1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3 *
4 * Copyright (C) 2012-2016 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
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31
32 #include "opt_nvme.h"
33
34 #include <sys/param.h>
35 #include <sys/bus.h>
36 #include <sys/sysctl.h>
37
38 #include "nvme_private.h"
39
40 #ifndef NVME_USE_NVD
41 #define NVME_USE_NVD 1
42 #endif
43
44 int nvme_use_nvd = NVME_USE_NVD;
45 bool nvme_verbose_cmd_dump = false;
46
47 SYSCTL_NODE(_hw, OID_AUTO, nvme, CTLFLAG_RD, 0, "NVMe sysctl tunables");
48 SYSCTL_INT(_hw_nvme, OID_AUTO, use_nvd, CTLFLAG_RDTUN,
49 &nvme_use_nvd, 1, "1 = Create NVD devices, 0 = Create NDA devices");
50 SYSCTL_BOOL(_hw_nvme, OID_AUTO, verbose_cmd_dump, CTLFLAG_RWTUN,
51 &nvme_verbose_cmd_dump, 0,
52 "enable verbose command printting when a command fails");
53
54 /*
55 * CTLTYPE_S64 and sysctl_handle_64 were added in r217616. Define these
56 * explicitly here for older kernels that don't include the r217616
57 * changeset.
58 */
59 #ifndef CTLTYPE_S64
60 #define CTLTYPE_S64 CTLTYPE_QUAD
61 #define sysctl_handle_64 sysctl_handle_quad
62 #endif
63
64 static void
nvme_dump_queue(struct nvme_qpair * qpair)65 nvme_dump_queue(struct nvme_qpair *qpair)
66 {
67 struct nvme_completion *cpl;
68 struct nvme_command *cmd;
69 int i;
70
71 printf("id:%04Xh phase:%d\n", qpair->id, qpair->phase);
72
73 printf("Completion queue:\n");
74 for (i = 0; i < qpair->num_entries; i++) {
75 cpl = &qpair->cpl[i];
76 printf("%05d: ", i);
77 nvme_dump_completion(cpl);
78 }
79
80 printf("Submission queue:\n");
81 for (i = 0; i < qpair->num_entries; i++) {
82 cmd = &qpair->cmd[i];
83 printf("%05d: ", i);
84 nvme_dump_command(cmd);
85 }
86 }
87
88
89 static int
nvme_sysctl_dump_debug(SYSCTL_HANDLER_ARGS)90 nvme_sysctl_dump_debug(SYSCTL_HANDLER_ARGS)
91 {
92 struct nvme_qpair *qpair = arg1;
93 uint32_t val = 0;
94
95 int error = sysctl_handle_int(oidp, &val, 0, req);
96
97 if (error)
98 return (error);
99
100 if (val != 0)
101 nvme_dump_queue(qpair);
102
103 return (0);
104 }
105
106 static int
nvme_sysctl_int_coal_time(SYSCTL_HANDLER_ARGS)107 nvme_sysctl_int_coal_time(SYSCTL_HANDLER_ARGS)
108 {
109 struct nvme_controller *ctrlr = arg1;
110 uint32_t oldval = ctrlr->int_coal_time;
111 int error = sysctl_handle_int(oidp, &ctrlr->int_coal_time, 0,
112 req);
113
114 if (error)
115 return (error);
116
117 if (oldval != ctrlr->int_coal_time)
118 nvme_ctrlr_cmd_set_interrupt_coalescing(ctrlr,
119 ctrlr->int_coal_time, ctrlr->int_coal_threshold, NULL,
120 NULL);
121
122 return (0);
123 }
124
125 static int
nvme_sysctl_int_coal_threshold(SYSCTL_HANDLER_ARGS)126 nvme_sysctl_int_coal_threshold(SYSCTL_HANDLER_ARGS)
127 {
128 struct nvme_controller *ctrlr = arg1;
129 uint32_t oldval = ctrlr->int_coal_threshold;
130 int error = sysctl_handle_int(oidp, &ctrlr->int_coal_threshold, 0,
131 req);
132
133 if (error)
134 return (error);
135
136 if (oldval != ctrlr->int_coal_threshold)
137 nvme_ctrlr_cmd_set_interrupt_coalescing(ctrlr,
138 ctrlr->int_coal_time, ctrlr->int_coal_threshold, NULL,
139 NULL);
140
141 return (0);
142 }
143
144 static int
nvme_sysctl_timeout_period(SYSCTL_HANDLER_ARGS)145 nvme_sysctl_timeout_period(SYSCTL_HANDLER_ARGS)
146 {
147 struct nvme_controller *ctrlr = arg1;
148 uint32_t oldval = ctrlr->timeout_period;
149 int error = sysctl_handle_int(oidp, &ctrlr->timeout_period, 0, req);
150
151 if (error)
152 return (error);
153
154 if (ctrlr->timeout_period > NVME_MAX_TIMEOUT_PERIOD ||
155 ctrlr->timeout_period < NVME_MIN_TIMEOUT_PERIOD) {
156 ctrlr->timeout_period = oldval;
157 return (EINVAL);
158 }
159
160 return (0);
161 }
162
163 static void
