST 0x02000000 /* Broadcast at L2 */ +/* Unused 0x04000000 */ +/* Unused 0x08000000 */ + +/* Below starts with 0x10000000 */ +#define IRAF_IS_IPV4 IAF_IS_IPV4 +#define IRAF_TRUSTED_ICMP IAF_TRUSTED_ICMP +#define IRAF_NO_LOOP_ZONEID_SET IAF_NO_LOOP_ZONEID_SET +#define IRAF_LOOPBACK_COPY IAF_LOOPBACK_COPY + +/* + * Normally these fields do not have a hold. But in some cases they do, for + * instance when we've gone through ip_*_attr_to/from_mblk. + * We use ira_free_flags to indicate that they have a hold and need to be + * released on cleanup. + */ +#define IRA_FREE_CRED 0x00000001 /* ira_cred needs to be rele */ +#define IRA_FREE_TSL 0x00000002 /* ira_tsl needs to be rele */ + +/* + * Optional destination cache entry for path MTU information, + * and ULP metrics. + */ +struct dce_s { + uint_t dce_generation; /* Changed since cached? */ + uint_t dce_flags; /* See below */ + uint_t dce_ipversion; /* IPv4/IPv6 version */ + uint32_t dce_pmtu; /* Path MTU if DCEF_PMTU */ + uint32_t dce_ident; /* Per destination IP ident. */ + iulp_t dce_uinfo; /* Metrics if DCEF_UINFO */ + + struct dce_s *dce_next; + struct dce_s **dce_ptpn; + struct dcb_s *dce_bucket; + + union { + in6_addr_t dceu_v6addr; + ipaddr_t dceu_v4addr; + } dce_u; +#define dce_v4addr dce_u.dceu_v4addr +#define dce_v6addr dce_u.dceu_v6addr + /* Note that for IPv6+IPMP we use the ifindex for the upper interface */ + uint_t dce_ifindex; /* For IPv6 link-locals */ + + kmutex_t dce_lock; + uint_t dce_refcnt; + uint64_t dce_last_change_time; /* Path MTU. In seconds */ + + ip_stack_t *dce_ipst; /* Does not have a netstack_hold */ +}; + +/* + * Values for dce_generation. + * + * If a DCE has DCE_GENERATION_CONDEMNED, the last dce_refrele should delete + * it. + * + * DCE_GENERATION_VERIFY is never stored in dce_generation but it is + * stored in places that cache DCE (such as ixa_dce_generation). + * It is used as a signal that the cache is stale and needs to be reverified. + */ +#define DCE_GENERATION_CONDEMNED 0 +#define DCE_GENERATION_VERIFY 1 +#define DCE_GENERATION_INITIAL 2 +#define DCE_IS_CONDEMNED(dce) \ + ((dce)->dce_generation == DCE_GENERATION_CONDEMNED) + + +/* + * Values for ips_src_generation. + * + * SRC_GENERATION_VERIFY is never stored in ips_src_generation but it is + * stored in places that cache IREs (ixa_src_generation). It is used as a + * signal that the cache is stale and needs to be reverified. + */ +#define SRC_GENERATION_VERIFY 0 +#define SRC_GENERATION_INITIAL 1 + /* * The kernel stores security attributes of all gateways in a database made * up of one or more tsol_gcdb_t elements. Each tsol_gcdb_t contains the @@ -2453,183 +2504,28 @@ extern kmutex_t gcgrp_lock; */ struct tsol_tnrhc; -typedef struct tsol_ire_gw_secattr_s { +struct tsol_ire_gw_secattr_s { kmutex_t igsa_lock; /* lock to protect following */ struct tsol_tnrhc *igsa_rhc; /* host entry for gateway */ tsol_gc_t *igsa_gc; /* for prefix IREs */ - tsol_gcgrp_t *igsa_gcgrp; /* for cache IREs */ -} tsol_ire_gw_secattr_t; - -/* - * Following are the macros to increment/decrement the reference - * count of the IREs and IRBs (ire bucket). - * - * 1) We bump up the reference count of an IRE to make sure that - * it does not get deleted and freed while we are using it. - * Typically all the lookup functions hold the bucket lock, - * and look for the IRE. If it finds an IRE, it bumps up the - * reference count before dropping the lock. Sometimes we *may* want - * to bump up the reference count after we *looked* up i.e without - * holding the bucket lock. So, the IRE_REFHOLD macro does not assert - * on the bucket lock being held. Any thread trying to delete from - * the hash bucket can still do so but cannot free the IRE if - * ire_refcnt is not 0. - * - * 2) We bump up the reference count on the bucket where the IRE resides - * (IRB), when we want to prevent the IREs getting deleted from a given - * hash bucket. This makes life easier for ire_walk type functions which - * wants to walk the IRE list, call a function, but needs to drop - * the bucket lock to prevent recursive rw_enters. While the - * lock is dropped, the list could be changed by other threads or - * the same thread could end up deleting the ire or the ire pointed by - * ire_next. IRE_REFHOLDing the ire or ire_next is not sufficient as - * a delete will still remove the ire from the bucket while we have - * dropped the lock and hence the ire_next would be NULL. Thus, we - * need a mechanism to prevent deletions from a given bucket. - * - * To prevent deletions, we bump up the reference count on the - * bucket. If the bucket is held, ire_delete just marks IRE_MARK_CONDEMNED - * both on the ire's ire_marks and the bucket's irb_marks. When the - * reference count on the bucket drops to zero, all the CONDEMNED ires - * are deleted. We don't have to bump up the reference count on the - * bucket if we are walking the bucket and never have to drop the bucket - * lock. Note that IRB_REFHOLD does not prevent addition of new ires - * in the list. It is okay because addition of new ires will not cause - * ire_next to point to freed memory. We do IRB_REFHOLD only when - * all of the 3 conditions are true : - * - * 1) The code needs to walk the IRE bucket from start to end. - * 2) It may have to drop the bucket lock sometimes while doing (1) - * 3) It does not want any ires to be deleted meanwhile. - */ - -/* - * Bump up the reference count on the IRE. We cannot assert that the - * bucket lock is being held as it is legal to bump up the reference - * count after the first lookup has returned the IRE without - * holding the lock. Currently ip_wput does this for caching IRE_CACHEs. - */ - -#ifdef DEBUG -#define IRE_UNTRACE_REF(ire) ire_untrace_ref(ire); -#define IRE_TRACE_REF(ire) ire_trace_ref(ire); -#else -#define IRE_UNTRACE_REF(ire) -#define IRE_TRACE_REF(ire) -#endif - -#define IRE_REFHOLD_NOTR(ire) { \ - atomic_add_32(&(ire)->ire_refcnt, 1); \ - ASSERT((ire)->ire_refcnt != 0); \ -} - -#define IRE_REFHOLD(ire) { \ - IRE_REFHOLD_NOTR(ire); \ - IRE_TRACE_REF(ire); \ -} - -#define IRE_REFHOLD_LOCKED(ire) { \ - IRE_TRACE_REF(ire); \ - (ire)->ire_refcnt++; \ -} - -/* - * Decrement the reference count on the IRE. - * In architectures e.g sun4u, where atomic_add_32_nv is just - * a cas, we need to maintain the right memory barrier semantics - * as that of mutex_exit i.e all the loads and stores should complete - * before the cas is executed. membar_exit() does that here. - * - * NOTE : This macro is used only in places where we want performance. - * To avoid bloating the code, we use the function "ire_refrele" - * which essentially calls the macro. - */ -#define IRE_REFRELE_NOTR(ire) { \ - ASSERT((ire)->ire_refcnt != 0); \ - membar_exit(); \ - if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0) \ - ire_inactive(ire); \ -} - -#define IRE_REFRELE(ire) { \ - if (ire->ire_bucket != NULL) { \ - IRE_UNTRACE_REF(ire); \ - } \ - IRE_REFRELE_NOTR(ire); \ -} - -/* - * Bump up the reference count on the hash bucket - IRB to - * prevent ires from being deleted in this bucket. - */ -#define IRB_REFHOLD(irb) { \ - rw_enter(&(irb)->irb_lock, RW_WRITER); \ - (irb)->irb_refcnt++; \ - ASSERT((irb)->irb_refcnt != 0); \ - rw_exit(&(irb)->irb_lock); \ -} -#define IRB_REFHOLD_LOCKED(irb) { \ - ASSERT(RW_WRITE_HELD(&(irb)->irb_lock)); \ - (irb)->irb_refcnt++; \ - ASSERT((irb)->irb_refcnt != 0); \ -} +}; void irb_refrele_ftable(irb_t *); -/* - * Note: when IRB_MARK_FTABLE (i.e., IRE_CACHETABLE entry), the irb_t - * is statically allocated, so that when the irb_refcnt goes to 0, - * we simply clean up the ire list and continue. - */ -#define IRB_REFRELE(irb) { \ - if ((irb)->irb_marks & IRB_MARK_FTABLE) { \ - irb_refrele_ftable((irb)); \ - } else { \ - rw_enter(&(irb)->irb_lock, RW_WRITER); \ - ASSERT((irb)->irb_refcnt != 0); \ - if (--(irb)->irb_refcnt == 0 && \ - ((irb)->irb_marks & IRE_MARK_CONDEMNED)) { \ - ire_t *ire_list; \ - \ - ire_list = ire_unlink(irb); \ - rw_exit(&(irb)->irb_lock); \ - ASSERT(ire_list != NULL); \ - ire_cleanup(ire_list); \ - } else { \ - rw_exit(&(irb)->irb_lock); \ - } \ - } \ -} extern struct kmem_cache *rt_entry_cache; -/* - * Lock the fast path mp for access, since the fp_mp can be deleted - * due a DL_NOTE_FASTPATH_FLUSH in the case of IRE_BROADCAST - */ - -#define LOCK_IRE_FP_MP(ire) { \ - if ((ire)->ire_type == IRE_BROADCAST) \ - mutex_enter(&ire->ire_nce->nce_lock); \ - } -#define UNLOCK_IRE_FP_MP(ire) { \ - if ((ire)->ire_type == IRE_BROADCAST) \ - mutex_exit(&ire->ire_nce->nce_lock); \ - } - typedef struct ire4 { - ipaddr_t ire4_src_addr; /* Source address to use. */ ipaddr_t ire4_mask; /* Mask for matching this IRE. */ ipaddr_t ire4_addr; /* Address this IRE represents. */ - ipaddr_t ire4_gateway_addr; /* Gateway if IRE_CACHE/IRE_OFFSUBNET */ - ipaddr_t ire4_cmask; /* Mask from parent prefix route */ + ipaddr_t ire4_gateway_addr; /* Gateway including for IRE_ONLINK */ + ipaddr_t ire4_setsrc_addr; /* RTF_SETSRC */ } ire4_t; typedef struct ire6 { - in6_addr_t ire6_src_addr; /* Source address to use. */ in6_addr_t ire6_mask; /* Mask for matching this IRE. */ in6_addr_t ire6_addr; /* Address this IRE represents. */ - in6_addr_t ire6_gateway_addr; /* Gateway if IRE_CACHE/IRE_OFFSUBNET */ - in6_addr_t ire6_cmask; /* Mask from parent prefix route */ + in6_addr_t ire6_gateway_addr; /* Gateway including for IRE_ONLINK */ + in6_addr_t ire6_setsrc_addr; /* RTF_SETSRC */ } ire6_t; typedef union ire_addr { @@ -2637,115 +2533,131 @@ typedef union ire_addr { ire4_t ire4_u; } ire_addr_u_t; -/* Internet Routing Entry */ -typedef struct ire_s { +/* + * Internet Routing Entry + * When we have multiple identical IREs we logically add them by manipulating + * ire_identical_ref and ire_delete first decrements + * that and when it reaches 1 we know it is the last IRE. + * "identical" is defined as being the same for: + * ire_addr, ire_netmask, ire_gateway, ire_ill, ire_zoneid, and ire_type + * For instance, multiple IRE_BROADCASTs for the same subnet number are + * viewed as identical, and so are the IRE_INTERFACEs when there are + * multiple logical interfaces (on the same ill) with the same subnet prefix. + */ +struct ire_s { struct ire_s *ire_next; /* The hash chain must be first. */ struct ire_s **ire_ptpn; /* Pointer to previous next. */ uint32_t ire_refcnt; /* Number of references */ - mblk_t *ire_mp; /* Non-null if allocated as mblk */ - queue_t *ire_rfq; /* recv from this queue */ - queue_t *ire_stq; /* send to this queue */ - union { - uint_t *max_fragp; /* Used only during ire creation */ - uint_t max_frag; /* MTU (next hop or path). */ - } imf_u; -#define ire_max_frag imf_u.max_frag -#define ire_max_fragp imf_u.max_fragp - uint32_t ire_frag_flag; /* IPH_DF or zero. */ - uint32_t ire_ident; /* Per IRE IP ident. */ - uint32_t ire_tire_mark; /* Used for reclaim of unused. */ + ill_t *ire_ill; + uint32_t ire_identical_ref; /* IRE_INTERFACE, IRE_BROADCAST */ uchar_t ire_ipversion; /* IPv4/IPv6 version */ - uchar_t ire_marks; /* IRE_MARK_CONDEMNED etc. */ ushort_t ire_type; /* Type of IRE */ + uint_t ire_generation; /* Generation including CONDEMNED */ uint_t ire_ib_pkt_count; /* Inbound packets for ire_addr */ uint_t ire_ob_pkt_count; /* Outbound packets to ire_addr */ - uint_t ire_ll_hdr_length; /* Non-zero if we do M_DATA prepends */ time_t ire_create_time; /* Time (in secs) IRE was created. */ - uint32_t ire_phandle; /* Associate prefix IREs to cache */ - uint32_t ire_ihandle; /* Associate interface IREs to cache */ - ipif_t *ire_ipif; /* the interface that this ire uses */ uint32_t ire_flags; /* flags related to route (RTF_*) */ /* - * Neighbor Cache Entry for IPv6; arp info for IPv4 + * ire_testhidden is TRUE for INTERFACE IREs of IS_UNDER_IPMP(ill) + * interfaces */ - struct nce_s *ire_nce; + boolean_t ire_testhidden; + pfirerecv_t ire_recvfn; /* Receive side handling */ + pfiresend_t ire_sendfn; /* Send side handling */ + pfirepostfrag_t ire_postfragfn; /* Bottom end of send handling */ + uint_t ire_masklen; /* # bits in ire_mask{,_v6} */ ire_addr_u_t ire_u; /* IPv4/IPv6 address info. */ irb_t *ire_bucket; /* Hash bucket when ire_ptphn is set */ - iulp_t ire_uinfo; /* Upper layer protocol info. */ - /* - * Protects ire_uinfo, ire_max_frag, and ire_frag_flag. - */ kmutex_t ire_lock; - uint_t ire_ipif_seqid; /* ipif_seqid of ire_ipif */ - uint_t ire_ipif_ifindex; /* ifindex associated with ipif */ - clock_t ire_last_used_time; /* Last used time */ + clock_t ire_last_used_time; /* For IRE_LOCAL reception */ tsol_ire_gw_secattr_t *ire_gw_secattr; /* gateway security attributes */ - zoneid_t ire_zoneid; /* for local address discrimination */ + zoneid_t ire_zoneid; + + /* + * Cached information of where to send packets that match this route. + * The ire_dep_* information is used to determine when ire_nce_cache + * needs to be updated. + * ire_nce_cache is the fastpath for the Neighbor Cache Entry + * for IPv6; arp info for IPv4 + * Since this is a cache setup and torn down independently of + * applications we need to use nce_ref{rele,hold}_notr for it. + */ + nce_t *ire_nce_cache; + + /* + * Quick check whether the ire_type and ire_masklen indicates + * that the IRE can have ire_nce_cache set i.e., whether it is + * IRE_ONLINK and for a single destination. + */ + boolean_t ire_nce_capable; + /* - * ire's that are embedded inside mblk_t and sent to the external - * resolver use the ire_stq_ifindex to track the ifindex of the - * ire_stq, so that the ill (if it exists) can be correctly recovered - * for cleanup in the esbfree routine when arp failure occurs. - * Similarly, the ire_stackid is used to recover the ip_stack_t. + * Dependency tracking so we can safely cache IRE and NCE pointers + * in offlink and onlink IREs. + * These are locked under the ips_ire_dep_lock rwlock. Write held + * when modifying the linkage. + * ire_dep_parent (Also chain towards IRE for nexthop) + * ire_dep_parent_generation: ire_generation of ire_dep_parent + * ire_dep_children (From parent to first child) + * ire_dep_sib_next (linked list of siblings) + * ire_dep_sib_ptpn (linked list of siblings) + * + * The parent has a ire_refhold on each child, and each child has + * an ire_refhold on its parent. + * Since ire_dep_parent is a cache setup and torn down independently of + * applications we need to use ire_ref{rele,hold}_notr for it. */ - uint_t ire_stq_ifindex; - netstackid_t