aboutsummaryrefslogtreecommitdiffstats
path: root/resources/libraries/python/CpuUtils.py
blob: 5d8b8a64a3c2db6eeaa0d9bfda45e0151b3b938d (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
# Copyright (c) 2018 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.

"""CPU utilities library."""

from resources.libraries.python.ssh import SSH

__all__ = ["CpuUtils"]


class CpuUtils(object):
    """CPU utilities"""

    # Number of threads per core.
    NR_OF_THREADS = 2

    @staticmethod
    def __str2int(string):
        """Conversion from string to integer, 0 in case of empty string.

        :param string: Input string.
        :type string: str
        :returns: Integer converted from string, 0 in case of ValueError.
        :rtype: int
        """
        try:
            return int(string)
        except ValueError:
            return 0

    @staticmethod
    def is_smt_enabled(cpu_info):
        """Uses CPU mapping to find out if SMT is enabled or not. If SMT is
        enabled, the L1d,L1i,L2,L3 setting is the same for two processors. These
        two processors are two threads of one core.

        :param cpu_info: CPU info, the output of "lscpu -p".
        :type cpu_info: list
        :returns: True if SMT is enabled, False if SMT is disabled.
        :rtype: bool
        """

        cpu_mems = [item[-4:] for item in cpu_info]
        cpu_mems_len = len(cpu_mems) / CpuUtils.NR_OF_THREADS
        count = 0
        for cpu_mem in cpu_mems[:cpu_mems_len]:
            if cpu_mem in cpu_mems[cpu_mems_len:]:
                count += 1
        return bool(count == cpu_mems_len)

    @staticmethod
    def get_cpu_layout_from_all_nodes(nodes):
        """Retrieve cpu layout from all nodes, assuming all nodes
           are Linux nodes.

        :param nodes: DICT__nodes from Topology.DICT__nodes.
        :type nodes: dict
        :raises RuntimeError: If the ssh command "lscpu -p" fails.
        """
        ssh = SSH()
        for node in nodes.values():
            ssh.connect(node)
            cmd = "lscpu -p"
            ret, stdout, stderr = ssh.exec_command(cmd)
#           parsing of "lscpu -p" output:
#           # CPU,Core,Socket,Node,,L1d,L1i,L2,L3
#           0,0,0,0,,0,0,0,0
#           1,1,0,0,,1,1,1,0
            if ret != 0:
                raise RuntimeError(
                    "Failed to execute ssh command, ret: {} err: {}".format(
                        ret, stderr))
            node['cpuinfo'] = list()
            for line in stdout.split("\n"):
                if len(line) > 0 and line[0] != "#":
                    node['cpuinfo'].append([CpuUtils.__str2int(x) for x in
                                            line.split(",")])

    @staticmethod
    def cpu_node_count(node):
        """Return count of numa nodes.

        :param node: Targeted node.
        :type node: dict
        :returns: Count of numa nodes.
        :rtype: int
        :raises RuntimeError: If node cpuinfo is not available.
        """
        cpu_info = node.get("cpuinfo")
        if cpu_info is not None:
            return node["cpuinfo"][-1][3] + 1
        else:
            raise RuntimeError("Node cpuinfo not available.")

    @staticmethod
    def cpu_list_per_node(node, cpu_node, smt_used=False):
        """Return node related list of CPU numbers.

        :param node: Node dictionary with cpuinfo.
        :param cpu_node: Numa node number.
        :param smt_used: True - we want to use SMT, otherwise false.
        :type node: dict
        :type cpu_node: int
        :type smt_used: bool
        :returns: List of cpu numbers related to numa from argument.
        :rtype: list of int
        :raises RuntimeError: If node cpuinfo is not available
            or if SMT is not enabled.
        """

        cpu_node = int(cpu_node)
        cpu_info = node.get("cpuinfo")
        if cpu_info is None:
            raise RuntimeError("Node cpuinfo not available.")

        smt_enabled = CpuUtils.is_smt_enabled(cpu_info)
        if not smt_enabled and smt_used:
            raise RuntimeError("SMT is not enabled.")

        cpu_list = []
        for cpu in cpu_info:
            if cpu[3] == cpu_node:
                cpu_list.append(cpu[0])

        if not smt_enabled or smt_enabled and smt_used:
            pass

        if smt_enabled and not smt_used:
            cpu_list_len = len(cpu_list)
            cpu_list = cpu_list[:cpu_list_len / CpuUtils.NR_OF_THREADS]

        return cpu_list

    @staticmethod
    def cpu_slice_of_list_per_node(node, cpu_node, skip_cnt=0, cpu_cnt=0,
                                   smt_used=False):
        """Return string of node related list of CPU numbers.

        :param node: Node dictionary with cpuinfo.
        :param cpu_node: Numa node number.
        :param skip_cnt: Skip first "skip_cnt" CPUs.
        :param cpu_cnt: Count of cpus to return, if 0 then return all.
        :param smt_used: True - we want to use SMT, otherwise false.
        :type node: dict
        :type cpu_node: int
        :type skip_cnt: int
        :type cpu_cnt: int
        :type smt_used: bool
        :returns: Cpu numbers related to numa from argument.
        :rtype: list
        :raises RuntimeError: If we require more cpus than available.
        """

        cpu_list = CpuUtils.cpu_list_per_node(node, cpu_node, smt_used)

        cpu_list_len = len(cpu_list)
        if cpu_cnt + skip_cnt > cpu_list_len:
            raise RuntimeError("cpu_cnt + skip_cnt > length(cpu list).")

        if cpu_cnt == 0:
            cpu_cnt = cpu_list_len - skip_cnt

        if smt_used:
            cpu_list_0 = cpu_list[:cpu_list_len / CpuUtils.NR_OF_THREADS]
            cpu_list_1 = cpu_list[cpu_list_len / CpuUtils.NR_OF_THREADS:]
            cpu_list = [cpu for cpu in cpu_list_0[skip_cnt:skip_cnt + cpu_cnt]]
            cpu_list_ex = [cpu for cpu in
                           cpu_list_1[skip_cnt:skip_cnt + cpu_cnt]]
            cpu_list.extend(cpu_list_ex)
        else:
            cpu_list = [cpu for cpu in cpu_list[skip_cnt:skip_cnt + cpu_cnt]]

        return cpu_list

    @staticmethod
    def cpu_list_per_node_str(node, cpu_node, skip_cnt=0, cpu_cnt=0, sep=",",
                              smt_used=False):
        """Return string of node related list of CPU numbers.

        :param node: Node dictionary with cpuinfo.
        :param cpu_node: Numa node number.
        :param skip_cnt: Skip first "skip_cnt" CPUs.
        :param cpu_cnt: Count of cpus to return, if 0 then return all.
        :param sep: Separator, default: 1,2,3,4,....
        :param smt_used: True - we want to use SMT, otherwise false.
        :type node: dict
        :type cpu_node: int
        :type skip_cnt: int
        :type cpu_cnt: int
        :type sep: str
        :type smt_used: bool
        :returns: Cpu numbers related to numa from argument.
        :rtype: str
        """

        cpu_list = CpuUtils.cpu_slice_of_list_per_node(node, cpu_node,
                                                       skip_cnt=skip_cnt,
                                                       cpu_cnt=cpu_cnt,
                                                       smt_used=smt_used)
        return sep.join(str(cpu) for cpu in cpu_list)

    @staticmethod
    def cpu_range_per_node_str(node, cpu_node, skip_cnt=0, cpu_cnt=0, sep="-",
                               smt_used=False):
        """Return string of node related range of CPU numbers, e.g. 0-4.

