/*************************************************************************** * _ _ ____ _ * Project ___| | | | _ \| | * / __| | | | |_) | | * | (__| |_| | _ <| |___ * \___|\___/|_| \_\_____| * * * Trivial file transfer protocol server. * * This code includes many modifications by Jim Guyton * * This source file was started based on netkit-tftpd 0.17 * Heavily modified for curl's test suite */ /* * Copyright (C) Daniel Stenberg, , et al. * Copyright (c) 1983, Regents of the University of California. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * SPDX-License-Identifier: BSD-4-Clause-UC */ #include "server_setup.h" #ifdef HAVE_SYS_IOCTL_H #include #endif #include #ifdef HAVE_FCNTL_H #include #endif #ifdef HAVE_NETINET_IN_H #include #endif #ifdef HAVE_ARPA_INET_H #include #endif #ifdef HAVE_NETDB_H #include #endif #ifdef HAVE_SYS_FILIO_H /* FIONREAD on Solaris 7 */ #include #endif #include #ifdef HAVE_PWD_H #include #endif #include #include "curlx.h" /* from the private lib dir */ #include "getpart.h" #include "util.h" #include "server_sockaddr.h" #include "tftp.h" /* include memdebug.h last */ #include "memdebug.h" /***************************************************************************** * STRUCT DECLARATIONS AND DEFINES * *****************************************************************************/ #ifndef PKTSIZE #define PKTSIZE (SEGSIZE + 4) /* SEGSIZE defined in arpa/tftp.h */ #endif struct testcase { char *buffer; /* holds the file data to send to the client */ size_t bufsize; /* size of the data in buffer */ char *rptr; /* read pointer into the buffer */ size_t rcount; /* amount of data left to read of the file */ long testno; /* test case number */ int ofile; /* file descriptor for output file when uploading to us */ int writedelay; /* number of seconds between each packet */ }; struct formats { const char *f_mode; int f_convert; }; struct errmsg { int e_code; const char *e_msg; }; typedef union { struct tftphdr hdr; char storage[PKTSIZE]; } tftphdr_storage_t; /* * bf.counter values in range [-1 .. SEGSIZE] represents size of data in the * bf.buf buffer. Additionally it can also hold flags BF_ALLOC or BF_FREE. */ struct bf { int counter; /* size of data in buffer, or flag */ tftphdr_storage_t buf; /* room for data packet */ }; #define BF_ALLOC -3 /* alloc'd but not yet filled */ #define BF_FREE -2 /* free */ #define opcode_RRQ 1 #define opcode_WRQ 2 #define opcode_DATA 3 #define opcode_ACK 4 #define opcode_ERROR 5 #define TIMEOUT 5 #undef MIN #define MIN(x,y) ((x)<(y)?(x):(y)) #ifndef DEFAULT_LOGFILE #define DEFAULT_LOGFILE "log/tftpd.log" #endif #define REQUEST_DUMP "server.input" #define DEFAULT_PORT 8999 /* UDP */ /***************************************************************************** * GLOBAL VARIABLES * *****************************************************************************/ static struct errmsg errmsgs[] = { { EUNDEF, "Undefined error code" }, { ENOTFOUND, "File not found" }, { EACCESS, "Access violation" }, { ENOSPACE, "Disk full or allocation exceeded" }, { EBADOP, "Illegal TFTP operation" }, { EBADID, "Unknown transfer ID" }, { EEXISTS, "File already exists" }, { ENOUSER, "No such user" }, { -1, 0 } }; static const struct formats formata[] = { { "netascii", 1 }, { "octet", 0 }, { NULL, 0 } }; static struct bf bfs[2]; static int nextone; /* index of next buffer to use */ static int current; /* index of buffer in use */ /* control flags for crlf conversions */ static int newline = 0; /* fillbuf: in middle of newline expansion */ static int prevchar = -1; /* putbuf: previous char (cr check) */ static tftphdr_storage_t buf; static tftphdr_storage_t ackbuf; static srvr_sockaddr_union_t from; static curl_socklen_t fromlen; static curl_socket_t peer = CURL_SOCKET_BAD; static unsigned int timeout; static unsigned int maxtimeout = 5 * TIMEOUT; #ifdef USE_IPV6 static bool use_ipv6 = FALSE; #endif static const char *ipv_inuse = "IPv4"; const char *serverlogfile = DEFAULT_LOGFILE; static const char *logdir = "log"; static char loglockfile[256]; static const char *pidname = ".tftpd.pid"; static const char *portname = NULL; /* none by default */ static int serverlogslocked = 0; static int wrotepidfile = 0; static int wroteportfile = 0; #ifdef HAVE_SIGSETJMP static sigjmp_buf timeoutbuf; #endif #if defined(HAVE_ALARM) && defined(SIGALRM) static const unsigned int rexmtval = TIMEOUT; #endif /***************************************************************************** * FUNCTION PROTOTYPES * *****************************************************************************/ static struct tftphdr *rw_init(int); static struct tftphdr *w_init(void); static struct tftphdr *r_init(void); static void read_ahead(struct testcase *test, int convert); static ssize_t write_behind(struct testcase *test, int convert); static int synchnet(curl_socket_t); static int do_tftp(struct testcase *test, struct tftphdr *tp, ssize_t size); static int validate_access(struct testcase *test, const char *filename, unsigned short mode); static void sendtftp(struct testcase *test, const struct formats *pf); static void recvtftp(struct testcase *test, const struct formats *pf); static void nak(int error); #if defined(HAVE_ALARM) && defined(SIGALRM) static void mysignal(int sig, void (*handler)(int)); static void timer(int signum); static void justtimeout(int signum); #endif /* HAVE_ALARM && SIGALRM */ /***************************************************************************** * FUNCTION IMPLEMENTATIONS * *****************************************************************************/ #if defined(HAVE_ALARM) && defined(SIGALRM) /* * Like signal(), but with well-defined semantics. */ static void mysignal(int sig, void (*handler)(int)) { struct sigaction sa; memset(&sa, 0, sizeof(sa)); sa.sa_handler = handler; sigaction(sig, &sa, NULL); } #ifdef HAVE_SIGSETJMP CURL_NORETURN #endif static void timer(int signum) { (void)signum; logmsg("alarm!"); timeout += rexmtval; if(timeout >= maxtimeout) { if(wrotepidfile) { wrotepidfile = 0; unlink(pidname); } if(wroteportfile) { wroteportfile = 0; unlink(portname); } if(serverlogslocked) { serverlogslocked = 0; clear_advisor_read_lock(loglockfile); } exit(1); } #ifdef HAVE_SIGSETJMP siglongjmp(timeoutbuf, 1); #endif } static void justtimeout(int signum) { (void)signum; } #endif /* HAVE_ALARM && SIGALRM */ /* * init for either read-ahead or write-behind. * zero for write-behind, one for read-head. */ static struct tftphdr *rw_init(int x) { newline = 0; /* init crlf flag */ prevchar = -1; bfs[0].counter = BF_ALLOC; /* pass out the first buffer */ current = 0; bfs[1].counter = BF_FREE; nextone = x; /* ahead or behind? */ return &bfs[0].buf.hdr; } static struct tftphdr *w_init(void) { return rw_init(0); /* write-behind */ } static struct tftphdr *r_init(void) { return rw_init(1); /* read-ahead */ } /* Have emptied current buffer by sending to net and getting ack. Free it and return next buffer filled with data. */ static int readit(struct testcase *test, struct tftphdr **dpp, int convert /* if true, convert to ASCII */) { struct bf *b; bfs[current].counter = BF_FREE; /* free old one */ current = !current; /* "incr" current */ b = &bfs[current]; /* look at new buffer */ if(b->counter == BF_FREE) /* if it's empty */ read_ahead(test, convert); /* fill it */ *dpp = &b->buf.hdr; /* set caller's ptr */ return b->counter; } /* * fill the input buffer, doing ASCII conversions if requested * conversions are lf -> cr, lf and cr -> cr, nul */ static void read_ahead(struct testcase *test, int convert /* if true, convert to ASCII */) { int i; char *p; int c; struct bf *b; struct tftphdr *dp; b = &bfs[nextone]; /* look at "next" buffer */ if(b->counter != BF_FREE) /* nop if not free */ return; nextone = !nextone; /* "incr" next buffer ptr */ dp = &b->buf.hdr; if(convert == 0) { /* The former file reading code did this: b->counter = read(fileno(file), dp->th_data, SEGSIZE); */ size_t copy_n = MIN(SEGSIZE, test->rcount); memcpy(dp->th_data, test->rptr, copy_n); /* decrease amount, advance pointer */ test->rcount -= copy_n; test->rptr += copy_n; b->counter = (int)copy_n; return; } p = dp->th_data; for(i = 0 ; i < SEGSIZE; i++) { if(newline) { if(prevchar == '\n') c = '\n'; /* lf to cr,lf */ else c = '\0'; /* cr to cr,nul */ newline = 0; } else { if(test->rcount) { c = test->rptr[0]; test->rptr++; test->rcount--; } else break; if(c == '\n' || c == '\r') { prevchar = c; c = '\r'; newline = 1; } } *p++ = (char)c; } b->counter = (int)(p - dp->th_data); } /* Update count associated with the buffer, get new buffer from the queue. Calls write_behind only if next buffer not available. */ static int writeit(struct testcase *test, struct tftphdr * volatile *dpp, int ct, int convert) { bfs[current].counter = ct; /* set size of data to write */ current = !current; /* switch to other buffer */ if(bfs[current].counter != BF_FREE) /* if not free */ write_behind(test, convert); /* flush it */ bfs[current].counter = BF_ALLOC; /* mark as alloc'd */ *dpp = &bfs[current].buf.hdr; return ct; /* this is a lie of course */ } /* * Output a buffer to a file, converting from netascii if requested. * CR, NUL -> CR and CR, LF => LF. * Note spec is undefined if we get CR as last byte of file or a * CR followed by anything else. In this case we leave it alone. */ static ssize_t write_behind(struct testcase *test, int convert) { char *writebuf; int count; int ct; char *p; int c; /* current character */ struct bf *b; struct tftphdr *dp; b = &bfs[nextone]; if(b->counter < -1) /* anything to flush? */ return 0; /* just nop if nothing to do */ if(!test->ofile) { char outfile[256]; msnprintf(outfile, sizeof(outfile), "%s/upload.%ld", logdir, test->testno); #ifdef _WIN32 test->ofile = open(outfile, O_CREAT|O_RDWR|O_BINARY, 0777); #else test->ofile = open(outfile, O_CREAT|O_RDWR, 0777); #endif if(test->ofile == -1) { logmsg("Couldn't create and/or open file %s for upload!", outfile); return -1; /* failure! */ } } count = b->counter; /* remember byte count */ b->counter = BF_FREE; /* reset flag */ dp = &b->buf.hdr; nextone = !nextone; /* incr for next time */ writebuf = dp->th_data; if(count <= 0) return -1; /* nak logic? */ if(convert == 0) return write(test->ofile, writebuf, count); p = writebuf; ct = count; while(ct--) { /* loop over the buffer */ c = *p++; /* pick up a character */ if(prevchar == '\r') { /* if prev char was cr */ if(c == '\n') /* if have cr,lf then just */ lseek(test->ofile, -1, SEEK_CUR); /* smash lf on top of the cr */ else if(c == '\0') /* if have cr,nul then */ goto skipit; /* just skip over the putc */ /* else just fall through and allow it */ } /* formerly