4 * XenLinux virtual block device driver.
6 * Copyright (c) 2003-2004, Keir Fraser & Steve Hand
7 * Modifications by Mark A. Williamson are (c) Intel Research Cambridge
8 * Copyright (c) 2004, Christian Limpach
9 * Copyright (c) 2004, Andrew Warfield
10 * Copyright (c) 2005, Christopher Clark
11 * Copyright (c) 2005, XenSource Ltd
13 * This program is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU General Public License version 2
15 * as published by the Free Software Foundation; or, when distributed
16 * separately from the Linux kernel or incorporated into other
17 * software packages, subject to the following license:
19 * Permission is hereby granted, free of charge, to any person obtaining a copy
20 * of this source file (the "Software"), to deal in the Software without
21 * restriction, including without limitation the rights to use, copy, modify,
22 * merge, publish, distribute, sublicense, and/or sell copies of the Software,
23 * and to permit persons to whom the Software is furnished to do so, subject to
24 * the following conditions:
26 * The above copyright notice and this permission notice shall be included in
27 * all copies or substantial portions of the Software.
29 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
30 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
31 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
32 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
33 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
34 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
38 #include <linux/interrupt.h>
39 #include <linux/blkdev.h>
40 #include <linux/hdreg.h>
41 #include <linux/cdrom.h>
42 #include <linux/module.h>
43 #include <linux/scatterlist.h>
45 #include <xen/xenbus.h>
46 #include <xen/grant_table.h>
47 #include <xen/events.h>
50 #include <xen/interface/grant_table.h>
51 #include <xen/interface/io/blkif.h>
52 #include <xen/interface/io/protocols.h>
54 #include <asm/xen/hypervisor.h>
57 BLKIF_STATE_DISCONNECTED,
58 BLKIF_STATE_CONNECTED,
59 BLKIF_STATE_SUSPENDED,
63 struct blkif_request req;
64 unsigned long request;
65 unsigned long frame[BLKIF_MAX_SEGMENTS_PER_REQUEST];
68 static struct block_device_operations xlvbd_block_fops;
70 #define BLK_RING_SIZE __RING_SIZE((struct blkif_sring *)0, PAGE_SIZE)
73 * We have one of these per vbd, whether ide, scsi or 'other'. They
74 * hang in private_data off the gendisk structure. We may end up
75 * putting all kinds of interesting stuff here :-)
79 struct xenbus_device *xbdev;
83 enum blkif_state connected;
85 struct blkif_front_ring ring;
86 struct scatterlist sg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
87 unsigned int evtchn, irq;
88 struct request_queue *rq;
89 struct work_struct work;
90 struct gnttab_free_callback callback;
91 struct blk_shadow shadow[BLK_RING_SIZE];
92 unsigned long shadow_free;
97 * The number of people holding this device open. We won't allow a
98 * hot-unplug unless this is 0.
103 static DEFINE_SPINLOCK(blkif_io_lock);
105 #define MAXIMUM_OUTSTANDING_BLOCK_REQS \
106 (BLKIF_MAX_SEGMENTS_PER_REQUEST * BLK_RING_SIZE)
107 #define GRANT_INVALID_REF 0
109 #define PARTS_PER_DISK 16
110 #define PARTS_PER_EXT_DISK 256
112 #define BLKIF_MAJOR(dev) ((dev)>>8)
113 #define BLKIF_MINOR(dev) ((dev) & 0xff)
116 #define EXTENDED (1<<EXT_SHIFT)
117 #define VDEV_IS_EXTENDED(dev) ((dev)&(EXTENDED))
118 #define BLKIF_MINOR_EXT(dev) ((dev)&(~EXTENDED))
120 #define DEV_NAME "xvd" /* name in /dev */
122 static int get_id_from_freelist(struct blkfront_info *info)
124 unsigned long free = info->shadow_free;
125 BUG_ON(free > BLK_RING_SIZE);
