Home | History | Annotate | Download | only in inet
      1 /*
      2  * CDDL HEADER START
      3  *
      4  * The contents of this file are subject to the terms of the
      5  * Common Development and Distribution License (the "License").
      6  * You may not use this file except in compliance with the License.
      7  *
      8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
      9  * or http://www.opensolaris.org/os/licensing.
     10  * See the License for the specific language governing permissions
     11  * and limitations under the License.
     12  *
     13  * When distributing Covered Code, include this CDDL HEADER in each
     14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
     15  * If applicable, add the following below this CDDL HEADER, with the
     16  * fields enclosed by brackets "[]" replaced with your own identifying
     17  * information: Portions Copyright [yyyy] [name of copyright owner]
     18  *
     19  * CDDL HEADER END
     20  */
     21 /*
     22  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
     23  * Use is subject to license terms.
     24  */
     25 
     26 #ifndef	_INET_IP_IMPL_H
     27 #define	_INET_IP_IMPL_H
     28 
     29 #pragma ident	"%Z%%M%	%I%	%E% SMI"
     30 
     31 /*
     32  * IP implementation private declarations.  These interfaces are
     33  * used to build the IP module and are not meant to be accessed
     34  * by any modules except IP itself.  They are undocumented and are
     35  * subject to change without notice.
     36  */
     37 
     38 #ifdef	__cplusplus
     39 extern "C" {
     40 #endif
     41 
     42 #ifdef _KERNEL
     43 
     44 #include <sys/sdt.h>
     45 
     46 #define	IP_MOD_ID		5701
     47 
     48 #ifdef	_BIG_ENDIAN
     49 #define	IP_HDR_CSUM_TTL_ADJUST	256
     50 #define	IP_TCP_CSUM_COMP	IPPROTO_TCP
     51 #define	IP_UDP_CSUM_COMP	IPPROTO_UDP
     52 #else
     53 #define	IP_HDR_CSUM_TTL_ADJUST	1
     54 #define	IP_TCP_CSUM_COMP	(IPPROTO_TCP << 8)
     55 #define	IP_UDP_CSUM_COMP	(IPPROTO_UDP << 8)
     56 #endif
     57 
     58 #define	TCP_CHECKSUM_OFFSET	16
     59 #define	TCP_CHECKSUM_SIZE	2
     60 
     61 #define	UDP_CHECKSUM_OFFSET	6
     62 #define	UDP_CHECKSUM_SIZE	2
     63 
     64 #define	IPH_TCPH_CHECKSUMP(ipha, hlen)	\
     65 	((uint16_t *)(((uchar_t *)(ipha)) + ((hlen) + TCP_CHECKSUM_OFFSET)))
     66 
     67 #define	IPH_UDPH_CHECKSUMP(ipha, hlen)	\
     68 	((uint16_t *)(((uchar_t *)(ipha)) + ((hlen) + UDP_CHECKSUM_OFFSET)))
     69 
     70 #define	ILL_HCKSUM_CAPABLE(ill)		\
     71 	(((ill)->ill_capabilities & ILL_CAPAB_HCKSUM) != 0)
     72 /*
     73  * Macro that performs software checksum calculation on the IP header.
     74  */
     75 #define	IP_HDR_CKSUM(ipha, sum, v_hlen_tos_len, ttl_protocol) {		\
     76 	(sum) += (ttl_protocol) + (ipha)->ipha_ident +			\
     77 	    ((v_hlen_tos_len) >> 16) +					\
     78 	    ((v_hlen_tos_len) & 0xFFFF) +				\
     79 	    (ipha)->ipha_fragment_offset_and_flags;			\
     80 	(sum) = (((sum) & 0xFFFF) + ((sum) >> 16));			\
     81 	(sum) = ~((sum) + ((sum) >> 16));				\
     82 	(ipha)->ipha_hdr_checksum = (uint16_t)(sum);			\
     83 }
     84 
     85 #define	IS_IP_HDR_HWCKSUM(ipsec, mp, ill)				\
     86 	((!ipsec) && (DB_CKSUMFLAGS(mp) & HCK_IPV4_HDRCKSUM) &&		\
     87 	ILL_HCKSUM_CAPABLE(ill) && dohwcksum)
     88 
     89 /*
     90  * This macro acts as a wrapper around IP_CKSUM_XMIT_FAST, and it performs
     91  * several checks on the IRE and ILL (among other things) in order to see
     92  * whether or not hardware checksum offload is allowed for the outgoing
     93  * packet.  It assumes that the caller has held a reference to the IRE.
