1 /* $OpenBSD: shf.c,v 1.16 2013/04/19 17:36:09 millert Exp $ */
2
3 /*-
4 * Copyright (c) 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2011,
5 * 2012, 2013, 2015, 2016, 2017, 2018
6 * mirabilos <m@mirbsd.org>
7 * Copyright (c) 2015
8 * Daniel Richard G. <skunk@iSKUNK.ORG>
9 *
10 * Provided that these terms and disclaimer and all copyright notices
11 * are retained or reproduced in an accompanying document, permission
12 * is granted to deal in this work without restriction, including un-
13 * limited rights to use, publicly perform, distribute, sell, modify,
14 * merge, give away, or sublicence.
15 *
16 * This work is provided "AS IS" and WITHOUT WARRANTY of any kind, to
17 * the utmost extent permitted by applicable law, neither express nor
18 * implied; without malicious intent or gross negligence. In no event
19 * may a licensor, author or contributor be held liable for indirect,
20 * direct, other damage, loss, or other issues arising in any way out
21 * of dealing in the work, even if advised of the possibility of such
22 * damage or existence of a defect, except proven that it results out
23 * of said person's immediate fault when using the work as intended.
24 *-
25 * Use %zX instead of %p and floating point isn't supported at all.
26 */
27
28 #include "sh.h"
29
30 __RCSID("$MirOS: src/bin/mksh/shf.c,v 1.97 2018/01/14 01:28:16 tg Exp $");
31
32 /* flags to shf_emptybuf() */
33 #define EB_READSW 0x01 /* about to switch to reading */
34 #define EB_GROW 0x02 /* grow buffer if necessary (STRING+DYNAMIC) */
35
36 /*
37 * Replacement stdio routines. Stdio is too flakey on too many machines
38 * to be useful when you have multiple processes using the same underlying
39 * file descriptors.
40 */
41
42 static int shf_fillbuf(struct shf *);
43 static int shf_emptybuf(struct shf *, int);
44
45 /*
46 * Open a file. First three args are for open(), last arg is flags for
47 * this package. Returns NULL if file could not be opened, or if a dup
48 * fails.
49 */
50 struct shf *
shf_open(const char * name,int oflags,int mode,int sflags)51 shf_open(const char *name, int oflags, int mode, int sflags)
52 {
53 struct shf *shf;
54 ssize_t bsize =
55 /* at most 512 */
56 sflags & SHF_UNBUF ? (sflags & SHF_RD ? 1 : 0) : SHF_BSIZE;
57 int fd, eno;
58
59 /* Done before open so if alloca fails, fd won't be lost. */
60 shf = alloc(sizeof(struct shf) + bsize, ATEMP);
61 shf->areap = ATEMP;
62 shf->buf = (unsigned char *)&shf[1];
63 shf->bsize = bsize;
64 shf->flags = SHF_ALLOCS;
65 /* Rest filled in by reopen. */
66
67 fd = binopen3(name, oflags, mode);
68 if (fd < 0) {
69 eno = errno;
70 afree(shf, shf->areap);
71 errno = eno;
72 return (NULL);
73 }
74 if ((sflags & SHF_MAPHI) && fd < FDBASE) {
75 int nfd;
76
77 nfd = fcntl(fd, F_DUPFD, FDBASE);
78 eno = errno;
79 close(fd);
80 if (nfd < 0) {
81 afree(shf, shf->areap);
82 errno = eno;
83 return (NULL);
84 }
85 fd = nfd;
86 }
87 sflags &= ~SHF_ACCMODE;
88 sflags |= (oflags & O_ACCMODE) == O_RDONLY ? SHF_RD :
89 ((oflags & O_ACCMODE) == O_WRONLY ? SHF_WR : SHF_RDWR);
90
91 return (shf_reopen(fd, sflags, shf));
92 }
93
94 /* helper function for shf_fdopen and shf_reopen */
95 static void
shf_open_hlp(int fd,int * sflagsp,const char * where)96 shf_open_hlp(int fd, int *sflagsp, const char *where)
97 {
98 int sflags = *sflagsp;
99
100 /* use fcntl() to figure out correct read/write flags */
101 if (sflags & SHF_GETFL) {
102 int flags = fcntl(fd, F_GETFL, 0);
103
104 if (flags < 0)
105 /* will get an error on first read/write */
106 sflags |= SHF_RDWR;
107 else {
108 switch (flags & O_ACCMODE) {
109 case O_RDONLY:
110 sflags |= SHF_RD;
111 break;
112 case O_WRONLY:
113 sflags |= SHF_WR;
114 break;
115 case O_RDWR:
116 sflags |= SHF_RDWR;
117 break;
118 }
119 }
120 *sflagsp = sflags;
121 }
122
123 if (!(sflags & (SHF_RD | SHF_WR)))
124 internal_errorf(Tf_sD_s, where, "missing read/write");
125 }
126
127 /* Set up the shf structure for a file descriptor. Doesn't fail. */
128 struct shf *
shf_fdopen(int fd,int sflags,struct shf * shf)129 shf_fdopen(int fd, int sflags, struct shf *shf)
130 {
131 ssize_t bsize =
132 /* at most 512 */
133 sflags & SHF_UNBUF ? (sflags & SHF_RD ? 1 : 0) : SHF_BSIZE;
134
135 shf_open_hlp(fd, &sflags, "shf_fdopen");
136 if (shf) {
137 if (bsize) {
138 shf->buf = alloc(bsize, ATEMP);
139 sflags |= SHF_ALLOCB;
140 } else
141 shf->buf = NULL;
142 } else {
143 shf = alloc(sizeof(struct shf) + bsize, ATEMP);
144 shf->buf = (unsigned char *)&shf[1];
145 sflags |= SHF_ALLOCS;
146 }
147 shf->areap = ATEMP;
148 shf->fd = fd;
149 shf->rp = shf->wp = shf->buf;
150 shf->rnleft = 0;
151 shf->rbsize = bsize;
152 shf->wnleft = 0; /* force call to shf_emptybuf() */
153 shf->wbsize = sflags & SHF_UNBUF ? 0 : bsize;
154 shf->flags = sflags;
155 shf->errnosv = 0;
