1 /*
2 * Copyright (c) 2012-2014 The DragonFly Project. All rights reserved.
3 *
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@dragonflybsd.org>
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 *
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
16 * distribution.
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 */
34 /*
35 * This module allows disk devices to be created and associated with a
36 * communications pipe or socket. You open the device and issue an
37 * ioctl() to install a new disk along with its communications descriptor.
38 *
39 * All further communication occurs via the descriptor using the DMSG
40 * LNK_CONN, LNK_SPAN, and BLOCK protocols. The descriptor can be a
41 * direct connection to a remote machine's disk (in-kernenl), to a remote
42 * cluster controller, to the local cluster controller, etc.
43 *
44 * /dev/xdisk is the control device, issue ioctl()s to create the /dev/xa%d
45 * devices. These devices look like raw disks to the system.
46 */
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/buf.h>
50 #include <sys/conf.h>
51 #include <sys/device.h>
52 #include <sys/devicestat.h>
53 #include <sys/disk.h>
54 #include <sys/kernel.h>
55 #include <sys/malloc.h>
56 #include <sys/sysctl.h>
57 #include <sys/proc.h>
58 #include <sys/queue.h>
59 #include <sys/tree.h>
60 #include <sys/udev.h>
61 #include <sys/uuid.h>
62 #include <sys/kern_syscall.h>
63
64 #include <sys/dmsg.h>
65 #include <sys/xdiskioctl.h>
66
67 #include <sys/buf2.h>
68
69 struct xa_softc;
70 struct xa_softc_tree;
71 RB_HEAD(xa_softc_tree, xa_softc);
72 RB_PROTOTYPE(xa_softc_tree, xa_softc, rbnode, xa_softc_cmp);
73
74 static int xa_active;
75 SYSCTL_INT(_debug, OID_AUTO, xa_active, CTLFLAG_RW, &xa_active, 0,
76 "Number of active xdisk IOs");
77 static uint64_t xa_last;
78 SYSCTL_ULONG(_debug, OID_AUTO, xa_last, CTLFLAG_RW, &xa_last, 0,
79 "Offset of last xdisk IO");
80 static int xa_debug = 1;
81 SYSCTL_INT(_debug, OID_AUTO, xa_debug, CTLFLAG_RW, &xa_debug, 0,
82 "xdisk debugging");
83
84 /*
85 * Track a BIO tag
86 */
87 struct xa_tag {
88 TAILQ_ENTRY(xa_tag) entry;
89 struct xa_softc *sc;
90 dmsg_blk_error_t status;
91 kdmsg_state_t *state;
92 struct bio *bio;
93 int waiting;
94 int async;
95 int done;
96 };
97
98 typedef struct xa_tag xa_tag_t;
99
100 /*
101 * Track devices.
102 */
103 struct xa_softc {
104 struct kdmsg_state_list spanq;
105 RB_ENTRY(xa_softc) rbnode;
106 cdev_t dev;
107 struct devstat stats;
108 struct disk_info info;
109 struct disk disk;
110 uuid_t peer_id;
111 int unit;
112 int opencnt;
113 int spancnt;
114 uint64_t keyid;
115 int serializing;
116 int last_error;
117 int terminating;
118 char peer_label[64]; /* from LNK_SPAN host/dev */
119 char pfs_label[64]; /* from LNK_SPAN serno */
120 xa_tag_t *open_tag;
121 TAILQ_HEAD(, bio) bioq; /* pending BIOs */
122 TAILQ_HEAD(, xa_tag) tag_freeq; /* available I/O tags */
123 TAILQ_HEAD(, xa_tag) tag_pendq; /* running I/O tags */
124 struct lock lk;
125 };
126
127 typedef struct xa_softc xa_softc_t;
128
129 struct xa_iocom {
130 TAILQ_ENTRY(xa_iocom) entry;
131 kdmsg_iocom_t iocom;
132 xa_softc_t dummysc;
133 };
134
135 typedef struct xa_iocom xa_iocom_t;
136
137 static int xa_softc_cmp(xa_softc_t *sc1, xa_softc_t *sc2);
138 RB_GENERATE(xa_softc_tree, xa_softc, rbnode, xa_softc_cmp);
139 static struct xa_softc_tree xa_device_tree;
140
141 #define MAXTAGS 64 /* no real limit */
142
143 static int xdisk_attach(struct xdisk_attach_ioctl *xaioc);
144 static int xdisk_detach(struct xdisk_attach_ioctl *xaioc);
145 static void xaio_exit(kdmsg_iocom_t *iocom);
146 static int xaio_rcvdmsg(kdmsg_msg_t *msg);
147
148 static void xa_terminate_check(struct xa_softc *sc);
149
150 static xa_tag_t *xa_setup_cmd(xa_softc_t *sc, struct bio *bio);
151 static void xa_start(xa_tag_t *tag, kdmsg_msg_t *msg, int async);
152 static void xa_done(xa_tag_t *tag, int wasbio);
153 static void xa_release(xa_tag_t *tag, int wasbio);
154 static uint32_t xa_wait(xa_tag_t *tag);
155 static int xa_sync_completion(kdmsg_state_t *state, kdmsg_msg_t *msg);
156 static int xa_bio_completion(kdmsg_state_t *state, kdmsg_msg_t *msg);
157 static void xa_restart_deferred(xa_softc_t *sc);
158
159 #define xa_printf(level, ctl, ...) \
160 if (xa_debug >= (level)) kprintf("xdisk: " ctl, __VA_ARGS__)
161
162 MALLOC_DEFINE(M_XDISK, "Networked disk client", "Network Disks");
163
164 /*
165 * Control device, issue ioctls to create xa devices.
