drivers/nvme/host/lightnvm.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * nvme-lightnvm.c - LightNVM NVMe device
   4  *
   5  * Copyright (C) 2014-2015 IT University of Copenhagen
   6  * Initial release: Matias Bjorling <[email protected]>
   7  */
   8
   9 #include "nvme.h"
  10
  11 #include <linux/nvme.h>
  12 #include <linux/bitops.h>
  13 #include <linux/lightnvm.h>
  14 #include <linux/vmalloc.h>
  15 #include <linux/sched/sysctl.h>
  16 #include <uapi/linux/lightnvm.h>
  17
  18 enum nvme_nvm_admin_opcode {
  19         nvme_nvm_admin_identity         = 0xe2,
  20         nvme_nvm_admin_get_bb_tbl       = 0xf2,
  21         nvme_nvm_admin_set_bb_tbl       = 0xf1,
  22 };
  23
  24 enum nvme_nvm_log_page {
  25         NVME_NVM_LOG_REPORT_CHUNK       = 0xca,
  26 };
  27
  28 struct nvme_nvm_ph_rw {
  29         __u8                    opcode;
  30         __u8                    flags;
  31         __u16                   command_id;
  32         __le32                  nsid;
  33         __u64                   rsvd2;
  34         __le64                  metadata;
  35         __le64                  prp1;
  36         __le64                  prp2;
  37         __le64                  spba;
  38         __le16                  length;
  39         __le16                  control;
  40         __le32                  dsmgmt;
  41         __le64                  resv;
  42 };
  43
  44 struct nvme_nvm_erase_blk {
  45         __u8                    opcode;
  46         __u8                    flags;
  47         __u16                   command_id;
  48         __le32                  nsid;
  49         __u64                   rsvd[2];
  50         __le64                  prp1;
  51         __le64                  prp2;
  52         __le64                  spba;
  53         __le16                  length;
  54         __le16                  control;
  55         __le32                  dsmgmt;
  56         __le64                  resv;
  57 };
  58
  59 struct nvme_nvm_identity {
  60         __u8                    opcode;
  61         __u8                    flags;
  62         __u16                   command_id;
  63         __le32                  nsid;
  64         __u64                   rsvd[2];
  65         __le64                  prp1;
  66         __le64                  prp2;
  67         __u32                   rsvd11[6];
  68 };
  69
  70 struct nvme_nvm_getbbtbl {
  71         __u8                    opcode;
  72         __u8                    flags;
  73         __u16                   command_id;
  74         __le32                  nsid;
  75         __u64                   rsvd[2];
  76         __le64                  prp1;
  77         __le64                  prp2;
  78         __le64                  spba;
  79         __u32                   rsvd4[4];
  80 };
  81
  82 struct nvme_nvm_setbbtbl {
  83         __u8                    opcode;
  84         __u8                    flags;
  85         __u16                   command_id;
  86         __le32                  nsid;
  87         __le64                  rsvd[2];
  88         __le64                  prp1;
  89         __le64                  prp2;
  90         __le64                  spba;
  91         __le16                  nlb;
  92         __u8                    value;
  93         __u8                    rsvd3;
  94         __u32                   rsvd4[3];
  95 };
  96
  97 struct nvme_nvm_command {
  98         union {
  99                 struct nvme_common_command common;
 100                 struct nvme_nvm_ph_rw ph_rw;
 101                 struct nvme_nvm_erase_blk erase;
 102                 struct nvme_nvm_identity identity;
 103                 struct nvme_nvm_getbbtbl get_bb;
 104                 struct nvme_nvm_setbbtbl set_bb;
 105         };
 106 };
 107
 108 struct nvme_nvm_id12_grp {
 109         __u8                    mtype;
 110         __u8                    fmtype;
 111         __le16                  res16;
 112         __u8                    num_ch;
 113         __u8                    num_lun;
 114         __u8                    num_pln;
 115         __u8                    rsvd1;
 116         __le16                  num_chk;
 117         __le16                  num_pg;
 118         __le16                  fpg_sz;
 119         __le16                  csecs;
 120         __le16                  sos;
 121         __le16                  rsvd2;
 122         __le32                  trdt;
 123         __le32                  trdm;
 124         __le32                  tprt;
 125         __le32                  tprm;
 126         __le32                  tbet;
 127         __le32                  tbem;
 128         __le32                  mpos;
 129         __le32                  mccap;
 130         __le16                  cpar;
 131         __u8                    reserved[906];
 132 } __packed;
 133
 134 struct nvme_nvm_id12_addrf {
 135         __u8                    ch_offset;
 136         __u8                    ch_len;
 137         __u8                    lun_offset;
 138         __u8                    lun_len;
 139         __u8                    pln_offset;
 140         __u8                    pln_len;
 141         __u8                    blk_offset;
 142         __u8                    blk_len;
 143         __u8                    pg_offset;
 144         __u8                    pg_len;
 145         __u8                    sec_offset;
 146         __u8                    sec_len;
 147         __u8                    res[4];
 148 } __packed;
 149
 150 struct nvme_nvm_id12 {
 151         __u8                    ver_id;
 152         __u8                    vmnt;
 153         __u8                    cgrps;
 154         __u8                    res;
 155         __le32                  cap;
 156         __le32                  dom;
 157         struct nvme_nvm_id12_addrf ppaf;
 158         __u8                    resv[228];
 159         struct nvme_nvm_id12_grp grp;
 160         __u8                    resv2[2880];
 161 } __packed;
 162
 163 struct nvme_nvm_bb_tbl {
 164         __u8    tblid[4];
 165         __le16  verid;
 166         __le16  revid;
 167         __le32  rvsd1;
 168         __le32  tblks;
 169         __le32  tfact;
 170         __le32  tgrown;