156 shf->bsize = bsize;
157 if (sflags & SHF_CLEXEC)
158 fcntl(fd, F_SETFD, FD_CLOEXEC);
159 return (shf);
160 }
161
162 /* Set up an existing shf (and buffer) to use the given fd */
163 struct shf *
shf_reopen(int fd,int sflags,struct shf * shf)164 shf_reopen(int fd, int sflags, struct shf *shf)
165 {
166 ssize_t bsize =
167 /* at most 512 */
168 sflags & SHF_UNBUF ? (sflags & SHF_RD ? 1 : 0) : SHF_BSIZE;
169
170 shf_open_hlp(fd, &sflags, "shf_reopen");
171 if (!shf || !shf->buf || shf->bsize < bsize)
172 internal_errorf(Tf_sD_s, "shf_reopen", Tbad_bsize);
173
174 /* assumes shf->buf and shf->bsize already set up */
175 shf->fd = fd;
176 shf->rp = shf->wp = shf->buf;
177 shf->rnleft = 0;
178 shf->rbsize = bsize;
179 shf->wnleft = 0; /* force call to shf_emptybuf() */
180 shf->wbsize = sflags & SHF_UNBUF ? 0 : bsize;
181 shf->flags = (shf->flags & (SHF_ALLOCS | SHF_ALLOCB)) | sflags;
182 shf->errnosv = 0;
183 if (sflags & SHF_CLEXEC)
184 fcntl(fd, F_SETFD, FD_CLOEXEC);
185 return (shf);
186 }
187
188 /*
189 * Open a string for reading or writing. If reading, bsize is the number
190 * of bytes that can be read. If writing, bsize is the maximum number of
191 * bytes that can be written. If shf is not NULL, it is filled in and
192 * returned, if it is NULL, shf is allocated. If writing and buf is NULL
193 * and SHF_DYNAMIC is set, the buffer is allocated (if bsize > 0, it is
194 * used for the initial size). Doesn't fail.
195 * When writing, a byte is reserved for a trailing NUL - see shf_sclose().
196 */
197 struct shf *
shf_sopen(char * buf,ssize_t bsize,int sflags,struct shf * shf)198 shf_sopen(char *buf, ssize_t bsize, int sflags, struct shf *shf)
199 {
200 /* can't have a read+write string */
201 if (!(!(sflags & SHF_RD) ^ !(sflags & SHF_WR)))
202 internal_errorf(Tf_flags, "shf_sopen",
203 (unsigned int)sflags);
204
205 if (!shf) {
206 shf = alloc(sizeof(struct shf), ATEMP);
207 sflags |= SHF_ALLOCS;
208 }
209 shf->areap = ATEMP;
210 if (!buf && (sflags & SHF_WR) && (sflags & SHF_DYNAMIC)) {
211 if (bsize <= 0)
212 bsize = 64;
213 sflags |= SHF_ALLOCB;
214 buf = alloc(bsize, shf->areap);
215 }
216 shf->fd = -1;
217 shf->buf = shf->rp = shf->wp = (unsigned char *)buf;
218 shf->rnleft = bsize;
219 shf->rbsize = bsize;
220 shf->wnleft = bsize - 1; /* space for a '\0' */
221 shf->wbsize = bsize;
222 shf->flags = sflags | SHF_STRING;
223 shf->errnosv = 0;
224 shf->bsize = bsize;
225
226 return (shf);
227 }
228
229 /* Flush and close file descriptor, free the shf structure */
230 int
shf_close(struct shf * shf)231 shf_close(struct shf *shf)
232 {
233 int ret = 0;
234
235 if (shf->fd >= 0) {
236 ret = shf_flush(shf);
237 if (close(shf->fd) < 0)
238 ret = -1;
239 }
240 if (shf->flags & SHF_ALLOCS)
241 afree(shf, shf->areap);
242 else if (shf->flags & SHF_ALLOCB)
243 afree(shf->buf, shf->areap);
244
245 return (ret);
246 }
247
248 /* Flush and close file descriptor, don't free file structure */
249 int
shf_fdclose(struct shf * shf)250 shf_fdclose(struct shf *shf)
251 {
252 int ret = 0;
253
254 if (shf->fd >= 0) {
255 ret = shf_flush(shf);
256 if (close(shf->fd) < 0)
257 ret = -1;
258 shf->rnleft = 0;
259 shf->rp = shf->buf;
260 shf->wnleft = 0;
261 shf->fd = -1;
262 }
263
264 return (ret);
265 }
266
267 /*
268 * Close a string - if it was opened for writing, it is NUL terminated;
269 * returns a pointer to the string and frees shf if it was allocated
270 * (does not free string if it was allocated).
271 */
272 char *
shf_sclose(struct shf * shf)273 shf_sclose(struct shf *shf)
274 {
275 unsigned char *s = shf->buf;
276
277 /* NUL terminate */
278 if (shf->flags & SHF_WR) {
279 shf->wnleft++;
280 shf_putc('\0', shf);
281 }
282 if (shf->flags & SHF_ALLOCS)
283 afree(shf, shf->areap);
284 return ((char *)s);
285 }
286
287 /*
288 * Un-read what has been read but not examined, or write what has been
289 * buffered. Returns 0 for success, -1 for (write) error.
290 */
291 int
shf_flush(struct shf * shf)292 shf_flush(struct shf *shf)
293 {
294 int rv = 0;
295
296 if (shf->flags & SHF_STRING)
297 rv = (shf->flags & SHF_WR) ? -1 : 0;
298 else if (shf->fd < 0)
299 internal_errorf(Tf_sD_s, "shf_flush", "no fd");
300 else if (shf->flags & SHF_ERROR) {
301 errno = shf->errnosv;
302 rv = -1;
303 } else if (shf->flags & SHF_READING) {
304 shf->flags &= ~(SHF_EOF | SHF_READING);
305 if (shf->rnleft > 0) {
306 if (lseek(shf->fd, (off_t)-shf->rnleft,
307 SEEK_CUR) == -1) {
308 shf->flags |= SHF_ERROR;
309 shf->errnosv = errno;
310 rv = -1;
311 }
312 shf->rnleft = 0;
313 shf->rp = shf->buf;
314 }
315 } else if (shf->flags & SHF_WRITING)
316 rv = shf_emptybuf(shf, 0);
317
318 return (rv);
319 }
320
321 /*
322 * Write out any buffered data. If currently reading, flushes the read
323 * buffer. Returns 0 for success, -1 for (write) error.
