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  * Copyright (c) 2016 by Delphix. All rights reserved.
25  */
26 
27 /*
28  * SunOS 5.x Multithreaded STREAMS DLPI FCIP Module
29  * This is a pseudo driver module to handle encapsulation of IP and ARP
30  * datagrams over FibreChannel interfaces. FCIP is a cloneable STREAMS
31  * driver module which interfaces with IP/ARP using DLPI. This module
32  * is a Style-2 DLS provider.
33  *
34  * The implementation of this module is based on RFC 2625 which gives
35  * details on the encapsulation of IP/ARP data over FibreChannel.
36  * The fcip module needs to resolve an IP address to a port address before
37  * sending data to a destination port. A FC device port has 2 addresses
38  * associated with it: A 8 byte World Wide unique Port Name and a 3 byte
39  * volatile Port number or Port_ID.
40  *
41  * The mapping between a IP address and the World Wide Port Name is handled
42  * by the ARP layer since the IP over FC draft requires the MAC address to
43  * be the least significant six bytes of the WorldWide Port Names. The
44  * fcip module however needs to identify the destination port uniquely when
45  * the destination FC device has multiple FC ports.
46  *
47  * The FC layer mapping between the World Wide Port Name and the Port_ID
48  * will be handled through the use of a fabric name server or through the
49  * use of the FARP ELS command as described in the draft. Since the Port_IDs
50  * are volatile, the mapping between the World Wide Port Name and Port_IDs
51  * must be maintained and validated before use each time a datagram
52  * needs to be sent to the destination ports. The FC transport module
53  * informs the fcip module of all changes to states of ports on the
54  * fabric through registered callbacks. This enables the fcip module
55  * to maintain the WW_PN to Port_ID mappings current.
56  *
57  * For details on how this module interfaces with the FibreChannel Transport
58  * modules, refer to PSARC/1997/385. Chapter 3 of the FibreChannel Transport
59  * Programming guide details the APIs between ULPs and the Transport.
60  *
61  * Now for some Caveats:
62  *
63  * RFC 2625 requires that a FibreChannel Port name (the Port WWN) have
64  * the NAA bits set to '0001' indicating a IEEE 48bit address which
65  * corresponds to a ULA (Universal LAN MAC address). But with FibreChannel
66  * adapters containing 2 or more ports, IEEE naming cannot identify the
67  * ports on an adapter uniquely so we will in the first implementation
68  * be operating only on Port 0 of each adapter.
69  */
70 
71 #include	<sys/types.h>
72 #include	<sys/errno.h>
73 #include	<sys/debug.h>
74 #include	<sys/time.h>
75 #include	<sys/sysmacros.h>
76 #include	<sys/systm.h>
77 #include	<sys/user.h>
78 #include	<sys/stropts.h>
79 #include	<sys/stream.h>
80 #include	<sys/strlog.h>
81 #include	<sys/strsubr.h>
82 #include	<sys/cmn_err.h>
83 #include	<sys/cpu.h>
84 #include	<sys/kmem.h>
85 #include	<sys/conf.h>
86 #include	<sys/ddi.h>
87 #include	<sys/sunddi.h>
88 #include	<sys/ksynch.h>
89 #include	<sys/stat.h>
90 #include	<sys/kstat.h>
91 #include	<sys/vtrace.h>
92 #include	<sys/strsun.h>
93 #include	<sys/varargs.h>
94 #include	<sys/modctl.h>
95 #include 	<sys/thread.h>
96 #include 	<sys/var.h>
97 #include 	<sys/proc.h>
98 #include	<inet/common.h>
99 #include	<netinet/ip6.h>
100 #include	<inet/ip.h>
101 #include	<inet/arp.h>
102 #include	<inet/mi.h>
103 #include	<inet/nd.h>
104 #include	<sys/dlpi.h>
105 #include	<sys/ethernet.h>
106 #include	<sys/file.h>
107 #include	<sys/syslog.h>
108 #include	<sys/disp.h>
109 #include	<sys/taskq.h>
110 
111 /*
112  * Leadville includes
113  */
114 
115 #include	<sys/fibre-channel/fc.h>
116 #include	<sys/fibre-channel/impl/fc_ulpif.h>
117 #include	<sys/fibre-channel/ulp/fcip.h>
118 
119 /*
120  * TNF Probe/trace facility include
121  */
122 #if defined(lint) || defined(FCIP_TNF_ENABLED)
123 #include <sys/tnf_probe.h>
124 #endif
125 
126 #define	FCIP_ESBALLOC
127 
128 /*
129  * Function prototypes
130  */
131 
132 /* standard loadable modules entry points */
133 static int	fcip_attach(dev_info_t *, ddi_attach_cmd_t);
134 static int 	fcip_detach(dev_info_t *, ddi_detach_cmd_t);
135 static void 	fcip_dodetach(struct fcipstr *slp);
136 static int fcip_getinfo(dev_info_t *dip, ddi_info_cmd_t cmd,
137     void *arg, void **result);
138 
139 
140 /* streams specific */
141 static void fcip_setipq(struct fcip *fptr);
142 static int fcip_wput(queue_t *, mblk_t *);
143 static int fcip_wsrv(queue_t *);
144 static void fcip_proto(queue_t *, mblk_t *);
145 static void fcip_ioctl(queue_t *, mblk_t *);
146 static int fcip_open(queue_t *wq, dev_t *devp, int flag,
147 		int sflag, cred_t *credp);
148 static int fcip_close(queue_t *rq, int flag, cred_t *credp);
149 static int fcip_start(queue_t *wq, mblk_t *mp, struct fcip *fptr,
150     struct fcip_dest *fdestp, int flags);
151 static void fcip_sendup(struct fcip *fptr, mblk_t *mp,
152     struct fcipstr *(*acceptfunc)());
153 static struct fcipstr *fcip_accept(struct fcipstr *slp, struct fcip *fptr,
154     int type, la_wwn_t *dhostp);
155 static mblk_t *fcip_addudind(struct fcip *fptr, mblk_t *mp,
156     fcph_network_hdr_t *nhdr, int type);
157 static int fcip_setup_mac_addr(struct fcip *fptr);
158 static void fcip_kstat_init(struct fcip *fptr);
159 static int fcip_stat_update(kstat_t *, int);
160 
161 
162 /* dlpi specific */
163 static void fcip_spareq(queue_t *wq, mblk_t *mp);
164 static void fcip_pareq(queue_t *wq, mblk_t *mp);
165 static void fcip_ubreq(queue_t *wq, mblk_t *mp);
166 static void fcip_breq(queue_t *wq, mblk_t *mp);
167 static void fcip_dreq(queue_t *wq, mblk_t *mp);
168 static void fcip_areq(queue_t *wq, mblk_t *mp);
169 static void fcip_udreq(queue_t *wq, mblk_t *mp);
170 static void fcip_ireq(queue_t *wq, mblk_t *mp);
171 static void fcip_dl_ioc_hdr_info(queue_t *wq, mblk_t *mp);
172 
173 
174 /* solaris sundry, DR/CPR etc */
175 static int fcip_cache_constructor(void *buf, void *arg, int size);
176 static void fcip_cache_destructor(void *buf, void *size);
177 static int fcip_handle_suspend(fcip_port_info_t *fport, fc_detach_cmd_t cmd);
178 static int fcip_handle_resume(fcip_port_info_t *fport,
179     fc_ulp_port_info_t *port_info, fc_attach_cmd_t cmd);
180 static fcip_port_info_t *fcip_softstate_free(fcip_port_info_t *fport);
181 static int fcip_port_attach_handler(struct fcip *fptr);
182 
183 
184 /*
185  * ulp - transport interface function prototypes
186  */
187 static int fcip_port_attach(opaque_t ulp_handle, fc_ulp_port_info_t *,
188     fc_attach_cmd_t cmd, uint32_t sid);
189 static int fcip_port_detach(opaque_t ulp_handle, fc_ulp_port_info_t *,
190     fc_detach_cmd_t cmd);
191 static int fcip_port_ioctl(opaque_t ulp_handle,  opaque_t port_handle,
192     dev_t dev, int cmd, intptr_t data, int mode, cred_t *credp, int *rval,
193     uint32_t claimed);
194 static void fcip_statec_cb(opaque_t ulp_handle, opaque_t phandle,
195     uint32_t port_state, uint32_t port_top, fc_portmap_t changelist[],
196     uint32_t listlen, uint32_t sid);
197 static int fcip_els_cb(opaque_t ulp_handle, opaque_t phandle,
198     fc_unsol_buf_t *buf, uint32_t claimed);
199 static int fcip_data_cb(opaque_t ulp_handle, opaque_t phandle,
200     fc_unsol_buf_t *payload, uint32_t claimed);
201 
202 
203 /* Routing table specific */
204 static void fcip_handle_topology(struct fcip *fptr);
205 static int fcip_init_port(struct fcip *fptr);
206 struct fcip_routing_table *fcip_lookup_rtable(struct fcip *fptr,
207     la_wwn_t *pwwn, int matchflag);
208 static void fcip_rt_update(struct fcip *fptr, fc_portmap_t *devlist,
209     uint32_t listlen);
210 static void fcip_rt_flush(struct fcip *fptr);
211 static void fcip_rte_remove_deferred(void *arg);
212 static int fcip_do_plogi(struct fcip *fptr, struct fcip_routing_table *frp);
213 
214 
215 /* dest table specific */
216 static struct fcip_dest *fcip_get_dest(struct fcip *fptr,
217     la_wwn_t *dlphys);
218 static struct fcip_dest *fcip_add_dest(struct fcip *fptr,
219     struct fcip_routing_table *frp);
220 static int fcip_dest_add_broadcast_entry(struct fcip *fptr, int new_flag);
221 static uint32_t fcip_get_broadcast_did(struct fcip *fptr);
222 static void fcip_cleanup_dest(struct fcip *fptr);
223 
224 
225 /* helper functions */
226 static fcip_port_info_t *fcip_get_port(opaque_t phandle);
227 static int fcip_wwn_compare(la_wwn_t *wwn1, la_wwn_t *wwn2, int flag);
228 static void fcip_ether_to_str(struct ether_addr *e, caddr_t s);
229 static int fcip_port_get_num_pkts(struct fcip *fptr);
230 static int fcip_check_port_busy(struct fcip *fptr);
231 static void fcip_check_remove_minor_node(void);
232 static int fcip_set_wwn(la_wwn_t *pwwn);
233 static int fcip_plogi_in_progress(struct fcip *fptr);
234 static int fcip_check_port_exists(struct fcip *fptr);
235 static int fcip_is_supported_fc_topology(int fc_topology);
236 
237 
238 /* pkt specific */
239 static fcip_pkt_t *fcip_pkt_alloc(struct fcip *fptr, mblk_t *bp,
240     int flags, int datalen);
241 static void fcip_pkt_free(struct fcip_pkt *fcip_pkt, int flags);
242 static fcip_pkt_t *fcip_ipkt_alloc(struct fcip *fptr, int cmdlen,
243     int resplen, opaque_t pd, int flags);
244 static void fcip_ipkt_free(fcip_pkt_t *fcip_pkt);
245 static void fcip_ipkt_callback(fc_packet_t *fc_pkt);
246 static void fcip_free_pkt_dma(fcip_pkt_t *fcip_pkt);
247 static void fcip_pkt_callback(fc_packet_t *fc_pkt);
248 static void fcip_init_unicast_pkt(fcip_pkt_t *fcip_pkt, fc_portid_t sid,
249     fc_portid_t did, void (*comp) ());
250 static int fcip_transport(fcip_pkt_t *fcip_pkt);
251 static void fcip_pkt_timeout(void *arg);
252 static void fcip_timeout(void *arg);
253 static void fcip_fdestp_enqueue_pkt(struct fcip_dest *fdestp,
254     fcip_pkt_t *fcip_pkt);
255 static int fcip_fdestp_dequeue_pkt(struct fcip_dest *fdestp,
256     fcip_pkt_t *fcip_pkt);
257 static int fcip_sendup_constructor(void *buf, void *arg, int flags);
258 static void fcip_sendup_thr(void *arg);
259 static int fcip_sendup_alloc_enque(struct fcip *ftpr, mblk_t *mp,
260     struct fcipstr *(*f)());
261 
262 /*
263  * zero copy inbound data handling
264  */
265 #ifdef FCIP_ESBALLOC
266 static void fcip_ubfree(char *arg);
267 #endif /* FCIP_ESBALLOC */
268 
269 #if !defined(FCIP_ESBALLOC)
270 static void *fcip_allocb(size_t size, uint_t pri);
271 #endif
272 
273 
274 /* FCIP FARP support functions */
275 static struct fcip_dest *fcip_do_farp(struct fcip *fptr, la_wwn_t *pwwn,
276     char *ip_addr, size_t ip_addr_len, int flags);
277 static void fcip_init_broadcast_pkt(fcip_pkt_t *fcip_pkt, void (*comp) (),
278     int is_els);
279 static int fcip_handle_farp_request(struct fcip *fptr, la_els_farp_t *fcmd);
280 static int fcip_handle_farp_response(struct fcip *fptr, la_els_farp_t *fcmd);
281 static void fcip_cache_arp_broadcast(struct fcip *ftpr, fc_unsol_buf_t *buf);
282 static void fcip_port_ns(void *arg);
283 
284 #ifdef DEBUG
285 
286 #include <sys/debug.h>
287 
288 #define	FCIP_DEBUG_DEFAULT	0x1
289 #define	FCIP_DEBUG_ATTACH	0x2
290 #define	FCIP_DEBUG_INIT		0x4
291 #define	FCIP_DEBUG_DETACH	0x8
292 #define	FCIP_DEBUG_DLPI		0x10
293 #define	FCIP_DEBUG_ELS		0x20
294 #define	FCIP_DEBUG_DOWNSTREAM	0x40
295 #define	FCIP_DEBUG_UPSTREAM	0x80
296 #define	FCIP_DEBUG_MISC		0x100
297 
298 #define	FCIP_DEBUG_STARTUP	(FCIP_DEBUG_ATTACH|FCIP_DEBUG_INIT)
299 #define	FCIP_DEBUG_DATAOUT	(FCIP_DEBUG_DLPI|FCIP_DEBUG_DOWNSTREAM)
300 #define	FCIP_DEBUG_DATAIN	(FCIP_DEBUG_ELS|FCIP_DEBUG_UPSTREAM)
301 
302 static int fcip_debug = FCIP_DEBUG_DEFAULT;
303 
304 #define	FCIP_DEBUG(level, args)	\
305 	if (fcip_debug & (level))	cmn_err args;
306 
307 #else	/* DEBUG */
308 
309 #define	FCIP_DEBUG(level, args)		/* do nothing */
310 
311 #endif	/* DEBUG */
312 
313 #define	KIOIP	KSTAT_INTR_PTR(fcip->fcip_intrstats)
314 
315 /*
316  * Endian independent ethernet to WWN copy
317  */
318 #define	ether_to_wwn(E, W)	\
319 	bzero((void *)(W), sizeof (la_wwn_t)); \
320 	bcopy((void *)(E), (void *)&((W)->raw_wwn[2]), ETHERADDRL); \
321 	(W)->raw_wwn[0] |= 0x10
322 
323 /*
324  * wwn_to_ether : Endian independent, copies a WWN to struct ether_addr.
325  * The args to the macro are pointers to WWN and ether_addr structures
326  */
327 #define	wwn_to_ether(W, E)	\
328 	bcopy((void *)&((W)->raw_wwn[2]), (void *)E, ETHERADDRL)
329 
330 /*
331  * The module_info structure contains identification and limit values.
332  * All queues associated with a certain driver share the same module_info
333  * structures. This structure defines the characteristics of that driver/
334  * module's queues. The module name must be unique. The max and min packet
335  * sizes limit the no. of characters in M_DATA messages. The Hi and Lo
336  * water marks are for flow control when a module has a service procedure.
337  */
338 static struct module_info	fcipminfo = {
339 	FCIPIDNUM,	/* mi_idnum : Module ID num */
340 	FCIPNAME, 	/* mi_idname: Module Name */
341 	FCIPMINPSZ,	/* mi_minpsz: Min packet size */
342 	FCIPMAXPSZ,	/* mi_maxpsz: Max packet size */
343 	FCIPHIWAT,	/* mi_hiwat : High water mark */
344 	FCIPLOWAT	/* mi_lowat : Low water mark */
345 };
346 
347 /*
348  * The qinit structres contain the module put, service. open and close
349  * procedure pointers. All modules and drivers with the same streamtab
350  * file (i.e same fmodsw or cdevsw entry points) point to the same
351  * upstream (read) and downstream (write) qinit structs.
352  */
353 static struct qinit	fcip_rinit = {
354 	NULL,		/* qi_putp */
355 	NULL,		/* qi_srvp */
356 	fcip_open,	/* qi_qopen */
357 	fcip_close,	/* qi_qclose */
358 	NULL,		/* qi_qadmin */
359 	&fcipminfo,	/* qi_minfo */
360 	NULL		/* qi_mstat */
361 };
362 
363 static struct qinit	fcip_winit = {
364 	fcip_wput,	/* qi_putp */
365 	fcip_wsrv,	/* qi_srvp */
366 	NULL,		/* qi_qopen */
367 	NULL,		/* qi_qclose */
368 	NULL,		/* qi_qadmin */
369 	&fcipminfo,	/* qi_minfo */
370 	NULL		/* qi_mstat */
371 };
372 
373 /*
374  * streamtab contains pointers to the read and write qinit structures
375  */
376 
377 static struct streamtab fcip_info = {
378 	&fcip_rinit,	/* st_rdinit */
379 	&fcip_winit,	/* st_wrinit */
380 	NULL,		/* st_muxrinit */
381 	NULL,		/* st_muxwrinit */
382 };
383 
384 static struct cb_ops  fcip_cb_ops = {
385 	nodev,				/* open */
386 	nodev,				/* close */
387 	nodev,				/* strategy */
388 	nodev,				/* print */
389 	nodev,				/* dump */
390 	nodev,				/* read */
391 	nodev,				/* write */
392 	nodev,				/* ioctl */
393 	nodev,				/* devmap */
394 	nodev,				/* mmap */
395 	nodev,				/* segmap */
396 	nochpoll,			/* poll */
397 	ddi_prop_op,			/* cb_prop_op */
398 	&fcip_info,			/* streamtab  */
399 	D_MP | D_HOTPLUG,		/* Driver compatibility flag */
400 	CB_REV,				/* rev */
401 	nodev,				/* int (*cb_aread)() */
402 	nodev				/* int (*cb_awrite)() */
403 };
404 
405 /*
406  * autoconfiguration routines.
