]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/writeback.h
arm/arm64: smccc: Implement SMCCC v1.1 inline primitive
[mirror_ubuntu-artful-kernel.git] / include / linux / writeback.h
CommitLineData
1da177e4 1/*
f30c2269 2 * include/linux/writeback.h
1da177e4
LT
3 */
4#ifndef WRITEBACK_H
5#define WRITEBACK_H
6
e8edc6e0 7#include <linux/sched.h>
a27bb332 8#include <linux/workqueue.h>
f5ff8422 9#include <linux/fs.h>
380c27ca 10#include <linux/flex_proportions.h>
21c6321f 11#include <linux/backing-dev-defs.h>
7637241e 12#include <linux/blk_types.h>
e8edc6e0 13
2f8b5444
CH
14struct bio;
15
54848d73
WF
16DECLARE_PER_CPU(int, dirty_throttle_leaks);
17
ffd1f609 18/*
1a12d8bd
WF
19 * The 1/4 region under the global dirty thresh is for smooth dirty throttling:
20 *
21 * (thresh - thresh/DIRTY_FULL_SCOPE, thresh)
22 *
ffd1f609
WF
23 * Further beyond, all dirtier tasks will enter a loop waiting (possibly long
24 * time) for the dirty pages to drop, unless written enough pages.
25 *
26 * The global dirty threshold is normally equal to the global dirty limit,
27 * except when the system suddenly allocates a lot of anonymous memory and
28 * knocks down the global dirty threshold quickly, in which case the global
29 * dirty limit will follow down slowly to prevent livelocking all dirtier tasks.
30 */
1a12d8bd
WF
31#define DIRTY_SCOPE 8
32#define DIRTY_FULL_SCOPE (DIRTY_SCOPE / 2)
ffd1f609 33
1da177e4
LT
34struct backing_dev_info;
35
1da177e4
LT
36/*
37 * fs/fs-writeback.c
38 */
39enum writeback_sync_modes {
40 WB_SYNC_NONE, /* Don't wait on anything */
41 WB_SYNC_ALL, /* Wait on every mapping */
1da177e4
LT
42};
43
0e175a18
CW
44/*
45 * why some writeback work was initiated
46 */
47enum wb_reason {
48 WB_REASON_BACKGROUND,
726d061f 49 WB_REASON_VMSCAN,
0e175a18
CW
50 WB_REASON_SYNC,
51 WB_REASON_PERIODIC,
52 WB_REASON_LAPTOP_TIMER,
53 WB_REASON_FREE_MORE_MEM,
54 WB_REASON_FS_FREE_SPACE,
fc6df808
WL
55 /*
56 * There is no bdi forker thread any more and works are done
57 * by emergency worker, however, this is TPs userland visible
58 * and we'll be exposing exactly the same information,
59 * so it has a mismatch name.
60 */
0e175a18
CW
61 WB_REASON_FORKER_THREAD,
62
63 WB_REASON_MAX,
64};
0e175a18 65
1da177e4
LT
66/*
67 * A control structure which tells the writeback code what to do. These are
68 * always on the stack, and hence need no locking. They are always initialised
69 * in a manner such that unspecified fields are set to zero.
70 */
71struct writeback_control {
1da177e4
LT
72 long nr_to_write; /* Write this many pages, and decrement
73 this for each page written */
74 long pages_skipped; /* Pages which were not written */
75
76 /*
95468fd4 77 * For a_ops->writepages(): if start or end are non-zero then this is
1da177e4
LT
78 * a hint that the filesystem need only write out the pages inside that
79 * byterange. The byte at `end' is included in the writeout request.