324 */
325 static int
shf_emptybuf(struct shf * shf,int flags)326 shf_emptybuf(struct shf *shf, int flags)
327 {
328 int ret = 0;
329
330 if (!(shf->flags & SHF_STRING) && shf->fd < 0)
331 internal_errorf(Tf_sD_s, "shf_emptybuf", "no fd");
332
333 if (shf->flags & SHF_ERROR) {
334 errno = shf->errnosv;
335 return (-1);
336 }
337
338 if (shf->flags & SHF_READING) {
339 if (flags & EB_READSW)
340 /* doesn't happen */
341 return (0);
342 ret = shf_flush(shf);
343 shf->flags &= ~SHF_READING;
344 }
345 if (shf->flags & SHF_STRING) {
346 unsigned char *nbuf;
347
348 /*
349 * Note that we assume SHF_ALLOCS is not set if
350 * SHF_ALLOCB is set... (changing the shf pointer could
351 * cause problems)
352 */
353 if (!(flags & EB_GROW) || !(shf->flags & SHF_DYNAMIC) ||
354 !(shf->flags & SHF_ALLOCB))
355 return (-1);
356 /* allocate more space for buffer */
357 nbuf = aresize2(shf->buf, 2, shf->wbsize, shf->areap);
358 shf->rp = nbuf + (shf->rp - shf->buf);
359 shf->wp = nbuf + (shf->wp - shf->buf);
360 shf->rbsize += shf->wbsize;
361 shf->wnleft += shf->wbsize;
362 shf->wbsize <<= 1;
363 shf->buf = nbuf;
364 } else {
365 if (shf->flags & SHF_WRITING) {
366 ssize_t n, ntowrite = shf->wp - shf->buf;
367 unsigned char *buf = shf->buf;
368
369 while (ntowrite > 0) {
370 n = write(shf->fd, buf, ntowrite);
371 if (n < 0) {
372 if (errno == EINTR &&
373 !(shf->flags & SHF_INTERRUPT))
374 continue;
375 shf->flags |= SHF_ERROR;
376 shf->errnosv = errno;
377 shf->wnleft = 0;
378 if (buf != shf->buf) {
379 /*
380 * allow a second flush
381 * to work
382 */
383 memmove(shf->buf, buf,
384 ntowrite);
385 shf->wp = shf->buf + ntowrite;
386 }
387 return (-1);
388 }
389 buf += n;
390 ntowrite -= n;
391 }
392 if (flags & EB_READSW) {
393 shf->wp = shf->buf;
394 shf->wnleft = 0;
395 shf->flags &= ~SHF_WRITING;
396 return (0);
397 }
398 }
399 shf->wp = shf->buf;
400 shf->wnleft = shf->wbsize;
401 }
402 shf->flags |= SHF_WRITING;
403
404 return (ret);
405 }
406
407 /* Fill up a read buffer. Returns -1 for a read error, 0 otherwise. */
408 static int
shf_fillbuf(struct shf * shf)409 shf_fillbuf(struct shf *shf)
410 {
411 ssize_t n;
412
413 if (shf->flags & SHF_STRING)
414 return (0);
415
416 if (shf->fd < 0)
417 internal_errorf(Tf_sD_s, "shf_fillbuf", "no fd");
418
419 if (shf->flags & (SHF_EOF | SHF_ERROR)) {
420 if (shf->flags & SHF_ERROR)
421 errno = shf->errnosv;
422 return (-1);
423 }
424
425 if ((shf->flags & SHF_WRITING) && shf_emptybuf(shf, EB_READSW) == -1)
426 return (-1);
427
428 shf->flags |= SHF_READING;
429
430 shf->rp = shf->buf;
431 while (/* CONSTCOND */ 1) {
432 n = blocking_read(shf->fd, (char *)shf->buf, shf->rbsize);
433 if (n < 0 && errno == EINTR && !(shf->flags & SHF_INTERRUPT))
434 continue;
435 break;
436 }
437 if (n < 0) {
438 shf->flags |= SHF_ERROR;
439 shf->errnosv = errno;
440 shf->rnleft = 0;
441 shf->rp = shf->buf;
442 return (-1);
443 }
444 if ((shf->rnleft = n) == 0)
445 shf->flags |= SHF_EOF;
446 return (0);
447 }
448
449 /*
450 * Read a buffer from shf. Returns the number of bytes read into buf, if
451 * no bytes were read, returns 0 if end of file was seen, -1 if a read
452 * error occurred.
453 */
454 ssize_t
shf_read(char * buf,ssize_t bsize,struct shf * shf)455 shf_read(char *buf, ssize_t bsize, struct shf *shf)
456 {
457 ssize_t ncopy, orig_bsize = bsize;
458
459 if (!(shf->flags & SHF_RD))
460 internal_errorf(Tf_flags, Tshf_read,
461 (unsigned int)shf->flags);
462
463 if (bsize <= 0)
464 internal_errorf(Tf_szs, Tshf_read, bsize, Tbsize);
465
466 while (bsize > 0) {
467 if (shf->rnleft == 0 &&
468 (shf_fillbuf(shf) == -1 || shf->rnleft == 0))
469 break;
470 ncopy = shf->rnleft;
471 if (ncopy > bsize)
472 ncopy = bsize;
473 memcpy(buf, shf->rp, ncopy);
474 buf += ncopy;
475 bsize -= ncopy;
476 shf->rp += ncopy;
477 shf->rnleft -= ncopy;
478 }
479 /* Note: fread(3S) returns 0 for errors - this doesn't */
480 return (orig_bsize == bsize ? (shf_error(shf) ? -1 : 0) :
481 orig_bsize - bsize);
482 }
483
484 /*
485 * Read up to a newline or -1. The newline is put in buf; buf is always
486 * NUL terminated. Returns NULL on read error or if nothing was read
487 * before end of file, returns a pointer to the NUL byte in buf
488 * otherwise.
