2 * nvme-lightnvm.c - LightNVM NVMe device
4 * Copyright (C) 2014-2015 IT University of Copenhagen
5 * Initial release: Matias Bjorling <mb@lightnvm.io>
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.
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.
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,
25 #include <linux/nvme.h>
26 #include <linux/bitops.h>
27 #include <linux/lightnvm.h>
28 #include <linux/vmalloc.h>
30 enum nvme_nvm_admin_opcode {
31 nvme_nvm_admin_identity = 0xe2,
32 nvme_nvm_admin_get_l2p_tbl = 0xea,
33 nvme_nvm_admin_get_bb_tbl = 0xf2,
34 nvme_nvm_admin_set_bb_tbl = 0xf1,
37 struct nvme_nvm_hb_rw {
53 struct nvme_nvm_ph_rw {
69 struct nvme_nvm_identity {
81 struct nvme_nvm_l2ptbl {
94 struct nvme_nvm_getbbtbl {
106 struct nvme_nvm_setbbtbl {
121 struct nvme_nvm_erase_blk {
136 struct nvme_nvm_command {
138 struct nvme_common_command common;
139 struct nvme_nvm_identity identity;
140 struct nvme_nvm_hb_rw hb_rw;
141 struct nvme_nvm_ph_rw ph_rw;
142 struct nvme_nvm_l2ptbl l2p;
143 struct nvme_nvm_getbbtbl get_bb;
144 struct nvme_nvm_setbbtbl set_bb;
145 struct nvme_nvm_erase_blk erase;
149 #define NVME_NVM_LP_MLC_PAIRS 886
150 struct nvme_nvm_lp_mlc {
152 __u8 pairs[NVME_NVM_LP_MLC_PAIRS];
155 struct nvme_nvm_lp_tbl {
157 struct nvme_nvm_lp_mlc mlc;
160 struct nvme_nvm_id_group {
184 struct nvme_nvm_lp_tbl lptbl;
187 struct nvme_nvm_addr_format {
210 struct nvme_nvm_addr_format ppaf;
212 struct nvme_nvm_id_group groups[4];
215 struct nvme_nvm_bb_tbl {
230 * Check we didn't inadvertently grow the command struct
232 static inline void _nvme_nvm_check_size(void)
234 BUILD_BUG_ON(sizeof(struct nvme_nvm_identity) != 64);
235 BUILD_BUG_ON(sizeof(struct nvme_nvm_hb_rw) != 64);
236 BUILD_BUG_ON(sizeof(struct nvme_nvm_ph_rw) != 64);
237 BUILD_BUG_ON(sizeof(struct nvme_nvm_getbbtbl) != 64);
238 BUILD_BUG_ON(sizeof(struct nvme_nvm_setbbtbl) != 64);
239 BUILD_BUG_ON(sizeof(struct nvme_nvm_l2ptbl) != 64);
240 BUILD_BUG_ON(sizeof(struct nvme_nvm_erase_blk) != 64);
241 BUILD_BUG_ON(sizeof(struct nvme_nvm_id_group) != 960);
242 BUILD_BUG_ON(sizeof(struct nvme_nvm_addr_format) != 128);
243 BUILD_BUG_ON(sizeof(struct nvme_nvm_id) != 4096);
244 BUILD_BUG_ON(sizeof(struct nvme_nvm_bb_tbl) != 512);
247 static int init_grps(struct nvm_id *nvm_id, struct nvme_nvm_id *nvme_nvm_id)
249 struct nvme_nvm_id_group *src;
250 struct nvm_id_group *dst;
253 end = min_t(u32, 4, nvm_id->cgrps);
255 for (i = 0; i < end; i++) {
256 src = &nvme_nvm_id->groups[i];
257 dst = &nvm_id->groups[i];
259 dst->mtype = src->mtype;
260 dst->fmtype = src->fmtype;
261 dst->num_ch = src->num_ch;
262 dst->num_lun = src->num_lun;
263 dst->num_pln = src->num_pln;
265 dst->num_pg = le16_to_cpu(src->num_pg);
266 dst->num_blk = le16_to_cpu(src->num_blk);
