]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - sys/dev/nvme/nvme.h
MFV r324714:
[FreeBSD/FreeBSD.git] / sys / dev / nvme / nvme.h
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (C) 2012-2013 Intel Corporation
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  *
28  * $FreeBSD$
29  */
30
31 #ifndef __NVME_H__
32 #define __NVME_H__
33
34 #ifdef _KERNEL
35 #include <sys/types.h>
36 #endif
37
38 #include <sys/param.h>
39 #include <sys/endian.h>
40
41 #define NVME_PASSTHROUGH_CMD            _IOWR('n', 0, struct nvme_pt_command)
42 #define NVME_RESET_CONTROLLER           _IO('n', 1)
43
44 #define NVME_IO_TEST                    _IOWR('n', 100, struct nvme_io_test)
45 #define NVME_BIO_TEST                   _IOWR('n', 101, struct nvme_io_test)
46
47 /*
48  * Macros to deal with NVME revisions, as defined VS register
49  */
50 #define NVME_REV(x, y)                  (((x) << 16) | ((y) << 8))
51 #define NVME_MAJOR(r)                   (((r) >> 16) & 0xffff)
52 #define NVME_MINOR(r)                   (((r) >> 8) & 0xff)
53
54 /*
55  * Use to mark a command to apply to all namespaces, or to retrieve global
56  *  log pages.
57  */
58 #define NVME_GLOBAL_NAMESPACE_TAG       ((uint32_t)0xFFFFFFFF)
59
60 /* Cap nvme to 1MB transfers driver explodes with larger sizes */
61 #define NVME_MAX_XFER_SIZE              (MAXPHYS < (1<<20) ? MAXPHYS : (1<<20))
62
63 /* Register field definitions */
64 #define NVME_CAP_LO_REG_MQES_SHIFT                      (0)
65 #define NVME_CAP_LO_REG_MQES_MASK                       (0xFFFF)
66 #define NVME_CAP_LO_REG_CQR_SHIFT                       (16)
67 #define NVME_CAP_LO_REG_CQR_MASK                        (0x1)
68 #define NVME_CAP_LO_REG_AMS_SHIFT                       (17)
69 #define NVME_CAP_LO_REG_AMS_MASK                        (0x3)
70 #define NVME_CAP_LO_REG_TO_SHIFT                        (24)
71 #define NVME_CAP_LO_REG_TO_MASK                         (0xFF)
72
73 #define NVME_CAP_HI_REG_DSTRD_SHIFT                     (0)
74 #define NVME_CAP_HI_REG_DSTRD_MASK                      (0xF)
75 #define NVME_CAP_HI_REG_CSS_NVM_SHIFT                   (5)
76 #define NVME_CAP_HI_REG_CSS_NVM_MASK                    (0x1)
77 #define NVME_CAP_HI_REG_MPSMIN_SHIFT                    (16)
78 #define NVME_CAP_HI_REG_MPSMIN_MASK                     (0xF)
79 #define NVME_CAP_HI_REG_MPSMAX_SHIFT                    (20)
80 #define NVME_CAP_HI_REG_MPSMAX_MASK                     (0xF)
81
82 #define NVME_CC_REG_EN_SHIFT                            (0)
83 #define NVME_CC_REG_EN_MASK                             (0x1)
84 #define NVME_CC_REG_CSS_SHIFT                           (4)
85 #define NVME_CC_REG_CSS_MASK                            (0x7)
86 #define NVME_CC_REG_MPS_SHIFT                           (7)
87 #define NVME_CC_REG_MPS_MASK                            (0xF)
88 #define NVME_CC_REG_AMS_SHIFT                           (11)
89 #define NVME_CC_REG_AMS_MASK                            (0x7)
90 #define NVME_CC_REG_SHN_SHIFT                           (14)
91 #define NVME_CC_REG_SHN_MASK                            (0x3)
92 #define NVME_CC_REG_IOSQES_SHIFT                        (16)
93 #define NVME_CC_REG_IOSQES_MASK                         (0xF)
94 #define NVME_CC_REG_IOCQES_SHIFT                        (20)
95 #define NVME_CC_REG_IOCQES_MASK                         (0xF)
96
97 #define NVME_CSTS_REG_RDY_SHIFT                         (0)
98 #define NVME_CSTS_REG_RDY_MASK                          (0x1)
99 #define NVME_CSTS_REG_CFS_SHIFT                         (1)
100 #define NVME_CSTS_REG_CFS_MASK                          (0x1)
101 #define NVME_CSTS_REG_SHST_SHIFT                        (2)
102 #define NVME_CSTS_REG_SHST_MASK                         (0x3)
103
104 #define NVME_CSTS_GET_SHST(csts)                        (((csts) >> NVME_CSTS_REG_SHST_SHIFT) & NVME_CSTS_REG_SHST_MASK)
105
106 #define NVME_AQA_REG_ASQS_SHIFT                         (0)
107 #define NVME_AQA_REG_ASQS_MASK                          (0xFFF)
108 #define NVME_AQA_REG_ACQS_SHIFT                         (16)
109 #define NVME_AQA_REG_ACQS_MASK                          (0xFFF)
110
111 /* Command field definitions */
112
113 #define NVME_CMD_OPC_SHIFT                              (0)
114 #define NVME_CMD_OPC_MASK                               (0xFF)
115 #define NVME_CMD_FUSE_SHIFT                             (8)
116 #define NVME_CMD_FUSE_MASK                              (0x3)
117
118 #define NVME_CMD_SET_OPC(opc)                           (htole16(((uint16_t)(opc) & NVME_CMD_OPC_MASK) << NVME_CMD_OPC_SHIFT))
119
120 #define NVME_STATUS_P_SHIFT                             (0)
121 #define NVME_STATUS_P_MASK                              (0x1)
122 #define NVME_STATUS_SC_SHIFT                            (1)
123 #define NVME_STATUS_SC_MASK                             (0xFF)
124 #define NVME_STATUS_SCT_SHIFT                           (9)
125 #define NVME_STATUS_SCT_MASK                            (0x7)
126 #define NVME_STATUS_M_SHIFT                             (14)
127 #define NVME_STATUS_M_MASK                              (0x1)
128 #define NVME_STATUS_DNR_SHIFT                           (15)
129 #define NVME_STATUS_DNR_MASK                            (0x1)
130
131 #define NVME_STATUS_GET_P(st)                           (((st) >> NVME_STATUS_P_SHIFT) & NVME_STATUS_P_MASK)
132 #define NVME_STATUS_GET_SC(st)                          (((st) >> NVME_STATUS_SC_SHIFT) & NVME_STATUS_SC_MASK)
133 #define NVME_STATUS_GET_SCT(st)                         (((st) >> NVME_STATUS_SCT_SHIFT) & NVME_STATUS_SCT_MASK)
134 #define NVME_STATUS_GET_M(st)                           (((st) >> NVME_STATUS_M_SHIFT) & NVME_STATUS_M_MASK)
135 #define NVME_STATUS_GET_DNR(st)                         (((st) >> NVME_STATUS_DNR_SHIFT) & NVME_STATUS_DNR_MASK)
136
137 #define NVME_PWR_ST_MPS_SHIFT                           (0)
138 #define NVME_PWR_ST_MPS_MASK                            (0x1)
139 #define NVME_PWR_ST_NOPS_SHIFT                          (1)
140 #define NVME_PWR_ST_NOPS_MASK                           (0x1)
141 #define NVME_PWR_ST_RRT_SHIFT                           (0)
142 #define NVME_PWR_ST_RRT_MASK                            (0x1F)
143 #define NVME_PWR_ST_RRL_SHIFT                           (0)
144 #define NVME_PWR_ST_RRL_MASK                            (0x1F)
145 #define NVME_PWR_ST_RWT_SHIFT                           (0)
146 #define NVME_PWR_ST_RWT_MASK                            (0x1F)
147 #define NVME_PWR_ST_RWL_SHIFT                           (0)
148 #define NVME_PWR_ST_RWL_MASK                            (0x1F)
149 #define NVME_PWR_ST_IPS_SHIFT                           (6)
150 #define NVME_PWR_ST_IPS_MASK                            (0x3)
151 #define NVME_PWR_ST_APW_SHIFT                           (0)
152 #define NVME_PWR_ST_APW_MASK                            (0x7)
153 #define NVME_PWR_ST_APS_SHIFT                           (6)
154 #define NVME_PWR_ST_APS_MASK                            (0x3)
155
156 /** Controller Multi-path I/O and Namespace Sharing Capabilities */
157 /* More then one port */
158 #define NVME_CTRLR_DATA_MIC_MPORTS_SHIFT                (0)
159 #define NVME_CTRLR_DATA_MIC_MPORTS_MASK                 (0x1)
160 /* More then one controller */
161 #define NVME_CTRLR_DATA_MIC_MCTRLRS_SHIFT               (1)
162 #define NVME_CTRLR_DATA_MIC_MCTRLRS_MASK                (0x1)
163 /* SR-IOV Virtual Function */
164 #define NVME_CTRLR_DATA_MIC_SRIOVVF_SHIFT               (2)
165 #define NVME_CTRLR_DATA_MIC_SRIOVVF_MASK                (0x1)
