Commit | Line | Data |
---|---|---|
16b374ca AA |
1 | /* |
2 | * Device operations for the pnfs nfs4 file layout driver. | |
3 | * | |
4 | * Copyright (c) 2002 | |
5 | * The Regents of the University of Michigan | |
6 | * All Rights Reserved | |
7 | * | |
8 | * Dean Hildebrand <dhildebz@umich.edu> | |
9 | * Garth Goodson <Garth.Goodson@netapp.com> | |
10 | * | |
11 | * Permission is granted to use, copy, create derivative works, and | |
12 | * redistribute this software and such derivative works for any purpose, | |
13 | * so long as the name of the University of Michigan is not used in | |
14 | * any advertising or publicity pertaining to the use or distribution | |
15 | * of this software without specific, written prior authorization. If | |
16 | * the above copyright notice or any other identification of the | |
17 | * University of Michigan is included in any copy of any portion of | |
18 | * this software, then the disclaimer below must also be included. | |
19 | * | |
20 | * This software is provided as is, without representation or warranty | |
21 | * of any kind either express or implied, including without limitation | |
22 | * the implied warranties of merchantability, fitness for a particular | |
23 | * purpose, or noninfringement. The Regents of the University of | |
24 | * Michigan shall not be liable for any damages, including special, | |
25 | * indirect, incidental, or consequential damages, with respect to any | |
26 | * claim arising out of or in connection with the use of the software, | |
27 | * even if it has been or is hereafter advised of the possibility of | |
28 | * such damages. | |
29 | */ | |
30 | ||
31 | #include <linux/nfs_fs.h> | |
32 | #include <linux/vmalloc.h> | |
98fc685a | 33 | #include <linux/module.h> |
16b374ca AA |
34 | |
35 | #include "internal.h" | |
73e39aaa | 36 | #include "nfs4session.h" |
16b374ca AA |
37 | #include "nfs4filelayout.h" |
38 | ||
39 | #define NFSDBG_FACILITY NFSDBG_PNFS_LD | |
40 | ||
98fc685a AA |
41 | static unsigned int dataserver_timeo = NFS4_DEF_DS_TIMEO; |
42 | static unsigned int dataserver_retrans = NFS4_DEF_DS_RETRANS; | |
43 | ||
16b374ca AA |
44 | /* |
45 | * Data server cache | |
46 | * | |
47 | * Data servers can be mapped to different device ids. | |
48 | * nfs4_pnfs_ds reference counting | |
49 | * - set to 1 on allocation | |
50 | * - incremented when a device id maps a data server already in the cache. | |
51 | * - decremented when deviceid is removed from the cache. | |
52 | */ | |
17280175 | 53 | static DEFINE_SPINLOCK(nfs4_ds_cache_lock); |
16b374ca AA |
54 | static LIST_HEAD(nfs4_data_server_cache); |
55 | ||
56 | /* Debug routines */ | |
57 | void | |
58 | print_ds(struct nfs4_pnfs_ds *ds) | |
59 | { | |
60 | if (ds == NULL) { | |
61 | printk("%s NULL device\n", __func__); | |
62 | return; | |
63 | } | |
c9895cb6 | 64 | printk(" ds %s\n" |
16b374ca AA |
65 | " ref count %d\n" |
66 | " client %p\n" | |
67 | " cl_exchange_flags %x\n", | |
c9895cb6 | 68 | ds->ds_remotestr, |
16b374ca AA |
69 | atomic_read(&ds->ds_count), ds->ds_clp, |
70 | ds->ds_clp ? ds->ds_clp->cl_exchange_flags : 0); | |
71 | } | |
72 | ||
14f9a607 WAA |
73 | static bool |
74 | same_sockaddr(struct sockaddr *addr1, struct sockaddr *addr2) | |
16b374ca | 75 | { |
c9895cb6 WAA |
76 | struct sockaddr_in *a, *b; |
77 | struct sockaddr_in6 *a6, *b6; | |
16b374ca | 78 | |
14f9a607 WAA |
79 | if (addr1->sa_family != addr2->sa_family) |
80 | return false; | |
81 | ||
82 | switch (addr1->sa_family) { | |
83 | case AF_INET: | |
84 | a = (struct sockaddr_in *)addr1; | |
85 | b = (struct sockaddr_in *)addr2; | |
86 | ||
87 | if (a->sin_addr.s_addr == b->sin_addr.s_addr && | |
88 | a->sin_port == b->sin_port) | |
89 | return true; | |
90 | break; | |
91 | ||
92 | case AF_INET6: | |
