/* m_getfld.c - read/parse a message */ #include "../h/mh.h" #include #include "../zotnet/mts.h" #include /* This module has a long and checkered history. First, it didn't burst maildrops correctly because it considered two CTRL-A:s in a row to be an inter-message delimiter. It really is four CTRL-A:s followed by a newline. Unfortunately, MMDF will convert this delimiter *inside* a message to a CTRL-B followed by three CTRL-A:s and a newline. This caused the old version of m_getfld() to declare eom prematurely. The fix was a lot slower than c == '\001' && peekc (iob) == '\001' but it worked, and to increase generality, UUCP style maildrops could be parsed as well. Unfortunately the speed issue finally caught up with us since this routine is at the very heart of MH. To speed things up considerably, the routine Eom() was made an auxilary function called by the macro eom(). Unless we are bursting a maildrop, the eom() macro returns FALSE saying we aren't at the end of the message. The next thing to do is to read the mtstailor file and initialize delimiter[] and delimlen accordingly... After mhl was made a built-in in msh, m_getfld() worked just fine (using m_unknown() at startup). Until one day: a message which was the result of a bursting was shown. Then, since the burst boundaries aren't CTRL-A:s, m_getfld() would blinding plunge on past the boundary. Very sad. The solution: introduce m_eomsbr(). This hook gets called after the end of each line (since testing for eom involves an fseek()). This worked fine, until one day: a message with no body portion arrived. Then the while (eom (c = Getc (iob), iob)) continue; loop caused m_getfld() to return FMTERR. So, that logic was changed to check for (*eom_action) and act accordingly. This worked fine, until one day: someone didn't use four CTRL:A's as their delimiters. So, the bullet got bit and we read mts.h and continue to struggle on. It's not that bad though, since the only time the code gets executed is when inc (or msh) calls it, and both of these have already called mts_init(). ------------------------ (Written by Van Jacobson for the mh6 m_getfld, January, 1986): This routine was accounting for 60% of the cpu time used by most mh programs. I spent a bit of time tuning and it now accounts for <10% of the time used. Like any heavily tuned routine, it's a bit complex and you want to be sure you understand everything that it's doing before you start hacking on it. Let me try to emphasize that: every line in this atrocity depends on every other line, sometimes in subtle ways. You should understand it all, in detail, before trying to change any part. If you do change it, test the result thoroughly (I use a hand-constructed test file that exercises all the ways a header name, header body, header continuation, header-body separator, body line and body eom can align themselves with respect to a buffer boundary). "Minor" bugs in this routine result in garbaged or lost mail. If you hack on this and slow it down, I, my children and my children's children will curse you. This routine gets used on three different types of files: normal, single msg files, "packed" unix or mmdf mailboxs (when used by inc) and packed, directoried bulletin board files (when used by msh). The biggest impact of different file types is in "eom" testing. The code has been carefully organized to test for eom at appropriate times and at no other times (since the check is quite expensive). I have tried to arrange things so that the eom check need only be done on entry to this routine. Since an eom can only occur after a newline, this is easy to manage for header fields. For the msg body, we try to efficiently search the input buffer to see if contains the eom delimiter. If it does, we take up to the delimiter, otherwise we take everything in the buffer. (The change to the body eom/copy processing produced the most noticeable performance difference, particularly for "inc" and "show".) There are three qualitatively different things this routine busts out of a message: field names, field text and msg bodies. Field names are typically short (~8 char) and the loop that extracts them might terminate on a colon, newline or max width. I considered using a Vax "scanc" to locate the end of the field followed by a "bcopy" but the routine call overhead on a Vax is too large for this to work on short names. If Berkeley ever makes "inline" part of