1 | ///////////////////////////////////////////////////////////////////////////////////////////// |
---|
2 | // File : main.c (for classif application) |
---|
3 | // Date : november 2014 |
---|
4 | // author : Alain Greiner |
---|
5 | ///////////////////////////////////////////////////////////////////////////////////////////// |
---|
6 | // This multi-threaded application takes a Gigabit Ethernet packets stream, and |
---|
7 | // makes a packet analysis and classification, based on the source MAC address. |
---|
8 | // It uses the NIC peripheral, and the chained buffer build by the CMA component |
---|
9 | // to consume packets on the Gigabit Ethernet port. |
---|
10 | // |
---|
11 | // It is described as a TCG (Task and Communication Graph) containing |
---|
12 | // - one "load" task per cluster. |
---|
13 | // - from one to three "analyse" tasks per cluster. |
---|
14 | // In each cluster, the "load" task communicates with the local "analyse" tasks through |
---|
15 | // a local MWMR fifo containing NB_PROCS_MAX containers (one item = one container). |
---|
16 | // The MWMR fifo descriptor and the data buffer containing the containers are defined |
---|
17 | // as global variables distributed in (up to) 16 clusters. |
---|
18 | // |
---|
19 | // Initialisation is done in two steps by the "load" tasks: |
---|
20 | // - Task "load" in cluster[0][0] initialises NIC & CMA channel, and initialises |
---|
21 | // the barrier between all "load" tasks. Other "load" tasks are waiting on the |
---|
22 | // global_sync synchronisation variable. |
---|
23 | // - In each cluster[x][y], the load" task allocates the MWMR fifo descriptor & the data |
---|
24 | // buffer in the local heap, and store the pointers on a global array of pointers. |
---|
25 | // The "analyse" tasks are waiting on the sync[x][y] variables. |
---|
26 | // |
---|
27 | // When initialisation is completed, all "load and "analyse" tasks loop on containers: |
---|
28 | // 1) The "load" task transfer containers from the kernel chbuf associated to the |
---|
29 | // NIC_RX channel (in cluster[0][0]), to the local MWMR fifo (in cluster[x][y]), |
---|
30 | // after an in termediate copy in a private stack buffer. |
---|
31 | // Each "load" task loads CONTAINERS_MAX containers before exit, and the |
---|
32 | // task in cluster[0,0] displays the results stored in global counters filled |
---|
33 | // by the "analyse" tasks when all "load" tasks reach the barrier. |
---|
34 | // |
---|
35 | // 2) The "analyse" task transfer one container from the local MWMR fifo to a private |
---|
36 | // local buffer. It analyse each packet contained in the container, compute the |
---|
37 | // packet type, depending on the 4 MSB bits of the source MAC address, |
---|
38 | // and increment the corresponding counters. |
---|
39 | // |
---|
40 | // It uses the he following hardware parameters, defined in the hard_config.h file: |
---|
41 | // - X_SIZE : number of clusters in a row |
---|
42 | // - Y_SIZE : number of clusters in a column |
---|
43 | // - NB_PROCS_MAX : number of processors per cluster |
---|
44 | ///////////////////////////////////////////////////////////////////////////////////////////// |
---|
45 | |
---|
46 | #include "stdio.h" |
---|
47 | #include "barrier.h" |
