drivers/nvme/host/lightnvm.c

   1 /*
   2  * nvme-lightnvm.c - LightNVM NVMe device
   3  *
   4  * Copyright (C) 2014-2015 IT University of Copenhagen
   5  * Initial release: Matias Bjorling <mb@lightnvm.io>
   6  *
   7  * This program is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU General Public License version
   9  * 2 as published by the Free Software Foundation.
  10  *
  11  * This program is distributed in the hope that it will be useful, but
  12  * WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * along with this program; see the file COPYING.  If not, write to
  18  * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139,
  19  * USA.
  20  *
  21  */
  22
  23 #include "nvme.h"
  24
  25 #include <linux/nvme.h>
  26 #include <linux/bitops.h>
  27 #include <linux/lightnvm.h>
  28 #include <linux/vmalloc.h>
  29 #include <linux/sched/sysctl.h>
  30 #include <uapi/linux/lightnvm.h>
  31
  32 enum nvme_nvm_admin_opcode {
  33         nvme_nvm_admin_identity         = 0xe2,
  34         nvme_nvm_admin_get_bb_tbl       = 0xf2,
  35         nvme_nvm_admin_set_bb_tbl       = 0xf1,
  36 };
  37
  38 struct nvme_nvm_ph_rw {
  39         __u8                    opcode;
  40         __u8                    flags;
  41         __u16                   command_id;
  42         __le32                  nsid;
  43         __u64                   rsvd2;
  44         __le64                  metadata;
  45         __le64                  prp1;
  46         __le64                  prp2;
  47         __le64                  spba;
  48         __le16                  length;
  49         __le16                  control;
  50         __le32                  dsmgmt;
  51         __le64                  resv;
  52 };
  53
  54 struct nvme_nvm_identity {
  55         __u8                    opcode;
  56         __u8                    flags;
  57         __u16                   command_id;
  58         __le32                  nsid;
  59         __u64                   rsvd[2];
  60         __le64                  prp1;
  61         __le64                  prp2;
  62         __le32                  chnl_off;
  63         __u32                   rsvd11[5];
  64 };
  65
  66 struct nvme_nvm_getbbtbl {
  67         __u8                    opcode;
  68         __u8                    flags;
  69         __u16                   command_id;
  70         __le32                  nsid;
  71         __u64                   rsvd[2];
  72         __le64                  prp1;
  73         __le64                  prp2;
  74         __le64                  spba;
  75         __u32                   rsvd4[4];
  76 };
  77
  78 struct nvme_nvm_setbbtbl {
  79         __u8                    opcode;
  80         __u8                    flags;
  81         __u16                   command_id;
  82         __le32                  nsid;
  83         __le64                  rsvd[2];
  84         __le64                  prp1;
  85         __le64                  prp2;
  86         __le64                  spba;
  87         __le16                  nlb;
  88         __u8                    value;
  89         __u8                    rsvd3;
  90         __u32                   rsvd4[3];
  91 };
  92
  93 struct nvme_nvm_erase_blk {
  94         __u8                    opcode;
  95         __u8                    flags;
  96         __u16                   command_id;
  97         __le32                  nsid;
  98         __u64                   rsvd[2];
  99         __le64                  prp1;
 100         __le64                  prp2;
 101         __le64                  spba;
 102         __le16                  length;
 103         __le16                  control;
 104         __le32                  dsmgmt;
 105         __le64                  resv;
 106 };
 107
 108 struct nvme_nvm_command {
 109         union {
 110                 struct nvme_common_command common;
 111                 struct nvme_nvm_identity identity;
 112                 struct nvme_nvm_ph_rw ph_rw;
 113                 struct nvme_nvm_getbbtbl get_bb;
 114                 struct nvme_nvm_setbbtbl set_bb;
 115                 struct nvme_nvm_erase_blk erase;
 116         };
 117 };
 118
 119 #define NVME_NVM_LP_MLC_PAIRS 886
 120 struct nvme_nvm_lp_mlc {
 121         __le16                  num_pairs;
 122         __u8                    pairs[NVME_NVM_LP_MLC_PAIRS];