80 */
111ebb6e
OH
81 loff_t range_start;
82 loff_t range_end;
1da177e4 83
4cd9069a
RK
84 enum writeback_sync_modes sync_mode;
85
22905f77 86 unsigned for_kupdate:1; /* A kupdate writeback */
b17621fe 87 unsigned for_background:1; /* A background writeback */
6e6938b6 88 unsigned tagged_writepages:1; /* tag-and-write to avoid livelock */
22905f77 89 unsigned for_reclaim:1; /* Invoked from the page allocator */
111ebb6e 90 unsigned range_cyclic:1; /* range_start is cyclic */
7747bd4b 91 unsigned for_sync:1; /* sync(2) WB_SYNC_ALL writeback */
b16b1deb
TH
92#ifdef CONFIG_CGROUP_WRITEBACK
93 struct bdi_writeback *wb; /* wb this writeback is issued under */
2a814908
TH
94 struct inode *inode; /* inode being written out */
95
96 /* foreign inode detection, see wbc_detach_inode() */
97 int wb_id; /* current wb id */
98 int wb_lcand_id; /* last foreign candidate wb id */
99 int wb_tcand_id; /* this foreign candidate wb id */
100 size_t wb_bytes; /* bytes written by current wb */
101 size_t wb_lcand_bytes; /* bytes written by last candidate */
102 size_t wb_tcand_bytes; /* bytes written by this candidate */
b16b1deb 103#endif
1da177e4
LT
104};
105
7637241e
JA
106static inline int wbc_to_write_flags(struct writeback_control *wbc)
107{
108 if (wbc->sync_mode == WB_SYNC_ALL)
109 return REQ_SYNC;
13edd5e7
JA
110 else if (wbc->for_kupdate || wbc->for_background)
111 return REQ_BACKGROUND;
7637241e
JA
112
113 return 0;
114}
115
380c27ca
TH
116/*
117 * A wb_domain represents a domain that wb's (bdi_writeback's) belong to
118 * and are measured against each other in. There always is one global
119 * domain, global_wb_domain, that every wb in the system is a member of.
120 * This allows measuring the relative bandwidth of each wb to distribute
121 * dirtyable memory accordingly.
122 */
123struct wb_domain {
dcc25ae7
TH
124 spinlock_t lock;
125
380c27ca
TH
126 /*
127 * Scale the writeback cache size proportional to the relative
128 * writeout speed.
129 *
130 * We do this by keeping a floating proportion between BDIs, based
131 * on page writeback completions [end_page_writeback()]. Those
132 * devices that write out pages fastest will get the larger share,
133 * while the slower will get a smaller share.
134 *
135 * We use page writeout completions because we are interested in
136 * getting rid of dirty pages. Having them written out is the
137 * primary goal.
138 *
139 * We introduce a concept of time, a period over which we measure
140 * these events, because demand can/will vary over time. The length
141 * of this period itself is measured in page writeback completions.
142 */
143 struct fprop_global completions;
144 struct timer_list period_timer; /* timer for aging of completions */
145 unsigned long period_time;
dcc25ae7
TH
146
147 /*
148 * The dirtyable memory and dirty threshold could be suddenly
149 * knocked down by a large amount (eg. on the startup of KVM in a
150 * swapless system). This may throw the system into deep dirty
151 * exceeded state and throttle heavy/light dirtiers alike. To
152 * retain good responsiveness, maintain global_dirty_limit for
153 * tracking slowly down to the knocked down dirty threshold.
154 *
155 * Both fields are protected by ->lock.
156 */
157 unsigned long dirty_limit_tstamp;
158 unsigned long dirty_limit;
380c27ca
TH
159};
160
2529bb3a
TH
161/**
162 * wb_domain_size_changed - memory available to a wb_domain has changed
163 * @dom: wb_domain of interest
164 *
165 * This function should be called when the amount of memory available to
166 * @dom has changed. It resets @dom's dirty limit parameters to prevent
167 * the past values which don't match the current configuration from skewing
168 * dirty throttling. Without this, when memory size of a wb_domain is
169 * greatly reduced, the dirty throttling logic may allow too many pages to
170 * be dirtied leading to consecutive unnecessary OOMs and may get stuck in
171 * that situation.
