xref: /openbmc/linux/fs/nfs/pnfs_nfs.c (revision 9b9c2cd4)
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 		if (bucket->clseg == NULL)
128 			bucket->clseg = pnfs_get_lseg(bucket->wlseg);
129 		if (list_empty(src)) {
130 			pnfs_put_lseg_locked(bucket->wlseg);
131 			bucket->wlseg = NULL;
132 		}
133 	}
134 	return ret;
135 }
136 
137 /* Move reqs from written to committing lists, returning count
138  * of number moved.
139  */
140 int pnfs_generic_scan_commit_lists(struct nfs_commit_info *cinfo,
141 				   int max)
142 {
143 	int i, rv = 0, cnt;
144 
145 	lockdep_assert_held(cinfo->lock);
146 	for (i = 0; i < cinfo->ds->nbuckets && max != 0; i++) {
147 		cnt = pnfs_generic_scan_ds_commit_list(&cinfo->ds->buckets[i],
148 						       cinfo, max);
149 		max -= cnt;
150 		rv += cnt;
151 	}
152 	return rv;
153 }
154 EXPORT_SYMBOL_GPL(pnfs_generic_scan_commit_lists);
155 
156 /* Pull everything off the committing lists and dump into @dst.  */
157 void pnfs_generic_recover_commit_reqs(struct list_head *dst,
158 				      struct nfs_commit_info *cinfo)
159 {
160 	struct pnfs_commit_bucket *b;
161 	struct pnfs_layout_segment *freeme;
162 	int i;
163 
164 	lockdep_assert_held(cinfo->lock);
165 restart:
166 	for (i = 0, b = cinfo->ds->buckets; i < cinfo->ds->nbuckets; i++, b++) {
167 		if (pnfs_generic_transfer_commit_list(&b->written, dst,
168 						      cinfo, 0)) {
169 			freeme = b->wlseg;
170 			b->wlseg = NULL;
171 			spin_unlock(cinfo->lock);
172 			pnfs_put_lseg(freeme);
173 			spin_lock(cinfo->lock);
174 			goto restart;
175 		}
176 	}
177 	cinfo->ds->nwritten = 0;
178 }
179 EXPORT_SYMBOL_GPL(pnfs_generic_recover_commit_reqs);
180 
181 static void pnfs_generic_retry_commit(struct nfs_commit_info *cinfo, int idx)
182 {
183 	struct pnfs_ds_commit_info *fl_cinfo = cinfo->ds;
184 	struct pnfs_commit_bucket *bucket;
185 	struct pnfs_layout_segment *freeme;
186 	LIST_HEAD(pages);
187 	int i;
188 
189 	spin_lock(cinfo->lock);
190 	for (i = idx; i < fl_cinfo->nbuckets; i++) {
191 		bucket = &fl_cinfo->buckets[i];
192 		if (list_empty(&bucket->committing))
193 			continue;
194 		freeme = bucket->clseg;
195 		bucket->clseg = NULL;
196 		list_splice_init(&bucket->committing, &pages);
197 		spin_unlock(cinfo->lock);
198 		nfs_retry_commit(&pages, freeme, cinfo, i);
199 		pnfs_put_lseg(freeme);
200 		spin_lock(cinfo->lock);
201 	}
202 	spin_unlock(cinfo->lock);
203 }
204 
205 static unsigned int
206 pnfs_generic_alloc_ds_commits(struct nfs_commit_info *cinfo,
207 			      struct list_head *list)
208 {
209 	struct pnfs_ds_commit_info *fl_cinfo;
210 	struct pnfs_commit_bucket *bucket;
211 	struct nfs_commit_data *data;
212 	int i;
213 	unsigned int nreq = 0;
214 
215 	fl_cinfo = cinfo->ds;
216 	bucket = fl_cinfo->buckets;