93 | a6 = (struct sockaddr_in6 *)addr1; | |
94 | b6 = (struct sockaddr_in6 *)addr2; | |
95 | ||
96 | /* LINKLOCAL addresses must have matching scope_id */ | |
97 | if (ipv6_addr_scope(&a6->sin6_addr) == | |
98 | IPV6_ADDR_SCOPE_LINKLOCAL && | |
99 | a6->sin6_scope_id != b6->sin6_scope_id) | |
100 | return false; | |
101 | ||
102 | if (ipv6_addr_equal(&a6->sin6_addr, &b6->sin6_addr) && | |
103 | a6->sin6_port == b6->sin6_port) | |
104 | return true; | |
105 | break; | |
106 | ||
107 | default: | |
108 | dprintk("%s: unhandled address family: %u\n", | |
109 | __func__, addr1->sa_family); | |
110 | return false; | |
111 | } | |
112 | ||
113 | return false; | |
114 | } | |
115 | ||
17280175 | 116 | static bool |
2d3fe01c WAA |
117 | _same_data_server_addrs_locked(const struct list_head *dsaddrs1, |
118 | const struct list_head *dsaddrs2) | |
14f9a607 | 119 | { |
14f9a607 WAA |
120 | struct nfs4_pnfs_ds_addr *da1, *da2; |
121 | ||
2d3fe01c WAA |
122 | /* step through both lists, comparing as we go */ |
123 | for (da1 = list_first_entry(dsaddrs1, typeof(*da1), da_node), | |
124 | da2 = list_first_entry(dsaddrs2, typeof(*da2), da_node); | |
125 | da1 != NULL && da2 != NULL; | |
126 | da1 = list_entry(da1->da_node.next, typeof(*da1), da_node), | |
127 | da2 = list_entry(da2->da_node.next, typeof(*da2), da_node)) { | |
128 | if (!same_sockaddr((struct sockaddr *)&da1->da_addr, | |
129 | (struct sockaddr *)&da2->da_addr)) | |
130 | return false; | |
16b374ca | 131 | } |
2d3fe01c WAA |
132 | if (da1 == NULL && da2 == NULL) |
133 | return true; | |
134 | ||
135 | return false; | |
16b374ca AA |
136 | } |
137 | ||
14f9a607 | 138 | /* |
2d3fe01c | 139 | * Lookup DS by addresses. nfs4_ds_cache_lock is held |
14f9a607 | 140 | */ |
2d3fe01c WAA |
141 | static struct nfs4_pnfs_ds * |
142 | _data_server_lookup_locked(const struct list_head *dsaddrs) | |
14f9a607 | 143 | { |
2d3fe01c | 144 | struct nfs4_pnfs_ds *ds; |
14f9a607 | 145 | |
2d3fe01c WAA |
146 | list_for_each_entry(ds, &nfs4_data_server_cache, ds_node) |
147 | if (_same_data_server_addrs_locked(&ds->ds_addrs, dsaddrs)) | |
148 | return ds; | |
149 | return NULL; | |
14f9a607 WAA |
150 | } |
151 | ||
d83217c1 AA |
152 | /* |
153 | * Create an rpc connection to the nfs4_pnfs_ds data server | |
35dbbc99 | 154 | * Currently only supports IPv4 and IPv6 addresses |
d83217c1 AA |
155 | */ |
156 | static int | |
157 | nfs4_ds_connect(struct nfs_server *mds_srv, struct nfs4_pnfs_ds *ds) | |
158 | { | |
7e574f0d | 159 | struct nfs_client *clp = ERR_PTR(-EIO); |
14f9a607 | 160 | struct nfs4_pnfs_ds_addr *da; |
d83217c1 AA |
161 | int status = 0; |
162 | ||
14f9a607 | 163 | dprintk("--> %s DS %s au_flavor %d\n", __func__, ds->ds_remotestr, |
d83217c1 AA |
164 | mds_srv->nfs_client->cl_rpcclient->cl_auth->au_flavor); |
165 | ||
7e574f0d WAA |
166 | list_for_each_entry(da, &ds->ds_addrs, da_node) { |
167 | dprintk("%s: DS %s: trying address %s\n", | |
168 | __func__, ds->ds_remotestr, da->da_remotestr); | |
14f9a607 | 169 | |
7e574f0d | 170 | clp = nfs4_set_ds_client(mds_srv->nfs_client, |
98fc685a AA |
171 | (struct sockaddr *)&da->da_addr, |
172 | da->da_addrlen, IPPROTO_TCP, | |
173 | dataserver_timeo, dataserver_retrans); | |
7e574f0d WAA |
174 | if (!IS_ERR(clp)) |
175 | break; | |
176 | } | |
177 | ||
d83217c1 AA |
178 | if (IS_ERR(clp)) { |
179 | status = PTR_ERR(clp); | |
180 | goto out; | |
181 | } | |
182 | ||
7b38c368 | 183 | status = nfs4_init_ds_session(clp, mds_srv->nfs_client->cl_lease_time); |
d83217c1 AA |
184 | if (status) |
185 | goto out_put; | |
186 | ||
187 | ds->ds_clp = clp; | |
c9895cb6 | 188 | dprintk("%s [new] addr: %s\n", __func__, ds->ds_remotestr); |
d83217c1 AA |
189 | out: |
190 | return status; | |
191 | out_put: | |
192 | nfs_put_client(clp); | |
193 | goto out; | |
194 | } | |
