afs: Remove afs_vlserver->probe.have_result
[linux-2.6-block.git] / fs / afs / vl_rotate.c
CommitLineData
b4d0d230 1// SPDX-License-Identifier: GPL-2.0-or-later
0a5143f2
DH
2/* Handle vlserver selection and rotation.
3 *
4 * Copyright (C) 2018 Red Hat, Inc. All Rights Reserved.
5 * Written by David Howells (dhowells@redhat.com)
0a5143f2
DH
6 */
7
8#include <linux/kernel.h>
9#include <linux/sched.h>
10#include <linux/sched/signal.h>
11#include "internal.h"
12#include "afs_vl.h"
13
14/*
15 * Begin an operation on a volume location server.
16 */
17bool afs_begin_vlserver_operation(struct afs_vl_cursor *vc, struct afs_cell *cell,
18 struct key *key)
19{
20 memset(vc, 0, sizeof(*vc));
21 vc->cell = cell;
22 vc->key = key;
23 vc->error = -EDESTADDRREQ;
24 vc->ac.error = SHRT_MAX;
25
26 if (signal_pending(current)) {
27 vc->error = -EINTR;
28 vc->flags |= AFS_VL_CURSOR_STOP;
29 return false;
30 }
31
32 return true;
33}
34
35/*
36 * Begin iteration through a server list, starting with the last used server if
37 * possible, or the last recorded good server if not.
38 */
39static bool afs_start_vl_iteration(struct afs_vl_cursor *vc)
40{
41 struct afs_cell *cell = vc->cell;
d5c32c89 42 unsigned int dns_lookup_count;
0a5143f2 43
d5c32c89
DH
44 if (cell->dns_source == DNS_RECORD_UNAVAILABLE ||
45 cell->dns_expiry <= ktime_get_real_seconds()) {
46 dns_lookup_count = smp_load_acquire(&cell->dns_lookup_count);
47 set_bit(AFS_CELL_FL_DO_LOOKUP, &cell->flags);
48 queue_work(afs_wq, &cell->manager);
49
50 if (cell->dns_source == DNS_RECORD_UNAVAILABLE) {
51 if (wait_var_event_interruptible(
52 &cell->dns_lookup_count,
53 smp_load_acquire(&cell->dns_lookup_count)
54 != dns_lookup_count) < 0) {
55 vc->error = -ERESTARTSYS;
56 return false;
57 }
58 }
59
60 /* Status load is ordered after lookup counter load */
61 if (cell->dns_source == DNS_RECORD_UNAVAILABLE) {
62 vc->error = -EDESTADDRREQ;
63 return false;
64 }
0a5143f2
DH
65 }
66
67 read_lock(&cell->vl_servers_lock);
68 vc->server_list = afs_get_vlserverlist(
69 rcu_dereference_protected(cell->vl_servers,
70 lockdep_is_held(&cell->vl_servers_lock)));
71 read_unlock(&cell->vl_servers_lock);
ca1cbbdc 72 if (!vc->server_list->nr_servers)
0a5143f2
DH
73 return false;
74
3bf0fb6f
DH
75 vc->untried = (1UL << vc->server_list->nr_servers) - 1;
76 vc->index = -1;
0a5143f2
DH
77 return true;
78}
79
80/*
81 * Select the vlserver to use. May be called multiple times to rotate
82 * through the vlservers.
83 */
84bool afs_select_vlserver(struct afs_vl_cursor *vc)
85{
86 struct afs_addr_list *alist;
87 struct afs_vlserver *vlserver;
4584ae96 88 struct afs_error e;
3bf0fb6f 89 u32 rtt;
4584ae96 90 int error = vc->ac.error, i;
0a5143f2 91
3bf0fb6f
DH
92 _enter("%lx[%d],%lx[%d],%d,%d",
93 vc->untried, vc->index,
94 vc->ac.tried, vc->ac.index,
0a5143f2
DH
95 error, vc->ac.abort_code);
96
97 if (vc->flags & AFS_VL_CURSOR_STOP) {
98 _leave(" = f [stopped]");
99 return false;
100 }
101
744bcd71
DH
102 vc->nr_iterations++;
103
0a5143f2
DH
104 /* Evaluate the result of the previous operation, if there was one. */
105 switch (error) {
106 case SHRT_MAX:
107 goto start;
108
109 default:
110 case 0:
111 /* Success or local failure. Stop. */
112 vc->error = error;
113 vc->flags |= AFS_VL_CURSOR_STOP;
114 _leave(" = f [okay/local %d]", vc->ac.error);
115 return false;
116
117 case -ECONNABORTED:
118 /* The far side rejected the operation on some grounds. This
119 * might involve the server being busy or the volume having been moved.
