Blame view
mm/readahead.c
13.8 KB
1da177e4c Linux-2.6.12-rc2 |
1 2 3 4 5 |
/* * mm/readahead.c - address_space-level file readahead. * * Copyright (C) 2002, Linus Torvalds * |
e1f8e8744 Remove Andrew Mor... |
6 |
* 09Apr2002 Andrew Morton |
1da177e4c Linux-2.6.12-rc2 |
7 8 9 10 11 12 13 14 15 |
* Initial version. */ #include <linux/kernel.h> #include <linux/fs.h> #include <linux/mm.h> #include <linux/module.h> #include <linux/blkdev.h> #include <linux/backing-dev.h> |
8bde37f08 [PATCH] io-accoun... |
16 |
#include <linux/task_io_accounting_ops.h> |
1da177e4c Linux-2.6.12-rc2 |
17 |
#include <linux/pagevec.h> |
f5ff8422b Fix warnings with... |
18 |
#include <linux/pagemap.h> |
1da177e4c Linux-2.6.12-rc2 |
19 |
|
1da177e4c Linux-2.6.12-rc2 |
20 21 22 23 24 25 26 27 |
/* * Initialise a struct file's readahead state. Assumes that the caller has * memset *ra to zero. */ void file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping) { ra->ra_pages = mapping->backing_dev_info->ra_pages; |
f4e6b498d readahead: combin... |
28 |
ra->prev_pos = -1; |
1da177e4c Linux-2.6.12-rc2 |
29 |
} |
d41cc702c [GFS2] Export fil... |
30 |
EXPORT_SYMBOL_GPL(file_ra_state_init); |
1da177e4c Linux-2.6.12-rc2 |
31 |
|
1da177e4c Linux-2.6.12-rc2 |
32 |
#define list_to_page(head) (list_entry((head)->prev, struct page, lru)) |
03fb3d2af FS-Cache: Release... |
33 34 |
/* * see if a page needs releasing upon read_cache_pages() failure |
266cf658e FS-Cache: Recruit... |
35 36 37 38 |
* - the caller of read_cache_pages() may have set PG_private or PG_fscache * before calling, such as the NFS fs marking pages that are cached locally * on disk, thus we need to give the fs a chance to clean up in the event of * an error |
03fb3d2af FS-Cache: Release... |
39 40 41 42 |
*/ static void read_cache_pages_invalidate_page(struct address_space *mapping, struct page *page) { |
266cf658e FS-Cache: Recruit... |
43 |
if (page_has_private(page)) { |
03fb3d2af FS-Cache: Release... |
44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 |
if (!trylock_page(page)) BUG(); page->mapping = mapping; do_invalidatepage(page, 0); page->mapping = NULL; unlock_page(page); } page_cache_release(page); } /* * release a list of pages, invalidating them first if need be */ static void read_cache_pages_invalidate_pages(struct address_space *mapping, struct list_head *pages) { struct page *victim; while (!list_empty(pages)) { victim = list_to_page(pages); list_del(&victim->lru); read_cache_pages_invalidate_page(mapping, victim); } } |
1da177e4c Linux-2.6.12-rc2 |
68 |
/** |
bd40cddae [PATCH] kernel-do... |
69 |
* read_cache_pages - populate an address space with some pages & start reads against them |
1da177e4c Linux-2.6.12-rc2 |
70 71 72 73 74 75 76 77 78 79 80 81 |
* @mapping: the address_space * @pages: The address of a list_head which contains the target pages. These * pages have their ->index populated and are otherwise uninitialised. * @filler: callback routine for filling a single page. * @data: private data for the callback routine. * * Hides the details of the LRU cache etc from the filesystems. */ int read_cache_pages(struct address_space *mapping, struct list_head *pages, int (*filler)(void *, struct page *), void *data) { struct page *page; |
1da177e4c Linux-2.6.12-rc2 |
82 |
int ret = 0; |
1da177e4c Linux-2.6.12-rc2 |
83 84 85 |
