~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/include/linux/writeback.h

Version: ~ [ linux-5.13-rc7 ] ~ [ linux-5.12.12 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.45 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.127 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.195 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.237 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.273 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.273 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /* SPDX-License-Identifier: GPL-2.0 */
  2 /*
  3  * include/linux/writeback.h
  4  */
  5 #ifndef WRITEBACK_H
  6 #define WRITEBACK_H
  7 
  8 #include <linux/sched.h>
  9 #include <linux/workqueue.h>
 10 #include <linux/fs.h>
 11 #include <linux/flex_proportions.h>
 12 #include <linux/backing-dev-defs.h>
 13 #include <linux/blk_types.h>
 14 #include <linux/blk-cgroup.h>
 15 
 16 struct bio;
 17 
 18 DECLARE_PER_CPU(int, dirty_throttle_leaks);
 19 
 20 /*
 21  * The 1/4 region under the global dirty thresh is for smooth dirty throttling:
 22  *
 23  *      (thresh - thresh/DIRTY_FULL_SCOPE, thresh)
 24  *
 25  * Further beyond, all dirtier tasks will enter a loop waiting (possibly long
 26  * time) for the dirty pages to drop, unless written enough pages.
 27  *
 28  * The global dirty threshold is normally equal to the global dirty limit,
 29  * except when the system suddenly allocates a lot of anonymous memory and
 30  * knocks down the global dirty threshold quickly, in which case the global
 31  * dirty limit will follow down slowly to prevent livelocking all dirtier tasks.
 32  */
 33 #define DIRTY_SCOPE             8
 34 #define DIRTY_FULL_SCOPE        (DIRTY_SCOPE / 2)
 35 
 36 struct backing_dev_info;
 37 
 38 /*
 39  * fs/fs-writeback.c
 40  */
 41 enum writeback_sync_modes {
 42         WB_SYNC_NONE,   /* Don't wait on anything */
 43         WB_SYNC_ALL,    /* Wait on every mapping */
 44 };
 45 
 46 /*
 47  * A control structure which tells the writeback code what to do.  These are
 48  * always on the stack, and hence need no locking.  They are always initialised
 49  * in a manner such that unspecified fields are set to zero.
 50  */
 51 struct writeback_control {
 52         long nr_to_write;               /* Write this many pages, and decrement
 53                                            this for each page written */
 54         long pages_skipped;             /* Pages which were not written */
 55 
 56         /*
 57          * For a_ops->writepages(): if start or end are non-zero then this is
 58          * a hint that the filesystem need only write out the pages inside that
 59          * byterange.  The byte at `end' is included in the writeout request.
 60          */
 61         loff_t range_start;
 62         loff_t range_end;
 63 
 64         enum writeback_sync_modes sync_mode;
 65 
 66         unsigned for_kupdate:1;         /* A kupdate writeback */
 67         unsigned for_background:1;      /* A background writeback */
 68         unsigned tagged_writepages:1;   /* tag-and-write to avoid livelock */
 69         unsigned for_reclaim:1;         /* Invoked from the page allocator */
 70         unsigned range_cyclic:1;        /* range_start is cyclic */
 71         unsigned for_sync:1;            /* sync(2) WB_SYNC_ALL writeback */
 72 
 73         /*
 74          * When writeback IOs are bounced through async layers, only the
 75          * initial synchronous phase should be accounted towards inode
 76          * cgroup ownership arbitration to avoid confusion.  Later stages
 77          * can set the following flag to disable the accounting.
