1 /*
2 * Copyright (C) 2008 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 * http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17 #include <string.h>
18 #include <sys/types.h>
19 #include <sys/socket.h>
20 #include <poll.h>
21 #include <sys/wait.h>
22 #include <stdio.h>
23 #include <stdlib.h>
24 #include <unistd.h>
25 #include <errno.h>
26 #include <fcntl.h>
27 #include <libgen.h>
28 #include <stdbool.h>
29 #include <pthread.h>
30
31 #include <logwrap/logwrap.h>
32 #include "private/android_filesystem_config.h"
33 #include "cutils/log.h"
34 #include <cutils/klog.h>
35
36 #define ARRAY_SIZE(x) (sizeof(x) / sizeof(*(x)))
37 #define MIN(a,b) (((a)<(b))?(a):(b))
38
39 static pthread_mutex_t fd_mutex = PTHREAD_MUTEX_INITIALIZER;
40
41 #define ERROR(fmt, args...) \
42 do { \
43 fprintf(stderr, fmt, ## args); \
44 ALOG(LOG_ERROR, "logwrapper", fmt, ## args); \
45 } while(0)
46
47 #define FATAL_CHILD(fmt, args...) \
48 do { \
49 ERROR(fmt, ## args); \
50 _exit(-1); \
51 } while(0)
52
53 #define MAX_KLOG_TAG 16
54
55 /* This is a simple buffer that holds up to the first beginning_buf->buf_size
56 * bytes of output from a command.
57 */
58 #define BEGINNING_BUF_SIZE 0x1000
59 struct beginning_buf {
60 char *buf;
61 size_t alloc_len;
62 /* buf_size is the usable space, which is one less than the allocated size */
63 size_t buf_size;
64 size_t used_len;
65 };
66
67 /* This is a circular buf that holds up to the last ending_buf->buf_size bytes
68 * of output from a command after the first beginning_buf->buf_size bytes
69 * (which are held in beginning_buf above).
70 */
71 #define ENDING_BUF_SIZE 0x1000
72 struct ending_buf {
73 char *buf;
74 ssize_t alloc_len;
75 /* buf_size is the usable space, which is one less than the allocated size */
76 ssize_t buf_size;
77 ssize_t used_len;
78 /* read and write offsets into the circular buffer */
79 int read;
80 int write;
81 };
82
83 /* A structure to hold all the abbreviated buf data */
84 struct abbr_buf {
85 struct beginning_buf b_buf;
86 struct ending_buf e_buf;
87 int beginning_buf_full;
88 };
89
90 /* Collect all the various bits of info needed for logging in one place. */
91 struct log_info {
92 int log_target;
93 char klog_fmt[MAX_KLOG_TAG * 2];
94 char *btag;
95 bool abbreviated;
96 FILE *fp;
97 struct abbr_buf a_buf;
98 };
99
100 /* Forware declaration */
101 static void add_line_to_abbr_buf(struct abbr_buf *a_buf, char *linebuf, int linelen);
102
103 /* Return 0 on success, and 1 when full */
add_line_to_linear_buf(struct beginning_buf * b_buf,char * line,ssize_t line_len)104 static int add_line_to_linear_buf(struct beginning_buf *b_buf,
105 char *line, ssize_t line_len)
106 {
107 int full = 0;
108
109 if ((line_len + b_buf->used_len) > b_buf->buf_size) {
110 full = 1;
111 } else {
112 /* Add to the end of the buf */
113 memcpy(b_buf->buf + b_buf->used_len, line, line_len);
114 b_buf->used_len += line_len;
115 }
116
117 return full;
118 }
119
add_line_to_circular_buf(struct ending_buf * e_buf,char * line,ssize_t line_len)120 static void add_line_to_circular_buf(struct ending_buf *e_buf,
121 char *line, ssize_t line_len)
122 {
123 ssize_t free_len;
124 ssize_t needed_space;
125 int cnt;
126
127 if (e_buf->buf == NULL) {
128 return;
129 }
130
131 if (line_len > e_buf->buf_size) {
132 return;
133 }
134
135 free_len = e_buf->buf_size - e_buf->used_len;
136
137 if (line_len > free_len) {
138 /* remove oldest entries at read, and move read to make
139 * room for the new string */
140 needed_space = line_len - free_len;
141 e_buf->read = (e_buf->read + needed_space) % e_buf->buf_size;
142 e_buf->used_len -= needed_space;
143 }
144
145 /* Copy the line into the circular buffer, dealing with possible
146 * wraparound.
