6  *  Swap reorganised 29.12.95, Stephen Tweedie
17 #include <linux/swap.h>
50 #include "swap.h"
60  * Some modules use swappable objects and may try to swap them out under
62  * check to see if any swap space is available.
73 static const char Bad_file[] = "Bad swap file entry ";
74 static const char Unused_file[] = "Unused swap file entry ";
75 static const char Bad_offset[] = "Bad swap offset entry ";
76 static const char Unused_offset[] = "Unused swap offset entry ";
122 /* Reclaim the swap entry anyway if possible */
125  * Reclaim the swap entry if there are no more mappings of the
129 /* Reclaim the swap entry if swap is getting full*/
132 /* returns 1 if swap entry is freed */
174  * swapon tell device that all the old swap contents can be discarded,
175  * to allow the swap device to optimize its wear-levelling.
184 	/* Do not discard the swap header page! */  in discard_swap()
244  * swap allocation tell device that a cluster of swap can now be discarded,
245  * to allow the swap device to optimize its wear-levelling.
465 	 * taken by scan_swap_map_slots(), mark the swap entries bad (occupied).  in swap_cluster_schedule_discard()
545 	 * If the swap is discardable, prepare discard the cluster  in free_cluster()
623  * Try to get a swap entry from current cpu's swap entry pool (a cluster). This
767 	 * Cross the swap address space size aligned trunk, choose  in set_cluster_next()
768 	 * another trunk randomly to avoid lock contention on swap  in set_cluster_next()
773 		/* No free swap slots available */  in set_cluster_next()
812 	 * We try to cluster swap pages by allocating them sequentially  in scan_swap_map_slots()
813 	 * in swap.  Once we've allocated SWAPFILE_CLUSTER pages this  in scan_swap_map_slots()
815 	 * a new cluster.  This prevents us from scattering swap pages  in scan_swap_map_slots()
816 	 * all over the entire swap partition, so that we reduce  in scan_swap_map_slots()
817 	 * overall disk seek times between swap pages.  -- sct  in scan_swap_map_slots()
819 	 * And we let swap pages go all over an SSD partition.  Hugh  in scan_swap_map_slots()
825 	 * cluster and swap cache.  For HDD, sequential access is more  in scan_swap_map_slots()
848 		 * start of partition, to minimize the span of allocated swap.  in scan_swap_map_slots()
896 	/* reuse swap entry of cache-only swap if not busy. */  in scan_swap_map_slots()
1010 	 * page swap is disabled.  Warn and fail the allocation.  in swap_alloc_cluster()
1222  * When we get a swap entry, if there aren't some other ways to
1223  * prevent swapoff, such as the folio in swap cache is locked, page
1224  * table lock is held, etc., the swap entry may become invalid because
1225  * of swapoff.  Then, we need to enclose all swap related functions
1226  * with get_swap_device() and put_swap_device(), unless the swap
1229  * Check whether swap entry is valid in the swap device.  If so,
1230  * return pointer to swap_info_struct, and keep the swap entry valid
1231  * via preventing the swap device from being swapoff, until
1250  * changing partly because the specified swap entry may be for another
1251  * swap device which has been swapoff.  And in do_swap_page(), after
1252  * the page is read from the swap device, the PTE is verified not
1253  * changed with the page table locked to check whether the swap device
1325  * Caller has made sure that the swap device corresponding to entry
1338  * Called after dropping swapcache to decrease refcnt to swap entries.
1423 	 * Sort swap entries by swap device, so each lock is only taken once.  in swapcache_free_entries()
1449  * This does not give an exact answer when swap count is continued,
1539 	swp_entry_t entry = folio->swap;  in folio_swapped()
1552  * folio_free_swap() - Free the swap space used for this folio.
1555  * If swap is getting full, or if there are no more mappings of this folio,
1556  * then call folio_free_swap to free its swap space.
1558  * Return: true if we were able to release the swap space.
1575 	 * hibernation is allocating its own swap pages for the image,  in folio_free_swap()
1577 	 * the swap from a folio which has already been recorded in the  in folio_free_swap()
1578 	 * image as a clean swapcache folio, and then reuse its swap for  in folio_free_swap()
1581 	 * later read back in from swap, now with the wrong data.  in folio_free_swap()
1595  * Free the swap entry like above, but also try to
1627 	/* This is called for allocating swap entry, not cache */  in get_swap_page_of_type()
1637  * Find the swap type that corresponds to given device (if any).
1640  * from 0, in which the swap header is expected to be located.
1691  * corresponding to given index in swap_info (swap type).
1705  * Return either the total number of swap pages of given type, or the number
1737  * No need to decide whether this PTE shares the swap entry with others,
1784 	 * when reading from swap. This metadata may be indexed by swap entry  in unuse_pte()
2101 		 * swap cache just before we acquired the page lock. The folio  in try_to_unuse()
2102 		 * might even be back in swap cache on another swap area. But  in try_to_unuse()
2113 	 * Lets check again to see if there are still swap entries in the map.  in try_to_unuse()
2115 	 * Under global memory pressure, swap entries can be reinserted back  in try_to_unuse()
2119 	 * above fails, that mm is likely to be freeing swap from  in try_to_unuse()
2122 	 * folio_alloc_swap(), temporarily hiding that swap.  It's easy  in try_to_unuse()
2135  * After a successful try_to_unuse, if no swap is now in use, we know
2225  * A `swap extent' is a simple thing which maps a contiguous range of pages
2226  * onto a contiguous range of disk blocks.  A rbtree of swap extents is
2232  * swap files identically.