     94  */
     95 #define	IP_CKSUM_XMIT(ill, ire, mp, ihp, up, proto, start, end,		\
     96 	    max_frag, ipsec_len, pseudo) {				\
     97 	uint32_t _hck_flags;						\
     98 	/*								\
     99 	 * We offload checksum calculation to hardware when IPsec isn't	\
    100 	 * present and if fragmentation isn't required.  We also check	\
    101 	 * if M_DATA fastpath is safe to be used on the	corresponding	\
    102 	 * IRE; this check is performed without grabbing ire_lock but	\
    103 	 * instead by holding a reference to it.  This is sufficient	\
    104 	 * for IRE_CACHE; for IRE_BROADCAST on non-Ethernet links, the	\
    105 	 * DL_NOTE_FASTPATH_FLUSH indication could come up from the	\
    106 	 * driver and trigger the IRE (hence fp_mp) deletion.  This is	\
    107 	 * why only IRE_CACHE type is eligible for offload.		\
    108 	 *								\
    109 	 * The presense of IP options also forces the network stack to	\
    110 	 * calculate the checksum in software.  This is because:	\
    111 	 *								\
    112 	 * Wrap around: certain partial-checksum NICs (eri, ce) limit	\
    113 	 * the size of "start offset" width to 6-bit.  This effectively	\
    114 	 * sets the largest value of the offset to 64-bytes, starting	\
    115 	 * from the MAC header.  When the cumulative MAC and IP headers	\
    116 	 * exceed such limit, the offset will wrap around.  This causes	\
    117 	 * the checksum to be calculated at the wrong place.		\
    118 	 *								\
    119 	 * IPv4 source routing: none of the full-checksum capable NICs	\
    120 	 * is capable of correctly handling the	IPv4 source-routing	\
    121 	 * option for purposes of calculating the pseudo-header; the	\
    122 	 * actual destination is different from the destination in the	\
    123 	 * header which is that of the next-hop.  (This case may not be	\
    124 	 * true for NICs which can parse IPv6 extension headers, but	\
    125 	 * we choose to simplify the implementation by not offloading	\
    126 	 * checksum when they are present.)				\
    127 	 *								\
    128 	 */								\
    129 	if ((ill) != NULL && ILL_HCKSUM_CAPABLE(ill) &&			\
    130 	    !((ire)->ire_flags & RTF_MULTIRT) &&			\
    131 	    (!((ire)->ire_type & IRE_BROADCAST) ||			\
    132 	    (ill)->ill_type == IFT_ETHER) &&			\
    133 	    (ipsec_len) == 0 &&						\
    134 	    (((ire)->ire_ipversion == IPV4_VERSION &&			\
    135 	    (start) == IP_SIMPLE_HDR_LENGTH &&				\
    136 	    ((ire)->ire_nce != NULL &&					\
    137 	    (ire)->ire_nce->nce_fp_mp != NULL &&	\
    138 	    MBLKHEAD(mp) >= MBLKL((ire)->ire_nce->nce_fp_mp))) ||	\
    139 	    ((ire)->ire_ipversion == IPV6_VERSION &&			\
    140 	    (start) == IPV6_HDR_LEN &&					\
    141 	    (ire)->ire_nce->nce_fp_mp != NULL &&			\
    142 	    MBLKHEAD(mp) >= MBLKL((ire)->ire_nce->nce_fp_mp))) &&	\
    143 	    (max_frag) >= (uint_t)((end) + (ipsec_len)) &&		\
    144 	    dohwcksum) {						\
    145 		_hck_flags = (ill)->ill_hcksum_capab->ill_hcksum_txflags; \
    146 	} else {							\
    147 		_hck_flags = 0;						\
    148 	}								\
    149 	IP_CKSUM_XMIT_FAST((ire)->ire_ipversion, _hck_flags, mp, ihp,	\
    150 	    up, proto, start, end, pseudo);				\
    151 }
    152 
    153 /*
    154  * Based on the device capabilities, this macro either marks an outgoing
    155  * packet with hardware checksum offload information or calculate the
    156  * checksum in software.  If the latter is performed, the checksum field
    157  * of the dblk is cleared; otherwise it will be non-zero and contain the
    158  * necessary flag(s) for the driver.
