0f43811dc1
Only accept 'net' and 'pxe' devices as underlying transport in tftp.c on x86. Prior to this change tftp code would attempt to send packets over any boot device, including zfs one with predictably sad results. Approved by: re (gjb) MFC After: 1 month
750 lines
18 KiB
C
750 lines
18 KiB
C
/* $NetBSD: tftp.c,v 1.4 1997/09/17 16:57:07 drochner Exp $ */
|
|
|
|
/*
|
|
* Copyright (c) 1996
|
|
* Matthias Drochner. 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 for the NetBSD Project
|
|
* by Matthias Drochner.
|
|
* 4. The name of the author may not be used to endorse or promote products
|
|
* derived from this software without specific prior written permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 AUTHOR 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.
|
|
*/
|
|
|
|
#include <sys/cdefs.h>
|
|
__FBSDID("$FreeBSD$");
|
|
|
|
/*
|
|
* Simple TFTP implementation for libsa.
|
|
* Assumes:
|
|
* - socket descriptor (int) at open_file->f_devdata
|
|
* - server host IP in global servip
|
|
* Restrictions:
|
|
* - read only
|
|
* - lseek only with SEEK_SET or SEEK_CUR
|
|
* - no big time differences between transfers (<tftp timeout)
|
|
*/
|
|
|
|
#include <sys/types.h>
|
|
#include <sys/stat.h>
|
|
#include <netinet/in.h>
|
|
#include <netinet/udp.h>
|
|
#include <netinet/in_systm.h>
|
|
#include <arpa/tftp.h>
|
|
|
|
#include <string.h>
|
|
|
|
#include "stand.h"
|
|
#include "net.h"
|
|
#include "netif.h"
|
|
|
|
#include "tftp.h"
|
|
|
|
struct tftp_handle;
|
|
|
|
static int tftp_open(const char *path, struct open_file *f);
|
|
static int tftp_close(struct open_file *f);
|
|
static int tftp_parse_oack(struct tftp_handle *h, char *buf, size_t len);
|
|
static int tftp_read(struct open_file *f, void *buf, size_t size, size_t *resid);
|
|
static int tftp_write(struct open_file *f, void *buf, size_t size, size_t *resid);
|
|
static off_t tftp_seek(struct open_file *f, off_t offset, int where);
|
|
static int tftp_set_blksize(struct tftp_handle *h, const char *str);
|
|
static int tftp_stat(struct open_file *f, struct stat *sb);
|
|
static ssize_t sendrecv_tftp(struct tftp_handle *h,
|
|
ssize_t (*sproc)(struct iodesc *, void *, size_t),
|
|
void *sbuf, size_t ssize,
|
|
ssize_t (*rproc)(struct tftp_handle *h, void *, ssize_t, time_t, unsigned short *),
|
|
void *rbuf, size_t rsize, unsigned short *rtype);
|
|
|
|
struct fs_ops tftp_fsops = {
|
|
"tftp",
|
|
tftp_open,
|
|
tftp_close,
|
|
tftp_read,
|
|
tftp_write,
|
|
tftp_seek,
|
|
tftp_stat,
|
|
null_readdir
|
|
};
|
|
|
|
extern struct in_addr servip;
|
|
|
|
static int tftpport = 2000;
|
|
static int is_open = 0;
|
|
|
|
/*
|
|
* The legacy TFTP_BLKSIZE value was SEGSIZE(512).
|
|
* TFTP_REQUESTED_BLKSIZE of 1428 is (Ethernet MTU, less the TFTP, UDP and
|
|
* IP header lengths).
|
|
*/
|
|
#define TFTP_REQUESTED_BLKSIZE 1428
|
|
|
|
/*
|
|
* Choose a blksize big enough so we can test with Ethernet
|
|
* Jumbo frames in the future.