407  */
408 static struct dev_ops fcip_ops = {
409 	DEVO_REV,		/* devo_rev, */
410 	0,			/* refcnt  */
411 	fcip_getinfo,		/* info */
412 	nulldev,		/* identify */
413 	nulldev,		/* probe */
414 	fcip_attach,		/* attach */
415 	fcip_detach,		/* detach */
416 	nodev,			/* RESET */
417 	&fcip_cb_ops,		/* driver operations */
418 	NULL,			/* bus operations */
419 	ddi_power		/* power management */
420 };
421 
422 #define	FCIP_VERSION	"1.61"
423 #define	FCIP_NAME	"SunFC FCIP v" FCIP_VERSION
424 
425 #define	PORT_DRIVER	"fp"
426 
427 #define	GETSTRUCT(struct, number)	\
428 	((struct *)kmem_zalloc((size_t)(sizeof (struct) * (number)), \
429 		KM_SLEEP))
430 
431 static struct modldrv modldrv = {
432 	&mod_driverops,			/* Type of module - driver */
433 	FCIP_NAME,			/* Name of module */
434 	&fcip_ops,			/* driver ops */
435 };
436 
437 static struct modlinkage modlinkage = {
438 	MODREV_1, (void *)&modldrv, NULL
439 };
440 
441 
442 /*
443  * Now for some global statics
444  */
445 static uint32_t	fcip_ub_nbufs = FCIP_UB_NBUFS;
446 static uint32_t fcip_ub_size = FCIP_UB_SIZE;
447 static int fcip_pkt_ttl_ticks = FCIP_PKT_TTL;
448 static int fcip_tick_incr = 1;
449 static int fcip_wait_cmds = FCIP_WAIT_CMDS;
450 static int fcip_num_attaching = 0;
451 static int fcip_port_attach_pending = 0;
452 static int fcip_create_nodes_on_demand = 1;	/* keep it similar to fcp */
453 static int fcip_cache_on_arp_broadcast = 0;
454 static int fcip_farp_supported = 0;
455 static int fcip_minor_node_created = 0;
456 
457 /*
458  * Supported FCAs
459  */
460 #define	QLC_PORT_1_ID_BITS		0x100
461 #define	QLC_PORT_2_ID_BITS		0x101
462 #define	QLC_PORT_NAA			0x2
463 #define	QLC_MODULE_NAME			"qlc"
464 #define	IS_QLC_PORT(port_dip)		\
465 			(strcmp(ddi_driver_name(ddi_get_parent((port_dip))),\
466 			QLC_MODULE_NAME) == 0)
467 
468 
469 /*
470  * fcip softstate structures head.
471  */
472 
473 static void *fcip_softp = NULL;
474 
475 /*
476  * linked list of active (inuse) driver streams
477  */
478 
479 static int fcip_num_instances = 0;
480 static dev_info_t *fcip_module_dip = (dev_info_t *)0;
481 
482 
483 /*
484  * Ethernet broadcast address: Broadcast addressing in IP over fibre
485  * channel should be the IEEE ULA (also the low 6 bytes of the Port WWN).
486  *
487  * The broadcast addressing varies for differing topologies a node may be in:
488  *	- On a private loop the ARP broadcast is a class 3 sequence sent
489  *	  using OPNfr (Open Broadcast Replicate primitive) followed by
490  *	  the ARP frame to D_ID 0xFFFFFF
491  *
492  *	- On a public Loop the broadcast sequence is sent to AL_PA 0x00
493  *	  (no OPNfr primitive).
494  *
495  *	- For direct attach and point to point topologies we just send
496  *	  the frame to D_ID 0xFFFFFF
497  *
498  * For public loop the handling would probably be different - for now
499  * I'll just declare this struct - It can be deleted if not necessary.
500  *
501  */
502 
503 
504 /*
505  * DL_INFO_ACK template for the fcip module. The dl_info_ack_t structure is
506  * returned as a part of an  DL_INFO_ACK message which is a M_PCPROTO message
507  * returned in response to a DL_INFO_REQ message sent to us from a DLS user
508  * Let us fake an ether header as much as possible.
509  *
510  * dl_addr_length is the Provider's DLSAP addr which is SAP addr +
511  *                Physical addr of the provider. We set this to
512  *                ushort_t + sizeof (la_wwn_t) for Fibre Channel ports.
513  * dl_mac_type    Lets just use DL_ETHER - we can try using DL_IPFC, a new
514  *		  dlpi.h define later.
515  * dl_sap_length  -2 indicating the SAP address follows the Physical addr
516  *		  component in the DLSAP addr.
517  * dl_service_mode: DLCLDS - connectionless data link service.
518  *
519  */
520 
521 static dl_info_ack_t fcip_infoack = {
522 	DL_INFO_ACK,				/* dl_primitive */
523 	FCIPMTU,				/* dl_max_sdu */
524 	0,					/* dl_min_sdu */
525 	FCIPADDRL,				/* dl_addr_length */
526 	DL_ETHER,				/* dl_mac_type */
527 	0,					/* dl_reserved */
528 	0,					/* dl_current_state */
529 	-2,					/* dl_sap_length */
530 	DL_CLDLS,				/* dl_service_mode */
531 	0,					/* dl_qos_length */
532 	0,					/* dl_qos_offset */
533 	0,					/* dl_range_length */
534 	0,					/* dl_range_offset */
535 	DL_STYLE2,				/* dl_provider_style */
536 	sizeof (dl_info_ack_t),			/* dl_addr_offset */
537 	DL_VERSION_2,				/* dl_version */
538 	ETHERADDRL,				/* dl_brdcst_addr_length */
539 	sizeof (dl_info_ack_t) + FCIPADDRL,	/* dl_brdcst_addr_offset */
540 	0					/* dl_growth */
541 };
542 
543 /*
544  * FCIP broadcast address definition.
545  */
546 static	struct ether_addr	fcipnhbroadcastaddr = {
547 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
548 };
549 
550 /*
551  * RFC2625 requires the broadcast ARP address in the ARP data payload to
552  * be set to 0x00 00 00 00 00 00 for ARP broadcast packets
553  */
554 static	struct ether_addr	fcip_arpbroadcast_addr = {
555 	0x00, 0x00, 0x00, 0x00, 0x00, 0x00
556 };
557 
558 
559 #define	ether_bcopy(src, dest)	bcopy((src), (dest), ETHERADDRL);
560 
561 /*
562  * global kernel locks
563  */
564 static kcondvar_t	fcip_global_cv;
565 static kmutex_t		fcip_global_mutex;
566 
567 /*
568  * fctl external defines
569  */
570 extern int fc_ulp_add(fc_ulp_modinfo_t *);
571 
572 /*
573  * fctl data structures
574  */
575 
576 #define	FCIP_REV	0x07
577 
578 /* linked list of port info structures */
579 static fcip_port_info_t *fcip_port_head = NULL;
580 
581 /* linked list of fcip structures */
582 static struct fcipstr	*fcipstrup = NULL;
583 static krwlock_t	fcipstruplock;
584 
585 
586 /*
587  * Module information structure. This structure gives the FC Transport modules
588  * information about an ULP that registers with it.
589  */
590 static fc_ulp_modinfo_t	fcip_modinfo = {
591 	0,			/* for xref checks? */
592 	FCTL_ULP_MODREV_4,	/* FCIP revision */
593 	FC_TYPE_IS8802_SNAP,	/* type 5 for SNAP encapsulated datagrams */
594 	FCIP_NAME,		/* module name as in the modldrv struct */
595 	0x0,			/* get all statec callbacks for now */
596 	fcip_port_attach,	/* port attach callback */
597 	fcip_port_detach,	/* port detach callback */
598 	fcip_port_ioctl,	/* port ioctl callback */
599 	fcip_els_cb,		/* els callback */
600 	fcip_data_cb,		/* data callback */
601 	fcip_statec_cb		/* state change callback */
602 };
603 
604 
605 /*
606  * Solaris 9 and up, the /kernel/drv/fp.conf file will have the following entry
607  *
608  * ddi-forceattach=1;
609  *
610  * This will ensure that fp is loaded at bootup. No additional checks are needed
611  */
612 int
_init(void)613 _init(void)
614 {
615 	int	rval;
616 
617 	FCIP_TNF_LOAD();
618 
619 	/*
620 	 * Initialize the mutexs used by port attach and other callbacks.
621 	 * The transport can call back into our port_attach_callback
622 	 * routine even before _init() completes and bad things can happen.
623 	 */
624 	mutex_init(&fcip_global_mutex, NULL, MUTEX_DRIVER, NULL);
625 	cv_init(&fcip_global_cv, NULL, CV_DRIVER, NULL);
626 	rw_init(&fcipstruplock, NULL, RW_DRIVER, NULL);
627 
628 	mutex_enter(&fcip_global_mutex);
629 	fcip_port_attach_pending = 1;
630 	mutex_exit(&fcip_global_mutex);
631 
632 	/*
633 	 * Now attempt to register fcip with the transport.
634 	 * If fc_ulp_add fails, fcip module will not be loaded.
635 	 */
636 	rval = fc_ulp_add(&fcip_modinfo);
637 	if (rval != FC_SUCCESS) {
638 		mutex_destroy(&fcip_global_mutex);
639 		cv_destroy(&fcip_global_cv);
640 		rw_destroy(&fcipstruplock);
641 		switch (rval) {
642 		case FC_ULP_SAMEMODULE:
643 			FCIP_DEBUG(FCIP_DEBUG_DEFAULT, (CE_WARN,
644 			    "!fcip: module is already registered with"
645 			    " transport"));
646 			rval = EEXIST;
647 			break;
648 		case FC_ULP_SAMETYPE:
649 			FCIP_DEBUG(FCIP_DEBUG_DEFAULT, (CE_WARN,
650 			    "!fcip: Another module of the same ULP type 0x%x"
651 			    " is already registered with the transport",
652 			    fcip_modinfo.ulp_type));
653 			rval = EEXIST;
654 			break;
655 		case FC_BADULP:
656 			FCIP_DEBUG(FCIP_DEBUG_DEFAULT, (CE_WARN,
657 			    "!fcip: Current fcip version 0x%x does not match"
658 			    " fctl version",
659 			    fcip_modinfo.ulp_rev));
660 			rval = ENODEV;
661 			break;
662 		default:
663 			FCIP_DEBUG(FCIP_DEBUG_DEFAULT, (CE_WARN,
664 			    "!fcip: fc_ulp_add failed with status 0x%x", rval));
665 			rval = ENODEV;
666 			break;
667 		}
668 		FCIP_TNF_UNLOAD(&modlinkage);
669 		return (rval);
670 	}
671 
672 	if ((rval = ddi_soft_state_init(&fcip_softp, sizeof (struct fcip),
673 			FCIP_NUM_INSTANCES)) != 0) {
674 		mutex_destroy(&fcip_global_mutex);
675 		cv_destroy(&fcip_global_cv);
676 		rw_destroy(&fcipstruplock);
677 		(void) fc_ulp_remove(&fcip_modinfo);
678 		FCIP_TNF_UNLOAD(&modlinkage);
679 		return (rval);
680 	}
681 
682 	if ((rval = mod_install(&modlinkage)) != 0) {
683 		FCIP_TNF_UNLOAD(&modlinkage);
684 		(void) fc_ulp_remove(&fcip_modinfo);
685 		mutex_destroy(&fcip_global_mutex);
686 		cv_destroy(&fcip_global_cv);
687 		rw_destroy(&fcipstruplock);
688 		ddi_soft_state_fini(&fcip_softp);
689 	}
690 	return (rval);
691 }
692 
693 /*
694  * Unload the port driver if this was the only ULP loaded and then
695  * deregister with the transport.
696  */
697 int
_fini(void)698 _fini(void)
699 {
700 	int	rval;
701 	int	rval1;
702 
703 	/*
704 	 * Do not permit the module to be unloaded before a port
705 	 * attach callback has happened.
706 	 */
707 	mutex_enter(&fcip_global_mutex);
708 	if (fcip_num_attaching || fcip_port_attach_pending) {
709 		mutex_exit(&fcip_global_mutex);
710 		return (EBUSY);
711 	}
712 	mutex_exit(&fcip_global_mutex);
713 
714 	if ((rval = mod_remove(&modlinkage)) != 0) {
715 		return (rval);
716 	}
717 
718 	/*
719 	 * unregister with the transport layer
720 	 */
721 	rval1 = fc_ulp_remove(&fcip_modinfo);
722 
723 	/*
724 	 * If the ULP was not registered with the transport, init should
725 	 * have failed. If transport has no knowledge of our existence
726 	 * we should simply bail out and succeed
727 	 */
728 #ifdef DEBUG
729 	if (rval1 == FC_BADULP) {
730 		FCIP_DEBUG(FCIP_DEBUG_DEFAULT, (CE_WARN,
731 		"fcip: ULP was never registered with the transport"));
732 		rval = ENODEV;
733 	} else if (rval1 == FC_BADTYPE) {
734 		FCIP_DEBUG(FCIP_DEBUG_DEFAULT, (CE_WARN,
735 			"fcip: No ULP of this type 0x%x was registered with "
736 			"transport", fcip_modinfo.ulp_type));
737 		rval = ENODEV;
738 	}
739 #endif /* DEBUG */
740 
741 	mutex_destroy(&fcip_global_mutex);
742 	rw_destroy(&fcipstruplock);
743 	cv_destroy(&fcip_global_cv);
744 	ddi_soft_state_fini(&fcip_softp);
745 
746 	FCIP_TNF_UNLOAD(&modlinkage);
747 
748 	return (rval);
749 }
750 
751 /*
752  * Info about this loadable module
753  */
754 int
_info(struct modinfo * modinfop)755 _info(struct modinfo *modinfop)
756 {
757 	return (mod_info(&modlinkage, modinfop));
758 }
759 
760 /*
761  * The port attach callback is invoked by the port driver when a FCA
762  * port comes online and binds with the transport layer. The transport
763  * then callsback into all ULP modules registered with it. The Port attach
764  * call back will also provide the ULP module with the Port's WWN and S_ID
765  */
766 /* ARGSUSED */
767 static int
fcip_port_attach(opaque_t ulp_handle,fc_ulp_port_info_t * port_info,fc_attach_cmd_t cmd,uint32_t sid)768 fcip_port_attach(opaque_t ulp_handle, fc_ulp_port_info_t *port_info,
769     fc_attach_cmd_t cmd, uint32_t sid)
770 {
771 	int 			rval = FC_FAILURE;
772 	int 			instance;
773 	struct fcip		*fptr;
774 	fcip_port_info_t	*fport = NULL;
775 	fcip_port_info_t	*cur_fport;
776 	fc_portid_t		src_id;
777 
778 	switch (cmd) {
779 	case FC_CMD_ATTACH: {
780 		la_wwn_t	*ww_pn = NULL;
781 		/*
782 		 * It was determined that, as per spec, the lower 48 bits of
783 		 * the port-WWN will always be unique. This will make the MAC
784 		 * address (i.e the lower 48 bits of the WWN), that IP/ARP
785 		 * depend on, unique too. Hence we should be able to remove the
786 		 * restriction of attaching to only one of the ports of
787 		 * multi port FCAs.
788 		 *
789 		 * Earlier, fcip used to attach only to qlc module and fail
790 		 * silently for attach failures resulting from unknown FCAs or
791 		 * unsupported FCA ports. Now, we'll do no such checks.
792 		 */
793 		ww_pn = &port_info->port_pwwn;
794 
795 		FCIP_TNF_PROBE_2((fcip_port_attach, "fcip io", /* CSTYLED */,
796 			tnf_string, msg, "port id bits",
797 			tnf_opaque, nport_id, ww_pn->w.nport_id));
798 		FCIP_DEBUG(FCIP_DEBUG_ATTACH, (CE_NOTE,
799 		    "port id bits: 0x%x", ww_pn->w.nport_id));
800 		/*
801 		 * A port has come online
802 		 */
803 		mutex_enter(&fcip_global_mutex);
804 		fcip_num_instances++;
805 		fcip_num_attaching++;
806 
807 		if (fcip_port_head == NULL) {
808 			/* OK to sleep here ? */
809 			fport = kmem_zalloc(sizeof (fcip_port_info_t),
810 						KM_NOSLEEP);
811 			if (fport == NULL) {
812 				fcip_num_instances--;
813 				fcip_num_attaching--;
814 				ASSERT(fcip_num_attaching >= 0);
815 				mutex_exit(&fcip_global_mutex);
816 				rval = FC_FAILURE;
817 				cmn_err(CE_WARN, "!fcip(%d): port attach "
818 				    "failed: alloc failed",
819 				    ddi_get_instance(port_info->port_dip));
820 				goto done;
821 			}
822 			fcip_port_head = fport;
823 		} else {
824 			/*
825 			 * traverse the port list and also check for
826 			 * duplicate port attaches - Nothing wrong in being
827 			 * paranoid Heh Heh.
828 			 */
829 			cur_fport = fcip_port_head;
830 			while (cur_fport != NULL) {
831 				if (cur_fport->fcipp_handle ==
832 				    port_info->port_handle) {
833 					fcip_num_instances--;
834 					fcip_num_attaching--;
835 					ASSERT(fcip_num_attaching >= 0);
836 					mutex_exit(&fcip_global_mutex);
837 					FCIP_DEBUG(FCIP_DEBUG_ATTACH, (CE_WARN,
838 					    "!fcip(%d): port already "
839 					    "attached!!", ddi_get_instance(
840 					    port_info->port_dip)));
841 					rval = FC_FAILURE;
842 					goto done;
843 				}
844 				cur_fport = cur_fport->fcipp_next;
845 			}
846 			fport = kmem_zalloc(sizeof (fcip_port_info_t),
847 						KM_NOSLEEP);
848 			if (fport == NULL) {
849 				rval = FC_FAILURE;
850 				fcip_num_instances--;
851 				fcip_num_attaching--;
852 				ASSERT(fcip_num_attaching >= 0);
853 				mutex_exit(&fcip_global_mutex);
854 				cmn_err(CE_WARN, "!fcip(%d): port attach "
855 				    "failed: alloc failed",
856 				    ddi_get_instance(port_info->port_dip));
857 				goto done;
858 			}
859 			fport->fcipp_next = fcip_port_head;
860 			fcip_port_head = fport;
861 		}
862 
863 		mutex_exit(&fcip_global_mutex);
864 
865 		/*
866 		 * now fill in the details about the port itself
867 		 */
868 		fport->fcipp_linkage = *port_info->port_linkage;
869 		fport->fcipp_handle = port_info->port_handle;
870 		fport->fcipp_dip = port_info->port_dip;
871 		fport->fcipp_topology = port_info->port_flags;
872 		fport->fcipp_pstate = port_info->port_state;
873 		fport->fcipp_naa = port_info->port_pwwn.w.naa_id;
874 		bcopy(&port_info->port_pwwn, &fport->fcipp_pwwn,
875 		    sizeof (la_wwn_t));
876 		bcopy(&port_info->port_nwwn, &fport->fcipp_nwwn,
877 		    sizeof (la_wwn_t));
878 		fport->fcipp_fca_pkt_size = port_info->port_fca_pkt_size;
879 		fport->fcipp_cmd_dma_attr = *port_info->port_cmd_dma_attr;
880 		fport->fcipp_resp_dma_attr = *port_info->port_resp_dma_attr;
881 		fport->fcipp_fca_acc_attr = *port_info->port_acc_attr;
882 		src_id.port_id = sid;
883 		src_id.priv_lilp_posit = 0;
884 		fport->fcipp_sid = src_id;
885 
886 		/*
887 		 * allocate soft state for this instance
888 		 */
889 		instance = ddi_get_instance(fport->fcipp_dip);
890 		if (ddi_soft_state_zalloc(fcip_softp,
891 		    instance) != DDI_SUCCESS) {
892 			rval = FC_FAILURE;
893 			cmn_err(CE_WARN, "!fcip(%d): port attach failed: "
894 			    "soft state alloc failed", instance);
895 			goto failure;
896 		}
897 
898 		fptr = ddi_get_soft_state(fcip_softp, instance);
899 
900 		if (fptr == NULL) {
901 			rval = FC_FAILURE;
902 			cmn_err(CE_WARN, "!fcip(%d): port attach failed: "
903 			    "failure to get soft state", instance);
904 			goto failure;
905 		}
906 
907 		/*
908 		 * initialize all mutexes and locks required for this module
909 		 */
910 		mutex_init(&fptr->fcip_mutex, NULL, MUTEX_DRIVER, NULL);
911 		mutex_init(&fptr->fcip_ub_mutex, NULL, MUTEX_DRIVER, NULL);
912 		mutex_init(&fptr->fcip_rt_mutex, NULL, MUTEX_DRIVER, NULL);
913 		mutex_init(&fptr->fcip_dest_mutex, NULL, MUTEX_DRIVER, NULL);
914 		mutex_init(&fptr->fcip_sendup_mutex, NULL, MUTEX_DRIVER, NULL);
915 		cv_init(&fptr->fcip_farp_cv, NULL, CV_DRIVER, NULL);
916 		cv_init(&fptr->fcip_sendup_cv, NULL, CV_DRIVER, NULL);
917 		cv_init(&fptr->fcip_ub_cv, NULL, CV_DRIVER, NULL);
918 
919 		mutex_enter(&fptr->fcip_mutex);
920 
921 		fptr->fcip_dip = fport->fcipp_dip;	/* parent's dip */
922 		fptr->fcip_instance = instance;
923 		fptr->fcip_ub_upstream = 0;
924 
925 		if (FC_PORT_STATE_MASK(port_info->port_state) ==
926 		    FC_STATE_ONLINE) {
927 			fptr->fcip_port_state = FCIP_PORT_ONLINE;
928 			if (fptr->fcip_flags & FCIP_LINK_DOWN) {
929 				fptr->fcip_flags &= ~FCIP_LINK_DOWN;
930 			}
931 		} else {
932 			fptr->fcip_port_state = FCIP_PORT_OFFLINE;
933 		}
934 
935 		fptr->fcip_flags |= FCIP_ATTACHING;
936 		fptr->fcip_port_info = fport;
937 
938 		/*
939 		 * Extract our MAC addr from our port's WWN. The lower 48
940 		 * bits will be our MAC address
941 		 */
942 		wwn_to_ether(&fport->fcipp_nwwn, &fptr->fcip_macaddr);
943 
944 		fport->fcipp_fcip = fptr;
945 
946 		FCIP_DEBUG(FCIP_DEBUG_ATTACH,
947 		    (CE_NOTE, "fcipdest : 0x%lx, rtable : 0x%lx",
948 		    (long)(sizeof (fptr->fcip_dest)),
949 		    (long)(sizeof (fptr->fcip_rtable))));
950 
951 		bzero(fptr->fcip_dest, sizeof (fptr->fcip_dest));
952 		bzero(fptr->fcip_rtable, sizeof (fptr->fcip_rtable));
953 
954 		/*
955 		 * create a taskq to handle sundry jobs for the driver
956 		 * This way we can have jobs run in parallel
957 		 */
958 		fptr->fcip_tq = taskq_create("fcip_tasks",
959 		    FCIP_NUM_THREADS, MINCLSYSPRI, FCIP_MIN_TASKS,
960 		    FCIP_MAX_TASKS, TASKQ_PREPOPULATE);
961 
962 		mutex_exit(&fptr->fcip_mutex);
963 
964 		/*
965 		 * create a separate thread to handle all unsolicited
966 		 * callback handling. This is because unsolicited_callback
967 		 * can happen from an interrupt context and the upstream
968 		 * modules can put new messages right back in the same
969 		 * thread context. This usually works fine, but sometimes
970 		 * we may have to block to obtain the dest struct entries
971 		 * for some remote ports.