147 */
148 cnt = MIN(line_len, e_buf->buf_size - e_buf->write);
149 memcpy(e_buf->buf + e_buf->write, line, cnt);
150 if (cnt < line_len) {
151 memcpy(e_buf->buf, line + cnt, line_len - cnt);
152 }
153 e_buf->used_len += line_len;
154 e_buf->write = (e_buf->write + line_len) % e_buf->buf_size;
155 }
156
157 /* Log directly to the specified log */
do_log_line(struct log_info * log_info,char * line)158 static void do_log_line(struct log_info *log_info, char *line) {
159 if (log_info->log_target & LOG_KLOG) {
160 klog_write(6, log_info->klog_fmt, line);
161 }
162 if (log_info->log_target & LOG_ALOG) {
163 ALOG(LOG_INFO, log_info->btag, "%s", line);
164 }
165 if (log_info->log_target & LOG_FILE) {
166 fprintf(log_info->fp, "%s\n", line);
167 }
168 }
169
170 /* Log to either the abbreviated buf, or directly to the specified log
171 * via do_log_line() above.
172 */
log_line(struct log_info * log_info,char * line,int len)173 static void log_line(struct log_info *log_info, char *line, int len) {
174 if (log_info->abbreviated) {
175 add_line_to_abbr_buf(&log_info->a_buf, line, len);
176 } else {
177 do_log_line(log_info, line);
178 }
179 }
180
181 /*
182 * The kernel will take a maximum of 1024 bytes in any single write to
183 * the kernel logging device file, so find and print each line one at
184 * a time. The allocated size for buf should be at least 1 byte larger
185 * than buf_size (the usable size of the buffer) to make sure there is
186 * room to temporarily stuff a null byte to terminate a line for logging.
187 */
print_buf_lines(struct log_info * log_info,char * buf,int buf_size)188 static void print_buf_lines(struct log_info *log_info, char *buf, int buf_size)
189 {
190 char *line_start;
191 char c;
192 int i;
193
194 line_start = buf;
195 for (i = 0; i < buf_size; i++) {
196 if (*(buf + i) == '\n') {
197 /* Found a line ending, print the line and compute new line_start */
198 /* Save the next char and replace with \0 */
199 c = *(buf + i + 1);
200 *(buf + i + 1) = '\0';
201 do_log_line(log_info, line_start);
202 /* Restore the saved char */
203 *(buf + i + 1) = c;
204 line_start = buf + i + 1;
205 } else if (*(buf + i) == '\0') {
206 /* The end of the buffer, print the last bit */
207 do_log_line(log_info, line_start);
208 break;
209 }
210 }
211 /* If the buffer was completely full, and didn't end with a newline, just
212 * ignore the partial last line.
213 */
214 }
215
init_abbr_buf(struct abbr_buf * a_buf)216 static void init_abbr_buf(struct abbr_buf *a_buf) {
217 char *new_buf;
218
219 memset(a_buf, 0, sizeof(struct abbr_buf));
220 new_buf = malloc(BEGINNING_BUF_SIZE);
221 if (new_buf) {
222 a_buf->b_buf.buf = new_buf;
223 a_buf->b_buf.alloc_len = BEGINNING_BUF_SIZE;
224 a_buf->b_buf.buf_size = BEGINNING_BUF_SIZE - 1;
225 }
226 new_buf = malloc(ENDING_BUF_SIZE);
227 if (new_buf) {
228 a_buf->e_buf.buf = new_buf;
229 a_buf->e_buf.alloc_len = ENDING_BUF_SIZE;
230 a_buf->e_buf.buf_size = ENDING_BUF_SIZE - 1;
231 }
232 }
233
free_abbr_buf(struct abbr_buf * a_buf)234 static void free_abbr_buf(struct abbr_buf *a_buf) {
235 free(a_buf->b_buf.buf);
236 free(a_buf->e_buf.buf);
237 }
238
add_line_to_abbr_buf(struct abbr_buf * a_buf,char * linebuf,int linelen)239 static void add_line_to_abbr_buf(struct abbr_buf *a_buf, char *linebuf, int linelen) {
240 if (!a_buf->beginning_buf_full) {
241 a_buf->beginning_buf_full =
242 add_line_to_linear_buf(&a_buf->b_buf, linebuf, linelen);
243 }
244 if (a_buf->beginning_buf_full) {
245 add_line_to_circular_buf(&a_buf->e_buf, linebuf, linelen);
246 }
247 }
248
print_abbr_buf(struct log_info * log_info)249 static void print_abbr_buf(struct log_info *log_info) {
250 struct abbr_buf *a_buf = &log_info->a_buf;
251
252 /* Add the abbreviated output to the kernel log */
253 if (a_buf->b_buf.alloc_len) {
254 print_buf_lines(log_info, a_buf->b_buf.buf, a_buf->b_buf.used_len);
255 }
256
257 /* Print an ellipsis to indicate that the buffer has wrapped or
258 * is full, and some data was not logged.
