xref: /illumos-gate/usr/src/uts/common/sys/nvme.h (revision 046911eb)
1 /*
2  * This file and its contents are supplied under the terms of the
3  * Common Development and Distribution License ("CDDL"), version 1.0.
4  * You may only use this file in accordance with the terms of version
5  * 1.0 of the CDDL.
6  *
7  * A full copy of the text of the CDDL should have accompanied this
8  * source.  A copy of the CDDL is also available via the Internet at
9  * http://www.illumos.org/license/CDDL.
10  */
11 
12 /*
13  * Copyright 2016 Nexenta Systems, Inc.
14  * Copyright 2020 Joyent, Inc.
15  * Copyright 2019 Western Digital Corporation
16  * Copyright 2024 Oxide Computer Company
17  * Copyright 2022 OmniOS Community Edition (OmniOSce) Association.
18  */
19 
20 #ifndef _SYS_NVME_H
21 #define	_SYS_NVME_H
22 
23 #include <sys/types.h>
24 #include <sys/debug.h>
25 #include <sys/stddef.h>
26 
27 #ifdef _KERNEL
28 #include <sys/types32.h>
29 #else
30 #include <sys/uuid.h>
31 #include <stdint.h>
32 #endif
33 
34 /*
35  * Declarations used for communication between nvmeadm(8) and nvme(4D)
36  */
37 
38 #ifdef __cplusplus
39 extern "C" {
40 #endif
41 
42 /*
43  * NVMe ioctl definitions
44  */
45 
46 #define	NVME_IOC			(('N' << 24) | ('V' << 16) | ('M' << 8))
47 #define	NVME_IOC_CTRL_INFO		(NVME_IOC | 0)
48 #define	NVME_IOC_IDENTIFY		(NVME_IOC | 1)
49 #define	NVME_IOC_GET_LOGPAGE		(NVME_IOC | 2)
50 #define	NVME_IOC_GET_FEATURE		(NVME_IOC | 3)
51 #define	NVME_IOC_FORMAT			(NVME_IOC | 4)
52 #define	NVME_IOC_DETACH			(NVME_IOC | 5)
53 #define	NVME_IOC_ATTACH			(NVME_IOC | 6)
54 #define	NVME_IOC_FIRMWARE_DOWNLOAD	(NVME_IOC | 7)
55 #define	NVME_IOC_FIRMWARE_COMMIT	(NVME_IOC | 8)
56 #define	NVME_IOC_PASSTHRU		(NVME_IOC | 9)
57 #define	NVME_IOC_NS_INFO		(NVME_IOC | 10)
58 #define	NVME_IOC_LOCK			(NVME_IOC | 11)
59 #define	NVME_IOC_UNLOCK			(NVME_IOC | 12)
60 #define	NVME_IOC_MAX			NVME_IOC_NS_INFO
61 
62 #define	IS_NVME_IOC(x)			((x) > NVME_IOC && (x) <= NVME_IOC_MAX)
63 #define	NVME_IOC_CMD(x)			((x) & 0xff)
64 
65 /*
66  * This represents the set of all possible errors that can be returned from an
67  * ioctl. Our general rule of thumb is that we only will use an errno value to
68  * indicate that certain processing failed: a lack of privileges, bad minor, or
69  * failure to copy in and out the initial ioctl structure. However, if we get
70  * far enough that there is any other failure (including a failure to copy in
71  * and out nested data such as the identify command payload) then we will issue
72  * an error here. Put differently, our basic promise is that there should be a
73  * single straightforward meaning for any errno returned and instead all the
74  * nuance is here. Our goal is that no one should guess what of two dozen things
75  * an EINVAL might have referred to.
76  *
77  * When we are dealing with field parameters, there are three general classes of
78  * errors that we define that are common across all request structures:
79  *
80  *   <REQ>_<FIELD>_RANGE	RANGE class errors indicate that the value
81  *				passed in is outside the range that the device
82  *				supports. The range may vary based on the
83  *				specification. This is used both for issues like
84  *				bad alignment in a value (e.g. not 4-byte
85  *				aligned) or a value that is larger than the
86  *				maximum possible size. Because the namespace ID
87  *				is shared in every request in the controller and
88  *				is part of our standard ioctl handling, we use a
89  *				single set of errors for that.
90  *
91  *   <REQ>_<FIELD>_UNSUP	This indicates that the controller cannot
92  *				support any value in the given field. This is
93  *				either because the field was introduced in an
94  *				NVMe specification later than the controller
95  *				supports or because there is an explicit feature
96  *				bit that indicates whether or not this field is
97  *				valid. Entries here may or may not have a
98  *				namespace unsupported entry due to the fact that
99  *				this is command specific.
100  *
101  *  <REQ>_<FIELD>_UNUSE		This class is perhaps the weirdest. This
102  *				represents a case where a given field cannot be
103  *				set because it is not used based on the
104  *				specifics of the request. For example, if you're
105  *				getting the health log page, you may not set the
106  *				LSP or LSI for that log page, even if you have
107  *				an NVMe 1.4 controller that supports both fields
108  *				because they have no meaning. A similar example
109  *				would be setting a controller ID when it has no
110  *				meaning in a particular identify request.
111  *
112  * While every field will have a RANGE class error, some fields will not have an
113  * UNSUP or UNUSE class error depending on the specifics. A field that has
114  * always been present since NVMe 1.0 and is always valid, such as say the log
115  * page ID field for a get log page request or the length of a firmware download
116  * request, currently are always valid. It is possible that future revisions to
117  * the specification or our logic may change this.
118  */
119 typedef enum {
120 	/*
121 	 * Indicates that the command actually completed successfully.
122 	 */
123 	NVME_IOCTL_E_OK	= 0,
124 	/*
125 	 * Indicates that the controller failed the command and the controller
126 	 * specific (SC/SCT) are available. For all other errors, those fields
127 	 * are reserved.
128 	 */
129 	NVME_IOCTL_E_CTRL_ERROR,
130 	/*
131 	 * Indicates that the controller is considered "dead" by the system and
132 	 * therefore is unusable. Separately, the controller may have been
133 	 * removed from the system due to hotplug or related. In that case, the
134 	 * gone variant is used to distinguish this.
135 	 */
136 	NVME_IOCTL_E_CTRL_DEAD,
137 	NVME_IOCTL_E_CTRL_GONE,
138 	/*
139 	 * Indicates that a bad namespace was requested. This would generally
140 	 * happen when referring to a namespace that is outside of controller's
141 	 * range.
142 	 */
143 	NVME_IOCTL_E_NS_RANGE,
144 	/*
145 	 * Indicates that a namespace is not usable in this context.
146 	 */
147 	NVME_IOCTL_E_NS_UNUSE,
148 	/*
149 	 * Indicates that the requested namespace could not be used because we
150 	 * are operating on a namespace minor and asked to operate on a
151 	 * different namespace.
152 	 */
153 	NVME_IOCTL_E_MINOR_WRONG_NS,
154 	/*
155 	 * Indicates that the requested ioctl can only operate on the controller
156 	 * minor and we were on a namespace minor. This is not used for when a
157 	 * namespace is incorrectly requested otherwise.
158 	 */
159 	NVME_IOCTL_E_NOT_CTRL,
160 	/*
161 	 * Indicates that we were asked to operate on the broadcast namespace
162 	 * either because it was specified or that was how the request was
163 	 * transformed and the broadcast namespace is not supported for this
164 	 * operation.
165 	 */
166 	NVME_IOCTL_E_NO_BCAST_NS,
167 	/*
168 	 * Indicates that the operation failed because the operation requires a
169 	 * controller or namespace write lock and the caller did not have it.
170 	 */
171 	NVME_IOCTL_E_NEED_CTRL_WRLOCK,
172 	NVME_IOCTL_E_NEED_NS_WRLOCK,
173 	/*
174 	 * Indicates that the operation could not proceed because someone else
175 	 * has exclusive access currently to the controller or namespace and
176 	 * therefore this request (which does not require exclusive access)
177 	 * could not proceed.
178 	 */
179 	NVME_IOCTL_E_CTRL_LOCKED,
180 	NVME_IOCTL_E_NS_LOCKED,
181 	/*
182 	 * Indicates that a standard log page was requested that the kernel
183 	 * doesn't know about.
184 	 */
185 	NVME_IOCTL_E_UNKNOWN_LOG_PAGE,
186 	/*
187 	 * Indicates that the controller does not support the requested log
188 	 * page; however, the kernel knows about it.
189 	 */
190 	NVME_IOCTL_E_UNSUP_LOG_PAGE,
191 	/*
192 	 * Indicates that the log page's scope requires operating on something
193 	 * that isn't what was requested. For example, trying to request the
194 	 * firmware information page on a namespace.
195 	 */
196 	NVME_IOCTL_E_BAD_LOG_SCOPE,
197 	/*
198 	 * Log page fields with bad values.
199 	 */
200 	NVME_IOCTL_E_LOG_CSI_RANGE,
201 	NVME_IOCTL_E_LOG_LID_RANGE,
202 	NVME_IOCTL_E_LOG_LSP_RANGE,
203 	NVME_IOCTL_E_LOG_LSI_RANGE,
204 	NVME_IOCTL_E_LOG_RAE_RANGE,
205 	NVME_IOCTL_E_LOG_SIZE_RANGE,
206 	NVME_IOCTL_E_LOG_OFFSET_RANGE,
207 	/*
208 	 * Log page fields that may not be supported.
209 	 */
210 	NVME_IOCTL_E_LOG_CSI_UNSUP,
211 	NVME_IOCTL_E_LOG_LSP_UNSUP,
212 	NVME_IOCTL_E_LOG_LSI_UNSUP,
213 	NVME_IOCTL_E_LOG_RAE_UNSUP,
214 	NVME_IOCTL_E_LOG_OFFSET_UNSUP,
215 	/*
216 	 * Log page fields that may not be usable, depending on context.
217 	 */
218 	NVME_IOCTL_E_LOG_LSP_UNUSE,
219 	NVME_IOCTL_E_LOG_LSI_UNUSE,
220 	NVME_IOCTL_E_LOG_RAE_UNUSE,
221 	/*
222 	 * Indicates that no DMA memory was available for a request.
223 	 */
224 	NVME_IOCTL_E_NO_DMA_MEM,
225 	/*
226 	 * Indicates that there was no kernel memory avilable for the request.
227 	 */
228 	NVME_IOCTL_E_NO_KERN_MEM,
229 	/*
230 	 * Indicates that an error occurred while trying to fill out the DMA PRP
231 	 */
232 	NVME_IOCTL_E_BAD_PRP,
233 	/*
234 	 * Indicates that a pointer to user data to read from or write to was
235 	 * not valid and generated a fault. Specifically this is for items that
236 	 * an ioctl structure points to.
237 	 */
238 	NVME_IOCTL_E_BAD_USER_DATA,
239 	/*
240 	 * Indicates that the kernel does not know about the requested identify
241 	 * command.
242 	 */
243 	NVME_IOCTL_E_UNKNOWN_IDENTIFY,
244 	/*
245 	 * Indicates that the controller does not support the requested identify
246 	 * command.
247 	 */
248 	NVME_IOCTL_E_UNSUP_IDENTIFY,
249 	/*
250 	 * The following errors indicate either a bad value for a given identify
251 	 * argument. This would happen because the value is outside the
252 	 * supported range. There is no CNS or below as those are the
253 	 * higher-level errors right above this.
254 	 */
255 	NVME_IOCTL_E_IDENTIFY_CTRLID_RANGE,
256 	/*
257 	 * Next, we have the unsupported and unusable pieces. The nsid was
258 	 * supported starting in NVMe 1.0, therefore it is never unsupported.
259 	 * However, the controller ID both requires controller support and is
260 	 * not usable in several requests.
261 	 */
262 	NVME_IOCTL_E_IDENTIFY_CTRLID_UNSUP,
263 	NVME_IOCTL_E_IDENTIFY_CTRLID_UNUSE,
264 	/*
265 	 * Indicates that the controller does not support the NVMe spec's
266 	 * general vendor unique command format.
267 	 */
268 	NVME_IOCTL_E_CTRL_VUC_UNSUP,
269 	/*
270 	 * The following indicate bad values for given NVMe vendor unique
271 	 * command fields. All of the cdw1[2-5] fields are not part of this
272 	 * because there is nothing that we can validate.
273 	 */
274 	NVME_IOCTL_E_VUC_TIMEOUT_RANGE,
275 	NVME_IOCTL_E_VUC_OPCODE_RANGE,
276 	NVME_IOCTL_E_VUC_FLAGS_RANGE,
277 	NVME_IOCTL_E_VUC_IMPACT_RANGE,
278 	NVME_IOCTL_E_VUC_NDT_RANGE,
279 	/*
280 	 * These indicate that the VUC data and that the corresponding pair of
281 	 * fields do not agree with each other.
282 	 */
283 	NVME_IOCTL_E_INCONSIST_VUC_FLAGS_NDT,
284 	NVME_IOCTL_E_INCONSIST_VUC_BUF_NDT,
285 	/*
286 	 * Indicates that the operation in question did not succeed because
287 	 * blkdev failed to detach. Most often this happens because the device
288 	 * node is busy. Reasons the device node could be busy include that the
289 	 * device is in a zpool, a file system is mounted, a process has the
290 	 * block device open, etc.
291 	 */
292 	NVME_IOCTL_E_BLKDEV_DETACH,
293 	/*
294 	 * Indicates that the operation in question failed because we were
295 	 * unable to create and online a new blkdev child.
