3  * Memory Migration functionality - linux/mm/migrate.c
7  * Page migration was first developed in the context of the memory hotplug
8  * project. The main authors of the migration code are:
88 	 * compaction threads can race against page migration functions  in isolate_movable_page()
92 	 * being (wrongly) re-isolated while it is under migration,  in isolate_movable_page()
140  * from where they were once taken off for compaction/migration.
181  * Restore a potential migration pte to a working pte entry
205 		/* PMD-mapped THP migration entry */  in remove_migration_pte()
219 		 * Recheck VMA as permissions can change since migration started  in remove_migration_pte()
269  * Get rid of all migration entries and replace them by
286  * Something used the pte of a page under migration. We need to
287  * get to the page and wait until migration is finished.
310 	 * Once page cache replacement of page migration started, page_count  in __migration_entry_wait()
672  *                    Migration functions
725 			 * async migration. Release the taken locks  in buffer_migrate_lock_buffers()
821  * Migration function for pages with buffers. This function can only be used
870 	 * migration. Writeout may mean we loose the lock and the  in writeout()
872 	 * At this point we know that the migration attempt cannot  in writeout()
887  * Default handling if a filesystem does not provide a migration function.
893 		/* Only writeback pages in full synchronous migration */  in fallback_migrate_page()
947 			 * for page migration.  in move_to_new_page()
957 		 * isolation step. In that case, we shouldn't try migration.  in move_to_new_page()
1039 		 * Only in the case of a full synchronous migration is it  in __unmap_and_move()
1061 	 * of migration. File cache pages are no problem because of page_lock()  in __unmap_and_move()
1062 	 * File Caches may use write_page() or lock_page() in migration, then,  in __unmap_and_move()
1109 		/* Establish migration ptes */  in __unmap_and_move()
1132 	 * If migration is successful, decrease refcount of the newpage  in __unmap_and_move()
1207 	 * If migration is successful, releases reference grabbed during  in unmap_and_move()
1243  * Counterpart of unmap_and_move_page() for hugepage migration.
1246  * because there is no race between I/O and migration for hugepage.
1254  * hugepage migration fails without data corruption.
1256  * There is also no race when direct I/O is issued on the page under migration,
1257  * because then pte is replaced with migration swap entry and direct I/O code
1258  * will wait in the page fault for migration to complete.
1273 	 * This check is necessary because some callers of hugepage migration  in unmap_and_move_huge_page()
1276 	 * kicking migration.  in unmap_and_move_huge_page()
1368 	 * If migration was not successful and there's a freeing callback, use  in unmap_and_move_huge_page()
1382  *		   supplied as the target for the page migration
1386  *			as the target of the page migration.
1387  * @put_new_page:	The function used to free target pages if migration
1390  * @mode:		The migration mode that specifies the constraints for
1391  *			page migration, if any.
1392  * @reason:		The reason for page migration.
1433 			 * during migration.  in migrate_pages()
1451 				 * THP migration might be unsupported or the  in migrate_pages()
1496 				 * removed from migration page list and not  in migrate_pages()
1552 		 * clear __GFP_RECLAIM to make the migration callback  in alloc_migration_target()
1759 			/* The page is successfully queued for migration */  in do_pages_move()
1973  * Returns true if this is a safe migration target node for misplaced NUMA
2027 	 * migrate_misplaced_transhuge_page() skips page migration's usual  in numamigrate_isolate_page()
2029 	 * has been isolated: a GUP pin, or any other pin, prevents migration.  in numamigrate_isolate_page()
2045 	 * disappearing underneath us during migration.  in numamigrate_isolate_page()
2143 	/* Prepare a page as a migration target */  in migrate_misplaced_transhuge_page()
2408 		 * any kind of migration. Side effect is that it "freezes" the  in migrate_vma_collect_pmd()
2421 		 * set up a special migration page table entry now.  in migrate_vma_collect_pmd()
2429 			/* Setup special migration page table entry */  in migrate_vma_collect_pmd()
2480  * @migrate: migrate struct containing all migration information
2512  * migrate_page_move_mapping(), except that here we allow migration of a
2536 		 * GUP will fail for those. Yet if there is a pending migration  in migrate_vma_check_page()
2537 		 * a thread might try to wait on the pte migration entry and  in migrate_vma_check_page()
2539 		 * differentiate a regular pin from migration wait. Hence to  in migrate_vma_check_page()
2541 		 * infinite loop (one stoping migration because the other is  in migrate_vma_check_page()
2542 		 * waiting on pte migration entry). We always return true here.  in migrate_vma_check_page()
2562  * @migrate: migrate struct containing all migration information
2588 			 * a deadlock between 2 concurrent migration where each  in migrate_vma_prepare()
2669  * migrate_vma_unmap() - replace page mapping with special migration pte entry
2670  * @migrate: migrate struct containing all migration information
2672  * Replace page mapping (CPU page table pte) with a special migration pte entry
2728  * @args: contains the vma, start, and pfns arrays for the migration
2749  * with MIGRATE_PFN_MIGRATE flag in src array unless this is a migration from
2759  * properly set the destination entry like for regular migration.  Note that
2761  * migration was successful for those entries after calling migrate_vma_pages()
2762  * just like for regular migration.
2972  * @migrate: migrate struct containing all migration information
2975  * struct page. This effectively finishes the migration from source page to the
3055  * @migrate: migrate struct containing all migration information
3057  * This replaces the special migration pte entry with either a mapping to the
3058  * new page if migration was successful for that page, or to the original page