166 */
167 static d_open_t xdisk_open;
168 static d_close_t xdisk_close;
169 static d_ioctl_t xdisk_ioctl;
170
171 static struct dev_ops xdisk_ops = {
172 { "xdisk", 0, D_MPSAFE | D_TRACKCLOSE },
173 .d_open = xdisk_open,
174 .d_close = xdisk_close,
175 .d_ioctl = xdisk_ioctl
176 };
177
178 /*
179 * XA disk devices
180 */
181 static d_open_t xa_open;
182 static d_close_t xa_close;
183 static d_ioctl_t xa_ioctl;
184 static d_strategy_t xa_strategy;
185 static d_psize_t xa_size;
186
187 static struct dev_ops xa_ops = {
188 { "xa", 0, D_DISK | D_CANFREE | D_MPSAFE | D_TRACKCLOSE },
189 .d_open = xa_open,
190 .d_close = xa_close,
191 .d_ioctl = xa_ioctl,
192 .d_read = physread,
193 .d_write = physwrite,
194 .d_strategy = xa_strategy,
195 .d_psize = xa_size
196 };
197
198 static int xdisk_opencount;
199 static cdev_t xdisk_dev;
200 struct lock xdisk_lk;
201 static TAILQ_HEAD(, xa_iocom) xaiocomq;
202
203 /*
204 * Module initialization
205 */
206 static int
xdisk_modevent(module_t mod,int type,void * data)207 xdisk_modevent(module_t mod, int type, void *data)
208 {
209 switch (type) {
210 case MOD_LOAD:
211 TAILQ_INIT(&xaiocomq);
212 RB_INIT(&xa_device_tree);
213 lockinit(&xdisk_lk, "xdisk", 0, 0);
214 xdisk_dev = make_dev(&xdisk_ops, 0,
215 UID_ROOT, GID_WHEEL, 0600, "xdisk");
216 break;
217 case MOD_UNLOAD:
218 case MOD_SHUTDOWN:
219 if (!RB_EMPTY(&xa_device_tree))
220 return (EBUSY);
221 if (xdisk_opencount || TAILQ_FIRST(&xaiocomq))
222 return (EBUSY);
223 if (xdisk_dev) {
224 destroy_dev(xdisk_dev);
225 xdisk_dev = NULL;
226 }
227 dev_ops_remove_all(&xdisk_ops);
228 dev_ops_remove_all(&xa_ops);
229 break;
230 default:
231 break;
232 }
233 return 0;
234 }
235
236 DEV_MODULE(xdisk, xdisk_modevent, 0);
237
238 static int
xa_softc_cmp(xa_softc_t * sc1,xa_softc_t * sc2)239 xa_softc_cmp(xa_softc_t *sc1, xa_softc_t *sc2)
240 {
241 return(strcmp(sc1->pfs_label, sc2->pfs_label));
242 }
243
244 /*
245 * Control device
246 */
247 static int
xdisk_open(struct dev_open_args * ap)248 xdisk_open(struct dev_open_args *ap)
249 {
250 lockmgr(&xdisk_lk, LK_EXCLUSIVE);
251 ++xdisk_opencount;
252 lockmgr(&xdisk_lk, LK_RELEASE);
253 return(0);
254 }
255
256 static int
xdisk_close(struct dev_close_args * ap)257 xdisk_close(struct dev_close_args *ap)
258 {
259 lockmgr(&xdisk_lk, LK_EXCLUSIVE);
260 --xdisk_opencount;
261 lockmgr(&xdisk_lk, LK_RELEASE);
262 return(0);
263 }
264
265 static int
xdisk_ioctl(struct dev_ioctl_args * ap)266 xdisk_ioctl(struct dev_ioctl_args *ap)
267 {
268 int error;
269
270 switch(ap->a_cmd) {
271 case XDISKIOCATTACH:
272 error = xdisk_attach((void *)ap->a_data);
273 break;
274 case XDISKIOCDETACH:
275 error = xdisk_detach((void *)ap->a_data);
276 break;
277 default:
278 error = ENOTTY;
279 break;
280 }
281 return error;
282 }
283
284 /************************************************************************
285 * DMSG INTERFACE *
286 ************************************************************************/
287
288 static int
xdisk_attach(struct xdisk_attach_ioctl * xaioc)289 xdisk_attach(struct xdisk_attach_ioctl *xaioc)
290 {
291 xa_iocom_t *xaio;
292 struct file *fp;
293
294 /*
295 * Normalize ioctl params
296 */
297 fp = holdfp(curthread, xaioc->fd, -1);
298 if (fp == NULL)
299 return EINVAL;
300 xa_printf(1, "xdisk_attach fp=%p\n", fp);
301
302 /*
303 * See if the serial number is already present. If we are
304 * racing a termination the disk subsystem may still have
305 * duplicate entries not yet removed so we wait a bit and
306 * retry.
307 */
308 lockmgr(&xdisk_lk, LK_EXCLUSIVE);
309
310 xaio = kmalloc(sizeof(*xaio), M_XDISK, M_WAITOK | M_ZERO);
311 kdmsg_iocom_init(&xaio->iocom, xaio,
312 KDMSG_IOCOMF_AUTOCONN,
313 M_XDISK, xaio_rcvdmsg);
314 xaio->iocom.exit_func = xaio_exit;
315
316 kdmsg_iocom_reconnect(&xaio->iocom, fp, "xdisk");
317
318 /*
319 * Setup our LNK_CONN advertisement for autoinitiate.
320 *
321 * Our filter is setup to only accept PEER_BLOCK advertisements.
322 * XXX no peer_id filter.
323 *
324 * We need a unique pfs_fsid to avoid confusion.
325 */
326 xaio->iocom.auto_lnk_conn.peer_type = DMSG_PEER_CLIENT;
327 xaio->iocom.auto_lnk_conn.proto_version = DMSG_SPAN_PROTO_1;
328 xaio->iocom.auto_lnk_conn.peer_mask = 1LLU << DMSG_PEER_BLOCK;
329 ksnprintf(xaio->iocom.auto_lnk_conn.peer_label,
330 sizeof(xaio->iocom.auto_lnk_conn.peer_label),
331 "%s/xdisk",
332 hostname);
333 /* kern_uuidgen(&xaio->iocom.auto_lnk_conn.pfs_fsid, 1); */
334
335 /*
336 * Setup our LNK_SPAN advertisement for autoinitiate
337 */
338 TAILQ_INSERT_TAIL(&xaiocomq, xaio, entry);
339 kdmsg_iocom_autoinitiate(&xaio->iocom, NULL);
340
341 lockmgr(&xdisk_lk, LK_RELEASE);
342
343 return 0;
344 }
345
346 static int
xdisk_detach(struct xdisk_attach_ioctl * xaioc)347 xdisk_detach(struct xdisk_attach_ioctl *xaioc)
348 {
349 return EINVAL;
350 }
351
352 /*
353 * Called from iocom core transmit thread upon disconnect.