putc(c, file); */ if(1 != write(test->ofile, &c, 1)) break; skipit: prevchar = c; } return count; } /* When an error has occurred, it is possible that the two sides are out of * synch. Ie: that what I think is the other side's response to packet N is * really their response to packet N-1. * * So, to try to prevent that, we flush all the input queued up for us on the * network connection on our host. * * We return the number of packets we flushed (mostly for reporting when trace * is active). */ static int synchnet(curl_socket_t f /* socket to flush */) { #if defined(HAVE_IOCTLSOCKET) unsigned long i; #else int i; #endif int j = 0; char rbuf[PKTSIZE]; srvr_sockaddr_union_t fromaddr; curl_socklen_t fromaddrlen; for(;;) { #if defined(HAVE_IOCTLSOCKET) (void) ioctlsocket(f, FIONREAD, &i); #else (void) ioctl(f, FIONREAD, &i); #endif if(i) { j++; #ifdef USE_IPV6 if(!use_ipv6) #endif fromaddrlen = sizeof(fromaddr.sa4); #ifdef USE_IPV6 else fromaddrlen = sizeof(fromaddr.sa6); #endif (void) recvfrom(f, rbuf, sizeof(rbuf), 0, &fromaddr.sa, &fromaddrlen); } else break; } return j; } int main(int argc, char **argv) { srvr_sockaddr_union_t me; struct tftphdr *tp; ssize_t n = 0; int arg = 1; unsigned short port = DEFAULT_PORT; curl_socket_t sock = CURL_SOCKET_BAD; int flag; int rc; int error; struct testcase test; int result = 0; memset(&test, 0, sizeof(test)); while(argc > arg) { if(!strcmp("--version", argv[arg])) { printf("tftpd IPv4%s\n", #ifdef USE_IPV6 "/IPv6" #else "" #endif ); return 0; } else if(!strcmp("--pidfile", argv[arg])) { arg++; if(argc > arg) pidname = argv[arg++]; } else if(!strcmp("--portfile", argv[arg])) { arg++; if(argc > arg) portname = argv[arg++]; } else if(!strcmp("--logfile", argv[arg])) { arg++; if(argc > arg) serverlogfile = argv[arg++]; } else if(!strcmp("--logdir", argv[arg])) { arg++; if(argc > arg) logdir = argv[arg++]; } else if(!strcmp("--ipv4", argv[arg])) { #ifdef USE_IPV6 ipv_inuse = "IPv4"; use_ipv6 = FALSE; #endif arg++; } else if(!strcmp("--ipv6", argv[arg])) { #ifdef USE_IPV6 ipv_inuse = "IPv6"; use_ipv6 = TRUE; #endif arg++; } else if(!strcmp("--port", argv[arg])) { arg++; if(argc > arg) { char *endptr; unsigned long ulnum = strtoul(argv[arg], &endptr, 10); port = curlx_ultous(ulnum); arg++; } } else if(!strcmp("--srcdir", argv[arg])) { arg++; if(argc > arg) { path = argv[arg]; arg++; } } else { puts("Usage: tftpd [option]\n" " --version\n" " --logfile [file]\n" " --logdir [directory]\n" " --pidfile [file]\n" " --portfile [file]\n" " --ipv4\n" " --ipv6\n" " --port [port]\n" " --srcdir [path]"); return 0; } } msnprintf(loglockfile, sizeof(loglockfile), "%s/%s/tftp-%s.lock", logdir, SERVERLOGS_LOCKDIR, ipv_inuse); #ifdef _WIN32 win32_init(); atexit(win32_cleanup); #endif install_signal_handlers(true); #ifdef USE_IPV6 if(!use_ipv6) #endif sock = socket(AF_INET, SOCK_DGRAM, 0); #ifdef USE_IPV6 else sock = socket(AF_INET6, SOCK_DGRAM, 0); #endif if(CURL_SOCKET_BAD == sock) { error = SOCKERRNO; logmsg("Error creating socket: (%d) %s", error, sstrerror(error)); result = 1; goto tftpd_cleanup; } flag = 1; if(0 != setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, (void *)&flag, sizeof(flag))) { error = SOCKERRNO; logmsg("setsockopt(SO_REUSEADDR) failed with error: (%d) %s", error, sstrerror(error)); result = 1; goto tftpd_cleanup; } #ifdef USE_IPV6 if(!use_ipv6) { #endif memset(&me.sa4, 0, sizeof(me.sa4)); me.sa4.sin_family = AF_INET; me.sa4.sin_addr.s_addr = INADDR_ANY; me.sa4.sin_port = htons(port); rc = bind(sock, &me.sa, sizeof(me.sa4)); #ifdef USE_IPV6 } else { memset(&me.sa6, 0, sizeof(me.sa6)); me.sa6.sin6_family = AF_INET6; me.sa6.sin6_addr = in6addr_any; me.sa6.sin6_port = htons(port); rc = bind(sock, &me.sa, sizeof(me.sa6)); } #endif /* USE_IPV6 */ if(0 != rc) { error = SOCKERRNO; logmsg("Error binding socket on port %hu: (%d) %s", port, error, sstrerror(error)); result = 1; goto tftpd_cleanup; } if(!port) { /* The system was supposed to choose a port number, figure out which port we actually got and update the listener port value with it. */ curl_socklen_t la_size; srvr_sockaddr_union_t localaddr; #ifdef USE_IPV6 if(!use_ipv6) #endif la_size = sizeof(localaddr.sa4); #ifdef USE_IPV6 else la_size = sizeof(localaddr.sa6); #endif memset(&localaddr.sa, 0, (size_t)la_size); if(getsockname(sock, &localaddr.sa, &la_size) < 0) { error = SOCKERRNO; logmsg("getsockname() failed with error: (%d) %s", error, sstrerror(error)); sclose(sock); goto tftpd_cleanup; } switch(localaddr.sa.sa_family) { case AF_INET: port = ntohs(localaddr.sa4.sin_port); break; #ifdef USE_IPV6 case AF_INET6: port = ntohs(localaddr.sa6.sin6_port); break; #endif default: break; } if(!port) { /* Real failure, listener port shall not be zero beyond this point. */ logmsg("Apparently getsockname() succeeded, with listener port zero."); logmsg("A valid reason for this failure is a binary built without"); logmsg("proper network library linkage. This might not be the only"); logmsg("reason, but double check it before anything else."); result = 2; goto tftpd_cleanup; } } wrotepidfile = write_pidfile(pidname); if(!wrotepidfile) { result = 1; goto tftpd_cleanup; } if(portname) { wroteportfile = write_portfile(portname, port); if(!wroteportfile) { result = 1; goto tftpd_cleanup; } } logmsg("Running %s version on port UDP/%d", ipv_inuse, (int)port); for(;;) { fromlen = sizeof(from); #ifdef USE_IPV6 if(!use_ipv6) #endif fromlen = sizeof(from.sa4); #ifdef USE_IPV6 else fromlen = sizeof(from.sa6); #endif n = (ssize_t)recvfrom(sock, &buf.storage[0], sizeof(buf.storage), 0, &from.sa, &fromlen); if(got_exit_signal) break; if(n < 0) { logmsg("recvfrom"); result = 3; break; } set_advisor_read_lock(loglockfile); serverlogslocked = 1; #ifdef USE_IPV6 if(!use_ipv6) { #endif from.sa4.sin_family = AF_INET; peer = socket(AF_INET, SOCK_DGRAM, 0); if(CURL_SOCKET_BAD == peer) { logmsg("socket"); result = 2; break; } if(connect(peer, &from.sa, sizeof(from.sa4)) < 0) { logmsg("connect: fail"); result = 1; break; } #ifdef USE_IPV6 } else { from.sa6.sin6_family = AF_INET6; peer = socket(AF_INET6, SOCK_DGRAM, 0); if(CURL_SOCKET_BAD == peer) { logmsg("socket"); result = 2; break; } if(connect(peer, &from.sa, sizeof(from.sa6)) < 0) { logmsg("connect: fail"); result = 1; break; } } #endif maxtimeout = 5*TIMEOUT; tp = &buf.hdr; tp->th_opcode = ntohs(tp->th_opcode); if(tp->th_opcode == opcode_RRQ || tp->th_opcode == opcode_WRQ) { memset(&test, 0, sizeof(test)); if(do_tftp(&test, tp, n) < 0) break; free(test.buffer); } sclose(peer); peer = CURL_SOCKET_BAD; if(got_exit_signal) break; if(serverlogslocked) { serverlogslocked = 0; clear_advisor_read_lock(loglockfile); } logmsg("end of one transfer"); } tftpd_cleanup: if(test.ofile > 0) close(test.ofile); if((peer != sock) && (peer != CURL_SOCKET_BAD)) sclose(peer); if(sock != CURL_SOCKET_BAD) sclose(sock); if(got_exit_signal) logmsg("signalled to die"); if(wrotepidfile) unlink(pidname); if(wroteportfile) unlink(portname); if(serverlogslocked) { serverlogslocked = 0; clear_advisor_read_lock(loglockfile); } restore_signal_handlers(true); if(got_exit_signal) { logmsg("========> %s