nvme_qpair_reset_stats(struct nvme_qpair * qpair)164 nvme_qpair_reset_stats(struct nvme_qpair *qpair)
165 {
166
167 qpair->num_cmds = 0;
168 qpair->num_intr_handler_calls = 0;
169 qpair->num_retries = 0;
170 qpair->num_failures = 0;
171 }
172
173 static int
nvme_sysctl_num_cmds(SYSCTL_HANDLER_ARGS)174 nvme_sysctl_num_cmds(SYSCTL_HANDLER_ARGS)
175 {
176 struct nvme_controller *ctrlr = arg1;
177 int64_t num_cmds = 0;
178 int i;
179
180 num_cmds = ctrlr->adminq.num_cmds;
181
182 for (i = 0; i < ctrlr->num_io_queues; i++)
183 num_cmds += ctrlr->ioq[i].num_cmds;
184
185 return (sysctl_handle_64(oidp, &num_cmds, 0, req));
186 }
187
188 static int
nvme_sysctl_num_intr_handler_calls(SYSCTL_HANDLER_ARGS)189 nvme_sysctl_num_intr_handler_calls(SYSCTL_HANDLER_ARGS)
190 {
191 struct nvme_controller *ctrlr = arg1;
192 int64_t num_intr_handler_calls = 0;
193 int i;
194
195 num_intr_handler_calls = ctrlr->adminq.num_intr_handler_calls;
196
197 for (i = 0; i < ctrlr->num_io_queues; i++)
198 num_intr_handler_calls += ctrlr->ioq[i].num_intr_handler_calls;
199
200 return (sysctl_handle_64(oidp, &num_intr_handler_calls, 0, req));
201 }
202
203 static int
nvme_sysctl_num_retries(SYSCTL_HANDLER_ARGS)204 nvme_sysctl_num_retries(SYSCTL_HANDLER_ARGS)
205 {
206 struct nvme_controller *ctrlr = arg1;
207 int64_t num_retries = 0;
208 int i;
209
210 num_retries = ctrlr->adminq.num_retries;
211
212 for (i = 0; i < ctrlr->num_io_queues; i++)
213 num_retries += ctrlr->ioq[i].num_retries;
214
215 return (sysctl_handle_64(oidp, &num_retries, 0, req));
216 }
217
218 static int
nvme_sysctl_num_failures(SYSCTL_HANDLER_ARGS)219 nvme_sysctl_num_failures(SYSCTL_HANDLER_ARGS)
220 {
221 struct nvme_controller *ctrlr = arg1;
222 int64_t num_failures = 0;
223 int i;
224
225 num_failures = ctrlr->adminq.num_failures;
226
227 for (i = 0; i < ctrlr->num_io_queues; i++)
228 num_failures += ctrlr->ioq[i].num_failures;
229
230 return (sysctl_handle_64(oidp, &num_failures, 0, req));
231 }
232
233 static int
nvme_sysctl_reset_stats(SYSCTL_HANDLER_ARGS)234 nvme_sysctl_reset_stats(SYSCTL_HANDLER_ARGS)
235 {
236 struct nvme_controller *ctrlr = arg1;
237 uint32_t i, val = 0;
238
239 int error = sysctl_handle_int(oidp, &val, 0, req);
240
241 if (error)
242 return (error);
243
244 if (val != 0) {
245 nvme_qpair_reset_stats(&ctrlr->adminq);
246
247 for (i = 0; i < ctrlr->num_io_queues; i++)
248 nvme_qpair_reset_stats(&ctrlr->ioq[i]);
249 }
250
251 return (0);
252 }
253
254
255 static void
nvme_sysctl_initialize_queue(struct nvme_qpair * qpair,struct sysctl_ctx_list * ctrlr_ctx,struct sysctl_oid * que_tree)256 nvme_sysctl_initialize_queue(struct nvme_qpair *qpair,
257 struct sysctl_ctx_list *ctrlr_ctx, struct sysctl_oid *que_tree)
258 {
259 struct sysctl_oid_list *que_list = SYSCTL_CHILDREN(que_tree);
260
261 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "num_entries",
262 CTLFLAG_RD, &qpair->num_entries, 0,
263 "Number of entries in hardware queue");
264 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "num_trackers",
265 CTLFLAG_RD, &qpair->num_trackers, 0,
266 "Number of trackers pre-allocated for this queue pair");
267 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "sq_head",
268 CTLFLAG_RD, &qpair->sq_head, 0,
269 "Current head of submission queue (as observed by driver)");
270 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "sq_tail",
271 CTLFLAG_RD, &qpair->sq_tail, 0,
272 "Current tail of submission queue (as observed by driver)");
273 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "cq_head",
274 CTLFLAG_RD, &qpair->cq_head, 0,
275 "Current head of completion queue (as observed by driver)");
276
277 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_cmds",
278 CTLFLAG_RD, &qpair->num_cmds, "Number of commands submitted");