166
167 /** OACS - optional admin command support */
168 /* supports security send/receive commands */
169 #define NVME_CTRLR_DATA_OACS_SECURITY_SHIFT             (0)
170 #define NVME_CTRLR_DATA_OACS_SECURITY_MASK              (0x1)
171 /* supports format nvm command */
172 #define NVME_CTRLR_DATA_OACS_FORMAT_SHIFT               (1)
173 #define NVME_CTRLR_DATA_OACS_FORMAT_MASK                (0x1)
174 /* supports firmware activate/download commands */
175 #define NVME_CTRLR_DATA_OACS_FIRMWARE_SHIFT             (2)
176 #define NVME_CTRLR_DATA_OACS_FIRMWARE_MASK              (0x1)
177 /* supports namespace management commands */
178 #define NVME_CTRLR_DATA_OACS_NSMGMT_SHIFT               (3)
179 #define NVME_CTRLR_DATA_OACS_NSMGMT_MASK                (0x1)
180 /* supports Device Self-test command */
181 #define NVME_CTRLR_DATA_OACS_SELFTEST_SHIFT             (4)
182 #define NVME_CTRLR_DATA_OACS_SELFTEST_MASK              (0x1)
183 /* supports Directives */
184 #define NVME_CTRLR_DATA_OACS_DIRECTIVES_SHIFT           (5)
185 #define NVME_CTRLR_DATA_OACS_DIRECTIVES_MASK            (0x1)
186 /* supports NVMe-MI Send/Receive */
187 #define NVME_CTRLR_DATA_OACS_NVMEMI_SHIFT               (6)
188 #define NVME_CTRLR_DATA_OACS_NVMEMI_MASK                (0x1)
189 /* supports Virtualization Management */
190 #define NVME_CTRLR_DATA_OACS_VM_SHIFT                   (7)
191 #define NVME_CTRLR_DATA_OACS_VM_MASK                    (0x1)
192 /* supports Doorbell Buffer Config */
193 #define NVME_CTRLR_DATA_OACS_DBBUFFER_SHIFT             (8)
194 #define NVME_CTRLR_DATA_OACS_DBBUFFER_MASK              (0x1)
195
196 /** firmware updates */
197 /* first slot is read-only */
198 #define NVME_CTRLR_DATA_FRMW_SLOT1_RO_SHIFT             (0)
199 #define NVME_CTRLR_DATA_FRMW_SLOT1_RO_MASK              (0x1)
200 /* number of firmware slots */
201 #define NVME_CTRLR_DATA_FRMW_NUM_SLOTS_SHIFT            (1)
202 #define NVME_CTRLR_DATA_FRMW_NUM_SLOTS_MASK             (0x7)
203
204 /** log page attributes */
205 /* per namespace smart/health log page */
206 #define NVME_CTRLR_DATA_LPA_NS_SMART_SHIFT              (0)
207 #define NVME_CTRLR_DATA_LPA_NS_SMART_MASK               (0x1)
208
209 /** AVSCC - admin vendor specific command configuration */
210 /* admin vendor specific commands use spec format */
211 #define NVME_CTRLR_DATA_AVSCC_SPEC_FORMAT_SHIFT         (0)
212 #define NVME_CTRLR_DATA_AVSCC_SPEC_FORMAT_MASK          (0x1)
213
214 /** Autonomous Power State Transition Attributes */
215 /* Autonomous Power State Transitions supported */
216 #define NVME_CTRLR_DATA_APSTA_APST_SUPP_SHIFT           (0)
217 #define NVME_CTRLR_DATA_APSTA_APST_SUPP_MASK            (0x1)
218
219 /** submission queue entry size */
220 #define NVME_CTRLR_DATA_SQES_MIN_SHIFT                  (0)
221 #define NVME_CTRLR_DATA_SQES_MIN_MASK                   (0xF)
222 #define NVME_CTRLR_DATA_SQES_MAX_SHIFT                  (4)
223 #define NVME_CTRLR_DATA_SQES_MAX_MASK                   (0xF)
224
225 /** completion queue entry size */
226 #define NVME_CTRLR_DATA_CQES_MIN_SHIFT                  (0)
227 #define NVME_CTRLR_DATA_CQES_MIN_MASK                   (0xF)
228 #define NVME_CTRLR_DATA_CQES_MAX_SHIFT                  (4)
229 #define NVME_CTRLR_DATA_CQES_MAX_MASK                   (0xF)
230
231 /** optional nvm command support */
232 #define NVME_CTRLR_DATA_ONCS_COMPARE_SHIFT              (0)
233 #define NVME_CTRLR_DATA_ONCS_COMPARE_MASK               (0x1)
234 #define NVME_CTRLR_DATA_ONCS_WRITE_UNC_SHIFT            (1)
235 #define NVME_CTRLR_DATA_ONCS_WRITE_UNC_MASK             (0x1)
236 #define NVME_CTRLR_DATA_ONCS_DSM_SHIFT                  (2)
237 #define NVME_CTRLR_DATA_ONCS_DSM_MASK                   (0x1)
238 #define NVME_CTRLR_DATA_ONCS_WRZERO_SHIFT               (3)
239 #define NVME_CTRLR_DATA_ONCS_WRZERO_MASK                (0x1)
240 #define NVME_CTRLR_DATA_ONCS_SAVEFEAT_SHIFT             (4)
241 #define NVME_CTRLR_DATA_ONCS_SAVEFEAT_MASK              (0x1)
242 #define NVME_CTRLR_DATA_ONCS_RESERV_SHIFT               (5)
243 #define NVME_CTRLR_DATA_ONCS_RESERV_MASK                (0x1)
244 #define NVME_CTRLR_DATA_ONCS_TIMESTAMP_SHIFT            (6)
245 #define NVME_CTRLR_DATA_ONCS_TIMESTAMP_MASK             (0x1)
246
247 /** Fused Operation Support */
248 #define NVME_CTRLR_DATA_FUSES_CNW_SHIFT         (0)
249 #define NVME_CTRLR_DATA_FUSES_CNW_MASK          (0x1)
250
251 /** Format NVM Attributes */
252 #define NVME_CTRLR_DATA_FNA_FORMAT_ALL_SHIFT            (0)
253 #define NVME_CTRLR_DATA_FNA_FORMAT_ALL_MASK             (0x1)
254 #define NVME_CTRLR_DATA_FNA_ERASE_ALL_SHIFT             (1)
255 #define NVME_CTRLR_DATA_FNA_ERASE_ALL_MASK              (0x1)
256 #define NVME_CTRLR_DATA_FNA_CRYPTO_ERASE_SHIFT          (2)
257 #define NVME_CTRLR_DATA_FNA_CRYPTO_ERASE_MASK           (0x1)
258
259 /** volatile write cache */
260 #define NVME_CTRLR_DATA_VWC_PRESENT_SHIFT               (0)
261 #define NVME_CTRLR_DATA_VWC_PRESENT_MASK                (0x1)
262
263 /** namespace features */
264 /* thin provisioning */
265 #define NVME_NS_DATA_NSFEAT_THIN_PROV_SHIFT             (0)
266 #define NVME_NS_DATA_NSFEAT_THIN_PROV_MASK              (0x1)
267 /* NAWUN, NAWUPF, and NACWU fields are valid */
268 #define NVME_NS_DATA_NSFEAT_NA_FIELDS_SHIFT             (1)
269 #define NVME_NS_DATA_NSFEAT_NA_FIELDS_MASK              (0x1)
270 /* Deallocated or Unwritten Logical Block errors supported */
271 #define NVME_NS_DATA_NSFEAT_DEALLOC_SHIFT               (2)
272 #define NVME_NS_DATA_NSFEAT_DEALLOC_MASK                (0x1)
273 /* NGUID and EUI64 fields are not reusable */
274 #define NVME_NS_DATA_NSFEAT_NO_ID_REUSE_SHIFT           (3)
275 #define NVME_NS_DATA_NSFEAT_NO_ID_REUSE_MASK            (0x1)
276
277 /** formatted lba size */
278 #define NVME_NS_DATA_FLBAS_FORMAT_SHIFT                 (0)
279 #define NVME_NS_DATA_FLBAS_FORMAT_MASK                  (0xF)
280 #define NVME_NS_DATA_FLBAS_EXTENDED_SHIFT               (4)
281 #define NVME_NS_DATA_FLBAS_EXTENDED_MASK                (0x1)
282
283 /** metadata capabilities */
284 /* metadata can be transferred as part of data prp list */
285 #define NVME_NS_DATA_MC_EXTENDED_SHIFT                  (0)
286 #define NVME_NS_DATA_MC_EXTENDED_MASK                   (0x1)
287 /* metadata can be transferred with separate metadata pointer */
288 #define NVME_NS_DATA_MC_POINTER_SHIFT                   (1)
289 #define NVME_NS_DATA_MC_POINTER_MASK                    (0x1)
290
291 /** end-to-end data protection capabilities */
292 /* protection information type 1 */
293 #define NVME_NS_DATA_DPC_PIT1_SHIFT                     (0)
294 #define NVME_NS_DATA_DPC_PIT1_MASK                      (0x1)
295 /* protection information type 2 */
296 #define NVME_NS_DATA_DPC_PIT2_SHIFT                     (1)
297 #define NVME_NS_DATA_DPC_PIT2_MASK                      (0x1)
298 /* protection information type 3 */
299 #define NVME_NS_DATA_DPC_PIT3_SHIFT                     (2)
300 #define NVME_NS_DATA_DPC_PIT3_MASK                      (0x1)
301 /* first eight bytes of metadata */
302 #define NVME_NS_DATA_DPC_MD_START_SHIFT                 (3)
303 #define NVME_NS_DATA_DPC_MD_START_MASK                  (0x1)
304 /* last eight bytes of metadata */
305 #define NVME_NS_DATA_DPC_MD_END_SHIFT                   (4)
306 #define NVME_NS_DATA_DPC_MD_END_MASK                    (0x1)
307
308 /** end-to-end data protection type settings */
309 /* protection information type */
310 #define NVME_NS_DATA_DPS_PIT_SHIFT                      (0)
