blob: 0706a7282a8ca6ce26b39fd7d7d7c930d6891c94 [file] [log] [blame]
Lee Schermerhornfa07e782008-10-18 20:26:47 -07001
2This document describes the Linux memory management "Unevictable LRU"
3infrastructure and the use of this infrastructure to manage several types
4of "unevictable" pages. The document attempts to provide the overall
5rationale behind this mechanism and the rationale for some of the design
6decisions that drove the implementation. The latter design rationale is
7discussed in the context of an implementation description. Admittedly, one
8can obtain the implementation details--the "what does it do?"--by reading the
9code. One hopes that the descriptions below add value by provide the answer
10to "why does it do that?".
11
12Unevictable LRU Infrastructure:
13
14The Unevictable LRU adds an additional LRU list to track unevictable pages
15and to hide these pages from vmscan. This mechanism is based on a patch by
16Larry Woodman of Red Hat to address several scalability problems with page
17reclaim in Linux. The problems have been observed at customer sites on large
18memory x86_64 systems. For example, a non-numal x86_64 platform with 128GB
19of main memory will have over 32 million 4k pages in a single zone. When a
20large fraction of these pages are not evictable for any reason [see below],
21vmscan will spend a lot of time scanning the LRU lists looking for the small
22fraction of pages that are evictable. This can result in a situation where
23all cpus are spending 100% of their time in vmscan for hours or days on end,
24with the system completely unresponsive.
25
26The Unevictable LRU infrastructure addresses the following classes of
27unevictable pages:
28
29+ page owned by ramfs
30+ page mapped into SHM_LOCKed shared memory regions
31+ page mapped into VM_LOCKED [mlock()ed] vmas
32
33The infrastructure might be able to handle other conditions that make pages
34unevictable, either by definition or by circumstance, in the future.
35
36
37The Unevictable LRU List
38
39The Unevictable LRU infrastructure consists of an additional, per-zone, LRU list
40called the "unevictable" list and an associated page flag, PG_unevictable, to
41indicate that the page is being managed on the unevictable list. The
42PG_unevictable flag is analogous to, and mutually exclusive with, the PG_active
43flag in that it indicates on which LRU list a page resides when PG_lru is set.
44The unevictable LRU list is source configurable based on the UNEVICTABLE_LRU
45Kconfig option.
46
47The Unevictable LRU infrastructure maintains unevictable pages on an additional
48LRU list for a few reasons:
49
501) We get to "treat unevictable pages just like we treat other pages in the
51 system, which means we get to use the same code to manipulate them, the
52 same code to isolate them (for migrate, etc.), the same code to keep track
53 of the statistics, etc..." [Rik van Riel]
54
552) We want to be able to migrate unevictable pages between nodes--for memory
56 defragmentation, workload management and memory hotplug. The linux kernel
57 can only migrate pages that it can successfully isolate from the lru lists.
58 If we were to maintain pages elsewise than on an lru-like list, where they
59 can be found by isolate_lru_page(), we would prevent their migration, unless
60 we reworked migration code to find the unevictable pages.
61
62
63The unevictable LRU list does not differentiate between file backed and swap
64backed [anon] pages. This differentiation is only important while the pages
65are, in fact, evictable.
66
67The unevictable LRU list benefits from the "arrayification" of the per-zone
68LRU lists and statistics originally proposed and posted by Christoph Lameter.
69
70The unevictable list does not use the lru pagevec mechanism. Rather,
71unevictable pages are placed directly on the page's zone's unevictable
72list under the zone lru_lock. The reason for this is to prevent stranding
73of pages on the unevictable list when one task has the page isolated from the
74lru and other tasks are changing the "evictability" state of the page.
75
76
77Unevictable LRU and Memory Controller Interaction
78
79The memory controller data structure automatically gets a per zone unevictable
80lru list as a result of the "arrayification" of the per-zone LRU lists. The
81memory controller tracks the movement of pages to and from the unevictable list.
82When a memory control group comes under memory pressure, the controller will
83not attempt to reclaim pages on the unevictable list. This has a couple of
84effects. Because the pages are "hidden" from reclaim on the unevictable list,
85the reclaim process can be more efficient, dealing only with pages that have
86a chance of being reclaimed. On the other hand, if too many of the pages
87charged to the control group are unevictable, the evictable portion of the
88working set of the tasks in the control group may not fit into the available
89memory. This can cause the control group to thrash or to oom-kill tasks.
