xref: /linux/fs/nfs/pnfs_nfs.c (revision 0da85d1e38b4c186d2248f0239b72d465f5b406c)
1 /*
2  * Common NFS I/O  operations for the pnfs file based
3  * layout drivers.
4  *
5  * Copyright (c) 2014, Primary Data, Inc. All rights reserved.
6  *
7  * Tom Haynes <loghyr@primarydata.com>
8  */
9 
10 #include <linux/nfs_fs.h>
11 #include <linux/nfs_page.h>
12 #include <linux/sunrpc/addr.h>
13 #include <linux/module.h>
14 
15 #include "nfs4session.h"
16 #include "internal.h"
17 #include "pnfs.h"
18 
19 #define NFSDBG_FACILITY		NFSDBG_PNFS
20 
21 void pnfs_generic_rw_release(void *data)
22 {
23 	struct nfs_pgio_header *hdr = data;
24 
25 	nfs_put_client(hdr->ds_clp);
26 	hdr->mds_ops->rpc_release(data);
27 }
28 EXPORT_SYMBOL_GPL(pnfs_generic_rw_release);
29 
30 /* Fake up some data that will cause nfs_commit_release to retry the writes. */
31 void pnfs_generic_prepare_to_resend_writes(struct nfs_commit_data *data)
32 {
33 	struct nfs_page *first = nfs_list_entry(data->pages.next);
34 
35 	data->task.tk_status = 0;
36 	memcpy(&data->verf.verifier, &first->wb_verf,
37 	       sizeof(data->verf.verifier));
38 	data->verf.verifier.data[0]++; /* ensure verifier mismatch */
39 }
40 EXPORT_SYMBOL_GPL(pnfs_generic_prepare_to_resend_writes);
41 
42 void pnfs_generic_write_commit_done(struct rpc_task *task, void *data)
43 {
44 	struct nfs_commit_data *wdata = data;
45 
46 	/* Note this may cause RPC to be resent */
47 	wdata->mds_ops->rpc_call_done(task, data);
48 }
49 EXPORT_SYMBOL_GPL(pnfs_generic_write_commit_done);
50 
51 void pnfs_generic_commit_release(void *calldata)
52 {
53 	struct nfs_commit_data *data = calldata;
54 
55 	data->completion_ops->completion(data);
56 	pnfs_put_lseg(data->lseg);
57 	nfs_put_client(data->ds_clp);
58 	nfs_commitdata_release(data);
59 }
60 EXPORT_SYMBOL_GPL(pnfs_generic_commit_release);
61 
62 /* The generic layer is about to remove the req from the commit list.
63  * If this will make the bucket empty, it will need to put the lseg reference.
64  * Note this must be called holding the inode (/cinfo) lock
65  */
66 void
67 pnfs_generic_clear_request_commit(struct nfs_page *req,
68 				  struct nfs_commit_info *cinfo)
69 {
70 	struct pnfs_layout_segment *freeme = NULL;
71 
72 	if (!test_and_clear_bit(PG_COMMIT_TO_DS, &req->wb_flags))
73 		goto out;
74 	cinfo->ds->nwritten--;
75 	if (list_is_singular(&req->wb_list)) {
76 		struct pnfs_commit_bucket *bucket;
77 
78 		bucket = list_first_entry(&req->wb_list,
79 					  struct pnfs_commit_bucket,
80 					  written);
81 		freeme = bucket->wlseg;
82 		bucket->wlseg = NULL;
83 	}
84 out:
85 	nfs_request_remove_commit_list(req, cinfo);
86 	pnfs_put_lseg_locked(freeme);
87 }
88 EXPORT_SYMBOL_GPL(pnfs_generic_clear_request_commit);
89 
90 static int
91 pnfs_generic_transfer_commit_list(struct list_head *src, struct list_head *dst,
92 				  struct nfs_commit_info *cinfo, int max)
93 {
94 	struct nfs_page *req, *tmp;
95 	int ret = 0;
96 
97 	list_for_each_entry_safe(req, tmp, src, wb_list) {
98 		if (!nfs_lock_request(req))
99 			continue;
100 		kref_get(&req->wb_kref);
101 		if (cond_resched_lock(cinfo->lock))
102 			list_safe_reset_next(req, tmp, wb_list);
103 		nfs_request_remove_commit_list(req, cinfo);
104 		clear_bit(PG_COMMIT_TO_DS, &req->wb_flags);
105 		nfs_list_add_request(req, dst);
106 		ret++;
107 		if ((ret == max) && !cinfo->dreq)
108 			break;
109 	}
110 	return ret;
111 }
112 
113 static int
114 pnfs_generic_scan_ds_commit_list(struct pnfs_commit_bucket *bucket,
115 				 struct nfs_commit_info *cinfo,
116 				 int max)
117 {
118 	struct list_head *src = &bucket->written;
119 	struct list_head *dst = &bucket->committing;
120 	int ret;
121 
122 	lockdep_assert_held(cinfo->lock);
123 	ret = pnfs_generic_transfer_commit_list(src, dst, cinfo, max);
124 	if (ret) {
125 		cinfo->ds->nwritten -= ret;
126 		cinfo->ds->ncommitting += ret;
127 		bucket->clseg = bucket->wlseg;
128 		if (list_empty(src))
129 			bucket->wlseg = NULL;
130 		else
131 			pnfs_get_lseg(bucket->clseg);
132 	}
133 	return ret;
134 }
135 
136 /* Move reqs from written to committing lists, returning count
137  * of number moved.
