1 // Copyright 2017 The Fuchsia Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
4
5 #include <arpa/inet.h>
6 #include <netinet/in.h>
7 #include <sys/socket.h>
8 #include <sys/stat.h>
9 #include <sys/time.h>
10
11 #include <fcntl.h>
12 #include <inttypes.h>
13 #include <stdbool.h>
14 #include <stdio.h>
15 #include <stdlib.h>
16 #include <string.h>
17 #include <time.h>
18 #include <unistd.h>
19
20 #include <errno.h>
21 #include <stdint.h>
22
23 #include <zircon/boot/netboot.h>
24
25 #include "bootserver.h"
26
27 #define DEFAULT_US_BETWEEN_PACKETS 20
28
29 static uint32_t cookie = 1;
30 static const int MAX_READ_RETRIES = 10;
31 static const int MAX_SEND_RETRIES = 10000;
32
io_rcv(int s,nbmsg * msg,nbmsg * ack)33 static int io_rcv(int s, nbmsg* msg, nbmsg* ack) {
34 for (int i = 0; i < MAX_READ_RETRIES; i++) {
35 bool retry_allowed = i + 1 < MAX_READ_RETRIES;
36
37 ssize_t r = read(s, ack, 2048);
38 if (r < 0) {
39 if (retry_allowed && errno == EAGAIN) {
40 continue;
41 }
42 fprintf(stderr, "\n%s: error: Socket read error %d\n", appname, errno);
43 return -1;
44 }
45 if ((size_t)r < sizeof(nbmsg)) {
46 fprintf(stderr, "\n%s: error: Read too short\n", appname);
47 return -1;
48 }
49 #ifdef DEBUG
50 fprintf(stdout, " < magic = %08x, cookie = %08x, cmd = %08x, arg = %08x\n",
51 ack->magic, ack->cookie, ack->cmd, ack->arg);
52 #endif
53
54 if (ack->magic != NB_MAGIC) {
55 fprintf(stderr, "\n%s: error: Bad magic\n", appname);
56 return 0;
57 }
58 if (msg) {
59 if (ack->cookie > msg->cookie) {
60 fprintf(stderr, "\n%s: error: Bad cookie\n", appname);
61 return 0;
62 }
63 }
64
65 if (ack->cmd == NB_ACK || ack->cmd == NB_FILE_RECEIVED) {
66 return 0;
67 }
68
69 switch (ack->cmd) {
70 case NB_ERROR:
71 fprintf(stderr, "\n%s: error: Generic error\n", appname);
72 break;
73 case NB_ERROR_BAD_CMD:
74 fprintf(stderr, "\n%s: error: Bad command\n", appname);
75 break;
76 case NB_ERROR_BAD_PARAM:
77 fprintf(stderr, "\n%s: error: Bad parameter\n", appname);
78 break;
79 case NB_ERROR_TOO_LARGE:
80 fprintf(stderr, "\n%s: error: File too large\n", appname);
81 break;
82 case NB_ERROR_BAD_FILE:
83 fprintf(stderr, "\n%s: error: Bad file\n", appname);
84 break;
85 default:
86 fprintf(stderr, "\n%s: error: Unknown command 0x%08X\n", appname, ack->cmd);
87 }
88 return -1;
89 }
90 fprintf(stderr, "\n%s: error: Unexpected code path\n", appname);
91 return -1;
92 }
93
io_send(int s,nbmsg * msg,size_t len)94 static int io_send(int s, nbmsg* msg, size_t len) {
95 for (int i = 0; i < MAX_SEND_RETRIES; i++) {
96 #if defined(__APPLE__)
97 bool retry_allowed = i + 1 < MAX_SEND_RETRIES;
98 #endif
99
100 int r = write(s, msg, len);
101 if (r < 0) {
102 #if defined(__APPLE__)
103 if (retry_allowed && errno == ENOBUFS) {
104 // On Darwin we manage to overflow the ethernet driver, so retry
105 struct timespec reqtime;
106 reqtime.tv_sec = 0;
107 reqtime.tv_nsec = 50 * 1000;
108 nanosleep(&reqtime, NULL);
109 continue;
110 }
111 #endif
112 fprintf(stderr, "\n%s: error: Socket write error %d\n", appname, errno);
113 return -1;
114 }
115 return 0;
116 }
117 fprintf(stderr, "\n%s: error: Unexpected code path\n", appname);
118 return -1;
119 }
120
io(int s,nbmsg * msg,size_t len,nbmsg * ack,bool wait_reply)121 static int io(int s, nbmsg* msg, size_t len, nbmsg* ack, bool wait_reply) {
122 int r, n;
123 struct timeval tv;
124 fd_set reads, writes;
125 fd_set* ws = NULL;
126 fd_set* rs = NULL;
127
128 ack->cookie = 0;
129 ack->cmd = 0;
130 ack->arg = 0;
131
132 FD_ZERO(&reads);