296 	 */
297 	NVME_IOCTL_E_BLKDEV_ATTACH,
298 	/*
299 	 * Indicates that the namespace requested for an attach is not supported
300 	 * by the system. This would happen due to properties of the namespace
301 	 * itself (e.g. utilizing metadata sectors).
302 	 */
303 	NVME_IOCTL_E_UNSUP_ATTACH_NS,
304 	/*
305 	 * Indicates that the format operation is not supported by the
306 	 * controller at all.
307 	 */
308 	NVME_IOCTL_E_CTRL_FORMAT_UNSUP,
309 	/*
310 	 * Indicates that the controller does not support the ability to perform
311 	 * a cryptographic secure erase.
312 	 */
313 	NVME_IOCTL_E_CTRL_CRYPTO_SE_UNSUP,
314 	/*
315 	 * Indicates that a format operation is targeting a namespace, but
316 	 * cannot be performed because it does not support formatting an
317 	 * individual namespace or performing a secure-erase of an individual
318 	 * namespace respectively.
319 	 */
320 	NVME_IOCTL_E_CTRL_NS_FORMAT_UNSUP,
321 	NVME_IOCTL_E_CTRL_NS_SE_UNSUP,
322 	/*
323 	 * The following indicate bad values for a format NVM request.
324 	 */
325 	NVME_IOCTL_E_FORMAT_LBAF_RANGE,
326 	NVME_IOCTL_E_FORMAT_SES_RANGE,
327 	/*
328 	 * Indicates that the requested LBA format is not supported due to its
329 	 * use of metadata.
330 	 */
331 	NVME_IOCTL_E_UNSUP_LBAF_META,
332 	/*
333 	 * Indicates that the firmware commands are not supported by the
334 	 * controller at all.
335 	 */
336 	NVME_IOCTL_E_CTRL_FW_UNSUP,
337 	/*
338 	 * Indicates that the controller has reported a firmware update
339 	 * granularity that exceeds the calculated / driver supported maximum
340 	 * DMA transfer size. As such we cannot perform this operation.
341 	 */
342 	NVME_IOCTL_E_FW_LOAD_IMPOS_GRAN,
343 	/*
344 	 * The following indicate bad values for a firmware load's length and
345 	 * offset.
346 	 */
347 	NVME_IOCTL_E_FW_LOAD_LEN_RANGE,
348 	NVME_IOCTL_E_FW_LOAD_OFFSET_RANGE,
349 	/*
350 	 * The following indicate bad values for a firmware commit's slot and
351 	 * action.
352 	 */
353 	NVME_IOCTL_E_FW_COMMIT_SLOT_RANGE,
354 	NVME_IOCTL_E_FW_COMMIT_ACTION_RANGE,
355 	/*
356 	 * Indicates that an explicit attempt was made to download an image into
357 	 * a read-only slot. Note, some instances of this cannot be caught prior
358 	 * to issuing a command to the controller (commit action 0b11 as it can
359 	 * be used whether there is or isn't a staged image) and will result in
360 	 * a controller error.
361 	 */
362 	NVME_IOCTL_E_RO_FW_SLOT,
363 	/*
364 	 * Indicates that the kernel doesn't know about the NVMe feature in
365 	 * question and therefore cannot proceed.
366 	 */
367 	NVME_IOCTL_E_UNKNOWN_FEATURE,
368 	/*
369 	 * Indicates that while the system knows about the feature in question,
370 	 * it is not supported by the controller.
371 	 */
372 	NVME_IOCTL_E_UNSUP_FEATURE,
373 	/*
374 	 * The following errors indicate a bad value for a given get feature
375 	 * field. This would happen because the value is outside the supported
376 	 * range.
377 	 */
378 	NVME_IOCTL_E_GET_FEAT_SEL_RANGE,
379 	NVME_IOCTL_E_GET_FEAT_CDW11_RANGE,
380 	NVME_IOCTL_E_GET_FEAT_DATA_RANGE,
381 	/*
382 	 * This set of errors indicate that the field is not supported. This can
383 	 * happen because a given get feature command doesn't support setting
384 	 * this value, the field isn't supported in this revision of the
385 	 * controller, or similar issues.
386 	 */
387 	NVME_IOCTL_E_GET_FEAT_SEL_UNSUP,
388 	/*
389 	 * Fields that may be circumstantially unusable.
390 	 */
391 	NVME_IOCTL_E_GET_FEAT_CDW11_UNUSE,
392 	NVME_IOCTL_E_GET_FEAT_DATA_UNUSE,
393 	/*
394 	 * The following errors indicate a bad lock type.
395 	 */
396 	NVME_IOCTL_E_BAD_LOCK_ENTITY,
397 	NVME_IOCTL_E_BAD_LOCK_LEVEL,
398 	NVME_IOCTL_E_BAD_LOCK_FLAGS,
399 	/*
400 	 * Indicates that a namespace open cannot lock or unlock a controller.
401 	 */
402 	NVME_IOCTL_E_NS_CANNOT_LOCK_CTRL,
403 	NVME_IOCTL_E_NS_CANNOT_UNLOCK_CTRL,
404 	/*
405 	 * Indicates that this lock is already held by the caller.
406 	 */
407 	NVME_IOCTL_E_LOCK_ALREADY_HELD,
408 	/*
409 	 * Indicates that we cannot take the controller lock, because the
410 	 * caller already has an active namespace lock.
411 	 */
412 	NVME_IOCTL_E_LOCK_NO_CTRL_WITH_NS,
413 	/*
414 	 * Indicates that we cannot take a namespace lock because a controller
415 	 * write lock already exists.
416 	 */
417 	NVME_IOCTL_LOCK_NO_NS_WITH_CTRL_WRLOCK,
418 	/*
419 	 * Indicates that we cannot take a namespace lock because we already
420 	 * have one.
421 	 */
422 	NVME_IOCTL_E_LOCK_NO_2ND_NS,
423 	/*
424 	 * Indicate that a blocking wait for a lock was interrupted due to a
425 	 * signal.
426 	 */
427 	NVME_IOCTL_E_LOCK_WAIT_SIGNAL,
428 	/*
429 	 * Indicates that the lock could not be acquired because it was already
430 	 * held and we were asked not to block on the lock.
431 	 */
432 	NVME_IOCTL_E_LOCK_WOULD_BLOCK,
433 	/*
434 	 * Indicates that the lock operation could not proceed because the minor
435 	 * is already blocking on another lock operation.
436 	 */
437 	NVME_IOCTL_E_LOCK_PENDING,
438 	/*
439 	 * Indicates that the requested lock could not be unlocked because it is
440 	 * not held. The minor may not hold the lock or it may be blocking for
441 	 * acquisition.
442 	 */
443 	NVME_IOCTL_E_LOCK_NOT_HELD,
444 	/*
445 	 * Indicates that the requested lock could not be unlocked because the
446 	 * namespace requested is not the namespace that is currently locked.
447 	 */
448 	NVME_IOCTL_E_LOCK_WRONG_NS,
449 	/*
450 	 * Indicates that the request could not proceed because a namespace is
451 	 * attached to blkdev. This would block a format operation, a vendor
452 	 * unique command that indicated that it would impact all namespaces,
453 	 * etc.
454 	 */
455 	NVME_IOCTL_E_NS_BLKDEV_ATTACH,
456 	/*
457 	 * Indicates that the blkdev address somehow would have overflowed our
458 	 * internal buffer.
459 	 */
460 	NVME_IOCTL_E_BD_ADDR_OVER
461 } nvme_ioctl_errno_t;
462 
463 /*
464  * This structure is embedded as the first item of every ioctl. It is also used
465  * directly for the attach (NVME_IOC_ATTACH) and detach (NVME_IOC_DETACH)
466  * ioctls.
467  */
468 typedef struct {
469 	/*
470 	 * This allows one to specify the namespace ID that the ioctl may
471 	 * target, if it supports it. This field may be left to zero to indicate
472 	 * that the current open device (whether the controller or a namespace)
473 	 * should be targeted. If a namespace is open, a value other than 0 or
474 	 * the current namespace's ID is invalid.
475 	 */
476 	uint32_t nioc_nsid;
477 	/*
478 	 * These next three values represent a possible error that may have
479 	 * occurred. On every ioctl nioc_drv_err is set to a value from the
480 	 * nvme_ioctl_errno_t enumeration. Anything other than NVME_IOCTL_E_OK
481 	 * indicates a failure of some kind. Some error values will put
482 	 * supplemental information in sct and sc. For example,
483 	 * NVME_IOCTL_E_CTRL_ERROR uses that as a way to return the raw error
484 	 * values from the controller for someone to inspect. Others may use
485 	 * this for their own well-defined supplemental information.
486 	 */
487 	uint32_t nioc_drv_err;
488 	uint32_t nioc_ctrl_sct;
489 	uint32_t nioc_ctrl_sc;
490 } nvme_ioctl_common_t;
491 
492 /*
493  * NVMe Identify Command (NVME_IOC_IDENTIFY).
494  */
495 typedef struct {
496 	nvme_ioctl_common_t nid_common;
497 	uint32_t nid_cns;
498 	uint32_t nid_ctrlid;
499 	uintptr_t nid_data;
500 } nvme_ioctl_identify_t;
501 
502 /*
503  * The following constants describe the maximum values that may be used in
504  * various identify requests.
505  */
506 #define	NVME_IDENTIFY_MAX_CTRLID	0xffff
507 #define	NVME_IDENTIFY_MAX_NSID		0xffffffff
508 #define	NVME_IDENTIFY_MAX_CNS_1v2	0xff
509 #define	NVME_IDENTIFY_MAX_CNS_1v1	0x3
510 #define	NVME_IDENTIFY_MAX_CNS		0x1
511 
512 /*
513  * Get a specific feature (NVME_IOC_GET_FEATURE).
514  */
515 typedef struct {
516 	nvme_ioctl_common_t nigf_common;
517 	uint32_t nigf_fid;
518 	uint32_t nigf_sel;
519 	uint32_t nigf_cdw11;
520 	uintptr_t nigf_data;
521 	uint64_t nigf_len;
522 	uint32_t nigf_cdw0;
523 } nvme_ioctl_get_feature_t;
524 
525 /*
526  * Feature maximums.
527  */
528 #define	NVME_FEAT_MAX_FID	0xff
529 #define	NVME_FEAT_MAX_SEL	0x3
530 
531 /*
532  * Get a specific log page (NVME_IOC_GET_LOGPAGE). By default, unused fields
533  * should be left at zero.  the input data length is specified by nigl_len, in
534  * bytes. The NVMe specification does not provide a way for a controller to
535  * write less bytes than requested for a log page. It is undefined behavior if a
536  * log page read requests more data than is supported. If this is successful,
537  * nigl_len bytes will be copied out.
538  */
539 typedef struct {
540 	nvme_ioctl_common_t nigl_common;
541 	uint32_t nigl_csi;
542 	uint32_t nigl_lid;
543 	uint32_t nigl_lsp;
544 	uint32_t nigl_lsi;
545 	uint32_t nigl_rae;
546 	uint64_t nigl_len;
547 	uint64_t nigl_offset;
548 	uintptr_t nigl_data;
549 } nvme_ioctl_get_logpage_t;
550 
551 /*
552  * The following constants describe the maximum values for fields that used in
553  * the log page request. Note, some of these change with the version. These
554  * values are inclusive. The default max is the lowest common value. Larger
555  * values are included here. While these values are what the command set
556  * maximums are, the device driver may support smaller minimums (e.g. for size).
557  */
558 #define	NVME_LOG_MAX_LID	0xff
559 #define	NVME_LOG_MAX_LSP	0x0f
560 #define	NVME_LOG_MAX_LSP_2v0	0x7f
561 #define	NVME_LOG_MAX_LSI	0xffff
562 #define	NVME_LOG_MAX_UUID	0x7f
563 #define	NVME_LOG_MAX_CSI	0xff
564 #define	NVME_LOG_MAX_RAE	0x1
565 #define	NVME_LOG_MAX_OFFSET	UINT64_MAX
566 
567 /*
568  * These maximum size values are inclusive like the others. The fields are 12
569  * and 32-bits wide respectively, but are zero based. That is accounted for by
570  * the shifts below.
571  */
572 #define	NVME_LOG_MAX_SIZE	((1ULL << 12ULL) * 4ULL)
573 #define	NVME_LOG_MAX_SIZE_1v2	((1ULL << 32ULL) * 4ULL)
574 
575 /*
576  * Inject a vendor-specific admin command (NVME_IOC_PASSTHRU).
577  */
578 typedef struct {
579 	nvme_ioctl_common_t npc_common;	/* NSID and status */
580 	uint32_t npc_opcode;	/* Command opcode. */
581 	uint32_t npc_timeout;	/* Command timeout, in seconds. */
582 	uint32_t npc_flags;	/* Flags for the command. */
583 	uint32_t npc_impact;	/* Impact information */
584 	uint32_t npc_cdw0;	/* Command-specific result DWord 0 */
585 	uint32_t npc_cdw12;	/* Command-specific DWord 12 */
586 	uint32_t npc_cdw13;	/* Command-specific DWord 13 */
587 	uint32_t npc_cdw14;	/* Command-specific DWord 14 */
588 	uint32_t npc_cdw15;	/* Command-specific DWord 15 */
589 	uint64_t npc_buflen;	/* Size of npc_buf. */
590 	uintptr_t npc_buf;	/* I/O source or destination */
591 } nvme_ioctl_passthru_t;
592 
593 /*
594  * Constants for the passthru admin commands. Because the timeout is a kernel
595  * property, we don't include that here.