311 #define NVME_NS_DATA_DPS_PIT_MASK                       (0x7)
312 /* 1 == protection info transferred at start of metadata */
313 /* 0 == protection info transferred at end of metadata */
314 #define NVME_NS_DATA_DPS_MD_START_SHIFT                 (3)
315 #define NVME_NS_DATA_DPS_MD_START_MASK                  (0x1)
316
317 /** Namespace Multi-path I/O and Namespace Sharing Capabilities */
318 /* the namespace may be attached to two or more controllers */
319 #define NVME_NS_DATA_NMIC_MAY_BE_SHARED_SHIFT           (0)
320 #define NVME_NS_DATA_NMIC_MAY_BE_SHARED_MASK            (0x1)
321
322 /** Reservation Capabilities */
323 /* Persist Through Power Loss */
324 #define NVME_NS_DATA_RESCAP_PTPL_SHIFT          (0)
325 #define NVME_NS_DATA_RESCAP_PTPL_MASK           (0x1)
326 /* supports the Write Exclusive */
327 #define NVME_NS_DATA_RESCAP_WR_EX_SHIFT         (1)
328 #define NVME_NS_DATA_RESCAP_WR_EX_MASK          (0x1)
329 /* supports the Exclusive Access */
330 #define NVME_NS_DATA_RESCAP_EX_AC_SHIFT         (2)
331 #define NVME_NS_DATA_RESCAP_EX_AC_MASK          (0x1)
332 /* supports the Write Exclusive â€“ Registrants Only */
333 #define NVME_NS_DATA_RESCAP_WR_EX_RO_SHIFT      (3)
334 #define NVME_NS_DATA_RESCAP_WR_EX_RO_MASK       (0x1)
335 /* supports the Exclusive Access - Registrants Only */
336 #define NVME_NS_DATA_RESCAP_EX_AC_RO_SHIFT      (4)
337 #define NVME_NS_DATA_RESCAP_EX_AC_RO_MASK       (0x1)
338 /* supports the Write Exclusive â€“ All Registrants */
339 #define NVME_NS_DATA_RESCAP_WR_EX_AR_SHIFT      (5)
340 #define NVME_NS_DATA_RESCAP_WR_EX_AR_MASK       (0x1)
341 /* supports the Exclusive Access - All Registrants */
342 #define NVME_NS_DATA_RESCAP_EX_AC_AR_SHIFT      (6)
343 #define NVME_NS_DATA_RESCAP_EX_AC_AR_MASK       (0x1)
344 /* Ignore Existing Key is used as defined in revision 1.3 or later */
345 #define NVME_NS_DATA_RESCAP_IEKEY13_SHIFT       (7)
346 #define NVME_NS_DATA_RESCAP_IEKEY13_MASK        (0x1)
347
348 /** Format Progress Indicator */
349 /* percentage of the Format NVM command that remains to be completed */
350 #define NVME_NS_DATA_FPI_PERC_SHIFT             (0)
351 #define NVME_NS_DATA_FPI_PERC_MASK              (0x7f)
352 /* namespace supports the Format Progress Indicator */
353 #define NVME_NS_DATA_FPI_SUPP_SHIFT             (7)
354 #define NVME_NS_DATA_FPI_SUPP_MASK              (0x1)
355
356 /** lba format support */
357 /* metadata size */
358 #define NVME_NS_DATA_LBAF_MS_SHIFT                      (0)
359 #define NVME_NS_DATA_LBAF_MS_MASK                       (0xFFFF)
360 /* lba data size */
361 #define NVME_NS_DATA_LBAF_LBADS_SHIFT                   (16)
362 #define NVME_NS_DATA_LBAF_LBADS_MASK                    (0xFF)
363 /* relative performance */
364 #define NVME_NS_DATA_LBAF_RP_SHIFT                      (24)
365 #define NVME_NS_DATA_LBAF_RP_MASK                       (0x3)
366
367 enum nvme_critical_warning_state {
368         NVME_CRIT_WARN_ST_AVAILABLE_SPARE               = 0x1,
369         NVME_CRIT_WARN_ST_TEMPERATURE                   = 0x2,
370         NVME_CRIT_WARN_ST_DEVICE_RELIABILITY            = 0x4,
371         NVME_CRIT_WARN_ST_READ_ONLY                     = 0x8,
372         NVME_CRIT_WARN_ST_VOLATILE_MEMORY_BACKUP        = 0x10,
373 };
374 #define NVME_CRIT_WARN_ST_RESERVED_MASK                 (0xE0)
375
376 /* slot for current FW */
377 #define NVME_FIRMWARE_PAGE_AFI_SLOT_SHIFT               (0)
378 #define NVME_FIRMWARE_PAGE_AFI_SLOT_MASK                (0x7)
379
380 /* CC register SHN field values */
381 enum shn_value {
382         NVME_SHN_NORMAL         = 0x1,
383         NVME_SHN_ABRUPT         = 0x2,
384 };
385
386 /* CSTS register SHST field values */
387 enum shst_value {
388         NVME_SHST_NORMAL        = 0x0,
389         NVME_SHST_OCCURRING     = 0x1,
390         NVME_SHST_COMPLETE      = 0x2,
391 };
392
393 struct nvme_registers
394 {
395         /** controller capabilities */
396         uint32_t                cap_lo;
397         uint32_t                cap_hi;
398
399         uint32_t                vs;     /* version */
400         uint32_t                intms;  /* interrupt mask set */
401         uint32_t                intmc;  /* interrupt mask clear */
402
403         /** controller configuration */
404         uint32_t                cc;
405
406         uint32_t                reserved1;
407
408         /** controller status */
409         uint32_t                csts;
410
411         uint32_t                reserved2;
412
413         /** admin queue attributes */
414         uint32_t                aqa;
415
416         uint64_t                asq;    /* admin submission queue base addr */
417         uint64_t                acq;    /* admin completion queue base addr */
418         uint32_t                reserved3[0x3f2];
419
420         struct {
421             uint32_t            sq_tdbl; /* submission queue tail doorbell */
422             uint32_t            cq_hdbl; /* completion queue head doorbell */
423         } doorbell[1] __packed;
424 } __packed;
425
426 _Static_assert(sizeof(struct nvme_registers) == 0x1008, "bad size for nvme_registers");
427
428 struct nvme_command
429 {
430         /* dword 0 */
431         uint16_t opc_fuse;      /* opcode, fused operation */
432         uint16_t cid;           /* command identifier */
433
434         /* dword 1 */
435         uint32_t nsid;          /* namespace identifier */
436
437         /* dword 2-3 */
438         uint32_t rsvd2;
439         uint32_t rsvd3;
440
441         /* dword 4-5 */
442         uint64_t mptr;          /* metadata pointer */
443
444         /* dword 6-7 */
445         uint64_t prp1;          /* prp entry 1 */
446
447         /* dword 8-9 */
448         uint64_t prp2;          /* prp entry 2 */
449
450         /* dword 10-15 */
451         uint32_t cdw10;         /* command-specific */
452         uint32_t cdw11;         /* command-specific */
453         uint32_t cdw12;         /* command-specific */
454         uint32_t cdw13;         /* command-specific */
455         uint32_t cdw14;         /* command-specific */
456         uint32_t cdw15;         /* command-specific */
457 } __packed;
458
459 _Static_assert(sizeof(struct nvme_command) == 16 * 4, "bad size for nvme_command");
460
461 struct nvme_completion {
462
463         /* dword 0 */
464         uint32_t                cdw0;   /* command-specific */
465
466         /* dword 1 */
467         uint32_t                rsvd1;
468
469         /* dword 2 */
470         uint16_t                sqhd;   /* submission queue head pointer */
471         uint16_t                sqid;   /* submission queue identifier */
472
473         /* dword 3 */
474         uint16_t                cid;    /* command identifier */
475         uint16_t                status;
476 } __packed;
477
478 _Static_assert(sizeof(struct nvme_completion) == 4 * 4, "bad size for nvme_completion");
479
480 struct nvme_dsm_range {
481         uint32_t attributes;
482         uint32_t length;
483         uint64_t starting_lba;
484 } __packed;
485
486 /* Largest DSM Trim that can be done */
487 #define NVME_MAX_DSM_TRIM               4096
488
489 _Static_assert(sizeof(struct nvme_dsm_range) == 16, "bad size for nvme_dsm_ranage");
490
491 /* status code types */
492 enum nvme_status_code_type {
493         NVME_SCT_GENERIC                = 0x0,
494         NVME_SCT_COMMAND_SPECIFIC       = 0x1,
495         NVME_SCT_MEDIA_ERROR            = 0x2,
496         /* 0x3-0x6 - reserved */
497         NVME_SCT_VENDOR_SPECIFIC        = 0x7,