90
91
92Unevictable LRU: Detecting Unevictable Pages
93
94The function page_evictable(page, vma) in vmscan.c determines whether a
95page is evictable or not. For ramfs pages and pages in SHM_LOCKed regions,
96page_evictable() tests a new address space flag, AS_UNEVICTABLE, in the page's
97address space using a wrapper function. Wrapper functions are used to set,
98clear and test the flag to reduce the requirement for #ifdef's throughout the
99source code. AS_UNEVICTABLE is set on ramfs inode/mapping when it is created.
100This flag remains for the life of the inode.
101
102For shared memory regions, AS_UNEVICTABLE is set when an application
103successfully SHM_LOCKs the region and is removed when the region is
104SHM_UNLOCKed. Note that shmctl(SHM_LOCK, ...) does not populate the page
105tables for the region as does, for example, mlock(). So, we make no special
106effort to push any pages in the SHM_LOCKed region to the unevictable list.
107Vmscan will do this when/if it encounters the pages during reclaim. On
108SHM_UNLOCK, shmctl() scans the pages in the region and "rescues" them from the
109unevictable list if no other condition keeps them unevictable. If a SHM_LOCKed
110region is destroyed, the pages are also "rescued" from the unevictable list in
111the process of freeing them.
112
113page_evictable() detects mlock()ed pages by testing an additional page flag,
114PG_mlocked via the PageMlocked() wrapper. If the page is NOT mlocked, and a
115non-NULL vma is supplied, page_evictable() will check whether the vma is
116VM_LOCKED via is_mlocked_vma(). is_mlocked_vma() will SetPageMlocked() and
117update the appropriate statistics if the vma is VM_LOCKED. This method allows
118efficient "culling" of pages in the fault path that are being faulted in to
119VM_LOCKED vmas.
120
121
122Unevictable Pages and Vmscan [shrink_*_list()]
123
124If unevictable pages are culled in the fault path, or moved to the unevictable
125list at mlock() or mmap() time, vmscan will never encounter the pages until
126they have become evictable again, for example, via munlock() and have been
127"rescued" from the unevictable list. However, there may be situations where we
128decide, for the sake of expediency, to leave a unevictable page on one of the
129regular active/inactive LRU lists for vmscan to deal with. Vmscan checks for
130such pages in all of the shrink_{active|inactive|page}_list() functions and
131will "cull" such pages that it encounters--that is, it diverts those pages to
132the unevictable list for the zone being scanned.
133
134There may be situations where a page is mapped into a VM_LOCKED vma, but the
135page is not marked as PageMlocked. Such pages will make it all the way to
136shrink_page_list() where they will be detected when vmscan walks the reverse
137map in try_to_unmap(). If try_to_unmap() returns SWAP_MLOCK, shrink_page_list()
138will cull the page at that point.
139
Lee Schermerhornfa07e782008-10-18 20:26:47 -0700140To "cull" an unevictable page, vmscan simply puts the page back on the lru
141list using putback_lru_page()--the inverse operation to isolate_lru_page()--
142after dropping the page lock. Because the condition which makes the page
143unevictable may change once the page is unlocked, putback_lru_page() will
144recheck the unevictable state of a page that it places on the unevictable lru
145list. If the page has become unevictable, putback_lru_page() removes it from
146the list and retries, including the page_unevictable() test. Because such a
147race is a rare event and movement of pages onto the unevictable list should be
148rare, these extra evictabilty checks should not occur in the majority of calls
149to putback_lru_page().
150
151
152Mlocked Page: Prior Work
153
154The "Unevictable Mlocked Pages" infrastructure is based on work originally
155posted by Nick Piggin in an RFC patch entitled "mm: mlocked pages off LRU".
156Nick posted his patch as an alternative to a patch posted by Christoph
157Lameter to achieve the same objective--hiding mlocked pages from vmscan.