 78          */
 79         unsigned no_cgroup_owner:1;
 80 
 81         unsigned punt_to_cgroup:1;      /* cgrp punting, see __REQ_CGROUP_PUNT */
 82 
 83 #ifdef CONFIG_CGROUP_WRITEBACK
 84         struct bdi_writeback *wb;       /* wb this writeback is issued under */
 85         struct inode *inode;            /* inode being written out */
 86 
 87         /* foreign inode detection, see wbc_detach_inode() */
 88         int wb_id;                      /* current wb id */
 89         int wb_lcand_id;                /* last foreign candidate wb id */
 90         int wb_tcand_id;                /* this foreign candidate wb id */
 91         size_t wb_bytes;                /* bytes written by current wb */
 92         size_t wb_lcand_bytes;          /* bytes written by last candidate */
 93         size_t wb_tcand_bytes;          /* bytes written by this candidate */
 94 #endif
 95 };
 96 
 97 static inline int wbc_to_write_flags(struct writeback_control *wbc)
 98 {
 99         int flags = 0;
100 
101         if (wbc->punt_to_cgroup)
102                 flags = REQ_CGROUP_PUNT;
103 
104         if (wbc->sync_mode == WB_SYNC_ALL)
105                 flags |= REQ_SYNC;
106         else if (wbc->for_kupdate || wbc->for_background)
107                 flags |= REQ_BACKGROUND;
108 
109         return flags;
110 }
111 
112 static inline struct cgroup_subsys_state *
113 wbc_blkcg_css(struct writeback_control *wbc)
114 {
115 #ifdef CONFIG_CGROUP_WRITEBACK
116         if (wbc->wb)
117                 return wbc->wb->blkcg_css;
118 #endif
119         return blkcg_root_css;
120 }
121 
122 /*
123  * A wb_domain represents a domain that wb's (bdi_writeback's) belong to
124  * and are measured against each other in.  There always is one global
125  * domain, global_wb_domain, that every wb in the system is a member of.
126  * This allows measuring the relative bandwidth of each wb to distribute
127  * dirtyable memory accordingly.
128  */
129 struct wb_domain {
130         spinlock_t lock;
131 
132         /*
133          * Scale the writeback cache size proportional to the relative
134          * writeout speed.
135          *
136          * We do this by keeping a floating proportion between BDIs, based
137          * on page writeback completions [end_page_writeback()]. Those
138          * devices that write out pages fastest will get the larger share,
139          * while the slower will get a smaller share.
140          *
141          * We use page writeout completions because we are interested in
142          * getting rid of dirty pages. Having them written out is the
143          * primary goal.
144          *
145          * We introduce a concept of time, a period over which we measure
146          * these events, because demand can/will vary over time. The length
147          * of this period itself is measured in page writeback completions.
148          */
149         struct fprop_global completions;
150         struct timer_list period_timer; /* timer for aging of completions */
151         unsigned long period_time;
152 
153         /*
154          * The dirtyable memory and dirty threshold could be suddenly
155          * knocked down by a large amount (eg. on the startup of KVM in a
156          * swapless system). This may throw the system into deep dirty
157          * exceeded state and throttle heavy/light dirtiers alike. To
158          * retain good responsiveness, maintain global_dirty_limit for
159          * tracking slowly down to the knocked down dirty threshold.
160          *
161          * Both fields are protected by ->lock.
162          */
163         unsigned long dirty_limit_tstamp;
164         unsigned long dirty_limit;
165 };
166 
167 /**
168  * wb_domain_size_changed - memory available to a wb_domain has changed
169  * @dom: wb_domain of interest
170  *
171  * This function should be called when the amount of memory available to
172  * @dom has changed.  It resets @dom's dirty limit parameters to prevent
173  * the past values which don't match the current configuration from skewing
174  * dirty throttling.  Without this, when memory size of a wb_domain is
175  * greatly reduced, the dirty throttling logic may allow too many pages to
176  * be dirtied leading to consecutive unnecessary OOMs and may get stuck in
177  * that situation.