596  */
597 #define	NVME_PASSTHRU_MIN_ADMIN_OPC	0xc0
598 #define	NVME_PASSTHRU_MAX_ADMIN_OPC	0xff
599 
600 /* Flags for NVMe passthru commands. */
601 #define	NVME_PASSTHRU_READ	0x1 /* Read from device */
602 #define	NVME_PASSTHRU_WRITE	0x2 /* Write to device */
603 
604 /*
605  * Impact information for NVMe passthru commands. The current impact flags are
606  * defined as follows:
607  *
608  * NVME_IMPACT_NS	This implies that one or all of the namespaces may be
609  *			changed. This command will rescan all namespace after
610  *			this occurs and update our state as a result. However,
611  *			this requires that all such namespaces not be attached
612  *			to blkdev to continue.
613  */
614 #define	NVME_IMPACT_NS		0x01
615 
616 
617 /*
618  * Firmware download (NVME_IOC_FIRMWARE_DOWNLOAD).
619  */
620 typedef struct {
621 	nvme_ioctl_common_t fwl_common;
622 	uintptr_t fwl_buf;
623 	uint64_t fwl_len;
624 	uint64_t fwl_off;
625 } nvme_ioctl_fw_load_t;
626 
627 /*
628  * Firmware commit (NVME_IOC_FIRMWARE_COMMIT). This was previously called
629  * firmware activate in earlier specification revisions.
630  */
631 typedef struct {
632 	nvme_ioctl_common_t fwc_common;
633 	uint32_t fwc_slot;
634 	uint32_t fwc_action;
635 } nvme_ioctl_fw_commit_t;
636 
637 /*
638  * Format NVM command (NVME_IOC_FORMAT)
639  */
640 typedef struct {
641 	nvme_ioctl_common_t nif_common;
642 	uint32_t nif_lbaf;
643 	uint32_t nif_ses;
644 } nvme_ioctl_format_t;
645 
646 typedef enum {
647 	NVME_LOCK_E_CTRL = 1,
648 	NVME_LOCK_E_NS
649 } nvme_lock_ent_t;
650 
651 typedef enum {
652 	NVME_LOCK_L_READ	= 1,
653 	NVME_LOCK_L_WRITE
654 } nvme_lock_level_t;
655 
656 typedef enum {
657 	NVME_LOCK_F_DONT_BLOCK	= 1 << 0
658 } nvme_lock_flags_t;
659 
660 /*
661  * Lock structure (NVME_IOC_LOCK).
662  */
663 typedef struct {
664 	nvme_ioctl_common_t nil_common;
665 	nvme_lock_ent_t nil_ent;
666 	nvme_lock_level_t nil_level;
667 	nvme_lock_flags_t nil_flags;
668 } nvme_ioctl_lock_t;
669 
670 /*
671  * Unlock structure (NVME_IOC_UNLOCK).
672  */
673 typedef struct {
674 	nvme_ioctl_common_t niu_common;
675 	nvme_lock_ent_t niu_ent;
676 } nvme_ioctl_unlock_t;
677 
678 /*
679  * 32-bit ioctl structures. These must be packed to be 4 bytes to get the proper
680  * ILP32 sizing.
681  */
682 #if defined(_KERNEL) && defined(_SYSCALL32)
683 #pragma pack(4)
684 typedef struct {
685 	nvme_ioctl_common_t nid_common;
686 	uint32_t nid_cns;
687 	uint32_t nid_ctrlid;
688 	uintptr32_t nid_data;
689 } nvme_ioctl_identify32_t;
690 
691 typedef struct {
692 	nvme_ioctl_common_t nigf_common;
693 	uint32_t nigf_fid;
694 	uint32_t nigf_sel;
695 	uint32_t nigf_cdw11;
696 	uintptr32_t nigf_data;
697 	uint64_t nigf_len;
698 	uint32_t nigf_cdw0;
699 } nvme_ioctl_get_feature32_t;
700 
701 typedef struct {
702 	nvme_ioctl_common_t nigl_common;
703 	uint32_t nigl_csi;
704 	uint32_t nigl_lid;
705 	uint32_t nigl_lsp;
706 	uint32_t nigl_lsi;
707 	uint32_t nigl_rae;
708 	uint64_t nigl_len;
709 	uint64_t nigl_offset;
710 	uintptr32_t nigl_data;
711 } nvme_ioctl_get_logpage32_t;
712 
713 typedef struct {
714 	nvme_ioctl_common_t npc_common;	/* NSID and status */
715 	uint32_t npc_opcode;	/* Command opcode. */
716 	uint32_t npc_timeout;	/* Command timeout, in seconds. */
717 	uint32_t npc_flags;	/* Flags for the command. */
718 	uint32_t npc_impact;	/* Impact information */
719 	uint32_t npc_cdw0;	/* Command-specific result DWord 0 */
720 	uint32_t npc_cdw12;	/* Command-specific DWord 12 */
721 	uint32_t npc_cdw13;	/* Command-specific DWord 13 */
722 	uint32_t npc_cdw14;	/* Command-specific DWord 14 */
723 	uint32_t npc_cdw15;	/* Command-specific DWord 15 */
724 	uint64_t npc_buflen;	/* Size of npc_buf. */
725 	uintptr32_t npc_buf;	/* I/O source or destination */
726 } nvme_ioctl_passthru32_t;
727 
728 typedef struct {
729 	nvme_ioctl_common_t fwl_common;
730 	uintptr32_t fwl_buf;
731 	uint64_t fwl_len;
732 	uint64_t fwl_off;
733 } nvme_ioctl_fw_load32_t;
734 #pragma pack()	/* pack(4) */
735 #endif	/* _KERNEL && _SYSCALL32 */
736 
737 /*
738  * NVMe capabilities. This is a set of fields that come from the controller's
739  * PCIe register space.
740  */
741 typedef struct {
742 	uint32_t cap_mpsmax;		/* Memory Page Size Maximum */
743 	uint32_t cap_mpsmin;		/* Memory Page Size Minimum */
744 } nvme_capabilities_t;
745 
746 /*
747  * NVMe version
748  */
749 typedef struct {
750 	uint16_t v_minor;
751 	uint16_t v_major;
752 } nvme_version_t;
753 
754 #define	NVME_VERSION_ATLEAST(v, maj, min) \
755 	(((v)->v_major) > (maj) || \
756 	((v)->v_major == (maj) && (v)->v_minor >= (min)))
757 
758 #define	NVME_VERSION_HIGHER(v, maj, min) \
759 	(((v)->v_major) > (maj) || \
760 	((v)->v_major == (maj) && (v)->v_minor > (min)))
761 
762 /*
763  * NVMe Namespace related constants. The maximum NSID is determined by the
764  * identify controller data structure.
765  */
766 #define	NVME_NSID_MIN	1
767 #define	NVME_NSID_BCAST	0xffffffff
768 
769 #pragma pack(1)
770 
771 typedef struct {
772 	uint64_t lo;
773 	uint64_t hi;
774 } nvme_uint128_t;
775 
776 /*
777  * NVMe Identify data structures
778  */
779 
780 #define	NVME_IDENTIFY_BUFSIZE	4096	/* buffer size for Identify */
781 
782 /* NVMe Identify parameters (cdw10) */
783 #define	NVME_IDENTIFY_NSID		0x0	/* Identify Namespace */
784 #define	NVME_IDENTIFY_CTRL		0x1	/* Identify Controller */
785 #define	NVME_IDENTIFY_NSID_LIST		0x2	/* List Active Namespaces */
786 #define	NVME_IDENTIFY_NSID_DESC		0x3	/* Namespace ID Descriptors */
787 
788 #define	NVME_IDENTIFY_NSID_ALLOC_LIST	0x10	/* List Allocated NSID */
789 #define	NVME_IDENTIFY_NSID_ALLOC	0x11	/* Identify Allocated NSID */
790 #define	NVME_IDENTIFY_NSID_CTRL_LIST	0x12	/* List Controllers on NSID */
791 #define	NVME_IDENTIFY_CTRL_LIST		0x13	/* Controller List */
792 #define	NVME_IDENTIFY_PRIMARY_CAPS	0x14	/* Primary Controller Caps */
793 
794 
795 /* NVMe Queue Entry Size bitfield */
796 typedef struct {
797 	uint8_t qes_min:4;		/* minimum entry size */
798 	uint8_t qes_max:4;		/* maximum entry size */
799 } nvme_idctl_qes_t;
800 
801 /* NVMe Power State Descriptor */
802 typedef struct {
803 	uint16_t psd_mp;		/* Maximum Power */
804 	uint8_t psd_rsvd1;
805 	uint8_t psd_mps:1;		/* Max Power Scale (1.1) */
806 	uint8_t psd_nops:1;		/* Non-Operational State (1.1) */
807 	uint8_t psd_rsvd2:6;
808 	uint32_t psd_enlat;		/* Entry Latency */
809 	uint32_t psd_exlat;		/* Exit Latency */
810 	uint8_t psd_rrt:5;		/* Relative Read Throughput */
811 	uint8_t psd_rsvd3:3;
812 	uint8_t psd_rrl:5;		/* Relative Read Latency */
813 	uint8_t psd_rsvd4:3;
814 	uint8_t psd_rwt:5;		/* Relative Write Throughput */
815 	uint8_t	psd_rsvd5:3;
816 	uint8_t psd_rwl:5;		/* Relative Write Latency */
817 	uint8_t psd_rsvd6:3;
818 	uint16_t psd_idlp;		/* Idle Power (1.2) */
819 	uint8_t psd_rsvd7:6;
820 	uint8_t psd_ips:2;		/* Idle Power Scale (1.2) */
821 	uint8_t psd_rsvd8;
822 	uint16_t psd_actp;		/* Active Power (1.2) */
823 	uint8_t psd_apw:3;		/* Active Power Workload (1.2) */
824 	uint8_t psd_rsvd9:3;
825 	uint8_t psd_aps:2;		/* Active Power Scale */
826 	uint8_t psd_rsvd10[9];
827 } nvme_idctl_psd_t;
828 
829 #define	NVME_SERIAL_SZ	20
830 #define	NVME_MODEL_SZ	40
831 #define	NVME_FWVER_SZ	8
832 
833 /* NVMe Identify Controller Data Structure */
834 typedef struct {
835 	/* Controller Capabilities & Features */
836 	uint16_t id_vid;		/* PCI vendor ID */
837 	uint16_t id_ssvid;		/* PCI subsystem vendor ID */
838 	char id_serial[NVME_SERIAL_SZ];	/* Serial Number */
839 	char id_model[NVME_MODEL_SZ];	/* Model Number */
840 	char id_fwrev[NVME_FWVER_SZ];	/* Firmware Revision */
841 	uint8_t id_rab;			/* Recommended Arbitration Burst */
842 	uint8_t id_oui[3];		/* vendor IEEE OUI */
843 	struct {			/* Multi-Interface Capabilities */
844 		uint8_t m_multi_pci:1;	/* HW has multiple PCIe interfaces */
845 		uint8_t m_multi_ctrl:1; /* HW has multiple controllers (1.1) */
846 		uint8_t m_sr_iov:1;	/* Controller is SR-IOV virt fn (1.1) */
847 		uint8_t m_anar_sup:1;	/* ANA Reporting Supported (1.4) */
848 		uint8_t m_rsvd:4;
849 	} id_mic;
850 	uint8_t	id_mdts;		/* Maximum Data Transfer Size */
851 	uint16_t id_cntlid;		/* Unique Controller Identifier (1.1) */
852 	/* Added in NVMe 1.2 */
853 	uint32_t id_ver;		/* Version (1.2) */
854 	uint32_t id_rtd3r;		/* RTD3 Resume Latency (1.2) */
855 	uint32_t id_rtd3e;		/* RTD3 Entry Latency (1.2) */
856 	struct {
857 		uint32_t oaes_rsvd0:8;
858 		uint32_t oaes_nsan:1;	/* Namespace Attribute Notices (1.2) */
859 		uint32_t oaes_fwact:1;	/* Firmware Activation Notices (1.2) */
860 		uint32_t oaes_rsvd1:1;
861 		uint32_t oaes_ansacn:1;	/* Asymmetric NS Access Change (1.4) */
862 		uint32_t oaes_plat:1;	/* Predictable Lat Event Agg. (1.4) */
863 		uint32_t oaes_lbasi:1;	/* LBA Status Information (1.4) */
864 		uint32_t oaes_egeal:1;	/* Endurance Group Event Agg. (1.4) */
865 		uint32_t oaes_rsvd2:17;
866 	} id_oaes;
867 	struct {
868 		uint32_t ctrat_hid:1;	/* 128-bit Host Identifier (1.2)  */
869 		uint32_t ctrat_nops:1;	/* Non-Operational Power State (1.3) */
870 		uint32_t ctrat_nvmset:1; /* NVMe Sets (1.4) */
871 		uint32_t ctrat_rrl:1;	/* Read Recovery Levels (1.4) */
872 		uint32_t ctrat_engrp:1; /* Endurance Groups (1.4) */
873 		uint32_t ctrat_plm:1;	/* Predictable Latency Mode (1.4) */
874 		uint32_t ctrat_tbkas:1;	/* Traffic Based Keep Alive (1.4) */
