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
50 #include <linux/nfs_fs.h>
51 #include <linux/nfs_page.h>
52 #include <linux/sunrpc/clnt.h>
53
54 #include <asm/uaccess.h>
55 #include <linux/atomic.h>
56
57 #include "internal.h"
58 #include "iostat.h"
59
60 #define NFSDBG_FACILITY NFSDBG_VFS
61
62 static struct kmem_cache *nfs_direct_cachep;
63
64 /*
65 * This represents a set of asynchronous requests that we're waiting on
66 */
67 struct nfs_direct_req {
68 struct kref kref; /* release manager */
69
70 /* I/O parameters */
71 struct nfs_open_context *ctx; /* file open context info */
72 struct nfs_lock_context *l_ctx; /* Lock context info */
73 struct kiocb * iocb; /* controlling i/o request */
74 struct inode * inode; /* target file of i/o */
75
76 /* completion state */
77 atomic_t io_count; /* i/os we're waiting for */
78 spinlock_t lock; /* protect completion state */
79 ssize_t count, /* bytes actually processed */
80 error; /* any reported error */
81 struct completion completion; /* wait for i/o completion */
82
83 /* commit state */
84 struct list_head rewrite_list; /* saved nfs_write_data structs */
85 struct nfs_write_data * commit_data; /* special write_data for commits */
86 int flags;
87 #define NFS_ODIRECT_DO_COMMIT (1) /* an unstable reply was received */
88 #define NFS_ODIRECT_RESCHED_WRITES (2) /* write verification failed */
89 struct nfs_writeverf verf; /* unstable write verifier */
90 };
91
92 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode);
93 static const struct rpc_call_ops nfs_write_direct_ops;
94
get_dreq(struct nfs_direct_req * dreq)95 static inline void get_dreq(struct nfs_direct_req *dreq)
96 {
97 atomic_inc(&dreq->io_count);
98 }
99
put_dreq(struct nfs_direct_req * dreq)100 static inline int put_dreq(struct nfs_direct_req *dreq)
101 {
102 return atomic_dec_and_test(&dreq->io_count);
103 }
104
105 /**
106 * nfs_direct_IO - NFS address space operation for direct I/O
107 * @rw: direction (read or write)
108 * @iocb: target I/O control block
109 * @iov: array of vectors that define I/O buffer
110 * @pos: offset in file to begin the operation
111 * @nr_segs: size of iovec array
112 *
113 * The presence of this routine in the address space ops vector means
114 * the NFS client supports direct I/O. However, we shunt off direct
115 * read and write requests before the VFS gets them, so this method
116 * should never be called.
117 */
nfs_direct_IO(int rw,struct kiocb * iocb,const struct iovec * iov,loff_t pos,unsigned long nr_segs)118 ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t pos, unsigned long nr_segs)
119 {
120 dprintk("NFS: nfs_direct_IO (%s) off/no(%Ld/%lu) EINVAL\n",
121 iocb->ki_filp->f_path.dentry->d_name.name,
122 (long long) pos, nr_segs);
123
124 return -EINVAL;
125 }
126
nfs_direct_dirty_pages(struct page ** pages,unsigned int pgbase,size_t count)127 static void nfs_direct_dirty_pages(struct page **pages, unsigned int pgbase, size_t count)
128 {
129 unsigned int npages;
130 unsigned int i;
131
132 if (count == 0)
133 return;
134 pages += (pgbase >> PAGE_SHIFT);
135 npages = (count + (pgbase & ~PAGE_MASK) + PAGE_SIZE - 1) >> PAGE_SHIFT;
136 for (i = 0; i < npages; i++) {
137 struct page *page = pages[i];
138 if (!PageCompound(page))
139 set_page_dirty(page);
140 }
141 }
142
nfs_direct_release_pages(struct page ** pages,unsigned int npages)143 static void nfs_direct_release_pages(struct page **pages, unsigned int npages)
144 {
145 unsigned int i;
146 for (i = 0; i < npages; i++)
147 page_cache_release(pages[i]);
148 }
149
nfs_direct_req_alloc(void)150 static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
151 {
152 struct nfs_direct_req *dreq;
153
154 dreq = kmem_cache_alloc(nfs_direct_cachep, GFP_KERNEL);
155 if (!dreq)
156 return NULL;
157
158 kref_init(&dreq->kref);
159 kref_get(&dreq->kref);
160 init_completion(&dreq->completion);
161 INIT_LIST_HEAD(&dreq->rewrite_list);
162 dreq->iocb = NULL;
163 dreq->ctx = NULL;
164 dreq->l_ctx = NULL;
165 spin_lock_init(&dreq->lock);
166 atomic_set(&dreq->io_count, 0);
167 dreq->count = 0;
168 dreq->error = 0;
169 dreq->flags = 0;
170
171 return dreq;
172 }
173
nfs_direct_req_free(struct kref * kref)174 static void nfs_direct_req_free(struct kref *kref)
175 {
176 struct nfs_direct_req *dreq = container_of(kref, struct nfs_direct_req, kref);
177
178 if (dreq->l_ctx != NULL)
179 nfs_put_lock_context(dreq->l_ctx);
180 if (dreq->ctx != NULL)
181 put_nfs_open_context(dreq->ctx);
182 kmem_cache_free(nfs_direct_cachep, dreq);
183 }
184
nfs_direct_req_release(struct nfs_direct_req * dreq)185 static void nfs_direct_req_release(struct nfs_direct_req *dreq)
186 {
187 kref_put(&dreq->kref, nfs_direct_req_free);
188 }
189
190 /*
191 * Collects and returns the final error value/byte-count.
