diff options
-rw-r--r-- | include/linux/swap.h | 2 | ||||
-rw-r--r-- | mm/swapfile.c | 155 |
2 files changed, 65 insertions, 92 deletions
diff --git a/include/linux/swap.h b/include/linux/swap.h index 109dfe794237..c9d8870892b8 100644 --- a/include/linux/swap.h +++ b/include/linux/swap.h | |||
@@ -154,7 +154,7 @@ enum { | |||
154 | #define SWAP_MAP_MAX 0x7ffe | 154 | #define SWAP_MAP_MAX 0x7ffe |
155 | #define SWAP_MAP_BAD 0x7fff | 155 | #define SWAP_MAP_BAD 0x7fff |
156 | #define SWAP_HAS_CACHE 0x8000 /* There is a swap cache of entry. */ | 156 | #define SWAP_HAS_CACHE 0x8000 /* There is a swap cache of entry. */ |
157 | #define SWAP_COUNT_MASK (~SWAP_HAS_CACHE) | 157 | |
158 | /* | 158 | /* |
159 | * The in-memory structure used to track swap areas. | 159 | * The in-memory structure used to track swap areas. |
160 | */ | 160 | */ |
diff --git a/mm/swapfile.c b/mm/swapfile.c index fa5f10b9c28b..52497490a7ca 100644 --- a/mm/swapfile.c +++ b/mm/swapfile.c | |||
@@ -53,30 +53,9 @@ static struct swap_info_struct *swap_info[MAX_SWAPFILES]; | |||
53 | 53 | ||
54 | static DEFINE_MUTEX(swapon_mutex); | 54 | static DEFINE_MUTEX(swapon_mutex); |
55 | 55 | ||
56 | /* For reference count accounting in swap_map */ | ||
57 | /* enum for swap_map[] handling. internal use only */ | ||
58 | enum { | ||
59 | SWAP_MAP = 0, /* ops for reference from swap users */ | ||
60 | SWAP_CACHE, /* ops for reference from swap cache */ | ||
61 | }; | ||
62 | |||
63 | static inline int swap_count(unsigned short ent) | 56 | static inline int swap_count(unsigned short ent) |
64 | { | 57 | { |
65 | return ent & SWAP_COUNT_MASK; | 58 | return ent & ~SWAP_HAS_CACHE; |
66 | } | ||
67 | |||
68 | static inline bool swap_has_cache(unsigned short ent) | ||
69 | { | ||
70 | return !!(ent & SWAP_HAS_CACHE); | ||
71 | } | ||
72 | |||
73 | static inline unsigned short encode_swapmap(int count, bool has_cache) | ||
74 | { | ||
75 | unsigned short ret = count; | ||
76 | |||
77 | if (has_cache) | ||
78 | return SWAP_HAS_CACHE | ret; | ||
79 | return ret; | ||
80 | } | 59 | } |
81 | 60 | ||
82 | /* returns 1 if swap entry is freed */ | 61 | /* returns 1 if swap entry is freed */ |
@@ -224,7 +203,7 @@ static int wait_for_discard(void *word) | |||
224 | #define LATENCY_LIMIT 256 | 203 | #define LATENCY_LIMIT 256 |
225 | 204 | ||
226 | static inline unsigned long scan_swap_map(struct swap_info_struct *si, | 205 | static inline unsigned long scan_swap_map(struct swap_info_struct *si, |
227 | int cache) | 206 | unsigned short usage) |
228 | { | 207 | { |
229 | unsigned long offset; | 208 | unsigned long offset; |
230 | unsigned long scan_base; | 209 | unsigned long scan_base; |
@@ -355,10 +334,7 @@ checks: | |||
355 | si->lowest_bit = si->max; | 334 | si->lowest_bit = si->max; |
356 | si->highest_bit = 0; | 335 | si->highest_bit = 0; |
357 | } | 336 | } |
358 | if (cache == SWAP_CACHE) /* at usual swap-out via vmscan.c */ | 337 | si->swap_map[offset] = usage; |
359 | si->swap_map[offset] = encode_swapmap(0, true); | ||
360 | else /* at suspend */ | ||
361 | si->swap_map[offset] = encode_swapmap(1, false); | ||
362 | si->cluster_next = offset + 1; | 338 | si->cluster_next = offset + 1; |
363 | si->flags -= SWP_SCANNING; | 339 | si->flags -= SWP_SCANNING; |
364 | 340 | ||
@@ -483,7 +459,7 @@ swp_entry_t get_swap_page(void) | |||
483 | 459 | ||
484 | swap_list.next = next; | 460 | swap_list.next = next; |
485 | /* This is called for allocating swap entry for cache */ | 461 | /* This is called for allocating swap entry for cache */ |
