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
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,
35 };
36
37 struct nvme_nvm_hb_rw {
38 __u8 opcode;
39 __u8 flags;
40 __u16 command_id;
41 __le32 nsid;
42 __u64 rsvd2;
43 __le64 metadata;
44 __le64 prp1;
45 __le64 prp2;
46 __le64 spba;
47 __le16 length;
48 __le16 control;
49 __le32 dsmgmt;
50 __le64 slba;
51 };
52
53 struct nvme_nvm_ph_rw {
54 __u8 opcode;
55 __u8 flags;
56 __u16 command_id;
57 __le32 nsid;
58 __u64 rsvd2;
59 __le64 metadata;
60 __le64 prp1;
61 __le64 prp2;
62 __le64 spba;
63 __le16 length;
64 __le16 control;
65 __le32 dsmgmt;
66 __le64 resv;
67 };
68
69 struct nvme_nvm_identity {
70 __u8 opcode;
71 __u8 flags;
72 __u16 command_id;
73 __le32 nsid;
74 __u64 rsvd[2];
75 __le64 prp1;
76 __le64 prp2;
77 __le32 chnl_off;
78 __u32 rsvd11[5];
79 };
80
81 struct nvme_nvm_l2ptbl {
82 __u8 opcode;
83 __u8 flags;
84 __u16 command_id;
85 __le32 nsid;
86 __le32 cdw2[4];
87 __le64 prp1;
88 __le64 prp2;
89 __le64 slba;
90 __le32 nlb;
91 __le16 cdw14[6];
92 };
93
94 struct nvme_nvm_getbbtbl {
95 __u8 opcode;
96 __u8 flags;
97 __u16 command_id;
98 __le32 nsid;
99 __u64 rsvd[2];
100 __le64 prp1;
101 __le64 prp2;
102 __le64 spba;
103 __u32 rsvd4[4];
104 };
105
106 struct nvme_nvm_setbbtbl {
107 __u8 opcode;
108 __u8 flags;
109 __u16 command_id;
110 __le32 nsid;
111 __le64 rsvd[2];
112 __le64 prp1;
113 __le64 prp2;
114 __le64 spba;
115 __le16 nlb;
116 __u8 value;
117 __u8 rsvd3;
118 __u32 rsvd4[3];
119 };
120
121 struct nvme_nvm_erase_blk {
122 __u8 opcode;
123 __u8 flags;
124 __u16 command_id;
125 __le32 nsid;
126 __u64 rsvd[2];
127 __le64 prp1;
128 __le64 prp2;
129 __le64 spba;
130 __le16 length;
131 __le16 control;
132 __le32 dsmgmt;
133 __le64 resv;
134 };
135
136 struct nvme_nvm_command {
137 union {
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;
146 };
147 };
148
149 struct nvme_nvm_id_group {
150 __u8 mtype;
151 __u8 fmtype;
152 __le16 res16;
153 __u8 num_ch;
154 __u8 num_lun;
155 __u8 num_pln;
156 __u8 rsvd1;
157 __le16 num_blk;
158 __le16 num_pg;
159 __le16 fpg_sz;
160 __le16 csecs;
161 __le16 sos;
162 __le16 rsvd2;
163 __le32 trdt;
164 __le32 trdm;
165 __le32 tprt;
166 __le32 tprm;
167 __le32 tbet;
168 __le32 tbem;
169 __le32 mpos;
170 __le32 mccap;
171 __le16 cpar;
172 __u8 reserved[906];
173 } __packed;
174
175 struct nvme_nvm_addr_format {
176 __u8 ch_offset;
177 __u8 ch_len;
178 __u8 lun_offset;
179 __u8 lun_len;
180 __u8 pln_offset;
181 __u8 pln_len;
182 __u8 blk_offset;
183 __u8 blk_len;
184 __u8 pg_offset;
185 __u8 pg_len;