498 };
499
500 /* generic command status codes */
501 enum nvme_generic_command_status_code {
502         NVME_SC_SUCCESS                         = 0x00,
503         NVME_SC_INVALID_OPCODE                  = 0x01,
504         NVME_SC_INVALID_FIELD                   = 0x02,
505         NVME_SC_COMMAND_ID_CONFLICT             = 0x03,
506         NVME_SC_DATA_TRANSFER_ERROR             = 0x04,
507         NVME_SC_ABORTED_POWER_LOSS              = 0x05,
508         NVME_SC_INTERNAL_DEVICE_ERROR           = 0x06,
509         NVME_SC_ABORTED_BY_REQUEST              = 0x07,
510         NVME_SC_ABORTED_SQ_DELETION             = 0x08,
511         NVME_SC_ABORTED_FAILED_FUSED            = 0x09,
512         NVME_SC_ABORTED_MISSING_FUSED           = 0x0a,
513         NVME_SC_INVALID_NAMESPACE_OR_FORMAT     = 0x0b,
514         NVME_SC_COMMAND_SEQUENCE_ERROR          = 0x0c,
515         NVME_SC_INVALID_SGL_SEGMENT_DESCR       = 0x0d,
516         NVME_SC_INVALID_NUMBER_OF_SGL_DESCR     = 0x0e,
517         NVME_SC_DATA_SGL_LENGTH_INVALID         = 0x0f,
518         NVME_SC_METADATA_SGL_LENGTH_INVALID     = 0x10,
519         NVME_SC_SGL_DESCRIPTOR_TYPE_INVALID     = 0x11,
520         NVME_SC_INVALID_USE_OF_CMB              = 0x12,
521         NVME_SC_PRP_OFFET_INVALID               = 0x13,
522         NVME_SC_ATOMIC_WRITE_UNIT_EXCEEDED      = 0x14,
523         NVME_SC_OPERATION_DENIED                = 0x15,
524         NVME_SC_SGL_OFFSET_INVALID              = 0x16,
525         /* 0x17 - reserved */
526         NVME_SC_HOST_ID_INCONSISTENT_FORMAT     = 0x18,
527         NVME_SC_KEEP_ALIVE_TIMEOUT_EXPIRED      = 0x19,
528         NVME_SC_KEEP_ALIVE_TIMEOUT_INVALID      = 0x1a,
529         NVME_SC_ABORTED_DUE_TO_PREEMPT          = 0x1b,
530         NVME_SC_SANITIZE_FAILED                 = 0x1c,
531         NVME_SC_SANITIZE_IN_PROGRESS            = 0x1d,
532         NVME_SC_SGL_DATA_BLOCK_GRAN_INVALID     = 0x1e,
533         NVME_SC_NOT_SUPPORTED_IN_CMB            = 0x1f,
534
535         NVME_SC_LBA_OUT_OF_RANGE                = 0x80,
536         NVME_SC_CAPACITY_EXCEEDED               = 0x81,
537         NVME_SC_NAMESPACE_NOT_READY             = 0x82,
538         NVME_SC_RESERVATION_CONFLICT            = 0x83,
539         NVME_SC_FORMAT_IN_PROGRESS              = 0x84,
540 };
541
542 /* command specific status codes */
543 enum nvme_command_specific_status_code {
544         NVME_SC_COMPLETION_QUEUE_INVALID        = 0x00,
545         NVME_SC_INVALID_QUEUE_IDENTIFIER        = 0x01,
546         NVME_SC_MAXIMUM_QUEUE_SIZE_EXCEEDED     = 0x02,
547         NVME_SC_ABORT_COMMAND_LIMIT_EXCEEDED    = 0x03,
548         /* 0x04 - reserved */
549         NVME_SC_ASYNC_EVENT_REQUEST_LIMIT_EXCEEDED = 0x05,
550         NVME_SC_INVALID_FIRMWARE_SLOT           = 0x06,
551         NVME_SC_INVALID_FIRMWARE_IMAGE          = 0x07,
552         NVME_SC_INVALID_INTERRUPT_VECTOR        = 0x08,
553         NVME_SC_INVALID_LOG_PAGE                = 0x09,
554         NVME_SC_INVALID_FORMAT                  = 0x0a,
555         NVME_SC_FIRMWARE_REQUIRES_RESET         = 0x0b,
556         NVME_SC_INVALID_QUEUE_DELETION          = 0x0c,
557         NVME_SC_FEATURE_NOT_SAVEABLE            = 0x0d,
558         NVME_SC_FEATURE_NOT_CHANGEABLE          = 0x0e,
559         NVME_SC_FEATURE_NOT_NS_SPECIFIC         = 0x0f,
560         NVME_SC_FW_ACT_REQUIRES_NVMS_RESET      = 0x10,
561         NVME_SC_FW_ACT_REQUIRES_RESET           = 0x11,
562         NVME_SC_FW_ACT_REQUIRES_TIME            = 0x12,
563         NVME_SC_FW_ACT_PROHIBITED               = 0x13,
564         NVME_SC_OVERLAPPING_RANGE               = 0x14,
565         NVME_SC_NS_INSUFFICIENT_CAPACITY        = 0x15,
566         NVME_SC_NS_ID_UNAVAILABLE               = 0x16,
567         /* 0x17 - reserved */
568         NVME_SC_NS_ALREADY_ATTACHED             = 0x18,
569         NVME_SC_NS_IS_PRIVATE                   = 0x19,
570         NVME_SC_NS_NOT_ATTACHED                 = 0x1a,
571         NVME_SC_THIN_PROV_NOT_SUPPORTED         = 0x1b,
572         NVME_SC_CTRLR_LIST_INVALID              = 0x1c,
573         NVME_SC_SELT_TEST_IN_PROGRESS           = 0x1d,
574         NVME_SC_BOOT_PART_WRITE_PROHIB          = 0x1e,
575         NVME_SC_INVALID_CTRLR_ID                = 0x1f,
576         NVME_SC_INVALID_SEC_CTRLR_STATE         = 0x20,
577         NVME_SC_INVALID_NUM_OF_CTRLR_RESRC      = 0x21,
578         NVME_SC_INVALID_RESOURCE_ID             = 0x22,
579
580         NVME_SC_CONFLICTING_ATTRIBUTES          = 0x80,
581         NVME_SC_INVALID_PROTECTION_INFO         = 0x81,
582         NVME_SC_ATTEMPTED_WRITE_TO_RO_PAGE      = 0x82,
583 };
584
585 /* media error status codes */
586 enum nvme_media_error_status_code {
587         NVME_SC_WRITE_FAULTS                    = 0x80,
588         NVME_SC_UNRECOVERED_READ_ERROR          = 0x81,
589         NVME_SC_GUARD_CHECK_ERROR               = 0x82,
590         NVME_SC_APPLICATION_TAG_CHECK_ERROR     = 0x83,
591         NVME_SC_REFERENCE_TAG_CHECK_ERROR       = 0x84,
592         NVME_SC_COMPARE_FAILURE                 = 0x85,
593         NVME_SC_ACCESS_DENIED                   = 0x86,
594         NVME_SC_DEALLOCATED_OR_UNWRITTEN        = 0x87,
595 };
596
597 /* admin opcodes */
598 enum nvme_admin_opcode {
599         NVME_OPC_DELETE_IO_SQ                   = 0x00,
600         NVME_OPC_CREATE_IO_SQ                   = 0x01,
601         NVME_OPC_GET_LOG_PAGE                   = 0x02,
602         /* 0x03 - reserved */
603         NVME_OPC_DELETE_IO_CQ                   = 0x04,
604         NVME_OPC_CREATE_IO_CQ                   = 0x05,
605         NVME_OPC_IDENTIFY                       = 0x06,
606         /* 0x07 - reserved */
607         NVME_OPC_ABORT                          = 0x08,
608         NVME_OPC_SET_FEATURES                   = 0x09,
609         NVME_OPC_GET_FEATURES                   = 0x0a,
610         /* 0x0b - reserved */
611         NVME_OPC_ASYNC_EVENT_REQUEST            = 0x0c,
612         NVME_OPC_NAMESPACE_MANAGEMENT           = 0x0d,
613         /* 0x0e-0x0f - reserved */
614         NVME_OPC_FIRMWARE_ACTIVATE              = 0x10,
615         NVME_OPC_FIRMWARE_IMAGE_DOWNLOAD        = 0x11,
616         NVME_OPC_DEVICE_SELF_TEST               = 0x14,
617         NVME_OPC_NAMESPACE_ATTACHMENT           = 0x15,
618         NVME_OPC_KEEP_ALIVE                     = 0x18,
619         NVME_OPC_DIRECTIVE_SEND                 = 0x19,
620         NVME_OPC_DIRECTIVE_RECEIVE              = 0x1a,
621         NVME_OPC_VIRTUALIZATION_MANAGEMENT      = 0x1c,
622         NVME_OPC_NVME_MI_SEND                   = 0x1d,
623         NVME_OPC_NVME_MI_RECEIVE                = 0x1e,
624         NVME_OPC_DOORBELL_BUFFER_CONFIG         = 0x7c,
625
626         NVME_OPC_FORMAT_NVM                     = 0x80,
627         NVME_OPC_SECURITY_SEND                  = 0x81,
628         NVME_OPC_SECURITY_RECEIVE               = 0x82,
629         NVME_OPC_SANITIZE                       = 0x84,
630 };
631
632 /* nvme nvm opcodes */
633 enum nvme_nvm_opcode {
634         NVME_OPC_FLUSH                          = 0x00,
635         NVME_OPC_WRITE                          = 0x01,
636         NVME_OPC_READ                           = 0x02,
637         /* 0x03 - reserved */
638         NVME_OPC_WRITE_UNCORRECTABLE            = 0x04,
639         NVME_OPC_COMPARE                        = 0x05,
640         /* 0x06 - reserved */
641         NVME_OPC_WRITE_ZEROES                   = 0x08,
642         /* 0x07 - reserved */
643         NVME_OPC_DATASET_MANAGEMENT             = 0x09,