259 */
260 if (a_buf->e_buf.used_len == a_buf->e_buf.buf_size) {
261 do_log_line(log_info, "...\n");
262 }
263
264 if (a_buf->e_buf.used_len == 0) {
265 return;
266 }
267
268 /* Simplest way to print the circular buffer is allocate a second buf
269 * of the same size, and memcpy it so it's a simple linear buffer,
270 * and then cal print_buf_lines on it */
271 if (a_buf->e_buf.read < a_buf->e_buf.write) {
272 /* no wrap around, just print it */
273 print_buf_lines(log_info, a_buf->e_buf.buf + a_buf->e_buf.read,
274 a_buf->e_buf.used_len);
275 } else {
276 /* The circular buffer will always have at least 1 byte unused,
277 * so by allocating alloc_len here we will have at least
278 * 1 byte of space available as required by print_buf_lines().
279 */
280 char * nbuf = malloc(a_buf->e_buf.alloc_len);
281 if (!nbuf) {
282 return;
283 }
284 int first_chunk_len = a_buf->e_buf.buf_size - a_buf->e_buf.read;
285 memcpy(nbuf, a_buf->e_buf.buf + a_buf->e_buf.read, first_chunk_len);
286 /* copy second chunk */
287 memcpy(nbuf + first_chunk_len, a_buf->e_buf.buf, a_buf->e_buf.write);
288 print_buf_lines(log_info, nbuf, first_chunk_len + a_buf->e_buf.write);
289 free(nbuf);
290 }
291 }
292
parent(const char * tag,int parent_read,pid_t pid,int * chld_sts,int log_target,bool abbreviated,char * file_path,const struct AndroidForkExecvpOption * opts,size_t opts_len)293 static int parent(const char *tag, int parent_read, pid_t pid,
294 int *chld_sts, int log_target, bool abbreviated, char *file_path,
295 const struct AndroidForkExecvpOption* opts, size_t opts_len) {
296 int status = 0;
297 char buffer[4096];
298 struct pollfd poll_fds[] = {
299 [0] = {
300 .fd = parent_read,
301 .events = POLLIN,
302 },
303 };
304 int rc = 0;
305 int fd;
306
307 struct log_info log_info;
308
309 int a = 0; // start index of unprocessed data
310 int b = 0; // end index of unprocessed data
311 int sz;
312 bool found_child = false;
313 char tmpbuf[256];
314
315 log_info.btag = basename(tag);
316 if (!log_info.btag) {
317 log_info.btag = (char*) tag;
318 }
319
320 if (abbreviated && (log_target == LOG_NONE)) {
321 abbreviated = 0;
322 }
323 if (abbreviated) {
324 init_abbr_buf(&log_info.a_buf);
325 }
326
327 if (log_target & LOG_KLOG) {
328 snprintf(log_info.klog_fmt, sizeof(log_info.klog_fmt),
329 "<6>%.*s: %%s\n", MAX_KLOG_TAG, log_info.btag);
330 }
331
332 if ((log_target & LOG_FILE) && !file_path) {
333 /* No file_path specified, clear the LOG_FILE bit */
334 log_target &= ~LOG_FILE;
335 }
336
337 if (log_target & LOG_FILE) {
338 fd = open(file_path, O_WRONLY | O_CREAT, 0664);
339 if (fd < 0) {
340 ERROR("Cannot log to file %s\n", file_path);
341 log_target &= ~LOG_FILE;
342 } else {
343 lseek(fd, 0, SEEK_END);
344 log_info.fp = fdopen(fd, "a");
345 }
346 }
347
348 log_info.log_target = log_target;
349 log_info.abbreviated = abbreviated;
350
351 while (!found_child) {
352 if (TEMP_FAILURE_RETRY(poll(poll_fds, ARRAY_SIZE(poll_fds), -1)) < 0) {
353 ERROR("poll failed\n");
354 rc = -1;
355 goto err_poll;
356 }
357
358 if (poll_fds[0].revents & POLLIN) {
359 sz = TEMP_FAILURE_RETRY(
360 read(parent_read, &buffer[b], sizeof(buffer) - 1 - b));
361
362 for (size_t i = 0; sz > 0 && i < opts_len; ++i) {
363 if (opts[i].opt_type == FORK_EXECVP_OPTION_CAPTURE_OUTPUT) {
364 opts[i].opt_capture_output.on_output(
365 (uint8_t*)&buffer[b], sz, opts[i].opt_capture_output.user_pointer);
366 }
367 }
368
369 sz += b;
370 // Log one line at a time
371 for (b = 0; b < sz; b++) {
372 if (buffer[b] == '\r') {
373 if (abbreviated) {
374 /* The abbreviated logging code uses newline as
375 * the line separator. Lucikly, the pty layer
376 * helpfully cooks the output of the command
377 * being run and inserts a CR before NL. So
378 * I just change it to NL here when doing
379 * abbreviated logging.
