1 /*
2 * Copyright (c) 2013-2017, Mellanox Technologies. All rights reserved.
3 *
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
9 *
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
12 * conditions are met:
13 *
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
16 * disclaimer.
17 *
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
22 *
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30 * SOFTWARE.
31 */
32
33 #include <linux/debugfs.h>
34
35 #include "mlx5_ib.h"
36 #include "cmd.h"
37
38 enum mlx5_ib_cong_node_type {
39 MLX5_IB_RROCE_ECN_RP = 1,
40 MLX5_IB_RROCE_ECN_NP = 2,
41 MLX5_IB_RROCE_GENERAL = 3,
42 };
43
44 static const char * const mlx5_ib_dbg_cc_name[] = {
45 "rp_clamp_tgt_rate",
46 "rp_clamp_tgt_rate_ati",
47 "rp_time_reset",
48 "rp_byte_reset",
49 "rp_threshold",
50 "rp_ai_rate",
51 "rp_max_rate",
52 "rp_hai_rate",
53 "rp_min_dec_fac",
54 "rp_min_rate",
55 "rp_rate_to_set_on_first_cnp",
56 "rp_dce_tcp_g",
57 "rp_dce_tcp_rtt",
58 "rp_rate_reduce_monitor_period",
59 "rp_initial_alpha_value",
60 "rp_gd",
61 "np_min_time_between_cnps",
62 "np_cnp_dscp",
63 "np_cnp_prio_mode",
64 "np_cnp_prio",
65 "rtt_resp_dscp_valid",
66 "rtt_resp_dscp",
67 };
68
69 #define MLX5_IB_RP_CLAMP_TGT_RATE_ATTR BIT(1)
70 #define MLX5_IB_RP_CLAMP_TGT_RATE_ATI_ATTR BIT(2)
71 #define MLX5_IB_RP_TIME_RESET_ATTR BIT(3)
72 #define MLX5_IB_RP_BYTE_RESET_ATTR BIT(4)
73 #define MLX5_IB_RP_THRESHOLD_ATTR BIT(5)
74 #define MLX5_IB_RP_MAX_RATE_ATTR BIT(6)
75 #define MLX5_IB_RP_AI_RATE_ATTR BIT(7)
76 #define MLX5_IB_RP_HAI_RATE_ATTR BIT(8)
77 #define MLX5_IB_RP_MIN_DEC_FAC_ATTR BIT(9)
78 #define MLX5_IB_RP_MIN_RATE_ATTR BIT(10)
79 #define MLX5_IB_RP_RATE_TO_SET_ON_FIRST_CNP_ATTR BIT(11)
80 #define MLX5_IB_RP_DCE_TCP_G_ATTR BIT(12)
81 #define MLX5_IB_RP_DCE_TCP_RTT_ATTR BIT(13)
82 #define MLX5_IB_RP_RATE_REDUCE_MONITOR_PERIOD_ATTR BIT(14)
83 #define MLX5_IB_RP_INITIAL_ALPHA_VALUE_ATTR BIT(15)
84 #define MLX5_IB_RP_GD_ATTR BIT(16)
85
86 #define MLX5_IB_NP_MIN_TIME_BETWEEN_CNPS_ATTR BIT(2)
87 #define MLX5_IB_NP_CNP_DSCP_ATTR BIT(3)
88 #define MLX5_IB_NP_CNP_PRIO_MODE_ATTR BIT(4)
89
90 #define MLX5_IB_GENERAL_RTT_RESP_DSCP_ATTR BIT(0)
91
92 static enum mlx5_ib_cong_node_type
mlx5_ib_param_to_node(enum mlx5_ib_dbg_cc_types param_offset)93 mlx5_ib_param_to_node(enum mlx5_ib_dbg_cc_types param_offset)
94 {
95 if (param_offset <= MLX5_IB_DBG_CC_RP_GD)
96 return MLX5_IB_RROCE_ECN_RP;
97
98 if (param_offset <= MLX5_IB_DBG_CC_NP_CNP_PRIO)
99 return MLX5_IB_RROCE_ECN_NP;
100
101 return MLX5_IB_RROCE_GENERAL;
102 }
103
mlx5_get_cc_param_val(void * field,int offset)104 static u32 mlx5_get_cc_param_val(void *field, int offset)
105 {
106 switch (offset) {
107 case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE:
108 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
109 clamp_tgt_rate);
110 case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE_ATI:
111 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
112 clamp_tgt_rate_after_time_inc);
113 case MLX5_IB_DBG_CC_RP_TIME_RESET:
114 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
115 rpg_time_reset);
116 case MLX5_IB_DBG_CC_RP_BYTE_RESET:
117 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
118 rpg_byte_reset);
119 case MLX5_IB_DBG_CC_RP_THRESHOLD:
120 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
121 rpg_threshold);
122 case MLX5_IB_DBG_CC_RP_AI_RATE:
123 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
124 rpg_ai_rate);
125 case MLX5_IB_DBG_CC_RP_MAX_RATE:
126 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
127 rpg_max_rate);
128 case MLX5_IB_DBG_CC_RP_HAI_RATE:
129 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
130 rpg_hai_rate);
131 case MLX5_IB_DBG_CC_RP_MIN_DEC_FAC:
132 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
133 rpg_min_dec_fac);
134 case MLX5_IB_DBG_CC_RP_MIN_RATE:
135 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
136 rpg_min_rate);
137 case MLX5_IB_DBG_CC_RP_RATE_TO_SET_ON_FIRST_CNP:
138 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
139 rate_to_set_on_first_cnp);
140 case MLX5_IB_DBG_CC_RP_DCE_TCP_G:
141 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
142 dce_tcp_g);
143 case MLX5_IB_DBG_CC_RP_DCE_TCP_RTT:
144 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
145 dce_tcp_rtt);
146 case MLX5_IB_DBG_CC_RP_RATE_REDUCE_MONITOR_PERIOD:
147 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
148 rate_reduce_monitor_period);
149 case MLX5_IB_DBG_CC_RP_INITIAL_ALPHA_VALUE:
150 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
151 initial_alpha_value);
152 case MLX5_IB_DBG_CC_RP_GD:
153 return MLX5_GET(cong_control_r_roce_ecn_rp, field,
154 rpg_gd);
155 case MLX5_IB_DBG_CC_NP_MIN_TIME_BETWEEN_CNPS:
156 return MLX5_GET(cong_control_r_roce_ecn_np, field,
157 min_time_between_cnps);
158 case MLX5_IB_DBG_CC_NP_CNP_DSCP:
159 return MLX5_GET(cong_control_r_roce_ecn_np, field,
160 cnp_dscp);
161 case MLX5_IB_DBG_CC_NP_CNP_PRIO_MODE:
162 return MLX5_GET(cong_control_r_roce_ecn_np, field,
163 cnp_prio_mode);
164 case MLX5_IB_DBG_CC_NP_CNP_PRIO:
165 return MLX5_GET(cong_control_r_roce_ecn_np, field,
166 cnp_802p_prio);
167 case MLX5_IB_DBG_CC_GENERAL_RTT_RESP_DSCP_VALID:
168 return MLX5_GET(cong_control_r_roce_general, field,
169 rtt_resp_dscp_valid);
170 case MLX5_IB_DBG_CC_GENERAL_RTT_RESP_DSCP:
171 return MLX5_GET(cong_control_r_roce_general, field,
172 rtt_resp_dscp);
173 default:
174 return 0;
175 }
176 }
177
mlx5_ib_set_cc_param_mask_val(void * field,int offset,u32 var,u32 * attr_mask)178 static void mlx5_ib_set_cc_param_mask_val(void *field, int offset,
179 u32 var, u32 *attr_mask)
180 {
181 switch (offset) {
182 case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE:
183 *attr_mask |= MLX5_IB_RP_CLAMP_TGT_RATE_ATTR;
184 MLX5_SET(cong_control_r_roce_ecn_rp, field,
185 clamp_tgt_rate, var);
186 break;
187 case MLX5_IB_DBG_CC_RP_CLAMP_TGT_RATE_ATI:
188 *attr_mask |= MLX5_IB_RP_CLAMP_TGT_RATE_ATI_ATTR;
189 MLX5_SET(cong_control_r_roce_ecn_rp, field,
190 clamp_tgt_rate_after_time_inc, var);
191 break;
192 case MLX5_IB_DBG_CC_RP_TIME_RESET:
193 *attr_mask |= MLX5_IB_RP_TIME_RESET_ATTR;
194 MLX5_SET(cong_control_r_roce_ecn_rp, field,
195 rpg_time_reset, var);
196 break;
197 case MLX5_IB_DBG_CC_RP_BYTE_RESET:
