1 /*
2 * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved.
3 *
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
8 * license below:
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 *
14 * Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 *
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.
21 *
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
25 * permission.
26 *
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.
38 *
39 * Author: Tom Tucker <tom@opengridcomputing.com>
40 */
41
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>
47
48 #define RPCDBG_FACILITY RPCDBG_SVCXPRT
49
50 /*
51 * Decodes a read chunk list. The expected format is as follows:
52 * descrim : xdr_one
53 * position : __be32 offset into XDR stream
54 * handle : __be32 RKEY
55 * . . .
56 * end-of-list: xdr_zero
57 */
decode_read_list(__be32 * va,__be32 * vaend)58 static __be32 *decode_read_list(__be32 *va, __be32 *vaend)
59 {
60 struct rpcrdma_read_chunk *ch = (struct rpcrdma_read_chunk *)va;
61
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);
66 return NULL;
67 }
68 ch++;
69 }
70 return &ch->rc_position;
71 }
72
73 /*
74 * Decodes a write chunk list. The expected format is as follows:
75 * descrim : xdr_one
76 * nchunks : <count>
77 * handle : __be32 RKEY ---+
78 * length : __be32 <len of segment> |
79 * offset : remove va + <count>
80 * . . . |
81 * ---+
82 */
decode_write_list(__be32 * va,__be32 * vaend)83 static __be32 *decode_write_list(__be32 *va, __be32 *vaend)
84 {
85 unsigned long start, end;
86 int nchunks;
87
88 struct rpcrdma_write_array *ary =
89 (struct rpcrdma_write_array *)va;
90
91 /* Check for not write-array */
92 if (ary->wc_discrim == xdr_zero)
93 return &ary->wc_nchunks;
94
95 if ((unsigned long)ary + sizeof(struct rpcrdma_write_array) >
96 (unsigned long)vaend) {
97 dprintk("svcrdma: ary=%p, vaend=%p\n", ary, vaend);
98 return NULL;
99 }
100 nchunks = be32_to_cpu(ary->wc_nchunks);
101
102 start = (unsigned long)&ary->wc_array[0];
103 end = (unsigned long)vaend;
104 if (nchunks < 0 ||
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);
109 return NULL;
110 }
111 /*
112 * rs_length is the 2nd 4B field in wc_target and taking its
113 * address skips the list terminator
114 */
115 return &ary->wc_array[nchunks].wc_target.rs_length;
116 }
117
decode_reply_array(__be32 * va,__be32 * vaend)118 static __be32 *decode_reply_array(__be32 *va, __be32 *vaend)
119 {
120 unsigned long start, end;
121 int nchunks;
122 struct rpcrdma_write_array *ary =
123 (struct rpcrdma_write_array *)va;
124
125 /* Check for no reply-array */
126 if (ary->wc_discrim == xdr_zero)
127 return &ary->wc_nchunks;
128
129 if ((unsigned long)ary + sizeof(struct rpcrdma_write_array) >
130 (unsigned long)vaend) {
131 dprintk("svcrdma: ary=%p, vaend=%p\n", ary, vaend);
132 return NULL;
133 }
134 nchunks = be32_to_cpu(ary->wc_nchunks);
135
136 start = (unsigned long)&ary->wc_array[0];
137 end = (unsigned long)vaend;
138 if (nchunks < 0 ||
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);
143 return NULL;
144 }
145 return (__be32 *)&ary->wc_array[nchunks];
146 }
147
svc_rdma_xdr_decode_req(struct rpcrdma_msg ** rdma_req,struct svc_rqst * rqstp)148 int svc_rdma_xdr_decode_req(struct rpcrdma_msg **rdma_req,
149 struct svc_rqst *rqstp)
150 {
151 struct rpcrdma_msg *rmsgp = NULL;
152 __be32 *va, *vaend;
153 u32 hdr_len;
154
155 rmsgp = (struct rpcrdma_msg *)rqstp->rq_arg.head[0].iov_base;
156
157 /* Verify that there's enough bytes for header + something */
158 if (rqstp->rq_arg.len <= RPCRDMA_HDRLEN_MIN) {
159 dprintk("svcrdma: header too short = %d\n",
160 rqstp->rq_arg.len);
161 return -EINVAL;
162 }
163
164 if (rmsgp->rm_vers != rpcrdma_version)
165 return -ENOSYS;
166
167 /* Pull in the extra for the padded case and bump our pointer */
168 if (rmsgp->rm_type == rdma_msgp) {
169 int hdrlen;
170
171 rmsgp->rm_body.rm_padded.rm_align =
172 be32_to_cpu(rmsgp->rm_body.rm_padded.rm_align);
173 rmsgp->rm_body.rm_padded.rm_thresh =
174 be32_to_cpu(rmsgp->rm_body.rm_padded.rm_thresh);
175
176 va = &rmsgp->rm_body.rm_padded.rm_pempty[4];
177 rqstp->rq_arg.head[0].iov_base = va;
178 hdrlen = (u32)((unsigned long)va - (unsigned long)rmsgp);
179 rqstp->rq_arg.head[0].iov_len -= hdrlen;
180 if (hdrlen > rqstp->rq_arg.len)
181 return -EINVAL;
182 return hdrlen;
183 }
184
185 /* The chunk list may contain either a read chunk list or a write
186 * chunk list and a reply chunk list.
