Line data Source code
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 : #include <sys/socket.h>
22 : #include <sys/ioctl.h>
23 : #include <net/if.h>
24 : #include <net/if_arp.h>
25 :
26 : #include <vlib/vlib.h>
27 : #include <vlib/unix/unix.h>
28 : #include <vnet/ip/ip.h>
29 : #include <vnet/ethernet/ethernet.h>
30 : #include <vnet/ip/ip4_packet.h>
31 : #include <vnet/ip/ip6_packet.h>
32 : #include <vnet/ip/ip_psh_cksum.h>
33 : #include <vnet/tcp/tcp_packet.h>
34 : #include <vnet/udp/udp_packet.h>
35 :
36 : #include <af_packet/af_packet.h>
37 : #include <vnet/devices/virtio/virtio_std.h>
38 : #include <vnet/devices/netlink.h>
39 :
40 : #define foreach_af_packet_tx_func_error \
41 : _(FRAME_NOT_READY, "tx frame not ready") \
42 : _(TXRING_EAGAIN, "tx sendto temporary failure") \
43 : _(TXRING_FATAL, "tx sendto fatal failure") \
44 : _(TXRING_OVERRUN, "tx ring overrun")
45 :
46 : typedef enum
47 : {
48 : #define _(f,s) AF_PACKET_TX_ERROR_##f,
49 : foreach_af_packet_tx_func_error
50 : #undef _
51 : AF_PACKET_TX_N_ERROR,
52 : } af_packet_tx_func_error_t;
53 :
54 : static char *af_packet_tx_func_error_strings[] = {
55 : #define _(n,s) s,
56 : foreach_af_packet_tx_func_error
57 : #undef _
58 : };
59 :
60 : typedef struct
61 : {
62 : u32 buffer_index;
63 : u32 hw_if_index;
64 : u16 queue_id;
65 : u8 is_v2;
66 : union
67 : {
68 : tpacket2_hdr_t tph2;
69 : tpacket3_hdr_t tph3;
70 : };
71 : vnet_virtio_net_hdr_t vnet_hdr;
72 : vlib_buffer_t buffer;
73 : } af_packet_tx_trace_t;
74 :
75 : #ifndef CLIB_MARCH_VARIANT
76 : u8 *
77 504 : format_af_packet_device_name (u8 * s, va_list * args)
78 : {
79 504 : u32 i = va_arg (*args, u32);
80 504 : af_packet_main_t *apm = &af_packet_main;
81 504 : af_packet_if_t *apif = pool_elt_at_index (apm->interfaces, i);
82 :
83 504 : s = format (s, "host-%s", apif->host_if_name);
84 504 : return s;
85 : }
86 : #endif /* CLIB_MARCH_VARIANT */
87 :
88 : static u8 *
89 0 : format_af_packet_device (u8 * s, va_list * args)
90 : {
91 0 : u32 dev_instance = va_arg (*args, u32);
92 0 : u32 indent = format_get_indent (s);
93 0 : int __clib_unused verbose = va_arg (*args, int);
94 :
95 0 : af_packet_main_t *apm = &af_packet_main;
96 0 : af_packet_if_t *apif = pool_elt_at_index (apm->interfaces, dev_instance);
97 0 : af_packet_queue_t *rx_queue = 0;
98 0 : af_packet_queue_t *tx_queue = 0;
99 :
100 0 : s = format (s, "Linux PACKET socket interface %s",
101 0 : (apif->version == TPACKET_V2) ? "v2" : "v3");
102 0 : s = format (s, "\n%UFEATURES:", format_white_space, indent);
103 0 : if (apif->is_qdisc_bypass_enabled)
104 0 : s = format (s, "\n%Uqdisc-bpass-enabled", format_white_space, indent + 2);
105 0 : if (apif->is_cksum_gso_enabled)
106 0 : s = format (s, "\n%Ucksum-gso-enabled", format_white_space, indent + 2);
107 0 : if (apif->is_fanout_enabled)
108 0 : s = format (s, "\n%Ufanout-enabled", format_white_space, indent + 2);
109 :
110 0 : vec_foreach (rx_queue, apif->rx_queues)
111 : {
112 0 : u32 rx_block_size = rx_queue->rx_req->req.tp_block_size;
113 0 : u32 rx_frame_size = rx_queue->rx_req->req.tp_frame_size;
114 0 : u32 rx_frame_nr = rx_queue->rx_req->req.tp_frame_nr;
115 0 : u32 rx_block_nr = rx_queue->rx_req->req.tp_block_nr;
116 :
117 0 : s = format (s, "\n%URX Queue %u:", format_white_space, indent,
118 0 : rx_queue->queue_id);
119 0 : s = format (s, "\n%Ublock size:%d nr:%d frame size:%d nr:%d",
120 : format_white_space, indent + 2, rx_block_size, rx_block_nr,
121 : rx_frame_size, rx_frame_nr);
122 0 : if (apif->version == TPACKET_V2)
