aoeblk.c revision 02e031cbc843b010e72fcc05c76113c688b2860f
1/* Copyright (c) 2007 Coraid, Inc.  See COPYING for GPL terms. */
2/*
3 * aoeblk.c
4 * block device routines
5 */
6
7#include <linux/kernel.h>
8#include <linux/hdreg.h>
9#include <linux/blkdev.h>
10#include <linux/backing-dev.h>
11#include <linux/fs.h>
12#include <linux/ioctl.h>
13#include <linux/slab.h>
14#include <linux/ratelimit.h>
15#include <linux/genhd.h>
16#include <linux/netdevice.h>
17#include <linux/mutex.h>
18#include "aoe.h"
19
20static DEFINE_MUTEX(aoeblk_mutex);
21static struct kmem_cache *buf_pool_cache;
22
23static ssize_t aoedisk_show_state(struct device *dev,
24				  struct device_attribute *attr, char *page)
25{
26	struct gendisk *disk = dev_to_disk(dev);
27	struct aoedev *d = disk->private_data;
28
29	return snprintf(page, PAGE_SIZE,
30			"%s%s\n",
31			(d->flags & DEVFL_UP) ? "up" : "down",
32			(d->flags & DEVFL_KICKME) ? ",kickme" :
33			(d->nopen && !(d->flags & DEVFL_UP)) ? ",closewait" : "");
34	/* I'd rather see nopen exported so we can ditch closewait */
35}
36static ssize_t aoedisk_show_mac(struct device *dev,
37				struct device_attribute *attr, char *page)
38{
39	struct gendisk *disk = dev_to_disk(dev);
40	struct aoedev *d = disk->private_data;
41	struct aoetgt *t = d->targets[0];
42
43	if (t == NULL)
44		return snprintf(page, PAGE_SIZE, "none\n");
45	return snprintf(page, PAGE_SIZE, "%pm\n", t->addr);
46}
47static ssize_t aoedisk_show_netif(struct device *dev,
48				  struct device_attribute *attr, char *page)
49{
50	struct gendisk *disk = dev_to_disk(dev);
51	struct aoedev *d = disk->private_data;
52	struct net_device *nds[8], **nd, **nnd, **ne;
53	struct aoetgt **t, **te;
54	struct aoeif *ifp, *e;
55	char *p;
56
57	memset(nds, 0, sizeof nds);
58	nd = nds;
59	ne = nd + ARRAY_SIZE(nds);
60	t = d->targets;
61	te = t + NTARGETS;
62	for (; t < te && *t; t++) {
63		ifp = (*t)->ifs;
64		e = ifp + NAOEIFS;
65		for (; ifp < e && ifp->nd; ifp++) {
66			for (nnd = nds; nnd < nd; nnd++)
67				if (*nnd == ifp->nd)
68					break;
69			if (nnd == nd && nd != ne)
70				*nd++ = ifp->nd;
71		}
72	}
73
74	ne = nd;
75	nd = nds;
76	if (*nd == NULL)
77		return snprintf(page, PAGE_SIZE, "none\n");
78	for (p = page; nd < ne; nd++)
79		p += snprintf(p, PAGE_SIZE - (p-page), "%s%s",
80			p == page ? "" : ",", (*nd)->name);
81	p += snprintf(p, PAGE_SIZE - (p-page), "\n");
82	return p-page;
83}
84/* firmware version */
85static ssize_t aoedisk_show_fwver(struct device *dev,
86				  struct device_attribute *attr, char *page)
87{
88	struct gendisk *disk = dev_to_disk(dev);
89	struct aoedev *d = disk->private_data;
90
91	return snprintf(page, PAGE_SIZE, "0x%04x\n", (unsigned int) d->fw_ver);
92}
93
94static DEVICE_ATTR(state, S_IRUGO, aoedisk_show_state, NULL);
95static DEVICE_ATTR(mac, S_IRUGO, aoedisk_show_mac, NULL);
96static DEVICE_ATTR(netif, S_IRUGO, aoedisk_show_netif, NULL);