354 */
355 static
356 void
xaio_exit(kdmsg_iocom_t * iocom)357 xaio_exit(kdmsg_iocom_t *iocom)
358 {
359 xa_iocom_t *xaio = iocom->handle;
360
361 lockmgr(&xdisk_lk, LK_EXCLUSIVE);
362 xa_printf(1, "%s", "xdisk_detach [xaio_exit()]\n");
363 TAILQ_REMOVE(&xaiocomq, xaio, entry);
364 lockmgr(&xdisk_lk, LK_RELEASE);
365
366 kdmsg_iocom_uninit(&xaio->iocom);
367
368 kfree(xaio, M_XDISK);
369 }
370
371 /*
372 * Called from iocom core to handle messages that the iocom core does not
373 * handle itself and for which a state function callback has not yet been
374 * established.
375 *
376 * We primarily care about LNK_SPAN transactions here.
377 */
378 static int
xaio_rcvdmsg(kdmsg_msg_t * msg)379 xaio_rcvdmsg(kdmsg_msg_t *msg)
380 {
381 kdmsg_state_t *state = msg->state;
382 xa_iocom_t *xaio = state->iocom->handle;
383 xa_softc_t *sc;
384
385 if (state) {
386 xa_printf(4,
387 "xdisk - rcvmsg state=%p rx=%08x tx=%08x msgcmd=%08x\n",
388 state, state->rxcmd, state->txcmd,
389 msg->any.head.cmd);
390 }
391 lockmgr(&xdisk_lk, LK_EXCLUSIVE);
392
393 switch(msg->tcmd) {
394 case DMSG_LNK_SPAN | DMSGF_CREATE | DMSGF_DELETE:
395 /*
396 * A LNK_SPAN transaction which is opened and closed
397 * degenerately is not useful to us, just ignore it.
398 */
399 kdmsg_msg_reply(msg, 0);
400 break;
401 case DMSG_LNK_SPAN | DMSGF_CREATE:
402 /*
403 * Manage the tracking node for the remote LNK_SPAN.
404 *
405 * Return a streaming result, leaving the transaction open
406 * in both directions to allow sub-transactions.
407 */
408 bcopy(msg->any.lnk_span.peer_label, xaio->dummysc.peer_label,
409 sizeof(xaio->dummysc.peer_label));
410 xaio->dummysc.peer_label[
411 sizeof(xaio->dummysc.peer_label) - 1] = 0;
412
413 bcopy(msg->any.lnk_span.pfs_label, xaio->dummysc.pfs_label,
414 sizeof(xaio->dummysc.pfs_label));
415 xaio->dummysc.pfs_label[
416 sizeof(xaio->dummysc.pfs_label) - 1] = 0;
417
418 xa_printf(3, "LINK_SPAN state %p create for %s\n",
419 msg->state, msg->any.lnk_span.pfs_label);
420
421 sc = RB_FIND(xa_softc_tree, &xa_device_tree, &xaio->dummysc);
422 if (sc == NULL) {
423 xa_softc_t *sctmp;
424 xa_tag_t *tag;
425 cdev_t dev;
426 int unit;
427 int n;
428
429 sc = kmalloc(sizeof(*sc), M_XDISK, M_WAITOK | M_ZERO);
430 bcopy(msg->any.lnk_span.peer_label, sc->peer_label,
431 sizeof(sc->peer_label));
432 sc->peer_label[sizeof(sc->peer_label) - 1] = 0;
433 bcopy(msg->any.lnk_span.pfs_label, sc->pfs_label,
434 sizeof(sc->pfs_label));
435 sc->pfs_label[sizeof(sc->pfs_label) - 1] = 0;
436
437 /* XXX FIXME O(N^2) */
438 unit = -1;
439 do {
440 ++unit;
441 RB_FOREACH(sctmp, xa_softc_tree,
442 &xa_device_tree) {
443 if (sctmp->unit == unit)
444 break;
445 }
446 } while (sctmp);
447
448 sc->unit = unit;
449 sc->serializing = 1;
450 sc->spancnt = 1;
451 lockinit(&sc->lk, "xalk", 0, 0);
452 TAILQ_INIT(&sc->spanq);
453 TAILQ_INIT(&sc->bioq);
454 TAILQ_INIT(&sc->tag_freeq);
455 TAILQ_INIT(&sc->tag_pendq);
456
457 lockmgr(&sc->lk, LK_EXCLUSIVE);
458 RB_INSERT(xa_softc_tree, &xa_device_tree, sc);
459 TAILQ_INSERT_TAIL(&sc->spanq, msg->state, user_entry);
460 msg->state->any.xa_sc = sc;
461
462 /*
463 * Setup block device
464 */
465 for (n = 0; n < MAXTAGS; ++n) {
466 tag = kmalloc(sizeof(*tag),
467 M_XDISK, M_WAITOK|M_ZERO);
468 tag->sc = sc;
469 TAILQ_INSERT_TAIL(&sc->tag_freeq, tag, entry);
470 }
471
472 if (sc->dev == NULL) {
473 dev = disk_create(unit, &sc->disk, &xa_ops);
474 dev->si_drv1 = sc;
475 sc->dev = dev;
476 devstat_add_entry(&sc->stats, "xa", unit,
477 DEV_BSIZE,
478 DEVSTAT_NO_ORDERED_TAGS,
479 DEVSTAT_TYPE_DIRECT |
480 DEVSTAT_TYPE_IF_OTHER,
481 DEVSTAT_PRIORITY_OTHER);
482 }
483
484 sc->info.d_media_blksize =
485 msg->any.lnk_span.media.block.blksize;
486 if (sc->info.d_media_blksize <= 0)
487 sc->info.d_media_blksize = 1;
488 sc->info.d_media_blocks =
489 msg->any.lnk_span.media.block.bytes /
490 sc->info.d_media_blksize;
491 sc->info.d_dsflags = DSO_MBRQUIET | DSO_RAWPSIZE;
492 sc->info.d_secpertrack = 32;
493 sc->info.d_nheads = 64;
494 sc->info.d_secpercyl = sc->info.d_secpertrack *
495 sc->info.d_nheads;
496 sc->info.d_ncylinders = 0;
497 if (sc->pfs_label[0])
498 sc->info.d_serialno = sc->pfs_label;
499 /*
500 * WARNING! disk_setdiskinfo() must be asynchronous
501 * because we are in the rxmsg thread. If
502 * it is synchronous and issues more disk
503 * I/Os, we will deadlock.