158In Nick's patch, he used one of the struct page lru list link fields as a count
159of VM_LOCKED vmas that map the page. This use of the link field for a count
160prevented the management of the pages on an LRU list. Thus, mlocked pages were
161not migratable as isolate_lru_page() could not find them and the lru list link
162field was not available to the migration subsystem. Nick resolved this by
163putting mlocked pages back on the lru list before attempting to isolate them,
164thus abandoning the count of VM_LOCKED vmas. When Nick's patch was integrated
165with the Unevictable LRU work, the count was replaced by walking the reverse
166map to determine whether any VM_LOCKED vmas mapped the page. More on this
167below.
168
169
170Mlocked Pages: Basic Management
171
172Mlocked pages--pages mapped into a VM_LOCKED vma--represent one class of
173unevictable pages. When such a page has been "noticed" by the memory
174management subsystem, the page is marked with the PG_mlocked [PageMlocked()]
175flag. A PageMlocked() page will be placed on the unevictable LRU list when
176it is added to the LRU. Pages can be "noticed" by memory management in
177several places:
178
1791) in the mlock()/mlockall() system call handlers.
1802) in the mmap() system call handler when mmap()ing a region with the
181 MAP_LOCKED flag, or mmap()ing a region in a task that has called
182 mlockall() with the MCL_FUTURE flag. Both of these conditions result
183 in the VM_LOCKED flag being set for the vma.
1843) in the fault path, if mlocked pages are "culled" in the fault path,
185 and when a VM_LOCKED stack segment is expanded.
Hugh Dickins63d6c5a2009-01-06 14:39:38 -08001864) as mentioned above, in vmscan:shrink_page_list() when attempting to
187 reclaim a page in a VM_LOCKED vma via try_to_unmap().
Lee Schermerhornfa07e782008-10-18 20:26:47 -0700188
189Mlocked pages become unlocked and rescued from the unevictable list when:
190
1911) mapped in a range unlocked via the munlock()/munlockall() system calls.
1922) munmapped() out of the last VM_LOCKED vma that maps the page, including
193 unmapping at task exit.
1943) when the page is truncated from the last VM_LOCKED vma of an mmap()ed file.
1954) before a page is COWed in a VM_LOCKED vma.
196
197
198Mlocked Pages: mlock()/mlockall() System Call Handling
199
200Both [do_]mlock() and [do_]mlockall() system call handlers call mlock_fixup()
201for each vma in the range specified by the call. In the case of mlockall(),
202this is the entire active address space of the task. Note that mlock_fixup()
203is used for both mlock()ing and munlock()ing a range of memory. A call to
204mlock() an already VM_LOCKED vma, or to munlock() a vma that is not VM_LOCKED
205is treated as a no-op--mlock_fixup() simply returns.
206
207If the vma passes some filtering described in "Mlocked Pages: Filtering Vmas"
208below, mlock_fixup() will attempt to merge the vma with its neighbors or split
209off a subset of the vma if the range does not cover the entire vma. Once the
210vma has been merged or split or neither, mlock_fixup() will call
211__mlock_vma_pages_range() to fault in the pages via get_user_pages() and
212to mark the pages as mlocked via mlock_vma_page().
213
214Note that the vma being mlocked might be mapped with PROT_NONE. In this case,
215get_user_pages() will be unable to fault in the pages. That's OK. If pages
216do end up getting faulted into this VM_LOCKED vma, we'll handle them in the
217fault path or in vmscan.
218
219Also note that a page returned by get_user_pages() could be truncated or
220migrated out from under us, while we're trying to mlock it. To detect
221this, __mlock_vma_pages_range() tests the page_mapping after acquiring
222the page lock. If the page is still associated with its mapping, we'll
223go ahead and call mlock_vma_page(). If the mapping is gone, we just
224unlock the page and move on. Worse case, this results in page mapped
225in a VM_LOCKED vma remaining on a normal LRU list without being
226PageMlocked(). Again, vmscan will detect and cull such pages.
227
228mlock_vma_page(), called with the page locked [N.B., not "mlocked"], will
229TestSetPageMlocked() for each page returned by get_user_pages(). We use
230TestSetPageMlocked() because the page might already be mlocked by another
231task/vma and we don't want to do extra work. We especially do not want to
232count an mlocked page more than once in the statistics. If the page was
233already mlocked, mlock_vma_page() is done.
