xref: /openbmc/linux/fs/nfs/direct.c (revision bc000245)
1 /*
2  * linux/fs/nfs/direct.c
3  *
4  * Copyright (C) 2003 by Chuck Lever <cel@netapp.com>
5  *
6  * High-performance uncached I/O for the Linux NFS client
7  *
8  * There are important applications whose performance or correctness
9  * depends on uncached access to file data.  Database clusters
10  * (multiple copies of the same instance running on separate hosts)
11  * implement their own cache coherency protocol that subsumes file
12  * system cache protocols.  Applications that process datasets
13  * considerably larger than the client's memory do not always benefit
14  * from a local cache.  A streaming video server, for instance, has no
15  * need to cache the contents of a file.
16  *
17  * When an application requests uncached I/O, all read and write requests
18  * are made directly to the server; data stored or fetched via these
19  * requests is not cached in the Linux page cache.  The client does not
20  * correct unaligned requests from applications.  All requested bytes are
21  * held on permanent storage before a direct write system call returns to
22  * an application.
23  *
24  * Solaris implements an uncached I/O facility called directio() that
25  * is used for backups and sequential I/O to very large files.  Solaris
26  * also supports uncaching whole NFS partitions with "-o forcedirectio,"
27  * an undocumented mount option.
28  *
29  * Designed by Jeff Kimmel, Chuck Lever, and Trond Myklebust, with
30  * help from Andrew Morton.
31  *
32  * 18 Dec 2001	Initial implementation for 2.4  --cel
33  * 08 Jul 2002	Version for 2.4.19, with bug fixes --trondmy
34  * 08 Jun 2003	Port to 2.5 APIs  --cel
35  * 31 Mar 2004	Handle direct I/O without VFS support  --cel
36  * 15 Sep 2004	Parallel async reads  --cel
37  * 04 May 2005	support O_DIRECT with aio  --cel
38  *
39  */
40 
41 #include <linux/errno.h>
42 #include <linux/sched.h>
43 #include <linux/kernel.h>
44 #include <linux/file.h>
45 #include <linux/pagemap.h>
46 #include <linux/kref.h>
47 #include <linux/slab.h>
48 #include <linux/task_io_accounting_ops.h>
49 #include <linux/module.h>
50 
51 #include <linux/nfs_fs.h>
52 #include <linux/nfs_page.h>
53 #include <linux/sunrpc/clnt.h>
54 
55 #include <asm/uaccess.h>
56 #include <linux/atomic.h>
57 
58 #include "internal.h"
59 #include "iostat.h"
60 #include "pnfs.h"
61 
62 #define NFSDBG_FACILITY		NFSDBG_VFS
63 
64 static struct kmem_cache *nfs_direct_cachep;
65 
66 /*
67  * This represents a set of asynchronous requests that we're waiting on
68  */
69 struct nfs_direct_req {
70 	struct kref		kref;		/* release manager */
71 
72 	/* I/O parameters */
73 	struct nfs_open_context	*ctx;		/* file open context info */
74 	struct nfs_lock_context *l_ctx;		/* Lock context info */
75 	struct kiocb *		iocb;		/* controlling i/o request */
76 	struct inode *		inode;		/* target file of i/o */
77 
78 	/* completion state */
79 	atomic_t		io_count;	/* i/os we're waiting for */
80 	spinlock_t		lock;		/* protect completion state */
81 	ssize_t			count,		/* bytes actually processed */
82 				bytes_left,	/* bytes left to be sent */
83 				error;		/* any reported error */
84 	struct completion	completion;	/* wait for i/o completion */
85 
86 	/* commit state */
87 	struct nfs_mds_commit_info mds_cinfo;	/* Storage for cinfo */
88 	struct pnfs_ds_commit_info ds_cinfo;	/* Storage for cinfo */
89 	struct work_struct	work;
90 	int			flags;
91 #define NFS_ODIRECT_DO_COMMIT		(1)	/* an unstable reply was received */
92 #define NFS_ODIRECT_RESCHED_WRITES	(2)	/* write verification failed */
93 	struct nfs_writeverf	verf;		/* unstable write verifier */
94 };
95 
96 static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops;
97 static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops;
98 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode);
99 static void nfs_direct_write_schedule_work(struct work_struct *work);
100 
101 static inline void get_dreq(struct nfs_direct_req *dreq)
102 {
103 	atomic_inc(&dreq->io_count);
104 }
105 
106 static inline int put_dreq(struct nfs_direct_req *dreq)
107 {
108 	return atomic_dec_and_test(&dreq->io_count);
109 }
110 
111 /**
112  * nfs_direct_IO - NFS address space operation for direct I/O
113  * @rw: direction (read or write)
114  * @iocb: target I/O control block
115  * @iov: array of vectors that define I/O buffer
116  * @pos: offset in file to begin the operation
117  * @nr_segs: size of iovec array
118  *
119  * The presence of this routine in the address space ops vector means
120  * the NFS client supports direct I/O. However, for most direct IO, we
121  * shunt off direct read and write requests before the VFS gets them,
122  * so this method is only ever called for swap.