 123 };
 124
 125 struct nvme_nvm_lp_tbl {
 126         __u8                    id[8];
 127         struct nvme_nvm_lp_mlc  mlc;
 128 };
 129
 130 struct nvme_nvm_id_group {
 131         __u8                    mtype;
 132         __u8                    fmtype;
 133         __le16                  res16;
 134         __u8                    num_ch;
 135         __u8                    num_lun;
 136         __u8                    num_pln;
 137         __u8                    rsvd1;
 138         __le16                  num_chk;
 139         __le16                  num_pg;
 140         __le16                  fpg_sz;
 141         __le16                  csecs;
 142         __le16                  sos;
 143         __le16                  rsvd2;
 144         __le32                  trdt;
 145         __le32                  trdm;
 146         __le32                  tprt;
 147         __le32                  tprm;
 148         __le32                  tbet;
 149         __le32                  tbem;
 150         __le32                  mpos;
 151         __le32                  mccap;
 152         __le16                  cpar;
 153         __u8                    reserved[10];
 154         struct nvme_nvm_lp_tbl lptbl;
 155 } __packed;
 156
 157 struct nvme_nvm_addr_format {
 158         __u8                    ch_offset;
 159         __u8                    ch_len;
 160         __u8                    lun_offset;
 161         __u8                    lun_len;
 162         __u8                    pln_offset;
 163         __u8                    pln_len;
 164         __u8                    blk_offset;
 165         __u8                    blk_len;
 166         __u8                    pg_offset;
 167         __u8                    pg_len;
 168         __u8                    sect_offset;
 169         __u8                    sect_len;
 170         __u8                    res[4];
 171 } __packed;
 172
 173 struct nvme_nvm_id {
 174         __u8                    ver_id;
 175         __u8                    vmnt;
 176         __u8                    cgrps;
 177         __u8                    res;
 178         __le32                  cap;
 179         __le32                  dom;
 180         struct nvme_nvm_addr_format ppaf;
 181         __u8                    resv[228];
 182         struct nvme_nvm_id_group groups[4];
 183 } __packed;
 184
 185 struct nvme_nvm_bb_tbl {
 186         __u8    tblid[4];
 187         __le16  verid;
 188         __le16  revid;
 189         __le32  rvsd1;
 190         __le32  tblks;
 191         __le32  tfact;
 192         __le32  tgrown;
 193         __le32  tdresv;
 194         __le32  thresv;
 195         __le32  rsvd2[8];
 196         __u8    blk[0];
 197 };
 198
 199 /*
 200  * Check we didn't inadvertently grow the command struct
 201  */
 202 static inline void _nvme_nvm_check_size(void)
 203 {
 204         BUILD_BUG_ON(sizeof(struct nvme_nvm_identity) != 64);
 205         BUILD_BUG_ON(sizeof(struct nvme_nvm_ph_rw) != 64);
 206         BUILD_BUG_ON(sizeof(struct nvme_nvm_getbbtbl) != 64);
 207         BUILD_BUG_ON(sizeof(struct nvme_nvm_setbbtbl) != 64);
 208         BUILD_BUG_ON(sizeof(struct nvme_nvm_erase_blk) != 64);
 209         BUILD_BUG_ON(sizeof(struct nvme_nvm_id_group) != 960);
 210         BUILD_BUG_ON(sizeof(struct nvme_nvm_addr_format) != 16);
 211         BUILD_BUG_ON(sizeof(struct nvme_nvm_id) != NVME_IDENTIFY_DATA_SIZE);
 212         BUILD_BUG_ON(sizeof(struct nvme_nvm_bb_tbl) != 64);
 213 }
 214
 215 static int init_grps(struct nvm_id *nvm_id, struct nvme_nvm_id *nvme_nvm_id)
 216 {
 217         struct nvme_nvm_id_group *src;
 218         struct nvm_id_group *grp;
 219         int sec_per_pg, sec_per_pl, pg_per_blk;
 220
 221         if (nvme_nvm_id->cgrps != 1)
 222                 return -EINVAL;
 223
 224         src = &nvme_nvm_id->groups[0];
 225         grp = &nvm_id->grp;
 226
 227         grp->mtype = src->mtype;
 228         grp->fmtype = src->fmtype;
 229
 230         grp->num_ch = src->num_ch;
 231         grp->num_lun = src->num_lun;
 232
 233         grp->num_chk = le16_to_cpu(src->num_chk);
 234         grp->csecs = le16_to_cpu(src->csecs);
 235         grp->sos = le16_to_cpu(src->sos);
 236
 237         pg_per_blk = le16_to_cpu(src->num_pg);
 238         sec_per_pg = le16_to_cpu(src->fpg_sz) / grp->csecs;
 239         sec_per_pl = sec_per_pg * src->num_pln;