486 | offset = scan_swap_map(si, SWAP_CACHE); | 462 | offset = scan_swap_map(si, SWAP_HAS_CACHE); |
487 | if (offset) { | 463 | if (offset) { |
488 | spin_unlock(&swap_lock); | 464 | spin_unlock(&swap_lock); |
489 | return swp_entry(type, offset); | 465 | return swp_entry(type, offset); |
@@ -508,7 +484,7 @@ swp_entry_t get_swap_page_of_type(int type) | |||
508 | if (si && (si->flags & SWP_WRITEOK)) { | 484 | if (si && (si->flags & SWP_WRITEOK)) { |
509 | nr_swap_pages--; | 485 | nr_swap_pages--; |
510 | /* This is called for allocating swap entry, not cache */ | 486 | /* This is called for allocating swap entry, not cache */ |
511 | offset = scan_swap_map(si, SWAP_MAP); | 487 | offset = scan_swap_map(si, 1); |
512 | if (offset) { | 488 | if (offset) { |
513 | spin_unlock(&swap_lock); | 489 | spin_unlock(&swap_lock); |
514 | return swp_entry(type, offset); | 490 | return swp_entry(type, offset); |
@@ -555,29 +531,31 @@ out: | |||
555 | return NULL; | 531 | return NULL; |
556 | } | 532 | } |
557 | 533 | ||
558 | static int swap_entry_free(struct swap_info_struct *p, | 534 | static unsigned short swap_entry_free(struct swap_info_struct *p, |
559 | swp_entry_t ent, int cache) | 535 | swp_entry_t entry, unsigned short usage) |
560 | { | 536 | { |
561 | unsigned long offset = swp_offset(ent); | 537 | unsigned long offset = swp_offset(entry); |
562 | int count = swap_count(p->swap_map[offset]); | 538 | unsigned short count; |
563 | bool has_cache; | 539 | unsigned short has_cache; |
564 | 540 | ||
565 | has_cache = swap_has_cache(p->swap_map[offset]); | 541 | count = p->swap_map[offset]; |
542 | has_cache = count & SWAP_HAS_CACHE; | ||
543 | count &= ~SWAP_HAS_CACHE; | ||
566 | 544 | ||
567 | if (cache == SWAP_MAP) { /* dropping usage count of swap */ | 545 | if (usage == SWAP_HAS_CACHE) { |
568 | if (count < SWAP_MAP_MAX) { | ||
569 | count--; | ||
570 | p->swap_map[offset] = encode_swapmap(count, has_cache); | ||
571 | } | ||
572 | } else { /* dropping swap cache flag */ | ||
573 | VM_BUG_ON(!has_cache); | 546 | VM_BUG_ON(!has_cache); |
574 | p->swap_map[offset] = encode_swapmap(count, false); | 547 | has_cache = 0; |
548 | } else if (count < SWAP_MAP_MAX) | ||
549 | count--; | ||
550 | |||
551 | if (!count) | ||
552 | mem_cgroup_uncharge_swap(entry); | ||
553 | |||
554 | usage = count | has_cache; | ||
555 | p->swap_map[offset] = usage; | ||
575 | 556 | ||
576 | } | ||
577 | /* return code. */ | ||
578 | count = p->swap_map[offset]; | ||
579 | /* free if no reference */ | 557 | /* free if no reference */ |
580 | if (!count) { | 558 | if (!usage) { |
581 | if (offset < p->lowest_bit) | 559 | if (offset < p->lowest_bit) |
582 | p->lowest_bit = offset; | 560 | p->lowest_bit = offset; |
583 | if (offset > p->highest_bit) | 561 | if (offset > p->highest_bit) |
@@ -588,9 +566,8 @@ static int swap_entry_free(struct swap_info_struct *p, | |||
588 | nr_swap_pages++; | 566 | nr_swap_pages++; |
589 | p->inuse_pages--; | 567 | p->inuse_pages--; |
590 | } | 568 | } |
591 | if (!swap_count(count)) | 569 | |
592 | mem_cgroup_uncharge_swap(ent); | 570 | return usage; |
593 | return count; | ||
594 | } | 571 | } |
595 | 572 | ||
596 | /* | 573 | /* |
@@ -603,7 +580,7 @@ void swap_free(swp_entry_t entry) | |||
603 | 580 | ||
604 | p = swap_info_get(entry); | 581 | p = swap_info_get(entry); |
605 | if (p) { | 582 | if (p) { |
606 | swap_entry_free(p, entry, SWAP_MAP); | 583 | swap_entry_free(p, entry, 1); |
607 | spin_unlock(&swap_lock); | 584 | spin_unlock(&swap_lock); |