217 	for (i = 0; i < fl_cinfo->nbuckets; i++, bucket++) {
218 		if (list_empty(&bucket->committing))
219 			continue;
220 		data = nfs_commitdata_alloc();
221 		if (!data)
222 			break;
223 		data->ds_commit_index = i;
224 		list_add(&data->pages, list);
225 		nreq++;
226 	}
227 
228 	/* Clean up on error */
229 	pnfs_generic_retry_commit(cinfo, i);
230 	return nreq;
231 }
232 
233 static inline
234 void pnfs_fetch_commit_bucket_list(struct list_head *pages,
235 		struct nfs_commit_data *data,
236 		struct nfs_commit_info *cinfo)
237 {
238 	struct pnfs_commit_bucket *bucket;
239 
240 	bucket = &cinfo->ds->buckets[data->ds_commit_index];
241 	spin_lock(cinfo->lock);
242 	list_splice_init(&bucket->committing, pages);
243 	data->lseg = bucket->clseg;
244 	bucket->clseg = NULL;
245 	spin_unlock(cinfo->lock);
246 
247 }
248 
249 /* This follows nfs_commit_list pretty closely */
250 int
251 pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages,
252 			     int how, struct nfs_commit_info *cinfo,
253 			     int (*initiate_commit)(struct nfs_commit_data *data,
254 						    int how))
255 {
256 	struct nfs_commit_data *data, *tmp;
257 	LIST_HEAD(list);
258 	unsigned int nreq = 0;
259 
260 	if (!list_empty(mds_pages)) {
261 		data = nfs_commitdata_alloc();
262 		if (data != NULL) {
263 			data->ds_commit_index = -1;
264 			list_add(&data->pages, &list);
265 			nreq++;
266 		} else {
267 			nfs_retry_commit(mds_pages, NULL, cinfo, 0);
268 			pnfs_generic_retry_commit(cinfo, 0);
269 			cinfo->completion_ops->error_cleanup(NFS_I(inode));
270 			return -ENOMEM;
271 		}
272 	}
273 
274 	nreq += pnfs_generic_alloc_ds_commits(cinfo, &list);
275 
276 	if (nreq == 0) {
277 		cinfo->completion_ops->error_cleanup(NFS_I(inode));
278 		goto out;
279 	}
280 
281 	atomic_add(nreq, &cinfo->mds->rpcs_out);
282 
283 	list_for_each_entry_safe(data, tmp, &list, pages) {
284 		list_del_init(&data->pages);
285 		if (data->ds_commit_index < 0) {
286 			nfs_init_commit(data, mds_pages, NULL, cinfo);
287 			nfs_initiate_commit(NFS_CLIENT(inode), data,
288 					    NFS_PROTO(data->inode),
289 					    data->mds_ops, how, 0);
290 		} else {
291 			LIST_HEAD(pages);
292 
293 			pnfs_fetch_commit_bucket_list(&pages, data, cinfo);
294 			nfs_init_commit(data, &pages, data->lseg, cinfo);
295 			initiate_commit(data, how);
296 		}
297 	}
298 out:
299 	cinfo->ds->ncommitting = 0;
300 	return PNFS_ATTEMPTED;
301 }
302 EXPORT_SYMBOL_GPL(pnfs_generic_commit_pagelist);
303 
304 /*
305  * Data server cache
306  *
307  * Data servers can be mapped to different device ids.
308  * nfs4_pnfs_ds reference counting
309  *   - set to 1 on allocation
310  *   - incremented when a device id maps a data server already in the cache.
311  *   - decremented when deviceid is removed from the cache.