while (!list_empty(pages)) { page = list_to_page(pages); list_del(&page->lru); |
eb2be1893 mm: buffered writ... |
86 87 |
if (add_to_page_cache_lru(page, mapping, page->index, GFP_KERNEL)) { |
03fb3d2af FS-Cache: Release... |
88 |
read_cache_pages_invalidate_page(mapping, page); |
1da177e4c Linux-2.6.12-rc2 |
89 90 |
continue; } |
eb2be1893 mm: buffered writ... |
91 |
page_cache_release(page); |
1da177e4c Linux-2.6.12-rc2 |
92 |
ret = filler(data, page); |
eb2be1893 mm: buffered writ... |
93 |
if (unlikely(ret)) { |
03fb3d2af FS-Cache: Release... |
94 |
read_cache_pages_invalidate_pages(mapping, pages); |
1da177e4c Linux-2.6.12-rc2 |
95 96 |
break; } |
8bde37f08 [PATCH] io-accoun... |
97 |
task_io_account_read(PAGE_CACHE_SIZE); |
1da177e4c Linux-2.6.12-rc2 |
98 |
} |
1da177e4c Linux-2.6.12-rc2 |
99 100 101 102 103 104 105 106 107 |
return ret; } EXPORT_SYMBOL(read_cache_pages); static int read_pages(struct address_space *mapping, struct file *filp, struct list_head *pages, unsigned nr_pages) { unsigned page_idx; |
994fc28c7 [PATCH] add AOP_T... |
108 |
int ret; |
1da177e4c Linux-2.6.12-rc2 |
109 110 111 |
if (mapping->a_ops->readpages) { ret = mapping->a_ops->readpages(filp, mapping, pages, nr_pages); |
029e332ea [PATCH] Cleanup r... |
112 113 |
/* Clean up the remaining pages */ put_pages_list(pages); |
1da177e4c Linux-2.6.12-rc2 |
114 115 |
goto out; } |
1da177e4c Linux-2.6.12-rc2 |
116 117 118 |
for (page_idx = 0; page_idx < nr_pages; page_idx++) { struct page *page = list_to_page(pages); list_del(&page->lru); |
eb2be1893 mm: buffered writ... |
119 |
if (!add_to_page_cache_lru(page, mapping, |
1da177e4c Linux-2.6.12-rc2 |
120 |
page->index, GFP_KERNEL)) { |
9f1a3cfcf [PATCH] AOP_TRUNC... |
121 |
mapping->a_ops->readpage(filp, page); |
eb2be1893 mm: buffered writ... |
122 123 |
} page_cache_release(page); |
1da177e4c Linux-2.6.12-rc2 |
124 |
} |
994fc28c7 [PATCH] add AOP_T... |
125 |
ret = 0; |
1da177e4c Linux-2.6.12-rc2 |
126 127 128 129 130 |
out: return ret; } /* |
1da177e4c Linux-2.6.12-rc2 |
131 132 133 134 135 136 137 138 139 140 141 142 |
* do_page_cache_readahead actually reads a chunk of disk. It allocates all * the pages first, then submits them all for I/O. This avoids the very bad * behaviour which would occur if page allocations are causing VM writeback. * We really don't want to intermingle reads and writes like that. * * Returns the number of pages requested, or the maximum amount of I/O allowed. * * do_page_cache_readahead() returns -1 if it encountered request queue * congestion. */ static int __do_page_cache_readahead(struct address_space *mapping, struct file *filp, |
46fc3e7b4 readahead: add lo... |
143 144 |
pgoff_t offset, unsigned long nr_to_read, unsigned long lookahead_size) |
1da177e4c Linux-2.6.12-rc2 |
145 146 147 148 149 150 151 152 153 154 155 |
{ struct inode *inode = mapping->host; struct page *page; unsigned long end_index; /* The last page we want to read */ LIST_HEAD(page_pool); int page_idx; int ret = 0; loff_t isize = i_size_read(inode); if (isize == 0) goto out; |
46fc3e7b4 readahead: add lo... |
156 |
end_index = ((isize - 1) >> PAGE_CACHE_SHIFT); |
1da177e4c Linux-2.6.12-rc2 |
157 158 159 160 |
/* * Preallocate as many pages as we will need. */ |
1da177e4c Linux-2.6.12-rc2 |
161 |
for (page_idx = 0; page_idx < nr_to_read; page_idx++) { |
7361f4d8c [PATCH] readahead... |
162 |
