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> |
5976687a | 34 | #include <linux/sunrpc/addr.h> |
16b374ca AA |
35 | |
36 | #include "internal.h" | |
73e39aaa | 37 | #include "nfs4session.h" |
16b374ca AA |
38 | #include "nfs4filelayout.h" |
39 | ||
40 | #define NFSDBG_FACILITY NFSDBG_PNFS_LD | |
41 | ||
98fc685a AA |
42 | static unsigned int dataserver_timeo = NFS4_DEF_DS_TIMEO; |
43 | static unsigned int dataserver_retrans = NFS4_DEF_DS_RETRANS; | |
44 | ||
16b374ca AA |
45 | /* |
46 | * Data server cache | |
47 | * | |
48 | * Data servers can be mapped to different device ids. | |
49 | * nfs4_pnfs_ds reference counting | |
50 | * - set to 1 on allocation | |
51 | * - incremented when a device id maps a data server already in the cache. | |
52 | * - decremented when deviceid is removed from the cache. | |
53 | */ | |
17280175 | 54 | static DEFINE_SPINLOCK(nfs4_ds_cache_lock); |
16b374ca AA |
55 | static LIST_HEAD(nfs4_data_server_cache); |
56 | ||
57 | /* Debug routines */ | |
58 | void | |
59 | print_ds(struct nfs4_pnfs_ds *ds) | |
60 | { | |
61 | if (ds == NULL) { | |
62 | printk("%s NULL device\n", __func__); | |
63 | return; | |
64 | } | |
c9895cb6 | 65 | printk(" ds %s\n" |
16b374ca AA |
66 | " ref count %d\n" |
67 | " client %p\n" | |
68 | " cl_exchange_flags %x\n", | |
c9895cb6 | 69 | ds->ds_remotestr, |
16b374ca AA |
70 | atomic_read(&ds->ds_count), ds->ds_clp, |
71 | ds->ds_clp ? ds->ds_clp->cl_exchange_flags : 0); | |
72 | } | |
73 | ||
14f9a607 WAA |
74 | static bool |
75 | same_sockaddr(struct sockaddr *addr1, struct sockaddr *addr2) | |
16b374ca | 76 | { |
c9895cb6 WAA |
77 | struct sockaddr_in *a, *b; |
78 | struct sockaddr_in6 *a6, *b6; | |
16b374ca | 79 | |
14f9a607 WAA |
80 | if (addr1->sa_family != addr2->sa_family) |
81 | return false; | |
82 | ||
83 | switch (addr1->sa_family) { | |
84 | case AF_INET: | |
85 | a = (struct sockaddr_in *)addr1; | |
86 | b = (struct sockaddr_in *)addr2; | |
87 | ||
88 | if (a->sin_addr.s_addr == b->sin_addr.s_addr && | |
89 | a->sin_port == b->sin_port) | |
90 | return true; | |
91 | break; | |
92 | ||
93 | case AF_INET6: | |
94 | a6 = (struct sockaddr_in6 *)addr1; | |
95 | b6 = (struct sockaddr_in6 *)addr2; | |
96 | ||
97 | /* LINKLOCAL addresses must have matching scope_id */ | |
98 | if (ipv6_addr_scope(&a6->sin6_addr) == | |
99 | IPV6_ADDR_SCOPE_LINKLOCAL && | |
100 | a6->sin6_scope_id != b6->sin6_scope_id) | |
101 | return false; | |
102 | ||
103 | if (ipv6_addr_equal(&a6->sin6_addr, &b6->sin6_addr) && | |
104 | a6->sin6_port == b6->sin6_port) | |
105 | return true; | |
106 | break; | |
107 | ||
108 | default: | |
109 | dprintk("%s: unhandled address family: %u\n", | |
110 | __func__, addr1->sa_family); | |
111 | return false; | |
112 | } | |
113 | ||
114 | return false; | |
115 | } | |
116 | ||
17280175 | 117 | static bool |
2d3fe01c WAA |
118 | _same_data_server_addrs_locked(const struct list_head *dsaddrs1, |
119 | const struct list_head *dsaddrs2) | |
14f9a607 | 120 | { |
14f9a607 WAA |
121 | struct nfs4_pnfs_ds_addr *da1, *da2; |
122 | ||
2d3fe01c WAA |
123 | /* step through both lists, comparing as we go */ |
124 | for (da1 = list_first_entry(dsaddrs1, typeof(*da1), da_node), | |
125 | da2 = list_first_entry(dsaddrs2, typeof(*da2), da_node); | |
126 | da1 != NULL && da2 != NULL; | |
127 | da1 = list_entry(da1->da_node.next, typeof(*da1), da_node), | |
128 | da2 = list_entry(da2->da_node.next, typeof(*da2), da_node)) { | |
129 | if (!same_sockaddr((struct sockaddr *)&da1->da_addr, | |
130 | (struct sockaddr *)&da2->da_addr)) | |
131 | return false; | |
