node.c revision 7ca5aaac
1/*
2 *------------------------------------------------------------------
3 * af_packet.c - linux kernel packet interface
4 *
5 * Copyright (c) 2016 Cisco and/or its affiliates.
6 * Licensed under the Apache License, Version 2.0 (the "License");
7 * you may not use this file except in compliance with the License.
8 * You may obtain a copy of the License at:
9 *
10 *     http://www.apache.org/licenses/LICENSE-2.0
11 *
12 * Unless required by applicable law or agreed to in writing, software
13 * distributed under the License is distributed on an "AS IS" BASIS,
14 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
15 * See the License for the specific language governing permissions and
16 * limitations under the License.
17 *------------------------------------------------------------------
18 */
19
20#include <linux/if_packet.h>
21
22#include <vlib/vlib.h>
23#include <vlib/unix/unix.h>
24#include <vnet/ip/ip.h>
25#include <vnet/ethernet/ethernet.h>
26#include <vnet/devices/devices.h>
27#include <vnet/feature/feature.h>
28#include <vnet/ethernet/packet.h>
29
30#include <vnet/devices/af_packet/af_packet.h>
31
32#define foreach_af_packet_input_error \
33  _(PARTIAL_PKT, "partial packet")
34
35typedef enum
36{
37#define _(f,s) AF_PACKET_INPUT_ERROR_##f,
38  foreach_af_packet_input_error
39#undef _
40    AF_PACKET_INPUT_N_ERROR,
41} af_packet_input_error_t;
42
43static char *af_packet_input_error_strings[] = {
44#define _(n,s) s,
45  foreach_af_packet_input_error
46#undef _
47};
48
49typedef struct
50{
51  u32 next_index;
52  u32 hw_if_index;
53  int block;
54  struct tpacket2_hdr tph;
55} af_packet_input_trace_t;
56
57static u8 *
58format_af_packet_input_trace (u8 * s, va_list * args)
59{
60  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
61  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
62  af_packet_input_trace_t *t = va_arg (*args, af_packet_input_trace_t *);
63  u32 indent = format_get_indent (s);
64
65  s = format (s, "af_packet: hw_if_index %d next-index %d",
66	      t->hw_if_index, t->next_index);
67
68  s =
69    format (s,
70	    "\n%Utpacket2_hdr:\n%Ustatus 0x%x len %u snaplen %u mac %u net %u"
71	    "\n%Usec 0x%x nsec 0x%x vlan %U"
72#ifdef TP_STATUS_VLAN_TPID_VALID
73	    " vlan_tpid %u"
74#endif
75	    ,
76	    format_white_space, indent + 2,
77	    format_white_space, indent + 4,
78	    t->tph.tp_status,
79	    t->tph.tp_len,
80	    t->tph.tp_snaplen,
81	    t->tph.tp_mac,
82	    t->tph.tp_net,
83	    format_white_space, indent + 4,
84	    t->tph.tp_sec,
85	    t->tph.tp_nsec, format_ethernet_vlan_tci, t->tph.tp_vlan_tci
86#ifdef TP_STATUS_VLAN_TPID_VALID
87	    , t->tph.tp_vlan_tpid
88#endif
89    );
90  return s;
91}
92
93always_inline void
94buffer_add_to_chain (vlib_main_t * vm, u32 bi, u32 first_bi, u32 prev_bi)
95{
96  vlib_buffer_t *b = vlib_get_buffer (vm, bi);
97  vlib_buffer_t *first_b = vlib_get_buffer (vm, first_bi);
98  vlib_buffer_t *prev_b = vlib_get_buffer (vm, prev_bi);
99
100  /* update first buffer */
101  first_b->total_length_not_including_first_buffer += b->current_length;
102
103  /* update previous buffer */
104  prev_b->next_buffer = bi;
105  prev_b->flags |= VLIB_BUFFER_NEXT_PRESENT;
106
107  /* update current buffer */
