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,
27 #include <linux/nvme.h>
28 #include <linux/bitops.h>
29 #include <linux/lightnvm.h>
30 #include <linux/vmalloc.h>
32 enum nvme_nvm_admin_opcode {
33 nvme_nvm_admin_identity = 0xe2,
34 nvme_nvm_admin_get_l2p_tbl = 0xea,
35 nvme_nvm_admin_get_bb_tbl = 0xf2,
36 nvme_nvm_admin_set_bb_tbl = 0xf1,
39 struct nvme_nvm_hb_rw {
55 struct nvme_nvm_ph_rw {
71 struct nvme_nvm_identity {
83 struct nvme_nvm_l2ptbl {
96 struct nvme_nvm_bbtbl {
110 struct nvme_nvm_erase_blk {
125 struct nvme_nvm_command {
127 struct nvme_common_command common;
128 struct nvme_nvm_identity identity;
129 struct nvme_nvm_hb_rw hb_rw;
130 struct nvme_nvm_ph_rw ph_rw;
131 struct nvme_nvm_l2ptbl l2p;
132 struct nvme_nvm_bbtbl get_bb;
133 struct nvme_nvm_bbtbl set_bb;
134 struct nvme_nvm_erase_blk erase;
138 struct nvme_nvm_id_group {
161 struct nvme_nvm_addr_format {
184 struct nvme_nvm_addr_format ppaf;
187 struct nvme_nvm_id_group groups[4];
191 * Check we didn't inadvertently grow the command struct
193 static inline void _nvme_nvm_check_size(void)
195 BUILD_BUG_ON(sizeof(struct nvme_nvm_identity) != 64);
196 BUILD_BUG_ON(sizeof(struct nvme_nvm_hb_rw) != 64);
197 BUILD_BUG_ON(sizeof(struct nvme_nvm_ph_rw) != 64);
198 BUILD_BUG_ON(sizeof(struct nvme_nvm_bbtbl) != 64);
199 BUILD_BUG_ON(sizeof(struct nvme_nvm_l2ptbl) != 64);
200 BUILD_BUG_ON(sizeof(struct nvme_nvm_erase_blk) != 64);
201 BUILD_BUG_ON(sizeof(struct nvme_nvm_id_group) != 960);
202 BUILD_BUG_ON(sizeof(struct nvme_nvm_addr_format) != 128);
203 BUILD_BUG_ON(sizeof(struct nvme_nvm_id) != 4096);
206 static int init_grps(struct nvm_id *nvm_id, struct nvme_nvm_id *nvme_nvm_id)
208 struct nvme_nvm_id_group *src;
209 struct nvm_id_group *dst;
212 end = min_t(u32, 4, nvm_id->cgrps);
214 for (i = 0; i < end; i++) {
215 src = &nvme_nvm_id->groups[i];
216 dst = &nvm_id->groups[i];
218 dst->mtype = src->mtype;
219 dst->fmtype = src->fmtype;
220 dst->num_ch = src->num_ch;
221 dst->num_lun = src->num_lun;
222 dst->num_pln = src->num_pln;
224 dst->num_pg = le16_to_cpu(src->num_pg);
225 dst->num_blk = le16_to_cpu(src->num_blk);
226 dst->fpg_sz = le16_to_cpu(src->fpg_sz);
227 dst->csecs = le16_to_cpu(src->csecs);
228 dst->sos = le16_to_cpu(src->sos);
230 dst->trdt = le32_to_cpu(src->trdt);
231 dst->trdm = le32_to_cpu(src->trdm);
232 dst->tprt = le32_to_cpu(src->tprt);
233 dst->tprm = le32_to_cpu(src->tprm);
234 dst->tbet = le32_to_cpu(src->tbet);
235 dst->tbem = le32_to_cpu(src->tbem);
236 dst->mpos = le32_to_cpu(src->mpos);
238 dst->cpar = le16_to_cpu(src->cpar);
244 static int nvme_nvm_identity(struct request_queue *q, struct nvm_id *nvm_id)
246 struct nvme_ns *ns = q->queuedata;
247 struct nvme_nvm_id *nvme_nvm_id;
248 struct nvme_nvm_command c = {};
251 c.identity.opcode = nvme_nvm_admin_identity;
252 c.identity.nsid = cpu_to_le32(ns->ns_id);
253 c.identity.chnl_off = 0;
255 nvme_nvm_id = kmalloc(sizeof(struct nvme_nvm_id), GFP_KERNEL);
