1 /*
2 * Filename: dev.c
3 *
4 *
5 * Authors: Joshua Morris <josh.h.morris@us.ibm.com>
6 * Philip Kelleher <pjk1939@linux.vnet.ibm.com>
7 *
8 * (C) Copyright 2013 IBM Corporation
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License as
12 * published by the Free Software Foundation; either version 2 of the
13 * License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful, but
16 * WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public License
21 * along with this program; if not, write to the Free Software Foundation,
22 * Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23 */
24
25 #include <linux/kernel.h>
26 #include <linux/interrupt.h>
27 #include <linux/module.h>
28 #include <linux/pci.h>
29 #include <linux/slab.h>
30
31 #include <linux/hdreg.h>
32 #include <linux/genhd.h>
33 #include <linux/blkdev.h>
34 #include <linux/bio.h>
35
36 #include <linux/fs.h>
37
38 #include "rsxx_priv.h"
39
40 static unsigned int blkdev_minors = 64;
41 module_param(blkdev_minors, uint, 0444);
42 MODULE_PARM_DESC(blkdev_minors, "Number of minors(partitions)");
43
44 /*
45 * For now I'm making this tweakable in case any applications hit this limit.
46 * If you see a "bio too big" error in the log you will need to raise this
47 * value.
48 */
49 static unsigned int blkdev_max_hw_sectors = 1024;
50 module_param(blkdev_max_hw_sectors, uint, 0444);
51 MODULE_PARM_DESC(blkdev_max_hw_sectors, "Max hw sectors for a single BIO");
52
53 static unsigned int enable_blkdev = 1;
54 module_param(enable_blkdev , uint, 0444);
55 MODULE_PARM_DESC(enable_blkdev, "Enable block device interfaces");
56
57
58 struct rsxx_bio_meta {
59 struct bio *bio;
60 atomic_t pending_dmas;
61 atomic_t error;
62 unsigned long start_time;
63 };
64
65 static struct kmem_cache *bio_meta_pool;
66
67 /*----------------- Block Device Operations -----------------*/
rsxx_blkdev_ioctl(struct block_device * bdev,fmode_t mode,unsigned int cmd,unsigned long arg)68 static int rsxx_blkdev_ioctl(struct block_device *bdev,
69 fmode_t mode,
70 unsigned int cmd,
71 unsigned long arg)
72 {
73 struct rsxx_cardinfo *card = bdev->bd_disk->private_data;
74
75 switch (cmd) {
76 case RSXX_GETREG:
77 return rsxx_reg_access(card, (void __user *)arg, 1);
78 case RSXX_SETREG:
79 return rsxx_reg_access(card, (void __user *)arg, 0);
80 }
81
82 return -ENOTTY;
83 }
84
rsxx_getgeo(struct block_device * bdev,struct hd_geometry * geo)85 static int rsxx_getgeo(struct block_device *bdev, struct hd_geometry *geo)
86 {
87 struct rsxx_cardinfo *card = bdev->bd_disk->private_data;
88 u64 blocks = card->size8 >> 9;
89
90 /*
91 * get geometry: Fake it. I haven't found any drivers that set
92 * geo->start, so we won't either.