195 | ||
16b374ca AA |
196 | static void |
197 | destroy_ds(struct nfs4_pnfs_ds *ds) | |
198 | { | |
14f9a607 WAA |
199 | struct nfs4_pnfs_ds_addr *da; |
200 | ||
16b374ca AA |
201 | dprintk("--> %s\n", __func__); |
202 | ifdebug(FACILITY) | |
203 | print_ds(ds); | |
204 | ||
205 | if (ds->ds_clp) | |
206 | nfs_put_client(ds->ds_clp); | |
14f9a607 WAA |
207 | |
208 | while (!list_empty(&ds->ds_addrs)) { | |
209 | da = list_first_entry(&ds->ds_addrs, | |
210 | struct nfs4_pnfs_ds_addr, | |
211 | da_node); | |
212 | list_del_init(&da->da_node); | |
213 | kfree(da->da_remotestr); | |
214 | kfree(da); | |
215 | } | |
216 | ||
c9895cb6 | 217 | kfree(ds->ds_remotestr); |
16b374ca AA |
218 | kfree(ds); |
219 | } | |
220 | ||
1775bc34 | 221 | void |
16b374ca AA |
222 | nfs4_fl_free_deviceid(struct nfs4_file_layout_dsaddr *dsaddr) |
223 | { | |
224 | struct nfs4_pnfs_ds *ds; | |
225 | int i; | |
226 | ||
a1eaecbc | 227 | nfs4_print_deviceid(&dsaddr->id_node.deviceid); |
16b374ca AA |
228 | |
229 | for (i = 0; i < dsaddr->ds_num; i++) { | |
230 | ds = dsaddr->ds_list[i]; | |
231 | if (ds != NULL) { | |
232 | if (atomic_dec_and_lock(&ds->ds_count, | |
233 | &nfs4_ds_cache_lock)) { | |
234 | list_del_init(&ds->ds_node); | |
235 | spin_unlock(&nfs4_ds_cache_lock); | |
236 | destroy_ds(ds); | |
237 | } | |
238 | } | |
239 | } | |
240 | kfree(dsaddr->stripe_indices); | |
241 | kfree(dsaddr); | |
242 | } | |
243 | ||
c9895cb6 WAA |
244 | /* |
245 | * Create a string with a human readable address and port to avoid | |
246 | * complicated setup around many dprinks. | |
247 | */ | |
248 | static char * | |
14f9a607 | 249 | nfs4_pnfs_remotestr(struct list_head *dsaddrs, gfp_t gfp_flags) |
c9895cb6 | 250 | { |
14f9a607 | 251 | struct nfs4_pnfs_ds_addr *da; |
c9895cb6 | 252 | char *remotestr; |
c9895cb6 | 253 | size_t len; |
14f9a607 | 254 | char *p; |
c9895cb6 | 255 | |
14f9a607 WAA |
256 | len = 3; /* '{', '}' and eol */ |
257 | list_for_each_entry(da, dsaddrs, da_node) { | |
258 | len += strlen(da->da_remotestr) + 1; /* string plus comma */ | |
c9895cb6 WAA |
259 | } |
260 | ||
14f9a607 WAA |
261 | remotestr = kzalloc(len, gfp_flags); |
262 | if (!remotestr) | |
c9895cb6 | 263 | return NULL; |
c9895cb6 | 264 | |
14f9a607 WAA |
265 | p = remotestr; |
266 | *(p++) = '{'; | |
267 | len--; | |
268 | list_for_each_entry(da, dsaddrs, da_node) { | |
269 | size_t ll = strlen(da->da_remotestr); | |
c9895cb6 | 270 | |
14f9a607 WAA |
271 | if (ll > len) |
272 | goto out_err; | |
c9895cb6 | 273 | |
14f9a607 WAA |
274 | memcpy(p, da->da_remotestr, ll); |
275 | p += ll; | |
276 | len -= ll; | |
c9895cb6 | 277 | |
14f9a607 WAA |
278 | if (len < 1) |
279 | goto out_err; | |
280 | (*p++) = ','; | |
281 | len--; | |
282 | } | |
283 | if (len < 2) | |
284 | goto out_err; | |
285 | *(p++) = '}'; | |
286 | *p = '\0'; | |
c9895cb6 | 287 | return remotestr; |
14f9a607 WAA |
288 | out_err: |
289 | kfree(remotestr); | |
290 | return NULL; | |
c9895cb6 WAA |
291 | } |
292 | ||
16b374ca | 293 | static struct nfs4_pnfs_ds * |
14f9a607 | 294 | nfs4_pnfs_ds_add(struct list_head *dsaddrs, gfp_t gfp_flags) |
16b374ca | 295 | { |
c9895cb6 WAA |
296 | struct nfs4_pnfs_ds *tmp_ds, *ds = NULL; |
297 | char *remotestr; | |
16b374ca | 298 | |
14f9a607 WAA |
299 | if (list_empty(dsaddrs)) { |
300 | dprintk("%s: no addresses defined\n", __func__); | |
301 | goto out; | |
302 | } | |
303 | ||
304 | ds = kzalloc(sizeof(*ds), gfp_flags); | |
16b374ca AA |
305 | if (!ds) |
306 | goto out; | |
307 | ||
c9895cb6 | 308 | /* this is only used for debugging, so it's ok if its NULL */ |
14f9a607 | 309 | remotestr = nfs4_pnfs_remotestr(dsaddrs, gfp_flags); |
c9895cb6 | 310 | |
16b374ca | 311 | spin_lock(&nfs4_ds_cache_lock); |
14f9a607 | 312 | tmp_ds = _data_server_lookup_locked(dsaddrs); |
