aboutsummaryrefslogtreecommitdiffstats
path: root/bootstrap-DMM.sh
blob: 7117205f10e435a076589b4bae27e96104704fdf (plain)
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
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
#!/bin/bash
# Copyright (c) 2018 Huawei Technologies Co.,Ltd.
# 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.

set -x

cat /etc/hostname
cat /etc/hosts

JOB_ARCHIVE_ARTIFACTS=(log.html output.xml report.html *.log)
LOG_ARCHIVE_ARTIFACTS=(log.html output.xml report.html *.log)
JOB_ARCHIVE_DIR="archive"
LOG_ARCHIVE_DIR="$WORKSPACE/archives"
mkdir -p ${JOB_ARCHIVE_DIR}
mkdir -p ${LOG_ARCHIVE_DIR}
SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
export PYTHONPATH=${SCRIPT_DIR}

if [ -f "/etc/redhat-release" ]; then
    DISTRO="CENTOS"
    sudo yum install -y python-devel python-virtualenv
    DPDK_ARTIFACTS=""
    VIRL_TOPOLOGY=$(cat ${SCRIPT_DIR}/VIRL_TOPOLOGY_CENTOS)
    VIRL_RELEASE=$(cat ${SCRIPT_DIR}/VIRL_RELEASE_CENTOS)
else
    DISTRO="UBUNTU"
    export DEBIAN_FRONTEND=noninteractive
    sudo apt-get -y update
    sudo apt-get -y install libpython2.7-dev python-virtualenv
    DPDK_ARTIFACTS=""
    VIRL_TOPOLOGY=$(cat ${SCRIPT_DIR}/VIRL_TOPOLOGY_UBUNTU)
    VIRL_RELEASE=$(cat ${SCRIPT_DIR}/VIRL_RELEASE_UBUNTU)
fi

VIRL_SERVERS=("10.30.51.28" "10.30.51.29" "10.30.51.30")
IPS_PER_VIRL=( "10.30.51.28:252"
               "10.30.51.29:252"
               "10.30.51.30:252" )
SIMS_PER_VIRL=( "10.30.51.28:13"
               "10.30.51.29:13"
               "10.30.51.30:13" )
IPS_PER_SIMULATION=5

function get_max_ip_nr() {
    virl_server=$1
    IP_VALUE="0"
    for item in "${IPS_PER_VIRL[@]}" ; do
        if [ "${item%%:*}" == "${virl_server}" ]
        then
            IP_VALUE=${item#*:}
            break
        fi
    done
    echo "$IP_VALUE"
}

function get_max_sim_nr() {
    virl_server=$1
    SIM_VALUE="0"
    for item in "${SIMS_PER_VIRL[@]}" ; do
        if [ "${item%%:*}" == "${virl_server}" ]
        then
            SIM_VALUE=${item#*:}
            break
        fi
    done
    echo "$SIM_VALUE"
}

VIRL_USERNAME=jenkins-in
VIRL_PKEY=priv_key
VIRL_SERVER_STATUS_FILE="status"
VIRL_SERVER_EXPECTED_STATUS="PRODUCTION"

SSH_OPTIONS="-i ${VIRL_PKEY} -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null -o BatchMode=yes -o LogLevel=error"

TEST_GROUPS=("func")
SUITE_PATH="tests.dmm"
SKIP_PATCH="SKIP_PATCH"

# Create tmp dir
mkdir ${SCRIPT_DIR}/tmp

# Use tmp dir to store log files
LOG_PATH="${SCRIPT_DIR}/tmp"

# Use tmp dir for tarballs
export TMPDIR="${SCRIPT_DIR}/tmp"

function ssh_do() {
    echo
    echo "### "  ssh $@
    ssh ${SSH_OPTIONS} $@
}

rm -f ${VIRL_PKEY}
cat > ${VIRL_PKEY} <<EOF
-----BEGIN RSA PRIVATE KEY-----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-----END RSA PRIVATE KEY-----
EOF
chmod 600 ${VIRL_PKEY}

#
# The server must be reachable and have a "status" file with
# the content "PRODUCTION" to be selected.
#
# If the server is not reachable or does not have the correct
# status remove it from the array and start again.
#
# Abort if there are no more servers left in the array.
#
VIRL_PROD_SERVERS=()
for index in "${!VIRL_SERVERS[@]}"; do
    virl_server_status=$(ssh ${SSH_OPTIONS} ${VIRL_USERNAME}@${VIRL_SERVERS[$index]} cat $VIRL_SERVER_STATUS_FILE 2>&1)
    echo VIRL HOST ${VIRL_SERVERS[$index]} status is \"$virl_server_status\"
    if [ "$virl_server_status" == "$VIRL_SERVER_EXPECTED_STATUS" ]
    then
        # Candidate is in good status. Add to array.
        VIRL_PROD_SERVERS+=(${VIRL_SERVERS[$index]})
    fi
done

VIRL_SERVERS=("${VIRL_PROD_SERVERS[@]}")
echo "VIRL servers in production: ${VIRL_SERVERS[@]}"
num_hosts=${#VIRL_SERVERS[@]}
if [ $num_hosts == 0 ]
then
    echo "No more VIRL candidate hosts available, failing."
    exit 127
fi

# Get the LOAD of each server based on number of active simulations (testcases)
VIRL_SERVER_LOAD=()
for index in "${!VIRL_SERVERS[@]}"; do
    VIRL_SERVER_LOAD[${index}]=$(ssh ${SSH_OPTIONS} ${VIRL_USERNAME}@${VIRL_SERVERS[$index]} "list-testcases | grep session | wc -l")
done

# Pick for each TEST_GROUP least loaded server
VIRL_SERVER=()
for index in "${!TEST_GROUPS[@]}"; do
    least_load_server_idx=$(echo "${VIRL_SERVER_LOAD[*]}" | tr -s ' ' '\n' | awk '{print($0" "NR)}' | sort -g -k1,1 | head -1 | cut -f2 -d' ')
    least_load_server=${VIRL_SERVERS[$least_load_server_idx-1]}
    VIRL_SERVER+=($least_load_server)
    # Adjusting load as we are not going run simulation immediately
    VIRL_SERVER_LOAD[$least_load_server_idx-1]=$((VIRL_SERVER_LOAD[$least_load_server_idx-1]+1))
done

echo "Selected VIRL servers: ${VIRL_SERVER[@]}"

# Temporarily download  DPDK packages
DMM_TAR_FILE="dmm_depends.tar.gz"

cd dmm/scripts/
./build.sh all
cd -

DPDK_DOWNLOAD_PATH=$(cat dmm/scripts/build_dpdk.sh | grep DPDK_DOWNLOAD_PATH=  | cut -d "=" -f2)
mv $DPDK_DOWNLOAD_PATH/dpdk-18.02.tar.xz .

