2 * xdr_rec.c, Implements TCP/IP based XDR streams with a "record marking"
3 * layer above tcp (for rpc's use).
5 * Copyright (c) 2010, Oracle America, Inc.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions are
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above
14 * copyright notice, this list of conditions and the following
15 * disclaimer in the documentation and/or other materials
16 * provided with the distribution.
17 * * Neither the name of the "Oracle America, Inc." nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
26 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE
28 * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
29 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
30 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
31 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 * These routines interface XDRSTREAMS to a tcp/ip connection.
35 * There is a record marking layer between the xdr stream
36 * and the tcp transport level. A record is composed on one or more
37 * record fragments. A record fragment is a thirty-two bit header followed
38 * by n bytes of data, where n is contained in the header. The header
39 * is represented as a htonl(u_long). The high order bit encodes
40 * whether or not the fragment is the last fragment of the record
41 * (1 => fragment is last, 0 => more fragments to follow.
42 * The other 31 bits encode the byte length of the fragment.
52 #include <libio/iolibio.h>
53 #include <shlib-compat.h>
55 static bool_t
xdrrec_getlong (XDR
*, long *);
56 static bool_t
xdrrec_putlong (XDR
*, const long *);
57 static bool_t
xdrrec_getbytes (XDR
*, caddr_t
, u_int
);
58 static bool_t
xdrrec_putbytes (XDR
*, const char *, u_int
);
59 static u_int
xdrrec_getpos (const XDR
*);
60 static bool_t
xdrrec_setpos (XDR
*, u_int
);
61 static int32_t *xdrrec_inline (XDR
*, u_int
);
62 static void xdrrec_destroy (XDR
*);
63 static bool_t
xdrrec_getint32 (XDR
*, int32_t *);
64 static bool_t
xdrrec_putint32 (XDR
*, const int32_t *);
66 static const struct xdr_ops xdrrec_ops
= {
80 * A record is composed of one or more record fragments.
81 * A record fragment is a two-byte header followed by zero to
82 * 2**32-1 bytes. The header is treated as a long unsigned and is
83 * encode/decoded to the network via htonl/ntohl. The low order 31 bits
84 * are a byte count of the fragment. The highest order bit is a boolean:
85 * 1 => this fragment is the last fragment of the record,
86 * 0 => this fragment is followed by more fragment(s).
88 * The fragment/record machinery is not general; it is constructed to
89 * meet the needs of xdr and rpc based on tcp.
92 #define LAST_FRAG (1UL << 31)
94 typedef struct rec_strm
101 int (*writeit
) (char *, char *, int);
102 caddr_t out_base
; /* output buffer (points to frag header) */
103 caddr_t out_finger
; /* next output position */
104 caddr_t out_boundry
; /* data cannot up to this address */
105 uint32_t *frag_header
; /* beginning of current fragment */
106 bool_t frag_sent
; /* true if buffer sent in middle of record */
110 int (*readit
) (char *, char *, int);
111 u_long in_size
; /* fixed size of the input buffer */
113 caddr_t in_finger
; /* location of next byte to be had */
114 caddr_t in_boundry
; /* can read up to this location */
115 long fbtbc
; /* fragment bytes to be consumed */
122 static u_int
fix_buf_size (u_int
);
123 static bool_t
skip_input_bytes (RECSTREAM
*, long);
124 static bool_t
flush_out (RECSTREAM
*, bool_t
);
125 static bool_t
set_input_fragment (RECSTREAM
*);
126 static bool_t
get_input_bytes (RECSTREAM
*, caddr_t
, int);
129 * Create an xdr handle for xdrrec
130 * xdrrec_create fills in xdrs. Sendsize and recvsize are
131 * send and recv buffer sizes (0 => use default).
132 * tcp_handle is an opaque handle that is passed as the first parameter to
133 * the procedures readit and writeit. Readit and writeit are read and
134 * write respectively. They are like the system
135 * calls expect that they take an opaque handle rather than an fd.