133 if (!wait_reply) {
134 FD_SET(s, &reads);
135 rs = &reads;
136 }
137
138 FD_ZERO(&writes);
139 if (msg && len > 0) {
140 msg->magic = NB_MAGIC;
141 msg->cookie = cookie++;
142
143 FD_SET(s, &writes);
144 ws = &writes;
145 }
146
147 if (rs || ws) {
148 n = s + 1;
149 tv.tv_sec = 10;
150 tv.tv_usec = 500000;
151 int rv = select(n, rs, ws, NULL, &tv);
152 if (rv == -1) {
153 fprintf(stderr, "\n%s: error: Select failed %d\n", appname, errno);
154 return -1;
155 } else if (rv == 0) {
156 // Timed-out
157 fprintf(stderr, "\n%s: error: Select timed out\n", appname);
158 return -1;
159 } else {
160 r = 0;
161 if (FD_ISSET(s, &reads)) {
162 r = io_rcv(s, msg, ack);
163 }
164
165 // If we got an ack, don't bother sending anything - go handle the ack first
166 if (!r && FD_ISSET(s, &writes) && (ack->cookie == 0 || ack->cmd != NB_ACK)) {
167 r = io_send(s, msg, len);
168 }
169
170 if (r || !wait_reply) {
171 return r;
172 }
173 }
174 } else if (!wait_reply) { // no-op
175 return 0;
176 }
177
178 if (wait_reply) {
179 return io_rcv(s, msg, ack);
180 }
181 fprintf(stderr, "\n%s: error: Select triggered without events\n", appname);
182 return -1;
183 }
184
185 typedef struct {
186 FILE* fp;
187 const char* data;
188 size_t datalen;
189 const char* ptr;
190 size_t avail;
191 } xferdata;
192
xread(xferdata * xd,void * data,size_t len)193 static ssize_t xread(xferdata* xd, void* data, size_t len) {
194 if (xd->fp == NULL) {
195 if (len > xd->avail) {
196 len = xd->avail;
197 }
198 memcpy(data, xd->ptr, len);
199 xd->avail -= len;
200 xd->ptr += len;
201 return len;
202 } else {
203 ssize_t r = fread(data, 1, len, xd->fp);
204 if (r == 0) {
205 return ferror(xd->fp) ? -1 : 0;
206 }
207 return r;
208 }
209 }
210
xseek(xferdata * xd,size_t off)211 static int xseek(xferdata* xd, size_t off) {
212 if (xd->fp == NULL) {
213 if (off > xd->datalen) {
214 return -1;
215 }
216 xd->ptr = xd->data + off;
217 xd->avail = xd->datalen - off;
218 return 0;
219 } else {
220 return fseek(xd->fp, off, SEEK_SET);
221 }
222 }
223
224 // UDP6_MAX_PAYLOAD (ETH_MTU - ETH_HDR_LEN - IP6_HDR_LEN - UDP_HDR_LEN)
225 // 1452 1514 - 14 - 40 - 8
226 // nbfile is PAYLOAD_SIZE + 2 * sizeof(size_t)
227
228 // Some EFI network stacks have problems with larger packets
229 // 1280 is friendlier
230 #define PAYLOAD_SIZE 1280
231
netboot_xfer(struct sockaddr_in6 * addr,const char * fn,const char * name)232 int netboot_xfer(struct sockaddr_in6* addr, const char* fn, const char* name) {
233 xferdata xd;
234 char msgbuf[2048];
235 char ackbuf[2048];
236 char tmp[INET6_ADDRSTRLEN];
237 struct timeval tv;
238 nbmsg* msg = (void*)msgbuf;
239 nbmsg* ack = (void*)ackbuf;
240 int s;
241 int status = -1;
242 size_t current_pos = 0;
243 size_t sz = 0;
244
245 if (!strcmp(fn, "(cmdline)")) {
246 xd.fp = NULL;
247 xd.data = name;
248 xd.datalen = strlen(name) + 1;
249 xd.ptr = xd.data;
250 xd.avail = xd.datalen;
251 name = NB_CMDLINE_FILENAME;
252 sz = xd.datalen;
253 } else {
254 if ((xd.fp = fopen(fn, "rb")) == NULL) {
255 fprintf(stderr, "%s: error: Could not open file %s\n", appname, fn);
256 return -1;
257 }
258 if (fseek(xd.fp, 0L, SEEK_END)) {
259 fprintf(stderr, "%s: error: Could not determine size of %s\n", appname, fn);
260 } else if ((sz = ftell(xd.fp)) < 0) {
261 fprintf(stderr, "%s: error: Could not determine size of %s\n", appname, fn);
262 sz = 0;
263 } else if (fseek(xd.fp, 0L, SEEK_SET)) {
264 fprintf(stderr, "%s: error: Failed to rewind %s\n", appname, fn);
265 return -1;
266 }
267 }