16b374ca | 132 | } |
2d3fe01c WAA |
133 | if (da1 == NULL && da2 == NULL) |
134 | return true; | |
135 | ||
136 | return false; | |
16b374ca AA |
137 | } |
138 | ||
14f9a607 | 139 | /* |
2d3fe01c | 140 | * Lookup DS by addresses. nfs4_ds_cache_lock is held |
14f9a607 | 141 | */ |
2d3fe01c WAA |
142 | static struct nfs4_pnfs_ds * |
143 | _data_server_lookup_locked(const struct list_head *dsaddrs) | |
14f9a607 | 144 | { |
2d3fe01c | 145 | struct nfs4_pnfs_ds *ds; |
14f9a607 | 146 | |
2d3fe01c WAA |
147 | list_for_each_entry(ds, &nfs4_data_server_cache, ds_node) |
148 | if (_same_data_server_addrs_locked(&ds->ds_addrs, dsaddrs)) | |
149 | return ds; | |
150 | return NULL; | |
14f9a607 WAA |
151 | } |
152 | ||
d83217c1 AA |
153 | /* |
154 | * Create an rpc connection to the nfs4_pnfs_ds data server | |
35dbbc99 | 155 | * Currently only supports IPv4 and IPv6 addresses |
d83217c1 AA |
156 | */ |
157 | static int | |
158 | nfs4_ds_connect(struct nfs_server *mds_srv, struct nfs4_pnfs_ds *ds) | |
159 | { | |
7e574f0d | 160 | struct nfs_client *clp = ERR_PTR(-EIO); |
14f9a607 | 161 | struct nfs4_pnfs_ds_addr *da; |
d83217c1 AA |
162 | int status = 0; |
163 | ||
14f9a607 | 164 | dprintk("--> %s DS %s au_flavor %d\n", __func__, ds->ds_remotestr, |
d83217c1 AA |
165 | mds_srv->nfs_client->cl_rpcclient->cl_auth->au_flavor); |
166 | ||
7e574f0d WAA |
167 | list_for_each_entry(da, &ds->ds_addrs, da_node) { |
168 | dprintk("%s: DS %s: trying address %s\n", | |
169 | __func__, ds->ds_remotestr, da->da_remotestr); | |
14f9a607 | 170 | |
7e574f0d | 171 | clp = nfs4_set_ds_client(mds_srv->nfs_client, |
98fc685a AA |
172 | (struct sockaddr *)&da->da_addr, |
173 | da->da_addrlen, IPPROTO_TCP, | |
174 | dataserver_timeo, dataserver_retrans); | |
7e574f0d WAA |
175 | if (!IS_ERR(clp)) |
176 | break; | |
177 | } | |
178 | ||
d83217c1 AA |
179 | if (IS_ERR(clp)) { |
180 | status = PTR_ERR(clp); | |
181 | goto out; | |
182 | } | |
183 | ||
7b38c368 | 184 | status = nfs4_init_ds_session(clp, mds_srv->nfs_client->cl_lease_time); |
d83217c1 AA |
185 | if (status) |
186 | goto out_put; | |
187 | ||
188 | ds->ds_clp = clp; | |
c9895cb6 | 189 | dprintk("%s [new] addr: %s\n", __func__, ds->ds_remotestr); |
d83217c1 AA |
190 | out: |
191 | return status; | |
192 | out_put: | |
193 | nfs_put_client(clp); | |
194 | goto out; | |
195 | } | |
196 | ||
16b374ca AA |
197 | static void |
198 | destroy_ds(struct nfs4_pnfs_ds *ds) | |
199 | { | |
14f9a607 WAA |
200 | struct nfs4_pnfs_ds_addr *da; |
201 | ||
16b374ca AA |
202 | dprintk("--> %s\n", __func__); |
203 | ifdebug(FACILITY) | |
204 | print_ds(ds); | |
205 | ||
206 | if (ds->ds_clp) | |
207 | nfs_put_client(ds->ds_clp); | |
14f9a607 WAA |
208 | |
209 | while (!list_empty(&ds->ds_addrs)) { | |
210 | da = list_first_entry(&ds->ds_addrs, | |
211 | struct nfs4_pnfs_ds_addr, | |
212 | da_node); | |
213 | list_del_init(&da->da_node); | |
214 | kfree(da->da_remotestr); | |
215 | kfree(da); | |
216 | } | |
217 | ||
c9895cb6 | 218 | kfree(ds->ds_remotestr); |
16b374ca AA |
219 | kfree(ds); |
220 | } | |
221 | ||
1775bc34 | 222 | void |
16b374ca AA |
223 | nfs4_fl_free_deviceid(struct nfs4_file_layout_dsaddr *dsaddr) |
224 | { | |
225 | struct nfs4_pnfs_ds *ds; | |
226 | int i; | |
227 | ||
a1eaecbc | 228 | nfs4_print_deviceid(&dsaddr->id_node.deviceid); |
16b374ca AA |
229 | |
230 | for (i = 0; i < dsaddr->ds_num; i++) { | |
231 | ds = dsaddr->ds_list[i]; | |
232 | if (ds != NULL) { | |
233 | if (atomic_dec_and_lock(&ds->ds_count, | |
234 | &nfs4_ds_cache_lock)) { | |
235 | list_del_init(&ds->ds_node); | |