120 */
121 switch (vc->ac.abort_code) {
122 case AFSVL_IO:
123 case AFSVL_BADVOLOPER:
124 case AFSVL_NOMEM:
125 /* The server went weird. */
126 vc->error = -EREMOTEIO;
127 //write_lock(&vc->cell->vl_servers_lock);
128 //vc->server_list->weird_mask |= 1 << vc->index;
129 //write_unlock(&vc->cell->vl_servers_lock);
130 goto next_server;
131
132 default:
133 vc->error = afs_abort_to_error(vc->ac.abort_code);
134 goto failed;
135 }
136
4584ae96
DH
137 case -ERFKILL:
138 case -EADDRNOTAVAIL:
0a5143f2
DH
139 case -ENETUNREACH:
140 case -EHOSTUNREACH:
4584ae96 141 case -EHOSTDOWN:
0a5143f2
DH
142 case -ECONNREFUSED:
143 case -ETIMEDOUT:
144 case -ETIME:
145 _debug("no conn %d", error);
146 vc->error = error;
147 goto iterate_address;
148
149 case -ECONNRESET:
150 _debug("call reset");
151 vc->error = error;
152 vc->flags |= AFS_VL_CURSOR_RETRY;
153 goto next_server;
6dfdf536
DH
154
155 case -EOPNOTSUPP:
156 _debug("notsupp");
157 goto next_server;
0a5143f2
DH
158 }
159
160restart_from_beginning:
161 _debug("restart");
162 afs_end_cursor(&vc->ac);
163 afs_put_vlserverlist(vc->cell->net, vc->server_list);
164 vc->server_list = NULL;
165 if (vc->flags & AFS_VL_CURSOR_RETRIED)
166 goto failed;
167 vc->flags |= AFS_VL_CURSOR_RETRIED;
168start:
169 _debug("start");
170
0a5143f2
DH
171 if (!afs_start_vl_iteration(vc))
172 goto failed;
173
3bf0fb6f
DH
174 error = afs_send_vl_probes(vc->cell->net, vc->key, vc->server_list);
175 if (error < 0)
176 goto failed_set_error;
177
178pick_server:
179 _debug("pick [%lx]", vc->untried);
180
181 error = afs_wait_for_vl_probes(vc->server_list, vc->untried);
182 if (error < 0)
183 goto failed_set_error;
184
185 /* Pick the untried server with the lowest RTT. */
186 vc->index = vc->server_list->preferred;
187 if (test_bit(vc->index, &vc->untried))
188 goto selected_server;
189
190 vc->index = -1;
191 rtt = U32_MAX;
192 for (i = 0; i < vc->server_list->nr_servers; i++) {
193 struct afs_vlserver *s = vc->server_list->servers[i].server;
194
195 if (!test_bit(i, &vc->untried) || !s->probe.responded)
196 continue;
197 if (s->probe.rtt < rtt) {
198 vc->index = i;
199 rtt = s->probe.rtt;
200 }
201 }
202
203 if (vc->index == -1)
204 goto no_more_servers;
205
206selected_server:
207 _debug("use %d", vc->index);
208 __clear_bit(vc->index, &vc->untried);
209
0a5143f2
DH
210 /* We're starting on a different vlserver from the list. We need to
211 * check it, find its address list and probe its capabilities before we
212 * use it.
213 */
214 ASSERTCMP(vc->ac.alist, ==, NULL);
215 vlserver = vc->server_list->servers[vc->index].server;
3bf0fb6f 216 vc->server = vlserver;
0a5143f2
DH
217
218 _debug("USING VLSERVER: %s", vlserver->name);
219
220 read_lock(&vlserver->lock);
221 alist = rcu_dereference_protected(vlserver->addresses,
222 lockdep_is_held(&vlserver->lock));
223 afs_get_addrlist(alist);
224 read_unlock(&vlserver->lock);
225
226 memset(&vc->ac, 0, sizeof(vc->ac));
227
0a5143f2
DH
228 if (!vc->ac.alist)
229 vc->ac.alist = alist;
230 else
231 afs_put_addrlist(alist);
232
3bf0fb6f 233 vc->ac.index = -1;
0a5143f2
DH
234
235iterate_address:
236 ASSERT(vc->ac.alist);
0a5143f2
DH
237 /* Iterate over the current server's address list to try and find an
238 * address on which it will respond to us.