644         /* 0x0a-0x0c - reserved */
645         NVME_OPC_RESERVATION_REGISTER           = 0x0d,
646         NVME_OPC_RESERVATION_REPORT             = 0x0e,
647         /* 0x0f-0x10 - reserved */
648         NVME_OPC_RESERVATION_ACQUIRE            = 0x11,
649         /* 0x12-0x14 - reserved */
650         NVME_OPC_RESERVATION_RELEASE            = 0x15,
651 };
652
653 enum nvme_feature {
654         /* 0x00 - reserved */
655         NVME_FEAT_ARBITRATION                   = 0x01,
656         NVME_FEAT_POWER_MANAGEMENT              = 0x02,
657         NVME_FEAT_LBA_RANGE_TYPE                = 0x03,
658         NVME_FEAT_TEMPERATURE_THRESHOLD         = 0x04,
659         NVME_FEAT_ERROR_RECOVERY                = 0x05,
660         NVME_FEAT_VOLATILE_WRITE_CACHE          = 0x06,
661         NVME_FEAT_NUMBER_OF_QUEUES              = 0x07,
662         NVME_FEAT_INTERRUPT_COALESCING          = 0x08,
663         NVME_FEAT_INTERRUPT_VECTOR_CONFIGURATION = 0x09,
664         NVME_FEAT_WRITE_ATOMICITY               = 0x0A,
665         NVME_FEAT_ASYNC_EVENT_CONFIGURATION     = 0x0B,
666         NVME_FEAT_AUTONOMOUS_POWER_STATE_TRANSITION = 0x0C,
667         NVME_FEAT_HOST_MEMORY_BUFFER            = 0x0D,
668         NVME_FEAT_TIMESTAMP                     = 0x0E,
669         NVME_FEAT_KEEP_ALIVE_TIMER              = 0x0F,
670         NVME_FEAT_HOST_CONTROLLED_THERMAL_MGMT  = 0x10,
671         NVME_FEAT_NON_OP_POWER_STATE_CONFIG     = 0x11,
672         /* 0x12-0x77 - reserved */
673         /* 0x78-0x7f - NVMe Management Interface */
674         NVME_FEAT_SOFTWARE_PROGRESS_MARKER      = 0x80,
675         /* 0x81-0xBF - command set specific (reserved) */
676         /* 0xC0-0xFF - vendor specific */
677 };
678
679 enum nvme_dsm_attribute {
680         NVME_DSM_ATTR_INTEGRAL_READ             = 0x1,
681         NVME_DSM_ATTR_INTEGRAL_WRITE            = 0x2,
682         NVME_DSM_ATTR_DEALLOCATE                = 0x4,
683 };
684
685 enum nvme_activate_action {
686         NVME_AA_REPLACE_NO_ACTIVATE             = 0x0,
687         NVME_AA_REPLACE_ACTIVATE                = 0x1,
688         NVME_AA_ACTIVATE                        = 0x2,
689 };
690
691 struct nvme_power_state {
692         /** Maximum Power */
693         uint16_t        mp;                     /* Maximum Power */
694         uint8_t         ps_rsvd1;
695         uint8_t         mps_nops;               /* Max Power Scale, Non-Operational State */
696
697         uint32_t        enlat;                  /* Entry Latency */
698         uint32_t        exlat;                  /* Exit Latency */
699
700         uint8_t         rrt;                    /* Relative Read Throughput */
701         uint8_t         rrl;                    /* Relative Read Latency */
702         uint8_t         rwt;                    /* Relative Write Throughput */
703         uint8_t         rwl;                    /* Relative Write Latency */
704
705         uint16_t        idlp;                   /* Idle Power */
706         uint8_t         ips;                    /* Idle Power Scale */
707         uint8_t         ps_rsvd8;
708
709         uint16_t        actp;                   /* Active Power */
710         uint8_t         apw_aps;                /* Active Power Workload, Active Power Scale */
711         uint8_t         ps_rsvd10[9];
712 } __packed;
713
714 _Static_assert(sizeof(struct nvme_power_state) == 32, "bad size for nvme_power_state");
715
716 #define NVME_SERIAL_NUMBER_LENGTH       20
717 #define NVME_MODEL_NUMBER_LENGTH        40
718 #define NVME_FIRMWARE_REVISION_LENGTH   8
719
720 struct nvme_controller_data {
721
722         /* bytes 0-255: controller capabilities and features */
723
724         /** pci vendor id */
725         uint16_t                vid;
726
727         /** pci subsystem vendor id */
728         uint16_t                ssvid;
729
730         /** serial number */
731         uint8_t                 sn[NVME_SERIAL_NUMBER_LENGTH];
732
733         /** model number */
734         uint8_t                 mn[NVME_MODEL_NUMBER_LENGTH];
735
736         /** firmware revision */
737         uint8_t                 fr[NVME_FIRMWARE_REVISION_LENGTH];
738
739         /** recommended arbitration burst */
740         uint8_t                 rab;
741
742         /** ieee oui identifier */
743         uint8_t                 ieee[3];
744
745         /** multi-interface capabilities */
746         uint8_t                 mic;
747
748         /** maximum data transfer size */
749         uint8_t                 mdts;
750
751         /** Controller ID */
752         uint16_t                ctrlr_id;
753
754         /** Version */
755         uint32_t                ver;
756
757         /** RTD3 Resume Latency */
758         uint32_t                rtd3r;
759
760         /** RTD3 Enter Latency */
761         uint32_t                rtd3e;
762
763         /** Optional Asynchronous Events Supported */
764         uint32_t                oaes;   /* bitfield really */
765
766         /** Controller Attributes */
767         uint32_t                ctratt; /* bitfield really */
768
769         uint8_t                 reserved1[12];
770
771         /** FRU Globally Unique Identifier */
772         uint8_t                 fguid[16];
773
774         uint8_t                 reserved2[128];
775
776         /* bytes 256-511: admin command set attributes */
777
778         /** optional admin command support */
779         uint16_t                oacs;
780
781         /** abort command limit */
782         uint8_t                 acl;
783
784         /** asynchronous event request limit */
785         uint8_t                 aerl;
786
787         /** firmware updates */
788         uint8_t                 frmw;
789
790         /** log page attributes */
791         uint8_t                 lpa;
792
793         /** error log page entries */
794         uint8_t                 elpe;
795
796         /** number of power states supported */
797         uint8_t                 npss;
798
799         /** admin vendor specific command configuration */
800         uint8_t                 avscc;
801
802         /** Autonomous Power State Transition Attributes */
803         uint8_t                 apsta;
804
805         /** Warning Composite Temperature Threshold */
806         uint16_t                wctemp;
807
808         /** Critical Composite Temperature Threshold */
809         uint16_t                cctemp;
810
811         /** Maximum Time for Firmware Activation */
812         uint16_t                mtfa;
813
814         /** Host Memory Buffer Preferred Size */
815         uint32_t                hmpre;
816
817         /** Host Memory Buffer Minimum Size */
818         uint32_t                hmmin;
819
820         /** Name space capabilities  */
821         struct {
822                 /* if nsmgmt, report tnvmcap and unvmcap */
823                 uint8_t    tnvmcap[16];
824                 uint8_t    unvmcap[16];
825         } __packed untncap;
826
827         /** Replay Protected Memory Block Support */
828         uint32_t                rpmbs; /* Really a bitfield */
829
830         /** Extended Device Self-test Time */
831         uint16_t                edstt;
832
833         /** Device Self-test Options */
834         uint8_t                 dsto; /* Really a bitfield */
835
836         /** Firmware Update Granularity */
837         uint8_t                 fwug;
838
839         /** Keep Alive Support */
840         uint16_t                kas;
841
842         /** Host Controlled Thermal Management Attributes */