236 | spin_unlock(&nfs4_ds_cache_lock); | |
237 | destroy_ds(ds); | |
238 | } | |
239 | } | |
240 | } | |
241 | kfree(dsaddr->stripe_indices); | |
242 | kfree(dsaddr); | |
243 | } | |
244 | ||
c9895cb6 WAA |
245 | /* |
246 | * Create a string with a human readable address and port to avoid | |
247 | * complicated setup around many dprinks. | |
248 | */ | |
249 | static char * | |
14f9a607 | 250 | nfs4_pnfs_remotestr(struct list_head *dsaddrs, gfp_t gfp_flags) |
c9895cb6 | 251 | { |
14f9a607 | 252 | struct nfs4_pnfs_ds_addr *da; |
c9895cb6 | 253 | char *remotestr; |
c9895cb6 | 254 | size_t len; |
14f9a607 | 255 | char *p; |
c9895cb6 | 256 | |
14f9a607 WAA |
257 | len = 3; /* '{', '}' and eol */ |
258 | list_for_each_entry(da, dsaddrs, da_node) { | |
259 | len += strlen(da->da_remotestr) + 1; /* string plus comma */ | |
c9895cb6 WAA |
260 | } |
261 | ||
14f9a607 WAA |
262 | remotestr = kzalloc(len, gfp_flags); |
263 | if (!remotestr) | |
c9895cb6 | 264 | return NULL; |
c9895cb6 | 265 | |
14f9a607 WAA |
266 | p = remotestr; |
267 | *(p++) = '{'; | |
268 | len--; | |
269 | list_for_each_entry(da, dsaddrs, da_node) { | |
270 | size_t ll = strlen(da->da_remotestr); | |
c9895cb6 | 271 | |
14f9a607 WAA |
272 | if (ll > len) |
273 | goto out_err; | |
c9895cb6 | 274 | |
14f9a607 WAA |
275 | memcpy(p, da->da_remotestr, ll); |
276 | p += ll; | |
277 | len -= ll; | |
c9895cb6 | 278 | |
14f9a607 WAA |
279 | if (len < 1) |
280 | goto out_err; | |
281 | (*p++) = ','; | |
282 | len--; | |
283 | } | |
284 | if (len < 2) | |
285 | goto out_err; | |
286 | *(p++) = '}'; | |
287 | *p = '\0'; | |
c9895cb6 | 288 | return remotestr; |
14f9a607 WAA |
289 | out_err: |
290 | kfree(remotestr); | |
291 | return NULL; | |
c9895cb6 WAA |
292 | } |
293 | ||
16b374ca | 294 | static struct nfs4_pnfs_ds * |
14f9a607 | 295 | nfs4_pnfs_ds_add(struct list_head *dsaddrs, gfp_t gfp_flags) |
16b374ca | 296 | { |
c9895cb6 WAA |
297 | struct nfs4_pnfs_ds *tmp_ds, *ds = NULL; |
298 | char *remotestr; | |
16b374ca | 299 | |
14f9a607 WAA |
300 | if (list_empty(dsaddrs)) { |
301 | dprintk("%s: no addresses defined\n", __func__); | |
302 | goto out; | |
303 | } | |
304 | ||
305 | ds = kzalloc(sizeof(*ds), gfp_flags); | |
16b374ca AA |
306 | if (!ds) |
307 | goto out; | |
308 | ||
c9895cb6 | 309 | /* this is only used for debugging, so it's ok if its NULL */ |
14f9a607 | 310 | remotestr = nfs4_pnfs_remotestr(dsaddrs, gfp_flags); |
c9895cb6 | 311 | |
16b374ca | 312 | spin_lock(&nfs4_ds_cache_lock); |
14f9a607 | 313 | tmp_ds = _data_server_lookup_locked(dsaddrs); |
16b374ca | 314 | if (tmp_ds == NULL) { |
14f9a607 WAA |
315 | INIT_LIST_HEAD(&ds->ds_addrs); |
316 | list_splice_init(dsaddrs, &ds->ds_addrs); | |
c9895cb6 | 317 | ds->ds_remotestr = remotestr; |
16b374ca AA |
318 | atomic_set(&ds->ds_count, 1); |
319 | INIT_LIST_HEAD(&ds->ds_node); | |
320 | ds->ds_clp = NULL; | |
321 | list_add(&ds->ds_node, &nfs4_data_server_cache); | |
c9895cb6 WAA |
322 | dprintk("%s add new data server %s\n", __func__, |
323 | ds->ds_remotestr); | |
16b374ca | 324 | } else { |
c9895cb6 | 325 | kfree(remotestr); |
16b374ca AA |
326 | kfree(ds); |
327 | atomic_inc(&tmp_ds->ds_count); | |
c9895cb6 WAA |
328 | dprintk("%s data server %s found, inc'ed ds_count to %d\n", |
329 | __func__, tmp_ds->ds_remotestr, | |
16b374ca AA |
330 | atomic_read(&tmp_ds->ds_count)); |
331 | ds = tmp_ds; | |
332 | } | |
333 | spin_unlock(&nfs4_ds_cache_lock); | |
334 | out: | |
335 | return ds; | |
336 | } | |
337 | ||
338 | /* | |
c9895cb6 | 339 | * Currently only supports ipv4, ipv6 and one multi-path address. |
16b374ca | 340 | */ |