192 */
nfs_direct_wait(struct nfs_direct_req * dreq)193 static ssize_t nfs_direct_wait(struct nfs_direct_req *dreq)
194 {
195 ssize_t result = -EIOCBQUEUED;
196
197 /* Async requests don't wait here */
198 if (dreq->iocb)
199 goto out;
200
201 result = wait_for_completion_killable(&dreq->completion);
202
203 if (!result)
204 result = dreq->error;
205 if (!result)
206 result = dreq->count;
207
208 out:
209 return (ssize_t) result;
210 }
211
212 /*
213 * Synchronous I/O uses a stack-allocated iocb. Thus we can't trust
214 * the iocb is still valid here if this is a synchronous request.
215 */
nfs_direct_complete(struct nfs_direct_req * dreq)216 static void nfs_direct_complete(struct nfs_direct_req *dreq)
217 {
218 if (dreq->iocb) {
219 long res = (long) dreq->error;
220 if (!res)
221 res = (long) dreq->count;
222 aio_complete(dreq->iocb, res, 0);
223 }
224 complete_all(&dreq->completion);
225
226 nfs_direct_req_release(dreq);
227 }
228
229 /*
230 * We must hold a reference to all the pages in this direct read request
231 * until the RPCs complete. This could be long *after* we are woken up in
232 * nfs_direct_wait (for instance, if someone hits ^C on a slow server).
233 */
nfs_direct_read_result(struct rpc_task * task,void * calldata)234 static void nfs_direct_read_result(struct rpc_task *task, void *calldata)
235 {
236 struct nfs_read_data *data = calldata;
237
238 nfs_readpage_result(task, data);
239 }
240
nfs_direct_read_release(void * calldata)241 static void nfs_direct_read_release(void *calldata)
242 {
243
244 struct nfs_read_data *data = calldata;
245 struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
246 int status = data->task.tk_status;
247
248 spin_lock(&dreq->lock);
249 if (unlikely(status < 0)) {
250 dreq->error = status;
251 spin_unlock(&dreq->lock);
252 } else {
253 dreq->count += data->res.count;
254 spin_unlock(&dreq->lock);
255 nfs_direct_dirty_pages(data->pagevec,
256 data->args.pgbase,
257 data->res.count);
258 }
259 nfs_direct_release_pages(data->pagevec, data->npages);
260
261 if (put_dreq(dreq))
262 nfs_direct_complete(dreq);
263 nfs_readdata_free(data);
264 }
265
266 static const struct rpc_call_ops nfs_read_direct_ops = {
267 .rpc_call_prepare = nfs_read_prepare,
268 .rpc_call_done = nfs_direct_read_result,
269 .rpc_release = nfs_direct_read_release,
270 };
271
272 /*
273 * For each rsize'd chunk of the user's buffer, dispatch an NFS READ
274 * operation. If nfs_readdata_alloc() or get_user_pages() fails,
275 * bail and stop sending more reads. Read length accounting is
276 * handled automatically by nfs_direct_read_result(). Otherwise, if
277 * no requests have been sent, just return an error.