138  */
139 int pnfs_generic_scan_commit_lists(struct nfs_commit_info *cinfo,
140 				   int max)
141 {
142 	int i, rv = 0, cnt;
143 
144 	lockdep_assert_held(cinfo->lock);
145 	for (i = 0; i < cinfo->ds->nbuckets && max != 0; i++) {
146 		cnt = pnfs_generic_scan_ds_commit_list(&cinfo->ds->buckets[i],
147 						       cinfo, max);
148 		max -= cnt;
149 		rv += cnt;
150 	}
151 	return rv;
152 }
153 EXPORT_SYMBOL_GPL(pnfs_generic_scan_commit_lists);
154 
155 /* Pull everything off the committing lists and dump into @dst.  */
156 void pnfs_generic_recover_commit_reqs(struct list_head *dst,
157 				      struct nfs_commit_info *cinfo)
158 {
159 	struct pnfs_commit_bucket *b;
160 	struct pnfs_layout_segment *freeme;
161 	int i;
162 
163 	lockdep_assert_held(cinfo->lock);
164 restart:
165 	for (i = 0, b = cinfo->ds->buckets; i < cinfo->ds->nbuckets; i++, b++) {
166 		if (pnfs_generic_transfer_commit_list(&b->written, dst,
167 						      cinfo, 0)) {
168 			freeme = b->wlseg;
169 			b->wlseg = NULL;
170 			spin_unlock(cinfo->lock);
171 			pnfs_put_lseg(freeme);
172 			spin_lock(cinfo->lock);
173 			goto restart;
174 		}
175 	}
176 	cinfo->ds->nwritten = 0;
177 }
178 EXPORT_SYMBOL_GPL(pnfs_generic_recover_commit_reqs);
179 
180 static void pnfs_generic_retry_commit(struct nfs_commit_info *cinfo, int idx)
181 {
182 	struct pnfs_ds_commit_info *fl_cinfo = cinfo->ds;
183 	struct pnfs_commit_bucket *bucket;
184 	struct pnfs_layout_segment *freeme;
185 	int i;
186 
187 	for (i = idx; i < fl_cinfo->nbuckets; i++) {
188 		bucket = &fl_cinfo->buckets[i];
189 		if (list_empty(&bucket->committing))
190 			continue;
191 		nfs_retry_commit(&bucket->committing, bucket->clseg, cinfo, i);
192 		spin_lock(cinfo->lock);
193 		freeme = bucket->clseg;
194 		bucket->clseg = NULL;
195 		spin_unlock(cinfo->lock);
196 		pnfs_put_lseg(freeme);
197 	}
198 }
199 
200 static unsigned int
201 pnfs_generic_alloc_ds_commits(struct nfs_commit_info *cinfo,
202 			      struct list_head *list)
203 {
204 	struct pnfs_ds_commit_info *fl_cinfo;
205 	struct pnfs_commit_bucket *bucket;
206 	struct nfs_commit_data *data;
207 	int i;
208 	unsigned int nreq = 0;
209 
210 	fl_cinfo = cinfo->ds;
211 	bucket = fl_cinfo->buckets;
212 	for (i = 0; i < fl_cinfo->nbuckets; i++, bucket++) {
213 		if (list_empty(&bucket->committing))
214 			continue;
215 		data = nfs_commitdata_alloc();
