]> code.citadel.org Git - citadel.git/blobdiff - citadel/mime_parser.c
* Changed the comments at the beginning of each file to a consistent format
[citadel.git] / citadel / mime_parser.c
index a882bc4ebcd68218a8b4440f16b6292c274f745c..34decdd0211b798209210f1d6f292fe0e560108a 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * mime_parser.c
+ * $Id$
  *
  * This is a really bad attempt at writing a parser to handle MIME-encoded
  * messages.
@@ -9,6 +9,7 @@
  *
  */
 
+#include "sysdep.h"
 #include <stdlib.h>
 #include <unistd.h>
 #include <stdio.h>
 #include <sys/types.h>
 #include <ctype.h>
 #include <string.h>
+#include <sys/stat.h>
+#include <errno.h>
+#include "citadel.h"
 #include "mime_parser.h"
+#include "sysdep_decls.h"
+#include "server.h"
 
 
 
-void extract_key(char *target, char *source, char *key) {
+void extract_key(char *target, char *source, char *key)
+{
        int a, b;
 
        strcpy(target, source);
-       for (a=0; a<strlen(target); ++a) {
+       for (a = 0; a < strlen(target); ++a) {
                if ((!strncasecmp(&target[a], key, strlen(key)))
-                  && (target[a+strlen(key)]=='=')) {
-                       strcpy(target, &target[a+strlen(key)+1]);
-                       if (target[0]==34) strcpy(target, &target[1]);
-                       for (b=0; b<strlen(target); ++b)
-                               if (target[b]==34) target[b]=0;
+                   && (target[a + strlen(key)] == '=')) {
+                       strcpy(target, &target[a + strlen(key) + 1]);
+                       if (target[0] == 34)
+                               strcpy(target, &target[1]);
+                       for (b = 0; b < strlen(target); ++b)
+                               if (target[b] == 34)
+                                       target[b] = 0;
                        return;
-                       }
                }
-       strcpy(target, "");
        }
+       strcpy(target, "");
+}
 
 
 
-       /**** OTHERWISE, HERE'S WHERE WE HANDLE THE STUFF!! *****
-
-       CallBack(name, filename, "", content, content_type, length);
-
-       **** END OF STUFF-HANDLER ****/
-
-
 /* 
  * Utility function to "readline" from memory
  * (returns new pointer)
  */
-char *memreadline(char *start, char *buf, int maxlen) {
+char *memreadline(char *start, char *buf, int maxlen)
+{
        char ch;
        char *ptr;
 
        ptr = start;
-       bzero(buf, maxlen);
+       memset(buf, 0, maxlen);
 
-       while(1) {
+       while (1) {
                ch = *ptr++;
-               if ((ch==10)||(ch==0)) {
-                       if (strlen(buf)>0)
-                               if (buf[strlen(buf)-1]==13)
-                                       buf[strlen(buf)-1] = 0;
+               if ((ch == 10) || (ch == 0)) {
+                       if (strlen(buf) > 0)
+                               if (buf[strlen(buf) - 1] == 13)
+                                       buf[strlen(buf) - 1] = 0;
                        return ptr;
-                       }
-               if (strlen(buf) < (maxlen-1)) {
-                       buf[strlen(buf)+1] = 0;
+               }
+               if (strlen(buf) < (maxlen - 1)) {
+                       buf[strlen(buf) + 1] = 0;
                        buf[strlen(buf)] = ch;
-                       }
                }
        }
+}
+
+/*
+ * Given a message or message-part body and a length, handle any necessary
+ * decoding and pass the request up the stack.
+ */
+void mime_decode(char *partnum,
+                char *part_start, size_t length,
+                char *content_type, char *encoding,
+                char *disposition,
+                char *name, char *filename,
+                void (*CallBack)
+                 (char *cbname,
+                  char *cbfilename,
+                  char *cbpartnum,
+                  char *cbdisp,
+                  void *cbcontent,
+                  char *cbtype,
+                  size_t cblength)
+)
+{
 
+       char *decoded;
+       struct stat statbuf;
+       int sendpipe[2];
+       int recvpipe[2];
+       int childpid;
+       size_t bytes_sent = 0;
+       size_t bytes_recv = 0;
+       size_t blocksize;
+       int write_error = 0;
 
