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 40static unsigned int blkdev_minors = 64; 41module_param(blkdev_minors, uint, 0444); 42MODULE_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 */ 49static unsigned int blkdev_max_hw_sectors = 1024; 50module_param(blkdev_max_hw_sectors, uint, 0444); 51MODULE_PARM_DESC(blkdev_max_hw_sectors, "Max hw sectors for a single BIO"); 52 53static unsigned int enable_blkdev = 1; 54module_param(enable_blkdev , uint, 0444); 55MODULE_PARM_DESC(enable_blkdev, "Enable block device interfaces"); 56 57 58struct rsxx_bio_meta { 59 struct bio *bio; 60 atomic_t pending_dmas; 61 atomic_t error; 62 unsigned long start_time; 63}; 64 65static struct kmem_cache *bio_meta_pool; 66 67/*----------------- Block Device Operations -----------------*/ 68static 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 85static 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 107static const struct block_device_operations rsxx_fops = { 108 .owner = THIS_MODULE, 109 .getgeo = rsxx_getgeo, 110 .ioctl = rsxx_blkdev_ioctl, 111}; 112 113static void disk_stats_start(struct rsxx_cardinfo *card, struct bio *bio) 114{ 115 struct hd_struct *part0 = &card->gendisk->part0; 116 int rw = bio_data_dir(bio); 117 int cpu; 118 119 cpu = part_stat_lock(); 120 121 part_round_stats(cpu, part0); 122 part_inc_in_flight(part0, rw); 123 124 part_stat_unlock(); 125} 126 127static void disk_stats_complete(struct rsxx_cardinfo *card, 128 struct bio *bio, 129 unsigned long start_time) 130{ 131 struct hd_struct *part0 = &card->gendisk->part0; 132 unsigned long duration = jiffies - start_time; 133 int rw = bio_data_dir(bio); 134 int cpu; 135 136 cpu = part_stat_lock(); 137 138 part_stat_add(cpu, part0, sectors[rw], bio_sectors(bio)); 139 part_stat_inc(cpu, part0, ios[rw]); 140 part_stat_add(cpu, part0, ticks[rw], duration); 141 142 part_round_stats(cpu, part0); 143 part_dec_in_flight(part0, rw); 144 145 part_stat_unlock(); 146} 147 148static void bio_dma_done_cb(struct rsxx_cardinfo *card, 149 void *cb_data, 150 unsigned int error) 151{ 152 struct rsxx_bio_meta *meta = cb_data; 153 154 if (error) 155 atomic_set(&meta->error, 1); 156 157 if (atomic_dec_and_test(&meta->pending_dmas)) { 158 if (!card->eeh_state && card->gendisk) 159 disk_stats_complete(card, meta->bio, meta->start_time); 160 161 bio_endio(meta->bio, atomic_read(&meta->error) ? -EIO : 0); 162 kmem_cache_free(bio_meta_pool, meta); 163 } 164} 165 166static void rsxx_make_request(struct request_queue *q, struct bio *bio) 167{ 168 struct rsxx_cardinfo *card = q->queuedata; 169 struct rsxx_bio_meta *bio_meta; 170 int st = -EINVAL; 171 172 might_sleep(); 173 174 if (!card) 175 goto req_err; 176 177 if (bio_end_sector(bio) > get_capacity(card->gendisk)) 178 goto req_err; 179 180 if (unlikely(card->halt)) { 181 st = -EFAULT; 182 goto req_err; 183 } 184 185 if (unlikely(card->dma_fault)) { 186 st = (-EFAULT); 187 goto req_err; 188 } 189 190 if (bio->bi_iter.bi_size == 0) { 191 dev_err(CARD_TO_DEV(card), "size zero BIO!\n"); 192 goto req_err; 193 } 194 195 bio_meta = kmem_cache_alloc(bio_meta_pool, GFP_KERNEL); 196 if (!bio_meta) { 197 st = -ENOMEM; 198 goto req_err; 199 } 200 201 bio_meta->bio = bio; 202 atomic_set(&bio_meta->error, 0); 203 atomic_set(&bio_meta->pending_dmas, 0); 204 bio_meta->start_time = jiffies; 205 206 if (!unlikely(card->halt)) 207 disk_stats_start(card, bio); 208 209 dev_dbg(CARD_TO_DEV(card), "BIO[%c]: meta: %p addr8: x%llx size: %d\n", 210 bio_data_dir(bio) ? 