972 		 */
973 		mutex_enter(&fptr->fcip_sendup_mutex);
974 		if (thread_create(NULL, DEFAULTSTKSZ,
975 		    (void (*)())fcip_sendup_thr, (caddr_t)fptr, 0, &p0,
976 		    TS_RUN, minclsyspri) == NULL) {
977 			mutex_exit(&fptr->fcip_sendup_mutex);
978 			cmn_err(CE_WARN,
979 			    "!unable to create fcip sendup thread for "
980 			    " instance: 0x%x", instance);
981 			rval = FC_FAILURE;
982 			goto done;
983 		}
984 		fptr->fcip_sendup_thr_initted = 1;
985 		fptr->fcip_sendup_head = fptr->fcip_sendup_tail = NULL;
986 		mutex_exit(&fptr->fcip_sendup_mutex);
987 
988 
989 		/* Let the attach handler do the rest */
990 		if (fcip_port_attach_handler(fptr) != FC_SUCCESS) {
991 			/*
992 			 * We have already cleaned up so return
993 			 */
994 			rval = FC_FAILURE;
995 			cmn_err(CE_WARN, "!fcip(%d): port attach failed",
996 			    instance);
997 			goto done;
998 		}
999 
1000 		FCIP_DEBUG(FCIP_DEBUG_ATTACH, (CE_CONT,
1001 		    "!fcip attach for port instance (0x%x) successful",
1002 		    instance));
1003 
1004 		rval = FC_SUCCESS;
1005 		goto done;
1006 	}
1007 	case FC_CMD_POWER_UP:
1008 	/* FALLTHROUGH */
1009 	case FC_CMD_RESUME:
1010 		mutex_enter(&fcip_global_mutex);
1011 		fport = fcip_port_head;
1012 		while (fport != NULL) {
1013 			if (fport->fcipp_handle == port_info->port_handle) {
1014 				break;
1015 			}
1016 			fport = fport->fcipp_next;
1017 		}
1018 		if (fport == NULL) {
1019 			rval = FC_SUCCESS;
1020 			mutex_exit(&fcip_global_mutex);
1021 			goto done;
1022 		}
1023 		rval = fcip_handle_resume(fport, port_info, cmd);
1024 		mutex_exit(&fcip_global_mutex);
1025 		goto done;
1026 
1027 	default:
1028 		FCIP_TNF_PROBE_2((fcip_port_attach, "fcip io", /* CSTYLED */,
1029 			tnf_string, msg, "unknown command type",
1030 			tnf_uint, cmd, cmd));
1031 		FCIP_DEBUG(FCIP_DEBUG_ATTACH, (CE_WARN,
1032 		    "unknown cmd type 0x%x in port_attach", cmd));
1033 		rval = FC_FAILURE;
1034 		goto done;
1035 	}
1036 
1037 failure:
1038 	if (fport) {
1039 		mutex_enter(&fcip_global_mutex);
1040 		fcip_num_attaching--;
1041 		ASSERT(fcip_num_attaching >= 0);
1042 		(void) fcip_softstate_free(fport);
1043 		fcip_port_attach_pending = 0;
1044 		mutex_exit(&fcip_global_mutex);
1045 	}
1046 	return (rval);
1047 
1048 done:
1049 	mutex_enter(&fcip_global_mutex);
1050 	fcip_port_attach_pending = 0;
1051 	mutex_exit(&fcip_global_mutex);
1052 	return (rval);
1053 }
1054 
1055 /*
1056  * fcip_port_attach_handler : Completes the port attach operation after
1057  * the ulp_port_attach routine has completed its ground work. The job
1058  * of this function among other things is to obtain and handle topology
1059  * specifics, initialize a port, setup broadcast address entries in
1060  * the fcip tables etc. This routine cleans up behind itself on failures.
1061  * Returns FC_SUCCESS or FC_FAILURE.
1062  */
1063 static int
fcip_port_attach_handler(struct fcip * fptr)1064 fcip_port_attach_handler(struct fcip *fptr)
1065 {
1066 	fcip_port_info_t		*fport = fptr->fcip_port_info;
1067 	int				rval = FC_FAILURE;
1068 
1069 	ASSERT(fport != NULL);
1070 
1071 	mutex_enter(&fcip_global_mutex);
1072 
1073 	FCIP_DEBUG(FCIP_DEBUG_ATTACH, (CE_NOTE,
1074 	    "fcip module dip: %p instance: %d",
1075 	    (void *)fcip_module_dip, ddi_get_instance(fptr->fcip_dip)));
1076 
1077 	if (fcip_module_dip == NULL) {
1078 		clock_t		fcip_lbolt;
1079 
1080 		fcip_lbolt = ddi_get_lbolt();
1081 		/*
1082 		 * we need to use the fcip devinfo for creating
1083 		 * the clone device node, but the fcip attach
1084 		 * (from its conf file entry claiming to be a
1085 		 * child of pseudo) may not have happened yet.
1086 		 * wait here for 10 seconds and fail port attach
1087 		 * if the fcip devinfo is not attached yet
1088 		 */
1089 		fcip_lbolt += drv_usectohz(FCIP_INIT_DELAY);
1090 
1091 		FCIP_DEBUG(FCIP_DEBUG_ATTACH,
1092 		    (CE_WARN, "cv_timedwait lbolt %lx", fcip_lbolt));
1093 
1094 		(void) cv_timedwait(&fcip_global_cv, &fcip_global_mutex,
1095 		    fcip_lbolt);
1096 
1097 		if (fcip_module_dip == NULL) {
1098 			mutex_exit(&fcip_global_mutex);
1099 
1100 			FCIP_DEBUG(FCIP_DEBUG_ATTACH, (CE_WARN,
1101 				"fcip attach did not happen"));
1102 			goto port_attach_cleanup;
1103 		}
1104 	}
1105 
1106 	if ((!fcip_minor_node_created) &&
1107 	    fcip_is_supported_fc_topology(fport->fcipp_topology)) {
1108 		/*
1109 		 * Checking for same topologies which are considered valid
1110 		 * by fcip_handle_topology(). Dont create a minor node if
1111 		 * nothing is hanging off the FC port.
1112 		 */
1113 		if (ddi_create_minor_node(fcip_module_dip, "fcip", S_IFCHR,
1114 		    ddi_get_instance(fptr->fcip_dip), DDI_PSEUDO,
1115 		    CLONE_DEV) == DDI_FAILURE) {
1116 			mutex_exit(&fcip_global_mutex);
1117 			FCIP_DEBUG(FCIP_DEBUG_ATTACH, (CE_WARN,
1118 			    "failed to create minor node for fcip(%d)",
1119 			    ddi_get_instance(fptr->fcip_dip)));
1120 			goto port_attach_cleanup;
1121 		}
1122 		fcip_minor_node_created++;
1123 	}
1124 	mutex_exit(&fcip_global_mutex);
1125 
1126 	/*
1127 	 * initialize port for traffic
1128 	 */
1129 	if (fcip_init_port(fptr) != FC_SUCCESS) {
1130 		/* fcip_init_port has already cleaned up its stuff */
1131 
1132 		mutex_enter(&fcip_global_mutex);
1133 
1134 		if ((fcip_num_instances == 1) &&
1135 		    (fcip_minor_node_created == 1)) {
1136 			/* Remove minor node iff this is the last instance */
1137 			ddi_remove_minor_node(fcip_module_dip, NULL);
1138 		}
1139 
1140 		mutex_exit(&fcip_global_mutex);
1141 
1142 		goto port_attach_cleanup;
1143 	}
1144 
1145 	mutex_enter(&fptr->fcip_mutex);
1146 	fptr->fcip_flags &= ~FCIP_ATTACHING;
1147 	fptr->fcip_flags |= FCIP_INITED;
1148 	fptr->fcip_timeout_ticks = 0;
1149 
1150 	/*
1151 	 * start the timeout threads
1152 	 */
1153 	fptr->fcip_timeout_id = timeout(fcip_timeout, fptr,
1154 	    drv_usectohz(1000000));
1155 
1156 	mutex_exit(&fptr->fcip_mutex);
1157 	mutex_enter(&fcip_global_mutex);
1158 	fcip_num_attaching--;
1159 	ASSERT(fcip_num_attaching >= 0);
1160 	mutex_exit(&fcip_global_mutex);
1161 	rval = FC_SUCCESS;
1162 	return (rval);
1163 
1164 port_attach_cleanup:
1165 	mutex_enter(&fcip_global_mutex);
1166 	(void) fcip_softstate_free(fport);
1167 	fcip_num_attaching--;
1168 	ASSERT(fcip_num_attaching >= 0);
1169 	mutex_exit(&fcip_global_mutex);
1170 	rval = FC_FAILURE;
1171 	return (rval);
1172 }
1173 
1174 
1175 /*
1176  * Handler for DDI_RESUME operations. Port must be ready to restart IP
1177  * traffic on resume
1178  */
1179 static int
fcip_handle_resume(fcip_port_info_t * fport,fc_ulp_port_info_t * port_info,fc_attach_cmd_t cmd)1180 fcip_handle_resume(fcip_port_info_t *fport, fc_ulp_port_info_t *port_info,
1181     fc_attach_cmd_t cmd)
1182 {
1183 	int 		rval = FC_SUCCESS;
1184 	struct fcip	*fptr = fport->fcipp_fcip;
1185 	struct fcipstr	*tslp;
1186 	int		index;
1187 
1188 
1189 	ASSERT(fptr != NULL);
1190 
1191 	mutex_enter(&fptr->fcip_mutex);
1192 
1193 	if (cmd == FC_CMD_POWER_UP) {
1194 		fptr->fcip_flags &= ~(FCIP_POWER_DOWN);
1195 		if (fptr->fcip_flags & FCIP_SUSPENDED) {
1196 			mutex_exit(&fptr->fcip_mutex);
1197 			return (FC_SUCCESS);
1198 		}
1199 	} else if (cmd == FC_CMD_RESUME) {
1200 		fptr->fcip_flags &= ~(FCIP_SUSPENDED);
1201 	} else {
1202 		mutex_exit(&fptr->fcip_mutex);
1203 		return (FC_FAILURE);
1204 	}
1205 
1206 	/*
1207 	 * set the current port state and topology
1208 	 */
1209 	fport->fcipp_topology = port_info->port_flags;
1210 	fport->fcipp_pstate = port_info->port_state;
1211 
1212 	rw_enter(&fcipstruplock, RW_READER);
1213 	for (tslp = fcipstrup; tslp; tslp = tslp->sl_nextp) {
1214 		if (tslp->sl_fcip == fptr) {
1215 			break;
1216 		}
1217 	}
1218 	rw_exit(&fcipstruplock);
1219 
1220 	/*
1221 	 * No active streams on this port
1222 	 */
1223 	if (tslp == NULL) {
1224 		rval = FC_SUCCESS;
1225 		goto done;
1226 	}
1227 
1228 	mutex_enter(&fptr->fcip_rt_mutex);
1229 	for (index = 0; index < FCIP_RT_HASH_ELEMS; index++) {
1230 		struct fcip_routing_table 	*frp;
1231 
1232 		frp = fptr->fcip_rtable[index];
1233 		while (frp) {
1234 			uint32_t		did;
1235 			/*
1236 			 * Mark the broadcast RTE available again. It
1237 			 * was marked SUSPENDED during SUSPEND.
1238 			 */
1239 			did = fcip_get_broadcast_did(fptr);
1240 			if (frp->fcipr_d_id.port_id == did) {
1241 				frp->fcipr_state = 0;
1242 				index = FCIP_RT_HASH_ELEMS;
1243 				break;
1244 			}
1245 			frp = frp->fcipr_next;
1246 		}
1247 	}
1248 	mutex_exit(&fptr->fcip_rt_mutex);
1249 
1250 	/*
1251 	 * fcip_handle_topology will update the port entries in the
1252 	 * routing table.
1253 	 * fcip_handle_topology also takes care of resetting the
1254 	 * fcipr_state field in the routing table structure. The entries
1255 	 * were set to RT_INVALID during suspend.
1256 	 */
1257 	fcip_handle_topology(fptr);
1258 
1259 done:
1260 	/*
1261 	 * Restart the timeout thread
1262 	 */
1263 	fptr->fcip_timeout_id = timeout(fcip_timeout, fptr,
1264 	    drv_usectohz(1000000));
1265 	mutex_exit(&fptr->fcip_mutex);
1266 	return (rval);
1267 }
1268 
1269 
1270 /*
1271  * Insert a destination port entry into the routing table for
1272  * this port
1273  */
1274 static void
fcip_rt_update(struct fcip * fptr,fc_portmap_t * devlist,uint32_t listlen)1275 fcip_rt_update(struct fcip *fptr, fc_portmap_t *devlist, uint32_t listlen)
1276 {
1277 	struct fcip_routing_table	*frp;
1278 	fcip_port_info_t		*fport = fptr->fcip_port_info;
1279 	int				hash_bucket, i;
1280 	fc_portmap_t			*pmap;
1281 	char				wwn_buf[20];
1282 
1283 	FCIP_TNF_PROBE_2((fcip_rt_update, "fcip io", /* CSTYLED */,
1284 		tnf_string, msg, "enter",
1285 		tnf_int, listlen, listlen));
1286 
1287 	ASSERT(!mutex_owned(&fptr->fcip_mutex));
1288 	mutex_enter(&fptr->fcip_rt_mutex);
1289 
1290 	for (i = 0; i < listlen; i++) {
1291 		pmap = &(devlist[i]);
1292 
1293 		frp = fcip_lookup_rtable(fptr, &(pmap->map_pwwn),
1294 		    FCIP_COMPARE_PWWN);
1295 		/*
1296 		 * If an entry for a port in the devlist exists in the
1297 		 * in the per port routing table, make sure the data
1298 		 * is current. We need to do this irrespective of the
1299 		 * underlying port topology.
1300 		 */
1301 		switch (pmap->map_type) {
1302 		/* FALLTHROUGH */
1303 		case PORT_DEVICE_NOCHANGE:
1304 		/* FALLTHROUGH */
1305 		case PORT_DEVICE_USER_LOGIN:
1306 		/* FALLTHROUGH */
1307 		case PORT_DEVICE_CHANGED:
1308 		/* FALLTHROUGH */
1309 		case PORT_DEVICE_NEW:
1310 			if (frp == NULL) {
1311 				goto add_new_entry;
1312 			} else if (frp) {
1313 				goto update_entry;
1314 			} else {
1315 				continue;
1316 			}
1317 
1318 		case PORT_DEVICE_OLD:
1319 		/* FALLTHROUGH */
1320 		case PORT_DEVICE_USER_LOGOUT:
1321 			/*
1322 			 * Mark entry for removal from Routing Table if
1323 			 * one exists. Let the timeout thread actually
1324 			 * remove the entry after we've given up hopes
1325 			 * of the port ever showing up.
1326 			 */
1327 			if (frp) {
1328 				uint32_t		did;
1329 
1330 				/*
1331 				 * Mark the routing table as invalid to bail
1332 				 * the packets early that are in transit
1333 				 */
1334 				did = fptr->fcip_broadcast_did;
1335 				if (frp->fcipr_d_id.port_id != did) {
1336 					frp->fcipr_pd = NULL;
1337 					frp->fcipr_state = FCIP_RT_INVALID;
1338 					frp->fcipr_invalid_timeout =
1339 					    fptr->fcip_timeout_ticks +
1340 					    FCIP_RTE_TIMEOUT;
1341 				}
1342 			}
1343 			continue;
1344 
1345 		default:
1346 			FCIP_DEBUG(FCIP_DEBUG_INIT, (CE_WARN,
1347 			    "unknown map flags in rt_update"));
1348 			continue;
1349 		}
1350 add_new_entry:
1351 		ASSERT(frp == NULL);
1352 		hash_bucket = FCIP_RT_HASH(pmap->map_pwwn.raw_wwn);
1353 
1354 		ASSERT(hash_bucket < FCIP_RT_HASH_ELEMS);
1355 
1356 		FCIP_TNF_PROBE_2((fcip_rt_update, "cfip io", /* CSTYLED */,
1357 			tnf_string, msg,
1358 			"add new entry",
1359 			tnf_int, hashbucket, hash_bucket));
1360 
1361 		frp = (struct fcip_routing_table *)
1362 		    kmem_zalloc(sizeof (struct fcip_routing_table), KM_SLEEP);
1363 		/* insert at beginning of hash bucket */
1364 		frp->fcipr_next = fptr->fcip_rtable[hash_bucket];
1365 		fptr->fcip_rtable[hash_bucket] = frp;
1366 		fc_wwn_to_str(&pmap->map_pwwn, wwn_buf);
1367 		FCIP_DEBUG(FCIP_DEBUG_ATTACH, (CE_NOTE,
1368 		    "added entry for pwwn %s and d_id 0x%x",
1369 		    wwn_buf, pmap->map_did.port_id));
1370 update_entry:
1371 		bcopy((void *)&pmap->map_pwwn,
1372 		    (void *)&frp->fcipr_pwwn, sizeof (la_wwn_t));
1373 		bcopy((void *)&pmap->map_nwwn, (void *)&frp->fcipr_nwwn,
1374 		    sizeof (la_wwn_t));
1375 		frp->fcipr_d_id = pmap->map_did;
1376 		frp->fcipr_state = pmap->map_state;
1377 		frp->fcipr_pd = pmap->map_pd;
1378 
1379 		/*
1380 		 * If there is no pd for a destination port that is not
1381 		 * a broadcast entry, the port is pretty much unusable - so
1382 		 * mark the port for removal so we can try adding back the
1383 		 * entry again.