608 | } | 585 | } |
609 | } | 586 | } |
@@ -614,19 +591,13 @@ void swap_free(swp_entry_t entry) | |||
614 | void swapcache_free(swp_entry_t entry, struct page *page) | 591 | void swapcache_free(swp_entry_t entry, struct page *page) |
615 | { | 592 | { |
616 | struct swap_info_struct *p; | 593 | struct swap_info_struct *p; |
617 | int ret; | 594 | unsigned short count; |
618 | 595 | ||
619 | p = swap_info_get(entry); | 596 | p = swap_info_get(entry); |
620 | if (p) { | 597 | if (p) { |
621 | ret = swap_entry_free(p, entry, SWAP_CACHE); | 598 | count = swap_entry_free(p, entry, SWAP_HAS_CACHE); |
622 | if (page) { | 599 | if (page) |
623 | bool swapout; | 600 | mem_cgroup_uncharge_swapcache(page, entry, count != 0); |
624 | if (ret) | ||
625 | swapout = true; /* the end of swap out */ | ||
626 | else | ||
627 | swapout = false; /* no more swap users! */ | ||
628 | mem_cgroup_uncharge_swapcache(page, entry, swapout); | ||
629 | } | ||
630 | spin_unlock(&swap_lock); | 601 | spin_unlock(&swap_lock); |
631 | } | 602 | } |
632 | } | 603 | } |
@@ -705,7 +676,7 @@ int free_swap_and_cache(swp_entry_t entry) | |||
705 | 676 | ||
706 | p = swap_info_get(entry); | 677 | p = swap_info_get(entry); |
707 | if (p) { | 678 | if (p) { |
708 | if (swap_entry_free(p, entry, SWAP_MAP) == SWAP_HAS_CACHE) { | 679 | if (swap_entry_free(p, entry, 1) == SWAP_HAS_CACHE) { |
709 | page = find_get_page(&swapper_space, entry.val); | 680 | page = find_get_page(&swapper_space, entry.val); |
710 | if (page && !trylock_page(page)) { | 681 | if (page && !trylock_page(page)) { |
711 | page_cache_release(page); | 682 | page_cache_release(page); |
@@ -1212,7 +1183,7 @@ static int try_to_unuse(unsigned int type) | |||
1212 | 1183 | ||
1213 | if (swap_count(*swap_map) == SWAP_MAP_MAX) { | 1184 | if (swap_count(*swap_map) == SWAP_MAP_MAX) { |
1214 | spin_lock(&swap_lock); | 1185 | spin_lock(&swap_lock); |
1215 | *swap_map = encode_swapmap(0, true); | 1186 | *swap_map = SWAP_HAS_CACHE; |
1216 | spin_unlock(&swap_lock); | 1187 | spin_unlock(&swap_lock); |
1217 | reset_overflow = 1; | 1188 | reset_overflow = 1; |
1218 | } | 1189 | } |
@@ -2111,16 +2082,16 @@ void si_swapinfo(struct sysinfo *val) | |||
2111 | * - swap-cache reference is requested but there is already one. -> EEXIST | 2082 | * - swap-cache reference is requested but there is already one. -> EEXIST |
2112 | * - swap-cache reference is requested but the entry is not used. -> ENOENT | 2083 | * - swap-cache reference is requested but the entry is not used. -> ENOENT |
2113 | */ | 2084 | */ |
2114 | static int __swap_duplicate(swp_entry_t entry, bool cache) | 2085 | static int __swap_duplicate(swp_entry_t entry, unsigned short usage) |
2115 | { | 2086 | { |
2116 | struct swap_info_struct *p; | 2087 | struct swap_info_struct *p; |
2117 | unsigned long offset, type; | 2088 | unsigned long offset, type; |
2118 | int result = -EINVAL; | 2089 | unsigned short count; |
2119 | int count; | 2090 | unsigned short has_cache; |
2120 | bool has_cache; | 2091 | int err = -EINVAL; |
2121 | 2092 | ||
2122 | if (non_swap_entry(entry)) | 2093 | if (non_swap_entry(entry)) |
2123 | return -EINVAL; | 2094 | goto out; |
2124 | 2095 | ||
2125 | type = swp_type(entry); | 2096 | type = swp_type(entry); |
2126 | if (type >= nr_swapfiles) | 2097 | if (type >= nr_swapfiles) |
@@ -2129,54 +2100,56 @@ static int __swap_duplicate(swp_entry_t entry, bool cache) | |||
2129 | offset = swp_offset(entry); | 2100 | offset = swp_offset(entry); |
2130 | 2101 | ||
