aboutsummaryrefslogtreecommitdiff
path: root/mm/frontswap.c
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2013-04-30 17:37:43 -0700
committerLinus Torvalds <torvalds@linux-foundation.org>2013-04-30 17:37:43 -0700
commit5f56886521d6ddd3648777fae44d82382dd8c87f (patch)
treeaa0db6331cdb01c23f1884439840aadd31bbcca4 /mm/frontswap.c
parentf1e9a236e5ddab6c349611ee86f54291916f226c (diff)
parente2a8b0a779787314eca1061308a8182e6c5bfabd (diff)
downloadlinux-master.tar.gz
Merge branch 'akpm' (incoming from Andrew)HEADmaster
Merge third batch of fixes from Andrew Morton: "Most of the rest. I still have two large patchsets against AIO and IPC, but they're a bit stuck behind other trees and I'm about to vanish for six days. - random fixlets - inotify - more of the MM queue - show_stack() cleanups - DMI update - kthread/workqueue things - compat cleanups - epoll udpates - binfmt updates - nilfs2 - hfs - hfsplus - ptrace - kmod - coredump - kexec - rbtree - pids - pidns - pps - semaphore tweaks - some w1 patches - relay updates - core Kconfig changes - sysrq tweaks" * emailed patches from Andrew Morton <akpm@linux-foundation.org>: (109 commits) Documentation/sysrq: fix inconstistent help message of sysrq key ethernet/emac/sysrq: fix inconstistent help message of sysrq key sparc/sysrq: fix inconstistent help message of sysrq key powerpc/xmon/sysrq: fix inconstistent help message of sysrq key ARM/etm/sysrq: fix inconstistent help message of sysrq key power/sysrq: fix inconstistent help message of sysrq key kgdb/sysrq: fix inconstistent help message of sysrq key lib/decompress.c: fix initconst notifier-error-inject: fix module names in Kconfig kernel/sys.c: make prctl(PR_SET_MM) generally available UAPI: remove empty Kbuild files menuconfig: print more info for symbol without prompts init/Kconfig: re-order CONFIG_EXPERT options to fix menuconfig display kconfig menu: move Virtualization drivers near other virtualization options Kconfig: consolidate CONFIG_DEBUG_STRICT_USER_COPY_CHECKS relay: use macro PAGE_ALIGN instead of FIX_SIZE kernel/relay.c: move FIX_SIZE macro into relay.c kernel/relay.c: remove unused function argument actor drivers/w1/slaves/w1_ds2760.c: fix the error handling in w1_ds2760_add_slave() drivers/w1/slaves/w1_ds2781.c: fix the error handling in w1_ds2781_add_slave() ...
Diffstat (limited to 'mm/frontswap.c')
-rw-r--r--mm/frontswap.c156
1 files changed, 123 insertions, 33 deletions
diff --git a/mm/frontswap.c b/mm/frontswap.c
index 2890e67d602..538367ef137 100644
--- a/mm/frontswap.c
+++ b/mm/frontswap.c
@@ -24,15 +24,7 @@
* frontswap_ops is set by frontswap_register_ops to contain the pointers
* to the frontswap "backend" implementation functions.
*/
-static struct frontswap_ops frontswap_ops __read_mostly;
-
-/*
- * This global enablement flag reduces overhead on systems where frontswap_ops
- * has not been registered, so is preferred to the slower alternative: a
- * function call that checks a non-global.
- */
-bool frontswap_enabled __read_mostly;
-EXPORT_SYMBOL(frontswap_enabled);
+static struct frontswap_ops *frontswap_ops __read_mostly;
/*
* If enabled, frontswap_store will return failure even on success. As
@@ -80,16 +72,70 @@ static inline void inc_frontswap_succ_stores(void) { }
static inline void inc_frontswap_failed_stores(void) { }
static inline void inc_frontswap_invalidates(void) { }
#endif
+
+/*
+ * Due to the asynchronous nature of the backends loading potentially
+ * _after_ the swap system has been activated, we have chokepoints
+ * on all frontswap functions to not call the backend until the backend
+ * has registered.