123 0 : s = format (s, " next frame:%d", rx_queue->next_rx_frame);
124 : else
125 0 : s = format (s, " next block:%d", rx_queue->next_rx_block);
126 0 : if (rx_queue->is_rx_pending)
127 : {
128 0 : s = format (
129 : s, "\n%UPending Request: num-rx-pkts:%d next-frame-offset:%d",
130 0 : format_white_space, indent + 2, rx_queue->num_rx_pkts,
131 : rx_queue->rx_frame_offset);
132 : }
133 : }
134 :
135 0 : vec_foreach (tx_queue, apif->tx_queues)
136 : {
137 0 : clib_spinlock_lock (&tx_queue->lockp);
138 0 : u32 tx_block_sz = tx_queue->tx_req->req.tp_block_size;
139 0 : u32 tx_frame_sz = tx_queue->tx_req->req.tp_frame_size;
140 0 : u32 tx_frame_nr = tx_queue->tx_req->req.tp_frame_nr;
141 0 : u32 tx_block_nr = tx_queue->tx_req->req.tp_block_nr;
142 0 : int block = 0;
143 0 : int n_send_req = 0, n_avail = 0, n_sending = 0, n_tot = 0, n_wrong = 0;
144 0 : u8 *tx_block_start = tx_queue->tx_ring[block];
145 0 : u32 tx_frame = tx_queue->next_tx_frame;
146 : tpacket3_hdr_t *tph3;
147 : tpacket2_hdr_t *tph2;
148 :
149 0 : s = format (s, "\n%UTX Queue %u:", format_white_space, indent,
150 0 : tx_queue->queue_id);
151 0 : s = format (s, "\n%Ublock size:%d nr:%d frame size:%d nr:%d",
152 : format_white_space, indent + 2, tx_block_sz, tx_block_nr,
153 : tx_frame_sz, tx_frame_nr);
154 0 : s = format (s, " next frame:%d", tx_queue->next_tx_frame);
155 0 : if (apif->version & TPACKET_V3)
156 : do
157 : {
158 0 : tph3 =
159 0 : (tpacket3_hdr_t *) (tx_block_start + tx_frame * tx_frame_sz);
160 0 : tx_frame = (tx_frame + 1) % tx_frame_nr;
161 0 : if (tph3->tp_status == 0)
162 0 : n_avail++;
163 0 : else if (tph3->tp_status & TP_STATUS_SEND_REQUEST)
164 0 : n_send_req++;
165 0 : else if (tph3->tp_status & TP_STATUS_SENDING)
166 0 : n_sending++;
167 : else
168 0 : n_wrong++;
169 0 : n_tot++;
170 : }
171 0 : while (tx_frame != tx_queue->next_tx_frame);
172 : else
173 : do
174 : {
175 0 : tph2 =
176 0 : (tpacket2_hdr_t *) (tx_block_start + tx_frame * tx_frame_sz);
177 0 : tx_frame = (tx_frame + 1) % tx_frame_nr;
178 0 : if (tph2->tp_status == 0)
179 0 : n_avail++;
180 0 : else if (tph2->tp_status & TP_STATUS_SEND_REQUEST)
181 0 : n_send_req++;
182 0 : else if (tph2->tp_status & TP_STATUS_SENDING)
183 0 : n_sending++;
184 : else
185 0 : n_wrong++;
186 0 : n_tot++;
187 : }
188 0 : while (tx_frame != tx_queue->next_tx_frame);
189 : s =
190 0 : format (s, "\n%Uavailable:%d request:%d sending:%d wrong:%d total:%d",
191 : format_white_space, indent + 2, n_avail, n_send_req, n_sending,
192 : n_wrong, n_tot);
193 0 : clib_spinlock_unlock (&tx_queue->lockp);
194 : }
195 0 : return s;
196 : }
197 :
198 : static u8 *
199 0 : format_af_packet_tx_trace (u8 *s, va_list *va)
200 : {
201 0 : CLIB_UNUSED (vlib_main_t * vm) = va_arg (*va, vlib_main_t *);
202 0 : CLIB_UNUSED (vlib_node_t * node) = va_arg (*va, vlib_node_t *);
203 0 : af_packet_tx_trace_t *t = va_arg (*va, af_packet_tx_trace_t *);
204 0 : u32 indent = format_get_indent (s);
205 :
206 0 : s = format (s, "af_packet: hw_if_index %u tx-queue %u", t->hw_if_index,
207 0 : t->queue_id);
208 :
209 0 : if (t->is_v2)
210 : {
211 0 : s = format (
212 : s,
213 : "\n%Utpacket2_hdr:\n%Ustatus 0x%x len %u snaplen %u mac %u net %u"
214 : "\n%Usec 0x%x nsec 0x%x vlan %U"
215 : #ifdef TP_STATUS_VLAN_TPID_VALID
216 : " vlan_tpid %u"
217 : #endif
218 : ,
219 : format_white_space, indent + 2, format_white_space, indent + 4,
220 0 : t->tph2.tp_status, t->tph2.tp_len, t->tph2.tp_snaplen, t->tph2.tp_mac,
221 0 : t->tph2.tp_net, format_white_space, indent + 4, t->tph2.tp_sec,
