1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
|
/*
* Copyright (c) 2015 Cisco and/or its affiliates.
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at:
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef __POLICE_H__
#define __POLICE_H__
typedef enum
{
POLICE_CONFORM = 0,
POLICE_EXCEED = 1,
POLICE_VIOLATE = 2,
} policer_result_e;
#define NUM_POLICE_RESULTS 3
typedef enum
{
QOS_ACTION_DROP = 0,
QOS_ACTION_TRANSMIT,
QOS_ACTION_MARK_AND_TRANSMIT
} __clib_packed qos_action_type_en;
// This is the hardware representation of the policer.
// To be multithread-safe, the policer is accessed through a spin-lock
// on the lock field. (For a policer update operation, 24B needs to be
// modified and this would be a challenge to do with atomic instructions.)
// The structure is padded so that no other data is put into the same
// 64B cache-line. This reduces cache-thrashing between threads.
//
// A note on scale:
// The HW TSC tick is roughly one CPU clock cycle.
// This is shifted to create a larger period, with a goal to be around 50usec.
// The period time will vary based on CPU clock speed.
// CPU speeds of 1Ghz to 8Ghz are targetted.
// The shift amount is a constant 17 bits, resulting in a period between
// 16usec (8Ghz CPU) and 131usec (1Ghz CPU).
// The token_per_period computation takes into account the clock speed.
//
// The 32-bit bucket/limit supports about 850ms of burst on a 40GE port,
// or 340ms on a 100GE port. If a larger burst is configured, then the
// programmed value is simply capped at 2^32-1. If we needed to support
// more than that, the bucket and limit fields could be expanded.
//
// tokens_per_period should be > 1000 to support 0.1% granularity.
// To support lower rates (which would not meet this requirement), the packet
// length, bucket, and limit values can be scaled. The scale is a power of 2
// so the multiplication can be implemented as a shift. The control plane
// computes the shift amount be the largest possible that still supports the
// burst size. This makes the rate accuracy as high as possible.
//
// The 64-bit last_update_time supports a 4Ghz CPU without rollover for 100
// years
//
// The lock field should be used for a spin-lock on the struct. Alternatively,
// a thread index field is provided so that policed packets may be handed
// off to a single worker thread.
#define POLICER_TICKS_PER_PERIOD_SHIFT 17
#define POLICER_TICKS_PER_PERIOD (1 << POLICER_TICKS_PER_PERIOD_SHIFT)
typedef struct
{
u32 lock; // for exclusive access to the struct
u32 single_rate; // 1 = single rate policer, 0 = two rate policer
u32 color_aware; // for hierarchical policing
u32 scale; // power-of-2 shift amount for lower rates
qos_action_type_en action[3];
ip_dscp_t mark_dscp[3];
u8 pad[2];
// Fields are marked as 2R if they are only used for a 2-rate policer,
// and MOD if they are modified as part of the update operation.
// 1 token = 1 byte.
u32 cir_tokens_per_period; // # of tokens for each period
u32 pir_tokens_per_period; // 2R
u32 current_limit;
u32 current_bucket; // MOD
u32 extended_limit;
u32 extended_bucket; // MOD
u64 last_update_time; // MOD
u32 thread_index; // Tie policer to a thread, rather than lock
u32 pad32;
} policer_read_response_type_st;
static inline policer_result_e
vnet_police_packet (policer_read_response_type_st * policer,
u32 packet_length,
policer_result_e packet_color, u64 time)
{
u64 n_periods;
u64 current_tokens, extended_tokens;
policer_result_e result;
// Scale packet length to support a wide range of speeds
packet_length = packet_length << policer->scale;
// Compute the number of policer periods that have passed since the last
// operation.
n_periods = time - policer->last_update_time;
policer->last_update_time = time;
// Since there is no background last-update-time adjustment, n_periods
// could grow large if the policer is idle for a long time. This could
// cause a 64-bit overflow when computing tokens_per_period * num_periods.
// It will overflow if log2(n_periods) + log2(tokens_per_period) > 64.
//
// To mitigate this, the policer configuration algorithm insures that
// tokens_per_period is less than 2^22, i.e. this is a 22 bit value not
// a 32-bit value. Thus overflow will only occur if n_periods > 64-22 or
// 42. 2^42 min-sized periods is 16us * 2^42, or 2 years. So this can
// rarely occur. If overflow does happen, the only effect will be that
// fewer tokens than the max burst will be added to the bucket for this
// packet. This constraint on tokens_per_period lets the ucode omit
// code to dynamically check for or prevent the overflow.
if (policer->single_rate)
{
// Compute number of tokens for this time period
current_tokens =
policer->current_bucket + n_periods * policer->cir_tokens_per_period;
if (current_tokens > policer->current_limit)
{
current_tokens = policer->current_limit;
}
extended_tokens =
policer->extended_bucket + n_periods * policer->cir_tokens_per_period;
if (extended_tokens > policer->extended_limit)
{
extended_tokens = policer->extended_limit;
}
// Determine color
if ((!policer->color_aware || (packet_color == POLICE_CONFORM))
&& (current_tokens >= packet_length))
{
policer->current_bucket = current_tokens - packet_length;
policer->extended_bucket = extended_tokens - packet_length;
result = POLICE_CONFORM;
}
else if ((!policer->color_aware || (packet_color != POLICE_VIOLATE))
&& (extended_tokens >= packet_length))
{
policer->current_bucket = current_tokens;
policer->extended_bucket = extended_tokens - packet_length;
result = POLICE_EXCEED;
}
else
{
policer->current_bucket = current_tokens;
policer->extended_bucket = extended_tokens;
result = POLICE_VIOLATE;
}
}
else
{
// Two-rate policer
// Compute number of tokens for this time period
current_tokens =
policer->current_bucket + n_periods * policer->cir_tokens_per_period;
extended_tokens =
policer->extended_bucket + n_periods * policer->pir_tokens_per_period;
if (current_tokens > policer->current_limit)
{
current_tokens = policer->current_limit;
}
if (extended_tokens > policer->extended_limit)
{
extended_tokens = policer->extended_limit;
}
// Determine color
if ((policer->color_aware && (packet_color == POLICE_VIOLATE))
|| (extended_tokens < packet_length))
{
policer->current_bucket = current_tokens;
policer->extended_bucket = extended_tokens;
result = POLICE_VIOLATE;
}
else if ((policer->color_aware && (packet_color == POLICE_EXCEED))
|| (current_tokens < packet_length))
{
policer->current_bucket = current_tokens;
policer->extended_bucket = extended_tokens - packet_length;
result = POLICE_EXCEED;
}
else
{
policer->current_bucket = current_tokens - packet_length;
policer->extended_bucket = extended_tokens - packet_length;
result = POLICE_CONFORM;
}
}
return result;
}
#endif // __POLICE_H__
/*
* fd.io coding-style-patch-verification: ON
*
* Local Variables:
* eval: (c-set-style "gnu")
* End:
*/
|
