drivers/nvdimm/blk.c

   1 /*
   2  * NVDIMM Block Window Driver
   3  * Copyright (c) 2014, Intel Corporation.
   4  *
   5  * This program is free software; you can redistribute it and/or modify it
   6  * under the terms and conditions of the GNU General Public License,
   7  * version 2, as published by the Free Software Foundation.
   8  *
   9  * This program is distributed in the hope it will be useful, but WITHOUT
  10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  12  * more details.
  13  */
  14
  15 #include <linux/blkdev.h>
  16 #include <linux/fs.h>
  17 #include <linux/genhd.h>
  18 #include <linux/module.h>
  19 #include <linux/moduleparam.h>
  20 #include <linux/nd.h>
  21 #include <linux/sizes.h>
  22 #include "nd.h"
  23
  24 static u32 nsblk_meta_size(struct nd_namespace_blk *nsblk)
  25 {
  26         return nsblk->lbasize - ((nsblk->lbasize >= 4096) ? 4096 : 512);
  27 }
  28
  29 static u32 nsblk_internal_lbasize(struct nd_namespace_blk *nsblk)
  30 {
  31         return roundup(nsblk->lbasize, INT_LBASIZE_ALIGNMENT);
  32 }
  33
  34 static u32 nsblk_sector_size(struct nd_namespace_blk *nsblk)
  35 {
  36         return nsblk->lbasize - nsblk_meta_size(nsblk);
  37 }
  38
  39 static resource_size_t to_dev_offset(struct nd_namespace_blk *nsblk,
  40                                 resource_size_t ns_offset, unsigned int len)
  41 {
  42         int i;
  43
  44         for (i = 0; i < nsblk->num_resources; i++) {
  45                 if (ns_offset < resource_size(nsblk->res[i])) {
  46                         if (ns_offset + len > resource_size(nsblk->res[i])) {
  47                                 dev_WARN_ONCE(&nsblk->common.dev, 1,
  48                                         "illegal request\n");
  49                                 return SIZE_MAX;
  50                         }
  51                         return nsblk->res[i]->start + ns_offset;
  52                 }
  53                 ns_offset -= resource_size(nsblk->res[i]);
  54         }
  55
  56         dev_WARN_ONCE(&nsblk->common.dev, 1, "request out of range\n");
  57         return SIZE_MAX;
  58 }
  59
  60 static struct nd_blk_region *to_ndbr(struct nd_namespace_blk *nsblk)
  61 {
  62         struct nd_region *nd_region;
  63         struct device *parent;
  64
  65         parent = nsblk->common.dev.parent;
  66         nd_region = container_of(parent, struct nd_region, dev);
  67         return container_of(nd_region, struct nd_blk_region, nd_region);
  68 }
  69
  70 #ifdef CONFIG_BLK_DEV_INTEGRITY
  71 static int nd_blk_rw_integrity(struct nd_namespace_blk *nsblk,
  72                 struct bio_integrity_payload *bip, u64 lba, int rw)
  73 {
  74         struct nd_blk_region *ndbr = to_ndbr(nsblk);
  75         unsigned int len = nsblk_meta_size(nsblk);
  76         resource_size_t dev_offset, ns_offset;
  77         u32 internal_lbasize, sector_size;
  78         int err = 0;
  79
  80         internal_lbasize = nsblk_internal_lbasize(nsblk);
  81         sector_size = nsblk_sector_size(nsblk);
  82         ns_offset = lba * internal_lbasize + sector_size;
  83         dev_offset = to_dev_offset(nsblk, ns_offset, len);
  84         if (dev_offset == SIZE_MAX)
  85                 return -EIO;
  86
  87         while (len) {
  88                 unsigned int cur_len;
  89                 struct bio_vec bv;
  90                 void *iobuf;
  91
  92                 bv = bvec_iter_bvec(bip->bip_vec, bip->bip_iter);
  93                 /*
  94                  * The 'bv' obtained from bvec_iter_bvec has its .bv_len and
  95                  * .bv_offset already adjusted for iter->bi_bvec_done, and we
  96                  * can use those directly
  97                  */
  98
  99                 cur_len = min(len, bv.bv_len);
 100                 iobuf = kmap_atomic(bv.bv_page);
 101                 err = ndbr->do_io(ndbr, dev_offset, iobuf + bv.bv_offset,
 102                                 cur_len, rw);
 103                 kunmap_atomic(iobuf);
 104                 if (err)
 105                         return err;
 106
 107                 len -= cur_len;
 108                 dev_offset += cur_len;
 109                 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, cur_len);
 110         }
 111
 112         return err;
 113 }
 114
 115 #else /* CONFIG_BLK_DEV_INTEGRITY */
 116 static int nd_blk_rw_integrity(struct nd_namespace_blk *nsblk,
 117                 struct bio_integrity_payload *bip, u64 lba, int rw)
 118 {
 119         return 0;
 120 }
 121 #endif
 122
 123 static int nsblk_do_bvec(struct nd_namespace_blk *nsblk,
 124                 struct bio_integrity_payload *bip, struct page *page,
 125                 unsigned int len, unsigned int off, int rw, sector_t sector)
 126 {
 127         struct nd_blk_region *ndbr = to_ndbr(nsblk);
 128         resource_size_t dev_offset, ns_offset;
 129         u32 internal_lbasize, sector_size;
 130         int err = 0;
 131         void *iobuf;
 132         u64 lba;
 133
 134         internal_lbasize = nsblk_internal_lbasize(nsblk);
 135         sector_size = nsblk_sector_size(nsblk);
 136         while (len) {
 137                 unsigned int cur_len;
 138
 139                 /*
 140                  * If we don't have an integrity payload, we don't have to
 141                  * split the bvec into sectors, as this would cause unnecessary
 142                  * Block Window setup/move steps. the do_io routine is capable
 143                  * of handling len <= PAGE_SIZE.