|
|
*/
|
|
#define TFTP_MAX_BLKSIZE 9008
|
|
|
|
struct tftp_handle {
|
|
struct iodesc *iodesc;
|
|
int currblock; /* contents of lastdata */
|
|
int islastblock; /* flag */
|
|
int validsize;
|
|
int off;
|
|
char *path; /* saved for re-requests */
|
|
unsigned int tftp_blksize;
|
|
unsigned long tftp_tsize;
|
|
struct {
|
|
u_char header[HEADER_SIZE];
|
|
struct tftphdr t;
|
|
u_char space[TFTP_MAX_BLKSIZE];
|
|
} __packed __aligned(4) lastdata;
|
|
};
|
|
|
|
#define TFTP_MAX_ERRCODE EOPTNEG
|
|
static const int tftperrors[TFTP_MAX_ERRCODE + 1] = {
|
|
0, /* ??? */
|
|
ENOENT,
|
|
EPERM,
|
|
ENOSPC,
|
|
EINVAL, /* ??? */
|
|
EINVAL, /* ??? */
|
|
EEXIST,
|
|
EINVAL, /* ??? */
|
|
EINVAL, /* Option negotiation failed. */
|
|
};
|
|
|
|
static int tftp_getnextblock(struct tftp_handle *h);
|
|
|
|
/* send error message back. */
|
|
static void
|
|
tftp_senderr(struct tftp_handle *h, u_short errcode, const char *msg)
|
|
{
|
|
struct {
|
|
u_char header[HEADER_SIZE];
|
|
struct tftphdr t;
|
|
u_char space[63]; /* +1 from t */
|
|
} __packed __aligned(4) wbuf;
|
|
char *wtail;
|
|
int len;
|
|
|
|
len = strlen(msg);
|
|
if (len > sizeof(wbuf.space))
|
|
len = sizeof(wbuf.space);
|
|
|
|
wbuf.t.th_opcode = htons((u_short) ERROR);
|
|
wbuf.t.th_code = htons(errcode);
|
|
|
|
wtail = wbuf.t.th_msg;
|
|
bcopy(msg, wtail, len);
|
|
wtail[len] = '\0';
|
|
wtail += len + 1;
|
|
|
|
sendudp(h->iodesc, &wbuf.t, wtail - (char *) &wbuf.t);
|
|
}
|
|
|
|
static void
|
|
tftp_sendack(struct tftp_handle *h)
|
|
{
|
|
struct {
|
|
u_char header[HEADER_SIZE];
|
|
struct tftphdr t;
|
|
} __packed __aligned(4) wbuf;
|
|
char *wtail;
|
|
|
|
wbuf.t.th_opcode = htons((u_short) ACK);
|
|
wtail = (char *) &wbuf.t.th_block;
|
|
wbuf.t.th_block = htons((u_short) h->currblock);
|
|
wtail += 2;
|
|
|
|
sendudp(h->iodesc, &wbuf.t, wtail - (char *) &wbuf.t);
|
|
}
|
|
|
|
static ssize_t
|
|
recvtftp(struct tftp_handle *h, void *pkt, ssize_t len, time_t tleft,
|
|
unsigned short *rtype)
|
|
{
|
|
struct iodesc *d = h->iodesc;
|
|
struct tftphdr *t;
|
|
|
|
errno = 0;
|
|
|
|
len = readudp(d, pkt, len, tleft);
|
|
|
|
if (len < 4)
|
|
return (-1);
|
|
|
|
t = (struct tftphdr *) pkt;
|
|
*rtype = ntohs(t->th_opcode);
|
|
switch (ntohs(t->th_opcode)) {
|
|
case DATA: {
|
|
int got;
|
|
|
|
if (htons(t->th_block) != d->xid) {
|
|
/*
|
|
* Expected block?
|
|
*/
|
|
return (-1);
|
|
}
|
|
if (d->xid == 1) {
|
|
/*
|
|
* First data packet from new port.