 171         __le32  tdresv;
 172         __le32  thresv;
 173         __le32  rsvd2[8];
 174         __u8    blk[0];
 175 };
 176
 177 struct nvme_nvm_id20_addrf {
 178         __u8                    grp_len;
 179         __u8                    pu_len;
 180         __u8                    chk_len;
 181         __u8                    lba_len;
 182         __u8                    resv[4];
 183 };
 184
 185 struct nvme_nvm_id20 {
 186         __u8                    mjr;
 187         __u8                    mnr;
 188         __u8                    resv[6];
 189
 190         struct nvme_nvm_id20_addrf lbaf;
 191
 192         __le32                  mccap;
 193         __u8                    resv2[12];
 194
 195         __u8                    wit;
 196         __u8                    resv3[31];
 197
 198         /* Geometry */
 199         __le16                  num_grp;
 200         __le16                  num_pu;
 201         __le32                  num_chk;
 202         __le32                  clba;
 203         __u8                    resv4[52];
 204
 205         /* Write data requirements */
 206         __le32                  ws_min;
 207         __le32                  ws_opt;
 208         __le32                  mw_cunits;
 209         __le32                  maxoc;
 210         __le32                  maxocpu;
 211         __u8                    resv5[44];
 212
 213         /* Performance related metrics */
 214         __le32                  trdt;
 215         __le32                  trdm;
 216         __le32                  twrt;
 217         __le32                  twrm;
 218         __le32                  tcrst;
 219         __le32                  tcrsm;
 220         __u8                    resv6[40];
 221
 222         /* Reserved area */
 223         __u8                    resv7[2816];
 224
 225         /* Vendor specific */
 226         __u8                    vs[1024];
 227 };
 228
 229 struct nvme_nvm_chk_meta {
 230         __u8    state;
 231         __u8    type;
 232         __u8    wi;
 233         __u8    rsvd[5];
 234         __le64  slba;
 235         __le64  cnlb;
 236         __le64  wp;
 237 };
 238
 239 /*
 240  * Check we didn't inadvertently grow the command struct
 241  */
 242 static inline void _nvme_nvm_check_size(void)
 243 {
 244         BUILD_BUG_ON(sizeof(struct nvme_nvm_identity) != 64);
 245         BUILD_BUG_ON(sizeof(struct nvme_nvm_ph_rw) != 64);
 246         BUILD_BUG_ON(sizeof(struct nvme_nvm_erase_blk) != 64);
 247         BUILD_BUG_ON(sizeof(struct nvme_nvm_getbbtbl) != 64);
 248         BUILD_BUG_ON(sizeof(struct nvme_nvm_setbbtbl) != 64);
 249         BUILD_BUG_ON(sizeof(struct nvme_nvm_id12_grp) != 960);
 250         BUILD_BUG_ON(sizeof(struct nvme_nvm_id12_addrf) != 16);
 251         BUILD_BUG_ON(sizeof(struct nvme_nvm_id12) != NVME_IDENTIFY_DATA_SIZE);
 252         BUILD_BUG_ON(sizeof(struct nvme_nvm_bb_tbl) != 64);
 253         BUILD_BUG_ON(sizeof(struct nvme_nvm_id20_addrf) != 8);
 254         BUILD_BUG_ON(sizeof(struct nvme_nvm_id20) != NVME_IDENTIFY_DATA_SIZE);
 255         BUILD_BUG_ON(sizeof(struct nvme_nvm_chk_meta) != 32);
 256         BUILD_BUG_ON(sizeof(struct nvme_nvm_chk_meta) !=
 257                                                 sizeof(struct nvm_chk_meta));
 258 }
 259
 260 static void nvme_nvm_set_addr_12(struct nvm_addrf_12 *dst,
 261                                  struct nvme_nvm_id12_addrf *src)
 262 {
 263         dst->ch_len = src->ch_len;
 264         dst->lun_len = src->lun_len;
 265         dst->blk_len = src->blk_len;
 266         dst->pg_len = src->pg_len;
 267         dst->pln_len = src->pln_len;
 268         dst->sec_len = src->sec_len;
 269
 270         dst->ch_offset = src->ch_offset;
 271         dst->lun_offset = src->lun_offset;
 272         dst->blk_offset = src->blk_offset;
 273         dst->pg_offset = src->pg_offset;
 274         dst->pln_offset = src->pln_offset;
 275         dst->sec_offset = src->sec_offset;
 276
 277         dst->ch_mask = ((1ULL << dst->ch_len) - 1) << dst->ch_offset;
 278         dst->lun_mask = ((1ULL << dst->lun_len) - 1) << dst->lun_offset;
 279         dst->blk_mask = ((1ULL << dst->blk_len) - 1) << dst->blk_offset;
 280         dst->pg_mask = ((1ULL << dst->pg_len) - 1) << dst->pg_offset;
 281         dst->pln_mask = ((1ULL << dst->pln_len) - 1) << dst->pln_offset;
 282         dst->sec_mask = ((1ULL << dst->sec_len) - 1) << dst->sec_offset;
 283 }
 284
 285 static int nvme_nvm_setup_12(struct nvme_nvm_id12 *id,
 286                              struct nvm_geo *geo)
 287 {
 288         struct nvme_nvm_id12_grp *src;
 289         int sec_per_pg, sec_per_pl, pg_per_blk;
 290
 291         if (id->cgrps != 1)
 292                 return -EINVAL;
 293
 294         src = &id->grp;
 295
 296         if (src->mtype != 0) {
 297                 pr_err("nvm: memory type not supported\n");
 298                 return -EINVAL;
 299         }
 300
 301         /* 1.2 spec. only reports a single version id - unfold */
 302         geo->major_ver_id = id->ver_id;
 303         geo->minor_ver_id = 2;
 304
 305         /* Set compacted version for upper layers */
 306         geo->version = NVM_OCSSD_SPEC_12;
 307
 308         geo->num_ch = src->num_ch;
 309         geo->num_lun = src->num_lun;
 310         geo->all_luns = geo->num_ch * geo->num_lun;
 311
 312         geo->num_chk = le16_to_cpu(src->num_chk);
 313
 314         geo->csecs = le16_to_cpu(src->csecs);
 315         geo->sos = le16_to_cpu(src->sos);
 316
 317         pg_per_blk = le16_to_cpu(src->num_pg);
 318         sec_per_pg = le16_to_cpu(src->fpg_sz) / geo->csecs;
 319         sec_per_pl = sec_per_pg * src->num_pln;
 320         geo->clba = sec_per_pl * pg_per_blk;
 321
 322         geo->all_chunks = geo->all_luns * geo->num_chk;
 323         geo->total_secs = geo->clba * geo->all_chunks;
 324
 325         geo->ws_min = sec_per_pg;
 326         geo->ws_opt = sec_per_pg;
 327         geo->mw_cunits = geo->ws_opt << 3;      /* default to MLC safe values */
 328
 329         /* Do not impose values for maximum number of open blocks as it is
 330          * unspecified in 1.2. Users of 1.2 must be aware of this and eventually
 331          * specify these values through a quirk if restrictions apply.