 240         grp->clba = sec_per_pl * pg_per_blk;
 241         grp->ws_per_chk = pg_per_blk;
 242
 243         grp->mpos = le32_to_cpu(src->mpos);
 244         grp->cpar = le16_to_cpu(src->cpar);
 245         grp->mccap = le32_to_cpu(src->mccap);
 246
 247         grp->ws_opt = grp->ws_min = sec_per_pg;
 248         grp->ws_seq = NVM_IO_SNGL_ACCESS;
 249
 250         if (grp->mpos & 0x020202) {
 251                 grp->ws_seq = NVM_IO_DUAL_ACCESS;
 252                 grp->ws_opt <<= 1;
 253         } else if (grp->mpos & 0x040404) {
 254                 grp->ws_seq = NVM_IO_QUAD_ACCESS;
 255                 grp->ws_opt <<= 2;
 256         }
 257
 258         grp->trdt = le32_to_cpu(src->trdt);
 259         grp->trdm = le32_to_cpu(src->trdm);
 260         grp->tprt = le32_to_cpu(src->tprt);
 261         grp->tprm = le32_to_cpu(src->tprm);
 262         grp->tbet = le32_to_cpu(src->tbet);
 263         grp->tbem = le32_to_cpu(src->tbem);
 264
 265         /* 1.2 compatibility */
 266         grp->num_pln = src->num_pln;
 267         grp->num_pg = le16_to_cpu(src->num_pg);
 268         grp->fpg_sz = le16_to_cpu(src->fpg_sz);
 269
 270         return 0;
 271 }
 272
 273 static int nvme_nvm_identity(struct nvm_dev *nvmdev, struct nvm_id *nvm_id)
 274 {
 275         struct nvme_ns *ns = nvmdev->q->queuedata;
 276         struct nvme_nvm_id *nvme_nvm_id;
 277         struct nvme_nvm_command c = {};
 278         int ret;
 279
 280         c.identity.opcode = nvme_nvm_admin_identity;
 281         c.identity.nsid = cpu_to_le32(ns->head->ns_id);
 282         c.identity.chnl_off = 0;
 283
 284         nvme_nvm_id = kmalloc(sizeof(struct nvme_nvm_id), GFP_KERNEL);
 285         if (!nvme_nvm_id)
 286                 return -ENOMEM;
 287
 288         ret = nvme_submit_sync_cmd(ns->ctrl->admin_q, (struct nvme_command *)&c,
 289                                 nvme_nvm_id, sizeof(struct nvme_nvm_id));
 290         if (ret) {
 291                 ret = -EIO;
 292                 goto out;
 293         }
 294
 295         nvm_id->ver_id = nvme_nvm_id->ver_id;
 296         nvm_id->vmnt = nvme_nvm_id->vmnt;
 297         nvm_id->cap = le32_to_cpu(nvme_nvm_id->cap);
 298         nvm_id->dom = le32_to_cpu(nvme_nvm_id->dom);
 299         memcpy(&nvm_id->ppaf, &nvme_nvm_id->ppaf,
 300                                         sizeof(struct nvm_addr_format));
 301
 302         ret = init_grps(nvm_id, nvme_nvm_id);
 303 out:
 304         kfree(nvme_nvm_id);
 305         return ret;
 306 }
 307
 308 static int nvme_nvm_get_bb_tbl(struct nvm_dev *nvmdev, struct ppa_addr ppa,
 309                                                                 u8 *blks)
 310 {
 311         struct request_queue *q = nvmdev->q;
 312         struct nvm_geo *geo = &nvmdev->geo;
 313         struct nvme_ns *ns = q->queuedata;
 314         struct nvme_ctrl *ctrl = ns->ctrl;
 315         struct nvme_nvm_command c = {};
 316         struct nvme_nvm_bb_tbl *bb_tbl;
 317         int nr_blks = geo->nr_chks * geo->plane_mode;
 318         int tblsz = sizeof(struct nvme_nvm_bb_tbl) + nr_blks;
 319         int ret = 0;
 320
 321         c.get_bb.opcode = nvme_nvm_admin_get_bb_tbl;
 322         c.get_bb.nsid = cpu_to_le32(ns->head->ns_id);
 323         c.get_bb.spba = cpu_to_le64(ppa.ppa);
 324
 325         bb_tbl = kzalloc(tblsz, GFP_KERNEL);
 326         if (!bb_tbl)
 327                 return -ENOMEM;
 328
 329         ret = nvme_submit_sync_cmd(ctrl->admin_q, (struct nvme_command *)&c,
 330                                                                 bb_tbl, tblsz);
 331         if (ret) {
 332                 dev_err(ctrl->device, "get bad block table failed (%d)\n", ret);
 333                 ret = -EIO;
 334                 goto out;
 335         }
 336
 337         if (bb_tbl->tblid[0] != 'B' || bb_tbl->tblid[1] != 'B' ||
 338                 bb_tbl->tblid[2] != 'L' || bb_tbl->tblid[3] != 'T') {
 339                 dev_err(ctrl->device, "bbt format mismatch\n");
 340                 ret = -EINVAL;
 341                 goto out;
 342         }
 343
 344         if (le16_to_cpu(bb_tbl->verid) != 1) {
 345                 ret = -EINVAL;
 346                 dev_err(ctrl->device, "bbt version not supported\n");
 347                 goto out;
 348         }
 349
 350         if (le32_to_cpu(bb_tbl->tblks) != nr_blks) {
 351                 ret = -EINVAL;