186 __u8 sect_offset;
187 __u8 sect_len;
188 __u8 res[4];
189 } __packed;
190
191 struct nvme_nvm_id {
192 __u8 ver_id;
193 __u8 vmnt;
194 __u8 cgrps;
195 __u8 res;
196 __le32 cap;
197 __le32 dom;
198 struct nvme_nvm_addr_format ppaf;
199 __u8 resv[228];
200 struct nvme_nvm_id_group groups[4];
201 } __packed;
202
203 struct nvme_nvm_bb_tbl {
204 __u8 tblid[4];
205 __le16 verid;
206 __le16 revid;
207 __le32 rvsd1;
208 __le32 tblks;
209 __le32 tfact;
210 __le32 tgrown;
211 __le32 tdresv;
212 __le32 thresv;
213 __le32 rsvd2[8];
214 __u8 blk[0];
215 };
216
217 /*
218 * Check we didn't inadvertently grow the command struct
219 */
_nvme_nvm_check_size(void)220 static inline void _nvme_nvm_check_size(void)
221 {
222 BUILD_BUG_ON(sizeof(struct nvme_nvm_identity) != 64);
223 BUILD_BUG_ON(sizeof(struct nvme_nvm_hb_rw) != 64);
224 BUILD_BUG_ON(sizeof(struct nvme_nvm_ph_rw) != 64);
225 BUILD_BUG_ON(sizeof(struct nvme_nvm_getbbtbl) != 64);
226 BUILD_BUG_ON(sizeof(struct nvme_nvm_setbbtbl) != 64);
227 BUILD_BUG_ON(sizeof(struct nvme_nvm_l2ptbl) != 64);
228 BUILD_BUG_ON(sizeof(struct nvme_nvm_erase_blk) != 64);
229 BUILD_BUG_ON(sizeof(struct nvme_nvm_id_group) != 960);
230 BUILD_BUG_ON(sizeof(struct nvme_nvm_addr_format) != 128);
231 BUILD_BUG_ON(sizeof(struct nvme_nvm_id) != 4096);
232 BUILD_BUG_ON(sizeof(struct nvme_nvm_bb_tbl) != 512);
233 }
234
init_grps(struct nvm_id * nvm_id,struct nvme_nvm_id * nvme_nvm_id)235 static int init_grps(struct nvm_id *nvm_id, struct nvme_nvm_id *nvme_nvm_id)
236 {
237 struct nvme_nvm_id_group *src;
238 struct nvm_id_group *dst;
239 int i, end;
240
241 end = min_t(u32, 4, nvm_id->cgrps);
242
243 for (i = 0; i < end; i++) {
244 src = &nvme_nvm_id->groups[i];
245 dst = &nvm_id->groups[i];
246
247 dst->mtype = src->mtype;
248 dst->fmtype = src->fmtype;
249 dst->num_ch = src->num_ch;
250 dst->num_lun = src->num_lun;
251 dst->num_pln = src->num_pln;
252
253 dst->num_pg = le16_to_cpu(src->num_pg);
254 dst->num_blk = le16_to_cpu(src->num_blk);
255 dst->fpg_sz = le16_to_cpu(src->fpg_sz);
256 dst->csecs = le16_to_cpu(src->csecs);
257 dst->sos = le16_to_cpu(src->sos);
258
259 dst->trdt = le32_to_cpu(src->trdt);
260 dst->trdm = le32_to_cpu(src->trdm);
261 dst->tprt = le32_to_cpu(src->tprt);
262 dst->tprm = le32_to_cpu(src->tprm);
263 dst->tbet = le32_to_cpu(src->tbet);
264 dst->tbem = le32_to_cpu(src->tbem);
265 dst->mpos = le32_to_cpu(src->mpos);
266 dst->mccap = le32_to_cpu(src->mccap);
267
268 dst->cpar = le16_to_cpu(src->cpar);
269 }
270
271 return 0;
272 }
273