ire_stackid; + ire_t *ire_dep_parent; + ire_t *ire_dep_children; + ire_t *ire_dep_sib_next; + ire_t **ire_dep_sib_ptpn; /* Pointer to previous next */ + uint_t ire_dep_parent_generation; + + uint_t ire_badcnt; /* Number of times ND_UNREACHABLE */ + uint64_t ire_last_badcnt; /* In seconds */ + + /* ire_defense* and ire_last_used_time are only used on IRE_LOCALs */ uint_t ire_defense_count; /* number of ARP conflicts */ uint_t ire_defense_time; /* last time defended (secs) */ + boolean_t ire_trace_disable; /* True when alloc fails */ ip_stack_t *ire_ipst; /* Does not have a netstack_hold */ -} ire_t; + iulp_t ire_metrics; +}; /* IPv4 compatibility macros */ -#define ire_src_addr ire_u.ire4_u.ire4_src_addr #define ire_mask ire_u.ire4_u.ire4_mask #define ire_addr ire_u.ire4_u.ire4_addr #define ire_gateway_addr ire_u.ire4_u.ire4_gateway_addr -#define ire_cmask ire_u.ire4_u.ire4_cmask +#define ire_setsrc_addr ire_u.ire4_u.ire4_setsrc_addr -#define ire_src_addr_v6 ire_u.ire6_u.ire6_src_addr #define ire_mask_v6 ire_u.ire6_u.ire6_mask #define ire_addr_v6 ire_u.ire6_u.ire6_addr #define ire_gateway_addr_v6 ire_u.ire6_u.ire6_gateway_addr -#define ire_cmask_v6 ire_u.ire6_u.ire6_cmask - -/* Convenient typedefs for sockaddrs */ -typedef struct sockaddr_in sin_t; -typedef struct sockaddr_in6 sin6_t; - -/* Address structure used for internal bind with IP */ -typedef struct ipa_conn_s { - ipaddr_t ac_laddr; - ipaddr_t ac_faddr; - uint16_t ac_fport; - uint16_t ac_lport; -} ipa_conn_t; - -typedef struct ipa6_conn_s { - in6_addr_t ac6_laddr; - in6_addr_t ac6_faddr; - uint16_t ac6_fport; - uint16_t ac6_lport; -} ipa6_conn_t; +#define ire_setsrc_addr_v6 ire_u.ire6_u.ire6_setsrc_addr /* - * Using ipa_conn_x_t or ipa6_conn_x_t allows us to modify the behavior of IP's - * bind handler. + * Values for ire_generation. + * + * If an IRE is marked with IRE_IS_CONDEMNED, the last walker of + * the bucket should delete this IRE from this bucket. + * + * IRE_GENERATION_VERIFY is never stored in ire_generation but it is + * stored in places that cache IREs (such as ixa_ire_generation and + * ire_dep_parent_generation). It is used as a signal that the cache is + * stale and needs to be reverified. */ -typedef struct ipa_conn_extended_s { - uint64_t acx_flags; - ipa_conn_t acx_conn; -} ipa_conn_x_t; +#define IRE_GENERATION_CONDEMNED 0 +#define IRE_GENERATION_VERIFY 1 +#define IRE_GENERATION_INITIAL 2 +#define IRE_IS_CONDEMNED(ire) \ + ((ire)->ire_generation == IRE_GENERATION_CONDEMNED) -typedef struct ipa6_conn_extended_s { - uint64_t ac6x_flags; - ipa6_conn_t ac6x_conn; -} ipa6_conn_x_t; - -/* flag values for ipa_conn_x_t and ipa6_conn_x_t. */ -#define ACX_VERIFY_DST 0x1ULL /* verify destination address is reachable */ +/* Convenient typedefs for sockaddrs */ +typedef struct sockaddr_in sin_t; +typedef struct sockaddr_in6 sin6_t; /* Name/Value Descriptor. */ typedef struct nv_s { @@ -2784,110 +2696,83 @@ extern uint_t ip_max_frag_dups; * to support the needs of such tools and private definitions moved to * private headers. */ -struct ip6_pkt_s { +struct ip_pkt_s { uint_t ipp_fields; /* Which fields are valid */ - uint_t ipp_sticky_ignored; /* sticky fields to ignore */ - uint_t ipp_ifindex; /* pktinfo ifindex */ in6_addr_t ipp_addr; /* pktinfo src/dst addr */ - uint_t ipp_unicast_hops; /* IPV6_UNICAST_HOPS */ - uint_t ipp_multicast_hops; /* IPV6_MULTICAST_HOPS */ +#define ipp_addr_v4 V4_PART_OF_V6(ipp_addr) + uint_t ipp_unicast_hops; /* IPV6_UNICAST_HOPS, IP_TTL */ uint_t ipp_hoplimit; /* IPV6_HOPLIMIT */ uint_t ipp_hopoptslen; - uint_t ipp_rtdstoptslen; + uint_t ipp_rthdrdstoptslen; uint_t ipp_rthdrlen; uint_t ipp_dstoptslen; - uint_t ipp_pathmtulen; uint_t ipp_fraghdrlen; ip6_hbh_t *ipp_hopopts; - ip6_dest_t *ipp_rtdstopts; + ip6_dest_t *ipp_rthdrdstopts; ip6_rthdr_t *ipp_rthdr; ip6_dest_t *ipp_dstopts; ip6_frag_t *ipp_fraghdr; - struct ip6_mtuinfo *ipp_pathmtu; - in6_addr_t ipp_nexthop; /* Transmit only */ - uint8_t ipp_tclass; - int8_t ipp_use_min_mtu; + uint8_t ipp_tclass; /* IPV6_TCLASS */ + uint8_t ipp_type_of_service; /* IP_TOS */ + uint_t ipp_ipv4_options_len; /* Len of IPv4 options */ + uint8_t *ipp_ipv4_options; /* Ptr to IPv4 options */ + uint_t ipp_label_len_v4; /* Len of TX label for IPv4 */ + uint8_t *ipp_label_v4; /* TX label for IPv4 */ + uint_t ipp_label_len_v6; /* Len of TX label for IPv6 */ + uint8_t *ipp_label_v6; /* TX label for IPv6 */ }; -typedef struct ip6_pkt_s ip6_pkt_t; - -extern void ip6_pkt_free(ip6_pkt_t *); /* free storage inside ip6_pkt_t */ - -/* - * This struct is used by ULP_opt_set() functions to return value of IPv4 - * ancillary options. Currently this is only used by udp and icmp and only - * IP_PKTINFO option is supported. - */ -typedef struct ip4_pkt_s { - uint_t ip4_ill_index; /* interface index */ - ipaddr_t ip4_addr; /* source address */ -} ip4_pkt_t; - -/* - * Used by ULP's to pass options info to ip_output - * currently only IP_PKTINFO is supported. - */ -typedef struct ip_opt_info_s { - uint_t ip_opt_ill_index; - uint_t ip_opt_flags; -} ip_opt_info_t; - -/* - * value for ip_opt_flags - */ -#define IP_VERIFY_SRC 0x1 +typedef struct ip_pkt_s ip_pkt_t; -/* - * This structure is used to convey information from IP and the ULP. - * Currently used for the IP_RECVSLLA, IP_RECVIF and IP_RECVPKTINFO options. - * The type of information field is set to IN_PKTINFO (i.e inbound pkt info) - */ -typedef struct ip_pktinfo { - uint32_t ip_pkt_ulp_type; /* type of info sent */ - uint32_t ip_pkt_flags; /* what is sent up by IP */ - uint32_t ip_pkt_ifindex; /* inbound interface index */ - struct sockaddr_dl ip_pkt_slla; /* has source link layer addr */ - struct in_addr ip_pkt_match_addr; /* matched address */ -} ip_pktinfo_t; - -/* - * flags to tell UDP what IP is sending; in_pkt_flags - */ -#define IPF_RECVIF 0x01 /* inbound interface index */ -#define IPF_RECVSLLA 0x02 /* source link layer address */ -/* - * Inbound interface index + matched address. - * Used only by IPV4. - */ -#define IPF_RECVADDR 0x04 +extern void ip_pkt_free(ip_pkt_t *); /* free storage inside ip_pkt_t */ +extern ipaddr_t ip_pkt_source_route_v4(const ip_pkt_t *); +extern in6_addr_t *ip_pkt_source_route_v6(const ip_pkt_t *); +extern int ip_pkt_copy(ip_pkt_t *, ip_pkt_t *, int); +extern void ip_pkt_source_route_reverse_v4(ip_pkt_t *); /* ipp_fields values */ -#define IPPF_IFINDEX 0x0001 /* Part of in6_pktinfo: ifindex */ -#define IPPF_ADDR 0x0002 /* Part of in6_pktinfo: src/dst addr */ -#define IPPF_SCOPE_ID 0x0004 /* Add xmit ip6i_t for sin6_scope_id */ -#define IPPF_NO_CKSUM 0x0008 /* Add xmit ip6i_t for IP6I_NO_*_CKSUM */ - -#define IPPF_RAW_CKSUM 0x0010 /* Add xmit ip6i_t for IP6I_RAW_CHECKSUM */ -#define IPPF_HOPLIMIT 0x0020 -#define IPPF_HOPOPTS 0x0040 -#define IPPF_RTHDR 0x0080 - -#define IPPF_RTDSTOPTS 0x0100 -#define IPPF_DSTOPTS 0x0200 -#define IPPF_NEXTHOP 0x0400 -#define IPPF_PATHMTU 0x0800 - -#define IPPF_TCLASS 0x1000 -#define IPPF_DONTFRAG 0x2000 -#define IPPF_USE_MIN_MTU 0x04000 -#define IPPF_MULTICAST_HOPS 0x08000 - -#define IPPF_UNICAST_HOPS 0x10000 -#define IPPF_FRAGHDR 0x20000 - -#define IPPF_HAS_IP6I \ - (IPPF_IFINDEX|IPPF_ADDR|IPPF_NEXTHOP|IPPF_SCOPE_ID| \ - IPPF_NO_CKSUM|IPPF_RAW_CKSUM|IPPF_HOPLIMIT|IPPF_DONTFRAG| \ - IPPF_USE_MIN_MTU|IPPF_MULTICAST_HOPS|IPPF_UNICAST_HOPS) +#define IPPF_ADDR 0x0001 /* Part of in6_pktinfo: src/dst addr */ +#define IPPF_HOPLIMIT 0x0002 /* Overrides unicast and multicast */ +#define IPPF_TCLASS 0x0004 /* Overrides class in sin6_flowinfo */ + +#define IPPF_HOPOPTS 0x0010 /* ipp_hopopts set */ +#define IPPF_RTHDR 0x0020 /* ipp_rthdr set */ +#define IPPF_RTHDRDSTOPTS 0x0040 /* ipp_rthdrdstopts set */ +#define IPPF_DSTOPTS 0x0080 /* ipp_dstopts set */ + +#define IPPF_IPV4_OPTIONS 0x0100 /* ipp_ipv4_options set */ +#define IPPF_LABEL_V4 0x0200 /* ipp_label_v4 set */ +#define IPPF_LABEL_V6 0x0400 /* ipp_label_v6 set */ + +#define IPPF_FRAGHDR 0x0800 /* Used for IPsec receive side */ + +/* + * Data structure which is passed to conn_opt_get/set. + * The conn_t is included even though it can be inferred from queue_t. + * setsockopt and getsockopt use conn_ixa and conn_xmit_ipp. However, + * when handling ancillary data we use separate ixa and ipps. + */ +typedef struct conn_opt_arg_s { + conn_t *coa_connp; + ip_xmit_attr_t *coa_ixa; + ip_pkt_t *coa_ipp; + boolean_t coa_ancillary; /* Ancillary data and not setsockopt */ + uint_t coa_changed; /* See below */ +} conn_opt_arg_t; + +/* + * Flags for what changed. + * If we want to be more efficient in the future we can have more fine + * grained flags e.g., a flag for just IP_TOS changing. + * For now we either call ip_set_destination (for "route changed") + * and/or conn_build_hdr_template/conn_prepend_hdr (for "header changed"). + */ +#define COA_HEADER_CHANGED 0x0001 +#define COA_ROUTE_CHANGED 0x0002 +#define COA_RCVBUF_CHANGED 0x0004 /* SO_RCVBUF */ +#define COA_SNDBUF_CHANGED 0x0008 /* SO_SNDBUF */ +#define COA_WROFF_CHANGED 0x0010 /* Header size changed */ +#define COA_ICMP_BIND_NEEDED 0x0020 +#define COA_OOBINLINE_CHANGED 0x0040 #define TCP_PORTS_OFFSET 0 #define UDP_PORTS_OFFSET 0 @@ -2902,32 +2787,21 @@ typedef struct ip_pktinfo { #define IPIF_LOOKUP_FAILED 2 /* Used as error code */ #define ILL_CAN_LOOKUP(ill) \ - (!((ill)->ill_state_flags & (ILL_CONDEMNED | ILL_CHANGING)) || \ + (!((ill)->ill_state_flags & ILL_CONDEMNED) || \ IAM_WRITER_ILL(ill)) -#define ILL_CAN_WAIT(ill, q) \ - (((q) != NULL) && !((ill)->ill_state_flags & (ILL_CONDEMNED))) +#define ILL_IS_CONDEMNED(ill) \ + ((ill)->ill_state_flags & ILL_CONDEMNED) #define IPIF_CAN_LOOKUP(ipif) \ - (!((ipif)->ipif_state_flags & (IPIF_CONDEMNED | IPIF_CHANGING)) || \ + (!((ipif)->ipif_state_flags & IPIF_CONDEMNED) || \ IAM_WRITER_IPIF(ipif)) -/* - * If the parameter 'q' is NULL, the caller is not interested in wait and - * restart of the operation if the ILL or IPIF cannot be looked up when it is - * marked as 'CHANGING'. Typically a thread that tries to send out data will - * end up passing NULLs as the last 4 parameters to ill_lookup_on_ifindex and - * in this case 'q' is NULL - */ -#define IPIF_CAN_WAIT(ipif, q) \ - (((q) != NULL) && !((ipif)->ipif_state_flags & (IPIF_CONDEMNED))) - -#define IPIF_CAN_LOOKUP_WALKER(ipif) \ - (!((ipif)->ipif_state_flags & (IPIF_CONDEMNED)) || \ - IAM_WRITER_IPIF(ipif)) +#define IPIF_IS_CONDEMNED(ipif) \ + ((ipif)->ipif_state_flags & IPIF_CONDEMNED) -#define ILL_UNMARK_CHANGING(ill) \ - (ill)->ill_state_flags &= ~ILL_CHANGING; +#define IPIF_IS_CHANGING(ipif) \ + ((ipif)->ipif_state_flags & IPIF_CHANGING) /* Macros used to assert that this thread is a writer */ #define IAM_WRITER_IPSQ(ipsq) ((ipsq)->ipsq_xop->ipx_writer == curthread) @@ -2956,9 +2830,9 @@ typedef struct ip_pktinfo { #define RELEASE_ILL_LOCKS(ill_1, ill_2) \ { \ if (ill_1 != NULL) \ - mutex_exit(&(ill_1)->ill_lock); \ + mutex_exit(&(ill_1)->ill_lock); \ if (ill_2 != NULL && ill_2 != ill_1) \ - mutex_exit(&(ill_2)->ill_lock); \ + mutex_exit(&(ill_2)->ill_lock); \ } /* Get the other protocol instance ill */ @@ -2975,20 +2849,13 @@ typedef struct cmd_info_s struct lifreq *ci_lifr; /* the lifreq struct passed down */ } cmd_info_t; -/* - * List of AH and ESP IPsec acceleration capable ills - */ -typedef struct ipsec_capab_ill_s { - uint_t ill_index; - boolean_t ill_isv6; - struct ipsec_capab_ill_s *next; -} ipsec_capab_ill_t; - extern struct kmem_cache *ire_cache; extern ipaddr_t ip_g_all_ones; -extern uint_t ip_loopback_mtu; /* /etc/system */ +extern uint_t ip_loopback_mtu; /* /etc/system */ +extern uint_t ip_loopback_mtuplus; +extern uint_t ip_loopback_mtu_v6plus; extern vmem_t *ip_minor_arena_sa; extern vmem_t *ip_minor_arena_la; @@ -3014,18 +2881,18 @@ extern vmem_t *ip_minor_arena_la; #define ips_ip_g_send_redirects ips_param_arr[5].ip_param_value #define ips_ip_g_forward_directed_bcast ips_param_arr[6].ip_param_value #define ips_ip_mrtdebug ips_param_arr[7].ip_param_value -#define ips_ip_timer_interval ips_param_arr[8].ip_param_value -#define ips_ip_ire_arp_interval ips_param_arr[9].ip_param_value -#define ips_ip_ire_redir_interval ips_param_arr[10].ip_param_value +#define ips_ip_ire_reclaim_fraction ips_param_arr[8].ip_param_value +#define ips_ip_nce_reclaim_fraction ips_param_arr[9].ip_param_value +#define ips_ip_dce_reclaim_fraction ips_param_arr[10].ip_param_value #define ips_ip_def_ttl ips_param_arr[11].ip_param_value #define ips_ip_forward_src_routed ips_param_arr[12].ip_param_value #define ips_ip_wroff_extra ips_param_arr[13].ip_param_value -#define ips_ip_ire_pathmtu_interval ips_param_arr[14].ip_param_value +#define ips_ip_pathmtu_interval ips_param_arr[14].ip_param_value #define ips_ip_icmp_return ips_param_arr[15].ip_param_value #define ips_ip_path_mtu_discovery ips_param_arr[16].ip_param_value -#define ips_ip_ignore_delete_time ips_param_arr[17].ip_param_value +#define ips_ip_pmtu_min ips_param_arr[17].ip_param_value #define ips_ip_ignore_redirect ips_param_arr[18].ip_param_value -#define ips_ip_output_queue ips_param_arr[19].ip_param_value +#define ips_ip_arp_icmp_error ips_param_arr[19].ip_param_value #define ips_ip_broadcast_ttl ips_param_arr[20].ip_param_value #define ips_ip_icmp_err_interval ips_param_arr[21].ip_param_value #define ips_ip_icmp_err_burst ips_param_arr[22].ip_param_value @@ -3046,7 +2913,7 @@ extern vmem_t *ip_minor_arena_la; #define ips_ipv6_send_redirects ips_param_arr[35].ip_param_value #define ips_ipv6_ignore_redirect ips_param_arr[36].ip_param_value #define ips_ipv6_strict_dst_multihoming ips_param_arr[37].ip_param_value -#define ips_ip_ire_reclaim_fraction ips_param_arr[38].ip_param_value +#define ips_src_check ips_param_arr[38].ip_param_value #define ips_ipsec_policy_log_interval ips_param_arr[39].ip_param_value #define ips_pim_accept_clear_messages ips_param_arr[40].ip_param_value #define ips_ip_ndp_unsolicit_interval ips_param_arr[41].ip_param_value @@ -3055,21 +2922,37 @@ extern vmem_t *ip_minor_arena_la; /* Misc IP configuration knobs */ #define ips_ip_policy_mask ips_param_arr[44].ip_param_value -#define ips_ip_multirt_resolution_interval ips_param_arr[45].ip_param_value +#define ips_ip_ecmp_behavior ips_param_arr[45].ip_param_value #define ips_ip_multirt_ttl ips_param_arr[46].ip_param_value -#define ips_ip_multidata_outbound ips_param_arr[47].ip_param_value -#define ips_ip_ndp_defense_interval ips_param_arr[48].ip_param_value -#define ips_ip_max_temp_idle ips_param_arr[49].ip_param_value -#define ips_ip_max_temp_defend ips_param_arr[50].ip_param_value -#define ips_ip_max_defend ips_param_arr[51].ip_param_value -#define ips_ip_defend_interval ips_param_arr[52].ip_param_value -#define ips_ip_dup_recovery ips_param_arr[53].ip_param_value -#define ips_ip_restrict_interzone_loopback ips_param_arr[54].ip_param_value -#define ips_ip_lso_outbound ips_param_arr[55].ip_param_value -#define ips_igmp_max_version ips_param_arr[56].ip_param_value -#define ips_mld_max_version ips_param_arr[57].ip_param_value -#define ips_ip_pmtu_min ips_param_arr[58].ip_param_value -#define ips_ipv6_drop_inbound_icmpv6 ips_param_arr[59].ip_param_value +#define ips_ip_ire_badcnt_lifetime ips_param_arr[47].ip_param_value +#define ips_ip_max_temp_idle ips_param_arr[48].ip_param_value +#define ips_ip_max_temp_defend ips_param_arr[49].ip_param_value +#define ips_ip_max_defend ips_param_arr[50].ip_param_value +#define ips_ip_defend_interval ips_param_arr[51].ip_param_value +#define ips_ip_dup_recovery ips_param_arr[52].ip_param_value +#define ips_ip_restrict_interzone_loopback ips_param_arr[53].ip_param_value +#define ips_ip_lso_outbound ips_param_arr[54].ip_param_value +#define ips_igmp_max_version ips_param_arr[55].ip_param_value +#define ips_mld_max_version ips_param_arr[56].ip_param_value +#define ips_ipv6_drop_inbound_icmpv6 ips_param_arr[57].ip_param_value +#define ips_arp_probe_delay ips_param_arr[58].ip_param_value +#define ips_arp_fastprobe_delay ips_param_arr[59].ip_param_value +#define ips_arp_probe_interval ips_param_arr[60].ip_param_value +#define ips_arp_fastprobe_interval ips_param_arr[61].ip_param_value +#define ips_arp_probe_count ips_param_arr[62].ip_param_value +#define ips_arp_fastprobe_count ips_param_arr[63].ip_param_value +#define ips_ipv4_dad_announce_interval ips_param_arr[64].ip_param_value +#define ips_ipv6_dad_announce_interval ips_param_arr[65].ip_param_value +#define ips_arp_defend_interval ips_param_arr[66].ip_param_value +#define ips_arp_defend_rate ips_param_arr[67].ip_param_value +#define ips_ndp_defend_interval ips_param_arr[68].ip_param_value +#define ips_ndp_defend_rate ips_param_arr[69].ip_param_value +#define ips_arp_defend_period ips_param_arr[70].ip_param_value +#define ips_ndp_defend_period ips_param_arr[71].ip_param_value +#define ips_ipv4_icmp_return_pmtu ips_param_arr[72].ip_param_value +#define ips_ipv6_icmp_return_pmtu ips_param_arr[73].ip_param_value +#define ips_ip_arp_publish_count ips_param_arr[74].ip_param_value +#define ips_ip_arp_publish_interval ips_param_arr[75].ip_param_value extern int dohwcksum; /* use h/w cksum if supported by the h/w */ #ifdef ZC_TEST @@ -3102,13 +2985,13 @@ extern struct module_info ip_mod_info; ((ipst)->ips_ip4_loopback_out_event.he_interested) #define HOOKS6_INTERESTED_LOOPBACK_OUT(ipst) \ ((ipst)->ips_ip6_loopback_out_event.he_interested) - /* - * Hooks macros used inside of ip + * Hooks marcos used inside of ip + * The callers use the above INTERESTED macros first, hence + * the he_interested check is superflous. */ -#define FW_HOOKS(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst) \ - \ - if ((_hook).he_interested) { \ +#define FW_HOOKS(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst, _err) \ + if ((_hook).he_interested) { \ hook_pkt_event_t info; \ \ _NOTE(CONSTCOND) \ @@ -3121,12 +3004,15 @@ extern struct module_info ip_mod_info; info.hpe_mp = &(_fm); \ info.hpe_mb = _m; \ info.hpe_flags = _llm; \ - if (hook_run(ipst->ips_ipv4_net_data->netd_hooks, \ - _event, (hook_data_t)&info) != 0) { \ + _err = hook_run(ipst->ips_ipv4_net_data->netd_hooks, \ + _event, (hook_data_t)&info); \ + if (_err != 0) { \ ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\ (_hook).he_name, (void *)_fm, (void *)_m)); \ - freemsg(_fm); \ - _fm = NULL; \ + if (_fm != NULL) { \ + freemsg(_fm); \ + _fm = NULL; \ + } \ _iph = NULL; \ _m = NULL; \ } else { \ @@ -3135,9 +3021,8 @@ extern struct module_info ip_mod_info; } \ } -#define FW_HOOKS6(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst) \ - \ - if ((_hook).he_interested) { \ +#define FW_HOOKS6(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst, _err) \ + if ((_hook).he_interested) { \ hook_pkt_event_t info; \ \ _NOTE(CONSTCOND) \ @@ -3150,12 +3035,15 @@ extern struct module_info ip_mod_info; info.hpe_mp = &(_fm); \ info.hpe_mb = _m; \ info.hpe_flags = _llm; \ - if (hook_run(ipst->ips_ipv6_net_data->netd_hooks, \ - _event, (hook_data_t)&info) != 0) { \ + _err = hook_run(ipst->ips_ipv6_net_data->netd_hooks, \ + _event, (hook_data_t)&info); \ + if (_err != 0) { \ ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\ (_hook).he_name, (void *)_fm, (void *)_m)); \ - freemsg(_fm); \ - _fm = NULL; \ + if (_fm != NULL) { \ + freemsg(_fm); \ + _fm = NULL; \ + } \ _iph = NULL; \ _m = NULL; \ } else { \ @@ -3194,24 +3082,6 @@ extern struct module_info ip_mod_info; #define IP_LOOPBACK_ADDR(addr) \ (((addr) & N_IN_CLASSA_NET == N_IN_LOOPBACK_NET)) -#ifdef DEBUG -/* IPsec HW acceleration debugging support */ - -#define IPSECHW_CAPAB 0x0001 /* capability negotiation */ -#define IPSECHW_SADB 0x0002 /* SADB exchange */ -#define IPSECHW_PKT 0x0004 /* general packet flow */ -#define IPSECHW_PKTIN 0x0008 /* driver in pkt processing details */ -#define IPSECHW_PKTOUT 0x0010 /* driver out pkt processing details */ - -#define IPSECHW_DEBUG(f, x) if (ipsechw_debug & (f)) { (void) printf x; } -#define IPSECHW_CALL(f, r, x) if (ipsechw_debug & (f)) { (void) r x; } - -extern uint32_t ipsechw_debug; -#else -#define IPSECHW_DEBUG(f, x) {} -#define IPSECHW_CALL(f, r, x) {} -#endif - extern int ip_debug; extern uint_t ip_thread_data; extern krwlock_t ip_thread_rwlock; @@ -3235,8 +3105,6 @@ extern list_t ip_thread_list; /* Default MAC-layer address string length for mac_colon_addr */ #define MAC_STR_LEN 128 -struct ipsec_out_s; - struct mac_header_info_s; extern void ill_frag_timer(void *); @@ -3252,86 +3120,173 @@ extern char *ip_dot_addr(ipaddr_t, char *); extern const char *mac_colon_addr(const uint8_t *, size_t, char *, size_t); extern void ip_lwput(queue_t *, mblk_t *); extern boolean_t icmp_err_rate_limit(ip_stack_t *); -extern void icmp_time_exceeded(queue_t *, mblk_t *, uint8_t, zoneid_t, - ip_stack_t *); -extern void icmp_unreachable(queue_t *, mblk_t *, uint8_t, zoneid_t, - ip_stack_t *); -extern mblk_t *ip_add_info(mblk_t *, ill_t *, uint_t, zoneid_t, ip_stack_t *); -cred_t *ip_best_cred(mblk_t *, conn_t *, pid_t *); -extern mblk_t *ip_bind_v4(queue_t *, mblk_t *, conn_t *); -extern boolean_t ip_bind_ipsec_policy_set(conn_t *, mblk_t *); -extern int ip_bind_laddr_v4(conn_t *, mblk_t **, uint8_t, ipaddr_t, - uint16_t, boolean_t); -extern int ip_proto_bind_laddr_v4(conn_t *, mblk_t **, uint8_t, ipaddr_t, - uint16_t, boolean_t); -extern int ip_proto_bind_connected_v4(conn_t *, mblk_t **, - uint8_t, ipaddr_t *, uint16_t, ipaddr_t, uint16_t, boolean_t, boolean_t, - cred_t *); -extern int ip_bind_connected_v4(conn_t *, mblk_t **, uint8_t, ipaddr_t *, - uint16_t, ipaddr_t, uint16_t, boolean_t, boolean_t, cred_t *); +extern void icmp_frag_needed(mblk_t *, int, ip_recv_attr_t *); +extern mblk_t *icmp_inbound_v4(mblk_t *, ip_recv_attr_t *); +extern void icmp_time_exceeded(mblk_t *, uint8_t, ip_recv_attr_t *); +extern void icmp_unreachable(mblk_t *, uint8_t, ip_recv_attr_t *); +extern boolean_t ip_ipsec_policy_inherit(conn_t *, conn_t *, ip_recv_attr_t *); +extern void *ip_pullup(mblk_t *, ssize_t, ip_recv_attr_t *); +extern void ip_setl2src(mblk_t *, ip_recv_attr_t *, ill_t *); +extern mblk_t *ip_check_and_align_header(mblk_t *, uint_t, ip_recv_attr_t *); +extern mblk_t *ip_check_length(mblk_t *, uchar_t *, ssize_t, uint_t, uint_t, + ip_recv_attr_t *); +extern mblk_t *ip_check_optlen(mblk_t *, ipha_t *, uint_t, uint_t, + ip_recv_attr_t *); +extern mblk_t *ip_fix_dbref(mblk_t *, ip_recv_attr_t *); extern uint_t ip_cksum(mblk_t *, int, uint32_t); extern int ip_close(queue_t *, int); extern uint16_t ip_csum_hdr(ipha_t *); -extern void ip_proto_not_sup(queue_t *, mblk_t *, uint_t, zoneid_t, - ip_stack_t *); +extern void ip_forward_xmit_v4(nce_t *, ill_t *, mblk_t *, ipha_t *, + ip_recv_attr_t *, uint32_t, uint32_t); +extern boolean_t ip_forward_options(mblk_t *, ipha_t *, ill_t *, + ip_recv_attr_t *); +extern int ip_fragment_v4(mblk_t *, nce_t *, iaflags_t, uint_t, uint32_t, + uint32_t, zoneid_t, zoneid_t, pfirepostfrag_t postfragfn, + uintptr_t *cookie); +extern void ip_proto_not_sup(mblk_t *, ip_recv_attr_t *); extern void ip_ire_g_fini(void); extern void ip_ire_g_init(void); extern void ip_ire_fini(ip_stack_t *); extern void ip_ire_init(ip_stack_t *); +extern void ip_mdata_to_mhi(ill_t *, mblk_t *, struct mac_header_info_s *); extern int ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp); extern int ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp); extern int ip_reassemble(mblk_t *, ipf_t *, uint_t, boolean_t, ill_t *, size_t); -extern int ip_opt_set_ill(conn_t *, int, boolean_t, boolean_t, - int, int, mblk_t *); extern void ip_rput(queue_t *, mblk_t *); extern void ip_input(ill_t *, ill_rx_ring_t *, mblk_t *, struct mac_header_info_s *); +extern void ip_input_v6(ill_t *, ill_rx_ring_t *, mblk_t *, + struct mac_header_info_s *); +extern mblk_t *ip_input_common_v4(ill_t *, ill_rx_ring_t *, mblk_t *, + struct mac_header_info_s *, squeue_t *, mblk_t **, uint_t *); +extern mblk_t *ip_input_common_v6(ill_t *, ill_rx_ring_t *, mblk_t *, + struct mac_header_info_s *, squeue_t *, mblk_t **, uint_t *); +extern void ill_input_full_v4(mblk_t *, void *, void *, + ip_recv_attr_t *, rtc_t *); +extern void ill_input_short_v4(mblk_t *, void *, void *, + ip_recv_attr_t *, rtc_t *); +extern void ill_input_full_v6(mblk_t *, void *, void *, + ip_recv_attr_t *, rtc_t *); +extern void ill_input_short_v6(mblk_t *, void *, void *, + ip_recv_attr_t *, rtc_t *); +extern ipaddr_t ip_input_options(ipha_t *, ipaddr_t, mblk_t *, + ip_recv_attr_t *, int *); +extern boolean_t ip_input_local_options(mblk_t *, ipha_t *, ip_recv_attr_t *); +extern mblk_t *ip_input_fragment(mblk_t *, ipha_t *, ip_recv_attr_t *); +extern mblk_t *ip_input_fragment_v6(mblk_t *, ip6_t *, ip6_frag_t *, uint_t, + ip_recv_attr_t *); +extern void ip_input_post_ipsec(mblk_t *, ip_recv_attr_t *); +extern void ip_fanout_v4(mblk_t *, ipha_t *, ip_recv_attr_t *); +extern void ip_fanout_v6(mblk_t *, ip6_t *, ip_recv_attr_t *); +extern void ip_fanout_proto_conn(conn_t *, mblk_t *, ipha_t *, ip6_t *, + ip_recv_attr_t *); +extern void ip_fanout_proto_v4(mblk_t *, ipha_t *, ip_recv_attr_t *); +extern void ip_fanout_send_icmp_v4(mblk_t *, uint_t, uint_t, + ip_recv_attr_t *); +extern void ip_fanout_udp_conn(conn_t *, mblk_t *, ipha_t *, ip6_t *, + ip_recv_attr_t *); +extern void ip_fanout_udp_multi_v4(mblk_t *, ipha_t *, uint16_t, uint16_t, + ip_recv_attr_t *); +extern mblk_t *zero_spi_check(mblk_t *, ip_recv_attr_t *); +extern void ip_build_hdrs_v4(uchar_t *, uint_t, const ip_pkt_t *, uint8_t); +extern int ip_find_hdr_v4(ipha_t *, ip_pkt_t *, boolean_t); +extern int ip_total_hdrs_len_v4(const ip_pkt_t *); + extern mblk_t *ip_accept_tcp(ill_t *, ill_rx_ring_t *, squeue_t *, mblk_t *, mblk_t **, uint_t *cnt); -extern void ip_rput_dlpi(queue_t *, mblk_t *); -extern void ip_rput_forward(ire_t *, ipha_t *, mblk_t *, ill_t *); -extern void ip_rput_forward_multicast(ipaddr_t, mblk_t *, ipif_t *); +extern void ip_rput_dlpi(ill_t *, mblk_t *); +extern void ip_rput_notdata(ill_t *, mblk_t *); extern void ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *, mib2_ipIfStatsEntry_t *); extern void ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *, mib2_ipv6IfIcmpEntry_t *); -extern void ip_udp_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *); -extern void ip_proto_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *, - uint32_t); extern void ip_rput_other(ipsq_t *, queue_t *, mblk_t *, void *); extern ire_t *ip_check_multihome(void *, ire_t *, ill_t *); -extern void ip_setpktversion(conn_t *, boolean_t, boolean_t, ip_stack_t *); -extern void ip_trash_ire_reclaim(void *); -extern void ip_trash_timer_expire(void *); -extern void ip_wput(queue_t *, mblk_t *); -extern void ip_output(void *, mblk_t *, void *, int); -extern void ip_output_options(void *, mblk_t *, void *, int, - ip_opt_info_t *); - -extern void ip_wput_ire(queue_t *, mblk_t *, ire_t *, conn_t *, int, - zoneid_t); -extern void ip_wput_local(queue_t *, ill_t *, ipha_t *, mblk_t *, ire_t *, - int, zoneid_t); -extern void ip_wput_multicast(queue_t *, mblk_t *, ipif_t *, zoneid_t); -extern void ip_wput_nondata(ipsq_t *, queue_t *, mblk_t *, void *); +extern void ip_send_potential_redirect_v4(mblk_t *, ipha_t *, ire_t *, + ip_recv_attr_t *); +extern int ip_set_destination_v4(ipaddr_t *, ipaddr_t, ipaddr_t, + ip_xmit_attr_t *, iulp_t *, uint32_t, uint_t); +extern int ip_set_destination_v6(in6_addr_t *, const in6_addr_t *, + const in6_addr_t *, ip_xmit_attr_t *, iulp_t *, uint32_t, uint_t); + +extern int ip_output_simple(mblk_t *, ip_xmit_attr_t *); +extern int ip_output_simple_v4(mblk_t *, ip_xmit_attr_t *); +extern int ip_output_simple_v6(mblk_t *, ip_xmit_attr_t *); +extern int ip_output_options(mblk_t *, ipha_t *, ip_xmit_attr_t *, + ill_t *); +extern void ip_output_local_options(ipha_t *, ip_stack_t *); + +extern ip_xmit_attr_t *conn_get_ixa(conn_t *, boolean_t); +extern ip_xmit_attr_t *conn_get_ixa_tryhard(conn_t *, boolean_t); +extern ip_xmit_attr_t *conn_replace_ixa(conn_t *, ip_xmit_attr_t *); +extern ip_xmit_attr_t *conn_get_ixa_exclusive(conn_t *); +extern ip_xmit_attr_t *ip_xmit_attr_duplicate(ip_xmit_attr_t *); +extern void ip_xmit_attr_replace_tsl(ip_xmit_attr_t *, ts_label_t *); +extern void ip_xmit_attr_restore_tsl(ip_xmit_attr_t *, cred_t *); +boolean_t ip_recv_attr_replace_label(ip_recv_attr_t *, ts_label_t *); +extern void ixa_inactive(ip_xmit_attr_t *); +extern void ixa_refrele(ip_xmit_attr_t *); +extern boolean_t ixa_check_drain_insert(conn_t *, ip_xmit_attr_t *); +extern void ixa_cleanup(ip_xmit_attr_t *); +extern void ira_cleanup(ip_recv_attr_t *, boolean_t); +extern void ixa_safe_copy(ip_xmit_attr_t *, ip_xmit_attr_t *); + +extern int conn_ip_output(mblk_t *, ip_xmit_attr_t *); +extern boolean_t ip_output_verify_local(ip_xmit_attr_t *); +extern mblk_t *ip_output_process_local(mblk_t *, ip_xmit_attr_t *, boolean_t, + boolean_t, conn_t *); + +extern int conn_opt_get(conn_opt_arg_t *, t_scalar_t, t_scalar_t, + uchar_t *); +extern int conn_opt_set(conn_opt_arg_t *, t_scalar_t, t_scalar_t, uint_t, + uchar_t *, boolean_t, cred_t *); +extern boolean_t conn_same_as_last_v4(conn_t *, sin_t *); +extern boolean_t conn_same_as_last_v6(conn_t *, sin6_t *); +extern int conn_update_label(const conn_t *, const ip_xmit_attr_t *, + const in6_addr_t *, ip_pkt_t *); + +extern int ip_opt_set_multicast_group(conn_t *, t_scalar_t, + uchar_t *, boolean_t, boolean_t); +extern int ip_opt_set_multicast_sources(conn_t *, t_scalar_t, + uchar_t *, boolean_t, boolean_t); +extern int conn_getsockname(conn_t *, struct sockaddr *, uint_t *); +extern int conn_getpeername(conn_t *, struct sockaddr *, uint_t *); + +extern int conn_build_hdr_template(conn_t *, uint_t, uint_t, + const in6_addr_t *, const in6_addr_t *, uint32_t); +extern mblk_t *conn_prepend_hdr(ip_xmit_attr_t *, const ip_pkt_t *, + const in6_addr_t *, const in6_addr_t *, uint8_t, uint32_t, uint_t, + mblk_t *, uint_t, uint_t, uint32_t *, int *); +extern void ip_attr_newdst(ip_xmit_attr_t *); +extern void ip_attr_nexthop(const ip_pkt_t *, const ip_xmit_attr_t *, + const in6_addr_t *, in6_addr_t *); +extern int conn_connect(conn_t *, iulp_t *, uint32_t); +extern int ip_attr_connect(const conn_t *, ip_xmit_attr_t *, + const in6_addr_t *, const in6_addr_t *, const in6_addr_t *, in_port_t, + in6_addr_t *, iulp_t *, uint32_t); +extern int conn_inherit_parent(conn_t *, conn_t *); + +extern void conn_ixa_cleanup(conn_t *connp, void *arg); + +extern boolean_t conn_wantpacket(conn_t *, ip_recv_attr_t *, ipha_t *); +extern uint_t ip_type_v4(ipaddr_t, ip_stack_t *); +extern uint_t ip_type_v6(const in6_addr_t *, ip_stack_t *); + +extern void ip_wput_nondata(queue_t *, mblk_t *); extern void ip_wsrv(queue_t *); extern char *ip_nv_lookup(nv_t *, int); extern boolean_t ip_local_addr_ok_v6(const in6_addr_t *, const in6_addr_t *); extern boolean_t ip_remote_addr_ok_v6(const in6_addr_t *, const in6_addr_t *); extern ipaddr_t ip_massage_options(ipha_t *, netstack_t *); extern ipaddr_t ip_net_mask(ipaddr_t); -extern void ip_newroute(queue_t *, mblk_t *, ipaddr_t, conn_t *, zoneid_t, - ip_stack_t *); -extern ipxmit_state_t ip_xmit_v4(mblk_t *, ire_t *, struct ipsec_out_s *, - boolean_t, conn_t *); -extern int ip_hdr_complete(ipha_t *, zoneid_t, ip_stack_t *); +extern void arp_bringup_done(ill_t *, int); +extern void arp_replumb_done(ill_t *, int); extern struct qinit iprinitv6; -extern struct qinit ipwinitv6; extern void ipmp_init(ip_stack_t *); extern void ipmp_destroy(ip_stack_t *); @@ -3347,12 +3302,11 @@ extern ill_t *ipmp_illgrp_add_ipif(ipmp_illgrp_t *, ipif_t *); extern void ipmp_illgrp_del_ipif(ipmp_illgrp_t *, ipif_t *); extern ill_t *ipmp_illgrp_next_ill(ipmp_illgrp_t *); extern ill_t *ipmp_illgrp_hold_next_ill(ipmp_illgrp_t *); -extern ill_t *ipmp_illgrp_cast_ill(ipmp_illgrp_t *); extern ill_t *ipmp_illgrp_hold_cast_ill(ipmp_illgrp_t *); extern ill_t *ipmp_illgrp_ipmp_ill(ipmp_illgrp_t *); extern void ipmp_illgrp_refresh_mtu(ipmp_illgrp_t *); -extern ipmp_arpent_t *ipmp_illgrp_create_arpent(ipmp_illgrp_t *, mblk_t *, - boolean_t); +extern ipmp_arpent_t *ipmp_illgrp_create_arpent(ipmp_illgrp_t *, + boolean_t, ipaddr_t, uchar_t *, size_t, uint16_t); extern void ipmp_illgrp_destroy_arpent(ipmp_illgrp_t *, ipmp_arpent_t *); extern ipmp_arpent_t *ipmp_illgrp_lookup_arpent(ipmp_illgrp_t *, ipaddr_t *); extern void ipmp_illgrp_refresh_arpent(ipmp_illgrp_t *); @@ -3373,19 +3327,25 @@ extern ill_t *ipmp_ipif_bound_ill(const ipif_t *); extern ill_t *ipmp_ipif_hold_bound_ill(const ipif_t *); extern boolean_t ipmp_ipif_is_dataaddr(const ipif_t *); extern boolean_t ipmp_ipif_is_stubaddr(const ipif_t *); +extern boolean_t ipmp_packet_is_probe(mblk_t *, ill_t *); +extern ill_t *ipmp_ill_get_xmit_ill(ill_t *, boolean_t); +extern void ipmp_ncec_flush_nce(ncec_t *); +extern void ipmp_ncec_fastpath(ncec_t *, ill_t *); extern void conn_drain_insert(conn_t *, idl_tx_list_t *); +extern void conn_setqfull(conn_t *, boolean_t *); +extern void conn_clrqfull(conn_t *, boolean_t *); extern int conn_ipsec_length(conn_t *); -extern void ip_wput_ipsec_out(queue_t *, mblk_t *, ipha_t *, ill_t *, - ire_t *); extern ipaddr_t ip_get_dst(ipha_t *); -extern int ipsec_out_extra_length(mblk_t *); -extern int ipsec_in_extra_length(mblk_t *); -extern mblk_t *ipsec_in_alloc(boolean_t, netstack_t *); -extern boolean_t ipsec_in_is_secure(mblk_t *); -extern void ipsec_out_process(queue_t *, mblk_t *, ire_t *, uint_t); -extern void ipsec_out_to_in(mblk_t *); -extern void ip_fanout_proto_again(mblk_t *, ill_t *, ill_t *, ire_t *); +extern uint_t ip_get_pmtu(ip_xmit_attr_t *); +extern uint_t ip_get_base_mtu(ill_t *, ire_t *); +extern mblk_t *ip_output_attach_policy(mblk_t *, ipha_t *, ip6_t *, + const conn_t *, ip_xmit_attr_t *); +extern int ipsec_out_extra_length(ip_xmit_attr_t *); +extern int ipsec_out_process(mblk_t *, ip_xmit_attr_t *); +extern int ip_output_post_ipsec(mblk_t *, ip_xmit_attr_t *); +extern void ipsec_out_to_in(ip_xmit_attr_t *, ill_t *ill, + ip_recv_attr_t *); extern void ire_cleanup(ire_t *); extern void ire_inactive(ire_t *); @@ -3407,14 +3367,13 @@ extern uint_t ip_srcid_find_addr(const in6_addr_t *, zoneid_t, netstack_t *); extern uint8_t ipoptp_next(ipoptp_t *); extern uint8_t ipoptp_first(ipoptp_t *, ipha_t *); -extern int ip_opt_get_user(const ipha_t *, uchar_t *); +extern int ip_opt_get_user(conn_t *, uchar_t *); extern int ipsec_req_from_conn(conn_t *, ipsec_req_t *, int); extern int ip_snmp_get(queue_t *q, mblk_t *mctl, int level); extern int ip_snmp_set(queue_t *q, int, int, uchar_t *, int); extern void ip_process_ioctl(ipsq_t *, queue_t *, mblk_t *, void *); extern void ip_quiesce_conn(conn_t *); extern void ip_reprocess_ioctl(ipsq_t *, queue_t *, mblk_t *, void *); -extern void ip_restart_optmgmt(ipsq_t *, queue_t *, mblk_t *, void *); extern void ip_ioctl_finish(queue_t *, mblk_t *, int, int, ipsq_t *); extern boolean_t ip_cmpbuf(const void *, uint_t, boolean_t, const void *, @@ -3425,32 +3384,36 @@ extern void ip_savebuf(void **, uint_t *, boolean_t, const void *, uint_t); extern boolean_t ipsq_pending_mp_cleanup(ill_t *, conn_t *); extern void conn_ioctl_cleanup(conn_t *); -extern ill_t *conn_get_held_ill(conn_t *, ill_t **, int *); - -struct tcp_stack; -extern void ip_xmit_reset_serialize(mblk_t *, int, zoneid_t, struct tcp_stack *, - conn_t *); - -struct multidata_s; -struct pdesc_s; - -extern mblk_t *ip_mdinfo_alloc(ill_mdt_capab_t *); -extern mblk_t *ip_mdinfo_return(ire_t *, conn_t *, char *, ill_mdt_capab_t *); -extern mblk_t *ip_lsoinfo_alloc(ill_lso_capab_t *); -extern mblk_t *ip_lsoinfo_return(ire_t *, conn_t *, char *, - ill_lso_capab_t *); -extern uint_t ip_md_cksum(struct pdesc_s *, int, uint_t); -extern boolean_t ip_md_addr_attr(struct multidata_s *, struct pdesc_s *, - const mblk_t *); -extern boolean_t ip_md_hcksum_attr(struct multidata_s *, struct pdesc_s *, - uint32_t, uint32_t, uint32_t, uint32_t); -extern boolean_t ip_md_zcopy_attr(struct multidata_s *, struct pdesc_s *, - uint_t); + extern void ip_unbind(conn_t *); extern void tnet_init(void); extern void tnet_fini(void); +/* + * Hook functions to enable cluster