267 dst->fpg_sz = le16_to_cpu(src->fpg_sz);
268 dst->csecs = le16_to_cpu(src->csecs);
269 dst->sos = le16_to_cpu(src->sos);
271 dst->trdt = le32_to_cpu(src->trdt);
272 dst->trdm = le32_to_cpu(src->trdm);
273 dst->tprt = le32_to_cpu(src->tprt);
274 dst->tprm = le32_to_cpu(src->tprm);
275 dst->tbet = le32_to_cpu(src->tbet);
276 dst->tbem = le32_to_cpu(src->tbem);
277 dst->mpos = le32_to_cpu(src->mpos);
278 dst->mccap = le32_to_cpu(src->mccap);
280 dst->cpar = le16_to_cpu(src->cpar);
282 if (dst->fmtype == NVM_ID_FMTYPE_MLC) {
283 memcpy(dst->lptbl.id, src->lptbl.id, 8);
284 dst->lptbl.mlc.num_pairs =
285 le16_to_cpu(src->lptbl.mlc.num_pairs);
287 if (dst->lptbl.mlc.num_pairs > NVME_NVM_LP_MLC_PAIRS) {
288 pr_err("nvm: number of MLC pairs not supported\n");
292 memcpy(dst->lptbl.mlc.pairs, src->lptbl.mlc.pairs,
293 dst->lptbl.mlc.num_pairs);
300 static int nvme_nvm_identity(struct nvm_dev *nvmdev, struct nvm_id *nvm_id)
302 struct nvme_ns *ns = nvmdev->q->queuedata;
303 struct nvme_nvm_id *nvme_nvm_id;
304 struct nvme_nvm_command c = {};
307 c.identity.opcode = nvme_nvm_admin_identity;
308 c.identity.nsid = cpu_to_le32(ns->ns_id);
309 c.identity.chnl_off = 0;
311 nvme_nvm_id = kmalloc(sizeof(struct nvme_nvm_id), GFP_KERNEL);
315 ret = nvme_submit_sync_cmd(ns->ctrl->admin_q, (struct nvme_command *)&c,
316 nvme_nvm_id, sizeof(struct nvme_nvm_id));
322 nvm_id->ver_id = nvme_nvm_id->ver_id;
323 nvm_id->vmnt = nvme_nvm_id->vmnt;
324 nvm_id->cgrps = nvme_nvm_id->cgrps;
325 nvm_id->cap = le32_to_cpu(nvme_nvm_id->cap);
326 nvm_id->dom = le32_to_cpu(nvme_nvm_id->dom);
327 memcpy(&nvm_id->ppaf, &nvme_nvm_id->ppaf,
328 sizeof(struct nvme_nvm_addr_format));
330 ret = init_grps(nvm_id, nvme_nvm_id);
336 static int nvme_nvm_get_l2p_tbl(struct nvm_dev *nvmdev, u64 slba, u32 nlb,
337 nvm_l2p_update_fn *update_l2p, void *priv)
339 struct nvme_ns *ns = nvmdev->q->queuedata;
340 struct nvme_nvm_command c = {};
341 u32 len = queue_max_hw_sectors(ns->ctrl->admin_q) << 9;
342 u32 nlb_pr_rq = len / sizeof(u64);
347 c.l2p.opcode = nvme_nvm_admin_get_l2p_tbl;
348 c.l2p.nsid = cpu_to_le32(ns->ns_id);
349 entries = kmalloc(len, GFP_KERNEL);
354 u32 cmd_nlb = min(nlb_pr_rq, nlb);
356 c.l2p.slba = cpu_to_le64(cmd_slba);
357 c.l2p.nlb = cpu_to_le32(cmd_nlb);
359 ret = nvme_submit_sync_cmd(ns->ctrl->admin_q,
360 (struct nvme_command *)&c, entries, len);
362 dev_err(ns->ctrl->dev, "L2P table transfer failed (%d)\n",
368 if (update_l2p(cmd_slba, cmd_nlb, entries, priv)) {
382 static int nvme_nvm_get_bb_tbl(struct nvm_dev *nvmdev, struct ppa_addr ppa,
383 int nr_blocks, nvm_bb_update_fn *update_bbtbl,
386 struct request_queue *q = nvmdev->q;
387 struct nvme_ns *ns = q->queuedata;
388 struct nvme_ctrl *ctrl = ns->ctrl;