843         uint16_t                hctma; /* Really a bitfield */
844
845         /** Minimum Thermal Management Temperature */
846         uint16_t                mntmt;
847
848         /** Maximum Thermal Management Temperature */
849         uint16_t                mxtmt;
850
851         /** Sanitize Capabilities */
852         uint32_t                sanicap; /* Really a bitfield */
853
854         uint8_t                 reserved3[180];
855         /* bytes 512-703: nvm command set attributes */
856
857         /** submission queue entry size */
858         uint8_t                 sqes;
859
860         /** completion queue entry size */
861         uint8_t                 cqes;
862
863         /** Maximum Outstanding Commands */
864         uint16_t                maxcmd;
865
866         /** number of namespaces */
867         uint32_t                nn;
868
869         /** optional nvm command support */
870         uint16_t                oncs;
871
872         /** fused operation support */
873         uint16_t                fuses;
874
875         /** format nvm attributes */
876         uint8_t                 fna;
877
878         /** volatile write cache */
879         uint8_t                 vwc;
880
881         /** Atomic Write Unit Normal */
882         uint16_t                awun;
883
884         /** Atomic Write Unit Power Fail */
885         uint16_t                awupf;
886
887         /** NVM Vendor Specific Command Configuration */
888         uint8_t                 nvscc;
889         uint8_t                 reserved5;
890
891         /** Atomic Compare & Write Unit */
892         uint16_t                acwu;
893         uint16_t                reserved6;
894
895         /** SGL Support */
896         uint32_t                sgls;
897
898         /* bytes 540-767: Reserved */
899         uint8_t                 reserved7[228];
900
901         /** NVM Subsystem NVMe Qualified Name */
902         uint8_t                 subnqn[256];
903
904         /* bytes 1024-1791: Reserved */
905         uint8_t                 reserved8[768];
906
907         /* bytes 1792-2047: NVMe over Fabrics specification */
908         uint8_t                 reserved9[256];
909
910         /* bytes 2048-3071: power state descriptors */
911         struct nvme_power_state power_state[32];
912
913         /* bytes 3072-4095: vendor specific */
914         uint8_t                 vs[1024];
915 } __packed __aligned(4);
916
917 _Static_assert(sizeof(struct nvme_controller_data) == 4096, "bad size for nvme_controller_data");
918
919 struct nvme_namespace_data {
920
921         /** namespace size */
922         uint64_t                nsze;
923
924         /** namespace capacity */
925         uint64_t                ncap;
926
927         /** namespace utilization */
928         uint64_t                nuse;
929
930         /** namespace features */
931         uint8_t                 nsfeat;
932
933         /** number of lba formats */
934         uint8_t                 nlbaf;
935
936         /** formatted lba size */
937         uint8_t                 flbas;
938
939         /** metadata capabilities */
940         uint8_t                 mc;
941
942         /** end-to-end data protection capabilities */
943         uint8_t                 dpc;
944
945         /** end-to-end data protection type settings */
946         uint8_t                 dps;
947
948         /** Namespace Multi-path I/O and Namespace Sharing Capabilities */
949         uint8_t                 nmic;
950
951         /** Reservation Capabilities */
952         uint8_t                 rescap;
953
954         /** Format Progress Indicator */
955         uint8_t                 fpi;
956
957         /** Deallocate Logical Block Features */
958         uint8_t                 dlfeat;
959
960         /** Namespace Atomic Write Unit Normal  */
961         uint16_t                nawun;
962
963         /** Namespace Atomic Write Unit Power Fail */
964         uint16_t                nawupf;
965
966         /** Namespace Atomic Compare & Write Unit */
967         uint16_t                nacwu;
968
969         /** Namespace Atomic Boundary Size Normal */
970         uint16_t                nabsn;
971
972         /** Namespace Atomic Boundary Offset */
973         uint16_t                nabo;
974
975         /** Namespace Atomic Boundary Size Power Fail */
976         uint16_t                nabspf;
977
978         /** Namespace Optimal IO Boundary */
979         uint16_t                noiob;
980
981         /** NVM Capacity */
982         uint8_t                 nvmcap[16];
983
984         /* bytes 64-103: Reserved */
985         uint8_t                 reserved5[40];
986
987         /** Namespace Globally Unique Identifier */
988         uint8_t                 nguid[16];
989
990         /** IEEE Extended Unique Identifier */
991         uint8_t                 eui64[8];
992
993         /** lba format support */
994         uint32_t                lbaf[16];
995
996         uint8_t                 reserved6[192];
997
998         uint8_t                 vendor_specific[3712];
999 } __packed __aligned(4);
1000
1001 _Static_assert(sizeof(struct nvme_namespace_data) == 4096, "bad size for nvme_namepsace_data");
1002
1003 enum nvme_log_page {
1004
1005         /* 0x00 - reserved */
1006         NVME_LOG_ERROR                  = 0x01,
1007         NVME_LOG_HEALTH_INFORMATION     = 0x02,
1008         NVME_LOG_FIRMWARE_SLOT          = 0x03,
1009         NVME_LOG_CHANGED_NAMESPACE      = 0x04,
1010         NVME_LOG_COMMAND_EFFECT         = 0x05,
1011         /* 0x06-0x7F - reserved */
1012         /* 0x80-0xBF - I/O command set specific */
1013         NVME_LOG_RES_NOTIFICATION       = 0x80,
1014         /* 0xC0-0xFF - vendor specific */
1015
1016         /*
1017          * The following are Intel Specific log pages, but they seem
1018          * to be widely implemented.
1019          */
1020         INTEL_LOG_READ_LAT_LOG          = 0xc1,
1021         INTEL_LOG_WRITE_LAT_LOG         = 0xc2,
1022         INTEL_LOG_TEMP_STATS            = 0xc5,
1023         INTEL_LOG_ADD_SMART             = 0xca,
1024         INTEL_LOG_DRIVE_MKT_NAME        = 0xdd,
1025
1026         /*
1027          * HGST log page, with lots ofs sub pages.
1028          */
1029         HGST_INFO_LOG                   = 0xc1,
1030 };
1031
1032 struct nvme_error_information_entry {
1033
1034         uint64_t                error_count;
1035         uint16_t                sqid;
1036         uint16_t                cid;
1037         uint16_t                status;
1038         uint16_t                error_location;
1039         uint64_t                lba;
1040         uint32_t                nsid;
1041         uint8_t                 vendor_specific;
1042         uint8_t                 reserved[35];
1043 } __packed __aligned(4);
1044
1045 _Static_assert(sizeof(struct nvme_error_information_entry) == 64, "bad size for nvme_error_information_entry");
1046
1047 struct nvme_health_information_page {
1048
1049         uint8_t                 critical_warning;
1050         uint16_t                temperature;
1051         uint8_t                 available_spare;
1052         uint8_t                 available_spare_threshold;
1053         uint8_t                 percentage_used;
1054
1055         uint8_t                 reserved[26];
1056
1057         /*
1058          * Note that the following are 128-bit values, but are
1059          *  defined as an array of 2 64-bit values.