    159  */
    160 #define	IP_CKSUM_XMIT_FAST(ipver, hck_flags, mp, ihp, up, proto, start,	\
    161 	    end, pseudo) {						\
    162 	uint32_t _sum;							\
    163 	/*								\
    164 	 * Underlying interface supports hardware checksum offload for	\
    165 	 * the payload; leave the payload checksum for the hardware to	\
    166 	 * calculate.  N.B: We only need to set up checksum info on the	\
    167 	 * first mblk.							\
    168 	 */								\
    169 	DB_CKSUMFLAGS(mp) = 0;						\
    170 	if (((ipver) == IPV4_VERSION &&					\
    171 	    ((hck_flags) & HCKSUM_INET_FULL_V4)) ||			\
    172 	    ((ipver) == IPV6_VERSION &&					\
    173 	    ((hck_flags) & HCKSUM_INET_FULL_V6))) {			\
    174 		/*							\
    175 		 * Hardware calculates pseudo-header, header and the	\
    176 		 * payload checksums, so clear the checksum field in	\
    177 		 * the protocol header.					\
    178 		 */							\
    179 		*(up) = 0;						\
    180 		DB_CKSUMFLAGS(mp) |= HCK_FULLCKSUM;			\
    181 	} else if ((hck_flags) & HCKSUM_INET_PARTIAL)  {		\
    182 		/*							\
    183 		 * Partial checksum offload has been enabled.  Fill	\
    184 		 * the checksum field in the protocl header with the	\
    185 		 * pseudo-header checksum value.			\
    186 		 */							\
    187 		_sum = ((proto) == IPPROTO_UDP) ?			\
    188 		    IP_UDP_CSUM_COMP : IP_TCP_CSUM_COMP;		\
    189 		_sum += *(up) + (pseudo);				\
    190 		_sum = (_sum & 0xFFFF) + (_sum >> 16);			\
    191 		*(up) = (_sum & 0xFFFF) + (_sum >> 16);			\
    192 		/*							\
    193 		 * Offsets are relative to beginning of IP header.	\
    194 		 */							\
    195 		DB_CKSUMSTART(mp) = (start);				\
    196 		DB_CKSUMSTUFF(mp) = ((proto) == IPPROTO_UDP) ?		\
    197 		    (start) + UDP_CHECKSUM_OFFSET :			\
    198 		    (start) + TCP_CHECKSUM_OFFSET;			\
    199 		DB_CKSUMEND(mp) = (end);				\
    200 		DB_CKSUMFLAGS(mp) |= HCK_PARTIALCKSUM;			\
    201 	} else {							\
    202 		/*							\
    203 		 * Software checksumming.				\
    204 		 */							\
    205 		_sum = ((proto) == IPPROTO_UDP) ?			\
    206 		    IP_UDP_CSUM_COMP : IP_TCP_CSUM_COMP;		\
    207 		_sum += (pseudo);					\
    208 		_sum = IP_CSUM(mp, start, _sum);			\
    209 		*(up) = (uint16_t)(((proto) == IPPROTO_UDP) ?		\
    210 		    (_sum ? _sum : ~_sum) : _sum);			\
    211 	}								\
    212 	/*								\
    213 	 * Hardware supports IP header checksum offload; clear the	\
    214 	 * contents of IP header checksum field as expected by NIC.	\
    215 	 * Do this only if we offloaded either full or partial sum.	\
    216 	 */								\
    217 	if ((ipver) == IPV4_VERSION && DB_CKSUMFLAGS(mp) != 0 &&	\
    218 	    ((hck_flags) & HCKSUM_IPHDRCKSUM)) {			\
    219 		DB_CKSUMFLAGS(mp) |= HCK_IPV4_HDRCKSUM;			\
    220 		((ipha_t *)(ihp))->ipha_hdr_checksum = 0;		\
    221 	}								\
    222 }
    223 
    224 /*
    225  * Macro to inspect the checksum of a fully-reassembled incoming datagram.