pgoff_t page_offset = offset + page_idx; |
c743d96b6 readahead: remove... |
163 |
|
1da177e4c Linux-2.6.12-rc2 |
164 165 |
if (page_offset > end_index) break; |
001281881 mm: use lockless ... |
166 |
rcu_read_lock(); |
1da177e4c Linux-2.6.12-rc2 |
167 |
page = radix_tree_lookup(&mapping->page_tree, page_offset); |
001281881 mm: use lockless ... |
168 |
rcu_read_unlock(); |
1da177e4c Linux-2.6.12-rc2 |
169 170 |
if (page) continue; |
1da177e4c Linux-2.6.12-rc2 |
171 |
page = page_cache_alloc_cold(mapping); |
1da177e4c Linux-2.6.12-rc2 |
172 173 174 175 |
if (!page) break; page->index = page_offset; list_add(&page->lru, &page_pool); |
46fc3e7b4 readahead: add lo... |
176 177 |
if (page_idx == nr_to_read - lookahead_size) SetPageReadahead(page); |
1da177e4c Linux-2.6.12-rc2 |
178 179 |
ret++; } |
1da177e4c Linux-2.6.12-rc2 |
180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 |
/* * Now start the IO. We ignore I/O errors - if the page is not * uptodate then the caller will launch readpage again, and * will then handle the error. */ if (ret) read_pages(mapping, filp, &page_pool, ret); BUG_ON(!list_empty(&page_pool)); out: return ret; } /* * Chunk the readahead into 2 megabyte units, so that we don't pin too much * memory at once. */ int force_page_cache_readahead(struct address_space *mapping, struct file *filp, |
7361f4d8c [PATCH] readahead... |
198 |
pgoff_t offset, unsigned long nr_to_read) |
1da177e4c Linux-2.6.12-rc2 |
199 200 201 202 203 204 205 206 207 208 209 210 211 212 |
{ int ret = 0; if (unlikely(!mapping->a_ops->readpage && !mapping->a_ops->readpages)) return -EINVAL; while (nr_to_read) { int err; unsigned long this_chunk = (2 * 1024 * 1024) / PAGE_CACHE_SIZE; if (this_chunk > nr_to_read) this_chunk = nr_to_read; err = __do_page_cache_readahead(mapping, filp, |
46fc3e7b4 readahead: add lo... |
213 |
offset, this_chunk, 0); |
1da177e4c Linux-2.6.12-rc2 |
214 215 216 217 218 219 220 221 222 223 224 225 |
if (err < 0) { ret = err; break; } ret += err; offset += this_chunk; nr_to_read -= this_chunk; } return ret; } /* |
1da177e4c Linux-2.6.12-rc2 |
226 227 228 229 230 231 232 |
* This version skips the IO if the queue is read-congested, and will tell the * block layer to abandon the readahead if request allocation would block. * * force_page_cache_readahead() will ignore queue congestion and will block on * request queues. */ int do_page_cache_readahead(struct address_space *mapping, struct file *filp, |
7361f4d8c [PATCH] readahead... |
233 |
pgoff_t offset, unsigned long nr_to_read) |
1da177e4c Linux-2.6.12-rc2 |
234 235 236 |
{ if (bdi_read_congested(mapping->backing_dev_info)) return -1; |
46fc3e7b4 readahead: add lo... |
237 |
return __do_page_cache_readahead(mapping, filp, offset, nr_to_read, 0); |
1da177e4c Linux-2.6.12-rc2 |
238 239 240 |
} /* |
1da177e4c Linux-2.6.12-rc2 |
241 242 243 244 245 |
* Given a desired number of PAGE_CACHE_SIZE readahead pages, return a * sensible upper limit. */ unsigned long max_sane_readahead(unsigned long nr) { |
4f98a2fee vmscan: split LRU... |
246 |
return min(nr, (node_page_state(numa_node_id(), NR_INACTIVE_FILE) |
05a0416be [PATCH] Drop __ge... |
247 |
+ node_page_state(numa_node_id(), NR_FREE_PAGES)) / 2); |
1da177e4c Linux-2.6.12-rc2 |
248 |
} |
5ce1110b9 readahead: data s... |
249 250 251 252 |
/* * Submit IO for the read-ahead request in file_ra_state. */ |