93 */
94 if (card->size8) {
95 geo->heads = 64;
96 geo->sectors = 16;
97 do_div(blocks, (geo->heads * geo->sectors));
98 geo->cylinders = blocks;
99 } else {
100 geo->heads = 0;
101 geo->sectors = 0;
102 geo->cylinders = 0;
103 }
104 return 0;
105 }
106
107 static const struct block_device_operations rsxx_fops = {
108 .owner = THIS_MODULE,
109 .getgeo = rsxx_getgeo,
110 .ioctl = rsxx_blkdev_ioctl,
111 };
112
disk_stats_start(struct rsxx_cardinfo * card,struct bio * bio)113 static void disk_stats_start(struct rsxx_cardinfo *card, struct bio *bio)
114 {
115 generic_start_io_acct(bio_data_dir(bio), bio_sectors(bio),
116 &card->gendisk->part0);
117 }
118
disk_stats_complete(struct rsxx_cardinfo * card,struct bio * bio,unsigned long start_time)119 static void disk_stats_complete(struct rsxx_cardinfo *card,
120 struct bio *bio,
121 unsigned long start_time)
122 {
123 generic_end_io_acct(bio_data_dir(bio), &card->gendisk->part0,
124 start_time);
125 }
126
bio_dma_done_cb(struct rsxx_cardinfo * card,void * cb_data,unsigned int error)127 static void bio_dma_done_cb(struct rsxx_cardinfo *card,
128 void *cb_data,
129 unsigned int error)
130 {
131 struct rsxx_bio_meta *meta = cb_data;
132
133 if (error)
134 atomic_set(&meta->error, 1);
135
136 if (atomic_dec_and_test(&meta->pending_dmas)) {
137 if (!card->eeh_state && card->gendisk)
138 disk_stats_complete(card, meta->bio, meta->start_time);
139
140 bio_endio(meta->bio, atomic_read(&meta->error) ? -EIO : 0);
141 kmem_cache_free(bio_meta_pool, meta);
142 }
143 }
144
rsxx_make_request(struct request_queue * q,struct bio * bio)145 static void rsxx_make_request(struct request_queue *q, struct bio *bio)
146 {
147 struct rsxx_cardinfo *card = q->queuedata;
148 struct rsxx_bio_meta *bio_meta;
149 int st = -EINVAL;
150
151 might_sleep();
152
153 if (!card)
154 goto req_err;
155
156 if (bio_end_sector(bio) > get_capacity(card->gendisk))
157 goto req_err;
158
159 if (unlikely(card->halt)) {
160 st = -EFAULT;
161 goto req_err;
162 }
163
164 if (unlikely(card->dma_fault)) {
165 st = (-EFAULT);
166 goto req_err;
167 }
168
169 if (bio->bi_iter.bi_size == 0) {
170 dev_err(CARD_TO_DEV(card), "size zero BIO!\n");
171 goto req_err;
172 }
173
174 bio_meta = kmem_cache_alloc(bio_meta_pool, GFP_KERNEL);
175 if (!bio_meta) {
176 st = -ENOMEM;
177 goto req_err;
178 }
179
180 bio_meta->bio = bio;
181 atomic_set(&bio_meta->error, 0);
182 atomic_set(&bio_meta->pending_dmas, 0);
183 bio_meta->start_time = jiffies;
184
185 if (!unlikely(card->halt))
186 disk_stats_start(card, bio);
187
188 dev_dbg(CARD_TO_DEV(card), "BIO[%c]: meta: %p addr8: x%llx size: %d\n",
189 bio_data_dir(bio) ? 'W' : 'R', bio_meta,
190 (u64)bio->bi_iter.bi_sector << 9, bio->bi_iter.bi_size);
191
192 st = rsxx_dma_queue_bio(card, bio, &bio_meta->pending_dmas,
193 bio_dma_done_cb, bio_meta);
194 if (st)
195 goto queue_err;
196
197 return;
198
199 queue_err:
200 kmem_cache_free(bio_meta_pool, bio_meta);
201 req_err:
202 bio_endio(bio, st);
203 }
204
205 /*----------------- Device Setup -------------------*/
rsxx_discard_supported(struct rsxx_cardinfo * card)206 static bool rsxx_discard_supported(struct rsxx_cardinfo *card)
207 {
208 unsigned char pci_rev;
209
210 pci_read_config_byte(card->dev, PCI_REVISION_ID, &pci_rev);
211
212 return (pci_rev >= RSXX_DISCARD_SUPPORT);
213 }
214
rsxx_attach_dev(struct rsxx_cardinfo * card)215 int rsxx_attach_dev(struct rsxx_cardinfo *card)
216 {
217 mutex_lock(&card->dev_lock);
218