216 		if (!data)
217 			break;
218 		data->ds_commit_index = i;
219 		spin_lock(cinfo->lock);
220 		data->lseg = bucket->clseg;
221 		bucket->clseg = NULL;
222 		spin_unlock(cinfo->lock);
223 		list_add(&data->pages, list);
224 		nreq++;
225 	}
226 
227 	/* Clean up on error */
228 	pnfs_generic_retry_commit(cinfo, i);
229 	return nreq;
230 }
231 
232 /* This follows nfs_commit_list pretty closely */
233 int
234 pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages,
235 			     int how, struct nfs_commit_info *cinfo,
236 			     int (*initiate_commit)(struct nfs_commit_data *data,
237 						    int how))
238 {
239 	struct nfs_commit_data *data, *tmp;
240 	LIST_HEAD(list);
241 	unsigned int nreq = 0;
242 
243 	if (!list_empty(mds_pages)) {
244 		data = nfs_commitdata_alloc();
245 		if (data != NULL) {
246 			data->lseg = NULL;
247 			list_add(&data->pages, &list);
248 			nreq++;
249 		} else {
250 			nfs_retry_commit(mds_pages, NULL, cinfo, 0);
251 			pnfs_generic_retry_commit(cinfo, 0);
252 			cinfo->completion_ops->error_cleanup(NFS_I(inode));
253 			return -ENOMEM;
254 		}
255 	}
256 
257 	nreq += pnfs_generic_alloc_ds_commits(cinfo, &list);
258 
259 	if (nreq == 0) {
260 		cinfo->completion_ops->error_cleanup(NFS_I(inode));
261 		goto out;
262 	}
263 
264 	atomic_add(nreq, &cinfo->mds->rpcs_out);
265 
266 	list_for_each_entry_safe(data, tmp, &list, pages) {
267 		list_del_init(&data->pages);
268 		if (!data->lseg) {
269 			nfs_init_commit(data, mds_pages, NULL, cinfo);
270 			nfs_initiate_commit(NFS_CLIENT(inode), data,
271 					    NFS_PROTO(data->inode),
272 					    data->mds_ops, how, 0);
273 		} else {
274 			struct pnfs_commit_bucket *buckets;
275 
276 			buckets = cinfo->ds->buckets;
277 			nfs_init_commit(data,
278 					&buckets[data->ds_commit_index].committing,
279 					data->lseg,
280 					cinfo);
281 			initiate_commit(data, how);
282 		}
283 	}
284 out:
285 	cinfo->ds->ncommitting = 0;
286 	return PNFS_ATTEMPTED;
287 }
288 EXPORT_SYMBOL_GPL(pnfs_generic_commit_pagelist);
289 
290 /*
291  * Data server cache
292  *
293  * Data servers can be mapped to different device ids.
294  * nfs4_pnfs_ds reference counting
295  *   - set to 1 on allocation
296  *   - incremented when a device id maps a data server already in the cache.
297  *   - decremented when deviceid is removed from the cache.