123  */
124 ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t pos, unsigned long nr_segs)
125 {
126 #ifndef CONFIG_NFS_SWAP
127 	dprintk("NFS: nfs_direct_IO (%pD) off/no(%Ld/%lu) EINVAL\n",
128 			iocb->ki_filp, (long long) pos, nr_segs);
129 
130 	return -EINVAL;
131 #else
132 	VM_BUG_ON(iocb->ki_nbytes != PAGE_SIZE);
133 
134 	if (rw == READ || rw == KERNEL_READ)
135 		return nfs_file_direct_read(iocb, iov, nr_segs, pos,
136 				rw == READ ? true : false);
137 	return nfs_file_direct_write(iocb, iov, nr_segs, pos,
138 				rw == WRITE ? true : false);
139 #endif /* CONFIG_NFS_SWAP */
140 }
141 
142 static void nfs_direct_release_pages(struct page **pages, unsigned int npages)
143 {
144 	unsigned int i;
145 	for (i = 0; i < npages; i++)
146 		page_cache_release(pages[i]);
147 }
148 
149 void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo,
150 			      struct nfs_direct_req *dreq)
151 {
152 	cinfo->lock = &dreq->lock;
153 	cinfo->mds = &dreq->mds_cinfo;
154 	cinfo->ds = &dreq->ds_cinfo;
155 	cinfo->dreq = dreq;
156 	cinfo->completion_ops = &nfs_direct_commit_completion_ops;
157 }
158 
159 static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
160 {
161 	struct nfs_direct_req *dreq;
162 
163 	dreq = kmem_cache_zalloc(nfs_direct_cachep, GFP_KERNEL);
164 	if (!dreq)
165 		return NULL;
166 
167 	kref_init(&dreq->kref);
168 	kref_get(&dreq->kref);
169 	init_completion(&dreq->completion);
170 	INIT_LIST_HEAD(&dreq->mds_cinfo.list);
171 	INIT_WORK(&dreq->work, nfs_direct_write_schedule_work);
172 	spin_lock_init(&dreq->lock);
173 
174 	return dreq;
175 }
176 
177 static void nfs_direct_req_free(struct kref *kref)
178 {
179 	struct nfs_direct_req *dreq = container_of(kref, struct nfs_direct_req, kref);
180 
181 	if (dreq->l_ctx != NULL)
182 		nfs_put_lock_context(dreq->l_ctx);
183 	if (dreq->ctx != NULL)
184 		put_nfs_open_context(dreq->ctx);
185 	kmem_cache_free(nfs_direct_cachep, dreq);
186 }
187 
188 static void nfs_direct_req_release(struct nfs_direct_req *dreq)
189 {
190 	kref_put(&dreq->kref, nfs_direct_req_free);
191 }
192 
193 ssize_t nfs_dreq_bytes_left(struct nfs_direct_req *dreq)
194 {
195 	return dreq->bytes_left;
196 }
197 EXPORT_SYMBOL_GPL(nfs_dreq_bytes_left);
198 
199 /*
200  * Collects and returns the final error value/byte-count.
201  */
202 static ssize_t nfs_direct_wait(struct nfs_direct_req *dreq)
203 {
204 	ssize_t result = -EIOCBQUEUED;
205 
206 	/* Async requests don't wait here */
207 	if (dreq->iocb)
208 		goto out;
209 
210 	result = wait_for_completion_killable(&dreq->completion);
211 
212 	if (!result)
213 		result = dreq->error;
214 	if (!result)
215 		result = dreq->count;
216 
217 out:
218 	return (ssize_t) result;
219 }
220 
221 /*
222  * Synchronous I/O uses a stack-allocated iocb.  Thus we can't trust
223  * the iocb is still valid here if this is a synchronous request.