138 xdrrec_create (XDR
*xdrs
, u_int sendsize
,
139 u_int recvsize
, caddr_t tcp_handle
,
140 int (*readit
) (char *, char *, int),
141 int (*writeit
) (char *, char *, int))
143 RECSTREAM
*rstrm
= (RECSTREAM
*) mem_alloc (sizeof (RECSTREAM
));
147 sendsize
= fix_buf_size (sendsize
);
148 recvsize
= fix_buf_size (recvsize
);
149 buf
= mem_alloc (sendsize
+ recvsize
+ BYTES_PER_XDR_UNIT
);
151 if (rstrm
== NULL
|| buf
== NULL
)
153 (void) __fxprintf (NULL
, "%s: %s", __func__
, _("out of memory\n"));
154 mem_free (rstrm
, sizeof (RECSTREAM
));
155 mem_free (buf
, sendsize
+ recvsize
+ BYTES_PER_XDR_UNIT
);
157 * This is bad. Should rework xdrrec_create to
158 * return a handle, and in this case return NULL
163 * adjust sizes and allocate buffer quad byte aligned
165 rstrm
->sendsize
= sendsize
;
166 rstrm
->recvsize
= recvsize
;
167 rstrm
->the_buffer
= buf
;
168 tmp
= rstrm
->the_buffer
;
169 if ((size_t)tmp
% BYTES_PER_XDR_UNIT
)
170 tmp
+= BYTES_PER_XDR_UNIT
- (size_t)tmp
% BYTES_PER_XDR_UNIT
;
171 rstrm
->out_base
= tmp
;
172 rstrm
->in_base
= tmp
+ sendsize
;
176 /* We have to add the cast since the `struct xdr_ops' in `struct XDR'
178 xdrs
->x_ops
= (struct xdr_ops
*) &xdrrec_ops
;
179 xdrs
->x_private
= (caddr_t
) rstrm
;
180 rstrm
->tcp_handle
= tcp_handle
;
181 rstrm
->readit
= readit
;
182 rstrm
->writeit
= writeit
;
183 rstrm
->out_finger
= rstrm
->out_boundry
= rstrm
->out_base
;
184 rstrm
->frag_header
= (uint32_t *) rstrm
->out_base
;
185 rstrm
->out_finger
+= 4;
186 rstrm
->out_boundry
+= sendsize
;
187 rstrm
->frag_sent
= FALSE
;
188 rstrm
->in_size
= recvsize
;
189 rstrm
->in_boundry
= rstrm
->in_base
;
190 rstrm
->in_finger
= (rstrm
->in_boundry
+= recvsize
);
192 rstrm
->last_frag
= TRUE
;
194 libc_hidden_nolink_sunrpc (xdrrec_create
, GLIBC_2_0
)
198 * The routines defined below are the xdr ops which will go into the
199 * xdr handle filled in by xdrrec_create.
203 xdrrec_getlong (XDR
*xdrs
, long *lp
)
205 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
206 int32_t *buflp
= (int32_t *) rstrm
->in_finger
;
209 /* first try the inline, fast case */
210 if (rstrm
->fbtbc
>= BYTES_PER_XDR_UNIT
&&
211 rstrm
->in_boundry
- (char *) buflp
>= BYTES_PER_XDR_UNIT
)
213 *lp
= (int32_t) ntohl (*buflp
);
214 rstrm
->fbtbc
-= BYTES_PER_XDR_UNIT
;
215 rstrm
->in_finger
+= BYTES_PER_XDR_UNIT
;
219 if (!xdrrec_getbytes (xdrs
, (caddr_t
) & mylong
,
222 *lp
= (int32_t) ntohl (mylong
);
228 xdrrec_putlong (XDR
*xdrs
, const long *lp
)
230 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
231 int32_t *dest_lp
= (int32_t *) rstrm
->out_finger
;
233 if ((rstrm
->out_finger
+= BYTES_PER_XDR_UNIT
) > rstrm
->out_boundry
)
236 * this case should almost never happen so the code is
239 rstrm
->out_finger
-= BYTES_PER_XDR_UNIT
;
240 rstrm
->frag_sent
= TRUE
;
241 if (!flush_out (rstrm
, FALSE
))
243 dest_lp
= (int32_t *) rstrm
->out_finger
;
244 rstrm
->out_finger
+= BYTES_PER_XDR_UNIT
;
246 *dest_lp
= htonl (*lp
);
250 static bool_t
/* must manage buffers, fragments, and records */
251 xdrrec_getbytes (XDR
*xdrs
, caddr_t addr
, u_int len
)
253 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
258 current
= rstrm
->fbtbc
;
261 if (rstrm
->last_frag
)
263 if (!set_input_fragment (rstrm
))
267 current
= (len
< current
) ? len
: current
;
268 if (!get_input_bytes (rstrm
, addr
, current
))
271 rstrm
->fbtbc
-= current
;
278 xdrrec_putbytes (XDR
*xdrs
, const char *addr
, u_int len
)
280 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
285 current
= rstrm
->out_boundry
- rstrm
->out_finger
;
286 current
= (len
< current
) ? len
: current
;
287 memcpy (rstrm
->out_finger
, addr
, current
);
288 rstrm
->out_finger
+= current
;
291 if (rstrm
->out_finger
== rstrm
->out_boundry