234
235If the page was NOT already mlocked, mlock_vma_page() attempts to isolate the
236page from the LRU, as it is likely on the appropriate active or inactive list
237at that time. If the isolate_lru_page() succeeds, mlock_vma_page() will
238putback the page--putback_lru_page()--which will notice that the page is now
239mlocked and divert the page to the zone's unevictable LRU list. If
240mlock_vma_page() is unable to isolate the page from the LRU, vmscan will handle
241it later if/when it attempts to reclaim the page.
242
243
244Mlocked Pages: Filtering Special Vmas
245
246mlock_fixup() filters several classes of "special" vmas:
247
2481) vmas with VM_IO|VM_PFNMAP set are skipped entirely. The pages behind
249 these mappings are inherently pinned, so we don't need to mark them as
250 mlocked. In any case, most of the pages have no struct page in which to
251 so mark the page. Because of this, get_user_pages() will fail for these
252 vmas, so there is no sense in attempting to visit them.
253
2542) vmas mapping hugetlbfs page are already effectively pinned into memory.
255 We don't need nor want to mlock() these pages. However, to preserve the
Hugh Dickins63d6c5a2009-01-06 14:39:38 -0800256 prior behavior of mlock()--before the unevictable/mlock changes--
257 mlock_fixup() will call make_pages_present() in the hugetlbfs vma range
258 to allocate the huge pages and populate the ptes.
Lee Schermerhornfa07e782008-10-18 20:26:47 -0700259
2603) vmas with VM_DONTEXPAND|VM_RESERVED are generally user space mappings of
261 kernel pages, such as the vdso page, relay channel pages, etc. These pages
262 are inherently unevictable and are not managed on the LRU lists.
263 mlock_fixup() treats these vmas the same as hugetlbfs vmas. It calls
264 make_pages_present() to populate the ptes.
265
266Note that for all of these special vmas, mlock_fixup() does not set the
267VM_LOCKED flag. Therefore, we won't have to deal with them later during
268munlock() or munmap()--for example, at task exit. Neither does mlock_fixup()
269account these vmas against the task's "locked_vm".
270
271Mlocked Pages: Downgrading the Mmap Semaphore.
272
273mlock_fixup() must be called with the mmap semaphore held for write, because
274it may have to merge or split vmas. However, mlocking a large region of
275memory can take a long time--especially if vmscan must reclaim pages to
276satisfy the regions requirements. Faulting in a large region with the mmap
277semaphore held for write can hold off other faults on the address space, in
278the case of a multi-threaded task. It can also hold off scans of the task's
279address space via /proc. While testing under heavy load, it was observed that
280the ps(1) command could be held off for many minutes while a large segment was
281mlock()ed down.
282
283To address this issue, and to make the system more responsive during mlock()ing
284of large segments, mlock_fixup() downgrades the mmap semaphore to read mode
285during the call to __mlock_vma_pages_range(). This works fine. However, the
286callers of mlock_fixup() expect the semaphore to be returned in write mode.
287So, mlock_fixup() "upgrades" the semphore to write mode. Linux does not
288support an atomic upgrade_sem() call, so mlock_fixup() must drop the semaphore
289and reacquire it in write mode. In a multi-threaded task, it is possible for
290the task memory map to change while the semaphore is dropped. Therefore,
291mlock_fixup() looks up the vma at the range start address after reacquiring
292the semaphore in write mode and verifies that it still covers the original
293range. If not, mlock_fixup() returns an error [-EAGAIN]. All callers of
294mlock_fixup() have been changed to deal with this new error condition.
295
296Note: when munlocking a region, all of the pages should already be resident--
297unless we have racing threads mlocking() and munlocking() regions. So,
298unlocking should not have to wait for page allocations nor faults of any kind.
299Therefore mlock_fixup() does not downgrade the semaphore for munlock().
300
301
302Mlocked Pages: munlock()/munlockall() System Call Handling
303
304The munlock() and munlockall() system calls are handled by the same functions--
305do_mlock[all]()--as the mlock() and mlockall() system calls with the unlock
306vs lock operation indicated by an argument. So, these system calls are also
307handled by mlock_fixup(). Again, if called for an already munlock()ed vma,
308mlock_fixup() simply returns. Because of the vma filtering discussed above,
309VM_LOCKED will not be set in any "special" vmas. So, these vmas will be
310ignored for munlock.