224  */
225 static void nfs_direct_complete(struct nfs_direct_req *dreq)
226 {
227 	if (dreq->iocb) {
228 		long res = (long) dreq->error;
229 		if (!res)
230 			res = (long) dreq->count;
231 		aio_complete(dreq->iocb, res, 0);
232 	}
233 	complete_all(&dreq->completion);
234 
235 	nfs_direct_req_release(dreq);
236 }
237 
238 static void nfs_direct_readpage_release(struct nfs_page *req)
239 {
240 	dprintk("NFS: direct read done (%s/%lld %d@%lld)\n",
241 		req->wb_context->dentry->d_inode->i_sb->s_id,
242 		(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
243 		req->wb_bytes,
244 		(long long)req_offset(req));
245 	nfs_release_request(req);
246 }
247 
248 static void nfs_direct_read_completion(struct nfs_pgio_header *hdr)
249 {
250 	unsigned long bytes = 0;
251 	struct nfs_direct_req *dreq = hdr->dreq;
252 
253 	if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
254 		goto out_put;
255 
256 	spin_lock(&dreq->lock);
257 	if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && (hdr->good_bytes == 0))
258 		dreq->error = hdr->error;
259 	else
260 		dreq->count += hdr->good_bytes;
261 	spin_unlock(&dreq->lock);
262 
263 	while (!list_empty(&hdr->pages)) {
264 		struct nfs_page *req = nfs_list_entry(hdr->pages.next);
265 		struct page *page = req->wb_page;
266 
267 		if (!PageCompound(page) && bytes < hdr->good_bytes)
268 			set_page_dirty(page);
269 		bytes += req->wb_bytes;
270 		nfs_list_remove_request(req);
271 		nfs_direct_readpage_release(req);
272 	}
273 out_put:
274 	if (put_dreq(dreq))
275 		nfs_direct_complete(dreq);
276 	hdr->release(hdr);
277 }
278 
279 static void nfs_read_sync_pgio_error(struct list_head *head)
280 {
281 	struct nfs_page *req;
282 
283 	while (!list_empty(head)) {
284 		req = nfs_list_entry(head->next);
285 		nfs_list_remove_request(req);
286 		nfs_release_request(req);
287 	}
288 }
289 
290 static void nfs_direct_pgio_init(struct nfs_pgio_header *hdr)
291 {
292 	get_dreq(hdr->dreq);
293 }
294 
295 static const struct nfs_pgio_completion_ops nfs_direct_read_completion_ops = {
296 	.error_cleanup = nfs_read_sync_pgio_error,
297 	.init_hdr = nfs_direct_pgio_init,
298 	.completion = nfs_direct_read_completion,
299 };
300 
301 /*
302  * For each rsize'd chunk of the user's buffer, dispatch an NFS READ
303  * operation.  If nfs_readdata_alloc() or get_user_pages() fails,
304  * bail and stop sending more reads.  Read length accounting is
305  * handled automatically by nfs_direct_read_result().  Otherwise, if
306  * no requests have been sent, just return an error.
307  */
308 static ssize_t nfs_direct_read_schedule_segment(struct nfs_pageio_descriptor *desc,
309 						const struct iovec *iov,
310 						loff_t pos, bool uio)
311 {
312 	struct nfs_direct_req *dreq = desc->pg_dreq;
313 	struct nfs_open_context *ctx = dreq->ctx;
314 	struct inode *inode = ctx->dentry->d_inode;
315 	unsigned long user_addr = (unsigned long)iov->iov_base;
316 	size_t count = iov->iov_len;
317 	size_t rsize = NFS_SERVER(inode)->rsize;
318 	unsigned int pgbase;
319 	int result;
320 	ssize_t started = 0;
321 	struct page **pagevec = NULL;
322 	unsigned int npages;
323 
324 	do {
325 		size_t bytes;
326 		int i;
327 
328 		pgbase = user_addr & ~PAGE_MASK;
329 		bytes = min(max_t(size_t, rsize, PAGE_SIZE), count);
330 
331 		result = -ENOMEM;
332 		npages = nfs_page_array_len(pgbase, bytes);
333 		if (!pagevec)
334 			pagevec = kmalloc(npages * sizeof(struct page *),
335 					  GFP_KERNEL);
336 		if (!pagevec)
337 			break;
338 		if (uio) {
339 			down_read(&current->mm->mmap_sem);
340 			result = get_user_pages(current, current->mm, user_addr,
341 					npages, 1, 0, pagevec, NULL);
342 			up_read(&current->mm->mmap_sem);
343 			if (result < 0)
344 				break;
345 		} else {
346 			WARN_ON(npages != 1);
347 			result = get_kernel_page(user_addr, 1, pagevec);
348 			if (WARN_ON(result != 1))
349 				break;
350 		}
351 
352 		if ((unsigned)result < npages) {
353 			bytes = result * PAGE_SIZE;
354 			if (bytes <= pgbase) {
355 				nfs_direct_release_pages(pagevec, result);
356 				break;
357 			}
358 			bytes -= pgbase;
359 			npages = result;