f9acc8c7b readahead: sanify... |
253 |
static unsigned long ra_submit(struct file_ra_state *ra, |
5ce1110b9 readahead: data s... |
254 255 |
struct address_space *mapping, struct file *filp) { |
5ce1110b9 readahead: data s... |
256 |
int actual; |
5ce1110b9 readahead: data s... |
257 |
actual = __do_page_cache_readahead(mapping, filp, |
f9acc8c7b readahead: sanify... |
258 |
ra->start, ra->size, ra->async_size); |
5ce1110b9 readahead: data s... |
259 260 261 |
return actual; } |
122a21d11 readahead: on-dem... |
262 263 |
/* |
c743d96b6 readahead: remove... |
264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 |
* Set the initial window size, round to next power of 2 and square * for small size, x 4 for medium, and x 2 for large * for 128k (32 page) max ra * 1-8 page = 32k initial, > 8 page = 128k initial */ static unsigned long get_init_ra_size(unsigned long size, unsigned long max) { unsigned long newsize = roundup_pow_of_two(size); if (newsize <= max / 32) newsize = newsize * 4; else if (newsize <= max / 4) newsize = newsize * 2; else newsize = max; return newsize; } /* |
122a21d11 readahead: on-dem... |
284 285 286 |
* Get the previous window size, ramp it up, and * return it as the new window size. */ |
c743d96b6 readahead: remove... |
287 |
static unsigned long get_next_ra_size(struct file_ra_state *ra, |
122a21d11 readahead: on-dem... |
288 289 |
unsigned long max) { |
f9acc8c7b readahead: sanify... |
290 |
unsigned long cur = ra->size; |
122a21d11 readahead: on-dem... |
291 292 293 |
unsigned long newsize; if (cur < max / 16) |
c743d96b6 readahead: remove... |
294 |
newsize = 4 * cur; |
122a21d11 readahead: on-dem... |
295 |
else |
c743d96b6 readahead: remove... |
296 |
newsize = 2 * cur; |
122a21d11 readahead: on-dem... |
297 298 299 300 301 302 303 304 305 306 |
return min(newsize, max); } /* * On-demand readahead design. * * The fields in struct file_ra_state represent the most-recently-executed * readahead attempt: * |
f9acc8c7b readahead: sanify... |
307 308 309 310 |
* |<----- async_size ---------| * |------------------- size -------------------->| * |==================#===========================| * ^start ^page marked with PG_readahead |
122a21d11 readahead: on-dem... |
311 312 313 314 |
* * To overlap application thinking time and disk I/O time, we do * `readahead pipelining': Do not wait until the application consumed all * readahead pages and stalled on the missing page at readahead_index; |
f9acc8c7b readahead: sanify... |
315 316 317 |
* Instead, submit an asynchronous readahead I/O as soon as there are * only async_size pages left in the readahead window. Normally async_size * will be equal to size, for maximum pipelining. |
122a21d11 readahead: on-dem... |
318 319 320 |
* * In interleaved sequential reads, concurrent streams on the same fd can * be invalidating each other's readahead state. So we flag the new readahead |
f9acc8c7b readahead: sanify... |
321 |
* page at (start+size-async_size) with PG_readahead, and use it as readahead |
122a21d11 readahead: on-dem... |
322 323 324 |
* indicator. The flag won't be set on already cached pages, to avoid the * readahead-for-nothing fuss, saving pointless page cache lookups. * |
f4e6b498d readahead: combin... |
325 |
* prev_pos tracks the last visited byte in the _previous_ read request. |
122a21d11 readahead: on-dem... |