&& len
> 0)
293 rstrm
->frag_sent
= TRUE
;
294 if (!flush_out (rstrm
, FALSE
))
302 xdrrec_getpos (const XDR
*xdrs
)
304 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
307 pos
= __lseek ((int) (long) rstrm
->tcp_handle
, (long) 0, 1);
313 pos
+= rstrm
->out_finger
- rstrm
->out_base
;
317 pos
-= rstrm
->in_boundry
- rstrm
->in_finger
;
328 xdrrec_setpos (XDR
*xdrs
, u_int pos
)
330 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
331 u_int currpos
= xdrrec_getpos (xdrs
);
332 int delta
= currpos
- pos
;
335 if ((int) currpos
!= -1)
340 newpos
= rstrm
->out_finger
- delta
;
341 if (newpos
> (caddr_t
) rstrm
->frag_header
&&
342 newpos
< rstrm
->out_boundry
)
344 rstrm
->out_finger
= newpos
;
350 newpos
= rstrm
->in_finger
- delta
;
351 if ((delta
< (int) (rstrm
->fbtbc
)) &&
352 (newpos
<= rstrm
->in_boundry
) &&
353 (newpos
>= rstrm
->in_base
))
355 rstrm
->in_finger
= newpos
;
356 rstrm
->fbtbc
-= delta
;
368 xdrrec_inline (XDR
*xdrs
, u_int len
)
370 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
377 if ((rstrm
->out_finger
+ len
) <= rstrm
->out_boundry
)
379 buf
= (int32_t *) rstrm
->out_finger
;
380 rstrm
->out_finger
+= len
;
385 if ((len
<= rstrm
->fbtbc
) &&
386 ((rstrm
->in_finger
+ len
) <= rstrm
->in_boundry
))
388 buf
= (int32_t *) rstrm
->in_finger
;
390 rstrm
->in_finger
+= len
;
401 xdrrec_destroy (XDR
*xdrs
)
403 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
405 mem_free (rstrm
->the_buffer
,
406 rstrm
->sendsize
+ rstrm
->recvsize
+ BYTES_PER_XDR_UNIT
);
407 mem_free ((caddr_t
) rstrm
, sizeof (RECSTREAM
));
411 xdrrec_getint32 (XDR
*xdrs
, int32_t *ip
)
413 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
414 int32_t *bufip
= (int32_t *) rstrm
->in_finger
;
417 /* first try the inline, fast case */
418 if (rstrm
->fbtbc
>= BYTES_PER_XDR_UNIT
&&
419 rstrm
->in_boundry
- (char *) bufip
>= BYTES_PER_XDR_UNIT
)
421 *ip
= ntohl (*bufip
);
422 rstrm
->fbtbc
-= BYTES_PER_XDR_UNIT
;
423 rstrm
->in_finger
+= BYTES_PER_XDR_UNIT
;
427 if (!xdrrec_getbytes (xdrs
, (caddr_t
) &mylong
,
430 *ip
= ntohl (mylong
);
436 xdrrec_putint32 (XDR
*xdrs
, const int32_t *ip
)
438 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
439 int32_t *dest_ip
= (int32_t *) rstrm
->out_finger
;
441 if ((rstrm
->out_finger
+= BYTES_PER_XDR_UNIT
) > rstrm
->out_boundry
)
444 * this case should almost never happen so the code is
447 rstrm
->out_finger
-= BYTES_PER_XDR_UNIT
;
448 rstrm
->frag_sent
= TRUE
;
449 if (!flush_out (rstrm
, FALSE
))
451 dest_ip
= (int32_t *) rstrm
->out_finger
;
452 rstrm
->out_finger
+= BYTES_PER_XDR_UNIT
;
454 *dest_ip
= htonl (*ip
);
459 * Exported routines to manage xdr records
463 * Before reading (deserializing from the stream, one should always call
464 * this procedure to guarantee proper record alignment.
467 xdrrec_skiprecord (XDR
*xdrs
)
469 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
471 while (rstrm
->fbtbc
> 0 || (!rstrm
->last_frag
))
473 if (!skip_input_bytes (rstrm
, rstrm
->fbtbc
))
476 if ((!rstrm
->last_frag
) && (!set_input_fragment (rstrm
)))
479 rstrm
->last_frag
= FALSE
;
482 libc_hidden_nolink_sunrpc (xdrrec_skiprecord
, GLIBC_2_0
)
485 * Lookahead function.
486 * Returns TRUE iff there is no more input in the buffer
487 * after consuming the rest of the current record.
490 xdrrec_eof (XDR
*xdrs
)
492 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
494 while (rstrm
->fbtbc
> 0 || (!rstrm
->last_frag
))
496 if (!skip_input_bytes (rstrm
, rstrm
->fbtbc
))
499 if ((!rstrm
->last_frag
) && (!set_input_fragment (rstrm
)))
502 if (rstrm
->in_finger
== rstrm
->in_boundry
)
506 libc_hidden_nolink_sunrpc (xdrrec_eof
, GLIBC_2_0
)
509 * The client must tell the package when an end-of-record has occurred.
510 * The second parameter tells whether the record should be flushed to the
511 * (output) tcp stream. (This lets the package support batched or
512 * pipelined procedure calls.) TRUE => immediate flush to tcp connection.
515 xdrrec_endofrecord (XDR
*xdrs
, bool_t sendnow
)
517 RECSTREAM
*rstrm
= (RECSTREAM
*) xdrs
->x_private
;
518 u_long len
; /* fragment length */
520 if (sendnow
|| rstrm
->frag_sent
521 || rstrm
->out_finger
+ BYTES_PER_XDR_UNIT
>= rstrm
->out_boundry
)
523 rstrm
->frag_sent
= FALSE
;
524 return flush_out (rstrm
, TRUE
);
526 len
= (rstrm
->out_finger
- (char *) rstrm
->frag_header
527 - BYTES_PER_XDR_UNIT
);
528 *rstrm
->frag_header
= htonl ((u_long
) len
| LAST_FRAG
);
529 rstrm
->frag_header
= (uint32_t *) rstrm
->out_finger
;
530 rstrm
->out_finger
+= BYTES_PER_XDR_UNIT
;
533 libc_hidden_nolink_sunrpc (xdrrec_endofrecord
, GLIBC_2_0
)
537 * Internal useful routines
540 flush_out (RECSTREAM
*rstrm
, bool_t eor
)
542 u_long eormask
= (eor
== TRUE
) ? LAST_FRAG
: 0;
543 u_long len
= (rstrm
->out_finger
- (char *) rstrm
->frag_header
544 - BYTES_PER_XDR_UNIT
);
546 *rstrm
->frag_header
= htonl (len
| eormask
);
547 len
= rstrm
->out_finger
- rstrm
->out_base
;
548 if ((*(rstrm
->writeit
)) (rstrm
->tcp_handle
, rstrm
->out_base
, (int) len
)
551 rstrm
->frag_header
= (uint32_t *) rstrm
->out_base
;
552 rstrm
->out_finger
= (caddr_t
) rstrm
->out_base
+ BYTES_PER_XDR_UNIT
;
556 static bool_t
/* knows nothing about records! Only about input buffers */
557 fill_input_buf (RECSTREAM
*rstrm
)
563 where
= rstrm
->in_base
;
564 i
= (size_t) rstrm
->in_boundry
% BYTES_PER_XDR_UNIT
;
566 len
= rstrm
->in_size
- i
;
567 if ((len
= (*(rstrm
->readit
)) (rstrm
->tcp_handle
, where
, len
)) == -1)
569 rstrm
->in_finger
= where
;
571 rstrm
->in_boundry
= where
;
575 static bool_t
/* knows nothing about records! Only about input buffers */
576 get_input_bytes (RECSTREAM
*rstrm
, caddr_t addr
, int len
)
582 current
= rstrm
->in_boundry
- rstrm
->in_finger
;
585 if (!fill_input_buf (rstrm
))
589 current
= (len
< current
) ? len
: current
;
590 memcpy (addr
, rstrm
->in_finger
, current
);
591 rstrm
->in_finger
+= current
;
598 static bool_t
/* next two bytes of the input stream are treated as a header */
599 set_input_fragment (RECSTREAM
*rstrm
)
603 if (! get_input_bytes (rstrm
, (caddr_t
)&header
, BYTES_PER_XDR_UNIT
))
605 header
= ntohl (header
);
606 rstrm
->last_frag
= ((header
& LAST_FRAG
) == 0) ? FALSE
: TRUE
;
608 * Sanity check. Try not to accept wildly incorrect fragment
609 * sizes. Unfortunately, only a size of zero can be identified as
610 * 'wildely incorrect', and this only, if it is not the last
611 * fragment of a message. Ridiculously large fragment sizes may look
612 * wrong, but we don't have any way to be certain that they aren't
613 * what the client actually intended to send us. Many existing RPC
614 * implementations may sent a fragment of size zero as the last
615 * fragment of a message.
619 rstrm
->fbtbc
= header
& ~LAST_FRAG
;
623 static bool_t
/* consumes input bytes; knows nothing about records! */
624 skip_input_bytes (RECSTREAM
*rstrm
, long cnt
)
630 current
= rstrm
->in_boundry
- rstrm
->in_finger
;
633 if (!fill_input_buf (rstrm
))
637 current
= (cnt
< current
) ? cnt
: current
;
638 rstrm
->in_finger
+= current
;
645 fix_buf_size (u_int s
)