nvme_nvm_identity(struct nvm_dev * nvmdev,struct nvm_id * nvm_id)274 static int nvme_nvm_identity(struct nvm_dev *nvmdev, struct nvm_id *nvm_id)
275 {
276 struct nvme_ns *ns = nvmdev->q->queuedata;
277 struct nvme_dev *dev = ns->dev;
278 struct nvme_nvm_id *nvme_nvm_id;
279 struct nvme_nvm_command c = {};
280 int ret;
281
282 c.identity.opcode = nvme_nvm_admin_identity;
283 c.identity.nsid = cpu_to_le32(ns->ns_id);
284 c.identity.chnl_off = 0;
285
286 nvme_nvm_id = kmalloc(sizeof(struct nvme_nvm_id), GFP_KERNEL);
287 if (!nvme_nvm_id)
288 return -ENOMEM;
289
290 ret = nvme_submit_sync_cmd(dev->admin_q, (struct nvme_command *)&c,
291 nvme_nvm_id, sizeof(struct nvme_nvm_id));
292 if (ret) {
293 ret = -EIO;
294 goto out;
295 }
296
297 nvm_id->ver_id = nvme_nvm_id->ver_id;
298 nvm_id->vmnt = nvme_nvm_id->vmnt;
299 nvm_id->cgrps = nvme_nvm_id->cgrps;
300 nvm_id->cap = le32_to_cpu(nvme_nvm_id->cap);
301 nvm_id->dom = le32_to_cpu(nvme_nvm_id->dom);
302 memcpy(&nvm_id->ppaf, &nvme_nvm_id->ppaf,
303 sizeof(struct nvme_nvm_addr_format));
304
305 ret = init_grps(nvm_id, nvme_nvm_id);
306 out:
307 kfree(nvme_nvm_id);
308 return ret;
309 }
310
nvme_nvm_get_l2p_tbl(struct nvm_dev * nvmdev,u64 slba,u32 nlb,nvm_l2p_update_fn * update_l2p,void * priv)311 static int nvme_nvm_get_l2p_tbl(struct nvm_dev *nvmdev, u64 slba, u32 nlb,
312 nvm_l2p_update_fn *update_l2p, void *priv)
313 {
314 struct nvme_ns *ns = nvmdev->q->queuedata;
315 struct nvme_dev *dev = ns->dev;
316 struct nvme_nvm_command c = {};
317 u32 len = queue_max_hw_sectors(dev->admin_q) << 9;
318 u32 nlb_pr_rq = len / sizeof(u64);
319 u64 cmd_slba = slba;
320 void *entries;
321 int ret = 0;
322
323 c.l2p.opcode = nvme_nvm_admin_get_l2p_tbl;
324 c.l2p.nsid = cpu_to_le32(ns->ns_id);
325 entries = kmalloc(len, GFP_KERNEL);
326 if (!entries)
327 return -ENOMEM;
328
329 while (nlb) {
330 u32 cmd_nlb = min(nlb_pr_rq, nlb);
331
332 c.l2p.slba = cpu_to_le64(cmd_slba);
333 c.l2p.nlb = cpu_to_le32(cmd_nlb);
334
335 ret = nvme_submit_sync_cmd(dev->admin_q,
336 (struct nvme_command *)&c, entries, len);
337 if (ret) {
338 dev_err(dev->dev, "L2P table transfer failed (%d)\n",
339 ret);
340 ret = -EIO;
341 goto out;
342 }
343
344 if (update_l2p(cmd_slba, cmd_nlb, entries, priv)) {
345 ret = -EINTR;
346 goto out;
347 }
348
349 cmd_slba += cmd_nlb;
350 nlb -= cmd_nlb;
351 }
352
353 out:
354 kfree(entries);
355 return ret;
356 }
357
nvme_nvm_get_bb_tbl(struct nvm_dev * nvmdev,struct ppa_addr ppa,int nr_blocks,nvm_bb_update_fn * update_bbtbl,void * priv)358 static int nvme_nvm_get_bb_tbl(struct nvm_dev *nvmdev, struct ppa_addr ppa,