 352                 dev_err(ctrl->device,
 353                                 "bbt unsuspected blocks returned (%u!=%u)",
 354                                 le32_to_cpu(bb_tbl->tblks), nr_blks);
 355                 goto out;
 356         }
 357
 358         memcpy(blks, bb_tbl->blk, geo->nr_chks * geo->plane_mode);
 359 out:
 360         kfree(bb_tbl);
 361         return ret;
 362 }
 363
 364 static int nvme_nvm_set_bb_tbl(struct nvm_dev *nvmdev, struct ppa_addr *ppas,
 365                                                         int nr_ppas, int type)
 366 {
 367         struct nvme_ns *ns = nvmdev->q->queuedata;
 368         struct nvme_nvm_command c = {};
 369         int ret = 0;
 370
 371         c.set_bb.opcode = nvme_nvm_admin_set_bb_tbl;
 372         c.set_bb.nsid = cpu_to_le32(ns->head->ns_id);
 373         c.set_bb.spba = cpu_to_le64(ppas->ppa);
 374         c.set_bb.nlb = cpu_to_le16(nr_ppas - 1);
 375         c.set_bb.value = type;
 376
 377         ret = nvme_submit_sync_cmd(ns->ctrl->admin_q, (struct nvme_command *)&c,
 378                                                                 NULL, 0);
 379         if (ret)
 380                 dev_err(ns->ctrl->device, "set bad block table failed (%d)\n",
 381                                                                         ret);
 382         return ret;
 383 }
 384
 385 static inline void nvme_nvm_rqtocmd(struct nvm_rq *rqd, struct nvme_ns *ns,
 386                                     struct nvme_nvm_command *c)
 387 {
 388         c->ph_rw.opcode = rqd->opcode;
 389         c->ph_rw.nsid = cpu_to_le32(ns->head->ns_id);
 390         c->ph_rw.spba = cpu_to_le64(rqd->ppa_addr.ppa);
 391         c->ph_rw.metadata = cpu_to_le64(rqd->dma_meta_list);
 392         c->ph_rw.control = cpu_to_le16(rqd->flags);
 393         c->ph_rw.length = cpu_to_le16(rqd->nr_ppas - 1);
 394 }
 395
 396 static void nvme_nvm_end_io(struct request *rq, blk_status_t status)
 397 {
 398         struct nvm_rq *rqd = rq->end_io_data;
 399
 400         rqd->ppa_status = le64_to_cpu(nvme_req(rq)->result.u64);
 401         rqd->error = nvme_req(rq)->status;
 402         nvm_end_io(rqd);
 403
 404         kfree(nvme_req(rq)->cmd);
 405         blk_mq_free_request(rq);
 406 }
 407
 408 static struct request *nvme_nvm_alloc_request(struct request_queue *q,
 409                                               struct nvm_rq *rqd,
 410                                               struct nvme_nvm_command *cmd)
 411 {
 412         struct nvme_ns *ns = q->queuedata;
 413         struct request *rq;
 414
 415         nvme_nvm_rqtocmd(rqd, ns, cmd);
 416
 417         rq = nvme_alloc_request(q, (struct nvme_command *)cmd, 0, NVME_QID_ANY);
 418         if (IS_ERR(rq))
 419                 return rq;
 420
 421         rq->cmd_flags &= ~REQ_FAILFAST_DRIVER;
 422
 423         if (rqd->bio) {
 424                 blk_init_request_from_bio(rq, rqd->bio);
 425         } else {
 426                 rq->ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, IOPRIO_NORM);
 427                 rq->__data_len = 0;
 428         }
 429
 430         return rq;
 431 }
 432
 433 static int nvme_nvm_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
 434 {
 435         struct request_queue *q = dev->q;
 436         struct nvme_nvm_command *cmd;
 437         struct request *rq;
 438
 439         cmd = kzalloc(sizeof(struct nvme_nvm_command), GFP_KERNEL);
 440         if (!cmd)
 441                 return -ENOMEM;
 442
 443         rq = nvme_nvm_alloc_request(q, rqd, cmd);
 444         if (IS_ERR(rq)) {
 445                 kfree(cmd);
 446                 return PTR_ERR(rq);
 447         }
 448
 449         rq->end_io_data = rqd;
 450
 451         blk_execute_rq_nowait(q, NULL, rq, 0, nvme_nvm_end_io);
 452
 453         return 0;
 454 }
 455
 456 static int nvme_nvm_submit_io_sync(struct nvm_dev *dev, struct nvm_rq *rqd)
 457 {
 458         struct request_queue *q = dev->q;
 459         struct request *rq;
 460         struct nvme_nvm_command cmd;
 461         int ret = 0;
 462
 463         memset(&cmd, 0, sizeof(struct nvme_nvm_command));
 464
 465         rq = nvme_nvm_alloc_request(q, rqd, &cmd);
 466         if (IS_ERR(rq))
 467                 return PTR_ERR(rq);
 468
 469         /* I/Os can fail and the error is signaled through rqd. Callers must
 470          * handle the error accordingly.