2131 | spin_lock(&swap_lock); | 2102 | spin_lock(&swap_lock); |
2132 | |||
2133 | if (unlikely(offset >= p->max)) | 2103 | if (unlikely(offset >= p->max)) |
2134 | goto unlock_out; | 2104 | goto unlock_out; |
2135 | 2105 | ||
2136 | count = swap_count(p->swap_map[offset]); | 2106 | count = p->swap_map[offset]; |
2137 | has_cache = swap_has_cache(p->swap_map[offset]); | 2107 | has_cache = count & SWAP_HAS_CACHE; |
2108 | count &= ~SWAP_HAS_CACHE; | ||
2109 | err = 0; | ||
2138 | 2110 | ||
2139 | if (cache == SWAP_CACHE) { /* called for swapcache/swapin-readahead */ | 2111 | if (usage == SWAP_HAS_CACHE) { |
2140 | 2112 | ||
2141 | /* set SWAP_HAS_CACHE if there is no cache and entry is used */ | 2113 | /* set SWAP_HAS_CACHE if there is no cache and entry is used */ |
2142 | if (!has_cache && count) { | 2114 | if (!has_cache && count) |
2143 | p->swap_map[offset] = encode_swapmap(count, true); | 2115 | has_cache = SWAP_HAS_CACHE; |
2144 | result = 0; | 2116 | else if (has_cache) /* someone else added cache */ |
2145 | } else if (has_cache) /* someone added cache */ | 2117 | err = -EEXIST; |
2146 | result = -EEXIST; | 2118 | else /* no users remaining */ |
2147 | else if (!count) /* no users */ | 2119 | err = -ENOENT; |
2148 | result = -ENOENT; | ||
2149 | 2120 | ||
2150 | } else if (count || has_cache) { | 2121 | } else if (count || has_cache) { |
2151 | if (count < SWAP_MAP_MAX - 1) { | 2122 | |
2152 | p->swap_map[offset] = encode_swapmap(count + 1, | 2123 | if (count < SWAP_MAP_MAX - 1) |
2153 | has_cache); | 2124 | count++; |
2154 | result = 0; | 2125 | else if (count <= SWAP_MAP_MAX) { |
2155 | } else if (count <= SWAP_MAP_MAX) { | ||
2156 | if (swap_overflow++ < 5) | 2126 | if (swap_overflow++ < 5) |
2157 | printk(KERN_WARNING | 2127 | printk(KERN_WARNING |
2158 | "swap_dup: swap entry overflow\n"); | 2128 | "swap_dup: swap entry overflow\n"); |
2159 | p->swap_map[offset] = encode_swapmap(SWAP_MAP_MAX, | 2129 | count = SWAP_MAP_MAX; |
2160 | has_cache); | 2130 | } else |
2161 | result = 0; | 2131 | err = -EINVAL; |
2162 | } | ||
2163 | } else | 2132 | } else |
2164 | result = -ENOENT; /* unused swap entry */ | 2133 | err = -ENOENT; /* unused swap entry */ |
2134 | |||
2135 | p->swap_map[offset] = count | has_cache; | ||
2136 | |||
2165 | unlock_out: | 2137 | unlock_out: |
2166 | spin_unlock(&swap_lock); | 2138 | spin_unlock(&swap_lock); |
2167 | out: | 2139 | out: |
2168 | return result; | 2140 | return err; |
2169 | 2141 | ||
2170 | bad_file: | 2142 | bad_file: |
2171 | printk(KERN_ERR "swap_dup: %s%08lx\n", Bad_file, entry.val); | 2143 | printk(KERN_ERR "swap_dup: %s%08lx\n", Bad_file, entry.val); |
2172 | goto out; | 2144 | goto out; |
2173 | } | 2145 | } |
2146 | |||
2174 | /* | 2147 | /* |
2175 | * increase reference count of swap entry by 1. | 2148 | * increase reference count of swap entry by 1. |
2176 | */ | 2149 | */ |
2177 | void swap_duplicate(swp_entry_t entry) | 2150 | void swap_duplicate(swp_entry_t entry) |
2178 | { | 2151 | { |
2179 | __swap_duplicate(entry, SWAP_MAP); | 2152 | __swap_duplicate(entry, 1); |
2180 | } | 2153 | } |
2181 | 2154 | ||
2182 | /* | 2155 | /* |
@@ -2189,7 +2162,7 @@ void swap_duplicate(swp_entry_t entry) | |||
2189 | */ | 2162 | */ |
2190 | int swapcache_prepare(swp_entry_t entry) | 2163 | int swapcache_prepare(swp_entry_t entry) |
2191 | { | 2164 | { |
2192 | return __swap_duplicate(entry, SWAP_CACHE); | 2165 | return __swap_duplicate(entry, SWAP_HAS_CACHE); |
2193 | } | 2166 | } |
2194 | 2167 | ||
2195 | /* | 2168 | /* |