380 */
381 buffer[b] = '\n';
382 } else {
383 buffer[b] = '\0';
384 }
385 } else if (buffer[b] == '\n') {
386 buffer[b] = '\0';
387 log_line(&log_info, &buffer[a], b - a);
388 a = b + 1;
389 }
390 }
391
392 if (a == 0 && b == sizeof(buffer) - 1) {
393 // buffer is full, flush
394 buffer[b] = '\0';
395 log_line(&log_info, &buffer[a], b - a);
396 b = 0;
397 } else if (a != b) {
398 // Keep left-overs
399 b -= a;
400 memmove(buffer, &buffer[a], b);
401 a = 0;
402 } else {
403 a = 0;
404 b = 0;
405 }
406 }
407
408 if (poll_fds[0].revents & POLLHUP) {
409 int ret;
410
411 ret = waitpid(pid, &status, WNOHANG);
412 if (ret < 0) {
413 rc = errno;
414 ALOG(LOG_ERROR, "logwrap", "waitpid failed with %s\n", strerror(errno));
415 goto err_waitpid;
416 }
417 if (ret > 0) {
418 found_child = true;
419 }
420 }
421 }
422
423 if (chld_sts != NULL) {
424 *chld_sts = status;
425 } else {
426 if (WIFEXITED(status))
427 rc = WEXITSTATUS(status);
428 else
429 rc = -ECHILD;
430 }
431
432 // Flush remaining data
433 if (a != b) {
434 buffer[b] = '\0';
435 log_line(&log_info, &buffer[a], b - a);
436 }
437
438 /* All the output has been processed, time to dump the abbreviated output */
439 if (abbreviated) {
440 print_abbr_buf(&log_info);
441 }
442
443 if (WIFEXITED(status)) {
444 if (WEXITSTATUS(status)) {
445 snprintf(tmpbuf, sizeof(tmpbuf),
446 "%s terminated by exit(%d)\n", log_info.btag, WEXITSTATUS(status));
447 do_log_line(&log_info, tmpbuf);
448 }
449 } else {
450 if (WIFSIGNALED(status)) {
451 snprintf(tmpbuf, sizeof(tmpbuf),
452 "%s terminated by signal %d\n", log_info.btag, WTERMSIG(status));
453 do_log_line(&log_info, tmpbuf);
454 } else if (WIFSTOPPED(status)) {
455 snprintf(tmpbuf, sizeof(tmpbuf),
456 "%s stopped by signal %d\n", log_info.btag, WSTOPSIG(status));
457 do_log_line(&log_info, tmpbuf);
458 }
459 }
460
461 err_waitpid:
462 err_poll:
463 if (log_target & LOG_FILE) {
464 fclose(log_info.fp); /* Also closes underlying fd */
465 }
466 if (abbreviated) {
467 free_abbr_buf(&log_info.a_buf);
468 }
469 return rc;
470 }
471
child(int argc,char * argv[])472 static void child(int argc, char* argv[]) {
473 // create null terminated argv_child array
474 char* argv_child[argc + 1];
475 memcpy(argv_child, argv, argc * sizeof(char *));
476 argv_child[argc] = NULL;
477
478 if (execvp(argv_child[0], argv_child)) {
479 FATAL_CHILD("executing %s failed: %s\n", argv_child[0],
480 strerror(errno));
481 }
482 }
483
android_fork_execvp_ext(int argc,char * argv[],int * status,bool ignore_int_quit,int log_target,bool abbreviated,char * file_path,const struct AndroidForkExecvpOption * opts,size_t opts_len)484 int android_fork_execvp_ext(int argc, char* argv[], int *status, bool ignore_int_quit,
485 int log_target, bool abbreviated, char *file_path,
486 const struct AndroidForkExecvpOption* opts, size_t opts_len) {