 471          */
 472         blk_execute_rq(q, NULL, rq, 0);
 473         if (nvme_req(rq)->flags & NVME_REQ_CANCELLED)
 474                 ret = -EINTR;
 475
 476         rqd->ppa_status = le64_to_cpu(nvme_req(rq)->result.u64);
 477         rqd->error = nvme_req(rq)->status;
 478
 479         blk_mq_free_request(rq);
 480
 481         return ret;
 482 }
 483
 484 static void *nvme_nvm_create_dma_pool(struct nvm_dev *nvmdev, char *name)
 485 {
 486         struct nvme_ns *ns = nvmdev->q->queuedata;
 487
 488         return dma_pool_create(name, ns->ctrl->dev, PAGE_SIZE, PAGE_SIZE, 0);
 489 }
 490
 491 static void nvme_nvm_destroy_dma_pool(void *pool)
 492 {
 493         struct dma_pool *dma_pool = pool;
 494
 495         dma_pool_destroy(dma_pool);
 496 }
 497
 498 static void *nvme_nvm_dev_dma_alloc(struct nvm_dev *dev, void *pool,
 499                                     gfp_t mem_flags, dma_addr_t *dma_handler)
 500 {
 501         return dma_pool_alloc(pool, mem_flags, dma_handler);
 502 }
 503
 504 static void nvme_nvm_dev_dma_free(void *pool, void *addr,
 505                                                         dma_addr_t dma_handler)
 506 {
 507         dma_pool_free(pool, addr, dma_handler);
 508 }
 509
 510 static struct nvm_dev_ops nvme_nvm_dev_ops = {
 511         .identity               = nvme_nvm_identity,
 512
 513         .get_bb_tbl             = nvme_nvm_get_bb_tbl,
 514         .set_bb_tbl             = nvme_nvm_set_bb_tbl,
 515
 516         .submit_io              = nvme_nvm_submit_io,
 517         .submit_io_sync         = nvme_nvm_submit_io_sync,
 518
 519         .create_dma_pool        = nvme_nvm_create_dma_pool,
 520         .destroy_dma_pool       = nvme_nvm_destroy_dma_pool,
 521         .dev_dma_alloc          = nvme_nvm_dev_dma_alloc,
 522         .dev_dma_free           = nvme_nvm_dev_dma_free,
 523
 524         .max_phys_sect          = 64,
 525 };
 526
 527 static int nvme_nvm_submit_user_cmd(struct request_queue *q,
 528                                 struct nvme_ns *ns,
 529                                 struct nvme_nvm_command *vcmd,
 530                                 void __user *ubuf, unsigned int bufflen,
 531                                 void __user *meta_buf, unsigned int meta_len,
 532                                 void __user *ppa_buf, unsigned int ppa_len,
 533                                 u32 *result, u64 *status, unsigned int timeout)
 534 {
 535         bool write = nvme_is_write((struct nvme_command *)vcmd);
 536         struct nvm_dev *dev = ns->ndev;
 537         struct gendisk *disk = ns->disk;
 538         struct request *rq;
 539         struct bio *bio = NULL;
 540         __le64 *ppa_list = NULL;
 541         dma_addr_t ppa_dma;
 542         __le64 *metadata = NULL;
 543         dma_addr_t metadata_dma;
 544         DECLARE_COMPLETION_ONSTACK(wait);
 545         int ret = 0;
 546
 547         rq = nvme_alloc_request(q, (struct nvme_command *)vcmd, 0,
 548                         NVME_QID_ANY);
 549         if (IS_ERR(rq)) {
 550                 ret = -ENOMEM;
 551                 goto err_cmd;
 552         }
 553
 554         rq->timeout = timeout ? timeout : ADMIN_TIMEOUT;
 555
 556         if (ppa_buf && ppa_len) {
 557                 ppa_list = dma_pool_alloc(dev->dma_pool, GFP_KERNEL, &ppa_dma);
 558                 if (!ppa_list) {
 559                         ret = -ENOMEM;
 560                         goto err_rq;
 561                 }
 562                 if (copy_from_user(ppa_list, (void __user *)ppa_buf,
 563                                                 sizeof(u64) * (ppa_len + 1))) {
 564                         ret = -EFAULT;
 565                         goto err_ppa;
 566                 }
 567                 vcmd->ph_rw.spba = cpu_to_le64(ppa_dma);
 568         } else {
 569                 vcmd->ph_rw.spba = cpu_to_le64((uintptr_t)ppa_buf);
 570         }
 571
 572         if (ubuf && bufflen) {
 573                 ret = blk_rq_map_user(q, rq, NULL, ubuf, bufflen, GFP_KERNEL);
 574                 if (ret)
 575                         goto err_ppa;
 576                 bio = rq->bio;
 577
 578                 if (meta_buf && meta_len) {
 579                         metadata = dma_pool_alloc(dev->dma_pool, GFP_KERNEL,
 580                                                                 &metadata_dma);
 581                         if (!metadata) {
 582                                 ret = -ENOMEM;
 583                                 goto err_map;
 584                         }
 585
 586                         if (write) {
 587                                 if (copy_from_user(metadata,