278 */
nfs_direct_read_schedule_segment(struct nfs_direct_req * dreq,const struct iovec * iov,loff_t pos)279 static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq,
280 const struct iovec *iov,
281 loff_t pos)
282 {
283 struct nfs_open_context *ctx = dreq->ctx;
284 struct inode *inode = ctx->dentry->d_inode;
285 unsigned long user_addr = (unsigned long)iov->iov_base;
286 size_t count = iov->iov_len;
287 size_t rsize = NFS_SERVER(inode)->rsize;
288 struct rpc_task *task;
289 struct rpc_message msg = {
290 .rpc_cred = ctx->cred,
291 };
292 struct rpc_task_setup task_setup_data = {
293 .rpc_client = NFS_CLIENT(inode),
294 .rpc_message = &msg,
295 .callback_ops = &nfs_read_direct_ops,
296 .workqueue = nfsiod_workqueue,
297 .flags = RPC_TASK_ASYNC,
298 };
299 unsigned int pgbase;
300 int result;
301 ssize_t started = 0;
302
303 do {
304 struct nfs_read_data *data;
305 size_t bytes;
306
307 pgbase = user_addr & ~PAGE_MASK;
308 bytes = min(rsize,count);
309
310 result = -ENOMEM;
311 data = nfs_readdata_alloc(nfs_page_array_len(pgbase, bytes));
312 if (unlikely(!data))
313 break;
314
315 down_read(¤t->mm->mmap_sem);
316 result = get_user_pages(current, current->mm, user_addr,
317 data->npages, 1, 0, data->pagevec, NULL);
318 up_read(¤t->mm->mmap_sem);
319 if (result < 0) {
320 nfs_readdata_free(data);
321 break;
322 }
323 if ((unsigned)result < data->npages) {
324 bytes = result * PAGE_SIZE;
325 if (bytes <= pgbase) {
326 nfs_direct_release_pages(data->pagevec, result);
327 nfs_readdata_free(data);
328 break;
329 }
330 bytes -= pgbase;
331 data->npages = result;
332 }
333
334 get_dreq(dreq);
335
336 data->req = (struct nfs_page *) dreq;
337 data->inode = inode;
338 data->cred = msg.rpc_cred;
339 data->args.fh = NFS_FH(inode);
340 data->args.context = ctx;
341 data->args.lock_context = dreq->l_ctx;
342 data->args.offset = pos;
343 data->args.pgbase = pgbase;
344 data->args.pages = data->pagevec;
345 data->args.count = bytes;
346 data->res.fattr = &data->fattr;
347 data->res.eof = 0;
348 data->res.count = bytes;
349 nfs_fattr_init(&data->fattr);
350 msg.rpc_argp = &data->args;
351 msg.rpc_resp = &data->res;
352
353 task_setup_data.task = &data->task;
354 task_setup_data.callback_data = data;
355 NFS_PROTO(inode)->read_setup(data, &msg);
356
357 task = rpc_run_task(&task_setup_data);
358 if (IS_ERR(task))
359 break;
360 rpc_put_task(task);
361
362 dprintk("NFS: %5u initiated direct read call "
363 "(req %s/%Ld, %zu bytes @ offset %Lu)\n",
364 data->task.tk_pid,
365 inode->i_sb->s_id,
366 (long long)NFS_FILEID(inode),
367 bytes,
368 (unsigned long long)data->args.offset);
369
370 started += bytes;
371 user_addr += bytes;
372 pos += bytes;
373 /* FIXME: Remove this unnecessary math from final patch */
374 pgbase += bytes;
375 pgbase &= ~PAGE_MASK;
376 BUG_ON(pgbase != (user_addr & ~PAGE_MASK));
377
378 count -= bytes;
379 } while (count != 0);
380
381 if (started)
382 return started;
383 return result < 0 ? (ssize_t) result : -EFAULT;
384 }
385
nfs_direct_read_schedule_iovec(struct nfs_direct_req * dreq,const struct iovec * iov,unsigned long nr_segs,loff_t pos)386 static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq,
387 const struct iovec *iov,
388 unsigned long nr_segs,
389 loff_t pos)
390 {
391 ssize_t result = -EINVAL;
392 size_t requested_bytes = 0;
393 unsigned long seg;
394
395 get_dreq(dreq);
396
397 for (seg = 0; seg < nr_segs; seg++) {
398 const struct iovec *vec = &iov[seg];
399 result = nfs_direct_read_schedule_segment(dreq, vec, pos);
400 if (result < 0)
401 break;
402 requested_bytes += result;
403 if ((size_t)result < vec->iov_len)
404 break;
405 pos += vec->iov_len;
406 }
407
408 /*
409 * If no bytes were started, return the error, and let the
410 * generic layer handle the completion.