222 0 : t->tph2.tp_nsec, format_ethernet_vlan_tci, t->tph2.tp_vlan_tci
223 : #ifdef TP_STATUS_VLAN_TPID_VALID
224 : ,
225 0 : t->tph2.tp_vlan_tpid
226 : #endif
227 : );
228 : }
229 : else
230 : {
231 0 : s = format (
232 : s,
233 : "\n%Utpacket3_hdr:\n%Ustatus 0x%x len %u snaplen %u mac %u net %u"
234 : "\n%Usec 0x%x nsec 0x%x vlan %U"
235 : #ifdef TP_STATUS_VLAN_TPID_VALID
236 : " vlan_tpid %u"
237 : #endif
238 : ,
239 : format_white_space, indent + 2, format_white_space, indent + 4,
240 0 : t->tph3.tp_status, t->tph3.tp_len, t->tph3.tp_snaplen, t->tph3.tp_mac,
241 0 : t->tph3.tp_net, format_white_space, indent + 4, t->tph3.tp_sec,
242 : t->tph3.tp_nsec, format_ethernet_vlan_tci, t->tph3.hv1.tp_vlan_tci
243 : #ifdef TP_STATUS_VLAN_TPID_VALID
244 : ,
245 0 : t->tph3.hv1.tp_vlan_tpid
246 : #endif
247 : );
248 : }
249 0 : s = format (s,
250 : "\n%Uvnet-hdr:\n%Uflags 0x%02x gso_type 0x%02x hdr_len %u"
251 : "\n%Ugso_size %u csum_start %u csum_offset %u",
252 : format_white_space, indent + 2, format_white_space, indent + 4,
253 0 : t->vnet_hdr.flags, t->vnet_hdr.gso_type, t->vnet_hdr.hdr_len,
254 0 : format_white_space, indent + 4, t->vnet_hdr.gso_size,
255 0 : t->vnet_hdr.csum_start, t->vnet_hdr.csum_offset);
256 :
257 0 : s = format (s, "\n%Ubuffer 0x%x:\n%U%U", format_white_space, indent + 2,
258 : t->buffer_index, format_white_space, indent + 4,
259 : format_vnet_buffer_no_chain, &t->buffer);
260 0 : s = format (s, "\n%U%U", format_white_space, indent + 2,
261 0 : format_ethernet_header_with_length, t->buffer.pre_data,
262 : sizeof (t->buffer.pre_data));
263 0 : return s;
264 : }
265 :
266 : static void
267 0 : af_packet_tx_trace (vlib_main_t *vm, vlib_node_runtime_t *node,
268 : vlib_buffer_t *b0, u32 bi, void *tph,
269 : vnet_virtio_net_hdr_t *vnet_hdr, u32 hw_if_index,
270 : u16 queue_id, u8 is_v2)
271 : {
272 : af_packet_tx_trace_t *t;
273 0 : t = vlib_add_trace (vm, node, b0, sizeof (t[0]));
274 0 : t->hw_if_index = hw_if_index;
275 0 : t->queue_id = queue_id;
276 0 : t->buffer_index = bi;
277 0 : t->is_v2 = is_v2;
278 :
279 0 : if (is_v2)
280 0 : clib_memcpy_fast (&t->tph2, (tpacket2_hdr_t *) tph,
281 : sizeof (tpacket2_hdr_t));
282 : else
283 0 : clib_memcpy_fast (&t->tph3, (tpacket3_hdr_t *) tph,
284 : sizeof (tpacket3_hdr_t));
285 0 : clib_memcpy_fast (&t->vnet_hdr, vnet_hdr, sizeof (*vnet_hdr));
286 0 : clib_memcpy_fast (&t->buffer, b0, sizeof (*b0) - sizeof (b0->pre_data));
287 0 : clib_memcpy_fast (t->buffer.pre_data, vlib_buffer_get_current (b0),
288 : sizeof (t->buffer.pre_data));
289 0 : }
290 :
291 : static_always_inline void
292 697292 : fill_gso_offload (vlib_buffer_t *b0, vnet_virtio_net_hdr_t *vnet_hdr)
293 : {
294 697292 : vnet_buffer_oflags_t oflags = vnet_buffer (b0)->oflags;
295 697292 : i16 l4_hdr_offset = vnet_buffer (b0)->l4_hdr_offset - b0->current_data;
296 697292 : if (b0->flags & VNET_BUFFER_F_IS_IP4)
297 : {
298 : ip4_header_t *ip4;
299 374977 : vnet_hdr->gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
300 374977 : vnet_hdr->gso_size = vnet_buffer2 (b0)->gso_size;
301 374977 : vnet_hdr->hdr_len = l4_hdr_offset + vnet_buffer2 (b0)->gso_l4_hdr_sz;
302 374977 : vnet_hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
303 374977 : vnet_hdr->csum_start = l4_hdr_offset; // 0x22;
304 374977 : vnet_hdr->csum_offset = STRUCT_OFFSET_OF (tcp_header_t, checksum);
305 374977 : ip4 = (ip4_header_t *) (b0->data + vnet_buffer (b0)->l3_hdr_offset);
306 374977 : if (oflags & VNET_BUFFER_OFFLOAD_F_IP_CKSUM)