16b374ca | 313 | if (tmp_ds == NULL) { |
14f9a607 WAA |
314 | INIT_LIST_HEAD(&ds->ds_addrs); |
315 | list_splice_init(dsaddrs, &ds->ds_addrs); | |
c9895cb6 | 316 | ds->ds_remotestr = remotestr; |
16b374ca AA |
317 | atomic_set(&ds->ds_count, 1); |
318 | INIT_LIST_HEAD(&ds->ds_node); | |
319 | ds->ds_clp = NULL; | |
320 | list_add(&ds->ds_node, &nfs4_data_server_cache); | |
c9895cb6 WAA |
321 | dprintk("%s add new data server %s\n", __func__, |
322 | ds->ds_remotestr); | |
16b374ca | 323 | } else { |
c9895cb6 | 324 | kfree(remotestr); |
16b374ca AA |
325 | kfree(ds); |
326 | atomic_inc(&tmp_ds->ds_count); | |
c9895cb6 WAA |
327 | dprintk("%s data server %s found, inc'ed ds_count to %d\n", |
328 | __func__, tmp_ds->ds_remotestr, | |
16b374ca AA |
329 | atomic_read(&tmp_ds->ds_count)); |
330 | ds = tmp_ds; | |
331 | } | |
332 | spin_unlock(&nfs4_ds_cache_lock); | |
333 | out: | |
334 | return ds; | |
335 | } | |
336 | ||
337 | /* | |
c9895cb6 | 338 | * Currently only supports ipv4, ipv6 and one multi-path address. |
16b374ca | 339 | */ |
14f9a607 | 340 | static struct nfs4_pnfs_ds_addr * |
17094272 | 341 | decode_ds_addr(struct net *net, struct xdr_stream *streamp, gfp_t gfp_flags) |
16b374ca | 342 | { |
14f9a607 | 343 | struct nfs4_pnfs_ds_addr *da = NULL; |
c9895cb6 | 344 | char *buf, *portstr; |
13fff2f3 | 345 | __be16 port; |
c9895cb6 | 346 | int nlen, rlen; |
16b374ca | 347 | int tmp[2]; |
35124a09 | 348 | __be32 *p; |
c9895cb6 | 349 | char *netid, *match_netid; |
14f9a607 WAA |
350 | size_t len, match_netid_len; |
351 | char *startsep = ""; | |
352 | char *endsep = ""; | |
353 | ||
16b374ca AA |
354 | |
355 | /* r_netid */ | |
35124a09 WAA |
356 | p = xdr_inline_decode(streamp, 4); |
357 | if (unlikely(!p)) | |
358 | goto out_err; | |
16b374ca | 359 | nlen = be32_to_cpup(p++); |
16b374ca | 360 | |
35124a09 WAA |
361 | p = xdr_inline_decode(streamp, nlen); |
362 | if (unlikely(!p)) | |
363 | goto out_err; | |
16b374ca | 364 | |
c9895cb6 WAA |
365 | netid = kmalloc(nlen+1, gfp_flags); |
366 | if (unlikely(!netid)) | |
16b374ca | 367 | goto out_err; |
16b374ca | 368 | |
c9895cb6 WAA |
369 | netid[nlen] = '\0'; |
370 | memcpy(netid, p, nlen); | |
371 | ||
372 | /* r_addr: ip/ip6addr with port in dec octets - see RFC 5665 */ | |
35124a09 WAA |
373 | p = xdr_inline_decode(streamp, 4); |
374 | if (unlikely(!p)) | |
c9895cb6 | 375 | goto out_free_netid; |
35124a09 WAA |
376 | rlen = be32_to_cpup(p); |
377 | ||
378 | p = xdr_inline_decode(streamp, rlen); | |
379 | if (unlikely(!p)) | |
c9895cb6 | 380 | goto out_free_netid; |
35124a09 | 381 | |
c9895cb6 WAA |
382 | /* port is ".ABC.DEF", 8 chars max */ |
383 | if (rlen > INET6_ADDRSTRLEN + IPV6_SCOPE_ID_LEN + 8) { | |
ad3d2eed | 384 | dprintk("%s: Invalid address, length %d\n", __func__, |
16b374ca | 385 | rlen); |
c9895cb6 | 386 | goto out_free_netid; |
16b374ca | 387 | } |
a75b9df9 | 388 | buf = kmalloc(rlen + 1, gfp_flags); |
b9f81057 SF |
389 | if (!buf) { |
390 | dprintk("%s: Not enough memory\n", __func__); | |
c9895cb6 | 391 | goto out_free_netid; |
b9f81057 | 392 | } |
16b374ca | 393 | buf[rlen] = '\0'; |
35124a09 | 394 | memcpy(buf, p, rlen); |
16b374ca | 395 | |
c9895cb6 WAA |
396 | /* replace port '.' with '-' */ |
397 | portstr = strrchr(buf, '.'); | |
398 | if (!portstr) { | |
399 | dprintk("%s: Failed finding expected dot in port\n", | |
400 | __func__); | |
401 | goto out_free_buf; | |
402 | } | |
403 | *portstr = '-'; | |
404 | ||
405 | /* find '.' between address and port */ | |
406 | portstr = strrchr(buf, '.'); | |
407 | if (!portstr) { | |
408 | dprintk("%s: Failed finding expected dot between address and " | |
409 | "port\n", __func__); | |
410 | goto out_free_buf; | |
16b374ca | 411 | } |
c9895cb6 | 412 | *portstr = '\0'; |
16b374ca | 413 | |
14f9a607 WAA |