 332          */
 333         geo->maxoc = geo->all_luns * geo->num_chk;
 334         geo->maxocpu = geo->num_chk;
 335
 336         geo->mccap = le32_to_cpu(src->mccap);
 337
 338         geo->trdt = le32_to_cpu(src->trdt);
 339         geo->trdm = le32_to_cpu(src->trdm);
 340         geo->tprt = le32_to_cpu(src->tprt);
 341         geo->tprm = le32_to_cpu(src->tprm);
 342         geo->tbet = le32_to_cpu(src->tbet);
 343         geo->tbem = le32_to_cpu(src->tbem);
 344
 345         /* 1.2 compatibility */
 346         geo->vmnt = id->vmnt;
 347         geo->cap = le32_to_cpu(id->cap);
 348         geo->dom = le32_to_cpu(id->dom);
 349
 350         geo->mtype = src->mtype;
 351         geo->fmtype = src->fmtype;
 352
 353         geo->cpar = le16_to_cpu(src->cpar);
 354         geo->mpos = le32_to_cpu(src->mpos);
 355
 356         geo->pln_mode = NVM_PLANE_SINGLE;
 357
 358         if (geo->mpos & 0x020202) {
 359                 geo->pln_mode = NVM_PLANE_DOUBLE;
 360                 geo->ws_opt <<= 1;
 361         } else if (geo->mpos & 0x040404) {
 362                 geo->pln_mode = NVM_PLANE_QUAD;
 363                 geo->ws_opt <<= 2;
 364         }
 365
 366         geo->num_pln = src->num_pln;
 367         geo->num_pg = le16_to_cpu(src->num_pg);
 368         geo->fpg_sz = le16_to_cpu(src->fpg_sz);
 369
 370         nvme_nvm_set_addr_12((struct nvm_addrf_12 *)&geo->addrf, &id->ppaf);
 371
 372         return 0;
 373 }
 374
 375 static void nvme_nvm_set_addr_20(struct nvm_addrf *dst,
 376                                  struct nvme_nvm_id20_addrf *src)
 377 {
 378         dst->ch_len = src->grp_len;
 379         dst->lun_len = src->pu_len;
 380         dst->chk_len = src->chk_len;
 381         dst->sec_len = src->lba_len;
 382
 383         dst->sec_offset = 0;
 384         dst->chk_offset = dst->sec_len;
 385         dst->lun_offset = dst->chk_offset + dst->chk_len;
 386         dst->ch_offset = dst->lun_offset + dst->lun_len;
 387
 388         dst->ch_mask = ((1ULL << dst->ch_len) - 1) << dst->ch_offset;
 389         dst->lun_mask = ((1ULL << dst->lun_len) - 1) << dst->lun_offset;
 390         dst->chk_mask = ((1ULL << dst->chk_len) - 1) << dst->chk_offset;
 391         dst->sec_mask = ((1ULL << dst->sec_len) - 1) << dst->sec_offset;
 392 }
 393
 394 static int nvme_nvm_setup_20(struct nvme_nvm_id20 *id,
 395                              struct nvm_geo *geo)
 396 {
 397         geo->major_ver_id = id->mjr;
 398         geo->minor_ver_id = id->mnr;
 399
 400         /* Set compacted version for upper layers */
 401         geo->version = NVM_OCSSD_SPEC_20;
 402
 403         geo->num_ch = le16_to_cpu(id->num_grp);
 404         geo->num_lun = le16_to_cpu(id->num_pu);
 405         geo->all_luns = geo->num_ch * geo->num_lun;
 406
 407         geo->num_chk = le32_to_cpu(id->num_chk);
 408         geo->clba = le32_to_cpu(id->clba);
 409
 410         geo->all_chunks = geo->all_luns * geo->num_chk;
 411         geo->total_secs = geo->clba * geo->all_chunks;
 412
 413         geo->ws_min = le32_to_cpu(id->ws_min);
 414         geo->ws_opt = le32_to_cpu(id->ws_opt);
 415         geo->mw_cunits = le32_to_cpu(id->mw_cunits);
 416         geo->maxoc = le32_to_cpu(id->maxoc);
 417         geo->maxocpu = le32_to_cpu(id->maxocpu);
 418
 419         geo->trdt = le32_to_cpu(id->trdt);
 420         geo->trdm = le32_to_cpu(id->trdm);
 421         geo->tprt = le32_to_cpu(id->twrt);
 422         geo->tprm = le32_to_cpu(id->twrm);
 423         geo->tbet = le32_to_cpu(id->tcrst);
 424         geo->tbem = le32_to_cpu(id->tcrsm);
 425
 426         nvme_nvm_set_addr_20(&geo->addrf, &id->lbaf);
 427
 428         return 0;
 429 }
 430
 431 static int nvme_nvm_identity(struct nvm_dev *nvmdev)
 432 {
 433         struct nvme_ns *ns = nvmdev->q->queuedata;
 434         struct nvme_nvm_id12 *id;
 435         struct nvme_nvm_command c = {};
 436         int ret;
 437
 438         c.identity.opcode = nvme_nvm_admin_identity;
 439         c.identity.nsid = cpu_to_le32(ns->head->ns_id);
 440
 441         id = kmalloc(sizeof(struct nvme_nvm_id12), GFP_KERNEL);
 442         if (!id)
 443                 return -ENOMEM;
 444
 445         ret = nvme_submit_sync_cmd(ns->ctrl->admin_q, (struct nvme_command *)&c,
 446                                 id, sizeof(struct nvme_nvm_id12));
 447         if (ret) {
 448                 ret = -EIO;
 449                 goto out;
 450         }
 451
 452         /*
 453          * The 1.2 and 2.0 specifications share the first byte in their geometry
 454          * command to make it possible to know what version a device implements.
 455          */
 456         switch (id->ver_id) {
 457         case 1:
 458                 ret = nvme_nvm_setup_12(id, &nvmdev->geo);
 459                 break;
 460         case 2:
 461                 ret = nvme_nvm_setup_20((struct nvme_nvm_id20 *)id,
 462                                                         &nvmdev->geo);
 463                 break;
 464         default:
 465                 dev_err(ns->ctrl->device, "OCSSD revision not supported (%d)\n",
 466                                                         id->ver_id);
 467                 ret = -EINVAL;
 468         }
 469
 470 out:
 471         kfree(id);
 472         return ret;
 473 }
 474
 475 static int nvme_nvm_get_bb_tbl(struct nvm_dev *nvmdev, struct ppa_addr ppa,
 476                                                                 u8 *blks)
 477 {
 478         struct request_queue *q = nvmdev->q;
 479         struct nvm_geo *geo = &nvmdev->geo;
 480         struct nvme_ns *ns = q->queuedata;
 481         struct nvme_ctrl *ctrl = ns->ctrl;
 482         struct nvme_nvm_command c = {};
 483         struct nvme_nvm_bb_tbl *bb_tbl;
 484         int nr_blks = geo->num_chk * geo->num_pln;
 485         int tblsz = sizeof(struct nvme_nvm_bb_tbl) + nr_blks;
 486         int ret = 0;
 487
 488         c.get_bb.opcode = nvme_nvm_admin_get_bb_tbl;
 489         c.get_bb.nsid = cpu_to_le32(ns->head->ns_id);
 490         c.get_bb.spba = cpu_to_le64(ppa.ppa);
 491
 492         bb_tbl = kzalloc(tblsz, GFP_KERNEL);
 493         if (!bb_tbl)
 494                 return -ENOMEM;
 495
 496         ret = nvme_submit_sync_cmd(ctrl->admin_q, (struct nvme_command *)&c,
 497                                                                 bb_tbl, tblsz);
 498         if (ret) {
 499                 dev_err(ctrl->device, "get bad block table failed (%d)\n", ret);
 500                 ret = -EIO;
 501                 goto out;
 502         }
 503
 504         if (bb_tbl->tblid[0] != 'B' || bb_tbl->tblid[1] != 'B' ||
 505                 bb_tbl->tblid[2] != 'L' || bb_tbl->tblid[3] != 'T') {
 506                 dev_err(ctrl->device, "bbt format mismatch\n");
 507                 ret = -EINVAL;
 508                 goto out;
 509         }
 510
 511         if (le16_to_cpu(bb_tbl->verid) != 1) {
 512                 ret = -EINVAL;
 513                 dev_err(ctrl->device, "bbt version not supported\n");
 514                 goto out;
 515         }
 516
 517         if (le32_to_cpu(bb_tbl->tblks) != nr_blks) {
 518                 ret = -EINVAL;
 519                 dev_err(ctrl->device,
 520                                 "bbt unsuspected blocks returned (%u!=%u)",
 521                                 le32_to_cpu(bb_tbl->tblks), nr_blks);
 522                 goto out;
 523         }
 524
 525         memcpy(blks, bb_tbl->blk, geo->num_chk * geo->num_pln);
 526 out:
 527         kfree(bb_tbl);
 528         return ret;
 529 }
 530
 531 static int nvme_nvm_set_bb_tbl(struct nvm_dev *nvmdev, struct ppa_addr *ppas,
 532                                                         int nr_ppas, int type)
 533 {
 534         struct nvme_ns *ns = nvmdev->q->queuedata;
 535         struct nvme_nvm_command c = {};
 536         int ret = 0;
 537
 538         c.set_bb.opcode = nvme_nvm_admin_set_bb_tbl;
 539         c.set_bb.nsid = cpu_to_le32(ns->head->ns_id);
 540         c.set_bb.spba = cpu_to_le64(ppas->ppa);
 541         c.set_bb.nlb = cpu_to_le16(nr_ppas - 1);
 542         c.set_bb.value = type;
 543
 544         ret = nvme_submit_sync_cmd(ns->ctrl->admin_q, (struct nvme_command *)&c,
 545                                                                 NULL, 0);
 546         if (ret)
 547                 dev_err(ns->ctrl->device, "set bad block table failed (%d)\n",
 548                                                                         ret);
 549         return ret;
 550 }
 551
 552 /*
 553  * Expect the lba in device format
 554  */
 555 static int nvme_nvm_get_chk_meta(struct nvm_dev *ndev,
 556                                  sector_t slba, int nchks,
 557                                  struct nvm_chk_meta *meta)
 558 {
 559         struct nvm_geo *geo = &ndev->geo;
 560         struct nvme_ns *ns = ndev->q->queuedata;
 561         struct nvme_ctrl *ctrl = ns->ctrl;
 562         struct nvme_nvm_chk_meta *dev_meta, *dev_meta_off;
 563         struct ppa_addr ppa;
 564         size_t left = nchks * sizeof(struct nvme_nvm_chk_meta);
 565         size_t log_pos, offset, len;
 566         int i, max_len;
 567         int ret = 0;
 568
 569         /*
 570          * limit requests to maximum 256K to avoid issuing arbitrary large
 571          * requests when the device does not specific a maximum transfer size.
 572          */
 573         max_len = min_t(unsigned int, ctrl->max_hw_sectors << 9, 256 * 1024);
 574
 575         dev_meta = kmalloc(max_len, GFP_KERNEL);
 576         if (!dev_meta)
 577                 return -ENOMEM;
 578
 579         /* Normalize lba address space to obtain log offset */
 580         ppa.ppa = slba;
 581         ppa = dev_to_generic_addr(ndev, ppa);
 582
 583         log_pos = ppa.m.chk;
 584         log_pos += ppa.m.pu * geo->num_chk;
 585         log_pos += ppa.m.grp * geo->num_lun * geo->num_chk;
 586
 587         offset = log_pos * sizeof(struct nvme_nvm_chk_meta);
 588
 589         while (left) {
 590                 len = min_t(unsigned int, left, max_len);
 591
 592                 memset(dev_meta, 0, max_len);
 593                 dev_meta_off = dev_meta;
 594
 595                 ret = nvme_get_log(ctrl, ns->head->ns_id,
 596                                 NVME_NVM_LOG_REPORT_CHUNK, 0, dev_meta, len,
 597                                 offset);
 598                 if (ret) {
 599                         dev_err(ctrl->device, "Get REPORT CHUNK log error\n");
 600                         break;
 601                 }
 602
 603                 for (i = 0; i < len; i += sizeof(struct nvme_nvm_chk_meta)) {
 604                         meta->state = dev_meta_off->state;
 605                         meta->type = dev_meta_off->type;
 606                         meta->wi = dev_meta_off->wi;
 607                         meta->slba = le64_to_cpu(dev_meta_off->slba);
 608                         meta->cnlb = le64_to_cpu(dev_meta_off->cnlb);
 609                         meta->wp = le64_to_cpu(dev_meta_off->wp);
 610
 611                         meta++;
 612                         dev_meta_off++;
 613                 }
 614
 615                 offset += len;
 616                 left -= len;
 617         }
 618
 619         kfree(dev_meta);
 620
 621         return ret;
 622 }
 623
 624 static inline void nvme_nvm_rqtocmd(struct nvm_rq *rqd, struct nvme_ns *ns,
 625                                     struct nvme_nvm_command *c)
 626 {
 627         c->ph_rw.opcode = rqd->opcode;
 628         c->ph_rw.nsid = cpu_to_le32(ns->head->ns_id);
 629         c->ph_rw.spba = cpu_to_le64(rqd->ppa_addr.ppa);
 630         c->ph_rw.metadata = cpu_to_le64(rqd->dma_meta_list);
 631         c->ph_rw.control = cpu_to_le16(rqd->flags);
 632         c->ph_rw.length = cpu_to_le16(rqd->nr_ppas - 1);
 633 }
 634
 635 static void nvme_nvm_end_io(struct request *rq, blk_status_t status)
 636 {
 637         struct nvm_rq *rqd = rq->end_io_data;
 638
 639         rqd->ppa_status = le64_to_cpu(nvme_req(rq)->result.u64);
 640         rqd->error = nvme_req(rq)->status;
 641         nvm_end_io(rqd);
 642
 643         kfree(nvme_req(rq)->cmd);
 644         blk_mq_free_request(rq);
 645 }
 646
 647 static struct request *nvme_nvm_alloc_request(struct request_queue *q,
 648                                               struct nvm_rq *rqd,
 649                                               struct nvme_nvm_command *cmd)
 650 {
 651         struct nvme_ns *ns = q->queuedata;
 652         struct request *rq;
 653
 654         nvme_nvm_rqtocmd(rqd, ns, cmd);
 655
 656         rq = nvme_alloc_request(q, (struct nvme_command *)cmd, 0, NVME_QID_ANY);
 657         if (IS_ERR(rq))
 658                 return rq;
 659
 660         rq->cmd_flags &= ~REQ_FAILFAST_DRIVER;
 661
 662         if (rqd->bio)
 663                 blk_init_request_from_bio(rq, rqd->bio);
 664         else
 665                 rq->ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, IOPRIO_NORM);
 666
 667         return rq;
 668 }
 669
 670 static int nvme_nvm_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
 671 {
 672         struct request_queue *q = dev->q;
 673         struct nvme_nvm_command *cmd;
 674         struct request *rq;
 675
 676         cmd = kzalloc(sizeof(struct nvme_nvm_command), GFP_KERNEL);
 677         if (!cmd)
 678                 return -ENOMEM;
 679
 680         rq = nvme_nvm_alloc_request(q, rqd, cmd);
 681         if (IS_ERR(rq)) {
 682                 kfree(cmd);
 683                 return PTR_ERR(rq);
 684         }
 685
 686         rq->end_io_data = rqd;
 687
 688         blk_execute_rq_nowait(q, NULL, rq, 0, nvme_nvm_end_io);
 689
 690         return 0;
 691 }
 692
 693 static int nvme_nvm_submit_io_sync(struct nvm_dev *dev, struct nvm_rq *rqd)
 694 {
 695         struct request_queue *q = dev->q;
 696         struct request *rq;
 697         struct nvme_nvm_command cmd;
 698         int ret = 0;
 699
 700         memset(&cmd, 0, sizeof(struct nvme_nvm_command));
 701
 702         rq = nvme_nvm_alloc_request(q, rqd, &cmd);
 703         if (IS_ERR(rq))
 704                 return PTR_ERR(rq);
 705
 706         /* I/Os can fail and the error is signaled through rqd. Callers must
 707          * handle the error accordingly.
 708          */
 709         blk_execute_rq(q, NULL, rq, 0);
 710         if (nvme_req(rq)->flags & NVME_REQ_CANCELLED)
 711                 ret = -EINTR;
 712
 713         rqd->ppa_status = le64_to_cpu(nvme_req(rq)->result.u64);
 714         rqd->error = nvme_req(rq)->status;
 715
 716         blk_mq_free_request(rq);
 717
 718         return ret;
 719 }
 720
 721 static void *nvme_nvm_create_dma_pool(struct nvm_dev *nvmdev, char *name,
 722                                         int size)
 723 {
 724         struct nvme_ns *ns = nvmdev->q->queuedata;
 725
 726         return dma_pool_create(name, ns->ctrl->dev, size, PAGE_SIZE, 0);
 727 }
 728
 729 static void nvme_nvm_destroy_dma_pool(void *pool)
 730 {
 731         struct dma_pool *dma_pool = pool;
 732
 733         dma_pool_destroy(dma_pool);
 734 }
 735
 736 static void *nvme_nvm_dev_dma_alloc(struct nvm_dev *dev, void *pool,
 737                                     gfp_t mem_flags, dma_addr_t *dma_handler)
 738 {
 739         return dma_pool_alloc(pool, mem_flags, dma_handler);
 740 }
 741
 742 static void nvme_nvm_dev_dma_free(void *pool, void *addr,
 743                                                         dma_addr_t dma_handler)
 744 {
 745         dma_pool_free(pool, addr, dma_handler);
 746 }
 747
 748 static struct nvm_dev_ops nvme_nvm_dev_ops = {
 749         .identity               = nvme_nvm_identity,
 750
 751         .get_bb_tbl             = nvme_nvm_get_bb_tbl,
 752         .set_bb_tbl             = nvme_nvm_set_bb_tbl,
 753
 754         .get_chk_meta           = nvme_nvm_get_chk_meta,
 755
 756         .submit_io              = nvme_nvm_submit_io,
 757         .submit_io_sync         = nvme_nvm_submit_io_sync,
 758
 759         .create_dma_pool        = nvme_nvm_create_dma_pool,
 760         .destroy_dma_pool       = nvme_nvm_destroy_dma_pool,
 761         .dev_dma_alloc          = nvme_nvm_dev_dma_alloc,
 762         .dev_dma_free           = nvme_nvm_dev_dma_free,
 763 };
 764
 765 static int nvme_nvm_submit_user_cmd(struct request_queue *q,
 766                                 struct nvme_ns *ns,
 767                                 struct nvme_nvm_command *vcmd,
 768                                 void __user *ubuf, unsigned int bufflen,
 769                                 void __user *meta_buf, unsigned int meta_len,
 770                                 void __user *ppa_buf, unsigned int ppa_len,
 771                                 u32 *result, u64 *status, unsigned int timeout)
 772 {
 773         bool write = nvme_is_write((struct nvme_command *)vcmd);
 774         struct nvm_dev *dev = ns->ndev;
 775         struct gendisk *disk = ns->disk;
 776         struct request *rq;
 777         struct bio *bio = NULL;
 778         __le64 *ppa_list = NULL;
 779         dma_addr_t ppa_dma;
 780         __le64 *metadata = NULL;
 781         dma_addr_t metadata_dma;
 782         DECLARE_COMPLETION_ONSTACK(wait);
 783         int ret = 0;
 784
 785         rq = nvme_alloc_request(q, (struct nvme_command *)vcmd, 0,
 786                         NVME_QID_ANY);
 787         if (IS_ERR(rq)) {
 788                 ret = -ENOMEM;
 789                 goto err_cmd;
 790         }
 791
 792         rq->timeout = timeout ? timeout : ADMIN_TIMEOUT;
 793
 794         if (ppa_buf && ppa_len) {
 795                 ppa_list = dma_pool_alloc(dev->dma_pool, GFP_KERNEL, &ppa_dma);
 796                 if (!ppa_list) {
 797                         ret = -ENOMEM;
 798                         goto err_rq;
 799                 }
 800                 if (copy_from_user(ppa_list, (void __user *)ppa_buf,
 801                                                 sizeof(u64) * (ppa_len + 1))) {
 802                         ret = -EFAULT;
 803                         goto err_ppa;
 804                 }
 805                 vcmd->ph_rw.spba = cpu_to_le64(ppa_dma);
 806         } else {
 807                 vcmd->ph_rw.spba = cpu_to_le64((uintptr_t)ppa_buf);
 808         }
 809
 810         if (ubuf && bufflen) {
 811                 ret = blk_rq_map_user(q, rq, NULL, ubuf, bufflen, GFP_KERNEL);
 812                 if (ret)
 813                         goto err_ppa;
 814                 bio = rq->bio;
 815
 816                 if (meta_buf && meta_len) {
 817                         metadata = dma_pool_alloc(dev->dma_pool, GFP_KERNEL,
 818                                                                 &metadata_dma);
 819                         if (!metadata) {
 820                                 ret = -ENOMEM;
 821                                 goto err_map;
 822                         }
 823
 824                         if (write) {
 825                                 if (copy_from_user(metadata,
 826                                                 (void __user *)meta_buf,
 827                                                 meta_len)) {
 828                                         ret = -EFAULT;
 829                                         goto err_meta;
 830                                 }
 831                         }
 832                         vcmd->ph_rw.metadata = cpu_to_le64(metadata_dma);
 833                 }
 834
 835                 bio->bi_disk = disk;
 836         }
 837
 838         blk_execute_rq(q, NULL, rq, 0);
 839
 840         if (nvme_req(rq)->flags & NVME_REQ_CANCELLED)
 841                 ret = -EINTR;
 842         else if (nvme_req(rq)->status & 0x7ff)
 843                 ret = -EIO;
 844         if (result)
 845                 *result = nvme_req(rq)->status & 0x7ff;
 846         if (status)
 847                 *status = le64_to_cpu(nvme_req(rq)->result.u64);
 848
 849         if (metadata && !ret && !write) {
 850                 if (copy_to_user(meta_buf, (void *)metadata, meta_len))
 851                         ret = -EFAULT;
 852         }
 853 err_meta:
 854         if (meta_buf && meta_len)
 855                 dma_pool_free(dev->dma_pool, metadata, metadata_dma);
 856 err_map:
 857         if (bio)
 858                 blk_rq_unmap_user(bio);
 859 err_ppa:
 860         if (ppa_buf && ppa_len)
 861                 dma_pool_free(dev->dma_pool, ppa_list, ppa_dma);
 862 err_rq:
 863         blk_mq_free_request(rq);
 864 err_cmd:
 865         return ret;
 866 }
 867
 868 static int nvme_nvm_submit_vio(struct nvme_ns *ns,
 869                                         struct nvm_user_vio __user *uvio)
 870 {
 871         struct nvm_user_vio vio;
 872         struct nvme_nvm_command c;
 873         unsigned int length;
 874         int ret;
 875
 876         if (copy_from_user(&vio, uvio, sizeof(vio)))
 877                 return -EFAULT;
 878         if (vio.flags)
 879                 return -EINVAL;
 880
 881         memset(&c, 0, sizeof(c));
 882         c.ph_rw.opcode = vio.opcode;
 883         c.ph_rw.nsid = cpu_to_le32(ns->head->ns_id);
 884         c.ph_rw.control = cpu_to_le16(vio.control);
 885         c.ph_rw.length = cpu_to_le16(vio.nppas);
 886
 887         length = (vio.nppas + 1) << ns->lba_shift;
 888
 889         ret = nvme_nvm_submit_user_cmd(ns->queue, ns, &c,
 890                         (void __user *)(uintptr_t)vio.addr, length,
 891                         (void __user *)(uintptr_t)vio.metadata,
 892                                                         vio.metadata_len,
 893                         (void __user *)(uintptr_t)vio.ppa_list, vio.nppas,
 894                         &vio.result, &vio.status, 0);
 895
 896         if (ret && copy_to_user(uvio, &vio, sizeof(vio)))
 897                 return -EFAULT;
 898
 899         return ret;
 900 }
 901
 902 static int nvme_nvm_user_vcmd(struct nvme_ns *ns, int admin,
 903                                         struct nvm_passthru_vio __user *uvcmd)
 904 {
 905         struct nvm_passthru_vio vcmd;
 906         struct nvme_nvm_command c;
 907         struct request_queue *q;
 908         unsigned int timeout = 0;
 909         int ret;
 910
 911         if (copy_from_user(&vcmd, uvcmd, sizeof(vcmd)))
 912                 return -EFAULT;
 913         if ((vcmd.opcode != 0xF2) && (!capable(CAP_SYS_ADMIN)))
 914                 return -EACCES;
 915         if (vcmd.flags)
 916                 return -EINVAL;
 917
 918         memset(&c, 0, sizeof(c));
 919         c.common.opcode = vcmd.opcode;
 920         c.common.nsid = cpu_to_le32(ns->head->ns_id);
 921         c.common.cdw2[0] = cpu_to_le32(vcmd.cdw2);
 922         c.common.cdw2[1] = cpu_to_le32(vcmd.cdw3);
 923         /* cdw11-12 */
 924         c.ph_rw.length = cpu_to_le16(vcmd.nppas);
 925         c.ph_rw.control  = cpu_to_le16(vcmd.control);
 926         c.common.cdw13 = cpu_to_le32(vcmd.cdw13);
 927         c.common.cdw14 = cpu_to_le32(vcmd.cdw14);
 928         c.common.cdw15 = cpu_to_le32(vcmd.cdw15);
 929
 930         if (vcmd.timeout_ms)
 931                 timeout = msecs_to_jiffies(vcmd.timeout_ms);
 932
 933         q = admin ? ns->ctrl->admin_q : ns->queue;
 934
 935         ret = nvme_nvm_submit_user_cmd(q, ns,
 936                         (struct nvme_nvm_command *)&c,
 937                         (void __user *)(uintptr_t)vcmd.addr, vcmd.data_len,
 938                         (void __user *)(uintptr_t)vcmd.metadata,
 939                                                         vcmd.metadata_len,
 940                         (void __user *)(uintptr_t)vcmd.ppa_list, vcmd.nppas,
 941                         &vcmd.result, &vcmd.status, timeout);
 942
 943         if (ret && copy_to_user(uvcmd, &vcmd, sizeof(vcmd)))
 944                 return -EFAULT;
 945
 946         return ret;
 947 }
 948
 949 int nvme_nvm_ioctl(struct nvme_ns *ns, unsigned int cmd, unsigned long arg)
 950 {
 951         switch (cmd) {
 952         case NVME_NVM_IOCTL_ADMIN_VIO:
 953                 return nvme_nvm_user_vcmd(ns, 1, (void __user *)arg);
 954         case NVME_NVM_IOCTL_IO_VIO:
 955                 return nvme_nvm_user_vcmd(ns, 0, (void __user *)arg);
 956         case NVME_NVM_IOCTL_SUBMIT_VIO:
 957                 return nvme_nvm_submit_vio(ns, (void __user *)arg);
 958         default:
 959                 return -ENOTTY;
 960         }
 961 }
 962
 963 int nvme_nvm_register(struct nvme_ns *ns, char *disk_name, int node)
 964 {
 965         struct request_queue *q = ns->queue;
 966         struct nvm_dev *dev;
 967         struct nvm_geo *geo;
 968
 969         _nvme_nvm_check_size();
 970
 971         dev = nvm_alloc_dev(node);
 972         if (!dev)
 973                 return -ENOMEM;
 974
 975         /* Note that csecs and sos will be overridden if it is a 1.2 drive. */
 976         geo = &dev->geo;
 977         geo->csecs = 1 << ns->lba_shift;
 978         geo->sos = ns->ms;
 979         geo->ext = ns->ext;
 980
 981         dev->q = q;
 982         memcpy(dev->name, disk_name, DISK_NAME_LEN);
 983         dev->ops = &nvme_nvm_dev_ops;
 984         dev->private_data = ns;
 985         ns->ndev = dev;
 986
 987         return nvm_register(dev);
 988 }
 989
 990 void nvme_nvm_unregister(struct nvme_ns *ns)
 991 {
 992         nvm_unregister(ns->ndev);
 993 }
 994
 995 static ssize_t nvm_dev_attr_show(struct device *dev,
 996                 struct device_attribute *dattr, char *page)
 997 {
 998         struct nvme_ns *ns = nvme_get_ns_from_dev(dev);
 999         struct nvm_dev *ndev = ns->ndev;
1000         struct nvm_geo *geo = &ndev->geo;
1001         struct attribute *attr;
1002
1003         if (!ndev)
1004                 return 0;
1005
1006         attr = &dattr->attr;
1007
1008         if (strcmp(attr->name, "version") == 0) {
1009                 if (geo->major_ver_id == 1)
1010                         return scnprintf(page, PAGE_SIZE, "%u\n",
1011                                                 geo->major_ver_id);
1012                 else
1013                         return scnprintf(page, PAGE_SIZE, "%u.%u\n",
1014                                                 geo->major_ver_id,
1015                                                 geo->minor_ver_id);
1016         } else if (strcmp(attr->name, "capabilities") == 0) {
1017                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->cap);
1018         } else if (strcmp(attr->name, "read_typ") == 0) {
1019                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->trdt);
1020         } else if (strcmp(attr->name, "read_max") == 0) {
1021                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->trdm);
1022         } else {
1023                 return scnprintf(page,
1024                                  PAGE_SIZE,
1025                                  "Unhandled attr(%s) in `%s`\n",
1026                                  attr->name, __func__);
1027         }
1028 }
1029
1030 static ssize_t nvm_dev_attr_show_ppaf(struct nvm_addrf_12 *ppaf, char *page)
1031 {
1032         return scnprintf(page, PAGE_SIZE,
1033                 "0x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x\n",
1034                                 ppaf->ch_offset, ppaf->ch_len,
1035                                 ppaf->lun_offset, ppaf->lun_len,
1036                                 ppaf->pln_offset, ppaf->pln_len,
1037                                 ppaf->blk_offset, ppaf->blk_len,
1038                                 ppaf->pg_offset, ppaf->pg_len,
1039                                 ppaf->sec_offset, ppaf->sec_len);
1040 }
1041
1042 static ssize_t nvm_dev_attr_show_12(struct device *dev,
1043                 struct device_attribute *dattr, char *page)
1044 {
1045         struct nvme_ns *ns = nvme_get_ns_from_dev(dev);
1046         struct nvm_dev *ndev = ns->ndev;
1047         struct nvm_geo *geo = &ndev->geo;
1048         struct attribute *attr;
1049
1050         if (!ndev)
1051                 return 0;
1052
1053         attr = &dattr->attr;
1054
1055         if (strcmp(attr->name, "vendor_opcode") == 0) {
1056                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->vmnt);
1057         } else if (strcmp(attr->name, "device_mode") == 0) {
1058                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->dom);
1059         /* kept for compatibility */
1060         } else if (strcmp(attr->name, "media_manager") == 0) {
1061                 return scnprintf(page, PAGE_SIZE, "%s\n", "gennvm");
1062         } else if (strcmp(attr->name, "ppa_format") == 0) {
1063                 return nvm_dev_attr_show_ppaf((void *)&geo->addrf, page);
1064         } else if (strcmp(attr->name, "media_type") == 0) {     /* u8 */
1065                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->mtype);
1066         } else if (strcmp(attr->name, "flash_media_type") == 0) {
1067                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->fmtype);
1068         } else if (strcmp(attr->name, "num_channels") == 0) {
1069                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->num_ch);
1070         } else if (strcmp(attr->name, "num_luns") == 0) {
1071                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->num_lun);
1072         } else if (strcmp(attr->name, "num_planes") == 0) {
1073                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->num_pln);
1074         } else if (strcmp(attr->name, "num_blocks") == 0) {     /* u16 */
1075                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->num_chk);
1076         } else if (strcmp(attr->name, "num_pages") == 0) {
1077                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->num_pg);
1078         } else if (strcmp(attr->name, "page_size") == 0) {
1079                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->fpg_sz);
1080         } else if (strcmp(attr->name, "hw_sector_size") == 0) {
1081                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->csecs);
1082         } else if (strcmp(attr->name, "oob_sector_size") == 0) {/* u32 */
1083                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->sos);
1084         } else if (strcmp(attr->name, "prog_typ") == 0) {
1085                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->tprt);
1086         } else if (strcmp(attr->name, "prog_max") == 0) {
1087                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->tprm);
1088         } else if (strcmp(attr->name, "erase_typ") == 0) {
1089                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->tbet);
1090         } else if (strcmp(attr->name, "erase_max") == 0) {
1091                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->tbem);
1092         } else if (strcmp(attr->name, "multiplane_modes") == 0) {
1093                 return scnprintf(page, PAGE_SIZE, "0x%08x\n", geo->mpos);
1094         } else if (strcmp(attr->name, "media_capabilities") == 0) {
1095                 return scnprintf(page, PAGE_SIZE, "0x%08x\n", geo->mccap);
1096         } else if (strcmp(attr->name, "max_phys_secs") == 0) {
1097                 return scnprintf(page, PAGE_SIZE, "%u\n", NVM_MAX_VLBA);
1098         } else {
1099                 return scnprintf(page, PAGE_SIZE,
1100                         "Unhandled attr(%s) in `%s`\n",
1101                         attr->name, __func__);
1102         }
1103 }
1104
1105 static ssize_t nvm_dev_attr_show_20(struct device *dev,
1106                 struct device_attribute *dattr, char *page)
1107 {
1108         struct nvme_ns *ns = nvme_get_ns_from_dev(dev);
1109         struct nvm_dev *ndev = ns->ndev;
1110         struct nvm_geo *geo = &ndev->geo;
1111         struct attribute *attr;
1112
1113         if (!ndev)
1114                 return 0;
1115
1116         attr = &dattr->attr;
1117
1118         if (strcmp(attr->name, "groups") == 0) {
1119                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->num_ch);
1120         } else if (strcmp(attr->name, "punits") == 0) {
1121                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->num_lun);
1122         } else if (strcmp(attr->name, "chunks") == 0) {
1123                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->num_chk);
1124         } else if (strcmp(attr->name, "clba") == 0) {
1125                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->clba);
1126         } else if (strcmp(attr->name, "ws_min") == 0) {
1127                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->ws_min);
1128         } else if (strcmp(attr->name, "ws_opt") == 0) {
1129                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->ws_opt);
1130         } else if (strcmp(attr->name, "maxoc") == 0) {
1131                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->maxoc);
1132         } else if (strcmp(attr->name, "maxocpu") == 0) {