1384 		 */
1385 		if ((frp->fcipr_pd == NULL) &&
1386 		    (frp->fcipr_d_id.port_id != fptr->fcip_broadcast_did)) {
1387 			frp->fcipr_state = PORT_DEVICE_INVALID;
1388 			frp->fcipr_invalid_timeout = fptr->fcip_timeout_ticks +
1389 			    (FCIP_RTE_TIMEOUT / 2);
1390 		}
1391 		frp->fcipr_fca_dev =
1392 		    fc_ulp_get_fca_device(fport->fcipp_handle, pmap->map_did);
1393 
1394 		/*
1395 		 * login to the remote port. Don't worry about
1396 		 * plogi failures for now
1397 		 */
1398 		if (pmap->map_pd != NULL) {
1399 			(void) fcip_do_plogi(fptr, frp);
1400 		} else if (FC_TOP_EXTERNAL(fport->fcipp_topology)) {
1401 			fc_wwn_to_str(&frp->fcipr_pwwn, wwn_buf);
1402 			FCIP_DEBUG(FCIP_DEBUG_MISC, (CE_NOTE,
1403 			    "logging into pwwn %s, d_id 0x%x",
1404 			    wwn_buf, frp->fcipr_d_id.port_id));
1405 			(void) fcip_do_plogi(fptr, frp);
1406 		}
1407 
1408 		FCIP_TNF_BYTE_ARRAY(fcip_rt_update, "fcip io", "detail",
1409 			"new wwn in rt", pwwn,
1410 			&frp->fcipr_pwwn, sizeof (la_wwn_t));
1411 	}
1412 	mutex_exit(&fptr->fcip_rt_mutex);
1413 }
1414 
1415 
1416 /*
1417  * return a matching routing table entry for a given fcip instance
1418  */
1419 struct fcip_routing_table *
fcip_lookup_rtable(struct fcip * fptr,la_wwn_t * wwn,int matchflag)1420 fcip_lookup_rtable(struct fcip *fptr, la_wwn_t *wwn, int matchflag)
1421 {
1422 	struct fcip_routing_table	*frp = NULL;
1423 	int				hash_bucket;
1424 
1425 
1426 	FCIP_TNF_PROBE_1((fcip_lookup_rtable, "fcip io", /* CSTYLED */,
1427 		tnf_string, msg, "enter"));
1428 	FCIP_TNF_BYTE_ARRAY(fcip_lookup_rtable, "fcip io", "detail",
1429 		"rtable lookup for", wwn,
1430 		&wwn->raw_wwn, sizeof (la_wwn_t));
1431 	FCIP_TNF_PROBE_2((fcip_lookup_rtable, "fcip io", /* CSTYLED */,
1432 		tnf_string, msg, "match by",
1433 		tnf_int, matchflag, matchflag));
1434 
1435 	ASSERT(mutex_owned(&fptr->fcip_rt_mutex));
1436 
1437 	hash_bucket = FCIP_RT_HASH(wwn->raw_wwn);
1438 	frp = fptr->fcip_rtable[hash_bucket];
1439 	while (frp != NULL) {
1440 
1441 		FCIP_TNF_BYTE_ARRAY(fcip_lookup_rtable, "fcip io", "detail",
1442 			"rtable entry", nwwn,
1443 			&(frp->fcipr_nwwn.raw_wwn), sizeof (la_wwn_t));
1444 
1445 		if (fcip_wwn_compare(&frp->fcipr_pwwn, wwn, matchflag) == 0) {
1446 			break;
1447 		}
1448 
1449 		frp = frp->fcipr_next;
1450 	}
1451 	FCIP_TNF_PROBE_2((fcip_lookup_rtable, "fcip io", /* CSTYLED */,
1452 		tnf_string, msg, "lookup result",
1453 		tnf_opaque, frp, frp));
1454 	return (frp);
1455 }
1456 
1457 /*
1458  * Attach of fcip under pseudo. The actual setup of the interface
1459  * actually happens in fcip_port_attach on a callback from the
1460  * transport. The port_attach callback however can proceed only
1461  * after the devinfo for fcip has been created under pseudo
1462  */
1463 static int
fcip_attach(dev_info_t * dip,ddi_attach_cmd_t cmd)1464 fcip_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
1465 {
1466 	switch ((int)cmd) {
1467 
1468 	case DDI_ATTACH: {
1469 		ASSERT(fcip_module_dip == NULL);
1470 		fcip_module_dip = dip;
1471 
1472 		/*
1473 		 * this call originates as a result of fcip's conf
1474 		 * file entry and will result in a fcip instance being
1475 		 * a child of pseudo. We should ensure here that the port
1476 		 * driver (fp) has been loaded and initted since we would
1477 		 * never get a port attach callback without fp being loaded.
1478 		 * If we are unable to succesfully load and initalize fp -
1479 		 * just fail this attach.
1480 		 */
1481 		mutex_enter(&fcip_global_mutex);
1482 
1483 		FCIP_DEBUG(FCIP_DEBUG_ATTACH,
1484 		    (CE_WARN, "global cv - signaling"));
1485 
1486 		cv_signal(&fcip_global_cv);
1487 
1488 		FCIP_DEBUG(FCIP_DEBUG_ATTACH,
1489 		    (CE_WARN, "global cv - signaled"));
1490 		mutex_exit(&fcip_global_mutex);
1491 		return (DDI_SUCCESS);
1492 	}
1493 	case DDI_RESUME:
1494 		/*
1495 		 * Resume appears trickier
1496 		 */
1497 		return (DDI_SUCCESS);
1498 	default:
1499 		return (DDI_FAILURE);
1500 	}
1501 }
1502 
1503 
1504 /*
1505  * The detach entry point to permit unloading fcip. We make sure
1506  * there are no active streams before we proceed with the detach
1507  */
1508 /* ARGSUSED */
1509 static int
fcip_detach(dev_info_t * dip,ddi_detach_cmd_t cmd)1510 fcip_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
1511 {
1512 	struct fcip		*fptr;
1513 	fcip_port_info_t	*fport;
1514 	int			detached;
1515 
1516 	switch (cmd) {
1517 	case DDI_DETACH: {
1518 		/*
1519 		 * If we got here, any active streams should have been
1520 		 * unplumbed but check anyway
1521 		 */
1522 		mutex_enter(&fcip_global_mutex);
1523 		if (fcipstrup != NULL) {
1524 			mutex_exit(&fcip_global_mutex);
1525 			return (DDI_FAILURE);
1526 		}
1527 
1528 		if (fcip_port_head != NULL) {
1529 			/*
1530 			 * Check to see if we have unattached/unbound
1531 			 * ports. If all the ports are unattached/unbound go
1532 			 * ahead and unregister with the transport
1533 			 */
1534 			fport = fcip_port_head;
1535 			while (fport != NULL) {
1536 				fptr = fport->fcipp_fcip;
1537 				if (fptr == NULL) {
1538 					continue;
1539 				}
1540 				mutex_enter(&fptr->fcip_mutex);
1541 				fptr->fcip_flags |= FCIP_DETACHING;
1542 				if (fptr->fcip_ipq ||
1543 				    fptr->fcip_flags & (FCIP_IN_TIMEOUT |
1544 				    FCIP_IN_CALLBACK | FCIP_ATTACHING |
1545 				    FCIP_SUSPENDED | FCIP_POWER_DOWN |
1546 				    FCIP_REG_INPROGRESS)) {
1547 					FCIP_TNF_PROBE_1((fcip_detach,
1548 					    "fcip io", /* CSTYLED */,
1549 					    tnf_string, msg,
1550 					    "fcip instance busy"));
1551 
1552 					mutex_exit(&fptr->fcip_mutex);
1553 					FCIP_DEBUG(FCIP_DEBUG_DETACH, (CE_WARN,
1554 					    "fcip instance busy"));
1555 					break;
1556 				}
1557 				/*
1558 				 * Check for any outstanding pkts. If yes
1559 				 * fail the detach
1560 				 */
1561 				mutex_enter(&fptr->fcip_dest_mutex);
1562 				if (fcip_port_get_num_pkts(fptr) > 0) {
1563 					mutex_exit(&fptr->fcip_dest_mutex);
1564 					mutex_exit(&fptr->fcip_mutex);
1565 					FCIP_DEBUG(FCIP_DEBUG_DETACH, (CE_WARN,
1566 					    "fcip instance busy - pkts "
1567 					    "pending"));
1568 					break;
1569 				}
1570 				mutex_exit(&fptr->fcip_dest_mutex);
1571 
1572 				mutex_enter(&fptr->fcip_rt_mutex);
1573 				if (fcip_plogi_in_progress(fptr)) {
1574 					mutex_exit(&fptr->fcip_rt_mutex);
1575 					mutex_exit(&fptr->fcip_mutex);
1576 					FCIP_DEBUG(FCIP_DEBUG_DETACH, (CE_WARN,
1577 					    "fcip instance busy - plogi in "
1578 					    "progress"));
1579 					break;
1580 				}
1581 				mutex_exit(&fptr->fcip_rt_mutex);
1582 
1583 				mutex_exit(&fptr->fcip_mutex);
1584 				fport = fport->fcipp_next;
1585 			}
1586 			/*
1587 			 * if fport is non NULL - we have active ports
1588 			 */
1589 			if (fport != NULL) {
1590 				/*
1591 				 * Remove the DETACHING flags on the ports
1592 				 */
1593 				fport = fcip_port_head;
1594 				while (fport != NULL) {
1595 					fptr = fport->fcipp_fcip;
1596 					mutex_enter(&fptr->fcip_mutex);
1597 					fptr->fcip_flags &= ~(FCIP_DETACHING);
1598 					mutex_exit(&fptr->fcip_mutex);
1599 					fport = fport->fcipp_next;
1600 				}
1601 				mutex_exit(&fcip_global_mutex);
1602 				return (DDI_FAILURE);
1603 			}
1604 		}
1605 
1606 		/*
1607 		 * free up all softstate structures
1608 		 */
1609 		fport = fcip_port_head;
1610 		while (fport != NULL) {
1611 			detached = 1;
1612 
1613 			fptr = fport->fcipp_fcip;
1614 			if (fptr) {
1615 				mutex_enter(&fptr->fcip_mutex);
1616 				/*
1617 				 * Check to see if somebody beat us to the
1618 				 * punch
1619 				 */
1620 				detached = fptr->fcip_flags & FCIP_DETACHED;
1621 				fptr->fcip_flags &= ~(FCIP_DETACHING);
1622 				fptr->fcip_flags |= FCIP_DETACHED;
1623 				mutex_exit(&fptr->fcip_mutex);
1624 			}
1625 
1626 			if (!detached) {
1627 				fport = fcip_softstate_free(fport);
1628 			} else {
1629 				/*
1630 				 * If the port was marked as detached
1631 				 * but it was still in the list, that
1632 				 * means another thread has marked it
1633 				 * but we got in while it released the
1634 				 * fcip_global_mutex in softstate_free.
1635 				 * Given that, we're still safe to use
1636 				 * fport->fcipp_next to find out what
1637 				 * the next port on the list is.
1638 				 */
1639 				fport = fport->fcipp_next;
1640 			}
1641 
1642 			FCIP_DEBUG(FCIP_DEBUG_DETACH,
1643 			    (CE_NOTE, "detaching port"));
1644 
1645 			FCIP_TNF_PROBE_1((fcip_detach,
1646 				"fcip io", /* CSTYLED */, tnf_string,
1647 				msg, "detaching port"));
1648 		}
1649 
1650 		/*
1651 		 * If we haven't removed all the port structures, we
1652 		 * aren't yet ready to be detached.
1653 		 */
1654 		if (fcip_port_head != NULL) {
1655 			mutex_exit(&fcip_global_mutex);
1656 			return (DDI_FAILURE);
1657 		}
1658 
1659 		fcip_num_instances = 0;
1660 		mutex_exit(&fcip_global_mutex);
1661 		fcip_module_dip = NULL;
1662 		return (DDI_SUCCESS);
1663 	}
1664 	case DDI_SUSPEND:
1665 		return (DDI_SUCCESS);
1666 	default:
1667 		return (DDI_FAILURE);
1668 	}
1669 }
1670 
1671 /*
1672  * The port_detach callback is called from the transport when a
1673  * FC port is being removed from the transport's control. This routine
1674  * provides fcip with an opportunity to cleanup all activities and
1675  * structures on the port marked for removal.
1676  */
1677 /* ARGSUSED */
1678 static int
fcip_port_detach(opaque_t ulp_handle,fc_ulp_port_info_t * port_info,fc_detach_cmd_t cmd)1679 fcip_port_detach(opaque_t ulp_handle, fc_ulp_port_info_t *port_info,
1680     fc_detach_cmd_t cmd)
1681 {
1682 	int 			rval = FC_FAILURE;
1683 	fcip_port_info_t	*fport;
1684 	struct fcip		*fptr;
1685 	struct fcipstr		*strp;
1686 
1687 	switch (cmd) {
1688 	case FC_CMD_DETACH: {
1689 		mutex_enter(&fcip_global_mutex);
1690 
1691 		if (fcip_port_head == NULL) {
1692 			/*
1693 			 * we are all done but our fini has not been
1694 			 * called yet!! Let's hope we have no active
1695 			 * fcip instances here. - strange secnario but
1696 			 * no harm in having this return a success.
1697 			 */
1698 			fcip_check_remove_minor_node();
1699 
1700 			mutex_exit(&fcip_global_mutex);
1701 			return (FC_SUCCESS);
1702 		} else {
1703 			/*
1704 			 * traverse the port list
1705 			 */
1706 			fport = fcip_port_head;
1707 			while (fport != NULL) {
1708 				if (fport->fcipp_handle ==
1709 				    port_info->port_handle) {
1710 					fptr = fport->fcipp_fcip;
1711 
1712 					/*
1713 					 * Fail the port detach if there is
1714 					 * still an attached, bound stream on
1715 					 * this interface.
1716 					 */
1717 
1718 					rw_enter(&fcipstruplock, RW_READER);
1719 
1720 					for (strp = fcipstrup; strp != NULL;
1721 					    strp = strp->sl_nextp) {
1722 						if (strp->sl_fcip == fptr) {
1723 							rw_exit(&fcipstruplock);
1724 							mutex_exit(
1725 							    &fcip_global_mutex);
1726 							return (FC_FAILURE);
1727 						}
1728 					}
1729 
1730 					rw_exit(&fcipstruplock);
1731 
1732 					/*
1733 					 * fail port detach if we are in
1734 					 * the middle of a deferred port attach
1735 					 * or if the port has outstanding pkts
1736 					 */
1737 					if (fptr != NULL) {
1738 						mutex_enter(&fptr->fcip_mutex);
1739 						if (fcip_check_port_busy
1740 						    (fptr) ||
1741 						    (fptr->fcip_flags &
1742 						    FCIP_DETACHED)) {
1743 							mutex_exit(
1744 							    &fptr->fcip_mutex);
1745 							mutex_exit(
1746 							    &fcip_global_mutex);
1747 							return (FC_FAILURE);
1748 						}
1749 
1750 						fptr->fcip_flags |=
1751 						    FCIP_DETACHED;
1752 						mutex_exit(&fptr->fcip_mutex);
1753 					}
1754 					(void) fcip_softstate_free(fport);
1755 
1756 					fcip_check_remove_minor_node();
1757 					mutex_exit(&fcip_global_mutex);
1758 					return (FC_SUCCESS);
1759 				}
1760 				fport = fport->fcipp_next;
1761 			}
1762 			ASSERT(fport == NULL);
1763 		}
1764 		mutex_exit(&fcip_global_mutex);
1765 		break;
1766 	}
1767 	case FC_CMD_POWER_DOWN:
1768 	/* FALLTHROUGH */
1769 	case FC_CMD_SUSPEND:
1770 		mutex_enter(&fcip_global_mutex);
1771 		fport = fcip_port_head;
1772 		while (fport != NULL) {
1773 			if (fport->fcipp_handle == port_info->port_handle) {
1774 				break;
1775 			}
1776 			fport = fport->fcipp_next;
1777 		}
1778 		if (fport == NULL) {
1779 			mutex_exit(&fcip_global_mutex);
1780 			break;
1781 		}
1782 		rval = fcip_handle_suspend(fport, cmd);
1783 		mutex_exit(&fcip_global_mutex);
1784 		break;
1785 	default:
1786 		FCIP_DEBUG(FCIP_DEBUG_DETACH,
1787 		    (CE_WARN, "unknown port detach command!!"));
1788 		break;
1789 	}
1790 	return (rval);
1791 }
1792 
1793 
1794 /*
1795  * Returns 0 if the port is not busy, else returns non zero.
1796  */
1797 static int
fcip_check_port_busy(struct fcip * fptr)1798 fcip_check_port_busy(struct fcip *fptr)
1799 {
1800 	int rval = 0, num_pkts = 0;
1801 
1802 	ASSERT(fptr != NULL);
1803 	ASSERT(MUTEX_HELD(&fptr->fcip_mutex));
1804 
1805 	mutex_enter(&fptr->fcip_dest_mutex);
1806 
1807 	if (fptr->fcip_flags & FCIP_PORT_BUSY ||
1808 	    ((num_pkts = fcip_port_get_num_pkts(fptr)) > 0) ||
1809 	    fptr->fcip_num_ipkts_pending) {
1810 		rval = 1;
1811 		FCIP_DEBUG(FCIP_DEBUG_DETACH,
1812 		    (CE_NOTE, "!fcip_check_port_busy: port is busy "
1813 		    "fcip_flags: 0x%x, num_pkts: 0x%x, ipkts_pending: 0x%lx!",
1814 		    fptr->fcip_flags, num_pkts, fptr->fcip_num_ipkts_pending));
1815 	}
1816 
1817 	mutex_exit(&fptr->fcip_dest_mutex);
1818 	return (rval);
1819 }
1820 
1821 /*
1822  * Helper routine to remove fcip's minor node
1823  * There is one minor node per system and it should be removed if there are no
1824  * other fcip instances (which has a 1:1 mapping for fp instances) present
1825  */
1826 static void
fcip_check_remove_minor_node(void)1827 fcip_check_remove_minor_node(void)
1828 {
1829 	ASSERT(MUTEX_HELD(&fcip_global_mutex));
1830 
1831 	/*
1832 	 * If there are no more fcip (fp) instances, remove the
1833 	 * minor node for fcip.
1834 	 * Reset fcip_minor_node_created to invalidate it.