        :param node: Node dictionary with cpuinfo.
        :param cpu_node: Numa node number.
        :param skip_cnt: Skip first "skip_cnt" CPUs.
        :param cpu_cnt: Count of cpus to return, if 0 then return all.
        :param sep: Separator, default: "-".
        :param smt_used: True - we want to use SMT, otherwise false.
        :type node: dict
        :type cpu_node: int
        :type skip_cnt: int
        :type cpu_cnt: int
        :type sep: str
        :type smt_used: bool
        :returns: String of node related range of CPU numbers.
        :rtype: str
        """

        cpu_list = CpuUtils.cpu_slice_of_list_per_node(node, cpu_node,
                                                       skip_cnt=skip_cnt,
                                                       cpu_cnt=cpu_cnt,
                                                       smt_used=smt_used)
        if smt_used:
            cpu_list_len = len(cpu_list)
            cpu_list_0 = cpu_list[:cpu_list_len / CpuUtils.NR_OF_THREADS]
            cpu_list_1 = cpu_list[cpu_list_len / CpuUtils.NR_OF_THREADS:]
            cpu_range = "{}{}{},{}{}{}".format(cpu_list_0[0], sep,
                                               cpu_list_0[-1],
                                               cpu_list_1[0], sep,
                                               cpu_list_1[-1])
        else:
            cpu_range = "{}{}{}".format(cpu_list[0], sep, cpu_list[-1])

        return cpu_range
> 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634
/*
 * 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.
 */
/*
 * pg_input.c: buffer generator input
 *
 * Copyright (c) 2008 Eliot Dresselhaus
 *
 * Permission is hereby granted, free of charge, to any person obtaining
 * a copy of this software and associated documentation files (the
 * "Software"), to deal in the Software without restriction, including
 * without limitation the rights to use, copy, modify, merge, publish,
 * distribute, sublicense, and/or sell copies of the Software, and to
 * permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 *
 * The above copyright notice and this permission notice shall be
 * included in all copies or substantial portions of the Software.
 *
 *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
 *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
 *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
 *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 */

#include <vlib/vlib.h>
#include <vnet/pg/pg.h>
#include <vnet/vnet.h>
#include <vnet/feature/feature.h>
#include <vnet/devices/devices.h>

static int
validate_buffer_data2 (vlib_buffer_t * b, pg_stream_t * s,
		       u32 data_offset, u32 n_bytes)
{
  u8 *bd, *pd, *pm;
  u32 i;

  bd = b->data;
  pd = s->fixed_packet_data + data_offset;
  pm = s->fixed_packet_data_mask + data_offset;

  if (pd + n_bytes >= vec_end (s->fixed_packet_data))
    n_bytes = (pd < vec_end (s->fixed_packet_data)
	       ? vec_end (s->fixed_packet_data) - pd : 0);

  for (i = 0; i < n_bytes; i++)
    if ((bd[i] & pm[i]) != pd[i])
      break;

  if (i >= n_bytes)
    return 1;

  clib_warning ("buffer %U", format_vnet_buffer, b);
  clib_warning ("differ at index %d", i);
  clib_warning ("is     %U", format_hex_bytes, bd, n_bytes);
  clib_warning ("mask   %U", format_hex_bytes, pm, n_bytes);
  clib_warning ("expect %U", format_hex_bytes, pd, n_bytes);
  return 0;
}

static int
validate_buffer_data (vlib_buffer_t * b, pg_stream_t * s)
{
  return validate_buffer_data2 (b, s, 0, s->buffer_bytes);
}

always_inline void
set_1 (void *a0,
       u64 v0, u64 v_min, u64 v_max, u32 n_bits, u32 is_net_byte_order)
{
  ASSERT (v0 >= v_min && v0 <= v_max);
  if (n_bits == BITS (u8))
    {
      ((u8 *) a0)[0] = v0;
    }
  else if (n_bits == BITS (u16))
    {
      if (is_net_byte_order)
	v0 = clib_host_to_net_u16 (v0);
      clib_mem_unaligned (a0, u16) = v0;
    }
  else if (n_bits == BITS (u32))
    {
      if (is_net_byte_order)
	v0 = clib_host_to_net_u32 (v0);
      clib_mem_unaligned (a0, u32) = v0;
    }
  else if (n_bits == BITS (u64))
    {
      if (is_net_byte_order)
	v0 = clib_host_to_net_u64 (v0);
      clib_mem_unaligned (a0, u64) = v0;
    }
}

always_inline void
set_2 (void *a0, void *a1,
       u64 v0, u64 v1,
       u64 v_min, u64 v_max,
       u32 n_bits, u32 is_net_byte_order, u32 is_increment)
{
  ASSERT (v0 >= v_min && v0 <= v_max);
  ASSERT (v1 >= v_min && v1 <= (v_max + is_increment));
  if (n_bits == BITS (u8))
    {
      ((u8 *) a0)[0] = v0;
      ((u8 *) a1)[0] = v1;
    }
  else if (n_bits == BITS (u16))
    {
      if (is_net_byte_order)
	{
	  v0 = clib_host_to_net_u16 (v0);
	  v1 = clib_host_to_net_u16 (v1);
	}
      clib_mem_unaligned (a0, u16) = v0;
      clib_mem_unaligned (a1, u16) = v1;
    }
  else if (n_bits == BITS (u32))
    {
      if (is_net_byte_order)
	{
	  v0 = clib_host_to_net_u32 (v0);
	  v1 = clib_host_to_net_u32 (v1);
	}
      clib_mem_unaligned (a0, u32) = v0;
      clib_mem_unaligned (a1, u32) = v1;
    }
  else if (n_bits == BITS (u64))
    {
      if (is_net_byte_order)
	{
	  v0 = clib_host_to_net_u64 (v0);
	  v1 = clib_host_to_net_u64 (v1);
	}
      clib_mem_unaligned (a0, u64) = v0;
      clib_mem_unaligned (a1, u64) = v1;
    }
}

static_always_inline void
do_set_fixed (pg_main_t * pg,
	      pg_stream_t * s,
	      u32 * buffers,
	      u32 n_buffers,
	      u32 n_bits,
	      u32 byte_offset, u32 is_net_byte_order, u64 v_min, u64 v_max)
{
  vlib_main_t *vm = vlib_get_main ();

  while (n_buffers >= 4)
    {
      vlib_buffer_t *b0, *b1, *b2, *b3;
      void *a0, *a1;

      b0 = vlib_get_buffer (vm, buffers[0]);
      b1 = vlib_get_buffer (vm, buffers[1]);
      b2 = vlib_get_buffer (vm, buffers[2]);
      b3 = vlib_get_buffer (vm, buffers[3]);
      buffers += 2;
      n_buffers -= 2;

      a0 = (void *) b0 + byte_offset;
      a1 = (void *) b1 + byte_offset;
      CLIB_PREFETCH ((void *) b2 + byte_offset, sizeof (v_min), WRITE);
      CLIB_PREFETCH ((void *) b3 + byte_offset, sizeof (v_min), WRITE);

      set_2 (a0, a1, v_min, v_min, v_min, v_max, n_bits, is_net_byte_order,
	     /* is_increment */ 0);

      ASSERT (validate_buffer_data (b0, s));
      ASSERT (validate_buffer_data (b1, s));
    }

  while (n_buffers > 0)
    {
      vlib_buffer_t *b0;
      void *a0;

      b0 = vlib_get_buffer (vm, buffers[0]);
      buffers += 1;
      n_buffers -= 1;

      a0 = (void *) b0 + byte_offset;

      set_1 (a0, v_min, v_min, v_max, n_bits, is_net_byte_order);