178  */
179 static inline void wb_domain_size_changed(struct wb_domain *dom)
180 {
181         spin_lock(&dom->lock);
182         dom->dirty_limit_tstamp = jiffies;
183         dom->dirty_limit = 0;
184         spin_unlock(&dom->lock);
185 }
186 
187 /*
188  * fs/fs-writeback.c
189  */     
190 struct bdi_writeback;
191 void writeback_inodes_sb(struct super_block *, enum wb_reason reason);
192 void writeback_inodes_sb_nr(struct super_block *, unsigned long nr,
193                                                         enum wb_reason reason);
194 void try_to_writeback_inodes_sb(struct super_block *sb, enum wb_reason reason);
195 void sync_inodes_sb(struct super_block *);
196 void wakeup_flusher_threads(enum wb_reason reason);
197 void wakeup_flusher_threads_bdi(struct backing_dev_info *bdi,
198                                 enum wb_reason reason);
199 void inode_wait_for_writeback(struct inode *inode);
200 
201 /* writeback.h requires fs.h; it, too, is not included from here. */
202 static inline void wait_on_inode(struct inode *inode)
203 {
204         might_sleep();
205         wait_on_bit(&inode->i_state, __I_NEW, TASK_UNINTERRUPTIBLE);
206 }
207 
208 #ifdef CONFIG_CGROUP_WRITEBACK
209 
210 #include <linux/cgroup.h>
211 #include <linux/bio.h>
212 
213 void __inode_attach_wb(struct inode *inode, struct page *page);
214 void wbc_attach_and_unlock_inode(struct writeback_control *wbc,
215                                  struct inode *inode)
216         __releases(&inode->i_lock);
217 void wbc_detach_inode(struct writeback_control *wbc);
218 void wbc_account_cgroup_owner(struct writeback_control *wbc, struct page *page,
219                               size_t bytes);
220 void cgroup_writeback_umount(void);
221 
222 /**
223  * inode_attach_wb - associate an inode with its wb
224  * @inode: inode of interest
225  * @page: page being dirtied (may be NULL)
226  *
227  * If @inode doesn't have its wb, associate it with the wb matching the
228  * memcg of @page or, if @page is NULL, %current.  May be called w/ or w/o
229  * @inode->i_lock.
230  */
231 static inline void inode_attach_wb(struct inode *inode, struct page *page)
232 {
233         if (!inode->i_wb)
234                 __inode_attach_wb(inode, page);
235 }
236 
237 /**
238  * inode_detach_wb - disassociate an inode from its wb
239  * @inode: inode of interest
240  *
241  * @inode is being freed.  Detach from its wb.
242  */
243 static inline void inode_detach_wb(struct inode *inode)
244 {
245         if (inode->i_wb) {
246                 WARN_ON_ONCE(!(inode->i_state & I_CLEAR));
247                 wb_put(inode->i_wb);
248                 inode->i_wb = NULL;
249         }
250 }
251 
252 /**
253  * wbc_attach_fdatawrite_inode - associate wbc and inode for fdatawrite
254  * @wbc: writeback_control of interest
255  * @inode: target inode
256  *
257  * This function is to be used by __filemap_fdatawrite_range(), which is an
258  * alternative entry point into writeback code, and first ensures @inode is
259  * associated with a bdi_writeback and attaches it to @wbc.
260  */
261 static inline void wbc_attach_fdatawrite_inode(struct writeback_control *wbc,
262                                                struct inode *inode)
263 {
264         spin_lock(&inode->i_lock);
265         inode_attach_wb(inode, NULL);
266         wbc_attach_and_unlock_inode(wbc, inode);
267 }
268 
269 /**
270  * wbc_init_bio - writeback specific initializtion of bio
271  * @wbc: writeback_control for the writeback in progress
272  * @bio: bio to be initialized
273  *
274  * @bio is a part of the writeback in progress controlled by @wbc.  Perform
275  * writeback specific initialization.  This is used to apply the cgroup
276  * writeback context.  Must be called after the bio has been associated with
277  * a device.
278  */
279 static inline void wbc_init_bio(struct writeback_control *wbc, struct bio *bio)
280 {
281         /*
282          * pageout() path doesn't attach @wbc to the inode being written
283          * out.  This is intentional as we don't want the function to block
284          * behind a slow cgroup.  Ultimately, we want pageout() to kick off
285          * regular writeback instead of writing things out itself.