14f9a607 | 341 | static struct nfs4_pnfs_ds_addr * |
17094272 | 342 | decode_ds_addr(struct net *net, struct xdr_stream *streamp, gfp_t gfp_flags) |
16b374ca | 343 | { |
14f9a607 | 344 | struct nfs4_pnfs_ds_addr *da = NULL; |
c9895cb6 | 345 | char *buf, *portstr; |
13fff2f3 | 346 | __be16 port; |
c9895cb6 | 347 | int nlen, rlen; |
16b374ca | 348 | int tmp[2]; |
35124a09 | 349 | __be32 *p; |
c9895cb6 | 350 | char *netid, *match_netid; |
14f9a607 WAA |
351 | size_t len, match_netid_len; |
352 | char *startsep = ""; | |
353 | char *endsep = ""; | |
354 | ||
16b374ca AA |
355 | |
356 | /* r_netid */ | |
35124a09 WAA |
357 | p = xdr_inline_decode(streamp, 4); |
358 | if (unlikely(!p)) | |
359 | goto out_err; | |
16b374ca | 360 | nlen = be32_to_cpup(p++); |
16b374ca | 361 | |
35124a09 WAA |
362 | p = xdr_inline_decode(streamp, nlen); |
363 | if (unlikely(!p)) | |
364 | goto out_err; | |
16b374ca | 365 | |
c9895cb6 WAA |
366 | netid = kmalloc(nlen+1, gfp_flags); |
367 | if (unlikely(!netid)) | |
16b374ca | 368 | goto out_err; |
16b374ca | 369 | |
c9895cb6 WAA |
370 | netid[nlen] = '\0'; |
371 | memcpy(netid, p, nlen); | |
372 | ||
373 | /* r_addr: ip/ip6addr with port in dec octets - see RFC 5665 */ | |
35124a09 WAA |
374 | p = xdr_inline_decode(streamp, 4); |
375 | if (unlikely(!p)) | |
c9895cb6 | 376 | goto out_free_netid; |
35124a09 WAA |
377 | rlen = be32_to_cpup(p); |
378 | ||
379 | p = xdr_inline_decode(streamp, rlen); | |
380 | if (unlikely(!p)) | |
c9895cb6 | 381 | goto out_free_netid; |
35124a09 | 382 | |
c9895cb6 WAA |
383 | /* port is ".ABC.DEF", 8 chars max */ |
384 | if (rlen > INET6_ADDRSTRLEN + IPV6_SCOPE_ID_LEN + 8) { | |
ad3d2eed | 385 | dprintk("%s: Invalid address, length %d\n", __func__, |
16b374ca | 386 | rlen); |
c9895cb6 | 387 | goto out_free_netid; |
16b374ca | 388 | } |
a75b9df9 | 389 | buf = kmalloc(rlen + 1, gfp_flags); |
b9f81057 SF |
390 | if (!buf) { |
391 | dprintk("%s: Not enough memory\n", __func__); | |
c9895cb6 | 392 | goto out_free_netid; |
b9f81057 | 393 | } |
16b374ca | 394 | buf[rlen] = '\0'; |
35124a09 | 395 | memcpy(buf, p, rlen); |
16b374ca | 396 | |
c9895cb6 WAA |
397 | /* replace port '.' with '-' */ |
398 | portstr = strrchr(buf, '.'); | |
399 | if (!portstr) { | |
400 | dprintk("%s: Failed finding expected dot in port\n", | |
401 | __func__); | |
402 | goto out_free_buf; | |
403 | } | |
404 | *portstr = '-'; | |
405 | ||
406 | /* find '.' between address and port */ | |
407 | portstr = strrchr(buf, '.'); | |
408 | if (!portstr) { | |
409 | dprintk("%s: Failed finding expected dot between address and " | |
410 | "port\n", __func__); | |
411 | goto out_free_buf; | |
16b374ca | 412 | } |
c9895cb6 | 413 | *portstr = '\0'; |
16b374ca | 414 | |
14f9a607 WAA |
415 | da = kzalloc(sizeof(*da), gfp_flags); |
416 | if (unlikely(!da)) | |
c9895cb6 | 417 | goto out_free_buf; |
14f9a607 WAA |
418 | |
419 | INIT_LIST_HEAD(&da->da_node); | |
420 | ||
17094272 | 421 | if (!rpc_pton(net, buf, portstr-buf, (struct sockaddr *)&da->da_addr, |
14f9a607 WAA |
422 | sizeof(da->da_addr))) { |
423 | dprintk("%s: error parsing address %s\n", __func__, buf); | |
424 | goto out_free_da; | |
16b374ca AA |
425 | } |
426 | ||
c9895cb6 WAA |
427 | portstr++; |
428 | sscanf(portstr, "%d-%d", &tmp[0], &tmp[1]); | |
16b374ca AA |
429 | port = htons((tmp[0] << 8) | (tmp[1])); |
430 | ||
14f9a607 | 431 | switch (da->da_addr.ss_family) { |
c9895cb6 | 432 | case AF_INET: |
14f9a607 WAA |
433 | ((struct sockaddr_in *)&da->da_addr)->sin_port = port; |
434 | da->da_addrlen = sizeof(struct sockaddr_in); | |
c9895cb6 WAA |