311
312If the vma is VM_LOCKED, mlock_fixup() again attempts to merge or split off
313the specified range. The range is then munlocked via the function
314__mlock_vma_pages_range()--the same function used to mlock a vma range--
315passing a flag to indicate that munlock() is being performed.
316
317Because the vma access protections could have been changed to PROT_NONE after
Hugh Dickins63d6c5a2009-01-06 14:39:38 -0800318faulting in and mlocking pages, get_user_pages() was unreliable for visiting
Lee Schermerhornfa07e782008-10-18 20:26:47 -0700319these pages for munlocking. Because we don't want to leave pages mlocked(),
320get_user_pages() was enhanced to accept a flag to ignore the permissions when
321fetching the pages--all of which should be resident as a result of previous
322mlock()ing.
323
324For munlock(), __mlock_vma_pages_range() unlocks individual pages by calling
325munlock_vma_page(). munlock_vma_page() unconditionally clears the PG_mlocked
326flag using TestClearPageMlocked(). As with mlock_vma_page(), munlock_vma_page()
327use the Test*PageMlocked() function to handle the case where the page might
328have already been unlocked by another task. If the page was mlocked,
329munlock_vma_page() updates that zone statistics for the number of mlocked
330pages. Note, however, that at this point we haven't checked whether the page
331is mapped by other VM_LOCKED vmas.
332
333We can't call try_to_munlock(), the function that walks the reverse map to check
334for other VM_LOCKED vmas, without first isolating the page from the LRU.
335try_to_munlock() is a variant of try_to_unmap() and thus requires that the page
336not be on an lru list. [More on these below.] However, the call to
337isolate_lru_page() could fail, in which case we couldn't try_to_munlock().
338So, we go ahead and clear PG_mlocked up front, as this might be the only chance
339we have. If we can successfully isolate the page, we go ahead and
340try_to_munlock(), which will restore the PG_mlocked flag and update the zone
341page statistics if it finds another vma holding the page mlocked. If we fail
342to isolate the page, we'll have left a potentially mlocked page on the LRU.
343This is fine, because we'll catch it later when/if vmscan tries to reclaim the
344page. This should be relatively rare.
345
346Mlocked Pages: Migrating Them...
347
348A page that is being migrated has been isolated from the lru lists and is
349held locked across unmapping of the page, updating the page's mapping
350[address_space] entry and copying the contents and state, until the
351page table entry has been replaced with an entry that refers to the new
352page. Linux supports migration of mlocked pages and other unevictable
353pages. This involves simply moving the PageMlocked and PageUnevictable states
354from the old page to the new page.
355
356Note that page migration can race with mlocking or munlocking of the same
357page. This has been discussed from the mlock/munlock perspective in the
358respective sections above. Both processes [migration, m[un]locking], hold
359the page locked. This provides the first level of synchronization. Page
360migration zeros out the page_mapping of the old page before unlocking it,
361so m[un]lock can skip these pages by testing the page mapping under page
362lock.
363
364When completing page migration, we place the new and old pages back onto the
365lru after dropping the page lock. The "unneeded" page--old page on success,
366new page on failure--will be freed when the reference count held by the
367migration process is released. To ensure that we don't strand pages on the
368unevictable list because of a race between munlock and migration, page
369migration uses the putback_lru_page() function to add migrated pages back to
370the lru.
371
372
373Mlocked Pages: mmap(MAP_LOCKED) System Call Handling
374
375In addition the the mlock()/mlockall() system calls, an application can request
376that a region of memory be mlocked using the MAP_LOCKED flag with the mmap()
377call. Furthermore, any mmap() call or brk() call that expands the heap by a
378task that has previously called mlockall() with the MCL_FUTURE flag will result
379in the newly mapped memory being mlocked. Before the unevictable/mlock changes,
380the kernel simply called make_pages_present() to allocate pages and populate
381the page table.
382
383To mlock a range of memory under the unevictable/mlock infrastructure, the
384mmap() handler and task address space expansion functions call
385mlock_vma_pages_range() specifying the vma and the address range to mlock.
