2 * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved.
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the BSD-type
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
14 * Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
17 * Redistributions in binary form must reproduce the above
18 * copyright notice, this list of conditions and the following
19 * disclaimer in the documentation and/or other materials provided
20 * with the distribution.
22 * Neither the name of the Network Appliance, Inc. nor the names of
23 * its contributors may be used to endorse or promote products
24 * derived from this software without specific prior written
27 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
28 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
29 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
30 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
31 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
32 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
33 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
34 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
35 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
36 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
37 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
39 * Author: Tom Tucker <tom@opengridcomputing.com>
42 #include <linux/sunrpc/xdr.h>
43 #include <linux/sunrpc/debug.h>
44 #include <asm/unaligned.h>
45 #include <linux/sunrpc/rpc_rdma.h>
46 #include <linux/sunrpc/svc_rdma.h>
48 #define RPCDBG_FACILITY RPCDBG_SVCXPRT
51 * Decodes a read chunk list. The expected format is as follows:
53 * position : __be32 offset into XDR stream
54 * handle : __be32 RKEY
56 * end-of-list: xdr_zero
58 static __be32 *decode_read_list(__be32 *va, __be32 *vaend)
60 struct rpcrdma_read_chunk *ch = (struct rpcrdma_read_chunk *)va;
62 while (ch->rc_discrim != xdr_zero) {
63 if (((unsigned long)ch + sizeof(struct rpcrdma_read_chunk)) >
64 (unsigned long)vaend) {
65 dprintk("svcrdma: vaend=%p, ch=%p\n", vaend, ch);
70 return &ch->rc_position;
74 * Decodes a write chunk list. The expected format is as follows:
77 * handle : __be32 RKEY ---+
78 * length : __be32 <len of segment> |
79 * offset : remove va + <count>
83 static __be32 *decode_write_list(__be32 *va, __be32 *vaend)
85 unsigned long start, end;
88 struct rpcrdma_write_array *ary =
89 (struct rpcrdma_write_array *)va;
91 /* Check for not write-array */
92 if (ary->wc_discrim == xdr_zero)
93 return &ary->wc_nchunks;
95 if ((unsigned long)ary + sizeof(struct rpcrdma_write_array) >
96 (unsigned long)vaend) {
97 dprintk("svcrdma: ary=%p, vaend=%p\n", ary, vaend);
100 nchunks = be32_to_cpu(ary->wc_nchunks);
102 start = (unsigned long)&ary->wc_array[0];
103 end = (unsigned long)vaend;
105 nchunks > (SIZE_MAX - start) / sizeof(struct rpcrdma_write_chunk) ||
106 (start + (sizeof(struct rpcrdma_write_chunk) * nchunks)) > end) {
107 dprintk("svcrdma: ary=%p, wc_nchunks=%d, vaend=%p\n",
108 ary, nchunks, vaend);
112 * rs_length is the 2nd 4B field in wc_target and taking its
113 * address skips the list terminator
115 return &ary->wc_array[nchunks].wc_target.rs_length;
118 static __be32 *decode_reply_array(__be32 *va, __be32 *vaend)
120 unsigned long start, end;
122 struct rpcrdma_write_array *ary =
123 (struct rpcrdma_write_array *)va;
125 /* Check for no reply-array */
126 if (ary->wc_discrim == xdr_zero)
127 return &ary->wc_nchunks;
129 if ((unsigned long)ary + sizeof(struct rpcrdma_write_array) >
130 (unsigned long)vaend) {
131 dprintk("svcrdma: ary=%p, vaend=%p\n", ary, vaend);
134 nchunks = be32_to_cpu(ary->wc_nchunks);
136 start = (unsigned long)&ary->wc_array[0];
137 end = (unsigned long)vaend;
139 nchunks > (SIZE_MAX - start) / sizeof(struct rpcrdma_write_chunk) ||
140 (start + (sizeof(struct rpcrdma_write_chunk) * nchunks)) > end) {
141 dprintk("svcrdma: ary=%p, wc_nchunks=%d, vaend=%p\n",
142 ary, nchunks, vaend);
145 return (__be32 *)&ary->wc_array[nchunks];
149 * svc_rdma_xdr_decode_req - Parse incoming RPC-over-RDMA header
150 * @rq_arg: Receive buffer
152 * On entry, xdr->head[0].iov_base points to first byte in the
153 * RPC-over-RDMA header.
155 * On successful exit, head[0] points to first byte past the
156 * RPC-over-RDMA header. For RDMA_MSG, this is the RPC message.
157 * The length of the RPC-over-RDMA header is returned.
159 int svc_rdma_xdr_decode_req(struct xdr_buf *rq_arg)
161 struct rpcrdma_msg *rmsgp;
166 /* Verify that there's enough bytes for header + something */
167 if (rq_arg->len <= RPCRDMA_HDRLEN_ERR) {
168 dprintk("svcrdma: header too short = %d\n",
173 rmsgp = (struct rpcrdma_msg *)rq_arg->head[0].iov_base;
174 if (rmsgp->rm_vers != rpcrdma_version) {
175 dprintk("%s: bad version %u\n", __func__,
176 be32_to_cpu(rmsgp->rm_vers));
177 return -EPROTONOSUPPORT;
180 switch (be32_to_cpu(rmsgp->rm_type)) {
187 dprintk("svcrdma: dropping RDMA_DONE message\n");
191 /* Possible if this is a backchannel reply.
