]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/nfs/read.c | |
3 | * | |
4 | * Block I/O for NFS | |
5 | * | |
6 | * Partial copy of Linus' read cache modifications to fs/nfs/file.c | |
7 | * modified for async RPC by okir@monad.swb.de | |
1da177e4 LT |
8 | */ |
9 | ||
1da177e4 LT |
10 | #include <linux/time.h> |
11 | #include <linux/kernel.h> | |
12 | #include <linux/errno.h> | |
13 | #include <linux/fcntl.h> | |
14 | #include <linux/stat.h> | |
15 | #include <linux/mm.h> | |
16 | #include <linux/slab.h> | |
17 | #include <linux/pagemap.h> | |
18 | #include <linux/sunrpc/clnt.h> | |
19 | #include <linux/nfs_fs.h> | |
20 | #include <linux/nfs_page.h> | |
64419a9b | 21 | #include <linux/module.h> |
1da177e4 | 22 | |
bae724ef | 23 | #include "pnfs.h" |
1da177e4 | 24 | |
f11c88af | 25 | #include "nfs4_fs.h" |
49a70f27 | 26 | #include "internal.h" |
91d5b470 | 27 | #include "iostat.h" |
9a9fc1c0 | 28 | #include "fscache.h" |
91d5b470 | 29 | |
1da177e4 LT |
30 | #define NFSDBG_FACILITY NFSDBG_PAGECACHE |
31 | ||
1751c363 | 32 | static const struct nfs_pageio_ops nfs_pageio_read_ops; |
ec06c096 TM |
33 | static const struct rpc_call_ops nfs_read_partial_ops; |
34 | static const struct rpc_call_ops nfs_read_full_ops; | |
1da177e4 | 35 | |
e18b890b | 36 | static struct kmem_cache *nfs_rdata_cachep; |
1da177e4 | 37 | |
8d5658c9 | 38 | struct nfs_read_data *nfs_readdata_alloc(unsigned int pagecount) |
3feb2d49 | 39 | { |
b6ee8cd2 | 40 | struct nfs_read_data *p; |
3feb2d49 | 41 | |
b6ee8cd2 | 42 | p = kmem_cache_zalloc(nfs_rdata_cachep, GFP_KERNEL); |
3feb2d49 | 43 | if (p) { |
3feb2d49 | 44 | INIT_LIST_HEAD(&p->pages); |
e9f7bee1 | 45 | p->npages = pagecount; |
0d0b5cb3 CL |
46 | if (pagecount <= ARRAY_SIZE(p->page_array)) |
47 | p->pagevec = p->page_array; | |
3feb2d49 | 48 | else { |
93870d76 | 49 | p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_KERNEL); |
0d0b5cb3 | 50 | if (!p->pagevec) { |
b6ee8cd2 | 51 | kmem_cache_free(nfs_rdata_cachep, p); |
3feb2d49 TM |
52 | p = NULL; |
53 | } | |
54 | } | |
55 | } | |
56 | return p; | |
57 | } | |
58 | ||
1ae88b2e | 59 | void nfs_readdata_free(struct nfs_read_data *p) |
3feb2d49 TM |
60 | { |
61 | if (p && (p->pagevec != &p->page_array[0])) | |
62 | kfree(p->pagevec); | |
b6ee8cd2 | 63 | kmem_cache_free(nfs_rdata_cachep, p); |
3feb2d49 TM |
64 | } |
65 | ||
493292dd | 66 | void nfs_readdata_release(struct nfs_read_data *rdata) |
1da177e4 | 67 | { |
383ba719 TM |
68 | put_nfs_open_context(rdata->args.context); |
69 | nfs_readdata_free(rdata); | |
1da177e4 LT |
70 | } |
71 | ||
1da177e4 LT |
72 | static |
73 | int nfs_return_empty_page(struct page *page) | |
74 | { | |
eebd2aa3 | 75 | zero_user(page, 0, PAGE_CACHE_SIZE); |
1da177e4 LT |
76 | SetPageUptodate(page); |
77 | unlock_page(page); | |
78 | return 0; | |
79 | } | |
80 | ||
1de3fc12 TM |
81 | static void nfs_readpage_truncate_uninitialised_page(struct nfs_read_data *data) |
82 | { | |
83 | unsigned int remainder = data->args.count - data->res.count; | |
84 | unsigned int base = data->args.pgbase + data->res.count; | |
85 | unsigned int pglen; | |
86 | struct page **pages; | |
87 | ||
88 | if (data->res.eof == 0 || remainder == 0) | |
89 | return; | |
90 | /* | |
91 | * Note: "remainder" can never be negative, since we check for | |
92 | * this in the XDR code. | |
93 | */ | |