+       lprintf(9, "mime_decode() called\n");
+
+       /* Some encodings aren't really encodings */
+       if (!strcasecmp(encoding, "7bit"))
+               strcpy(encoding, "");
+       if (!strcasecmp(encoding, "8bit"))
+               strcpy(encoding, "");
+       if (!strcasecmp(encoding, "binary"))
+               strcpy(encoding, "");
+
+       /* If this part is not encoded, send as-is */
+       if (strlen(encoding) == 0) {
+               CallBack(name, filename, partnum, disposition, part_start,
+                        content_type, length);
+               return;
+       }
+       if ((strcasecmp(encoding, "base64"))
+           && (strcasecmp(encoding, "quoted-printable"))) {
+               lprintf(5, "ERROR: unknown MIME encoding '%s'\n", encoding);
+               return;
+       }
+       /*
+        * Allocate a buffer for the decoded data.  The output buffer is the
+        * same size as the input buffer; this assumes that the decoded data
+        * will never be larger than the encoded data.  This is a safe
+        * assumption with base64, uuencode, and quoted-printable.  Just to
+        * be safe, we still pad the buffer a bit.
+        */
+       decoded = mallok(length + 1024);
+       if (decoded == NULL) {
+               lprintf(5, "ERROR: cannot allocate memory.\n");
+               return;
+       }
+       if (pipe(sendpipe) != 0)
+               return;
+       if (pipe(recvpipe) != 0)
+               return;
+
+       childpid = fork();
+       if (childpid < 0) {
+               phree(decoded);
+               return;
+       }
+       if (childpid == 0) {
+               close(2);
+               /* send stdio to the pipes */
+               if (dup2(sendpipe[0], 0) < 0)
+                       lprintf(5, "ERROR dup2()\n");
+               if (dup2(recvpipe[1], 1) < 0)
+                       lprintf(5, "ERROR dup2()\n");
+               close(sendpipe[1]);     /* Close the ends we're not using */
+               close(recvpipe[0]);
+               if (!strcasecmp(encoding, "base64"))
+                       execlp("./base64", "base64", "-d", NULL);
+               else if (!strcasecmp(encoding, "quoted-printable"))
+                       execlp("./qpdecode", "qpdecode", NULL);
+               lprintf(5, "ERROR: cannot exec decoder for %s\n", encoding);
+               exit(1);
+       }
+       close(sendpipe[0]);     /* Close the ends we're not using  */
+       close(recvpipe[1]);
+
+       while ((bytes_sent < length) && (write_error == 0)) {
+               /* Empty the input pipe FIRST */
+               while (fstat(recvpipe[0], &statbuf), (statbuf.st_size > 0)) {
+                       blocksize = read(recvpipe[0], &decoded[bytes_recv],
+                                        statbuf.st_size);
+                       if (blocksize < 0)
+                               lprintf(5, "ERROR: cannot read from pipe\n");
+                       else
+                               bytes_recv = bytes_recv + blocksize;
+               }
+               /* Then put some data into the output pipe */
+               blocksize = length - bytes_sent;
+               if (blocksize > 2048)
+                       blocksize = 2048;
+               if (write(sendpipe[1], &part_start[bytes_sent], blocksize) < 0) {
+                       lprintf(5, "ERROR: cannot write to pipe: %s\n",
+                               strerror(errno));
+                       write_error = 1;
+               }
+               bytes_sent = bytes_sent + blocksize;
+       }
+       close(sendpipe[1]);
+       /* Empty the input pipe */
+       while ((blocksize = read(recvpipe[0], &decoded[bytes_recv], 1)),
+              (blocksize > 0)) {
+               bytes_recv = bytes_recv + blocksize;
+       }
+
+       if (bytes_recv > 0)
+               CallBack(name, filename, partnum, disposition, decoded,
+                        content_type, bytes_recv);
+
+       phree(decoded);
+}
 
 /*
  * Break out the components of a multipart message
  * (This function expects to be fed HEADERS + CONTENT)
+ * Note: NULL can be supplied as content_end; in this case, the message is
+ * considered to have ended when the parser encounters a 0x00 byte.
  */
-void mime_parser(char *content,
-               void (*CallBack)
-                       (char *cbname,
-                       char *cbfilename,
-                       char *cbencoding,
-                       void *cbcontent,
-                       char *cbtype,
-                       size_t cblength)
-               ) {
+void the_mime_parser(char *partnum,
+                    char *content_start, char *content_end,
+                    void (*CallBack)
+                     (char *cbname,
+                      char *cbfilename,
+                      char *cbpartnum,
+                      char *cbdisp,
+                      void *cbcontent,
+                      char *cbtype,
+                      size_t cblength)
+)
+{
 
        char *ptr;
        char *part_start, *part_end;
        char buf[256];
        char header[256];
        char boundary[256];
+       char startary[256];
+       char endary[256];
        char content_type[256];
        char encoding[256];
-       int content_length;
+       char disposition[256];
+       char name[256];
+       char filename[256];
+       int is_multipart;
+       int part_seq = 0;
        int i;
+       size_t length;
+       char nested_partnum[256];
 
-       ptr = content;
-       bzero(boundary, sizeof boundary);
-       bzero(content_type, sizeof content_type);
-       bzero(encoding, sizeof encoding);
-       content_length = 0;
+       lprintf(9, "the_mime_parser() called\n");
+       ptr = content_start;
+       memset(boundary, 0, sizeof boundary);
+       memset(content_type, 0, sizeof content_type);
+       memset(encoding, 0, sizeof encoding);
+       memset(name, 0, sizeof name);
+       memset(filename, 0, sizeof filename);
+       memset(disposition, 0, sizeof disposition);
 