    226  */
    227 #define	IP_CKSUM_RECV_REASS(hck_flags, off, pseudo, sum, err) {		\
    228 	(err) = B_FALSE;						\
    229 	if ((hck_flags) & HCK_FULLCKSUM) {				\
    230 		/*							\
    231 		 * The sum of all fragment checksums should		\
    232 		 * result in -0 (0xFFFF) or otherwise invalid.		\
    233 		 */							\
    234 		if ((sum) != 0xFFFF)					\
    235 			(err) = B_TRUE;					\
    236 	} else if ((hck_flags) & HCK_PARTIALCKSUM) {			\
    237 		(sum) += (pseudo);					\
    238 		(sum) = ((sum) & 0xFFFF) + ((sum) >> 16);		\
    239 		(sum) = ((sum) & 0xFFFF) + ((sum) >> 16);		\
    240 		if (~(sum) & 0xFFFF)					\
    241 			(err) = B_TRUE;					\
    242 	} else if (((sum) = IP_CSUM(mp, off, pseudo)) != 0) {		\
    243 		(err) = B_TRUE;						\
    244 	}								\
    245 }
    246 
    247 /*
    248  * This macro inspects an incoming packet to see if the checksum value
    249  * contained in it is valid; if the hardware has provided the information,
    250  * the value is verified, otherwise it performs software checksumming.
    251  * The checksum value is returned to caller.
    252  */
    253 #define	IP_CKSUM_RECV(hck_flags, sum, cksum_start, ulph_off, mp, mp1, err) { \
    254 	int32_t _len;							\
    255 									\
    256 	(err) = B_FALSE;						\
    257 	if ((hck_flags) & HCK_FULLCKSUM) {				\
    258 		/*							\
    259 		 * Full checksum has been computed by the hardware	\
    260 		 * and has been attached.  If the driver wants us to	\
    261 		 * verify the correctness of the attached value, in	\
    262 		 * order to protect against faulty hardware, compare	\
    263 		 * it against -0 (0xFFFF) to see if it's valid.		\
    264 		 */							\
    265 		(sum) = DB_CKSUM16(mp);					\
    266 		if (!((hck_flags) & HCK_FULLCKSUM_OK) && (sum) != 0xFFFF) \
    267 			(err) = B_TRUE;					\
    268 	} else if (((hck_flags) & HCK_PARTIALCKSUM) &&			\
    269 	    ((mp1) == NULL || (mp1)->b_cont == NULL) &&			\
    270 	    (ulph_off) >= DB_CKSUMSTART(mp) &&				\
    271 	    ((_len = (ulph_off) - DB_CKSUMSTART(mp)) & 1) == 0) {	\
    272 		uint32_t _adj;						\
    273 		/*							\
    274 		 * Partial checksum has been calculated by hardware	\
    275 		 * and attached to the packet; in addition, any		\
    276 		 * prepended extraneous data is even byte aligned,	\
    277 		 * and there are at most two mblks associated with	\
    278 		 * the packet.  If any such data exists, we adjust	\
    279 		 * the checksum; also take care any postpended data.	\
    280 		 */							\
    281 		IP_ADJCKSUM_PARTIAL(cksum_start, mp, mp1, _len, _adj);	\
    282 		/*							\
    283 		 * One's complement subtract extraneous checksum	\
    284 		 */							\
    285 		(sum) += DB_CKSUM16(mp);				\
    286 		if (_adj >= (sum))					\
    287 			(sum) = ~(_adj - (sum)) & 0xFFFF;		\
    288 		else							\
    289 			(sum) -= _adj;					\
    290 		(sum) = ((sum) & 0xFFFF) + ((int)(sum) >> 16);		\
    291 		(sum) = ((sum) & 0xFFFF) + ((int)(sum) >> 16);		\
    292 		if (~(sum) & 0xFFFF)					\
    293 			(err) = B_TRUE;					\
    294 	} else if (((sum) = IP_CSUM(mp, ulph_off, sum)) != 0) {		\
    295 		(err) = B_TRUE;						\
    296 	}								\
    297 }
    298 
    299 /*
    300  * Macro to adjust a given checksum value depending on any prepended
    301  * or postpended data on the packet.  It expects the start offset to
    302  * begin at an even boundary and that the packet consists of at most
    303  * two mblks.