94 | pages = &data->args.pages[base >> PAGE_CACHE_SHIFT]; | |
95 | base &= ~PAGE_CACHE_MASK; | |
96 | pglen = PAGE_CACHE_SIZE - base; | |
79558f36 TM |
97 | for (;;) { |
98 | if (remainder <= pglen) { | |
eebd2aa3 | 99 | zero_user(*pages, base, remainder); |
79558f36 TM |
100 | break; |
101 | } | |
eebd2aa3 | 102 | zero_user(*pages, base, pglen); |
79558f36 TM |
103 | pages++; |
104 | remainder -= pglen; | |
105 | pglen = PAGE_CACHE_SIZE; | |
106 | base = 0; | |
107 | } | |
1de3fc12 TM |
108 | } |
109 | ||
62e4a769 | 110 | void nfs_pageio_init_read_mds(struct nfs_pageio_descriptor *pgio, |
1751c363 TM |
111 | struct inode *inode) |
112 | { | |
113 | nfs_pageio_init(pgio, inode, &nfs_pageio_read_ops, | |
114 | NFS_SERVER(inode)->rsize, 0); | |
115 | } | |
116 | ||
493292dd TM |
117 | void nfs_pageio_reset_read_mds(struct nfs_pageio_descriptor *pgio) |
118 | { | |
119 | pgio->pg_ops = &nfs_pageio_read_ops; | |
120 | pgio->pg_bsize = NFS_SERVER(pgio->pg_inode)->rsize; | |
121 | } | |
1f945357 | 122 | EXPORT_SYMBOL_GPL(nfs_pageio_reset_read_mds); |
493292dd | 123 | |
1751c363 TM |
124 | static void nfs_pageio_init_read(struct nfs_pageio_descriptor *pgio, |
125 | struct inode *inode) | |
126 | { | |
127 | if (!pnfs_pageio_init_read(pgio, inode)) | |
128 | nfs_pageio_init_read_mds(pgio, inode); | |
129 | } | |
130 | ||
f42b293d DH |
131 | int nfs_readpage_async(struct nfs_open_context *ctx, struct inode *inode, |
132 | struct page *page) | |
1da177e4 | 133 | { |
1da177e4 LT |
134 | struct nfs_page *new; |
135 | unsigned int len; | |
c76069bd | 136 | struct nfs_pageio_descriptor pgio; |
1da177e4 | 137 | |
49a70f27 | 138 | len = nfs_page_length(page); |
1da177e4 LT |
139 | if (len == 0) |
140 | return nfs_return_empty_page(page); | |
141 | new = nfs_create_request(ctx, inode, page, 0, len); | |
142 | if (IS_ERR(new)) { | |
143 | unlock_page(page); | |
144 | return PTR_ERR(new); | |
145 | } | |
146 | if (len < PAGE_CACHE_SIZE) | |
eebd2aa3 | 147 | zero_user_segment(page, len, PAGE_CACHE_SIZE); |
1da177e4 | 148 | |
1751c363 | 149 | nfs_pageio_init_read(&pgio, inode); |
d8007d4d | 150 | nfs_pageio_add_request(&pgio, new); |
1751c363 | 151 | nfs_pageio_complete(&pgio); |
1da177e4 LT |
152 | return 0; |
153 | } | |
154 | ||
155 | static void nfs_readpage_release(struct nfs_page *req) | |
156 | { | |
3d4ff43d | 157 | struct inode *d_inode = req->wb_context->dentry->d_inode; |
7f8e05f6 DH |
158 | |
159 | if (PageUptodate(req->wb_page)) | |
160 | nfs_readpage_to_fscache(d_inode, req->wb_page, 0); | |
161 | ||
1da177e4 LT |
162 | unlock_page(req->wb_page); |
163 | ||
1da177e4 | 164 | dprintk("NFS: read done (%s/%Ld %d@%Ld)\n", |
3d4ff43d AV |
165 | req->wb_context->dentry->d_inode->i_sb->s_id, |
166 | (long long)NFS_FILEID(req->wb_context->dentry->d_inode), | |
1da177e4 LT |
167 | req->wb_bytes, |
168 | (long long)req_offset(req)); | |
10d2c46f | 169 | nfs_release_request(req); |
1da177e4 LT |
170 | } |
171 | ||
dc70d7b3 | 172 | int nfs_initiate_read(struct nfs_read_data *data, struct rpc_clnt *clnt, |
64419a9b | 173 | const struct rpc_call_ops *call_ops) |
1da177e4 | 174 | { |
64419a9b | 175 | struct inode *inode = data->inode; |
84115e1c | 176 | int swap_flags = IS_SWAPFILE(inode) ? NFS_RPC_SWAPFLAGS : 0; |
07737691 | 177 | struct rpc_task *task; |
bdc7f021 TM |
178 | struct rpc_message msg = { |
179 | .rpc_argp = &data->args, | |