504 */
505 disk_setdiskinfo(&sc->disk, &sc->info);
506 xa_restart_deferred(sc); /* eats serializing */
507 lockmgr(&sc->lk, LK_RELEASE);
508 } else {
509 lockmgr(&sc->lk, LK_EXCLUSIVE);
510 ++sc->spancnt;
511 TAILQ_INSERT_TAIL(&sc->spanq, msg->state, user_entry);
512 msg->state->any.xa_sc = sc;
513 if (sc->serializing == 0 && sc->open_tag == NULL) {
514 sc->serializing = 1;
515 xa_restart_deferred(sc); /* eats serializing */
516 }
517 lockmgr(&sc->lk, LK_RELEASE);
518 if (sc->dev && sc->dev->si_disk) {
519 xa_printf(1, "reprobe disk: %s\n",
520 sc->pfs_label);
521 disk_msg_send(DISK_DISK_REPROBE,
522 sc->dev->si_disk,
523 NULL);
524 }
525 }
526 xa_printf(2, "sc %p spancnt %d\n", sc, sc->spancnt);
527 kdmsg_msg_result(msg, 0);
528 break;
529 case DMSG_LNK_SPAN | DMSGF_DELETE:
530 /*
531 * Manage the tracking node for the remote LNK_SPAN.
532 *
533 * Return a final result, closing our end of the transaction.
534 */
535 sc = msg->state->any.xa_sc;
536 xa_printf(3, "LINK_SPAN state %p delete for %s (sc=%p)\n",
537 msg->state, (sc ? sc->pfs_label : "(null)"), sc);
538 lockmgr(&sc->lk, LK_EXCLUSIVE);
539 msg->state->any.xa_sc = NULL;
540 TAILQ_REMOVE(&sc->spanq, msg->state, user_entry);
541 --sc->spancnt;
542
543 xa_printf(2, "sc %p spancnt %d\n", sc, sc->spancnt);
544
545 /*
546 * Spans can come and go as the graph stabilizes, so if
547 * we lose a span along with sc->open_tag we may be able
548 * to restart the I/Os on a different span.
549 */
550 if (sc->spancnt &&
551 sc->serializing == 0 && sc->open_tag == NULL) {
552 sc->serializing = 1;
553 xa_restart_deferred(sc);
554 }
555 lockmgr(&sc->lk, LK_RELEASE);
556 kdmsg_msg_reply(msg, 0);
557
558 #if 0
559 /*
560 * Termination
561 */
562 if (sc->spancnt == 0)
563 xa_terminate_check(sc);
564 #endif
565 break;
566 case DMSG_LNK_SPAN | DMSGF_DELETE | DMSGF_REPLY:
567 /*
568 * Ignore unimplemented streaming replies on our LNK_SPAN
569 * transaction.
570 */
571 xa_printf(3, "LINK_SPAN state %p delete+reply\n",
572 msg->state);
573 break;
574 case DMSG_LNK_SPAN | DMSGF_REPLY:
575 /*
576 * Ignore unimplemented streaming replies on our LNK_SPAN
577 * transaction.
578 */
579 xa_printf(3, "LINK_SPAN state %p reply\n",
580 msg->state);
581 break;
582 case DMSG_DBG_SHELL:
583 /*
584 * Execute shell command (not supported atm).
585 *
586 * This is a one-way packet but if not (e.g. if part of
587 * a streaming transaction), we will have already closed
588 * our end.
589 */
590 kdmsg_msg_reply(msg, DMSG_ERR_NOSUPP);
591 break;
592 case DMSG_DBG_SHELL | DMSGF_REPLY:
593 /*
594 * Receive one or more replies to a shell command
595 * that we sent. Just dump it to the console.
596 *
597 * This is a one-way packet but if not (e.g. if
598 * part of a streaming transaction), we will have
599 * already closed our end.
600 */
601 if (msg->aux_data) {
602 msg->aux_data[msg->aux_size - 1] = 0;
603 xa_printf(0, "DEBUGMSG: %s\n", msg->aux_data);
604 }
605 break;
606 default:
607 /*
608 * Unsupported one-way message, streaming message, or
609 * transaction.
610 *
611 * Terminate any unsupported transactions with an error
612 * and ignore any unsupported streaming messages.
613 *
614 * NOTE: This case also includes DMSG_LNK_ERROR messages
615 * which might be one-way, replying to those would
616 * cause an infinite ping-pong.
617 */
618 if (msg->any.head.cmd & DMSGF_CREATE)
619 kdmsg_msg_reply(msg, DMSG_ERR_NOSUPP);
620 break;
621 }
622 lockmgr(&xdisk_lk, LK_RELEASE);
623
624 return 0;
625 }
626
627 /*
628 * Determine if we can destroy the xa_softc.
629 *
630 * Called with xdisk_lk held.
631 */
632 static
633 void
xa_terminate_check(struct xa_softc * sc)634 xa_terminate_check(struct xa_softc *sc)
635 {
636 xa_tag_t *tag;
637
638 /*
639 * Determine if we can destroy the softc.
640 */
641 xa_printf(1, "Terminate check xa%d (%d,%d,%d) sc=%p ",
642 sc->unit,
643 sc->opencnt, sc->serializing, sc->spancnt,
644 sc);
645
646 if (sc->opencnt || sc->serializing || sc->spancnt ||
647 TAILQ_FIRST(&sc->bioq) || TAILQ_FIRST(&sc->tag_pendq)) {
648 xa_printf(1, "%s", "(leave intact)\n");
649 return;
650 }
651
652 /*
653 * Remove from device tree, a race with a new incoming span
654 * will create a new softc and disk.
655 */
656 RB_REMOVE(xa_softc_tree, &xa_device_tree, sc);
657 sc->terminating = 1;
658
659 /*
660 * Device has to go first to prevent device ops races.