307 0 : ip4->checksum = ip4_header_checksum (ip4);
308 374977 : tcp_header_t *tcp =
309 374977 : (tcp_header_t *) (b0->data + vnet_buffer (b0)->l4_hdr_offset);
310 374977 : tcp->checksum = ip4_pseudo_header_cksum (ip4);
311 : }
312 322315 : else if (b0->flags & VNET_BUFFER_F_IS_IP6)
313 : {
314 : ip6_header_t *ip6;
315 322315 : vnet_hdr->gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
316 322315 : vnet_hdr->gso_size = vnet_buffer2 (b0)->gso_size;
317 322315 : vnet_hdr->hdr_len = l4_hdr_offset + vnet_buffer2 (b0)->gso_l4_hdr_sz;
318 322315 : vnet_hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
319 322315 : vnet_hdr->csum_start = l4_hdr_offset; // 0x36;
320 322315 : vnet_hdr->csum_offset = STRUCT_OFFSET_OF (tcp_header_t, checksum);
321 322315 : ip6 = (ip6_header_t *) (b0->data + vnet_buffer (b0)->l3_hdr_offset);
322 322315 : tcp_header_t *tcp =
323 322315 : (tcp_header_t *) (b0->data + vnet_buffer (b0)->l4_hdr_offset);
324 322315 : tcp->checksum = ip6_pseudo_header_cksum (ip6);
325 : }
326 697292 : }
327 :
328 : static_always_inline void
329 623548 : fill_cksum_offload (vlib_buffer_t *b0, vnet_virtio_net_hdr_t *vnet_hdr)
330 : {
331 623548 : vnet_buffer_oflags_t oflags = vnet_buffer (b0)->oflags;
332 623548 : i16 l4_hdr_offset = vnet_buffer (b0)->l4_hdr_offset - b0->current_data;
333 623548 : if (b0->flags & VNET_BUFFER_F_IS_IP4)
334 : {
335 : ip4_header_t *ip4;
336 338263 : ip4 = (ip4_header_t *) (b0->data + vnet_buffer (b0)->l3_hdr_offset);
337 338263 : if (oflags & VNET_BUFFER_OFFLOAD_F_IP_CKSUM)
338 0 : ip4->checksum = ip4_header_checksum (ip4);
339 338263 : vnet_hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
340 338263 : vnet_hdr->csum_start = l4_hdr_offset;
341 338263 : if (oflags & VNET_BUFFER_OFFLOAD_F_TCP_CKSUM)
342 : {
343 338263 : tcp_header_t *tcp =
344 338263 : (tcp_header_t *) (b0->data + vnet_buffer (b0)->l4_hdr_offset);
345 338263 : tcp->checksum = ip4_pseudo_header_cksum (ip4);
346 338263 : vnet_hdr->csum_offset = STRUCT_OFFSET_OF (tcp_header_t, checksum);
347 338263 : vnet_hdr->hdr_len = l4_hdr_offset + tcp_header_bytes (tcp);
348 : }
349 0 : else if (oflags & VNET_BUFFER_OFFLOAD_F_UDP_CKSUM)
350 : {
351 0 : udp_header_t *udp =
352 0 : (udp_header_t *) (b0->data + vnet_buffer (b0)->l4_hdr_offset);
353 0 : udp->checksum = ip4_pseudo_header_cksum (ip4);
354 0 : vnet_hdr->csum_offset = STRUCT_OFFSET_OF (udp_header_t, checksum);
355 0 : vnet_hdr->hdr_len = l4_hdr_offset + sizeof (udp_header_t);
356 : }
357 : }
358 285285 : else if (b0->flags & VNET_BUFFER_F_IS_IP6)
359 : {
360 : ip6_header_t *ip6;
361 285285 : vnet_hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
362 285285 : vnet_hdr->csum_start = l4_hdr_offset;
363 285285 : ip6 = (ip6_header_t *) (b0->data + vnet_buffer (b0)->l3_hdr_offset);
364 285285 : if (oflags & VNET_BUFFER_OFFLOAD_F_TCP_CKSUM)
365 : {
366 285285 : tcp_header_t *tcp =
367 285285 : (tcp_header_t *) (b0->data + vnet_buffer (b0)->l4_hdr_offset);
368 285285 : tcp->checksum = ip6_pseudo_header_cksum (ip6);
369 285285 : vnet_hdr->csum_offset = STRUCT_OFFSET_OF (tcp_header_t, checksum);
370 285285 : vnet_hdr->hdr_len = l4_hdr_offset + tcp_header_bytes (tcp);
371 : }
372 0 : else if (oflags & VNET_BUFFER_OFFLOAD_F_UDP_CKSUM)
373 : {
374 0 : udp_header_t *udp =
375 0 : (udp_header_t *) (b0->data + vnet_buffer (b0)->l4_hdr_offset);
376 0 : udp->checksum = ip6_pseudo_header_cksum (ip6);
377 0 : vnet_hdr->csum_offset = STRUCT_OFFSET_OF (udp_header_t, checksum);
378 0 : vnet_hdr->hdr_len = l4_hdr_offset + sizeof (udp_header_t);
379 : }