180 | .rpc_resp = &data->res, | |
64419a9b | 181 | .rpc_cred = data->cred, |
bdc7f021 | 182 | }; |
84115e1c | 183 | struct rpc_task_setup task_setup_data = { |
07737691 | 184 | .task = &data->task, |
64419a9b | 185 | .rpc_client = clnt, |
bdc7f021 | 186 | .rpc_message = &msg, |
84115e1c TM |
187 | .callback_ops = call_ops, |
188 | .callback_data = data, | |
101070ca | 189 | .workqueue = nfsiod_workqueue, |
84115e1c TM |
190 | .flags = RPC_TASK_ASYNC | swap_flags, |
191 | }; | |
1da177e4 | 192 | |
64419a9b AA |
193 | /* Set up the initial task struct. */ |
194 | NFS_PROTO(inode)->read_setup(data, &msg); | |
195 | ||
196 | dprintk("NFS: %5u initiated read call (req %s/%lld, %u bytes @ " | |
197 | "offset %llu)\n", | |
198 | data->task.tk_pid, | |
199 | inode->i_sb->s_id, | |
200 | (long long)NFS_FILEID(inode), | |
201 | data->args.count, | |
202 | (unsigned long long)data->args.offset); | |
203 | ||
204 | task = rpc_run_task(&task_setup_data); | |
205 | if (IS_ERR(task)) | |
206 | return PTR_ERR(task); | |
207 | rpc_put_task(task); | |
208 | return 0; | |
209 | } | |
dc70d7b3 | 210 | EXPORT_SYMBOL_GPL(nfs_initiate_read); |
64419a9b AA |
211 | |
212 | /* | |
213 | * Set up the NFS read request struct | |
214 | */ | |
6e4efd56 TM |
215 | static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data, |
216 | unsigned int count, unsigned int offset) | |
64419a9b | 217 | { |
3d4ff43d | 218 | struct inode *inode = req->wb_context->dentry->d_inode; |
64419a9b | 219 | |
1da177e4 | 220 | data->req = req; |
84115e1c | 221 | data->inode = inode; |
64419a9b | 222 | data->cred = req->wb_context->cred; |
1da177e4 LT |
223 | |
224 | data->args.fh = NFS_FH(inode); | |
225 | data->args.offset = req_offset(req) + offset; | |
226 | data->args.pgbase = req->wb_pgbase + offset; | |
227 | data->args.pages = data->pagevec; | |
228 | data->args.count = count; | |
383ba719 | 229 | data->args.context = get_nfs_open_context(req->wb_context); |
f11ac8db | 230 | data->args.lock_context = req->wb_lock_context; |
1da177e4 LT |
231 | |
232 | data->res.fattr = &data->fattr; | |
233 | data->res.count = count; | |
234 | data->res.eof = 0; | |
0e574af1 | 235 | nfs_fattr_init(&data->fattr); |
6e4efd56 | 236 | } |
1da177e4 | 237 | |
6e4efd56 | 238 | static int nfs_do_read(struct nfs_read_data *data, |
493292dd | 239 | const struct rpc_call_ops *call_ops) |
6e4efd56 | 240 | { |
5f00bcb3 | 241 | struct inode *inode = data->args.context->dentry->d_inode; |
6e4efd56 | 242 | |
64419a9b | 243 | return nfs_initiate_read(data, NFS_CLIENT(inode), call_ops); |
1da177e4 LT |
244 | } |
245 | ||
275acaaf TM |
246 | static int |
247 | nfs_do_multiple_reads(struct list_head *head, | |
493292dd | 248 | const struct rpc_call_ops *call_ops) |
275acaaf TM |
249 | { |
250 | struct nfs_read_data *data; | |
251 | int ret = 0; | |
252 | ||
253 | while (!list_empty(head)) { | |
254 | int ret2; | |
255 | ||
256 | data = list_entry(head->next, struct nfs_read_data, list); | |
257 | list_del_init(&data->list); | |
258 | ||
493292dd | 259 | ret2 = nfs_do_read(data, call_ops); |
275acaaf TM |
260 | if (ret == 0) |
261 | ret = ret2; | |
262 | } | |
263 | return ret; | |
264 | } | |
265 | ||
1da177e4 LT |
266 | static void |
267 | nfs_async_read_error(struct list_head *head) | |
268 | { | |
269 | struct nfs_page *req; | |
270 | ||