219 /* The block device requires the stripe size from the config. */
220 if (enable_blkdev) {
221 if (card->config_valid)
222 set_capacity(card->gendisk, card->size8 >> 9);
223 else
224 set_capacity(card->gendisk, 0);
225 add_disk(card->gendisk);
226
227 card->bdev_attached = 1;
228 }
229
230 mutex_unlock(&card->dev_lock);
231
232 return 0;
233 }
234
rsxx_detach_dev(struct rsxx_cardinfo * card)235 void rsxx_detach_dev(struct rsxx_cardinfo *card)
236 {
237 mutex_lock(&card->dev_lock);
238
239 if (card->bdev_attached) {
240 del_gendisk(card->gendisk);
241 card->bdev_attached = 0;
242 }
243
244 mutex_unlock(&card->dev_lock);
245 }
246
rsxx_setup_dev(struct rsxx_cardinfo * card)247 int rsxx_setup_dev(struct rsxx_cardinfo *card)
248 {
249 unsigned short blk_size;
250
251 mutex_init(&card->dev_lock);
252
253 if (!enable_blkdev)
254 return 0;
255
256 card->major = register_blkdev(0, DRIVER_NAME);
257 if (card->major < 0) {
258 dev_err(CARD_TO_DEV(card), "Failed to get major number\n");
259 return -ENOMEM;
260 }
261
262 card->queue = blk_alloc_queue(GFP_KERNEL);
263 if (!card->queue) {
264 dev_err(CARD_TO_DEV(card), "Failed queue alloc\n");
265 unregister_blkdev(card->major, DRIVER_NAME);
266 return -ENOMEM;
267 }
268
269 card->gendisk = alloc_disk(blkdev_minors);
270 if (!card->gendisk) {
271 dev_err(CARD_TO_DEV(card), "Failed disk alloc\n");
272 blk_cleanup_queue(card->queue);
273 unregister_blkdev(card->major, DRIVER_NAME);
274 return -ENOMEM;
275 }
276
277 if (card->config_valid) {
278 blk_size = card->config.data.block_size;
279 blk_queue_dma_alignment(card->queue, blk_size - 1);
280 blk_queue_logical_block_size(card->queue, blk_size);
281 }
282
283 blk_queue_make_request(card->queue, rsxx_make_request);
284 blk_queue_bounce_limit(card->queue, BLK_BOUNCE_ANY);
285 blk_queue_max_hw_sectors(card->queue, blkdev_max_hw_sectors);
286 blk_queue_physical_block_size(card->queue, RSXX_HW_BLK_SIZE);
287
288 queue_flag_set_unlocked(QUEUE_FLAG_NONROT, card->queue);
289 queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, card->queue);
290 if (rsxx_discard_supported(card)) {
291 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, card->queue);
292 blk_queue_max_discard_sectors(card->queue,
293 RSXX_HW_BLK_SIZE >> 9);
294 card->queue->limits.discard_granularity = RSXX_HW_BLK_SIZE;
295 card->queue->limits.discard_alignment = RSXX_HW_BLK_SIZE;
296 card->queue->limits.discard_zeroes_data = 1;
297 }
298
299 card->queue->queuedata = card;
300
301 snprintf(card->gendisk->disk_name, sizeof(card->gendisk->disk_name),
302 "rsxx%d", card->disk_id);
303 card->gendisk->driverfs_dev = &card->dev->dev;
304 card->gendisk->major = card->major;
305 card->gendisk->first_minor = 0;
306 card->gendisk->fops = &rsxx_fops;
307 card->gendisk->private_data = card;
308 card->gendisk->queue = card->queue;
309
310 return 0;
311 }
312
rsxx_destroy_dev(struct rsxx_cardinfo * card)313 void rsxx_destroy_dev(struct rsxx_cardinfo *card)
314 {
315 if (!enable_blkdev)
316 return;
317
318 put_disk(card->gendisk);
319 card->gendisk = NULL;
320
321 blk_cleanup_queue(card->queue);
322 card->queue->queuedata = NULL;
323 unregister_blkdev(card->major, DRIVER_NAME);
324 }
325
rsxx_dev_init(void)326 int rsxx_dev_init(void)
327 {
328 bio_meta_pool = KMEM_CACHE(rsxx_bio_meta, SLAB_HWCACHE_ALIGN);
329 if (!bio_meta_pool)
330 return -ENOMEM;
331
332 return 0;
333 }
334
rsxx_dev_cleanup(void)335 void rsxx_dev_cleanup(void)
336 {
337 kmem_cache_destroy(bio_meta_pool);
338 }
339
340
341