      ASSERT (validate_buffer_data (b0, s));
    }
}

static_always_inline u64
do_set_increment (pg_main_t * pg,
		  pg_stream_t * s,
		  u32 * buffers,
		  u32 n_buffers,
		  u32 n_bits,
		  u32 byte_offset,
		  u32 is_net_byte_order,
		  u32 want_sum, u64 * sum_result, u64 v_min, u64 v_max, u64 v)
{
  vlib_main_t *vm = vlib_get_main ();
  u64 sum = 0;

  ASSERT (v >= v_min && v <= v_max);

  while (n_buffers >= 4)
    {
      vlib_buffer_t *b0, *b1, *b2, *b3;
      void *a0, *a1;
      u64 v_old;

      b0 = vlib_get_buffer (vm, buffers[0]);
      b1 = vlib_get_buffer (vm, buffers[1]);
      b2 = vlib_get_buffer (vm, buffers[2]);
      b3 = vlib_get_buffer (vm, buffers[3]);
      buffers += 2;
      n_buffers -= 2;

      a0 = (void *) b0 + byte_offset;
      a1 = (void *) b1 + byte_offset;
      CLIB_PREFETCH ((void *) b2 + byte_offset, sizeof (v_min), WRITE);
      CLIB_PREFETCH ((void *) b3 + byte_offset, sizeof (v_min), WRITE);

      v_old = v;
      v = v_old + 2;
      v = v > v_max ? v_min : v;
      set_2 (a0, a1,
	     v_old + 0, v_old + 1, v_min, v_max, n_bits, is_net_byte_order,
	     /* is_increment */ 1);

      if (want_sum)
	sum += 2 * v_old + 1;

      if (PREDICT_FALSE (v_old + 1 > v_max))
	{
	  if (want_sum)
	    sum -= 2 * v_old + 1;

	  v = v_old;
	  set_1 (a0, v + 0, v_min, v_max, n_bits, is_net_byte_order);
	  if (want_sum)
	    sum += v;
	  v += 1;

	  v = v > v_max ? v_min : v;
	  set_1 (a1, v + 0, v_min, v_max, n_bits, is_net_byte_order);
	  if (want_sum)
	    sum += v;
	  v += 1;
	}

      ASSERT (validate_buffer_data (b0, s));
      ASSERT (validate_buffer_data (b1, s));
    }

  while (n_buffers > 0)
    {
      vlib_buffer_t *b0;
      void *a0;
      u64 v_old;

      b0 = vlib_get_buffer (vm, buffers[0]);
      buffers += 1;
      n_buffers -= 1;

      a0 = (void *) b0 + byte_offset;

      v_old = v;
      if (want_sum)
	sum += v_old;
      v += 1;
      v = v > v_max ? v_min : v;

      ASSERT (v_old >= v_min && v_old <= v_max);
      set_1 (a0, v_old, v_min, v_max, n_bits, is_net_byte_order);

      ASSERT (validate_buffer_data (b0, s));
    }

  if (want_sum)
    *sum_result = sum;

  return v;
}

static_always_inline void
do_set_random (pg_main_t * pg,
	       pg_stream_t * s,
	       u32 * buffers,
	       u32 n_buffers,
	       u32 n_bits,
	       u32 byte_offset,
	       u32 is_net_byte_order,
	       u32 want_sum, u64 * sum_result, u64 v_min, u64 v_max)
{
  vlib_main_t *vm = vlib_get_main ();
  u64 v_diff = v_max - v_min + 1;
  u64 r_mask = max_pow2 (v_diff) - 1;
  u64 v0, v1;
  u64 sum = 0;
  void *random_data;

  random_data = clib_random_buffer_get_data
    (&vm->random_buffer, n_buffers * n_bits / BITS (u8));

  v0 = v1 = v_min;

  while (n_buffers >= 4)
    {
      vlib_buffer_t *b0, *b1, *b2, *b3;
      void *a0, *a1;
      u64 r0 = 0, r1 = 0;	/* warnings be gone */

      b0 = vlib_get_buffer (vm, buffers[0]);
      b1 = vlib_get_buffer (vm, buffers[1]);
      b2 = vlib_get_buffer (vm, buffers[2]);
      b3 = vlib_get_buffer (vm, buffers[3]);
      buffers += 2;
      n_buffers -= 2;

      a0 = (void *) b0 + byte_offset;
      a1 = (void *) b1 + byte_offset;
      CLIB_PREFETCH ((void *) b2 + byte_offset, sizeof (v_min), WRITE);
      CLIB_PREFETCH ((void *) b3 + byte_offset, sizeof (v_min), WRITE);

      switch (n_bits)
	{
#define _(n)					\
	  case BITS (u##n):			\
	    {					\
	      u##n * r = random_data;		\
	      r0 = r[0];			\
	      r1 = r[1];			\
	      random_data = r + 2;		\
	    }					\
	  break;

	  _(8);
	  _(16);
	  _(32);
	  _(64);

#undef _
	}

      /* Add power of 2 sized random number which may be out of range. */
      v0 += r0 & r_mask;
      v1 += r1 & r_mask;

      /* Twice should be enough to reduce to v_min .. v_max range. */
      v0 = v0 > v_max ? v0 - v_diff : v0;
      v1 = v1 > v_max ? v1 - v_diff : v1;
      v0 = v0 > v_max ? v0 - v_diff : v0;
      v1 = v1 > v_max ? v1 - v_diff : v1;

      if (want_sum)
	sum += v0 + v1;

      set_2 (a0, a1, v0, v1, v_min, v_max, n_bits, is_net_byte_order,
	     /* is_increment */ 0);

      ASSERT (validate_buffer_data (b0, s));
      ASSERT (validate_buffer_data (b1, s));
    }

  while (n_buffers > 0)
    {
      vlib_buffer_t *b0;
      void *a0;
      u64 r0 = 0;		/* warnings be gone */

      b0 = vlib_get_buffer (vm, buffers[0]);
      buffers += 1;
      n_buffers -= 1;

      a0 = (void *) b0 + byte_offset;

      switch (n_bits)
	{
#define _(n)					\
	  case BITS (u##n):			\
	    {					\
	      u##n * r = random_data;		\
	      r0 = r[0];			\
	      random_data = r + 1;		\
	    }					\
	  break;

	  _(8);
	  _(16);
	  _(32);
	  _(64);

#undef _
	}

      /* Add power of 2 sized random number which may be out of range. */
      v0 += r0 & r_mask;

      /* Twice should be enough to reduce to v_min .. v_max range. */
      v0 = v0 > v_max ? v0 - v_diff : v0;
      v0 = v0 > v_max ? v0 - v_diff : v0;

      if (want_sum)
	sum += v0;

      set_1 (a0, v0, v_min, v_max, n_bits, is_net_byte_order);

      ASSERT (validate_buffer_data (b0, s));
    }

  if (want_sum)
    *sum_result = sum;
}

#define _(i,t)							\
  clib_mem_unaligned (a##i, t) =				\
    clib_host_to_net_##t ((clib_net_to_host_mem_##t (a##i) &~ mask)	\
			  | (v##i << shift))

always_inline void
setbits_1 (void *a0,
	   u64 v0,
	   u64 v_min, u64 v_max,
	   u32 max_bits, u32 n_bits, u64 mask, u32 shift)
{
  ASSERT (v0 >= v_min && v0 <= v_max);
  if (max_bits == BITS (u8))
    ((u8 *) a0)[0] = (((u8 *) a0)[0] & ~mask) | (v0 << shift);

  else if (max_bits == BITS (u16))
    {
      _(0, u16);
    }
  else if (max_bits == BITS (u32))
    {
      _(0, u32);
    }
  else if (max_bits == BITS (u64))
    {
      _(0, u64);
    }
}

always_inline void
setbits_2 (void *a0, void *a1,
	   u64 v0, u64 v1,
	   u64 v_min, u64 v_max,
	   u32 max_bits, u32 n_bits, u64 mask, u32 shift, u32 is_increment)
{
  ASSERT (v0 >= v_min && v0 <= v_max);
  ASSERT (v1 >= v_min && v1 <= v_max + is_increment);
  if (max_bits == BITS (u8))
    {
      ((u8 *) a0)[0] = (((u8 *) a0)[0] & ~mask) | (v0 << shift);
      ((u8 *) a1)[0] = (((u8 *) a1)[0] & ~mask) | (v1 << shift);
    }

  else if (max_bits == BITS (u16))
    {
      _(0, u16);
      _(1, u16);
    }
  else if (max_bits == BITS (u32))
    {
      _(0, u32);
      _(1, u32);
    }
  else if (max_bits == BITS (u64))
    {
      _(0, u64);
      _(1, u64);
    }
}