126 info->shadow_free = info->shadow[free].req.id;
127 info->shadow[free].req.id = 0x0fffffee; /* debug */
131 static void add_id_to_freelist(struct blkfront_info *info,
134 info->shadow[id].req.id = info->shadow_free;
135 info->shadow[id].request = 0;
136 info->shadow_free = id;
139 static void blkif_restart_queue_callback(void *arg)
141 struct blkfront_info *info = (struct blkfront_info *)arg;
142 schedule_work(&info->work);
145 static int blkif_getgeo(struct block_device *bd, struct hd_geometry *hg)
147 /* We don't have real geometry info, but let's at least return
148 values consistent with the size of the device */
149 sector_t nsect = get_capacity(bd->bd_disk);
150 sector_t cylinders = nsect;
154 sector_div(cylinders, hg->heads * hg->sectors);
155 hg->cylinders = cylinders;
156 if ((sector_t)(hg->cylinders + 1) * hg->heads * hg->sectors < nsect)
157 hg->cylinders = 0xffff;
161 static int blkif_ioctl(struct block_device *bdev, fmode_t mode,
162 unsigned command, unsigned long argument)
164 struct blkfront_info *info = bdev->bd_disk->private_data;
167 dev_dbg(&info->xbdev->dev, "command: 0x%x, argument: 0x%lx\n",
168 command, (long)argument);
171 case CDROMMULTISESSION:
172 dev_dbg(&info->xbdev->dev, "FIXME: support multisession CDs later\n");
173 for (i = 0; i < sizeof(struct cdrom_multisession); i++)
174 if (put_user(0, (char __user *)(argument + i)))
178 case CDROM_GET_CAPABILITY: {
179 struct gendisk *gd = info->gd;
180 if (gd->flags & GENHD_FL_CD)
186 /*printk(KERN_ALERT "ioctl %08x not supported by Xen blkdev\n",
188 return -EINVAL; /* same return as native Linux */
195 * blkif_queue_request
199 * id: for guest use only.
200 * operation: BLKIF_OP_{READ,WRITE,PROBE}
201 * buffer: buffer to read/write into. this should be a
202 * virtual address in the guest os.
204 static int blkif_queue_request(struct request *req)
206 struct blkfront_info *info = req->rq_disk->private_data;
207 unsigned long buffer_mfn;
208 struct blkif_request *ring_req;
210 unsigned int fsect, lsect;
212 grant_ref_t gref_head;
213 struct scatterlist *sg;
215 if (unlikely(info->connected != BLKIF_STATE_CONNECTED))
218 if (gnttab_alloc_grant_references(
219 BLKIF_MAX_SEGMENTS_PER_REQUEST, &gref_head) < 0) {
220 gnttab_request_free_callback(
222 blkif_restart_queue_callback,
224 BLKIF_MAX_SEGMENTS_PER_REQUEST);
228 /* Fill out a communications ring structure. */
229 ring_req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
230 id = get_id_from_freelist(info);
231 info->shadow[id].request = (unsigned long)req;
234 ring_req->sector_number = (blkif_sector_t)blk_rq_pos(req);
235 ring_req->handle = info->handle;
237 ring_req->operation = rq_data_dir(req) ?
238 BLKIF_OP_WRITE : BLKIF_OP_READ;
239 if (blk_barrier_rq(req))
240 ring_req->operation = BLKIF_OP_WRITE_BARRIER;
242 ring_req->nr_segments = blk_rq_map_sg(req->q, req, info->sg);
243 BUG_ON(ring_req->nr_segments > BLKIF_MAX_SEGMENTS_PER_REQUEST);
245 for_each_sg(info->sg, sg, ring_req->nr_segments, i) {
246 buffer_mfn = pfn_to_mfn(page_to_pfn(sg_page(sg)));
247 fsect = sg->offset >> 9;
248 lsect = fsect + (sg->length >> 9) - 1;
249 /* install a grant reference. */
250 ref = gnttab_claim_grant_reference(&gref_head);
251 BUG_ON(ref == -ENOSPC);
253 gnttab_grant_foreign_access_ref(
255 info->xbdev->otherend_id,
259 info->shadow[id].frame[i] = mfn_to_pfn(buffer_mfn);
261 (struct blkif_request_segment) {
264 .last_sect = lsect };
267 info->ring.req_prod_pvt++;
269 /* Keep a private copy so we can reissue requests when recovering. */
270 info->shadow[id].req = *ring_req;
272 gnttab_free_grant_references(gref_head);