198 *attr_mask |= MLX5_IB_RP_BYTE_RESET_ATTR;
199 MLX5_SET(cong_control_r_roce_ecn_rp, field,
200 rpg_byte_reset, var);
201 break;
202 case MLX5_IB_DBG_CC_RP_THRESHOLD:
203 *attr_mask |= MLX5_IB_RP_THRESHOLD_ATTR;
204 MLX5_SET(cong_control_r_roce_ecn_rp, field,
205 rpg_threshold, var);
206 break;
207 case MLX5_IB_DBG_CC_RP_AI_RATE:
208 *attr_mask |= MLX5_IB_RP_AI_RATE_ATTR;
209 MLX5_SET(cong_control_r_roce_ecn_rp, field,
210 rpg_ai_rate, var);
211 break;
212 case MLX5_IB_DBG_CC_RP_MAX_RATE:
213 *attr_mask |= MLX5_IB_RP_MAX_RATE_ATTR;
214 MLX5_SET(cong_control_r_roce_ecn_rp, field,
215 rpg_max_rate, var);
216 break;
217 case MLX5_IB_DBG_CC_RP_HAI_RATE:
218 *attr_mask |= MLX5_IB_RP_HAI_RATE_ATTR;
219 MLX5_SET(cong_control_r_roce_ecn_rp, field,
220 rpg_hai_rate, var);
221 break;
222 case MLX5_IB_DBG_CC_RP_MIN_DEC_FAC:
223 *attr_mask |= MLX5_IB_RP_MIN_DEC_FAC_ATTR;
224 MLX5_SET(cong_control_r_roce_ecn_rp, field,
225 rpg_min_dec_fac, var);
226 break;
227 case MLX5_IB_DBG_CC_RP_MIN_RATE:
228 *attr_mask |= MLX5_IB_RP_MIN_RATE_ATTR;
229 MLX5_SET(cong_control_r_roce_ecn_rp, field,
230 rpg_min_rate, var);
231 break;
232 case MLX5_IB_DBG_CC_RP_RATE_TO_SET_ON_FIRST_CNP:
233 *attr_mask |= MLX5_IB_RP_RATE_TO_SET_ON_FIRST_CNP_ATTR;
234 MLX5_SET(cong_control_r_roce_ecn_rp, field,
235 rate_to_set_on_first_cnp, var);
236 break;
237 case MLX5_IB_DBG_CC_RP_DCE_TCP_G:
238 *attr_mask |= MLX5_IB_RP_DCE_TCP_G_ATTR;
239 MLX5_SET(cong_control_r_roce_ecn_rp, field,
240 dce_tcp_g, var);
241 break;
242 case MLX5_IB_DBG_CC_RP_DCE_TCP_RTT:
243 *attr_mask |= MLX5_IB_RP_DCE_TCP_RTT_ATTR;
244 MLX5_SET(cong_control_r_roce_ecn_rp, field,
245 dce_tcp_rtt, var);
246 break;
247 case MLX5_IB_DBG_CC_RP_RATE_REDUCE_MONITOR_PERIOD:
248 *attr_mask |= MLX5_IB_RP_RATE_REDUCE_MONITOR_PERIOD_ATTR;
249 MLX5_SET(cong_control_r_roce_ecn_rp, field,
250 rate_reduce_monitor_period, var);
251 break;
252 case MLX5_IB_DBG_CC_RP_INITIAL_ALPHA_VALUE:
253 *attr_mask |= MLX5_IB_RP_INITIAL_ALPHA_VALUE_ATTR;
254 MLX5_SET(cong_control_r_roce_ecn_rp, field,
255 initial_alpha_value, var);
256 break;
257 case MLX5_IB_DBG_CC_RP_GD:
258 *attr_mask |= MLX5_IB_RP_GD_ATTR;
259 MLX5_SET(cong_control_r_roce_ecn_rp, field,
260 rpg_gd, var);
261 break;
262 case MLX5_IB_DBG_CC_NP_MIN_TIME_BETWEEN_CNPS:
263 *attr_mask |= MLX5_IB_NP_MIN_TIME_BETWEEN_CNPS_ATTR;
264 MLX5_SET(cong_control_r_roce_ecn_np, field,
265 min_time_between_cnps, var);
266 break;
267 case MLX5_IB_DBG_CC_NP_CNP_DSCP:
268 *attr_mask |= MLX5_IB_NP_CNP_DSCP_ATTR;
269 MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_dscp, var);
270 break;
271 case MLX5_IB_DBG_CC_NP_CNP_PRIO_MODE:
272 *attr_mask |= MLX5_IB_NP_CNP_PRIO_MODE_ATTR;
273 MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_prio_mode, var);
274 break;
275 case MLX5_IB_DBG_CC_NP_CNP_PRIO:
276 *attr_mask |= MLX5_IB_NP_CNP_PRIO_MODE_ATTR;
277 MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_prio_mode, 0);
278 MLX5_SET(cong_control_r_roce_ecn_np, field, cnp_802p_prio, var);
279 break;
280 case MLX5_IB_DBG_CC_GENERAL_RTT_RESP_DSCP_VALID:
281 *attr_mask |= MLX5_IB_GENERAL_RTT_RESP_DSCP_ATTR;