489 */
490 char *
shf_getse(char * buf,ssize_t bsize,struct shf * shf)491 shf_getse(char *buf, ssize_t bsize, struct shf *shf)
492 {
493 unsigned char *end;
494 ssize_t ncopy;
495 char *orig_buf = buf;
496
497 if (!(shf->flags & SHF_RD))
498 internal_errorf(Tf_flags, "shf_getse",
499 (unsigned int)shf->flags);
500
501 if (bsize <= 0)
502 return (NULL);
503
504 /* save room for NUL */
505 --bsize;
506 do {
507 if (shf->rnleft == 0) {
508 if (shf_fillbuf(shf) == -1)
509 return (NULL);
510 if (shf->rnleft == 0) {
511 *buf = '\0';
512 return (buf == orig_buf ? NULL : buf);
513 }
514 }
515 end = (unsigned char *)memchr((char *)shf->rp, '\n',
516 shf->rnleft);
517 ncopy = end ? end - shf->rp + 1 : shf->rnleft;
518 if (ncopy > bsize)
519 ncopy = bsize;
520 memcpy(buf, (char *) shf->rp, ncopy);
521 shf->rp += ncopy;
522 shf->rnleft -= ncopy;
523 buf += ncopy;
524 bsize -= ncopy;
525 #ifdef MKSH_WITH_TEXTMODE
526 if (end && buf > orig_buf + 1 && buf[-2] == '\r') {
527 buf--;
528 bsize++;
529 buf[-1] = '\n';
530 }
531 #endif
532 } while (!end && bsize);
533 #ifdef MKSH_WITH_TEXTMODE
534 if (!bsize && buf[-1] == '\r') {
535 int c = shf_getc(shf);
536 if (c == '\n')
537 buf[-1] = '\n';
538 else if (c != -1)
539 shf_ungetc(c, shf);
540 }
541 #endif
542 *buf = '\0';
543 return (buf);
544 }
545
546 /* Returns the char read. Returns -1 for error and end of file. */
547 int
shf_getchar(struct shf * shf)548 shf_getchar(struct shf *shf)
549 {
550 if (!(shf->flags & SHF_RD))
551 internal_errorf(Tf_flags, "shf_getchar",
552 (unsigned int)shf->flags);
553
554 if (shf->rnleft == 0 && (shf_fillbuf(shf) == -1 || shf->rnleft == 0))
555 return (-1);
556 --shf->rnleft;
557 return (ord(*shf->rp++));
558 }
559
560 /*
561 * Put a character back in the input stream. Returns the character if
562 * successful, -1 if there is no room.
563 */
564 int
shf_ungetc(int c,struct shf * shf)565 shf_ungetc(int c, struct shf *shf)
566 {
567 if (!(shf->flags & SHF_RD))
568 internal_errorf(Tf_flags, "shf_ungetc",
569 (unsigned int)shf->flags);
570
571 if ((shf->flags & SHF_ERROR) || c == -1 ||
572 (shf->rp == shf->buf && shf->rnleft))
573 return (-1);
574
575 if ((shf->flags & SHF_WRITING) && shf_emptybuf(shf, EB_READSW) == -1)
576 return (-1);
577
578 if (shf->rp == shf->buf)
579 shf->rp = shf->buf + shf->rbsize;
580 if (shf->flags & SHF_STRING) {
581 /*
582 * Can unget what was read, but not something different;
583 * we don't want to modify a string.
584 */
585 if ((int)(shf->rp[-1]) != c)
586 return (-1);
587 shf->flags &= ~SHF_EOF;
588 shf->rp--;
589 shf->rnleft++;
590 return (c);
591 }
592 shf->flags &= ~SHF_EOF;
593 *--(shf->rp) = c;
594 shf->rnleft++;
595 return (c);
596 }
597
598 /*
599 * Write a character. Returns the character if successful, -1 if the
600 * char could not be written.
601 */
602 int
shf_putchar(int c,struct shf * shf)603 shf_putchar(int c, struct shf *shf)
604 {
605 if (!(shf->flags & SHF_WR))
606 internal_errorf(Tf_flags, "shf_putchar",
607 (unsigned int)shf->flags);
608
609 if (c == -1)
610 return (-1);
611
612 if (shf->flags & SHF_UNBUF) {
613 unsigned char cc = (unsigned char)c;
614 ssize_t n;
615
616 if (shf->fd < 0)
617 internal_errorf(Tf_sD_s, "shf_putchar", "no fd");
618 if (shf->flags & SHF_ERROR) {
619 errno = shf->errnosv;
620 return (-1);
621 }
622 while ((n = write(shf->fd, &cc, 1)) != 1)
623 if (n < 0) {
624 if (errno == EINTR &&
625 !(shf->flags & SHF_INTERRUPT))
626 continue;
627 shf->flags |= SHF_ERROR;
628 shf->errnosv = errno;
629 return (-1);
630 }
631 } else {
632 /* Flush deals with strings and sticky errors */
633 if (shf->wnleft == 0 && shf_emptybuf(shf, EB_GROW) == -1)
634 return (-1);
635 shf->wnleft--;
636 *shf->wp++ = c;
637 }
638
639 return (c);
640 }
641
642 /*
643 * Write a string. Returns the length of the string if successful, -1
644 * if the string could not be written.