#undef _

static_always_inline void
do_setbits_fixed (pg_main_t * pg,
		  pg_stream_t * s,
		  u32 * buffers,
		  u32 n_buffers,
		  u32 max_bits,
		  u32 n_bits,
		  u32 byte_offset, u64 v_min, u64 v_max, u64 mask, u32 shift)
{
  vlib_main_t *vm = vlib_get_main ();

  while (n_buffers >= 4)
    {
      vlib_buffer_t *b0, *b1, *b2, *b3;
      void *a0, *a1;

      b0 = vlib_get_buffer (vm, buffers[0]);
      b1 = vlib_get_buffer (vm, buffers[1]);
      b2 = vlib_get_buffer (vm, buffers[2]);
      b3 = vlib_get_buffer (vm, buffers[3]);
      buffers += 2;
      n_buffers -= 2;

      a0 = (void *) b0 + byte_offset;
      a1 = (void *) b1 + byte_offset;
      CLIB_PREFETCH ((void *) b2 + byte_offset, sizeof (v_min), WRITE);
      CLIB_PREFETCH ((void *) b3 + byte_offset, sizeof (v_min), WRITE);

      setbits_2 (a0, a1,
		 v_min, v_min, v_min, v_max, max_bits, n_bits, mask, shift,
		 /* is_increment */ 0);

      ASSERT (validate_buffer_data (b0, s));
      ASSERT (validate_buffer_data (b1, s));
    }

  while (n_buffers > 0)
    {
      vlib_buffer_t *b0;
      void *a0;

      b0 = vlib_get_buffer (vm, buffers[0]);
      buffers += 1;
      n_buffers -= 1;

      a0 = (void *) b0 + byte_offset;

      setbits_1 (a0, v_min, v_min, v_max, max_bits, n_bits, mask, shift);
      ASSERT (validate_buffer_data (b0, s));
    }
}

static_always_inline u64
do_setbits_increment (pg_main_t * pg,
		      pg_stream_t * s,
		      u32 * buffers,
		      u32 n_buffers,
		      u32 max_bits,
		      u32 n_bits,
		      u32 byte_offset,
		      u64 v_min, u64 v_max, u64 v, u64 mask, u32 shift)
{
  vlib_main_t *vm = vlib_get_main ();

  ASSERT (v >= v_min && v <= v_max);

  while (n_buffers >= 4)
    {
      vlib_buffer_t *b0, *b1, *b2, *b3;
      void *a0, *a1;
      u64 v_old;

      b0 = vlib_get_buffer (vm, buffers[0]);
      b1 = vlib_get_buffer (vm, buffers[1]);
      b2 = vlib_get_buffer (vm, buffers[2]);
      b3 = vlib_get_buffer (vm, buffers[3]);
      buffers += 2;
      n_buffers -= 2;

      a0 = (void *) b0 + byte_offset;
      a1 = (void *) b1 + byte_offset;
      CLIB_PREFETCH ((void *) b2 + byte_offset, sizeof (v_min), WRITE);
      CLIB_PREFETCH ((void *) b3 + byte_offset, sizeof (v_min), WRITE);

      v_old = v;
      v = v_old + 2;
      v = v > v_max ? v_min : v;
      setbits_2 (a0, a1,
		 v_old + 0, v_old + 1,
		 v_min, v_max, max_bits, n_bits, mask, shift,
		 /* is_increment */ 1);

      if (PREDICT_FALSE (v_old + 1 > v_max))
	{
	  v = v_old;
	  setbits_1 (a0, v + 0, v_min, v_max, max_bits, n_bits, mask, shift);
	  v += 1;

	  v = v > v_max ? v_min : v;
	  setbits_1 (a1, v + 0, v_min, v_max, max_bits, n_bits, mask, shift);
	  v += 1;
	}
      ASSERT (validate_buffer_data (b0, s));
      ASSERT (validate_buffer_data (b1, s));
    }

  while (n_buffers > 0)
    {
      vlib_buffer_t *b0;
      void *a0;
      u64 v_old;

      b0 = vlib_get_buffer (vm, buffers[0]);
      buffers += 1;
      n_buffers -= 1;

      a0 = (void *) b0 + byte_offset;

      v_old = v;
      v = v_old + 1;
      v = v > v_max ? v_min : v;

      ASSERT (v_old >= v_min && v_old <= v_max);
      setbits_1 (a0, v_old, v_min, v_max, max_bits, n_bits, mask, shift);

      ASSERT (validate_buffer_data (b0, s));
    }

  return v;
}

static_always_inline void
do_setbits_random (pg_main_t * pg,
		   pg_stream_t * s,
		   u32 * buffers,
		   u32 n_buffers,
		   u32 max_bits,
		   u32 n_bits,
		   u32 byte_offset, u64 v_min, u64 v_max, u64 mask, u32 shift)
{
  vlib_main_t *vm = vlib_get_main ();
  u64 v_diff = v_max - v_min + 1;
  u64 r_mask = max_pow2 (v_diff) - 1;
  u64 v0, v1;
  void *random_data;

  random_data = clib_random_buffer_get_data
    (&vm->random_buffer, n_buffers * max_bits / BITS (u8));
  v0 = v1 = v_min;

  while (n_buffers >= 4)
    {
      vlib_buffer_t *b0, *b1, *b2, *b3;
      void *a0, *a1;
      u64 r0 = 0, r1 = 0;	/* warnings be gone */

      b0 = vlib_get_buffer (vm, buffers[0]);
      b1 = vlib_get_buffer (vm, buffers[1]);
      b2 = vlib_get_buffer (vm, buffers[2]);
      b3 = vlib_get_buffer (vm, buffers[3]);
      buffers += 2;
      n_buffers -= 2;

      a0 = (void *) b0 + byte_offset;
      a1 = (void *) b1 + byte_offset;
      CLIB_PREFETCH ((void *) b2 + byte_offset, sizeof (v_min), WRITE);
      CLIB_PREFETCH ((void *) b3 + byte_offset, sizeof (v_min), WRITE);

      switch (max_bits)
	{
#define _(n)					\
	  case BITS (u##n):			\
	    {					\
	      u##n * r = random_data;		\
	      r0 = r[0];			\
	      r1 = r[1];			\
	      random_data = r + 2;		\
	    }					\
	  break;

	  _(8);
	  _(16);
	  _(32);
	  _(64);

#undef _
	}

      /* Add power of 2 sized random number which may be out of range. */
      v0 += r0 & r_mask;
      v1 += r1 & r_mask;

      /* Twice should be enough to reduce to v_min .. v_max range. */
      v0 = v0 > v_max ? v0 - v_diff : v0;
      v1 = v1 > v_max ? v1 - v_diff : v1;
      v0 = v0 > v_max ? v0 - v_diff : v0;
      v1 = v1 > v_max ? v1 - v_diff : v1;

      setbits_2 (a0, a1, v0, v1, v_min, v_max, max_bits, n_bits, mask, shift,
		 /* is_increment */ 0);