259 ret = nvme_submit_sync_cmd(q, (struct nvme_command *)&c, nvme_nvm_id,
260 sizeof(struct nvme_nvm_id));
266 nvm_id->ver_id = nvme_nvm_id->ver_id;
267 nvm_id->vmnt = nvme_nvm_id->vmnt;
268 nvm_id->cgrps = nvme_nvm_id->cgrps;
269 nvm_id->cap = le32_to_cpu(nvme_nvm_id->cap);
270 nvm_id->dom = le32_to_cpu(nvme_nvm_id->dom);
272 ret = init_grps(nvm_id, nvme_nvm_id);
278 static int nvme_nvm_get_l2p_tbl(struct request_queue *q, u64 slba, u32 nlb,
279 nvm_l2p_update_fn *update_l2p, void *priv)
281 struct nvme_ns *ns = q->queuedata;
282 struct nvme_dev *dev = ns->dev;
283 struct nvme_nvm_command c = {};
284 u32 len = queue_max_hw_sectors(q) << 9;
285 u32 nlb_pr_rq = len / sizeof(u64);
290 c.l2p.opcode = nvme_nvm_admin_get_l2p_tbl;
291 c.l2p.nsid = cpu_to_le32(ns->ns_id);
292 entries = kmalloc(len, GFP_KERNEL);
297 u32 cmd_nlb = min(nlb_pr_rq, nlb);
299 c.l2p.slba = cpu_to_le64(cmd_slba);
300 c.l2p.nlb = cpu_to_le32(cmd_nlb);
302 ret = nvme_submit_sync_cmd(q, (struct nvme_command *)&c,
305 dev_err(dev->dev, "L2P table transfer failed (%d)\n",
311 if (update_l2p(cmd_slba, cmd_nlb, entries, priv)) {
325 static int nvme_nvm_get_bb_tbl(struct request_queue *q, int lunid,
326 unsigned int nr_blocks,
327 nvm_bb_update_fn *update_bbtbl, void *priv)
329 struct nvme_ns *ns = q->queuedata;
330 struct nvme_dev *dev = ns->dev;
331 struct nvme_nvm_command c = {};
336 c.get_bb.opcode = nvme_nvm_admin_get_bb_tbl;
337 c.get_bb.nsid = cpu_to_le32(ns->ns_id);
338 c.get_bb.lbb = cpu_to_le32(lunid);
339 bb_bitmap_size = ((nr_blocks >> 15) + 1) * PAGE_SIZE;
340 bb_bitmap = kmalloc(bb_bitmap_size, GFP_KERNEL);
344 bitmap_zero(bb_bitmap, nr_blocks);
346 ret = nvme_submit_sync_cmd(q, (struct nvme_command *)&c, bb_bitmap,
349 dev_err(dev->dev, "get bad block table failed (%d)\n", ret);
354 ret = update_bbtbl(lunid, bb_bitmap, nr_blocks, priv);
365 static inline void nvme_nvm_rqtocmd(struct request *rq, struct nvm_rq *rqd,
366 struct nvme_ns *ns, struct nvme_nvm_command *c)
368 c->ph_rw.opcode = rqd->opcode;
369 c->ph_rw.nsid = cpu_to_le32(ns->ns_id);
370 c->ph_rw.spba = cpu_to_le64(rqd->ppa_addr.ppa);
371 c->ph_rw.control = cpu_to_le16(rqd->flags);
372 c->ph_rw.length = cpu_to_le16(rqd->nr_pages - 1);
374 if (rqd->opcode == NVM_OP_HBWRITE || rqd->opcode == NVM_OP_HBREAD)
375 c->hb_rw.slba = cpu_to_le64(nvme_block_nr(ns,
376 rqd->bio->bi_iter.bi_sector));
379 static void nvme_nvm_end_io(struct request *rq, int error)
381 struct nvm_rq *rqd = rq->end_io_data;
382 struct nvm_dev *dev = rqd->dev;
384 if (dev->mt->end_io(rqd, error))
385 pr_err("nvme: err status: %x result: %lx\n",
386 rq->errors, (unsigned long)rq->special);
389 blk_mq_free_request(rq);
392 static int nvme_nvm_submit_io(struct request_queue *q, struct nvm_rq *rqd)
394 struct nvme_ns *ns = q->queuedata;
396 struct bio *bio = rqd->bio;
397 struct nvme_nvm_command *cmd;
399 rq = blk_mq_alloc_request(q, bio_rw(bio), GFP_KERNEL, 0);