298  */
299 static DEFINE_SPINLOCK(nfs4_ds_cache_lock);
300 static LIST_HEAD(nfs4_data_server_cache);
301 
302 /* Debug routines */
303 static void
304 print_ds(struct nfs4_pnfs_ds *ds)
305 {
306 	if (ds == NULL) {
307 		printk(KERN_WARNING "%s NULL device\n", __func__);
308 		return;
309 	}
310 	printk(KERN_WARNING "        ds %s\n"
311 		"        ref count %d\n"
312 		"        client %p\n"
313 		"        cl_exchange_flags %x\n",
314 		ds->ds_remotestr,
315 		atomic_read(&ds->ds_count), ds->ds_clp,
316 		ds->ds_clp ? ds->ds_clp->cl_exchange_flags : 0);
317 }
318 
319 static bool
320 same_sockaddr(struct sockaddr *addr1, struct sockaddr *addr2)
321 {
322 	struct sockaddr_in *a, *b;
323 	struct sockaddr_in6 *a6, *b6;
324 
325 	if (addr1->sa_family != addr2->sa_family)
326 		return false;
327 
328 	switch (addr1->sa_family) {
329 	case AF_INET:
330 		a = (struct sockaddr_in *)addr1;
331 		b = (struct sockaddr_in *)addr2;
332 
333 		if (a->sin_addr.s_addr == b->sin_addr.s_addr &&
334 		    a->sin_port == b->sin_port)
335 			return true;
336 		break;
337 
338 	case AF_INET6:
339 		a6 = (struct sockaddr_in6 *)addr1;
340 		b6 = (struct sockaddr_in6 *)addr2;
341 
342 		/* LINKLOCAL addresses must have matching scope_id */
343 		if (ipv6_addr_src_scope(&a6->sin6_addr) ==
344 		    IPV6_ADDR_SCOPE_LINKLOCAL &&
345 		    a6->sin6_scope_id != b6->sin6_scope_id)
346 			return false;
347 
348 		if (ipv6_addr_equal(&a6->sin6_addr, &b6->sin6_addr) &&
349 		    a6->sin6_port == b6->sin6_port)
350 			return true;
351 		break;
352 
353 	default:
354 		dprintk("%s: unhandled address family: %u\n",
355 			__func__, addr1->sa_family);
356 		return false;
357 	}
358 
359 	return false;
360 }
361 
362 static bool
363 _same_data_server_addrs_locked(const struct list_head *dsaddrs1,
364 			       const struct list_head *dsaddrs2)
365 {
366 	struct nfs4_pnfs_ds_addr *da1, *da2;
367 
368 	/* step through both lists, comparing as we go */
369 	for (da1 = list_first_entry(dsaddrs1, typeof(*da1), da_node),
370 	     da2 = list_first_entry(dsaddrs2, typeof(*da2), da_node);
371 	     da1 != NULL && da2 != NULL;
372 	     da1 = list_entry(da1->da_node.next, typeof(*da1), da_node),
373 	     da2 = list_entry(da2->da_node.next, typeof(*da2), da_node)) {
374 		if (!same_sockaddr((struct sockaddr *)&da1->da_addr,
375 				   (struct sockaddr *)&da2->da_addr))
376 			return false;
377 	}
378 	if (da1 == NULL && da2 == NULL)
379 		return true;
380 
381 	return false;
382 }
383 
384 /*
385  * Lookup DS by addresses.  nfs4_ds_cache_lock is held
386  */
387 static struct nfs4_pnfs_ds *
388 _data_server_lookup_locked(const struct list_head *dsaddrs)
389 {
390 	struct nfs4_pnfs_ds *ds;
391 
392 	list_for_each_entry(ds, &nfs4_data_server_cache, ds_node)
393 		if (_same_data_server_addrs_locked(&ds->ds_addrs, dsaddrs))
394 			return ds;
395 	return NULL;
396 }
397 
398 static void destroy_ds(struct nfs4_pnfs_ds *ds)
399 {
400 	struct nfs4_pnfs_ds_addr *da;
401 
402 	dprintk("--> %s\n", __func__);
403 	ifdebug(FACILITY)
404 		print_ds(ds);
405 
406 	nfs_put_client(ds->ds_clp);
407 
408 	while (!list_empty(&ds->ds_addrs)) {
409 		da = list_first_entry(&ds->ds_addrs,
410 				      struct nfs4_pnfs_ds_addr,
411 				      da_node);
412 		list_del_init(&da->da_node);
413 		kfree(da->da_remotestr);
414 		kfree(da);
415 	}
416 
417 	kfree(ds->ds_remotestr);
418 	kfree(ds);
419 }
420 
421 void nfs4_pnfs_ds_put(struct nfs4_pnfs_ds *ds)
422 {
423 	if (atomic_dec_and_lock(&ds->ds_count,
424 				&nfs4_ds_cache_lock)) {
425 		list_del_init(&ds->ds_node);
426 		spin_unlock(&nfs4_ds_cache_lock);
427 		destroy_ds(ds);
428 	}
429 }
430 EXPORT_SYMBOL_GPL(nfs4_pnfs_ds_put);
431 
432 /*
433  * Create a string with a human readable address and port to avoid
434  * complicated setup around many dprinks.