411 */
412 if (requested_bytes == 0) {
413 nfs_direct_req_release(dreq);
414 return result < 0 ? result : -EIO;
415 }
416
417 if (put_dreq(dreq))
418 nfs_direct_complete(dreq);
419 return 0;
420 }
421
nfs_direct_read(struct kiocb * iocb,const struct iovec * iov,unsigned long nr_segs,loff_t pos)422 static ssize_t nfs_direct_read(struct kiocb *iocb, const struct iovec *iov,
423 unsigned long nr_segs, loff_t pos)
424 {
425 ssize_t result = -ENOMEM;
426 struct inode *inode = iocb->ki_filp->f_mapping->host;
427 struct nfs_direct_req *dreq;
428
429 dreq = nfs_direct_req_alloc();
430 if (dreq == NULL)
431 goto out;
432
433 dreq->inode = inode;
434 dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
435 dreq->l_ctx = nfs_get_lock_context(dreq->ctx);
436 if (dreq->l_ctx == NULL)
437 goto out_release;
438 if (!is_sync_kiocb(iocb))
439 dreq->iocb = iocb;
440
441 result = nfs_direct_read_schedule_iovec(dreq, iov, nr_segs, pos);
442 if (!result)
443 result = nfs_direct_wait(dreq);
444 out_release:
445 nfs_direct_req_release(dreq);
446 out:
447 return result;
448 }
449
nfs_direct_free_writedata(struct nfs_direct_req * dreq)450 static void nfs_direct_free_writedata(struct nfs_direct_req *dreq)
451 {
452 while (!list_empty(&dreq->rewrite_list)) {
453 struct nfs_write_data *data = list_entry(dreq->rewrite_list.next, struct nfs_write_data, pages);
454 list_del(&data->pages);
455 nfs_direct_release_pages(data->pagevec, data->npages);
456 nfs_writedata_free(data);
457 }
458 }
459
460 #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
nfs_direct_write_reschedule(struct nfs_direct_req * dreq)461 static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
462 {
463 struct inode *inode = dreq->inode;
464 struct list_head *p;
465 struct nfs_write_data *data;
466 struct rpc_task *task;
467 struct rpc_message msg = {
468 .rpc_cred = dreq->ctx->cred,
469 };
470 struct rpc_task_setup task_setup_data = {
471 .rpc_client = NFS_CLIENT(inode),
472 .rpc_message = &msg,
473 .callback_ops = &nfs_write_direct_ops,
474 .workqueue = nfsiod_workqueue,
475 .flags = RPC_TASK_ASYNC,
476 };
477
478 dreq->count = 0;
479 get_dreq(dreq);
480
481 list_for_each(p, &dreq->rewrite_list) {
482 data = list_entry(p, struct nfs_write_data, pages);
483
484 get_dreq(dreq);
485
486 /* Use stable writes */
487 data->args.stable = NFS_FILE_SYNC;
488
489 /*
490 * Reset data->res.
491 */
492 nfs_fattr_init(&data->fattr);
493 data->res.count = data->args.count;
494 memset(&data->verf, 0, sizeof(data->verf));
495
496 /*
497 * Reuse data->task; data->args should not have changed
498 * since the original request was sent.
499 */
500 task_setup_data.task = &data->task;
501 task_setup_data.callback_data = data;
502 msg.rpc_argp = &data->args;
503 msg.rpc_resp = &data->res;
504 NFS_PROTO(inode)->write_setup(data, &msg);
505
506 /*
507 * We're called via an RPC callback, so BKL is already held.