360 		}
361 
362 		for (i = 0; i < npages; i++) {
363 			struct nfs_page *req;
364 			unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase);
365 			/* XXX do we need to do the eof zeroing found in async_filler? */
366 			req = nfs_create_request(dreq->ctx, dreq->inode,
367 						 pagevec[i],
368 						 pgbase, req_len);
369 			if (IS_ERR(req)) {
370 				result = PTR_ERR(req);
371 				break;
372 			}
373 			req->wb_index = pos >> PAGE_SHIFT;
374 			req->wb_offset = pos & ~PAGE_MASK;
375 			if (!nfs_pageio_add_request(desc, req)) {
376 				result = desc->pg_error;
377 				nfs_release_request(req);
378 				break;
379 			}
380 			pgbase = 0;
381 			bytes -= req_len;
382 			started += req_len;
383 			user_addr += req_len;
384 			pos += req_len;
385 			count -= req_len;
386 			dreq->bytes_left -= req_len;
387 		}
388 		/* The nfs_page now hold references to these pages */
389 		nfs_direct_release_pages(pagevec, npages);
390 	} while (count != 0 && result >= 0);
391 
392 	kfree(pagevec);
393 
394 	if (started)
395 		return started;
396 	return result < 0 ? (ssize_t) result : -EFAULT;
397 }
398 
399 static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq,
400 					      const struct iovec *iov,
401 					      unsigned long nr_segs,
402 					      loff_t pos, bool uio)
403 {
404 	struct nfs_pageio_descriptor desc;
405 	ssize_t result = -EINVAL;
406 	size_t requested_bytes = 0;
407 	unsigned long seg;
408 
409 	NFS_PROTO(dreq->inode)->read_pageio_init(&desc, dreq->inode,
410 			     &nfs_direct_read_completion_ops);
411 	get_dreq(dreq);
412 	desc.pg_dreq = dreq;
413 
414 	for (seg = 0; seg < nr_segs; seg++) {
415 		const struct iovec *vec = &iov[seg];
416 		result = nfs_direct_read_schedule_segment(&desc, vec, pos, uio);
417 		if (result < 0)
418 			break;
419 		requested_bytes += result;
420 		if ((size_t)result < vec->iov_len)
421 			break;
422 		pos += vec->iov_len;
423 	}
424 
425 	nfs_pageio_complete(&desc);
426 
427 	/*
428 	 * If no bytes were started, return the error, and let the
429 	 * generic layer handle the completion.
430 	 */
431 	if (requested_bytes == 0) {
432 		nfs_direct_req_release(dreq);
433 		return result < 0 ? result : -EIO;
434 	}
435 
436 	if (put_dreq(dreq))
437 		nfs_direct_complete(dreq);
438 	return 0;
439 }
440 
441 static ssize_t nfs_direct_read(struct kiocb *iocb, const struct iovec *iov,
442 			       unsigned long nr_segs, loff_t pos, bool uio)
443 {
444 	ssize_t result = -ENOMEM;
445 	struct inode *inode = iocb->ki_filp->f_mapping->host;
446 	struct nfs_direct_req *dreq;
447 	struct nfs_lock_context *l_ctx;
448 
449 	dreq = nfs_direct_req_alloc();
450 	if (dreq == NULL)
451 		goto out;
452 
453 	dreq->inode = inode;
454 	dreq->bytes_left = iov_length(iov, nr_segs);
455 	dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
456 	l_ctx = nfs_get_lock_context(dreq->ctx);
457 	if (IS_ERR(l_ctx)) {
458 		result = PTR_ERR(l_ctx);
459 		goto out_release;
460 	}
461 	dreq->l_ctx = l_ctx;
462 	if (!is_sync_kiocb(iocb))
463 		dreq->iocb = iocb;
464 
465 	NFS_I(inode)->read_io += iov_length(iov, nr_segs);
466 	result = nfs_direct_read_schedule_iovec(dreq, iov, nr_segs, pos, uio);
467 	if (!result)
468 		result = nfs_direct_wait(dreq);
469 out_release:
470 	nfs_direct_req_release(dreq);
471 out:
472 	return result;
473 }
474 
475 static void nfs_inode_dio_write_done(struct inode *inode)
476 {
477 	nfs_zap_mapping(inode, inode->i_mapping);
478 	inode_dio_done(inode);
479 }
480 
481 #if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
482 static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
483 {
484 	struct nfs_pageio_descriptor desc;
485 	struct nfs_page *req, *tmp;
486 	LIST_HEAD(reqs);
487 	struct nfs_commit_info cinfo;
488 	LIST_HEAD(failed);
489 
490 	nfs_init_cinfo_from_dreq(&cinfo, dreq);
491 	pnfs_recover_commit_reqs(dreq->inode, &reqs, &cinfo);
492 	spin_lock(cinfo.lock);
493 	nfs_scan_commit_list(&cinfo.mds->list, &reqs, &cinfo, 0);
494 	spin_unlock(cinfo.lock);
495 
496 	dreq->count = 0;
497 	get_dreq(dreq);
498 
499 	NFS_PROTO(dreq->inode)->write_pageio_init(&desc, dreq->inode, FLUSH_STABLE,