108  b->next_buffer = 0;
109}
110
111static_always_inline void
112mark_tcp_udp_cksum_calc (vlib_buffer_t * b)
113{
114  ethernet_header_t *eth = vlib_buffer_get_current (b);
115  if (clib_net_to_host_u16 (eth->type) == ETHERNET_TYPE_IP4)
116    {
117      ip4_header_t *ip4 =
118	(vlib_buffer_get_current (b) + sizeof (ethernet_header_t));
119      b->flags |= VNET_BUFFER_F_IS_IP4;
120      if (ip4->protocol == IP_PROTOCOL_TCP)
121	{
122	  b->flags |= VNET_BUFFER_F_OFFLOAD_TCP_CKSUM;
123	  ((tcp_header_t
124	    *) (vlib_buffer_get_current (b) +
125		sizeof (ethernet_header_t) +
126		ip4_header_bytes (ip4)))->checksum = 0;
127	}
128      else if (ip4->protocol == IP_PROTOCOL_UDP)
129	{
130	  b->flags |= VNET_BUFFER_F_OFFLOAD_UDP_CKSUM;
131	  ((udp_header_t
132	    *) (vlib_buffer_get_current (b) +
133		sizeof (ethernet_header_t) +
134		ip4_header_bytes (ip4)))->checksum = 0;
135	}
136      vnet_buffer (b)->l3_hdr_offset = sizeof (ethernet_header_t);
137      vnet_buffer (b)->l4_hdr_offset =
138	sizeof (ethernet_header_t) + ip4_header_bytes (ip4);
139    }
140  else if (clib_net_to_host_u16 (eth->type) == ETHERNET_TYPE_IP6)
141    {
142      ip6_header_t *ip6 =
143	(vlib_buffer_get_current (b) + sizeof (ethernet_header_t));
144      b->flags |= VNET_BUFFER_F_IS_IP6;
145      u16 ip6_hdr_len = sizeof (ip6_header_t);
146      if (ip6_ext_hdr (ip6->protocol))
147	{
148	  ip6_ext_header_t *p = (void *) (ip6 + 1);
149	  ip6_hdr_len += ip6_ext_header_len (p);
150	  while (ip6_ext_hdr (p->next_hdr))
151	    {
152	      ip6_hdr_len += ip6_ext_header_len (p);
153	      p = ip6_ext_next_header (p);
154	    }
155	}
156      if (ip6->protocol == IP_PROTOCOL_TCP)
157	{
158	  b->flags |= VNET_BUFFER_F_OFFLOAD_TCP_CKSUM;
159	  ((tcp_header_t
160	    *) (vlib_buffer_get_current (b) +
161		sizeof (ethernet_header_t) + ip6_hdr_len))->checksum = 0;
162	}
163      else if (ip6->protocol == IP_PROTOCOL_UDP)
164	{
165	  b->flags |= VNET_BUFFER_F_OFFLOAD_UDP_CKSUM;
166	  ((udp_header_t
167	    *) (vlib_buffer_get_current (b) +
168		sizeof (ethernet_header_t) + ip6_hdr_len))->checksum = 0;
169	}
170      vnet_buffer (b)->l3_hdr_offset = sizeof (ethernet_header_t);
171      vnet_buffer (b)->l4_hdr_offset =
172	sizeof (ethernet_header_t) + ip6_hdr_len;
173    }
174}
175
176always_inline uword
177af_packet_device_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
178			   vlib_frame_t * frame, af_packet_if_t * apif)
179{
180  af_packet_main_t *apm = &af_packet_main;
181  struct tpacket2_hdr *tph;
182  u32 next_index = VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT;
183  u32 block = 0;
184  u32 rx_frame;
185  u32 n_free_bufs;
186  u32 n_rx_packets = 0;
187  u32 n_rx_bytes = 0;
188  u32 *to_next = 0;
189  u32 block_size = apif->rx_req->tp_block_size;
190  u32 frame_size = apif->rx_req->tp_frame_size;
191  u32 frame_num = apif->rx_req->tp_frame_nr;
192  u8 *block_start = apif->rx_ring + block * block_size;
193  uword n_trace = vlib_get_trace_count (vm, node);
194  u32 thread_index = vm->thread_index;
195  u32 n_buffer_bytes = vlib_buffer_get_default_data_size (vm);
196  u32 min_bufs = apif->rx_req->tp_frame_size / n_buffer_bytes;
197
198  n_free_bufs = vec_len (apm->rx_buffers[thread_index]);