278 static inline void flush_requests(struct blkfront_info *info)
282 RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&info->ring, notify);
285 notify_remote_via_irq(info->irq);
290 * read a block; request is in a request queue
292 static void do_blkif_request(struct request_queue *rq)
294 struct blkfront_info *info = NULL;
298 pr_debug("Entered do_blkif_request\n");
302 while ((req = elv_next_request(rq)) != NULL) {
303 info = req->rq_disk->private_data;
304 if (!blk_fs_request(req)) {
305 __blk_end_request_cur(req, -EIO);
309 if (RING_FULL(&info->ring))
312 pr_debug("do_blk_req %p: cmd %p, sec %lx, "
313 "(%u/%u) buffer:%p [%s]\n",
314 req, req->cmd, (unsigned long)blk_rq_pos(req),
315 blk_rq_cur_sectors(req), blk_rq_sectors(req),
316 req->buffer, rq_data_dir(req) ? "write" : "read");
319 blkdev_dequeue_request(req);
320 if (blkif_queue_request(req)) {
321 blk_requeue_request(rq, req);
323 /* Avoid pointless unplugs. */
332 flush_requests(info);
335 static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size)
337 struct request_queue *rq;
339 rq = blk_init_queue(do_blkif_request, &blkif_io_lock);
343 queue_flag_set_unlocked(QUEUE_FLAG_VIRT, rq);
345 /* Hard sector size and max sectors impersonate the equiv. hardware. */
346 blk_queue_hardsect_size(rq, sector_size);
347 blk_queue_max_sectors(rq, 512);
349 /* Each segment in a request is up to an aligned page in size. */
350 blk_queue_segment_boundary(rq, PAGE_SIZE - 1);
351 blk_queue_max_segment_size(rq, PAGE_SIZE);
353 /* Ensure a merged request will fit in a single I/O ring slot. */
354 blk_queue_max_phys_segments(rq, BLKIF_MAX_SEGMENTS_PER_REQUEST);
355 blk_queue_max_hw_segments(rq, BLKIF_MAX_SEGMENTS_PER_REQUEST);
357 /* Make sure buffer addresses are sector-aligned. */
358 blk_queue_dma_alignment(rq, 511);
360 /* Make sure we don't use bounce buffers. */
361 blk_queue_bounce_limit(rq, BLK_BOUNCE_ANY);
369 static int xlvbd_barrier(struct blkfront_info *info)
373 err = blk_queue_ordered(info->rq,
374 info->feature_barrier ? QUEUE_ORDERED_DRAIN : QUEUE_ORDERED_NONE,
380 printk(KERN_INFO "blkfront: %s: barriers %s\n",
382 info->feature_barrier ? "enabled" : "disabled");
387 static int xlvbd_alloc_gendisk(blkif_sector_t capacity,
388 struct blkfront_info *info,
389 u16 vdisk_info, u16 sector_size)
398 BUG_ON(info->gd != NULL);
399 BUG_ON(info->rq != NULL);
401 if ((info->vdevice>>EXT_SHIFT) > 1) {
402 /* this is above the extended range; something is wrong */
403 printk(KERN_WARNING "blkfront: vdevice 0x%x is above the extended range; ignoring\n", info->vdevice);
407 if (!VDEV_IS_EXTENDED(info->vdevice)) {
408 minor = BLKIF_MINOR(info->vdevice);
409 nr_parts = PARTS_PER_DISK;
411 minor = BLKIF_MINOR_EXT(info->vdevice);
412 nr_parts = PARTS_PER_EXT_DISK;
415 if ((minor % nr_parts) == 0)
416 nr_minors = nr_parts;
418 gd = alloc_disk(nr_minors);
422 offset = minor / nr_parts;
426 sprintf(gd->disk_name, "%s%c", DEV_NAME, 'a' + offset);
428 sprintf(gd->disk_name, "%s%c%c", DEV_NAME,
429 'a' + ((offset / 26)-1), 'a' + (offset % 26));
432 sprintf(gd->disk_name, "%s%c%d", DEV_NAME,
434 minor & (nr_parts - 1));
436 sprintf(gd->disk_name, "%s%c%c%d", DEV_NAME,
437 'a' + ((offset / 26) - 1),
439 minor & (nr_parts - 1));
442 gd->major = XENVBD_MAJOR;
443 gd->first_minor = minor;
444 gd->fops = &xlvbd_block_fops;
445 gd->private_data = info;
446 gd->driverfs_dev = &(info->xbdev->dev);
447 set_capacity(gd, capacity);
449 if (xlvbd_init_blk_queue(gd, sector_size)) {
454 info->rq = gd->queue;
457 if (info->feature_barrier)
460 if (vdisk_info & VDISK_READONLY)
463 if (vdisk_info & VDISK_REMOVABLE)
464 gd->flags |= GENHD_FL_REMOVABLE;