435  */
436 static char *
437 nfs4_pnfs_remotestr(struct list_head *dsaddrs, gfp_t gfp_flags)
438 {
439 	struct nfs4_pnfs_ds_addr *da;
440 	char *remotestr;
441 	size_t len;
442 	char *p;
443 
444 	len = 3;        /* '{', '}' and eol */
445 	list_for_each_entry(da, dsaddrs, da_node) {
446 		len += strlen(da->da_remotestr) + 1;    /* string plus comma */
447 	}
448 
449 	remotestr = kzalloc(len, gfp_flags);
450 	if (!remotestr)
451 		return NULL;
452 
453 	p = remotestr;
454 	*(p++) = '{';
455 	len--;
456 	list_for_each_entry(da, dsaddrs, da_node) {
457 		size_t ll = strlen(da->da_remotestr);
458 
459 		if (ll > len)
460 			goto out_err;
461 
462 		memcpy(p, da->da_remotestr, ll);
463 		p += ll;
464 		len -= ll;
465 
466 		if (len < 1)
467 			goto out_err;
468 		(*p++) = ',';
469 		len--;
470 	}
471 	if (len < 2)
472 		goto out_err;
473 	*(p++) = '}';
474 	*p = '\0';
475 	return remotestr;
476 out_err:
477 	kfree(remotestr);
478 	return NULL;
479 }
480 
481 /*
482  * Given a list of multipath struct nfs4_pnfs_ds_addr, add it to ds cache if
483  * uncached and return cached struct nfs4_pnfs_ds.
484  */
485 struct nfs4_pnfs_ds *
486 nfs4_pnfs_ds_add(struct list_head *dsaddrs, gfp_t gfp_flags)
487 {
488 	struct nfs4_pnfs_ds *tmp_ds, *ds = NULL;
489 	char *remotestr;
490 
491 	if (list_empty(dsaddrs)) {
492 		dprintk("%s: no addresses defined\n", __func__);
493 		goto out;
494 	}
495 
496 	ds = kzalloc(sizeof(*ds), gfp_flags);
497 	if (!ds)
498 		goto out;
499 
500 	/* this is only used for debugging, so it's ok if its NULL */
501 	remotestr = nfs4_pnfs_remotestr(dsaddrs, gfp_flags);
502 
503 	spin_lock(&nfs4_ds_cache_lock);
504 	tmp_ds = _data_server_lookup_locked(dsaddrs);
505 	if (tmp_ds == NULL) {
506 		INIT_LIST_HEAD(&ds->ds_addrs);
507 		list_splice_init(dsaddrs, &ds->ds_addrs);
508 		ds->ds_remotestr = remotestr;
509 		atomic_set(&ds->ds_count, 1);
510 		INIT_LIST_HEAD(&ds->ds_node);
511 		ds->ds_clp = NULL;
512 		list_add(&ds->ds_node, &nfs4_data_server_cache);
513 		dprintk("%s add new data server %s\n", __func__,
514 			ds->ds_remotestr);
515 	} else {
516 		kfree(remotestr);
517 		kfree(ds);
518 		atomic_inc(&tmp_ds->ds_count);
519 		dprintk("%s data server %s found, inc'ed ds_count to %d\n",
520 			__func__, tmp_ds->ds_remotestr,
521 			atomic_read(&tmp_ds->ds_count));
522 		ds = tmp_ds;
523 	}
524 	spin_unlock(&nfs4_ds_cache_lock);
525 out:
526 	return ds;
527 }
528 EXPORT_SYMBOL_GPL(nfs4_pnfs_ds_add);
529 
530 static void nfs4_wait_ds_connect(struct nfs4_pnfs_ds *ds)
531 {
532 	might_sleep();
533 	wait_on_bit(&ds->ds_state, NFS4DS_CONNECTING,
534 			TASK_KILLABLE);
535 }
536 
537 static void nfs4_clear_ds_conn_bit(struct nfs4_pnfs_ds *ds)
538 {
539 	smp_mb__before_atomic();
540 	clear_bit(NFS4DS_CONNECTING, &ds->ds_state);
541 	smp_mb__after_atomic();
542 	wake_up_bit(&ds->ds_state, NFS4DS_CONNECTING);
543 }
544 
545 static struct nfs_client *(*get_v3_ds_connect)(
546 			struct nfs_client *mds_clp,
547 			const struct sockaddr *ds_addr,
548 			int ds_addrlen,
549 			int ds_proto,