1835 	 */
1836 	if (fcip_num_instances == 0 && (fcip_module_dip != NULL)) {
1837 		ddi_remove_minor_node(fcip_module_dip, NULL);
1838 		fcip_minor_node_created = 0;
1839 	}
1840 }
1841 
1842 /*
1843  * This routine permits the suspend operation during a CPR/System
1844  * power management operation. The routine basically quiesces I/Os
1845  * on all active interfaces
1846  */
1847 static int
fcip_handle_suspend(fcip_port_info_t * fport,fc_detach_cmd_t cmd)1848 fcip_handle_suspend(fcip_port_info_t *fport, fc_detach_cmd_t cmd)
1849 {
1850 	struct fcip	*fptr = fport->fcipp_fcip;
1851 	timeout_id_t	tid;
1852 	int 		index;
1853 	int		tryagain = 0;
1854 	int		count;
1855 	struct fcipstr	*tslp;
1856 
1857 
1858 	ASSERT(fptr != NULL);
1859 	mutex_enter(&fptr->fcip_mutex);
1860 
1861 	/*
1862 	 * Fail if we are in the middle of a callback. Don't use delay during
1863 	 * suspend since clock intrs are not available so busy wait
1864 	 */
1865 	count = 0;
1866 	while (count++ < 15 &&
1867 	    ((fptr->fcip_flags & FCIP_IN_CALLBACK) ||
1868 	    (fptr->fcip_flags & FCIP_IN_TIMEOUT))) {
1869 		mutex_exit(&fptr->fcip_mutex);
1870 		drv_usecwait(1000000);
1871 		mutex_enter(&fptr->fcip_mutex);
1872 	}
1873 
1874 	if (fptr->fcip_flags & FCIP_IN_CALLBACK ||
1875 	    fptr->fcip_flags & FCIP_IN_TIMEOUT) {
1876 		mutex_exit(&fptr->fcip_mutex);
1877 		return (FC_FAILURE);
1878 	}
1879 
1880 	if (cmd == FC_CMD_POWER_DOWN) {
1881 		if (fptr->fcip_flags & FCIP_SUSPENDED) {
1882 			fptr->fcip_flags |= FCIP_POWER_DOWN;
1883 			mutex_exit(&fptr->fcip_mutex);
1884 			goto success;
1885 		} else {
1886 			fptr->fcip_flags |= FCIP_POWER_DOWN;
1887 		}
1888 	} else if (cmd == FC_CMD_SUSPEND) {
1889 		fptr->fcip_flags |= FCIP_SUSPENDED;
1890 	} else {
1891 		mutex_exit(&fptr->fcip_mutex);
1892 		return (FC_FAILURE);
1893 	}
1894 
1895 	mutex_exit(&fptr->fcip_mutex);
1896 	/*
1897 	 * If no streams are plumbed - its the easiest case - Just
1898 	 * bail out without having to do much
1899 	 */
1900 
1901 	rw_enter(&fcipstruplock, RW_READER);
1902 	for (tslp = fcipstrup; tslp; tslp = tslp->sl_nextp) {
1903 		if (tslp->sl_fcip == fptr) {
1904 			break;
1905 		}
1906 	}
1907 	rw_exit(&fcipstruplock);
1908 
1909 	/*
1910 	 * No active streams on this port
1911 	 */
1912 	if (tslp == NULL) {
1913 		goto success;
1914 	}
1915 
1916 	/*
1917 	 * Walk through each Routing table structure and check if
1918 	 * the destination table has any outstanding commands. If yes
1919 	 * wait for the commands to drain. Since we go through each
1920 	 * routing table entry in succession, it may be wise to wait
1921 	 * only a few seconds for each entry.
1922 	 */
1923 	mutex_enter(&fptr->fcip_rt_mutex);
1924 	while (!tryagain) {
1925 
1926 		tryagain = 0;
1927 		for (index = 0; index < FCIP_RT_HASH_ELEMS; index++) {
1928 			struct fcip_routing_table 	*frp;
1929 			struct fcip_dest 		*fdestp;
1930 			la_wwn_t			*pwwn;
1931 			int				hash_bucket;
1932 
1933 			frp = fptr->fcip_rtable[index];
1934 			while (frp) {
1935 				/*
1936 				 * Mark the routing table as SUSPENDED. Even
1937 				 * mark the broadcast entry SUSPENDED to
1938 				 * prevent any ARP or other broadcasts. We
1939 				 * can reset the state of the broadcast
1940 				 * RTE when we resume.
1941 				 */
1942 				frp->fcipr_state = FCIP_RT_SUSPENDED;
1943 				pwwn = &frp->fcipr_pwwn;
1944 
1945 				/*
1946 				 * Get hold of destination pointer
1947 				 */
1948 				mutex_enter(&fptr->fcip_dest_mutex);
1949 
1950 				hash_bucket = FCIP_DEST_HASH(pwwn->raw_wwn);
1951 				ASSERT(hash_bucket < FCIP_DEST_HASH_ELEMS);
1952 
1953 				fdestp = fptr->fcip_dest[hash_bucket];
1954 				while (fdestp != NULL) {
1955 					mutex_enter(&fdestp->fcipd_mutex);
1956 					if (fdestp->fcipd_rtable) {
1957 						if (fcip_wwn_compare(pwwn,
1958 						    &fdestp->fcipd_pwwn,
1959 						    FCIP_COMPARE_PWWN) == 0) {
1960 							mutex_exit(
1961 							&fdestp->fcipd_mutex);
1962 							break;
1963 						}
1964 					}
1965 					mutex_exit(&fdestp->fcipd_mutex);
1966 					fdestp = fdestp->fcipd_next;
1967 				}
1968 
1969 				mutex_exit(&fptr->fcip_dest_mutex);
1970 				if (fdestp == NULL) {
1971 					frp = frp->fcipr_next;
1972 					continue;
1973 				}
1974 
1975 				/*
1976 				 * Wait for fcip_wait_cmds seconds for
1977 				 * the commands to drain.
1978 				 */
1979 				count = 0;
1980 				mutex_enter(&fdestp->fcipd_mutex);
1981 				while (fdestp->fcipd_ncmds &&
1982 				    count < fcip_wait_cmds) {
1983 					mutex_exit(&fdestp->fcipd_mutex);
1984 					mutex_exit(&fptr->fcip_rt_mutex);
1985 					drv_usecwait(1000000);
1986 					mutex_enter(&fptr->fcip_rt_mutex);
1987 					mutex_enter(&fdestp->fcipd_mutex);
1988 					count++;
1989 				}
1990 				/*
1991 				 * Check if we were able to drain all cmds
1992 				 * successfully. Else continue with other
1993 				 * ports and try during the second pass
1994 				 */
1995 				if (fdestp->fcipd_ncmds) {
1996 					tryagain++;
1997 				}
1998 				mutex_exit(&fdestp->fcipd_mutex);
1999 
2000 				frp = frp->fcipr_next;
2001 			}
2002 		}
2003 		if (tryagain == 0) {
2004 			break;
2005 		}
2006 	}
2007 	mutex_exit(&fptr->fcip_rt_mutex);
2008 
2009 	if (tryagain) {
2010 		mutex_enter(&fptr->fcip_mutex);
2011 		fptr->fcip_flags &= ~(FCIP_SUSPENDED | FCIP_POWER_DOWN);
2012 		mutex_exit(&fptr->fcip_mutex);
2013 		return (FC_FAILURE);
2014 	}
2015 
2016 success:
2017 	mutex_enter(&fptr->fcip_mutex);
2018 	tid = fptr->fcip_timeout_id;
2019 	fptr->fcip_timeout_id = NULL;
2020 	mutex_exit(&fptr->fcip_mutex);
2021 
2022 	(void) untimeout(tid);
2023 
2024 	return (FC_SUCCESS);
2025 }
2026 
2027 /*
2028  * the getinfo(9E) entry point
2029  */
2030 /* ARGSUSED */
2031 static int
fcip_getinfo(dev_info_t * dip,ddi_info_cmd_t cmd,void * arg,void ** result)2032 fcip_getinfo(dev_info_t *dip, ddi_info_cmd_t cmd, void *arg, void **result)
2033 {
2034 	int rval = DDI_FAILURE;
2035 
2036 	switch (cmd) {
2037 	case DDI_INFO_DEVT2DEVINFO:
2038 		*result = fcip_module_dip;
2039 		if (*result)
2040 			rval = DDI_SUCCESS;
2041 		break;
2042 
2043 	case DDI_INFO_DEVT2INSTANCE:
2044 		*result = (void *)0;
2045 		rval = DDI_SUCCESS;
2046 		break;
2047 	default:
2048 		break;
2049 	}
2050 
2051 	return (rval);
2052 }
2053 
2054 /*
2055  * called from fcip_attach to initialize kstats for the link
2056  */
2057 /* ARGSUSED */
2058 static void
fcip_kstat_init(struct fcip * fptr)2059 fcip_kstat_init(struct fcip *fptr)
2060 {
2061 	int instance;
2062 	char buf[16];
2063 	struct fcipstat	*fcipstatp;
2064 
2065 	ASSERT(mutex_owned(&fptr->fcip_mutex));
2066 
2067 	instance = ddi_get_instance(fptr->fcip_dip);
2068 	(void) sprintf(buf, "fcip%d", instance);
2069 
2070 #ifdef	kstat
2071 	fptr->fcip_kstatp = kstat_create("fcip", instance, buf, "net",
2072 	    KSTAT_TYPE_NAMED,
2073 	    (sizeof (struct fcipstat)/ sizeof (kstat_named_t)),
2074 	    KSTAT_FLAG_PERSISTENT);
2075 #else
2076 	fptr->fcip_kstatp = kstat_create("fcip", instance, buf, "net",
2077 	    KSTAT_TYPE_NAMED,
2078 	    (sizeof (struct fcipstat)/ sizeof (kstat_named_t)), 0);
2079 #endif
2080 	if (fptr->fcip_kstatp == NULL) {
2081 		FCIP_DEBUG(FCIP_DEBUG_INIT, (CE_WARN, "kstat created failed"));
2082 		return;
2083 	}
2084 
2085 	fcipstatp = (struct  fcipstat *)fptr->fcip_kstatp->ks_data;
2086 	kstat_named_init(&fcipstatp->fcips_ipackets,	"ipackets",
2087 		KSTAT_DATA_ULONG);
2088 	kstat_named_init(&fcipstatp->fcips_ierrors,	"ierrors",
2089 		KSTAT_DATA_ULONG);
2090 	kstat_named_init(&fcipstatp->fcips_opackets,	"opackets",
2091 		KSTAT_DATA_ULONG);
2092 	kstat_named_init(&fcipstatp->fcips_oerrors,	"oerrors",
2093 		KSTAT_DATA_ULONG);
2094 	kstat_named_init(&fcipstatp->fcips_collisions,	"collisions",
2095 		KSTAT_DATA_ULONG);
2096 	kstat_named_init(&fcipstatp->fcips_nocanput,	"nocanput",
2097 		KSTAT_DATA_ULONG);
2098 	kstat_named_init(&fcipstatp->fcips_allocbfail,	"allocbfail",
2099 		KSTAT_DATA_ULONG);
2100 
2101 	kstat_named_init(&fcipstatp->fcips_defer, "defer",
2102 		KSTAT_DATA_ULONG);
2103 	kstat_named_init(&fcipstatp->fcips_fram, "fram",
2104 		KSTAT_DATA_ULONG);
2105 	kstat_named_init(&fcipstatp->fcips_crc, "crc",
2106 		KSTAT_DATA_ULONG);
2107 	kstat_named_init(&fcipstatp->fcips_oflo, "oflo",
2108 		KSTAT_DATA_ULONG);
2109 	kstat_named_init(&fcipstatp->fcips_uflo, "uflo",
2110 		KSTAT_DATA_ULONG);
2111 	kstat_named_init(&fcipstatp->fcips_missed, "missed",
2112 		KSTAT_DATA_ULONG);
2113 	kstat_named_init(&fcipstatp->fcips_tlcol, "tlcol",
2114 		KSTAT_DATA_ULONG);
2115 	kstat_named_init(&fcipstatp->fcips_trtry, "trtry",
2116 		KSTAT_DATA_ULONG);
2117 	kstat_named_init(&fcipstatp->fcips_tnocar, "tnocar",
2118 		KSTAT_DATA_ULONG);
2119 	kstat_named_init(&fcipstatp->fcips_inits, "inits",
2120 		KSTAT_DATA_ULONG);
2121 	kstat_named_init(&fcipstatp->fcips_notbufs, "notbufs",
2122 		KSTAT_DATA_ULONG);
2123 	kstat_named_init(&fcipstatp->fcips_norbufs, "norbufs",
2124 		KSTAT_DATA_ULONG);
2125 	kstat_named_init(&fcipstatp->fcips_allocbfail, "allocbfail",
2126 		KSTAT_DATA_ULONG);
2127 
2128 	/*
2129 	 * required by kstat for MIB II objects(RFC 1213)
2130 	 */
2131 	kstat_named_init(&fcipstatp->fcips_rcvbytes, "fcips_rcvbytes",
2132 		KSTAT_DATA_ULONG);	/* # octets received */
2133 					/* MIB - ifInOctets */
2134 	kstat_named_init(&fcipstatp->fcips_xmtbytes, "fcips_xmtbytes",
2135 		KSTAT_DATA_ULONG);	/* # octets xmitted */
2136 					/* MIB - ifOutOctets */
2137 	kstat_named_init(&fcipstatp->fcips_multircv,	"fcips_multircv",
2138 		KSTAT_DATA_ULONG);	/* # multicast packets */
2139 					/* delivered to upper layer */
2140 					/* MIB - ifInNUcastPkts */
2141 	kstat_named_init(&fcipstatp->fcips_multixmt,	"fcips_multixmt",
2142 		KSTAT_DATA_ULONG);	/* # multicast packets */
2143 					/* requested to be sent */
2144 					/* MIB - ifOutNUcastPkts */
2145 	kstat_named_init(&fcipstatp->fcips_brdcstrcv, "fcips_brdcstrcv",
2146 		KSTAT_DATA_ULONG); /* # broadcast packets */
2147 					/* delivered to upper layer */
2148 					/* MIB - ifInNUcastPkts */
2149 	kstat_named_init(&fcipstatp->fcips_brdcstxmt, "fcips_brdcstxmt",
2150 		KSTAT_DATA_ULONG);	/* # broadcast packets */
2151 					/* requested to be sent */
2152 					/* MIB - ifOutNUcastPkts */
2153 	kstat_named_init(&fcipstatp->fcips_norcvbuf,	"fcips_norcvbuf",
2154 		KSTAT_DATA_ULONG);	/* # rcv packets discarded */
2155 					/* MIB - ifInDiscards */
2156 	kstat_named_init(&fcipstatp->fcips_noxmtbuf,	"fcips_noxmtbuf",
2157 		KSTAT_DATA_ULONG);	/* # xmt packets discarded */
2158 
2159 	fptr->fcip_kstatp->ks_update = fcip_stat_update;
2160 	fptr->fcip_kstatp->ks_private = (void *) fptr;
2161 	kstat_install(fptr->fcip_kstatp);
2162 }
2163 
2164 /*
2165  * Update the defined kstats for netstat et al to use
2166  */
2167 /* ARGSUSED */
2168 static int
fcip_stat_update(kstat_t * fcip_statp,int val)2169 fcip_stat_update(kstat_t *fcip_statp, int val)
2170 {
2171 	struct fcipstat	*fcipstatp;
2172 	struct fcip	*fptr;
2173 
2174 	fptr = (struct fcip *)fcip_statp->ks_private;
2175 	fcipstatp = (struct fcipstat *)fcip_statp->ks_data;
2176 
2177 	if (val == KSTAT_WRITE) {
2178 		fptr->fcip_ipackets	= fcipstatp->fcips_ipackets.value.ul;
2179 		fptr->fcip_ierrors	= fcipstatp->fcips_ierrors.value.ul;
2180 		fptr->fcip_opackets	= fcipstatp->fcips_opackets.value.ul;
2181 		fptr->fcip_oerrors	= fcipstatp->fcips_oerrors.value.ul;
2182 		fptr->fcip_collisions	= fcipstatp->fcips_collisions.value.ul;
2183 		fptr->fcip_defer	= fcipstatp->fcips_defer.value.ul;
2184 		fptr->fcip_fram	= fcipstatp->fcips_fram.value.ul;
2185 		fptr->fcip_crc	= fcipstatp->fcips_crc.value.ul;
2186 		fptr->fcip_oflo	= fcipstatp->fcips_oflo.value.ul;
2187 		fptr->fcip_uflo	= fcipstatp->fcips_uflo.value.ul;
2188 		fptr->fcip_missed	= fcipstatp->fcips_missed.value.ul;
2189 		fptr->fcip_tlcol	= fcipstatp->fcips_tlcol.value.ul;
2190 		fptr->fcip_trtry	= fcipstatp->fcips_trtry.value.ul;
2191 		fptr->fcip_tnocar	= fcipstatp->fcips_tnocar.value.ul;
2192 		fptr->fcip_inits	= fcipstatp->fcips_inits.value.ul;
2193 		fptr->fcip_notbufs	= fcipstatp->fcips_notbufs.value.ul;
2194 		fptr->fcip_norbufs	= fcipstatp->fcips_norbufs.value.ul;
2195 		fptr->fcip_nocanput	= fcipstatp->fcips_nocanput.value.ul;
2196 		fptr->fcip_allocbfail	= fcipstatp->fcips_allocbfail.value.ul;
2197 		fptr->fcip_rcvbytes	= fcipstatp->fcips_rcvbytes.value.ul;
2198 		fptr->fcip_xmtbytes	= fcipstatp->fcips_xmtbytes.value.ul;
2199 		fptr->fcip_multircv	= fcipstatp->fcips_multircv.value.ul;
2200 		fptr->fcip_multixmt	= fcipstatp->fcips_multixmt.value.ul;
2201 		fptr->fcip_brdcstrcv	= fcipstatp->fcips_brdcstrcv.value.ul;
2202 		fptr->fcip_norcvbuf	= fcipstatp->fcips_norcvbuf.value.ul;
2203 		fptr->fcip_noxmtbuf	= fcipstatp->fcips_noxmtbuf.value.ul;
2204 		fptr->fcip_allocbfail	= fcipstatp->fcips_allocbfail.value.ul;
2205 		fptr->fcip_allocbfail	= fcipstatp->fcips_allocbfail.value.ul;
2206 		fptr->fcip_allocbfail	= fcipstatp->fcips_allocbfail.value.ul;
2207 		fptr->fcip_allocbfail	= fcipstatp->fcips_allocbfail.value.ul;
2208 		fptr->fcip_allocbfail	= fcipstatp->fcips_allocbfail.value.ul;
2209 		fptr->fcip_allocbfail	= fcipstatp->fcips_allocbfail.value.ul;
2210 		fptr->fcip_allocbfail	= fcipstatp->fcips_allocbfail.value.ul;
2211 		fptr->fcip_allocbfail	= fcipstatp->fcips_allocbfail.value.ul;
2212 
2213 	} else {
2214 		fcipstatp->fcips_ipackets.value.ul	= fptr->fcip_ipackets;
2215 		fcipstatp->fcips_ierrors.value.ul	= fptr->fcip_ierrors;
2216 		fcipstatp->fcips_opackets.value.ul	= fptr->fcip_opackets;
2217 		fcipstatp->fcips_oerrors.value.ul	= fptr->fcip_oerrors;
2218 		fcipstatp->fcips_collisions.value.ul	= fptr->fcip_collisions;
2219 		fcipstatp->fcips_nocanput.value.ul	= fptr->fcip_nocanput;
2220 		fcipstatp->fcips_allocbfail.value.ul	= fptr->fcip_allocbfail;
2221 		fcipstatp->fcips_defer.value.ul	= fptr->fcip_defer;
2222 		fcipstatp->fcips_fram.value.ul	= fptr->fcip_fram;
2223 		fcipstatp->fcips_crc.value.ul	= fptr->fcip_crc;
2224 		fcipstatp->fcips_oflo.value.ul	= fptr->fcip_oflo;
2225 		fcipstatp->fcips_uflo.value.ul	= fptr->fcip_uflo;
2226 		fcipstatp->fcips_missed.value.ul	= fptr->fcip_missed;
2227 		fcipstatp->fcips_tlcol.value.ul	= fptr->fcip_tlcol;
2228 		fcipstatp->fcips_trtry.value.ul	= fptr->fcip_trtry;
2229 		fcipstatp->fcips_tnocar.value.ul	= fptr->fcip_tnocar;
2230 		fcipstatp->fcips_inits.value.ul	= fptr->fcip_inits;
2231 		fcipstatp->fcips_norbufs.value.ul	= fptr->fcip_norbufs;
2232 		fcipstatp->fcips_notbufs.value.ul	= fptr->fcip_notbufs;
2233 		fcipstatp->fcips_rcvbytes.value.ul	= fptr->fcip_rcvbytes;
2234 		fcipstatp->fcips_xmtbytes.value.ul	= fptr->fcip_xmtbytes;
2235 		fcipstatp->fcips_multircv.value.ul	= fptr->fcip_multircv;
2236 		fcipstatp->fcips_multixmt.value.ul	= fptr->fcip_multixmt;
2237 		fcipstatp->fcips_brdcstrcv.value.ul	= fptr->fcip_brdcstrcv;
2238 		fcipstatp->fcips_brdcstxmt.value.ul	= fptr->fcip_brdcstxmt;
2239 		fcipstatp->fcips_norcvbuf.value.ul	= fptr->fcip_norcvbuf;
2240 		fcipstatp->fcips_noxmtbuf.value.ul	= fptr->fcip_noxmtbuf;
2241 
2242 	}
2243 	return (0);
2244 }
2245 
2246 
2247 /*
2248  * fcip_statec_cb: handles all required state change callback notifications
2249  * it receives from the transport
2250  */
2251 /* ARGSUSED */
2252 static void
fcip_statec_cb(opaque_t ulp_handle,opaque_t phandle,uint32_t port_state,uint32_t port_top,fc_portmap_t changelist[],uint32_t listlen,uint32_t sid)2253 fcip_statec_cb(opaque_t ulp_handle, opaque_t phandle,
2254     uint32_t port_state, uint32_t port_top, fc_portmap_t changelist[],
2255     uint32_t listlen, uint32_t sid)
2256 {
2257 	fcip_port_info_t	*fport;
2258 	struct fcip 		*fptr;
2259 	struct fcipstr		*slp;
2260 	queue_t			*wrq;
2261 	int			instance;
2262 	int 			index;
2263 	struct fcip_routing_table 	*frtp;
2264 
2265 	fport = fcip_get_port(phandle);
2266 
2267 	if (fport == NULL) {
2268 		return;
2269 	}
2270 
2271 	fptr = fport->fcipp_fcip;
2272 	ASSERT(fptr != NULL);
2273 
2274 	if (fptr == NULL) {
2275 		return;
2276 	}
2277 
2278 	instance = ddi_get_instance(fport->fcipp_dip);
2279 
2280 	FCIP_TNF_PROBE_4((fcip_statec_cb, "fcip io", /* CSTYLED */,
2281 		tnf_string, msg, "state change callback",
2282 		tnf_uint, instance, instance,
2283 		tnf_uint, S_ID, sid,
2284 		tnf_int, count, listlen));
2285 	FCIP_DEBUG(FCIP_DEBUG_ELS,
2286 	    (CE_NOTE, "fcip%d, state change callback: state:0x%x, "
2287 	    "S_ID:0x%x, count:0x%x", instance, port_state, sid, listlen));
2288 
2289 	mutex_enter(&fptr->fcip_mutex);
2290 
2291 	if ((fptr->fcip_flags & (FCIP_DETACHING | FCIP_DETACHED)) ||
2292 	    (fptr->fcip_flags & (FCIP_SUSPENDED | FCIP_POWER_DOWN))) {
2293 		mutex_exit(&fptr->fcip_mutex);
2294 		return;
2295 	}
2296 
2297 	/*
2298 	 * set fcip flags to indicate we are in the middle of a
2299 	 * state change callback so we can wait till the statechange
2300 	 * is handled before succeeding/failing the SUSPEND/POWER DOWN.