      ASSERT (validate_buffer_data (b0, s));
      ASSERT (validate_buffer_data (b1, s));
    }

  while (n_buffers > 0)
    {
      vlib_buffer_t *b0;
      void *a0;
      u64 r0 = 0;		/* warnings be gone */

      b0 = vlib_get_buffer (vm, buffers[0]);
      buffers += 1;
      n_buffers -= 1;

      a0 = (void *) b0 + byte_offset;

      switch (max_bits)
	{
#define _(n)					\
	  case BITS (u##n):			\
	    {					\
	      u##n * r = random_data;		\
	      r0 = r[0];			\
	      random_data = r + 1;		\
	    }					\
	  break;

	  _(8);
	  _(16);
	  _(32);
	  _(64);

#undef _
	}

      /* Add power of 2 sized random number which may be out of range. */
      v0 += r0 & r_mask;

      /* Twice should be enough to reduce to v_min .. v_max range. */
      v0 = v0 > v_max ? v0 - v_diff : v0;
      v0 = v0 > v_max ? v0 - v_diff : v0;

      setbits_1 (a0, v0, v_min, v_max, max_bits, n_bits, mask, shift);

      ASSERT (validate_buffer_data (b0, s));
    }
}

static u64
do_it (pg_main_t * pg,
       pg_stream_t * s,
       u32 * buffers,
       u32 n_buffers,
       u32 lo_bit, u32 hi_bit,
       u64 v_min, u64 v_max, u64 v, pg_edit_type_t edit_type)
{
  u32 max_bits, l0, l1, h1, start_bit;

  if (v_min == v_max)
    edit_type = PG_EDIT_FIXED;

  l0 = lo_bit / BITS (u8);
  l1 = lo_bit % BITS (u8);
  h1 = hi_bit % BITS (u8);

  start_bit = l0 * BITS (u8);

  max_bits = hi_bit - start_bit;
  ASSERT (max_bits <= 64);

#define _(n)						\
  case (n):						\
    if (edit_type == PG_EDIT_INCREMENT)			\
      v = do_set_increment (pg, s, buffers, n_buffers,	\
			    BITS (u##n),		\
			    l0,				\
			    /* is_net_byte_order */ 1,	\
			    /* want sum */ 0, 0,	\
			    v_min, v_max,		\
			    v);				\
    else if (edit_type == PG_EDIT_RANDOM)		\
      do_set_random (pg, s, buffers, n_buffers,		\
		     BITS (u##n),			\
		     l0,				\
		     /* is_net_byte_order */ 1,		\
		     /* want sum */ 0, 0,		\
		     v_min, v_max);			\
    else /* edit_type == PG_EDIT_FIXED */		\
      do_set_fixed (pg, s, buffers, n_buffers,		\
		    BITS (u##n),			\
		    l0,					\
		    /* is_net_byte_order */ 1,		\
		    v_min, v_max);			\
  goto done;

  if (l1 == 0 && h1 == 0)
    {
      switch (max_bits)
	{
	  _(8);
	  _(16);
	  _(32);
	  _(64);
	}
    }

#undef _

  {
    u64 mask;
    u32 shift = l1;
    u32 n_bits = max_bits;

    max_bits = clib_max (max_pow2 (n_bits), 8);

    mask = ((u64) 1 << (u64) n_bits) - 1;
    mask &= ~(((u64) 1 << (u64) shift) - 1);

    mask <<= max_bits - n_bits;
    shift += max_bits - n_bits;

    switch (max_bits)
      {
#define _(n)								\
	case (n):							\
	  if (edit_type == PG_EDIT_INCREMENT)				\
	    v = do_setbits_increment (pg, s, buffers, n_buffers,	\
				      BITS (u##n), n_bits,		\
				      l0, v_min, v_max, v,		\
				      mask, shift);			\
	  else if (edit_type == PG_EDIT_RANDOM)				\
	    do_setbits_random (pg, s, buffers, n_buffers,		\
			       BITS (u##n), n_bits,			\
			       l0, v_min, v_max,			\
			       mask, shift);				\
	  else /* edit_type == PG_EDIT_FIXED */				\
	    do_setbits_fixed (pg, s, buffers, n_buffers,		\
			      BITS (u##n), n_bits,			\
			      l0, v_min, v_max,				\
			      mask, shift);				\
	goto done;

	_(8);
	_(16);
	_(32);
	_(64);

#undef _
      }
  }

done:
  return v;
}

static void
pg_generate_set_lengths (pg_main_t * pg,
			 pg_stream_t * s, u32 * buffers, u32 n_buffers)
{
  u64 v_min, v_max, length_sum;
  pg_edit_type_t edit_type;

  v_min = s->min_packet_bytes;
  v_max = s->max_packet_bytes;
  edit_type = s->packet_size_edit_type;

  if (edit_type == PG_EDIT_INCREMENT)
    s->last_increment_packet_size
      = do_set_increment (pg, s, buffers, n_buffers,
			  8 * STRUCT_SIZE_OF (vlib_buffer_t, current_length),
			  STRUCT_OFFSET_OF (vlib_buffer_t, current_length),
			  /* is_net_byte_order */ 0,
			  /* want sum */ 1, &length_sum,
			  v_min, v_max, s->last_increment_packet_size);

  else if (edit_type == PG_EDIT_RANDOM)
    do_set_random (pg, s, buffers, n_buffers,
		   8 * STRUCT_SIZE_OF (vlib_buffer_t, current_length),
		   STRUCT_OFFSET_OF (vlib_buffer_t, current_length),
		   /* is_net_byte_order */ 0,
		   /* want sum */ 1, &length_sum,
		   v_min, v_max);

  else				/* edit_type == PG_EDIT_FIXED */
    {
      do_set_fixed (pg, s, buffers, n_buffers,
		    8 * STRUCT_SIZE_OF (vlib_buffer_t, current_length),
		    STRUCT_OFFSET_OF (vlib_buffer_t, current_length),
		    /* is_net_byte_order */ 0,
		    v_min, v_max);
      length_sum = v_min * n_buffers;
    }

  {
    vnet_main_t *vnm = vnet_get_main ();
    vnet_interface_main_t *im = &vnm->interface_main;
    vnet_sw_interface_t *si =
      vnet_get_sw_interface (vnm, s->sw_if_index[VLIB_RX]);

    vlib_increment_combined_counter (im->combined_sw_if_counters
				     + VNET_INTERFACE_COUNTER_RX,
				     vlib_get_thread_index (),
				     si->sw_if_index, n_buffers, length_sum);
  }

}

static void
pg_generate_fix_multi_buffer_lengths (pg_main_t * pg,
				      pg_stream_t * s,
				      u32 * buffers, u32 n_buffers)
{
  vlib_main_t *vm = vlib_get_main ();
  pg_buffer_index_t *pbi;
  uword n_bytes_left;
  static u32 *unused_buffers = 0;

  while (n_buffers > 0)
    {
      vlib_buffer_t *b;
      u32 bi;

      bi = buffers[0];
      b = vlib_get_buffer (vm, bi);

      /* Current length here is length of whole packet. */
      n_bytes_left = b->current_length;

      pbi = s->buffer_indices;
      while (1)
	{
	  uword n = clib_min (n_bytes_left, s->buffer_bytes);

	  b->current_length = n;
	  n_bytes_left -= n;
	  if (n_bytes_left > 0)
	    b->flags |= VLIB_BUFFER_NEXT_PRESENT;
	  else
	    b->flags &= ~VLIB_BUFFER_NEXT_PRESENT;