435 | match_netid = "tcp"; |
436 | match_netid_len = 3; | |
437 | break; | |
438 | ||
439 | case AF_INET6: | |
14f9a607 WAA |
440 | ((struct sockaddr_in6 *)&da->da_addr)->sin6_port = port; |
441 | da->da_addrlen = sizeof(struct sockaddr_in6); | |
c9895cb6 WAA |
442 | match_netid = "tcp6"; |
443 | match_netid_len = 4; | |
14f9a607 WAA |
444 | startsep = "["; |
445 | endsep = "]"; | |
c9895cb6 WAA |
446 | break; |
447 | ||
448 | default: | |
449 | dprintk("%s: unsupported address family: %u\n", | |
14f9a607 WAA |
450 | __func__, da->da_addr.ss_family); |
451 | goto out_free_da; | |
c9895cb6 WAA |
452 | } |
453 | ||
454 | if (nlen != match_netid_len || strncmp(netid, match_netid, nlen)) { | |
455 | dprintk("%s: ERROR: r_netid \"%s\" != \"%s\"\n", | |
456 | __func__, netid, match_netid); | |
14f9a607 | 457 | goto out_free_da; |
c9895cb6 WAA |
458 | } |
459 | ||
14f9a607 WAA |
460 | /* save human readable address */ |
461 | len = strlen(startsep) + strlen(buf) + strlen(endsep) + 7; | |
462 | da->da_remotestr = kzalloc(len, gfp_flags); | |
463 | ||
464 | /* NULL is ok, only used for dprintk */ | |
465 | if (da->da_remotestr) | |
466 | snprintf(da->da_remotestr, len, "%s%s%s:%u", startsep, | |
467 | buf, endsep, ntohs(port)); | |
468 | ||
469 | dprintk("%s: Parsed DS addr %s\n", __func__, da->da_remotestr); | |
470 | kfree(buf); | |
471 | kfree(netid); | |
472 | return da; | |
473 | ||
474 | out_free_da: | |
475 | kfree(da); | |
c9895cb6 | 476 | out_free_buf: |
14f9a607 | 477 | dprintk("%s: Error parsing DS addr: %s\n", __func__, buf); |
16b374ca | 478 | kfree(buf); |
c9895cb6 WAA |
479 | out_free_netid: |
480 | kfree(netid); | |
16b374ca | 481 | out_err: |
14f9a607 | 482 | return NULL; |
16b374ca AA |
483 | } |
484 | ||
485 | /* Decode opaque device data and return the result */ | |
486 | static struct nfs4_file_layout_dsaddr* | |
a75b9df9 | 487 | decode_device(struct inode *ino, struct pnfs_device *pdev, gfp_t gfp_flags) |
16b374ca | 488 | { |
35124a09 | 489 | int i; |
16b374ca AA |
490 | u32 cnt, num; |
491 | u8 *indexp; | |
35124a09 WAA |
492 | __be32 *p; |
493 | u8 *stripe_indices; | |
494 | u8 max_stripe_index; | |
495 | struct nfs4_file_layout_dsaddr *dsaddr = NULL; | |
496 | struct xdr_stream stream; | |
f7da7a12 | 497 | struct xdr_buf buf; |
35124a09 | 498 | struct page *scratch; |
14f9a607 WAA |
499 | struct list_head dsaddrs; |
500 | struct nfs4_pnfs_ds_addr *da; | |
35124a09 WAA |
501 | |
502 | /* set up xdr stream */ | |
a75b9df9 | 503 | scratch = alloc_page(gfp_flags); |
35124a09 WAA |
504 | if (!scratch) |
505 | goto out_err; | |
506 | ||
f7da7a12 | 507 | xdr_init_decode_pages(&stream, &buf, pdev->pages, pdev->pglen); |
35124a09 | 508 | xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE); |
16b374ca AA |
509 | |
510 | /* Get the stripe count (number of stripe index) */ | |
35124a09 WAA |
511 | p = xdr_inline_decode(&stream, 4); |
512 | if (unlikely(!p)) | |
513 | goto out_err_free_scratch; | |
514 | ||
515 | cnt = be32_to_cpup(p); | |
16b374ca AA |
516 | dprintk("%s stripe count %d\n", __func__, cnt); |
517 | if (cnt > NFS4_PNFS_MAX_STRIPE_CNT) { | |
a030889a | 518 | printk(KERN_WARNING "NFS: %s: stripe count %d greater than " |
16b374ca AA |
519 | "supported maximum %d\n", __func__, |
520 | cnt, NFS4_PNFS_MAX_STRIPE_CNT); | |
35124a09 WAA |
521 | goto out_err_free_scratch; |
522 | } | |
523 | ||
524 | /* read stripe indices */ | |
a75b9df9 | 525 | stripe_indices = kcalloc(cnt, sizeof(u8), gfp_flags); |
35124a09 WAA |
526 | if (!stripe_indices) |
527 | goto out_err_free_scratch; | |
528 | ||