172 */
173static inline void wb_domain_size_changed(struct wb_domain *dom)
174{
175 spin_lock(&dom->lock);
176 dom->dirty_limit_tstamp = jiffies;
177 dom->dirty_limit = 0;
178 spin_unlock(&dom->lock);
179}
180
1da177e4
LT
181/*
182 * fs/fs-writeback.c
183 */
03ba3782 184struct bdi_writeback;
0e175a18
CW
185void writeback_inodes_sb(struct super_block *, enum wb_reason reason);
186void writeback_inodes_sb_nr(struct super_block *, unsigned long nr,
187 enum wb_reason reason);
f30a7d0c
TH
188bool try_to_writeback_inodes_sb(struct super_block *, enum wb_reason reason);
189bool try_to_writeback_inodes_sb_nr(struct super_block *, unsigned long nr,
190 enum wb_reason reason);
0dc83bd3 191void sync_inodes_sb(struct super_block *);
0e175a18 192void wakeup_flusher_threads(long nr_pages, enum wb_reason reason);
169ebd90 193void inode_wait_for_writeback(struct inode *inode);
1da177e4
LT
194
195/* writeback.h requires fs.h; it, too, is not included from here. */
196static inline void wait_on_inode(struct inode *inode)
197{
198 might_sleep();
74316201 199 wait_on_bit(&inode->i_state, __I_NEW, TASK_UNINTERRUPTIBLE);
1da177e4 200}
1c0eeaf5 201
21c6321f
TH
202#ifdef CONFIG_CGROUP_WRITEBACK
203
b16b1deb
TH
204#include <linux/cgroup.h>
205#include <linux/bio.h>
206
21c6321f 207void __inode_attach_wb(struct inode *inode, struct page *page);
b16b1deb
TH
208void wbc_attach_and_unlock_inode(struct writeback_control *wbc,
209 struct inode *inode)
210 __releases(&inode->i_lock);
211void wbc_detach_inode(struct writeback_control *wbc);
2a814908
TH
212void wbc_account_io(struct writeback_control *wbc, struct page *page,
213 size_t bytes);
a1a0e23e 214void cgroup_writeback_umount(void);
21c6321f
TH
215
216/**
217 * inode_attach_wb - associate an inode with its wb
218 * @inode: inode of interest
219 * @page: page being dirtied (may be NULL)
220 *
221 * If @inode doesn't have its wb, associate it with the wb matching the
222 * memcg of @page or, if @page is NULL, %current. May be called w/ or w/o
223 * @inode->i_lock.
224 */
225static inline void inode_attach_wb(struct inode *inode, struct page *page)
226{
227 if (!inode->i_wb)
228 __inode_attach_wb(inode, page);
229}
230
231/**
232 * inode_detach_wb - disassociate an inode from its wb
233 * @inode: inode of interest
234 *
235 * @inode is being freed. Detach from its wb.
236 */
237static inline void inode_detach_wb(struct inode *inode)
238{
239 if (inode->i_wb) {
f759741d 240 WARN_ON_ONCE(!(inode->i_state & I_CLEAR));
21c6321f
TH
241 wb_put(inode->i_wb);
242 inode->i_wb = NULL;
243 }
244}
245
b16b1deb
TH
246/**
247 * wbc_attach_fdatawrite_inode - associate wbc and inode for fdatawrite
248 * @wbc: writeback_control of interest
249 * @inode: target inode
250 *
251 * This function is to be used by __filemap_fdatawrite_range(), which is an
252 * alternative entry point into writeback code, and first ensures @inode is
253 * associated with a bdi_writeback and attaches it to @wbc.
254 */
255static inline void wbc_attach_fdatawrite_inode(struct writeback_control *wbc,
256 struct inode *inode)
257{
258 spin_lock(&inode->i_lock);
259 inode_attach_wb(inode, NULL);
260 wbc_attach_and_unlock_inode(wbc, inode);
261}
262
263/**
264 * wbc_init_bio - writeback specific initializtion of bio
265 * @wbc: writeback_control for the writeback in progress
266 * @bio: bio to be initialized
267 *
268 * @bio is a part of the writeback in progress controlled by @wbc. Perform
269 * writeback specific initialization. This is used to apply the cgroup
270 * writeback context.
271 */
272static inline void wbc_init_bio(struct writeback_control *wbc, struct bio *bio)
273{
274 /*
275 * pageout() path doesn't attach @wbc to the inode being written
276 * out. This is intentional as we don't want the function to block
277 * behind a slow cgroup. Ultimately, we want pageout() to kick off
278 * regular writeback instead of writing things out itself.