	  /* Return unused buffers to fifos. */
	  if (n == 0)
	    vec_add1 (unused_buffers, bi);

	  pbi++;
	  if (pbi >= vec_end (s->buffer_indices))
	    break;

	  bi = b->next_buffer;
	  b = vlib_get_buffer (vm, bi);
	}
      ASSERT (n_bytes_left == 0);

      buffers += 1;
      n_buffers -= 1;
    }

  if (vec_len (unused_buffers) > 0)
    {
      vlib_buffer_free_no_next (vm, unused_buffers, vec_len (unused_buffers));
      _vec_len (unused_buffers) = 0;
    }
}

static void
pg_generate_edit (pg_main_t * pg,
		  pg_stream_t * s, u32 * buffers, u32 n_buffers)
{
  pg_edit_t *e;

  vec_foreach (e, s->non_fixed_edits)
  {
    switch (e->type)
      {
      case PG_EDIT_RANDOM:
      case PG_EDIT_INCREMENT:
	{
	  u32 lo_bit, hi_bit;
	  u64 v_min, v_max;

	  v_min = pg_edit_get_value (e, PG_EDIT_LO);
	  v_max = pg_edit_get_value (e, PG_EDIT_HI);

	  hi_bit = (BITS (u8) * STRUCT_OFFSET_OF (vlib_buffer_t, data)
		    + BITS (u8) + e->lsb_bit_offset);
	  lo_bit = hi_bit - e->n_bits;

	  e->last_increment_value
	    = do_it (pg, s, buffers, n_buffers, lo_bit, hi_bit, v_min, v_max,
		     e->last_increment_value, e->type);
	}
	break;

      case PG_EDIT_UNSPECIFIED:
	break;

      default:
	/* Should not be any fixed edits left. */
	ASSERT (0);
	break;
      }
  }

  /* Call any edit functions to e.g. completely IP lengths, checksums, ... */
  {
    int i;
    for (i = vec_len (s->edit_groups) - 1; i >= 0; i--)
      {
	pg_edit_group_t *g = s->edit_groups + i;
	if (g->edit_function)
	  g->edit_function (pg, s, g, buffers, n_buffers);
      }
  }
}

static void
pg_set_next_buffer_pointers (pg_main_t * pg,
			     pg_stream_t * s,
			     u32 * buffers, u32 * next_buffers, u32 n_buffers)
{
  vlib_main_t *vm = vlib_get_main ();

  while (n_buffers >= 4)
    {
      u32 ni0, ni1;
      vlib_buffer_t *b0, *b1;

      b0 = vlib_get_buffer (vm, buffers[0]);
      b1 = vlib_get_buffer (vm, buffers[1]);
      ni0 = next_buffers[0];
      ni1 = next_buffers[1];

      vlib_prefetch_buffer_with_index (vm, buffers[2], WRITE);
      vlib_prefetch_buffer_with_index (vm, buffers[3], WRITE);

      b0->flags |= VLIB_BUFFER_NEXT_PRESENT;
      b1->flags |= VLIB_BUFFER_NEXT_PRESENT;
      b0->next_buffer = ni0;
      b1->next_buffer = ni1;

      buffers += 2;
      next_buffers += 2;
      n_buffers -= 2;
    }

  while (n_buffers > 0)
    {
      u32 ni0;
      vlib_buffer_t *b0;

      b0 = vlib_get_buffer (vm, buffers[0]);
      ni0 = next_buffers[0];
      buffers += 1;
      next_buffers += 1;
      n_buffers -= 1;

      b0->flags |= VLIB_BUFFER_NEXT_PRESENT;
      b0->next_buffer = ni0;
    }
}

static_always_inline void
init_replay_buffers_inline (vlib_main_t * vm,
			    pg_stream_t * s,
			    u32 * buffers,
			    u32 n_buffers, u32 data_offset, u32 n_data)
{
  u32 n_left, *b, i, l;

  n_left = n_buffers;
  b = buffers;
  i = s->current_replay_packet_index;
  l = vec_len (s->replay_packet_templates);

  while (n_left >= 1)
    {
      u32 bi0, n0;
      vlib_buffer_t *b0;
      u8 *d0;

      bi0 = b[0];
      b += 1;
      n_left -= 1;

      b0 = vlib_get_buffer (vm, bi0);

      vnet_buffer (b0)->sw_if_index[VLIB_RX] = s->sw_if_index[VLIB_RX];
      /* was s->sw_if_index[VLIB_TX]; */
      vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;

      d0 = vec_elt (s->replay_packet_templates, i);

      n0 = n_data;
      if (data_offset + n_data >= vec_len (d0))
	n0 = vec_len (d0) > data_offset ? vec_len (d0) - data_offset : 0;

      b0->current_length = n0;

      clib_memcpy (b0->data, d0 + data_offset, n0);
      i = i + 1 == l ? 0 : i + 1;
    }
}

static_always_inline void
init_buffers_inline (vlib_main_t * vm,
		     pg_stream_t * s,
		     u32 * buffers,
		     u32 n_buffers, u32 data_offset, u32 n_data, u32 set_data)
{
  u32 n_left, *b;
  u8 *data, *mask;

  if (vec_len (s->replay_packet_templates) > 0)
    return init_replay_buffers_inline (vm, s, buffers, n_buffers, data_offset,
				       n_data);

  data = s->fixed_packet_data + data_offset;
  mask = s->fixed_packet_data_mask + data_offset;
  if (data + n_data >= vec_end (s->fixed_packet_data))
    n_data = (data < vec_end (s->fixed_packet_data)
	      ? vec_end (s->fixed_packet_data) - data : 0);
  if (n_data > 0)
    {
      ASSERT (data + n_data <= vec_end (s->fixed_packet_data));
      ASSERT (mask + n_data <= vec_end (s->fixed_packet_data_mask));
    }

  n_left = n_buffers;
  b = buffers;

  while (n_left >= 4)
    {
      u32 bi0, bi1;
      vlib_buffer_t *b0, *b1;

      /* Prefetch next iteration. */
      vlib_prefetch_buffer_with_index (vm, b[2], STORE);
      vlib_prefetch_buffer_with_index (vm, b[3], STORE);

      bi0 = b[0];
      bi1 = b[1];
      b += 2;
      n_left -= 2;

      b0 = vlib_get_buffer (vm, bi0);
      b1 = vlib_get_buffer (vm, bi1);

      vnet_buffer (b0)->sw_if_index[VLIB_RX] =
	vnet_buffer (b1)->sw_if_index[VLIB_RX] = s->sw_if_index[VLIB_RX];

      vnet_buffer (b0)->sw_if_index[VLIB_TX] =
	vnet_buffer (b1)->sw_if_index[VLIB_TX] = (u32) ~ 0;

      if (set_data)
	{
	  clib_memcpy (b0->data, data, n_data);
	  clib_memcpy (b1->data, data, n_data);
	}
      else
	{
	  ASSERT (validate_buffer_data2 (b0, s, data_offset, n_data));
	  ASSERT (validate_buffer_data2 (b1, s, data_offset, n_data));
	}
    }

  while (n_left >= 1)
    {
      u32 bi0;
      vlib_buffer_t *b0;

      bi0 = b[0];
      b += 1;
      n_left -= 1;

      b0 = vlib_get_buffer (vm, bi0);
      vnet_buffer (b0)->sw_if_index[VLIB_RX] = s->sw_if_index[VLIB_RX];
      /* s->sw_if_index[VLIB_TX]; */
      vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;

      if (set_data)
	clib_memcpy (b0->data, data, n_data);
      else
	ASSERT (validate_buffer_data2 (b0, s, data_offset, n_data));
    }
}

static u32
pg_stream_fill_helper (pg_main_t * pg,
		       pg_stream_t * s,
		       pg_buffer_index_t * bi,
		       u32 * buffers, u32 * next_buffers, u32 n_alloc)
{
  vlib_main_t *vm = vlib_get_main ();
  uword is_start_of_packet = bi == s->buffer_indices;
  u32 n_allocated;

  n_allocated = vlib_buffer_alloc_from_free_list (vm,
						  buffers,
						  n_alloc,
						  bi->free_list_index);
  if (n_allocated == 0)
    return 0;