 588                                                 (void __user *)meta_buf,
 589                                                 meta_len)) {
 590                                         ret = -EFAULT;
 591                                         goto err_meta;
 592                                 }
 593                         }
 594                         vcmd->ph_rw.metadata = cpu_to_le64(metadata_dma);
 595                 }
 596
 597                 bio->bi_disk = disk;
 598         }
 599
 600         blk_execute_rq(q, NULL, rq, 0);
 601
 602         if (nvme_req(rq)->flags & NVME_REQ_CANCELLED)
 603                 ret = -EINTR;
 604         else if (nvme_req(rq)->status & 0x7ff)
 605                 ret = -EIO;
 606         if (result)
 607                 *result = nvme_req(rq)->status & 0x7ff;
 608         if (status)
 609                 *status = le64_to_cpu(nvme_req(rq)->result.u64);
 610
 611         if (metadata && !ret && !write) {
 612                 if (copy_to_user(meta_buf, (void *)metadata, meta_len))
 613                         ret = -EFAULT;
 614         }
 615 err_meta:
 616         if (meta_buf && meta_len)
 617                 dma_pool_free(dev->dma_pool, metadata, metadata_dma);
 618 err_map:
 619         if (bio)
 620                 blk_rq_unmap_user(bio);
 621 err_ppa:
 622         if (ppa_buf && ppa_len)
 623                 dma_pool_free(dev->dma_pool, ppa_list, ppa_dma);
 624 err_rq:
 625         blk_mq_free_request(rq);
 626 err_cmd:
 627         return ret;
 628 }
 629
 630 static int nvme_nvm_submit_vio(struct nvme_ns *ns,
 631                                         struct nvm_user_vio __user *uvio)
 632 {
 633         struct nvm_user_vio vio;
 634         struct nvme_nvm_command c;
 635         unsigned int length;
 636         int ret;
 637
 638         if (copy_from_user(&vio, uvio, sizeof(vio)))
 639                 return -EFAULT;
 640         if (vio.flags)
 641                 return -EINVAL;
 642
 643         memset(&c, 0, sizeof(c));
 644         c.ph_rw.opcode = vio.opcode;
 645         c.ph_rw.nsid = cpu_to_le32(ns->head->ns_id);
 646         c.ph_rw.control = cpu_to_le16(vio.control);
 647         c.ph_rw.length = cpu_to_le16(vio.nppas);
 648
 649         length = (vio.nppas + 1) << ns->lba_shift;
 650
 651         ret = nvme_nvm_submit_user_cmd(ns->queue, ns, &c,
 652                         (void __user *)(uintptr_t)vio.addr, length,
 653                         (void __user *)(uintptr_t)vio.metadata,
 654                                                         vio.metadata_len,
 655                         (void __user *)(uintptr_t)vio.ppa_list, vio.nppas,
 656                         &vio.result, &vio.status, 0);
 657
 658         if (ret && copy_to_user(uvio, &vio, sizeof(vio)))
 659                 return -EFAULT;
 660
 661         return ret;
 662 }
 663
 664 static int nvme_nvm_user_vcmd(struct nvme_ns *ns, int admin,
 665                                         struct nvm_passthru_vio __user *uvcmd)
 666 {
 667         struct nvm_passthru_vio vcmd;
 668         struct nvme_nvm_command c;
 669         struct request_queue *q;
 670         unsigned int timeout = 0;
 671         int ret;
 672
 673         if (copy_from_user(&vcmd, uvcmd, sizeof(vcmd)))
 674                 return -EFAULT;
 675         if ((vcmd.opcode != 0xF2) && (!capable(CAP_SYS_ADMIN)))
 676                 return -EACCES;
 677         if (vcmd.flags)
 678                 return -EINVAL;
 679
 680         memset(&c, 0, sizeof(c));
 681         c.common.opcode = vcmd.opcode;
 682         c.common.nsid = cpu_to_le32(ns->head->ns_id);
 683         c.common.cdw2[0] = cpu_to_le32(vcmd.cdw2);
 684         c.common.cdw2[1] = cpu_to_le32(vcmd.cdw3);
 685         /* cdw11-12 */
 686         c.ph_rw.length = cpu_to_le16(vcmd.nppas);
 687         c.ph_rw.control  = cpu_to_le16(vcmd.control);
 688         c.common.cdw10[3] = cpu_to_le32(vcmd.cdw13);
 689         c.common.cdw10[4] = cpu_to_le32(vcmd.cdw14);
 690         c.common.cdw10[5] = cpu_to_le32(vcmd.cdw15);
 691
 692         if (vcmd.timeout_ms)
 693                 timeout = msecs_to_jiffies(vcmd.timeout_ms);
 694
 695         q = admin ? ns->ctrl->admin_q : ns->queue;
 696
 697         ret = nvme_nvm_submit_user_cmd(q, ns,
 698                         (struct nvme_nvm_command *)&c,
 699                         (void __user *)(uintptr_t)vcmd.addr, vcmd.data_len,
 700                         (void __user *)(uintptr_t)vcmd.metadata,
 701                                                         vcmd.metadata_len,
 702                         (void __user *)(uintptr_t)vcmd.ppa_list, vcmd.nppas,
 703                         &vcmd.result, &vcmd.status, timeout);
 704
 705         if (ret && copy_to_user(uvcmd, &vcmd, sizeof(vcmd)))
 706                 return -EFAULT;
 707
 708         return ret;
 709 }
 710
 711 int nvme_nvm_ioctl(struct nvme_ns *ns, unsigned int cmd, unsigned long arg)
 712 {
 713         switch (cmd) {
 714         case NVME_NVM_IOCTL_ADMIN_VIO:
 715                 return nvme_nvm_user_vcmd(ns, 1, (void __user *)arg);
 716         case NVME_NVM_IOCTL_IO_VIO:
 717                 return nvme_nvm_user_vcmd(ns, 0, (void __user *)arg);
 718         case NVME_NVM_IOCTL_SUBMIT_VIO:
 719                 return nvme_nvm_submit_vio(ns, (void __user *)arg);
 720         default:
 721                 return -ENOTTY;
 722         }
 723 }
 724
 725 int nvme_nvm_register(struct nvme_ns *ns, char *disk_name, int node)
 726 {
 727         struct request_queue *q = ns->queue;
 728         struct nvm_dev *dev;
 729
 730         _nvme_nvm_check_size();
 731
 732         dev = nvm_alloc_dev(node);
 733         if (!dev)
 734                 return -ENOMEM;
 735
 736         dev->q = q;
 737         memcpy(dev->name, disk_name, DISK_NAME_LEN);
 738         dev->ops = &nvme_nvm_dev_ops;
 739         dev->private_data = ns;
 740         ns->ndev = dev;
 741
 742         return nvm_register(dev);
 743 }
 744
 745 void nvme_nvm_unregister(struct nvme_ns *ns)
 746 {
 747         nvm_unregister(ns->ndev);
 748 }
 749
 750 static ssize_t nvm_dev_attr_show(struct device *dev,
 751                                  struct device_attribute *dattr, char *page)
 752 {
 753         struct nvme_ns *ns = nvme_get_ns_from_dev(dev);
 754         struct nvm_dev *ndev = ns->ndev;
 755         struct nvm_id *id;
 756         struct nvm_id_group *grp;
 757         struct attribute *attr;
 758
 759         if (!ndev)
 760                 return 0;
 761
 762         id = &ndev->identity;
 763         grp = &id->grp;
 764         attr = &dattr->attr;
 765
 766         if (strcmp(attr->name, "version") == 0) {
 767                 return scnprintf(page, PAGE_SIZE, "%u\n", id->ver_id);
 768         } else if (strcmp(attr->name, "vendor_opcode") == 0) {
 769                 return scnprintf(page, PAGE_SIZE, "%u\n", id->vmnt);
 770         } else if (strcmp(attr->name, "capabilities") == 0) {
 771                 return scnprintf(page, PAGE_SIZE, "%u\n", id->cap);
 772         } else if (strcmp(attr->name, "device_mode") == 0) {
 773                 return scnprintf(page, PAGE_SIZE, "%u\n", id->dom);
 774         /* kept for compatibility */
 775         } else if (strcmp(attr->name, "media_manager") == 0) {
 776                 return scnprintf(page, PAGE_SIZE, "%s\n", "gennvm");
 777         } else if (strcmp(attr->name, "ppa_format") == 0) {
 778                 return scnprintf(page, PAGE_SIZE,
 779                         "0x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x\n",
 780                         id->ppaf.ch_offset, id->ppaf.ch_len,
 781                         id->ppaf.lun_offset, id->ppaf.lun_len,
 782                         id->ppaf.pln_offset, id->ppaf.pln_len,
 783                         id->ppaf.blk_offset, id->ppaf.blk_len,
 784                         id->ppaf.pg_offset, id->ppaf.pg_len,
 785                         id->ppaf.sect_offset, id->ppaf.sect_len);
 786         } else if (strcmp(attr->name, "media_type") == 0) {     /* u8 */
 787                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->mtype);
 788         } else if (strcmp(attr->name, "flash_media_type") == 0) {
 789                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->fmtype);
 790         } else if (strcmp(attr->name, "num_channels") == 0) {
 791                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->num_ch);
 792         } else if (strcmp(attr->name, "num_luns") == 0) {
 793                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->num_lun);
 794         } else if (strcmp(attr->name, "num_planes") == 0) {
 795                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->num_pln);
 796         } else if (strcmp(attr->name, "num_blocks") == 0) {     /* u16 */
 797                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->num_chk);
 798         } else if (strcmp(attr->name, "num_pages") == 0) {
 799                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->num_pg);
 800         } else if (strcmp(attr->name, "page_size") == 0) {
 801                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->fpg_sz);
 802         } else if (strcmp(attr->name, "hw_sector_size") == 0) {
 803                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->csecs);