'W' : 'R', bio_meta, 211 (u64)bio->bi_iter.bi_sector << 9, bio->bi_iter.bi_size); 212 213 st = rsxx_dma_queue_bio(card, bio, &bio_meta->pending_dmas, 214 bio_dma_done_cb, bio_meta); 215 if (st) 216 goto queue_err; 217 218 return; 219 220queue_err: 221 kmem_cache_free(bio_meta_pool, bio_meta); 222req_err: 223 bio_endio(bio, st); 224} 225 226/*----------------- Device Setup -------------------*/ 227static bool rsxx_discard_supported(struct rsxx_cardinfo *card) 228{ 229 unsigned char pci_rev; 230 231 pci_read_config_byte(card->dev, PCI_REVISION_ID, &pci_rev); 232 233 return (pci_rev >= RSXX_DISCARD_SUPPORT); 234} 235 236int rsxx_attach_dev(struct rsxx_cardinfo *card) 237{ 238 mutex_lock(&card->dev_lock); 239 240 /* The block device requires the stripe size from the config. */ 241 if (enable_blkdev) { 242 if (card->config_valid) 243 set_capacity(card->gendisk, card->size8 >> 9); 244 else 245 set_capacity(card->gendisk, 0); 246 add_disk(card->gendisk); 247 248 card->bdev_attached = 1; 249 } 250 251 mutex_unlock(&card->dev_lock); 252 253 return 0; 254} 255 256void rsxx_detach_dev(struct rsxx_cardinfo *card) 257{ 258 mutex_lock(&card->dev_lock); 259 260 if (card->bdev_attached) { 261 del_gendisk(card->gendisk); 262 card->bdev_attached = 0; 263 } 264 265 mutex_unlock(&card->dev_lock); 266} 267 268int rsxx_setup_dev(struct rsxx_cardinfo *card) 269{ 270 unsigned short blk_size; 271 272 mutex_init(&card->dev_lock); 273 274 if (!enable_blkdev) 275 return 0; 276 277 card->major = register_blkdev(0, DRIVER_NAME); 278 if (card->major < 0) { 279 dev_err(CARD_TO_DEV(card), "Failed to get major number\n"); 280 return -ENOMEM; 281 } 282 283 card->queue = blk_alloc_queue(GFP_KERNEL); 284 if (!card->queue) { 285 dev_err(CARD_TO_DEV(card), "Failed queue alloc\n"); 286 unregister_blkdev(card->major, DRIVER_NAME); 287 return -ENOMEM; 288 } 289 290 card->gendisk = alloc_disk(blkdev_minors); 291 if (!card->gendisk) { 292 dev_err(CARD_TO_DEV(card), "Failed disk alloc\n"); 293 blk_cleanup_queue(card->queue); 294 unregister_blkdev(card->major, DRIVER_NAME); 295 return -ENOMEM; 296 } 297 298 if (card->config_valid) { 299 blk_size = card->config.data.block_size; 300 blk_queue_dma_alignment(card->queue, blk_size - 1); 301 blk_queue_logical_block_size(card->queue, blk_size); 302 } 303 304 blk_queue_make_request(card->queue, rsxx_make_request); 305 blk_queue_bounce_limit(card->queue, BLK_BOUNCE_ANY); 306 blk_queue_max_hw_sectors(card->queue, blkdev_max_hw_sectors); 307 blk_queue_physical_block_size(card->queue, RSXX_HW_BLK_SIZE); 308 309 queue_flag_set_unlocked(QUEUE_FLAG_NONROT, card->queue); 310 queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, card->queue); 311 if (rsxx_discard_supported(card)) { 312 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, card->queue); 313 blk_queue_max_discard_sectors(card->queue, 314 RSXX_HW_BLK_SIZE >> 9); 315 card->queue->limits.discard_granularity = RSXX_HW_BLK_SIZE; 316 card->queue->limits.discard_alignment = RSXX_HW_BLK_SIZE; 317 card->queue->limits.discard_zeroes_data = 1; 318 } 319 320 card->queue->queuedata = card; 321 322 snprintf(card->gendisk->disk_name, sizeof(card->gendisk->disk_name), 323 "rsxx%d", card->disk_id); 324 card->gendisk->driverfs_dev = &card->dev->dev; 325 card->gendisk->major = card->major; 326 card->gendisk->first_minor = 0; 327 card->gendisk->fops = &rsxx_fops; 328 card->gendisk->private_data = card; 329 card->gendisk->queue = card->queue; 330 331 return 0; 332} 333 334void rsxx_destroy_dev(struct rsxx_cardinfo *card) 335{ 336 if (!enable_blkdev) 337 return; 338 339 put_disk(card->gendisk); 340 card->gendisk = NULL; 341 342 blk_cleanup_queue(card->queue); 343 card->queue->queuedata = NULL; 344 unregister_blkdev(card->major, DRIVER_NAME); 345} 346 347int rsxx_dev_init(void) 348{ 349 bio_meta_pool = KMEM_CACHE(rsxx_bio_meta, SLAB_HWCACHE_ALIGN); 350 if (!bio_meta_pool) 351 return -ENOMEM; 352 353 return 0; 354} 355 356void rsxx_dev_cleanup(void) 357{ 358 kmem_cache_destroy(bio_meta_pool); 359} 360 361 362