199  if (PREDICT_FALSE (n_free_bufs < VLIB_FRAME_SIZE))
200    {
201      vec_validate (apm->rx_buffers[thread_index],
202		    VLIB_FRAME_SIZE + n_free_bufs - 1);
203      n_free_bufs +=
204	vlib_buffer_alloc (vm, &apm->rx_buffers[thread_index][n_free_bufs],
205			   VLIB_FRAME_SIZE);
206      _vec_len (apm->rx_buffers[thread_index]) = n_free_bufs;
207    }
208
209  rx_frame = apif->next_rx_frame;
210  tph = (struct tpacket2_hdr *) (block_start + rx_frame * frame_size);
211  while ((tph->tp_status & TP_STATUS_USER) && (n_free_bufs > min_bufs))
212    {
213      vlib_buffer_t *b0 = 0, *first_b0 = 0;
214      u32 next0 = next_index;
215
216      u32 n_left_to_next;
217      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
218      while ((tph->tp_status & TP_STATUS_USER) && (n_free_bufs > min_bufs) &&
219	     n_left_to_next)
220	{
221	  u32 data_len = tph->tp_snaplen;
222	  u32 offset = 0;
223	  u32 bi0 = 0, first_bi0 = 0, prev_bi0;
224
225	  while (data_len)
226	    {
227	      /* grab free buffer */
228	      u32 last_empty_buffer =
229		vec_len (apm->rx_buffers[thread_index]) - 1;
230	      prev_bi0 = bi0;
231	      bi0 = apm->rx_buffers[thread_index][last_empty_buffer];
232	      b0 = vlib_get_buffer (vm, bi0);
233	      _vec_len (apm->rx_buffers[thread_index]) = last_empty_buffer;
234	      n_free_bufs--;
235
236	      /* copy data */
237	      u32 bytes_to_copy =
238		data_len > n_buffer_bytes ? n_buffer_bytes : data_len;
239	      u32 vlan_len = 0;
240	      u32 bytes_copied = 0;
241	      b0->current_data = 0;
242	      /* Kernel removes VLAN headers, so reconstruct VLAN */
243	      if (PREDICT_FALSE (tph->tp_status & TP_STATUS_VLAN_VALID))
244		{
245		  if (PREDICT_TRUE (offset == 0))
246		    {
247		      clib_memcpy_fast (vlib_buffer_get_current (b0),
248					(u8 *) tph + tph->tp_mac,
249					sizeof (ethernet_header_t));
250		      ethernet_header_t *eth = vlib_buffer_get_current (b0);
251		      ethernet_vlan_header_t *vlan =
252			(ethernet_vlan_header_t *) (eth + 1);
253		      vlan->priority_cfi_and_id =
254			clib_host_to_net_u16 (tph->tp_vlan_tci);
255		      vlan->type = eth->type;
256		      eth->type = clib_host_to_net_u16 (ETHERNET_TYPE_VLAN);
257		      vlan_len = sizeof (ethernet_vlan_header_t);
258		      bytes_copied = sizeof (ethernet_header_t);
259		    }
260		}
261	      clib_memcpy_fast (((u8 *) vlib_buffer_get_current (b0)) +
262				bytes_copied + vlan_len,
263				(u8 *) tph + tph->tp_mac + offset +
264				bytes_copied, (bytes_to_copy - bytes_copied));
265
266	      /* fill buffer header */
267	      b0->current_length = bytes_to_copy + vlan_len;
268
269	      if (offset == 0)
270		{
271		  b0->total_length_not_including_first_buffer = 0;
272		  b0->flags = VLIB_BUFFER_TOTAL_LENGTH_VALID;
273		  vnet_buffer (b0)->sw_if_index[VLIB_RX] = apif->sw_if_index;
274		  vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;
275		  first_bi0 = bi0;
276		  first_b0 = vlib_get_buffer (vm, first_bi0);
277		  if (tph->tp_status & TP_STATUS_CSUMNOTREADY)
278		    mark_tcp_udp_cksum_calc (first_b0);
279		}
280	      else
281		buffer_add_to_chain (vm, bi0, first_bi0, prev_bi0);
282
283	      offset += bytes_to_copy;
284	      data_len -= bytes_to_copy;
285	    }
286	  n_rx_packets++;