645 */
646 ssize_t
shf_puts(const char * s,struct shf * shf)647 shf_puts(const char *s, struct shf *shf)
648 {
649 if (!s)
650 return (-1);
651
652 return (shf_write(s, strlen(s), shf));
653 }
654
655 /* Write a buffer. Returns nbytes if successful, -1 if there is an error. */
656 ssize_t
shf_write(const char * buf,ssize_t nbytes,struct shf * shf)657 shf_write(const char *buf, ssize_t nbytes, struct shf *shf)
658 {
659 ssize_t n, ncopy, orig_nbytes = nbytes;
660
661 if (!(shf->flags & SHF_WR))
662 internal_errorf(Tf_flags, Tshf_write,
663 (unsigned int)shf->flags);
664
665 if (nbytes < 0)
666 internal_errorf(Tf_szs, Tshf_write, nbytes, Tbytes);
667
668 /* Don't buffer if buffer is empty and we're writting a large amount. */
669 if ((ncopy = shf->wnleft) &&
670 (shf->wp != shf->buf || nbytes < shf->wnleft)) {
671 if (ncopy > nbytes)
672 ncopy = nbytes;
673 memcpy(shf->wp, buf, ncopy);
674 nbytes -= ncopy;
675 buf += ncopy;
676 shf->wp += ncopy;
677 shf->wnleft -= ncopy;
678 }
679 if (nbytes > 0) {
680 if (shf->flags & SHF_STRING) {
681 /* resize buffer until there's enough space left */
682 while (nbytes > shf->wnleft)
683 if (shf_emptybuf(shf, EB_GROW) == -1)
684 return (-1);
685 /* then write everything into the buffer */
686 } else {
687 /* flush deals with sticky errors */
688 if (shf_emptybuf(shf, EB_GROW) == -1)
689 return (-1);
690 /* write chunks larger than window size directly */
691 if (nbytes > shf->wbsize) {
692 ncopy = nbytes;
693 if (shf->wbsize)
694 ncopy -= nbytes % shf->wbsize;
695 nbytes -= ncopy;
696 while (ncopy > 0) {
697 n = write(shf->fd, buf, ncopy);
698 if (n < 0) {
699 if (errno == EINTR &&
700 !(shf->flags & SHF_INTERRUPT))
701 continue;
702 shf->flags |= SHF_ERROR;
703 shf->errnosv = errno;
704 shf->wnleft = 0;
705 /*
706 * Note: fwrite(3) returns 0
707 * for errors - this doesn't
708 */
709 return (-1);
710 }
711 buf += n;
712 ncopy -= n;
713 }
714 }
715 /* ... and buffer the rest */
716 }
717 if (nbytes > 0) {
718 /* write remaining bytes to buffer */
719 memcpy(shf->wp, buf, nbytes);
720 shf->wp += nbytes;
721 shf->wnleft -= nbytes;
722 }
723 }
724
725 return (orig_nbytes);
726 }
727
728 ssize_t
shf_fprintf(struct shf * shf,const char * fmt,...)729 shf_fprintf(struct shf *shf, const char *fmt, ...)
730 {
731 va_list args;
732 ssize_t n;
733
734 va_start(args, fmt);
735 n = shf_vfprintf(shf, fmt, args);
736 va_end(args);
737
738 return (n);
739 }
740
741 ssize_t
shf_snprintf(char * buf,ssize_t bsize,const char * fmt,...)742 shf_snprintf(char *buf, ssize_t bsize, const char *fmt, ...)
743 {
744 struct shf shf;
745 va_list args;
746 ssize_t n;
747
748 if (!buf || bsize <= 0)
749 internal_errorf("shf_snprintf: buf %zX, bsize %zd",
750 (size_t)buf, bsize);
751
752 shf_sopen(buf, bsize, SHF_WR, &shf);
753 va_start(args, fmt);
754 n = shf_vfprintf(&shf, fmt, args);
755 va_end(args);
756 /* NUL terminates */
757 shf_sclose(&shf);
758 return (n);
759 }
760
761 char *
shf_smprintf(const char * fmt,...)762 shf_smprintf(const char *fmt, ...)
763 {
764 struct shf shf;
765 va_list args;
766
767 shf_sopen(NULL, 0, SHF_WR|SHF_DYNAMIC, &shf);
768 va_start(args, fmt);
769 shf_vfprintf(&shf, fmt, args);
770 va_end(args);
771 /* NUL terminates */
772 return (shf_sclose(&shf));
773 }
774
775 #define FL_HASH 0x001 /* '#' seen */
776 #define FL_PLUS 0x002 /* '+' seen */
777 #define FL_RIGHT 0x004 /* '-' seen */
778 #define FL_BLANK 0x008 /* ' ' seen */
779 #define FL_SHORT 0x010 /* 'h' seen */
780 #define FL_LONG 0x020 /* 'l' seen */
781 #define FL_ZERO 0x040 /* '0' seen */
782 #define FL_DOT 0x080 /* '.' seen */
783 #define FL_UPPER 0x100 /* format character was uppercase */
784 #define FL_NUMBER 0x200 /* a number was formated %[douxefg] */
785 #define FL_SIZET 0x400 /* 'z' seen */
786 #define FM_SIZES 0x430 /* h/l/z mask */
787
788 ssize_t
shf_vfprintf(struct shf * shf,const char * fmt,va_list args)789 shf_vfprintf(struct shf *shf, const char *fmt, va_list args)
790 {
791 const char *s;
792 char c, *cp;
793 int tmp = 0, flags;
794 size_t field, precision, len;
795 unsigned long lnum;
796 /* %#o produces the longest output */
797 char numbuf[(8 * sizeof(long) + 2) / 3 + 1 + /* NUL */ 1];
798 /* this stuff for dealing with the buffer */
799 ssize_t nwritten = 0;
800
801 #define VA(type) va_arg(args, type)
802
803 if (!fmt)
804 return (0);
805
806 while ((c = *fmt++)) {
807 if (c != '%') {
808 shf_putc(c, shf);
809 nwritten++;
810 continue;
811 }
812 /*
813 * This will accept flags/fields in any order - not just
814 * the order specified in printf(3), but this is the way
815 * _doprnt() seems to work (on BSD and SYSV). The only
816 * restriction is that the format character must come
817 * last :-).