wget http://security.ubuntu.com/ubuntu/pool/main/n/numactl/libnuma1_2.0.11-1ubuntu1.1_amd64.deb
wget http://security.ubuntu.com/ubuntu/pool/main/n/numactl/libnuma-dev_2.0.11-1ubuntu1.1_amd64.deb
wget http://security.ubuntu.com/ubuntu/pool/main/e/ethtool/ethtool_4.5-1_amd64.deb
wget http://security.ubuntu.com/ubuntu/pool/main/l/lsof/lsof_4.89+dfsg-0.1_amd64.deb

tar zcf ${DMM_TAR_FILE} dpdk-18.02.tar.xz  ./dmm/ libnuma*.deb

VIRL_DIR_LOC="/tmp"

cat ${VIRL_PKEY}

# Copy the files to VIRL hosts
DONE=""
for index in "${!VIRL_SERVER[@]}"; do
    # Do not copy files in case they have already been copied to the VIRL host
    [[ "${DONE[@]}" =~ "${VIRL_SERVER[${index}]}" ]] && copy=0 || copy=1

    if [ "${copy}" -eq "0" ]; then
        echo "DMM_TAR_FILE has already been copied to the VIRL host ${VIRL_SERVER[${index}]}"
    else
        scp ${SSH_OPTIONS} ${DMM_TAR_FILE} \
        ${VIRL_USERNAME}@${VIRL_SERVER[${index}]}:${VIRL_DIR_LOC}/

        result=$?
        if [ "${result}" -ne "0" ]; then
            echo "Failed to copy DMM_TAR_FILE to VIRL host ${VIRL_SERVER[${index}]}"
            echo ${result}
            exit ${result}
        else
            echo "DMM_TAR_FILE successfully copied to the VIRL host ${VIRL_SERVER[${index}]}"
        fi
        DONE+=(${VIRL_SERVER[${index}]})
    fi
done

# Start a simulation on VIRL server

function stop_virl_simulation {
    for index in "${!VIRL_SERVER[@]}"; do
        ssh ${SSH_OPTIONS} ${VIRL_USERNAME}@${VIRL_SERVER[${index}]}\
            "stop-testcase ${VIRL_SID[${index}]}"
    done
}

# Upon script exit, cleanup the simulation execution
trap stop_virl_simulation EXIT

for index in "${!VIRL_SERVER[@]}"; do
    echo "Starting simulation nr. ${index} on VIRL server ${VIRL_SERVER[${index}]}"
    # Get given VIRL server limits for max. number of VMs and IPs
    max_ips=$(get_max_ip_nr ${VIRL_SERVER[${index}]})
    max_ips_from_sims=$(($(get_max_sim_nr ${VIRL_SERVER[${index}]})*IPS_PER_SIMULATION))
    # Set quota to lower value
    IP_QUOTA=$([ $max_ips -le $max_ips_from_sims ] && echo "$max_ips" || echo "$max_ips_from_sims")
    # Start the simulation
    VIRL_SID[${index}]=$(ssh ${SSH_OPTIONS} \
        ${VIRL_USERNAME}@${VIRL_SERVER[${index}]} \
        "start-testcase-DMM -vv --quota ${IP_QUOTA} --copy ${VIRL_TOPOLOGY} \
        --release ${VIRL_RELEASE} ${VIRL_DIR_LOC}/${DMM_TAR_FILE}")
        # TODO: remove param ${DMM_TAR_FILE} when start-testcase script is
        # updated on all virl servers
    retval=$?
    if [ ${retval} -ne "0" ]; then
        echo "VIRL simulation start failed on ${VIRL_SERVER[${index}]}"
        exit ${retval}
    fi
    if [[ ! "${VIRL_SID[${index}]}" =~ session-[a-zA-Z0-9_]{6} ]]; then
        echo "No VIRL session ID reported."
        exit 127
    fi
    echo "VIRL simulation nr. ${index} started on ${VIRL_SERVER[${index}]}"

    ssh_do ${VIRL_USERNAME}@${VIRL_SERVER[${index}]}\
     cat /scratch/${VIRL_SID[${index}]}/topology.yaml

    # Download the topology file from VIRL session and rename it
    scp ${SSH_OPTIONS} \
        ${VIRL_USERNAME}@${VIRL_SERVER[${index}]}:/scratch/${VIRL_SID[${index}]}/topology.yaml \
        topologies/enabled/topology${index}.yaml

    retval=$?
    if [ ${retval} -ne "0" ]; then
        echo "Failed to copy topology file from VIRL simulation nr. ${index} on VIRL server ${VIRL_SERVER[${index}]}"
        exit ${retval}
    fi
done

echo ${VIRL_SID[@]}

virtualenv --system-site-packages env
. env/bin/activate

echo pip install
pip install -r ${SCRIPT_DIR}/requirements.txt

for index in "${!VIRL_SERVER[@]}"; do
    pykwalify -s ${SCRIPT_DIR}/resources/topology_schemas/3_node_topology.sch.yaml \
              -s ${SCRIPT_DIR}/resources/topology_schemas/topology.sch.yaml \
              -d ${SCRIPT_DIR}/topologies/enabled/topology${index}.yaml \
              -vvv
    if [ "$?" -ne "0" ]; then
        echo "Topology${index} schema validation failed."
        echo "However, the tests will start."
    fi
done

function run_test_set() {
    set +x
    OLDIFS=$IFS
    IFS=","
    nr=$(echo $1)
    rm -f ${LOG_PATH}/test_run${nr}.log
    exec &> >(while read line; do echo "$(date +'%H:%M:%S') $line" \
     >> ${LOG_PATH}/test_run${nr}.log; done;)
    suite_str=""
    for suite in ${TEST_GROUPS[${nr}]}; do
        suite_str="${suite_str} --suite ${SUITE_PATH}.${suite}"
    done
    IFS=$OLDIFS

    echo "PYTHONPATH=`pwd` pybot -L TRACE -W 136\
        -v TOPOLOGY_PATH:${SCRIPT_DIR}/topologies/enabled/topology${nr}.yaml \
        ${suite_str} \
        --include vm_envAND3_node_single_link_topo \
        --include vm_envAND3_node_double_link_topo \
        --exclude PERFTEST \
        --exclude ${SKIP_PATCH} \
        --noncritical EXPECTED_FAILING \
        --output ${LOG_PATH}/log_test_set_run${nr} \
        tests/"

    PYTHONPATH=`pwd` pybot -L TRACE -W 136\
        -v TOPOLOGY_PATH:${SCRIPT_DIR}/topologies/enabled/topology${nr}.yaml \
        ${suite_str} \
        --include vm_envAND3_node_single_link_topo \
        --include vm_envAND3_node_double_link_topo \
        --exclude PERFTEST \
        --exclude ${SKIP_PATCH} \
        --noncritical EXPECTED_FAILING \
        --output ${LOG_PATH}/log_test_set_run${nr} \
        tests/

    local local_run_rc=$?
    set -x
    echo ${local_run_rc} > ${LOG_PATH}/rc_test_run${nr}
}

set +x
# Send to background an instance of the run_test_set() function for each number,
# record the pid.
for index in "${!VIRL_SERVER[@]}"; do
    run_test_set ${index} &
    pid=$!
    echo "Sent to background: Test_set${index} (pid=$pid)"
    pids[$pid]=$index
done

echo
echo -n "Waiting..."