282 MLX5_SET(cong_control_r_roce_general, field, rtt_resp_dscp_valid, var);
283 break;
284 case MLX5_IB_DBG_CC_GENERAL_RTT_RESP_DSCP:
285 *attr_mask |= MLX5_IB_GENERAL_RTT_RESP_DSCP_ATTR;
286 MLX5_SET(cong_control_r_roce_general, field, rtt_resp_dscp_valid, 1);
287 MLX5_SET(cong_control_r_roce_general, field, rtt_resp_dscp, var);
288 break;
289 }
290 }
291
mlx5_ib_get_cc_params(struct mlx5_ib_dev * dev,u32 port_num,int offset,u32 * var)292 static int mlx5_ib_get_cc_params(struct mlx5_ib_dev *dev, u32 port_num,
293 int offset, u32 *var)
294 {
295 int outlen = MLX5_ST_SZ_BYTES(query_cong_params_out);
296 void *out;
297 void *field;
298 int err;
299 enum mlx5_ib_cong_node_type node;
300 struct mlx5_core_dev *mdev;
301
302 /* Takes a 1-based port number */
303 mdev = mlx5_ib_get_native_port_mdev(dev, port_num + 1, NULL);
304 if (!mdev)
305 return -ENODEV;
306
307 out = kvzalloc(outlen, GFP_KERNEL);
308 if (!out) {
309 err = -ENOMEM;
310 goto alloc_err;
311 }
312
313 node = mlx5_ib_param_to_node(offset);
314
315 err = mlx5_cmd_query_cong_params(mdev, node, out);
316 if (err)
317 goto free;
318
319 field = MLX5_ADDR_OF(query_cong_params_out, out, congestion_parameters);
320 *var = mlx5_get_cc_param_val(field, offset);
321
322 free:
323 kvfree(out);
324 alloc_err:
325 mlx5_ib_put_native_port_mdev(dev, port_num + 1);
326 return err;
327 }
328
mlx5_ib_set_cc_params(struct mlx5_ib_dev * dev,u32 port_num,int offset,u32 var)329 static int mlx5_ib_set_cc_params(struct mlx5_ib_dev *dev, u32 port_num,
330 int offset, u32 var)
331 {
332 int inlen = MLX5_ST_SZ_BYTES(modify_cong_params_in);
333 void *in;
334 void *field;
335 enum mlx5_ib_cong_node_type node;
336 struct mlx5_core_dev *mdev;
337 u32 attr_mask = 0;
338 int err;
339
340 /* Takes a 1-based port number */
341 mdev = mlx5_ib_get_native_port_mdev(dev, port_num + 1, NULL);
342 if (!mdev)
343 return -ENODEV;
344
345 in = kvzalloc(inlen, GFP_KERNEL);
346 if (!in) {
347 err = -ENOMEM;
348 goto alloc_err;
349 }
350
351 MLX5_SET(modify_cong_params_in, in, opcode,
352 MLX5_CMD_OP_MODIFY_CONG_PARAMS);
353
354 node = mlx5_ib_param_to_node(offset);
355 MLX5_SET(modify_cong_params_in, in, cong_protocol, node);
356
357 field = MLX5_ADDR_OF(modify_cong_params_in, in, congestion_parameters);
358 mlx5_ib_set_cc_param_mask_val(field, offset, var, &attr_mask);
359
360 field = MLX5_ADDR_OF(modify_cong_params_in, in, field_select);
361 MLX5_SET(field_select_r_roce_rp, field, field_select_r_roce_rp,
362 attr_mask);
363
364 err = mlx5_cmd_exec_in(dev->mdev, modify_cong_params, in);
365 kvfree(in);
366 alloc_err:
367 mlx5_ib_put_native_port_mdev(dev, port_num + 1);
368 return err;
369 }
370
set_param(struct file * filp,const char __user * buf,size_t count,loff_t * pos)371 static ssize_t set_param(struct file *filp, const char __user *buf,
372 size_t count, loff_t *pos)
373 {
374 struct mlx5_ib_dbg_param *param = filp->private_data;
375 int offset = param->offset;
376 char lbuf[11] = { };
377 u32 var;
378 int ret;
379
380 if (count > sizeof(lbuf))
381 return -EINVAL;
382
383 if (copy_from_user(lbuf, buf, count))
384 return -EFAULT;
385
386 lbuf[sizeof(lbuf) - 1] = '\0';
387