875 		uint32_t ctrat_nsg:1;	/* Namespace Granularity (1.4) */
876 		uint32_t ctrat_sqass:1;	/* SQ Associations (1.4) */
877 		uint32_t ctrat_uuid:1;	/* UUID List (1.4) */
878 		uint32_t ctrat_rsvd:22;
879 	} id_ctratt;
880 	uint16_t id_rrls;		/* Read Recovery Levels (1.4) */
881 	uint8_t id_rsvd_cc[111-102];
882 	uint8_t id_cntrltype;		/* Controller Type (1.4) */
883 	uint8_t id_frguid[16];		/* FRU GUID (1.3) */
884 	uint16_t id_crdt1;		/* Command Retry Delay Time 1 (1.4) */
885 	uint16_t id_crdt2;		/* Command Retry Delay Time 2 (1.4) */
886 	uint16_t id_crdt3;		/* Command Retry Delay Time 3 (1.4) */
887 	uint8_t id_rsvd2_cc[240 - 134];
888 	uint8_t id_rsvd_nvmemi[253 - 240];
889 	/* NVMe-MI region */
890 	struct {			/* NVMe Subsystem Report */
891 		uint8_t nvmsr_nvmesd:1;	/* NVMe Storage Device */
892 		uint8_t nvmsr_nvmee:1;	/* NVMe Enclosure */
893 		uint8_t nvmsr_rsvd:6;
894 	} id_nvmsr;
895 	struct {			/* VPD Write Cycle Information */
896 		uint8_t vwci_crem:7;	/* Write Cycles Remaining */
897 		uint8_t vwci_valid:1;	/* Write Cycles Remaining Valid */
898 	} id_vpdwc;
899 	struct {			/* Management Endpoint Capabilities */
900 		uint8_t mec_smbusme:1;	/* SMBus Port Management Endpoint */
901 		uint8_t mec_pcieme:1;	/* PCIe Port Management Endpoint */
902 		uint8_t mec_rsvd:6;
903 	} id_mec;
904 
905 	/* Admin Command Set Attributes */
906 	struct {			/* Optional Admin Command Support */
907 		uint16_t oa_security:1;	/* Security Send & Receive */
908 		uint16_t oa_format:1;	/* Format NVM */
909 		uint16_t oa_firmware:1;	/* Firmware Activate & Download */
910 		uint16_t oa_nsmgmt:1;	/* Namespace Management (1.2) */
911 		uint16_t oa_selftest:1;	/* Self Test (1.3) */
912 		uint16_t oa_direct:1;	/* Directives (1.3) */
913 		uint16_t oa_nvmemi:1;	/* MI-Send/Recv (1.3) */
914 		uint16_t oa_virtmgmt:1;	/* Virtualization Management (1.3) */
915 		uint16_t oa_doorbell:1;	/* Doorbell Buffer Config (1.3) */
916 		uint16_t oa_lbastat:1;	/* LBA Status (1.4) */
917 		uint16_t oa_rsvd:6;
918 	} id_oacs;
919 	uint8_t	id_acl;			/* Abort Command Limit */
920 	uint8_t id_aerl;		/* Asynchronous Event Request Limit */
921 	struct {			/* Firmware Updates */
922 		uint8_t fw_readonly:1;	/* Slot 1 is Read-Only */
923 		uint8_t	fw_nslot:3;	/* number of firmware slots */
924 		uint8_t fw_norst:1;	/* Activate w/o reset (1.2) */
925 		uint8_t fw_rsvd:3;
926 	} id_frmw;
927 	struct {			/* Log Page Attributes */
928 		uint8_t lp_smart:1;	/* SMART/Health information per NS */
929 		uint8_t lp_cmdeff:1;	/* Command Effects (1.2) */
930 		uint8_t lp_extsup:1;	/* Extended Get Log Page (1.2) */
931 		uint8_t lp_telemetry:1;	/* Telemetry Log Pages (1.3) */
932 		uint8_t lp_persist:1;	/* Persistent Log Page (1.4) */
933 		uint8_t lp_rsvd:3;
934 	} id_lpa;
935 	uint8_t id_elpe;		/* Error Log Page Entries */
936 	uint8_t	id_npss;		/* Number of Power States */
937 	struct {			/* Admin Vendor Specific Command Conf */
938 		uint8_t av_spec:1;	/* use format from spec */
939 		uint8_t av_rsvd:7;
940 	} id_avscc;
941 	struct {			/* Autonomous Power State Trans (1.1) */
942 		uint8_t ap_sup:1;	/* APST supported (1.1) */
943 		uint8_t ap_rsvd:7;
944 	} id_apsta;
945 	uint16_t ap_wctemp;		/* Warning Composite Temp. (1.2) */
946 	uint16_t ap_cctemp;		/* Critical Composite Temp. (1.2) */
947 	uint16_t ap_mtfa;		/* Maximum Firmware Activation (1.2) */
948 	uint32_t ap_hmpre;		/* Host Memory Buf Pref Size (1.2) */
949 	uint32_t ap_hmmin;		/* Host Memory Buf Min Size (1.2) */
950 	nvme_uint128_t ap_tnvmcap;	/* Total NVM Capacity in Bytes (1.2) */
951 	nvme_uint128_t ap_unvmcap;	/* Unallocated NVM Capacity (1.2) */
952 	struct {			/* Replay Protected Mem. Block (1.2) */
953 		uint32_t rpmbs_units:3;	/* Number of targets */
954 		uint32_t rpmbs_auth:3;	/* Auth method */
955 		uint32_t rpmbs_rsvd:10;
956 		uint32_t rpmbs_tot:8;	/* Total size in 128KB */
957 		uint32_t rpmbs_acc:8;	/* Access size in 512B */
958 	} ap_rpmbs;
959 	/* Added in NVMe 1.3 */
960 	uint16_t ap_edstt;		/* Ext. Device Self-test time (1.3) */
961 	struct {			/* Device Self-test Options */
962 		uint8_t dsto_sub:1;	/* Subsystem level self-test (1.3) */
963 		uint8_t dsto_rsvd:7;
964 	} ap_dsto;
965 	uint8_t ap_fwug;		/* Firmware Update Granularity (1.3) */
966 	uint16_t ap_kas;		/* Keep Alive Support (1.2) */
967 	struct {			/* Host Thermal Management (1.3) */
968 		uint16_t hctma_hctm:1;	/* Host Controlled (1.3) */
969 		uint16_t hctma_rsvd:15;
970 	} ap_hctma;
971 	uint16_t ap_mntmt;		/* Minimum Thermal Temperature (1.3) */
972 	uint16_t ap_mxtmt;		/* Maximum Thermal Temperature (1.3) */
973 	struct {			/* Sanitize Caps */
974 		uint32_t san_ces:1;	/* Crypto Erase Support (1.3) */
975 		uint32_t san_bes:1;	/* Block Erase Support (1.3) */
976 		uint32_t san_ows:1;	/* Overwite Support (1.3) */
977 		uint32_t san_rsvd:26;
978 		uint32_t san_ndi:1;	/* No-deallocate Inhibited (1.4) */
979 		uint32_t san_nodmmas:2;	/* No-Deallocate Modifies Media (1.4) */
980 	} ap_sanitize;
981 	uint32_t ap_hmminds;		/* Host Mem Buf Min Desc Entry (1.4) */
982 	uint16_t ap_hmmaxd;		/* How Mem Max Desc Entries (1.4) */
983 	uint16_t ap_nsetidmax;		/* Max NVMe set identifier (1.4) */
984 	uint16_t ap_engidmax;		/* Max Endurance Group ID (1.4) */
985 	uint8_t ap_anatt;		/* ANA Transition Time (1.4) */
986 	struct {			/* Asymmetric Namespace Access Caps */
987 		uint8_t anacap_opt:1;	/* Optimized State (1.4) */
988 		uint8_t anacap_unopt:1;	/* Un-optimized State (1.4) */
989 		uint8_t anacap_inacc:1;	/* Inaccessible State (1.4) */
990 		uint8_t anacap_ploss:1;	/* Persistent Loss (1.4) */
991 		uint8_t anacap_chg:1;	/* Change State (1.4 ) */
992 		uint8_t anacap_rsvd:1;
993 		uint8_t anacap_grpns:1;	/* ID Changes with NS Attach (1.4) */
994 		uint8_t anacap_grpid:1;	/* Supports Group ID (1.4) */
995 	} ap_anacap;
996 	uint32_t ap_anagrpmax;		/* ANA Group ID Max (1.4) */
997 	uint32_t ap_nanagrpid;		/* Number of ANA Group IDs (1.4) */
998 	uint32_t ap_pels;		/* Persistent Event Log Size (1.4) */
999 	uint8_t id_rsvd_ac[512 - 356];
1000 
1001 	/* NVM Command Set Attributes */
1002 	nvme_idctl_qes_t id_sqes;	/* Submission Queue Entry Size */
1003 	nvme_idctl_qes_t id_cqes;	/* Completion Queue Entry Size */
1004 	uint16_t id_maxcmd;		/* Max Outstanding Commands (1.3) */
1005 	uint32_t id_nn;			/* Number of Namespaces */
1006 	struct {			/* Optional NVM Command Support */
1007 		uint16_t on_compare:1;	/* Compare */
1008 		uint16_t on_wr_unc:1;	/* Write Uncorrectable */
1009 		uint16_t on_dset_mgmt:1; /* Dataset Management */
1010 		uint16_t on_wr_zero:1;	/* Write Zeros (1.1) */
1011 		uint16_t on_save:1;	/* Save/Select in Get/Set Feat (1.1) */
1012 		uint16_t on_reserve:1;	/* Reservations (1.1) */
1013 		uint16_t on_ts:1;	/* Timestamp (1.3) */
1014 		uint16_t on_verify:1;	/* Verify (1.4) */
1015 		uint16_t on_rsvd:8;
1016 	} id_oncs;
1017 	struct {			/* Fused Operation Support */
1018 		uint16_t f_cmp_wr:1;	/* Compare and Write */
1019 		uint16_t f_rsvd:15;
1020 	} id_fuses;
1021 	struct {			/* Format NVM Attributes */
1022 		uint8_t fn_format:1;	/* Format applies to all NS */
1023 		uint8_t fn_sec_erase:1;	/* Secure Erase applies to all NS */
1024 		uint8_t fn_crypt_erase:1; /* Cryptographic Erase supported */
1025 		uint8_t fn_rsvd:5;
1026 	} id_fna;
1027 	struct {			/* Volatile Write Cache */
1028 		uint8_t vwc_present:1;	/* Volatile Write Cache present */
1029 		uint8_t vwc_nsflush:2;	/* Flush with NS ffffffff (1.4) */
1030 		uint8_t rsvd:5;
1031 	} id_vwc;
1032 	uint16_t id_awun;		/* Atomic Write Unit Normal */
1033 	uint16_t id_awupf;		/* Atomic Write Unit Power Fail */
1034 	struct {			/* NVM Vendor Specific Command Conf */
1035 		uint8_t nv_spec:1;	/* use format from spec */
1036 		uint8_t nv_rsvd:7;
1037 	} id_nvscc;
1038 	struct {			/* Namespace Write Protection Caps */
1039 		uint8_t nwpc_base:1;	/* Base support (1.4) */
1040 		uint8_t nwpc_wpupc:1;	/* Write prot until power cycle (1.4) */
1041 		uint8_t nwpc_permwp:1;	/* Permanent write prot (1.4) */
1042 		uint8_t nwpc_rsvd:5;
1043 	} id_nwpc;
1044 	uint16_t id_acwu;		/* Atomic Compare & Write Unit (1.1) */
1045 	uint16_t id_rsvd_nc_3;
1046 	struct {			/* SGL Support (1.1) */
1047 		uint16_t sgl_sup:2;	/* SGL Supported in NVM cmds (1.3) */
1048 		uint16_t sgl_keyed:1;	/* Keyed SGL Support (1.2) */
1049 		uint16_t sgl_rsvd1:13;
1050 		uint16_t sgl_bucket:1;	/* SGL Bit Bucket supported (1.1) */
1051 		uint16_t sgl_balign:1;	/* SGL Byte Aligned (1.2) */
1052 		uint16_t sgl_sglgtd:1;	/* SGL Length Longer than Data (1.2) */
1053 		uint16_t sgl_mptr:1;	/* SGL MPTR w/ SGL (1.2) */
1054 		uint16_t sgl_offset:1;	/* SGL Address is offset (1.2) */
1055 		uint16_t sgl_tport:1;	/* Transport SGL Data Block (1.4) */
1056 		uint16_t sgl_rsvd2:10;
1057 	} id_sgls;
1058 	uint32_t id_mnan;		/* Maximum Number of Allowed NSes */
1059 	uint8_t id_rsvd_nc_4[768 - 544];
1060 
1061 	/* I/O Command Set Attributes */
1062 	uint8_t id_subnqn[1024 - 768];	/* Subsystem Qualified Name (1.2.1+) */
1063 	uint8_t id_rsvd_ioc[1792 - 1024];
1064 	uint8_t id_nvmof[2048 - 1792];	/* NVMe over Fabrics */
1065 
1066 	/* Power State Descriptors */
1067 	nvme_idctl_psd_t id_psd[32];
1068 
1069 	/* Vendor Specific */
1070 	uint8_t id_vs[1024];
1071 } nvme_identify_ctrl_t;
1072 
1073 /*
1074  * NVMe Controller Types
1075  */
1076 #define	NVME_CNTRLTYPE_RSVD	0