279 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_intr_handler_calls",
280 CTLFLAG_RD, &qpair->num_intr_handler_calls,
281 "Number of times interrupt handler was invoked (will typically be "
282 "less than number of actual interrupts generated due to "
283 "coalescing)");
284 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_retries",
285 CTLFLAG_RD, &qpair->num_retries, "Number of commands retried");
286 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_failures",
287 CTLFLAG_RD, &qpair->num_failures,
288 "Number of commands ending in failure after all retries");
289
290 SYSCTL_ADD_PROC(ctrlr_ctx, que_list, OID_AUTO,
291 "dump_debug", CTLTYPE_UINT | CTLFLAG_RW, qpair, 0,
292 nvme_sysctl_dump_debug, "IU", "Dump debug data");
293 }
294
295 void
nvme_sysctl_initialize_ctrlr(struct nvme_controller * ctrlr)296 nvme_sysctl_initialize_ctrlr(struct nvme_controller *ctrlr)
297 {
298 struct sysctl_ctx_list *ctrlr_ctx;
299 struct sysctl_oid *ctrlr_tree, *que_tree;
300 struct sysctl_oid_list *ctrlr_list;
301 #define QUEUE_NAME_LENGTH 16
302 char queue_name[QUEUE_NAME_LENGTH];
303 int i;
304
305 ctrlr_ctx = device_get_sysctl_ctx(ctrlr->dev);
306 ctrlr_tree = device_get_sysctl_tree(ctrlr->dev);
307 ctrlr_list = SYSCTL_CHILDREN(ctrlr_tree);
308
309 SYSCTL_ADD_UINT(ctrlr_ctx, ctrlr_list, OID_AUTO, "num_cpus_per_ioq",
310 CTLFLAG_RD, &ctrlr->num_cpus_per_ioq, 0,
311 "Number of CPUs assigned per I/O queue pair");
312
313 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO,
314 "int_coal_time", CTLTYPE_UINT | CTLFLAG_RW, ctrlr, 0,
315 nvme_sysctl_int_coal_time, "IU",
316 "Interrupt coalescing timeout (in microseconds)");
317
318 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO,
319 "int_coal_threshold", CTLTYPE_UINT | CTLFLAG_RW, ctrlr, 0,
320 nvme_sysctl_int_coal_threshold, "IU",
321 "Interrupt coalescing threshold");
322
323 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO,
324 "timeout_period", CTLTYPE_UINT | CTLFLAG_RW, ctrlr, 0,
325 nvme_sysctl_timeout_period, "IU",
326 "Timeout period (in seconds)");
327
328 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO,
329 "num_cmds", CTLTYPE_S64 | CTLFLAG_RD,
330 ctrlr, 0, nvme_sysctl_num_cmds, "IU",
331 "Number of commands submitted");
332
333 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO,
334 "num_intr_handler_calls", CTLTYPE_S64 | CTLFLAG_RD,
335 ctrlr, 0, nvme_sysctl_num_intr_handler_calls, "IU",
336 "Number of times interrupt handler was invoked (will "
337 "typically be less than number of actual interrupts "
338 "generated due to coalescing)");
339
340 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO,
341 "num_retries", CTLTYPE_S64 | CTLFLAG_RD,
342 ctrlr, 0, nvme_sysctl_num_retries, "IU",
343 "Number of commands retried");
344
345 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO,
346 "num_failures", CTLTYPE_S64 | CTLFLAG_RD,
347 ctrlr, 0, nvme_sysctl_num_failures, "IU",
348 "Number of commands ending in failure after all retries");
349
350 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO,
351 "reset_stats", CTLTYPE_UINT | CTLFLAG_RW, ctrlr, 0,
352 nvme_sysctl_reset_stats, "IU", "Reset statistics to zero");
353
354 que_tree = SYSCTL_ADD_NODE(ctrlr_ctx, ctrlr_list, OID_AUTO, "adminq",
355 CTLFLAG_RD, NULL, "Admin Queue");
356
357 nvme_sysctl_initialize_queue(&ctrlr->adminq, ctrlr_ctx, que_tree);
358
359 for (i = 0; i < ctrlr->num_io_queues; i++) {
360 snprintf(queue_name, QUEUE_NAME_LENGTH, "ioq%d", i);
361 que_tree = SYSCTL_ADD_NODE(ctrlr_ctx, ctrlr_list, OID_AUTO,
362 queue_name, CTLFLAG_RD, NULL, "IO Queue");
363 nvme_sysctl_initialize_queue(&ctrlr->ioq[i], ctrlr_ctx,
364 que_tree);
365 }
366 }
367