359 int nr_blocks, nvm_bb_update_fn *update_bbtbl,
360 void *priv)
361 {
362 struct request_queue *q = nvmdev->q;
363 struct nvme_ns *ns = q->queuedata;
364 struct nvme_dev *dev = ns->dev;
365 struct nvme_nvm_command c = {};
366 struct nvme_nvm_bb_tbl *bb_tbl;
367 int tblsz = sizeof(struct nvme_nvm_bb_tbl) + nr_blocks;
368 int ret = 0;
369
370 c.get_bb.opcode = nvme_nvm_admin_get_bb_tbl;
371 c.get_bb.nsid = cpu_to_le32(ns->ns_id);
372 c.get_bb.spba = cpu_to_le64(ppa.ppa);
373
374 bb_tbl = kzalloc(tblsz, GFP_KERNEL);
375 if (!bb_tbl)
376 return -ENOMEM;
377
378 ret = nvme_submit_sync_cmd(dev->admin_q, (struct nvme_command *)&c,
379 bb_tbl, tblsz);
380 if (ret) {
381 dev_err(dev->dev, "get bad block table failed (%d)\n", ret);
382 ret = -EIO;
383 goto out;
384 }
385
386 if (bb_tbl->tblid[0] != 'B' || bb_tbl->tblid[1] != 'B' ||
387 bb_tbl->tblid[2] != 'L' || bb_tbl->tblid[3] != 'T') {
388 dev_err(dev->dev, "bbt format mismatch\n");
389 ret = -EINVAL;
390 goto out;
391 }
392
393 if (le16_to_cpu(bb_tbl->verid) != 1) {
394 ret = -EINVAL;
395 dev_err(dev->dev, "bbt version not supported\n");
396 goto out;
397 }
398
399 if (le32_to_cpu(bb_tbl->tblks) != nr_blocks) {
400 ret = -EINVAL;
401 dev_err(dev->dev, "bbt unsuspected blocks returned (%u!=%u)",
402 le32_to_cpu(bb_tbl->tblks), nr_blocks);
403 goto out;
404 }
405
406 ppa = dev_to_generic_addr(nvmdev, ppa);
407 ret = update_bbtbl(ppa, nr_blocks, bb_tbl->blk, priv);
408 if (ret) {
409 ret = -EINTR;
410 goto out;
411 }
412
413 out:
414 kfree(bb_tbl);
415 return ret;
416 }
417
nvme_nvm_set_bb_tbl(struct nvm_dev * nvmdev,struct nvm_rq * rqd,int type)418 static int nvme_nvm_set_bb_tbl(struct nvm_dev *nvmdev, struct nvm_rq *rqd,
419 int type)
420 {
421 struct nvme_ns *ns = nvmdev->q->queuedata;
422 struct nvme_dev *dev = ns->dev;
423 struct nvme_nvm_command c = {};
424 int ret = 0;
425
426 c.set_bb.opcode = nvme_nvm_admin_set_bb_tbl;
427 c.set_bb.nsid = cpu_to_le32(ns->ns_id);
428 c.set_bb.spba = cpu_to_le64(rqd->ppa_addr.ppa);
429 c.set_bb.nlb = cpu_to_le16(rqd->nr_pages - 1);
430 c.set_bb.value = type;
431
432 ret = nvme_submit_sync_cmd(dev->admin_q, (struct nvme_command *)&c,
433 NULL, 0);
434 if (ret)
435 dev_err(dev->dev, "set bad block table failed (%d)\n", ret);
436 return ret;
437 }
438
nvme_nvm_rqtocmd(struct request * rq,struct nvm_rq * rqd,struct nvme_ns * ns,struct nvme_nvm_command * c)439 static inline void nvme_nvm_rqtocmd(struct request *rq, struct nvm_rq *rqd,
440 struct nvme_ns *ns, struct nvme_nvm_command *c)
441 {
442 c->ph_rw.opcode = rqd->opcode;