312  */
313 static DEFINE_SPINLOCK(nfs4_ds_cache_lock);
314 static LIST_HEAD(nfs4_data_server_cache);
315 
316 /* Debug routines */
317 static void
318 print_ds(struct nfs4_pnfs_ds *ds)
319 {
320 	if (ds == NULL) {
321 		printk(KERN_WARNING "%s NULL device\n", __func__);
322 		return;
323 	}
324 	printk(KERN_WARNING "        ds %s\n"
325 		"        ref count %d\n"
326 		"        client %p\n"
327 		"        cl_exchange_flags %x\n",
328 		ds->ds_remotestr,
329 		atomic_read(&ds->ds_count), ds->ds_clp,
330 		ds->ds_clp ? ds->ds_clp->cl_exchange_flags : 0);
331 }
332 
333 static bool
334 same_sockaddr(struct sockaddr *addr1, struct sockaddr *addr2)
335 {
336 	struct sockaddr_in *a, *b;
337 	struct sockaddr_in6 *a6, *b6;
338 
339 	if (addr1->sa_family != addr2->sa_family)
340 		return false;
341 
342 	switch (addr1->sa_family) {
343 	case AF_INET:
344 		a = (struct sockaddr_in *)addr1;
345 		b = (struct sockaddr_in *)addr2;
346 
347 		if (a->sin_addr.s_addr == b->sin_addr.s_addr &&
348 		    a->sin_port == b->sin_port)
349 			return true;
350 		break;
351 
352 	case AF_INET6:
353 		a6 = (struct sockaddr_in6 *)addr1;
354 		b6 = (struct sockaddr_in6 *)addr2;
355 
356 		/* LINKLOCAL addresses must have matching scope_id */
357 		if (ipv6_addr_src_scope(&a6->sin6_addr) ==
358 		    IPV6_ADDR_SCOPE_LINKLOCAL &&
359 		    a6->sin6_scope_id != b6->sin6_scope_id)
360 			return false;
361 
362 		if (ipv6_addr_equal(&a6->sin6_addr, &b6->sin6_addr) &&
363 		    a6->sin6_port == b6->sin6_port)
364 			return true;
365 		break;
366 
367 	default:
368 		dprintk("%s: unhandled address family: %u\n",
369 			__func__, addr1->sa_family);
370 		return false;
371 	}
372 
373 	return false;
374 }
375 
376 /*
377  * Checks if 'dsaddrs1' contains a subset of 'dsaddrs2'. If it does,
378  * declare a match.
379  */
380 static bool
381 _same_data_server_addrs_locked(const struct list_head *dsaddrs1,
382 			       const struct list_head *dsaddrs2)
383 {
384 	struct nfs4_pnfs_ds_addr *da1, *da2;
385 	struct sockaddr *sa1, *sa2;
386 	bool match = false;
387 
388 	list_for_each_entry(da1, dsaddrs1, da_node) {
389 		sa1 = (struct sockaddr *)&da1->da_addr;
390 		match = false;
391 		list_for_each_entry(da2, dsaddrs2, da_node) {
392 			sa2 = (struct sockaddr *)&da2->da_addr;
393 			match = same_sockaddr(sa1, sa2);
394 			if (match)
395 				break;
396 		}
397 		if (!match)
398 			break;
399 	}
400 	return match;
401 }
402 
403 /*
404  * Lookup DS by addresses.  nfs4_ds_cache_lock is held
405  */
406 static struct nfs4_pnfs_ds *
407 _data_server_lookup_locked(const struct list_head *dsaddrs)
408 {
409 	struct nfs4_pnfs_ds *ds;
410 
411 	list_for_each_entry(ds, &nfs4_data_server_cache, ds_node)
412 		if (_same_data_server_addrs_locked(&ds->ds_addrs, dsaddrs))
413 			return ds;
414 	return NULL;
415 }
416 
417 static void destroy_ds(struct nfs4_pnfs_ds *ds)
418 {
419 	struct nfs4_pnfs_ds_addr *da;
420 
421 	dprintk("--> %s\n", __func__);
422 	ifdebug(FACILITY)
423 		print_ds(ds);
424 
425 	nfs_put_client(ds->ds_clp);
426 
427 	while (!list_empty(&ds->ds_addrs)) {
428 		da = list_first_entry(&ds->ds_addrs,
429 				      struct nfs4_pnfs_ds_addr,
430 				      da_node);
431 		list_del_init(&da->da_node);
432 		kfree(da->da_remotestr);
433 		kfree(da);
434 	}
435 
436 	kfree(ds->ds_remotestr);
437 	kfree(ds);
438 }
439 
440 void nfs4_pnfs_ds_put(struct nfs4_pnfs_ds *ds)
441 {
442 	if (atomic_dec_and_lock(&ds->ds_count,
443 				&nfs4_ds_cache_lock)) {
444 		list_del_init(&ds->ds_node);
445 		spin_unlock(&nfs4_ds_cache_lock);
446 		destroy_ds(ds);
447 	}
448 }
449 EXPORT_SYMBOL_GPL(nfs4_pnfs_ds_put);
450 
451 /*
452  * Create a string with a human readable address and port to avoid
453  * complicated setup around many dprinks.
