2 * Copyright (c) 2013-2016, Mellanox Technologies. All rights reserved.
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:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
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.
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
33 #include <linux/kernel.h>
34 #include <linux/module.h>
35 #include <linux/mlx5/driver.h>
36 #include "mlx5_core.h"
38 /* Scheduling element fw management */
39 int mlx5_create_scheduling_element_cmd(struct mlx5_core_dev
*dev
, u8 hierarchy
,
40 void *ctx
, u32
*element_id
)
42 u32 out
[MLX5_ST_SZ_DW(create_scheduling_element_in
)] = {};
43 u32 in
[MLX5_ST_SZ_DW(create_scheduling_element_in
)] = {};
47 schedc
= MLX5_ADDR_OF(create_scheduling_element_in
, in
,
49 MLX5_SET(create_scheduling_element_in
, in
, opcode
,
50 MLX5_CMD_OP_CREATE_SCHEDULING_ELEMENT
);
51 MLX5_SET(create_scheduling_element_in
, in
, scheduling_hierarchy
,
53 memcpy(schedc
, ctx
, MLX5_ST_SZ_BYTES(scheduling_context
));
55 err
= mlx5_cmd_exec_inout(dev
, create_scheduling_element
, in
, out
);
59 *element_id
= MLX5_GET(create_scheduling_element_out
, out
,
60 scheduling_element_id
);
64 int mlx5_modify_scheduling_element_cmd(struct mlx5_core_dev
*dev
, u8 hierarchy
,
65 void *ctx
, u32 element_id
,
68 u32 in
[MLX5_ST_SZ_DW(modify_scheduling_element_in
)] = {};
71 schedc
= MLX5_ADDR_OF(modify_scheduling_element_in
, in
,
73 MLX5_SET(modify_scheduling_element_in
, in
, opcode
,
74 MLX5_CMD_OP_MODIFY_SCHEDULING_ELEMENT
);
75 MLX5_SET(modify_scheduling_element_in
, in
, scheduling_element_id
,
77 MLX5_SET(modify_scheduling_element_in
, in
, modify_bitmask
,
79 MLX5_SET(modify_scheduling_element_in
, in
, scheduling_hierarchy
,
81 memcpy(schedc
, ctx
, MLX5_ST_SZ_BYTES(scheduling_context
));
83 return mlx5_cmd_exec_in(dev
, modify_scheduling_element
, in
);
86 int mlx5_destroy_scheduling_element_cmd(struct mlx5_core_dev
*dev
, u8 hierarchy
,
89 u32 in
[MLX5_ST_SZ_DW(destroy_scheduling_element_in
)] = {};
91 MLX5_SET(destroy_scheduling_element_in
, in
, opcode
,
92 MLX5_CMD_OP_DESTROY_SCHEDULING_ELEMENT
);
93 MLX5_SET(destroy_scheduling_element_in
, in
, scheduling_element_id
,
95 MLX5_SET(destroy_scheduling_element_in
, in
, scheduling_hierarchy
,
98 return mlx5_cmd_exec_in(dev
, destroy_scheduling_element
, in
);
101 static bool mlx5_rl_are_equal_raw(struct mlx5_rl_entry
*entry
, void *rl_in
,
104 return (!memcmp(entry
->rl_raw
, rl_in
, sizeof(entry
->rl_raw
)) &&
108 /* Finds an entry where we can register the given rate
109 * If the rate already exists, return the entry where it is registered,
110 * otherwise return the first available entry.