500 			      &nfs_direct_write_completion_ops);
501 	desc.pg_dreq = dreq;
502 
503 	list_for_each_entry_safe(req, tmp, &reqs, wb_list) {
504 		if (!nfs_pageio_add_request(&desc, req)) {
505 			nfs_list_remove_request(req);
506 			nfs_list_add_request(req, &failed);
507 			spin_lock(cinfo.lock);
508 			dreq->flags = 0;
509 			dreq->error = -EIO;
510 			spin_unlock(cinfo.lock);
511 		}
512 		nfs_release_request(req);
513 	}
514 	nfs_pageio_complete(&desc);
515 
516 	while (!list_empty(&failed)) {
517 		req = nfs_list_entry(failed.next);
518 		nfs_list_remove_request(req);
519 		nfs_unlock_and_release_request(req);
520 	}
521 
522 	if (put_dreq(dreq))
523 		nfs_direct_write_complete(dreq, dreq->inode);
524 }
525 
526 static void nfs_direct_commit_complete(struct nfs_commit_data *data)
527 {
528 	struct nfs_direct_req *dreq = data->dreq;
529 	struct nfs_commit_info cinfo;
530 	struct nfs_page *req;
531 	int status = data->task.tk_status;
532 
533 	nfs_init_cinfo_from_dreq(&cinfo, dreq);
534 	if (status < 0) {
535 		dprintk("NFS: %5u commit failed with error %d.\n",
536 			data->task.tk_pid, status);
537 		dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
538 	} else if (memcmp(&dreq->verf, &data->verf, sizeof(data->verf))) {
539 		dprintk("NFS: %5u commit verify failed\n", data->task.tk_pid);
540 		dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
541 	}
542 
543 	dprintk("NFS: %5u commit returned %d\n", data->task.tk_pid, status);
544 	while (!list_empty(&data->pages)) {
545 		req = nfs_list_entry(data->pages.next);
546 		nfs_list_remove_request(req);
547 		if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) {
548 			/* Note the rewrite will go through mds */
549 			nfs_mark_request_commit(req, NULL, &cinfo);
550 		} else
551 			nfs_release_request(req);
552 		nfs_unlock_and_release_request(req);
553 	}
554 
555 	if (atomic_dec_and_test(&cinfo.mds->rpcs_out))
556 		nfs_direct_write_complete(dreq, data->inode);
557 }
558 
559 static void nfs_direct_error_cleanup(struct nfs_inode *nfsi)
560 {
561 	/* There is no lock to clear */
562 }
563 
564 static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops = {
565 	.completion = nfs_direct_commit_complete,
566 	.error_cleanup = nfs_direct_error_cleanup,
567 };
568 
569 static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
570 {
571 	int res;
572 	struct nfs_commit_info cinfo;
573 	LIST_HEAD(mds_list);
574 
575 	nfs_init_cinfo_from_dreq(&cinfo, dreq);
576 	nfs_scan_commit(dreq->inode, &mds_list, &cinfo);
577 	res = nfs_generic_commit_list(dreq->inode, &mds_list, 0, &cinfo);
578 	if (res < 0) /* res == -ENOMEM */
579 		nfs_direct_write_reschedule(dreq);
580 }
581 
582 static void nfs_direct_write_schedule_work(struct work_struct *work)
583 {
584 	struct nfs_direct_req *dreq = container_of(work, struct nfs_direct_req, work);
585 	int flags = dreq->flags;
586 
587 	dreq->flags = 0;
588 	switch (flags) {
589 		case NFS_ODIRECT_DO_COMMIT:
590 			nfs_direct_commit_schedule(dreq);
591 			break;
592 		case NFS_ODIRECT_RESCHED_WRITES:
593 			nfs_direct_write_reschedule(dreq);
594 			break;
595 		default:
596 			nfs_inode_dio_write_done(dreq->inode);
597 			nfs_direct_complete(dreq);
598 	}
599 }
600 
601 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
602 {
603 	schedule_work(&dreq->work); /* Calls nfs_direct_write_schedule_work */
604 }
605 
606 #else
607 static void nfs_direct_write_schedule_work(struct work_struct *work)
608 {
609 }
610 
611 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
612 {
613 	nfs_inode_dio_write_done(inode);
614 	nfs_direct_complete(dreq);
615 }
616 #endif
617 
618 /*
619  * NB: Return the value of the first error return code.  Subsequent
620  *     errors after the first one are ignored.
621  */
622 /*
623  * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE
624  * operation.  If nfs_writedata_alloc() or get_user_pages() fails,
625  * bail and stop sending more writes.  Write length accounting is
626  * handled automatically by nfs_direct_write_result().  Otherwise, if
627  * no requests have been sent, just return an error.
