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 if (atomic_read(&meta->error))
141 bio_io_error(meta->bio);
142 else
143 bio_endio(meta->bio);
144 kmem_cache_free(bio_meta_pool, meta);
145 }
146 }
147
rsxx_make_request(struct request_queue * q,struct bio * bio)148 static blk_qc_t rsxx_make_request(struct request_queue *q, struct bio *bio)
149 {
150 struct rsxx_cardinfo *card = q->queuedata;
151 struct rsxx_bio_meta *bio_meta;
152 int st = -EINVAL;
153
154 blk_queue_split(q, &bio, q->bio_split);
155
156 might_sleep();
157
158 if (!card)
159 goto req_err;
160
161 if (bio_end_sector(bio) > get_capacity(card->gendisk))
162 goto req_err;
163
164 if (unlikely(card->halt)) {
165 st = -EFAULT;
166 goto req_err;
167 }
168
169 if (unlikely(card->dma_fault)) {
170 st = (-EFAULT);
171 goto req_err;
172 }
173
174 if (bio->bi_iter.bi_size == 0) {
175 dev_err(CARD_TO_DEV(card), "size zero BIO!\n");
176 goto req_err;
177 }
178
179 bio_meta = kmem_cache_alloc(bio_meta_pool, GFP_KERNEL);
180 if (!bio_meta) {
181 st = -ENOMEM;
182 goto req_err;
183 }
184
185 bio_meta->bio = bio;
186 atomic_set(&bio_meta->error, 0);
187 atomic_set(&bio_meta->pending_dmas, 0);
188 bio_meta->start_time = jiffies;
189
190 if (!unlikely(card->halt))
191 disk_stats_start(card, bio);
192
193 dev_dbg(CARD_TO_DEV(card), "BIO[%c]: meta: %p addr8: x%llx size: %d\n",
194 bio_data_dir(bio) ? 'W' : 'R', bio_meta,
195 (u64)bio->bi_iter.bi_sector << 9, bio->bi_iter.bi_size);
196
197 st = rsxx_dma_queue_bio(card, bio, &bio_meta->pending_dmas,
198 bio_dma_done_cb, bio_meta);
199 if (st)
200 goto queue_err;
201
202 return BLK_QC_T_NONE;
203
204 queue_err:
205 kmem_cache_free(bio_meta_pool, bio_meta);
206 req_err:
207 if (st)
208 bio->bi_error = st;
209 bio_endio(bio);
210 return BLK_QC_T_NONE;
211 }
212
213 /*----------------- Device Setup -------------------*/
rsxx_discard_supported(struct rsxx_cardinfo * card)214 static bool rsxx_discard_supported(struct rsxx_cardinfo *card)
215 {
216 unsigned char pci_rev;
217
218 pci_read_config_byte(card->dev, PCI_REVISION_ID, &pci_rev);
219
220 return (pci_rev >= RSXX_DISCARD_SUPPORT);
221 }
222
rsxx_attach_dev(struct rsxx_cardinfo * card)223 int rsxx_attach_dev(struct rsxx_cardinfo *card)
224 {
225 mutex_lock(&card->dev_lock);
226
227 /* The block device requires the stripe size from the config. */
228 if (enable_blkdev) {
229 if (card->config_valid)
230 set_capacity(card->gendisk, card->size8 >> 9);
231 else
232 set_capacity(card->gendisk, 0);
233 add_disk(card->gendisk);
234
235 card->bdev_attached = 1;
236 }
237
238 mutex_unlock(&card->dev_lock);
239
240 return 0;
241 }
242
rsxx_detach_dev(struct rsxx_cardinfo * card)243 void rsxx_detach_dev(struct rsxx_cardinfo *card)
244 {
245 mutex_lock(&card->dev_lock);
246
247 if (card->bdev_attached) {
248 del_gendisk(card->gendisk);
249 card->bdev_attached = 0;
250 }
251
252 mutex_unlock(&card->dev_lock);
253 }
254
rsxx_setup_dev(struct rsxx_cardinfo * card)255 int rsxx_setup_dev(struct rsxx_cardinfo *card)
256 {
257 unsigned short blk_size;
258
259 mutex_init(&card->dev_lock);
260
261 if (!enable_blkdev)
262 return 0;
263
264 card->major = register_blkdev(0, DRIVER_NAME);
265 if (card->major < 0) {
266 dev_err(CARD_TO_DEV(card), "Failed to get major number\n");
267 return -ENOMEM;
268 }
269
270 card->queue = blk_alloc_queue(GFP_KERNEL);
271 if (!card->queue) {
272 dev_err(CARD_TO_DEV(card), "Failed queue alloc\n");
273 unregister_blkdev(card->major, DRIVER_NAME);
274 return -ENOMEM;
275 }
276
277 card->gendisk = alloc_disk(blkdev_minors);
278 if (!card->gendisk) {
279 dev_err(CARD_TO_DEV(card), "Failed disk alloc\n");
280 blk_cleanup_queue(card->queue);
281 unregister_blkdev(card->major, DRIVER_NAME);
282 return -ENOMEM;
283 }
284
285 if (card->config_valid) {
286 blk_size = card->config.data.block_size;
287 blk_queue_dma_alignment(card->queue, blk_size - 1);
288 blk_queue_logical_block_size(card->queue, blk_size);
289 }
290
291 blk_queue_make_request(card->queue, rsxx_make_request);
292 blk_queue_bounce_limit(card->queue, BLK_BOUNCE_ANY);
293 blk_queue_max_hw_sectors(card->queue, blkdev_max_hw_sectors);
294 blk_queue_physical_block_size(card->queue, RSXX_HW_BLK_SIZE);
295
296 queue_flag_set_unlocked(QUEUE_FLAG_NONROT, card->queue);
297 queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, card->queue);
298 if (rsxx_discard_supported(card)) {
299 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, card->queue);
300 blk_queue_max_discard_sectors(card->queue,
301 RSXX_HW_BLK_SIZE >> 9);
302 card->queue->limits.discard_granularity = RSXX_HW_BLK_SIZE;
303 card->queue->limits.discard_alignment = RSXX_HW_BLK_SIZE;
304 card->queue->limits.discard_zeroes_data = 1;
305 }
306
307 card->queue->queuedata = card;
308
309 snprintf(card->gendisk->disk_name, sizeof(card->gendisk->disk_name),
310 "rsxx%d", card->disk_id);
311 card->gendisk->driverfs_dev = &card->dev->dev;
312 card->gendisk->major = card->major;
313 card->gendisk->first_minor = 0;
314 card->gendisk->fops = &rsxx_fops;
315 card->gendisk->private_data = card;
316 card->gendisk->queue = card->queue;
317
318 return 0;
319 }
320
rsxx_destroy_dev(struct rsxx_cardinfo * card)321 void rsxx_destroy_dev(struct rsxx_cardinfo *card)
322 {
323 if (!enable_blkdev)
324 return;
325
326 put_disk(card->gendisk);
327 card->gendisk = NULL;
328
329 blk_cleanup_queue(card->queue);
330 card->queue->queuedata = NULL;
331 unregister_blkdev(card->major, DRIVER_NAME);
332 }
333
rsxx_dev_init(void)334 int rsxx_dev_init(void)
335 {
336 bio_meta_pool = KMEM_CACHE(rsxx_bio_meta, SLAB_HWCACHE_ALIGN);
337 if (!bio_meta_pool)
338 return -ENOMEM;
339
340 return 0;
341 }
342
rsxx_dev_cleanup(void)343 void rsxx_dev_cleanup(void)
344 {
345 kmem_cache_destroy(bio_meta_pool);
346 }
347
348
349