187 */
188 va = &rmsgp->rm_body.rm_chunks[0];
189 vaend = (__be32 *)((unsigned long)rmsgp + rqstp->rq_arg.len);
190 va = decode_read_list(va, vaend);
191 if (!va)
192 return -EINVAL;
193 va = decode_write_list(va, vaend);
194 if (!va)
195 return -EINVAL;
196 va = decode_reply_array(va, vaend);
197 if (!va)
198 return -EINVAL;
199
200 rqstp->rq_arg.head[0].iov_base = va;
201 hdr_len = (unsigned long)va - (unsigned long)rmsgp;
202 rqstp->rq_arg.head[0].iov_len -= hdr_len;
203
204 *rdma_req = rmsgp;
205 return hdr_len;
206 }
207
svc_rdma_xdr_encode_error(struct svcxprt_rdma * xprt,struct rpcrdma_msg * rmsgp,enum rpcrdma_errcode err,__be32 * va)208 int svc_rdma_xdr_encode_error(struct svcxprt_rdma *xprt,
209 struct rpcrdma_msg *rmsgp,
210 enum rpcrdma_errcode err, __be32 *va)
211 {
212 __be32 *startp = va;
213
214 *va++ = rmsgp->rm_xid;
215 *va++ = rmsgp->rm_vers;
216 *va++ = cpu_to_be32(xprt->sc_max_requests);
217 *va++ = rdma_error;
218 *va++ = cpu_to_be32(err);
219 if (err == ERR_VERS) {
220 *va++ = rpcrdma_version;
221 *va++ = rpcrdma_version;
222 }
223
224 return (int)((unsigned long)va - (unsigned long)startp);
225 }
226
svc_rdma_xdr_get_reply_hdr_len(struct rpcrdma_msg * rmsgp)227 int svc_rdma_xdr_get_reply_hdr_len(struct rpcrdma_msg *rmsgp)
228 {
229 struct rpcrdma_write_array *wr_ary;
230
231 /* There is no read-list in a reply */
232
233 /* skip write list */
234 wr_ary = (struct rpcrdma_write_array *)
235 &rmsgp->rm_body.rm_chunks[1];
236 if (wr_ary->wc_discrim)
237 wr_ary = (struct rpcrdma_write_array *)
238 &wr_ary->wc_array[be32_to_cpu(wr_ary->wc_nchunks)].
239 wc_target.rs_length;
240 else
241 wr_ary = (struct rpcrdma_write_array *)
242 &wr_ary->wc_nchunks;
243
244 /* skip reply array */
245 if (wr_ary->wc_discrim)
246 wr_ary = (struct rpcrdma_write_array *)
247 &wr_ary->wc_array[be32_to_cpu(wr_ary->wc_nchunks)];
248 else
249 wr_ary = (struct rpcrdma_write_array *)
250 &wr_ary->wc_nchunks;
251
252 return (unsigned long) wr_ary - (unsigned long) rmsgp;
253 }
254
svc_rdma_xdr_encode_write_list(struct rpcrdma_msg * rmsgp,int chunks)255 void svc_rdma_xdr_encode_write_list(struct rpcrdma_msg *rmsgp, int chunks)
256 {
257 struct rpcrdma_write_array *ary;
258
259 /* no read-list */
260 rmsgp->rm_body.rm_chunks[0] = xdr_zero;
261
262 /* write-array discrim */
263 ary = (struct rpcrdma_write_array *)
264 &rmsgp->rm_body.rm_chunks[1];
265 ary->wc_discrim = xdr_one;
266 ary->wc_nchunks = cpu_to_be32(chunks);
267
268 /* write-list terminator */
269 ary->wc_array[chunks].wc_target.rs_handle = xdr_zero;
270
271 /* reply-array discriminator */
272 ary->wc_array[chunks].wc_target.rs_length = xdr_zero;
273 }
274
svc_rdma_xdr_encode_reply_array(struct rpcrdma_write_array * ary,int chunks)275 void svc_rdma_xdr_encode_reply_array(struct rpcrdma_write_array *ary,
276 int chunks)
277 {
278 ary->wc_discrim = xdr_one;
279 ary->wc_nchunks = cpu_to_be32(chunks);
280 }
281
svc_rdma_xdr_encode_array_chunk(struct rpcrdma_write_array * ary,int chunk_no,__be32 rs_handle,__be64 rs_offset,u32 write_len)282 void svc_rdma_xdr_encode_array_chunk(struct rpcrdma_write_array *ary,
283 int chunk_no,
284 __be32 rs_handle,
285 __be64 rs_offset,
286 u32 write_len)
287 {
288 struct rpcrdma_segment *seg = &ary->wc_array[chunk_no].wc_target;
289 seg->rs_handle = rs_handle;
290 seg->rs_offset = rs_offset;
291 seg->rs_length = cpu_to_be32(write_len);
292 }
293
svc_rdma_xdr_encode_reply_header(struct svcxprt_rdma * xprt,struct rpcrdma_msg * rdma_argp,struct rpcrdma_msg * rdma_resp,enum rpcrdma_proc rdma_type)294 void svc_rdma_xdr_encode_reply_header(struct svcxprt_rdma *xprt,
295 struct rpcrdma_msg *rdma_argp,
296 struct rpcrdma_msg *rdma_resp,
297 enum rpcrdma_proc rdma_type)
298 {
299 rdma_resp->rm_xid = rdma_argp->rm_xid;
300 rdma_resp->rm_vers = rdma_argp->rm_vers;
301 rdma_resp->rm_credit = cpu_to_be32(xprt->sc_max_requests);
302 rdma_resp->rm_type = cpu_to_be32(rdma_type);
303
304 /* Encode <nul> chunks lists */
305 rdma_resp->rm_body.rm_chunks[0] = xdr_zero;
306 rdma_resp->rm_body.rm_chunks[1] = xdr_zero;
307 rdma_resp->rm_body.rm_chunks[2] = xdr_zero;
308 }
309