|
|
*/
|
|
struct udphdr *uh;
|
|
uh = (struct udphdr *) pkt - 1;
|
|
d->destport = uh->uh_sport;
|
|
} /* else check uh_sport has not changed??? */
|
|
got = len - (t->th_data - (char *) t);
|
|
return got;
|
|
}
|
|
case ERROR:
|
|
if ((unsigned) ntohs(t->th_code) > TFTP_MAX_ERRCODE) {
|
|
printf("illegal tftp error %d\n", ntohs(t->th_code));
|
|
errno = EIO;
|
|
} else {
|
|
#ifdef TFTP_DEBUG
|
|
printf("tftp-error %d\n", ntohs(t->th_code));
|
|
#endif
|
|
errno = tftperrors[ntohs(t->th_code)];
|
|
}
|
|
return (-1);
|
|
case OACK: {
|
|
struct udphdr *uh;
|
|
int tftp_oack_len;
|
|
|
|
/*
|
|
* Unexpected OACK. TFTP transfer already in progress.
|
|
* Drop the pkt.
|
|
*/
|
|
if (d->xid != 1) {
|
|
return (-1);
|
|
}
|
|
|
|
/*
|
|
* Remember which port this OACK came from, because we need
|
|
* to send the ACK or errors back to it.
|
|
*/
|
|
uh = (struct udphdr *) pkt - 1;
|
|
d->destport = uh->uh_sport;
|
|
|
|
/* Parse options ACK-ed by the server. */
|
|
tftp_oack_len = len - sizeof(t->th_opcode);
|
|
if (tftp_parse_oack(h, t->th_u.tu_stuff, tftp_oack_len) != 0) {
|
|
tftp_senderr(h, EOPTNEG, "Malformed OACK");
|
|
errno = EIO;
|
|
return (-1);
|
|
}
|
|
return (0);
|
|
}
|
|
default:
|
|
#ifdef TFTP_DEBUG
|
|
printf("tftp type %d not handled\n", ntohs(t->th_opcode));
|
|
#endif
|
|
return (-1);
|
|
}
|
|
}
|
|
|
|
/* send request, expect first block (or error) */
|
|
static int
|
|
tftp_makereq(struct tftp_handle *h)
|
|
{
|
|
struct {
|
|
u_char header[HEADER_SIZE];
|
|
struct tftphdr t;
|
|
u_char space[FNAME_SIZE + 6];
|
|
} __packed __aligned(4) wbuf;
|
|
char *wtail;
|
|
int l;
|
|
ssize_t res;
|
|
struct tftphdr *t;
|
|
char *tftp_blksize = NULL;
|
|
int blksize_l;
|
|
unsigned short rtype = 0;
|
|
|
|
/*
|
|
* Allow overriding default TFTP block size by setting
|
|
* a tftp.blksize environment variable.
|
|
*/
|
|
if ((tftp_blksize = getenv("tftp.blksize")) != NULL) {
|
|
tftp_set_blksize(h, tftp_blksize);
|
|
}
|
|
|
|
wbuf.t.th_opcode = htons((u_short) RRQ);
|
|
wtail = wbuf.t.th_stuff;
|
|
l = strlen(h->path);
|
|
#ifdef TFTP_PREPEND_PATH
|
|
if (l > FNAME_SIZE - (sizeof(TFTP_PREPEND_PATH) - 1))
|
|
return (ENAMETOOLONG);
|
|
bcopy(TFTP_PREPEND_PATH, wtail, sizeof(TFTP_PREPEND_PATH) - 1);
|
|
wtail += sizeof(TFTP_PREPEND_PATH) - 1;
|
|
#else
|
|
if (l > FNAME_SIZE)
|
|
return (ENAMETOOLONG);
|
|
#endif
|
|
bcopy(h->path, wtail, l + 1);
|
|
wtail += l + 1;
|
|
bcopy("octet", wtail, 6);
|
|
wtail += 6;
|
|
bcopy("blksize", wtail, 8);
|
|
wtail += 8;
|
|
blksize_l = sprintf(wtail, "%d", h->tftp_blksize);
|
|