529 | p = xdr_inline_decode(&stream, cnt << 2); | |
530 | if (unlikely(!p)) | |
531 | goto out_err_free_stripe_indices; | |
532 | ||
533 | indexp = &stripe_indices[0]; | |
534 | max_stripe_index = 0; | |
535 | for (i = 0; i < cnt; i++) { | |
536 | *indexp = be32_to_cpup(p++); | |
537 | max_stripe_index = max(max_stripe_index, *indexp); | |
538 | indexp++; | |
16b374ca AA |
539 | } |
540 | ||
541 | /* Check the multipath list count */ | |
35124a09 WAA |
542 | p = xdr_inline_decode(&stream, 4); |
543 | if (unlikely(!p)) | |
544 | goto out_err_free_stripe_indices; | |
545 | ||
546 | num = be32_to_cpup(p); | |
16b374ca AA |
547 | dprintk("%s ds_num %u\n", __func__, num); |
548 | if (num > NFS4_PNFS_MAX_MULTI_CNT) { | |
a030889a | 549 | printk(KERN_WARNING "NFS: %s: multipath count %d greater than " |
16b374ca AA |
550 | "supported maximum %d\n", __func__, |
551 | num, NFS4_PNFS_MAX_MULTI_CNT); | |
35124a09 | 552 | goto out_err_free_stripe_indices; |
16b374ca | 553 | } |
35124a09 WAA |
554 | |
555 | /* validate stripe indices are all < num */ | |
556 | if (max_stripe_index >= num) { | |
a030889a | 557 | printk(KERN_WARNING "NFS: %s: stripe index %u >= num ds %u\n", |
35124a09 WAA |
558 | __func__, max_stripe_index, num); |
559 | goto out_err_free_stripe_indices; | |
560 | } | |
561 | ||
16b374ca AA |
562 | dsaddr = kzalloc(sizeof(*dsaddr) + |
563 | (sizeof(struct nfs4_pnfs_ds *) * (num - 1)), | |
a75b9df9 | 564 | gfp_flags); |
16b374ca | 565 | if (!dsaddr) |
35124a09 | 566 | goto out_err_free_stripe_indices; |
16b374ca AA |
567 | |
568 | dsaddr->stripe_count = cnt; | |
35124a09 WAA |
569 | dsaddr->stripe_indices = stripe_indices; |
570 | stripe_indices = NULL; | |
16b374ca | 571 | dsaddr->ds_num = num; |
1775bc34 BH |
572 | nfs4_init_deviceid_node(&dsaddr->id_node, |
573 | NFS_SERVER(ino)->pnfs_curr_ld, | |
574 | NFS_SERVER(ino)->nfs_client, | |
a1eaecbc | 575 | &pdev->dev_id); |
16b374ca | 576 | |
14f9a607 WAA |
577 | INIT_LIST_HEAD(&dsaddrs); |
578 | ||
16b374ca AA |
579 | for (i = 0; i < dsaddr->ds_num; i++) { |
580 | int j; | |
35124a09 WAA |
581 | u32 mp_count; |
582 | ||
583 | p = xdr_inline_decode(&stream, 4); | |
584 | if (unlikely(!p)) | |
585 | goto out_err_free_deviceid; | |
16b374ca | 586 | |
35124a09 | 587 | mp_count = be32_to_cpup(p); /* multipath count */ |
35124a09 | 588 | for (j = 0; j < mp_count; j++) { |
73ea666c | 589 | da = decode_ds_addr(NFS_SERVER(ino)->nfs_client->cl_net, |
17094272 | 590 | &stream, gfp_flags); |
14f9a607 WAA |
591 | if (da) |
592 | list_add_tail(&da->da_node, &dsaddrs); | |
593 | } | |
594 | if (list_empty(&dsaddrs)) { | |
595 | dprintk("%s: no suitable DS addresses found\n", | |
596 | __func__); | |
597 | goto out_err_free_deviceid; | |
598 | } | |
599 | ||
600 | dsaddr->ds_list[i] = nfs4_pnfs_ds_add(&dsaddrs, gfp_flags); | |
601 | if (!dsaddr->ds_list[i]) | |
602 | goto out_err_drain_dsaddrs; | |
603 | ||
604 | /* If DS was already in cache, free ds addrs */ | |
605 | while (!list_empty(&dsaddrs)) { | |
606 | da = list_first_entry(&dsaddrs, | |
607 | struct nfs4_pnfs_ds_addr, | |
608 | da_node); | |
609 | list_del_init(&da->da_node); | |
610 | kfree(da->da_remotestr); | |
611 | kfree(da); | |
16b374ca AA |
612 | } |
613 | } | |
35124a09 WAA |
614 | |
615 | __free_page(scratch); | |
16b374ca AA |
616 | return dsaddr; |
617 | ||
14f9a607 WAA |
618 | out_err_drain_dsaddrs: |
619 | while (!list_empty(&dsaddrs)) { | |
620 | da = list_first_entry(&dsaddrs, struct nfs4_pnfs_ds_addr, | |
621 | da_node); | |
622 | list_del_init(&da->da_node); | |
623 | kfree(da->da_remotestr); | |
624 | kfree(da); | |
625 | } | |