tftpd (port: %d pid: %ld) exits with signal (%d)", ipv_inuse, (int)port, (long)Curl_getpid(), exit_signal); /* * To properly set the return status of the process we * must raise the same signal SIGINT or SIGTERM that we * caught and let the old handler take care of it. */ raise(exit_signal); } logmsg("========> tftpd quits"); return result; } /* * Handle initial connection protocol. */ static int do_tftp(struct testcase *test, struct tftphdr *tp, ssize_t size) { char *cp; int first = 1, ecode; const struct formats *pf; char *filename, *mode = NULL; #ifdef USE_WINSOCK DWORD recvtimeout, recvtimeoutbak; #endif const char *option = "mode"; /* mode is implicit */ int toggle = 1; FILE *server; char dumpfile[256]; msnprintf(dumpfile, sizeof(dumpfile), "%s/%s", logdir, REQUEST_DUMP); /* Open request dump file. */ server = fopen(dumpfile, "ab"); if(!server) { int error = errno; logmsg("fopen() failed with error: %d %s", error, strerror(error)); logmsg("Error opening file: %s", dumpfile); return -1; } /* store input protocol */ fprintf(server, "opcode = %x\n", tp->th_opcode); cp = (char *)&tp->th_stuff; filename = cp; do { bool endofit = true; while(cp < &buf.storage[size]) { if(*cp == '\0') { endofit = false; break; } cp++; } if(endofit) /* no more options */ break; /* before increasing pointer, make sure it is still within the legal space */ if((cp + 1) < &buf.storage[size]) { ++cp; if(first) { /* store the mode since we need it later */ mode = cp; first = 0; } if(toggle) /* name/value pair: */ fprintf(server, "%s = %s\n", option, cp); else { /* store the name pointer */ option = cp; } toggle ^= 1; } else /* No more options */ break; } while(1); if(*cp) { nak(EBADOP); fclose(server); return 3; } /* store input protocol */ fprintf(server, "filename = %s\n", filename); for(cp = mode; cp && *cp; cp++) if(ISUPPER(*cp)) *cp = (char)tolower((int)*cp); /* store input protocol */ fclose(server); for(pf = formata; pf->f_mode; pf++) if(strcmp(pf->f_mode, mode) == 0) break; if(!pf->f_mode) { nak(EBADOP); return 2; } ecode = validate_access(test, filename, tp->th_opcode); if(ecode) { nak(ecode); return 1; } #ifdef USE_WINSOCK recvtimeout = sizeof(recvtimeoutbak); getsockopt(peer, SOL_SOCKET, SO_RCVTIMEO, (char *)&recvtimeoutbak, (int *)&recvtimeout); recvtimeout = TIMEOUT*1000; setsockopt(peer, SOL_SOCKET, SO_RCVTIMEO, (const char *)&recvtimeout, sizeof(recvtimeout)); #endif if(tp->th_opcode == opcode_WRQ) recvtftp(test, pf); else sendtftp(test, pf); #ifdef USE_WINSOCK recvtimeout = recvtimeoutbak; setsockopt(peer, SOL_SOCKET, SO_RCVTIMEO, (const char *)&recvtimeout, sizeof(recvtimeout)); #endif return 0; } /* Based on the testno, parse the correct server commands. */ static int parse_servercmd(struct testcase *req) { FILE *stream; int error; stream = test2fopen(req->testno, logdir); if(!stream) { error = errno; logmsg("fopen() failed with error: %d %s", error, strerror(error)); logmsg(" Couldn't open test file %ld", req->testno); return 1; /* done */ } else { char *orgcmd = NULL; char *cmd = NULL; size_t cmdsize = 0; int num = 0; /* get the custom server control "commands" */ error = getpart(&orgcmd, &cmdsize, "reply", "servercmd", stream); fclose(stream); if(error) { logmsg("getpart() failed with error: %d", error); return 1; /* done */ } cmd = orgcmd; while(cmd && cmdsize) { char *check; if(1 == sscanf(cmd, "writedelay: %d", &num)) { logmsg("instructed to delay %d secs between packets", num); req->writedelay = num; } else { logmsg("Unknown instruction found: %s", cmd); } /* try to deal with CRLF or just LF */ check = strchr(cmd, '\r'); if(!check) check = strchr(cmd, '\n'); if(check) { /* get to the letter following the newline */ while((*check == '\r') || (*check == '\n')) check++; if(!