 144                  */
 145                 cur_len = bip ? min(len, sector_size) : len;
 146
 147                 lba = div_u64(sector << SECTOR_SHIFT, sector_size);
 148                 ns_offset = lba * internal_lbasize;
 149                 dev_offset = to_dev_offset(nsblk, ns_offset, cur_len);
 150                 if (dev_offset == SIZE_MAX)
 151                         return -EIO;
 152
 153                 iobuf = kmap_atomic(page);
 154                 err = ndbr->do_io(ndbr, dev_offset, iobuf + off, cur_len, rw);
 155                 kunmap_atomic(iobuf);
 156                 if (err)
 157                         return err;
 158
 159                 if (bip) {
 160                         err = nd_blk_rw_integrity(nsblk, bip, lba, rw);
 161                         if (err)
 162                                 return err;
 163                 }
 164                 len -= cur_len;
 165                 off += cur_len;
 166                 sector += sector_size >> SECTOR_SHIFT;
 167         }
 168
 169         return err;
 170 }
 171
 172 static blk_qc_t nd_blk_make_request(struct request_queue *q, struct bio *bio)
 173 {
 174         struct bio_integrity_payload *bip;
 175         struct nd_namespace_blk *nsblk;
 176         struct bvec_iter iter;
 177         unsigned long start;
 178         struct bio_vec bvec;
 179         int err = 0, rw;
 180         bool do_acct;
 181
 182         if (!bio_integrity_prep(bio))
 183                 return BLK_QC_T_NONE;
 184
 185         bip = bio_integrity(bio);
 186         nsblk = q->queuedata;
 187         rw = bio_data_dir(bio);
 188         do_acct = nd_iostat_start(bio, &start);
 189         bio_for_each_segment(bvec, bio, iter) {
 190                 unsigned int len = bvec.bv_len;
 191
 192                 BUG_ON(len > PAGE_SIZE);
 193                 err = nsblk_do_bvec(nsblk, bip, bvec.bv_page, len,
 194                                 bvec.bv_offset, rw, iter.bi_sector);
 195                 if (err) {
 196                         dev_dbg(&nsblk->common.dev,
 197                                         "io error in %s sector %lld, len %d,\n",
 198                                         (rw == READ) ? "READ" : "WRITE",
 199                                         (unsigned long long) iter.bi_sector, len);
 200                         bio->bi_status = errno_to_blk_status(err);
 201                         break;
 202                 }
 203         }
 204         if (do_acct)
 205                 nd_iostat_end(bio, start);
 206
 207         bio_endio(bio);
 208         return BLK_QC_T_NONE;
 209 }
 210
 211 static int nsblk_rw_bytes(struct nd_namespace_common *ndns,
 212                 resource_size_t offset, void *iobuf, size_t n, int rw,
 213                 unsigned long flags)
 214 {
 215         struct nd_namespace_blk *nsblk = to_nd_namespace_blk(&ndns->dev);
 216         struct nd_blk_region *ndbr = to_ndbr(nsblk);
 217         resource_size_t dev_offset;
 218
 219         dev_offset = to_dev_offset(nsblk, offset, n);
 220
 221         if (unlikely(offset + n > nsblk->size)) {
 222                 dev_WARN_ONCE(&ndns->dev, 1, "request out of range\n");
 223                 return -EFAULT;
 224         }
 225
 226         if (dev_offset == SIZE_MAX)
 227                 return -EIO;
 228
 229         return ndbr->do_io(ndbr, dev_offset, iobuf, n, rw);
 230 }
 231
 232 static const struct block_device_operations nd_blk_fops = {
 233         .owner = THIS_MODULE,
 234         .revalidate_disk = nvdimm_revalidate_disk,
 235 };
 236
 237 static void nd_blk_release_queue(void *q)
 238 {
 239         blk_cleanup_queue(q);
 240 }
 241
 242 static void nd_blk_release_disk(void *disk)