    304  */
    305 #define	IP_ADJCKSUM_PARTIAL(cksum_start, mp, mp1, len, adj) {		\
    306 	/*								\
    307 	 * Prepended extraneous data; adjust checksum.			\
    308 	 */								\
    309 	if ((len) > 0)							\
    310 		(adj) = IP_BCSUM_PARTIAL(cksum_start, len, 0);		\
    311 	else								\
    312 		(adj) = 0;						\
    313 	/*								\
    314 	 * len is now the total length of mblk(s)			\
    315 	 */								\
    316 	(len) = MBLKL(mp);						\
    317 	if ((mp1) == NULL)						\
    318 		(mp1) = (mp);						\
    319 	else								\
    320 		(len) += MBLKL(mp1);					\
    321 	/*								\
    322 	 * Postpended extraneous data; adjust checksum.			\
    323 	 */								\
    324 	if (((len) = (DB_CKSUMEND(mp) - len)) > 0) {			\
    325 		uint32_t _pad;						\
    326 									\
    327 		_pad = IP_BCSUM_PARTIAL((mp1)->b_wptr, len, 0);		\
    328 		/*							\
    329 		 * If the postpended extraneous data was odd		\
    330 		 * byte aligned, swap resulting checksum bytes.		\
    331 		 */							\
    332 		if ((uintptr_t)(mp1)->b_wptr & 1)			\
    333 			(adj) += ((_pad << 8) & 0xFFFF) | (_pad >> 8);	\
    334 		else							\
    335 			(adj) += _pad;					\
    336 		(adj) = ((adj) & 0xFFFF) + ((int)(adj) >> 16);		\
    337 	}								\
    338 }
    339 
    340 #define	ILL_MDT_CAPABLE(ill)		\
    341 	(((ill)->ill_capabilities & ILL_CAPAB_MDT) != 0)
    342 
    343 /*
    344  * ioctl identifier and structure for Multidata Transmit update
    345  * private M_CTL communication from IP to ULP.
    346  */
    347 #define	MDT_IOC_INFO_UPDATE	(('M' << 8) + 1020)
    348 
    349 typedef struct ip_mdt_info_s {
    350 	uint_t	mdt_info_id;	/* MDT_IOC_INFO_UPDATE */
    351 	ill_mdt_capab_t	mdt_capab; /* ILL MDT capabilities */
    352 } ip_mdt_info_t;
    353 
    354 /*
    355  * Macro that determines whether or not a given ILL is allowed for MDT.
    356  */
    357 #define	ILL_MDT_USABLE(ill)						\
    358 	(ILL_MDT_CAPABLE(ill) &&					\
    359 	ill->ill_mdt_capab != NULL &&					\
    360 	ill->ill_mdt_capab->ill_mdt_version == MDT_VERSION_2 &&		\
    361 	ill->ill_mdt_capab->ill_mdt_on != 0)
    362 
    363 #define	ILL_LSO_CAPABLE(ill)		\
    364 	(((ill)->ill_capabilities & ILL_CAPAB_LSO) != 0)
    365 
    366 /*
    367  * ioctl identifier and structure for Large Segment Offload
    368  * private M_CTL communication from IP to ULP.
    369  */
    370 #define	LSO_IOC_INFO_UPDATE	(('L' << 24) + ('S' << 16) + ('O' << 8))
    371 
    372 typedef struct ip_lso_info_s {
    373 	uint_t	lso_info_id;	/* LSO_IOC_INFO_UPDATE */
    374 	ill_lso_capab_t	lso_capab; /* ILL LSO capabilities */
    375 } ip_lso_info_t;
    376 
    377 /*
    378  * Macro that determines whether or not a given ILL is allowed for LSO.
    379  */
    380 #define	ILL_LSO_USABLE(ill)						\
    381 	(ILL_LSO_CAPABLE(ill) &&					\
    382 	ill->ill_lso_capab != NULL &&					\
    383 	ill->ill_lso_capab->ill_lso_version == LSO_VERSION_1 &&		\
    384 	ill->ill_lso_capab->ill_lso_on != 0)
    385 
    386 #define	ILL_LSO_TCP_USABLE(ill)						\
    387 	(ILL_LSO_USABLE(ill) &&						\
    388 	ill->ill_lso_capab->ill_lso_flags & LSO_TX_BASIC_TCP_IPV4)
    389 
    390 /*
    391  * Macro that determines whether or not a given CONN may be considered
    392  * for fast path prior to proceeding further with LSO or Multidata.