*check) /* if we reached a zero, get out */ break; cmd = check; } else break; } free(orgcmd); } return 0; /* OK! */ } /* * Validate file access. */ static int validate_access(struct testcase *test, const char *filename, unsigned short mode) { char *ptr; logmsg("trying to get file: %s mode %x", filename, mode); if(!strncmp("verifiedserver", filename, 14)) { char weare[128]; size_t count = msnprintf(weare, sizeof(weare), "WE ROOLZ: %" CURL_FORMAT_CURL_OFF_T "\r\n", our_getpid()); logmsg("Are-we-friendly question received"); test->buffer = strdup(weare); test->rptr = test->buffer; /* set read pointer */ test->bufsize = count; /* set total count */ test->rcount = count; /* set data left to read */ return 0; /* fine */ } /* find the last slash */ ptr = strrchr(filename, '/'); if(ptr) { char partbuf[80]="data"; long partno; long testno; FILE *stream; ptr++; /* skip the slash */ /* skip all non-numericals following the slash */ while(*ptr && !ISDIGIT(*ptr)) ptr++; /* get the number */ testno = strtol(ptr, &ptr, 10); if(testno > 10000) { partno = testno % 10000; testno /= 10000; } else partno = 0; logmsg("requested test number %ld part %ld", testno, partno); test->testno = testno; (void)parse_servercmd(test); stream = test2fopen(testno, logdir); if(0 != partno) msnprintf(partbuf, sizeof(partbuf), "data%ld", partno); if(!stream) { int error = errno; logmsg("fopen() failed with error: %d %s", error, strerror(error)); logmsg("Couldn't open test file for test: %ld", testno); return EACCESS; } else { size_t count; int error = getpart(&test->buffer, &count, "reply", partbuf, stream); fclose(stream); if(error) { logmsg("getpart() failed with error: %d", error); return EACCESS; } if(test->buffer) { test->rptr = test->buffer; /* set read pointer */ test->bufsize = count; /* set total count */ test->rcount = count; /* set data left to read */ } else return EACCESS; } } else { logmsg("no slash found in path"); return EACCESS; /* failure */ } logmsg("file opened and all is good"); return 0; } /* * Send the requested file. */ static void sendtftp(struct testcase *test, const struct formats *pf) { int size; ssize_t n; /* These are volatile to live through a siglongjmp */ volatile unsigned short sendblock; /* block count */ struct tftphdr * volatile sdp = r_init(); /* data buffer */ struct tftphdr * const sap = &ackbuf.hdr; /* ack buffer */ sendblock = 1; #if defined(HAVE_ALARM) && defined(SIGALRM) mysignal(SIGALRM, timer); #endif do { size = readit(test, (struct tftphdr **)&sdp, pf->f_convert); if(size < 0) { nak(errno + 100); return; } sdp->th_opcode = htons(opcode_DATA); sdp->th_block = htons(sendblock); timeout = 0; #ifdef HAVE_SIGSETJMP (void) sigsetjmp(timeoutbuf, 1); #endif if(test->writedelay) { logmsg("Pausing %d seconds before %d bytes", test->writedelay, size); wait_ms(1000*test->writedelay); } send_data: logmsg("write"); if(swrite(peer, sdp, size + 4) != size + 4) { logmsg("write: fail"); return; } read_ahead(test, pf->f_convert); for(;;) { #ifdef HAVE_ALARM alarm(rexmtval); /* read the ack */ #endif logmsg("read"); n = sread(peer, &ackbuf.storage[0], sizeof(ackbuf.storage)); logmsg("read: %zd", n); #ifdef HAVE_ALARM alarm(0); #endif if(got_exit_signal) return; if(n < 0) { logmsg("read: fail"); return; } sap->th_opcode = ntohs(sap->th_opcode); sap->th_block = ntohs(sap->th_block); if(sap->th_opcode == opcode_ERROR) { logmsg("got ERROR"); return; } if(sap->th_opcode == opcode_ACK) { if(sap->th_block == sendblock) { break; } /* Re-synchronize with the other side */ (void) synchnet(peer); if(sap->th_block == (sendblock-1)) { goto send_data; } } } sendblock++; } while(size == SEGSIZE); } /* * Receive a file. */ static void recvtftp(struct testcase *test, const struct formats *pf) { ssize_t n, size; /* These are volatile to live through a siglongjmp */ volatile unsigned short recvblock; /* block count */ struct tftphdr * volatile rdp; /* data buffer */ struct tftphdr *rap; /* ack buffer */ recvblock = 0; rdp = w_init(); #if defined(HAVE_ALARM) && defined(SIGALRM) mysignal(SIGALRM, timer); #endif rap = &ackbuf.hdr; do { timeout = 0; rap->th_opcode = htons(opcode_ACK); rap->th_block = htons(recvblock); recvblock++; #ifdef HAVE_SIGSETJMP (void) sigsetjmp(timeoutbuf, 1); #endif send_ack: logmsg("write"); if(swrite(peer, &ackbuf.storage[0], 4) != 4) { logmsg("write: fail"); goto abort; } write_behind(test, pf->f_convert); for(;;) { #ifdef HAVE_ALARM alarm(rexmtval); #endif logmsg("read"); n = sread(peer, rdp, PKTSIZE); logmsg("read: %zd", n); #ifdef HAVE_ALARM alarm(0); #endif if(got_exit_signal) goto abort; if(n < 0) { /* really? */ logmsg("read: fail"); goto abort; } rdp->th_opcode = ntohs(rdp->th_opcode); rdp->th_block = ntohs(rdp->th_block); if(rdp->th_opcode == opcode_ERROR) goto abort; if(rdp->th_opcode == opcode_DATA) { if(rdp->th_block == recvblock) { break; /* normal */ } /* Re-synchronize with the other side */ (void) synchnet(peer); if(rdp->th_block == (recvblock-1)) goto send_ack; /* rexmit */ } } size = writeit(test, &rdp, (int)(n - 4), pf->f_convert); if(size != (n-4)) { /* ahem */ if(size < 0) nak(errno + 100); else nak(ENOSPACE); goto abort; } } while(size == SEGSIZE); write_behind(test, pf->f_convert); /* close the output file as early as possible after upload completion */ if(test->ofile > 0) { close(test->ofile); test->ofile = 0; } rap->th_opcode = htons(opcode_ACK); /* send the "final" ack */ rap->th_block = htons(recvblock); (void) swrite(peer, &ackbuf.storage[0], 4); #if defined(HAVE_ALARM) && defined(SIGALRM) mysignal(SIGALRM, justtimeout); /* just abort read on timeout */ alarm(rexmtval); #endif /* normally times out and quits */ n = sread(peer, &buf.storage[0], sizeof(buf.storage)); #ifdef HAVE_ALARM alarm(0); #endif if(got_exit_signal) goto abort; if(n >= 4 && /* if read some data */ rdp->th_opcode == opcode_DATA && /* and got a data block */ recvblock == rdp->th_block) { /* then my last ack was lost */ (void) swrite(peer, &ackbuf.storage[0], 4); /* resend final ack */ } abort: /* make sure the output file is closed in case of abort */ if(test->ofile > 0) { close(test->ofile); test->ofile = 0; } return; } /* * Send a nak packet (error message). Error code passed in is one of the * standard TFTP codes, or a Unix errno offset by 100. */ static void nak(int error) { struct tftphdr *tp; int length; struct errmsg *pe; tp = &buf.hdr; tp->th_opcode = htons(opcode_ERROR); tp->th_code = htons((unsigned short)error); for(pe = errmsgs; pe->e_code >= 0; pe++) if(pe->e_code == error) break; if(pe->e_code < 0) { pe->e_msg = strerror(error - 100); tp->th_code = EUNDEF; /* set 'undef' errorcode */ } length = (int)strlen(pe->e_msg); /* we use memcpy() instead of strcpy() in order to avoid buffer overflow * report from glibc with FORTIFY_SOURCE */ memcpy(tp->th_msg, pe->e_msg, length + 1); length += 5; if(swrite(peer, &buf.storage[0], length) != length) logmsg("nak: fail\n"); }