  /*
   * We can't assume we got all the buffers we asked for...
   * This never worked until recently.
   */
  n_alloc = n_allocated;

  /* Reinitialize buffers */
  init_buffers_inline
    (vm, s,
     buffers,
     n_alloc, (bi - s->buffer_indices) * s->buffer_bytes /* data offset */ ,
     s->buffer_bytes,
     /* set_data */ 1);

  if (next_buffers)
    pg_set_next_buffer_pointers (pg, s, buffers, next_buffers, n_alloc);

  if (is_start_of_packet)
    {
      if (vec_len (s->replay_packet_templates) > 0)
	{
	  vnet_main_t *vnm = vnet_get_main ();
	  vnet_interface_main_t *im = &vnm->interface_main;
	  vnet_sw_interface_t *si =
	    vnet_get_sw_interface (vnm, s->sw_if_index[VLIB_RX]);
	  u32 l = 0;
	  u32 i;
	  for (i = 0; i < n_alloc; i++)
	    l += vlib_buffer_index_length_in_chain (vm, buffers[i]);
	  vlib_increment_combined_counter (im->combined_sw_if_counters
					   + VNET_INTERFACE_COUNTER_RX,
					   vlib_get_thread_index (),
					   si->sw_if_index, n_alloc, l);
	  s->current_replay_packet_index += n_alloc;
	  s->current_replay_packet_index %=
	    vec_len (s->replay_packet_templates);
	}
      else
	{
	  pg_generate_set_lengths (pg, s, buffers, n_alloc);
	  if (vec_len (s->buffer_indices) > 1)
	    pg_generate_fix_multi_buffer_lengths (pg, s, buffers, n_alloc);

	  pg_generate_edit (pg, s, buffers, n_alloc);
	}
    }

  return n_alloc;
}

static u32
pg_stream_fill (pg_main_t * pg, pg_stream_t * s, u32 n_buffers)
{
  pg_buffer_index_t *bi;
  word i, n_in_fifo, n_alloc, n_free, n_added;
  u32 *tail, *start, *end, *last_tail, *last_start;

  bi = s->buffer_indices;

  n_in_fifo = clib_fifo_elts (bi->buffer_fifo);
  if (n_in_fifo >= n_buffers)
    return n_in_fifo;

  n_alloc = n_buffers - n_in_fifo;

  /* Round up, but never generate more than limit. */
  n_alloc = clib_max (VLIB_FRAME_SIZE, n_alloc);

  if (s->n_packets_limit > 0
      && s->n_packets_generated + n_in_fifo + n_alloc >= s->n_packets_limit)
    {
      n_alloc = s->n_packets_limit - s->n_packets_generated - n_in_fifo;
      if (n_alloc < 0)
	n_alloc = 0;
    }

  /* All buffer fifos should have the same size. */
  if (CLIB_DEBUG > 0)
    {
      uword l = ~0, e;
      vec_foreach (bi, s->buffer_indices)
      {
	e = clib_fifo_elts (bi->buffer_fifo);
	if (bi == s->buffer_indices)
	  l = e;
	ASSERT (l == e);
      }
    }

  last_tail = last_start = 0;
  n_added = n_alloc;

  for (i = vec_len (s->buffer_indices) - 1; i >= 0; i--)
    {
      bi = vec_elt_at_index (s->buffer_indices, i);

      n_free = clib_fifo_free_elts (bi->buffer_fifo);
      if (n_free < n_alloc)
	clib_fifo_resize (bi->buffer_fifo, n_alloc - n_free);

      tail = clib_fifo_advance_tail (bi->buffer_fifo, n_alloc);
      start = bi->buffer_fifo;
      end = clib_fifo_end (bi->buffer_fifo);

      if (tail + n_alloc <= end)
	{
	  n_added =
	    pg_stream_fill_helper (pg, s, bi, tail, last_tail, n_alloc);
	}
      else
	{
	  u32 n = clib_min (end - tail, n_alloc);
	  n_added = pg_stream_fill_helper (pg, s, bi, tail, last_tail, n);

	  if (n_added == n && n_alloc > n_added)
	    {
	      n_added += pg_stream_fill_helper
		(pg, s, bi, start, last_start, n_alloc - n_added);
	    }
	}

      if (PREDICT_FALSE (n_added < n_alloc))
	tail = clib_fifo_advance_tail (bi->buffer_fifo, n_added - n_alloc);

      last_tail = tail;
      last_start = start;

      /* Verify that pkts in the fifo are properly allocated */
    }

  return n_in_fifo + n_added;
}

typedef struct
{
  u32 stream_index;

  u32 packet_length;
  u32 sw_if_index;

  /* Use pre data for packet data. */
  vlib_buffer_t buffer;
} pg_input_trace_t;

static u8 *
format_pg_input_trace (u8 * s, va_list * va)
{
  vlib_main_t *vm = va_arg (*va, vlib_main_t *);
  CLIB_UNUSED (vlib_node_t * node) = va_arg (*va, vlib_node_t *);
  pg_input_trace_t *t = va_arg (*va, pg_input_trace_t *);
  pg_main_t *pg = &pg_main;
  pg_stream_t *stream;
  vlib_node_t *n;
  u32 indent = format_get_indent (s);

  stream = 0;
  if (!pool_is_free_index (pg->streams, t->stream_index))
    stream = pool_elt_at_index (pg->streams, t->stream_index);

  if (stream)
    s = format (s, "stream %v", pg->streams[t->stream_index].name);
  else
    s = format (s, "stream %d", t->stream_index);

  s = format (s, ", %d bytes", t->packet_length);
  s = format (s, ", %d sw_if_index", t->sw_if_index);

  s = format (s, "\n%U%U",
	      format_white_space, indent, format_vnet_buffer, &t->buffer);

  s = format (s, "\n%U", format_white_space, indent);

  n = 0;
  if (stream)
    n = vlib_get_node (vm, stream->node_index);

  if (n && n->format_buffer)
    s = format (s, "%U", n->format_buffer,
		t->buffer.pre_data, sizeof (t->buffer.pre_data));
  else
    s = format (s, "%U",
		format_hex_bytes, t->buffer.pre_data,
		ARRAY_LEN (t->buffer.pre_data));
  return s;
}

static void
pg_input_trace (pg_main_t * pg,
		vlib_node_runtime_t * node,
		pg_stream_t * s, u32 * buffers, u32 n_buffers)
{
  vlib_main_t *vm = vlib_get_main ();
  u32 *b, n_left, stream_index, next_index;

  n_left = n_buffers;
  b = buffers;
  stream_index = s - pg->streams;
  next_index = s->next_index;

  while (n_left >= 2)
    {
      u32 bi0, bi1;
      vlib_buffer_t *b0, *b1;
      pg_input_trace_t *t0, *t1;

      bi0 = b[0];
      bi1 = b[1];
      b += 2;
      n_left -= 2;

      b0 = vlib_get_buffer (vm, bi0);
      b1 = vlib_get_buffer (vm, bi1);

      vlib_trace_buffer (vm, node, next_index, b0, /* follow_chain */ 1);
      vlib_trace_buffer (vm, node, next_index, b1, /* follow_chain */ 1);

      t0 = vlib_add_trace (vm, node, b0, sizeof (t0[0]));
      t1 = vlib_add_trace (vm, node, b1, sizeof (t1[0]));