35124a09 | 626 | out_err_free_deviceid: |
16b374ca | 627 | nfs4_fl_free_deviceid(dsaddr); |
35124a09 WAA |
628 | /* stripe_indicies was part of dsaddr */ |
629 | goto out_err_free_scratch; | |
630 | out_err_free_stripe_indices: | |
631 | kfree(stripe_indices); | |
632 | out_err_free_scratch: | |
633 | __free_page(scratch); | |
16b374ca AA |
634 | out_err: |
635 | dprintk("%s ERROR: returning NULL\n", __func__); | |
636 | return NULL; | |
637 | } | |
638 | ||
639 | /* | |
ea8eecdd CH |
640 | * Decode the opaque device specified in 'dev' and add it to the cache of |
641 | * available devices. | |
16b374ca | 642 | */ |
ea8eecdd | 643 | static struct nfs4_file_layout_dsaddr * |
a75b9df9 | 644 | decode_and_add_device(struct inode *inode, struct pnfs_device *dev, gfp_t gfp_flags) |
16b374ca | 645 | { |
a1eaecbc BH |
646 | struct nfs4_deviceid_node *d; |
647 | struct nfs4_file_layout_dsaddr *n, *new; | |
16b374ca | 648 | |
a75b9df9 | 649 | new = decode_device(inode, dev, gfp_flags); |
ea8eecdd | 650 | if (!new) { |
a030889a | 651 | printk(KERN_WARNING "NFS: %s: Could not decode or add device\n", |
16b374ca AA |
652 | __func__); |
653 | return NULL; | |
654 | } | |
655 | ||
a1eaecbc BH |
656 | d = nfs4_insert_deviceid_node(&new->id_node); |
657 | n = container_of(d, struct nfs4_file_layout_dsaddr, id_node); | |
658 | if (n != new) { | |
ea8eecdd | 659 | nfs4_fl_free_deviceid(new); |
a1eaecbc | 660 | return n; |
ea8eecdd CH |
661 | } |
662 | ||
ea8eecdd | 663 | return new; |
16b374ca AA |
664 | } |
665 | ||
666 | /* | |
667 | * Retrieve the information for dev_id, add it to the list | |
668 | * of available devices, and return it. | |
669 | */ | |
670 | struct nfs4_file_layout_dsaddr * | |
78e4e05c | 671 | filelayout_get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id, gfp_t gfp_flags) |
16b374ca AA |
672 | { |
673 | struct pnfs_device *pdev = NULL; | |
674 | u32 max_resp_sz; | |
675 | int max_pages; | |
676 | struct page **pages = NULL; | |
677 | struct nfs4_file_layout_dsaddr *dsaddr = NULL; | |
678 | int rc, i; | |
679 | struct nfs_server *server = NFS_SERVER(inode); | |
680 | ||
681 | /* | |
682 | * Use the session max response size as the basis for setting | |
683 | * GETDEVICEINFO's maxcount | |
684 | */ | |
685 | max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz; | |
e5265a0c | 686 | max_pages = nfs_page_array_len(0, max_resp_sz); |
16b374ca AA |
687 | dprintk("%s inode %p max_resp_sz %u max_pages %d\n", |
688 | __func__, inode, max_resp_sz, max_pages); | |
689 | ||
a75b9df9 | 690 | pdev = kzalloc(sizeof(struct pnfs_device), gfp_flags); |
16b374ca AA |
691 | if (pdev == NULL) |
692 | return NULL; | |
693 | ||
a75b9df9 | 694 | pages = kzalloc(max_pages * sizeof(struct page *), gfp_flags); |
16b374ca AA |
695 | if (pages == NULL) { |
696 | kfree(pdev); | |
697 | return NULL; | |
698 | } | |
699 | for (i = 0; i < max_pages; i++) { | |
a75b9df9 | 700 | pages[i] = alloc_page(gfp_flags); |
16b374ca AA |
701 | if (!pages[i]) |
702 | goto out_free; | |
703 | } | |
704 | ||
16b374ca AA |
705 | memcpy(&pdev->dev_id, dev_id, sizeof(*dev_id)); |
706 | pdev->layout_type = LAYOUT_NFSV4_1_FILES; | |
707 | pdev->pages = pages; | |
708 | pdev->pgbase = 0; | |
05bf14ad | 709 | pdev->pglen = max_resp_sz; |
16b374ca AA |
710 | pdev->mincount = 0; |
711 | ||
712 | rc = nfs4_proc_getdeviceinfo(server, pdev); | |
713 | dprintk("%s getdevice info returns %d\n", __func__, rc); | |
714 | if (rc) | |
715 | goto out_free; | |
716 | ||
717 | /* | |
718 | * Found new device, need to decode it and then add it to the | |