454  */
455 static char *
456 nfs4_pnfs_remotestr(struct list_head *dsaddrs, gfp_t gfp_flags)
457 {
458 	struct nfs4_pnfs_ds_addr *da;
459 	char *remotestr;
460 	size_t len;
461 	char *p;
462 
463 	len = 3;        /* '{', '}' and eol */
464 	list_for_each_entry(da, dsaddrs, da_node) {
465 		len += strlen(da->da_remotestr) + 1;    /* string plus comma */
466 	}
467 
468 	remotestr = kzalloc(len, gfp_flags);
469 	if (!remotestr)
470 		return NULL;
471 
472 	p = remotestr;
473 	*(p++) = '{';
474 	len--;
475 	list_for_each_entry(da, dsaddrs, da_node) {
476 		size_t ll = strlen(da->da_remotestr);
477 
478 		if (ll > len)
479 			goto out_err;
480 
481 		memcpy(p, da->da_remotestr, ll);
482 		p += ll;
483 		len -= ll;
484 
485 		if (len < 1)
486 			goto out_err;
487 		(*p++) = ',';
488 		len--;
489 	}
490 	if (len < 2)
491 		goto out_err;
492 	*(p++) = '}';
493 	*p = '\0';
494 	return remotestr;
495 out_err:
496 	kfree(remotestr);
497 	return NULL;
498 }
499 
500 /*
501  * Given a list of multipath struct nfs4_pnfs_ds_addr, add it to ds cache if
502  * uncached and return cached struct nfs4_pnfs_ds.
503  */
504 struct nfs4_pnfs_ds *
505 nfs4_pnfs_ds_add(struct list_head *dsaddrs, gfp_t gfp_flags)
506 {
507 	struct nfs4_pnfs_ds *tmp_ds, *ds = NULL;
508 	char *remotestr;
509 
510 	if (list_empty(dsaddrs)) {
511 		dprintk("%s: no addresses defined\n", __func__);
512 		goto out;
513 	}
514 
515 	ds = kzalloc(sizeof(*ds), gfp_flags);
516 	if (!ds)
517 		goto out;
518 
519 	/* this is only used for debugging, so it's ok if its NULL */
520 	remotestr = nfs4_pnfs_remotestr(dsaddrs, gfp_flags);
521 
522 	spin_lock(&nfs4_ds_cache_lock);
523 	tmp_ds = _data_server_lookup_locked(dsaddrs);
524 	if (tmp_ds == NULL) {
525 		INIT_LIST_HEAD(&ds->ds_addrs);
526 		list_splice_init(dsaddrs, &ds->ds_addrs);
527 		ds->ds_remotestr = remotestr;
528 		atomic_set(&ds->ds_count, 1);
529 		INIT_LIST_HEAD(&ds->ds_node);
530 		ds->ds_clp = NULL;
531 		list_add(&ds->ds_node, &nfs4_data_server_cache);
532 		dprintk("%s add new data server %s\n", __func__,
533 			ds->ds_remotestr);
534 	} else {
535 		kfree(remotestr);
536 		kfree(ds);
537 		atomic_inc(&tmp_ds->ds_count);
538 		dprintk("%s data server %s found, inc'ed ds_count to %d\n",
539 			__func__, tmp_ds->ds_remotestr,
540 			atomic_read(&tmp_ds->ds_count));
541 		ds = tmp_ds;
542 	}
543 	spin_unlock(&nfs4_ds_cache_lock);
544 out:
545 	return ds;
546 }
547 EXPORT_SYMBOL_GPL(nfs4_pnfs_ds_add);
548 
549 static void nfs4_wait_ds_connect(struct nfs4_pnfs_ds *ds)
550 {
551 	might_sleep();
552 	wait_on_bit(&ds->ds_state, NFS4DS_CONNECTING,
553 			TASK_KILLABLE);
554 }
555 
556 static void nfs4_clear_ds_conn_bit(struct nfs4_pnfs_ds *ds)
557 {
558 	smp_mb__before_atomic();
559 	clear_bit(NFS4DS_CONNECTING, &ds->ds_state);
560 	smp_mb__after_atomic();
561 	wake_up_bit(&ds->ds_state, NFS4DS_CONNECTING);
562 }
563 
564 static struct nfs_client *(*get_v3_ds_connect)(
565 			struct nfs_client *mds_clp,
566 			const struct sockaddr *ds_addr,
567 			int ds_addrlen,
568 			int ds_proto,