487 pid_t pid;
488 int parent_ptty;
489 int child_ptty;
490 struct sigaction intact;
491 struct sigaction quitact;
492 sigset_t blockset;
493 sigset_t oldset;
494 int rc = 0;
495
496 rc = pthread_mutex_lock(&fd_mutex);
497 if (rc) {
498 ERROR("failed to lock signal_fd mutex\n");
499 goto err_lock;
500 }
501
502 /* Use ptty instead of socketpair so that STDOUT is not buffered */
503 parent_ptty = TEMP_FAILURE_RETRY(open("/dev/ptmx", O_RDWR));
504 if (parent_ptty < 0) {
505 ERROR("Cannot create parent ptty\n");
506 rc = -1;
507 goto err_open;
508 }
509
510 char child_devname[64];
511 if (grantpt(parent_ptty) || unlockpt(parent_ptty) ||
512 ptsname_r(parent_ptty, child_devname, sizeof(child_devname)) != 0) {
513 ERROR("Problem with /dev/ptmx\n");
514 rc = -1;
515 goto err_ptty;
516 }
517
518 child_ptty = TEMP_FAILURE_RETRY(open(child_devname, O_RDWR));
519 if (child_ptty < 0) {
520 ERROR("Cannot open child_ptty\n");
521 rc = -1;
522 goto err_child_ptty;
523 }
524
525 sigemptyset(&blockset);
526 sigaddset(&blockset, SIGINT);
527 sigaddset(&blockset, SIGQUIT);
528 pthread_sigmask(SIG_BLOCK, &blockset, &oldset);
529
530 pid = fork();
531 if (pid < 0) {
532 close(child_ptty);
533 ERROR("Failed to fork\n");
534 rc = -1;
535 goto err_fork;
536 } else if (pid == 0) {
537 pthread_mutex_unlock(&fd_mutex);
538 pthread_sigmask(SIG_SETMASK, &oldset, NULL);
539 close(parent_ptty);
540
541 // redirect stdin, stdout and stderr
542 for (size_t i = 0; i < opts_len; ++i) {
543 if (opts[i].opt_type == FORK_EXECVP_OPTION_INPUT) {
544 dup2(child_ptty, 0);
545 break;
546 }
547 }
548 dup2(child_ptty, 1);
549 dup2(child_ptty, 2);
550 close(child_ptty);
551
552 child(argc, argv);
553 } else {
554 close(child_ptty);
555 if (ignore_int_quit) {
556 struct sigaction ignact;
557
558 memset(&ignact, 0, sizeof(ignact));
559 ignact.sa_handler = SIG_IGN;
560 sigaction(SIGINT, &ignact, &intact);
561 sigaction(SIGQUIT, &ignact, &quitact);
562 }
563
564 for (size_t i = 0; i < opts_len; ++i) {
565 if (opts[i].opt_type == FORK_EXECVP_OPTION_INPUT) {
566 size_t left = opts[i].opt_input.input_len;
567 const uint8_t* input = opts[i].opt_input.input;
568 while (left > 0) {
569 ssize_t res =
570 TEMP_FAILURE_RETRY(write(parent_ptty, input, left));
571 if (res < 0) {
572 break;
573 }
574 left -= res;
575 input += res;
576 }
577 }
578 }
579
580 rc = parent(argv[0], parent_ptty, pid, status, log_target,
581 abbreviated, file_path, opts, opts_len);
582 }
583
584 if (ignore_int_quit) {
585 sigaction(SIGINT, &intact, NULL);
586 sigaction(SIGQUIT, &quitact, NULL);
587 }
588 err_fork:
589 pthread_sigmask(SIG_SETMASK, &oldset, NULL);
590 err_child_ptty:
591 err_ptty:
592 close(parent_ptty);
593 err_open:
594 pthread_mutex_unlock(&fd_mutex);
595 err_lock:
596 return rc;
597 }
598