466 if (vdisk_info & VDISK_CDROM)
467 gd->flags |= GENHD_FL_CD;
475 static void kick_pending_request_queues(struct blkfront_info *info)
477 if (!RING_FULL(&info->ring)) {
478 /* Re-enable calldowns. */
479 blk_start_queue(info->rq);
480 /* Kick things off immediately. */
481 do_blkif_request(info->rq);
485 static void blkif_restart_queue(struct work_struct *work)
487 struct blkfront_info *info = container_of(work, struct blkfront_info, work);
489 spin_lock_irq(&blkif_io_lock);
490 if (info->connected == BLKIF_STATE_CONNECTED)
491 kick_pending_request_queues(info);
492 spin_unlock_irq(&blkif_io_lock);
495 static void blkif_free(struct blkfront_info *info, int suspend)
497 /* Prevent new requests being issued until we fix things up. */
498 spin_lock_irq(&blkif_io_lock);
499 info->connected = suspend ?
500 BLKIF_STATE_SUSPENDED : BLKIF_STATE_DISCONNECTED;
501 /* No more blkif_request(). */
503 blk_stop_queue(info->rq);
504 /* No more gnttab callback work. */
505 gnttab_cancel_free_callback(&info->callback);
506 spin_unlock_irq(&blkif_io_lock);
508 /* Flush gnttab callback work. Must be done with no locks held. */
509 flush_scheduled_work();
511 /* Free resources associated with old device channel. */
512 if (info->ring_ref != GRANT_INVALID_REF) {
513 gnttab_end_foreign_access(info->ring_ref, 0,
514 (unsigned long)info->ring.sring);
515 info->ring_ref = GRANT_INVALID_REF;
516 info->ring.sring = NULL;
519 unbind_from_irqhandler(info->irq, info);
520 info->evtchn = info->irq = 0;
524 static void blkif_completion(struct blk_shadow *s)
527 for (i = 0; i < s->req.nr_segments; i++)
528 gnttab_end_foreign_access(s->req.seg[i].gref, 0, 0UL);
531 static irqreturn_t blkif_interrupt(int irq, void *dev_id)
534 struct blkif_response *bret;
537 struct blkfront_info *info = (struct blkfront_info *)dev_id;
540 spin_lock_irqsave(&blkif_io_lock, flags);
542 if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) {
543 spin_unlock_irqrestore(&blkif_io_lock, flags);
548 rp = info->ring.sring->rsp_prod;
549 rmb(); /* Ensure we see queued responses up to 'rp'. */
551 for (i = info->ring.rsp_cons; i != rp; i++) {
554 bret = RING_GET_RESPONSE(&info->ring, i);
556 req = (struct request *)info->shadow[id].request;
558 blkif_completion(&info->shadow[id]);
560 add_id_to_freelist(info, id);
562 error = (bret->status == BLKIF_RSP_OKAY) ? 0 : -EIO;
563 switch (bret->operation) {
564 case BLKIF_OP_WRITE_BARRIER:
565 if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) {
566 printk(KERN_WARNING "blkfront: %s: write barrier op failed\n",
567 info->gd->disk_name);
569 info->feature_barrier = 0;
575 if (unlikely(bret->status != BLKIF_RSP_OKAY))
576 dev_dbg(&info->xbdev->dev, "Bad return from blkdev data "
577 "request: %x\n", bret->status);
579 __blk_end_request_all(req, error);
586 info->ring.rsp_cons = i;
588 if (i != info->ring.req_prod_pvt) {
590 RING_FINAL_CHECK_FOR_RESPONSES(&info->ring, more_to_do);
594 info->ring.sring->rsp_event = i + 1;
596 kick_pending_request_queues(info);
598 spin_unlock_irqrestore(&blkif_io_lock, flags);
604 static int setup_blkring(struct xenbus_device *dev,
605 struct blkfront_info *info)
607 struct blkif_sring *sring;
610 info->ring_ref = GRANT_INVALID_REF;
612 sring = (struct blkif_sring *)__get_free_page(GFP_NOIO | __GFP_HIGH);
614 xenbus_dev_fatal(dev, -ENOMEM, "allocating shared ring");
617 SHARED_RING_INIT(sring);
618 FRONT_RING_INIT(&info->ring, sring, PAGE_SIZE);
620 sg_init_table(info->sg, BLKIF_MAX_SEGMENTS_PER_REQUEST);
622 err = xenbus_grant_ring(dev, virt_to_mfn(info->ring.sring));
624 free_page((unsigned long)sring);
625 info->ring.sring = NULL;