389 struct nvme_nvm_command c = {};
390 struct nvme_nvm_bb_tbl *bb_tbl;
391 int tblsz = sizeof(struct nvme_nvm_bb_tbl) + nr_blocks;
394 c.get_bb.opcode = nvme_nvm_admin_get_bb_tbl;
395 c.get_bb.nsid = cpu_to_le32(ns->ns_id);
396 c.get_bb.spba = cpu_to_le64(ppa.ppa);
398 bb_tbl = kzalloc(tblsz, GFP_KERNEL);
402 ret = nvme_submit_sync_cmd(ctrl->admin_q, (struct nvme_command *)&c,
405 dev_err(ctrl->dev, "get bad block table failed (%d)\n", ret);
410 if (bb_tbl->tblid[0] != 'B' || bb_tbl->tblid[1] != 'B' ||
411 bb_tbl->tblid[2] != 'L' || bb_tbl->tblid[3] != 'T') {
412 dev_err(ctrl->dev, "bbt format mismatch\n");
417 if (le16_to_cpu(bb_tbl->verid) != 1) {
419 dev_err(ctrl->dev, "bbt version not supported\n");
423 if (le32_to_cpu(bb_tbl->tblks) != nr_blocks) {
425 dev_err(ctrl->dev, "bbt unsuspected blocks returned (%u!=%u)",
426 le32_to_cpu(bb_tbl->tblks), nr_blocks);
430 ppa = dev_to_generic_addr(nvmdev, ppa);
431 ret = update_bbtbl(ppa, nr_blocks, bb_tbl->blk, priv);
437 static int nvme_nvm_set_bb_tbl(struct nvm_dev *nvmdev, struct nvm_rq *rqd,
440 struct nvme_ns *ns = nvmdev->q->queuedata;
441 struct nvme_nvm_command c = {};
444 c.set_bb.opcode = nvme_nvm_admin_set_bb_tbl;
445 c.set_bb.nsid = cpu_to_le32(ns->ns_id);
446 c.set_bb.spba = cpu_to_le64(rqd->ppa_addr.ppa);
447 c.set_bb.nlb = cpu_to_le16(rqd->nr_pages - 1);
448 c.set_bb.value = type;
450 ret = nvme_submit_sync_cmd(ns->ctrl->admin_q, (struct nvme_command *)&c,
453 dev_err(ns->ctrl->dev, "set bad block table failed (%d)\n", ret);
457 static inline void nvme_nvm_rqtocmd(struct request *rq, struct nvm_rq *rqd,
458 struct nvme_ns *ns, struct nvme_nvm_command *c)
460 c->ph_rw.opcode = rqd->opcode;
461 c->ph_rw.nsid = cpu_to_le32(ns->ns_id);
462 c->ph_rw.spba = cpu_to_le64(rqd->ppa_addr.ppa);
463 c->ph_rw.control = cpu_to_le16(rqd->flags);
464 c->ph_rw.length = cpu_to_le16(rqd->nr_pages - 1);
466 if (rqd->opcode == NVM_OP_HBWRITE || rqd->opcode == NVM_OP_HBREAD)
467 c->hb_rw.slba = cpu_to_le64(nvme_block_nr(ns,
468 rqd->bio->bi_iter.bi_sector));
471 static void nvme_nvm_end_io(struct request *rq, int error)
473 struct nvm_rq *rqd = rq->end_io_data;
475 nvm_end_io(rqd, error);
478 blk_mq_free_request(rq);
481 static int nvme_nvm_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
483 struct request_queue *q = dev->q;
484 struct nvme_ns *ns = q->queuedata;
486 struct bio *bio = rqd->bio;
487 struct nvme_nvm_command *cmd;
489 rq = blk_mq_alloc_request(q, bio_rw(bio), 0);
493 cmd = kzalloc(sizeof(struct nvme_nvm_command), GFP_KERNEL);
495 blk_mq_free_request(rq);
499 rq->cmd_type = REQ_TYPE_DRV_PRIV;
500 rq->ioprio = bio_prio(bio);
502 if (bio_has_data(bio))
503 rq->nr_phys_segments = bio_phys_segments(q, bio);
505 rq->__data_len = bio->bi_iter.bi_size;