628  */
629 static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *desc,
630 						 const struct iovec *iov,
631 						 loff_t pos, bool uio)
632 {
633 	struct nfs_direct_req *dreq = desc->pg_dreq;
634 	struct nfs_open_context *ctx = dreq->ctx;
635 	struct inode *inode = ctx->dentry->d_inode;
636 	unsigned long user_addr = (unsigned long)iov->iov_base;
637 	size_t count = iov->iov_len;
638 	size_t wsize = NFS_SERVER(inode)->wsize;
639 	unsigned int pgbase;
640 	int result;
641 	ssize_t started = 0;
642 	struct page **pagevec = NULL;
643 	unsigned int npages;
644 
645 	do {
646 		size_t bytes;
647 		int i;
648 
649 		pgbase = user_addr & ~PAGE_MASK;
650 		bytes = min(max_t(size_t, wsize, PAGE_SIZE), count);
651 
652 		result = -ENOMEM;
653 		npages = nfs_page_array_len(pgbase, bytes);
654 		if (!pagevec)
655 			pagevec = kmalloc(npages * sizeof(struct page *), GFP_KERNEL);
656 		if (!pagevec)
657 			break;
658 
659 		if (uio) {
660 			down_read(&current->mm->mmap_sem);
661 			result = get_user_pages(current, current->mm, user_addr,
662 						npages, 0, 0, pagevec, NULL);
663 			up_read(&current->mm->mmap_sem);
664 			if (result < 0)
665 				break;
666 		} else {
667 			WARN_ON(npages != 1);
668 			result = get_kernel_page(user_addr, 0, pagevec);
669 			if (WARN_ON(result != 1))
670 				break;
671 		}
672 
673 		if ((unsigned)result < npages) {
674 			bytes = result * PAGE_SIZE;
675 			if (bytes <= pgbase) {
676 				nfs_direct_release_pages(pagevec, result);
677 				break;
678 			}
679 			bytes -= pgbase;
680 			npages = result;
681 		}
682 
683 		for (i = 0; i < npages; i++) {
684 			struct nfs_page *req;
685 			unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase);
686 
687 			req = nfs_create_request(dreq->ctx, dreq->inode,
688 						 pagevec[i],
689 						 pgbase, req_len);
690 			if (IS_ERR(req)) {
691 				result = PTR_ERR(req);
692 				break;
693 			}
694 			nfs_lock_request(req);
695 			req->wb_index = pos >> PAGE_SHIFT;
696 			req->wb_offset = pos & ~PAGE_MASK;
697 			if (!nfs_pageio_add_request(desc, req)) {
698 				result = desc->pg_error;
699 				nfs_unlock_and_release_request(req);
700 				break;
701 			}
702 			pgbase = 0;
703 			bytes -= req_len;
704 			started += req_len;
705 			user_addr += req_len;
706 			pos += req_len;
707 			count -= req_len;
708 			dreq->bytes_left -= req_len;
709 		}
710 		/* The nfs_page now hold references to these pages */
711 		nfs_direct_release_pages(pagevec, npages);
712 	} while (count != 0 && result >= 0);
713 
714 	kfree(pagevec);
715 
716 	if (started)
717 		return started;
718 	return result < 0 ? (ssize_t) result : -EFAULT;
719 }
720 
721 static void nfs_direct_write_completion(struct nfs_pgio_header *hdr)
722 {
723 	struct nfs_direct_req *dreq = hdr->dreq;
724 	struct nfs_commit_info cinfo;
725 	int bit = -1;
726 	struct nfs_page *req = nfs_list_entry(hdr->pages.next);
727 
728 	if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
729 		goto out_put;
730 
731 	nfs_init_cinfo_from_dreq(&cinfo, dreq);
732 
733 	spin_lock(&dreq->lock);
734 
735 	if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) {
736 		dreq->flags = 0;
737 		dreq->error = hdr->error;
738 	}
739 	if (dreq->error != 0)
740 		bit = NFS_IOHDR_ERROR;
741 	else {
742 		dreq->count += hdr->good_bytes;
743 		if (test_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags)) {
744 			dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
745 			bit = NFS_IOHDR_NEED_RESCHED;
746 		} else if (test_bit(NFS_IOHDR_NEED_COMMIT, &hdr->flags)) {
747 			if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES)
748 				bit = NFS_IOHDR_NEED_RESCHED;
749 			else if (dreq->flags == 0) {
750 				memcpy(&dreq->verf, hdr->verf,
751 				       sizeof(dreq->verf));
752 				bit = NFS_IOHDR_NEED_COMMIT;
753 				dreq->flags = NFS_ODIRECT_DO_COMMIT;
754 			} else if (dreq->flags == NFS_ODIRECT_DO_COMMIT) {
755 				if (memcmp(&dreq->verf, hdr->verf, sizeof(dreq->verf))) {
756 					dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
757 					bit = NFS_IOHDR_NEED_RESCHED;
758 				} else
759 					bit = NFS_IOHDR_NEED_COMMIT;
760 			}
761 		}
762 	}