192 * XXX: We should cancel this XID, though.
194 dprintk("svcrdma: dropping RDMA_ERROR message\n");
198 /* Pull in the extra for the padded case, bump our pointer */
199 rmsgp->rm_body.rm_padded.rm_align =
200 be32_to_cpu(rmsgp->rm_body.rm_padded.rm_align);
201 rmsgp->rm_body.rm_padded.rm_thresh =
202 be32_to_cpu(rmsgp->rm_body.rm_padded.rm_thresh);
204 va = &rmsgp->rm_body.rm_padded.rm_pempty[4];
205 rq_arg->head[0].iov_base = va;
206 len = (u32)((unsigned long)va - (unsigned long)rmsgp);
207 rq_arg->head[0].iov_len -= len;
208 if (len > rq_arg->len)
212 dprintk("svcrdma: bad rdma procedure (%u)\n",
213 be32_to_cpu(rmsgp->rm_type));
217 /* The chunk list may contain either a read chunk list or a write
218 * chunk list and a reply chunk list.
220 va = &rmsgp->rm_body.rm_chunks[0];
221 vaend = (__be32 *)((unsigned long)rmsgp + rq_arg->len);
222 va = decode_read_list(va, vaend);
224 dprintk("svcrdma: failed to decode read list\n");
227 va = decode_write_list(va, vaend);
229 dprintk("svcrdma: failed to decode write list\n");
232 va = decode_reply_array(va, vaend);
234 dprintk("svcrdma: failed to decode reply chunk\n");
238 rq_arg->head[0].iov_base = va;
239 hdr_len = (unsigned long)va - (unsigned long)rmsgp;
240 rq_arg->head[0].iov_len -= hdr_len;
244 int svc_rdma_xdr_encode_error(struct svcxprt_rdma *xprt,
245 struct rpcrdma_msg *rmsgp,
246 enum rpcrdma_errcode err, __be32 *va)
250 *va++ = rmsgp->rm_xid;
251 *va++ = rmsgp->rm_vers;
252 *va++ = cpu_to_be32(xprt->sc_max_requests);
254 *va++ = cpu_to_be32(err);
255 if (err == ERR_VERS) {
256 *va++ = rpcrdma_version;
257 *va++ = rpcrdma_version;
260 return (int)((unsigned long)va - (unsigned long)startp);
263 int svc_rdma_xdr_get_reply_hdr_len(struct rpcrdma_msg *rmsgp)
265 struct rpcrdma_write_array *wr_ary;
267 /* There is no read-list in a reply */
269 /* skip write list */
270 wr_ary = (struct rpcrdma_write_array *)
271 &rmsgp->rm_body.rm_chunks[1];
272 if (wr_ary->wc_discrim)
273 wr_ary = (struct rpcrdma_write_array *)
274 &wr_ary->wc_array[be32_to_cpu(wr_ary->wc_nchunks)].
277 wr_ary = (struct rpcrdma_write_array *)
280 /* skip reply array */
281 if (wr_ary->wc_discrim)
282 wr_ary = (struct rpcrdma_write_array *)
283 &wr_ary->wc_array[be32_to_cpu(wr_ary->wc_nchunks)];
285 wr_ary = (struct rpcrdma_write_array *)
288 return (unsigned long) wr_ary - (unsigned long) rmsgp;
291 void svc_rdma_xdr_encode_write_list(struct rpcrdma_msg *rmsgp, int chunks)
293 struct rpcrdma_write_array *ary;
296 rmsgp->rm_body.rm_chunks[0] = xdr_zero;
298 /* write-array discrim */
299 ary = (struct rpcrdma_write_array *)
300 &rmsgp->rm_body.rm_chunks[1];
301 ary->wc_discrim = xdr_one;
302 ary->wc_nchunks = cpu_to_be32(chunks);
304 /* write-list terminator */
305 ary->wc_array[chunks].wc_target.rs_handle = xdr_zero;
307 /* reply-array discriminator */
308 ary->wc_array[chunks].wc_target.rs_length = xdr_zero;
311 void svc_rdma_xdr_encode_reply_array(struct rpcrdma_write_array *ary,
314 ary->wc_discrim = xdr_one;
315 ary->wc_nchunks = cpu_to_be32(chunks);
318 void svc_rdma_xdr_encode_array_chunk(struct rpcrdma_write_array *ary,
324 struct rpcrdma_segment *seg = &ary->wc_array[chunk_no].wc_target;
325 seg->rs_handle = rs_handle;
326 seg->rs_offset = rs_offset;
327 seg->rs_length = cpu_to_be32(write_len);
330 void svc_rdma_xdr_encode_reply_header(struct svcxprt_rdma *xprt,
331 struct rpcrdma_msg *rdma_argp,
332 struct rpcrdma_msg *rdma_resp,
333 enum rpcrdma_proc rdma_type)
335 rdma_resp->rm_xid = rdma_argp->rm_xid;
336 rdma_resp->rm_vers = rdma_argp->rm_vers;
337 rdma_resp->rm_credit = cpu_to_be32(xprt->sc_max_requests);
338 rdma_resp->rm_type = cpu_to_be32(rdma_type);
340 /* Encode <nul> chunks lists */
341 rdma_resp->rm_body.rm_chunks[0] = xdr_zero;
342 rdma_resp->rm_body.rm_chunks[1] = xdr_zero;
343 rdma_resp->rm_body.rm_chunks[2] = xdr_zero;