# Watch the stable of background processes.
# If a pid goes away, remove it from the array.
while [ -n "${pids[*]}" ]; do
    for i in $(seq 0 9); do
        sleep 1
        echo -n "."
    done
    for pid in "${!pids[@]}"; do
        if ! ps "$pid" >/dev/null; then
            echo -e "\n"
            echo "Test_set${pids[$pid]} with PID $pid finished."
            unset pids[$pid]
        fi
    done
    if [ -z "${!pids[*]}" ]; then
        break
    fi
    echo -n -e "\nStill waiting for test set(s): ${pids[*]} ..."
done

echo
echo "All test set runs finished."
echo

set -x

RC=0
for index in "${!VIRL_SERVER[@]}"; do
    echo "Test_set${index} log:"
    cat ${LOG_PATH}/test_run${index}.log
    RC_PARTIAL_RUN=$(cat ${LOG_PATH}/rc_test_run${index})
    if [ -z "$RC_PARTIAL_RUN" ]; then
        echo "Failed to retrieve return code from test run ${index}"
        exit 1
    fi
    RC=$((RC+RC_PARTIAL_RUN))
    rm -f ${LOG_PATH}/rc_test_run${index}
    rm -f ${LOG_PATH}/test_run${index}.log
    echo
done

# Log the final result
if [ "${RC}" -eq "0" ]; then
    set +x
    echo
    echo "========================================================================================================================================"
    echo "Final result of all test loops:                                                                                                 | PASS |"
    echo "All critical tests have passed."
    echo "========================================================================================================================================"
    echo
    set -x
else
    if [ "${RC}" -eq "1" ]; then
        HLP_STR="test has"
    else
        HLP_STR="tests have"
    fi
    set +x
    echo
    echo "========================================================================================================================================"
    echo "Final result of all test loops:                                                                                                 | FAIL |"
    echo "${RC} critical ${HLP_STR} failed."
    echo "========================================================================================================================================"
    echo
    set -x
fi

echo Post-processing test data...

partial_logs=""
for index in "${!VIRL_SERVER[@]}"; do
    partial_logs="${partial_logs} ${LOG_PATH}/log_test_set_run${index}.xml"
done

# Rebot output post-processing
rebot --noncritical EXPECTED_FAILING \
      --output output.xml ${partial_logs}

# Remove unnecessary log files
rm -f ${partial_logs}

# Archive JOB artifacts in jenkins
for i in ${JOB_ARCHIVE_ARTIFACTS[@]}; do
    cp $( readlink -f ${i} | tr '\n' ' ' ) ${JOB_ARCHIVE_DIR}/
done
# Archive JOB artifacts to logs.fd.io
for i in ${LOG_ARCHIVE_ARTIFACTS[@]}; do
    cp $( readlink -f ${i} | tr '\n' ' ' ) ${LOG_ARCHIVE_DIR}/
done

echo Post-processing finished.