      t0->stream_index = stream_index;
      t1->stream_index = stream_index;

      t0->packet_length = vlib_buffer_length_in_chain (vm, b0);
      t1->packet_length = vlib_buffer_length_in_chain (vm, b1);

      t0->sw_if_index = vnet_buffer (b0)->sw_if_index[VLIB_RX];
      t1->sw_if_index = vnet_buffer (b1)->sw_if_index[VLIB_RX];

      clib_memcpy (&t0->buffer, b0, sizeof (b0[0]) - sizeof (b0->pre_data));
      clib_memcpy (&t1->buffer, b1, sizeof (b1[0]) - sizeof (b1->pre_data));

      clib_memcpy (t0->buffer.pre_data, b0->data,
		   sizeof (t0->buffer.pre_data));
      clib_memcpy (t1->buffer.pre_data, b1->data,
		   sizeof (t1->buffer.pre_data));
    }

  while (n_left >= 1)
    {
      u32 bi0;
      vlib_buffer_t *b0;
      pg_input_trace_t *t0;

      bi0 = b[0];
      b += 1;
      n_left -= 1;

      b0 = vlib_get_buffer (vm, bi0);

      vlib_trace_buffer (vm, node, next_index, b0, /* follow_chain */ 1);
      t0 = vlib_add_trace (vm, node, b0, sizeof (t0[0]));

      t0->stream_index = stream_index;
      t0->packet_length = vlib_buffer_length_in_chain (vm, b0);
      t0->sw_if_index = vnet_buffer (b0)->sw_if_index[VLIB_RX];
      clib_memcpy (&t0->buffer, b0, sizeof (b0[0]) - sizeof (b0->pre_data));
      clib_memcpy (t0->buffer.pre_data, b0->data,
		   sizeof (t0->buffer.pre_data));
    }
}

static uword
pg_generate_packets (vlib_node_runtime_t * node,
		     pg_main_t * pg,
		     pg_stream_t * s, uword n_packets_to_generate)
{
  vlib_main_t *vm = vlib_get_main ();
  u32 *to_next, n_this_frame, n_left, n_trace, n_packets_in_fifo;
  uword n_packets_generated;
  pg_buffer_index_t *bi, *bi0;
  u32 next_index = s->next_index;
  vnet_feature_main_t *fm = &feature_main;
  vnet_feature_config_main_t *cm;
  u8 feature_arc_index = fm->device_input_feature_arc_index;
  cm = &fm->feature_config_mains[feature_arc_index];
  u32 current_config_index = ~(u32) 0;
  int i;

  bi0 = s->buffer_indices;

  n_packets_in_fifo = pg_stream_fill (pg, s, n_packets_to_generate);
  n_packets_to_generate = clib_min (n_packets_in_fifo, n_packets_to_generate);
  n_packets_generated = 0;

  if (PREDICT_FALSE
      (vnet_have_features (feature_arc_index, s->sw_if_index[VLIB_RX])))
    {
      current_config_index =
	vec_elt (cm->config_index_by_sw_if_index, s->sw_if_index[VLIB_RX]);
      vnet_get_config_data (&cm->config_main, &current_config_index,
			    &next_index, 0);
    }

  while (n_packets_to_generate > 0)
    {
      u32 *head, *start, *end;

      vlib_get_next_frame (vm, node, next_index, to_next, n_left);

      n_this_frame = n_packets_to_generate;
      if (n_this_frame > n_left)
	n_this_frame = n_left;

      start = bi0->buffer_fifo;
      end = clib_fifo_end (bi0->buffer_fifo);
      head = clib_fifo_head (bi0->buffer_fifo);

      if (head + n_this_frame <= end)
	vlib_copy_buffers (to_next, head, n_this_frame);
      else
	{
	  u32 n = end - head;
	  vlib_copy_buffers (to_next + 0, head, n);
	  vlib_copy_buffers (to_next + n, start, n_this_frame - n);
	}

      vec_foreach (bi, s->buffer_indices)
	clib_fifo_advance_head (bi->buffer_fifo, n_this_frame);

      if (current_config_index != ~(u32) 0)
	for (i = 0; i < n_this_frame; i++)
	  {
	    vlib_buffer_t *b;
	    b = vlib_get_buffer (vm, to_next[i]);
	    vnet_buffer (b)->device_input_feat.saved_next_index =
	      s->next_index;
	    vnet_buffer (b)->device_input_feat.buffer_advance = 0;
	    b->current_config_index = current_config_index;
	    vnet_buffer (b)->feature_arc_index = feature_arc_index;
	  }

      n_trace = vlib_get_trace_count (vm, node);
      if (n_trace > 0)
	{
	  u32 n = clib_min (n_trace, n_this_frame);
	  pg_input_trace (pg, node, s, to_next, n);
	  vlib_set_trace_count (vm, node, n_trace - n);
	}
      n_packets_to_generate -= n_this_frame;
      n_packets_generated += n_this_frame;
      n_left -= n_this_frame;
      vlib_put_next_frame (vm, node, next_index, n_left);
    }

  return n_packets_generated;
}

static uword
pg_input_stream (vlib_node_runtime_t * node, pg_main_t * pg, pg_stream_t * s)
{
  vlib_main_t *vm = vlib_get_main ();
  uword n_packets;
  f64 time_now, dt;

  if (s->n_packets_limit > 0 && s->n_packets_generated >= s->n_packets_limit)
    {
      pg_stream_enable_disable (pg, s, /* want_enabled */ 0);
      return 0;
    }

  /* Apply rate limit. */
  time_now = vlib_time_now (vm);
  if (s->time_last_generate == 0)
    s->time_last_generate = time_now;

  dt = time_now - s->time_last_generate;
  s->time_last_generate = time_now;

  n_packets = VLIB_FRAME_SIZE;
  if (s->rate_packets_per_second > 0)
    {
      s->packet_accumulator += dt * s->rate_packets_per_second;
      n_packets = s->packet_accumulator;

      /* Never allow accumulator to grow if we get behind. */
      s->packet_accumulator -= n_packets;
    }

  /* Apply fixed limit. */
  if (s->n_packets_limit > 0
      && s->n_packets_generated + n_packets > s->n_packets_limit)
    n_packets = s->n_packets_limit - s->n_packets_generated;

  /* Generate up to one frame's worth of packets. */
  if (n_packets > VLIB_FRAME_SIZE)
    n_packets = VLIB_FRAME_SIZE;

  if (n_packets > 0)
    n_packets = pg_generate_packets (node, pg, s, n_packets);

  s->n_packets_generated += n_packets;

  return n_packets;
}

uword
pg_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
{
  uword i;
  pg_main_t *pg = &pg_main;
  uword n_packets = 0;
  u32 worker_index = 0;

  if (vlib_num_workers ())
    worker_index = vlib_get_current_worker_index ();

  /* *INDENT-OFF* */
  clib_bitmap_foreach (i, pg->enabled_streams[worker_index], ({
    pg_stream_t *s = vec_elt_at_index (pg->streams, i);
    n_packets += pg_input_stream (node, pg, s);
  }));
  /* *INDENT-ON* */

  return n_packets;
}

/* *INDENT-OFF* */
VLIB_REGISTER_NODE (pg_input_node) = {
  .function = pg_input,
  .name = "pg-input",
  .sibling_of = "device-input",
  .type = VLIB_NODE_TYPE_INPUT,

  .format_trace = format_pg_input_trace,

  /* Input node will be left disabled until a stream is active. */
  .state = VLIB_NODE_STATE_DISABLED,
};
/* *INDENT-ON* */

/*
 * fd.io coding-style-patch-verification: ON
 *
 * Local Variables:
 * eval: (c-set-style "gnu")
 * End:
 */