508 */
509 task = rpc_run_task(&task_setup_data);
510 if (!IS_ERR(task))
511 rpc_put_task(task);
512
513 dprintk("NFS: %5u rescheduled direct write call (req %s/%Ld, %u bytes @ offset %Lu)\n",
514 data->task.tk_pid,
515 inode->i_sb->s_id,
516 (long long)NFS_FILEID(inode),
517 data->args.count,
518 (unsigned long long)data->args.offset);
519 }
520
521 if (put_dreq(dreq))
522 nfs_direct_write_complete(dreq, inode);
523 }
524
nfs_direct_commit_result(struct rpc_task * task,void * calldata)525 static void nfs_direct_commit_result(struct rpc_task *task, void *calldata)
526 {
527 struct nfs_write_data *data = calldata;
528
529 /* Call the NFS version-specific code */
530 NFS_PROTO(data->inode)->commit_done(task, data);
531 }
532
nfs_direct_commit_release(void * calldata)533 static void nfs_direct_commit_release(void *calldata)
534 {
535 struct nfs_write_data *data = calldata;
536 struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
537 int status = data->task.tk_status;
538
539 if (status < 0) {
540 dprintk("NFS: %5u commit failed with error %d.\n",
541 data->task.tk_pid, status);
542 dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
543 } else if (memcmp(&dreq->verf, &data->verf, sizeof(data->verf))) {
544 dprintk("NFS: %5u commit verify failed\n", data->task.tk_pid);
545 dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
546 }
547
548 dprintk("NFS: %5u commit returned %d\n", data->task.tk_pid, status);
549 nfs_direct_write_complete(dreq, data->inode);
550 nfs_commit_free(data);
551 }
552
553 static const struct rpc_call_ops nfs_commit_direct_ops = {
554 .rpc_call_prepare = nfs_write_prepare,
555 .rpc_call_done = nfs_direct_commit_result,
556 .rpc_release = nfs_direct_commit_release,
557 };
558
nfs_direct_commit_schedule(struct nfs_direct_req * dreq)559 static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
560 {
561 struct nfs_write_data *data = dreq->commit_data;
562 struct rpc_task *task;
563 struct rpc_message msg = {
564 .rpc_argp = &data->args,
565 .rpc_resp = &data->res,
566 .rpc_cred = dreq->ctx->cred,
567 };
568 struct rpc_task_setup task_setup_data = {
569 .task = &data->task,
570 .rpc_client = NFS_CLIENT(dreq->inode),
571 .rpc_message = &msg,
572 .callback_ops = &nfs_commit_direct_ops,
573 .callback_data = data,
574 .workqueue = nfsiod_workqueue,
575 .flags = RPC_TASK_ASYNC,
576 };
577
578 data->inode = dreq->inode;
579 data->cred = msg.rpc_cred;
580
581 data->args.fh = NFS_FH(data->inode);
582 data->args.offset = 0;
583 data->args.count = 0;
584 data->args.context = dreq->ctx;
585 data->args.lock_context = dreq->l_ctx;
586 data->res.count = 0;
587 data->res.fattr = &data->fattr;
588 data->res.verf = &data->verf;
589 nfs_fattr_init(&data->fattr);
590
591 NFS_PROTO(data->inode)->commit_setup(data, &msg);
592
593 /* Note: task.tk_ops->rpc_release will free dreq->commit_data */
594 dreq->commit_data = NULL;
595
596 dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);
597
598 task = rpc_run_task(&task_setup_data);
599 if (!IS_ERR(task))
600 rpc_put_task(task);
601 }
602
nfs_direct_write_complete(struct nfs_direct_req * dreq,struct inode * inode)603 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
604 {
605 int flags = dreq->flags;
606
607 dreq->flags = 0;
608 switch (flags) {
609 case NFS_ODIRECT_DO_COMMIT:
610 nfs_direct_commit_schedule(dreq);
611 break;
612 case NFS_ODIRECT_RESCHED_WRITES:
613 nfs_direct_write_reschedule(dreq);
614 break;
615 default:
616 if (dreq->commit_data != NULL)
617 nfs_commit_free(dreq->commit_data);
618 nfs_direct_free_writedata(dreq);
619 nfs_zap_mapping(inode, inode->i_mapping);
620 nfs_direct_complete(dreq);
621 }
622 }
623
nfs_alloc_commit_data(struct nfs_direct_req * dreq)624 static void nfs_alloc_commit_data(struct nfs_direct_req *dreq)
625 {
626 dreq->commit_data = nfs_commitdata_alloc();
627 if (dreq->commit_data != NULL)
628 dreq->commit_data->req = (struct nfs_page *) dreq;
629 }
630 #else
nfs_alloc_commit_data(struct nfs_direct_req * dreq)631 static inline void nfs_alloc_commit_data(struct nfs_direct_req *dreq)
632 {
633 dreq->commit_data = NULL;
634 }
635
nfs_direct_write_complete(struct nfs_direct_req * dreq,struct inode * inode)636 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
637 {
638 nfs_direct_free_writedata(dreq);
639 nfs_zap_mapping(inode, inode->i_mapping);
640 nfs_direct_complete(dreq);
641 }
642 #endif
643
nfs_direct_write_result(struct rpc_task * task,void * calldata)644 static void nfs_direct_write_result(struct rpc_task *task, void *calldata)
645 {
646 struct nfs_write_data *data = calldata;
647
648 nfs_writeback_done(task, data);
649 }
650
651 /*
652 * NB: Return the value of the first error return code. Subsequent
653 * errors after the first one are ignored.
654 */
nfs_direct_write_release(void * calldata)655 static void nfs_direct_write_release(void *calldata)
656 {
657 struct nfs_write_data *data = calldata;
658 struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
659 int status = data->task.tk_status;
660
661 spin_lock(&dreq->lock);
662
663 if (unlikely(status < 0)) {
664 /* An error has occurred, so we should not commit */
665 dreq->flags = 0;
666 dreq->error = status;
667 }
668 if (unlikely(dreq->error != 0))
669 goto out_unlock;
670
671 dreq->count += data->res.count;
672
673 if (data->res.verf->committed != NFS_FILE_SYNC) {
674 switch (dreq->flags) {
675 case 0:
676 memcpy(&dreq->verf, &data->verf, sizeof(dreq->verf));
677 dreq->flags = NFS_ODIRECT_DO_COMMIT;
678 break;
679 case NFS_ODIRECT_DO_COMMIT:
680 if (memcmp(&dreq->verf, &data->verf, sizeof(dreq->verf))) {
681 dprintk("NFS: %5u write verify failed\n", data->task.tk_pid);
682 dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
683 }
684 }
685 }
686 out_unlock:
687 spin_unlock(&dreq->lock);
688
689 if (put_dreq(dreq))
690 nfs_direct_write_complete(dreq, data->inode);
691 }
692
693 static const struct rpc_call_ops nfs_write_direct_ops = {
694 .rpc_call_prepare = nfs_write_prepare,
695 .rpc_call_done = nfs_direct_write_result,
696 .rpc_release = nfs_direct_write_release,
697 };
698
699 /*
700 * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE
701 * operation. If nfs_writedata_alloc() or get_user_pages() fails,
702 * bail and stop sending more writes. Write length accounting is
703 * handled automatically by nfs_direct_write_result(). Otherwise, if
704 * no requests have been sent, just return an error.
705 */
nfs_direct_write_schedule_segment(struct nfs_direct_req * dreq,const struct iovec * iov,loff_t pos,int sync)706 static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq,
707 const struct iovec *iov,
708 loff_t pos, int sync)
709 {
710 struct nfs_open_context *ctx = dreq->ctx;
711 struct inode *inode = ctx->dentry->d_inode;
712 unsigned long user_addr = (unsigned long)iov->iov_base;
713 size_t count = iov->iov_len;
714 struct rpc_task *task;
715 struct rpc_message msg = {
716 .rpc_cred = ctx->cred,
717 };
718 struct rpc_task_setup task_setup_data = {
719 .rpc_client = NFS_CLIENT(inode),
720 .rpc_message = &msg,
721 .callback_ops = &nfs_write_direct_ops,
722 .workqueue = nfsiod_workqueue,
723 .flags = RPC_TASK_ASYNC,
724 };
725 size_t wsize = NFS_SERVER(inode)->wsize;
726 unsigned int pgbase;
727 int result;
728 ssize_t started = 0;
729
730 do {
731 struct nfs_write_data *data;
732 size_t bytes;
733
734 pgbase = user_addr & ~PAGE_MASK;
735 bytes = min(wsize,count);
736
737 result = -ENOMEM;
738 data = nfs_writedata_alloc(nfs_page_array_len(pgbase, bytes));
739 if (unlikely(!data))
740 break;
741
742 down_read(¤t->mm->mmap_sem);
743 result = get_user_pages(current, current->mm, user_addr,
744 data->npages, 0, 0, data->pagevec, NULL);
745 up_read(¤t->mm->mmap_sem);
746 if (result < 0) {
747 nfs_writedata_free(data);
748 break;
749 }
750 if ((unsigned)result < data->npages) {
751 bytes = result * PAGE_SIZE;
752 if (bytes <= pgbase) {
753 nfs_direct_release_pages(data->pagevec, result);
754 nfs_writedata_free(data);
755 break;
756 }
757 bytes -= pgbase;
758 data->npages = result;
759 }
760
761 get_dreq(dreq);
762
763 list_move_tail(&data->pages, &dreq->rewrite_list);
764
765 data->req = (struct nfs_page *) dreq;
766 data->inode = inode;
767 data->cred = msg.rpc_cred;
768 data->args.fh = NFS_FH(inode);
769 data->args.context = ctx;
770 data->args.lock_context = dreq->l_ctx;
771 data->args.offset = pos;
772 data->args.pgbase = pgbase;
773 data->args.pages = data->pagevec;
774 data->args.count = bytes;
775 data->args.stable = sync;
776 data->res.fattr = &data->fattr;
777 data->res.count = bytes;
778 data->res.verf = &data->verf;
779 nfs_fattr_init(&data->fattr);
780
781 task_setup_data.task = &data->task;
782 task_setup_data.callback_data = data;
783 msg.rpc_argp = &data->args;
784 msg.rpc_resp = &data->res;
785 NFS_PROTO(inode)->write_setup(data, &msg);
786
787 task = rpc_run_task(&task_setup_data);
788 if (IS_ERR(task))
789 break;
790 rpc_put_task(task);
791
792 dprintk("NFS: %5u initiated direct write call "
793 "(req %s/%Ld, %zu bytes @ offset %Lu)\n",
794 data->task.tk_pid,
795 inode->i_sb->s_id,
796 (long long)NFS_FILEID(inode),
797 bytes,
798 (unsigned long long)data->args.offset);
799
800 started += bytes;
801 user_addr += bytes;
802 pos += bytes;
803
804 /* FIXME: Remove this useless math from the final patch */
805 pgbase += bytes;
806 pgbase &= ~PAGE_MASK;
807 BUG_ON(pgbase != (user_addr & ~PAGE_MASK));
808
809 count -= bytes;
810 } while (count != 0);
811
812 if (started)
813 return started;
814 return result < 0 ? (ssize_t) result : -EFAULT;
815 }
816
nfs_direct_write_schedule_iovec(struct nfs_direct_req * dreq,const struct iovec * iov,unsigned long nr_segs,loff_t pos,int sync)817 static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
818 const struct iovec *iov,
819 unsigned long nr_segs,
820 loff_t pos, int sync)
821 {
822 ssize_t result = 0;
823 size_t requested_bytes = 0;
824 unsigned long seg;
825
826 get_dreq(dreq);
827
828 for (seg = 0; seg < nr_segs; seg++) {
829 const struct iovec *vec = &iov[seg];
830 result = nfs_direct_write_schedule_segment(dreq, vec,
831 pos, sync);
832 if (result < 0)
833 break;
834 requested_bytes += result;
835 if ((size_t)result < vec->iov_len)
836 break;
837 pos += vec->iov_len;
838 }
839
840 /*
841 * If no bytes were started, return the error, and let the
842 * generic layer handle the completion.
843 */
844 if (requested_bytes == 0) {
845 nfs_direct_req_release(dreq);
846 return result < 0 ? result : -EIO;
847 }
848
849 if (put_dreq(dreq))
850 nfs_direct_write_complete(dreq, dreq->inode);
851 return 0;
852 }
853
nfs_direct_write(struct kiocb * iocb,const struct iovec * iov,unsigned long nr_segs,loff_t pos,size_t count)854 static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov,
855 unsigned long nr_segs, loff_t pos,
856 size_t count)
857 {
858 ssize_t result = -ENOMEM;
859 struct inode *inode = iocb->ki_filp->f_mapping->host;
860 struct nfs_direct_req *dreq;
861 size_t wsize = NFS_SERVER(inode)->wsize;
862 int sync = NFS_UNSTABLE;
863
864 dreq = nfs_direct_req_alloc();
865 if (!dreq)
866 goto out;
867 nfs_alloc_commit_data(dreq);
868
869 if (dreq->commit_data == NULL || count <= wsize)
870 sync = NFS_FILE_SYNC;
871
872 dreq->inode = inode;
873 dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
874 dreq->l_ctx = nfs_get_lock_context(dreq->ctx);
875 if (dreq->l_ctx == NULL)
876 goto out_release;
877 if (!is_sync_kiocb(iocb))
878 dreq->iocb = iocb;
879
880 result = nfs_direct_write_schedule_iovec(dreq, iov, nr_segs, pos, sync);
881 if (!result)
882 result = nfs_direct_wait(dreq);
883 out_release:
884 nfs_direct_req_release(dreq);
885 out:
886 return result;
887 }
888
889 /**
890 * nfs_file_direct_read - file direct read operation for NFS files
891 * @iocb: target I/O control block
892 * @iov: vector of user buffers into which to read data
893 * @nr_segs: size of iov vector
894 * @pos: byte offset in file where reading starts
895 *
896 * We use this function for direct reads instead of calling
897 * generic_file_aio_read() in order to avoid gfar's check to see if
898 * the request starts before the end of the file. For that check
899 * to work, we must generate a GETATTR before each direct read, and
900 * even then there is a window between the GETATTR and the subsequent
901 * READ where the file size could change. Our preference is simply
902 * to do all reads the application wants, and the server will take
903 * care of managing the end of file boundary.