---|
48 | #include "malloc.h" |
---|
49 | #include "user_lock.h" |
---|
50 | #include "mwmr_channel.h" |
---|
51 | #include "hard_config.h" |
---|
52 | |
---|
53 | #define LOAD_VERBOSE 0 |
---|
54 | #define ANALYSE_VERBOSE 0 |
---|
55 | #define CONTAINERS_MAX 10 |
---|
56 | |
---|
57 | /////////////////////////////////////////////////////////////////////////////////////////// |
---|
58 | // Global variables |
---|
59 | // The communication channels are distributed in the clusters, |
---|
60 | // but the pointers arrays are global variables in cluster[0][0] |
---|
61 | /////////////////////////////////////////////////////////////////////////////////////////// |
---|
62 | |
---|
63 | mwmr_channel_t* mwmr[X_SIZE][Y_SIZE]; // distributed MWMR fifos pointers |
---|
64 | |
---|
65 | unsigned int local_sync[X_SIZE][Y_SIZE]; // distributed synchros "load" / "analyse" |
---|
66 | |
---|
67 | unsigned int global_sync = 0; // global synchro between "load" tasks |
---|
68 | |
---|
69 | unsigned int count[16]; // instrumentation counters |
---|
70 | |
---|
71 | giet_barrier_t barrier; // barrier between "load" (instrumentation) |
---|
72 | |
---|
73 | unsigned int nic_channel; // allocated NIC channel index |
---|
74 | |
---|
75 | /////////////////////////////////////////////////////////////////////////////////////////// |
---|
76 | __attribute__ ((constructor)) void load() |
---|
77 | /////////////////////////////////////////////////////////////////////////////////////////// |
---|
78 | { |
---|
79 | // get processor identifiers |
---|
80 | unsigned int x; |
---|
81 | unsigned int y; |
---|
82 | unsigned int l; |
---|
83 | giet_proc_xyp( &x, &y, &l ); |
---|
84 | |
---|
85 | if (X_SIZE > 4 ) giet_exit("The X_SIZE parameter cannot be larger than 4\n"); |
---|
86 | if (Y_SIZE > 4 ) giet_exit("The Y_SIZE parameter cannot be larger than 4\n"); |
---|
87 | |
---|
88 | // local buffer to store one container in private stack |
---|
89 | unsigned int temp[1024]; |
---|
90 | |
---|
91 | // giet_shr_printf("\n@@@ P[%d,%d,%d] enters load task at cycle %d\n", |
---|
92 | // x, y, l, giet_proctime() ); |
---|
93 | |
---|
94 | // allocates data buffer for MWMR fifo in local heap |
---|
95 | unsigned int* data = malloc( NB_PROCS_MAX<<12 ); |
---|
96 | |
---|
97 | //giet_shr_printf("\n@@@ P[%d,%d,%d] completes data malloc at cycle %d " |
---|
98 | // "/ &data = %x\n", x, y, l, giet_proctime(), (unsigned int)data ); |
---|
99 | |
---|
100 | // allocates MWMR fifo descriptor in local heap |
---|
101 | mwmr_channel_t* fifo = malloc( sizeof(mwmr_channel_t) ); |
---|
102 | |
---|
103 | //giet_shr_printf("\n@@@ P[%d,%d,%d] completes mwmr malloc at cycle %d " |
---|
104 | // "/ &mwmr = %x\n", x, y, l, giet_proctime(), (unsigned int)mwmr ); |
---|
105 | |
---|
106 | // makes copy of pointer in global array for "analyse" tasks |
---|
107 | mwmr[x][y] = fifo; |
---|
108 | |
---|
109 | // display status for cluster[X_SIZE-1][Y_SIZE-1] |
---|
110 | if ( (x==X_SIZE-1) && (y==Y_SIZE-1) ) |
---|
111 | giet_shr_printf("\n*** Task load starts on P[%d,%d,%d] at cycle %d\n" |
---|
112 | " &fifo = %x / &data = %x / &sync = %x\n" |
---|
113 | " x_size = %d / y_size = %d / nprocs = %d\n", |
---|
114 | x , y , l , giet_proctime() , |
---|
115 | (unsigned int)fifo , (unsigned int)data, (unsigned int)(&local_sync[x][y]) , |
---|
116 | X_SIZE, Y_SIZE, NB_PROCS_MAX ); |
---|
117 | |
---|
118 | // Task load on cluster[0,0] makes global initialisation: |
---|
119 | // - NIC & CMA channels allocation & initialisation. |
---|
120 | // - barrier for all load tasks initialisation. |
---|
121 | // Other load task wait completion. |
---|
122 | if ( (x==0) && (y==0) ) |
---|
123 | { |
---|
124 | // get NIC_RX channel |
---|
125 | nic_channel = giet_nic_rx_alloc(); |
---|
126 | |
---|
127 | // start CMA transfer |
---|
128 | giet_nic_rx_start(); |
---|
129 | |
---|
130 | // barrier init |
---|
131 | barrier_init( &barrier, X_SIZE * Y_SIZE ); |
---|
132 | |
---|
133 | // clear NIC RX channels stats |
---|
134 | giet_nic_rx_clear(); |
---|
135 | |
---|
136 | global_sync = 1; |
---|
137 | } |
---|
138 | else |
---|
139 | { |
---|
140 | while ( global_sync == 0 ) asm volatile ("nop"); |
---|
141 | } |
---|
142 | |
---|
143 | // Each load task initialises local MWMR fifo (width = 4kbytes / depth = NB_PROCS_MAX) |
---|
144 | mwmr_init( fifo , data , 1024 , NB_PROCS_MAX ); |
---|
145 | |
---|
146 | // signal MWMR fifo initialisation completion to analyse tasks |
---|
147 | local_sync[x][y] = 1; |
---|
148 | |
---|
149 | // main loop (on containers) |
---|
150 | unsigned int container = 0; |
---|
151 | while ( container < CONTAINERS_MAX ) |
---|
152 | { |
---|
153 | // get one container from kernel chbuf |
---|
154 | giet_nic_rx_move( nic_channel, temp ); |
---|
155 | |
---|
156 | // get packets number |
---|
157 | unsigned int npackets = temp[0] & 0x0000FFFF; |
---|
158 | unsigned int nwords = temp[0] >> 16; |
---|
159 | |
---|
160 | if ( (x==X_SIZE-1) && (y==Y_SIZE-1) ) |
---|
161 | giet_shr_printf("\nTask load on P[%d,%d,%d] get container %d at cycle %d" |
---|
162 | " : %d packets / %d words\n", |
---|
163 | x, y, l, container, giet_proctime(), npackets, nwords ); |
---|
164 | |
---|
165 | // put container to MWMR channel |
---|
166 | mwmr_write( fifo, temp, 1 ); |
---|
167 | |
---|
168 | container++; |
---|
169 | } |
---|
170 | |
---|
171 | // all load tasks synchronise before result display |
---|
172 | barrier_wait( &barrier ); |
---|
173 | |
---|
174 | // Task load in cluster[0,0] stops NIC and displays results |
---|
175 | if ( (x==0) && (y==0) ) |
---|
176 | { |
---|
177 | giet_nic_rx_stop(); |
---|
178 | |
---|
179 | giet_shr_printf("\n@@@@ Clasification Results @@@\n" |
---|
180 | " - TYPE 0 : %d packets\n" |
---|
181 | " - TYPE 1 : %d packets\n" |
---|
182 | " - TYPE 2 : %d packets\n" |
---|
183 | " - TYPE 3 : %d packets\n" |
---|
184 | " - TYPE 4 : %d packets\n" |
---|
185 | " - TYPE 5 : %d packets\n" |
---|
186 | " - TYPE 6 : %d packets\n" |
---|
187 | " - TYPE 7 : %d packets\n" |
---|
188 | " - TYPE 8 : %d packets\n" |
---|
189 | " - TYPE 9 : %d packets\n" |
---|
190 | " - TYPE A : %d packets\n" |
---|
191 | " - TYPE B : %d packets\n" |
---|
192 | " - TYPE C : %d packets\n" |
---|
193 | " - TYPE D : %d packets\n" |
---|
194 | " - TYPE E : %d packets\n" |
---|
195 | " - TYPE F : %d packets\n", |
---|
196 | count[0x0], count[0x1], count[0x2], count[0x3], |
---|
197 | count[0x4], count[0x5], count[0x6], count[0x7], |
---|
198 | count[0x8], count[0x9], count[0xA], count[0xB], |
---|
199 | count[0xC], count[0xD], count[0xE], count[0xF] ); |
---|
200 | |
---|
201 | giet_nic_rx_stats(); |
---|
202 | |
---|
203 | } |
---|
204 | |
---|
205 | // all load tasks synchronise before exit |
---|
206 | barrier_wait( &barrier ); |
---|
207 | |
---|
208 | giet_exit("Task completed"); |
---|
209 | |
---|
210 | } // end load() |
---|
211 | |
---|
212 | //////////////////////////////////////////// |
---|
213 | __attribute__ ((constructor)) void analyse() |
---|
214 | //////////////////////////////////////////// |
---|
215 | { |
---|
216 | // get processor identifiers |
---|
217 | unsigned int x; |
---|
218 | unsigned int y; |
---|
219 | unsigned int l; |
---|
220 | giet_proc_xyp( &x, &y, &l ); |
---|
221 | |
---|
222 | // local buffer to store one container |
---|
223 | unsigned int temp[1024]; |
---|
224 | |
---|
225 | // wait MWMR channel initialisation (done by task load) |
---|
226 | while ( local_sync[x][y] == 0 ) asm volatile ("nop"); |
---|
227 | |
---|
228 | // get pointer on MWMR channel descriptor |
---|
229 | mwmr_channel_t* fifo = mwmr[x][y]; |
---|
230 | |
---|
231 | // display status for cluster[X_SIZE-1][Y_SIZE-1] |
---|
232 | if ( (x==X_SIZE-1) && (y==Y_SIZE-1) ) |
---|
233 | giet_shr_printf("\n*** Task analyse starts on P[%d,%d,%d] at cycle %d\n" |
---|
234 | " &fifo = %x / &sync = %x\n", |
---|
235 | x, y, l, giet_proctime(), |
---|
236 | (unsigned int)fifo, (unsigned int)(&local_sync[x][y]) ); |
---|
237 | |
---|
238 | // main loop (on containers) |
---|
239 | unsigned int nwords; // number of words in container |
---|
240 | unsigned int npackets; // number of packets in container |
---|
241 | unsigned int length; // number of bytes in current packet |
---|
242 | unsigned int word; // current packet first word in container |
---|
243 | unsigned int type; // current packet type |
---|
244 | unsigned int p; // current packet index |
---|
245 | while ( 1 ) |
---|
246 | { |
---|
247 | // get one container from MWMR fifo |
---|
248 | mwmr_read( fifo, temp, 1 ); |
---|
249 | |
---|
250 | // get packets number |
---|
251 | npackets = temp[0] & 0x0000FFFF; |
---|
252 | nwords = temp[0] >> 16; |
---|
253 | |
---|
254 | if ( (x==X_SIZE-1) && (y==Y_SIZE-1) ) |
---|
255 | giet_shr_printf("\nTask analyse on P[%d,%d,%d] get container at cycle %d" |
---|
256 | " : %d packets / %d words\n", |
---|
257 | x, y, l, giet_proctime(), npackets, nwords ); |
---|
258 | |
---|
259 | // initialize word index in container |
---|
260 | word = 34; |
---|
261 | |
---|
262 | // loop on packets |
---|
263 | for( p = 0 ; p < npackets ; p++ ) |
---|
264 | { |
---|
265 | // get packet length from container header |
---|
266 | if ( (p & 0x1) == 0 ) length = temp[1+(p>>1)] >> 16; |
---|
267 | else length = temp[1+(p>>1)] & 0x0000FFFF; |
---|
268 | |
---|
269 | // get packet type (source mac address 4 MSB bits) |
---|
270 | type = (temp[word+1] & 0x0000F000) >> 12; |
---|
271 | |
---|
272 | // increment counter |
---|
273 | atomic_increment( &count[type], 1 ); |
---|
274 | |
---|
275 | // update word index |
---|
276 | if ( length & 0x3 ) word += (length>>2)+1; |
---|
277 | else word += (length>>2); |
---|
278 | } |
---|
279 | } |
---|
280 | } // end analyse() |
---|
281 | |
---|