1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* AFS vlserver list management.
3 *
4 * Copyright (C) 2018 Red Hat, Inc. All Rights Reserved.
5 * Written by David Howells (dhowells@redhat.com)
6 */
7
8 #include <linux/kernel.h>
9 #include <linux/slab.h>
10 #include "internal.h"
11
afs_alloc_vlserver(const char * name,size_t name_len,unsigned short port)12 struct afs_vlserver *afs_alloc_vlserver(const char *name, size_t name_len,
13 unsigned short port)
14 {
15 struct afs_vlserver *vlserver;
16
17 vlserver = kzalloc(struct_size(vlserver, name, name_len + 1),
18 GFP_KERNEL);
19 if (vlserver) {
20 atomic_set(&vlserver->usage, 1);
21 rwlock_init(&vlserver->lock);
22 init_waitqueue_head(&vlserver->probe_wq);
23 spin_lock_init(&vlserver->probe_lock);
24 vlserver->name_len = name_len;
25 vlserver->port = port;
26 memcpy(vlserver->name, name, name_len);
27 }
28 return vlserver;
29 }
30
afs_vlserver_rcu(struct rcu_head * rcu)31 static void afs_vlserver_rcu(struct rcu_head *rcu)
32 {
33 struct afs_vlserver *vlserver = container_of(rcu, struct afs_vlserver, rcu);
34
35 afs_put_addrlist(rcu_access_pointer(vlserver->addresses));
36 kfree_rcu(vlserver, rcu);
37 }
38
afs_put_vlserver(struct afs_net * net,struct afs_vlserver * vlserver)39 void afs_put_vlserver(struct afs_net *net, struct afs_vlserver *vlserver)
40 {
41 if (vlserver) {
42 unsigned int u = atomic_dec_return(&vlserver->usage);
43 //_debug("VL PUT %p{%u}", vlserver, u);
44
45 if (u == 0)
46 call_rcu(&vlserver->rcu, afs_vlserver_rcu);
47 }
48 }
49
afs_alloc_vlserver_list(unsigned int nr_servers)50 struct afs_vlserver_list *afs_alloc_vlserver_list(unsigned int nr_servers)
51 {
52 struct afs_vlserver_list *vllist;
53
54 vllist = kzalloc(struct_size(vllist, servers, nr_servers), GFP_KERNEL);
55 if (vllist) {
56 atomic_set(&vllist->usage, 1);
57 rwlock_init(&vllist->lock);
58 }
59
60 return vllist;
61 }
62
afs_put_vlserverlist(struct afs_net * net,struct afs_vlserver_list * vllist)63 void afs_put_vlserverlist(struct afs_net *net, struct afs_vlserver_list *vllist)
64 {
65 if (vllist) {
66 unsigned int u = atomic_dec_return(&vllist->usage);
67
68 //_debug("VLLS PUT %p{%u}", vllist, u);
69 if (u == 0) {
70 int i;
71
72 for (i = 0; i < vllist->nr_servers; i++) {
73 afs_put_vlserver(net, vllist->servers[i].server);
74 }
75 kfree_rcu(vllist, rcu);
76 }
77 }
78 }
79
afs_extract_le16(const u8 ** _b)80 static u16 afs_extract_le16(const u8 **_b)
81 {
82 u16 val;
83
84 val = (u16)*(*_b)++ << 0;
85 val |= (u16)*(*_b)++ << 8;
86 return val;
87 }
88
89 /*
90 * Build a VL server address list from a DNS queried server list.