569 			unsigned int ds_timeo,
570 			unsigned int ds_retrans,
571 			rpc_authflavor_t au_flavor);
572 
573 static bool load_v3_ds_connect(void)
574 {
575 	if (!get_v3_ds_connect) {
576 		get_v3_ds_connect = symbol_request(nfs3_set_ds_client);
577 		WARN_ON_ONCE(!get_v3_ds_connect);
578 	}
579 
580 	return(get_v3_ds_connect != NULL);
581 }
582 
583 void nfs4_pnfs_v3_ds_connect_unload(void)
584 {
585 	if (get_v3_ds_connect) {
586 		symbol_put(nfs3_set_ds_client);
587 		get_v3_ds_connect = NULL;
588 	}
589 }
590 EXPORT_SYMBOL_GPL(nfs4_pnfs_v3_ds_connect_unload);
591 
592 static int _nfs4_pnfs_v3_ds_connect(struct nfs_server *mds_srv,
593 				 struct nfs4_pnfs_ds *ds,
594 				 unsigned int timeo,
595 				 unsigned int retrans,
596 				 rpc_authflavor_t au_flavor)
597 {
598 	struct nfs_client *clp = ERR_PTR(-EIO);
599 	struct nfs4_pnfs_ds_addr *da;
600 	int status = 0;
601 
602 	dprintk("--> %s DS %s au_flavor %d\n", __func__,
603 		ds->ds_remotestr, au_flavor);
604 
605 	if (!load_v3_ds_connect())
606 		goto out;
607 
608 	list_for_each_entry(da, &ds->ds_addrs, da_node) {
609 		dprintk("%s: DS %s: trying address %s\n",
610 			__func__, ds->ds_remotestr, da->da_remotestr);
611 
612 		clp = get_v3_ds_connect(mds_srv->nfs_client,
613 					(struct sockaddr *)&da->da_addr,
614 					da->da_addrlen, IPPROTO_TCP,
615 					timeo, retrans, au_flavor);
616 		if (!IS_ERR(clp))
617 			break;
618 	}
619 
620 	if (IS_ERR(clp)) {
621 		status = PTR_ERR(clp);
622 		goto out;
623 	}
624 
625 	smp_wmb();
626 	ds->ds_clp = clp;
627 	dprintk("%s [new] addr: %s\n", __func__, ds->ds_remotestr);
628 out:
629 	return status;
630 }
631 
632 static int _nfs4_pnfs_v4_ds_connect(struct nfs_server *mds_srv,
633 				 struct nfs4_pnfs_ds *ds,
634 				 unsigned int timeo,
635 				 unsigned int retrans,
636 				 u32 minor_version,
637 				 rpc_authflavor_t au_flavor)
638 {
639 	struct nfs_client *clp = ERR_PTR(-EIO);
640 	struct nfs4_pnfs_ds_addr *da;
641 	int status = 0;
642 
643 	dprintk("--> %s DS %s au_flavor %d\n", __func__, ds->ds_remotestr,
644 		au_flavor);
645 
646 	list_for_each_entry(da, &ds->ds_addrs, da_node) {
647 		dprintk("%s: DS %s: trying address %s\n",
648 			__func__, ds->ds_remotestr, da->da_remotestr);
649 
650 		clp = nfs4_set_ds_client(mds_srv->nfs_client,
651 					(struct sockaddr *)&da->da_addr,
652 					da->da_addrlen, IPPROTO_TCP,
653 					timeo, retrans, minor_version,
654 					au_flavor);
655 		if (!IS_ERR(clp))
656 			break;
657 	}
658 
659 	if (IS_ERR(clp)) {
660 		status = PTR_ERR(clp);
661 		goto out;
662 	}
663 
664 	status = nfs4_init_ds_session(clp, mds_srv->nfs_client->cl_lease_time);
665 	if (status)
666 		goto out_put;
667 
668 	smp_wmb();
669 	ds->ds_clp = clp;
670 	dprintk("%s [new] addr: %s\n", __func__, ds->ds_remotestr);
671 out:
672 	return status;
673 out_put:
674 	nfs_put_client(clp);
675 	goto out;
676 }
677 
678 /*
679  * Create an rpc connection to the nfs4_pnfs_ds data server.
680  * Currently only supports IPv4 and IPv6 addresses.
681  * If connection fails, make devid unavailable.