2301 	 */
2302 	fptr->fcip_flags |= FCIP_IN_SC_CB;
2303 
2304 	fport->fcipp_pstate = port_state;
2305 
2306 	/*
2307 	 * Check if topology changed. If Yes - Modify the broadcast
2308 	 * RTE entries to understand the new broadcast D_IDs
2309 	 */
2310 	if (fport->fcipp_topology != port_top &&
2311 	    (port_top != FC_TOP_UNKNOWN)) {
2312 		/* REMOVE later */
2313 		FCIP_DEBUG(FCIP_DEBUG_ELS, (CE_NOTE,
2314 		    "topology changed: Old topology: 0x%x New topology 0x%x",
2315 		    fport->fcipp_topology, port_top));
2316 		/*
2317 		 * If topology changed - attempt a rediscovery of
2318 		 * devices. Helps specially in Fabric/Public loops
2319 		 * and if on_demand_node_creation is disabled
2320 		 */
2321 		fport->fcipp_topology = port_top;
2322 		fcip_handle_topology(fptr);
2323 	}
2324 
2325 	mutex_exit(&fptr->fcip_mutex);
2326 
2327 	switch (FC_PORT_STATE_MASK(port_state)) {
2328 	case FC_STATE_ONLINE:
2329 	/* FALLTHROUGH */
2330 	case FC_STATE_LIP:
2331 	/* FALLTHROUGH */
2332 	case FC_STATE_LIP_LBIT_SET:
2333 
2334 		/*
2335 		 * nothing to do here actually other than if we
2336 		 * were actually logged onto a port in the devlist
2337 		 * (which indicates active communication between
2338 		 * the host port and the port in the changelist).
2339 		 * If however we are in a private loop or point to
2340 		 * point mode, we need to check for any IP capable
2341 		 * ports and update our routing table.
2342 		 */
2343 		switch (port_top) {
2344 		case FC_TOP_FABRIC:
2345 			/*
2346 			 * This indicates a fabric port with a NameServer.
2347 			 * Check the devlist to see if we are in active
2348 			 * communication with a port on the devlist.
2349 			 */
2350 			FCIP_DEBUG(FCIP_DEBUG_ELS, (CE_NOTE,
2351 			    "Statec_cb: fabric topology"));
2352 			fcip_rt_update(fptr, changelist, listlen);
2353 			break;
2354 		case FC_TOP_NO_NS:
2355 			/*
2356 			 * No nameserver - so treat it like a Private loop
2357 			 * or point to point topology and get a map of
2358 			 * devices on the link and get IP capable ports to
2359 			 * to update the routing table.
2360 			 */
2361 			FCIP_DEBUG(FCIP_DEBUG_ELS,
2362 			    (CE_NOTE, "Statec_cb: NO_NS topology"));
2363 		/* FALLTHROUGH */
2364 		case FC_TOP_PRIVATE_LOOP:
2365 			FCIP_DEBUG(FCIP_DEBUG_ELS, (CE_NOTE,
2366 			    "Statec_cb: Pvt_Loop topology"));
2367 		/* FALLTHROUGH */
2368 		case FC_TOP_PT_PT:
2369 			/*
2370 			 * call get_port_map() and update routing table
2371 			 */
2372 			fcip_rt_update(fptr, changelist, listlen);
2373 			break;
2374 		default:
2375 			FCIP_DEBUG(FCIP_DEBUG_ELS,
2376 			    (CE_NOTE, "Statec_cb: Unknown topology"));
2377 		}
2378 
2379 		/*
2380 		 * We should now enable the Queues and permit I/Os
2381 		 * to flow through downstream. The update of routing
2382 		 * table should have flushed out any port entries that
2383 		 * don't exist or are not available after the state change
2384 		 */
2385 		mutex_enter(&fptr->fcip_mutex);
2386 		fptr->fcip_port_state = FCIP_PORT_ONLINE;
2387 		if (fptr->fcip_flags & FCIP_LINK_DOWN) {
2388 			fptr->fcip_flags &= ~FCIP_LINK_DOWN;
2389 		}
2390 		mutex_exit(&fptr->fcip_mutex);
2391 
2392 		/*
2393 		 * Enable write queues
2394 		 */
2395 		rw_enter(&fcipstruplock, RW_READER);
2396 		for (slp = fcipstrup; slp != NULL; slp = slp->sl_nextp) {
2397 			if (slp && slp->sl_fcip == fptr) {
2398 				wrq = WR(slp->sl_rq);
2399 				if (wrq->q_flag & QFULL) {
2400 					qenable(wrq);
2401 				}
2402 			}
2403 		}
2404 		rw_exit(&fcipstruplock);
2405 		break;
2406 	case FC_STATE_OFFLINE:
2407 		/*
2408 		 * mark the port_state OFFLINE and wait for it to
2409 		 * become online. Any new messages in this state will
2410 		 * simply be queued back up. If the port does not
2411 		 * come online in a short while, we can begin failing
2412 		 * messages and flush the routing table
2413 		 */
2414 		mutex_enter(&fptr->fcip_mutex);
2415 		fptr->fcip_mark_offline = fptr->fcip_timeout_ticks +
2416 		    FCIP_OFFLINE_TIMEOUT;
2417 		fptr->fcip_port_state = FCIP_PORT_OFFLINE;
2418 		mutex_exit(&fptr->fcip_mutex);
2419 
2420 		/*
2421 		 * Mark all Routing table entries as invalid to prevent
2422 		 * any commands from trickling through to ports that
2423 		 * have disappeared from under us
2424 		 */
2425 		mutex_enter(&fptr->fcip_rt_mutex);
2426 		for (index = 0; index < FCIP_RT_HASH_ELEMS; index++) {
2427 			frtp = fptr->fcip_rtable[index];
2428 			while (frtp) {
2429 				frtp->fcipr_state = PORT_DEVICE_INVALID;
2430 				frtp = frtp->fcipr_next;
2431 			}
2432 		}
2433 		mutex_exit(&fptr->fcip_rt_mutex);
2434 
2435 		break;
2436 
2437 	case FC_STATE_RESET_REQUESTED:
2438 		/*
2439 		 * Release all Unsolicited buffers back to transport/FCA.
2440 		 * This also means the port state is marked offline - so
2441 		 * we may have to do what OFFLINE state requires us to do.
2442 		 * Care must be taken to wait for any active unsolicited
2443 		 * buffer with the other Streams modules - so wait for
2444 		 * a freeb if the unsolicited buffer is passed back all
2445 		 * the way upstream.
2446 		 */
2447 		mutex_enter(&fptr->fcip_mutex);
2448 
2449 #ifdef FCIP_ESBALLOC
2450 		while (fptr->fcip_ub_upstream) {
2451 			cv_wait(&fptr->fcip_ub_cv, &fptr->fcip_mutex);
2452 		}
2453 #endif	/* FCIP_ESBALLOC */
2454 
2455 		fptr->fcip_mark_offline = fptr->fcip_timeout_ticks +
2456 		    FCIP_OFFLINE_TIMEOUT;
2457 		fptr->fcip_port_state = FCIP_PORT_OFFLINE;
2458 		mutex_exit(&fptr->fcip_mutex);
2459 		break;
2460 
2461 	case FC_STATE_DEVICE_CHANGE:
2462 		if (listlen) {
2463 			fcip_rt_update(fptr, changelist, listlen);
2464 		}
2465 		break;
2466 	case FC_STATE_RESET:
2467 		/*
2468 		 * Not much to do I guess - wait for port to become
2469 		 * ONLINE. If the port doesn't become online in a short
2470 		 * while, the upper layers abort any request themselves.
2471 		 * We can just putback the messages in the streams queues
2472 		 * if the link is offline
2473 		 */
2474 		break;
2475 	}
2476 	mutex_enter(&fptr->fcip_mutex);
2477 	fptr->fcip_flags &= ~(FCIP_IN_SC_CB);
2478 	mutex_exit(&fptr->fcip_mutex);
2479 }
2480 
2481 /*
2482  * Given a port handle, return the fcip_port_info structure corresponding
2483  * to that port handle. The transport allocates and communicates with
2484  * ULPs using port handles
2485  */
2486 static fcip_port_info_t *
fcip_get_port(opaque_t phandle)2487 fcip_get_port(opaque_t phandle)
2488 {
2489 	fcip_port_info_t *fport;
2490 
2491 	ASSERT(phandle != NULL);
2492 
2493 	mutex_enter(&fcip_global_mutex);
2494 	fport = fcip_port_head;
2495 
2496 	while (fport != NULL) {
2497 		if (fport->fcipp_handle == phandle) {
2498 			/* found */
2499 			break;
2500 		}
2501 		fport = fport->fcipp_next;
2502 	}
2503 
2504 	mutex_exit(&fcip_global_mutex);
2505 
2506 	return (fport);
2507 }
2508 
2509 /*
2510  * Handle inbound ELS requests received by the transport. We are only
2511  * intereseted in FARP/InARP mostly.
2512  */
2513 /* ARGSUSED */
2514 static int
fcip_els_cb(opaque_t ulp_handle,opaque_t phandle,fc_unsol_buf_t * buf,uint32_t claimed)2515 fcip_els_cb(opaque_t ulp_handle, opaque_t phandle,
2516     fc_unsol_buf_t *buf, uint32_t claimed)
2517 {
2518 	fcip_port_info_t	*fport;
2519 	struct fcip 		*fptr;
2520 	int			instance;
2521 	uchar_t			r_ctl;
2522 	uchar_t			ls_code;
2523 	la_els_farp_t		farp_cmd;
2524 	la_els_farp_t		*fcmd;
2525 	int			rval = FC_UNCLAIMED;
2526 
2527 	fport = fcip_get_port(phandle);
2528 	if (fport == NULL) {
2529 		return (FC_UNCLAIMED);
2530 	}
2531 
2532 	fptr = fport->fcipp_fcip;
2533 	ASSERT(fptr != NULL);
2534 	if (fptr == NULL) {
2535 		return (FC_UNCLAIMED);
2536 	}
2537 
2538 	instance = ddi_get_instance(fport->fcipp_dip);
2539 
2540 	mutex_enter(&fptr->fcip_mutex);
2541 	if ((fptr->fcip_flags & (FCIP_DETACHING | FCIP_DETACHED)) ||
2542 	    (fptr->fcip_flags & (FCIP_SUSPENDED | FCIP_POWER_DOWN))) {
2543 		mutex_exit(&fptr->fcip_mutex);
2544 		return (FC_UNCLAIMED);
2545 	}
2546 
2547 	/*
2548 	 * set fcip flags to indicate we are in the middle of a
2549 	 * ELS callback so we can wait till the statechange
2550 	 * is handled before succeeding/failing the SUSPEND/POWER DOWN.
2551 	 */
2552 	fptr->fcip_flags |= FCIP_IN_ELS_CB;
2553 	mutex_exit(&fptr->fcip_mutex);
2554 
2555 	FCIP_TNF_PROBE_2((fcip_els_cb, "fcip io", /* CSTYLED */,
2556 		tnf_string, msg, "ELS callback",
2557 		tnf_uint, instance, instance));
2558 
2559 	FCIP_DEBUG(FCIP_DEBUG_ELS,
2560 	    (CE_NOTE, "fcip%d, ELS callback , ", instance));
2561 
2562 	r_ctl = buf->ub_frame.r_ctl;
2563 	switch (r_ctl & R_CTL_ROUTING) {
2564 	case R_CTL_EXTENDED_SVC:
2565 		if (r_ctl == R_CTL_ELS_REQ) {
2566 			ls_code = buf->ub_buffer[0];
2567 			if (ls_code == LA_ELS_FARP_REQ) {
2568 				/*
2569 				 * Inbound FARP broadcast request
2570 				 */
2571 				if (buf->ub_bufsize != sizeof (la_els_farp_t)) {
2572 					FCIP_DEBUG(FCIP_DEBUG_ELS, (CE_WARN,
2573 					    "Invalid FARP req buffer size "
2574 					    "expected 0x%lx, got 0x%x",
2575 					    (long)(sizeof (la_els_farp_t)),
2576 					    buf->ub_bufsize));
2577 					rval = FC_UNCLAIMED;
2578 					goto els_cb_done;
2579 				}
2580 				fcmd = (la_els_farp_t *)buf;
2581 				if (fcip_wwn_compare(&fcmd->resp_nwwn,
2582 				    &fport->fcipp_nwwn,
2583 				    FCIP_COMPARE_NWWN) != 0) {
2584 					rval = FC_UNCLAIMED;
2585 					goto els_cb_done;
2586 				}
2587 				/*
2588 				 * copy the FARP request and release the
2589 				 * unsolicited buffer
2590 				 */
2591 				fcmd = &farp_cmd;
2592 				bcopy((void *)buf, (void *)fcmd,
2593 				    sizeof (la_els_farp_t));
2594 				(void) fc_ulp_ubrelease(fport->fcipp_handle, 1,
2595 				    &buf->ub_token);
2596 
2597 				if (fcip_farp_supported &&
2598 				    fcip_handle_farp_request(fptr, fcmd) ==
2599 				    FC_SUCCESS) {
2600 					/*
2601 					 * We successfully sent out a FARP
2602 					 * reply to the requesting port
2603 					 */
2604 					rval = FC_SUCCESS;
2605 					goto els_cb_done;
2606 				} else {
2607 					rval = FC_UNCLAIMED;
2608 					goto els_cb_done;
2609 				}
2610 			}
2611 		} else if (r_ctl == R_CTL_ELS_RSP) {
2612 			ls_code = buf->ub_buffer[0];
2613 			if (ls_code == LA_ELS_FARP_REPLY) {
2614 				/*
2615 				 * We received a REPLY to our FARP request
2616 				 */
2617 				if (buf->ub_bufsize != sizeof (la_els_farp_t)) {
2618 					FCIP_DEBUG(FCIP_DEBUG_ELS, (CE_WARN,
2619 					    "Invalid FARP req buffer size "
2620 					    "expected 0x%lx, got 0x%x",
2621 					    (long)(sizeof (la_els_farp_t)),
2622 					    buf->ub_bufsize));
2623 					rval = FC_UNCLAIMED;
2624 					goto els_cb_done;
2625 				}
2626 				fcmd = &farp_cmd;
2627 				bcopy((void *)buf, (void *)fcmd,
2628 				    sizeof (la_els_farp_t));
2629 				(void) fc_ulp_ubrelease(fport->fcipp_handle, 1,
2630 				    &buf->ub_token);
2631 				if (fcip_farp_supported &&
2632 				    fcip_handle_farp_response(fptr, fcmd) ==
2633 				    FC_SUCCESS) {
2634 					FCIP_DEBUG(FCIP_DEBUG_ELS, (CE_NOTE,
2635 					    "Successfully recevied a FARP "
2636 					    "response"));
2637 					mutex_enter(&fptr->fcip_mutex);
2638 					fptr->fcip_farp_rsp_flag = 1;
2639 					cv_signal(&fptr->fcip_farp_cv);
2640 					mutex_exit(&fptr->fcip_mutex);
2641 					rval = FC_SUCCESS;
2642 					goto els_cb_done;
2643 				} else {
2644 					FCIP_DEBUG(FCIP_DEBUG_ELS, (CE_WARN,
2645 					    "Unable to handle a FARP response "
2646 					    "receive"));
2647 					rval = FC_UNCLAIMED;
2648 					goto els_cb_done;
2649 				}
2650 			}
2651 		}
2652 		break;
2653 	default:
2654 		break;
2655 	}
2656 els_cb_done:
2657 	mutex_enter(&fptr->fcip_mutex);
2658 	fptr->fcip_flags &= ~(FCIP_IN_ELS_CB);
2659 	mutex_exit(&fptr->fcip_mutex);
2660 	return (rval);
2661 }
2662 
2663 
2664 /*
2665  * Handle inbound FARP requests
2666  */
2667 static int
fcip_handle_farp_request(struct fcip * fptr,la_els_farp_t * fcmd)2668 fcip_handle_farp_request(struct fcip *fptr, la_els_farp_t *fcmd)
2669 {
2670 	fcip_pkt_t		*fcip_pkt;
2671 	fc_packet_t		*fc_pkt;
2672 	fcip_port_info_t	*fport = fptr->fcip_port_info;
2673 	int			rval = FC_FAILURE;
2674 	opaque_t		fca_dev;
2675 	fc_portmap_t 		map;
2676 	struct fcip_routing_table *frp;
2677 	struct fcip_dest *fdestp;
2678 
2679 	/*
2680 	 * Add an entry for the remote port into our routing and destination
2681 	 * tables.