91 */
afs_extract_vl_addrs(const u8 ** _b,const u8 * end,u8 nr_addrs,u16 port)92 static struct afs_addr_list *afs_extract_vl_addrs(const u8 **_b, const u8 *end,
93 u8 nr_addrs, u16 port)
94 {
95 struct afs_addr_list *alist;
96 const u8 *b = *_b;
97 int ret = -EINVAL;
98
99 alist = afs_alloc_addrlist(nr_addrs, VL_SERVICE, port);
100 if (!alist)
101 return ERR_PTR(-ENOMEM);
102 if (nr_addrs == 0)
103 return alist;
104
105 for (; nr_addrs > 0 && end - b >= nr_addrs; nr_addrs--) {
106 struct dns_server_list_v1_address hdr;
107 __be32 x[4];
108
109 hdr.address_type = *b++;
110
111 switch (hdr.address_type) {
112 case DNS_ADDRESS_IS_IPV4:
113 if (end - b < 4) {
114 _leave(" = -EINVAL [short inet]");
115 goto error;
116 }
117 memcpy(x, b, 4);
118 afs_merge_fs_addr4(alist, x[0], port);
119 b += 4;
120 break;
121
122 case DNS_ADDRESS_IS_IPV6:
123 if (end - b < 16) {
124 _leave(" = -EINVAL [short inet6]");
125 goto error;
126 }
127 memcpy(x, b, 16);
128 afs_merge_fs_addr6(alist, x, port);
129 b += 16;
130 break;
131
132 default:
133 _leave(" = -EADDRNOTAVAIL [unknown af %u]",
134 hdr.address_type);
135 ret = -EADDRNOTAVAIL;
136 goto error;
137 }
138 }
139
140 /* Start with IPv6 if available. */
141 if (alist->nr_ipv4 < alist->nr_addrs)
142 alist->preferred = alist->nr_ipv4;
143
144 *_b = b;
145 return alist;
146
147 error:
148 *_b = b;
149 afs_put_addrlist(alist);
150 return ERR_PTR(ret);
151 }
152
153 /*
154 * Build a VL server list from a DNS queried server list.
155 */
afs_extract_vlserver_list(struct afs_cell * cell,const void * buffer,size_t buffer_size)156 struct afs_vlserver_list *afs_extract_vlserver_list(struct afs_cell *cell,
157 const void *buffer,
158 size_t buffer_size)
159 {
160 const struct dns_server_list_v1_header *hdr = buffer;
161 struct dns_server_list_v1_server bs;
162 struct afs_vlserver_list *vllist, *previous;
163 struct afs_addr_list *addrs;
164 struct afs_vlserver *server;
165 const u8 *b = buffer, *end = buffer + buffer_size;
166 int ret = -ENOMEM, nr_servers, i, j;
167
168 _enter("");
169
170 /* Check that it's a server list, v1 */
171 if (end - b < sizeof(*hdr) ||
172 hdr->hdr.content != DNS_PAYLOAD_IS_SERVER_LIST ||
173 hdr->hdr.version != 1) {
174 pr_notice("kAFS: Got DNS record [%u,%u] len %zu\n",
175 hdr->hdr.content, hdr->hdr.version, end - b);
176 ret = -EDESTADDRREQ;
177 goto dump;
178 }
179
180 nr_servers = hdr->nr_servers;
181
182 vllist = afs_alloc_vlserver_list(nr_servers);
183 if (!vllist)
184 return ERR_PTR(-ENOMEM);
185
186 vllist->source = (hdr->source < NR__dns_record_source) ?
187 hdr->source : NR__dns_record_source;
188 vllist->status = (hdr->status < NR__dns_lookup_status) ?