386mlock_vma_pages_range() filters vmas like mlock_fixup(), as described above in
387"Mlocked Pages: Filtering Vmas". It will clear the VM_LOCKED flag, which will
388have already been set by the caller, in filtered vmas. Thus these vma's need
389not be visited for munlock when the region is unmapped.
390
391For "normal" vmas, mlock_vma_pages_range() calls __mlock_vma_pages_range() to
392fault/allocate the pages and mlock them. Again, like mlock_fixup(),
393mlock_vma_pages_range() downgrades the mmap semaphore to read mode before
394attempting to fault/allocate and mlock the pages; and "upgrades" the semaphore
395back to write mode before returning.
396
397The callers of mlock_vma_pages_range() will have already added the memory
398range to be mlocked to the task's "locked_vm". To account for filtered vmas,
399mlock_vma_pages_range() returns the number of pages NOT mlocked. All of the
400callers then subtract a non-negative return value from the task's locked_vm.
401A negative return value represent an error--for example, from get_user_pages()
402attempting to fault in a vma with PROT_NONE access. In this case, we leave
403the memory range accounted as locked_vm, as the protections could be changed
404later and pages allocated into that region.
405
406
407Mlocked Pages: munmap()/exit()/exec() System Call Handling
408
409When unmapping an mlocked region of memory, whether by an explicit call to
410munmap() or via an internal unmap from exit() or exec() processing, we must
411munlock the pages if we're removing the last VM_LOCKED vma that maps the pages.
Hugh Dickins63d6c5a2009-01-06 14:39:38 -0800412Before the unevictable/mlock changes, mlocking did not mark the pages in any
413way, so unmapping them required no processing.
Lee Schermerhornfa07e782008-10-18 20:26:47 -0700414
415To munlock a range of memory under the unevictable/mlock infrastructure, the
416munmap() hander and task address space tear down function call
417munlock_vma_pages_all(). The name reflects the observation that one always
418specifies the entire vma range when munlock()ing during unmap of a region.
419Because of the vma filtering when mlocking() regions, only "normal" vmas that
420actually contain mlocked pages will be passed to munlock_vma_pages_all().
421
422munlock_vma_pages_all() clears the VM_LOCKED vma flag and, like mlock_fixup()
423for the munlock case, calls __munlock_vma_pages_range() to walk the page table
424for the vma's memory range and munlock_vma_page() each resident page mapped by
425the vma. This effectively munlocks the page, only if this is the last
426VM_LOCKED vma that maps the page.
427
428
429Mlocked Page: try_to_unmap()
430
431[Note: the code changes represented by this section are really quite small
432compared to the text to describe what happening and why, and to discuss the
433implications.]
434
435Pages can, of course, be mapped into multiple vmas. Some of these vmas may
436have VM_LOCKED flag set. It is possible for a page mapped into one or more
437VM_LOCKED vmas not to have the PG_mlocked flag set and therefore reside on one
438of the active or inactive LRU lists. This could happen if, for example, a
439task in the process of munlock()ing the page could not isolate the page from
440the LRU. As a result, vmscan/shrink_page_list() might encounter such a page
441as described in "Unevictable Pages and Vmscan [shrink_*_list()]". To
442handle this situation, try_to_unmap() has been enhanced to check for VM_LOCKED
443vmas while it is walking a page's reverse map.
444
445try_to_unmap() is always called, by either vmscan for reclaim or for page
446migration, with the argument page locked and isolated from the LRU. BUG_ON()
447assertions enforce this requirement. Separate functions handle anonymous and
448mapped file pages, as these types of pages have different reverse map
449mechanisms.
450
451 try_to_unmap_anon()
452
453To unmap anonymous pages, each vma in the list anchored in the anon_vma must be
454visited--at least until a VM_LOCKED vma is encountered. If the page is being
455unmapped for migration, VM_LOCKED vmas do not stop the process because mlocked
456pages are migratable. However, for reclaim, if the page is mapped into a
457VM_LOCKED vma, the scan stops. try_to_unmap() attempts to acquire the mmap
458semphore of the mm_struct to which the vma belongs in read mode. If this is
459successful, try_to_unmap() will mlock the page via mlock_vma_page()--we
460wouldn't have gotten to try_to_unmap() if the page were already mlocked--and
461will return SWAP_MLOCK, indicating that the page is unevictable. If the
462mmap semaphore cannot be acquired, we are not sure whether the page is really
463unevictable or not. In this case, try_to_unmap() will return SWAP_AGAIN.