661 */
662 if (sc->dev) {
663 disk_destroy(&sc->disk);
664 devstat_remove_entry(&sc->stats);
665 sc->dev->si_drv1 = NULL;
666 sc->dev = NULL;
667 }
668
669 xa_printf(1, "%s", "(remove from tree)\n");
670 sc->serializing = 1;
671 KKASSERT(sc->opencnt == 0);
672 KKASSERT(TAILQ_EMPTY(&sc->tag_pendq));
673
674 while ((tag = TAILQ_FIRST(&sc->tag_freeq)) != NULL) {
675 TAILQ_REMOVE(&sc->tag_freeq, tag, entry);
676 tag->sc = NULL;
677 kfree(tag, M_XDISK);
678 }
679
680 kfree(sc, M_XDISK);
681 }
682
683 /************************************************************************
684 * XA DEVICE INTERFACE *
685 ************************************************************************/
686
687 static int
xa_open(struct dev_open_args * ap)688 xa_open(struct dev_open_args *ap)
689 {
690 cdev_t dev = ap->a_head.a_dev;
691 xa_softc_t *sc;
692 int error;
693
694 dev->si_bsize_phys = 512;
695 dev->si_bsize_best = 32768;
696
697 /*
698 * Interlock open with opencnt, wait for attachment operations
699 * to finish.
700 */
701 lockmgr(&xdisk_lk, LK_EXCLUSIVE);
702 again:
703 sc = dev->si_drv1;
704 if (sc == NULL) {
705 lockmgr(&xdisk_lk, LK_RELEASE);
706 return ENXIO; /* raced destruction */
707 }
708 if (sc->serializing) {
709 tsleep(sc, 0, "xarace", hz / 10);
710 goto again;
711 }
712 if (sc->terminating) {
713 lockmgr(&xdisk_lk, LK_RELEASE);
714 return ENXIO; /* raced destruction */
715 }
716 sc->serializing = 1;
717
718 /*
719 * Serialize initial open
720 */
721 if (sc->opencnt++ > 0) {
722 sc->serializing = 0;
723 wakeup(sc);
724 lockmgr(&xdisk_lk, LK_RELEASE);
725 return(0);
726 }
727
728 /*
729 * Issue BLK_OPEN if necessary. ENXIO is returned if we have trouble.
730 */
731 if (sc->open_tag == NULL) {
732 lockmgr(&sc->lk, LK_EXCLUSIVE);
733 xa_restart_deferred(sc); /* eats serializing */
734 lockmgr(&sc->lk, LK_RELEASE);
735 } else {
736 sc->serializing = 0;
737 wakeup(sc);
738 }
739 lockmgr(&xdisk_lk, LK_RELEASE);
740
741 /*
742 * Wait for completion of the BLK_OPEN
743 */
744 lockmgr(&xdisk_lk, LK_EXCLUSIVE);
745 while (sc->serializing)
746 lksleep(sc, &xdisk_lk, 0, "xaopen", hz);
747
748 error = sc->last_error;
749 if (error) {
750 KKASSERT(sc->opencnt > 0);
751 --sc->opencnt;
752 xa_terminate_check(sc);
753 sc = NULL; /* sc may be invalid now */
754 }
755 lockmgr(&xdisk_lk, LK_RELEASE);
756
757 return (error);
758 }
759
760 static int
xa_close(struct dev_close_args * ap)761 xa_close(struct dev_close_args *ap)
762 {
763 cdev_t dev = ap->a_head.a_dev;
764 xa_softc_t *sc;
765 xa_tag_t *tag;
766
767 lockmgr(&xdisk_lk, LK_EXCLUSIVE);
768 sc = dev->si_drv1;
769 if (sc == NULL)
770 return ENXIO; /* raced destruction */
771 if (sc->terminating) {
772 lockmgr(&sc->lk, LK_RELEASE);
773 return ENXIO; /* raced destruction */
774 }
775 lockmgr(&sc->lk, LK_EXCLUSIVE);
776
777 /*
778 * NOTE: Clearing open_tag allows a concurrent open to re-open
779 * the device and prevents autonomous completion of the tag.
780 */
781 if (sc->opencnt == 1 && sc->open_tag) {
782 tag = sc->open_tag;
783 sc->open_tag = NULL;
784 lockmgr(&sc->lk, LK_RELEASE);
785 kdmsg_state_reply(tag->state, 0); /* close our side */
786 xa_wait(tag); /* wait on remote */
787 } else {
788 lockmgr(&sc->lk, LK_RELEASE);
789 }
790 KKASSERT(sc->opencnt > 0);
791 --sc->opencnt;
792 xa_terminate_check(sc);
793 lockmgr(&xdisk_lk, LK_RELEASE);
794
795 return(0);
796 }
797
798 static int
xa_strategy(struct dev_strategy_args * ap)799 xa_strategy(struct dev_strategy_args *ap)
800 {
801 xa_softc_t *sc = ap->a_head.a_dev->si_drv1;
802 xa_tag_t *tag;
803 struct bio *bio = ap->a_bio;
804
805 devstat_start_transaction(&sc->stats);
806 atomic_add_int(&xa_active, 1);
807 xa_last = bio->bio_offset;
808
809 /*
810 * If no tags are available NULL is returned and the bio is
811 * placed on sc->bioq.
812 */
813 lockmgr(&sc->lk, LK_EXCLUSIVE);
814 tag = xa_setup_cmd(sc, bio);
815 if (tag)
816 xa_start(tag, NULL, 1);
817 lockmgr(&sc->lk, LK_RELEASE);
818
819 return(0);
820 }
821
822 static int
xa_ioctl(struct dev_ioctl_args * ap)823 xa_ioctl(struct dev_ioctl_args *ap)
824 {
825 return(ENOTTY);
826 }
827
828 static int
xa_size(struct dev_psize_args * ap)829 xa_size(struct dev_psize_args *ap)
830 {
831 struct xa_softc *sc;
832
833 if ((sc = ap->a_head.a_dev->si_drv1) == NULL)
834 return (ENXIO);
835 ap->a_result = sc->info.d_media_blocks;
836 return (0);
837 }
838
839 /************************************************************************
840 * XA BLOCK PROTOCOL STATE MACHINE *
841 ************************************************************************
842 *
843 * Implement tag/msg setup and related functions.
844 * Called with sc->lk held.