 243 {
 244         del_gendisk(disk);
 245         put_disk(disk);
 246 }
 247
 248 static int nsblk_attach_disk(struct nd_namespace_blk *nsblk)
 249 {
 250         struct device *dev = &nsblk->common.dev;
 251         resource_size_t available_disk_size;
 252         struct request_queue *q;
 253         struct gendisk *disk;
 254         u64 internal_nlba;
 255
 256         internal_nlba = div_u64(nsblk->size, nsblk_internal_lbasize(nsblk));
 257         available_disk_size = internal_nlba * nsblk_sector_size(nsblk);
 258
 259         q = blk_alloc_queue(GFP_KERNEL);
 260         if (!q)
 261                 return -ENOMEM;
 262         if (devm_add_action_or_reset(dev, nd_blk_release_queue, q))
 263                 return -ENOMEM;
 264
 265         blk_queue_make_request(q, nd_blk_make_request);
 266         blk_queue_max_hw_sectors(q, UINT_MAX);
 267         blk_queue_logical_block_size(q, nsblk_sector_size(nsblk));
 268         queue_flag_set_unlocked(QUEUE_FLAG_NONROT, q);
 269         q->queuedata = nsblk;
 270
 271         disk = alloc_disk(0);
 272         if (!disk)
 273                 return -ENOMEM;
 274
 275         disk->first_minor       = 0;
 276         disk->fops              = &nd_blk_fops;
 277         disk->queue             = q;
 278         disk->flags             = GENHD_FL_EXT_DEVT;
 279         nvdimm_namespace_disk_name(&nsblk->common, disk->disk_name);
 280         set_capacity(disk, 0);
 281         device_add_disk(dev, disk);
 282
 283         if (devm_add_action_or_reset(dev, nd_blk_release_disk, disk))
 284                 return -ENOMEM;
 285
 286         if (nsblk_meta_size(nsblk)) {
 287                 int rc = nd_integrity_init(disk, nsblk_meta_size(nsblk));
 288
 289                 if (rc)
 290                         return rc;
 291         }
 292
 293         set_capacity(disk, available_disk_size >> SECTOR_SHIFT);
 294         revalidate_disk(disk);
 295         return 0;
 296 }
 297
 298 static int nd_blk_probe(struct device *dev)
 299 {
 300         struct nd_namespace_common *ndns;
 301         struct nd_namespace_blk *nsblk;
 302
 303         ndns = nvdimm_namespace_common_probe(dev);
 304         if (IS_ERR(ndns))
 305                 return PTR_ERR(ndns);
 306
 307         nsblk = to_nd_namespace_blk(&ndns->dev);
 308         nsblk->size = nvdimm_namespace_capacity(ndns);
 309         dev_set_drvdata(dev, nsblk);
 310
 311         ndns->rw_bytes = nsblk_rw_bytes;
 312         if (is_nd_btt(dev))
 313                 return nvdimm_namespace_attach_btt(ndns);
 314         else if (nd_btt_probe(dev, ndns) == 0) {
 315                 /* we'll come back as btt-blk */
 316                 return -ENXIO;
 317         } else
 318                 return nsblk_attach_disk(nsblk);
 319 }
 320
 321 static int nd_blk_remove(struct device *dev)
 322 {
 323         if (is_nd_btt(dev))
 324                 nvdimm_namespace_detach_btt(to_nd_btt(dev));
 325         return 0;
 326 }
 327
 328 static struct nd_device_driver nd_blk_driver = {
 329         .probe = nd_blk_probe,
 330         .remove = nd_blk_remove,
 331         .drv = {
 332                 .name = "nd_blk",
 333         },
 334         .type = ND_DRIVER_NAMESPACE_BLK,
 335 };
 336
 337 static int __init nd_blk_init(void)
 338 {
 339         return nd_driver_register(&nd_blk_driver);
 340 }
 341
 342 static void __exit nd_blk_exit(void)
 343 {
 344         driver_unregister(&nd_blk_driver.drv);
 345 }
 346
 347 MODULE_AUTHOR("Ross Zwisler <ross.zwisler@linux.intel.com>");
 348 MODULE_LICENSE("GPL v2");
 349 MODULE_ALIAS_ND_DEVICE(ND_DEVICE_NAMESPACE_BLK);
 350 module_init(nd_blk_init);
 351 module_exit(nd_blk_exit);