268
269 if (sz > 0) {
270 initialize_status(xd.data, sz);
271 }
272
273 if ((s = socket(AF_INET6, SOCK_DGRAM, IPPROTO_UDP)) < 0) {
274 fprintf(stderr, "%s: error: Cannot create socket %d\n", appname, errno);
275 goto done;
276 }
277 tv.tv_sec = 0;
278 tv.tv_usec = 250 * 1000;
279 setsockopt(s, SOL_SOCKET, SO_RCVTIMEO, &tv, sizeof(tv));
280 if (connect(s, (void*)addr, sizeof(*addr)) < 0) {
281 fprintf(stderr, "%s: error: Cannot connect to [%s]%d\n", appname,
282 inet_ntop(AF_INET6, &addr->sin6_addr, tmp, sizeof(tmp)),
283 ntohs(addr->sin6_port));
284 goto done;
285 }
286
287 msg->cmd = NB_SEND_FILE;
288 msg->arg = sz;
289 strcpy((void*)msg->data, name);
290 if (io(s, msg, sizeof(nbmsg) + strlen(name) + 1, ack, true)) {
291 fprintf(stderr, "%s: error: Failed to start transfer\n", appname);
292 goto done;
293 }
294
295 msg->cmd = NB_DATA;
296 msg->arg = 0;
297
298 bool completed = false;
299 do {
300 struct timeval packet_start_time;
301 gettimeofday(&packet_start_time, NULL);
302
303 ssize_t r = xread(&xd, msg->data, PAYLOAD_SIZE);
304 if (r < 0) {
305 fprintf(stderr, "\n%s: error: Reading '%s'\n", appname, fn);
306 goto done;
307 }
308
309 update_status(msg->arg);
310
311 if (r == 0) {
312 fprintf(stderr, "\n%s: Reached end of file, waiting for confirmation.\n", appname);
313 // Do not send anything, but keep waiting on incoming messages
314 if (io(s, NULL, 0, ack, true)) {
315 goto done;
316 }
317 } else {
318 if (current_pos + (size_t)r >= sz) {
319 msg->cmd = NB_LAST_DATA;
320 } else {
321 msg->cmd = NB_DATA;
322 }
323
324 if (io(s, msg, sizeof(nbmsg) + r, ack, false)) {
325 goto done;
326 }
327
328 // Some UEFI netstacks can lose back-to-back packets at max speed
329 // so throttle output.
330 // At 1280 bytes per packet, we should at least have 10 microseconds
331 // between packets, to be safe using 20 microseconds here.
332 // 1280 bytes * (1,000,000/10) seconds = 128,000,000 bytes/seconds = 122MB/s = 976Mb/s
333 // We wait as a busy wait as the context switching a sleep can cause
334 // will often degrade performance significantly.
335 int64_t us_since_last_packet;
336 do {
337 struct timeval now;
338 gettimeofday(&now, NULL);
339 us_since_last_packet = (int64_t)(now.tv_sec - packet_start_time.tv_sec) * 1000000 +
340 ((int64_t)now.tv_usec - (int64_t)packet_start_time.tv_usec);
341 } while (us_since_last_packet < us_between_packets);
342 }
343
344 // ACKs really are NACKs
345 if (ack->cookie > 0 && ack->cmd == NB_ACK) {
346 // ACKs tend to be generated in groups, since a dropped packet will cause ACKs for all
347 // outstanding packets. Therefore briefly sleep when we receive an ACK with a different
348 // position, to let things settle and prevent ourselves from fighting subsequent acks.
349 if (ack->arg != current_pos) {
350 fprintf(stderr, "\n%s: need to reset to %d from %zu\n",
351 appname, ack->arg, current_pos);
352 current_pos = ack->arg;
353
354 tv.tv_usec = 100000;
355 select(0, NULL, NULL, NULL, &tv);
356 }
357 if (xseek(&xd, current_pos)) {
358 fprintf(stderr, "\n%s: error: Failed to rewind '%s' to %zu\n",
359 appname, fn, current_pos);
360 goto done;
361 }
362 } else if (ack->cmd == NB_FILE_RECEIVED) {
363 current_pos += r;
364 completed = true;
365 } else {
366 current_pos += r;
367 }
368
369 msg->arg = current_pos;
370 } while (!completed);
371
372 status = 0;
373 update_status(msg->arg);
374 done:
375 if (s >= 0) {
376 close(s);
377 }
378 if (xd.fp != NULL) {
379 fclose(xd.fp);
380 }
381 return status;
382 }
383