506 rq->bio = rq->biotail = bio;
508 nvme_nvm_rqtocmd(rq, rqd, ns, cmd);
510 rq->cmd = (unsigned char *)cmd;
511 rq->cmd_len = sizeof(struct nvme_nvm_command);
512 rq->special = (void *)0;
514 rq->end_io_data = rqd;
516 blk_execute_rq_nowait(q, NULL, rq, 0, nvme_nvm_end_io);
521 static int nvme_nvm_erase_block(struct nvm_dev *dev, struct nvm_rq *rqd)
523 struct request_queue *q = dev->q;
524 struct nvme_ns *ns = q->queuedata;
525 struct nvme_nvm_command c = {};
527 c.erase.opcode = NVM_OP_ERASE;
528 c.erase.nsid = cpu_to_le32(ns->ns_id);
529 c.erase.spba = cpu_to_le64(rqd->ppa_addr.ppa);
530 c.erase.length = cpu_to_le16(rqd->nr_pages - 1);
532 return nvme_submit_sync_cmd(q, (struct nvme_command *)&c, NULL, 0);
535 static void *nvme_nvm_create_dma_pool(struct nvm_dev *nvmdev, char *name)
537 struct nvme_ns *ns = nvmdev->q->queuedata;
539 return dma_pool_create(name, ns->ctrl->dev, PAGE_SIZE, PAGE_SIZE, 0);
542 static void nvme_nvm_destroy_dma_pool(void *pool)
544 struct dma_pool *dma_pool = pool;
546 dma_pool_destroy(dma_pool);
549 static void *nvme_nvm_dev_dma_alloc(struct nvm_dev *dev, void *pool,
550 gfp_t mem_flags, dma_addr_t *dma_handler)
552 return dma_pool_alloc(pool, mem_flags, dma_handler);
555 static void nvme_nvm_dev_dma_free(void *pool, void *ppa_list,
556 dma_addr_t dma_handler)
558 dma_pool_free(pool, ppa_list, dma_handler);
561 static struct nvm_dev_ops nvme_nvm_dev_ops = {
562 .identity = nvme_nvm_identity,
564 .get_l2p_tbl = nvme_nvm_get_l2p_tbl,
566 .get_bb_tbl = nvme_nvm_get_bb_tbl,
567 .set_bb_tbl = nvme_nvm_set_bb_tbl,
569 .submit_io = nvme_nvm_submit_io,
570 .erase_block = nvme_nvm_erase_block,
572 .create_dma_pool = nvme_nvm_create_dma_pool,
573 .destroy_dma_pool = nvme_nvm_destroy_dma_pool,
574 .dev_dma_alloc = nvme_nvm_dev_dma_alloc,
575 .dev_dma_free = nvme_nvm_dev_dma_free,
580 int nvme_nvm_register(struct request_queue *q, char *disk_name)
582 return nvm_register(q, disk_name, &nvme_nvm_dev_ops);
585 void nvme_nvm_unregister(struct request_queue *q, char *disk_name)
587 nvm_unregister(disk_name);
590 /* move to shared place when used in multiple places. */
591 #define PCI_VENDOR_ID_CNEX 0x1d1d
592 #define PCI_DEVICE_ID_CNEX_WL 0x2807
593 #define PCI_DEVICE_ID_CNEX_QEMU 0x1f1f
595 int nvme_nvm_ns_supported(struct nvme_ns *ns, struct nvme_id_ns *id)
597 struct nvme_ctrl *ctrl = ns->ctrl;
598 /* XXX: this is poking into PCI structures from generic code! */
599 struct pci_dev *pdev = to_pci_dev(ctrl->dev);
601 /* QEMU NVMe simulator - PCI ID + Vendor specific bit */
602 if (pdev->vendor == PCI_VENDOR_ID_CNEX &&
603 pdev->device == PCI_DEVICE_ID_CNEX_QEMU &&
607 /* CNEX Labs - PCI ID + Vendor specific bit */
608 if (pdev->vendor == PCI_VENDOR_ID_CNEX &&
609 pdev->device == PCI_DEVICE_ID_CNEX_WL &&