682  */
683 void nfs4_pnfs_ds_connect(struct nfs_server *mds_srv, struct nfs4_pnfs_ds *ds,
684 			  struct nfs4_deviceid_node *devid, unsigned int timeo,
685 			  unsigned int retrans, u32 version,
686 			  u32 minor_version, rpc_authflavor_t au_flavor)
687 {
688 	if (test_and_set_bit(NFS4DS_CONNECTING, &ds->ds_state) == 0) {
689 		int err = 0;
690 
691 		if (version == 3) {
692 			err = _nfs4_pnfs_v3_ds_connect(mds_srv, ds, timeo,
693 						       retrans, au_flavor);
694 		} else if (version == 4) {
695 			err = _nfs4_pnfs_v4_ds_connect(mds_srv, ds, timeo,
696 						       retrans, minor_version,
697 						       au_flavor);
698 		} else {
699 			dprintk("%s: unsupported DS version %d\n", __func__,
700 				version);
701 			err = -EPROTONOSUPPORT;
702 		}
703 
704 		if (err)
705 			nfs4_mark_deviceid_unavailable(devid);
706 		nfs4_clear_ds_conn_bit(ds);
707 	} else {
708 		nfs4_wait_ds_connect(ds);
709 	}
710 }
711 EXPORT_SYMBOL_GPL(nfs4_pnfs_ds_connect);
712 
713 /*
714  * Currently only supports ipv4, ipv6 and one multi-path address.
715  */
716 struct nfs4_pnfs_ds_addr *
717 nfs4_decode_mp_ds_addr(struct net *net, struct xdr_stream *xdr, gfp_t gfp_flags)
718 {
719 	struct nfs4_pnfs_ds_addr *da = NULL;
720 	char *buf, *portstr;
721 	__be16 port;
722 	int nlen, rlen;
723 	int tmp[2];
724 	__be32 *p;
725 	char *netid, *match_netid;
726 	size_t len, match_netid_len;
727 	char *startsep = "";
728 	char *endsep = "";
729 
730 
731 	/* r_netid */
732 	p = xdr_inline_decode(xdr, 4);
733 	if (unlikely(!p))
734 		goto out_err;
735 	nlen = be32_to_cpup(p++);
736 
737 	p = xdr_inline_decode(xdr, nlen);
738 	if (unlikely(!p))
739 		goto out_err;
740 
741 	netid = kmalloc(nlen+1, gfp_flags);
742 	if (unlikely(!netid))
743 		goto out_err;
744 
745 	netid[nlen] = '\0';
746 	memcpy(netid, p, nlen);
747 
748 	/* r_addr: ip/ip6addr with port in dec octets - see RFC 5665 */
749 	p = xdr_inline_decode(xdr, 4);
750 	if (unlikely(!p))
751 		goto out_free_netid;
752 	rlen = be32_to_cpup(p);
753 
754 	p = xdr_inline_decode(xdr, rlen);
755 	if (unlikely(!p))
756 		goto out_free_netid;
757 
758 	/* port is ".ABC.DEF", 8 chars max */
759 	if (rlen > INET6_ADDRSTRLEN + IPV6_SCOPE_ID_LEN + 8) {
760 		dprintk("%s: Invalid address, length %d\n", __func__,
761 			rlen);
762 		goto out_free_netid;
763 	}
764 	buf = kmalloc(rlen + 1, gfp_flags);
765 	if (!buf) {
766 		dprintk("%s: Not enough memory\n", __func__);
767 		goto out_free_netid;
768 	}
769 	buf[rlen] = '\0';
770 	memcpy(buf, p, rlen);
771 
772 	/* replace port '.' with '-' */
773 	portstr = strrchr(buf, '.');
774 	if (!portstr) {
775 		dprintk("%s: Failed finding expected dot in port\n",
776 			__func__);
777 		goto out_free_buf;
778 	}
779 	*portstr = '-';
780 
781 	/* find '.' between address and port */
782 	portstr = strrchr(buf, '.');
783 	if (!portstr) {
784 		dprintk("%s: Failed finding expected dot between address and "
785 			"port\n", __func__);
786 		goto out_free_buf;
787 	}
788 	*portstr = '\0';
789 
790 	da = kzalloc(sizeof(*da), gfp_flags);
791 	if (unlikely(!da))
792 		goto out_free_buf;
793 
794 	INIT_LIST_HEAD(&da->da_node);
795 
796 	if (!rpc_pton(net, buf, portstr-buf, (struct sockaddr *)&da->da_addr,