2682 	 */
2683 	map.map_did = fcmd->req_id;
2684 	map.map_hard_addr.hard_addr = fcmd->req_id.port_id;
2685 	map.map_state = PORT_DEVICE_VALID;
2686 	map.map_type = PORT_DEVICE_NEW;
2687 	map.map_flags = 0;
2688 	map.map_pd = NULL;
2689 	bcopy((void *)&fcmd->req_pwwn, (void *)&map.map_pwwn,
2690 	    sizeof (la_wwn_t));
2691 	bcopy((void *)&fcmd->req_nwwn, (void *)&map.map_nwwn,
2692 	    sizeof (la_wwn_t));
2693 	fcip_rt_update(fptr, &map, 1);
2694 	mutex_enter(&fptr->fcip_rt_mutex);
2695 	frp = fcip_lookup_rtable(fptr, &fcmd->req_pwwn, FCIP_COMPARE_NWWN);
2696 	mutex_exit(&fptr->fcip_rt_mutex);
2697 
2698 	fdestp = fcip_add_dest(fptr, frp);
2699 
2700 	fcip_pkt = fcip_ipkt_alloc(fptr, sizeof (la_els_farp_t),
2701 	    sizeof (la_els_farp_t), NULL, KM_SLEEP);
2702 	if (fcip_pkt == NULL) {
2703 		rval = FC_FAILURE;
2704 		goto farp_done;
2705 	}
2706 	/*
2707 	 * Fill in our port's PWWN and NWWN
2708 	 */
2709 	fcmd->resp_pwwn = fport->fcipp_pwwn;
2710 	fcmd->resp_nwwn = fport->fcipp_nwwn;
2711 
2712 	fcip_init_unicast_pkt(fcip_pkt, fport->fcipp_sid,
2713 	    fcmd->req_id, NULL);
2714 
2715 	fca_dev =
2716 	    fc_ulp_get_fca_device(fport->fcipp_handle, fcmd->req_id);
2717 	fc_pkt = FCIP_PKT_TO_FC_PKT(fcip_pkt);
2718 	fc_pkt->pkt_cmd_fhdr.r_ctl = R_CTL_ELS_RSP;
2719 	fc_pkt->pkt_fca_device = fca_dev;
2720 	fcip_pkt->fcip_pkt_dest = fdestp;
2721 
2722 	/*
2723 	 * Attempt a PLOGI again
2724 	 */
2725 	if (fcmd->resp_flags & FARP_INIT_P_LOGI) {
2726 		if (fcip_do_plogi(fptr, frp) != FC_SUCCESS) {
2727 			/*
2728 			 * Login to the remote port failed. There is no
2729 			 * point continuing with the FARP request further
2730 			 * so bail out here.
2731 			 */
2732 			frp->fcipr_state = PORT_DEVICE_INVALID;
2733 			rval = FC_FAILURE;
2734 			goto farp_done;
2735 		}
2736 	}
2737 
2738 	FCIP_CP_OUT(fcmd, fc_pkt->pkt_cmd, fc_pkt->pkt_cmd_acc,
2739 	    sizeof (la_els_farp_t));
2740 
2741 	rval = fc_ulp_issue_els(fport->fcipp_handle, fc_pkt);
2742 	if (rval != FC_SUCCESS) {
2743 		FCIP_TNF_PROBE_2((fcip_handle_farp_request, "fcip io",
2744 		    /* CSTYLED */, tnf_string, msg,
2745 		    "fcip_transport of farp reply failed",
2746 		    tnf_uint, rval, rval));
2747 		FCIP_DEBUG(FCIP_DEBUG_ELS, (CE_WARN,
2748 		    "fcip_transport of farp reply failed 0x%x", rval));
2749 	}
2750 
2751 farp_done:
2752 	return (rval);
2753 }
2754 
2755 
2756 /*
2757  * Handle FARP responses to our FARP requests. When we receive a FARP
2758  * reply, we need to add the entry for the Port that replied into our
2759  * routing and destination hash tables. It is possible that the remote
2760  * port did not login into us (FARP responses can be received without
2761  * a PLOGI)
2762  */
2763 static int
fcip_handle_farp_response(struct fcip * fptr,la_els_farp_t * fcmd)2764 fcip_handle_farp_response(struct fcip *fptr, la_els_farp_t *fcmd)
2765 {
2766 	int			rval = FC_FAILURE;
2767 	fc_portmap_t 		map;
2768 	struct fcip_routing_table *frp;
2769 	struct fcip_dest *fdestp;
2770 
2771 	/*
2772 	 * Add an entry for the remote port into our routing and destination
2773 	 * tables.
2774 	 */
2775 	map.map_did = fcmd->dest_id;
2776 	map.map_hard_addr.hard_addr = fcmd->dest_id.port_id;
2777 	map.map_state = PORT_DEVICE_VALID;
2778 	map.map_type = PORT_DEVICE_NEW;
2779 	map.map_flags = 0;
2780 	map.map_pd = NULL;
2781 	bcopy((void *)&fcmd->resp_pwwn, (void *)&map.map_pwwn,
2782 	    sizeof (la_wwn_t));
2783 	bcopy((void *)&fcmd->resp_nwwn, (void *)&map.map_nwwn,
2784 	    sizeof (la_wwn_t));
2785 	fcip_rt_update(fptr, &map, 1);
2786 	mutex_enter(&fptr->fcip_rt_mutex);
2787 	frp = fcip_lookup_rtable(fptr, &fcmd->resp_pwwn, FCIP_COMPARE_NWWN);
2788 	mutex_exit(&fptr->fcip_rt_mutex);
2789 
2790 	fdestp = fcip_add_dest(fptr, frp);
2791 
2792 	if (fdestp != NULL) {
2793 		rval = FC_SUCCESS;
2794 	}
2795 	return (rval);
2796 }
2797 
2798 
2799 #define	FCIP_HDRS_LENGTH	\
2800 	sizeof (fcph_network_hdr_t)+sizeof (llc_snap_hdr_t)+sizeof (ipha_t)
2801 
2802 /*
2803  * fcip_data_cb is the heart of most IP operations. This routine is called
2804  * by the transport when any unsolicited IP data arrives at a port (which
2805  * is almost all IP data). This routine then strips off the Network header
2806  * from the payload (after authenticating the received payload ofcourse),
2807  * creates a message blk and sends the data upstream. You will see ugly
2808  * #defines because of problems with using esballoc() as opposed to
2809  * allocb to prevent an extra copy of data. We should probably move to
2810  * esballoc entirely when the MTU eventually will be larger than 1500 bytes
2811  * since copies will get more expensive then. At 1500 byte MTUs, there is
2812  * no noticable difference between using allocb and esballoc. The other
2813  * caveat is that the qlc firmware still cannot tell us accurately the
2814  * no. of valid bytes in the unsol buffer it DMA'ed so we have to resort
2815  * to looking into the IP header and hoping that the no. of bytes speficified
2816  * in the header was actually received.
2817  */
2818 /* ARGSUSED */
2819 static int
fcip_data_cb(opaque_t ulp_handle,opaque_t phandle,fc_unsol_buf_t * buf,uint32_t claimed)2820 fcip_data_cb(opaque_t ulp_handle, opaque_t phandle,
2821     fc_unsol_buf_t *buf, uint32_t claimed)
2822 {
2823 	fcip_port_info_t		*fport;
2824 	struct fcip 			*fptr;
2825 	fcph_network_hdr_t		*nhdr;
2826 	llc_snap_hdr_t			*snaphdr;
2827 	mblk_t				*bp;
2828 	uint32_t 			len;
2829 	uint32_t			hdrlen;
2830 	ushort_t			type;
2831 	ipha_t				*iphdr;
2832 	int				rval;
2833 
2834 #ifdef FCIP_ESBALLOC
2835 	frtn_t				*free_ubuf;
2836 	struct fcip_esballoc_arg	*fesb_argp;
2837 #endif /* FCIP_ESBALLOC */
2838 
2839 	fport = fcip_get_port(phandle);
2840 	if (fport == NULL) {
2841 		return (FC_UNCLAIMED);
2842 	}
2843 
2844 	fptr = fport->fcipp_fcip;
2845 	ASSERT(fptr != NULL);
2846 
2847 	if (fptr == NULL) {
2848 		return (FC_UNCLAIMED);
2849 	}
2850 
2851 	mutex_enter(&fptr->fcip_mutex);
2852 	if ((fptr->fcip_flags & (FCIP_DETACHING | FCIP_DETACHED)) ||
2853 	    (fptr->fcip_flags & (FCIP_SUSPENDED | FCIP_POWER_DOWN))) {
2854 		mutex_exit(&fptr->fcip_mutex);
2855 		rval = FC_UNCLAIMED;
2856 		goto data_cb_done;
2857 	}
2858 
2859 	/*
2860 	 * set fcip flags to indicate we are in the middle of a
2861 	 * data callback so we can wait till the statechange
2862 	 * is handled before succeeding/failing the SUSPEND/POWER DOWN.
2863 	 */
2864 	fptr->fcip_flags |= FCIP_IN_DATA_CB;
2865 	mutex_exit(&fptr->fcip_mutex);
2866 
2867 	FCIP_TNF_PROBE_2((fcip_data_cb, "fcip io", /* CSTYLED */,
2868 		tnf_string, msg, "data callback",
2869 		tnf_int, instance, ddi_get_instance(fport->fcipp_dip)));
2870 	FCIP_DEBUG(FCIP_DEBUG_UPSTREAM,
2871 	    (CE_NOTE, "fcip%d, data callback",
2872 	    ddi_get_instance(fport->fcipp_dip)));
2873 
2874 	/*
2875 	 * get to the network and snap headers in the payload
2876 	 */
2877 	nhdr = (fcph_network_hdr_t *)buf->ub_buffer;
2878 	snaphdr = (llc_snap_hdr_t *)(buf->ub_buffer +
2879 	    sizeof (fcph_network_hdr_t));
2880 
2881 	hdrlen = sizeof (fcph_network_hdr_t) + sizeof (llc_snap_hdr_t);
2882 
2883 	/*
2884 	 * get the IP header to obtain the no. of bytes we need to read
2885 	 * off from the unsol buffer. This obviously is because not all
2886 	 * data fills up the unsol buffer completely and the firmware
2887 	 * doesn't tell us how many valid bytes are in there as well
2888 	 */
2889 	iphdr = (ipha_t *)(buf->ub_buffer + hdrlen);
2890 	snaphdr->pid = BE_16(snaphdr->pid);
2891 	type = snaphdr->pid;
2892 
2893 	FCIP_DEBUG(FCIP_DEBUG_UPSTREAM,
2894 	    (CE_CONT, "SNAPHDR: dsap %x, ssap %x, ctrl %x\n",
2895 	    snaphdr->dsap, snaphdr->ssap, snaphdr->ctrl));
2896 
2897 	FCIP_DEBUG(FCIP_DEBUG_UPSTREAM,
2898 	    (CE_CONT, "oui[0] 0x%x oui[1] 0x%x oui[2] 0x%x pid 0x%x\n",
2899 	    snaphdr->oui[0], snaphdr->oui[1], snaphdr->oui[2], snaphdr->pid));
2900 
2901 	/* Authneticate, Authenticate */
2902 	if (type == ETHERTYPE_IP) {
2903 		len = hdrlen + BE_16(iphdr->ipha_length);
2904 	} else if (type == ETHERTYPE_ARP) {
2905 		len = hdrlen + 28;
2906 	} else {
2907 		len = buf->ub_bufsize;
2908 	}
2909 
2910 	FCIP_DEBUG(FCIP_DEBUG_UPSTREAM,
2911 	    (CE_CONT, "effective packet length is %d bytes.\n", len));
2912 
2913 	if (len < hdrlen || len > FCIP_UB_SIZE) {
2914 		FCIP_DEBUG(FCIP_DEBUG_UPSTREAM,
2915 		    (CE_NOTE, "Incorrect buffer size %d bytes", len));
2916 		rval = FC_UNCLAIMED;
2917 		goto data_cb_done;
2918 	}
2919 
2920 	if (buf->ub_frame.type != FC_TYPE_IS8802_SNAP) {
2921 		FCIP_DEBUG(FCIP_DEBUG_UPSTREAM, (CE_NOTE, "Not IP/ARP data"));
2922 		rval = FC_UNCLAIMED;
2923 		goto data_cb_done;
2924 	}
2925 
2926 	FCIP_DEBUG(FCIP_DEBUG_UPSTREAM, (CE_NOTE, "checking wwn"));
2927 
2928 	if ((fcip_wwn_compare(&nhdr->net_dest_addr, &fport->fcipp_pwwn,
2929 	    FCIP_COMPARE_NWWN) != 0) &&
2930 	    (!IS_BROADCAST_ADDR(&nhdr->net_dest_addr))) {
2931 		rval = FC_UNCLAIMED;
2932 		goto data_cb_done;
2933 	} else if (fcip_cache_on_arp_broadcast &&
2934 	    IS_BROADCAST_ADDR(&nhdr->net_dest_addr)) {
2935 		fcip_cache_arp_broadcast(fptr, buf);
2936 	}
2937 
2938 	FCIP_DEBUG(FCIP_DEBUG_UPSTREAM, (CE_NOTE, "Allocate streams block"));
2939 
2940 	/*
2941 	 * Using esballoc instead of allocb should be faster, atleast at
2942 	 * larger MTUs than 1500 bytes. Someday we'll get there :)
2943 	 */
2944 #if defined(FCIP_ESBALLOC)
2945 	/*
2946 	 * allocate memory for the frtn function arg. The Function
2947 	 * (fcip_ubfree) arg is a struct fcip_esballoc_arg type
2948 	 * which contains pointers to the unsol buffer and the
2949 	 * opaque port handle for releasing the unsol buffer back to
2950 	 * the FCA for reuse
2951 	 */
2952 	fesb_argp = (struct fcip_esballoc_arg *)
2953 	    kmem_zalloc(sizeof (struct fcip_esballoc_arg), KM_NOSLEEP);
2954 
2955 	if (fesb_argp == NULL) {
2956 		FCIP_DEBUG(FCIP_DEBUG_UPSTREAM,
2957 		    (CE_WARN, "esballoc of mblk failed in data_cb"));
2958 		rval = FC_UNCLAIMED;
2959 		goto data_cb_done;
2960 	}
2961 	/*
2962 	 * Check with KM_NOSLEEP
2963 	 */
2964 	free_ubuf = (frtn_t *)kmem_zalloc(sizeof (frtn_t), KM_NOSLEEP);
2965 	if (free_ubuf == NULL) {
2966 		kmem_free(fesb_argp, sizeof (struct fcip_esballoc_arg));
2967 		FCIP_DEBUG(FCIP_DEBUG_UPSTREAM,
2968 		    (CE_WARN, "esballoc of mblk failed in data_cb"));
2969 		rval = FC_UNCLAIMED;
2970 		goto data_cb_done;
2971 	}
2972 
2973 	fesb_argp->frtnp = free_ubuf;
2974 	fesb_argp->buf = buf;
2975 	fesb_argp->phandle = phandle;
2976 	free_ubuf->free_func = fcip_ubfree;
2977 	free_ubuf->free_arg = (char *)fesb_argp;
2978 	if ((bp = (mblk_t *)esballoc((unsigned char *)buf->ub_buffer,
2979 	    len, BPRI_MED, free_ubuf)) == NULL) {
2980 		kmem_free(fesb_argp, sizeof (struct fcip_esballoc_arg));
2981 		kmem_free(free_ubuf, sizeof (frtn_t));
2982 		FCIP_DEBUG(FCIP_DEBUG_UPSTREAM,
2983 		    (CE_WARN, "esballoc of mblk failed in data_cb"));
2984 		rval = FC_UNCLAIMED;
2985 		goto data_cb_done;
2986 	}
2987 #elif !defined(FCIP_ESBALLOC)
2988 	/*
2989 	 * allocate streams mblk and copy the contents of the
2990 	 * unsolicited buffer into this newly alloc'ed mblk
2991 	 */
2992 	if ((bp = (mblk_t *)fcip_allocb((size_t)len, BPRI_LO)) == NULL) {
2993 		FCIP_DEBUG(FCIP_DEBUG_UPSTREAM,
2994 		    (CE_WARN, "alloc of mblk failed in data_cb"));
2995 		rval = FC_UNCLAIMED;
2996 		goto data_cb_done;
2997 	}
2998 
2999 	/*
3000 	 * Unsolicited buffers handed up to us from the FCA must be
3001 	 * endian clean so just bcopy the data into our mblk. Else
3002 	 * we may have to either copy the data byte by byte or
3003 	 * use the ddi_rep_get* routines to do the copy for us.
3004 	 */
3005 	bcopy(buf->ub_buffer, bp->b_rptr, len);
3006 
3007 	/*
3008 	 * for esballoc'ed mblks - free the UB in the frtn function
3009 	 * along with the memory allocated for the function arg.
3010 	 * for allocb'ed mblk - release the unsolicited buffer here
3011 	 */
3012 	(void) fc_ulp_ubrelease(phandle, 1, &buf->ub_token);
3013 
3014 #endif	/* FCIP_ESBALLOC */
3015 
3016 	bp->b_wptr = bp->b_rptr + len;
3017 	fptr->fcip_ipackets++;
3018 
3019 	if (type == ETHERTYPE_IP) {
3020 		mutex_enter(&fptr->fcip_mutex);
3021 		fptr->fcip_ub_upstream++;
3022 		mutex_exit(&fptr->fcip_mutex);
3023 		bp->b_rptr += hdrlen;
3024 
3025 		/*
3026 		 * Check if ipq is valid in the sendup thread
3027 		 */
3028 		if (fcip_sendup_alloc_enque(fptr, bp, NULL) != FC_SUCCESS) {
3029 			freemsg(bp);
3030 		}
3031 	} else {
3032 		/*
3033 		 * We won't get ethernet 802.3 packets in FCIP but we may get
3034 		 * types other than ETHERTYPE_IP, such as ETHERTYPE_ARP. Let
3035 		 * fcip_sendup() do the matching.