326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 |
* It should be maintained by the caller, and will be used for detecting * small random reads. Note that the readahead algorithm checks loosely * for sequential patterns. Hence interleaved reads might be served as * sequential ones. * * There is a special-case: if the first page which the application tries to * read happens to be the first page of the file, it is assumed that a linear * read is about to happen and the window is immediately set to the initial size * based on I/O request size and the max_readahead. * * The code ramps up the readahead size aggressively at first, but slow down as * it approaches max_readhead. */ /* * A minimal readahead algorithm for trivial sequential/random reads. */ static unsigned long ondemand_readahead(struct address_space *mapping, struct file_ra_state *ra, struct file *filp, |
cf914a7d6 readahead: split ... |
346 |
bool hit_readahead_marker, pgoff_t offset, |
122a21d11 readahead: on-dem... |
347 348 |
unsigned long req_size) { |
f4e6b498d readahead: combin... |
349 350 351 |
int max = ra->ra_pages; /* max readahead pages */ pgoff_t prev_offset; int sequential; |
122a21d11 readahead: on-dem... |
352 353 |
/* |
f9acc8c7b readahead: sanify... |
354 |
* It's the expected callback offset, assume sequential access. |
122a21d11 readahead: on-dem... |
355 356 |
* Ramp up sizes, and push forward the readahead window. */ |
f9acc8c7b readahead: sanify... |
357 358 359 360 361 362 |
if (offset && (offset == (ra->start + ra->size - ra->async_size) || offset == (ra->start + ra->size))) { ra->start += ra->size; ra->size = get_next_ra_size(ra, max); ra->async_size = ra->size; goto readit; |
122a21d11 readahead: on-dem... |
363 |
} |
f4e6b498d readahead: combin... |
364 365 |
prev_offset = ra->prev_pos >> PAGE_CACHE_SHIFT; sequential = offset - prev_offset <= 1UL || req_size > max; |
122a21d11 readahead: on-dem... |
366 367 368 369 |
/* * Standalone, small read. * Read as is, and do not pollute the readahead state. */ |
cf914a7d6 readahead: split ... |
370 |
if (!hit_readahead_marker && !sequential) { |
122a21d11 readahead: on-dem... |
371 372 373 374 375 |
return __do_page_cache_readahead(mapping, filp, offset, req_size, 0); } /* |
6b10c6c9f readahead: basic ... |
376 377 378 379 380 381 382 |
* Hit a marked page without valid readahead state. * E.g. interleaved reads. * Query the pagecache for async_size, which normally equals to * readahead size. Ramp it up and use it as the new readahead size. */ if (hit_readahead_marker) { pgoff_t start; |
30002ed2e mm: readahead sca... |
383 384 385 |
rcu_read_lock(); start = radix_tree_next_hole(&mapping->page_tree, offset,max+1); rcu_read_unlock(); |
6b10c6c9f readahead: basic ... |
386 387 388 389 390 391 392 393 394 395 396 397 |
if (!start || start - offset > max) return 0; ra->start = start; ra->size = start - offset; /* old async_size */ ra->size = get_next_ra_size(ra, max); ra->async_size = ra->size; goto readit; } /* |
122a21d11 readahead: on-dem... |
398 399 400 401 402 403 |
* It may be one of * - first read on start of file * - sequential cache miss * - oversize random read * Start readahead for it. */ |
f9acc8c7b readahead: sanify... |
404 405 406 |
ra->start = offset; ra->size = get_init_ra_size(req_size, max); ra->async_size = ra->size > req_size ? ra->size - req_size : ra->size; |
