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->rtt = UINT_MAX;
25 vlserver->name_len = name_len;
26 vlserver->port = port;
27 memcpy(vlserver->name, name, name_len);
28 }
29 return vlserver;
30 }
31
afs_vlserver_rcu(struct rcu_head * rcu)32 static void afs_vlserver_rcu(struct rcu_head *rcu)
33 {
34 struct afs_vlserver *vlserver = container_of(rcu, struct afs_vlserver, rcu);
35
36 afs_put_addrlist(rcu_access_pointer(vlserver->addresses));
37 kfree_rcu(vlserver, rcu);
38 }
39
afs_put_vlserver(struct afs_net * net,struct afs_vlserver * vlserver)40 void afs_put_vlserver(struct afs_net *net, struct afs_vlserver *vlserver)
41 {
42 if (vlserver) {
43 unsigned int u = atomic_dec_return(&vlserver->usage);
44 //_debug("VL PUT %p{%u}", vlserver, u);
45
46 if (u == 0)
47 call_rcu(&vlserver->rcu, afs_vlserver_rcu);
48 }
49 }
50
afs_alloc_vlserver_list(unsigned int nr_servers)51 struct afs_vlserver_list *afs_alloc_vlserver_list(unsigned int nr_servers)
52 {
53 struct afs_vlserver_list *vllist;
54
55 vllist = kzalloc(struct_size(vllist, servers, nr_servers), GFP_KERNEL);
56 if (vllist) {
57 atomic_set(&vllist->usage, 1);
58 rwlock_init(&vllist->lock);
59 }
60
61 return vllist;
62 }
63
afs_put_vlserverlist(struct afs_net * net,struct afs_vlserver_list * vllist)64 void afs_put_vlserverlist(struct afs_net *net, struct afs_vlserver_list *vllist)
65 {
66 if (vllist) {
67 unsigned int u = atomic_dec_return(&vllist->usage);
68
69 //_debug("VLLS PUT %p{%u}", vllist, u);
70 if (u == 0) {
71 int i;
72
73 for (i = 0; i < vllist->nr_servers; i++) {
74 afs_put_vlserver(net, vllist->servers[i].server);
75 }
76 kfree_rcu(vllist, rcu);
77 }
78 }
79 }
80
afs_extract_le16(const u8 ** _b)81 static u16 afs_extract_le16(const u8 **_b)
82 {
83 u16 val;
84
85 val = (u16)*(*_b)++ << 0;
86 val |= (u16)*(*_b)++ << 8;
87 return val;
88 }
89
90 /*
91 * Build a VL server address list from a DNS queried server list.
92 */
afs_extract_vl_addrs(const u8 ** _b,const u8 * end,u8 nr_addrs,u16 port)93 static struct afs_addr_list *afs_extract_vl_addrs(const u8 **_b, const u8 *end,
94 u8 nr_addrs, u16 port)
95 {
96 struct afs_addr_list *alist;
97 const u8 *b = *_b;
98 int ret = -EINVAL;
99
100 alist = afs_alloc_addrlist(nr_addrs, VL_SERVICE, port);
101 if (!alist)
102 return ERR_PTR(-ENOMEM);
103 if (nr_addrs == 0)
104 return alist;
105
106 for (; nr_addrs > 0 && end - b >= nr_addrs; nr_addrs--) {
107 struct dns_server_list_v1_address hdr;
108 __be32 x[4];
109
110 hdr.address_type = *b++;
111
112 switch (hdr.address_type) {
113 case DNS_ADDRESS_IS_IPV4:
114 if (end - b < 4) {
115 _leave(" = -EINVAL [short inet]");
116 goto error;
117 }
118 memcpy(x, b, 4);
119 afs_merge_fs_addr4(alist, x[0], port);
120 b += 4;
121 break;
122
123 case DNS_ADDRESS_IS_IPV6:
124 if (end - b < 16) {
125 _leave(" = -EINVAL [short inet6]");
126 goto error;
127 }
128 memcpy(x, b, 16);
129 afs_merge_fs_addr6(alist, x, port);
130 b += 16;
131 break;
132
133 default:
134 _leave(" = -EADDRNOTAVAIL [unknown af %u]",
135 hdr.address_type);
136 ret = -EADDRNOTAVAIL;
137 goto error;
138 }
139 }
140
141 /* Start with IPv6 if available. */
142 if (alist->nr_ipv4 < alist->nr_addrs)
143 alist->preferred = alist->nr_ipv4;
144
145 *_b = b;
146 return alist;
147
148 error:
149 *_b = b;
150 afs_put_addrlist(alist);
151 return ERR_PTR(ret);
152 }
153
154 /*
155 * Build a VL server list from a DNS queried server list.