111 * If the table is full, return NULL
113 static struct mlx5_rl_entry
*find_rl_entry(struct mlx5_rl_table
*table
,
114 void *rl_in
, u16 uid
, bool dedicated
)
116 struct mlx5_rl_entry
*ret_entry
= NULL
;
117 bool empty_found
= false;
120 for (i
= 0; i
< table
->max_size
; i
++) {
122 if (!table
->rl_entry
[i
].refcount
)
123 return &table
->rl_entry
[i
];
127 if (table
->rl_entry
[i
].refcount
) {
128 if (table
->rl_entry
[i
].dedicated
)
130 if (mlx5_rl_are_equal_raw(&table
->rl_entry
[i
], rl_in
,
132 return &table
->rl_entry
[i
];
133 } else if (!empty_found
) {
135 ret_entry
= &table
->rl_entry
[i
];
142 static int mlx5_set_pp_rate_limit_cmd(struct mlx5_core_dev
*dev
,
143 struct mlx5_rl_entry
*entry
, bool set
)
145 u32 in
[MLX5_ST_SZ_DW(set_pp_rate_limit_in
)] = {};
148 pp_context
= MLX5_ADDR_OF(set_pp_rate_limit_in
, in
, ctx
);
149 MLX5_SET(set_pp_rate_limit_in
, in
, opcode
,
150 MLX5_CMD_OP_SET_PP_RATE_LIMIT
);
151 MLX5_SET(set_pp_rate_limit_in
, in
, uid
, entry
->uid
);
152 MLX5_SET(set_pp_rate_limit_in
, in
, rate_limit_index
, entry
->index
);
154 memcpy(pp_context
, entry
->rl_raw
, sizeof(entry
->rl_raw
));
155 return mlx5_cmd_exec_in(dev
, set_pp_rate_limit
, in
);
158 bool mlx5_rl_is_in_range(struct mlx5_core_dev
*dev
, u32 rate
)
160 struct mlx5_rl_table
*table
= &dev
->priv
.rl_table
;
162 return (rate
<= table
->max_rate
&& rate
>= table
->min_rate
);
164 EXPORT_SYMBOL(mlx5_rl_is_in_range
);
166 bool mlx5_rl_are_equal(struct mlx5_rate_limit
*rl_0
,
167 struct mlx5_rate_limit
*rl_1
)
169 return ((rl_0
->rate
== rl_1
->rate
) &&
170 (rl_0
->max_burst_sz
== rl_1
->max_burst_sz
) &&
171 (rl_0
->typical_pkt_sz
== rl_1
->typical_pkt_sz
));
173 EXPORT_SYMBOL(mlx5_rl_are_equal
);
175 int mlx5_rl_add_rate_raw(struct mlx5_core_dev
*dev
, void *rl_in
, u16 uid
,
176 bool dedicated_entry
, u16
*index
)
178 struct mlx5_rl_table
*table
= &dev
->priv
.rl_table
;
179 struct mlx5_rl_entry
*entry
;
183 rate
= MLX5_GET(set_pp_rate_limit_context
, rl_in
, rate_limit
);
184 mutex_lock(&table
->rl_lock
);
186 if (!rate
|| !mlx5_rl_is_in_range(dev
, rate
)) {
187 mlx5_core_err(dev
, "Invalid rate: %u, should be %u to %u\n",
188 rate
, table
->min_rate
, table
->max_rate
);
193 entry
= find_rl_entry(table
, rl_in
, uid
, dedicated_entry
);
195 mlx5_core_err(dev
, "Max number of %u rates reached\n",
200 if (entry
->refcount
) {
201 /* rate already configured */
204 memcpy(entry
->rl_raw
, rl_in
, sizeof(entry
->rl_raw
));
207 err
= mlx5_set_pp_rate_limit_cmd(dev
, entry
, true);
211 "Failed configuring rate limit(err %d): rate %u, max_burst_sz %u, typical_pkt_sz %u\n",
213 MLX5_GET(set_pp_rate_limit_context
, rl_in
,
215 MLX5_GET(set_pp_rate_limit_context
, rl_in
,
216 typical_packet_size
));
221 entry
->dedicated
= dedicated_entry
;
223 *index
= entry
->index
;
226 mutex_unlock(&table
->rl_lock
);
229 EXPORT_SYMBOL(mlx5_rl_add_rate_raw
);
231 void mlx5_rl_remove_rate_raw(struct mlx5_core_dev
*dev
, u16 index
)
233 struct mlx5_rl_table
*table
= &dev
->priv
.rl_table
;
234 struct mlx5_rl_entry
*entry
;
236 mutex_lock(&table
->rl_lock
);
237 entry
= &table
->rl_entry
[index
- 1];
239 if (!entry
->refcount
)
240 /* need to remove rate */
241 mlx5_set_pp_rate_limit_cmd(dev
, entry
, false);
242 mutex_unlock(&table
->rl_lock
);
244 EXPORT_SYMBOL(mlx5_rl_remove_rate_raw
);
246 int mlx5_rl_add_rate(struct mlx5_core_dev
*dev
, u16
*index
,
247 struct mlx5_rate_limit
*rl
)
249 u8 rl_raw
[MLX5_ST_SZ_BYTES(set_pp_rate_limit_context
)] = {};
251 MLX5_SET(set_pp_rate_limit_context
, rl_raw
, rate_limit
, rl
->rate
);
252 MLX5_SET(set_pp_rate_limit_context
, rl_raw
, burst_upper_bound
,
254 MLX5_SET(set_pp_rate_limit_context
, rl_raw
, typical_packet_size
,
257 return mlx5_rl_add_rate_raw(dev
, rl_raw
,
258 MLX5_CAP_QOS(dev
, packet_pacing_uid
) ?