239 */
240 if (!afs_iterate_addresses(&vc->ac))
241 goto next_server;
242
3bf0fb6f
DH
243 _debug("VL address %d/%d", vc->ac.index, vc->ac.alist->nr_addrs);
244
2feeaf84 245 _leave(" = t %pISpc", &vc->ac.alist->addrs[vc->ac.index].transport);
0a5143f2
DH
246 return true;
247
248next_server:
249 _debug("next");
250 afs_end_cursor(&vc->ac);
3bf0fb6f 251 goto pick_server;
0a5143f2 252
3bf0fb6f 253no_more_servers:
0a5143f2
DH
254 /* That's all the servers poked to no good effect. Try again if some
255 * of them were busy.
256 */
257 if (vc->flags & AFS_VL_CURSOR_RETRY)
258 goto restart_from_beginning;
259
4584ae96
DH
260 e.error = -EDESTADDRREQ;
261 e.responded = false;
3bf0fb6f
DH
262 for (i = 0; i < vc->server_list->nr_servers; i++) {
263 struct afs_vlserver *s = vc->server_list->servers[i].server;
3bf0fb6f 264
4584ae96
DH
265 afs_prioritise_error(&e, READ_ONCE(s->probe.error),
266 s->probe.abort_code);
3bf0fb6f
DH
267 }
268
3bf0fb6f
DH
269failed_set_error:
270 vc->error = error;
0a5143f2
DH
271failed:
272 vc->flags |= AFS_VL_CURSOR_STOP;
273 afs_end_cursor(&vc->ac);
274 _leave(" = f [failed %d]", vc->error);
275 return false;
276}
277
744bcd71
DH
278/*
279 * Dump cursor state in the case of the error being EDESTADDRREQ.
280 */
281static void afs_vl_dump_edestaddrreq(const struct afs_vl_cursor *vc)
282{
283 static int count;
284 int i;
285
286 if (!IS_ENABLED(CONFIG_AFS_DEBUG_CURSOR) || count > 3)
287 return;
288 count++;
289
290 rcu_read_lock();
291 pr_notice("EDESTADDR occurred\n");
3bf0fb6f
DH
292 pr_notice("VC: ut=%lx ix=%u ni=%hu fl=%hx err=%hd\n",
293 vc->untried, vc->index, vc->nr_iterations, vc->flags, vc->error);
744bcd71
DH
294
295 if (vc->server_list) {
296 const struct afs_vlserver_list *sl = vc->server_list;
297 pr_notice("VC: SL nr=%u ix=%u\n",
298 sl->nr_servers, sl->index);
299 for (i = 0; i < sl->nr_servers; i++) {
300 const struct afs_vlserver *s = sl->servers[i].server;
3bf0fb6f
DH
301 pr_notice("VC: server %s+%hu fl=%lx E=%hd\n",
302 s->name, s->port, s->flags, s->probe.error);
744bcd71
DH
303 if (s->addresses) {
304 const struct afs_addr_list *a =
305 rcu_dereference(s->addresses);
3bf0fb6f 306 pr_notice("VC: - nr=%u/%u/%u pf=%u\n",
744bcd71 307 a->nr_ipv4, a->nr_addrs, a->max_addrs,
3bf0fb6f 308 a->preferred);
be59167c
DH
309 pr_notice("VC: - R=%lx F=%lx\n",
310 a->responded, a->failed);
744bcd71
DH
311 if (a == vc->ac.alist)
312 pr_notice("VC: - current\n");
313 }
314 }
315 }
316
3bf0fb6f
DH
317 pr_notice("AC: t=%lx ax=%u ac=%d er=%d r=%u ni=%u\n",
318 vc->ac.tried, vc->ac.index, vc->ac.abort_code, vc->ac.error,
319 vc->ac.responded, vc->ac.nr_iterations);
744bcd71
DH
320 rcu_read_unlock();
321}
322
0a5143f2
DH
323/*
324 * Tidy up a volume location server cursor and unlock the vnode.
325 */
326int afs_end_vlserver_operation(struct afs_vl_cursor *vc)
327{
328 struct afs_net *net = vc->cell->net;
329
744bcd71 330 if (vc->error == -EDESTADDRREQ ||
4584ae96 331 vc->error == -EADDRNOTAVAIL ||
744bcd71
DH
332 vc->error == -ENETUNREACH ||
333 vc->error == -EHOSTUNREACH)
334 afs_vl_dump_edestaddrreq(vc);
335
0a5143f2
DH
336 afs_end_cursor(&vc->ac);
337 afs_put_vlserverlist(net, vc->server_list);
338
339 if (vc->error == -ECONNABORTED)
340 vc->error = afs_abort_to_error(vc->ac.abort_code);
341
342 return vc->error;
343}