1077 #define	NVME_CNTRLTYPE_IO	1
1078 #define	NVME_CNTRLTYPE_DISC	2
1079 #define	NVME_CNTRLTYPE_ADMIN	3
1080 
1081 /*
1082  * RPMBS Authentication Types
1083  */
1084 #define	NVME_RPMBS_AUTH_HMAC_SHA256	0
1085 
1086 /*
1087  * NODMMAS Values
1088  */
1089 #define	NVME_NODMMAS_UNDEF	0x00
1090 #define	NVME_NODMMAS_NOMOD	0x01
1091 #define	NVME_NODMMAS_DOMOD	0x02
1092 
1093 /*
1094  * VWC NSID flushes
1095  */
1096 #define	NVME_VWCNS_UNKNOWN	0x00
1097 #define	NVME_VWCNS_UNSUP	0x02
1098 #define	NVME_VWCNS_SUP		0x03
1099 
1100 /*
1101  * SGL Support Values
1102  */
1103 #define	NVME_SGL_UNSUP		0x00
1104 #define	NVME_SGL_SUP_UNALIGN	0x01
1105 #define	NVME_SGL_SUP_ALIGN	0x02
1106 
1107 /* NVMe Identify Namespace LBA Format */
1108 typedef struct {
1109 	uint16_t lbaf_ms;		/* Metadata Size */
1110 	uint8_t lbaf_lbads;		/* LBA Data Size */
1111 	uint8_t lbaf_rp:2;		/* Relative Performance */
1112 	uint8_t lbaf_rsvd1:6;
1113 } nvme_idns_lbaf_t;
1114 
1115 #define	NVME_MAX_LBAF	16
1116 
1117 /* NVMe Identify Namespace Data Structure */
1118 typedef struct {
1119 	uint64_t id_nsize;		/* Namespace Size */
1120 	uint64_t id_ncap;		/* Namespace Capacity */
1121 	uint64_t id_nuse;		/* Namespace Utilization */
1122 	struct {			/* Namespace Features */
1123 		uint8_t f_thin:1;	/* Thin Provisioning */
1124 		uint8_t f_nsabp:1;	/* Namespace atomics (1.2) */
1125 		uint8_t f_dae:1;	/* Deallocated errors supported (1.2) */
1126 		uint8_t f_uidreuse:1;	/* GUID reuse impossible (1.3) */
1127 		uint8_t f_optperf:1;	/* Namespace I/O opt (1.4) */
1128 		uint8_t f_rsvd:3;
1129 	} id_nsfeat;
1130 	uint8_t id_nlbaf;		/* Number of LBA formats */
1131 	struct {			/* Formatted LBA size */
1132 		uint8_t lba_format:4;	/* LBA format */
1133 		uint8_t lba_extlba:1;	/* extended LBA (includes metadata) */
1134 		uint8_t lba_rsvd:3;
1135 	} id_flbas;
1136 	struct {			/* Metadata Capabilities */
1137 		uint8_t mc_extlba:1;	/* extended LBA transfers */
1138 		uint8_t mc_separate:1;	/* separate metadata transfers */
1139 		uint8_t mc_rsvd:6;
1140 	} id_mc;
1141 	struct {			/* Data Protection Capabilities */
1142 		uint8_t dp_type1:1;	/* Protection Information Type 1 */
1143 		uint8_t dp_type2:1;	/* Protection Information Type 2 */
1144 		uint8_t dp_type3:1;	/* Protection Information Type 3 */
1145 		uint8_t dp_first:1;	/* first 8 bytes of metadata */
1146 		uint8_t dp_last:1;	/* last 8 bytes of metadata */
1147 		uint8_t dp_rsvd:3;
1148 	} id_dpc;
1149 	struct {			/* Data Protection Settings */
1150 		uint8_t dp_pinfo:3;	/* Protection Information enabled */
1151 		uint8_t dp_first:1;	/* first 8 bytes of metadata */
1152 		uint8_t dp_rsvd:4;
1153 	} id_dps;
1154 	struct {			/* NS Multi-Path/Sharing Cap (1.1) */
1155 		uint8_t nm_shared:1;	/* NS is shared (1.1) */
1156 		uint8_t nm_rsvd:7;
1157 	} id_nmic;
1158 	struct {			/* Reservation Capabilities (1.1) */
1159 		uint8_t rc_persist:1;	/* Persist Through Power Loss (1.1) */
1160 		uint8_t rc_wr_excl:1;	/* Write Exclusive (1.1) */
1161 		uint8_t rc_excl:1;	/* Exclusive Access (1.1) */
1162 		uint8_t rc_wr_excl_r:1;	/* Wr Excl - Registrants Only (1.1) */
1163 		uint8_t rc_excl_r:1;	/* Excl Acc - Registrants Only (1.1) */
1164 		uint8_t rc_wr_excl_a:1;	/* Wr Excl - All Registrants (1.1) */
1165 		uint8_t rc_excl_a:1;	/* Excl Acc - All Registrants (1.1) */
1166 		uint8_t rc_ign_ekey:1;	/* Ignore Existing Key (1.3) */
1167 	} id_rescap;
1168 	struct {			/* Format Progress Indicator (1.2) */
1169 		uint8_t fpi_remp:7;	/* Percent NVM Format Remaining (1.2) */
1170 		uint8_t fpi_sup:1;	/* Supported (1.2) */
1171 	} id_fpi;
1172 	uint8_t id_dfleat;		/* Deallocate Log. Block (1.3) */
1173 	uint16_t id_nawun;		/* Atomic Write Unit Normal (1.2) */
1174 	uint16_t id_nawupf;		/* Atomic Write Unit Power Fail (1.2) */
1175 	uint16_t id_nacwu;		/* Atomic Compare & Write Unit (1.2) */
1176 	uint16_t id_nabsn;		/* Atomic Boundary Size Normal (1.2) */
1177 	uint16_t id_nbao;		/* Atomic Boundary Offset (1.2) */
1178 	uint16_t id_nabspf;		/* Atomic Boundary Size Fail (1.2) */
1179 	uint16_t id_noiob;		/* Optimal I/O Bondary (1.3) */
1180 	nvme_uint128_t id_nvmcap;	/* NVM Capacity */
1181 	uint16_t id_npwg;		/* NS Pref. Write Gran. (1.4) */
1182 	uint16_t id_npwa;		/* NS Pref. Write Align. (1.4) */
1183 	uint16_t id_npdg;		/* NS Pref. Deallocate Gran. (1.4) */
1184 	uint16_t id_npda;		/* NS Pref. Deallocate Align. (1.4) */
1185 	uint16_t id_nows;		/* NS. Optimal Write Size (1.4) */
1186 	uint8_t id_rsvd1[92 - 74];
1187 	uint32_t id_anagrpid;		/* ANA Group Identifier (1.4) */
1188 	uint8_t id_rsvd2[99 - 96];
1189 	struct {
1190 		uint8_t nsa_wprot:1;	/* Write Protected (1.4) */
1191 		uint8_t nsa_rsvd:7;
1192 	} id_nsattr;
1193 	uint16_t id_nvmsetid;		/* NVM Set Identifier (1.4) */
1194 	uint16_t id_endgid;		/* Endurance Group Identifier (1.4) */
1195 	uint8_t id_nguid[16];		/* Namespace GUID (1.2) */
1196 	uint8_t id_eui64[8];		/* IEEE Extended Unique Id (1.1) */
1197 	nvme_idns_lbaf_t id_lbaf[NVME_MAX_LBAF];	/* LBA Formats */
1198 
1199 	uint8_t id_rsvd3[384 - 192];
1200 
1201 	uint8_t id_vs[4096 - 384];	/* Vendor Specific */
1202 } nvme_identify_nsid_t;
1203 
1204 /* NVMe Identify Namespace ID List */
1205 typedef struct {
1206 					/* Ordered list of Namespace IDs */
1207 	uint32_t nl_nsid[NVME_IDENTIFY_BUFSIZE / sizeof (uint32_t)];
1208 } nvme_identify_nsid_list_t;
1209 
1210 /* NVME Identify Controller ID List */
1211 typedef struct {
1212 	uint16_t	cl_nid;		/* Number of controller entries */
1213 					/* unique controller identifiers */
1214 	uint16_t	cl_ctlid[NVME_IDENTIFY_BUFSIZE / sizeof (uint16_t) - 1];
1215 } nvme_identify_ctrl_list_t;
1216 
1217 /* NVMe Identify Namespace Descriptor */
1218 typedef struct {
1219 	uint8_t nd_nidt;		/* Namespace Identifier Type */
1220 	uint8_t nd_nidl;		/* Namespace Identifier Length */
1221 	uint8_t nd_resv[2];
1222 	uint8_t nd_nid[];		/* Namespace Identifier */
1223 } nvme_identify_nsid_desc_t;
1224 
1225 #define	NVME_NSID_DESC_EUI64	1
1226 #define	NVME_NSID_DESC_NGUID	2
1227 #define	NVME_NSID_DESC_NUUID	3
1228 #define	NVME_NSID_DESC_MIN	NVME_NSID_DESC_EUI64
1229 #define	NVME_NSID_DESC_MAX	NVME_NSID_DESC_NUUID
1230 
1231 #define	NVME_NSID_DESC_LEN_EUI64	8
1232 #define	NVME_NSID_DESC_LEN_NGUID	16
1233 #define	NVME_NSID_DESC_LEN_NUUID	UUID_LEN
1234 
1235 /* NVMe Identify Primary Controller Capabilities */
1236 typedef struct {
1237 	uint16_t	nipc_cntlid;	/* Controller ID */
1238 	uint16_t	nipc_portid;	/* Port Identifier */
1239 	uint8_t		nipc_crt;	/* Controller Resource Types */
1240 	uint8_t		nipc_rsvd0[32 - 5];
1241 	uint32_t	nipc_vqfrt;	/* VQ Resources Flexible Total */
1242 	uint32_t	nipc_vqrfa;	/* VQ Resources Flexible Assigned */
1243 	uint16_t	nipc_vqrfap;	/* VQ Resources to Primary */
1244 	uint16_t	nipc_vqprt;	/* VQ Resources Private Total */
1245 	uint16_t	nipc_vqfrsm;	/* VQ Resources Secondary Max */
1246 	uint16_t	nipc_vqgran;	/* VQ Flexible Resource Gran */
1247 	uint8_t		nipc_rvsd1[64 - 48];
1248 	uint32_t	nipc_vifrt;	/* VI Flexible total */
1249 	uint32_t	nipc_virfa;	/* VI Flexible Assigned */
1250 	uint16_t	nipc_virfap;	/* VI Flexible Allocated to Primary */
1251 	uint16_t	nipc_viprt;	/* VI Resources Private Total */
1252 	uint16_t	nipc_vifrsm;	/* VI Resources Secondary Max */
1253 	uint16_t	nipc_vigran;	/* VI Flexible Granularity */
1254 	uint8_t		nipc_rsvd2[4096 - 80];
1255 } nvme_identify_primary_caps_t;
1256 
1257 /*
1258  * NVMe completion queue entry status field
1259  */
1260 typedef struct {
1261 	uint16_t sf_p:1;		/* Phase Tag */
1262 	uint16_t sf_sc:8;		/* Status Code */
1263 	uint16_t sf_sct:3;		/* Status Code Type */
1264 	uint16_t sf_rsvd2:2;
1265 	uint16_t sf_m:1;		/* More */
1266 	uint16_t sf_dnr:1;		/* Do Not Retry */
1267 } nvme_cqe_sf_t;
1268 
1269 
1270 /*
1271  * NVMe Get Log Page
1272  */
1273 #define	NVME_LOGPAGE_SUP	0x00	/* Supported Logs (2.0) */
1274 #define	NVME_LOGPAGE_ERROR	0x01	/* Error Information */
1275 #define	NVME_LOGPAGE_HEALTH	0x02	/* SMART/Health Information */
1276 #define	NVME_LOGPAGE_FWSLOT	0x03	/* Firmware Slot Information */
1277 #define	NVME_LOGPAGE_NSCHANGE	0x04	/* Changed namespace (1.2) */
1278 #define	NVME_LOGPAGE_CMDSUP	0x05	/* Cmds. Supported and Effects (1.3) */
1279 #define	NVME_LOGPAGE_SELFTEST	0x06	/* Device self-test (1.3) */
1280 #define	NVME_LOGPAGE_TELMHOST	0x07	/* Telemetry Host-Initiated */
1281 #define	NVME_LOGPAGE_TELMCTRL	0x08	/* Telemetry Controller-Initiated */
1282 #define	NVME_LOGPAGE_ENDGRP	0x09	/* Endurance Group Information (1.4) */
1283 #define	NVME_LOGPAGE_PLATSET	0x0a	/* Predictable Lat. per NVM Set (1.4) */
1284 #define	NVME_LOGPAGE_PLATAGG	0x0b	/* Predictable Lat. Event Agg (1.4) */
1285 #define	NVME_LOGPAGE_ASYMNS	0x0c	/* Asymmetric Namespace Access (1.4) */
1286 #define	NVME_LOGPAGE_PEVLOG	0x0d	/* Persistent Event Log (1.4) */
1287 #define	NVME_LOGPAGE_LBASTS	0x0e	/* LBA Status Information (1.4) */
1288 #define	NVME_LOGPAGE_ENDAGG	0x0f	/* Endurance Group Event Agg. (1.4) */
1289 
1290 #define	NVME_LOGPAGE_VEND_MIN	0xc0
1291 #define	NVME_LOGPAGE_VEND_MAX	0xff
1292 
1293 /*
1294  * Supported Log Pages (2.0). There is one entry of an nvme_logsup_t that then
1295  * exists on a per-log basis.
1296  */
1297 
1298 /*
1299  * The NVMe Log Identifier specific parameter field. Currently there is only one
1300  * defined field for the persistent event log (pel).