122a21d11 readahead: on-dem... |
407 |
|
f9acc8c7b readahead: sanify... |
408 |
readit: |
122a21d11 readahead: on-dem... |
409 410 411 412 |
return ra_submit(ra, mapping, filp); } /** |
cf914a7d6 readahead: split ... |
413 |
* page_cache_sync_readahead - generic file readahead |
122a21d11 readahead: on-dem... |
414 415 416 |
* @mapping: address_space which holds the pagecache and I/O vectors * @ra: file_ra_state which holds the readahead state * @filp: passed on to ->readpage() and ->readpages() |
cf914a7d6 readahead: split ... |
417 |
* @offset: start offset into @mapping, in pagecache page-sized units |
122a21d11 readahead: on-dem... |
418 |
* @req_size: hint: total size of the read which the caller is performing in |
cf914a7d6 readahead: split ... |
419 |
* pagecache pages |
122a21d11 readahead: on-dem... |
420 |
* |
cf914a7d6 readahead: split ... |
421 422 423 424 |
* page_cache_sync_readahead() should be called when a cache miss happened: * it will submit the read. The readahead logic may decide to piggyback more * pages onto the read request if access patterns suggest it will improve * performance. |
122a21d11 readahead: on-dem... |
425 |
*/ |
cf914a7d6 readahead: split ... |
426 427 428 |
void page_cache_sync_readahead(struct address_space *mapping, struct file_ra_state *ra, struct file *filp, pgoff_t offset, unsigned long req_size) |
122a21d11 readahead: on-dem... |
429 430 431 |
{ /* no read-ahead */ if (!ra->ra_pages) |
cf914a7d6 readahead: split ... |
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 |
return; /* do read-ahead */ ondemand_readahead(mapping, ra, filp, false, offset, req_size); } EXPORT_SYMBOL_GPL(page_cache_sync_readahead); /** * page_cache_async_readahead - file readahead for marked pages * @mapping: address_space which holds the pagecache and I/O vectors * @ra: file_ra_state which holds the readahead state * @filp: passed on to ->readpage() and ->readpages() * @page: the page at @offset which has the PG_readahead flag set * @offset: start offset into @mapping, in pagecache page-sized units * @req_size: hint: total size of the read which the caller is performing in * pagecache pages * * page_cache_async_ondemand() should be called when a page is used which |
f7850d932 mm/readahead: fix... |
450 |
* has the PG_readahead flag; this is a marker to suggest that the application |
cf914a7d6 readahead: split ... |
451 |
* has used up enough of the readahead window that we should start pulling in |
f7850d932 mm/readahead: fix... |
452 453 |
* more pages. */ |
cf914a7d6 readahead: split ... |
454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 |
void page_cache_async_readahead(struct address_space *mapping, struct file_ra_state *ra, struct file *filp, struct page *page, pgoff_t offset, unsigned long req_size) { /* no read-ahead */ if (!ra->ra_pages) return; /* * Same bit is used for PG_readahead and PG_reclaim. */ if (PageWriteback(page)) return; ClearPageReadahead(page); /* * Defer asynchronous read-ahead on IO congestion. */ if (bdi_read_congested(mapping->backing_dev_info)) return; |
122a21d11 readahead: on-dem... |
477 478 |
/* do read-ahead */ |
cf914a7d6 readahead: split ... |
479 |
ondemand_readahead(mapping, ra, filp, true, offset, req_size); |
122a21d11 readahead: on-dem... |
480 |
} |
cf914a7d6 readahead: split ... |
481 |
EXPORT_SYMBOL_GPL(page_cache_async_readahead); |