271 | while (!list_empty(head)) { | |
272 | req = nfs_list_entry(head->next); | |
273 | nfs_list_remove_request(req); | |
1da177e4 LT |
274 | nfs_readpage_release(req); |
275 | } | |
276 | } | |
277 | ||
1da177e4 LT |
278 | /* |
279 | * Generate multiple requests to fill a single page. | |
280 | * | |
281 | * We optimize to reduce the number of read operations on the wire. If we | |
282 | * detect that we're reading a page, or an area of a page, that is past the | |
283 | * end of file, we do not generate NFS read operations but just clear the | |
284 | * parts of the page that would have come back zero from the server anyway. | |
285 | * | |
286 | * We rely on the cached value of i_size to make this determination; another | |
287 | * client can fill pages on the server past our cached end-of-file, but we | |
288 | * won't see the new data until our attribute cache is updated. This is more | |
289 | * or less conventional NFS client behavior. | |
290 | */ | |
275acaaf | 291 | static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc, struct list_head *res) |
1da177e4 | 292 | { |
c76069bd | 293 | struct nfs_page *req = nfs_list_entry(desc->pg_list.next); |
1da177e4 LT |
294 | struct page *page = req->wb_page; |
295 | struct nfs_read_data *data; | |
d097971d | 296 | size_t rsize = desc->pg_bsize, nbytes; |
e9f7bee1 | 297 | unsigned int offset; |
1da177e4 | 298 | int requests = 0; |
dbae4c73 | 299 | int ret = 0; |
1da177e4 LT |
300 | |
301 | nfs_list_remove_request(req); | |
302 | ||
275acaaf | 303 | offset = 0; |
c76069bd | 304 | nbytes = desc->pg_count; |
e9f7bee1 TM |
305 | do { |
306 | size_t len = min(nbytes,rsize); | |
307 | ||
8d5658c9 | 308 | data = nfs_readdata_alloc(1); |
1da177e4 LT |
309 | if (!data) |
310 | goto out_bad; | |
275acaaf TM |
311 | data->pagevec[0] = page; |
312 | nfs_read_rpcsetup(req, data, len, offset); | |
313 | list_add(&data->list, res); | |
1da177e4 | 314 | requests++; |
e9f7bee1 | 315 | nbytes -= len; |
275acaaf | 316 | offset += len; |
e9f7bee1 | 317 | } while(nbytes != 0); |
1da177e4 | 318 | atomic_set(&req->wb_complete, requests); |
50828d7e | 319 | desc->pg_rpc_callops = &nfs_read_partial_ops; |
dbae4c73 | 320 | return ret; |
1da177e4 | 321 | out_bad: |
275acaaf TM |
322 | while (!list_empty(res)) { |
323 | data = list_entry(res->next, struct nfs_read_data, list); | |
6e4efd56 | 324 | list_del(&data->list); |
73fb7bc7 | 325 | nfs_readdata_release(data); |
1da177e4 | 326 | } |
1da177e4 LT |
327 | nfs_readpage_release(req); |
328 | return -ENOMEM; | |
329 | } | |
330 | ||
275acaaf | 331 | static int nfs_pagein_one(struct nfs_pageio_descriptor *desc, struct list_head *res) |
1da177e4 LT |
332 | { |
333 | struct nfs_page *req; | |
334 | struct page **pages; | |
335 | struct nfs_read_data *data; | |
c76069bd | 336 | struct list_head *head = &desc->pg_list; |
3b609184 | 337 | int ret = 0; |
1da177e4 | 338 | |
c76069bd FI |
339 | data = nfs_readdata_alloc(nfs_page_array_len(desc->pg_base, |
340 | desc->pg_count)); | |
bae724ef FI |
341 | if (!data) { |
342 | nfs_async_read_error(head); | |
3b609184 | 343 | ret = -ENOMEM; |
bae724ef FI |
344 | goto out; |
345 | } | |
1da177e4 | 346 | |
1da177e4 | 347 | pages = data->pagevec; |
1da177e4 LT |
348 | while (!list_empty(head)) { |
349 | req = nfs_list_entry(head->next); | |
350 | nfs_list_remove_request(req); | |
351 | nfs_list_add_request(req, &data->pages); | |
1da177e4 | 352 | *pages++ = req->wb_page; |
1da177e4 LT |
353 | } |
354 | req = nfs_list_entry(data->pages.next); | |
355 | ||
6e4efd56 | 356 | nfs_read_rpcsetup(req, data, desc->pg_count, 0); |
275acaaf | 357 | list_add(&data->list, res); |
50828d7e | 358 | desc->pg_rpc_callops = &nfs_read_full_ops; |
bae724ef | 359 | out: |
dbae4c73 | 360 | return ret; |
1da177e4 LT |
361 | } |
362 | ||
493292dd TM |
363 | int nfs_generic_pagein(struct nfs_pageio_descriptor *desc, struct list_head *head) |
364 | { | |
365 | if (desc->pg_bsize < PAGE_CACHE_SIZE) | |
366 | return nfs_pagein_multi(desc, head); | |
367 | return nfs_pagein_one(desc, head); | |
368 | } | |
369 | ||
370 | static int nfs_generic_pg_readpages(struct nfs_pageio_descriptor *desc) | |
1751c363 | 371 | { |
275acaaf TM |
372 | LIST_HEAD(head); |
373 | int ret; | |
374 | ||
493292dd | 375 | ret = nfs_generic_pagein(desc, &head); |
50828d7e | 376 | if (ret == 0) |
493292dd | 377 | ret = nfs_do_multiple_reads(&head, desc->pg_rpc_callops); |
275acaaf | 378 | return ret; |
1751c363 | 379 | } |
1751c363 TM |
380 | |
381 | static const struct nfs_pageio_ops nfs_pageio_read_ops = { | |
382 | .pg_test = nfs_generic_pg_test, | |
383 | .pg_doio = nfs_generic_pg_readpages, | |
384 | }; | |
385 | ||
0b671301 TM |
386 | /* |
387 | * This is the callback from RPC telling us whether a reply was | |
388 | * received or some error occurred (timeout or socket shutdown). | |
389 | */ | |
390 | int nfs_readpage_result(struct rpc_task *task, struct nfs_read_data *data) | |
391 | { | |
392 | int status; | |
393 | ||
3110ff80 | 394 | dprintk("NFS: %s: %5u, (status %d)\n", __func__, task->tk_pid, |
0b671301 TM |
395 | task->tk_status); |
396 | ||
397 | status = NFS_PROTO(data->inode)->read_done(task, data); | |
398 | if (status != 0) | |
399 | return status; | |
400 | ||
401 | nfs_add_stats(data->inode, NFSIOS_SERVERREADBYTES, data->res.count); | |
402 | ||
403 | if (task->tk_status == -ESTALE) { | |
3a10c30a | 404 | set_bit(NFS_INO_STALE, &NFS_I(data->inode)->flags); |
0b671301 TM |
405 | nfs_mark_for_revalidate(data->inode); |
406 | } | |
0b671301 TM |
407 | return 0; |
408 | } | |
409 | ||
fdd1e74c | 410 | static void nfs_readpage_retry(struct rpc_task *task, struct nfs_read_data *data) |
0b671301 TM |
411 | { |
412 | struct nfs_readargs *argp = &data->args; | |
413 | struct nfs_readres *resp = &data->res; | |
414 | ||
415 | if (resp->eof || resp->count == argp->count) | |
d61e612a | 416 | return; |
0b671301 TM |
417 | |
418 | /* This is a short read! */ | |
419 | nfs_inc_stats(data->inode, NFSIOS_SHORTREAD); | |
420 | /* Has the server at least made some progress? */ | |
421 | if (resp->count == 0) | |
d61e612a | 422 | return; |
0b671301 TM |
423 | |
424 | /* Yes, so retry the read at the end of the data */ | |
cbdabc7f | 425 | data->mds_offset += resp->count; |
0b671301 TM |
426 | argp->offset += resp->count; |
427 | argp->pgbase += resp->count; | |
428 | argp->count -= resp->count; | |
d00c5d43 | 429 | rpc_restart_call_prepare(task); |
0b671301 TM |
430 | } |
431 | ||
1da177e4 LT |
432 | /* |
433 | * Handle a read reply that fills part of a page. | |
434 | */ | |
ec06c096 | 435 | static void nfs_readpage_result_partial(struct rpc_task *task, void *calldata) |
1da177e4 | 436 | { |
ec06c096 | 437 | struct nfs_read_data *data = calldata; |
1da177e4 | 438 | |
ec06c096 TM |
439 | if (nfs_readpage_result(task, data) != 0) |
440 | return; | |
fdd1e74c TM |
441 | if (task->tk_status < 0) |
442 | return; | |
0b671301 | 443 | |
fdd1e74c TM |
444 | nfs_readpage_truncate_uninitialised_page(data); |
445 | nfs_readpage_retry(task, data); | |
446 | } | |
447 | ||
448 | static void nfs_readpage_release_partial(void *calldata) | |
449 | { | |
450 | struct nfs_read_data *data = calldata; | |
451 | struct nfs_page *req = data->req; | |
452 | struct page *page = req->wb_page; | |
453 | int status = data->task.tk_status; | |
454 | ||
455 | if (status < 0) | |
fba73005 | 456 | set_bit(PG_PARTIAL_READ_FAILED, &req->wb_flags); |
fdd1e74c | 457 | |
1da177e4 | 458 | if (atomic_dec_and_test(&req->wb_complete)) { |
fba73005 | 459 | if (!test_bit(PG_PARTIAL_READ_FAILED, &req->wb_flags)) |
1da177e4 LT |
460 | SetPageUptodate(page); |
461 | nfs_readpage_release(req); | |
462 | } | |
fdd1e74c | 463 | nfs_readdata_release(calldata); |
1da177e4 LT |
464 | } |
465 | ||
f11c88af AA |
466 | void nfs_read_prepare(struct rpc_task *task, void *calldata) |
467 | { | |
468 | struct nfs_read_data *data = calldata; | |
ea7c3303 | 469 | NFS_PROTO(data->inode)->read_rpc_prepare(task, data); |
f11c88af | 470 | } |
f11c88af | 471 | |
ec06c096 | 472 | static const struct rpc_call_ops nfs_read_partial_ops = { |
f11c88af | 473 | .rpc_call_prepare = nfs_read_prepare, |
ec06c096 | 474 | .rpc_call_done = nfs_readpage_result_partial, |
fdd1e74c | 475 | .rpc_release = nfs_readpage_release_partial, |
ec06c096 TM |
476 | }; |
477 | ||
1de3fc12 TM |
478 | static void nfs_readpage_set_pages_uptodate(struct nfs_read_data *data) |
479 | { | |
480 | unsigned int count = data->res.count; | |
481 | unsigned int base = data->args.pgbase; | |
482 | struct page **pages; | |
483 | ||
79558f36 TM |
484 | if (data->res.eof) |
485 | count = data->args.count; | |
1de3fc12 TM |
486 | if (unlikely(count == 0)) |
487 | return; | |
488 | pages = &data->args.pages[base >> PAGE_CACHE_SHIFT]; | |
489 | base &= ~PAGE_CACHE_MASK; | |
490 | count += base; | |
491 | for (;count >= PAGE_CACHE_SIZE; count -= PAGE_CACHE_SIZE, pages++) | |
492 | SetPageUptodate(*pages); | |
0b671301 TM |
493 | if (count == 0) |
494 | return; | |
495 | /* Was this a short read? */ | |
496 | if (data->res.eof || data->res.count == data->args.count) | |
1de3fc12 TM |
497 | SetPageUptodate(*pages); |
498 | } | |
499 | ||
1da177e4 LT |
500 | /* |
501 | * This is the callback from RPC telling us whether a reply was | |
502 | * received or some error occurred (timeout or socket shutdown). | |
503 | */ | |
ec06c096 | 504 | static void nfs_readpage_result_full(struct rpc_task *task, void *calldata) |
1da177e4 | 505 | { |
ec06c096 | 506 | struct nfs_read_data *data = calldata; |
1da177e4 | 507 | |
0b671301 TM |
508 | if (nfs_readpage_result(task, data) != 0) |
509 | return; | |
fdd1e74c TM |
510 | if (task->tk_status < 0) |
511 | return; | |
1de3fc12 | 512 | /* |
0b671301 | 513 | * Note: nfs_readpage_retry may change the values of |
1de3fc12 | 514 | * data->args. In the multi-page case, we therefore need |
0b671301 TM |
515 | * to ensure that we call nfs_readpage_set_pages_uptodate() |
516 | * first. | |
1de3fc12 | 517 | */ |
fdd1e74c TM |
518 | nfs_readpage_truncate_uninitialised_page(data); |