259 MLX5_SHARED_RESOURCE_UID
: 0,
262 EXPORT_SYMBOL(mlx5_rl_add_rate
);
264 void mlx5_rl_remove_rate(struct mlx5_core_dev
*dev
, struct mlx5_rate_limit
*rl
)
266 u8 rl_raw
[MLX5_ST_SZ_BYTES(set_pp_rate_limit_context
)] = {};
267 struct mlx5_rl_table
*table
= &dev
->priv
.rl_table
;
268 struct mlx5_rl_entry
*entry
= NULL
;
270 /* 0 is a reserved value for unlimited rate */
274 MLX5_SET(set_pp_rate_limit_context
, rl_raw
, rate_limit
, rl
->rate
);
275 MLX5_SET(set_pp_rate_limit_context
, rl_raw
, burst_upper_bound
,
277 MLX5_SET(set_pp_rate_limit_context
, rl_raw
, typical_packet_size
,
280 mutex_lock(&table
->rl_lock
);
281 entry
= find_rl_entry(table
, rl_raw
,
282 MLX5_CAP_QOS(dev
, packet_pacing_uid
) ?
283 MLX5_SHARED_RESOURCE_UID
: 0, false);
284 if (!entry
|| !entry
->refcount
) {
285 mlx5_core_warn(dev
, "Rate %u, max_burst_sz %u typical_pkt_sz %u are not configured\n",
286 rl
->rate
, rl
->max_burst_sz
, rl
->typical_pkt_sz
);
291 if (!entry
->refcount
)
292 /* need to remove rate */
293 mlx5_set_pp_rate_limit_cmd(dev
, entry
, false);
296 mutex_unlock(&table
->rl_lock
);
298 EXPORT_SYMBOL(mlx5_rl_remove_rate
);
300 int mlx5_init_rl_table(struct mlx5_core_dev
*dev
)
302 struct mlx5_rl_table
*table
= &dev
->priv
.rl_table
;
305 mutex_init(&table
->rl_lock
);
306 if (!MLX5_CAP_GEN(dev
, qos
) || !MLX5_CAP_QOS(dev
, packet_pacing
)) {
311 /* First entry is reserved for unlimited rate */
312 table
->max_size
= MLX5_CAP_QOS(dev
, packet_pacing_rate_table_size
) - 1;
313 table
->max_rate
= MLX5_CAP_QOS(dev
, packet_pacing_max_rate
);
314 table
->min_rate
= MLX5_CAP_QOS(dev
, packet_pacing_min_rate
);
316 table
->rl_entry
= kcalloc(table
->max_size
, sizeof(struct mlx5_rl_entry
),
318 if (!table
->rl_entry
)
321 /* The index represents the index in HW rate limit table
322 * Index 0 is reserved for unlimited rate
324 for (i
= 0; i
< table
->max_size
; i
++)
325 table
->rl_entry
[i
].index
= i
+ 1;
327 /* Index 0 is reserved */
328 mlx5_core_info(dev
, "Rate limit: %u rates are supported, range: %uMbps to %uMbps\n",
330 table
->min_rate
>> 10,
331 table
->max_rate
>> 10);
336 void mlx5_cleanup_rl_table(struct mlx5_core_dev
*dev
)
338 struct mlx5_rl_table
*table
= &dev
->priv
.rl_table
;
341 /* Clear all configured rates */
342 for (i
= 0; i
< table
->max_size
; i
++)
343 if (table
->rl_entry
[i
].refcount
)
344 mlx5_set_pp_rate_limit_cmd(dev
, &table
->rl_entry
[i
],
347 kfree(dev
->priv
.rl_table
.rl_entry
);