719 | * list of known devices for this mountpoint. | |
720 | */ | |
a75b9df9 | 721 | dsaddr = decode_and_add_device(inode, pdev, gfp_flags); |
16b374ca | 722 | out_free: |
16b374ca AA |
723 | for (i = 0; i < max_pages; i++) |
724 | __free_page(pages[i]); | |
725 | kfree(pages); | |
726 | kfree(pdev); | |
727 | dprintk("<-- %s dsaddr %p\n", __func__, dsaddr); | |
728 | return dsaddr; | |
729 | } | |
730 | ||
ea8eecdd CH |
731 | void |
732 | nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr) | |
16b374ca | 733 | { |
1775bc34 | 734 | nfs4_put_deviceid_node(&dsaddr->id_node); |
16b374ca | 735 | } |
cfe7f412 FI |
736 | |
737 | /* | |
738 | * Want res = (offset - layout->pattern_offset)/ layout->stripe_unit | |
739 | * Then: ((res + fsi) % dsaddr->stripe_count) | |
740 | */ | |
741 | u32 | |
742 | nfs4_fl_calc_j_index(struct pnfs_layout_segment *lseg, loff_t offset) | |
743 | { | |
744 | struct nfs4_filelayout_segment *flseg = FILELAYOUT_LSEG(lseg); | |
745 | u64 tmp; | |
746 | ||
747 | tmp = offset - flseg->pattern_offset; | |
748 | do_div(tmp, flseg->stripe_unit); | |
749 | tmp += flseg->first_stripe_index; | |
750 | return do_div(tmp, flseg->dsaddr->stripe_count); | |
751 | } | |
752 | ||
753 | u32 | |
754 | nfs4_fl_calc_ds_index(struct pnfs_layout_segment *lseg, u32 j) | |
755 | { | |
756 | return FILELAYOUT_LSEG(lseg)->dsaddr->stripe_indices[j]; | |
757 | } | |
758 | ||
759 | struct nfs_fh * | |
760 | nfs4_fl_select_ds_fh(struct pnfs_layout_segment *lseg, u32 j) | |
761 | { | |
762 | struct nfs4_filelayout_segment *flseg = FILELAYOUT_LSEG(lseg); | |
763 | u32 i; | |
764 | ||
765 | if (flseg->stripe_type == STRIPE_SPARSE) { | |
766 | if (flseg->num_fh == 1) | |
767 | i = 0; | |
768 | else if (flseg->num_fh == 0) | |
769 | /* Use the MDS OPEN fh set in nfs_read_rpcsetup */ | |
770 | return NULL; | |
771 | else | |
772 | i = nfs4_fl_calc_ds_index(lseg, j); | |
773 | } else | |
774 | i = j; | |
775 | return flseg->fh_array[i]; | |
776 | } | |
777 | ||
778 | struct nfs4_pnfs_ds * | |
779 | nfs4_fl_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx) | |
780 | { | |
781 | struct nfs4_file_layout_dsaddr *dsaddr = FILELAYOUT_LSEG(lseg)->dsaddr; | |
782 | struct nfs4_pnfs_ds *ds = dsaddr->ds_list[ds_idx]; | |
554d458d AA |
783 | struct nfs4_deviceid_node *devid = FILELAYOUT_DEVID_NODE(lseg); |
784 | ||
1dfed273 | 785 | if (filelayout_test_devid_unavailable(devid)) |
554d458d | 786 | return NULL; |
cfe7f412 FI |
787 | |
788 | if (ds == NULL) { | |
a030889a | 789 | printk(KERN_ERR "NFS: %s: No data server for offset index %d\n", |
cfe7f412 | 790 | __func__, ds_idx); |
1dfed273 TM |
791 | filelayout_mark_devid_invalid(devid); |
792 | return NULL; | |
cfe7f412 FI |
793 | } |
794 | ||
795 | if (!ds->ds_clp) { | |
568e8c49 | 796 | struct nfs_server *s = NFS_SERVER(lseg->pls_layout->plh_inode); |
cfe7f412 FI |
797 | int err; |
798 | ||
568e8c49 | 799 | err = nfs4_ds_connect(s, ds); |
1dfed273 TM |
800 | if (err) { |
801 | nfs4_mark_deviceid_unavailable(devid); | |
802 | return NULL; | |
803 | } | |
cfe7f412 FI |
804 | } |
805 | return ds; | |
806 | } | |
98fc685a AA |
807 | |
808 | module_param(dataserver_retrans, uint, 0644); | |
809 | MODULE_PARM_DESC(dataserver_retrans, "The number of times the NFSv4.1 client " | |
810 | "retries a request before it attempts further " | |
811 | " recovery action."); | |
812 | module_param(dataserver_timeo, uint, 0644); | |
813 | MODULE_PARM_DESC(dataserver_timeo, "The time (in tenths of a second) the " | |
814 | "NFSv4.1 client waits for a response from a " | |
815 | " data server before it retries an NFS request."); |