the C optimiser (so things like "scanc" turn into inline instructions) a change here would be worthwhile. Field text is typically 60 - 100 characters so there's (barely) a win in doing a routine call to something that does a "locc" followed by a "bmove". About 30% of the fields have continuations (usually the 822 "received:" lines) and each continuation generates another routine call. "Inline" would be a big win here, as well. Messages, as of this writing, seem to come in two flavors: small (~1K) and long (>2K). Most messages have 400 - 600 bytes of headers so message bodies average at least a few hundred characters. Assuming your system uses reasonably sized stdio buffers (1K or more), this routine should be able to remove the body in large (>500 byte) chunks. The makes the cost of a call to "bcopy" small but there is a premium on checking for the eom in packed maildrops. The eom pattern is always a simple string so we can construct an efficient pattern matcher for it (e.g., a Vax "matchc" instruction). Some thought went into recognizing the start of an eom that has been split across two buffers. This routine wants to deal with large chunks of data so, rather than "getc" into a local buffer, it uses stdio's buffer. If you try to use it on a non-buffered file, you'll get what you deserve. This routine "knows" that struct FILEs have a _ptr and a _cnt to describe the current state of the buffer and it knows that _filbuf ignores the _ptr & _cnt and simply fills the buffer. If stdio on your system doesn't work this way, you may have to make small changes in this routine. This routine also "knows" that an EOF indication on a stream is "sticky" (i.e., you will keep getting EOF until you reposition the stream). If your system doesn't work this way it is broken and you should complain to the vendor. As a consequence of the sticky EOF, this routine will never return any kind of EOF status when there is data in "name" or "buf"). */ #define Getc(iob) getc(iob) #define eom(c,iob) (msg_style != MS_DEFAULT && \ (((c) == *msg_delim && m_Eom(c,iob)) ||\ (eom_action && (*eom_action)(c)))) static char *matchc(); static char *locc(); static char **pat_map; int msg_count = 0; /* disgusting hack for "inc" so it can * know how many characters were stuffed * in the buffer on the last call (see * comments in uip/scansbr.c) */ int msg_style = MS_DEFAULT; /* * The "full" delimiter string for a packed maildrop consists * of a newline followed by the actual delimiter. E.g., the * full string for a Unix maildrop would be: "\n\nFrom ". * "Fdelim" points to the start of the full string and is used * in the BODY case of the main routine to search the buffer for * a possible eom. Msg_delim points to the first character of * the actual delim. string (i.e., fdelim+1). Edelim * points to the 2nd character of actual delimiter string. It * is used in m_Eom because the first character of the string * has been read and matched before m_Eom is called. */ char *msg_delim = ""; static char *fdelim; static char *delimend; static int fdelimlen; static char *edelim; static int edelimlen; static int (*eom_action) () = NULL; /* */ m_getfld (state, name, buf, bufsz, iob) int state; int bufsz; char *name, *buf; register FILE *iob; { register char *cp; register char *bp; register int cnt; register int c; register int i; register int j; register char *ep; register char *sp; if ((c = Getc(iob)) < 0) { msg_count = 0; *buf = 0; return FILEEOF; } if (eom (c, iob)) { if (! eom_action) { /* flush null messages */ while ((c = Getc(iob)) >= 0 && eom (c, iob)) ; if (c >= 0) (void) ungetc(c, iob); } msg_count = 0; *buf = 0; return FILEEOF; } switch (state) { case FLDEOF: case BODYEOF: case FLD: if (c == '\n' || c == '-') { /* we hit the header/body separator */ while (c != '\n' && (c = Getc(iob)) >= 0) ; if (c < 0 || (c = Getc(iob)) < 0 || eom (c, iob)) { if (! eom_action) { /* flush null messages */ while ((c = Getc(iob)) >= 0 && eom (c, iob)) ; if (c >= 0) (void) ungetc(c, iob); } msg_count = 0; *buf = 0; return FILEEOF; } state = BODY; goto body; } /* * get the name of this component. take characters up * to a ':', a newline or NAMESZ-1 characters, whichever * comes first. */ cp = name; i = NAMESZ - 1; for (;;) { bp = sp = iob->_ptr - 1; j = (cnt = iob->_cnt+1) < i ? cnt : i; while ((c = *bp++) != ':' && c != '\n' && --j >= 0) *cp++ = c; j = bp - sp; if ((cnt -= j) <= 0) { if (_filbuf(iob) == EOF) { *cp = *buf = NULL; advise (NULLCP, "eof encountered in field \"%s\"", name); return FMTERR; } } else { iob->_ptr = bp + 1; iob->_cnt = cnt - 1; } if (c == ':') break; /* * something went wrong. possibilities are: * . hit a newline (error) * . got more than namesz chars. (error) * . hit the end of the buffer. (loop) */ if (c == '\n') { *cp = *buf = NULL; advise (NULLCP, "eol encountered in field \"%s\"", name); state = FMTERR; goto finish; } if ((i -= j) <= 0) { *cp = *buf = NULL; advise (NULLCP, "field name \"%s\" exceeds %d bytes", name, NAMESZ - 1); state = LENERR; goto finish; } } while (isspace (*--cp) && cp >= name) ; *++cp = NULL; /* fall through */ case FLDPLUS: /* * get (more of) the text of a field. take * characters up to the end of this field (newline * followed by non-blank) or bufsz-1 characters. */ cp = buf; i = bufsz-1; for (;;) { cnt = iob->_cnt++; bp = --iob->_ptr; c = cnt < i ? cnt : i; while (ep = locc( c, bp, '\n' )) { /* * if we hit the end of this field, return. */ if ((j = *++ep) != ' ' && j != '\t') { j = ep - iob->_ptr; (void) bcopy( iob->_ptr, cp, j); iob->_ptr = ep; iob->_cnt -= j; cp += j; state = FLD; goto finish; } c -= ep - bp; bp = ep; } /* * end of input or dest buffer - copy what we've found. */ c += bp - iob->_ptr; (void) bcopy( iob->_ptr, cp, c); i -= c; cp += c; if (i <= 0) { /* the dest buffer is full */ iob->_cnt -= c; iob->_ptr += c; state = FLDPLUS; break; } /* * There's one character left in the input buffer. * Copy it & fill the buffer. If the last char * was a newline and the next char is not whitespace, * this is the end of the field. Otherwise loop. */ --i; *cp++ = j = *(iob->_ptr + c); c = _filbuf(iob); if (j == '\n' && c != ' ' && c != '\t') { if (c != EOF) --iob->_ptr, ++iob->_cnt; state = FLD; break; } } break; case BODY: body: /* * get the message body up to bufsz characters or the * end of the message. Sleazy hack: if bufsz is negative * we assume that we were called to copy directly into * the output buffer and we don't add an eos. */ i = (bufsz < 0) ? -bufsz : bufsz-1; bp = --iob->_ptr; cnt = ++iob->_cnt; c = (cnt < i ? cnt : i); if (msg_style != MS_DEFAULT && c > 1) { /* * packed maildrop - only take up to the (possible) * start of the next message. This "matchc" should * probably be a Boyer-Moore matcher for non-vaxen, * particularly since we have the alignment table * all built for the end-of-buffer test (next). * But our vax timings indicate that the "matchc" * instruction is 50% faster than a carefully coded * B.M. matcher for most strings. (So much for elegant * algorithms vs. brute force.) Since I (currently) * run MH on a vax, we use the matchc instruction. --vj */ if (ep = matchc( fdelimlen, fdelim, c, bp ) ) c = ep - bp + 1; else { /* * There's no delim in the buffer but there may be * a partial one at the end. If so, we want to leave * it so the "eom" check on the next call picks it up. * Use a modified Boyer-Moore matcher to make this * check relatively cheap. The first "while" figures * out what position in the pattern matches the last * character in the buffer. The inner "while" matches * the pattern against the buffer, backwards starting * at that position. Note that unless the buffer * ends with one of the characters in the pattern * (excluding the first and last), we do only one test. */ sp = delimend; ep = bp + c - 1; while ((cp = pat_map[*ep]) < sp) { ep = bp + c - 1; sp = cp; while (*--ep == *--cp && cp > fdelim) ; if (cp == fdelim) { if (*ep == *cp && ep > bp) c = (ep - bp) + 1; break; } } } } (void) bcopy( bp, buf, c ); iob->_cnt -= c; iob->_ptr += c; if (bufsz < 0) { msg_count = c; return (state); } cp = buf + c; break; default: adios (NULLCP, "m_getfld() called with bogus state of %d", state); } finish:; *cp = NULL; msg_count = cp - buf; return (state); } /* */ #ifdef RPATHS static char unixbuf[BUFSIZ] = ""; #endif RPATHS void m_unknown (iob) register FILE *iob; { register int c; register long pos; char text[10]; register char *cp; msg_style = MS_UNKNOWN; /* Figure out what the message delimitter string is for this * maildrop. (This used to be part of m_Eom but I didn't like * the idea of an "if" statement that could only succeed on the * first call to m_Eom getting executed on each call, i.e., at * every newline in the message). * * If the first line of the maildrop is a Unix "from" line, we say the * style is UUCP and eat the rest of the line. Otherwise we say the