287	  n_rx_bytes += tph->tp_snaplen;
288	  to_next[0] = first_bi0;
289	  to_next += 1;
290	  n_left_to_next--;
291
292	  /* drop partial packets */
293	  if (PREDICT_FALSE (tph->tp_len != tph->tp_snaplen))
294	    {
295	      next0 = VNET_DEVICE_INPUT_NEXT_DROP;
296	      first_b0->error =
297		node->errors[AF_PACKET_INPUT_ERROR_PARTIAL_PKT];
298	    }
299	  else
300	    {
301	      next0 = VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT;
302
303	      if (PREDICT_FALSE (apif->per_interface_next_index != ~0))
304		next0 = apif->per_interface_next_index;
305
306	      /* redirect if feature path enabled */
307	      vnet_feature_start_device_input_x1 (apif->sw_if_index, &next0,
308						  first_b0);
309	    }
310
311	  /* trace */
312	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (first_b0);
313	  if (PREDICT_FALSE (n_trace > 0))
314	    {
315	      af_packet_input_trace_t *tr;
316	      vlib_trace_buffer (vm, node, next0, first_b0,	/* follow_chain */
317				 0);
318	      vlib_set_trace_count (vm, node, --n_trace);
319	      tr = vlib_add_trace (vm, node, first_b0, sizeof (*tr));
320	      tr->next_index = next0;
321	      tr->hw_if_index = apif->hw_if_index;
322	      clib_memcpy_fast (&tr->tph, tph, sizeof (struct tpacket2_hdr));
323	    }
324
325	  /* enque and take next packet */
326	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
327					   n_left_to_next, first_bi0, next0);
328
329	  /* next packet */
330	  tph->tp_status = TP_STATUS_KERNEL;
331	  rx_frame = (rx_frame + 1) % frame_num;
332	  tph = (struct tpacket2_hdr *) (block_start + rx_frame * frame_size);
333	}
334
335      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
336    }
337
338  apif->next_rx_frame = rx_frame;
339
340  vlib_increment_combined_counter
341    (vnet_get_main ()->interface_main.combined_sw_if_counters
342     + VNET_INTERFACE_COUNTER_RX,
343     vlib_get_thread_index (), apif->hw_if_index, n_rx_packets, n_rx_bytes);
344
345  vnet_device_increment_rx_packets (thread_index, n_rx_packets);
346  return n_rx_packets;
347}
348
349VLIB_NODE_FN (af_packet_input_node) (vlib_main_t * vm,
350				     vlib_node_runtime_t * node,
351				     vlib_frame_t * frame)
352{
353  u32 n_rx_packets = 0;
354  af_packet_main_t *apm = &af_packet_main;
355  vnet_device_input_runtime_t *rt = (void *) node->runtime_data;
356  vnet_device_and_queue_t *dq;
357
358  foreach_device_and_queue (dq, rt->devices_and_queues)
359  {
360    af_packet_if_t *apif;
361    apif = vec_elt_at_index (apm->interfaces, dq->dev_instance);
362    if (apif->is_admin_up)
363      n_rx_packets += af_packet_device_input_fn (vm, node, frame, apif);
364  }
365
366  return n_rx_packets;
367}
368
369/* *INDENT-OFF* */
370VLIB_REGISTER_NODE (af_packet_input_node) = {
371  .name = "af-packet-input",
372  .flags = VLIB_NODE_FLAG_TRACE_SUPPORTED,
373  .sibling_of = "device-input",
374  .format_trace = format_af_packet_input_trace,
375  .type = VLIB_NODE_TYPE_INPUT,
376  .state = VLIB_NODE_STATE_INTERRUPT,
377  .n_errors = AF_PACKET_INPUT_N_ERROR,
378  .error_strings = af_packet_input_error_strings,
379};
380/* *INDENT-ON* */
381
382
383/*
384 * fd.io coding-style-patch-verification: ON
385 *
386 * Local Variables:
387 * eval: (c-set-style "gnu")
388 * End:
389 */
390