189 hdr->status : NR__dns_lookup_status;
190
191 read_lock(&cell->vl_servers_lock);
192 previous = afs_get_vlserverlist(
193 rcu_dereference_protected(cell->vl_servers,
194 lockdep_is_held(&cell->vl_servers_lock)));
195 read_unlock(&cell->vl_servers_lock);
196
197 b += sizeof(*hdr);
198 while (end - b >= sizeof(bs)) {
199 bs.name_len = afs_extract_le16(&b);
200 bs.priority = afs_extract_le16(&b);
201 bs.weight = afs_extract_le16(&b);
202 bs.port = afs_extract_le16(&b);
203 bs.source = *b++;
204 bs.status = *b++;
205 bs.protocol = *b++;
206 bs.nr_addrs = *b++;
207
208 _debug("extract %u %u %u %u %u %u %*.*s",
209 bs.name_len, bs.priority, bs.weight,
210 bs.port, bs.protocol, bs.nr_addrs,
211 bs.name_len, bs.name_len, b);
212
213 if (end - b < bs.name_len)
214 break;
215
216 ret = -EPROTONOSUPPORT;
217 if (bs.protocol == DNS_SERVER_PROTOCOL_UNSPECIFIED) {
218 bs.protocol = DNS_SERVER_PROTOCOL_UDP;
219 } else if (bs.protocol != DNS_SERVER_PROTOCOL_UDP) {
220 _leave(" = [proto %u]", bs.protocol);
221 goto error;
222 }
223
224 if (bs.port == 0)
225 bs.port = AFS_VL_PORT;
226 if (bs.source > NR__dns_record_source)
227 bs.source = NR__dns_record_source;
228 if (bs.status > NR__dns_lookup_status)
229 bs.status = NR__dns_lookup_status;
230
231 /* See if we can update an old server record */
232 server = NULL;
233 for (i = 0; i < previous->nr_servers; i++) {
234 struct afs_vlserver *p = previous->servers[i].server;
235
236 if (p->name_len == bs.name_len &&
237 p->port == bs.port &&
238 strncasecmp(b, p->name, bs.name_len) == 0) {
239 server = afs_get_vlserver(p);
240 break;
241 }
242 }
243
244 if (!server) {
245 ret = -ENOMEM;
246 server = afs_alloc_vlserver(b, bs.name_len, bs.port);
247 if (!server)
248 goto error;
249 }
250
251 b += bs.name_len;
252
253 /* Extract the addresses - note that we can't skip this as we
254 * have to advance the payload pointer.
255 */
256 addrs = afs_extract_vl_addrs(&b, end, bs.nr_addrs, bs.port);
257 if (IS_ERR(addrs)) {
258 ret = PTR_ERR(addrs);
259 goto error_2;
260 }
261
262 if (vllist->nr_servers >= nr_servers) {
263 _debug("skip %u >= %u", vllist->nr_servers, nr_servers);
264 afs_put_addrlist(addrs);
265 afs_put_vlserver(cell->net, server);
266 continue;
267 }
268
269 addrs->source = bs.source;
270 addrs->status = bs.status;
271
272 if (addrs->nr_addrs == 0) {
273 afs_put_addrlist(addrs);
274 if (!rcu_access_pointer(server->addresses)) {
275 afs_put_vlserver(cell->net, server);
276 continue;
277 }
278 } else {
279 struct afs_addr_list *old = addrs;
280
281 write_lock(&server->lock);
282 rcu_swap_protected(server->addresses, old,
283 lockdep_is_held(&server->lock));
284 write_unlock(&server->lock);
285 afs_put_addrlist(old);
286 }
287
288
289 /* TODO: Might want to check for duplicates */
290
291 /* Insertion-sort by priority and weight */
292 for (j = 0; j < vllist->nr_servers; j++) {
293 if (bs.priority < vllist->servers[j].priority)
294 break; /* Lower preferable */
295 if (bs.priority == vllist->servers[j].priority &&
296 bs.weight > vllist->servers[j].weight)
297 break; /* Higher preferable */
298 }
299
300 if (j < vllist->nr_servers) {
301 memmove(vllist->servers + j + 1,
302 vllist->servers + j,
303 (vllist->nr_servers - j) * sizeof(struct afs_vlserver_entry));
304 }
305
306 clear_bit(AFS_VLSERVER_FL_PROBED, &server->flags);
307
308 vllist->servers[j].priority = bs.priority;
309 vllist->servers[j].weight = bs.weight;
310 vllist->servers[j].server = server;
311 vllist->nr_servers++;
312 }
313
314 if (b != end) {
315 _debug("parse error %zd", b - end);
316 goto error;
317 }
318
319 afs_put_vlserverlist(cell->net, previous);
320 _leave(" = ok [%u]", vllist->nr_servers);
321 return vllist;
322
323 error_2:
324 afs_put_vlserver(cell->net, server);
325 error:
326 afs_put_vlserverlist(cell->net, vllist);
327 afs_put_vlserverlist(cell->net, previous);
328 dump:
329 if (ret != -ENOMEM) {
330 printk(KERN_DEBUG "DNS: at %zu\n", (const void *)b - buffer);
331 print_hex_dump_bytes("DNS: ", DUMP_PREFIX_NONE, buffer, buffer_size);
332 }
333 return ERR_PTR(ret);
334 }
335