380 : }
381 623548 : }
382 :
383 817322 : VNET_DEVICE_CLASS_TX_FN (af_packet_device_class) (vlib_main_t * vm,
384 : vlib_node_runtime_t * node,
385 : vlib_frame_t * frame)
386 : {
387 815086 : af_packet_main_t *apm = &af_packet_main;
388 815086 : vnet_hw_if_tx_frame_t *tf = vlib_frame_scalar_args (frame);
389 815086 : u32 *buffers = vlib_frame_vector_args (frame);
390 815086 : u32 n_left = frame->n_vectors;
391 815086 : u32 n_sent = 0;
392 815086 : vnet_interface_output_runtime_t *rd = (void *) node->runtime_data;
393 815086 : af_packet_if_t *apif =
394 815086 : pool_elt_at_index (apm->interfaces, rd->dev_instance);
395 815086 : u16 queue_id = tf->queue_id;
396 815086 : af_packet_queue_t *tx_queue = vec_elt_at_index (apif->tx_queues, queue_id);
397 815086 : u32 block = 0, frame_size = 0, frame_num = 0, tx_frame = 0;
398 815086 : u8 *block_start = 0;
399 815086 : tpacket3_hdr_t *tph3 = 0;
400 815086 : tpacket2_hdr_t *tph2 = 0;
401 815086 : u32 frame_not_ready = 0;
402 815086 : u8 is_cksum_gso_enabled = (apif->is_cksum_gso_enabled == 1) ? 1 : 0;
403 815086 : u32 tpacket_align = 0;
404 815086 : u8 is_v2 = (apif->version == TPACKET_V2) ? 1 : 0;
405 :
406 815086 : if (tf->shared_queue)
407 0 : clib_spinlock_lock (&tx_queue->lockp);
408 :
409 815086 : frame_size = tx_queue->tx_req->req.tp_frame_size;
410 815086 : frame_num = tx_queue->tx_req->req.tp_frame_nr;
411 815086 : block_start = tx_queue->tx_ring[block];
412 815086 : tx_frame = tx_queue->next_tx_frame;
413 815086 : if (is_v2)
414 : {
415 271212 : tpacket_align = TPACKET_ALIGN (sizeof (tpacket2_hdr_t));
416 6264150 : while (n_left)
417 : {
418 : u32 len;
419 5992940 : vnet_virtio_net_hdr_t *vnet_hdr = 0;
420 5992940 : u32 offset = 0;
421 5992940 : vlib_buffer_t *b0 = 0, *b0_first = 0;
422 : u32 bi, bi_first;
423 :
424 5992940 : bi = bi_first = buffers[0];
425 5992940 : n_left--;
426 5992940 : buffers++;
427 :
428 5992940 : tph2 = (tpacket2_hdr_t *) (block_start + tx_frame * frame_size);
429 5992940 : if (PREDICT_FALSE (tph2->tp_status &
430 : (TP_STATUS_SEND_REQUEST | TP_STATUS_SENDING)))
431 : {
432 0 : frame_not_ready++;
433 0 : goto nextv2;
434 : }
435 :
436 5992940 : b0_first = b0 = vlib_get_buffer (vm, bi);
437 :
438 5992940 : if (PREDICT_TRUE (is_cksum_gso_enabled))
439 : {
440 3968290 : vnet_hdr =
441 3968290 : (vnet_virtio_net_hdr_t *) ((u8 *) tph2 + tpacket_align);
442 :
443 3968290 : clib_memset_u8 (vnet_hdr, 0, sizeof (vnet_virtio_net_hdr_t));
444 3968290 : offset = sizeof (vnet_virtio_net_hdr_t);
445 :
446 3968290 : if (b0->flags & VNET_BUFFER_F_GSO)
447 516765 : fill_gso_offload (b0, vnet_hdr);
448 3451530 : else if (b0->flags & VNET_BUFFER_F_OFFLOAD)
449 411179 : fill_cksum_offload (b0, vnet_hdr);
450 : }
451 :
452 5992940 : len = b0->current_length;
453 5992940 : clib_memcpy_fast ((u8 *) tph2 + tpacket_align + offset,
454 5992940 : vlib_buffer_get_current (b0), len);
455 5992940 : offset += len;
456 :
457 24344300 : while (b0->flags & VLIB_BUFFER_NEXT_PRESENT)
458 : {
459 18351300 : b0 = vlib_get_buffer (vm, b0->next_buffer);
460 18351300 : len = b0->current_length;
461 18351300 : clib_memcpy_fast ((u8 *) tph2 + tpacket_align + offset,
462 18351300 : vlib_buffer_get_current (b0), len);
463 18351300 : offset += len;
464 : }
465 :
466 5992940 : tph2->tp_len = tph2->tp_snaplen = offset;
467 5992940 : tph2->tp_status = TP_STATUS_SEND_REQUEST;
468 5992940 : n_sent++;
469 :
470 5992940 : if (PREDICT_FALSE (b0_first->flags & VLIB_BUFFER_IS_TRACED))
471 : {
472 0 : if (PREDICT_TRUE (is_cksum_gso_enabled))