2234  * Whether the swapdev is an S_ISREG file or an S_ISBLK blockdev, the swap
2244  * For all swap devices we set S_SWAPFILE across the life of the swapon.  This
2245  * prevents users from writing to the swap device, which will corrupt memory.
2247  * The amount of disk space which a single swap extent represents varies.
2304 	 * low-to-high, while swap ordering is high-to-low  in setup_swap_info()
2335 	 * which allocates swap pages from the highest available priority  in _enable_swap_info()
2340 	/* add to available list iff swap device is not full */  in _enable_swap_info()
2357 	 * Finished initializing swap device, now it's safe to reference it.  in enable_swap_info()
2466 		/* re-insert swap space back into swap_list */  in SYSCALL_DEFINE1()
2475 	 * Wait for swap operations protected by get/put_swap_device()  in SYSCALL_DEFINE1()
2479 	 * the swap cache data structure.  in SYSCALL_DEFINE1()
2528 	/* Destroy swap account information */  in SYSCALL_DEFINE1()
2581 static void *swap_start(struct seq_file *swap, loff_t *pos)  in swap_start()  argument
2602 static void *swap_next(struct seq_file *swap, void *v, loff_t *pos)  in swap_next()  argument
2622 static void swap_stop(struct seq_file *swap, void *v)  in swap_stop()  argument
2627 static int swap_show(struct seq_file *swap, void *v)  in swap_show()  argument
2635 		seq_puts(swap, "Filename\t\t\t\tType\t\tSize\t\tUsed\t\tPriority\n");  in swap_show()
2643 	len = seq_file_path(swap, file, " \t\n\\");  in swap_show()
2644 	seq_printf(swap, "%*s%s\t%lu\t%s%lu\t%s%d\n",  in swap_show()
2795  * Find out how many pages are allowed for a single swap device. There
2797  * 1) the number of bits for the swap offset in the swp_entry_t type, and
2798  * 2) the number of bits in the swap pte, as defined by the different
2801  * In order to find the largest possible bit mask, a swap entry with
2802  * swap type 0 and swap offset ~0UL is created, encoded to a swap pte,
2803  * decoded to a swp_entry_t again, and finally the swap offset is
2808  * of a swap pte.
2832 		pr_err("Unable to find swap-space signature\n");  in read_swap_header()
2836 	/* swap partition endianness hack... */  in read_swap_header()
2846 	/* Check the swap header's sub-version */  in read_swap_header()
2848 		pr_warn("Unable to handle swap header version %d\n",  in read_swap_header()
2860 		pr_warn("Empty swap-file\n");  in read_swap_header()
2864 		pr_warn("Truncating oversized swap area, only using %luk out of %luk\n",  in read_swap_header()
2879 		pr_warn("Swap area shorter than signature indicates\n");  in read_swap_header()
2950 		pr_warn("Empty swap-file\n");  in setup_swap_map_and_extents()
3045 	 * Read the swap header.  in SYSCALL_DEFINE2()
3064 	/* OK, set up the swap map and apply the bad block list */  in SYSCALL_DEFINE2()
3136 		 * When discard is enabled for swap with no particular  in SYSCALL_DEFINE2()
3137 		 * policy flagged, we set all swap discard flags here in  in SYSCALL_DEFINE2()
3147 		 * perform discards for released swap page-clusters.  in SYSCALL_DEFINE2()
3170 	 * swap device.  in SYSCALL_DEFINE2()
3186 	pr_info("Adding %uk swap on %s.  Priority:%d extents:%d across:%lluk %s%s%s%s\n",  in SYSCALL_DEFINE2()
3258  * Verify that a swap entry is valid and increment its swap map count.
3264  * - swap-cache reference is requested but there is already one. -> EEXIST
3265  * - swap-cache reference is requested but the entry is not used. -> ENOENT
3266  * - swap-mapped reference requested but needs continued swap count. -> ENOMEM
3285 	 * swapin_readahead() doesn't check if a swap entry is valid, so the  in __swap_duplicate()
3286 	 * swap entry could be SWAP_MAP_BAD. Check here with lock held.  in __swap_duplicate()
3318 		err = -ENOENT;			/* unused swap entry */  in __swap_duplicate()
3328  * Help swapoff by noting that swap entry belongs to shmem/tmpfs
3337  * Increase reference count of swap entry by 1.
3353  * @entry: swap entry for which we allocate swap cache.
3355  * Called when allocating swap cache for existing swap entry,
3357  * -EEXIST means there is a swap cache.
3400 	swp_entry_t swap = page_swap_entry(page);  in __page_file_index()  local
3401 	return swp_offset(swap);  in __page_file_index()
3406  * add_swap_count_continuation - called when a swap count is duplicated
3409  * (for that entry and for its neighbouring PAGE_SIZE swap entries).  Called
3441 		 * __swap_duplicate(): the swap device may be swapoff  in add_swap_count_continuation()
3455 		 * The higher the swap count, the more likely it is that tasks  in add_swap_count_continuation()
3456 		 * will race to add swap count continuation: we need to avoid  in add_swap_count_continuation()
3524  * Called while __swap_duplicate() or swap_entry_free() holds swap or cluster
3642 	 * We've already scheduled a throttle, avoid taking the global swap  in __folio_throttle_swaprate()
3667 		pr_emerg("Not enough memory for swap heads, swap is disabled\n");  in swapfile_init()