 804         } else if (strcmp(attr->name, "oob_sector_size") == 0) {/* u32 */
 805                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->sos);
 806         } else if (strcmp(attr->name, "read_typ") == 0) {
 807                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->trdt);
 808         } else if (strcmp(attr->name, "read_max") == 0) {
 809                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->trdm);
 810         } else if (strcmp(attr->name, "prog_typ") == 0) {
 811                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->tprt);
 812         } else if (strcmp(attr->name, "prog_max") == 0) {
 813                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->tprm);
 814         } else if (strcmp(attr->name, "erase_typ") == 0) {
 815                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->tbet);
 816         } else if (strcmp(attr->name, "erase_max") == 0) {
 817                 return scnprintf(page, PAGE_SIZE, "%u\n", grp->tbem);
 818         } else if (strcmp(attr->name, "multiplane_modes") == 0) {
 819                 return scnprintf(page, PAGE_SIZE, "0x%08x\n", grp->mpos);
 820         } else if (strcmp(attr->name, "media_capabilities") == 0) {
 821                 return scnprintf(page, PAGE_SIZE, "0x%08x\n", grp->mccap);
 822         } else if (strcmp(attr->name, "max_phys_secs") == 0) {
 823                 return scnprintf(page, PAGE_SIZE, "%u\n",
 824                                 ndev->ops->max_phys_sect);
 825         } else {
 826                 return scnprintf(page,
 827                                  PAGE_SIZE,
 828                                  "Unhandled attr(%s) in `nvm_dev_attr_show`\n",
 829                                  attr->name);
 830         }
 831 }
 832
 833 #define NVM_DEV_ATTR_RO(_name)                                          \
 834         DEVICE_ATTR(_name, S_IRUGO, nvm_dev_attr_show, NULL)
 835
 836 static NVM_DEV_ATTR_RO(version);
 837 static NVM_DEV_ATTR_RO(vendor_opcode);
 838 static NVM_DEV_ATTR_RO(capabilities);
 839 static NVM_DEV_ATTR_RO(device_mode);
 840 static NVM_DEV_ATTR_RO(ppa_format);
 841 static NVM_DEV_ATTR_RO(media_manager);
 842
 843 static NVM_DEV_ATTR_RO(media_type);
 844 static NVM_DEV_ATTR_RO(flash_media_type);
 845 static NVM_DEV_ATTR_RO(num_channels);
 846 static NVM_DEV_ATTR_RO(num_luns);
 847 static NVM_DEV_ATTR_RO(num_planes);
 848 static NVM_DEV_ATTR_RO(num_blocks);
 849 static NVM_DEV_ATTR_RO(num_pages);
 850 static NVM_DEV_ATTR_RO(page_size);
 851 static NVM_DEV_ATTR_RO(hw_sector_size);
 852 static NVM_DEV_ATTR_RO(oob_sector_size);
 853 static NVM_DEV_ATTR_RO(read_typ);
 854 static NVM_DEV_ATTR_RO(read_max);
 855 static NVM_DEV_ATTR_RO(prog_typ);
 856 static NVM_DEV_ATTR_RO(prog_max);
 857 static NVM_DEV_ATTR_RO(erase_typ);
 858 static NVM_DEV_ATTR_RO(erase_max);
 859 static NVM_DEV_ATTR_RO(multiplane_modes);
 860 static NVM_DEV_ATTR_RO(media_capabilities);
 861 static NVM_DEV_ATTR_RO(max_phys_secs);
 862
 863 static struct attribute *nvm_dev_attrs[] = {
 864         &dev_attr_version.attr,
 865         &dev_attr_vendor_opcode.attr,
 866         &dev_attr_capabilities.attr,
 867         &dev_attr_device_mode.attr,
 868         &dev_attr_media_manager.attr,
 869
 870         &dev_attr_ppa_format.attr,
 871         &dev_attr_media_type.attr,
 872         &dev_attr_flash_media_type.attr,
 873         &dev_attr_num_channels.attr,
 874         &dev_attr_num_luns.attr,
 875         &dev_attr_num_planes.attr,
 876         &dev_attr_num_blocks.attr,
 877         &dev_attr_num_pages.attr,
 878         &dev_attr_page_size.attr,
 879         &dev_attr_hw_sector_size.attr,
 880         &dev_attr_oob_sector_size.attr,
 881         &dev_attr_read_typ.attr,
 882         &dev_attr_read_max.attr,
 883         &dev_attr_prog_typ.attr,
 884         &dev_attr_prog_max.attr,
 885         &dev_attr_erase_typ.attr,
 886         &dev_attr_erase_max.attr,
 887         &dev_attr_multiplane_modes.attr,
 888         &dev_attr_media_capabilities.attr,
 889         &dev_attr_max_phys_secs.attr,
 890         NULL,
 891 };
 892
 893 static const struct attribute_group nvm_dev_attr_group = {
 894         .name           = "lightnvm",
 895         .attrs          = nvm_dev_attrs,
 896 };
 897
 898 int nvme_nvm_register_sysfs(struct nvme_ns *ns)
 899 {
 900         return sysfs_create_group(&disk_to_dev(ns->disk)->kobj,
 901                                         &nvm_dev_attr_group);
 902 }
 903
 904 void nvme_nvm_unregister_sysfs(struct nvme_ns *ns)
 905 {
 906         sysfs_remove_group(&disk_to_dev(ns->disk)->kobj,
 907                                         &nvm_dev_attr_group);
 908 }