473 0 : af_packet_tx_trace (vm, node, b0_first, bi_first, tph2,
474 : vnet_hdr, apif->hw_if_index, queue_id, 1);
475 : else
476 : {
477 0 : vnet_virtio_net_hdr_t vnet_hdr2 = {};
478 0 : af_packet_tx_trace (vm, node, b0_first, bi_first, tph2,
479 : &vnet_hdr2, apif->hw_if_index, queue_id,
480 : 1);
481 : }
482 : }
483 5992940 : tx_frame = (tx_frame + 1) % frame_num;
484 :
485 5992940 : nextv2:
486 : /* check if we've exhausted the ring */
487 5992940 : if (PREDICT_FALSE (frame_not_ready + n_sent == frame_num))
488 0 : break;
489 : }
490 : }
491 : else
492 : {
493 543874 : tpacket_align = TPACKET_ALIGN (sizeof (tpacket3_hdr_t));
494 :
495 9495590 : while (n_left)
496 : {
497 : u32 len;
498 8951710 : vnet_virtio_net_hdr_t *vnet_hdr = 0;
499 8951710 : u32 offset = 0;
500 8951710 : vlib_buffer_t *b0 = 0, *b0_first = 0;
501 : u32 bi, bi_first;
502 :
503 8951710 : bi = bi_first = buffers[0];
504 8951710 : n_left--;
505 8951710 : buffers++;
506 :
507 8951710 : tph3 = (tpacket3_hdr_t *) (block_start + tx_frame * frame_size);
508 8951710 : if (PREDICT_FALSE (tph3->tp_status &
509 : (TP_STATUS_SEND_REQUEST | TP_STATUS_SENDING)))
510 : {
511 0 : frame_not_ready++;
512 0 : goto nextv3;
513 : }
514 :
515 8951710 : b0_first = b0 = vlib_get_buffer (vm, bi);
516 :
517 8951710 : if (PREDICT_TRUE (is_cksum_gso_enabled))
518 : {
519 1163250 : vnet_hdr =
520 1163250 : (vnet_virtio_net_hdr_t *) ((u8 *) tph3 + tpacket_align);
521 :
522 1163250 : clib_memset_u8 (vnet_hdr, 0, sizeof (vnet_virtio_net_hdr_t));
523 1163250 : offset = sizeof (vnet_virtio_net_hdr_t);
524 :
525 1163250 : if (b0->flags & VNET_BUFFER_F_GSO)
526 180527 : fill_gso_offload (b0, vnet_hdr);
527 982725 : else if (b0->flags & VNET_BUFFER_F_OFFLOAD)
528 212369 : fill_cksum_offload (b0, vnet_hdr);
529 : }
530 :
531 8951710 : len = b0->current_length;
532 8951710 : clib_memcpy_fast ((u8 *) tph3 + tpacket_align + offset,
533 8951710 : vlib_buffer_get_current (b0), len);
534 8951710 : offset += len;
535 :
536 17471800 : while (b0->flags & VLIB_BUFFER_NEXT_PRESENT)
537 : {
538 8520090 : b0 = vlib_get_buffer (vm, b0->next_buffer);
539 8520090 : len = b0->current_length;
540 8520090 : clib_memcpy_fast ((u8 *) tph3 + tpacket_align + offset,
541 8520090 : vlib_buffer_get_current (b0), len);
542 8520090 : offset += len;
543 : }
544 :
545 8951710 : tph3->tp_len = tph3->tp_snaplen = offset;
546 8951710 : tph3->tp_status = TP_STATUS_SEND_REQUEST;
547 8951710 : n_sent++;
548 :
549 8951710 : if (PREDICT_FALSE (b0_first->flags & VLIB_BUFFER_IS_TRACED))
550 : {
551 0 : if (PREDICT_TRUE (is_cksum_gso_enabled))
552 0 : af_packet_tx_trace (vm, node, b0_first, bi_first, tph3,
553 : vnet_hdr, apif->hw_if_index, queue_id, 0);
554 : else
555 : {
556 0 : vnet_virtio_net_hdr_t vnet_hdr2 = {};
557 0 : af_packet_tx_trace (vm, node, b0_first, bi_first, tph3,
558 : &vnet_hdr2, apif->hw_if_index, queue_id,
559 : 0);
560 : }
561 : }
562 8951710 : tx_frame = (tx_frame + 1) % frame_num;
563 :
564 8951710 : nextv3:
565 : /* check if we've exhausted the ring */
566 8951710 : if (PREDICT_FALSE (frame_not_ready + n_sent == frame_num))
567 0 : break;
568 : }
569 : }
570 815086 : CLIB_MEMORY_BARRIER ();
571 :
572 815086 : if (PREDICT_TRUE (n_sent || tx_queue->is_tx_pending))
573 : {
574 815086 : tx_queue->next_tx_frame = tx_frame;
575 815086 : tx_queue->is_tx_pending = 0;
576 :
577 815086 : if (PREDICT_FALSE (
578 : sendto (tx_queue->fd, NULL, 0, MSG_DONTWAIT, NULL, 0) == -1))
579 : {
580 : /* Uh-oh, drop & move on, but count whether it was fatal or not.