wtail += blksize_l + 1;
|
|
bcopy("tsize", wtail, 6);
|
|
wtail += 6;
|
|
bcopy("0", wtail, 2);
|
|
wtail += 2;
|
|
|
|
t = &h->lastdata.t;
|
|
|
|
/* h->iodesc->myport = htons(--tftpport); */
|
|
h->iodesc->myport = htons(tftpport + (getsecs() & 0x3ff));
|
|
h->iodesc->destport = htons(IPPORT_TFTP);
|
|
h->iodesc->xid = 1; /* expected block */
|
|
|
|
h->currblock = 0;
|
|
h->islastblock = 0;
|
|
h->validsize = 0;
|
|
|
|
res = sendrecv_tftp(h, &sendudp, &wbuf.t, wtail - (char *) &wbuf.t,
|
|
&recvtftp, t, sizeof(*t) + h->tftp_blksize, &rtype);
|
|
|
|
if (rtype == OACK)
|
|
return (tftp_getnextblock(h));
|
|
|
|
/* Server ignored our blksize request, revert to TFTP default. */
|
|
h->tftp_blksize = SEGSIZE;
|
|
|
|
switch (rtype) {
|
|
case DATA: {
|
|
h->currblock = 1;
|
|
h->validsize = res;
|
|
h->islastblock = 0;
|
|
if (res < h->tftp_blksize) {
|
|
h->islastblock = 1; /* very short file */
|
|
tftp_sendack(h);
|
|
}
|
|
return (0);
|
|
}
|
|
case ERROR:
|
|
default:
|
|
return (errno);
|
|
}
|
|
|
|
}
|
|
|
|
/* ack block, expect next */
|
|
static int
|
|
tftp_getnextblock(struct tftp_handle *h)
|
|
{
|
|
struct {
|
|
u_char header[HEADER_SIZE];
|
|
struct tftphdr t;
|
|
} __packed __aligned(4) wbuf;
|
|
char *wtail;
|
|
int res;
|
|
struct tftphdr *t;
|
|
unsigned short rtype = 0;
|
|
wbuf.t.th_opcode = htons((u_short) ACK);
|
|
wtail = (char *) &wbuf.t.th_block;
|
|
wbuf.t.th_block = htons((u_short) h->currblock);
|
|
wtail += 2;
|
|
|
|
t = &h->lastdata.t;
|
|
|
|
h->iodesc->xid = h->currblock + 1; /* expected block */
|
|
|
|
res = sendrecv_tftp(h, &sendudp, &wbuf.t, wtail - (char *) &wbuf.t,
|
|
&recvtftp, t, sizeof(*t) + h->tftp_blksize, &rtype);
|
|
|
|
if (res == -1) /* 0 is OK! */
|
|
return (errno);
|
|
|
|
h->currblock++;
|
|
h->validsize = res;
|
|
if (res < h->tftp_blksize)
|
|
h->islastblock = 1; /* EOF */
|
|
|
|
if (h->islastblock == 1) {
|
|
/* Send an ACK for the last block */
|
|
wbuf.t.th_block = htons((u_short) h->currblock);
|
|
sendudp(h->iodesc, &wbuf.t, wtail - (char *)&wbuf.t);
|
|
}
|
|
|
|
return (0);
|
|
}
|
|
|
|
static int
|
|
tftp_open(const char *path, struct open_file *f)
|
|
{
|
|
struct tftp_handle *tftpfile;
|
|
struct iodesc *io;
|
|
int res;
|
|
|
|
if (strcmp(f->f_dev->dv_name, "net") != 0) {
|
|
#ifdef __i386__
|
|
if (strcmp(f->f_dev->dv_name, "pxe") != 0)
|
|
return (EINVAL);
|
|
#else
|
|
return (EINVAL);
|
|
#endif
|
|
}
|
|
|
|
if (is_open)
|
|
return (EBUSY);
|
|
|
|
tftpfile = (struct tftp_handle *) malloc(sizeof(*tftpfile));
|
|
if (!tftpfile)
|
|
return (ENOMEM);
|
|
|
|
memset(tftpfile, 0, sizeof(*tftpfile));
|
|
tftpfile->tftp_blksize = TFTP_REQUESTED_BLKSIZE;
|
|
tftpfile->iodesc = io = socktodesc(*(int *) (f->f_devdata));
|
|
if (io == NULL)
|
|
return (EINVAL);
|
|
|
|
io->destip = servip;
|
|
tftpfile->off = 0;
|
|
tftpfile->path = strdup(path);
|
|
if (tftpfile->path == NULL) {
|
|
free(tftpfile);
|
|
return(ENOMEM);
|
|
}
|
|
|
|
res = tftp_makereq(tftpfile);
|
|
|
|
if (res) {
|
|
free(tftpfile->path);
|
|
free(tftpfile);
|
|
return (res);
|
|
}
|
|
f->f_fsdata = (void *) tftpfile;
|
|
is_open = 1;
|
|
return (0);
|
|
}
|
|
|
|
static int
|
|
tftp_read(struct open_file *f, void *addr, size_t size,
|
|
size_t *resid /* out */)
|
|
{
|
|
struct tftp_handle *tftpfile;
|
|
static int tc = 0;
|
|
tftpfile = (struct tftp_handle *) f->f_fsdata;
|
|
|
|
while (size > 0) {
|
|
int needblock, count;
|
|
|
|
if (!(tc++ % 16))
|
|
twiddle();
|
|
|
|
needblock = tftpfile->off / tftpfile->tftp_blksize + 1;
|
|
|
|
if (tftpfile->currblock > needblock) { /* seek backwards */
|
|
tftp_senderr(tftpfile, 0, "No error: read aborted");
|
|
tftp_makereq(tftpfile); /* no error check, it worked
|
|
* for open */
|
|
}
|
|
|
|
while (tftpfile->currblock < needblock) {
|
|
int res;
|
|
|
|
res = tftp_getnextblock(tftpfile);
|
|
if (res) { /* no answer */
|
|
#ifdef TFTP_DEBUG
|
|
printf("tftp: read error\n");
|
|
#endif
|
|
return (res);
|
|
}
|
|
if (tftpfile->islastblock)
|
|
break;
|
|
}
|
|
|
|
if (tftpfile->currblock == needblock) {
|
|
int offinblock, inbuffer;
|
|
|
|
offinblock = tftpfile->off % tftpfile->tftp_blksize;
|
|
|
|
inbuffer = tftpfile->validsize - offinblock;
|
|
if (inbuffer < 0) {
|
|
#ifdef TFTP_DEBUG
|
|
printf("tftp: invalid offset %d\n",
|
|
tftpfile->off);
|
|
#endif
|
|
return (EINVAL);
|
|
}
|
|
count = (size < inbuffer ? size : inbuffer);
|
|
bcopy(tftpfile->lastdata.t.th_data + offinblock,
|
|
addr, count);
|
|
|
|
addr = (char *)addr + count;
|
|
tftpfile->off += count;
|
|
size -= count;
|
|
|
|
if ((tftpfile->islastblock) && (count == inbuffer))
|
|
break; /* EOF */
|
|
} else {
|
|
#ifdef TFTP_DEBUG
|
|
printf("tftp: block %d not found\n", needblock);
|
|
#endif
|
|
return (EINVAL);
|
|
}
|
|
|
|
}
|
|
|
|
if (resid)
|
|
*resid = size;
|
|
return (0);
|
|
}
|
|
|
|
static int
|
|
tftp_close(struct open_file *f)
|
|
{
|
|
struct tftp_handle *tftpfile;
|
|
tftpfile = (struct tftp_handle *) f->f_fsdata;
|
|
|
|
/* let it time out ... */
|
|
|
|
if (tftpfile) {
|
|
free(tftpfile->path);
|
|
free(tftpfile);
|
|
}
|
|
is_open = 0;
|
|
return (0);
|
|
}
|
|
|
|
static int
|
|
tftp_write(struct open_file *f __unused, void *start __unused, size_t size __unused,
|
|
size_t *resid __unused /* out */)
|
|
{
|
|
return (EROFS);
|
|
}
|
|
|
|
static int
|
|
tftp_stat(struct open_file *f, struct stat *sb)
|
|
{
|
|
struct tftp_handle *tftpfile;
|
|
tftpfile = (struct tftp_handle *) f->f_fsdata;
|
|
|
|
sb->st_mode = 0444 | S_IFREG;
|
|
sb->st_nlink = 1;
|
|
sb->st_uid = 0;
|
|
sb->st_gid = 0;
|
|
sb->st_size = -1;
|
|
return (0);
|
|
}
|
|
|
|
static off_t
|
|
tftp_seek(struct open_file *f, off_t offset, int where)
|
|
{
|
|
struct tftp_handle *tftpfile;
|
|
tftpfile = (struct tftp_handle *) f->f_fsdata;
|
|
|
|
switch (where) {
|
|
case SEEK_SET:
|
|
tftpfile->off = offset;
|
|
break;
|
|
case SEEK_CUR:
|
|
tftpfile->off += offset;
|
|
break;
|
|
default:
|
|
errno = EOFFSET;
|
|
return (-1);
|
|
}
|
|
return (tftpfile->off);
|
|
}
|
|
|
|
static ssize_t
|
|
sendrecv_tftp(struct tftp_handle *h,
|
|
ssize_t (*sproc)(struct iodesc *, void *, size_t),
|
|
void *sbuf, size_t ssize,
|
|
ssize_t (*rproc)(struct tftp_handle *, void *, ssize_t, time_t, unsigned short *),
|
|
void *rbuf, size_t rsize, unsigned short *rtype)
|
|
{
|
|
struct iodesc *d = h->iodesc;
|
|
ssize_t cc;
|
|
time_t t, t1, tleft;
|
|
|
|
#ifdef TFTP_DEBUG
|
|
if (debug)
|
|
printf("sendrecv: called\n");
|
|
#endif
|
|
|
|
tleft = MINTMO;
|
|
t = t1 = getsecs();
|
|
for (;;) {
|
|
if ((getsecs() - t) > MAXTMO) {
|
|
errno = ETIMEDOUT;
|
|
return -1;
|
|
}
|
|
|
|
cc = (*sproc)(d, sbuf, ssize);
|
|
if (cc != -1 && cc < ssize)
|
|
panic("sendrecv: short write! (%zd < %zu)",
|
|
cc, ssize);
|
|
|
|
if (cc == -1) {
|
|
/* Error on transmit; wait before retrying */
|
|
while ((getsecs() - t1) < tleft);
|
|
continue;
|
|
}
|
|
|
|
recvnext:
|
|
/* Try to get a packet and process it. */
|
|
cc = (*rproc)(h, rbuf, rsize, tleft, rtype);
|
|
/* Return on data, EOF or real error. */
|
|
if (cc != -1 || errno != 0)
|
|
return (cc);
|
|
if ((getsecs() - t1) < tleft) {
|
|
goto recvnext;
|
|
}
|
|
|
|
/* Timed out or didn't get the packet we're waiting for */
|
|
tleft += MINTMO;
|
|
if (tleft > (2 * MINTMO)) {
|
|
tleft = (2 * MINTMO);
|
|
}
|
|
t1 = getsecs();
|
|
}
|
|
}
|
|
|
|
static int
|
|
tftp_set_blksize(struct tftp_handle *h, const char *str)
|
|
{
|
|
char *endptr;
|
|
int new_blksize;
|
|
int ret = 0;
|
|
|
|
if (h == NULL || str == NULL)
|
|
return (ret);
|
|
|
|
new_blksize =
|
|
(unsigned int)strtol(str, &endptr, 0);
|
|
|
|
/*
|
|
* Only accept blksize value if it is numeric.
|
|
* RFC2348 specifies that acceptable values are 8-65464.
|
|
* Let's choose a limit less than MAXRSPACE.