763 	spin_unlock(&dreq->lock);
764 
765 	while (!list_empty(&hdr->pages)) {
766 		req = nfs_list_entry(hdr->pages.next);
767 		nfs_list_remove_request(req);
768 		switch (bit) {
769 		case NFS_IOHDR_NEED_RESCHED:
770 		case NFS_IOHDR_NEED_COMMIT:
771 			kref_get(&req->wb_kref);
772 			nfs_mark_request_commit(req, hdr->lseg, &cinfo);
773 		}
774 		nfs_unlock_and_release_request(req);
775 	}
776 
777 out_put:
778 	if (put_dreq(dreq))
779 		nfs_direct_write_complete(dreq, hdr->inode);
780 	hdr->release(hdr);
781 }
782 
783 static void nfs_write_sync_pgio_error(struct list_head *head)
784 {
785 	struct nfs_page *req;
786 
787 	while (!list_empty(head)) {
788 		req = nfs_list_entry(head->next);
789 		nfs_list_remove_request(req);
790 		nfs_unlock_and_release_request(req);
791 	}
792 }
793 
794 static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = {
795 	.error_cleanup = nfs_write_sync_pgio_error,
796 	.init_hdr = nfs_direct_pgio_init,
797 	.completion = nfs_direct_write_completion,
798 };
799 
800 static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
801 					       const struct iovec *iov,
802 					       unsigned long nr_segs,
803 					       loff_t pos, bool uio)
804 {
805 	struct nfs_pageio_descriptor desc;
806 	struct inode *inode = dreq->inode;
807 	ssize_t result = 0;
808 	size_t requested_bytes = 0;
809 	unsigned long seg;
810 
811 	NFS_PROTO(inode)->write_pageio_init(&desc, inode, FLUSH_COND_STABLE,
812 			      &nfs_direct_write_completion_ops);
813 	desc.pg_dreq = dreq;
814 	get_dreq(dreq);
815 	atomic_inc(&inode->i_dio_count);
816 
817 	NFS_I(dreq->inode)->write_io += iov_length(iov, nr_segs);
818 	for (seg = 0; seg < nr_segs; seg++) {
819 		const struct iovec *vec = &iov[seg];
820 		result = nfs_direct_write_schedule_segment(&desc, vec, pos, uio);
821 		if (result < 0)
822 			break;
823 		requested_bytes += result;
824 		if ((size_t)result < vec->iov_len)
825 			break;
826 		pos += vec->iov_len;
827 	}
828 	nfs_pageio_complete(&desc);
829 
830 	/*
831 	 * If no bytes were started, return the error, and let the
832 	 * generic layer handle the completion.
833 	 */
834 	if (requested_bytes == 0) {
835 		inode_dio_done(inode);
836 		nfs_direct_req_release(dreq);
837 		return result < 0 ? result : -EIO;
838 	}
839 
840 	if (put_dreq(dreq))
841 		nfs_direct_write_complete(dreq, dreq->inode);
842 	return 0;
843 }
844 
845 static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov,
846 				unsigned long nr_segs, loff_t pos,
847 				size_t count, bool uio)
848 {
849 	ssize_t result = -ENOMEM;
850 	struct inode *inode = iocb->ki_filp->f_mapping->host;
851 	struct nfs_direct_req *dreq;
852 	struct nfs_lock_context *l_ctx;
853 
854 	dreq = nfs_direct_req_alloc();
855 	if (!dreq)
856 		goto out;
857 
858 	dreq->inode = inode;
859 	dreq->bytes_left = count;
860 	dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
861 	l_ctx = nfs_get_lock_context(dreq->ctx);
862 	if (IS_ERR(l_ctx)) {
863 		result = PTR_ERR(l_ctx);
864 		goto out_release;
865 	}
866 	dreq->l_ctx = l_ctx;
867 	if (!is_sync_kiocb(iocb))
868 		dreq->iocb = iocb;
869 
870 	result = nfs_direct_write_schedule_iovec(dreq, iov, nr_segs, pos, uio);
871 	if (!result)
872 		result = nfs_direct_wait(dreq);
873 out_release:
874 	nfs_direct_req_release(dreq);
875 out:
876 	return result;
877 }
878 
879 /**
880  * nfs_file_direct_read - file direct read operation for NFS files
881  * @iocb: target I/O control block
882  * @iov: vector of user buffers into which to read data
883  * @nr_segs: size of iov vector
884  * @pos: byte offset in file where reading starts
885  *
886  * We use this function for direct reads instead of calling
887  * generic_file_aio_read() in order to avoid gfar's check to see if
888  * the request starts before the end of the file.  For that check
889  * to work, we must generate a GETATTR before each direct read, and
890  * even then there is a window between the GETATTR and the subsequent
891  * READ where the file size could change.  Our preference is simply
892  * to do all reads the application wants, and the server will take
893  * care of managing the end of file boundary.
