1 /*
2 * Copyright (c) 2006 Oracle. 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
8 * OpenIB.org BSD license below:
9 *
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
12 * conditions are met:
13 *
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
16 * disclaimer.
17 *
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
22 *
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30 * SOFTWARE.
31 *
32 */
33 #include <linux/percpu.h>
34 #include <linux/seq_file.h>
35 #include <linux/slab.h>
36 #include <linux/proc_fs.h>
37 #include <linux/export.h>
38
39 #include "rds.h"
40
41 /*
42 * This file implements a getsockopt() call which copies a set of fixed
43 * sized structs into a user-specified buffer as a means of providing
44 * read-only information about RDS.
45 *
46 * For a given information source there are a given number of fixed sized
47 * structs at a given time. The structs are only copied if the user-specified
48 * buffer is big enough. The destination pages that make up the buffer
49 * are pinned for the duration of the copy.
50 *
51 * This gives us the following benefits:
52 *
53 * - simple implementation, no copy "position" across multiple calls
54 * - consistent snapshot of an info source
55 * - atomic copy works well with whatever locking info source has
56 * - one portable tool to get rds info across implementations
57 * - long-lived tool can get info without allocating
58 *
59 * at the following costs:
60 *
61 * - info source copy must be pinned, may be "large"
62 */
63
64 struct rds_info_iterator {
65 struct page **pages;
66 void *addr;
67 unsigned long offset;
68 };
69
70 static DEFINE_SPINLOCK(rds_info_lock);
71 static rds_info_func rds_info_funcs[RDS_INFO_LAST - RDS_INFO_FIRST + 1];
72
rds_info_register_func(int optname,rds_info_func func)73 void rds_info_register_func(int optname, rds_info_func func)
74 {
75 int offset = optname - RDS_INFO_FIRST;
76
77 BUG_ON(optname < RDS_INFO_FIRST || optname > RDS_INFO_LAST);
78
79 spin_lock(&rds_info_lock);
80 BUG_ON(rds_info_funcs[offset]);
81 rds_info_funcs[offset] = func;
82 spin_unlock(&rds_info_lock);
83 }
84 EXPORT_SYMBOL_GPL(rds_info_register_func);
85
rds_info_deregister_func(int optname,rds_info_func func)86 void rds_info_deregister_func(int optname, rds_info_func func)
87 {
88 int offset = optname - RDS_INFO_FIRST;
89
90 BUG_ON(optname < RDS_INFO_FIRST || optname > RDS_INFO_LAST);
91
92 spin_lock(&rds_info_lock);
93 BUG_ON(rds_info_funcs[offset] != func);
94 rds_info_funcs[offset] = NULL;
95 spin_unlock(&rds_info_lock);
96 }
97 EXPORT_SYMBOL_GPL(rds_info_deregister_func);
98
99 /*
100 * Typically we hold an atomic kmap across multiple rds_info_copy() calls
101 * because the kmap is so expensive. This must be called before using blocking
102 * operations while holding the mapping and as the iterator is torn down.
103 */
rds_info_iter_unmap(struct rds_info_iterator * iter)104 void rds_info_iter_unmap(struct rds_info_iterator *iter)
105 {
106 if (iter->addr) {
107 kunmap_atomic(iter->addr);
108 iter->addr = NULL;
109 }
110 }
111
112 /*
113 * get_user_pages() called flush_dcache_page() on the pages for us.
114 */
rds_info_copy(struct rds_info_iterator * iter,void * data,unsigned long bytes)115 void rds_info_copy(struct rds_info_iterator *iter, void *data,
116 unsigned long bytes)
117 {
118 unsigned long this;
119
120 while (bytes) {
121 if (!iter->addr)
122 iter->addr = kmap_atomic(*iter->pages);
123
124 this = min(bytes, PAGE_SIZE - iter->offset);
125
126 rdsdebug("page %p addr %p offset %lu this %lu data %p "
127 "bytes %lu\n", *iter->pages, iter->addr,
128 iter->offset, this, data, bytes);
129
130 memcpy(iter->addr + iter->offset, data, this);
131
132 data += this;
133 bytes -= this;
134 iter->offset += this;
135
136 if (iter->offset == PAGE_SIZE) {
137 kunmap_atomic(iter->addr);
138 iter->addr = NULL;
139 iter->offset = 0;
140 iter->pages++;
141 }
142 }
143 }
144 EXPORT_SYMBOL_GPL(rds_info_copy);
145
146 /*
147 * @optval points to the userspace buffer that the information snapshot
148 * will be copied into.
149 *
150 * @optlen on input is the size of the buffer in userspace. @optlen
151 * on output is the size of the requested snapshot in bytes.
152 *
153 * This function returns -errno if there is a failure, particularly -ENOSPC
154 * if the given userspace buffer was not large enough to fit the snapshot.
155 * On success it returns the positive number of bytes of each array element
156 * in the snapshot.
157 */
rds_info_getsockopt(struct socket * sock,int optname,char __user * optval,int __user * optlen)158 int rds_info_getsockopt(struct socket *sock, int optname, char __user *optval,
159 int __user *optlen)
160 {
161 struct rds_info_iterator iter;
162 struct rds_info_lengths lens;
163 unsigned long nr_pages = 0;
164 unsigned long start;
165 unsigned long i;
166 rds_info_func func;
167 struct page **pages = NULL;
168 int ret;
169 int len;
170 int total;
171
172 if (get_user(len, optlen)) {
173 ret = -EFAULT;
174 goto out;
175 }
176
177 /* check for all kinds of wrapping and the like */
178 start = (unsigned long)optval;
179 if (len < 0 || len > INT_MAX - PAGE_SIZE + 1 || start + len < start) {
180 ret = -EINVAL;
181 goto out;
182 }
183
184 /* a 0 len call is just trying to probe its length */
185 if (len == 0)
186 goto call_func;
187
188 nr_pages = (PAGE_ALIGN(start + len) - (start & PAGE_MASK))
189 >> PAGE_SHIFT;
190
191 pages = kmalloc_array(nr_pages, sizeof(struct page *), GFP_KERNEL);
192 if (!pages) {
193 ret = -ENOMEM;
194 goto out;
195 }
196 ret = get_user_pages_fast(start, nr_pages, FOLL_WRITE, pages);
197 if (ret != nr_pages) {
198 if (ret > 0)
199 nr_pages = ret;
200 else
201 nr_pages = 0;
202 ret = -EAGAIN; /* XXX ? */
203 goto out;
204 }
205
206 rdsdebug("len %d nr_pages %lu\n", len, nr_pages);
207
208 call_func:
209 func = rds_info_funcs[optname - RDS_INFO_FIRST];
210 if (!func) {
211 ret = -ENOPROTOOPT;
212 goto out;
213 }
214
215 iter.pages = pages;
216 iter.addr = NULL;
217 iter.offset = start & (PAGE_SIZE - 1);
218
219 func(sock, len, &iter, &lens);
220 BUG_ON(lens.each == 0);
221
222 total = lens.nr * lens.each;
223
224 rds_info_iter_unmap(&iter);
225
226 if (total > len) {
227 len = total;
228 ret = -ENOSPC;
229 } else {
230 len = total;
231 ret = lens.each;
232 }
233
234 if (put_user(len, optlen))
235 ret = -EFAULT;
236
237 out:
238 for (i = 0; pages && i < nr_pages; i++)
239 put_page(pages[i]);
240 kfree(pages);
241
242 return ret;
243 }
244