443 c->ph_rw.nsid = cpu_to_le32(ns->ns_id);
444 c->ph_rw.spba = cpu_to_le64(rqd->ppa_addr.ppa);
445 c->ph_rw.control = cpu_to_le16(rqd->flags);
446 c->ph_rw.length = cpu_to_le16(rqd->nr_pages - 1);
447
448 if (rqd->opcode == NVM_OP_HBWRITE || rqd->opcode == NVM_OP_HBREAD)
449 c->hb_rw.slba = cpu_to_le64(nvme_block_nr(ns,
450 rqd->bio->bi_iter.bi_sector));
451 }
452
nvme_nvm_end_io(struct request * rq,int error)453 static void nvme_nvm_end_io(struct request *rq, int error)
454 {
455 struct nvm_rq *rqd = rq->end_io_data;
456 struct nvm_dev *dev = rqd->dev;
457
458 if (dev->mt && dev->mt->end_io(rqd, error))
459 pr_err("nvme: err status: %x result: %lx\n",
460 rq->errors, (unsigned long)rq->special);
461
462 kfree(rq->cmd);
463 blk_mq_free_request(rq);
464 }
465
nvme_nvm_submit_io(struct nvm_dev * dev,struct nvm_rq * rqd)466 static int nvme_nvm_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
467 {
468 struct request_queue *q = dev->q;
469 struct nvme_ns *ns = q->queuedata;
470 struct request *rq;
471 struct bio *bio = rqd->bio;
472 struct nvme_nvm_command *cmd;
473
474 rq = blk_mq_alloc_request(q, bio_rw(bio), GFP_KERNEL, 0);
475 if (IS_ERR(rq))
476 return -ENOMEM;
477
478 cmd = kzalloc(sizeof(struct nvme_nvm_command), GFP_KERNEL);
479 if (!cmd) {
480 blk_mq_free_request(rq);
481 return -ENOMEM;
482 }
483
484 rq->cmd_type = REQ_TYPE_DRV_PRIV;
485 rq->ioprio = bio_prio(bio);
486
487 if (bio_has_data(bio))
488 rq->nr_phys_segments = bio_phys_segments(q, bio);
489
490 rq->__data_len = bio->bi_iter.bi_size;
491 rq->bio = rq->biotail = bio;
492
493 nvme_nvm_rqtocmd(rq, rqd, ns, cmd);
494
495 rq->cmd = (unsigned char *)cmd;
496 rq->cmd_len = sizeof(struct nvme_nvm_command);
497 rq->special = (void *)0;
498
499 rq->end_io_data = rqd;
500
501 blk_execute_rq_nowait(q, NULL, rq, 0, nvme_nvm_end_io);
502
503 return 0;
504 }
505
nvme_nvm_erase_block(struct nvm_dev * dev,struct nvm_rq * rqd)506 static int nvme_nvm_erase_block(struct nvm_dev *dev, struct nvm_rq *rqd)
507 {
508 struct request_queue *q = dev->q;
509 struct nvme_ns *ns = q->queuedata;
510 struct nvme_nvm_command c = {};
511
512 c.erase.opcode = NVM_OP_ERASE;
513 c.erase.nsid = cpu_to_le32(ns->ns_id);
514 c.erase.spba = cpu_to_le64(rqd->ppa_addr.ppa);
515 c.erase.length = cpu_to_le16(rqd->nr_pages - 1);
516
517 return nvme_submit_sync_cmd(q, (struct nvme_command *)&c, NULL, 0);
518 }
519
nvme_nvm_create_dma_pool(struct nvm_dev * nvmdev,char * name)520 static void *nvme_nvm_create_dma_pool(struct nvm_dev *nvmdev, char *name)
521 {
522 struct nvme_ns *ns = nvmdev->q->queuedata;