networking + * On non-clustered systems these vectors must always be NULL. + */ +extern int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol, + sa_family_t addr_family, uint8_t *laddrp, void *args); +extern uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol, + sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp, + void *args); +extern int (*cl_inet_connect2)(netstackid_t stack_id, uint8_t protocol, + boolean_t is_outgoing, sa_family_t addr_family, uint8_t *laddrp, + in_port_t lport, uint8_t *faddrp, in_port_t fport, void *args); +extern void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t, + void *); +extern void (*cl_inet_getspi)(netstackid_t stack_id, uint8_t protocol, + uint8_t *ptr, size_t len, void *args); +extern int (*cl_inet_checkspi)(netstackid_t stack_id, uint8_t protocol, + uint32_t spi, void *args); +extern void (*cl_inet_deletespi)(netstackid_t stack_id, uint8_t protocol, + uint32_t spi, void *args); +extern void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t, + sa_family_t, in6_addr_t, in6_addr_t, void *); + + /* Hooks for CGTP (multirt routes) filtering module */ #define CGTP_FILTER_REV_1 1 #define CGTP_FILTER_REV_2 2 @@ -3491,73 +3454,6 @@ extern int ip_cgtp_filter_register(netstackid_t, cgtp_filter_ops_t *); extern int ip_cgtp_filter_unregister(netstackid_t); extern int ip_cgtp_filter_is_registered(netstackid_t); -/* Flags for ire_multirt_lookup() */ - -#define MULTIRT_USESTAMP 0x0001 -#define MULTIRT_SETSTAMP 0x0002 -#define MULTIRT_CACHEGW 0x0004 - -/* Debug stuff for multirt route resolution. */ -#if defined(DEBUG) && !defined(__lint) -/* Our "don't send, rather drop" flag. */ -#define MULTIRT_DEBUG_FLAG 0x8000 - -#define MULTIRT_TRACE(x) ip2dbg(x) - -#define MULTIRT_DEBUG_TAG(mblk) \ - do { \ - ASSERT(mblk != NULL); \ - MULTIRT_TRACE(("%s[%d]: tagging mblk %p, tag was %d\n", \ - __FILE__, __LINE__, \ - (void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \ - (mblk)->b_flag |= MULTIRT_DEBUG_FLAG; \ - } while (0) - -#define MULTIRT_DEBUG_UNTAG(mblk) \ - do { \ - ASSERT(mblk != NULL); \ - MULTIRT_TRACE(("%s[%d]: untagging mblk %p, tag was %d\n", \ - __FILE__, __LINE__, \ - (void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \ - (mblk)->b_flag &= ~MULTIRT_DEBUG_FLAG; \ - } while (0) - -#define MULTIRT_DEBUG_TAGGED(mblk) \ - (((mblk)->b_flag & MULTIRT_DEBUG_FLAG) ? B_TRUE : B_FALSE) -#else -#define MULTIRT_DEBUG_TAG(mblk) ASSERT(mblk != NULL) -#define MULTIRT_DEBUG_UNTAG(mblk) ASSERT(mblk != NULL) -#define MULTIRT_DEBUG_TAGGED(mblk) B_FALSE -#endif - -/* - * Per-ILL Multidata Transmit capabilities. - */ -struct ill_mdt_capab_s { - uint_t ill_mdt_version; /* interface version */ - uint_t ill_mdt_on; /* on/off switch for MDT on this ILL */ - uint_t ill_mdt_hdr_head; /* leading header fragment extra space */ - uint_t ill_mdt_hdr_tail; /* trailing header fragment extra space */ - uint_t ill_mdt_max_pld; /* maximum payload buffers per Multidata */ - uint_t ill_mdt_span_limit; /* maximum payload span per packet */ -}; - -struct ill_hcksum_capab_s { - uint_t ill_hcksum_version; /* interface version */ - uint_t ill_hcksum_txflags; /* capabilities on transmit */ -}; - -struct ill_zerocopy_capab_s { - uint_t ill_zerocopy_version; /* interface version */ - uint_t ill_zerocopy_flags; /* capabilities */ -}; - -struct ill_lso_capab_s { - uint_t ill_lso_on; /* on/off switch for LSO on this ILL */ - uint_t ill_lso_flags; /* capabilities */ - uint_t ill_lso_max; /* maximum size of payload */ -}; - /* * rr_ring_state cycles in the order shown below from RR_FREE through * RR_FREE_IN_PROG and back to RR_FREE. @@ -3669,18 +3565,61 @@ extern void ip_squeue_clean_ring(ill_t *, ill_rx_ring_t *); extern void ip_squeue_quiesce_ring(ill_t *, ill_rx_ring_t *); extern void ip_squeue_restart_ring(ill_t *, ill_rx_ring_t *); extern void ip_squeue_clean_all(ill_t *); +extern boolean_t ip_source_routed(ipha_t *, ip_stack_t *); extern void tcp_wput(queue_t *, mblk_t *); -extern int ip_fill_mtuinfo(struct in6_addr *, in_port_t, - struct ip6_mtuinfo *, netstack_t *); -extern ipif_t *conn_get_held_ipif(conn_t *, ipif_t **, int *); +extern int ip_fill_mtuinfo(conn_t *, ip_xmit_attr_t *, + struct ip6_mtuinfo *); extern hook_t *ipobs_register_hook(netstack_t *, pfv_t); extern void ipobs_unregister_hook(netstack_t *, hook_t *); extern void ipobs_hook(mblk_t *, int, zoneid_t, zoneid_t, const ill_t *, ip_stack_t *); typedef void (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *); +extern void dce_g_init(void); +extern void dce_g_destroy(void); +extern void dce_stack_init(ip_stack_t *); +extern void dce_stack_destroy(ip_stack_t *); +extern void dce_cleanup(uint_t, ip_stack_t *); +extern dce_t *dce_get_default(ip_stack_t *); +extern dce_t *dce_lookup_pkt(mblk_t *, ip_xmit_attr_t *, uint_t *); +extern dce_t *dce_lookup_v4(ipaddr_t, ip_stack_t *, uint_t *); +extern dce_t *dce_lookup_v6(const in6_addr_t *, uint_t, ip_stack_t *, + uint_t *); +extern dce_t *dce_lookup_and_add_v4(ipaddr_t, ip_stack_t *); +extern dce_t *dce_lookup_and_add_v6(const in6_addr_t *, uint_t, + ip_stack_t *); +extern int dce_update_uinfo_v4(ipaddr_t, iulp_t *, ip_stack_t *); +extern int dce_update_uinfo_v6(const in6_addr_t *, uint_t, iulp_t *, + ip_stack_t *); +extern int dce_update_uinfo(const in6_addr_t *, uint_t, iulp_t *, + ip_stack_t *); +extern void dce_increment_generation(dce_t *); +extern void dce_increment_all_generations(boolean_t, ip_stack_t *); +extern void dce_refrele(dce_t *); +extern void dce_refhold(dce_t *); +extern void dce_refrele_notr(dce_t *); +extern void dce_refhold_notr(dce_t *); +mblk_t *ip_snmp_get_mib2_ip_dce(queue_t *, mblk_t *, ip_stack_t *ipst); + +extern ip_laddr_t ip_laddr_verify_v4(ipaddr_t, zoneid_t, + ip_stack_t *, boolean_t); +extern ip_laddr_t ip_laddr_verify_v6(const in6_addr_t *, zoneid_t, + ip_stack_t *, boolean_t, uint_t); +extern int ip_laddr_fanout_insert(conn_t *); + +extern boolean_t ip_verify_src(mblk_t *, ip_xmit_attr_t *, uint_t *); +extern int ip_verify_ire(mblk_t *, ip_xmit_attr_t *); + +extern mblk_t *ip_xmit_attr_to_mblk(ip_xmit_attr_t *); +extern boolean_t ip_xmit_attr_from_mblk(mblk_t *, ip_xmit_attr_t *); +extern mblk_t *ip_xmit_attr_free_mblk(mblk_t *); +extern mblk_t *ip_recv_attr_to_mblk(ip_recv_attr_t *); +extern boolean_t ip_recv_attr_from_mblk(mblk_t *, ip_recv_attr_t *); +extern mblk_t *ip_recv_attr_free_mblk(mblk_t *); +extern boolean_t ip_recv_attr_is_mblk(mblk_t *); + /* * Squeue tags. Tags only need to be unique when the callback function is the * same to distinguish between different calls, but we use unique tags for @@ -3729,16 +3668,8 @@ typedef void (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *); #define SQTAG_CONNECT_FINISH 41 #define SQTAG_SYNCHRONOUS_OP 42 #define SQTAG_TCP_SHUTDOWN_OUTPUT 43 -#define SQTAG_XMIT_EARLY_RESET 44 - -#define NOT_OVER_IP(ip_wq) \ - (ip_wq->q_next != NULL || \ - (ip_wq->q_qinfo->qi_minfo->mi_idname) == NULL || \ - strcmp(ip_wq->q_qinfo->qi_minfo->mi_idname, \ - IP_MOD_NAME) != 0 || \ - ip_wq->q_qinfo->qi_minfo->mi_idnum != IP_MOD_ID) +#define SQTAG_TCP_IXA_CLEANUP 44 -#define PROTO_FLOW_CNTRLD(connp) (connp->conn_flow_cntrld) #endif /* _KERNEL */ #ifdef __cplusplus |