if [ ${RC} -eq 0 ]; then
    RETURN_STATUS=0
else
    RETURN_STATUS=1
fi

exit ${RETURN_STATUS}
s="p">; } void session_free (session_t * s) { pool_put (session_manager_main.wrk[s->thread_index].sessions, s); if (CLIB_DEBUG) clib_memset (s, 0xFA, sizeof (*s)); } void session_free_w_fifos (session_t * s) { segment_manager_dealloc_fifos (s->svm_segment_index, s->rx_fifo, s->tx_fifo); session_free (s); } /** * Cleans up session and lookup table. * * Transport connection must still be valid. */ static void session_delete (session_t * s) { int rv; /* Delete from the main lookup table. */ if ((rv = session_lookup_del_session (s))) clib_warning ("hash delete error, rv %d", rv); session_free_w_fifos (s); } static session_t * session_alloc_for_connection (transport_connection_t * tc) { session_t *s; u32 thread_index = tc->thread_index; ASSERT (thread_index == vlib_get_thread_index () || transport_protocol_is_cl (tc->proto)); s = session_alloc (thread_index); s->session_type = session_type_from_proto_and_ip (tc->proto, tc->is_ip4); s->enqueue_epoch = (u64) ~ 0; s->session_state = SESSION_STATE_CLOSED; /* Attach transport to session and vice versa */ s->connection_index = tc->c_index; tc->s_index = s->session_index; return s; } /** * Discards bytes from buffer chain * * It discards n_bytes_to_drop starting at first buffer after chain_b */ always_inline void session_enqueue_discard_chain_bytes (vlib_main_t * vm, vlib_buffer_t * b, vlib_buffer_t ** chain_b, u32 n_bytes_to_drop) { vlib_buffer_t *next = *chain_b; u32 to_drop = n_bytes_to_drop; ASSERT (b->flags & VLIB_BUFFER_NEXT_PRESENT); while (to_drop && (next->flags & VLIB_BUFFER_NEXT_PRESENT)) { next = vlib_get_buffer (vm, next->next_buffer); if (next->current_length > to_drop) { vlib_buffer_advance (next, to_drop); to_drop = 0; } else { to_drop -= next->current_length; next->current_length = 0; } } *chain_b = next; if (to_drop == 0) b->total_length_not_including_first_buffer -= n_bytes_to_drop; } /** * Enqueue buffer chain tail */ always_inline int session_enqueue_chain_tail (session_t * s, vlib_buffer_t * b, u32 offset, u8 is_in_order) { vlib_buffer_t *chain_b; u32 chain_bi, len, diff; vlib_main_t *vm = vlib_get_main (); u8 *data; u32 written = 0; int rv = 0; if (is_in_order && offset) { diff = offset - b->current_length; if (diff > b->total_length_not_including_first_buffer) return 0; chain_b = b; session_enqueue_discard_chain_bytes (vm, b, &chain_b, diff); chain_bi = vlib_get_buffer_index (vm, chain_b); } else chain_bi = b->next_buffer; do { chain_b = vlib_get_buffer (vm, chain_bi); data = vlib_buffer_get_current (chain_b); len = chain_b->current_length; if (!len) continue; if (is_in_order) { rv = svm_fifo_enqueue_nowait (s->rx_fifo, len, data); if (rv == len) { written += rv; } else if (rv < len) { return (rv > 0) ? (written + rv) : written; } else if (rv > len) { written += rv; /* written more than what was left in chain */ if (written > b->total_length_not_including_first_buffer) return written; /* drop the bytes that have already been delivered */ session_enqueue_discard_chain_bytes (vm, b, &chain_b, rv - len); } } else { rv = svm_fifo_enqueue_with_offset (s->rx_fifo, offset, len, data); if (rv) { clib_warning ("failed to enqueue multi-buffer seg"); return -1; } offset += len; } } while ((chain_bi = (chain_b->flags & VLIB_BUFFER_NEXT_PRESENT) ? chain_b->next_buffer : 0)); if (is_in_order) return written; return 0; } /* * Enqueue data for delivery to session peer. Does not notify peer of enqueue * event but on request can queue notification events for later delivery by * calling stream_server_flush_enqueue_events(). * * @param tc Transport connection which is to be enqueued data * @param b Buffer to be enqueued * @param offset Offset at which to start enqueueing if out-of-order * @param queue_event Flag to indicate if peer is to be notified or if event * is to be queued. The former is useful when more data is * enqueued and only one event is to be generated. * @param is_in_order Flag to indicate if data is in order * @return Number of bytes enqueued or a negative value if enqueueing failed. */ int session_enqueue_stream_connection (transport_connection_t * tc, vlib_buffer_t * b, u32 offset, u8 queue_event, u8 is_in_order) { session_t *s; int enqueued = 0, rv, in_order_off; s = session_get (tc->s_index, tc->thread_index); if (is_in_order) { enqueued = svm_fifo_enqueue_nowait (s->rx_fifo, b->current_length, vlib_buffer_get_current (b)); if (PREDICT_FALSE ((b->flags & VLIB_BUFFER_NEXT_PRESENT) && enqueued >= 0)) { in_order_off = enqueued > b->current_length ? enqueued : 0; rv = session_enqueue_chain_tail (s, b, in_order_off, 1); if (rv > 0) enqueued += rv; } } else { rv = svm_fifo_enqueue_with_offset (s->rx_fifo, offset, b->current_length, vlib_buffer_get_current (b)); if (PREDICT_FALSE ((b->flags & VLIB_BUFFER_NEXT_PRESENT) && !rv)) session_enqueue_chain_tail (s, b, offset + b->current_length, 0); /* if something was enqueued, report even this as success for ooo * segment handling */ return rv; } if (queue_event) { /* Queue RX event on this fifo. Eventually these will need to be flushed * by calling stream_server_flush_enqueue_events () */ session_manager_worker_t *wrk; wrk = session_manager_get_worker (s->thread_index); if (s->enqueue_epoch != wrk->current_enqueue_epoch[tc->proto]) { s->enqueue_epoch = wrk->current_enqueue_epoch[tc->proto]; vec_add1 (wrk->session_to_enqueue[tc->proto], s->session_index); } } return enqueued; } int session_enqueue_dgram_connection (session_t * s, session_dgram_hdr_t * hdr, vlib_buffer_t * b, u8 proto, u8 queue_event) { int enqueued = 0, rv, in_order_off; ASSERT (svm_fifo_max_enqueue (s->rx_fifo) >= b->current_length + sizeof (*hdr)); svm_fifo_enqueue_nowait (s->rx_fifo, sizeof (session_dgram_hdr_t), (u8 *) hdr); enqueued = svm_fifo_enqueue_nowait (s->rx_fifo, b->current_length, vlib_buffer_get_current (b)); if (PREDICT_FALSE ((b->flags & VLIB_BUFFER_NEXT_PRESENT) && enqueued >= 0)) { in_order_off = enqueued > b->current_length ? enqueued : 0; rv = session_enqueue_chain_tail (s, b, in_order_off, 1); if (rv > 0) enqueued += rv; } if (queue_event) { /* Queue RX event on this fifo. Eventually these will need to be flushed * by calling stream_server_flush_enqueue_events () */ session_manager_worker_t *wrk; wrk = session_manager_get_worker (s->thread_index); if (s->enqueue_epoch != wrk->current_enqueue_epoch[proto]) { s->enqueue_epoch = wrk->current_enqueue_epoch[proto]; vec_add1 (wrk->session_to_enqueue[proto], s->session_index); } } return enqueued; } /** Check if we have space in rx fifo to push more bytes */ u8 stream_session_no_space (transport_connection_t * tc, u32 thread_index, u16 data_len) { session_t *s = session_get (tc->s_index, thread_index); if (PREDICT_FALSE (s->session_state != SESSION_STATE_READY)) return 1; if (data_len > svm_fifo_max_enqueue (s->rx_fifo)) return 1; return 0; } u32 session_tx_fifo_max_dequeue (transport_connection_t * tc) { session_t *s = session_get (tc->s_index, tc->thread_index); if (!s->tx_fifo) return 0; return svm_fifo_max_dequeue (s->tx_fifo); } int stream_session_peek_bytes (transport_connection_t * tc, u8 * buffer, u32 offset, u32 max_bytes) { session_t *s = session_get (tc->s_index, tc->thread_index); return svm_fifo_peek (s->tx_fifo, offset, max_bytes, buffer); } u32 session_dequeue_drop (transport_connection_t * tc, u32 max_bytes) { session_t *s = session_get (tc->s_index, tc->thread_index); return svm_fifo_dequeue_drop (s->tx_fifo, max_bytes); } static inline int session_notify_subscribers (u32 app_index, session_t * s, svm_fifo_t * f, session_evt_type_t evt_type) { app_worker_t *app_wrk; application_t *app; int i; app = application_get (app_index); if (!app) return -1; for (i = 0; i < f->n_subscribers; i++) { app_wrk = application_get_worker (app, f->subscribers[i]); if (!app_wrk) continue; if (app_worker_lock_and_send_event (app_wrk, s, evt_type)) return -1; } return 0; } /** * Notify session peer that new data has been enqueued. * * @param s Stream session for which the event is to be generated. * @param lock Flag to indicate if call should lock message queue. * * @return 0 on success or negative number if failed to send notification. */ static inline int session_enqueue_notify (session_t * s) { app_worker_t *app_wrk; app_wrk = app_worker_get_if_valid (s->app_wrk_index); if (PREDICT_FALSE (!app_wrk)) { SESSION_DBG ("invalid s->app_index = %d", s->app_wrk_index); return 0; } /* *INDENT-OFF* */ SESSION_EVT_DBG(SESSION_EVT_ENQ, s, ({ ed->data[0] = FIFO_EVENT_APP_RX; ed->data[1] = svm_fifo_max_dequeue (s->rx_fifo); })); /* *INDENT-ON* */ if (PREDICT_FALSE (app_worker_lock_and_send_event (app_wrk, s, FIFO_EVENT_APP_RX))) return -1; if (PREDICT_FALSE (svm_fifo_n_subscribers (s->rx_fifo))) return session_notify_subscribers (app_wrk->app_index, s, s->rx_fifo, FIFO_EVENT_APP_RX); return 0; } int session_dequeue_notify (session_t * s) { app_worker_t *app_wrk; app_wrk = app_worker_get_if_valid (s->app_wrk_index); if (PREDICT_FALSE (!app_wrk)) return -1; if (PREDICT_FALSE (app_worker_lock_and_send_event (app_wrk, s, FIFO_EVENT_APP_TX))) return -1; if (PREDICT_FALSE (s->tx_fifo->n_subscribers)) return session_notify_subscribers (app_wrk->app_index, s, s->tx_fifo, FIFO_EVENT_APP_TX); svm_fifo_clear_tx_ntf (s->tx_fifo); return 0; } /** * Flushes queue of sessions that are to be notified of new data * enqueued events. * * @param thread_index Thread index for which the flush is to be performed. * @return 0 on success or a positive number indicating the number of * failures due to API queue being full. */ int session_manager_flush_enqueue_events (u8 transport_proto, u32 thread_index) { session_manager_worker_t *wrk = session_manager_get_worker (thread_index); session_t *s; int i, errors = 0; u32 *indices; indices = wrk->session_to_enqueue[transport_proto]; for (i = 0; i < vec_len (indices); i++) { s = session_get_if_valid (indices[i], thread_index); if (PREDICT_FALSE (!s)) { errors++; continue; } if (PREDICT_FALSE (session_enqueue_notify (s))) errors++; } vec_reset_length (indices); wrk->session_to_enqueue[transport_proto] = indices; wrk->current_enqueue_epoch[transport_proto]++; return errors; } int session_manager_flush_all_enqueue_events (u8 transport_proto) { vlib_thread_main_t *vtm = vlib_get_thread_main (); int i, errors = 0; for (i = 0; i < 1 + vtm->n_threads; i++) errors += session_manager_flush_enqueue_events (transport_proto, i); return errors; } /** * Init fifo tail and head pointers * * Useful if transport uses absolute offsets for tracking ooo segments. */ void stream_session_init_fifos_pointers (transport_connection_t * tc, u32 rx_pointer, u32 tx_pointer) { session_t *s; s = session_get (tc->s_index, tc->thread_index); svm_fifo_init_pointers (s->rx_fifo, rx_pointer); svm_fifo_init_pointers (s->tx_fifo, tx_pointer); } int session_stream_connect_notify (transport_connection_t * tc, u8 is_fail) { u32 opaque = 0, new_ti, new_si; app_worker_t *app_wrk; session_t *s = 0; u64 ho_handle; /* * Find connection handle and cleanup half-open table */ ho_handle = session_lookup_half_open_handle (tc); if (ho_handle == HALF_OPEN_LOOKUP_INVALID_VALUE) { SESSION_DBG ("half-open was removed!"); return -1; } session_lookup_del_half_open (tc); /* Get the app's index from the handle we stored when opening connection * and the opaque (api_context for external apps) from transport session * index */ app_wrk = app_worker_get_if_valid (ho_handle >> 32); if (!app_wrk) return -1; opaque = tc->s_index; if (is_fail) return app_worker_connect_notify (app_wrk, s, opaque); s = session_alloc_for_connection (tc); s->session_state = SESSION_STATE_CONNECTING; s->app_wrk_index = app_wrk->wrk_index; new_si = s->session_index; new_ti = s->thread_index; if (app_worker_init_connected (app_wrk, s)) { session_free (s); app_worker_connect_notify (app_wrk, 0, opaque); return -1; } if (app_worker_connect_notify (app_wrk, s, opaque)) { s = session_get (new_si, new_ti); session_free_w_fifos (s); return -1; } s = session_get (new_si, new_ti); s->session_state = SESSION_STATE_READY; session_lookup_add_connection (tc, session_handle (s)); return 0; } typedef struct _session_switch_pool_args { u32 session_index; u32 thread_index; u32 new_thread_index; u32 new_session_index; } session_switch_pool_args_t; static void session_switch_pool (void *cb_args) { session_switch_pool_args_t *args = (session_switch_pool_args_t *) cb_args; session_t *s; ASSERT (args->thread_index == vlib_get_thread_index ()); s = session_get (args->session_index, args->thread_index); s->tx_fifo->master_session_index = args->new_session_index; s->tx_fifo->master_thread_index = args->new_thread_index; transport_cleanup (session_get_transport_proto (s), s->connection_index, s->thread_index); session_free (s); clib_mem_free (cb_args); } /** * Move dgram session to the right thread */ int session_dgram_connect_notify (transport_connection_t * tc, u32 old_thread_index, session_t ** new_session) { session_t *new_s; session_switch_pool_args_t *rpc_args; /* * Clone half-open session to the right thread. */ new_s = session_clone_safe (tc->s_index, old_thread_index); new_s->connection_index = tc->c_index; new_s->rx_fifo->master_session_index = new_s->session_index; new_s->rx_fifo->master_thread_index = new_s->thread_index; new_s->session_state = SESSION_STATE_READY; session_lookup_add_connection (tc, session_handle (new_s)); /* * Ask thread owning the old session to clean it up and make us the tx * fifo owner */ rpc_args = clib_mem_alloc (sizeof (*rpc_args)); rpc_args->new_session_index = new_s->session_index; rpc_args->new_thread_index = new_s->thread_index; rpc_args->session_index = tc->s_index; rpc_args->thread_index = old_thread_index; session_send_rpc_evt_to_thread (rpc_args->thread_index, session_switch_pool, rpc_args); tc->s_index = new_s->session_index; new_s->connection_index = tc->c_index; *new_session = new_s; return 0; } /** * Notification from transport that connection is being closed. * * A disconnect is sent to application but state is not removed. Once * disconnect is acknowledged by application, session disconnect is called. * Ultimately this leads to close being called on transport (passive close). */ void session_transport_closing_notify (transport_connection_t * tc) { app_worker_t *app_wrk; application_t *app; session_t *s; s = session_get (tc->s_index, tc->thread_index); if (s->session_state >= SESSION_STATE_TRANSPORT_CLOSING) return; s->session_state = SESSION_STATE_TRANSPORT_CLOSING; app_wrk = app_worker_get_if_valid (s->app_wrk_index); if (!app_wrk) return; app = application_get (app_wrk->app_index); app->cb_fns.session_disconnect_callback (s); } /** * Notification from transport that connection is being deleted * * This removes the session if it is still valid. It should be called only on * previously fully established sessions. For instance failed connects should * call stream_session_connect_notify and indicate that the connect has * failed. */ void session_transport_delete_notify (transport_connection_t * tc) { session_t *s; /* App might've been removed already */ if (!