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 | |
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(lock: &rds_info_lock); |
80 | BUG_ON(rds_info_funcs[offset]); |
81 | rds_info_funcs[offset] = func; |
82 | spin_unlock(lock: &rds_info_lock); |
83 | } |
84 | EXPORT_SYMBOL_GPL(rds_info_register_func); |
85 | |
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(lock: &rds_info_lock); |
93 | BUG_ON(rds_info_funcs[offset] != func); |
94 | rds_info_funcs[offset] = NULL; |
95 | spin_unlock(lock: &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 | */ |
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 | */ |
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(page: *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 | */ |
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 | rds_info_func func; |
166 | struct page **pages = NULL; |
167 | int ret; |
168 | int len; |
169 | int total; |
170 | |
171 | if (get_user(len, optlen)) { |
172 | ret = -EFAULT; |
173 | goto out; |
174 | } |
175 | |
176 | /* check for all kinds of wrapping and the like */ |
177 | start = (unsigned long)optval; |
178 | if (len < 0 || len > INT_MAX - PAGE_SIZE + 1 || start + len < start) { |
179 | ret = -EINVAL; |
180 | goto out; |
181 | } |
182 | |
183 | /* a 0 len call is just trying to probe its length */ |
184 | if (len == 0) |
185 | goto call_func; |
186 | |
187 | nr_pages = (PAGE_ALIGN(start + len) - (start & PAGE_MASK)) |
188 | >> PAGE_SHIFT; |
189 | |
190 | pages = kmalloc_array(n: nr_pages, size: sizeof(struct page *), GFP_KERNEL); |
191 | if (!pages) { |
192 | ret = -ENOMEM; |
193 | goto out; |
194 | } |
195 | ret = pin_user_pages_fast(start, nr_pages, gup_flags: FOLL_WRITE, pages); |
196 | if (ret != nr_pages) { |
197 | if (ret > 0) |
198 | nr_pages = ret; |
199 | else |
200 | nr_pages = 0; |
201 | ret = -EAGAIN; /* XXX ? */ |
202 | goto out; |
203 | } |
204 | |
205 | rdsdebug("len %d nr_pages %lu\n" , len, nr_pages); |
206 | |
207 | call_func: |
208 | func = rds_info_funcs[optname - RDS_INFO_FIRST]; |
209 | if (!func) { |
210 | ret = -ENOPROTOOPT; |
211 | goto out; |
212 | } |
213 | |
214 | iter.pages = pages; |
215 | iter.addr = NULL; |
216 | iter.offset = start & (PAGE_SIZE - 1); |
217 | |
218 | func(sock, len, &iter, &lens); |
219 | BUG_ON(lens.each == 0); |
220 | |
221 | total = lens.nr * lens.each; |
222 | |
223 | rds_info_iter_unmap(iter: &iter); |
224 | |
225 | if (total > len) { |
226 | len = total; |
227 | ret = -ENOSPC; |
228 | } else { |
229 | len = total; |
230 | ret = lens.each; |
231 | } |
232 | |
233 | if (put_user(len, optlen)) |
234 | ret = -EFAULT; |
235 | |
236 | out: |
237 | if (pages) |
238 | unpin_user_pages(pages, npages: nr_pages); |
239 | kfree(objp: pages); |
240 | |
241 | return ret; |
242 | } |
243 | |