845 */
846 static xa_tag_t *
xa_setup_cmd(xa_softc_t * sc,struct bio * bio)847 xa_setup_cmd(xa_softc_t *sc, struct bio *bio)
848 {
849 xa_tag_t *tag;
850
851 /*
852 * Only get a tag if we have a valid virtual circuit to the server.
853 */
854 if ((tag = TAILQ_FIRST(&sc->tag_freeq)) != NULL) {
855 TAILQ_REMOVE(&sc->tag_freeq, tag, entry);
856 tag->bio = bio;
857 TAILQ_INSERT_TAIL(&sc->tag_pendq, tag, entry);
858 }
859
860 /*
861 * If we can't dispatch now and this is a bio, queue it for later.
862 */
863 if (tag == NULL && bio) {
864 TAILQ_INSERT_TAIL(&sc->bioq, bio, bio_act);
865 }
866
867 return (tag);
868 }
869
870 /*
871 * Called with sc->lk held
872 */
873 static void
xa_start(xa_tag_t * tag,kdmsg_msg_t * msg,int async)874 xa_start(xa_tag_t *tag, kdmsg_msg_t *msg, int async)
875 {
876 xa_softc_t *sc = tag->sc;
877
878 tag->done = 0;
879 tag->async = async;
880 tag->status.head.error = DMSG_ERR_IO; /* fallback error */
881
882 if (msg == NULL) {
883 struct bio *bio;
884 struct buf *bp;
885 kdmsg_state_t *trans;
886
887 if (sc->opencnt == 0 || sc->open_tag == NULL) {
888 TAILQ_FOREACH(trans, &sc->spanq, user_entry) {
889 if ((trans->rxcmd & DMSGF_DELETE) == 0)
890 break;
891 }
892 } else {
893 trans = sc->open_tag->state;
894 }
895 if (trans == NULL)
896 goto skip;
897
898 KKASSERT(tag->bio);
899 bio = tag->bio;
900 bp = bio->bio_buf;
901
902 switch(bp->b_cmd) {
903 case BUF_CMD_READ:
904 msg = kdmsg_msg_alloc(trans,
905 DMSG_BLK_READ |
906 DMSGF_CREATE |
907 DMSGF_DELETE,
908 xa_bio_completion, tag);
909 msg->any.blk_read.keyid = sc->keyid;
910 msg->any.blk_read.offset = bio->bio_offset;
911 msg->any.blk_read.bytes = bp->b_bcount;
912 break;
913 case BUF_CMD_WRITE:
914 msg = kdmsg_msg_alloc(trans,
915 DMSG_BLK_WRITE |
916 DMSGF_CREATE | DMSGF_DELETE,
917 xa_bio_completion, tag);
918 msg->any.blk_write.keyid = sc->keyid;
919 msg->any.blk_write.offset = bio->bio_offset;
920 msg->any.blk_write.bytes = bp->b_bcount;
921 msg->aux_data = bp->b_data;
922 msg->aux_size = bp->b_bcount;
923 break;
924 case BUF_CMD_FLUSH:
925 msg = kdmsg_msg_alloc(trans,
926 DMSG_BLK_FLUSH |
927 DMSGF_CREATE | DMSGF_DELETE,
928 xa_bio_completion, tag);
929 msg->any.blk_flush.keyid = sc->keyid;
930 msg->any.blk_flush.offset = bio->bio_offset;
931 msg->any.blk_flush.bytes = bp->b_bcount;
932 break;
933 case BUF_CMD_FREEBLKS:
934 msg = kdmsg_msg_alloc(trans,
935 DMSG_BLK_FREEBLKS |
936 DMSGF_CREATE | DMSGF_DELETE,
937 xa_bio_completion, tag);
938 msg->any.blk_freeblks.keyid = sc->keyid;
939 msg->any.blk_freeblks.offset = bio->bio_offset;
940 msg->any.blk_freeblks.bytes = bp->b_bcount;
941 break;
942 default:
943 bp->b_flags |= B_ERROR;
944 bp->b_error = EIO;
945 devstat_end_transaction_buf(&sc->stats, bp);
946 atomic_add_int(&xa_active, -1);
947 biodone(bio);
948 tag->bio = NULL;
949 break;
950 }
951 }
952
953 /*
954 * If no msg was allocated we likely could not find a good span.
955 */
956 skip:
957 if (msg) {
958 /*
959 * Message was passed in or constructed.
960 */
961 tag->state = msg->state;
962 lockmgr(&sc->lk, LK_RELEASE);
963 kdmsg_msg_write(msg);
964 lockmgr(&sc->lk, LK_EXCLUSIVE);
965 } else if (tag->bio &&
966 (tag->bio->bio_buf->b_flags & B_FAILONDIS) == 0) {
967 /*
968 * No spans available but BIO is not allowed to fail
969 * on connectivity problems. Requeue the BIO.
970 */
971 TAILQ_INSERT_TAIL(&sc->bioq, tag->bio, bio_act);
972 tag->bio = NULL;
973 lockmgr(&sc->lk, LK_RELEASE);
974 xa_done(tag, 1);
975 lockmgr(&sc->lk, LK_EXCLUSIVE);
976 } else {
977 /*
978 * No spans available, bio is allowed to fail.
979 */
980 lockmgr(&sc->lk, LK_RELEASE);
981 tag->status.head.error = DMSG_ERR_IO;
982 xa_done(tag, 1);
983 lockmgr(&sc->lk, LK_EXCLUSIVE);
984 }
985 }
986
987 static uint32_t
xa_wait(xa_tag_t * tag)988 xa_wait(xa_tag_t *tag)
989 {
990 xa_softc_t *sc = tag->sc;
991 uint32_t error;
992
993 lockmgr(&sc->lk, LK_EXCLUSIVE);
994 tag->waiting = 1;
995 while (tag->done == 0)
996 lksleep(tag, &sc->lk, 0, "xawait", 0);
997 lockmgr(&sc->lk, LK_RELEASE);
998
999 error = tag->status.head.error;
1000 tag->waiting = 0;
1001 xa_release(tag, 0);
1002
1003 return error;
1004 }
1005
1006 static void
xa_done(xa_tag_t * tag,int wasbio)1007 xa_done(xa_tag_t *tag, int wasbio)
1008 {
1009 KKASSERT(tag->bio == NULL);
1010
1011 tag->state = NULL;
1012 tag->done = 1;
1013 if (tag->waiting)
1014 wakeup(tag);
1015 if (tag->async)
1016 xa_release(tag, wasbio);
1017 }
1018
1019 /*
1020 * Release a tag. If everything looks ok and there are pending BIOs
1021 * (due to all tags in-use), we can use the tag to start the next BIO.
1022 * Do not try to restart if the connection is currently failed.
1023 */
1024 static
1025 void
xa_release(xa_tag_t * tag,int wasbio)1026 xa_release(xa_tag_t *tag, int wasbio)
1027 {
1028 xa_softc_t *sc = tag->sc;
1029 struct bio *bio;
1030
1031 if ((bio = tag->bio) != NULL) {
1032 struct buf *bp = bio->bio_buf;
1033
1034 bp->b_error = EIO;
1035 bp->b_flags |= B_ERROR;
1036 devstat_end_transaction_buf(&sc->stats, bp);
1037 atomic_add_int(&xa_active, -1);
1038 biodone(bio);
1039 tag->bio = NULL;
1040 }
1041
1042 lockmgr(&sc->lk, LK_EXCLUSIVE);
1043
1044 if (wasbio && sc->open_tag &&
1045 (bio = TAILQ_FIRST(&sc->bioq)) != NULL) {
1046 TAILQ_REMOVE(&sc->bioq, bio, bio_act);
1047 tag->bio = bio;
1048 xa_start(tag, NULL, 1);
1049 } else {
1050 TAILQ_REMOVE(&sc->tag_pendq, tag, entry);
1051 TAILQ_INSERT_TAIL(&sc->tag_freeq, tag, entry);
1052 }
1053 lockmgr(&sc->lk, LK_RELEASE);
1054 }
1055
1056 /*
1057 * Handle messages under the BLKOPEN transaction.
1058 */
1059 static int
xa_sync_completion(kdmsg_state_t * state,kdmsg_msg_t * msg)1060 xa_sync_completion(kdmsg_state_t *state, kdmsg_msg_t *msg)
1061 {
1062 xa_tag_t *tag = state->any.any;
1063 xa_softc_t *sc;
1064 struct bio *bio;
1065
1066 /*
1067 * If the tag has been cleaned out we already closed our side
1068 * of the transaction and we are waiting for the other side to
1069 * close.
1070 */
1071 xa_printf(1, "xa_sync_completion: tag %p msg %08x state %p\n",
1072 tag, msg->any.head.cmd, msg->state);
1073
1074 if (tag == NULL) {
1075 if (msg->any.head.cmd & DMSGF_CREATE)
1076 kdmsg_state_reply(state, DMSG_ERR_LOSTLINK);
1077 return 0;
1078 }
1079 sc = tag->sc;
1080
1081 /*
1082 * Validate the tag
1083 */
1084 lockmgr(&sc->lk, LK_EXCLUSIVE);
1085
1086 /*
1087 * Handle initial response to our open and restart any deferred
1088 * BIOs on success.
1089 *
1090 * NOTE: DELETE may also be set.
1091 */
1092 if (msg->any.head.cmd & DMSGF_CREATE) {
1093 switch(msg->any.head.cmd & DMSGF_CMDSWMASK) {
1094 case DMSG_LNK_ERROR | DMSGF_REPLY:
1095 bzero(&tag->status, sizeof(tag->status));
1096 tag->status.head = msg->any.head;
1097 break;
1098 case DMSG_BLK_ERROR | DMSGF_REPLY:
1099 tag->status = msg->any.blk_error;
1100 break;
1101 }
1102 sc->last_error = tag->status.head.error;
1103 xa_printf(1, "blk_open completion status %d\n",
1104 sc->last_error);
1105 if (sc->last_error == 0) {
1106 while ((bio = TAILQ_FIRST(&sc->bioq)) != NULL) {
1107 tag = xa_setup_cmd(sc, NULL);
1108 if (tag == NULL)
1109 break;
1110 TAILQ_REMOVE(&sc->bioq, bio, bio_act);
1111 tag->bio = bio;
1112 xa_start(tag, NULL, 1);
1113 }
1114 }
1115 sc->serializing = 0;
1116 wakeup(sc);
1117 }
1118
1119 /*
1120 * Handle unexpected termination (or lost comm channel) from other
1121 * side. Autonomous completion only if open_tag matches,
1122 * otherwise another thread is probably waiting on the tag.
1123 *
1124 * (see xa_close() for other interactions)
1125 */
1126 if (msg->any.head.cmd & DMSGF_DELETE) {
1127 kdmsg_state_reply(tag->state, 0);
1128 if (sc->open_tag == tag) {
1129 sc->open_tag = NULL;
1130 xa_done(tag, 0);
1131 } else {
1132 tag->async = 0;
1133 xa_done(tag, 0);
1134 }
1135 }
1136 lockmgr(&sc->lk, LK_RELEASE);
1137
1138 return (0);
1139 }
1140
1141 static int
xa_bio_completion(kdmsg_state_t * state,kdmsg_msg_t * msg)1142 xa_bio_completion(kdmsg_state_t *state, kdmsg_msg_t *msg)
1143 {
1144 xa_tag_t *tag = state->any.any;
1145 xa_softc_t *sc = tag->sc;
1146 struct bio *bio;
1147 struct buf *bp;
1148
1149 /*
1150 * Get the bio from the tag. If no bio is present we just do
1151 * 'done' handling.
1152 */
1153 if ((bio = tag->bio) == NULL)
1154 goto handle_done;
1155 bp = bio->bio_buf;
1156
1157 /*
1158 * Process return status
1159 */
1160 switch(msg->any.head.cmd & DMSGF_CMDSWMASK) {
1161 case DMSG_LNK_ERROR | DMSGF_REPLY:
1162 bzero(&tag->status, sizeof(tag->status));
1163 tag->status.head = msg->any.head;
1164 if (tag->status.head.error)
1165 tag->status.resid = bp->b_bcount;
1166 else
1167 tag->status.resid = 0;
1168 break;
1169 case DMSG_BLK_ERROR | DMSGF_REPLY:
1170 tag->status = msg->any.blk_error;
1171 break;
1172 }
1173
1174 /*
1175 * If the device is open stall the bio on DMSG errors. If an
1176 * actual I/O error occured on the remote device, DMSG_ERR_IO
1177 * will be returned.