581 : * Note that we have no reliable way to properly determine the
582 : * disposition of the packets we just enqueued for delivery.
583 : */
584 : uword counter;
585 :
586 0 : if (unix_error_is_fatal (errno))
587 : {
588 0 : counter = AF_PACKET_TX_ERROR_TXRING_FATAL;
589 : }
590 : else
591 : {
592 0 : counter = AF_PACKET_TX_ERROR_TXRING_EAGAIN;
593 : /* non-fatal error: kick again next time
594 : * note that you could still end up in a deadlock: if you do not
595 : * try to send new packets (ie reschedule this tx node), eg.
596 : * because your peer is waiting for the unsent packets to reply
597 : * to you but your waiting for its reply etc., you are not going
598 : * to kick again, and everybody is waiting for the other to talk
599 : * 1st... */
600 0 : tx_queue->is_tx_pending = 1;
601 : }
602 :
603 0 : vlib_error_count (vm, node->node_index, counter, 1);
604 : }
605 : }
606 :
607 815086 : if (tf->shared_queue)
608 0 : clib_spinlock_unlock (&tx_queue->lockp);
609 :
610 815086 : if (PREDICT_FALSE (frame_not_ready))
611 0 : vlib_error_count (vm, node->node_index,
612 : AF_PACKET_TX_ERROR_FRAME_NOT_READY, frame_not_ready);
613 :
614 815086 : if (PREDICT_FALSE (frame_not_ready + n_sent == frame_num))
615 0 : vlib_error_count (vm, node->node_index, AF_PACKET_TX_ERROR_TXRING_OVERRUN,
616 : n_left);
617 :
618 815086 : vlib_buffer_free (vm, vlib_frame_vector_args (frame), frame->n_vectors);
619 815086 : return frame->n_vectors;
620 : }
621 :
622 : static void
623 0 : af_packet_set_interface_next_node (vnet_main_t * vnm, u32 hw_if_index,
624 : u32 node_index)
625 : {
626 0 : af_packet_main_t *apm = &af_packet_main;
627 0 : vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index);
628 0 : af_packet_if_t *apif =
629 0 : pool_elt_at_index (apm->interfaces, hw->dev_instance);
630 :
631 : /* Shut off redirection */
632 0 : if (node_index == ~0)
633 : {
634 0 : apif->per_interface_next_index = node_index;
635 0 : return;
636 : }
637 :
638 0 : apif->per_interface_next_index =
639 0 : vlib_node_add_next (vlib_get_main (), af_packet_input_node.index,
640 : node_index);
641 : }
642 :
643 : static void
644 0 : af_packet_clear_hw_interface_counters (u32 instance)
645 : {
646 : /* Nothing for now */
647 0 : }
648 :
649 : static clib_error_t *
650 504 : af_packet_interface_admin_up_down (vnet_main_t * vnm, u32 hw_if_index,
651 : u32 flags)
652 : {
653 504 : af_packet_main_t *apm = &af_packet_main;
654 504 : vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index);
655 504 : af_packet_if_t *apif =
656 504 : pool_elt_at_index (apm->interfaces, hw->dev_instance);
657 : u32 hw_flags;
658 :
659 504 : if (apif->host_if_index < 0)
660 0 : return 0; /* no error */
661 :
662 504 : apif->is_admin_up = (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) != 0;
663 :
664 504 : if (apif->is_admin_up)
665 : {
666 252 : hw_flags = VNET_HW_INTERFACE_FLAG_LINK_UP;
667 252 : vnet_netlink_set_link_state (apif->host_if_index, 1);
668 : }
669 : else
670 : {
671 252 : hw_flags = 0;
672 252 : vnet_netlink_set_link_state (apif->host_if_index, 0);
673 : }
674 :
675 504 : vnet_hw_interface_set_flags (vnm, hw_if_index, hw_flags);
676 :
677 504 : return 0; /* no error */
678 : }
679 :
680 : static clib_error_t *
681 0 : af_packet_subif_add_del_function (vnet_main_t * vnm,
682 : u32 hw_if_index,
683 : struct vnet_sw_interface_t *st, int is_add)
684 : {
685 : /* Nothing for now */
686 0 : return 0;
687 : }
688 :
689 0 : static clib_error_t *af_packet_set_mac_address_function
690 : (struct vnet_hw_interface_t *hi, const u8 * old_address, const u8 * address)
691 : {