519 | nfs_readpage_set_pages_uptodate(data); | |
520 | nfs_readpage_retry(task, data); | |
521 | } | |
522 | ||
523 | static void nfs_readpage_release_full(void *calldata) | |
524 | { | |
525 | struct nfs_read_data *data = calldata; | |
526 | ||
1da177e4 LT |
527 | while (!list_empty(&data->pages)) { |
528 | struct nfs_page *req = nfs_list_entry(data->pages.next); | |
1da177e4 | 529 | |
1de3fc12 | 530 | nfs_list_remove_request(req); |
62e4a769 | 531 | nfs_readpage_release(req); |
1da177e4 | 532 | } |
fdd1e74c | 533 | nfs_readdata_release(calldata); |
1da177e4 LT |
534 | } |
535 | ||
ec06c096 | 536 | static const struct rpc_call_ops nfs_read_full_ops = { |
f11c88af | 537 | .rpc_call_prepare = nfs_read_prepare, |
ec06c096 | 538 | .rpc_call_done = nfs_readpage_result_full, |
fdd1e74c | 539 | .rpc_release = nfs_readpage_release_full, |
ec06c096 TM |
540 | }; |
541 | ||
1da177e4 LT |
542 | /* |
543 | * Read a page over NFS. | |
544 | * We read the page synchronously in the following case: | |
545 | * - The error flag is set for this page. This happens only when a | |
546 | * previous async read operation failed. | |
547 | */ | |
548 | int nfs_readpage(struct file *file, struct page *page) | |
549 | { | |
550 | struct nfs_open_context *ctx; | |
551 | struct inode *inode = page->mapping->host; | |
552 | int error; | |
553 | ||
554 | dprintk("NFS: nfs_readpage (%p %ld@%lu)\n", | |
555 | page, PAGE_CACHE_SIZE, page->index); | |
91d5b470 CL |
556 | nfs_inc_stats(inode, NFSIOS_VFSREADPAGE); |
557 | nfs_add_stats(inode, NFSIOS_READPAGES, 1); | |
558 | ||
1da177e4 LT |
559 | /* |
560 | * Try to flush any pending writes to the file.. | |
561 | * | |
562 | * NOTE! Because we own the page lock, there cannot | |
563 | * be any new pending writes generated at this point | |
564 | * for this page (other pages can be written to). | |
565 | */ | |
566 | error = nfs_wb_page(inode, page); | |
567 | if (error) | |
de05a0cc TM |
568 | goto out_unlock; |
569 | if (PageUptodate(page)) | |
570 | goto out_unlock; | |
1da177e4 | 571 | |
5f004cf2 TM |
572 | error = -ESTALE; |
573 | if (NFS_STALE(inode)) | |
de05a0cc | 574 | goto out_unlock; |
5f004cf2 | 575 | |
1da177e4 | 576 | if (file == NULL) { |
cf1308ff | 577 | error = -EBADF; |
d530838b | 578 | ctx = nfs_find_open_context(inode, NULL, FMODE_READ); |
1da177e4 | 579 | if (ctx == NULL) |
de05a0cc | 580 | goto out_unlock; |
1da177e4 | 581 | } else |
cd3758e3 | 582 | ctx = get_nfs_open_context(nfs_file_open_context(file)); |
1da177e4 | 583 | |
9a9fc1c0 DH |
584 | if (!IS_SYNC(inode)) { |
585 | error = nfs_readpage_from_fscache(ctx, inode, page); | |
586 | if (error == 0) | |
587 | goto out; | |
588 | } | |
589 | ||
8e0969f0 TM |
590 | error = nfs_readpage_async(ctx, inode, page); |
591 | ||
9a9fc1c0 | 592 | out: |
1da177e4 LT |
593 | put_nfs_open_context(ctx); |
594 | return error; | |
de05a0cc | 595 | out_unlock: |
1da177e4 LT |
596 | unlock_page(page); |
597 | return error; | |
598 | } | |
599 | ||
600 | struct nfs_readdesc { | |
8b09bee3 | 601 | struct nfs_pageio_descriptor *pgio; |
1da177e4 LT |
602 | struct nfs_open_context *ctx; |
603 | }; | |
604 | ||
605 | static int | |
606 | readpage_async_filler(void *data, struct page *page) | |
607 | { | |
608 | struct nfs_readdesc *desc = (struct nfs_readdesc *)data; | |
609 | struct inode *inode = page->mapping->host; | |
610 | struct nfs_page *new; | |