|
|
*/
|
|
if (*endptr == '\0' && new_blksize >= 8
|
|
&& new_blksize <= TFTP_MAX_BLKSIZE) {
|
|
h->tftp_blksize = new_blksize;
|
|
ret = 1;
|
|
}
|
|
|
|
return (ret);
|
|
}
|
|
|
|
/*
|
|
* In RFC2347, the TFTP Option Acknowledgement package (OACK)
|
|
* is used to acknowledge a client's option negotiation request.
|
|
* The format of an OACK packet is:
|
|
* +-------+---~~---+---+---~~---+---+---~~---+---+---~~---+---+
|
|
* | opc | opt1 | 0 | value1 | 0 | optN | 0 | valueN | 0 |
|
|
* +-------+---~~---+---+---~~---+---+---~~---+---+---~~---+---+
|
|
*
|
|
* opc
|
|
* The opcode field contains a 6, for Option Acknowledgment.
|
|
*
|
|
* opt1
|
|
* The first option acknowledgment, copied from the original
|
|
* request.
|
|
*
|
|
* value1
|
|
* The acknowledged value associated with the first option. If
|
|
* and how this value may differ from the original request is
|
|
* detailed in the specification for the option.
|
|
*
|
|
* optN, valueN
|
|
* The final option/value acknowledgment pair.
|
|
*/
|
|
static int
|
|
tftp_parse_oack(struct tftp_handle *h, char *buf, size_t len)
|
|
{
|
|
/*
|
|
* We parse the OACK strings into an array
|
|
* of name-value pairs.
|
|
*/
|
|
char *tftp_options[128] = { 0 };
|
|
char *val = buf;
|
|
int i = 0;
|
|
int option_idx = 0;
|
|
int blksize_is_set = 0;
|
|
int tsize = 0;
|
|
|
|
unsigned int orig_blksize;
|
|
|
|
while (option_idx < 128 && i < len) {
|
|
if (buf[i] == '\0') {
|
|
if (&buf[i] > val) {
|
|
tftp_options[option_idx] = val;
|
|
val = &buf[i] + 1;
|
|
++option_idx;
|
|
}
|
|
}
|
|
++i;
|
|
}
|
|
|
|
/* Save the block size we requested for sanity check later. */
|
|
orig_blksize = h->tftp_blksize;
|
|
|
|
/*
|
|
* Parse individual TFTP options.
|
|
* * "blksize" is specified in RFC2348.
|
|
* * "tsize" is specified in RFC2349.
|
|
*/
|
|
for (i = 0; i < option_idx; i += 2) {
|
|
if (strcasecmp(tftp_options[i], "blksize") == 0) {
|
|
if (i + 1 < option_idx)
|
|
blksize_is_set =
|
|
tftp_set_blksize(h, tftp_options[i + 1]);
|
|
} else if (strcasecmp(tftp_options[i], "tsize") == 0) {
|
|
if (i + 1 < option_idx)
|
|
tsize = strtol(tftp_options[i + 1], (char **)NULL, 10);
|
|
} else {
|
|
/* Do not allow any options we did not expect to be ACKed. */
|
|
printf("unexpected tftp option '%s'\n", tftp_options[i]);
|
|
return (-1);
|
|
}
|
|
}
|
|
|
|
if (!blksize_is_set) {
|
|
/*
|
|
* If TFTP blksize was not set, try defaulting
|
|
* to the legacy TFTP blksize of SEGSIZE(512)
|
|
*/
|
|
h->tftp_blksize = SEGSIZE;
|
|
} else if (h->tftp_blksize > orig_blksize) {
|
|
/*
|
|
* Server should not be proposing block sizes that
|
|
* exceed what we said we can handle.
|
|
*/
|
|
printf("unexpected blksize %u\n", h->tftp_blksize);
|
|
return (-1);
|
|
}
|
|
|
|
#ifdef TFTP_DEBUG
|
|
printf("tftp_blksize: %u\n", h->tftp_blksize);
|
|
printf("tftp_tsize: %lu\n", h->tftp_tsize);
|
|
#endif
|
|
return 0;
|
|
}
|