894  *
895  * This function also eliminates unnecessarily updating the file's
896  * atime locally, as the NFS server sets the file's atime, and this
897  * client must read the updated atime from the server back into its
898  * cache.
899  */
900 ssize_t nfs_file_direct_read(struct kiocb *iocb, const struct iovec *iov,
901 				unsigned long nr_segs, loff_t pos, bool uio)
902 {
903 	ssize_t retval = -EINVAL;
904 	struct file *file = iocb->ki_filp;
905 	struct address_space *mapping = file->f_mapping;
906 	size_t count;
907 
908 	count = iov_length(iov, nr_segs);
909 	nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count);
910 
911 	dfprintk(FILE, "NFS: direct read(%pD2, %zd@%Ld)\n",
912 		file, count, (long long) pos);
913 
914 	retval = 0;
915 	if (!count)
916 		goto out;
917 
918 	retval = nfs_sync_mapping(mapping);
919 	if (retval)
920 		goto out;
921 
922 	task_io_account_read(count);
923 
924 	retval = nfs_direct_read(iocb, iov, nr_segs, pos, uio);
925 	if (retval > 0)
926 		iocb->ki_pos = pos + retval;
927 
928 out:
929 	return retval;
930 }
931 
932 /**
933  * nfs_file_direct_write - file direct write operation for NFS files
934  * @iocb: target I/O control block
935  * @iov: vector of user buffers from which to write data
936  * @nr_segs: size of iov vector
937  * @pos: byte offset in file where writing starts
938  *
939  * We use this function for direct writes instead of calling
940  * generic_file_aio_write() in order to avoid taking the inode
941  * semaphore and updating the i_size.  The NFS server will set
942  * the new i_size and this client must read the updated size
943  * back into its cache.  We let the server do generic write
944  * parameter checking and report problems.
945  *
946  * We eliminate local atime updates, see direct read above.
947  *
948  * We avoid unnecessary page cache invalidations for normal cached
949  * readers of this file.
950  *
951  * Note that O_APPEND is not supported for NFS direct writes, as there
952  * is no atomic O_APPEND write facility in the NFS protocol.
953  */
954 ssize_t nfs_file_direct_write(struct kiocb *iocb, const struct iovec *iov,
955 				unsigned long nr_segs, loff_t pos, bool uio)
956 {
957 	ssize_t retval = -EINVAL;
958 	struct file *file = iocb->ki_filp;
959 	struct address_space *mapping = file->f_mapping;
960 	size_t count;
961 
962 	count = iov_length(iov, nr_segs);
963 	nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count);
964 
965 	dfprintk(FILE, "NFS: direct write(%pD2, %zd@%Ld)\n",
966 		file, count, (long long) pos);
967 
968 	retval = generic_write_checks(file, &pos, &count, 0);
969 	if (retval)
970 		goto out;
971 
972 	retval = -EINVAL;
973 	if ((ssize_t) count < 0)
974 		goto out;
975 	retval = 0;
976 	if (!count)
977 		goto out;
978 
979 	retval = nfs_sync_mapping(mapping);
980 	if (retval)
981 		goto out;
982 
983 	task_io_account_write(count);
984 
985 	retval = nfs_direct_write(iocb, iov, nr_segs, pos, count, uio);
986 	if (retval > 0) {
987 		struct inode *inode = mapping->host;
988 
989 		iocb->ki_pos = pos + retval;
990 		spin_lock(&inode->i_lock);
991 		if (i_size_read(inode) < iocb->ki_pos)
992 			i_size_write(inode, iocb->ki_pos);
993 		spin_unlock(&inode->i_lock);
994 	}
995 out:
996 	return retval;
997 }
998 
999 /**
1000  * nfs_init_directcache - create a slab cache for nfs_direct_req structures
1001  *
1002  */
1003 int __init nfs_init_directcache(void)
1004 {
1005 	nfs_direct_cachep = kmem_cache_create("nfs_direct_cache",
1006 						sizeof(struct nfs_direct_req),
1007 						0, (SLAB_RECLAIM_ACCOUNT|
1008 							SLAB_MEM_SPREAD),
1009 						NULL);
1010 	if (nfs_direct_cachep == NULL)
1011 		return -ENOMEM;
1012 
1013 	return 0;
1014 }
1015 
1016 /**
1017  * nfs_destroy_directcache - destroy the slab cache for nfs_direct_req structures
1018  *
1019  */
1020 void nfs_destroy_directcache(void)
1021 {
1022 	kmem_cache_destroy(nfs_direct_cachep);
1023 }
1024