414 | da = kzalloc(sizeof(*da), gfp_flags); |
415 | if (unlikely(!da)) | |
c9895cb6 | 416 | goto out_free_buf; |
14f9a607 WAA |
417 | |
418 | INIT_LIST_HEAD(&da->da_node); | |
419 | ||
17094272 | 420 | if (!rpc_pton(net, buf, portstr-buf, (struct sockaddr *)&da->da_addr, |
14f9a607 WAA |
421 | sizeof(da->da_addr))) { |
422 | dprintk("%s: error parsing address %s\n", __func__, buf); | |
423 | goto out_free_da; | |
16b374ca AA |
424 | } |
425 | ||
c9895cb6 WAA |
426 | portstr++; |
427 | sscanf(portstr, "%d-%d", &tmp[0], &tmp[1]); | |
16b374ca AA |
428 | port = htons((tmp[0] << 8) | (tmp[1])); |
429 | ||
14f9a607 | 430 | switch (da->da_addr.ss_family) { |
c9895cb6 | 431 | case AF_INET: |
14f9a607 WAA |
432 | ((struct sockaddr_in *)&da->da_addr)->sin_port = port; |
433 | da->da_addrlen = sizeof(struct sockaddr_in); | |
c9895cb6 WAA |
434 | match_netid = "tcp"; |
435 | match_netid_len = 3; | |
436 | break; | |
437 | ||
438 | case AF_INET6: | |
14f9a607 WAA |
439 | ((struct sockaddr_in6 *)&da->da_addr)->sin6_port = port; |
440 | da->da_addrlen = sizeof(struct sockaddr_in6); | |
c9895cb6 WAA |
441 | match_netid = "tcp6"; |
442 | match_netid_len = 4; | |
14f9a607 WAA |
443 | startsep = "["; |
444 | endsep = "]"; | |
c9895cb6 WAA |
445 | break; |
446 | ||
447 | default: | |
448 | dprintk("%s: unsupported address family: %u\n", | |
14f9a607 WAA |
449 | __func__, da->da_addr.ss_family); |
450 | goto out_free_da; | |
c9895cb6 WAA |
451 | } |
452 | ||
453 | if (nlen != match_netid_len || strncmp(netid, match_netid, nlen)) { | |
454 | dprintk("%s: ERROR: r_netid \"%s\" != \"%s\"\n", | |
455 | __func__, netid, match_netid); | |
14f9a607 | 456 | goto out_free_da; |
c9895cb6 WAA |
457 | } |
458 | ||
14f9a607 WAA |
459 | /* save human readable address */ |
460 | len = strlen(startsep) + strlen(buf) + strlen(endsep) + 7; | |
461 | da->da_remotestr = kzalloc(len, gfp_flags); | |
462 | ||
463 | /* NULL is ok, only used for dprintk */ | |
464 | if (da->da_remotestr) | |
465 | snprintf(da->da_remotestr, len, "%s%s%s:%u", startsep, | |
466 | buf, endsep, ntohs(port)); | |
467 | ||
468 | dprintk("%s: Parsed DS addr %s\n", __func__, da->da_remotestr); | |
469 | kfree(buf); | |
470 | kfree(netid); | |
471 | return da; | |
472 | ||
473 | out_free_da: | |
474 | kfree(da); | |
c9895cb6 | 475 | out_free_buf: |
14f9a607 | 476 | dprintk("%s: Error parsing DS addr: %s\n", __func__, buf); |
16b374ca | 477 | kfree(buf); |
c9895cb6 WAA |
478 | out_free_netid: |
479 | kfree(netid); | |
16b374ca | 480 | out_err: |
14f9a607 | 481 | return NULL; |
16b374ca AA |
482 | } |
483 | ||
484 | /* Decode opaque device data and return the result */ | |
485 | static struct nfs4_file_layout_dsaddr* | |
a75b9df9 | 486 | decode_device(struct inode *ino, struct pnfs_device *pdev, gfp_t gfp_flags) |
16b374ca | 487 | { |
35124a09 | 488 | int i; |
16b374ca AA |
489 | u32 cnt, num; |
490 | u8 *indexp; | |
35124a09 WAA |
491 | __be32 *p; |
492 | u8 *stripe_indices; | |
493 | u8 max_stripe_index; | |
494 | struct nfs4_file_layout_dsaddr *dsaddr = NULL; | |
495 | struct xdr_stream stream; | |
f7da7a12 | 496 | struct xdr_buf buf; |
35124a09 | 497 | struct page *scratch; |
14f9a607 WAA |
498 | struct list_head dsaddrs; |
499 | struct nfs4_pnfs_ds_addr *da; | |
35124a09 WAA |
500 | |
501 | /* set up xdr stream */ | |
a75b9df9 | 502 | scratch = alloc_page(gfp_flags); |
35124a09 WAA |
503 | if (!scratch) |
504 | goto out_err; | |
505 | ||
f7da7a12 | 506 | xdr_init_decode_pages(&stream, &buf, pdev->pages, pdev->pglen); |
35124a09 | 507 | xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE); |
16b374ca AA |
508 | |
509 | /* Get the stripe count (number of stripe index) */ | |
35124a09 WAA |
510 | p = xdr_inline_decode(&stream, 4); |
511 | if (unlikely(!p)) | |
512 | goto out_err_free_scratch; | |
513 | ||