388 if (kstrtou32(lbuf, 0, &var))
389 return -EINVAL;
390
391 ret = mlx5_ib_set_cc_params(param->dev, param->port_num, offset, var);
392 return ret ? ret : count;
393 }
394
get_param(struct file * filp,char __user * buf,size_t count,loff_t * pos)395 static ssize_t get_param(struct file *filp, char __user *buf, size_t count,
396 loff_t *pos)
397 {
398 struct mlx5_ib_dbg_param *param = filp->private_data;
399 int offset = param->offset;
400 u32 var = 0;
401 int ret;
402 char lbuf[11];
403
404 ret = mlx5_ib_get_cc_params(param->dev, param->port_num, offset, &var);
405 if (ret)
406 return ret;
407
408 ret = snprintf(lbuf, sizeof(lbuf), "%d\n", var);
409 if (ret < 0)
410 return ret;
411
412 return simple_read_from_buffer(buf, count, pos, lbuf, ret);
413 }
414
415 static const struct file_operations dbg_cc_fops = {
416 .owner = THIS_MODULE,
417 .open = simple_open,
418 .write = set_param,
419 .read = get_param,
420 };
421
mlx5_ib_cleanup_cong_debugfs(struct mlx5_ib_dev * dev,u32 port_num)422 void mlx5_ib_cleanup_cong_debugfs(struct mlx5_ib_dev *dev, u32 port_num)
423 {
424 if (!mlx5_debugfs_root ||
425 !dev->port[port_num].dbg_cc_params ||
426 !dev->port[port_num].dbg_cc_params->root)
427 return;
428
429 debugfs_remove_recursive(dev->port[port_num].dbg_cc_params->root);
430 kfree(dev->port[port_num].dbg_cc_params);
431 dev->port[port_num].dbg_cc_params = NULL;
432 }
433
mlx5_ib_init_cong_debugfs(struct mlx5_ib_dev * dev,u32 port_num)434 void mlx5_ib_init_cong_debugfs(struct mlx5_ib_dev *dev, u32 port_num)
435 {
436 struct mlx5_ib_dbg_cc_params *dbg_cc_params;
437 struct mlx5_core_dev *mdev;
438 int i;
439
440 if (!mlx5_debugfs_root)
441 return;
442
443 /* Takes a 1-based port number */
444 mdev = mlx5_ib_get_native_port_mdev(dev, port_num + 1, NULL);
445 if (!mdev)
446 return;
447
448 if (!MLX5_CAP_GEN(mdev, cc_query_allowed) ||
449 !MLX5_CAP_GEN(mdev, cc_modify_allowed))
450 goto put_mdev;
451
452 dbg_cc_params = kzalloc(sizeof(*dbg_cc_params), GFP_KERNEL);
453 if (!dbg_cc_params)
454 goto err;
455
456 dev->port[port_num].dbg_cc_params = dbg_cc_params;
457
458 dbg_cc_params->root = debugfs_create_dir("cc_params", mlx5_debugfs_get_dev_root(mdev));
459
460 for (i = 0; i < MLX5_IB_DBG_CC_MAX; i++) {
461 if ((i == MLX5_IB_DBG_CC_GENERAL_RTT_RESP_DSCP_VALID ||
462 i == MLX5_IB_DBG_CC_GENERAL_RTT_RESP_DSCP))
463 if (!MLX5_CAP_GEN(mdev, roce) ||
464 !MLX5_CAP_ROCE(mdev, roce_cc_general))
465 continue;
466
467 dbg_cc_params->params[i].offset = i;
468 dbg_cc_params->params[i].dev = dev;
469 dbg_cc_params->params[i].port_num = port_num;
470 dbg_cc_params->params[i].dentry =
471 debugfs_create_file(mlx5_ib_dbg_cc_name[i],
472 0600, dbg_cc_params->root,
473 &dbg_cc_params->params[i],
474 &dbg_cc_fops);
475 }
476
477 put_mdev:
478 mlx5_ib_put_native_port_mdev(dev, port_num + 1);
479 return;
480
481 err:
482 mlx5_ib_warn(dev, "cong debugfs failure\n");
483 mlx5_ib_cleanup_cong_debugfs(dev, port_num);
484 mlx5_ib_put_native_port_mdev(dev, port_num + 1);
485
486 /*
487 * We don't want to fail driver if debugfs failed to initialize,
488 * so we are not forwarding error to the user.
489 */
490 return;
491 }
492