1060          */
1061         /* Data Units Read is always in 512-byte units. */
1062         uint64_t                data_units_read[2];
1063         /* Data Units Written is always in 512-byte units. */
1064         uint64_t                data_units_written[2];
1065         /* For NVM command set, this includes Compare commands. */
1066         uint64_t                host_read_commands[2];
1067         uint64_t                host_write_commands[2];
1068         /* Controller Busy Time is reported in minutes. */
1069         uint64_t                controller_busy_time[2];
1070         uint64_t                power_cycles[2];
1071         uint64_t                power_on_hours[2];
1072         uint64_t                unsafe_shutdowns[2];
1073         uint64_t                media_errors[2];
1074         uint64_t                num_error_info_log_entries[2];
1075         uint32_t                warning_temp_time;
1076         uint32_t                error_temp_time;
1077         uint16_t                temp_sensor[8];
1078
1079         uint8_t                 reserved2[296];
1080 } __packed __aligned(4);
1081
1082 _Static_assert(sizeof(struct nvme_health_information_page) == 512, "bad size for nvme_health_information_page");
1083
1084 struct nvme_firmware_page {
1085
1086         uint8_t                 afi;
1087         uint8_t                 reserved[7];
1088         uint64_t                revision[7]; /* revisions for 7 slots */
1089         uint8_t                 reserved2[448];
1090 } __packed __aligned(4);
1091
1092 _Static_assert(sizeof(struct nvme_firmware_page) == 512, "bad size for nvme_firmware_page");
1093
1094 struct nvme_ns_list {
1095         uint32_t                ns[1024];
1096 } __packed __aligned(4);
1097
1098 _Static_assert(sizeof(struct nvme_ns_list) == 4096, "bad size for nvme_ns_list");
1099
1100 struct intel_log_temp_stats
1101 {
1102         uint64_t        current;
1103         uint64_t        overtemp_flag_last;
1104         uint64_t        overtemp_flag_life;
1105         uint64_t        max_temp;
1106         uint64_t        min_temp;
1107         uint64_t        _rsvd[5];
1108         uint64_t        max_oper_temp;
1109         uint64_t        min_oper_temp;
1110         uint64_t        est_offset;
1111 } __packed __aligned(4);
1112
1113 _Static_assert(sizeof(struct intel_log_temp_stats) == 13 * 8, "bad size for intel_log_temp_stats");
1114
1115 #define NVME_TEST_MAX_THREADS   128
1116
1117 struct nvme_io_test {
1118
1119         enum nvme_nvm_opcode    opc;
1120         uint32_t                size;
1121         uint32_t                time;   /* in seconds */
1122         uint32_t                num_threads;
1123         uint32_t                flags;
1124         uint64_t                io_completed[NVME_TEST_MAX_THREADS];
1125 };
1126
1127 enum nvme_io_test_flags {
1128
1129         /*
1130          * Specifies whether dev_refthread/dev_relthread should be
1131          *  called during NVME_BIO_TEST.  Ignored for other test
1132          *  types.
1133          */
1134         NVME_TEST_FLAG_REFTHREAD =      0x1,
1135 };
1136
1137 struct nvme_pt_command {
1138
1139         /*
1140          * cmd is used to specify a passthrough command to a controller or
1141          *  namespace.
1142          *
1143          * The following fields from cmd may be specified by the caller:
1144          *      * opc  (opcode)
1145          *      * nsid (namespace id) - for admin commands only
1146          *      * cdw10-cdw15
1147          *
1148          * Remaining fields must be set to 0 by the caller.
1149          */
1150         struct nvme_command     cmd;
1151
1152         /*
1153          * cpl returns completion status for the passthrough command
1154          *  specified by cmd.
1155          *
1156          * The following fields will be filled out by the driver, for
1157          *  consumption by the caller:
1158          *      * cdw0
1159          *      * status (except for phase)
1160          *
1161          * Remaining fields will be set to 0 by the driver.
1162          */
1163         struct nvme_completion  cpl;
1164
1165         /* buf is the data buffer associated with this passthrough command. */
1166         void *                  buf;
1167
1168         /*
1169          * len is the length of the data buffer associated with this
1170          *  passthrough command.
1171          */
1172         uint32_t                len;
1173
1174         /*
1175          * is_read = 1 if the passthrough command will read data into the
1176          *  supplied buffer from the controller.
1177          *
1178          * is_read = 0 if the passthrough command will write data from the
1179          *  supplied buffer to the controller.
1180          */
1181         uint32_t                is_read;
1182
1183         /*
1184          * driver_lock is used by the driver only.  It must be set to 0
1185          *  by the caller.
1186          */
1187         struct mtx *            driver_lock;
1188 };
1189
1190 #define nvme_completion_is_error(cpl)                                   \
1191         (NVME_STATUS_GET_SC((cpl)->status) != 0 || NVME_STATUS_GET_SCT((cpl)->status) != 0)
1192
1193 void    nvme_strvis(uint8_t *dst, const uint8_t *src, int dstlen, int srclen);
1194
1195 #ifdef _KERNEL
1196
1197 struct bio;
1198
1199 struct nvme_namespace;
1200 struct nvme_controller;
1201 struct nvme_consumer;
1202
1203 typedef void (*nvme_cb_fn_t)(void *, const struct nvme_completion *);
1204
1205 typedef void *(*nvme_cons_ns_fn_t)(struct nvme_namespace *, void *);
1206 typedef void *(*nvme_cons_ctrlr_fn_t)(struct nvme_controller *);
1207 typedef void (*nvme_cons_async_fn_t)(void *, const struct nvme_completion *,
1208                                      uint32_t, void *, uint32_t);
1209 typedef void (*nvme_cons_fail_fn_t)(void *);
1210
1211 enum nvme_namespace_flags {
1212         NVME_NS_DEALLOCATE_SUPPORTED    = 0x1,
1213         NVME_NS_FLUSH_SUPPORTED         = 0x2,
1214 };
1215
1216 int     nvme_ctrlr_passthrough_cmd(struct nvme_controller *ctrlr,
1217                                    struct nvme_pt_command *pt,
1218                                    uint32_t nsid, int is_user_buffer,
1219                                    int is_admin_cmd);
1220
1221 /* Admin functions */
1222 void    nvme_ctrlr_cmd_set_feature(struct nvme_controller *ctrlr,
1223                                    uint8_t feature, uint32_t cdw11,
1224                                    void *payload, uint32_t payload_size,
1225                                    nvme_cb_fn_t cb_fn, void *cb_arg);
1226 void    nvme_ctrlr_cmd_get_feature(struct nvme_controller *ctrlr,
1227                                    uint8_t feature, uint32_t cdw11,
1228                                    void *payload, uint32_t payload_size,
1229                                    nvme_cb_fn_t cb_fn, void *cb_arg);
1230 void    nvme_ctrlr_cmd_get_log_page(struct nvme_controller *ctrlr,
1231                                     uint8_t log_page, uint32_t nsid,
1232                                     void *payload, uint32_t payload_size,
1233                                     nvme_cb_fn_t cb_fn, void *cb_arg);
1234
1235 /* NVM I/O functions */
1236 int     nvme_ns_cmd_write(struct nvme_namespace *ns, void *payload,
1237                           uint64_t lba, uint32_t lba_count, nvme_cb_fn_t cb_fn,
1238                           void *cb_arg);
1239 int     nvme_ns_cmd_write_bio(struct nvme_namespace *ns, struct bio *bp,
1240                               nvme_cb_fn_t cb_fn, void *cb_arg);
1241 int     nvme_ns_cmd_read(struct nvme_namespace *ns, void *payload,
1242                          uint64_t lba, uint32_t lba_count, nvme_cb_fn_t cb_fn,
1243                          void *cb_arg);
1244 int     nvme_ns_cmd_read_bio(struct nvme_namespace *ns, struct bio *bp,
1245                               nvme_cb_fn_t cb_fn, void *cb_arg);
1246 int     nvme_ns_cmd_deallocate(struct nvme_namespace *ns, void *payload,
1247                                uint8_t num_ranges, nvme_cb_fn_t cb_fn,
1248                                void *cb_arg);
1249 int     nvme_ns_cmd_flush(struct nvme_namespace *ns, nvme_cb_fn_t cb_fn,
1250                           void *cb_arg);
1251 int     nvme_ns_dump(struct nvme_namespace *ns, void *virt, off_t offset,
1252                      size_t len);
1253
1254 /* Registration functions */
1255 struct nvme_consumer *  nvme_register_consumer(nvme_cons_ns_fn_t    ns_fn,
1256                                                nvme_cons_ctrlr_fn_t ctrlr_fn,
1257                                                nvme_cons_async_fn_t async_fn,
1258                                                nvme_cons_fail_fn_t  fail_fn);
1259 void            nvme_unregister_consumer(struct nvme_consumer *consumer);
1260
1261 /* Controller helper functions */
1262 device_t        nvme_ctrlr_get_device(struct nvme_controller *ctrlr);
1263 const struct nvme_controller_data *
1264                 nvme_ctrlr_get_data(struct nvme_controller *ctrlr);
1265
1266 /* Namespace helper functions */
1267 uint32_t        nvme_ns_get_max_io_xfer_size(struct nvme_namespace *ns);
1268 uint32_t        nvme_ns_get_sector_size(struct nvme_namespace *ns);
1269 uint64_t        nvme_ns_get_num_sectors(struct nvme_namespace *ns);
1270 uint64_t        nvme_ns_get_size(struct nvme_namespace *ns);
1271 uint32_t        nvme_ns_get_flags(struct nvme_namespace *ns);
1272 const char *    nvme_ns_get_serial_number(struct nvme_namespace *ns);
1273 const char *    nvme_ns_get_model_number(struct nvme_namespace *ns);
1274 const struct nvme_namespace_data *
1275                 nvme_ns_get_data(struct nvme_namespace *ns);
1276 uint32_t        nvme_ns_get_stripesize(struct nvme_namespace *ns);
1277
1278 int     nvme_ns_bio_process(struct nvme_namespace *ns, struct bio *bp,
1279                             nvme_cb_fn_t cb_fn);
1280
1281 /*
1282  * Command building helper functions -- shared with CAM
1283  * These functions assume allocator zeros out cmd structure
1284  * CAM's xpt_get_ccb and the request allocator for nvme both
1285  * do zero'd allocations.