628 info->ring_ref = err;
630 err = xenbus_alloc_evtchn(dev, &info->evtchn);
634 err = bind_evtchn_to_irqhandler(info->evtchn,
636 IRQF_SAMPLE_RANDOM, "blkif", info);
638 xenbus_dev_fatal(dev, err,
639 "bind_evtchn_to_irqhandler failed");
651 /* Common code used when first setting up, and when resuming. */
652 static int talk_to_backend(struct xenbus_device *dev,
653 struct blkfront_info *info)
655 const char *message = NULL;
656 struct xenbus_transaction xbt;
659 /* Create shared ring, alloc event channel. */
660 err = setup_blkring(dev, info);
665 err = xenbus_transaction_start(&xbt);
667 xenbus_dev_fatal(dev, err, "starting transaction");
668 goto destroy_blkring;
671 err = xenbus_printf(xbt, dev->nodename,
672 "ring-ref", "%u", info->ring_ref);
674 message = "writing ring-ref";
675 goto abort_transaction;
677 err = xenbus_printf(xbt, dev->nodename,
678 "event-channel", "%u", info->evtchn);
680 message = "writing event-channel";
681 goto abort_transaction;
683 err = xenbus_printf(xbt, dev->nodename, "protocol", "%s",
684 XEN_IO_PROTO_ABI_NATIVE);
686 message = "writing protocol";
687 goto abort_transaction;
690 err = xenbus_transaction_end(xbt, 0);
694 xenbus_dev_fatal(dev, err, "completing transaction");
695 goto destroy_blkring;
698 xenbus_switch_state(dev, XenbusStateInitialised);
703 xenbus_transaction_end(xbt, 1);
705 xenbus_dev_fatal(dev, err, "%s", message);
714 * Entry point to this code when a new device is created. Allocate the basic
715 * structures and the ring buffer for communication with the backend, and
716 * inform the backend of the appropriate details for those. Switch to
719 static int blkfront_probe(struct xenbus_device *dev,
720 const struct xenbus_device_id *id)
723 struct blkfront_info *info;
725 /* FIXME: Use dynamic device id if this is not set. */
726 err = xenbus_scanf(XBT_NIL, dev->nodename,
727 "virtual-device", "%i", &vdevice);
729 /* go looking in the extended area instead */
730 err = xenbus_scanf(XBT_NIL, dev->nodename, "virtual-device-ext",
733 xenbus_dev_fatal(dev, err, "reading virtual-device");
738 info = kzalloc(sizeof(*info), GFP_KERNEL);
740 xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure");
745 info->vdevice = vdevice;
746 info->connected = BLKIF_STATE_DISCONNECTED;
747 INIT_WORK(&info->work, blkif_restart_queue);
749 for (i = 0; i < BLK_RING_SIZE; i++)
750 info->shadow[i].req.id = i+1;
751 info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
753 /* Front end dir is a number, which is used as the id. */
754 info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0);
755 dev->dev.driver_data = info;
757 err = talk_to_backend(dev, info);
760 dev->dev.driver_data = NULL;
768 static int blkif_recover(struct blkfront_info *info)
771 struct blkif_request *req;
772 struct blk_shadow *copy;
775 /* Stage 1: Make a safe copy of the shadow state. */
776 copy = kmalloc(sizeof(info->shadow),
777 GFP_NOIO | __GFP_REPEAT | __GFP_HIGH);
780 memcpy(copy, info->shadow, sizeof(info->shadow));
782 /* Stage 2: Set up free list. */
783 memset(&info->shadow, 0, sizeof(info->shadow));
784 for (i = 0; i < BLK_RING_SIZE; i++)
785 info->shadow[i].req.id = i+1;
786 info->shadow_free = info->ring.req_prod_pvt;
787 info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
789 /* Stage 3: Find pending requests and requeue them. */
790 for (i = 0; i < BLK_RING_SIZE; i++) {
792 if (copy[i].request == 0)
795 /* Grab a request slot and copy shadow state into it. */
796 req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
799 /* We get a new request id, and must reset the shadow state. */
800 req->id = get_id_from_freelist(info);