97static struct device_attribute dev_attr_firmware_version = {
98	.attr = { .name = "firmware-version", .mode = S_IRUGO },
99	.show = aoedisk_show_fwver,
100};
101
102static struct attribute *aoe_attrs[] = {
103	&dev_attr_state.attr,
104	&dev_attr_mac.attr,
105	&dev_attr_netif.attr,
106	&dev_attr_firmware_version.attr,
107	NULL,
108};
109
110static const struct attribute_group attr_group = {
111	.attrs = aoe_attrs,
112};
113
114static int
115aoedisk_add_sysfs(struct aoedev *d)
116{
117	return sysfs_create_group(&disk_to_dev(d->gd)->kobj, &attr_group);
118}
119void
120aoedisk_rm_sysfs(struct aoedev *d)
121{
122	sysfs_remove_group(&disk_to_dev(d->gd)->kobj, &attr_group);
123}
124
125static int
126aoeblk_open(struct block_device *bdev, fmode_t mode)
127{
128	struct aoedev *d = bdev->bd_disk->private_data;
129	ulong flags;
130
131	mutex_lock(&aoeblk_mutex);
132	spin_lock_irqsave(&d->lock, flags);
133	if (d->flags & DEVFL_UP) {
134		d->nopen++;
135		spin_unlock_irqrestore(&d->lock, flags);
136		mutex_unlock(&aoeblk_mutex);
137		return 0;
138	}
139	spin_unlock_irqrestore(&d->lock, flags);
140	mutex_unlock(&aoeblk_mutex);
141	return -ENODEV;
142}
143
144static int
145aoeblk_release(struct gendisk *disk, fmode_t mode)
146{
147	struct aoedev *d = disk->private_data;
148	ulong flags;
149
150	spin_lock_irqsave(&d->lock, flags);
151
152	if (--d->nopen == 0) {
153		spin_unlock_irqrestore(&d->lock, flags);
154		aoecmd_cfg(d->aoemajor, d->aoeminor);
155		return 0;
156	}
157	spin_unlock_irqrestore(&d->lock, flags);
158
159	return 0;
160}
161
162static int
163aoeblk_make_request(struct request_queue *q, struct bio *bio)
164{
165	struct sk_buff_head queue;
166	struct aoedev *d;
167	struct buf *buf;
168	ulong flags;
169
170	blk_queue_bounce(q, &bio);
171
172	if (bio == NULL) {
173		printk(KERN_ERR "aoe: bio is NULL\n");
174		BUG();
175		return 0;
176	}
177	d = bio->bi_bdev->bd_disk->private_data;
178	if (d == NULL) {
179		printk(KERN_ERR "aoe: bd_disk->private_data is NULL\n");
180		BUG();
181		bio_endio(bio, -ENXIO);
182		return 0;
183	} else if (bio->bi_io_vec == NULL) {
184		printk(KERN_ERR "aoe: bi_io_vec is NULL\n");
185		BUG();
186		bio_endio(bio, -ENXIO);
187		return 0;
188	}
189	buf = mempool_alloc(d->bufpool, GFP_NOIO);
190	if (buf == NULL) {
191		printk(KERN_INFO "aoe: buf allocation failure\n");
192		bio_endio(bio, -ENOMEM);
193		return 0;
194	}
195	memset(buf, 0, sizeof(*buf));
196	INIT_LIST_HEAD(&buf->bufs);
197	buf->stime = jiffies;
198	buf->bio = bio;
199	buf->resid = bio->bi_size;
200	buf->sector = bio->bi_sector;
201	buf->bv = &bio->bi_io_vec[bio->bi_idx];
202	buf->bv_resid = buf->bv->bv_len;
203	WARN_ON(buf->bv_resid == 0);
204	buf->bv_off = buf->bv->bv_offset;
205
206	spin_lock_irqsave(&d->lock, flags);
207
208	if ((d->flags & DEVFL_UP) == 0) {
209		pr_info_ratelimited("aoe: device %ld.%d is not up\n",
210			d->aoemajor, d->aoeminor);