1301  */
1302 typedef union {
1303 	uint16_t nsl_lidsp;		/* Raw Value */
1304 	struct {			/* Persistent Event Log */
1305 		uint16_t nsl_ec512:1;
1306 		uint16_t nsl_pel_rsvd0p1:15;
1307 	} nsl_pel;
1308 } nvme_suplog_lidsp_t;
1309 
1310 typedef struct {
1311 	uint16_t ns_lsupp:1;
1312 	uint16_t ns_ios:1;
1313 	uint16_t ns_rsvd0p2:14;
1314 	nvme_suplog_lidsp_t ns_lidsp;
1315 } nvme_suplog_t;
1316 
1317 CTASSERT(sizeof (nvme_suplog_lidsp_t) == 2);
1318 CTASSERT(sizeof (nvme_suplog_t) == 4);
1319 
1320 typedef struct {
1321 	nvme_suplog_t	nl_logs[256];
1322 } nvme_suplog_log_t;
1323 
1324 CTASSERT(sizeof (nvme_suplog_log_t) == 1024);
1325 
1326 /*
1327  * SMART / Health information
1328  */
1329 typedef struct {
1330 	uint64_t el_count;		/* Error Count */
1331 	uint16_t el_sqid;		/* Submission Queue ID */
1332 	uint16_t el_cid;		/* Command ID */
1333 	nvme_cqe_sf_t el_sf;		/* Status Field */
1334 	uint8_t	el_byte;		/* Parameter Error Location byte */
1335 	uint8_t	el_bit:3;		/* Parameter Error Location bit */
1336 	uint8_t el_rsvd1:5;
1337 	uint64_t el_lba;		/* Logical Block Address */
1338 	uint32_t el_nsid;		/* Namespace ID */
1339 	uint8_t	el_vendor;		/* Vendor Specific Information avail */
1340 	uint8_t el_rsvd2[64 - 29];
1341 } nvme_error_log_entry_t;
1342 
1343 typedef struct {
1344 	struct {			/* Critical Warning */
1345 		uint8_t cw_avail:1;	/* available space too low */
1346 		uint8_t cw_temp:1;	/* temperature too high */
1347 		uint8_t cw_reliab:1;	/* degraded reliability */
1348 		uint8_t cw_readonly:1;	/* media is read-only */
1349 		uint8_t cw_volatile:1;	/* volatile memory backup failed */
1350 		uint8_t cw_rsvd:3;
1351 	} hl_crit_warn;
1352 	uint16_t hl_temp;		/* Temperature */
1353 	uint8_t hl_avail_spare;		/* Available Spare */
1354 	uint8_t hl_avail_spare_thr;	/* Available Spare Threshold */
1355 	uint8_t hl_used;		/* Percentage Used */
1356 	uint8_t hl_rsvd1[32 - 6];
1357 	nvme_uint128_t hl_data_read;	/* Data Units Read */
1358 	nvme_uint128_t hl_data_write;	/* Data Units Written */
1359 	nvme_uint128_t hl_host_read;	/* Host Read Commands */
1360 	nvme_uint128_t hl_host_write;	/* Host Write Commands */
1361 	nvme_uint128_t hl_ctrl_busy;	/* Controller Busy Time */
1362 	nvme_uint128_t hl_power_cycles;	/* Power Cycles */
1363 	nvme_uint128_t hl_power_on_hours; /* Power On Hours */
1364 	nvme_uint128_t hl_unsafe_shutdn; /* Unsafe Shutdowns */
1365 	nvme_uint128_t hl_media_errors;	/* Media Errors */
1366 	nvme_uint128_t hl_errors_logged; /* Number of errors logged */
1367 	/* Added in NVMe 1.2 */
1368 	uint32_t hl_warn_temp_time;	/* Warning Composite Temp Time */
1369 	uint32_t hl_crit_temp_time;	/* Critical Composite Temp Time */
1370 	uint16_t hl_temp_sensor_1;	/* Temperature Sensor 1 */
1371 	uint16_t hl_temp_sensor_2;	/* Temperature Sensor 2 */
1372 	uint16_t hl_temp_sensor_3;	/* Temperature Sensor 3 */
1373 	uint16_t hl_temp_sensor_4;	/* Temperature Sensor 4 */
1374 	uint16_t hl_temp_sensor_5;	/* Temperature Sensor 5 */
1375 	uint16_t hl_temp_sensor_6;	/* Temperature Sensor 6 */
1376 	uint16_t hl_temp_sensor_7;	/* Temperature Sensor 7 */
1377 	uint16_t hl_temp_sensor_8;	/* Temperature Sensor 8 */
1378 	/* Added in NVMe 1.3 */
1379 	uint32_t hl_tmtemp_1_tc;	/* Thermal Mgmt Temp 1 Transition # */
1380 	uint32_t hl_tmtemp_2_tc;	/* Thermal Mgmt Temp 1 Transition # */
1381 	uint32_t hl_tmtemp_1_time;	/* Time in Thermal Mgmt Temp 1 */
1382 	uint32_t hl_tmtemp_2_time;	/* Time in Thermal Mgmt Temp 2 */
1383 	uint8_t hl_rsvd2[512 - 232];
1384 } nvme_health_log_t;
1385 
1386 /*
1387  * The NVMe spec allows for up to seven firmware slots.
1388  */
1389 #define	NVME_MAX_FWSLOTS	7
1390 
1391 typedef struct {
1392 	/* Active Firmware Slot */
1393 	uint8_t fw_afi:3;
1394 	uint8_t fw_rsvd1:1;
1395 	/* Next Active Firmware Slot */
1396 	uint8_t fw_next:3;
1397 	uint8_t fw_rsvd2:1;
1398 	uint8_t fw_rsvd3[7];
1399 	/* Firmware Revision / Slot */
1400 	char fw_frs[NVME_MAX_FWSLOTS][NVME_FWVER_SZ];
1401 	uint8_t fw_rsvd4[512 - 64];
1402 } nvme_fwslot_log_t;
1403 
1404 /*
1405  * The NVMe spec specifies that the changed namespace list contains up to
1406  * 1024 entries.
1407  */
1408 #define	NVME_NSCHANGE_LIST_SIZE	1024
1409 
1410 typedef struct {
1411 	uint32_t	nscl_ns[NVME_NSCHANGE_LIST_SIZE];
1412 } nvme_nschange_list_t;
1413 
1414 /*
1415  * Commands Supported and Effects log page and information structure. This was
1416  * an optional log page added in NVMe 1.2.
1417  */
1418 typedef struct {
1419 	uint8_t cmd_csupp:1;	/* Command supported */
1420 	uint8_t cmd_lbcc:1;	/* Logical block content change */
1421 	uint8_t cmd_ncc:1;	/* Namespace capability change */
1422 	uint8_t cmd_nic:1;	/* Namespace inventory change */
1423 	uint8_t cmd_ccc:1;	/* Controller capability change */
1424 	uint8_t cmd_rsvd0p5:3;
1425 	uint8_t cmd_rsvd1;
1426 	uint16_t cmd_cse:3;	/* Command submission and execution */
1427 	uint16_t cmd_uuid:1;	/* UUID select supported, 1.4 */
1428 	uint16_t cmd_csp:12;	/* Command Scope, 2.0 */
1429 } nvme_cmdeff_t;
1430 
1431 CTASSERT(sizeof (nvme_cmdeff_t) == 4);
1432 
1433 typedef enum {
1434 	NVME_CMDEFF_CSP_NS		= 1 << 0,
1435 	NVME_CMDEFF_CSP_CTRL		= 1 << 1,
1436 	NVME_CMDEFF_CSP_SET		= 1 << 2,
1437 	NVME_CMDEFF_CSP_ENDURANCE	= 1 << 3,
1438 	NVME_CMDEFF_CSP_DOMAIN		= 1 << 4,
1439 	NVME_CMDEFF_CSP_NVM		= 1 << 5
1440 } nvme_cmdeff_csp_t;
1441 
1442 typedef enum {
1443 	NVME_CMDEFF_CSE_NONE	= 0,
1444 	NVME_CMDEFF_CSE_NS,
1445 	NVME_CMDEFF_CSE_CTRL
1446 } nvme_cmdeff_cse_t;
1447 
1448 typedef struct {
1449 	nvme_cmdeff_t	cme_admin[256];
1450 	nvme_cmdeff_t	cme_io[256];
1451 	uint8_t		cme_rsvd2048[2048];
1452 } nvme_cmdeff_log_t;
1453 
1454 CTASSERT(sizeof (nvme_cmdeff_log_t) == 4096);
1455 CTASSERT(offsetof(nvme_cmdeff_log_t, cme_rsvd2048) == 2048);
1456 
1457 /*
1458  * NVMe Format NVM
1459  */
1460 #define	NVME_FRMT_SES_NONE	0
1461 #define	NVME_FRMT_SES_USER	1
1462 #define	NVME_FRMT_SES_CRYPTO	2
1463 #define	NVME_FRMT_MAX_SES	2
1464 
1465 #define	NVME_FRMT_MAX_LBAF	15
1466 
1467 typedef union {
1468 	struct {
1469 		uint32_t fm_lbaf:4;		/* LBA Format */
1470 		uint32_t fm_ms:1;		/* Metadata Settings */
1471 		uint32_t fm_pi:3;		/* Protection Information */
1472 		uint32_t fm_pil:1;		/* Prot. Information Location */
1473 		uint32_t fm_ses:3;		/* Secure Erase Settings */
1474 		uint32_t fm_resvd:20;
1475 	} b;
1476 	uint32_t r;
1477 } nvme_format_nvm_t;
1478 
1479 
1480 /*
1481  * NVMe Get / Set Features
1482  */
1483 #define	NVME_FEAT_ARBITRATION	0x1	/* Command Arbitration */
1484 #define	NVME_FEAT_POWER_MGMT	0x2	/* Power Management */
1485 #define	NVME_FEAT_LBA_RANGE	0x3	/* LBA Range Type */
1486 #define	NVME_FEAT_TEMPERATURE	0x4	/* Temperature Threshold */
1487 #define	NVME_FEAT_ERROR		0x5	/* Error Recovery */
1488 #define	NVME_FEAT_WRITE_CACHE	0x6	/* Volatile Write Cache */
1489 #define	NVME_FEAT_NQUEUES	0x7	/* Number of Queues */
1490 #define	NVME_FEAT_INTR_COAL	0x8	/* Interrupt Coalescing */
1491 #define	NVME_FEAT_INTR_VECT	0x9	/* Interrupt Vector Configuration */
1492 #define	NVME_FEAT_WRITE_ATOM	0xa	/* Write Atomicity */
1493 #define	NVME_FEAT_ASYNC_EVENT	0xb	/* Asynchronous Event Configuration */
1494 #define	NVME_FEAT_AUTO_PST	0xc	/* Autonomous Power State Transition */
1495 					/* (1.1) */
1496 
1497 #define	NVME_FEAT_PROGRESS	0x80	/* Software Progress Marker */
1498 
1499 /*
1500  * This enumeration represents the capabilities in the Get Features select / Set
1501  * Features save options. This was introduced in NVMe 1.1 and the values below
1502  * match the specification. An optional feature in the identify controller data
1503  * structure is set to indicate that this is supported (id_oncs.on_save).
1504  */
1505 typedef enum {
1506 	NVME_FEATURE_SEL_CURRENT	= 0,
1507 	NVME_FEATURE_SEL_DEFAULT,
1508 	NVME_FEATURE_SEL_SAVED,
1509 	NVME_FEATURE_SEL_SUPPORTED
1510 } nvme_feature_sel_t;
1511 
1512 typedef union {
1513 	struct {
1514 		uint32_t gt_fid:8;	/* Feature ID */
1515 		uint32_t gt_sel:3;	/* Select */
1516 		uint32_t gt_rsvd:21;
1517 	} b;
1518 	uint32_t r;
1519 } nvme_get_features_dw10_t;
1520 
1521 /* Arbitration Feature */
1522 typedef union {
1523 	struct {
1524 		uint8_t arb_ab:3;	/* Arbitration Burst */
1525 		uint8_t arb_rsvd:5;
1526 		uint8_t arb_lpw;	/* Low Priority Weight */
1527 		uint8_t arb_mpw;	/* Medium Priority Weight */
1528 		uint8_t arb_hpw;	/* High Priority Weight */
1529 	} b;
1530 	uint32_t r;
1531 } nvme_arbitration_t;
1532 
1533 /* Power Management Feature */
1534 typedef union {
1535 	struct {
1536 		uint32_t pm_ps:5;	/* Power State */
1537 		uint32_t pm_rsvd:27;
1538 	} b;
1539 	uint32_t r;
1540 } nvme_power_mgmt_t;
1541 
1542 /* LBA Range Type Feature */
1543 typedef union {
1544 	struct {
1545 		uint32_t lr_num:6;	/* Number of LBA ranges */
1546 		uint32_t lr_rsvd:26;
1547 	} b;
1548 	uint32_t r;
1549 } nvme_lba_range_type_t;
1550 
1551 typedef struct {
1552 	uint8_t lr_type;		/* Type */
1553 	struct {			/* Attributes */
1554 		uint8_t lr_write:1;	/* may be overwritten */
1555 		uint8_t lr_hidden:1;	/* hidden from OS/EFI/BIOS */
1556 		uint8_t lr_rsvd1:6;
1557 	} lr_attr;
1558 	uint8_t lr_rsvd2[14];
1559 	uint64_t lr_slba;		/* Starting LBA */
1560 	uint64_t lr_nlb;		/* Number of Logical Blocks */
1561 	uint8_t lr_guid[16];		/* Unique Identifier */
1562 	uint8_t lr_rsvd3[16];
1563 } nvme_lba_range_t;
1564 
1565 #define	NVME_LBA_RANGE_BUFSIZE	4096
1566 
1567 /* Temperature Threshold Feature */
1568 typedef union {
1569 	struct {
1570 		uint16_t tt_tmpth;	/* Temperature Threshold */
1571 		uint16_t tt_tmpsel:4;	/* Temperature Select */
1572 		uint16_t tt_thsel:2;	/* Temperature Type */
1573 		uint16_t tt_resv:10;
1574 	} b;
1575 	uint32_t r;
1576 } nvme_temp_threshold_t;
1577 
1578 #define	NVME_TEMP_THRESH_MAX_SENSOR	8
1579 #define	NVME_TEMP_THRESH_ALL	0xf
1580 #define	NVME_TEMP_THRESH_OVER	0x00
1581 #define	NVME_TEMP_THRESH_UNDER	0x01
1582 
1583 /* Error Recovery Feature */
1584 typedef union {
1585 	struct {
1586 		uint16_t er_tler;	/* Time-Limited Error Recovery */
1587 		uint16_t er_rsvd;
1588 	} b;
1589 	uint32_t r;
1590 } nvme_error_recovery_t;
1591 
1592 /* Volatile Write Cache Feature */
1593 typedef union {
1594 	struct {
1595 		uint32_t wc_wce:1;	/* Volatile Write Cache Enable */
1596 		uint32_t wc_rsvd:31;
1597 	} b;
1598 	uint32_t r;
1599 } nvme_write_cache_t;
1600 
1601 /* Number of Queues Feature */
1602 typedef union {
1603 	struct {
1604 		uint16_t nq_nsq;	/* Number of Submission Queues */
1605 		uint16_t nq_ncq;	/* Number of Completion Queues */
1606 	} b;
1607 	uint32_t r;
1608 } nvme_nqueues_t;
1609 
1610 /* Interrupt Coalescing Feature */
1611 typedef union {
1612 	struct {
1613 		uint8_t ic_thr;		/* Aggregation Threshold */
1614 		uint8_t ic_time;	/* Aggregation Time */
1615 		uint16_t ic_rsvd;
1616 	} b;
1617 	uint32_t r;
1618 } nvme_intr_coal_t;
1619 
1620 /* Interrupt Configuration Features */
1621 typedef union {
1622 	struct {
1623 		uint16_t iv_iv;		/* Interrupt Vector */
1624 		uint16_t iv_cd:1;	/* Coalescing Disable */
1625 		uint16_t iv_rsvd:15;
1626 	} b;
1627 	uint32_t r;
1628 } nvme_intr_vect_t;
1629 
1630 /* Write Atomicity Feature */
1631 typedef union {
1632 	struct {
1633 		uint32_t wa_dn:1;	/* Disable Normal */
1634 		uint32_t wa_rsvd:31;
1635 	} b;
1636 	uint32_t r;
1637 } nvme_write_atomicity_t;
1638 
1639 /* Asynchronous Event Configuration Feature */
1640 typedef union {
1641 	struct {
1642 		uint8_t aec_avail:1;	/* Available space too low */
1643 		uint8_t aec_temp:1;	/* Temperature too high */
1644 		uint8_t aec_reliab:1;	/* Degraded reliability */
1645 		uint8_t aec_readonly:1;	/* Media is read-only */
1646 		uint8_t aec_volatile:1;	/* Volatile memory backup failed */
1647 		uint8_t aec_rsvd1:3;
1648 		uint8_t aec_nsan:1;	/* Namespace attribute notices (1.2) */
1649 		uint8_t aec_fwact:1;	/* Firmware activation notices (1.2) */
1650 		uint8_t aec_telln:1;	/* Telemetry log notices (1.3) */
1651 		uint8_t aec_ansacn:1;	/* Asymm. NS access change (1.4) */
1652 		uint8_t aec_plat:1;	/* Predictable latency ev. agg. (1.4) */
1653 		uint8_t aec_lbasi:1;	/* LBA status information (1.4) */
1654 		uint8_t aec_egeal:1;	/* Endurance group ev. agg. (1.4) */
1655 		uint8_t aec_rsvd2:1;
1656 		uint8_t aec_rsvd3[2];
1657 	} b;
1658 	uint32_t r;
1659 } nvme_async_event_conf_t;
1660 
1661 /* Autonomous Power State Transition Feature (1.1) */
1662 typedef union {
1663 	struct {
1664 		uint32_t apst_apste:1;	/* APST enabled */
1665 		uint32_t apst_rsvd:31;
1666 	} b;
1667 	uint32_t r;
1668 } nvme_auto_power_state_trans_t;
1669 
1670 typedef struct {
1671 	uint32_t apst_rsvd1:3;
1672 	uint32_t apst_itps:5;	/* Idle Transition Power State */
1673 	uint32_t apst_itpt:24;	/* Idle Time Prior to Transition */
1674 	uint32_t apst_rsvd2;
1675 } nvme_auto_power_state_t;
1676 
1677 #define	NVME_AUTO_PST_BUFSIZE	256
1678 
1679 /* Software Progress Marker Feature */
1680 typedef union {
1681 	struct {
1682 		uint8_t spm_pbslc;	/* Pre-Boot Software Load Count */
1683 		uint8_t spm_rsvd[3];
1684 	} b;
1685 	uint32_t r;
1686 } nvme_software_progress_marker_t;
1687 
1688 /*
1689  * Firmware Commit - Command Dword 10
1690  */
1691 #define	NVME_FWC_SAVE		0x0	/* Save image only */
1692 #define	NVME_FWC_SAVE_ACTIVATE	0x1	/* Save and activate at next reset */
1693 #define	NVME_FWC_ACTIVATE	0x2	/* Activate slot at next reset */
1694 #define	NVME_FWC_ACTIVATE_IMMED	0x3	/* Activate slot immediately */
1695 
1696 /*
1697  * Firmware slot number is only 3 bits, and zero is not allowed.
1698  * Valid range is 1 to 7.
1699  */
1700 #define	NVME_FW_SLOT_MIN	1U	/* lowest allowable slot number ... */
1701 #define	NVME_FW_SLOT_MAX	7U	/* ... and highest */
1702 
1703 /*
1704  * Some constants to make verification of DWORD variables and arguments easier.
1705  * A DWORD is 4 bytes.
1706  */
1707 #define	NVME_DWORD_SHIFT	2
1708 #define	NVME_DWORD_SIZE		(1 << NVME_DWORD_SHIFT)
1709 #define	NVME_DWORD_MASK		(NVME_DWORD_SIZE - 1)
1710 
1711 /*
1712  * Maximum offset a firmware image can be load at is the number of
1713  * DWORDS in a 32 bit field. Expressed in bytes its is:
1714  */
1715 #define	NVME_FW_OFFSETB_MAX	((u_longlong_t)UINT32_MAX << NVME_DWORD_SHIFT)
1716 
1717 typedef union {
1718 	struct {
1719 		uint32_t fc_slot:3;	/* Firmware slot */
1720 		uint32_t fc_action:3;	/* Commit action */
1721 		uint32_t fc_rsvd:26;
1722 	} b;
1723 	uint32_t r;
1724 } nvme_firmware_commit_dw10_t;
1725 
1726 #pragma pack() /* pack(1) */
1727 
1728 /* NVMe completion status code type */
1729 #define	NVME_CQE_SCT_GENERIC	0	/* Generic Command Status */
1730 #define	NVME_CQE_SCT_SPECIFIC	1	/* Command Specific Status */
1731 #define	NVME_CQE_SCT_INTEGRITY	2	/* Media and Data Integrity Errors */
1732 #define	NVME_CQE_SCT_PATH	3	/* Path Related Status (1.4) */
1733 #define	NVME_CQE_SCT_VENDOR	7	/* Vendor Specific */
1734 
1735 /*
1736  * Status code ranges
1737  */
1738 #define	NVME_CQE_SC_GEN_MIN		0x00
1739 #define	NVME_CQE_SC_GEN_MAX		0x7f
1740 #define	NVME_CQE_SC_CSI_MIN		0x80
1741 #define	NVME_CQE_SC_CSI_MAX		0xbf
1742 #define	NVME_CQE_SC_VEND_MIN		0xc0
1743 #define	NVME_CQE_SC_VEND_MAX		0xff
1744 
1745 /* NVMe completion status code (generic) */
1746 #define	NVME_CQE_SC_GEN_SUCCESS		0x0	/* Successful Completion */
1747 #define	NVME_CQE_SC_GEN_INV_OPC		0x1	/* Invalid Command Opcode */
1748 #define	NVME_CQE_SC_GEN_INV_FLD		0x2	/* Invalid Field in Command */
1749 #define	NVME_CQE_SC_GEN_ID_CNFL		0x3	/* Command ID Conflict */
1750 #define	NVME_CQE_SC_GEN_DATA_XFR_ERR	0x4	/* Data Transfer Error */
1751 #define	NVME_CQE_SC_GEN_ABORT_PWRLOSS	0x5	/* Cmds Aborted / Pwr Loss */
1752 #define	NVME_CQE_SC_GEN_INTERNAL_ERR	0x6	/* Internal Error */
1753 #define	NVME_CQE_SC_GEN_ABORT_REQUEST	0x7	/* Command Abort Requested */
1754 #define	NVME_CQE_SC_GEN_ABORT_SQ_DEL	0x8	/* Cmd Aborted / SQ deletion */
1755 #define	NVME_CQE_SC_GEN_ABORT_FUSE_FAIL	0x9	/* Cmd Aborted / Failed Fused */
1756 #define	NVME_CQE_SC_GEN_ABORT_FUSE_MISS	0xa	/* Cmd Aborted / Missing Fusd */
1757 #define	NVME_CQE_SC_GEN_INV_NS		0xb	/* Inval Namespace or Format */
1758 #define	NVME_CQE_SC_GEN_CMD_SEQ_ERR	0xc	/* Command Sequence Error */
1759 #define	NVME_CQE_SC_GEN_INV_SGL_LAST	0xd	/* Inval SGL Last Seg Desc */
1760 #define	NVME_CQE_SC_GEN_INV_SGL_NUM	0xe	/* Inval Number of SGL Desc */
1761 #define	NVME_CQE_SC_GEN_INV_DSGL_LEN	0xf	/* Data SGL Length Invalid */
1762 #define	NVME_CQE_SC_GEN_INV_MSGL_LEN	0x10	/* Metadata SGL Length Inval */
1763 #define	NVME_CQE_SC_GEN_INV_SGL_DESC	0x11	/* SGL Descriptor Type Inval */
1764 /* Added in NVMe 1.2 */
1765 #define	NVME_CQE_SC_GEN_INV_USE_CMB	0x12	/* Inval use of Ctrl Mem Buf */
1766 #define	NVME_CQE_SC_GEN_INV_PRP_OFF	0x13	/* PRP Offset Invalid */
1767 #define	NVME_CQE_SC_GEN_AWU_EXCEEDED	0x14	/* Atomic Write Unit Exceeded */
1768 #define	NVME_CQE_SC_GEN_OP_DENIED	0x15	/* Operation Denied */
1769 #define	NVME_CQE_SC_GEN_INV_SGL_OFF	0x16	/* SGL Offset Invalid */
1770 #define	NVME_CQE_SC_GEN_INV_SGL_ST	0x17	/* SGL Sub type Invalid */
1771 #define	NVME_CQE_SC_GEN_INCON_HOSTID	0x18	/* Host ID Inconsistent fmt */
1772 #define	NVME_CQE_SC_GEN_KA_EXP		0x19	/* Keep Alive Timer Expired */
1773 #define	NVME_CQE_SC_GEN_INV_KA_TO	0x1a	/* Keep Alive Timeout Invalid */
1774 /* Added in NVMe 1.3 */
1775 #define	NVME_CQE_SC_GEN_ABORT_PREEMPT	0x1b	/* Cmd aborted due to preempt */
1776 #define	NVME_CQE_SC_GEN_SANITIZE_FAIL	0x1c	/* Sanitize Failed */
1777 #define	NVME_CQE_SC_GEN_SANITIZING	0x1d	/* Sanitize in Progress */
1778 #define	NVME_CQE_SC_GEN_INV_SGL_GRAN	0x1e	/* SGL Data Block Gran. Inval */
1779 #define	NVME_CQE_SC_GEN_NO_CMD_Q_CMD	0x1f	/* Command not sup for CMB Q */
1780 /* Added in NVMe 1.4 */
1781 #define	NVME_CQE_SC_GEN_NS_RDONLY	0x20	/* Namespace is write prot. */
1782 #define	NVME_CQE_SC_GEN_CMD_INTR	0x21	/* Command Interrupted */
1783 #define	NVME_CQE_SC_GEN_TRANSIENT	0x22	/* Transient Transport Error */
1784 /* Added in NVMe 2.0 */
1785 #define	NVME_CQE_SC_GEN_CMD_LOCK	0x23	/* Command/Feature Lockdown */
1786 #define	NVME_CQE_SC_ADM_MEDIA_NR	0x24	/* Admin Cmd Media Not Ready */
1787 
1788 /* NVMe completion status code (generic NVM commands) */
1789 #define	NVME_CQE_SC_GEN_NVM_LBA_RANGE	0x80	/* LBA Out Of Range */
1790 #define	NVME_CQE_SC_GEN_NVM_CAP_EXC	0x81	/* Capacity Exceeded */
1791 #define	NVME_CQE_SC_GEN_NVM_NS_NOTRDY	0x82	/* Namespace Not Ready */
1792 #define	NVME_CQE_SC_GEN_NVM_RSV_CNFLCT	0x83	/* Reservation Conflict */
1793 #define	NVME_CQE_SC_GEN_NVM_FORMATTING	0x84	/* Format in progress (1.2) */
1794 /* Added in NVMe 2.0 */
1795 #define	NVME_CQE_SC_GEN_KEY_INV_VAL	0x85	/* Invalid value size */
1796 #define	NVME_CQE_SC_GEN_KEY_INV_KEY	0x86	/* Invalid key size */
1797 #define	NVME_CQE_SC_GEN_KEY_ENOENT	0x87	/* KV Key Does Not Exist */
1798 #define	NVME_CQE_SC_GEN_KEY_UNRECOV	0x88	/* Unrecovered Error */
1799 #define	NVME_CQE_SC_GEN_KEY_EXISTS	0x89	/* Key already exists */
1800 
1801 /* NVMe completion status code (command specific) */
1802 #define	NVME_CQE_SC_SPC_INV_CQ		0x0	/* Completion Queue Invalid */
1803 #define	NVME_CQE_SC_SPC_INV_QID		0x1	/* Invalid Queue Identifier */
1804 #define	NVME_CQE_SC_SPC_MAX_QSZ_EXC	0x2	/* Max Queue Size Exceeded */
1805 #define	NVME_CQE_SC_SPC_ABRT_CMD_EXC	0x3	/* Abort Cmd Limit Exceeded */
1806 #define	NVME_CQE_SC_SPC_ASYNC_EVREQ_EXC	0x5	/* Async Event Request Limit */
1807 #define	NVME_CQE_SC_SPC_INV_FW_SLOT	0x6	/* Invalid Firmware Slot */
1808 #define	NVME_CQE_SC_SPC_INV_FW_IMG	0x7	/* Invalid Firmware Image */
1809 #define	NVME_CQE_SC_SPC_INV_INT_VECT	0x8	/* Invalid Interrupt Vector */
1810 #define	NVME_CQE_SC_SPC_INV_LOG_PAGE	0x9	/* Invalid Log Page */
1811 #define	NVME_CQE_SC_SPC_INV_FORMAT	0xa	/* Invalid Format */
1812 #define	NVME_CQE_SC_SPC_FW_RESET	0xb	/* FW Application Reset Reqd */
1813 #define	NVME_CQE_SC_SPC_INV_Q_DEL	0xc	/* Invalid Queue Deletion */
1814 #define	NVME_CQE_SC_SPC_FEAT_SAVE	0xd	/* Feature Id Not Saveable */
1815 #define	NVME_CQE_SC_SPC_FEAT_CHG	0xe	/* Feature Not Changeable */
1816 #define	NVME_CQE_SC_SPC_FEAT_NS_SPEC	0xf	/* Feature Not Namespace Spec */
1817 /* Added in NVMe 1.2 */
1818 #define	NVME_CQE_SC_SPC_FW_NSSR		0x10	/* FW Application NSSR Reqd */
1819 #define	NVME_CQE_SC_SPC_FW_NEXT_RESET	0x11	/* FW Application Next Reqd */
1820 #define	NVME_CQE_SC_SPC_FW_MTFA		0x12	/* FW Application Exceed MTFA */
1821 #define	NVME_CQE_SC_SPC_FW_PROHIBITED	0x13	/* FW Application Prohibited */
1822 #define	NVME_CQE_SC_SPC_FW_OVERLAP	0x14	/* Overlapping FW ranges */
1823 #define	NVME_CQE_SC_SPC_NS_INSUF_CAP	0x15	/* NS Insufficient Capacity */
1824 #define	NVME_CQE_SC_SPC_NS_NO_ID	0x16	/* NS ID Unavailable */
1825 /* 0x17 is reserved */
1826 #define	NVME_CQE_SC_SPC_NS_ATTACHED	0x18	/* NS Already Attached */
1827 #define	NVME_CQE_SC_SPC_NS_PRIV		0x19	/* NS is private */
1828 #define	NVME_CQE_SC_SPC_NS_NOT_ATTACH	0x1a	/* NS Not Attached */
1829 #define	NVME_CQE_SC_SPC_THIN_ENOTSUP	0x1b	/* Thin Provisioning ENOTSUP */
1830 #define	NVME_CQE_SC_SPC_INV_CTRL_LIST	0x1c	/* Controller list invalid */
1831 /* Added in NVMe 1.3 */
1832 #define	NVME_CQE_SC_SPC_SELF_TESTING	0x1d	/* Self-test in progress */
1833 #define	NVME_CQE_SC_SPC_NO_BP_WRITE	0x1e	/* No Boot Partition Write */
1834 #define	NVME_CQE_SC_SPC_INV_CTRL_ID	0x1f	/* Invalid Controller Id */
1835 #define	NVME_CQE_SC_SPC_INV_SEC_CTRL	0x20	/* Invalid Sec. Ctrl state */
1836 #define	NVME_CQE_SC_SPC_INV_CTRL_NRSRC	0x21	/* Inv. # Ctrl Resources */
1837 #define	NVME_CQE_SC_SPC_INV_RSRC_ID	0x22	/* Inv. Resource ID */
1838 /* Added in NVMe 1.4 */
1839 #define	NVME_CQE_SC_SPC_NO_SAN_PMR	0x23	/* Sanitize prohib. w/ pmem */
1840 #define	NVME_CQE_SC_SPC_INV_ANA_GID	0x24	/* Invalid ANA group ID */
1841 #define	NVME_CQE_SC_SPC_ANA_ATTACH	0x25	/* ANA Attach Failed */
1842 /* Added in NVMe 2.0 */
1843 #define	NVME_CQE_SC_SPC_INSUF_CAP	0x26	/* Insufficient Capacity */
1844 #define	NVME_CQE_SC_SPC_NS_ATTACH_LIM	0x27	/* NS Attach Limit Exceeded */
1845 #define	NVME_CQE_SC_SPC_LOCKDOWN_UNSUP	0x28	/* Prohib Cmd Exec Not Sup */
1846 #define	NVME_CQE_SC_SPC_UNSUP_IO_CMD	0x29	/* I/O Command set not sup */
1847 #define	NVME_CQE_SC_SPC_DIS_IO_CMD	0x2a	/* I/O Command set not enab */
1848 #define	NVME_CQE_SC_SPC_INV_CMD_COMBO	0x2b	/* I/O command set combo rej */
1849 #define	NVME_CQE_SC_SPC_INV_IO_CMD	0x2c	/* Invalid I/O command set */
1850 #define	NVME_CQE_SC_SPC_UNAVAIL_ID	0x2d	/* Unavailable ID */
1851 
1852 
1853 /* NVMe completion status code (I/O command specific) */
1854 #define	NVME_CQE_SC_SPC_NVM_CNFL_ATTR	0x80	/* Conflicting Attributes */
1855 #define	NVME_CQE_SC_SPC_NVM_INV_PROT	0x81	/* Invalid Protection */
1856 #define	NVME_CQE_SC_SPC_NVM_READONLY	0x82	/* Write to Read Only Range */
1857 /* Added in 2.0 */
1858 #define	NVME_CQE_SC_SPC_IO_LIMIT	0x83	/* Cmd Size Limit Exceeded */
1859 /* 0x84 to 0xb7 are reserved */
1860 #define	NVME_CQE_SC_SPC_ZONE_BDRY_ERR	0xb8	/* Zoned Boundary Error */
1861 #define	NVME_CQE_SC_SPC_ZONE_FULL	0xb9	/* Zone is Full */
1862 #define	NVME_CQE_SC_SPC_ZONE_RDONLY	0xba	/* Zone is Read Only */
1863 #define	NVME_CQE_SC_SPC_ZONE_OFFLINE	0xbb	/* Zone is Offline */
1864 #define	NVME_CQE_SC_SPC_ZONE_INV_WRITE	0xbc	/* Zone Invalid Write */
1865 #define	NVME_CQE_SC_SPC_ZONE_ACT	0xbd	/* Too May Active Zones */
1866 #define	NVME_CQE_SC_SPC_ZONE_OPEN	0xbe	/* Too May Open Zones */
1867 #define	NVME_CQE_SC_SPC_INV_ZONE_TRANS	0xbf	/* Invalid Zone State Trans */
1868 
1869 /* NVMe completion status code (data / metadata integrity) */
1870 #define	NVME_CQE_SC_INT_NVM_WRITE	0x80	/* Write Fault */
1871 #define	NVME_CQE_SC_INT_NVM_READ	0x81	/* Unrecovered Read Error */
1872 #define	NVME_CQE_SC_INT_NVM_GUARD	0x82	/* Guard Check Error */
1873 #define	NVME_CQE_SC_INT_NVM_APPL_TAG	0x83	/* Application Tag Check Err */
1874 #define	NVME_CQE_SC_INT_NVM_REF_TAG	0x84	/* Reference Tag Check Err */
1875 #define	NVME_CQE_SC_INT_NVM_COMPARE	0x85	/* Compare Failure */
1876 #define	NVME_CQE_SC_INT_NVM_ACCESS	0x86	/* Access Denied */
1877 /* Added in 1.2 */
1878 #define	NVME_CQE_SC_INT_NVM_DEALLOC	0x87	/* Dealloc Log Block */
1879 /* Added in 2.0 */
1880 #define	NVME_CQE_SC_INT_NVM_TAG		0x88	/* End-to-End Storage Tag Err */
1881 
1882 /* NVMe completion status code (path related) */
1883 /* Added in NVMe 1.4 */
1884 #define	NVME_CQE_SC_PATH_INT_ERR	0x00	/* Internal Path Error */
1885 #define	NVME_CQE_SC_PATH_AA_PLOSS	0x01	/* Asym Access Pers Loss */
1886 #define	NVME_CQE_SC_PATH_AA_INACC	0x02	/* Asym Access Inaccessible */
1887 #define	NVME_CQE_SC_PATH_AA_TRANS	0x03	/* Asym Access Transition */
1888 #define	NVME_CQE_SC_PATH_CTRL_ERR	0x60	/* Controller Path Error */
1889 #define	NVME_CQE_SC_PATH_HOST_ERR	0x70	/* Host Path Error */
1890 #define	NVME_CQE_SC_PATH_HOST_ABRT	0x71	/* Cmd aborted by host */
1891 
1892 /*
1893  * Controller information (NVME_IOC_CTRL_INFO). This is a consolidation of misc.
1894  * information that we want to know about a controller.
1895  */
1896 typedef struct {
1897 	nvme_ioctl_common_t nci_common;
1898 	nvme_identify_ctrl_t nci_ctrl_id;
1899 	nvme_identify_nsid_t nci_common_ns;
1900 	nvme_version_t nci_vers;
1901 	nvme_capabilities_t nci_caps;
1902 	uint32_t nci_nintrs;
1903 } nvme_ioctl_ctrl_info_t;
1904 
1905 /*
1906  * NVME namespace state flags.
1907  *
1908  * The values are defined entirely by the driver. Some states correspond to
1909  * namespace states described by the NVMe specification r1.3 section 6.1, others
1910  * are specific to the implementation of this driver. These are present in the
1911  * nvme_ns_kinfo_t that is used with the NVME_IOC_NS_INFO ioctl.
1912  *
1913  * The states are as follows:
1914  * - ALLOCATED: the namespace exists in the controller as per the NVMe spec
1915  * - ACTIVE: the namespace exists and is attached to this controller as per the
1916  *   NVMe spec. Any namespace that is ACTIVE is also ALLOCATED. This must not be
1917  *   confused with the ATTACHED state.
1918  * - ATTACHED: the driver has attached a blkdev(4D) instance to this namespace.
1919  *   This state can be changed by userspace with the ioctls NVME_IOC_ATTACH and
1920  *   NVME_IOC_DETACH. A namespace can only be ATTACHED when it is not IGNORED.
1921  * - IGNORED: the driver ignores this namespace, it never attaches a blkdev(4D).
1922  *   Namespaces are IGNORED when they are not ACTIVE, or if they are ACTIVE but
1923  *   have certain properties that the driver cannot handle.
1924  */
1925 typedef enum {
1926 	NVME_NS_STATE_ALLOCATED	=	1 << 0,
1927 	NVME_NS_STATE_ACTIVE	=	1 << 1,
1928 	NVME_NS_STATE_ATTACHED	=	1 << 2,
1929 	NVME_NS_STATE_IGNORED	=	1 << 3
1930 } nvme_ns_state_t;
1931 
1932 /*
1933  * This is the maximum length of the NVMe namespace's blkdev address. This is
1934  * only valid in the structure with the NVME_NS_STATE_ATTACHED flag is set.
1935  * Otherwise the entry will be all zeros. This is useful when you need to
1936  * determine what the corresponding blkdev instance in libdevinfo for the
1937  * device.
1938  */
1939 #define	NVME_BLKDEV_NAMELEN	128
1940 
1941 /*
1942  * Namespace Information (NVME_IOC_NS_INFO).
1943  */
1944 typedef struct {
1945 	nvme_ioctl_common_t nni_common;
1946 	nvme_ns_state_t	nni_state;
1947 	char nni_addr[NVME_BLKDEV_NAMELEN];
1948 	nvme_identify_nsid_t nni_id;
1949 } nvme_ioctl_ns_info_t;
1950 
1951 /*
1952  * NVMe Command Set Identifiers. This was added in NVMe 2.0, but in all the
1953  * places it was required to be specified, the default value of 0 indicates the
1954  * traditional NVM command set.
1955  */
1956 typedef enum {
1957 	NVME_CSI_NVM	= 0,
1958 	NVME_CSI_KV,
1959 	NVME_CSI_ZNS
1960 } nvme_csi_t;
1961 
1962 #ifdef __cplusplus
1963 }
1964 #endif
1965 
1966 #endif /* _SYS_NVME_H */
1967