550 			unsigned int ds_timeo,
551 			unsigned int ds_retrans,
552 			rpc_authflavor_t au_flavor);
553 
554 static bool load_v3_ds_connect(void)
555 {
556 	if (!get_v3_ds_connect) {
557 		get_v3_ds_connect = symbol_request(nfs3_set_ds_client);
558 		WARN_ON_ONCE(!get_v3_ds_connect);
559 	}
560 
561 	return(get_v3_ds_connect != NULL);
562 }
563 
564 void nfs4_pnfs_v3_ds_connect_unload(void)
565 {
566 	if (get_v3_ds_connect) {
567 		symbol_put(nfs3_set_ds_client);
568 		get_v3_ds_connect = NULL;
569 	}
570 }
571 EXPORT_SYMBOL_GPL(nfs4_pnfs_v3_ds_connect_unload);
572 
573 static int _nfs4_pnfs_v3_ds_connect(struct nfs_server *mds_srv,
574 				 struct nfs4_pnfs_ds *ds,
575 				 unsigned int timeo,
576 				 unsigned int retrans,
577 				 rpc_authflavor_t au_flavor)
578 {
579 	struct nfs_client *clp = ERR_PTR(-EIO);
580 	struct nfs4_pnfs_ds_addr *da;
581 	int status = 0;
582 
583 	dprintk("--> %s DS %s au_flavor %d\n", __func__,
584 		ds->ds_remotestr, au_flavor);
585 
586 	if (!load_v3_ds_connect())
587 		goto out;
588 
589 	list_for_each_entry(da, &ds->ds_addrs, da_node) {
590 		dprintk("%s: DS %s: trying address %s\n",
591 			__func__, ds->ds_remotestr, da->da_remotestr);
592 
593 		clp = get_v3_ds_connect(mds_srv->nfs_client,
594 					(struct sockaddr *)&da->da_addr,
595 					da->da_addrlen, IPPROTO_TCP,
596 					timeo, retrans, au_flavor);
597 		if (!IS_ERR(clp))
598 			break;
599 	}
600 
601 	if (IS_ERR(clp)) {
602 		status = PTR_ERR(clp);
603 		goto out;
604 	}
605 
606 	smp_wmb();
607 	ds->ds_clp = clp;
608 	dprintk("%s [new] addr: %s\n", __func__, ds->ds_remotestr);
609 out:
610 	return status;
611 }
612 
613 static int _nfs4_pnfs_v4_ds_connect(struct nfs_server *mds_srv,
614 				 struct nfs4_pnfs_ds *ds,
615 				 unsigned int timeo,
616 				 unsigned int retrans,
617 				 u32 minor_version,
618 				 rpc_authflavor_t au_flavor)
619 {
620 	struct nfs_client *clp = ERR_PTR(-EIO);
621 	struct nfs4_pnfs_ds_addr *da;
622 	int status = 0;
623 
624 	dprintk("--> %s DS %s au_flavor %d\n", __func__, ds->ds_remotestr,
625 		au_flavor);
626 
627 	list_for_each_entry(da, &ds->ds_addrs, da_node) {
628 		dprintk("%s: DS %s: trying address %s\n",
629 			__func__, ds->ds_remotestr, da->da_remotestr);
630 
631 		clp = nfs4_set_ds_client(mds_srv->nfs_client,
632 					(struct sockaddr *)&da->da_addr,
633 					da->da_addrlen, IPPROTO_TCP,
634 					timeo, retrans, minor_version,
635 					au_flavor);
636 		if (!IS_ERR(clp))
637 			break;
638 	}
639 
640 	if (IS_ERR(clp)) {
641 		status = PTR_ERR(clp);
642 		goto out;
643 	}
644 
645 	status = nfs4_init_ds_session(clp, mds_srv->nfs_client->cl_lease_time);
646 	if (status)
647 		goto out_put;
648 
649 	smp_wmb();
650 	ds->ds_clp = clp;
651 	dprintk("%s [new] addr: %s\n", __func__, ds->ds_remotestr);
652 out:
653 	return status;
654 out_put:
655 	nfs_put_client(clp);
656 	goto out;
657 }
658 
659 /*
660  * Create an rpc connection to the nfs4_pnfs_ds data server.
661  * Currently only supports IPv4 and IPv6 addresses.
662  * If connection fails, make devid unavailable.