3036 		 */
3037 		mutex_enter(&fptr->fcip_mutex);
3038 		fptr->fcip_ub_upstream++;
3039 		mutex_exit(&fptr->fcip_mutex);
3040 		if (fcip_sendup_alloc_enque(fptr, bp,
3041 		    fcip_accept) != FC_SUCCESS) {
3042 			freemsg(bp);
3043 		}
3044 	}
3045 
3046 	rval = FC_SUCCESS;
3047 
3048 	/*
3049 	 * Unset fcip_flags to indicate we are out of callback and return
3050 	 */
3051 data_cb_done:
3052 	mutex_enter(&fptr->fcip_mutex);
3053 	fptr->fcip_flags &= ~(FCIP_IN_DATA_CB);
3054 	mutex_exit(&fptr->fcip_mutex);
3055 	return (rval);
3056 }
3057 
3058 #if !defined(FCIP_ESBALLOC)
3059 /*
3060  * Allocate a message block for the inbound data to be sent upstream.
3061  */
3062 static void *
fcip_allocb(size_t size,uint_t pri)3063 fcip_allocb(size_t size, uint_t pri)
3064 {
3065 	mblk_t	*mp;
3066 
3067 	if ((mp = allocb(size, pri)) == NULL) {
3068 		return (NULL);
3069 	}
3070 	return (mp);
3071 }
3072 
3073 #endif
3074 
3075 /*
3076  * This helper routine kmem cache alloc's a sendup element for enquing
3077  * into the sendup list for callbacks upstream from the dedicated sendup
3078  * thread. We enque the msg buf into the sendup list and cv_signal the
3079  * sendup thread to finish the callback for us.
3080  */
3081 static int
fcip_sendup_alloc_enque(struct fcip * fptr,mblk_t * mp,struct fcipstr * (* f)())3082 fcip_sendup_alloc_enque(struct fcip *fptr, mblk_t *mp, struct fcipstr *(*f)())
3083 {
3084 	struct fcip_sendup_elem 	*msg_elem;
3085 	int				rval = FC_FAILURE;
3086 
3087 	FCIP_TNF_PROBE_1((fcip_sendup_alloc_enque, "fcip io", /* CSTYLED */,
3088 		tnf_string, msg, "sendup msg enque"));
3089 	msg_elem = kmem_cache_alloc(fptr->fcip_sendup_cache, KM_NOSLEEP);
3090 	if (msg_elem == NULL) {
3091 		/* drop pkt to floor - update stats */
3092 		rval = FC_FAILURE;
3093 		goto sendup_alloc_done;
3094 	}
3095 	msg_elem->fcipsu_mp = mp;
3096 	msg_elem->fcipsu_func = f;
3097 
3098 	mutex_enter(&fptr->fcip_sendup_mutex);
3099 	if (fptr->fcip_sendup_head == NULL) {
3100 		fptr->fcip_sendup_head = fptr->fcip_sendup_tail = msg_elem;
3101 	} else {
3102 		fptr->fcip_sendup_tail->fcipsu_next = msg_elem;
3103 		fptr->fcip_sendup_tail = msg_elem;
3104 	}
3105 	fptr->fcip_sendup_cnt++;
3106 	cv_signal(&fptr->fcip_sendup_cv);
3107 	mutex_exit(&fptr->fcip_sendup_mutex);
3108 	rval = FC_SUCCESS;
3109 
3110 sendup_alloc_done:
3111 	return (rval);
3112 }
3113 
3114 /*
3115  * One of the ways of performing the WWN to D_ID mapping required for
3116  * IPFC data is to cache the unsolicited ARP broadcast messages received
3117  * and update the routing table to add entry for the destination port
3118  * if we are the intended recipient of the ARP broadcast message. This is
3119  * one of the methods recommended in the rfc to obtain the WWN to D_ID mapping
3120  * but is not typically used unless enabled. The driver prefers to use the
3121  * nameserver/lilp map to obtain this mapping.
3122  */
3123 static void
fcip_cache_arp_broadcast(struct fcip * fptr,fc_unsol_buf_t * buf)3124 fcip_cache_arp_broadcast(struct fcip *fptr, fc_unsol_buf_t *buf)
3125 {
3126 	fcip_port_info_t		*fport;
3127 	fcph_network_hdr_t		*nhdr;
3128 	struct fcip_routing_table	*frp;
3129 	fc_portmap_t			map;
3130 
3131 	fport = fptr->fcip_port_info;
3132 	if (fport == NULL) {
3133 		return;
3134 	}
3135 	ASSERT(fport != NULL);
3136 
3137 	nhdr = (fcph_network_hdr_t *)buf->ub_buffer;
3138 
3139 	mutex_enter(&fptr->fcip_rt_mutex);
3140 	frp = fcip_lookup_rtable(fptr, &nhdr->net_src_addr, FCIP_COMPARE_NWWN);
3141 	mutex_exit(&fptr->fcip_rt_mutex);
3142 	if (frp == NULL) {
3143 		map.map_did.port_id = buf->ub_frame.s_id;
3144 		map.map_hard_addr.hard_addr = buf->ub_frame.s_id;
3145 		map.map_state = PORT_DEVICE_VALID;
3146 		map.map_type = PORT_DEVICE_NEW;
3147 		map.map_flags = 0;
3148 		map.map_pd = NULL;
3149 		bcopy((void *)&nhdr->net_src_addr, (void *)&map.map_pwwn,
3150 		    sizeof (la_wwn_t));
3151 		bcopy((void *)&nhdr->net_src_addr, (void *)&map.map_nwwn,
3152 		    sizeof (la_wwn_t));
3153 		fcip_rt_update(fptr, &map, 1);
3154 		mutex_enter(&fptr->fcip_rt_mutex);
3155 		frp = fcip_lookup_rtable(fptr, &nhdr->net_src_addr,
3156 		    FCIP_COMPARE_NWWN);
3157 		mutex_exit(&fptr->fcip_rt_mutex);
3158 
3159 		(void) fcip_add_dest(fptr, frp);
3160 	}
3161 
3162 }
3163 
3164 /*
3165  * This is a dedicated thread to do callbacks from fcip's data callback
3166  * routines into the modules upstream. The reason for this thread is
3167  * the data callback function can be called from an interrupt context and
3168  * the upstream modules *can* make calls downstream in the same thread
3169  * context. If the call is to a fabric port which is not yet in our
3170  * routing tables, we may have to query the nameserver/fabric for the
3171  * MAC addr to Port_ID mapping which may be blocking calls.
3172  */
3173 static void
fcip_sendup_thr(void * arg)3174 fcip_sendup_thr(void *arg)
3175 {
3176 	struct fcip		*fptr = (struct fcip *)arg;
3177 	struct fcip_sendup_elem	*msg_elem;
3178 	queue_t			*ip4q = NULL;
3179 
3180 	CALLB_CPR_INIT(&fptr->fcip_cpr_info, &fptr->fcip_sendup_mutex,
3181 	    callb_generic_cpr, "fcip_sendup_thr");
3182 
3183 	mutex_enter(&fptr->fcip_sendup_mutex);
3184 	for (;;) {
3185 
3186 		while (fptr->fcip_sendup_thr_initted &&
3187 		    fptr->fcip_sendup_head == NULL) {
3188 			CALLB_CPR_SAFE_BEGIN(&fptr->fcip_cpr_info);
3189 			cv_wait(&fptr->fcip_sendup_cv,
3190 			    &fptr->fcip_sendup_mutex);
3191 			CALLB_CPR_SAFE_END(&fptr->fcip_cpr_info,
3192 			    &fptr->fcip_sendup_mutex);
3193 		}
3194 
3195 		if (fptr->fcip_sendup_thr_initted == 0) {
3196 			break;
3197 		}
3198 
3199 		FCIP_TNF_PROBE_1((fcip_sendup_thr, "fcip io", /* CSTYLED */,
3200 		    tnf_string, msg, "fcip sendup thr - new msg"));
3201 
3202 		msg_elem = fptr->fcip_sendup_head;
3203 		fptr->fcip_sendup_head = msg_elem->fcipsu_next;
3204 		msg_elem->fcipsu_next = NULL;
3205 		mutex_exit(&fptr->fcip_sendup_mutex);
3206 
3207 		if (msg_elem->fcipsu_func == NULL) {
3208 			/*
3209 			 * Message for ipq. Check to see if the ipq is
3210 			 * is still valid. Since the thread is asynchronous,
3211 			 * there could have been a close on the stream
3212 			 */
3213 			mutex_enter(&fptr->fcip_mutex);
3214 			if (fptr->fcip_ipq && canputnext(fptr->fcip_ipq)) {
3215 				ip4q = fptr->fcip_ipq;
3216 				mutex_exit(&fptr->fcip_mutex);
3217 				putnext(ip4q, msg_elem->fcipsu_mp);
3218 			} else {
3219 				mutex_exit(&fptr->fcip_mutex);
3220 				freemsg(msg_elem->fcipsu_mp);
3221 			}
3222 		} else {
3223 			fcip_sendup(fptr, msg_elem->fcipsu_mp,
3224 			    msg_elem->fcipsu_func);
3225 		}
3226 
3227 #if !defined(FCIP_ESBALLOC)
3228 		/*
3229 		 * for allocb'ed mblk - decrement upstream count here
3230 		 */
3231 		mutex_enter(&fptr->fcip_mutex);
3232 		ASSERT(fptr->fcip_ub_upstream > 0);
3233 		fptr->fcip_ub_upstream--;
3234 		mutex_exit(&fptr->fcip_mutex);
3235 #endif /* FCIP_ESBALLOC */
3236 
3237 		kmem_cache_free(fptr->fcip_sendup_cache, (void *)msg_elem);
3238 		mutex_enter(&fptr->fcip_sendup_mutex);
3239 		fptr->fcip_sendup_cnt--;
3240 	}
3241 
3242 
3243 #ifndef	__lock_lint
3244 	CALLB_CPR_EXIT(&fptr->fcip_cpr_info);
3245 #else
3246 	mutex_exit(&fptr->fcip_sendup_mutex);
3247 #endif /* __lock_lint */
3248 
3249 	/* Wake up fcip detach thread by the end */
3250 	cv_signal(&fptr->fcip_sendup_cv);
3251 
3252 	thread_exit();
3253 }
3254 
3255 #ifdef FCIP_ESBALLOC
3256 
3257 /*
3258  * called from the stream head when it is done using an unsolicited buffer.
3259  * We release this buffer then to the FCA for reuse.
3260  */
3261 static void
fcip_ubfree(char * arg)3262 fcip_ubfree(char *arg)
3263 {
3264 	struct fcip_esballoc_arg *fesb_argp = (struct fcip_esballoc_arg *)arg;
3265 	fc_unsol_buf_t	*ubuf;
3266 	frtn_t		*frtnp;
3267 	fcip_port_info_t		*fport;
3268 	struct fcip 			*fptr;
3269 
3270 
3271 	fport = fcip_get_port(fesb_argp->phandle);
3272 	fptr = fport->fcipp_fcip;
3273 
3274 	ASSERT(fesb_argp != NULL);
3275 	ubuf = fesb_argp->buf;
3276 	frtnp = fesb_argp->frtnp;
3277 
3278 
3279 	FCIP_DEBUG(FCIP_DEBUG_UPSTREAM,
3280 	    (CE_WARN, "freeing ubuf after esballoc in fcip_ubfree"));
3281 	(void) fc_ulp_ubrelease(fesb_argp->phandle, 1, &ubuf->ub_token);
3282 
3283 	mutex_enter(&fptr->fcip_mutex);
3284 	ASSERT(fptr->fcip_ub_upstream > 0);
3285 	fptr->fcip_ub_upstream--;
3286 	cv_signal(&fptr->fcip_ub_cv);
3287 	mutex_exit(&fptr->fcip_mutex);
3288 
3289 	kmem_free(frtnp, sizeof (frtn_t));
3290 	kmem_free(fesb_argp, sizeof (struct fcip_esballoc_arg));
3291 }
3292 
3293 #endif /* FCIP_ESBALLOC */
3294 
3295 /*
3296  * handle data other than that of type ETHERTYPE_IP and send it on its
3297  * way upstream to the right streams module to handle
3298  */
3299 static void
fcip_sendup(struct fcip * fptr,mblk_t * mp,struct fcipstr * (* acceptfunc)())3300 fcip_sendup(struct fcip *fptr, mblk_t *mp, struct fcipstr *(*acceptfunc)())
3301 {
3302 	struct fcipstr	*slp, *nslp;
3303 	la_wwn_t	*dhostp;
3304 	mblk_t		*nmp;
3305 	uint32_t 	isgroupaddr;
3306 	int 		type;
3307 	uint32_t	hdrlen;
3308 	fcph_network_hdr_t	*nhdr;
3309 	llc_snap_hdr_t		*snaphdr;
3310 
3311 	FCIP_TNF_PROBE_1((fcip_sendup, "fcip io", /* CSTYLED */,
3312 		tnf_string, msg, "fcip sendup"));
3313 	nhdr = (fcph_network_hdr_t *)mp->b_rptr;
3314 	snaphdr =
3315 	    (llc_snap_hdr_t *)(mp->b_rptr + sizeof (fcph_network_hdr_t));
3316 	dhostp = &nhdr->net_dest_addr;
3317 	type = snaphdr->pid;
3318 	hdrlen = sizeof (fcph_network_hdr_t) + sizeof (llc_snap_hdr_t);
3319 
3320 	/* No group address with fibre channel */
3321 	isgroupaddr = 0;
3322 
3323 	/*
3324 	 * While holding a reader lock on the linked list of streams structures,
3325 	 * attempt to match the address criteria for each stream
3326 	 * and pass up the raw M_DATA ("fastpath") or a DL_UNITDATA_IND.
3327 	 */
3328 
3329 	rw_enter(&fcipstruplock, RW_READER);
3330 
3331 	if ((slp = (*acceptfunc)(fcipstrup, fptr, type, dhostp)) == NULL) {
3332 		rw_exit(&fcipstruplock);
3333 		FCIP_TNF_PROBE_1((fcip_sendup, "fcip io", /* CSTYLED */,
3334 		    tnf_string, msg, "fcip sendup - no slp"));
3335 		freemsg(mp);
3336 		return;
3337 	}
3338 
3339 	/*
3340 	 * Loop on matching open streams until (*acceptfunc)() returns NULL.
3341 	 */
3342 	for (; nslp = (*acceptfunc)(slp->sl_nextp, fptr, type, dhostp);
3343 	    slp = nslp) {
3344 		if (canputnext(slp->sl_rq)) {
3345 			if (nmp = dupmsg(mp)) {
3346 				if ((slp->sl_flags & FCIP_SLFAST) &&
3347 							!isgroupaddr) {
3348 					nmp->b_rptr += hdrlen;
3349 					putnext(slp->sl_rq, nmp);
3350 				} else if (slp->sl_flags & FCIP_SLRAW) {
3351 					/* No headers when FCIP_SLRAW is set */
3352 					putnext(slp->sl_rq, nmp);
3353 				} else if ((nmp = fcip_addudind(fptr, nmp,
3354 				    nhdr, type))) {
3355 					putnext(slp->sl_rq, nmp);
3356 				}
3357 			}
3358 		}
3359 	}
3360 
3361 	/*
3362 	 * Do the last one.
3363 	 */
3364 	if (canputnext(slp->sl_rq)) {
3365 		if (slp->sl_flags & FCIP_SLFAST) {
3366 			mp->b_rptr += hdrlen;
3367 			putnext(slp->sl_rq, mp);
3368 		} else if (slp->sl_flags & FCIP_SLRAW) {
3369 			putnext(slp->sl_rq, mp);
3370 		} else if ((mp = fcip_addudind(fptr, mp, nhdr, type))) {
3371 			putnext(slp->sl_rq, mp);
3372 		}
3373 	} else {
3374 		freemsg(mp);
3375 	}
3376 	FCIP_TNF_PROBE_1((fcip_sendup, "fcip io", /* CSTYLED */,
3377 	    tnf_string, msg, "fcip sendup done"));
3378 
3379 	rw_exit(&fcipstruplock);
3380 }
3381 
3382 /*
3383  * Match the stream based on type and wwn if necessary.
3384  * Destination wwn dhostp is passed to this routine is reserved
3385  * for future usage. We don't need to use it right now since port
3386  * to fcip instance mapping is unique and wwn is already validated when
3387  * packet comes to fcip.
3388  */
3389 /* ARGSUSED */
3390 static struct fcipstr *
fcip_accept(struct fcipstr * slp,struct fcip * fptr,int type,la_wwn_t * dhostp)3391 fcip_accept(struct fcipstr *slp, struct fcip *fptr, int type, la_wwn_t *dhostp)
3392 {
3393 	t_uscalar_t 	sap;
3394 
3395 	FCIP_TNF_PROBE_1((fcip_accept, "fcip io", /* CSTYLED */,
3396 	    tnf_string, msg, "fcip accept"));
3397 
3398 	for (; slp; slp = slp->sl_nextp) {
3399 		sap = slp->sl_sap;
3400 		FCIP_DEBUG(FCIP_DEBUG_UPSTREAM, (CE_CONT,
3401 		    "fcip_accept: checking next sap = %x, type = %x",
3402 		    sap, type));
3403 
3404 		if ((slp->sl_fcip == fptr) && (type == sap)) {
3405 			return (slp);
3406 		}
3407 	}
3408 	return (NULL);
3409 }
3410 
3411 /*
3412  * Handle DL_UNITDATA_IND messages
3413  */
3414 static mblk_t *
fcip_addudind(struct fcip * fptr,mblk_t * mp,fcph_network_hdr_t * nhdr,int type)3415 fcip_addudind(struct fcip *fptr, mblk_t *mp, fcph_network_hdr_t *nhdr,
3416     int type)
3417 {
3418 	dl_unitdata_ind_t	*dludindp;
3419 	struct	fcipdladdr	*dlap;
3420 	mblk_t	*nmp;
3421 	int	size;
3422 	uint32_t hdrlen;
3423 	struct ether_addr	src_addr;
3424 	struct ether_addr	dest_addr;
3425 
3426 
3427 	hdrlen = (sizeof (llc_snap_hdr_t) + sizeof (fcph_network_hdr_t));
3428 	mp->b_rptr += hdrlen;
3429 
3430 	FCIP_TNF_PROBE_1((fcip_addudind, "fcip io", /* CSTYLED */,
3431 	    tnf_string, msg, "fcip addudind"));
3432 
3433 	/*
3434 	 * Allocate an M_PROTO mblk for the DL_UNITDATA_IND.
3435 	 */
3436 	size = sizeof (dl_unitdata_ind_t) + FCIPADDRL + FCIPADDRL;
3437 	if ((nmp = allocb(size, BPRI_LO)) == NULL) {
3438 		fptr->fcip_allocbfail++;
3439 		freemsg(mp);
3440 		return (NULL);
3441 	}
3442 	DB_TYPE(nmp) = M_PROTO;
3443 	nmp->b_wptr = nmp->b_datap->db_lim;
3444 	nmp->b_rptr = nmp->b_wptr - size;
3445 
3446 	/*
3447 	 * Construct a DL_UNITDATA_IND primitive.
3448 	 */
3449 	dludindp = (dl_unitdata_ind_t *)nmp->b_rptr;
3450 	dludindp->dl_primitive = DL_UNITDATA_IND;
3451 	dludindp->dl_dest_addr_length = FCIPADDRL;
3452 	dludindp->dl_dest_addr_offset = sizeof (dl_unitdata_ind_t);
3453 	dludindp->dl_src_addr_length = FCIPADDRL;
3454 	dludindp->dl_src_addr_offset = sizeof (dl_unitdata_ind_t) + FCIPADDRL;
3455 	dludindp->dl_group_addre