211		spin_unlock_irqrestore(&d->lock, flags);
212		mempool_free(buf, d->bufpool);
213		bio_endio(bio, -ENXIO);
214		return 0;
215	}
216
217	list_add_tail(&buf->bufs, &d->bufq);
218
219	aoecmd_work(d);
220	__skb_queue_head_init(&queue);
221	skb_queue_splice_init(&d->sendq, &queue);
222
223	spin_unlock_irqrestore(&d->lock, flags);
224	aoenet_xmit(&queue);
225
226	return 0;
227}
228
229static int
230aoeblk_getgeo(struct block_device *bdev, struct hd_geometry *geo)
231{
232	struct aoedev *d = bdev->bd_disk->private_data;
233
234	if ((d->flags & DEVFL_UP) == 0) {
235		printk(KERN_ERR "aoe: disk not up\n");
236		return -ENODEV;
237	}
238
239	geo->cylinders = d->geo.cylinders;
240	geo->heads = d->geo.heads;
241	geo->sectors = d->geo.sectors;
242	return 0;
243}
244
245static const struct block_device_operations aoe_bdops = {
246	.open = aoeblk_open,
247	.release = aoeblk_release,
248	.getgeo = aoeblk_getgeo,
249	.owner = THIS_MODULE,
250};
251
252/* alloc_disk and add_disk can sleep */
253void
254aoeblk_gdalloc(void *vp)
255{
256	struct aoedev *d = vp;
257	struct gendisk *gd;
258	ulong flags;
259
260	gd = alloc_disk(AOE_PARTITIONS);
261	if (gd == NULL) {
262		printk(KERN_ERR
263			"aoe: cannot allocate disk structure for %ld.%d\n",
264			d->aoemajor, d->aoeminor);
265		goto err;
266	}
267
268	d->bufpool = mempool_create_slab_pool(MIN_BUFS, buf_pool_cache);
269	if (d->bufpool == NULL) {
270		printk(KERN_ERR "aoe: cannot allocate bufpool for %ld.%d\n",
271			d->aoemajor, d->aoeminor);
272		goto err_disk;
273	}
274
275	d->blkq = blk_alloc_queue(GFP_KERNEL);
276	if (!d->blkq)
277		goto err_mempool;
278	blk_queue_make_request(d->blkq, aoeblk_make_request);
279	d->blkq->backing_dev_info.name = "aoe";
280	if (bdi_init(&d->blkq->backing_dev_info))
281		goto err_blkq;
282	spin_lock_irqsave(&d->lock, flags);
283	gd->major = AOE_MAJOR;
284	gd->first_minor = d->sysminor * AOE_PARTITIONS;
285	gd->fops = &aoe_bdops;
286	gd->private_data = d;
287	set_capacity(gd, d->ssize);
288	snprintf(gd->disk_name, sizeof gd->disk_name, "etherd/e%ld.%d",
289		d->aoemajor, d->aoeminor);
290
291	gd->queue = d->blkq;
292	d->gd = gd;
293	d->flags &= ~DEVFL_GDALLOC;
294	d->flags |= DEVFL_UP;
295
296	spin_unlock_irqrestore(&d->lock, flags);
297
298	add_disk(gd);
299	aoedisk_add_sysfs(d);
300	return;
301
302err_blkq:
303	blk_cleanup_queue(d->blkq);
304	d->blkq = NULL;
305err_mempool:
306	mempool_destroy(d->bufpool);
307err_disk:
308	put_disk(gd);
309err:
310	spin_lock_irqsave(&d->lock, flags);
311	d->flags &= ~DEVFL_GDALLOC;
312	spin_unlock_irqrestore(&d->lock, flags);
313}
314
315void
316aoeblk_exit(void)
317{
318	kmem_cache_destroy(buf_pool_cache);
319}
320
321int __init
322aoeblk_init(void)
323{
324	buf_pool_cache = kmem_cache_create("aoe_bufs",
325					   sizeof(struct buf),
326					   0, 0, NULL);
327	if (buf_pool_cache == NULL)
328		return -ENOMEM;
329
330	return 0;
331}
332
333