156 */
afs_extract_vlserver_list(struct afs_cell * cell,const void * buffer,size_t buffer_size)157 struct afs_vlserver_list *afs_extract_vlserver_list(struct afs_cell *cell,
158 const void *buffer,
159 size_t buffer_size)
160 {
161 const struct dns_server_list_v1_header *hdr = buffer;
162 struct dns_server_list_v1_server bs;
163 struct afs_vlserver_list *vllist, *previous;
164 struct afs_addr_list *addrs;
165 struct afs_vlserver *server;
166 const u8 *b = buffer, *end = buffer + buffer_size;
167 int ret = -ENOMEM, nr_servers, i, j;
168
169 _enter("");
170
171 /* Check that it's a server list, v1 */
172 if (end - b < sizeof(*hdr) ||
173 hdr->hdr.content != DNS_PAYLOAD_IS_SERVER_LIST ||
174 hdr->hdr.version != 1) {
175 pr_notice("kAFS: Got DNS record [%u,%u] len %zu\n",
176 hdr->hdr.content, hdr->hdr.version, end - b);
177 ret = -EDESTADDRREQ;
178 goto dump;
179 }
180
181 nr_servers = hdr->nr_servers;
182
183 vllist = afs_alloc_vlserver_list(nr_servers);
184 if (!vllist)
185 return ERR_PTR(-ENOMEM);
186
187 vllist->source = (hdr->source < NR__dns_record_source) ?
188 hdr->source : NR__dns_record_source;
189 vllist->status = (hdr->status < NR__dns_lookup_status) ?
190 hdr->status : NR__dns_lookup_status;
191
192 read_lock(&cell->vl_servers_lock);
193 previous = afs_get_vlserverlist(
194 rcu_dereference_protected(cell->vl_servers,
195 lockdep_is_held(&cell->vl_servers_lock)));
196 read_unlock(&cell->vl_servers_lock);
197
198 b += sizeof(*hdr);
199 while (end - b >= sizeof(bs)) {
200 bs.name_len = afs_extract_le16(&b);
201 bs.priority = afs_extract_le16(&b);
202 bs.weight = afs_extract_le16(&b);
203 bs.port = afs_extract_le16(&b);
204 bs.source = *b++;
205 bs.status = *b++;
206 bs.protocol = *b++;
207 bs.nr_addrs = *b++;
208
209 _debug("extract %u %u %u %u %u %u %*.*s",
210 bs.name_len, bs.priority, bs.weight,
211 bs.port, bs.protocol, bs.nr_addrs,
212 bs.name_len, bs.name_len, b);
213
214 if (end - b < bs.name_len)
215 break;
216
217 ret = -EPROTONOSUPPORT;
218 if (bs.protocol == DNS_SERVER_PROTOCOL_UNSPECIFIED) {
219 bs.protocol = DNS_SERVER_PROTOCOL_UDP;
220 } else if (bs.protocol != DNS_SERVER_PROTOCOL_UDP) {
221 _leave(" = [proto %u]", bs.protocol);
222 goto error;
223 }
224
225 if (bs.port == 0)
226 bs.port = AFS_VL_PORT;
227 if (bs.source > NR__dns_record_source)
228 bs.source = NR__dns_record_source;
229 if (bs.status > NR__dns_lookup_status)
230 bs.status = NR__dns_lookup_status;
231
232 /* See if we can update an old server record */
233 server = NULL;
234 for (i = 0; i < previous->nr_servers; i++) {
235 struct afs_vlserver *p = previous->servers[i].server;
236
237 if (p->name_len == bs.name_len &&
238 p->port == bs.port &&
239 strncasecmp(b, p->name, bs.name_len) == 0) {
240 server = afs_get_vlserver(p);
241 break;
242 }
243 }
244
245 if (!server) {
246 ret = -ENOMEM;
247 server = afs_alloc_vlserver(b, bs.name_len, bs.port);
248 if (!server)
249 goto error;
250 }
251
252 b += bs.name_len;
253
254 /* Extract the addresses - note that we can't skip this as we
255 * have to advance the payload pointer.