279 */
280 if (wbc->wb)
281 bio_associate_blkcg(bio, wbc->wb->blkcg_css);
282}
283
21c6321f
TH
284#else /* CONFIG_CGROUP_WRITEBACK */
285
286static inline void inode_attach_wb(struct inode *inode, struct page *page)
287{
288}
289
290static inline void inode_detach_wb(struct inode *inode)
291{
292}
293
b16b1deb
TH
294static inline void wbc_attach_and_unlock_inode(struct writeback_control *wbc,
295 struct inode *inode)
296 __releases(&inode->i_lock)
297{
298 spin_unlock(&inode->i_lock);
299}
300
301static inline void wbc_attach_fdatawrite_inode(struct writeback_control *wbc,
302 struct inode *inode)
303{
304}
305
306static inline void wbc_detach_inode(struct writeback_control *wbc)
307{
308}
309
310static inline void wbc_init_bio(struct writeback_control *wbc, struct bio *bio)
311{
312}
313
2a814908
TH
314static inline void wbc_account_io(struct writeback_control *wbc,
315 struct page *page, size_t bytes)
316{
317}
318
a1a0e23e
TH
319static inline void cgroup_writeback_umount(void)
320{
321}
322
21c6321f
TH
323#endif /* CONFIG_CGROUP_WRITEBACK */
324
1da177e4
LT
325/*
326 * mm/page-writeback.c
327 */
c2c4986e 328#ifdef CONFIG_BLOCK
31373d09 329void laptop_io_completion(struct backing_dev_info *info);
1da177e4 330void laptop_sync_completion(void);
31373d09
MG
331void laptop_mode_sync(struct work_struct *work);
332void laptop_mode_timer_fn(unsigned long data);
c2c4986e
JA
333#else
334static inline void laptop_sync_completion(void) { }
335#endif
281e3726 336bool node_dirty_ok(struct pglist_data *pgdat);
380c27ca 337int wb_domain_init(struct wb_domain *dom, gfp_t gfp);
841710aa
TH
338#ifdef CONFIG_CGROUP_WRITEBACK
339void wb_domain_exit(struct wb_domain *dom);
340#endif
1da177e4 341
dcc25ae7 342extern struct wb_domain global_wb_domain;
c42843f2 343
1da177e4
LT
344/* These are exported to sysctl. */
345extern int dirty_background_ratio;
2da02997 346extern unsigned long dirty_background_bytes;
1da177e4 347extern int vm_dirty_ratio;
2da02997 348extern unsigned long vm_dirty_bytes;
704503d8
AD
349extern unsigned int dirty_writeback_interval;
350extern unsigned int dirty_expire_interval;
1efff914 351extern unsigned int dirtytime_expire_interval;
195cf453 352extern int vm_highmem_is_dirtyable;
1da177e4
LT
353extern int block_dump;
354extern int laptop_mode;
355
2da02997 356extern int dirty_background_ratio_handler(struct ctl_table *table, int write,
8d65af78 357 void __user *buffer, size_t *lenp,
2da02997
DR
358 loff_t *ppos);
359extern int dirty_background_bytes_handler(struct ctl_table *table, int write,
8d65af78 360 void __user *buffer, size_t *lenp,
2da02997 361 loff_t *ppos);
04fbfdc1 362extern int dirty_ratio_handler(struct ctl_table *table, int write,
8d65af78 363 void __user *buffer, size_t *lenp,
04fbfdc1 364 loff_t *ppos);
2da02997 365extern int dirty_bytes_handler(struct ctl_table *table, int write,
8d65af78 366 void __user *buffer, size_t *lenp,
2da02997 367 loff_t *ppos);
1efff914
TT
368int dirtytime_interval_handler(struct ctl_table *table, int write,
369 void __user *buffer, size_t *lenp, loff_t *ppos);
04fbfdc1 370
1da177e4 371struct ctl_table;
8d65af78 372int dirty_writeback_centisecs_handler(struct ctl_table *, int,
1da177e4
LT
373 void __user *, size_t *, loff_t *);
374
16c4042f 375void global_dirty_limits(unsigned long *pbackground, unsigned long *pdirty);
0d960a38 376unsigned long wb_calc_thresh(struct bdi_writeback *wb, unsigned long thresh);
a88a341a 377
8a731799 378void wb_update_bandwidth(struct bdi_writeback *wb, unsigned long start_time);
d0e1d66b 379void balance_dirty_pages_ratelimited(struct address_space *mapping);
aa661bbe 380bool wb_over_bg_thresh(struct bdi_writeback *wb);
fa5a734e 381
0ea97180
MS
382typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc,
383 void *data);
384
0ea97180
MS
385int generic_writepages(struct address_space *mapping,
386 struct writeback_control *wbc);
5b41d924
ES
387void tag_pages_for_writeback(struct address_space *mapping,
388 pgoff_t start, pgoff_t end);
0ea97180
MS
389int write_cache_pages(struct address_space *mapping,
390 struct writeback_control *wbc, writepage_t writepage,
391 void *data);
1da177e4 392int do_writepages(struct address_space *mapping, struct writeback_control *wbc);
2d1d43f6 393void writeback_set_ratelimit(void);
92c09c04
NK
394void tag_pages_for_writeback(struct address_space *mapping,
395 pgoff_t start, pgoff_t end);
1da177e4 396
2f800fbd
WF
397void account_page_redirty(struct page *page);
398
6c60d2b5
DC
399void sb_mark_inode_writeback(struct inode *inode);
400void sb_clear_inode_writeback(struct inode *inode);
401
1da177e4 402#endif /* WRITEBACK_H */