523 struct nvme_dev *dev = ns->dev;
524
525 return dma_pool_create(name, dev->dev, PAGE_SIZE, PAGE_SIZE, 0);
526 }
527
nvme_nvm_destroy_dma_pool(void * pool)528 static void nvme_nvm_destroy_dma_pool(void *pool)
529 {
530 struct dma_pool *dma_pool = pool;
531
532 dma_pool_destroy(dma_pool);
533 }
534
nvme_nvm_dev_dma_alloc(struct nvm_dev * dev,void * pool,gfp_t mem_flags,dma_addr_t * dma_handler)535 static void *nvme_nvm_dev_dma_alloc(struct nvm_dev *dev, void *pool,
536 gfp_t mem_flags, dma_addr_t *dma_handler)
537 {
538 return dma_pool_alloc(pool, mem_flags, dma_handler);
539 }
540
nvme_nvm_dev_dma_free(void * pool,void * ppa_list,dma_addr_t dma_handler)541 static void nvme_nvm_dev_dma_free(void *pool, void *ppa_list,
542 dma_addr_t dma_handler)
543 {
544 dma_pool_free(pool, ppa_list, dma_handler);
545 }
546
547 static struct nvm_dev_ops nvme_nvm_dev_ops = {
548 .identity = nvme_nvm_identity,
549
550 .get_l2p_tbl = nvme_nvm_get_l2p_tbl,
551
552 .get_bb_tbl = nvme_nvm_get_bb_tbl,
553 .set_bb_tbl = nvme_nvm_set_bb_tbl,
554
555 .submit_io = nvme_nvm_submit_io,
556 .erase_block = nvme_nvm_erase_block,
557
558 .create_dma_pool = nvme_nvm_create_dma_pool,
559 .destroy_dma_pool = nvme_nvm_destroy_dma_pool,
560 .dev_dma_alloc = nvme_nvm_dev_dma_alloc,
561 .dev_dma_free = nvme_nvm_dev_dma_free,
562
563 .max_phys_sect = 64,
564 };
565
nvme_nvm_register(struct request_queue * q,char * disk_name)566 int nvme_nvm_register(struct request_queue *q, char *disk_name)
567 {
568 return nvm_register(q, disk_name, &nvme_nvm_dev_ops);
569 }
570
nvme_nvm_unregister(struct request_queue * q,char * disk_name)571 void nvme_nvm_unregister(struct request_queue *q, char *disk_name)
572 {
573 nvm_unregister(disk_name);
574 }
575
576 /* move to shared place when used in multiple places. */
577 #define PCI_VENDOR_ID_CNEX 0x1d1d
578 #define PCI_DEVICE_ID_CNEX_WL 0x2807
579 #define PCI_DEVICE_ID_CNEX_QEMU 0x1f1f
580
nvme_nvm_ns_supported(struct nvme_ns * ns,struct nvme_id_ns * id)581 int nvme_nvm_ns_supported(struct nvme_ns *ns, struct nvme_id_ns *id)
582 {
583 struct nvme_dev *dev = ns->dev;
584 struct pci_dev *pdev = to_pci_dev(dev->dev);
585
586 /* QEMU NVMe simulator - PCI ID + Vendor specific bit */
587 if (pdev->vendor == PCI_VENDOR_ID_CNEX &&
588 pdev->device == PCI_DEVICE_ID_CNEX_QEMU &&
589 id->vs[0] == 0x1)
590 return 1;
591
592 /* CNEX Labs - PCI ID + Vendor specific bit */
593 if (pdev->vendor == PCI_VENDOR_ID_CNEX &&
594 pdev->device == PCI_DEVICE_ID_CNEX_WL &&
595 id->vs[0] == 0x1)
596 return 1;
597
598 return 0;
599 }
600