256 */
257 addrs = afs_extract_vl_addrs(&b, end, bs.nr_addrs, bs.port);
258 if (IS_ERR(addrs)) {
259 ret = PTR_ERR(addrs);
260 goto error_2;
261 }
262
263 if (vllist->nr_servers >= nr_servers) {
264 _debug("skip %u >= %u", vllist->nr_servers, nr_servers);
265 afs_put_addrlist(addrs);
266 afs_put_vlserver(cell->net, server);
267 continue;
268 }
269
270 addrs->source = bs.source;
271 addrs->status = bs.status;
272
273 if (addrs->nr_addrs == 0) {
274 afs_put_addrlist(addrs);
275 if (!rcu_access_pointer(server->addresses)) {
276 afs_put_vlserver(cell->net, server);
277 continue;
278 }
279 } else {
280 struct afs_addr_list *old = addrs;
281
282 write_lock(&server->lock);
283 old = rcu_replace_pointer(server->addresses, old,
284 lockdep_is_held(&server->lock));
285 write_unlock(&server->lock);
286 afs_put_addrlist(old);
287 }
288
289
290 /* TODO: Might want to check for duplicates */
291
292 /* Insertion-sort by priority and weight */
293 for (j = 0; j < vllist->nr_servers; j++) {
294 if (bs.priority < vllist->servers[j].priority)
295 break; /* Lower preferable */
296 if (bs.priority == vllist->servers[j].priority &&
297 bs.weight > vllist->servers[j].weight)
298 break; /* Higher preferable */
299 }
300
301 if (j < vllist->nr_servers) {
302 memmove(vllist->servers + j + 1,
303 vllist->servers + j,
304 (vllist->nr_servers - j) * sizeof(struct afs_vlserver_entry));
305 }
306
307 clear_bit(AFS_VLSERVER_FL_PROBED, &server->flags);
308
309 vllist->servers[j].priority = bs.priority;
310 vllist->servers[j].weight = bs.weight;
311 vllist->servers[j].server = server;
312 vllist->nr_servers++;
313 }
314
315 if (b != end) {
316 _debug("parse error %zd", b - end);
317 goto error;
318 }
319
320 afs_put_vlserverlist(cell->net, previous);
321 _leave(" = ok [%u]", vllist->nr_servers);
322 return vllist;
323
324 error_2:
325 afs_put_vlserver(cell->net, server);
326 error:
327 afs_put_vlserverlist(cell->net, vllist);
328 afs_put_vlserverlist(cell->net, previous);
329 dump:
330 if (ret != -ENOMEM) {
331 printk(KERN_DEBUG "DNS: at %zu\n", (const void *)b - buffer);
332 print_hex_dump_bytes("DNS: ", DUMP_PREFIX_NONE, buffer, buffer_size);
333 }
334 return ERR_PTR(ret);
335 }
336