818 */
819 flags = 0;
820 field = precision = 0;
821 while ((c = *fmt++)) {
822 switch (c) {
823 case '#':
824 flags |= FL_HASH;
825 continue;
826
827 case '+':
828 flags |= FL_PLUS;
829 continue;
830
831 case '-':
832 flags |= FL_RIGHT;
833 continue;
834
835 case ' ':
836 flags |= FL_BLANK;
837 continue;
838
839 case '0':
840 if (!(flags & FL_DOT))
841 flags |= FL_ZERO;
842 continue;
843
844 case '.':
845 flags |= FL_DOT;
846 precision = 0;
847 continue;
848
849 case '*':
850 tmp = VA(int);
851 if (tmp < 0) {
852 if (flags & FL_DOT)
853 precision = 0;
854 else {
855 field = (unsigned int)-tmp;
856 flags |= FL_RIGHT;
857 }
858 } else if (flags & FL_DOT)
859 precision = (unsigned int)tmp;
860 else
861 field = (unsigned int)tmp;
862 continue;
863
864 case 'l':
865 flags &= ~FM_SIZES;
866 flags |= FL_LONG;
867 continue;
868
869 case 'h':
870 flags &= ~FM_SIZES;
871 flags |= FL_SHORT;
872 continue;
873
874 case 'z':
875 flags &= ~FM_SIZES;
876 flags |= FL_SIZET;
877 continue;
878 }
879 if (ctype(c, C_DIGIT)) {
880 bool overflowed = false;
881
882 tmp = ksh_numdig(c);
883 while (ctype((c = *fmt++), C_DIGIT))
884 if (notok2mul(2147483647, tmp, 10))
885 overflowed = true;
886 else
887 tmp = tmp * 10 + ksh_numdig(c);
888 --fmt;
889 if (overflowed)
890 tmp = 0;
891 if (flags & FL_DOT)
892 precision = (unsigned int)tmp;
893 else
894 field = (unsigned int)tmp;
895 continue;
896 }
897 break;
898 }
899
900 if (!c)
901 /* nasty format */
902 break;
903
904 if (ctype(c, C_UPPER)) {
905 flags |= FL_UPPER;
906 c = ksh_tolower(c);
907 }
908
909 switch (c) {
910 case 'd':
911 case 'i':
912 if (flags & FL_SIZET)
913 lnum = (long)VA(ssize_t);
914 else if (flags & FL_LONG)
915 lnum = VA(long);
916 else if (flags & FL_SHORT)
917 lnum = (long)(short)VA(int);
918 else
919 lnum = (long)VA(int);
920 goto integral;
921
922 case 'o':
923 case 'u':
924 case 'x':
925 if (flags & FL_SIZET)
926 lnum = VA(size_t);
927 else if (flags & FL_LONG)
928 lnum = VA(unsigned long);
929 else if (flags & FL_SHORT)
930 lnum = (unsigned long)(unsigned short)VA(int);
931 else
932 lnum = (unsigned long)VA(unsigned int);
933
934 integral:
935 flags |= FL_NUMBER;
936 cp = numbuf + sizeof(numbuf);
937 *--cp = '\0';
938
939 switch (c) {
940 case 'd':
941 case 'i':
942 if (0 > (long)lnum) {
943 lnum = -(long)lnum;
944 tmp = 1;
945 } else
946 tmp = 0;
947 /* FALLTHROUGH */
948 case 'u':
949 do {
950 *--cp = digits_lc[lnum % 10];
951 lnum /= 10;
952 } while (lnum);
953
954 if (c != 'u') {
955 if (tmp)
956 *--cp = '-';
957 else if (flags & FL_PLUS)
958 *--cp = '+';
959 else if (flags & FL_BLANK)
960 *--cp = ' ';
961 }
962 break;
963
964 case 'o':
965 do {
966 *--cp = digits_lc[lnum & 0x7];
967 lnum >>= 3;
968 } while (lnum);
969
970 if ((flags & FL_HASH) && *cp != '0')
971 *--cp = '0';
972 break;
973
974 case 'x': {
975 const char *digits = (flags & FL_UPPER) ?
976 digits_uc : digits_lc;
977 do {
978 *--cp = digits[lnum & 0xF];
979 lnum >>= 4;
980 } while (lnum);
981
982 if (flags & FL_HASH) {
983 *--cp = (flags & FL_UPPER) ? 'X' : 'x';
984 *--cp = '0';
985 }
986 }
987 }
988 len = numbuf + sizeof(numbuf) - 1 - (s = cp);
989 if (flags & FL_DOT) {
990 if (precision > len) {
991 field = precision;
992 flags |= FL_ZERO;
993 } else
994 /* no loss */
995 precision = len;
996 }
997 break;
998
999 case 's':
1000 if ((s = VA(const char *)) == NULL)
1001 s = "(null)";
1002 else if (flags & FL_HASH) {
1003 print_value_quoted(shf, s);
1004 continue;
1005 }
1006 len = utf_mbswidth(s);
1007 break;
1008
1009 case 'c':
1010 flags &= ~FL_DOT;
1011 c = (char)(VA(int));
1012 /* FALLTHROUGH */
1013
1014 case '%':
1015 default:
1016 numbuf[0] = c;
1017 numbuf[1] = 0;
1018 s = numbuf;
1019 len = 1;
1020 break;
1021 }
1022
1023 /*
1024 * At this point s should point to a string that is to be
1025 * formatted, and len should be the length of the string.
1026 */
1027 if (!(flags & FL_DOT) || len < precision)
1028 precision = len;
1029 if (field > precision) {
1030 field -= precision;
1031 if (!(flags & FL_RIGHT)) {
1032 /* skip past sign or 0x when padding with 0 */
1033 if ((flags & FL_ZERO) && (flags & FL_NUMBER)) {
1034 if (ctype(*s, C_SPC | C_PLUS | C_MINUS)) {
1035 shf_putc(*s, shf);
1036 s++;
1037 precision--;
1038 nwritten++;
1039 } else if (*s == '0') {
1040 shf_putc(*s, shf);
1041 s++;
1042 nwritten++;
1043 if (--precision &&
1044 ksh_eq(*s, 'X', 'x')) {
1045 shf_putc(*s, shf);
1046 s++;
1047 precision--;
1048 nwritten++;
1049 }
1050 }
1051 c = '0';
1052 } else
1053 c = flags & FL_ZERO ? '0' : ' ';
1054 nwritten += field;
1055 while (field--)
1056 shf_putc(c, shf);
1057 field = 0;
1058 } else
1059 c = ' ';
1060 } else
1061 field = 0;
1062
1063 nwritten += precision;
1064 precision = utf_skipcols(s, precision, &tmp) - s;
1065 while (precision--)
1066 shf_putc(*s++, shf);
1067
1068 nwritten += field;
1069 while (field--)
1070 shf_putc(c, shf);
1071 }
1072
1073 return (shf_error(shf) ? -1 : nwritten);
1074 }
1075
1076 #if defined(MKSH_SMALL) && !defined(MKSH_SMALL_BUT_FAST)
1077 int
shf_getc(struct shf * shf)1078 shf_getc(struct shf *shf)
1079 {
1080 return (shf_getc_i(shf));
1081 }
1082
1083 int
shf_putc(int c,struct shf * shf)1084 shf_putc(int c, struct shf *shf)
1085 {
1086 return (shf_putc_i(c, shf));
1087 }
1088 #endif
1089
1090 #ifdef DEBUG
1091 const char *
cstrerror(int errnum)1092 cstrerror(int errnum)
1093 {
1094 #undef strerror
1095 return (strerror(errnum));
1096 #define strerror dontuse_strerror /* poisoned */
1097 }
1098 #elif !HAVE_STRERROR
1099
1100 #if HAVE_SYS_ERRLIST
1101 #if !HAVE_SYS_ERRLIST_DECL
1102 extern const int sys_nerr;
1103 extern const char * const sys_errlist[];
1104 #endif
1105 #endif
1106
1107 const char *
cstrerror(int errnum)1108 cstrerror(int errnum)
1109 {
1110 /* "Unknown error: " + sign + rough estimate + NUL */
1111 static char errbuf[15 + 1 + (8 * sizeof(int) + 2) / 3 + 1];
1112
1113 #if HAVE_SYS_ERRLIST
1114 if (errnum > 0 && errnum < sys_nerr && sys_errlist[errnum])
1115 return (sys_errlist[errnum]);
1116 #endif
1117
1118 switch (errnum) {
1119 case 0:
1120 return ("Undefined error: 0");
1121 case EPERM:
1122 return ("Operation not permitted");
1123 case ENOENT:
1124 return ("No such file or directory");
1125 #ifdef ESRCH
1126 case ESRCH:
1127 return ("No such process");
1128 #endif
1129 #ifdef E2BIG
1130 case E2BIG:
1131 return ("Argument list too long");
1132 #endif
1133 case ENOEXEC:
1134 return ("Exec format error");
1135 case EBADF:
1136 return ("Bad file descriptor");
1137 #ifdef ENOMEM
1138 case ENOMEM:
1139 return ("Cannot allocate memory");
1140 #endif
1141 case EACCES:
1142 return ("Permission denied");
1143 case EEXIST:
1144 return ("File exists");
1145 case ENOTDIR:
1146 return ("Not a directory");
1147 #ifdef EINVAL
1148 case EINVAL:
1149 return ("Invalid argument");
1150 #endif
1151 #ifdef ELOOP
1152 case ELOOP:
1153 return ("Too many levels of symbolic links");
1154 #endif
1155 default:
1156 shf_snprintf(errbuf, sizeof(errbuf),
1157 "Unknown error: %d", errnum);
1158 return (errbuf);
1159 }
1160 }
1161 #endif
1162
1163 /* fast character classes */
1164 const uint32_t tpl_ctypes[128] = {
1165 /* 0x00 */
1166 CiNUL, CiCNTRL, CiCNTRL, CiCNTRL,
1167 CiCNTRL, CiCNTRL, CiCNTRL, CiCNTRL,
1168 CiCNTRL, CiTAB, CiNL, CiSPX,
1169 CiSPX, CiCR, CiCNTRL, CiCNTRL,
1170 /* 0x10 */
1171 CiCNTRL, CiCNTRL, CiCNTRL, CiCNTRL,
1172 CiCNTRL, CiCNTRL, CiCNTRL, CiCNTRL,
1173 CiCNTRL, CiCNTRL, CiCNTRL, CiCNTRL,
1174 CiCNTRL, CiCNTRL, CiCNTRL, CiCNTRL,
1175 /* 0x20 */
1176 CiSP, CiALIAS | CiVAR1, CiQC, CiHASH,
1177 CiSS, CiPERCT, CiQCL, CiQC,
1178 CiQCL, CiQCL, CiQCX | CiVAR1, CiPLUS,
1179 CiALIAS, CiMINUS, CiALIAS, CiQCM,
1180 /* 0x30 */
1181 CiOCTAL, CiOCTAL, CiOCTAL, CiOCTAL,
1182 CiOCTAL, CiOCTAL, CiOCTAL, CiOCTAL,
1183 CiDIGIT, CiDIGIT, CiCOLON, CiQCL,
1184 CiANGLE, CiEQUAL, CiANGLE, CiQUEST,
1185 /* 0x40 */
1186 CiALIAS | CiVAR1, CiUPPER | CiHEXLT,
1187 CiUPPER | CiHEXLT, CiUPPER | CiHEXLT,
1188 CiUPPER | CiHEXLT, CiUPPER | CiHEXLT,
1189 CiUPPER | CiHEXLT, CiUPPER,
1190 CiUPPER, CiUPPER, CiUPPER, CiUPPER,
1191 CiUPPER, CiUPPER, CiUPPER, CiUPPER,
1192 /* 0x50 */
1193 CiUPPER, CiUPPER, CiUPPER, CiUPPER,
1194 CiUPPER, CiUPPER, CiUPPER, CiUPPER,
1195 CiUPPER, CiUPPER, CiUPPER, CiQCX | CiBRACK,
1196 CiQCX, CiBRACK, CiQCM, CiUNDER,
1197 /* 0x60 */
1198 CiGRAVE, CiLOWER | CiHEXLT,
1199 CiLOWER | CiHEXLT, CiLOWER | CiHEXLT,
1200 CiLOWER | CiHEXLT, CiLOWER | CiHEXLT,
1201 CiLOWER | CiHEXLT, CiLOWER,
1202 CiLOWER, CiLOWER, CiLOWER, CiLOWER,
1203 CiLOWER, CiLOWER, CiLOWER, CiLOWER,
1204 /* 0x70 */
1205 CiLOWER, CiLOWER, CiLOWER, CiLOWER,
1206 CiLOWER, CiLOWER, CiLOWER, CiLOWER,
1207 CiLOWER, CiLOWER, CiLOWER, CiCURLY,
1208 CiQCL, CiCURLY, CiQCM, CiCNTRL
1209 };
1210
1211 void
set_ifs(const char * s)1212 set_ifs(const char *s)
1213 {
1214 #if defined(MKSH_EBCDIC) || defined(MKSH_FAUX_EBCDIC)
1215 int i = 256;
1216
1217 memset(ksh_ctypes, 0, sizeof(ksh_ctypes));
1218 while (i--)
1219 if (ebcdic_map[i] < 0x80U)
1220 ksh_ctypes[i] = tpl_ctypes[ebcdic_map[i]];
1221 #else
1222 memcpy(ksh_ctypes, tpl_ctypes, sizeof(tpl_ctypes));
1223 memset((char *)ksh_ctypes + sizeof(tpl_ctypes), '\0',
1224 sizeof(ksh_ctypes) - sizeof(tpl_ctypes));
1225 #endif
1226 ifs0 = *s;
1227 while (*s)
1228 ksh_ctypes[ord(*s++)] |= CiIFS;
1229 }
1230
1231 #if defined(MKSH_EBCDIC) || defined(MKSH_FAUX_EBCDIC)
1232 #include <locale.h>
1233
1234 /*
1235 * Many headaches with EBCDIC:
1236 * 1. There are numerous EBCDIC variants, and it is not feasible for us
1237 * to support them all. But we can support the EBCDIC code pages that
1238 * contain all (most?) of the characters in ASCII, and these
1239 * usually tend to agree on the code points assigned to the ASCII
1240 * subset. If you need a representative example, look at EBCDIC 1047,
1241 * which is first among equals in the IBM MVS development
1242 * environment: https://en.wikipedia.org/wiki/EBCDIC_1047
1243 * Unfortunately, the square brackets are not consistently mapped,
1244 * and for certain reasons, we need an unambiguous bijective
1245 * mapping between EBCDIC and "extended ASCII".
1246 * 2. Character ranges that are contiguous in ASCII, like the letters
1247 * in [A-Z], are broken up into segments (i.e. [A-IJ-RS-Z]), so we
1248 * can't implement e.g. islower() as { return c >= 'a' && c <= 'z'; }
1249 * because it will also return true for a handful of extraneous
1250 * characters (like the plus-minus sign at 0x8F in EBCDIC 1047, a
1251 * little after 'i'). But at least '_' is not one of these.
1252 * 3. The normal [0-9A-Za-z] characters are at codepoints beyond 0x80.
1253 * Not only do they require all 8 bits instead of 7, if chars are
1254 * signed, they will have negative integer values! Something like
1255 * (c - 'A') could actually become (c + 63)! Use the ord() macro to
1256 * ensure you're getting a value in [0, 255] (ORD for constants).
1257 * 4. '\n' is actually NL (0x15, U+0085) instead of LF (0x25, U+000A).
1258 * EBCDIC has a proper newline character instead of "emulating" one
1259 * with line feeds, although this is mapped to LF for our purposes.
1260 * 5. Note that it is possible to compile programs in ASCII mode on IBM
1261 * mainframe systems, using the -qascii option to the XL C compiler.
1262 * We can determine the build mode by looking at __CHARSET_LIB:
1263 * 0 == EBCDIC, 1 == ASCII
1264 */
1265
1266 void
ebcdic_init(void)1267 ebcdic_init(void)
1268 {
1269 int i = 256;
1270 unsigned char t;
1271 bool mapcache[256];
1272
1273 while (i--)
1274 ebcdic_rtt_toascii[i] = i;
1275 memset(ebcdic_rtt_fromascii, 0xFF, sizeof(ebcdic_rtt_fromascii));
1276 setlocale(LC_ALL, "");
1277 #ifdef MKSH_EBCDIC
1278 if (__etoa_l(ebcdic_rtt_toascii, 256) != 256) {
1279 write(2, "mksh: could not map EBCDIC to ASCII\n", 36);
1280 exit(255);
1281 }
1282 #endif
1283
1284 memset(mapcache, 0, sizeof(mapcache));
1285 i = 256;
1286 while (i--) {
1287 t = ebcdic_rtt_toascii[i];
1288 /* ensure unique round-trip capable mapping */
1289 if (mapcache[t]) {
1290 write(2, "mksh: duplicate EBCDIC to ASCII mapping\n", 40);
1291 exit(255);
1292 }
1293 /*
1294 * since there are 256 input octets, this also ensures
1295 * the other mapping direction is completely filled
1296 */
1297 mapcache[t] = true;
1298 /* fill the complete round-trip map */
1299 ebcdic_rtt_fromascii[t] = i;
1300 /*
1301 * Only use the converted value if it's in the range
1302 * [0x00; 0x7F], which I checked; the "extended ASCII"
1303 * characters can be any encoding, not just Latin1,
1304 * and the C1 control characters other than NEL are
1305 * hopeless, but we map EBCDIC NEL to ASCII LF so we
1306 * cannot even use C1 NEL.
1307 * If ever we map to Unicode, bump the table width to
1308 * an unsigned int, and or the raw unconverted EBCDIC
1309 * values with 0x01000000 instead.
1310 */
1311 if (t < 0x80U)
1312 ebcdic_map[i] = (unsigned short)ord(t);
1313 else
1314 ebcdic_map[i] = (unsigned short)(0x100U | ord(i));
1315 }
1316 if (ebcdic_rtt_toascii[0] || ebcdic_rtt_fromascii[0] || ebcdic_map[0]) {
1317 write(2, "mksh: NUL not at position 0\n", 28);
1318 exit(255);
1319 }
1320 }
1321 #endif
1322