286          */
287         if (wbc->wb)
288                 bio_associate_blkg_from_css(bio, wbc->wb->blkcg_css);
289 }
290 
291 #else   /* CONFIG_CGROUP_WRITEBACK */
292 
293 static inline void inode_attach_wb(struct inode *inode, struct page *page)
294 {
295 }
296 
297 static inline void inode_detach_wb(struct inode *inode)
298 {
299 }
300 
301 static inline void wbc_attach_and_unlock_inode(struct writeback_control *wbc,
302                                                struct inode *inode)
303         __releases(&inode->i_lock)
304 {
305         spin_unlock(&inode->i_lock);
306 }
307 
308 static inline void wbc_attach_fdatawrite_inode(struct writeback_control *wbc,
309                                                struct inode *inode)
310 {
311 }
312 
313 static inline void wbc_detach_inode(struct writeback_control *wbc)
314 {
315 }
316 
317 static inline void wbc_init_bio(struct writeback_control *wbc, struct bio *bio)
318 {
319 }
320 
321 static inline void wbc_account_cgroup_owner(struct writeback_control *wbc,
322                                             struct page *page, size_t bytes)
323 {
324 }
325 
326 static inline void cgroup_writeback_umount(void)
327 {
328 }
329 
330 #endif  /* CONFIG_CGROUP_WRITEBACK */
331 
332 /*
333  * mm/page-writeback.c
334  */
335 #ifdef CONFIG_BLOCK
336 void laptop_io_completion(struct backing_dev_info *info);
337 void laptop_sync_completion(void);
338 void laptop_mode_sync(struct work_struct *work);
339 void laptop_mode_timer_fn(struct timer_list *t);
340 #else
341 static inline void laptop_sync_completion(void) { }
342 #endif
343 bool node_dirty_ok(struct pglist_data *pgdat);
344 int wb_domain_init(struct wb_domain *dom, gfp_t gfp);
345 #ifdef CONFIG_CGROUP_WRITEBACK
346 void wb_domain_exit(struct wb_domain *dom);
347 #endif
348 
349 extern struct wb_domain global_wb_domain;
350 
351 /* These are exported to sysctl. */
352 extern int dirty_background_ratio;
353 extern unsigned long dirty_background_bytes;
354 extern int vm_dirty_ratio;
355 extern unsigned long vm_dirty_bytes;
356 extern unsigned int dirty_writeback_interval;
357 extern unsigned int dirty_expire_interval;
358 extern unsigned int dirtytime_expire_interval;
359 extern int vm_highmem_is_dirtyable;
360 extern int block_dump;
361 extern int laptop_mode;
362 
363 extern int dirty_background_ratio_handler(struct ctl_table *table, int write,
364                 void __user *buffer, size_t *lenp,
365                 loff_t *ppos);
366 extern int dirty_background_bytes_handler(struct ctl_table *table, int write,
367                 void __user *buffer, size_t *lenp,
368                 loff_t *ppos);
369 extern int dirty_ratio_handler(struct ctl_table *table, int write,
370                 void __user *buffer, size_t *lenp,
371                 loff_t *ppos);
372 extern int dirty_bytes_handler(struct ctl_table *table, int write,
373                 void __user *buffer, size_t *lenp,
374                 loff_t *ppos);
375 int dirtytime_interval_handler(struct ctl_table *table, int write,
376                                void __user *buffer, size_t *lenp, loff_t *ppos);
377 
378 struct ctl_table;
379 int dirty_writeback_centisecs_handler(struct ctl_table *, int,
380                                       void __user *, size_t *, loff_t *);
381 
382 void global_dirty_limits(unsigned long *pbackground, unsigned long *pdirty);
383 unsigned long wb_calc_thresh(struct bdi_writeback *wb, unsigned long thresh);
384 
385 void wb_update_bandwidth(struct bdi_writeback *wb, unsigned long start_time);
386 void balance_dirty_pages_ratelimited(struct address_space *mapping);
387 bool wb_over_bg_thresh(struct bdi_writeback *wb);
388 
389 typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc,
390                                 void *data);
391 
392 int generic_writepages(struct address_space *mapping,
393                        struct writeback_control *wbc);
394 void tag_pages_for_writeback(struct address_space *mapping,
395                              pgoff_t start, pgoff_t end);
396 int write_cache_pages(struct address_space *mapping,
397                       struct writeback_control *wbc, writepage_t writepage,
398                       void *data);
399 int do_writepages(struct address_space *mapping, struct writeback_control *wbc);
400 void writeback_set_ratelimit(void);
401 void tag_pages_for_writeback(struct address_space *mapping,
402                              pgoff_t start, pgoff_t end);
403 
404 void account_page_redirty(struct page *page);
405 
406 void sb_mark_inode_writeback(struct inode *inode);
407 void sb_clear_inode_writeback(struct inode *inode);
408 
409 #endif          /* WRITEBACK_H */
410 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp