xref: /freebsd/sys/dev/xen/blkfront/blkfront.c (revision 9ccc37e32070303fb293a2a1697ffa71eeb49b25)
1 /*
2  * XenBSD block device driver
3  *
4  * Copyright (c) 2009 Scott Long, Yahoo!
5  * Copyright (c) 2009 Frank Suchomel, Citrix
6  * Copyright (c) 2009 Doug F. Rabson, Citrix
7  * Copyright (c) 2005 Kip Macy
8  * Copyright (c) 2003-2004, Keir Fraser & Steve Hand
9  * Modifications by Mark A. Williamson are (c) Intel Research Cambridge
10  *
11  *
12  * Permission is hereby granted, free of charge, to any person obtaining a copy
13  * of this software and associated documentation files (the "Software"), to
14  * deal in the Software without restriction, including without limitation the
15  * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
16  * sell copies of the Software, and to permit persons to whom the Software is
17  * furnished to do so, subject to the following conditions:
18  *
19  * The above copyright notice and this permission notice shall be included in
20  * all copies or substantial portions of the Software.
21  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
22  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
23  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
24  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
25  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
26  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
27  * DEALINGS IN THE SOFTWARE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/malloc.h>
36 #include <sys/kernel.h>
37 #include <vm/vm.h>
38 #include <vm/pmap.h>
39 
40 #include <sys/bio.h>
41 #include <sys/bus.h>
42 #include <sys/conf.h>
43 #include <sys/module.h>
44 #include <sys/sysctl.h>
45 
46 #include <machine/bus.h>
47 #include <sys/rman.h>
48 #include <machine/resource.h>
49 #include <machine/intr_machdep.h>
50 #include <machine/vmparam.h>
51 #include <sys/bus_dma.h>
52 
53 #include <machine/_inttypes.h>
54 #include <machine/xen/xen-os.h>
55 #include <machine/xen/xenvar.h>
56 #include <machine/xen/xenfunc.h>
57 
58 #include <xen/hypervisor.h>
59 #include <xen/xen_intr.h>
60 #include <xen/evtchn.h>
61 #include <xen/gnttab.h>
62 #include <xen/interface/grant_table.h>
63 #include <xen/interface/io/protocols.h>
64 #include <xen/xenbus/xenbusvar.h>
65 
66 #include <geom/geom_disk.h>
67 
68 #include <dev/xen/blkfront/block.h>
69 
70 #include "xenbus_if.h"
71 
72 /* prototypes */
73 static void xb_free_command(struct xb_command *cm);
74 static void xb_startio(struct xb_softc *sc);
75 static void blkfront_connect(struct xb_softc *);
76 static void blkfront_closing(device_t);
77 static int blkfront_detach(device_t);
78 static int setup_blkring(struct xb_softc *);
79 static void blkif_int(void *);
80 static void blkfront_initialize(struct xb_softc *);
81 static int blkif_completion(struct xb_command *);
82 static void blkif_free(struct xb_softc *);
83 static void blkif_queue_cb(void *, bus_dma_segment_t *, int, int);
84 
85 static MALLOC_DEFINE(M_XENBLOCKFRONT, "xbd", "Xen Block Front driver data");
86 
87 #define GRANT_INVALID_REF 0
88 
89 /* Control whether runtime update of vbds is enabled. */
90 #define ENABLE_VBD_UPDATE 0
91 
92 #if ENABLE_VBD_UPDATE
93 static void vbd_update(void);
94 #endif
95 
96 #define BLKIF_STATE_DISCONNECTED 0
97 #define BLKIF_STATE_CONNECTED    1
98 #define BLKIF_STATE_SUSPENDED    2
99 
100 #ifdef notyet
101 static char *blkif_state_name[] = {
102 	[BLKIF_STATE_DISCONNECTED] = "disconnected",
103 	[BLKIF_STATE_CONNECTED]    = "connected",
104 	[BLKIF_STATE_SUSPENDED]    = "closed",
105 };
106 
107 static char * blkif_status_name[] = {
108 	[BLKIF_INTERFACE_STATUS_CLOSED]       = "closed",
109 	[BLKIF_INTERFACE_STATUS_DISCONNECTED] = "disconnected",
110 	[BLKIF_INTERFACE_STATUS_CONNECTED]    = "connected",
111 	[BLKIF_INTERFACE_STATUS_CHANGED]      = "changed",
112 };
113 #endif
114 
115 #if 0
116 #define DPRINTK(fmt, args...) printf("[XEN] %s:%d: " fmt ".\n", __func__, __LINE__, ##args)
117 #else
118 #define DPRINTK(fmt, args...)
119 #endif
120 
121 static int blkif_open(struct disk *dp);
122 static int blkif_close(struct disk *dp);
123 static int blkif_ioctl(struct disk *dp, u_long cmd, void *addr, int flag, struct thread *td);
124 static int blkif_queue_request(struct xb_softc *sc, struct xb_command *cm);
125 static void xb_strategy(struct bio *bp);
126 
127 // In order to quiesce the device during kernel dumps, outstanding requests to
128 // DOM0 for disk reads/writes need to be accounted for.
129 static	int	xb_dump(void *, void *, vm_offset_t, off_t, size_t);
130 
131 /* XXX move to xb_vbd.c when VBD update support is added */
132 #define MAX_VBDS 64
133 
134 #define XBD_SECTOR_SIZE		512	/* XXX: assume for now */
135 #define XBD_SECTOR_SHFT		9
136 
137 /*
138  * Translate Linux major/minor to an appropriate name and unit
139  * number. For HVM guests, this allows us to use the same drive names
140  * with blkfront as the emulated drives, easing transition slightly.
141  */
142 static void
143 blkfront_vdevice_to_unit(uint32_t vdevice, int *unit, const char **name)
144 {
145 	static struct vdev_info {
146 		int major;
147 		int shift;
148 		int base;
149 		const char *name;
150 	} info[] = {
151 		{3,	6,	0,	"ad"},	/* ide0 */
152 		{22,	6,	2,	"ad"},	/* ide1 */
153 		{33,	6,	4,	"ad"},	/* ide2 */
154 		{34,	6,	6,	"ad"},	/* ide3 */
155 		{56,	6,	8,	"ad"},	/* ide4 */
156 		{57,	6,	10,	"ad"},	/* ide5 */
157 		{88,	6,	12,	"ad"},	/* ide6 */
158 		{89,	6,	14,	"ad"},	/* ide7 */
159 		{90,	6,	16,	"ad"},	/* ide8 */
160 		{91,	6,	18,	"ad"},	/* ide9 */
161 
162 		{8,	4,	0,	"da"},	/* scsi disk0 */
163 		{65,	4,	16,	"da"},	/* scsi disk1 */
164 		{66,	4,	32,	"da"},	/* scsi disk2 */
165 		{67,	4,	48,	"da"},	/* scsi disk3 */
166 		{68,	4,	64,	"da"},	/* scsi disk4 */
167 		{69,	4,	80,	"da"},	/* scsi disk5 */
168 		{70,	4,	96,	"da"},	/* scsi disk6 */
169 		{71,	4,	112,	"da"},	/* scsi disk7 */
170 		{128,	4,	128,	"da"},	/* scsi disk8 */
171 		{129,	4,	144,	"da"},	/* scsi disk9 */
172 		{130,	4,	160,	"da"},	/* scsi disk10 */
173 		{131,	4,	176,	"da"},	/* scsi disk11 */
174 		{132,	4,	192,	"da"},	/* scsi disk12 */
175 		{133,	4,	208,	"da"},	/* scsi disk13 */
176 		{134,	4,	224,	"da"},	/* scsi disk14 */
177 		{135,	4,	240,	"da"},	/* scsi disk15 */
178 
179 		{202,	4,	0,	"xbd"},	/* xbd */
180 
181 		{0,	0,	0,	NULL},
182 	};
183 	int major = vdevice >> 8;
184 	int minor = vdevice & 0xff;
185 	int i;
186 
187 	if (vdevice & (1 << 28)) {
188 		*unit = (vdevice & ((1 << 28) - 1)) >> 8;
189 		*name = "xbd";
190 		return;
191 	}
192 
193 	for (i = 0; info[i].major; i++) {
194 		if (info[i].major == major) {
195 			*unit = info[i].base + (minor >> info[i].shift);
196 			*name = info[i].name;
197 			return;
198 		}
199 	}
200 
201 	*unit = minor >> 4;
202 	*name = "xbd";
203 }
204 
205 int
206 xlvbd_add(struct xb_softc *sc, blkif_sector_t sectors,
207     int vdevice, uint16_t vdisk_info, unsigned long sector_size)
208 {
209 	int	unit, error = 0;
210 	const char *name;
211 
212 	blkfront_vdevice_to_unit(vdevice, &unit, &name);
213 
214 	sc->xb_unit = unit;
215 
216 	if (strcmp(name, "xbd"))
217 		device_printf(sc->xb_dev, "attaching as %s%d\n", name, unit);
218 
219 	sc->xb_disk = disk_alloc();
220 	sc->xb_disk->d_unit = sc->xb_unit;
221 	sc->xb_disk->d_open = blkif_open;
222 	sc->xb_disk->d_close = blkif_close;
223 	sc->xb_disk->d_ioctl = blkif_ioctl;
224 	sc->xb_disk->d_strategy = xb_strategy;
225 	sc->xb_disk->d_dump = xb_dump;
226 	sc->xb_disk->d_name = name;
227 	sc->xb_disk->d_drv1 = sc;
228 	sc->xb_disk->d_sectorsize = sector_size;
229 
230 	sc->xb_disk->d_mediasize = sectors * sector_size;
231 	sc->xb_disk->d_maxsize = sc->max_request_size;
232 	sc->xb_disk->d_flags = 0;
233 	disk_create(sc->xb_disk, DISK_VERSION_00);
234 
235 	return error;
236 }
237 
238 /************************ end VBD support *****************/
239 
240 /*
241  * Read/write routine for a buffer.  Finds the proper unit, place it on
242  * the sortq and kick the controller.
243  */
244 static void
245 xb_strategy(struct bio *bp)
246 {
247 	struct xb_softc	*sc = (struct xb_softc *)bp->bio_disk->d_drv1;
248 
249 	/* bogus disk? */
250 	if (sc == NULL) {
251 		bp->bio_error = EINVAL;
252 		bp->bio_flags |= BIO_ERROR;
253 		bp->bio_resid = bp->bio_bcount;
254 		biodone(bp);
255 		return;
256 	}
257 
258 	/*
259 	 * Place it in the queue of disk activities for this disk
260 	 */
261 	mtx_lock(&sc->xb_io_lock);
262 
263 	xb_enqueue_bio(sc, bp);
264 	xb_startio(sc);
265 
266 	mtx_unlock(&sc->xb_io_lock);
267 	return;
268 }
269 
270 static void
271 xb_bio_complete(struct xb_softc *sc, struct xb_command *cm)
272 {
273 	struct bio *bp;
274 
275 	bp = cm->bp;
276 
277 	if ( unlikely(cm->status != BLKIF_RSP_OKAY) ) {
278 		disk_err(bp, "disk error" , -1, 0);
279 		printf(" status: %x\n", cm->status);
280 		bp->bio_flags |= BIO_ERROR;
281 	}
282 
283 	if (bp->bio_flags & BIO_ERROR)
284 		bp->bio_error = EIO;
285 	else
286 		bp->bio_resid = 0;
287 
288 	xb_free_command(cm);
289 	biodone(bp);
290 }
291 
292 // Quiesce the disk writes for a dump file before allowing the next buffer.
293 static void
294 xb_quiesce(struct xb_softc *sc)
295 {
296 	int		mtd;
297 
298 	// While there are outstanding requests
299 	while (!TAILQ_EMPTY(&sc->cm_busy)) {
300 		RING_FINAL_CHECK_FOR_RESPONSES(&sc->ring, mtd);
301 		if (mtd) {
302 			/* Recieved request completions, update queue. */
303 			blkif_int(sc);
304 		}
305 		if (!TAILQ_EMPTY(&sc->cm_busy)) {
306 			/*
307 			 * Still pending requests, wait for the disk i/o
308 			 * to complete.
309 			 */
310 			HYPERVISOR_yield();
311 		}
312 	}
313 }
314 
315 /* Kernel dump function for a paravirtualized disk device */
316 static void
317 xb_dump_complete(struct xb_command *cm)
318 {
319 
320 	xb_enqueue_complete(cm);
321 }
322 
323 static int
324 xb_dump(void *arg, void *virtual, vm_offset_t physical, off_t offset,
325         size_t length)
326 {
327 	struct	disk   	*dp = arg;
328 	struct xb_softc	*sc = (struct xb_softc *) dp->d_drv1;
329 	struct xb_command *cm;
330 	size_t		chunk;
331 	int		sbp;
332 	int		rc = 0;
333 
334 	if (length <= 0)
335 		return (rc);
336 
337 	xb_quiesce(sc);	/* All quiet on the western front. */
338 
339 	/*
340 	 * If this lock is held, then this module is failing, and a
341 	 * successful kernel dump is highly unlikely anyway.
342 	 */
343 	mtx_lock(&sc->xb_io_lock);
344 
345 	/* Split the 64KB block as needed */
346 	for (sbp=0; length > 0; sbp++) {
347 		cm = xb_dequeue_free(sc);
348 		if (cm == NULL) {
349 			mtx_unlock(&sc->xb_io_lock);
350 			device_printf(sc->xb_dev, "dump: no more commands?\n");
351 			return (EBUSY);
352 		}
353 
354 		if (gnttab_alloc_grant_references(sc->max_request_segments,
355 						  &cm->gref_head) != 0) {
356 			xb_free_command(cm);
357 			mtx_unlock(&sc->xb_io_lock);
358 			device_printf(sc->xb_dev, "no more grant allocs?\n");
359 			return (EBUSY);
360 		}
361 
362 		chunk = length > sc->max_request_size
363 		      ? sc->max_request_size : length;
364 		cm->data = virtual;
365 		cm->datalen = chunk;
366 		cm->operation = BLKIF_OP_WRITE;
367 		cm->sector_number = offset / dp->d_sectorsize;
368 		cm->cm_complete = xb_dump_complete;
369 
370 		xb_enqueue_ready(cm);
371 
372 		length -= chunk;
373 		offset += chunk;
374 		virtual = (char *) virtual + chunk;
375 	}
376 
377 	/* Tell DOM0 to do the I/O */
378 	xb_startio(sc);
379 	mtx_unlock(&sc->xb_io_lock);
380 
381 	/* Poll for the completion. */
382 	xb_quiesce(sc);	/* All quite on the eastern front */
383 
384 	/* If there were any errors, bail out... */
385 	while ((cm = xb_dequeue_complete(sc)) != NULL) {
386 		if (cm->status != BLKIF_RSP_OKAY) {
387 			device_printf(sc->xb_dev,
388 			    "Dump I/O failed at sector %jd\n",
389 			    cm->sector_number);
390 			rc = EIO;
391 		}
392 		xb_free_command(cm);
393 	}
394 
395 	return (rc);
396 }
397 
398 
399 static int
400 blkfront_probe(device_t dev)
401 {
402 
403 	if (!strcmp(xenbus_get_type(dev), "vbd")) {
404 		device_set_desc(dev, "Virtual Block Device");
405 		device_quiet(dev);
406 		return (0);
407 	}
408 
409 	return (ENXIO);
410 }
411 
412 static void
413 xb_setup_sysctl(struct xb_softc *xb)
414 {
415 	struct sysctl_ctx_list *sysctl_ctx = NULL;
416 	struct sysctl_oid      *sysctl_tree = NULL;
417 
418 	sysctl_ctx = device_get_sysctl_ctx(xb->xb_dev);
419 	if (sysctl_ctx == NULL)
420 		return;
421 
422 	sysctl_tree = device_get_sysctl_tree(xb->xb_dev);
423 	if (sysctl_tree == NULL)
424 		return;
425 
426 	SYSCTL_ADD_UINT(sysctl_ctx, SYSCTL_CHILDREN(sysctl_tree), OID_AUTO,
427 		        "max_requests", CTLFLAG_RD, &xb->max_requests, -1,
428 		        "maximum outstanding requests (negotiated)");
429 
430 	SYSCTL_ADD_UINT(sysctl_ctx, SYSCTL_CHILDREN(sysctl_tree), OID_AUTO,
431 		        "max_request_segments", CTLFLAG_RD,
432 		        &xb->max_request_segments, 0,
433 		        "maximum number of pages per requests (negotiated)");
434 
435 	SYSCTL_ADD_UINT(sysctl_ctx, SYSCTL_CHILDREN(sysctl_tree), OID_AUTO,
436 		        "max_request_size", CTLFLAG_RD,
437 		        &xb->max_request_size, 0,
438 		        "maximum size in bytes of a request (negotiated)");
439 
440 	SYSCTL_ADD_UINT(sysctl_ctx, SYSCTL_CHILDREN(sysctl_tree), OID_AUTO,
441 		        "ring_pages", CTLFLAG_RD,
442 		        &xb->ring_pages, 0,
443 		        "communication channel pages (negotiated)");
444 }
445 
446 /*
447  * Setup supplies the backend dir, virtual device.  We place an event
448  * channel and shared frame entries.  We watch backend to wait if it's
449  * ok.
450  */
451 static int
452 blkfront_attach(device_t dev)
453 {
454 	struct xb_softc *sc;
455 	const char *name;
456 	uint32_t vdevice;
457 	int error;
458 	int i;
459 	int unit;
460 
461 	/* FIXME: Use dynamic device id if this is not set. */
462 	error = xs_scanf(XST_NIL, xenbus_get_node(dev),
463 	    "virtual-device", NULL, "%" PRIu32, &vdevice);
464 	if (error) {
465 		xenbus_dev_fatal(dev, error, "reading virtual-device");
466 		device_printf(dev, "Couldn't determine virtual device.\n");
467 		return (error);
468 	}
469 
470 	blkfront_vdevice_to_unit(vdevice, &unit, &name);
471 	if (!strcmp(name, "xbd"))
472 		device_set_unit(dev, unit);
473 
474 	sc = device_get_softc(dev);
475 	mtx_init(&sc->xb_io_lock, "blkfront i/o lock", NULL, MTX_DEF);
476 	xb_initq_free(sc);
477 	xb_initq_busy(sc);
478 	xb_initq_ready(sc);
479 	xb_initq_complete(sc);
480 	xb_initq_bio(sc);
481 	for (i = 0; i < XBF_MAX_RING_PAGES; i++)
482 		sc->ring_ref[i] = GRANT_INVALID_REF;
483 
484 	sc->xb_dev = dev;
485 	sc->vdevice = vdevice;
486 	sc->connected = BLKIF_STATE_DISCONNECTED;
487 
488 	xb_setup_sysctl(sc);
489 
490 	/* Wait for backend device to publish its protocol capabilities. */
491 	xenbus_set_state(dev, XenbusStateInitialising);
492 
493 	return (0);
494 }
495 
496 static int
497 blkfront_suspend(device_t dev)
498 {
499 	struct xb_softc *sc = device_get_softc(dev);
500 	int retval;
501 	int saved_state;
502 
503 	/* Prevent new requests being issued until we fix things up. */
504 	mtx_lock(&sc->xb_io_lock);
505 	saved_state = sc->connected;
506 	sc->connected = BLKIF_STATE_SUSPENDED;
507 
508 	/* Wait for outstanding I/O to drain. */
509 	retval = 0;
510 	while (TAILQ_EMPTY(&sc->cm_busy) == 0) {
511 		if (msleep(&sc->cm_busy, &sc->xb_io_lock,
512 			   PRIBIO, "blkf_susp", 30 * hz) == EWOULDBLOCK) {
513 			retval = EBUSY;
514 			break;
515 		}
516 	}
517 	mtx_unlock(&sc->xb_io_lock);
518 
519 	if (retval != 0)
520 		sc->connected = saved_state;
521 
522 	return (retval);
523 }
524 
525 static int
526 blkfront_resume(device_t dev)
527 {
528 	struct xb_softc *sc = device_get_softc(dev);
529 
530 	DPRINTK("blkfront_resume: %s\n", xenbus_get_node(dev));
531 
532 	blkif_free(sc);
533 	blkfront_initialize(sc);
534 	return (0);
535 }
536 
537 static void
538 blkfront_initialize(struct xb_softc *sc)
539 {
540 	const char *otherend_path;
541 	const char *node_path;
542 	uint32_t max_ring_page_order;
543 	int error;
544 	int i;
545 
546 	if (xenbus_get_state(sc->xb_dev) != XenbusStateInitialising) {
547 		/* Initialization has already been performed. */
548 		return;
549 	}
550 
551 	/*
552 	 * Protocol defaults valid even if negotiation for a
553 	 * setting fails.
554 	 */
555 	max_ring_page_order = 0;
556 	sc->ring_pages = 1;
557 	sc->max_request_segments = BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK;
558 	sc->max_request_size = XBF_SEGS_TO_SIZE(sc->max_request_segments);
559 	sc->max_request_blocks = BLKIF_SEGS_TO_BLOCKS(sc->max_request_segments);
560 
561 	/*
562 	 * Protocol negotiation.
563 	 *
564 	 * \note xs_gather() returns on the first encountered error, so
565 	 *       we must use independant calls in order to guarantee
566 	 *       we don't miss information in a sparsly populated back-end
567 	 *       tree.
568 	 *
569 	 * \note xs_scanf() does not update variables for unmatched
570 	 *	 fields.
571 	 */
572 	otherend_path = xenbus_get_otherend_path(sc->xb_dev);
573 	node_path = xenbus_get_node(sc->xb_dev);
574 
575 	/* Support both backend schemes for relaying ring page limits. */
576 	(void)xs_scanf(XST_NIL, otherend_path,
577 		       "max-ring-page-order", NULL, "%" PRIu32,
578 		       &max_ring_page_order);
579 	sc->ring_pages = 1 << max_ring_page_order;
580 	(void)xs_scanf(XST_NIL, otherend_path,
581 		       "max-ring-pages", NULL, "%" PRIu32,
582 		       &sc->ring_pages);
583 	if (sc->ring_pages < 1)
584 		sc->ring_pages = 1;
585 
586 	sc->max_requests = BLKIF_MAX_RING_REQUESTS(sc->ring_pages * PAGE_SIZE);
587 	(void)xs_scanf(XST_NIL, otherend_path,
588 		       "max-requests", NULL, "%" PRIu32,
589 		       &sc->max_requests);
590 
591 	(void)xs_scanf(XST_NIL, otherend_path,
592 		       "max-request-segments", NULL, "%" PRIu32,
593 		       &sc->max_request_segments);
594 
595 	(void)xs_scanf(XST_NIL, otherend_path,
596 		       "max-request-size", NULL, "%" PRIu32,
597 		       &sc->max_request_size);
598 
599 	if (sc->ring_pages > XBF_MAX_RING_PAGES) {
600 		device_printf(sc->xb_dev, "Back-end specified ring-pages of "
601 			      "%u limited to front-end limit of %zu.\n",
602 			      sc->ring_pages, XBF_MAX_RING_PAGES);
603 		sc->ring_pages = XBF_MAX_RING_PAGES;
604 	}
605 
606 	if (powerof2(sc->ring_pages) == 0) {
607 		uint32_t new_page_limit;
608 
609 		new_page_limit = 0x01 << (fls(sc->ring_pages) - 1);
610 		device_printf(sc->xb_dev, "Back-end specified ring-pages of "
611 			      "%u is not a power of 2. Limited to %u.\n",
612 			      sc->ring_pages, new_page_limit);
613 		sc->ring_pages = new_page_limit;
614 	}
615 
616 	if (sc->max_requests > XBF_MAX_REQUESTS) {
617 		device_printf(sc->xb_dev, "Back-end specified max_requests of "
618 			      "%u limited to front-end limit of %u.\n",
619 			      sc->max_requests, XBF_MAX_REQUESTS);
620 		sc->max_requests = XBF_MAX_REQUESTS;
621 	}
622 
623 	if (sc->max_request_segments > XBF_MAX_SEGMENTS_PER_REQUEST) {
624 		device_printf(sc->xb_dev, "Back-end specified "
625 			      "max_request_segments of %u limited to "
626 			      "front-end limit of %u.\n",
627 			      sc->max_request_segments,
628 			      XBF_MAX_SEGMENTS_PER_REQUEST);
629 		sc->max_request_segments = XBF_MAX_SEGMENTS_PER_REQUEST;
630 	}
631 
632 	if (sc->max_request_size > XBF_MAX_REQUEST_SIZE) {
633 		device_printf(sc->xb_dev, "Back-end specified "
634 			      "max_request_size of %u limited to front-end "
635 			      "limit of %u.\n", sc->max_request_size,
636 			      XBF_MAX_REQUEST_SIZE);
637 		sc->max_request_size = XBF_MAX_REQUEST_SIZE;
638 	}
639 
640  	if (sc->max_request_size > XBF_SEGS_TO_SIZE(sc->max_request_segments)) {
641  		device_printf(sc->xb_dev, "Back-end specified "
642  			      "max_request_size of %u limited to front-end "
643  			      "limit of %u.  (Too few segments.)\n",
644  			      sc->max_request_size,
645  			      XBF_SEGS_TO_SIZE(sc->max_request_segments));
646  		sc->max_request_size =
647  		    XBF_SEGS_TO_SIZE(sc->max_request_segments);
648  	}
649 
650 	sc->max_request_blocks = BLKIF_SEGS_TO_BLOCKS(sc->max_request_segments);
651 
652 	/* Allocate datastructures based on negotiated values. */
653 	error = bus_dma_tag_create(NULL,		/* parent */
654 				   512, PAGE_SIZE,	/* algnmnt, boundary */
655 				   BUS_SPACE_MAXADDR,	/* lowaddr */
656 				   BUS_SPACE_MAXADDR,	/* highaddr */
657 				   NULL, NULL,		/* filter, filterarg */
658 				   sc->max_request_size,
659 				   sc->max_request_segments,
660 				   PAGE_SIZE,		/* maxsegsize */
661 				   BUS_DMA_ALLOCNOW,	/* flags */
662 				   busdma_lock_mutex,	/* lockfunc */
663 				   &sc->xb_io_lock,	/* lockarg */
664 				   &sc->xb_io_dmat);
665 	if (error != 0) {
666 		xenbus_dev_fatal(sc->xb_dev, error,
667 				 "Cannot allocate parent DMA tag\n");
668 		return;
669 	}
670 
671 	/* Per-transaction data allocation. */
672 	sc->shadow = malloc(sizeof(*sc->shadow) * sc->max_requests,
673 			    M_XENBLOCKFRONT, M_NOWAIT|M_ZERO);
674 	if (sc->shadow == NULL) {
675 		bus_dma_tag_destroy(sc->xb_io_dmat);
676 		xenbus_dev_fatal(sc->xb_dev, error,
677 				 "Cannot allocate request structures\n");
678 		return;
679 	}
680 
681 	for (i = 0; i < sc->max_requests; i++) {
682 		struct xb_command *cm;
683 
684 		cm = &sc->shadow[i];
685 		cm->sg_refs = malloc(sizeof(grant_ref_t)
686 				   * sc->max_request_segments,
687 				     M_XENBLOCKFRONT, M_NOWAIT);
688 		if (cm->sg_refs == NULL)
689 			break;
690 		cm->id = i;
691 		cm->cm_sc = sc;
692 		if (bus_dmamap_create(sc->xb_io_dmat, 0, &cm->map) != 0)
693 			break;
694 		xb_free_command(cm);
695 	}
696 
697 	if (setup_blkring(sc) != 0)
698 		return;
699 
700 	/* Support both backend schemes for relaying ring page limits. */
701 	error = xs_printf(XST_NIL, node_path,
702 			 "num-ring-pages","%u", sc->ring_pages);
703 	if (error) {
704 		xenbus_dev_fatal(sc->xb_dev, error,
705 				 "writing %s/num-ring-pages",
706 				 node_path);
707 		return;
708 	}
709 	error = xs_printf(XST_NIL, node_path,
710 			 "ring-page-order","%u", fls(sc->ring_pages) - 1);
711 	if (error) {
712 		xenbus_dev_fatal(sc->xb_dev, error,
713 				 "writing %s/ring-page-order",
714 				 node_path);
715 		return;
716 	}
717 
718 	error = xs_printf(XST_NIL, node_path,
719 			 "max-requests","%u", sc->max_requests);
720 	if (error) {
721 		xenbus_dev_fatal(sc->xb_dev, error,
722 				 "writing %s/max-requests",
723 				 node_path);
724 		return;
725 	}
726 
727 	error = xs_printf(XST_NIL, node_path,
728 			 "max-request-segments","%u", sc->max_request_segments);
729 	if (error) {
730 		xenbus_dev_fatal(sc->xb_dev, error,
731 				 "writing %s/max-request-segments",
732 				 node_path);
733 		return;
734 	}
735 
736 	error = xs_printf(XST_NIL, node_path,
737 			 "max-request-size","%u", sc->max_request_size);
738 	if (error) {
739 		xenbus_dev_fatal(sc->xb_dev, error,
740 				 "writing %s/max-request-size",
741 				 node_path);
742 		return;
743 	}
744 
745 	error = xs_printf(XST_NIL, node_path, "event-channel",
746 			  "%u", irq_to_evtchn_port(sc->irq));
747 	if (error) {
748 		xenbus_dev_fatal(sc->xb_dev, error,
749 				 "writing %s/event-channel",
750 				 node_path);
751 		return;
752 	}
753 
754 	error = xs_printf(XST_NIL, node_path,
755 			  "protocol", "%s", XEN_IO_PROTO_ABI_NATIVE);
756 	if (error) {
757 		xenbus_dev_fatal(sc->xb_dev, error,
758 				 "writing %s/protocol",
759 				 node_path);
760 		return;
761 	}
762 
763 	xenbus_set_state(sc->xb_dev, XenbusStateInitialised);
764 }
765 
766 static int
767 setup_blkring(struct xb_softc *sc)
768 {
769 	blkif_sring_t *sring;
770 	uintptr_t sring_page_addr;
771 	int error;
772 	int i;
773 
774 	sring = malloc(sc->ring_pages * PAGE_SIZE, M_XENBLOCKFRONT,
775 		       M_NOWAIT|M_ZERO);
776 	if (sring == NULL) {
777 		xenbus_dev_fatal(sc->xb_dev, ENOMEM, "allocating shared ring");
778 		return (ENOMEM);
779 	}
780 	SHARED_RING_INIT(sring);
781 	FRONT_RING_INIT(&sc->ring, sring, sc->ring_pages * PAGE_SIZE);
782 
783 	for (i = 0, sring_page_addr = (uintptr_t)sring;
784 	     i < sc->ring_pages;
785 	     i++, sring_page_addr += PAGE_SIZE) {
786 
787 		error = xenbus_grant_ring(sc->xb_dev,
788 		    (vtomach(sring_page_addr) >> PAGE_SHIFT), &sc->ring_ref[i]);
789 		if (error) {
790 			xenbus_dev_fatal(sc->xb_dev, error,
791 					 "granting ring_ref(%d)", i);
792 			return (error);
793 		}
794 	}
795 	if (sc->ring_pages == 1) {
796 		error = xs_printf(XST_NIL, xenbus_get_node(sc->xb_dev),
797 				  "ring-ref", "%u", sc->ring_ref[0]);
798 		if (error) {
799 			xenbus_dev_fatal(sc->xb_dev, error,
800 					 "writing %s/ring-ref",
801 					 xenbus_get_node(sc->xb_dev));
802 			return (error);
803 		}
804 	} else {
805 		for (i = 0; i < sc->ring_pages; i++) {
806 			char ring_ref_name[]= "ring_refXX";
807 
808 			snprintf(ring_ref_name, sizeof(ring_ref_name),
809 				 "ring-ref%u", i);
810 			error = xs_printf(XST_NIL, xenbus_get_node(sc->xb_dev),
811 					 ring_ref_name, "%u", sc->ring_ref[i]);
812 			if (error) {
813 				xenbus_dev_fatal(sc->xb_dev, error,
814 						 "writing %s/%s",
815 						 xenbus_get_node(sc->xb_dev),
816 						 ring_ref_name);
817 				return (error);
818 			}
819 		}
820 	}
821 
822 	error = bind_listening_port_to_irqhandler(
823 	    xenbus_get_otherend_id(sc->xb_dev),
824 	    "xbd", (driver_intr_t *)blkif_int, sc,
825 	    INTR_TYPE_BIO | INTR_MPSAFE, &sc->irq);
826 	if (error) {
827 		xenbus_dev_fatal(sc->xb_dev, error,
828 		    "bind_evtchn_to_irqhandler failed");
829 		return (error);
830 	}
831 
832 	return (0);
833 }
834 
835 /**
836  * Callback received when the backend's state changes.
837  */
838 static void
839 blkfront_backend_changed(device_t dev, XenbusState backend_state)
840 {
841 	struct xb_softc *sc = device_get_softc(dev);
842 
843 	DPRINTK("backend_state=%d\n", backend_state);
844 
845 	switch (backend_state) {
846 	case XenbusStateUnknown:
847 	case XenbusStateInitialising:
848 	case XenbusStateReconfigured:
849 	case XenbusStateReconfiguring:
850 	case XenbusStateClosed:
851 		break;
852 
853 	case XenbusStateInitWait:
854 	case XenbusStateInitialised:
855 		blkfront_initialize(sc);
856 		break;
857 
858 	case XenbusStateConnected:
859 		blkfront_initialize(sc);
860 		blkfront_connect(sc);
861 		break;
862 
863 	case XenbusStateClosing:
864 		if (sc->users > 0)
865 			xenbus_dev_error(dev, -EBUSY,
866 					 "Device in use; refusing to close");
867 		else
868 			blkfront_closing(dev);
869 		break;
870 	}
871 }
872 
873 /*
874 ** Invoked when the backend is finally 'ready' (and has published
875 ** the details about the physical device - #sectors, size, etc).
876 */
877 static void
878 blkfront_connect(struct xb_softc *sc)
879 {
880 	device_t dev = sc->xb_dev;
881 	unsigned long sectors, sector_size;
882 	unsigned int binfo;
883 	int err, feature_barrier;
884 
885 	if( (sc->connected == BLKIF_STATE_CONNECTED) ||
886 	    (sc->connected == BLKIF_STATE_SUSPENDED) )
887 		return;
888 
889 	DPRINTK("blkfront.c:connect:%s.\n", xenbus_get_otherend_path(dev));
890 
891 	err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev),
892 			"sectors", "%lu", &sectors,
893 			"info", "%u", &binfo,
894 			"sector-size", "%lu", &sector_size,
895 			NULL);
896 	if (err) {
897 		xenbus_dev_fatal(dev, err,
898 		    "reading backend fields at %s",
899 		    xenbus_get_otherend_path(dev));
900 		return;
901 	}
902 	err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev),
903 			"feature-barrier", "%lu", &feature_barrier,
904 			NULL);
905 	if (!err || feature_barrier)
906 		sc->xb_flags |= XB_BARRIER;
907 
908 	if (sc->xb_disk == NULL) {
909 		device_printf(dev, "%juMB <%s> at %s",
910 		    (uintmax_t) sectors / (1048576 / sector_size),
911 		    device_get_desc(dev),
912 		    xenbus_get_node(dev));
913 		bus_print_child_footer(device_get_parent(dev), dev);
914 
915 		xlvbd_add(sc, sectors, sc->vdevice, binfo, sector_size);
916 	}
917 
918 	(void)xenbus_set_state(dev, XenbusStateConnected);
919 
920 	/* Kick pending requests. */
921 	mtx_lock(&sc->xb_io_lock);
922 	sc->connected = BLKIF_STATE_CONNECTED;
923 	xb_startio(sc);
924 	sc->xb_flags |= XB_READY;
925 	mtx_unlock(&sc->xb_io_lock);
926 }
927 
928 /**
929  * Handle the change of state of the backend to Closing.  We must delete our
930  * device-layer structures now, to ensure that writes are flushed through to
931  * the backend.  Once this is done, we can switch to Closed in
932  * acknowledgement.
933  */
934 static void
935 blkfront_closing(device_t dev)
936 {
937 	struct xb_softc *sc = device_get_softc(dev);
938 
939 	xenbus_set_state(dev, XenbusStateClosing);
940 
941 	DPRINTK("blkfront_closing: %s removed\n", xenbus_get_node(dev));
942 
943 	if (sc->xb_disk != NULL) {
944 		disk_destroy(sc->xb_disk);
945 		sc->xb_disk = NULL;
946 	}
947 
948 	xenbus_set_state(dev, XenbusStateClosed);
949 }
950 
951 
952 static int
953 blkfront_detach(device_t dev)
954 {
955 	struct xb_softc *sc = device_get_softc(dev);
956 
957 	DPRINTK("blkfront_remove: %s removed\n", xenbus_get_node(dev));
958 
959 	blkif_free(sc);
960 	mtx_destroy(&sc->xb_io_lock);
961 
962 	return 0;
963 }
964 
965 
966 static inline void
967 flush_requests(struct xb_softc *sc)
968 {
969 	int notify;
970 
971 	RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&sc->ring, notify);
972 
973 	if (notify)
974 		notify_remote_via_irq(sc->irq);
975 }
976 
977 static void
978 blkif_restart_queue_callback(void *arg)
979 {
980 	struct xb_softc *sc = arg;
981 
982 	mtx_lock(&sc->xb_io_lock);
983 
984 	xb_startio(sc);
985 
986 	mtx_unlock(&sc->xb_io_lock);
987 }
988 
989 static int
990 blkif_open(struct disk *dp)
991 {
992 	struct xb_softc	*sc = (struct xb_softc *)dp->d_drv1;
993 
994 	if (sc == NULL) {
995 		printf("xb%d: not found", sc->xb_unit);
996 		return (ENXIO);
997 	}
998 
999 	sc->xb_flags |= XB_OPEN;
1000 	sc->users++;
1001 	return (0);
1002 }
1003 
1004 static int
1005 blkif_close(struct disk *dp)
1006 {
1007 	struct xb_softc	*sc = (struct xb_softc *)dp->d_drv1;
1008 
1009 	if (sc == NULL)
1010 		return (ENXIO);
1011 	sc->xb_flags &= ~XB_OPEN;
1012 	if (--(sc->users) == 0) {
1013 		/*
1014 		 * Check whether we have been instructed to close.  We will
1015 		 * have ignored this request initially, as the device was
1016 		 * still mounted.
1017 		 */
1018 		if (xenbus_get_otherend_state(sc->xb_dev) == XenbusStateClosing)
1019 			blkfront_closing(sc->xb_dev);
1020 	}
1021 	return (0);
1022 }
1023 
1024 static int
1025 blkif_ioctl(struct disk *dp, u_long cmd, void *addr, int flag, struct thread *td)
1026 {
1027 	struct xb_softc	*sc = (struct xb_softc *)dp->d_drv1;
1028 
1029 	if (sc == NULL)
1030 		return (ENXIO);
1031 
1032 	return (ENOTTY);
1033 }
1034 
1035 static void
1036 xb_free_command(struct xb_command *cm)
1037 {
1038 
1039 	KASSERT((cm->cm_flags & XB_ON_XBQ_MASK) == 0,
1040 	    ("Freeing command that is still on a queue\n"));
1041 
1042 	cm->cm_flags = 0;
1043 	cm->bp = NULL;
1044 	cm->cm_complete = NULL;
1045 	xb_enqueue_free(cm);
1046 }
1047 
1048 /*
1049  * blkif_queue_request
1050  *
1051  * request block io
1052  *
1053  * id: for guest use only.
1054  * operation: BLKIF_OP_{READ,WRITE,PROBE}
1055  * buffer: buffer to read/write into. this should be a
1056  *   virtual address in the guest os.
1057  */
1058 static struct xb_command *
1059 xb_bio_command(struct xb_softc *sc)
1060 {
1061 	struct xb_command *cm;
1062 	struct bio *bp;
1063 
1064 	if (unlikely(sc->connected != BLKIF_STATE_CONNECTED))
1065 		return (NULL);
1066 
1067 	bp = xb_dequeue_bio(sc);
1068 	if (bp == NULL)
1069 		return (NULL);
1070 
1071 	if ((cm = xb_dequeue_free(sc)) == NULL) {
1072 		xb_requeue_bio(sc, bp);
1073 		return (NULL);
1074 	}
1075 
1076 	if (gnttab_alloc_grant_references(sc->max_request_segments,
1077 	    &cm->gref_head) != 0) {
1078 		gnttab_request_free_callback(&sc->callback,
1079 			blkif_restart_queue_callback, sc,
1080 			sc->max_request_segments);
1081 		xb_requeue_bio(sc, bp);
1082 		xb_enqueue_free(cm);
1083 		sc->xb_flags |= XB_FROZEN;
1084 		return (NULL);
1085 	}
1086 
1087 	cm->bp = bp;
1088 	cm->data = bp->bio_data;
1089 	cm->datalen = bp->bio_bcount;
1090 	cm->operation = (bp->bio_cmd == BIO_READ) ? BLKIF_OP_READ :
1091 	    BLKIF_OP_WRITE;
1092 	cm->sector_number = (blkif_sector_t)bp->bio_pblkno;
1093 
1094 	return (cm);
1095 }
1096 
1097 static int
1098 blkif_queue_request(struct xb_softc *sc, struct xb_command *cm)
1099 {
1100 	int	error;
1101 
1102 	error = bus_dmamap_load(sc->xb_io_dmat, cm->map, cm->data, cm->datalen,
1103 	    blkif_queue_cb, cm, 0);
1104 	if (error == EINPROGRESS) {
1105 		printf("EINPROGRESS\n");
1106 		sc->xb_flags |= XB_FROZEN;
1107 		cm->cm_flags |= XB_CMD_FROZEN;
1108 		return (0);
1109 	}
1110 
1111 	return (error);
1112 }
1113 
1114 static void
1115 blkif_queue_cb(void *arg, bus_dma_segment_t *segs, int nsegs, int error)
1116 {
1117 	struct xb_softc *sc;
1118 	struct xb_command *cm;
1119 	blkif_request_t	*ring_req;
1120 	struct blkif_request_segment *sg;
1121 	struct blkif_request_segment *last_block_sg;
1122 	grant_ref_t *sg_ref;
1123 	vm_paddr_t buffer_ma;
1124 	uint64_t fsect, lsect;
1125 	int ref;
1126 	int op;
1127 	int block_segs;
1128 
1129 	cm = arg;
1130 	sc = cm->cm_sc;
1131 
1132 //printf("%s: Start\n", __func__);
1133 	if (error) {
1134 		printf("error %d in blkif_queue_cb\n", error);
1135 		cm->bp->bio_error = EIO;
1136 		biodone(cm->bp);
1137 		xb_free_command(cm);
1138 		return;
1139 	}
1140 
1141 	/* Fill out a communications ring structure. */
1142 	ring_req = RING_GET_REQUEST(&sc->ring, sc->ring.req_prod_pvt);
1143 	sc->ring.req_prod_pvt++;
1144 	ring_req->id = cm->id;
1145 	ring_req->operation = cm->operation;
1146 	ring_req->sector_number = cm->sector_number;
1147 	ring_req->handle = (blkif_vdev_t)(uintptr_t)sc->xb_disk;
1148 	ring_req->nr_segments = nsegs;
1149 	cm->nseg = nsegs;
1150 
1151 	block_segs    = MIN(nsegs, BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK);
1152 	sg            = ring_req->seg;
1153 	last_block_sg = sg + block_segs;
1154 	sg_ref        = cm->sg_refs;
1155 
1156 	while (1) {
1157 
1158 		while (sg < last_block_sg) {
1159 			buffer_ma = segs->ds_addr;
1160 			fsect = (buffer_ma & PAGE_MASK) >> XBD_SECTOR_SHFT;
1161 			lsect = fsect + (segs->ds_len  >> XBD_SECTOR_SHFT) - 1;
1162 
1163 			KASSERT(lsect <= 7, ("XEN disk driver data cannot "
1164 				"cross a page boundary"));
1165 
1166 			/* install a grant reference. */
1167 			ref = gnttab_claim_grant_reference(&cm->gref_head);
1168 
1169 			/*
1170 			 * GNTTAB_LIST_END == 0xffffffff, but it is private
1171 			 * to gnttab.c.
1172 			 */
1173 			KASSERT(ref != ~0, ("grant_reference failed"));
1174 
1175 			gnttab_grant_foreign_access_ref(
1176 				ref,
1177 				xenbus_get_otherend_id(sc->xb_dev),
1178 				buffer_ma >> PAGE_SHIFT,
1179 				ring_req->operation == BLKIF_OP_WRITE);
1180 
1181 			*sg_ref = ref;
1182 			*sg = (struct blkif_request_segment) {
1183 				.gref       = ref,
1184 				.first_sect = fsect,
1185 				.last_sect  = lsect };
1186 			sg++;
1187 			sg_ref++;
1188 			segs++;
1189 			nsegs--;
1190 		}
1191 		block_segs = MIN(nsegs, BLKIF_MAX_SEGMENTS_PER_SEGMENT_BLOCK);
1192 		if (block_segs == 0)
1193 			break;
1194 
1195 		sg = BLKRING_GET_SEG_BLOCK(&sc->ring, sc->ring.req_prod_pvt);
1196 		sc->ring.req_prod_pvt++;
1197 		last_block_sg = sg + block_segs;
1198 	}
1199 
1200 	if (cm->operation == BLKIF_OP_READ)
1201 		op = BUS_DMASYNC_PREREAD;
1202 	else if (cm->operation == BLKIF_OP_WRITE)
1203 		op = BUS_DMASYNC_PREWRITE;
1204 	else
1205 		op = 0;
1206 	bus_dmamap_sync(sc->xb_io_dmat, cm->map, op);
1207 
1208 	gnttab_free_grant_references(cm->gref_head);
1209 
1210 	xb_enqueue_busy(cm);
1211 
1212 	/*
1213 	 * This flag means that we're probably executing in the busdma swi
1214 	 * instead of in the startio context, so an explicit flush is needed.
1215 	 */
1216 	if (cm->cm_flags & XB_CMD_FROZEN)
1217 		flush_requests(sc);
1218 
1219 //printf("%s: Done\n", __func__);
1220 	return;
1221 }
1222 
1223 /*
1224  * Dequeue buffers and place them in the shared communication ring.
1225  * Return when no more requests can be accepted or all buffers have
1226  * been queued.
1227  *
1228  * Signal XEN once the ring has been filled out.
1229  */
1230 static void
1231 xb_startio(struct xb_softc *sc)
1232 {
1233 	struct xb_command *cm;
1234 	int error, queued = 0;
1235 
1236 	mtx_assert(&sc->xb_io_lock, MA_OWNED);
1237 
1238 	if (sc->connected != BLKIF_STATE_CONNECTED)
1239 		return;
1240 
1241 	while (RING_FREE_REQUESTS(&sc->ring) >= sc->max_request_blocks) {
1242 		if (sc->xb_flags & XB_FROZEN)
1243 			break;
1244 
1245 		cm = xb_dequeue_ready(sc);
1246 
1247 		if (cm == NULL)
1248 		    cm = xb_bio_command(sc);
1249 
1250 		if (cm == NULL)
1251 			break;
1252 
1253 		if ((error = blkif_queue_request(sc, cm)) != 0) {
1254 			printf("blkif_queue_request returned %d\n", error);
1255 			break;
1256 		}
1257 		queued++;
1258 	}
1259 
1260 	if (queued != 0)
1261 		flush_requests(sc);
1262 }
1263 
1264 static void
1265 blkif_int(void *xsc)
1266 {
1267 	struct xb_softc *sc = xsc;
1268 	struct xb_command *cm;
1269 	blkif_response_t *bret;
1270 	RING_IDX i, rp;
1271 	int op;
1272 
1273 	mtx_lock(&sc->xb_io_lock);
1274 
1275 	if (unlikely(sc->connected == BLKIF_STATE_DISCONNECTED)) {
1276 		mtx_unlock(&sc->xb_io_lock);
1277 		return;
1278 	}
1279 
1280  again:
1281 	rp = sc->ring.sring->rsp_prod;
1282 	rmb(); /* Ensure we see queued responses up to 'rp'. */
1283 
1284 	for (i = sc->ring.rsp_cons; i != rp;) {
1285 		bret = RING_GET_RESPONSE(&sc->ring, i);
1286 		cm   = &sc->shadow[bret->id];
1287 
1288 		xb_remove_busy(cm);
1289 		i += blkif_completion(cm);
1290 
1291 		if (cm->operation == BLKIF_OP_READ)
1292 			op = BUS_DMASYNC_POSTREAD;
1293 		else if (cm->operation == BLKIF_OP_WRITE)
1294 			op = BUS_DMASYNC_POSTWRITE;
1295 		else
1296 			op = 0;
1297 		bus_dmamap_sync(sc->xb_io_dmat, cm->map, op);
1298 		bus_dmamap_unload(sc->xb_io_dmat, cm->map);
1299 
1300 		/*
1301 		 * If commands are completing then resources are probably
1302 		 * being freed as well.  It's a cheap assumption even when
1303 		 * wrong.
1304 		 */
1305 		sc->xb_flags &= ~XB_FROZEN;
1306 
1307 		/*
1308 		 * Directly call the i/o complete routine to save an
1309 		 * an indirection in the common case.
1310 		 */
1311 		cm->status = bret->status;
1312 		if (cm->bp)
1313 			xb_bio_complete(sc, cm);
1314 		else if (cm->cm_complete)
1315 			(cm->cm_complete)(cm);
1316 		else
1317 			xb_free_command(cm);
1318 	}
1319 
1320 	sc->ring.rsp_cons = i;
1321 
1322 	if (i != sc->ring.req_prod_pvt) {
1323 		int more_to_do;
1324 		RING_FINAL_CHECK_FOR_RESPONSES(&sc->ring, more_to_do);
1325 		if (more_to_do)
1326 			goto again;
1327 	} else {
1328 		sc->ring.sring->rsp_event = i + 1;
1329 	}
1330 
1331 	xb_startio(sc);
1332 
1333 	if (unlikely(sc->connected == BLKIF_STATE_SUSPENDED))
1334 		wakeup(&sc->cm_busy);
1335 
1336 	mtx_unlock(&sc->xb_io_lock);
1337 }
1338 
1339 static void
1340 blkif_free(struct xb_softc *sc)
1341 {
1342 	uint8_t *sring_page_ptr;
1343 	int i;
1344 
1345 	/* Prevent new requests being issued until we fix things up. */
1346 	mtx_lock(&sc->xb_io_lock);
1347 	sc->connected = BLKIF_STATE_DISCONNECTED;
1348 	mtx_unlock(&sc->xb_io_lock);
1349 
1350 	/* Free resources associated with old device channel. */
1351 	if (sc->ring.sring != NULL) {
1352 		sring_page_ptr = (uint8_t *)sc->ring.sring;
1353 		for (i = 0; i < sc->ring_pages; i++) {
1354 			if (sc->ring_ref[i] != GRANT_INVALID_REF) {
1355 				gnttab_end_foreign_access_ref(sc->ring_ref[i]);
1356 				sc->ring_ref[i] = GRANT_INVALID_REF;
1357 			}
1358 			sring_page_ptr += PAGE_SIZE;
1359 		}
1360 		free(sc->ring.sring, M_XENBLOCKFRONT);
1361 		sc->ring.sring = NULL;
1362 	}
1363 
1364 	if (sc->shadow) {
1365 
1366 		for (i = 0; i < sc->max_requests; i++) {
1367 			struct xb_command *cm;
1368 
1369 			cm = &sc->shadow[i];
1370 			if (cm->sg_refs != NULL) {
1371 				free(cm->sg_refs, M_XENBLOCKFRONT);
1372 				cm->sg_refs = NULL;
1373 			}
1374 
1375 			bus_dmamap_destroy(sc->xb_io_dmat, cm->map);
1376 		}
1377 		free(sc->shadow, M_XENBLOCKFRONT);
1378 		sc->shadow = NULL;
1379 
1380 		bus_dma_tag_destroy(sc->xb_io_dmat);
1381 
1382 		xb_initq_free(sc);
1383 		xb_initq_ready(sc);
1384 		xb_initq_complete(sc);
1385 	}
1386 
1387 	if (sc->irq) {
1388 		unbind_from_irqhandler(sc->irq);
1389 		sc->irq = 0;
1390 	}
1391 }
1392 
1393 static int
1394 blkif_completion(struct xb_command *s)
1395 {
1396 //printf("%s: Req %p(%d)\n", __func__, s, s->nseg);
1397 	gnttab_end_foreign_access_references(s->nseg, s->sg_refs);
1398 	return (BLKIF_SEGS_TO_BLOCKS(s->nseg));
1399 }
1400 
1401 /* ** Driver registration ** */
1402 static device_method_t blkfront_methods[] = {
1403 	/* Device interface */
1404 	DEVMETHOD(device_probe,         blkfront_probe),
1405 	DEVMETHOD(device_attach,        blkfront_attach),
1406 	DEVMETHOD(device_detach,        blkfront_detach),
1407 	DEVMETHOD(device_shutdown,      bus_generic_shutdown),
1408 	DEVMETHOD(device_suspend,       blkfront_suspend),
1409 	DEVMETHOD(device_resume,        blkfront_resume),
1410 
1411 	/* Xenbus interface */
1412 	DEVMETHOD(xenbus_otherend_changed, blkfront_backend_changed),
1413 
1414 	{ 0, 0 }
1415 };
1416 
1417 static driver_t blkfront_driver = {
1418 	"xbd",
1419 	blkfront_methods,
1420 	sizeof(struct xb_softc),
1421 };
1422 devclass_t blkfront_devclass;
1423 
1424 DRIVER_MODULE(xbd, xenbusb_front, blkfront_driver, blkfront_devclass, 0, 0);
1425