514 | cnt = be32_to_cpup(p); | |
16b374ca AA |
515 | dprintk("%s stripe count %d\n", __func__, cnt); |
516 | if (cnt > NFS4_PNFS_MAX_STRIPE_CNT) { | |
a030889a | 517 | printk(KERN_WARNING "NFS: %s: stripe count %d greater than " |
16b374ca AA |
518 | "supported maximum %d\n", __func__, |
519 | cnt, NFS4_PNFS_MAX_STRIPE_CNT); | |
35124a09 WAA |
520 | goto out_err_free_scratch; |
521 | } | |
522 | ||
523 | /* read stripe indices */ | |
a75b9df9 | 524 | stripe_indices = kcalloc(cnt, sizeof(u8), gfp_flags); |
35124a09 WAA |
525 | if (!stripe_indices) |
526 | goto out_err_free_scratch; | |
527 | ||
528 | p = xdr_inline_decode(&stream, cnt << 2); | |
529 | if (unlikely(!p)) | |
530 | goto out_err_free_stripe_indices; | |
531 | ||
532 | indexp = &stripe_indices[0]; | |
533 | max_stripe_index = 0; | |
534 | for (i = 0; i < cnt; i++) { | |
535 | *indexp = be32_to_cpup(p++); | |
536 | max_stripe_index = max(max_stripe_index, *indexp); | |
537 | indexp++; | |
16b374ca AA |
538 | } |
539 | ||
540 | /* Check the multipath list count */ | |
35124a09 WAA |
541 | p = xdr_inline_decode(&stream, 4); |
542 | if (unlikely(!p)) | |
543 | goto out_err_free_stripe_indices; | |
544 | ||
545 | num = be32_to_cpup(p); | |
16b374ca AA |
546 | dprintk("%s ds_num %u\n", __func__, num); |
547 | if (num > NFS4_PNFS_MAX_MULTI_CNT) { | |
a030889a | 548 | printk(KERN_WARNING "NFS: %s: multipath count %d greater than " |
16b374ca AA |
549 | "supported maximum %d\n", __func__, |
550 | num, NFS4_PNFS_MAX_MULTI_CNT); | |
35124a09 | 551 | goto out_err_free_stripe_indices; |
16b374ca | 552 | } |
35124a09 WAA |
553 | |
554 | /* validate stripe indices are all < num */ | |
555 | if (max_stripe_index >= num) { | |
a030889a | 556 | printk(KERN_WARNING "NFS: %s: stripe index %u >= num ds %u\n", |
35124a09 WAA |
557 | __func__, max_stripe_index, num); |
558 | goto out_err_free_stripe_indices; | |
559 | } | |
560 | ||
16b374ca AA |
561 | dsaddr = kzalloc(sizeof(*dsaddr) + |
562 | (sizeof(struct nfs4_pnfs_ds *) * (num - 1)), | |
a75b9df9 | 563 | gfp_flags); |
16b374ca | 564 | if (!dsaddr) |
35124a09 | 565 | goto out_err_free_stripe_indices; |
16b374ca AA |
566 | |
567 | dsaddr->stripe_count = cnt; | |
35124a09 WAA |
568 | dsaddr->stripe_indices = stripe_indices; |
569 | stripe_indices = NULL; | |
16b374ca | 570 | dsaddr->ds_num = num; |
1775bc34 BH |
571 | nfs4_init_deviceid_node(&dsaddr->id_node, |
572 | NFS_SERVER(ino)->pnfs_curr_ld, | |
573 | NFS_SERVER(ino)->nfs_client, | |
a1eaecbc | 574 | &pdev->dev_id); |
16b374ca | 575 | |
14f9a607 WAA |
576 | INIT_LIST_HEAD(&dsaddrs); |
577 | ||
16b374ca AA |
578 | for (i = 0; i < dsaddr->ds_num; i++) { |
579 | int j; | |
35124a09 WAA |
580 | u32 mp_count; |
581 | ||
582 | p = xdr_inline_decode(&stream, 4); | |
583 | if (unlikely(!p)) | |
584 | goto out_err_free_deviceid; | |
16b374ca | 585 | |
35124a09 | 586 | mp_count = be32_to_cpup(p); /* multipath count */ |
35124a09 | 587 | for (j = 0; j < mp_count; j++) { |
73ea666c | 588 | da = decode_ds_addr(NFS_SERVER(ino)->nfs_client->cl_net, |
17094272 | 589 | &stream, gfp_flags); |
14f9a607 WAA |
590 | if (da) |
591 | list_add_tail(&da->da_node, &dsaddrs); | |
592 | } | |
593 | if (list_empty(&dsaddrs)) { | |
594 | dprintk("%s: no suitable DS addresses found\n", | |
595 | __func__); | |
596 | goto out_err_free_deviceid; | |
597 | } | |
598 | ||
599 | dsaddr->ds_list[i] = nfs4_pnfs_ds_add(&dsaddrs, gfp_flags); | |
600 | if (!dsaddr->ds_list[i]) | |
601 | goto out_err_drain_dsaddrs; | |
602 | ||
603 | /* If DS was already in cache, free ds addrs */ | |
604 | while (!list_empty(&dsaddrs)) { | |
605 | da = list_first_entry(&dsaddrs, | |
606 | struct nfs4_pnfs_ds_addr, | |
607 | da_node); | |
608 | list_del_init(&da->da_node); | |
609 | kfree(da->da_remotestr); | |
610 | kfree(da); | |
16b374ca AA |
611 | } |
612 | } | |
35124a09 WAA |
613 | |
614 | __free_page(scratch); | |
16b374ca AA |
615 | return dsaddr; |
616 | ||
14f9a607 WAA |
617 | out_err_drain_dsaddrs: |
618 | while (!list_empty(&dsaddrs)) { | |
619 | da = list_first_entry(&dsaddrs, struct nfs4_pnfs_ds_addr, | |
620 | da_node); | |
621 | list_del_init(&da->da_node); | |
622 | kfree(da->da_remotestr); | |
623 | kfree(da); | |
624 | } | |
35124a09 | 625 | out_err_free_deviceid: |
16b374ca | 626 | nfs4_fl_free_deviceid(dsaddr); |
35124a09 WAA |
627 | /* stripe_indicies was part of dsaddr */ |
628 | goto out_err_free_scratch; | |
629 | out_err_free_stripe_indices: | |
630 | kfree(stripe_indices); | |
631 | out_err_free_scratch: | |
632 | __free_page(scratch); | |
16b374ca AA |
633 | out_err: |
634 | dprintk("%s ERROR: returning NULL\n", __func__); | |
635 | return NULL; | |
636 | } | |
637 | ||
638 | /* | |
ea8eecdd CH |
639 | * Decode the opaque device specified in 'dev' and add it to the cache of |
640 | * available devices. | |
16b374ca | 641 | */ |
ea8eecdd | 642 | static struct nfs4_file_layout_dsaddr * |
a75b9df9 | 643 | decode_and_add_device(struct inode *inode, struct pnfs_device *dev, gfp_t gfp_flags) |
16b374ca | 644 | { |
a1eaecbc BH |
645 | struct nfs4_deviceid_node *d; |
646 | struct nfs4_file_layout_dsaddr *n, *new; | |
16b374ca | 647 | |
a75b9df9 | 648 | new = decode_device(inode, dev, gfp_flags); |
ea8eecdd | 649 | if (!new) { |
a030889a | 650 | printk(KERN_WARNING "NFS: %s: Could not decode or add device\n", |
16b374ca AA |
651 | __func__); |
652 | return NULL; | |
653 | } | |
654 | ||
a1eaecbc BH |
655 | d = nfs4_insert_deviceid_node(&new->id_node); |
656 | n = container_of(d, struct nfs4_file_layout_dsaddr, id_node); | |
657 | if (n != new) { | |
ea8eecdd | 658 | nfs4_fl_free_deviceid(new); |
a1eaecbc | 659 | return n; |
ea8eecdd CH |
660 | } |
661 | ||
ea8eecdd | 662 | return new; |
16b374ca AA |
663 | } |
664 | ||
665 | /* | |
666 | * Retrieve the information for dev_id, add it to the list | |
667 | * of available devices, and return it. | |
668 | */ | |
669 | struct nfs4_file_layout_dsaddr * | |
78e4e05c | 670 | filelayout_get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id, gfp_t gfp_flags) |
16b374ca AA |
671 | { |
672 | struct pnfs_device *pdev = NULL; | |
673 | u32 max_resp_sz; | |
674 | int max_pages; | |
675 | struct page **pages = NULL; | |
676 | struct nfs4_file_layout_dsaddr *dsaddr = NULL; | |
677 | int rc, i; | |
678 | struct nfs_server *server = NFS_SERVER(inode); | |
679 | ||
680 | /* | |
681 | * Use the session max response size as the basis for setting | |
682 | * GETDEVICEINFO's maxcount | |
683 | */ | |
684 | max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz; | |
e5265a0c | 685 | max_pages = nfs_page_array_len(0, max_resp_sz); |
16b374ca AA |
686 | dprintk("%s inode %p max_resp_sz %u max_pages %d\n", |
687 | __func__, inode, max_resp_sz, max_pages); | |
688 | ||
a75b9df9 | 689 | pdev = kzalloc(sizeof(struct pnfs_device), gfp_flags); |
16b374ca AA |
690 | if (pdev == NULL) |
691 | return NULL; | |
692 | ||
a75b9df9 | 693 | pages = kzalloc(max_pages * sizeof(struct page *), gfp_flags); |
16b374ca AA |
694 | if (pages == NULL) { |
695 | kfree(pdev); | |
696 | return NULL; | |
697 | } | |
698 | for (i = 0; i < max_pages; i++) { | |
a75b9df9 | 699 | pages[i] = alloc_page(gfp_flags); |
16b374ca AA |
700 | if (!pages[i]) |
701 | goto out_free; | |
702 | } | |
703 | ||
16b374ca AA |
704 | memcpy(&pdev->dev_id, dev_id, sizeof(*dev_id)); |
705 | pdev->layout_type = LAYOUT_NFSV4_1_FILES; | |
706 | pdev->pages = pages; | |
707 | pdev->pgbase = 0; | |
05bf14ad | 708 | pdev->pglen = max_resp_sz; |
16b374ca AA |
709 | pdev->mincount = 0; |
710 | ||
711 | rc = nfs4_proc_getdeviceinfo(server, pdev); | |
712 | dprintk("%s getdevice info returns %d\n", __func__, rc); | |
713 | if (rc) | |
714 | goto out_free; | |
715 | ||
716 | /* | |
717 | * Found new device, need to decode it and then add it to the | |
718 | * list of known devices for this mountpoint. | |
719 | */ | |
a75b9df9 | 720 | dsaddr = decode_and_add_device(inode, pdev, gfp_flags); |
16b374ca | 721 | out_free: |
16b374ca AA |
722 | for (i = 0; i < max_pages; i++) |
723 | __free_page(pages[i]); | |
724 | kfree(pages); | |
725 | kfree(pdev); | |
726 | dprintk("<-- %s dsaddr %p\n", __func__, dsaddr); | |
727 | return dsaddr; | |
728 | } | |
729 | ||
ea8eecdd CH |
730 | void |
731 | nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr) | |
16b374ca | 732 | { |
1775bc34 | 733 | nfs4_put_deviceid_node(&dsaddr->id_node); |
16b374ca | 734 | } |
cfe7f412 FI |
735 | |
736 | /* | |
737 | * Want res = (offset - layout->pattern_offset)/ layout->stripe_unit | |
738 | * Then: ((res + fsi) % dsaddr->stripe_count) | |
739 | */ | |
740 | u32 | |
741 | nfs4_fl_calc_j_index(struct pnfs_layout_segment *lseg, loff_t offset) | |
742 | { | |
743 | struct nfs4_filelayout_segment *flseg = FILELAYOUT_LSEG(lseg); | |
744 | u64 tmp; | |
745 | ||
746 | tmp = offset - flseg->pattern_offset; | |
747 | do_div(tmp, flseg->stripe_unit); | |
748 | tmp += flseg->first_stripe_index; | |
749 | return do_div(tmp, flseg->dsaddr->stripe_count); | |
750 | } | |
751 | ||
752 | u32 | |
753 | nfs4_fl_calc_ds_index(struct pnfs_layout_segment *lseg, u32 j) | |
754 | { | |
755 | return FILELAYOUT_LSEG(lseg)->dsaddr->stripe_indices[j]; | |
756 | } | |
757 | ||
758 | struct nfs_fh * | |
759 | nfs4_fl_select_ds_fh(struct pnfs_layout_segment *lseg, u32 j) | |
760 | { | |
761 | struct nfs4_filelayout_segment *flseg = FILELAYOUT_LSEG(lseg); | |
762 | u32 i; | |
763 | ||
764 | if (flseg->stripe_type == STRIPE_SPARSE) { | |
765 | if (flseg->num_fh == 1) | |
766 | i = 0; | |
767 | else if (flseg->num_fh == 0) | |
768 | /* Use the MDS OPEN fh set in nfs_read_rpcsetup */ | |
769 | return NULL; | |
770 | else | |
771 | i = nfs4_fl_calc_ds_index(lseg, j); | |
772 | } else | |
773 | i = j; | |
774 | return flseg->fh_array[i]; | |
775 | } | |
776 | ||
777 | struct nfs4_pnfs_ds * | |
778 | nfs4_fl_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx) | |
779 | { | |
780 | struct nfs4_file_layout_dsaddr *dsaddr = FILELAYOUT_LSEG(lseg)->dsaddr; | |
781 | struct nfs4_pnfs_ds *ds = dsaddr->ds_list[ds_idx]; | |
554d458d AA |
782 | struct nfs4_deviceid_node *devid = FILELAYOUT_DEVID_NODE(lseg); |
783 | ||
1dfed273 | 784 | if (filelayout_test_devid_unavailable(devid)) |
554d458d | 785 | return NULL; |
cfe7f412 FI |
786 | |
787 | if (ds == NULL) { | |
a030889a | 788 | printk(KERN_ERR "NFS: %s: No data server for offset index %d\n", |
cfe7f412 | 789 | __func__, ds_idx); |
1dfed273 TM |
790 | filelayout_mark_devid_invalid(devid); |
791 | return NULL; | |
cfe7f412 FI |
792 | } |
793 | ||
794 | if (!ds->ds_clp) { | |
568e8c49 | 795 | struct nfs_server *s = NFS_SERVER(lseg->pls_layout->plh_inode); |
cfe7f412 FI |
796 | int err; |
797 | ||
568e8c49 | 798 | err = nfs4_ds_connect(s, ds); |
1dfed273 TM |
799 | if (err) { |
800 | nfs4_mark_deviceid_unavailable(devid); | |
801 | return NULL; | |
802 | } | |
cfe7f412 FI |
803 | } |
804 | return ds; | |
805 | } | |
98fc685a AA |
806 | |
807 | module_param(dataserver_retrans, uint, 0644); | |
808 | MODULE_PARM_DESC(dataserver_retrans, "The number of times the NFSv4.1 client " | |
809 | "retries a request before it attempts further " | |
810 | " recovery action."); | |
811 | module_param(dataserver_timeo, uint, 0644); | |
812 | MODULE_PARM_DESC(dataserver_timeo, "The time (in tenths of a second) the " | |
813 | "NFSv4.1 client waits for a response from a " | |
814 | " data server before it retries an NFS request."); |