663  */
664 void nfs4_pnfs_ds_connect(struct nfs_server *mds_srv, struct nfs4_pnfs_ds *ds,
665 			  struct nfs4_deviceid_node *devid, unsigned int timeo,
666 			  unsigned int retrans, u32 version,
667 			  u32 minor_version, rpc_authflavor_t au_flavor)
668 {
669 	if (test_and_set_bit(NFS4DS_CONNECTING, &ds->ds_state) == 0) {
670 		int err = 0;
671 
672 		if (version == 3) {
673 			err = _nfs4_pnfs_v3_ds_connect(mds_srv, ds, timeo,
674 						       retrans, au_flavor);
675 		} else if (version == 4) {
676 			err = _nfs4_pnfs_v4_ds_connect(mds_srv, ds, timeo,
677 						       retrans, minor_version,
678 						       au_flavor);
679 		} else {
680 			dprintk("%s: unsupported DS version %d\n", __func__,
681 				version);
682 			err = -EPROTONOSUPPORT;
683 		}
684 
685 		if (err)
686 			nfs4_mark_deviceid_unavailable(devid);
687 		nfs4_clear_ds_conn_bit(ds);
688 	} else {
689 		nfs4_wait_ds_connect(ds);
690 	}
691 }
692 EXPORT_SYMBOL_GPL(nfs4_pnfs_ds_connect);
693 
694 /*
695  * Currently only supports ipv4, ipv6 and one multi-path address.
696  */
697 struct nfs4_pnfs_ds_addr *
698 nfs4_decode_mp_ds_addr(struct net *net, struct xdr_stream *xdr, gfp_t gfp_flags)
699 {
700 	struct nfs4_pnfs_ds_addr *da = NULL;
701 	char *buf, *portstr;
702 	__be16 port;
703 	int nlen, rlen;
704 	int tmp[2];
705 	__be32 *p;
706 	char *netid, *match_netid;
707 	size_t len, match_netid_len;
708 	char *startsep = "";
709 	char *endsep = "";
710 
711 
712 	/* r_netid */
713 	p = xdr_inline_decode(xdr, 4);
714 	if (unlikely(!p))
715 		goto out_err;
716 	nlen = be32_to_cpup(p++);
717 
718 	p = xdr_inline_decode(xdr, nlen);
719 	if (unlikely(!p))
720 		goto out_err;
721 
722 	netid = kmalloc(nlen+1, gfp_flags);
723 	if (unlikely(!netid))
724 		goto out_err;
725 
726 	netid[nlen] = '\0';
727 	memcpy(netid, p, nlen);
728 
729 	/* r_addr: ip/ip6addr with port in dec octets - see RFC 5665 */
730 	p = xdr_inline_decode(xdr, 4);
731 	if (unlikely(!p))
732 		goto out_free_netid;
733 	rlen = be32_to_cpup(p);
734 
735 	p = xdr_inline_decode(xdr, rlen);
736 	if (unlikely(!p))
737 		goto out_free_netid;
738 
739 	/* port is ".ABC.DEF", 8 chars max */
740 	if (rlen > INET6_ADDRSTRLEN + IPV6_SCOPE_ID_LEN + 8) {
741 		dprintk("%s: Invalid address, length %d\n", __func__,
742 			rlen);
743 		goto out_free_netid;
744 	}
745 	buf = kmalloc(rlen + 1, gfp_flags);
746 	if (!buf) {
747 		dprintk("%s: Not enough memory\n", __func__);
748 		goto out_free_netid;
749 	}
750 	buf[rlen] = '\0';
751 	memcpy(buf, p, rlen);
752 
753 	/* replace port '.' with '-' */
754 	portstr = strrchr(buf, '.');
755 	if (!portstr) {
756 		dprintk("%s: Failed finding expected dot in port\n",
757 			__func__);
758 		goto out_free_buf;
759 	}
760 	*portstr = '-';
761 
762 	/* find '.' between address and port */
763 	portstr = strrchr(buf, '.');
764 	if (!portstr) {
765 		dprintk("%s: Failed finding expected dot between address and "
766 			"port\n", __func__);
767 		goto out_free_buf;
768 	}
769 	*portstr = '\0';
770 
771 	da = kzalloc(sizeof(*da), gfp_flags);
772 	if (unlikely(!da))
773 		goto out_free_buf;
774 
775 	INIT_LIST_HEAD(&da->da_node);
776 