style * is MMDF & look for the delimiter string specified when MH was built * (or from the mtstailor file). */ pos = ftell (iob); if (fread (text, sizeof *text, 5, iob) == 5 && strncmp (text, "From ", 5) == 0) { msg_style = MS_UUCP; fdelim = "\n\nFrom "; #ifndef RPATHS while ((c = getc (iob)) != '\n' && c >= 0) ; #else RPATHS cp = unixbuf; while ((c = getc (iob)) != '\n') *cp++ = c; *cp = NULL; #endif RPATHS } else { /* not a Unix style maildrop */ (void) fseek (iob, pos, 0); if (mmdlm2 == NULLCP || *mmdlm2 == NULL) mmdlm2 = "\001\001\001\001\n"; fdelim = (char *)malloc((unsigned)strlen(mmdlm2)+2); *fdelim = '\n'; (void)strcpy(fdelim+1, mmdlm2); msg_style = MS_MMDF; } fdelimlen = strlen(fdelim); msg_delim = fdelim+1; edelim = msg_delim+1; edelimlen = fdelimlen-2; delimend = msg_delim + edelimlen; if (edelimlen <= 1) adios (NULLCP, "maildrop delimiter must be at least 2 bytes"); /* * build a Boyer-Moore end-position map for the matcher in m_getfld. * N.B. - we don't match just the first char (since it's the newline * separator) or the last char (since the matchc would have found it * if it was a real delim). */ pat_map = (char **) malloc( 256 * sizeof (char *)); for (c = 256; c--; ) pat_map[c] = delimend + 1; for (cp = fdelim + 1; cp < delimend; cp++ ) pat_map[*cp] = cp; if (msg_style == MS_MMDF) { /* flush extra msg hdrs */ while ((c = Getc(iob)) >= 0 && eom (c, iob)) ; if (c >= 0) (void) ungetc(c, iob); } } void m_eomsbr (action) int (*action) (); { if (eom_action = action) { msg_style = MS_MSH; *msg_delim = 0; fdelimlen = 1; delimend = fdelim; } else { msg_style = MS_MMDF; msg_delim = fdelim + 1; fdelimlen = strlen (fdelim); delimend = msg_delim + edelimlen; } } /* */ /* test for msg delimiter string */ int m_Eom (c, iob) register int c; register FILE *iob; { register long pos = 0L; register int i; char text[10]; #ifdef RPATHS register char *cp; #endif RPATHS pos = ftell (iob); if ((i = fread (text, sizeof *text, edelimlen, iob)) != edelimlen || strncmp (text, edelim, edelimlen)) { if (i == 0 && msg_style == MS_UUCP) /* the final newline in the (brain damaged) unix-format * maildrop is part of the delimitter - delete it. */ return 1; (void) fseek (iob, pos, 0); return 0; } if (msg_style == MS_UUCP) { #ifndef RPATHS while ((c = getc (iob)) != '\n') if (c < 0) break; #else RPATHS cp = unixbuf; while ((c = getc (iob)) != '\n' && c >= 0) *cp++ = c; *cp = NULL; #endif RPATHS } return 1; } /* */ #ifdef RPATHS char *unixline () { register char *cp, *dp, *pp; static char unixfrom[BUFSIZ]; pp = unixfrom; if (cp = dp = index (unixbuf, ' ')) { while (cp = index (cp + 1, 'r')) if (strncmp (cp, "remote from ", 12) == 0) { *cp = NULL; (void) sprintf (pp, "%s!", cp + 12); pp += strlen (pp); break; } if (cp == NULL) cp = unixbuf + strlen (unixbuf); if ((cp -= 25) >= dp) *cp = NULL; } (void) sprintf (pp, "%s\n", unixbuf); unixbuf[0] = NULL; return unixfrom; } #endif RPATHS /* */ #if (vax && !lint) asm(".align 1"); asm("_matchc: .word 0"); asm(" movq 4(ap),r0"); asm(" movq 12(ap),r2"); asm(" matchc r0,(r1),r2,(r3)"); asm(" beql 1f"); asm(" movl 4(ap),r3"); asm("1: subl3 4(ap),r3,r0"); asm(" ret"); #else static char * matchc( patln, pat, strln, str ) int patln; char *pat; int strln; register char *str; { register char *es = str + strln - patln; register char *sp; register char *pp; register char *ep = pat + patln; register char pc = *pat++; for(;;) { while (pc != *str++) if (str > es) return 0; sp = str; pp = pat; while (pp < ep && *sp++ == *pp++) ; if (pp >= ep) return (--str); } } #endif /* */ /* * Locate character "term" in the next "cnt" characters of "src". * If found, return its address, otherwise return 0. */ #if (vax && !lint) asm(".align 1"); asm("_locc: .word 0"); asm(" movq 4(ap),r0"); asm(" locc 12(ap),r0,(r1)"); asm(" beql 1f"); asm(" movl r1,r0"); asm("1: ret"); #else static char * locc( cnt, src, term ) register int cnt; register char *src; register char term; { while (*src++ != term && --cnt > 0); return (cnt > 0 ? --src : NULLCP); } #endif /* */ #if !defined (BSD42) && !defined (bcopy) int bcmp (b1, b2, length) register char *b1, *b2; register int length; { while (length-- > 0) if (*b1++ != *b2++) return 1; return 0; } bcopy (b1, b2, length) register char *b1, *b2; register int length; { while (length-- > 0) *b2++ = *b1++; } bzero (b, length) register char *b; register int length; { while (length-- > 0) *b++ = NULL; } #endif not BSD42 or SYS5