403 cmd = kzalloc(sizeof(struct nvme_nvm_command), GFP_KERNEL);
405 blk_mq_free_request(rq);
409 rq->cmd_type = REQ_TYPE_DRV_PRIV;
410 rq->ioprio = bio_prio(bio);
412 if (bio_has_data(bio))
413 rq->nr_phys_segments = bio_phys_segments(q, bio);
415 rq->__data_len = bio->bi_iter.bi_size;
416 rq->bio = rq->biotail = bio;
418 nvme_nvm_rqtocmd(rq, rqd, ns, cmd);
420 rq->cmd = (unsigned char *)cmd;
421 rq->cmd_len = sizeof(struct nvme_nvm_command);
422 rq->special = (void *)0;
424 rq->end_io_data = rqd;
426 blk_execute_rq_nowait(q, NULL, rq, 0, nvme_nvm_end_io);
431 static int nvme_nvm_erase_block(struct request_queue *q, struct nvm_rq *rqd)
433 struct nvme_ns *ns = q->queuedata;
434 struct nvme_nvm_command c = {};
436 c.erase.opcode = NVM_OP_ERASE;
437 c.erase.nsid = cpu_to_le32(ns->ns_id);
438 c.erase.spba = cpu_to_le64(rqd->ppa_addr.ppa);
439 c.erase.length = cpu_to_le16(rqd->nr_pages - 1);
441 return nvme_submit_sync_cmd(q, (struct nvme_command *)&c, NULL, 0);
444 static void *nvme_nvm_create_dma_pool(struct request_queue *q, char *name)
446 struct nvme_ns *ns = q->queuedata;
447 struct nvme_dev *dev = ns->dev;
449 return dma_pool_create(name, dev->dev, PAGE_SIZE, PAGE_SIZE, 0);
452 static void nvme_nvm_destroy_dma_pool(void *pool)
454 struct dma_pool *dma_pool = pool;
456 dma_pool_destroy(dma_pool);
459 static void *nvme_nvm_dev_dma_alloc(struct request_queue *q, void *pool,
460 gfp_t mem_flags, dma_addr_t *dma_handler)
462 return dma_pool_alloc(pool, mem_flags, dma_handler);
465 static void nvme_nvm_dev_dma_free(void *pool, void *ppa_list,
466 dma_addr_t dma_handler)
468 dma_pool_free(pool, ppa_list, dma_handler);
471 static struct nvm_dev_ops nvme_nvm_dev_ops = {
472 .identity = nvme_nvm_identity,
474 .get_l2p_tbl = nvme_nvm_get_l2p_tbl,
476 .get_bb_tbl = nvme_nvm_get_bb_tbl,
478 .submit_io = nvme_nvm_submit_io,
479 .erase_block = nvme_nvm_erase_block,
481 .create_dma_pool = nvme_nvm_create_dma_pool,
482 .destroy_dma_pool = nvme_nvm_destroy_dma_pool,
483 .dev_dma_alloc = nvme_nvm_dev_dma_alloc,
484 .dev_dma_free = nvme_nvm_dev_dma_free,
489 int nvme_nvm_register(struct request_queue *q, char *disk_name)
491 return nvm_register(q, disk_name, &nvme_nvm_dev_ops);
494 void nvme_nvm_unregister(struct request_queue *q, char *disk_name)
496 nvm_unregister(disk_name);
499 int nvme_nvm_ns_supported(struct nvme_ns *ns, struct nvme_id_ns *id)
501 struct nvme_dev *dev = ns->dev;
502 struct pci_dev *pdev = to_pci_dev(dev->dev);
504 /* QEMU NVMe simulator - PCI ID + Vendor specific bit */
505 if (pdev->vendor == PCI_VENDOR_ID_INTEL && pdev->device == 0x5845 &&
509 /* CNEX Labs - PCI ID + Vendor specific bit */
510 if (pdev->vendor == 0x1d1d && pdev->device == 0x2807 &&
517 int nvme_nvm_register(struct request_queue *q, char *disk_name)
521 void nvme_nvm_unregister(struct request_queue *q, char *disk_name) {};
522 int nvme_nvm_ns_supported(struct nvme_ns *ns, struct nvme_id_ns *id)
526 #endif /* CONFIG_NVM */