777 	if (!rpc_pton(net, buf, portstr-buf, (struct sockaddr *)&da->da_addr,
778 		      sizeof(da->da_addr))) {
779 		dprintk("%s: error parsing address %s\n", __func__, buf);
780 		goto out_free_da;
781 	}
782 
783 	portstr++;
784 	sscanf(portstr, "%d-%d", &tmp[0], &tmp[1]);
785 	port = htons((tmp[0] << 8) | (tmp[1]));
786 
787 	switch (da->da_addr.ss_family) {
788 	case AF_INET:
789 		((struct sockaddr_in *)&da->da_addr)->sin_port = port;
790 		da->da_addrlen = sizeof(struct sockaddr_in);
791 		match_netid = "tcp";
792 		match_netid_len = 3;
793 		break;
794 
795 	case AF_INET6:
796 		((struct sockaddr_in6 *)&da->da_addr)->sin6_port = port;
797 		da->da_addrlen = sizeof(struct sockaddr_in6);
798 		match_netid = "tcp6";
799 		match_netid_len = 4;
800 		startsep = "[";
801 		endsep = "]";
802 		break;
803 
804 	default:
805 		dprintk("%s: unsupported address family: %u\n",
806 			__func__, da->da_addr.ss_family);
807 		goto out_free_da;
808 	}
809 
810 	if (nlen != match_netid_len || strncmp(netid, match_netid, nlen)) {
811 		dprintk("%s: ERROR: r_netid \"%s\" != \"%s\"\n",
812 			__func__, netid, match_netid);
813 		goto out_free_da;
814 	}
815 
816 	/* save human readable address */
817 	len = strlen(startsep) + strlen(buf) + strlen(endsep) + 7;
818 	da->da_remotestr = kzalloc(len, gfp_flags);
819 
820 	/* NULL is ok, only used for dprintk */
821 	if (da->da_remotestr)
822 		snprintf(da->da_remotestr, len, "%s%s%s:%u", startsep,
823 			 buf, endsep, ntohs(port));
824 
825 	dprintk("%s: Parsed DS addr %s\n", __func__, da->da_remotestr);
826 	kfree(buf);
827 	kfree(netid);
828 	return da;
829 
830 out_free_da:
831 	kfree(da);
832 out_free_buf:
833 	dprintk("%s: Error parsing DS addr: %s\n", __func__, buf);
834 	kfree(buf);
835 out_free_netid:
836 	kfree(netid);
837 out_err:
838 	return NULL;
839 }
840 EXPORT_SYMBOL_GPL(nfs4_decode_mp_ds_addr);
841 
842 void
843 pnfs_layout_mark_request_commit(struct nfs_page *req,
844 				struct pnfs_layout_segment *lseg,
845 				struct nfs_commit_info *cinfo,
846 				u32 ds_commit_idx)
847 {
848 	struct list_head *list;
849 	struct pnfs_commit_bucket *buckets;
850 
851 	spin_lock(cinfo->lock);
852 	buckets = cinfo->ds->buckets;
853 	list = &buckets[ds_commit_idx].written;
854 	if (list_empty(list)) {
855 		/* Non-empty buckets hold a reference on the lseg.  That ref
856 		 * is normally transferred to the COMMIT call and released
857 		 * there.  It could also be released if the last req is pulled
858 		 * off due to a rewrite, in which case it will be done in
859 		 * pnfs_common_clear_request_commit
860 		 */
861 		WARN_ON_ONCE(buckets[ds_commit_idx].wlseg != NULL);
862 		buckets[ds_commit_idx].wlseg = pnfs_get_lseg(lseg);
863 	}
864 	set_bit(PG_COMMIT_TO_DS, &req->wb_flags);
865 	cinfo->ds->nwritten++;
866 	spin_unlock(cinfo->lock);
867 
868 	nfs_request_add_commit_list(req, list, cinfo);
869 }
870 EXPORT_SYMBOL_GPL(pnfs_layout_mark_request_commit);
871 
872 int
873 pnfs_nfs_generic_sync(struct inode *inode, bool datasync)
874 {
875 	if (datasync)
876 		return 0;
877 	return pnfs_layoutcommit_inode(inode, true);
878 }
879 EXPORT_SYMBOL_GPL(pnfs_nfs_generic_sync);
880 
881