797 		      sizeof(da->da_addr))) {
798 		dprintk("%s: error parsing address %s\n", __func__, buf);
799 		goto out_free_da;
800 	}
801 
802 	portstr++;
803 	sscanf(portstr, "%d-%d", &tmp[0], &tmp[1]);
804 	port = htons((tmp[0] << 8) | (tmp[1]));
805 
806 	switch (da->da_addr.ss_family) {
807 	case AF_INET:
808 		((struct sockaddr_in *)&da->da_addr)->sin_port = port;
809 		da->da_addrlen = sizeof(struct sockaddr_in);
810 		match_netid = "tcp";
811 		match_netid_len = 3;
812 		break;
813 
814 	case AF_INET6:
815 		((struct sockaddr_in6 *)&da->da_addr)->sin6_port = port;
816 		da->da_addrlen = sizeof(struct sockaddr_in6);
817 		match_netid = "tcp6";
818 		match_netid_len = 4;
819 		startsep = "[";
820 		endsep = "]";
821 		break;
822 
823 	default:
824 		dprintk("%s: unsupported address family: %u\n",
825 			__func__, da->da_addr.ss_family);
826 		goto out_free_da;
827 	}
828 
829 	if (nlen != match_netid_len || strncmp(netid, match_netid, nlen)) {
830 		dprintk("%s: ERROR: r_netid \"%s\" != \"%s\"\n",
831 			__func__, netid, match_netid);
832 		goto out_free_da;
833 	}
834 
835 	/* save human readable address */
836 	len = strlen(startsep) + strlen(buf) + strlen(endsep) + 7;
837 	da->da_remotestr = kzalloc(len, gfp_flags);
838 
839 	/* NULL is ok, only used for dprintk */
840 	if (da->da_remotestr)
841 		snprintf(da->da_remotestr, len, "%s%s%s:%u", startsep,
842 			 buf, endsep, ntohs(port));
843 
844 	dprintk("%s: Parsed DS addr %s\n", __func__, da->da_remotestr);
845 	kfree(buf);
846 	kfree(netid);
847 	return da;
848 
849 out_free_da:
850 	kfree(da);
851 out_free_buf:
852 	dprintk("%s: Error parsing DS addr: %s\n", __func__, buf);
853 	kfree(buf);
854 out_free_netid:
855 	kfree(netid);
856 out_err:
857 	return NULL;
858 }
859 EXPORT_SYMBOL_GPL(nfs4_decode_mp_ds_addr);
860 
861 void
862 pnfs_layout_mark_request_commit(struct nfs_page *req,
863 				struct pnfs_layout_segment *lseg,
864 				struct nfs_commit_info *cinfo,
865 				u32 ds_commit_idx)
866 {
867 	struct list_head *list;
868 	struct pnfs_commit_bucket *buckets;
869 
870 	spin_lock(cinfo->lock);
871 	buckets = cinfo->ds->buckets;
872 	list = &buckets[ds_commit_idx].written;
873 	if (list_empty(list)) {
874 		/* Non-empty buckets hold a reference on the lseg.  That ref
875 		 * is normally transferred to the COMMIT call and released
876 		 * there.  It could also be released if the last req is pulled
877 		 * off due to a rewrite, in which case it will be done in
878 		 * pnfs_common_clear_request_commit
879 		 */
880 		WARN_ON_ONCE(buckets[ds_commit_idx].wlseg != NULL);
881 		buckets[ds_commit_idx].wlseg = pnfs_get_lseg(lseg);
882 	}
883 	set_bit(PG_COMMIT_TO_DS, &req->wb_flags);
884 	cinfo->ds->nwritten++;
885 
886 	nfs_request_add_commit_list_locked(req, list, cinfo);
887 	spin_unlock(cinfo->lock);
888 	nfs_mark_page_unstable(req->wb_page, cinfo);
889 }
890 EXPORT_SYMBOL_GPL(pnfs_layout_mark_request_commit);
891 
892 int
893 pnfs_nfs_generic_sync(struct inode *inode, bool datasync)
894 {
895 	if (datasync)
896 		return 0;
897 	return pnfs_layoutcommit_inode(inode, true);
898 }
899 EXPORT_SYMBOL_GPL(pnfs_nfs_generic_sync);
900 
901