1286  */
1287 static inline
1288 void    nvme_ns_flush_cmd(struct nvme_command *cmd, uint32_t nsid)
1289 {
1290
1291         cmd->opc_fuse = NVME_CMD_SET_OPC(NVME_OPC_FLUSH);
1292         cmd->nsid = htole32(nsid);
1293 }
1294
1295 static inline
1296 void    nvme_ns_rw_cmd(struct nvme_command *cmd, uint32_t rwcmd, uint32_t nsid,
1297     uint64_t lba, uint32_t count)
1298 {
1299         cmd->opc_fuse = NVME_CMD_SET_OPC(rwcmd);
1300         cmd->nsid = htole32(nsid);
1301         cmd->cdw10 = htole32(lba & 0xffffffffu);
1302         cmd->cdw11 = htole32(lba >> 32);
1303         cmd->cdw12 = htole32(count-1);
1304 }
1305
1306 static inline
1307 void    nvme_ns_write_cmd(struct nvme_command *cmd, uint32_t nsid,
1308     uint64_t lba, uint32_t count)
1309 {
1310         nvme_ns_rw_cmd(cmd, NVME_OPC_WRITE, nsid, lba, count);
1311 }
1312
1313 static inline
1314 void    nvme_ns_read_cmd(struct nvme_command *cmd, uint32_t nsid,
1315     uint64_t lba, uint32_t count)
1316 {
1317         nvme_ns_rw_cmd(cmd, NVME_OPC_READ, nsid, lba, count);
1318 }
1319
1320 static inline
1321 void    nvme_ns_trim_cmd(struct nvme_command *cmd, uint32_t nsid,
1322     uint32_t num_ranges)
1323 {
1324         cmd->opc_fuse = NVME_CMD_SET_OPC(NVME_OPC_DATASET_MANAGEMENT);
1325         cmd->nsid = htole32(nsid);
1326         cmd->cdw10 = htole32(num_ranges - 1);
1327         cmd->cdw11 = htole32(NVME_DSM_ATTR_DEALLOCATE);
1328 }
1329
1330 extern int nvme_use_nvd;
1331
1332 #endif /* _KERNEL */
1333
1334 /* Endianess conversion functions for NVMe structs */
1335 static inline
1336 void    nvme_completion_swapbytes(struct nvme_completion *s)
1337 {
1338
1339         s->cdw0 = le32toh(s->cdw0);
1340         /* omit rsvd1 */
1341         s->sqhd = le16toh(s->sqhd);
1342         s->sqid = le16toh(s->sqid);
1343         /* omit cid */
1344         s->status = le16toh(s->status);
1345 }
1346
1347 static inline
1348 void    nvme_power_state_swapbytes(struct nvme_power_state *s)
1349 {
1350
1351         s->mp = le16toh(s->mp);
1352         s->enlat = le32toh(s->enlat);
1353         s->exlat = le32toh(s->exlat);
1354         s->idlp = le16toh(s->idlp);
1355         s->actp = le16toh(s->actp);
1356 }
1357
1358 static inline
1359 void    nvme_controller_data_swapbytes(struct nvme_controller_data *s)
1360 {
1361         int i;
1362
1363         s->vid = le16toh(s->vid);
1364         s->ssvid = le16toh(s->ssvid);
1365         s->ctrlr_id = le16toh(s->ctrlr_id);
1366         s->ver = le32toh(s->ver);
1367         s->rtd3r = le32toh(s->rtd3r);
1368         s->rtd3e = le32toh(s->rtd3e);
1369         s->oaes = le32toh(s->oaes);
1370         s->ctratt = le32toh(s->ctratt);
1371         s->oacs = le16toh(s->oacs);
1372         s->wctemp = le16toh(s->wctemp);
1373         s->cctemp = le16toh(s->cctemp);
1374         s->mtfa = le16toh(s->mtfa);
1375         s->hmpre = le32toh(s->hmpre);
1376         s->hmmin = le32toh(s->hmmin);
1377         s->rpmbs = le32toh(s->rpmbs);
1378         s->edstt = le16toh(s->edstt);
1379         s->kas = le16toh(s->kas);
1380         s->hctma = le16toh(s->hctma);
1381         s->mntmt = le16toh(s->mntmt);
1382         s->mxtmt = le16toh(s->mxtmt);
1383         s->sanicap = le32toh(s->sanicap);
1384         s->maxcmd = le16toh(s->maxcmd);
1385         s->nn = le32toh(s->nn);
1386         s->oncs = le16toh(s->oncs);
1387         s->fuses = le16toh(s->fuses);
1388         s->awun = le16toh(s->awun);
1389         s->awupf = le16toh(s->awupf);
1390         s->acwu = le16toh(s->acwu);
1391         s->sgls = le32toh(s->sgls);
1392         for (i = 0; i < 32; i++)
1393                 nvme_power_state_swapbytes(&s->power_state[i]);
1394 }
1395
1396 static inline
1397 void    nvme_namespace_data_swapbytes(struct nvme_namespace_data *s)
1398 {
1399         int i;
1400
1401         s->nsze = le64toh(s->nsze);
1402         s->ncap = le64toh(s->ncap);
1403         s->nuse = le64toh(s->nuse);
1404         s->nawun = le16toh(s->nawun);
1405         s->nawupf = le16toh(s->nawupf);
1406         s->nacwu = le16toh(s->nacwu);
1407         s->nabsn = le16toh(s->nabsn);
1408         s->nabo = le16toh(s->nabo);
1409         s->nabspf = le16toh(s->nabspf);
1410         s->noiob = le16toh(s->noiob);
1411         for (i = 0; i < 16; i++)
1412                 s->lbaf[i] = le32toh(s->lbaf[i]);
1413 }
1414
1415 static inline
1416 void    nvme_error_information_entry_swapbytes(struct nvme_error_information_entry *s)
1417 {
1418
1419         s->error_count = le64toh(s->error_count);
1420         s->sqid = le16toh(s->sqid);
1421         s->cid = le16toh(s->cid);
1422         s->status = le16toh(s->status);
1423         s->error_location = le16toh(s->error_location);
1424         s->lba = le64toh(s->lba);
1425         s->nsid = le32toh(s->nsid);
1426 }
1427
1428 static inline
1429 void    nvme_le128toh(void *p)
1430 {
1431 #if _BYTE_ORDER != _LITTLE_ENDIAN
1432         /* Swap 16 bytes in place */
1433         char *tmp = (char*)p;
1434         char b;
1435         int i;
1436         for (i = 0; i < 8; i++) {
1437                 b = tmp[i];
1438                 tmp[i] = tmp[15-i];
1439                 tmp[15-i] = b;
1440         }
1441 #else
1442         (void)p;
1443 #endif
1444 }
1445
1446 static inline
1447 void    nvme_health_information_page_swapbytes(struct nvme_health_information_page *s)
1448 {
1449         int i;
1450
1451         s->temperature = le16toh(s->temperature);
1452         nvme_le128toh((void *)s->data_units_read);
1453         nvme_le128toh((void *)s->data_units_written);
1454         nvme_le128toh((void *)s->host_read_commands);
1455         nvme_le128toh((void *)s->host_write_commands);
1456         nvme_le128toh((void *)s->controller_busy_time);
1457         nvme_le128toh((void *)s->power_cycles);
1458         nvme_le128toh((void *)s->power_on_hours);
1459         nvme_le128toh((void *)s->unsafe_shutdowns);
1460         nvme_le128toh((void *)s->media_errors);
1461         nvme_le128toh((void *)s->num_error_info_log_entries);
1462         s->warning_temp_time = le32toh(s->warning_temp_time);
1463         s->error_temp_time = le32toh(s->error_temp_time);
1464         for (i = 0; i < 8; i++)
1465                 s->temp_sensor[i] = le16toh(s->temp_sensor[i]);
1466 }
1467
1468
1469 static inline
1470 void    nvme_firmware_page_swapbytes(struct nvme_firmware_page *s)
1471 {
1472         int i;
1473
1474         for (i = 0; i < 7; i++)
1475                 s->revision[i] = le64toh(s->revision[i]);
1476 }
1477
1478 static inline
1479 void    nvme_ns_list_swapbytes(struct nvme_ns_list *s)
1480 {
1481         int i;
1482
1483         for (i = 0; i < 1024; i++)
1484                 s->ns[i] = le32toh(s->ns[i]);
1485 }
1486
1487 static inline
1488 void    intel_log_temp_stats_swapbytes(struct intel_log_temp_stats *s)
1489 {
1490
1491         s->current = le64toh(s->current);
1492         s->overtemp_flag_last = le64toh(s->overtemp_flag_last);
1493         s->overtemp_flag_life = le64toh(s->overtemp_flag_life);
1494         s->max_temp = le64toh(s->max_temp);
1495         s->min_temp = le64toh(s->min_temp);
1496         /* omit _rsvd[] */
1497         s->max_oper_temp = le64toh(s->max_oper_temp);
1498         s->min_oper_temp = le64toh(s->min_oper_temp);
1499         s->est_offset = le64toh(s->est_offset);
1500 }
1501
1502 #endif /* __NVME_H__ */