464
465 try_to_unmap_file() -- linear mappings
466
467Unmapping of a mapped file page works the same, except that the scan visits
468all vmas that maps the page's index/page offset in the page's mapping's
469reverse map priority search tree. It must also visit each vma in the page's
470mapping's non-linear list, if the list is non-empty. As for anonymous pages,
471on encountering a VM_LOCKED vma for a mapped file page, try_to_unmap() will
472attempt to acquire the associated mm_struct's mmap semaphore to mlock the page,
473returning SWAP_MLOCK if this is successful, and SWAP_AGAIN, if not.
474
475 try_to_unmap_file() -- non-linear mappings
476
477If a page's mapping contains a non-empty non-linear mapping vma list, then
478try_to_un{map|lock}() must also visit each vma in that list to determine
479whether the page is mapped in a VM_LOCKED vma. Again, the scan must visit
480all vmas in the non-linear list to ensure that the pages is not/should not be
481mlocked. If a VM_LOCKED vma is found in the list, the scan could terminate.
482However, there is no easy way to determine whether the page is actually mapped
483in a given vma--either for unmapping or testing whether the VM_LOCKED vma
484actually pins the page.
485
486So, try_to_unmap_file() handles non-linear mappings by scanning a certain
487number of pages--a "cluster"--in each non-linear vma associated with the page's
488mapping, for each file mapped page that vmscan tries to unmap. If this happens
489to unmap the page we're trying to unmap, try_to_unmap() will notice this on
490return--(page_mapcount(page) == 0)--and return SWAP_SUCCESS. Otherwise, it
491will return SWAP_AGAIN, causing vmscan to recirculate this page. We take
492advantage of the cluster scan in try_to_unmap_cluster() as follows:
493
494For each non-linear vma, try_to_unmap_cluster() attempts to acquire the mmap
495semaphore of the associated mm_struct for read without blocking. If this
496attempt is successful and the vma is VM_LOCKED, try_to_unmap_cluster() will
497retain the mmap semaphore for the scan; otherwise it drops it here. Then,
498for each page in the cluster, if we're holding the mmap semaphore for a locked
499vma, try_to_unmap_cluster() calls mlock_vma_page() to mlock the page. This
500call is a no-op if the page is already locked, but will mlock any pages in
501the non-linear mapping that happen to be unlocked. If one of the pages so
502mlocked is the page passed in to try_to_unmap(), try_to_unmap_cluster() will
503return SWAP_MLOCK, rather than the default SWAP_AGAIN. This will allow vmscan
504to cull the page, rather than recirculating it on the inactive list. Again,
505if try_to_unmap_cluster() cannot acquire the vma's mmap sem, it returns
506SWAP_AGAIN, indicating that the page is mapped by a VM_LOCKED vma, but
507couldn't be mlocked.
508
509
510Mlocked pages: try_to_munlock() Reverse Map Scan
511
512TODO/FIXME: a better name might be page_mlocked()--analogous to the
Hugh Dickins63d6c5a2009-01-06 14:39:38 -0800513page_referenced() reverse map walker.
Lee Schermerhornfa07e782008-10-18 20:26:47 -0700514
Hugh Dickins63d6c5a2009-01-06 14:39:38 -0800515When munlock_vma_page()--see "Mlocked Pages: munlock()/munlockall()
516System Call Handling" above--tries to munlock a page, it needs to
Lee Schermerhornfa07e782008-10-18 20:26:47 -0700517determine whether or not the page is mapped by any VM_LOCKED vma, without
518actually attempting to unmap all ptes from the page. For this purpose, the
519unevictable/mlock infrastructure introduced a variant of try_to_unmap() called
520try_to_munlock().