(s = session_get_if_valid (tc->s_index, tc->thread_index))) return; /* Make sure we don't try to send anything more */ svm_fifo_dequeue_drop_all (s->tx_fifo); switch (s->session_state) { case SESSION_STATE_CREATED: /* Session was created but accept notification was not yet sent to the * app. Cleanup everything. */ session_lookup_del_session (s); session_free_w_fifos (s); break; case SESSION_STATE_ACCEPTING: case SESSION_STATE_TRANSPORT_CLOSING: /* If transport finishes or times out before we get a reply * from the app, mark transport as closed and wait for reply * before removing the session. Cleanup session table in advance * because transport will soon be closed and closed sessions * are assumed to have been removed from the lookup table */ session_lookup_del_session (s); s->session_state = SESSION_STATE_TRANSPORT_CLOSED; break; case SESSION_STATE_CLOSING: case SESSION_STATE_CLOSED_WAITING: /* Cleanup lookup table as transport needs to still be valid. * Program transport close to ensure that all session events * have been cleaned up. Once transport close is called, the * session is just removed because both transport and app have * confirmed the close*/ session_lookup_del_session (s); s->session_state = SESSION_STATE_TRANSPORT_CLOSED; session_program_transport_close (s); break; case SESSION_STATE_TRANSPORT_CLOSED: break; case SESSION_STATE_CLOSED: session_delete (s); break; default: clib_warning ("session state %u", s->session_state); session_delete (s); break; } } /** * Notification from transport that session can be closed * * Should be called by transport only if it was closed with non-empty * tx fifo and once it decides to begin the closing procedure prior to * issuing a delete notify. This gives the chance to the session layer * to cleanup any outstanding events. */ void session_transport_closed_notify (transport_connection_t * tc) { session_t *s; if (!(s = session_get_if_valid (tc->s_index, tc->thread_index))) return; /* If app close has not been received or has not yet resulted in * a transport close, only mark the session transport as closed */ if (s->session_state <= SESSION_STATE_CLOSING) { session_lookup_del_session (s); s->session_state = SESSION_STATE_TRANSPORT_CLOSED; } else s->session_state = SESSION_STATE_CLOSED; } /** * Notify application that connection has been reset. */ void session_transport_reset_notify (transport_connection_t * tc) { session_t *s; app_worker_t *app_wrk; application_t *app; s = session_get (tc->s_index, tc->thread_index); svm_fifo_dequeue_drop_all (s->tx_fifo); if (s->session_state >= SESSION_STATE_TRANSPORT_CLOSING) return; s->session_state = SESSION_STATE_TRANSPORT_CLOSING; app_wrk = app_worker_get (s->app_wrk_index); app = application_get (app_wrk->app_index); app->cb_fns.session_reset_callback (s); } int session_stream_accept_notify (transport_connection_t * tc) { app_worker_t *app_wrk; session_t *s; s = session_get (tc->s_index, tc->thread_index); app_wrk = app_worker_get_if_valid (s->app_wrk_index); if (!app_wrk) return -1; s->session_state = SESSION_STATE_ACCEPTING; return app_worker_accept_notify (app_wrk, s); } /** * Accept a stream session. Optionally ping the server by callback. */ int session_stream_accept (transport_connection_t * tc, u32 listener_index, u8 notify) { session_t *s; int rv; s = session_alloc_for_connection (tc); s->listener_index = listener_index; s->session_state = SESSION_STATE_CREATED; if ((rv = app_worker_init_accepted (s))) return rv; session_lookup_add_connection (tc, session_handle (s)); /* Shoulder-tap the server */ if (notify) { app_worker_t *app_wrk = app_worker_get (s->app_wrk_index); return app_worker_accept_notify (app_wrk, s); } return 0; } int session_open_cl (u32 app_wrk_index, session_endpoint_t * rmt, u32 opaque) { transport_connection_t *tc; transport_endpoint_cfg_t *tep; app_worker_t *app_wrk; session_t *s; int rv; tep = session_endpoint_to_transport_cfg (rmt); rv = transport_connect (rmt->transport_proto, tep); if (rv < 0) { SESSION_DBG ("Transport failed to open connection."); return VNET_API_ERROR_SESSION_CONNECT; } tc = transport_get_half_open (rmt->transport_proto, (u32) rv); /* For dgram type of service, allocate session and fifos now */ app_wrk = app_worker_get (app_wrk_index); s = session_alloc_for_connection (tc); s->app_wrk_index = app_wrk->wrk_index; s->session_state = SESSION_STATE_OPENED; if (app_worker_init_connected (app_wrk, s)) { session_free (s); return -1; } return app_worker_connect_notify (app_wrk, s, opaque); } int session_open_vc (u32 app_wrk_index, session_endpoint_t * rmt, u32 opaque) { transport_connection_t *tc; transport_endpoint_cfg_t *tep; u64 handle; int rv; tep = session_endpoint_to_transport_cfg (rmt); rv = transport_connect (rmt->transport_proto, tep); if (rv < 0) { SESSION_DBG ("Transport failed to open connection."); return VNET_API_ERROR_SESSION_CONNECT; } tc = transport_get_half_open (rmt->transport_proto, (u32) rv); /* If transport offers a stream service, only allocate session once the * connection has been established. * Add connection to half-open table and save app and tc index. The * latter is needed to help establish the connection while the former * is needed when the connect notify comes and we have to notify the * external app */ handle = (((u64) app_wrk_index) << 32) | (u64) tc->c_index; session_lookup_add_half_open (tc, handle); /* Store api_context (opaque) for when the reply comes. Not the nicest * thing but better than allocating a separate half-open pool. */ tc->s_index = opaque; return 0; } int session_open_app (u32 app_wrk_index, session_endpoint_t * rmt, u32 opaque) { session_endpoint_cfg_t *sep = (session_endpoint_cfg_t *) rmt; transport_endpoint_cfg_t *tep_cfg = session_endpoint_to_transport_cfg (sep); sep->app_wrk_index = app_wrk_index; sep->opaque = opaque; return transport_connect (rmt->transport_proto, tep_cfg); } typedef int (*session_open_service_fn) (u32, session_endpoint_t *, u32); /* *INDENT-OFF* */ static session_open_service_fn session_open_srv_fns[TRANSPORT_N_SERVICES] = { session_open_vc, session_open_cl, session_open_app, }; /* *INDENT-ON* */ /** * Ask transport to open connection to remote transport endpoint. * * Stores handle for matching request with reply since the call can be * asynchronous. For instance, for TCP the 3-way handshake must complete * before reply comes. Session is only created once connection is established. * * @param app_index Index of the application requesting the connect * @param st Session type requested. * @param tep Remote transport endpoint * @param opaque Opaque data (typically, api_context) the application expects * on open completion. */ int session_open (u32 app_wrk_index, session_endpoint_t * rmt, u32 opaque) { transport_service_type_t tst; tst = transport_protocol_service_type (rmt->transport_proto); return session_open_srv_fns[tst] (app_wrk_index, rmt, opaque); } /** * Ask transport to listen on session endpoint. * * @param s Session for which listen will be called. Note that unlike * established sessions, listen sessions are not associated to a * thread. * @param sep Local endpoint to be listened on. */ int session_listen (session_t * ls, session_endpoint_cfg_t * sep) { transport_endpoint_t *tep; u32 tc_index, s_index; /* Transport bind/listen */ tep = session_endpoint_to_transport (sep); s_index = ls->session_index; tc_index = transport_start_listen (session_get_transport_proto (ls), s_index, tep); if (tc_index == (u32) ~ 0) return -1; /* Attach transport to session. Lookup tables are populated by the app * worker because local tables (for ct sessions) are not backed by a fib */ ls = listen_session_get (s_index); ls->connection_index = tc_index; return 0; } /** * Ask transport to stop listening on local transport endpoint. * * @param s Session to stop listening on. It must be in state LISTENING. */ int session_stop_listen (session_t * s) { transport_proto_t tp = session_get_transport_proto (s); transport_connection_t *tc; if (s->session_state != SESSION_STATE_LISTENING) return -1; tc = transport_get_listener (tp, s->connection_index); if (!tc) return VNET_API_ERROR_ADDRESS_NOT_IN_USE; session_lookup_del_connection (tc); transport_stop_listen (tp, s->connection_index); return 0; } /** * Initialize session closing procedure. * * Request is always sent to session node to ensure that all outstanding * requests are served before transport is notified. */ void session_close (session_t * s) { if (!s) return; if (s->session_state >= SESSION_STATE_CLOSING) { /* Session will only be removed once both app and transport * acknowledge the close */ if (s->session_state == SESSION_STATE_TRANSPORT_CLOSED) session_program_transport_close (s); /* Session already closed. Clear the tx fifo */ if (s->session_state == SESSION_STATE_CLOSED) svm_fifo_dequeue_drop_all (s->tx_fifo); return; } s->session_state = SESSION_STATE_CLOSING; session_program_transport_close (s); } /** * Notify transport the session can be disconnected. This should eventually * result in a delete notification that allows us to cleanup session state. * Called for both active/passive disconnects. * * Must be called from the session's thread. */ void session_transport_close (session_t * s) { /* If transport is already closed, just free the session */ if (s->session_state >= SESSION_STATE_TRANSPORT_CLOSED) { session_free_w_fifos (s); return; } /* If tx queue wasn't drained, change state to closed waiting for transport. * This way, the transport, if it so wishes, can continue to try sending the * outstanding data (in closed state it cannot). It MUST however at one * point, either after sending everything or after a timeout, call delete * notify. This will finally lead to the complete cleanup of the session. */ if (svm_fifo_max_dequeue (s->tx_fifo)) s->session_state = SESSION_STATE_CLOSED_WAITING; else s->session_state = SESSION_STATE_CLOSED; transport_close (session_get_transport_proto (s), s->connection_index, s->thread_index); } /** * Cleanup transport and session state. * * Notify transport of the cleanup and free the session. This should * be called only if transport reported some error and is already * closed. */ void session_transport_cleanup (session_t * s) { s->session_state = SESSION_STATE_CLOSED; /* Delete from main lookup table before we axe the the transport */ session_lookup_del_session (s); transport_cleanup (session_get_transport_proto (s), s->connection_index, s->thread_index); /* Since we called cleanup, no delete notification will come. So, make * sure the session is properly freed. */ session_free_w_fifos (s); } /** * Allocate event queues in the shared-memory segment * * That can either be a newly created memfd segment, that will need to be * mapped by all stack users, or the binary api's svm region. The latter is * assumed to be already mapped. NOTE that this assumption DOES NOT hold if * api clients bootstrap shm api over sockets (i.e. use memfd segments) and * vpp uses api svm region for event queues. */ void session_vpp_event_queues_allocate (session_manager_main_t * smm) { u32 evt_q_length = 2048, evt_size = sizeof (session_event_t); ssvm_private_t *eqs = &smm->evt_qs_segment; api_main_t *am = &api_main; uword eqs_size = 64 << 20; pid_t vpp_pid = getpid (); void *oldheap; int i; if (smm->configured_event_queue_length) evt_q_length = smm->configured_event_queue_length; if (smm->evt_qs_use_memfd_seg) { if (smm->evt_qs_segment_size) eqs_size = smm->evt_qs_segment_size; eqs->ssvm_size = eqs_size; eqs->i_am_master = 1; eqs->my_pid = vpp_pid; eqs->name = format (0, "%s%c", "evt-qs-segment", 0); eqs->requested_va = smm->session_baseva; if (ssvm_master_init (eqs, SSVM_SEGMENT_MEMFD)) { clib_warning ("failed to initialize queue segment"); return; } } if (smm->evt_qs_use_memfd_seg) oldheap = ssvm_push_heap (eqs->sh); else oldheap = svm_push_data_heap (am->vlib_rp); for (i = 0; i < vec_len (smm->wrk); i++) { svm_msg_q_cfg_t _cfg, *cfg = &_cfg; svm_msg_q_ring_cfg_t rc[SESSION_MQ_N_RINGS] = { {evt_q_length, evt_size, 0} , {evt_q_length >> 1, 256, 0} }; cfg->consumer_pid = 0; cfg->n_rings = 2; cfg->q_nitems = evt_q_length; cfg->ring_cfgs = rc; smm->wrk[i].vpp_event_queue = svm_msg_q_alloc (cfg); if (smm->evt_qs_use_memfd_seg) { if (svm_msg_q_alloc_consumer_eventfd (smm->wrk[i].vpp_event_queue)) clib_warning ("eventfd returned"); } } if (smm->evt_qs_use_memfd_seg) ssvm_pop_heap (oldheap); else svm_pop_heap (oldheap); } ssvm_private_t * session_manager_get_evt_q_segment (void) { session_manager_main_t *smm = &session_manager_main; if (smm->evt_qs_use_memfd_seg) return &smm->evt_qs_segment; return 0; } /* *INDENT-OFF* */ static session_fifo_rx_fn *session_tx_fns[TRANSPORT_TX_N_FNS] = { session_tx_fifo_peek_and_snd, session_tx_fifo_dequeue_and_snd, session_tx_fifo_dequeue_internal, session_tx_fifo_dequeue_and_snd }; /* *INDENT-ON* */ /** * Initialize session layer for given transport proto and ip version * * Allocates per session type (transport proto + ip version) data structures * and adds arc from session queue node to session type output node. */ void session_register_transport (transport_proto_t transport_proto, const transport_proto_vft_t * vft, u8 is_ip4, u32 output_node) { session_manager_main_t *smm = &session_manager_main; session_type_t session_type; u32 next_index = ~0; session_type = session_type_from_proto_and_ip (transport_proto, is_ip4); vec_validate (smm->session_type_to_next, session_type); vec_validate (smm->session_tx_fns, session_type); /* *INDENT-OFF* */ if (output_node != ~0) { foreach_vlib_main (({ next_index = vlib_node_add_next (this_vlib_main, session_queue_node.index, output_node); })); } /* *INDENT-ON* */ smm->session_type_to_next[session_type] = next_index; smm->session_tx_fns[session_type] = session_tx_fns[vft->tx_type]; } transport_connection_t * session_get_transport (session_t * s) { if (s->session_state != SESSION_STATE_LISTENING) return transport_get_connection (session_get_transport_proto (s), s->connection_index, s->thread_index); else