1133                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->maxocpu);
1134         } else if (strcmp(attr->name, "mw_cunits") == 0) {
1135                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->mw_cunits);
1136         } else if (strcmp(attr->name, "write_typ") == 0) {
1137                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->tprt);
1138         } else if (strcmp(attr->name, "write_max") == 0) {
1139                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->tprm);
1140         } else if (strcmp(attr->name, "reset_typ") == 0) {
1141                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->tbet);
1142         } else if (strcmp(attr->name, "reset_max") == 0) {
1143                 return scnprintf(page, PAGE_SIZE, "%u\n", geo->tbem);
1144         } else {
1145                 return scnprintf(page, PAGE_SIZE,
1146                         "Unhandled attr(%s) in `%s`\n",
1147                         attr->name, __func__);
1148         }
1149 }
1150
1151 #define NVM_DEV_ATTR_RO(_name)                                  \
1152         DEVICE_ATTR(_name, S_IRUGO, nvm_dev_attr_show, NULL)
1153 #define NVM_DEV_ATTR_12_RO(_name)                                       \
1154         DEVICE_ATTR(_name, S_IRUGO, nvm_dev_attr_show_12, NULL)
1155 #define NVM_DEV_ATTR_20_RO(_name)                                       \
1156         DEVICE_ATTR(_name, S_IRUGO, nvm_dev_attr_show_20, NULL)
1157
1158 /* general attributes */
1159 static NVM_DEV_ATTR_RO(version);
1160 static NVM_DEV_ATTR_RO(capabilities);
1161
1162 static NVM_DEV_ATTR_RO(read_typ);
1163 static NVM_DEV_ATTR_RO(read_max);
1164
1165 /* 1.2 values */
1166 static NVM_DEV_ATTR_12_RO(vendor_opcode);
1167 static NVM_DEV_ATTR_12_RO(device_mode);
1168 static NVM_DEV_ATTR_12_RO(ppa_format);
1169 static NVM_DEV_ATTR_12_RO(media_manager);
1170 static NVM_DEV_ATTR_12_RO(media_type);
1171 static NVM_DEV_ATTR_12_RO(flash_media_type);
1172 static NVM_DEV_ATTR_12_RO(num_channels);
1173 static NVM_DEV_ATTR_12_RO(num_luns);
1174 static NVM_DEV_ATTR_12_RO(num_planes);
1175 static NVM_DEV_ATTR_12_RO(num_blocks);
1176 static NVM_DEV_ATTR_12_RO(num_pages);
1177 static NVM_DEV_ATTR_12_RO(page_size);
1178 static NVM_DEV_ATTR_12_RO(hw_sector_size);
1179 static NVM_DEV_ATTR_12_RO(oob_sector_size);
1180 static NVM_DEV_ATTR_12_RO(prog_typ);
1181 static NVM_DEV_ATTR_12_RO(prog_max);
1182 static NVM_DEV_ATTR_12_RO(erase_typ);
1183 static NVM_DEV_ATTR_12_RO(erase_max);
1184 static NVM_DEV_ATTR_12_RO(multiplane_modes);
1185 static NVM_DEV_ATTR_12_RO(media_capabilities);
1186 static NVM_DEV_ATTR_12_RO(max_phys_secs);
1187
1188 /* 2.0 values */
1189 static NVM_DEV_ATTR_20_RO(groups);
1190 static NVM_DEV_ATTR_20_RO(punits);
1191 static NVM_DEV_ATTR_20_RO(chunks);
1192 static NVM_DEV_ATTR_20_RO(clba);
1193 static NVM_DEV_ATTR_20_RO(ws_min);
1194 static NVM_DEV_ATTR_20_RO(ws_opt);
1195 static NVM_DEV_ATTR_20_RO(maxoc);
1196 static NVM_DEV_ATTR_20_RO(maxocpu);
1197 static NVM_DEV_ATTR_20_RO(mw_cunits);
1198 static NVM_DEV_ATTR_20_RO(write_typ);
1199 static NVM_DEV_ATTR_20_RO(write_max);
1200 static NVM_DEV_ATTR_20_RO(reset_typ);
1201 static NVM_DEV_ATTR_20_RO(reset_max);
1202
1203 static struct attribute *nvm_dev_attrs[] = {
1204         /* version agnostic attrs */
1205         &dev_attr_version.attr,
1206         &dev_attr_capabilities.attr,
1207         &dev_attr_read_typ.attr,
1208         &dev_attr_read_max.attr,
1209
1210         /* 1.2 attrs */
1211         &dev_attr_vendor_opcode.attr,
1212         &dev_attr_device_mode.attr,
1213         &dev_attr_media_manager.attr,
1214         &dev_attr_ppa_format.attr,
1215         &dev_attr_media_type.attr,
1216         &dev_attr_flash_media_type.attr,
1217         &dev_attr_num_channels.attr,
1218         &dev_attr_num_luns.attr,
1219         &dev_attr_num_planes.attr,
1220         &dev_attr_num_blocks.attr,
1221         &dev_attr_num_pages.attr,
1222         &dev_attr_page_size.attr,
1223         &dev_attr_hw_sector_size.attr,
1224         &dev_attr_oob_sector_size.attr,
1225         &dev_attr_prog_typ.attr,
1226         &dev_attr_prog_max.attr,
1227         &dev_attr_erase_typ.attr,
1228         &dev_attr_erase_max.attr,
1229         &dev_attr_multiplane_modes.attr,
1230         &dev_attr_media_capabilities.attr,
1231         &dev_attr_max_phys_secs.attr,
1232
1233         /* 2.0 attrs */
1234         &dev_attr_groups.attr,
1235         &dev_attr_punits.attr,
1236         &dev_attr_chunks.attr,
1237         &dev_attr_clba.attr,
1238         &dev_attr_ws_min.attr,
1239         &dev_attr_ws_opt.attr,
1240         &dev_attr_maxoc.attr,
1241         &dev_attr_maxocpu.attr,
1242         &dev_attr_mw_cunits.attr,
1243
1244         &dev_attr_write_typ.attr,
1245         &dev_attr_write_max.attr,
1246         &dev_attr_reset_typ.attr,
1247         &dev_attr_reset_max.attr,
1248
1249         NULL,
1250 };
1251
1252 static umode_t nvm_dev_attrs_visible(struct kobject *kobj,
1253                                      struct attribute *attr, int index)
1254 {
1255         struct device *dev = container_of(kobj, struct device, kobj);
1256         struct gendisk *disk = dev_to_disk(dev);
1257         struct nvme_ns *ns = disk->private_data;
1258         struct nvm_dev *ndev = ns->ndev;
1259         struct device_attribute *dev_attr =
1260                 container_of(attr, typeof(*dev_attr), attr);
1261
1262         if (!ndev)
1263                 return 0;
1264
1265         if (dev_attr->show == nvm_dev_attr_show)
1266                 return attr->mode;
1267
1268         switch (ndev->geo.major_ver_id) {
1269         case 1:
1270                 if (dev_attr->show == nvm_dev_attr_show_12)
1271                         return attr->mode;
1272                 break;
1273         case 2:
1274                 if (dev_attr->show == nvm_dev_attr_show_20)
1275                         return attr->mode;
1276                 break;
1277         }
1278
1279         return 0;
1280 }
1281
1282 const struct attribute_group nvme_nvm_attr_group = {
1283         .name           = "lightnvm",
1284         .attrs          = nvm_dev_attrs,
1285         .is_visible     = nvm_dev_attrs_visible,
1286 };