1178 */
1179 if (tag->status.head.error &&
1180 (msg->any.head.cmd & DMSGF_DELETE) && sc->opencnt) {
1181 if (tag->status.head.error != DMSG_ERR_IO)
1182 goto handle_repend;
1183 }
1184
1185 /*
1186 * Process bio completion
1187 *
1188 * For reads any returned data is zero-extended if necessary, so
1189 * the server can short-cut any all-zeros reads if it desires.
1190 */
1191 switch(bp->b_cmd) {
1192 case BUF_CMD_READ:
1193 if (msg->aux_data && msg->aux_size) {
1194 if (msg->aux_size < bp->b_bcount) {
1195 bcopy(msg->aux_data, bp->b_data, msg->aux_size);
1196 bzero(bp->b_data + msg->aux_size,
1197 bp->b_bcount - msg->aux_size);
1198 } else {
1199 bcopy(msg->aux_data, bp->b_data, bp->b_bcount);
1200 }
1201 } else {
1202 bzero(bp->b_data, bp->b_bcount);
1203 }
1204 /* fall through */
1205 case BUF_CMD_WRITE:
1206 case BUF_CMD_FLUSH:
1207 case BUF_CMD_FREEBLKS:
1208 default:
1209 if (tag->status.resid > bp->b_bcount)
1210 tag->status.resid = bp->b_bcount;
1211 bp->b_resid = tag->status.resid;
1212 if (tag->status.head.error != 0) {
1213 bp->b_error = EIO;
1214 bp->b_flags |= B_ERROR;
1215 } else {
1216 bp->b_resid = 0;
1217 }
1218 devstat_end_transaction_buf(&sc->stats, bp);
1219 atomic_add_int(&xa_active, -1);
1220 biodone(bio);
1221 tag->bio = NULL;
1222 break;
1223 }
1224
1225 /*
1226 * Handle completion of the transaction. If the bioq is not empty
1227 * we can initiate another bio on the same tag.
1228 *
1229 * NOTE: Most of our transactions will be single-message
1230 * CREATE+DELETEs, so we won't have to terminate the
1231 * transaction separately, here. But just in case they
1232 * aren't be sure to terminate the transaction.
1233 */
1234 handle_done:
1235 if (msg->any.head.cmd & DMSGF_DELETE) {
1236 xa_done(tag, 1);
1237 if ((state->txcmd & DMSGF_DELETE) == 0)
1238 kdmsg_msg_reply(msg, 0);
1239 }
1240 return (0);
1241
1242 /*
1243 * Handle the case where the transaction failed due to a
1244 * connectivity issue. The tag is put away with wasbio=0
1245 * and we put the BIO back onto the bioq for a later restart.
1246 *
1247 * probe I/Os (where the device is not open) will be failed
1248 * instead of requeued.
1249 */
1250 handle_repend:
1251 tag->bio = NULL;
1252 if (bio->bio_buf->b_flags & B_FAILONDIS) {
1253 xa_printf(1, "xa_strategy: lost link, fail probe bp %p\n",
1254 bio->bio_buf);
1255 bio->bio_buf->b_error = ENXIO;
1256 bio->bio_buf->b_flags |= B_ERROR;
1257 biodone(bio);
1258 bio = NULL;
1259 } else {
1260 xa_printf(1, "xa_strategy: lost link, requeue bp %p\n",
1261 bio->bio_buf);
1262 }
1263 xa_done(tag, 0);
1264 if ((state->txcmd & DMSGF_DELETE) == 0)
1265 kdmsg_msg_reply(msg, 0);
1266
1267 /*
1268 * Requeue the bio
1269 */
1270 if (bio) {
1271 lockmgr(&sc->lk, LK_EXCLUSIVE);
1272 TAILQ_INSERT_TAIL(&sc->bioq, bio, bio_act);
1273 lockmgr(&sc->lk, LK_RELEASE);
1274 }
1275 return (0);
1276 }
1277
1278 /*
1279 * Restart as much deferred I/O as we can. The serializer is set and we
1280 * eat it (clear it) when done.
1281 *
1282 * Called with sc->lk held
1283 */
1284 static
1285 void
xa_restart_deferred(xa_softc_t * sc)1286 xa_restart_deferred(xa_softc_t *sc)
1287 {
1288 kdmsg_state_t *span;
1289 kdmsg_msg_t *msg;
1290 xa_tag_t *tag;
1291 int error;
1292
1293 KKASSERT(sc->serializing);
1294
1295 /*
1296 * Determine if a restart is needed.
1297 */
1298 if (sc->opencnt == 0) {
1299 /*
1300 * Device is not open, nothing to do, eat serializing.
1301 */
1302 sc->serializing = 0;
1303 wakeup(sc);
1304 } else if (sc->open_tag == NULL) {
1305 /*
1306 * BLK_OPEN required before we can restart any BIOs.
1307 * Select the best LNK_SPAN to issue the BLK_OPEN under.
1308 *
1309 * serializing interlocks waiting open()s.
1310 */
1311 error = 0;
1312 TAILQ_FOREACH(span, &sc->spanq, user_entry) {
1313 if ((span->rxcmd & DMSGF_DELETE) == 0)
1314 break;
1315 }
1316 if (span == NULL)
1317 error = ENXIO;
1318
1319 if (error == 0) {
1320 tag = xa_setup_cmd(sc, NULL);
1321 if (tag == NULL)
1322 error = ENXIO;
1323 }
1324 if (error == 0) {
1325 sc->open_tag = tag;
1326 msg = kdmsg_msg_alloc(span,
1327 DMSG_BLK_OPEN |
1328 DMSGF_CREATE,
1329 xa_sync_completion, tag);
1330 msg->any.blk_open.modes = DMSG_BLKOPEN_RD;
1331 xa_printf(1,
1332 "BLK_OPEN tag %p state %p "
1333 "span-state %p\n",
1334 tag, msg->state, span);
1335 xa_start(tag, msg, 0);
1336 }
1337 if (error) {
1338 sc->serializing = 0;
1339 wakeup(sc);
1340 }
1341 /* else leave serializing set until BLK_OPEN response */
1342 } else {
1343 /* nothing to do */
1344 sc->serializing = 0;
1345 wakeup(sc);
1346 }
1347 }
1348