return transport_get_listener (session_get_transport_proto (s), s->connection_index); } transport_connection_t * listen_session_get_transport (session_t * s) { return transport_get_listener (session_get_transport_proto (s), s->connection_index); } int listen_session_get_local_session_endpoint (session_t * listener, session_endpoint_t * sep) { transport_connection_t *tc; tc = listen_session_get_transport (listener); if (!tc) { clib_warning ("no transport"); return -1; } /* N.B. The ip should not be copied because this is the local endpoint */ sep->port = tc->lcl_port; sep->transport_proto = tc->proto; sep->is_ip4 = tc->is_ip4; return 0; } void session_flush_frames_main_thread (vlib_main_t * vm) { ASSERT (vlib_get_thread_index () == 0); vlib_process_signal_event_mt (vm, session_queue_process_node.index, SESSION_Q_PROCESS_FLUSH_FRAMES, 0); } static clib_error_t * session_manager_main_enable (vlib_main_t * vm) { segment_manager_main_init_args_t _sm_args = { 0 }, *sm_args = &_sm_args; session_manager_main_t *smm = &session_manager_main; vlib_thread_main_t *vtm = vlib_get_thread_main (); u32 num_threads, preallocated_sessions_per_worker; session_manager_worker_t *wrk; int i, j; num_threads = 1 /* main thread */ + vtm->n_threads; if (num_threads < 1) return clib_error_return (0, "n_thread_stacks not set"); /* Allocate cache line aligned worker contexts */ vec_validate_aligned (smm->wrk, num_threads - 1, CLIB_CACHE_LINE_BYTES); for (i = 0; i < TRANSPORT_N_PROTO; i++) { for (j = 0; j < num_threads; j++) smm->wrk[j].current_enqueue_epoch[i] = 1; } for (i = 0; i < num_threads; i++) { wrk = &smm->wrk[i]; vec_validate (wrk->free_event_vector, 128); _vec_len (wrk->free_event_vector) = 0; vec_validate (wrk->pending_event_vector, 128); _vec_len (wrk->pending_event_vector) = 0; vec_validate (wrk->pending_disconnects, 128); _vec_len (wrk->pending_disconnects) = 0; vec_validate (wrk->postponed_event_vector, 128); _vec_len (wrk->postponed_event_vector) = 0; wrk->last_vlib_time = vlib_time_now (vlib_mains[i]); wrk->dispatch_period = 500e-6; if (num_threads > 1) clib_rwlock_init (&smm->wrk[i].peekers_rw_locks); } #if SESSION_DEBUG vec_validate (smm->last_event_poll_by_thread, num_threads - 1); #endif /* Allocate vpp event queues segment and queue */ session_vpp_event_queues_allocate (smm); /* Initialize fifo segment main baseva and timeout */ sm_args->baseva = smm->session_baseva + smm->evt_qs_segment_size; sm_args->size = smm->session_va_space_size; segment_manager_main_init (sm_args); /* Preallocate sessions */ if (smm->preallocated_sessions) { if (num_threads == 1) { pool_init_fixed (smm->wrk[0].sessions, smm->preallocated_sessions); } else { int j; preallocated_sessions_per_worker = (1.1 * (f64) smm->preallocated_sessions / (f64) (num_threads - 1)); for (j = 1; j < num_threads; j++) { pool_init_fixed (smm->wrk[j].sessions, preallocated_sessions_per_worker); } } } session_lookup_init (); app_namespaces_init (); transport_init (); smm->is_enabled = 1; /* Enable transports */ transport_enable_disable (vm, 1); transport_init_tx_pacers_period (); return 0; } void session_node_enable_disable (u8 is_en) { u8 state = is_en ? VLIB_NODE_STATE_POLLING : VLIB_NODE_STATE_DISABLED; vlib_thread_main_t *vtm = vlib_get_thread_main (); u8 have_workers = vtm->n_threads != 0; /* *INDENT-OFF* */ foreach_vlib_main (({ if (have_workers && ii == 0) { vlib_node_set_state (this_vlib_main, session_queue_process_node.index, state); if (is_en) { vlib_node_t *n = vlib_get_node (this_vlib_main, session_queue_process_node.index); vlib_start_process (this_vlib_main, n->runtime_index); } else { vlib_process_signal_event_mt (this_vlib_main, session_queue_process_node.index, SESSION_Q_PROCESS_STOP, 0); } continue; } vlib_node_set_state (this_vlib_main, session_queue_node.index, state); })); /* *INDENT-ON* */ } clib_error_t * vnet_session_enable_disable (vlib_main_t * vm, u8 is_en) { clib_error_t *error = 0; if (is_en) { if (session_manager_main.is_enabled) return 0; session_node_enable_disable (is_en); error = session_manager_main_enable (vm); } else { session_manager_main.is_enabled = 0; session_node_enable_disable (is_en); } return error; } clib_error_t * session_manager_main_init (vlib_main_t * vm) { session_manager_main_t *smm = &session_manager_main; smm->session_baseva = HIGH_SEGMENT_BASEVA; #if (HIGH_SEGMENT_BASEVA > (4ULL << 30)) smm->session_va_space_size = 128ULL << 30; smm->evt_qs_segment_size = 64 << 20; #else smm->session_va_space_size = 128 << 20; smm->evt_qs_segment_size = 1 << 20; #endif smm->is_enabled = 0; return 0; } VLIB_INIT_FUNCTION (session_manager_main_init); static clib_error_t * session_config_fn (vlib_main_t * vm, unformat_input_t * input) { session_manager_main_t *smm = &session_manager_main; u32 nitems; uword tmp; while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { if (unformat (input, "event-queue-length %d", &nitems)) { if (nitems >= 2048) smm->configured_event_queue_length = nitems; else clib_warning ("event queue length %d too small, ignored", nitems); } else if (unformat (input, "preallocated-sessions %d", &smm->preallocated_sessions)) ; else if (unformat (input, "v4-session-table-buckets %d", &smm->configured_v4_session_table_buckets)) ; else if (unformat (input, "v4-halfopen-table-buckets %d", &smm->configured_v4_halfopen_table_buckets)) ; else if (unformat (input, "v6-session-table-buckets %d", &smm->configured_v6_session_table_buckets)) ; else if (unformat (input, "v6-halfopen-table-buckets %d", &smm->configured_v6_halfopen_table_buckets)) ; else if (unformat (input, "v4-session-table-memory %U", unformat_memory_size, &tmp)) { if (tmp >= 0x100000000) return clib_error_return (0, "memory size %llx (%lld) too large", tmp, tmp); smm->configured_v4_session_table_memory = tmp; } else if (unformat (input, "v4-halfopen-table-memory %U", unformat_memory_size, &tmp)) { if (tmp >= 0x100000000) return clib_error_return (0, "memory size %llx (%lld) too large", tmp, tmp); smm->configured_v4_halfopen_table_memory = tmp; } else if (unformat (input, "v6-session-table-memory %U", unformat_memory_size, &tmp)) { if (tmp >= 0x100000000) return clib_error_return (0, "memory size %llx (%lld) too large", tmp, tmp); smm->configured_v6_session_table_memory = tmp; } else if (unformat (input, "v6-halfopen-table-memory %U", unformat_memory_size, &tmp)) { if (tmp >= 0x100000000) return clib_error_return (0, "memory size %llx (%lld) too large", tmp, tmp); smm->configured_v6_halfopen_table_memory = tmp; } else if (unformat (input, "local-endpoints-table-memory %U", unformat_memory_size, &tmp)) { if (tmp >= 0x100000000) return clib_error_return (0, "memory size %llx (%lld) too large", tmp, tmp); smm->local_endpoints_table_memory = tmp; } else if (unformat (input, "local-endpoints-table-buckets %d", &smm->local_endpoints_table_buckets)) ; else if (unformat (input, "evt_qs_memfd_seg")) smm->evt_qs_use_memfd_seg = 1; else if (unformat (input, "evt_qs_seg_size %U", unformat_memory_size, &smm->evt_qs_segment_size)) ; else return clib_error_return (0, "unknown input `%U'", format_unformat_error, input); } return 0; } VLIB_CONFIG_FUNCTION (session_config_fn, "session"); /* * fd.io coding-style-patch-verification: ON * * Local Variables: * eval: (c-set-style "gnu") * End: */