521
522try_to_munlock() calls the same functions as try_to_unmap() for anonymous and
523mapped file pages with an additional argument specifing unlock versus unmap
524processing. Again, these functions walk the respective reverse maps looking
525for VM_LOCKED vmas. When such a vma is found for anonymous pages and file
526pages mapped in linear VMAs, as in the try_to_unmap() case, the functions
527attempt to acquire the associated mmap semphore, mlock the page via
528mlock_vma_page() and return SWAP_MLOCK. This effectively undoes the
Hugh Dickins63d6c5a2009-01-06 14:39:38 -0800529pre-clearing of the page's PG_mlocked done by munlock_vma_page.
Lee Schermerhornfa07e782008-10-18 20:26:47 -0700530
531If try_to_unmap() is unable to acquire a VM_LOCKED vma's associated mmap
532semaphore, it will return SWAP_AGAIN. This will allow shrink_page_list()
533to recycle the page on the inactive list and hope that it has better luck
534with the page next time.
535
536For file pages mapped into non-linear vmas, the try_to_munlock() logic works
537slightly differently. On encountering a VM_LOCKED non-linear vma that might
538map the page, try_to_munlock() returns SWAP_AGAIN without actually mlocking
539the page. munlock_vma_page() will just leave the page unlocked and let
540vmscan deal with it--the usual fallback position.
541
542Note that try_to_munlock()'s reverse map walk must visit every vma in a pages'
543reverse map to determine that a page is NOT mapped into any VM_LOCKED vma.
544However, the scan can terminate when it encounters a VM_LOCKED vma and can
545successfully acquire the vma's mmap semphore for read and mlock the page.
546Although try_to_munlock() can be called many [very many!] times when
547munlock()ing a large region or tearing down a large address space that has been
Hugh Dickins63d6c5a2009-01-06 14:39:38 -0800548mlocked via mlockall(), overall this is a fairly rare event.
Lee Schermerhornfa07e782008-10-18 20:26:47 -0700549
550Mlocked Page: Page Reclaim in shrink_*_list()
551
552shrink_active_list() culls any obviously unevictable pages--i.e.,
553!page_evictable(page, NULL)--diverting these to the unevictable lru
554list. However, shrink_active_list() only sees unevictable pages that
555made it onto the active/inactive lru lists. Note that these pages do not
556have PageUnevictable set--otherwise, they would be on the unevictable list and
557shrink_active_list would never see them.
558
559Some examples of these unevictable pages on the LRU lists are:
560
5611) ramfs pages that have been placed on the lru lists when first allocated.
562
5632) SHM_LOCKed shared memory pages. shmctl(SHM_LOCK) does not attempt to
564 allocate or fault in the pages in the shared memory region. This happens
565 when an application accesses the page the first time after SHM_LOCKing
566 the segment.
567
5683) Mlocked pages that could not be isolated from the lru and moved to the
569 unevictable list in mlock_vma_page().
570
5713) Pages mapped into multiple VM_LOCKED vmas, but try_to_munlock() couldn't
572 acquire the vma's mmap semaphore to test the flags and set PageMlocked.
573 munlock_vma_page() was forced to let the page back on to the normal
574 LRU list for vmscan to handle.
575
Hugh Dickins63d6c5a2009-01-06 14:39:38 -0800576shrink_inactive_list() also culls any unevictable pages that it finds on
577the inactive lists, again diverting them to the appropriate zone's unevictable
Lee Schermerhornfa07e782008-10-18 20:26:47 -0700578lru list. shrink_inactive_list() should only see SHM_LOCKed pages that became
579SHM_LOCKed after shrink_active_list() had moved them to the inactive list, or
580pages mapped into VM_LOCKED vmas that munlock_vma_page() couldn't isolate from
581the lru to recheck via try_to_munlock(). shrink_inactive_list() won't notice
582the latter, but will pass on to shrink_page_list().
583
584shrink_page_list() again culls obviously unevictable pages that it could
Hugh Dickins63d6c5a2009-01-06 14:39:38 -0800585encounter for similar reason to shrink_inactive_list(). Pages mapped into
Lee Schermerhornfa07e782008-10-18 20:26:47 -0700586VM_LOCKED vmas but without PG_mlocked set will make it all the way to
Hugh Dickins63d6c5a2009-01-06 14:39:38 -0800587try_to_unmap(). shrink_page_list() will divert them to the unevictable list
588when try_to_unmap() returns SWAP_MLOCK, as discussed above.