801 memcpy(&info->shadow[req->id], ©[i], sizeof(copy[i]));
803 /* Rewrite any grant references invalidated by susp/resume. */
804 for (j = 0; j < req->nr_segments; j++)
805 gnttab_grant_foreign_access_ref(
807 info->xbdev->otherend_id,
808 pfn_to_mfn(info->shadow[req->id].frame[j]),
811 info->shadow[req->id].request));
812 info->shadow[req->id].req = *req;
814 info->ring.req_prod_pvt++;
819 xenbus_switch_state(info->xbdev, XenbusStateConnected);
821 spin_lock_irq(&blkif_io_lock);
823 /* Now safe for us to use the shared ring */
824 info->connected = BLKIF_STATE_CONNECTED;
826 /* Send off requeued requests */
827 flush_requests(info);
829 /* Kick any other new requests queued since we resumed */
830 kick_pending_request_queues(info);
832 spin_unlock_irq(&blkif_io_lock);
838 * We are reconnecting to the backend, due to a suspend/resume, or a backend
839 * driver restart. We tear down our blkif structure and recreate it, but
840 * leave the device-layer structures intact so that this is transparent to the
841 * rest of the kernel.
843 static int blkfront_resume(struct xenbus_device *dev)
845 struct blkfront_info *info = dev->dev.driver_data;
848 dev_dbg(&dev->dev, "blkfront_resume: %s\n", dev->nodename);
850 blkif_free(info, info->connected == BLKIF_STATE_CONNECTED);
852 err = talk_to_backend(dev, info);
853 if (info->connected == BLKIF_STATE_SUSPENDED && !err)
854 err = blkif_recover(info);
861 * Invoked when the backend is finally 'ready' (and has told produced
862 * the details about the physical device - #sectors, size, etc).
864 static void blkfront_connect(struct blkfront_info *info)
866 unsigned long long sectors;
867 unsigned long sector_size;
871 if ((info->connected == BLKIF_STATE_CONNECTED) ||
872 (info->connected == BLKIF_STATE_SUSPENDED) )
875 dev_dbg(&info->xbdev->dev, "%s:%s.\n",
876 __func__, info->xbdev->otherend);
878 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
879 "sectors", "%llu", §ors,
880 "info", "%u", &binfo,
881 "sector-size", "%lu", §or_size,
884 xenbus_dev_fatal(info->xbdev, err,
885 "reading backend fields at %s",
886 info->xbdev->otherend);
890 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
891 "feature-barrier", "%lu", &info->feature_barrier,
894 info->feature_barrier = 0;
896 err = xlvbd_alloc_gendisk(sectors, info, binfo, sector_size);
898 xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s",
899 info->xbdev->otherend);
903 xenbus_switch_state(info->xbdev, XenbusStateConnected);
905 /* Kick pending requests. */
906 spin_lock_irq(&blkif_io_lock);
907 info->connected = BLKIF_STATE_CONNECTED;
908 kick_pending_request_queues(info);
909 spin_unlock_irq(&blkif_io_lock);
917 * Handle the change of state of the backend to Closing. We must delete our
918 * device-layer structures now, to ensure that writes are flushed through to
919 * the backend. Once is this done, we can switch to Closed in
922 static void blkfront_closing(struct xenbus_device *dev)
924 struct blkfront_info *info = dev->dev.driver_data;
927 dev_dbg(&dev->dev, "blkfront_closing: %s removed\n", dev->nodename);
929 if (info->rq == NULL)
932 spin_lock_irqsave(&blkif_io_lock, flags);
934 del_gendisk(info->gd);
936 /* No more blkif_request(). */
937 blk_stop_queue(info->rq);
939 /* No more gnttab callback work. */
940 gnttab_cancel_free_callback(&info->callback);
941 spin_unlock_irqrestore(&blkif_io_lock, flags);
943 /* Flush gnttab callback work. Must be done with no locks held. */
944 flush_scheduled_work();
946 blk_cleanup_queue(info->rq);
950 xenbus_frontend_closed(dev);
954 * Callback received when the backend's state changes.