692 0 : af_packet_main_t *apm = &af_packet_main;
693 0 : af_packet_if_t *apif =
694 0 : pool_elt_at_index (apm->interfaces, hi->dev_instance);
695 : int rv, fd;
696 : struct ifreq ifr;
697 :
698 0 : if (apif->mode == AF_PACKET_IF_MODE_IP)
699 : {
700 0 : vlib_log_warn (apm->log_class, "af_packet_%s interface is in IP mode",
701 : apif->host_if_name);
702 0 : return clib_error_return (0,
703 : " MAC update failed, interface is in IP mode");
704 : }
705 :
706 0 : fd = socket (AF_UNIX, SOCK_DGRAM, 0);
707 0 : if (0 > fd)
708 : {
709 0 : vlib_log_warn (apm->log_class, "af_packet_%s could not open socket",
710 : apif->host_if_name);
711 0 : return 0;
712 : }
713 :
714 : /* if interface is a bridge ignore */
715 0 : if (apif->host_if_index < 0)
716 0 : goto error; /* no error */
717 :
718 : /* use host_if_index in case host name has changed */
719 0 : ifr.ifr_ifindex = apif->host_if_index;
720 0 : if ((rv = ioctl (fd, SIOCGIFNAME, &ifr)) < 0)
721 : {
722 0 : vlib_log_warn
723 : (apm->log_class,
724 : "af_packet_%s ioctl could not retrieve eth name, error: %d",
725 : apif->host_if_name, rv);
726 0 : goto error;
727 : }
728 :
729 0 : clib_memcpy (ifr.ifr_hwaddr.sa_data, address, 6);
730 0 : ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
731 :
732 0 : if ((rv = ioctl (fd, SIOCSIFHWADDR, &ifr)) < 0)
733 : {
734 0 : vlib_log_warn (apm->log_class,
735 : "af_packet_%s ioctl could not set mac, error: %d",
736 : apif->host_if_name, rv);
737 0 : goto error;
738 : }
739 :
740 0 : error:
741 :
742 0 : if (0 <= fd)
743 0 : close (fd);
744 :
745 0 : return 0; /* no error */
746 : }
747 :
748 : static clib_error_t *
749 252 : af_packet_interface_rx_mode_change (vnet_main_t *vnm, u32 hw_if_index, u32 qid,
750 : vnet_hw_if_rx_mode mode)
751 : {
752 252 : af_packet_main_t *apm = &af_packet_main;
753 252 : vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index);
754 : af_packet_if_t *apif;
755 :
756 252 : apif = vec_elt_at_index (apm->interfaces, hw->dev_instance);
757 :
758 252 : if (mode == VNET_HW_IF_RX_MODE_ADAPTIVE)
759 : {
760 0 : vlib_log_err (apm->log_class,
761 : "af_packet_%s adaptive mode is not supported",
762 : apif->host_if_name);
763 0 : return clib_error_return (
764 : 0, "af_packet_%s adaptive mode is not supported", apif->host_if_name);
765 : }
766 :
767 252 : af_packet_queue_t *rx_queue = vec_elt_at_index (apif->rx_queues, qid);
768 :
769 252 : if (rx_queue->mode != mode)
770 : {
771 252 : rx_queue->mode = mode;
772 :
773 252 : if (mode == VNET_HW_IF_RX_MODE_POLLING)
774 0 : apm->polling_count++;
775 252 : else if (mode == VNET_HW_IF_RX_MODE_INTERRUPT && apm->polling_count > 0)
776 0 : apm->polling_count--;
777 : }
778 :
779 252 : return 0;
780 : }
781 :
782 10639 : VNET_DEVICE_CLASS (af_packet_device_class) = {
783 : .name = "af-packet",
784 : .format_device_name = format_af_packet_device_name,
785 : .format_device = format_af_packet_device,
786 : .format_tx_trace = format_af_packet_tx_trace,
787 : .tx_function_n_errors = AF_PACKET_TX_N_ERROR,
788 : .tx_function_error_strings = af_packet_tx_func_error_strings,
789 : .rx_redirect_to_node = af_packet_set_interface_next_node,
790 : .clear_counters = af_packet_clear_hw_interface_counters,
791 : .admin_up_down_function = af_packet_interface_admin_up_down,
792 : .subif_add_del_function = af_packet_subif_add_del_function,
793 : .mac_addr_change_function = af_packet_set_mac_address_function,
794 : .rx_mode_change_function = af_packet_interface_rx_mode_change,
795 : };
796 :
797 : /*
798 : * fd.io coding-style-patch-verification: ON
799 : *
800 : * Local Variables:
801 : * eval: (c-set-style "gnu")
802 : * End:
803 : */
|