        /* Learn interesting things from the headers */
        strcpy(header, "");
        do {
                ptr = memreadline(ptr, buf, sizeof buf);
-               for (i=0; i<strlen(buf); ++i)
-                       if (isspace(buf[i])) buf[i]=' ';
+               if (*ptr == 0)
+                       return; /* premature end of message */
+               if (content_end != NULL)
+                       if (ptr >= content_end)
+                               return;
+
+               for (i = 0; i < strlen(buf); ++i)
+                       if (isspace(buf[i]))
+                               buf[i] = ' ';
                if (!isspace(buf[0])) {
-                       if (!strncasecmp(header, "Content-type: ", 14))
+                       if (!strncasecmp(header, "Content-type: ", 14)) {
                                strcpy(content_type, &header[14]);
-                       if (!strncasecmp(header, "Content-length: ", 16))
-                               content_length = atoi(&header[16]);
+                               extract_key(name, content_type, "name");
+                       }
+                       if (!strncasecmp(header, "Content-Disposition: ", 21)) {
+                               strcpy(disposition, &header[21]);
+                               extract_key(filename, disposition, "filename");
+                       }
                        if (!strncasecmp(header,
-                               "Content-transfer-encoding: ", 27))
-                                       strcpy(encoding, &header[27]);
-                       if (strlen(boundary)==0)
+                                     "Content-transfer-encoding: ", 27))
+                               strcpy(encoding, &header[27]);
+                       if (strlen(boundary) == 0)
                                extract_key(boundary, header, "boundary");
                        strcpy(header, "");
-                       }
-               if ((strlen(header)+strlen(buf)+2)<sizeof(header))
+               }
+               if ((strlen(header) + strlen(buf) + 2) < sizeof(header))
                        strcat(header, buf);
-               } while ((strlen(buf) > 0) && (*ptr != 0));
+       } while ((strlen(buf) > 0) && (*ptr != 0));
 
-       cprintf("Content type is <%s>\n", content_type);
-       cprintf("Encoding is <%s>\n", encoding);
-       cprintf("Content length is %d\n", content_length);
-       cprintf("Boundary is <%s>\n", boundary);
+       for (i = 0; i < strlen(disposition); ++i)
+               if (disposition[i] == ';')
+                       disposition[i] = 0;
+       while (isspace(disposition[0]))
+               strcpy(disposition, &disposition[1]);
+       for (i = 0; i < strlen(content_type); ++i)
+               if (content_type[i] == ';')
+                       content_type[i] = 0;
+       while (isspace(content_type[0]))
+               strcpy(content_type, &content_type[1]);
 
-       if (*ptr == 0) return; /* premature end of message */
+       if (strlen(boundary) > 0) {
+               is_multipart = 1;
+       } else {
+               is_multipart = 0;
+       }
 
        /* If this is a multipart message, then recursively process it */
-       if (strlen(boundary)>0) {
-               }
-       
+       part_start = NULL;
+       if (is_multipart) {
+
+               /* Tell the client about this message's multipartedness */
+               CallBack("", "", partnum, "", NULL, content_type, 0);
+
+               /* Figure out where the boundaries are */
+               sprintf(startary, "--%s", boundary);
+               sprintf(endary, "--%s--", boundary);
+               do {
+                       part_end = ptr;
+                       ptr = memreadline(ptr, buf, sizeof buf);
+                       if (*ptr == 0)
+                               return;         /* premature end of message */
+                       if (content_end != NULL)
+                               if (ptr >= content_end)
+                                       return;
+                       if ((!strcasecmp(buf, startary))
+                           || (!strcasecmp(buf, endary))) {
+                               if (part_start != NULL) {
+                                       sprintf(nested_partnum, "%s.%d",
+                                               partnum, ++part_seq);
+                                       the_mime_parser(nested_partnum,
+                                                   part_start, part_end,
+                                                       CallBack);
+                               }
+                               part_start = ptr;
+                       }
+               } while (strcasecmp(buf, endary));
+       }
 
+       /* If it's not a multipart message, then do something with it */
+       if (!is_multipart) {
+               part_start = ptr;
+               length = 0;
+               while ((*ptr != 0)
+                     && ((content_end == NULL) || (ptr < content_end))) {
+                       ++length;
+                       part_end = ptr++;
+               }
+               mime_decode(partnum,
+                           part_start, length,
+                           content_type, encoding, disposition,
+                           name, filename, CallBack);
        }
+
+
+}
+
+
+
+/*
+ * Entry point for the MIME parser.
+ * (This function expects to be fed HEADERS + CONTENT)
+ * Note: NULL can be supplied as content_end; in this case, the message is
+ * considered to have ended when the parser encounters a 0x00 byte.
+ */
+void mime_parser(char *content_start, char *content_end,
+                void (*CallBack)
+                 (char *cbname,
+                  char *cbfilename,
+                  char *cbpartnum,
+                  char *cbdisp,
+                  void *cbcontent,
+                  char *cbtype,
+                  size_t cblength)
+)
+{
+
+       lprintf(9, "mime_parser() called\n");
+       the_mime_parser("1", content_start, content_end, CallBack);
+}