956 static void backend_changed(struct xenbus_device *dev,
957 enum xenbus_state backend_state)
959 struct blkfront_info *info = dev->dev.driver_data;
960 struct block_device *bd;
962 dev_dbg(&dev->dev, "blkfront:backend_changed.\n");
964 switch (backend_state) {
965 case XenbusStateInitialising:
966 case XenbusStateInitWait:
967 case XenbusStateInitialised:
968 case XenbusStateUnknown:
969 case XenbusStateClosed:
972 case XenbusStateConnected:
973 blkfront_connect(info);
976 case XenbusStateClosing:
977 if (info->gd == NULL)
978 xenbus_dev_fatal(dev, -ENODEV, "gd is NULL");
979 bd = bdget_disk(info->gd, 0);
981 xenbus_dev_fatal(dev, -ENODEV, "bdget failed");
983 mutex_lock(&bd->bd_mutex);
985 xenbus_dev_error(dev, -EBUSY,
986 "Device in use; refusing to close");
988 blkfront_closing(dev);
989 mutex_unlock(&bd->bd_mutex);
995 static int blkfront_remove(struct xenbus_device *dev)
997 struct blkfront_info *info = dev->dev.driver_data;
999 dev_dbg(&dev->dev, "blkfront_remove: %s removed\n", dev->nodename);
1001 blkif_free(info, 0);
1008 static int blkfront_is_ready(struct xenbus_device *dev)
1010 struct blkfront_info *info = dev->dev.driver_data;
1012 return info->is_ready;
1015 static int blkif_open(struct block_device *bdev, fmode_t mode)
1017 struct blkfront_info *info = bdev->bd_disk->private_data;
1022 static int blkif_release(struct gendisk *disk, fmode_t mode)
1024 struct blkfront_info *info = disk->private_data;
1026 if (info->users == 0) {
1027 /* Check whether we have been instructed to close. We will
1028 have ignored this request initially, as the device was
1030 struct xenbus_device *dev = info->xbdev;
1031 enum xenbus_state state = xenbus_read_driver_state(dev->otherend);
1033 if (state == XenbusStateClosing && info->is_ready)
1034 blkfront_closing(dev);
1039 static struct block_device_operations xlvbd_block_fops =
1041 .owner = THIS_MODULE,
1043 .release = blkif_release,
1044 .getgeo = blkif_getgeo,
1045 .locked_ioctl = blkif_ioctl,
1049 static struct xenbus_device_id blkfront_ids[] = {
1054 static struct xenbus_driver blkfront = {
1056 .owner = THIS_MODULE,
1057 .ids = blkfront_ids,
1058 .probe = blkfront_probe,
1059 .remove = blkfront_remove,
1060 .resume = blkfront_resume,
1061 .otherend_changed = backend_changed,
1062 .is_ready = blkfront_is_ready,
1065 static int __init xlblk_init(void)
1070 if (register_blkdev(XENVBD_MAJOR, DEV_NAME)) {
1071 printk(KERN_WARNING "xen_blk: can't get major %d with name %s\n",
1072 XENVBD_MAJOR, DEV_NAME);
1076 return xenbus_register_frontend(&blkfront);
1078 module_init(xlblk_init);
1081 static void __exit xlblk_exit(void)
1083 return xenbus_unregister_driver(&blkfront);
1085 module_exit(xlblk_exit);
1087 MODULE_DESCRIPTION("Xen virtual block device frontend");
1088 MODULE_LICENSE("GPL");
1089 MODULE_ALIAS_BLOCKDEV_MAJOR(XENVBD_MAJOR);
1090 MODULE_ALIAS("xen:vbd");
1091 MODULE_ALIAS("xenblk");