    393  */
    394 #define	CONN_IS_LSO_MD_FASTPATH(connp)	\
    395 	((connp)->conn_dontroute == 0 &&	/* SO_DONTROUTE */	\
    396 	!((connp)->conn_nexthop_set) &&		/* IP_NEXTHOP */	\
    397 	(connp)->conn_nofailover_ill == NULL &&	/* IPIF_NOFAILOVER */	\
    398 	(connp)->conn_outgoing_pill == NULL &&	/* IP{V6}_BOUND_PIF */	\
    399 	(connp)->conn_outgoing_ill == NULL)	/* IP{V6}_BOUND_IF */
    400 
    401 /* Definitons for fragmenting IP packets using MDT. */
    402 
    403 /*
    404  * Smaller and private version of pdescinfo_t used specifically for IP,
    405  * which allows for only a single payload span per packet.
    406  */
    407 typedef struct ip_pdescinfo_s PDESCINFO_STRUCT(2)	ip_pdescinfo_t;
    408 
    409 /*
    410  * Macro version of ip_can_frag_mdt() which avoids the function call if we
    411  * only examine a single message block.
    412  */
    413 #define	IP_CAN_FRAG_MDT(mp, hdr_len, len)			\
    414 	(((mp)->b_cont == NULL) ?				\
    415 	(MBLKL(mp) >= ((hdr_len) + ip_wput_frag_mdt_min)) :	\
    416 	ip_can_frag_mdt((mp), (hdr_len), (len)))
    417 
    418 /*
    419  * Macro that determines whether or not a given IPC requires
    420  * outbound IPSEC processing.
    421  */
    422 #define	CONN_IPSEC_OUT_ENCAPSULATED(connp)	\
    423 	((connp)->conn_out_enforce_policy ||	\
    424 	((connp)->conn_latch != NULL &&		\
    425 	(connp)->conn_latch->ipl_out_policy != NULL))
    426 
    427 /*
    428  * These are used by the synchronous streams code in tcp and udp.
    429  * When we set the flags for a wakeup from a synchronous stream we
    430  * always set RSLEEP in sd_wakeq, even if we have a read thread waiting
    431  * to do the io. This is in case the read thread gets interrupted
    432  * before completing the io. The RSLEEP flag in sd_wakeq is used to
    433  * indicate that there is data available at the synchronous barrier.
    434  * The assumption is that subsequent functions calls through rwnext()
    435  * will reset sd_wakeq appropriately.
    436  */
    437 #define	STR_WAKEUP_CLEAR(stp) {						\
    438 	mutex_enter(&stp->sd_lock);					\
    439 	stp->sd_wakeq &= ~RSLEEP;					\
    440 	mutex_exit(&stp->sd_lock);					\
    441 }
    442 
    443 #define	STR_WAKEUP_SET(stp) {						\
    444 	mutex_enter(&stp->sd_lock);					\
    445 	if (stp->sd_flag & RSLEEP) {					\
    446 		stp->sd_flag &= ~RSLEEP;				\
    447 		cv_broadcast(&_RD(stp->sd_wrq)->q_wait);		\
    448 	}								\
    449 	stp->sd_wakeq |= RSLEEP;					\
    450 	mutex_exit(&stp->sd_lock);					\
    451 }
    452 
    453 /*
    454  * Combined wakeup and sendsig to avoid dropping and reacquiring the
    455  * sd_lock. The list of messages waiting at the synchronous barrier is
    456  * supplied in order to determine whether a wakeup needs to occur. We
    457  * only send a wakeup to the application when necessary, i.e. during
    458  * the first enqueue when the received messages list will be NULL.