611 | unsigned int len; | |
de05a0cc TM |
612 | int error; |
613 | ||
49a70f27 | 614 | len = nfs_page_length(page); |
1da177e4 LT |
615 | if (len == 0) |
616 | return nfs_return_empty_page(page); | |
de05a0cc | 617 | |
1da177e4 | 618 | new = nfs_create_request(desc->ctx, inode, page, 0, len); |
de05a0cc TM |
619 | if (IS_ERR(new)) |
620 | goto out_error; | |
621 | ||
1da177e4 | 622 | if (len < PAGE_CACHE_SIZE) |
eebd2aa3 | 623 | zero_user_segment(page, len, PAGE_CACHE_SIZE); |
f8512ad0 FI |
624 | if (!nfs_pageio_add_request(desc->pgio, new)) { |
625 | error = desc->pgio->pg_error; | |
626 | goto out_unlock; | |
627 | } | |
1da177e4 | 628 | return 0; |
de05a0cc TM |
629 | out_error: |
630 | error = PTR_ERR(new); | |
de05a0cc TM |
631 | out_unlock: |
632 | unlock_page(page); | |
633 | return error; | |
1da177e4 LT |
634 | } |
635 | ||
636 | int nfs_readpages(struct file *filp, struct address_space *mapping, | |
637 | struct list_head *pages, unsigned nr_pages) | |
638 | { | |
8b09bee3 | 639 | struct nfs_pageio_descriptor pgio; |
1da177e4 | 640 | struct nfs_readdesc desc = { |
8b09bee3 | 641 | .pgio = &pgio, |
1da177e4 LT |
642 | }; |
643 | struct inode *inode = mapping->host; | |
8b09bee3 | 644 | unsigned long npages; |
5f004cf2 | 645 | int ret = -ESTALE; |
1da177e4 LT |
646 | |
647 | dprintk("NFS: nfs_readpages (%s/%Ld %d)\n", | |
648 | inode->i_sb->s_id, | |
649 | (long long)NFS_FILEID(inode), | |
650 | nr_pages); | |
91d5b470 | 651 | nfs_inc_stats(inode, NFSIOS_VFSREADPAGES); |
1da177e4 | 652 | |
5f004cf2 TM |
653 | if (NFS_STALE(inode)) |
654 | goto out; | |
655 | ||
1da177e4 | 656 | if (filp == NULL) { |
d530838b | 657 | desc.ctx = nfs_find_open_context(inode, NULL, FMODE_READ); |
1da177e4 LT |
658 | if (desc.ctx == NULL) |
659 | return -EBADF; | |
660 | } else | |
cd3758e3 | 661 | desc.ctx = get_nfs_open_context(nfs_file_open_context(filp)); |
9a9fc1c0 DH |
662 | |
663 | /* attempt to read as many of the pages as possible from the cache | |
664 | * - this returns -ENOBUFS immediately if the cookie is negative | |
665 | */ | |
666 | ret = nfs_readpages_from_fscache(desc.ctx, inode, mapping, | |
667 | pages, &nr_pages); | |
668 | if (ret == 0) | |
669 | goto read_complete; /* all pages were read */ | |
670 | ||
1751c363 | 671 | nfs_pageio_init_read(&pgio, inode); |
8b09bee3 | 672 | |
1da177e4 | 673 | ret = read_cache_pages(mapping, pages, readpage_async_filler, &desc); |
8b09bee3 TM |
674 | |
675 | nfs_pageio_complete(&pgio); | |
676 | npages = (pgio.pg_bytes_written + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; | |
677 | nfs_add_stats(inode, NFSIOS_READPAGES, npages); | |
9a9fc1c0 | 678 | read_complete: |
1da177e4 | 679 | put_nfs_open_context(desc.ctx); |
5f004cf2 | 680 | out: |
1da177e4 LT |
681 | return ret; |
682 | } | |
683 | ||
f7b422b1 | 684 | int __init nfs_init_readpagecache(void) |
1da177e4 LT |
685 | { |
686 | nfs_rdata_cachep = kmem_cache_create("nfs_read_data", | |
687 | sizeof(struct nfs_read_data), | |
688 | 0, SLAB_HWCACHE_ALIGN, | |
20c2df83 | 689 | NULL); |
1da177e4 LT |
690 | if (nfs_rdata_cachep == NULL) |
691 | return -ENOMEM; | |
692 | ||
1da177e4 LT |
693 | return 0; |
694 | } | |
695 | ||
266bee88 | 696 | void nfs_destroy_readpagecache(void) |
1da177e4 | 697 | { |
1a1d92c1 | 698 | kmem_cache_destroy(nfs_rdata_cachep); |
1da177e4 | 699 | } |