904 *
905 * This function also eliminates unnecessarily updating the file's
906 * atime locally, as the NFS server sets the file's atime, and this
907 * client must read the updated atime from the server back into its
908 * cache.
909 */
nfs_file_direct_read(struct kiocb * iocb,const struct iovec * iov,unsigned long nr_segs,loff_t pos)910 ssize_t nfs_file_direct_read(struct kiocb *iocb, const struct iovec *iov,
911 unsigned long nr_segs, loff_t pos)
912 {
913 ssize_t retval = -EINVAL;
914 struct file *file = iocb->ki_filp;
915 struct address_space *mapping = file->f_mapping;
916 size_t count;
917
918 count = iov_length(iov, nr_segs);
919 nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count);
920
921 dfprintk(FILE, "NFS: direct read(%s/%s, %zd@%Ld)\n",
922 file->f_path.dentry->d_parent->d_name.name,
923 file->f_path.dentry->d_name.name,
924 count, (long long) pos);
925
926 retval = 0;
927 if (!count)
928 goto out;
929
930 retval = nfs_sync_mapping(mapping);
931 if (retval)
932 goto out;
933
934 task_io_account_read(count);
935
936 retval = nfs_direct_read(iocb, iov, nr_segs, pos);
937 if (retval > 0)
938 iocb->ki_pos = pos + retval;
939
940 out:
941 return retval;
942 }
943
944 /**
945 * nfs_file_direct_write - file direct write operation for NFS files
946 * @iocb: target I/O control block
947 * @iov: vector of user buffers from which to write data
948 * @nr_segs: size of iov vector
949 * @pos: byte offset in file where writing starts
950 *
951 * We use this function for direct writes instead of calling
952 * generic_file_aio_write() in order to avoid taking the inode
953 * semaphore and updating the i_size. The NFS server will set
954 * the new i_size and this client must read the updated size
955 * back into its cache. We let the server do generic write
956 * parameter checking and report problems.
957 *
958 * We eliminate local atime updates, see direct read above.
959 *
960 * We avoid unnecessary page cache invalidations for normal cached
961 * readers of this file.
962 *
963 * Note that O_APPEND is not supported for NFS direct writes, as there
964 * is no atomic O_APPEND write facility in the NFS protocol.
965 */
nfs_file_direct_write(struct kiocb * iocb,const struct iovec * iov,unsigned long nr_segs,loff_t pos)966 ssize_t nfs_file_direct_write(struct kiocb *iocb, const struct iovec *iov,
967 unsigned long nr_segs, loff_t pos)
968 {
969 ssize_t retval = -EINVAL;
970 struct file *file = iocb->ki_filp;
971 struct address_space *mapping = file->f_mapping;
972 size_t count;
973
974 count = iov_length(iov, nr_segs);
975 nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count);
976
977 dfprintk(FILE, "NFS: direct write(%s/%s, %zd@%Ld)\n",
978 file->f_path.dentry->d_parent->d_name.name,
979 file->f_path.dentry->d_name.name,
980 count, (long long) pos);
981
982 retval = generic_write_checks(file, &pos, &count, 0);
983 if (retval)
984 goto out;
985
986 retval = -EINVAL;
987 if ((ssize_t) count < 0)
988 goto out;
989 retval = 0;
990 if (!count)
991 goto out;
992
993 retval = nfs_sync_mapping(mapping);
994 if (retval)
995 goto out;
996
997 task_io_account_write(count);
998
999 retval = nfs_direct_write(iocb, iov, nr_segs, pos, count);
1000
1001 if (retval > 0)
1002 iocb->ki_pos = pos + retval;
1003
1004 out:
1005 return retval;
1006 }
1007
1008 /**
1009 * nfs_init_directcache - create a slab cache for nfs_direct_req structures
1010 *
1011 */
nfs_init_directcache(void)1012 int __init nfs_init_directcache(void)
1013 {
1014 nfs_direct_cachep = kmem_cache_create("nfs_direct_cache",
1015 sizeof(struct nfs_direct_req),
1016 0, (SLAB_RECLAIM_ACCOUNT|
1017 SLAB_MEM_SPREAD),
1018 NULL);
1019 if (nfs_direct_cachep == NULL)
1020 return -ENOMEM;
1021
1022 return 0;
1023 }
1024
1025 /**
1026 * nfs_destroy_directcache - destroy the slab cache for nfs_direct_req structures
1027 *
1028 */
nfs_destroy_directcache(void)1029 void nfs_destroy_directcache(void)
1030 {
1031 kmem_cache_destroy(nfs_direct_cachep);
1032 }
1033