    459  */
    460 #define	STR_WAKEUP_SENDSIG(stp, rcv_list) {				\
    461 	int _events;							\
    462 	mutex_enter(&stp->sd_lock);					\
    463 	if (rcv_list == NULL) {						\
    464 		if (stp->sd_flag & RSLEEP) {				\
    465 			stp->sd_flag &= ~RSLEEP;			\
    466 			cv_broadcast(&_RD(stp->sd_wrq)->q_wait);	\
    467 		}							\
    468 		stp->sd_wakeq |= RSLEEP;				\
    469 	}								\
    470 	if ((_events = stp->sd_sigflags & (S_INPUT | S_RDNORM)) != 0)	\
    471 		strsendsig(stp->sd_siglist, _events, 0, 0);		\
    472 	if (stp->sd_rput_opt & SR_POLLIN) {				\
    473 		stp->sd_rput_opt &= ~SR_POLLIN;				\
    474 		mutex_exit(&stp->sd_lock);				\
    475 		pollwakeup(&stp->sd_pollist, POLLIN | POLLRDNORM);	\
    476 	} else {							\
    477 		mutex_exit(&stp->sd_lock);				\
    478 	}								\
    479 }
    480 
    481 #define	CONN_UDP_SYNCSTR(connp)						\
    482 	(IPCL_IS_UDP(connp) && (connp)->conn_udp->udp_direct_sockfs)
    483 
    484 /*
    485  * Macro that checks whether or not a particular UDP conn is
    486  * flow-controlling on the read-side.  If udp module is directly
    487  * above ip, check to see if the drain queue is full; note here
    488  * that we check this without any lock protection because this
    489  * is a coarse granularity inbound flow-control.  If the module
    490  * above ip is not udp, then use canputnext to determine the
    491  * flow-control.
    492  *
    493  * Note that these checks are done after the conn is found in
    494  * the UDP fanout table.
    495  * FIXME? Might be faster to check both udp_drain_qfull and canputnext.
    496  */
    497 #define	CONN_UDP_FLOWCTLD(connp)					\
    498 	(CONN_UDP_SYNCSTR(connp) ?					\
    499 	(connp)->conn_udp->udp_drain_qfull :				\
    500 	!canputnext((connp)->conn_rq))
    501 
    502 #define	ILL_DLS_CAPABLE(ill)	\
    503 	(((ill)->ill_capabilities &		\
    504 	(ILL_CAPAB_POLL|ILL_CAPAB_SOFT_RING)) != 0)
    505 
    506 /*
    507  * Macro that hands off one or more messages directly to DLD
    508  * when the interface is marked with ILL_CAPAB_POLL.
    509  */
    510 #define	IP_DLS_ILL_TX(ill, ipha, mp, ipst) {				\
    511 	ill_dls_capab_t *ill_dls = ill->ill_dls_capab;			\
    512 	ASSERT(ILL_DLS_CAPABLE(ill));					\
    513 	ASSERT(ill_dls != NULL);					\
    514 	ASSERT(ill_dls->ill_tx != NULL);				\
    515 	ASSERT(ill_dls->ill_tx_handle != NULL);				\
    516 	DTRACE_PROBE4(ip4__physical__out__start,			\
    517 	    ill_t *, NULL, ill_t *, ill,				\
    518 	    ipha_t *, ipha, mblk_t *, mp);				\
    519 	FW_HOOKS(ipst->ips_ip4_physical_out_event,			\
    520 	    ipst->ips_ipv4firewall_physical_out,			\
    521 	    NULL, ill, ipha, mp, mp, 0, ipst);				\
    522 	DTRACE_PROBE1(ip4__physical__out__end, mblk_t *, mp);		\
    523 	if (mp != NULL)	{						\
    524 		DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,		\
    525 		    void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,	\
    526 		    ipha_t *, ipha, ip6_t *, NULL, int,	0);		\
    527 		ill_dls->ill_tx(ill_dls->ill_tx_handle, mp);		\
    528 	}								\
    529 }
    530 
    531 extern int	ip_wput_frag_mdt_min;
    532 extern boolean_t ip_can_frag_mdt(mblk_t *, ssize_t, ssize_t);
    533 extern mblk_t   *ip_prepend_zoneid(mblk_t *, zoneid_t, ip_stack_t *);
    534 
    535 #endif	/* _KERNEL */
    536 
    537 #ifdef	__cplusplus
    538 }
    539 #endif
    540 
    541 #endif	/* _INET_IP_IMPL_H */
    542