+ *
+ * Specifically when no backend is registered (nobody called
+ * frontswap_register_ops) all calls to frontswap_init (which is done via
+ * swapon -> enable_swap_info -> frontswap_init) are registered and remembered
+ * (via the setting of need_init bitmap) but fail to create tmem_pools. When a
+ * backend registers with frontswap at some later point the previous
+ * calls to frontswap_init are executed (by iterating over the need_init
+ * bitmap) to create tmem_pools and set the respective poolids. All of that is
+ * guarded by us using atomic bit operations on the 'need_init' bitmap.
+ *
+ * This would not guards us against the user deciding to call swapoff right as
+ * we are calling the backend to initialize (so swapon is in action).
+ * Fortunatly for us, the swapon_mutex has been taked by the callee so we are
+ * OK. The other scenario where calls to frontswap_store (called via
+ * swap_writepage) is racing with frontswap_invalidate_area (called via
+ * swapoff) is again guarded by the swap subsystem.
+ *
+ * While no backend is registered all calls to frontswap_[store|load|
+ * invalidate_area|invalidate_page] are ignored or fail.
+ *
+ * The time between the backend being registered and the swap file system
+ * calling the backend (via the frontswap_* functions) is indeterminate as
+ * frontswap_ops is not atomic_t (or a value guarded by a spinlock).
+ * That is OK as we are comfortable missing some of these calls to the newly
+ * registered backend.
+ *
+ * Obviously the opposite (unloading the backend) must be done after all
+ * the frontswap_[store|load|invalidate_area|invalidate_page] start
+ * ignorning or failing the requests - at which point frontswap_ops
+ * would have to be made in some fashion atomic.
+ */
+static DECLARE_BITMAP(need_init, MAX_SWAPFILES);
+
/*
* Register operations for frontswap, returning previous thus allowing
* detection of multiple backends and possible nesting.
*/
-struct frontswap_ops frontswap_register_ops(struct frontswap_ops *ops)
+struct frontswap_ops *frontswap_register_ops(struct frontswap_ops *ops)
{
- struct frontswap_ops old = frontswap_ops;
-
- frontswap_ops = *ops;
- frontswap_enabled = true;
+ struct frontswap_ops *old = frontswap_ops;
+ int i;
+
+ for (i = 0; i < MAX_SWAPFILES; i++) {
+ if (test_and_clear_bit(i, need_init)) {
+ struct swap_info_struct *sis = swap_info[i];
+ /* __frontswap_init _should_ have set it! */
+ if (!sis->frontswap_map)
+ return ERR_PTR(-EINVAL);
+ ops->init(i);
+ }
+ }
+ /*
+ * We MUST have frontswap_ops set _after_ the frontswap_init's
+ * have been called. Otherwise __frontswap_store might fail. Hence
+ * the barrier to make sure compiler does not re-order us.
+ */
+ barrier();
+ frontswap_ops = ops;
return old;
}
EXPORT_SYMBOL(frontswap_register_ops);
@@ -115,20 +161,48 @@ EXPORT_SYMBOL(frontswap_tmem_exclusive_gets);
/*
* Called when a swap device is swapon'd.
*/
-void __frontswap_init(unsigned type)
+void __frontswap_init(unsigned type, unsigned long *map)
{
struct swap_info_struct *sis = swap_info[type];
BUG_ON(sis == NULL);
- if (sis->frontswap_map == NULL)
+
+ /*
+ * p->frontswap is a bitmap that we MUST have to figure out which page
+ * has gone in frontswap. Without it there is no point of continuing.
+ */
+ if (WARN_ON(!map))
return;
- frontswap_ops.init(type);
+ /*
+ * Irregardless of whether the frontswap backend has been loaded
+ * before this function or it will be later, we _MUST_ have the
+ * p->frontswap set to something valid to work properly.
+ */
+ frontswap_map_set(sis, map);
+ if (frontswap_ops)
+ frontswap_ops->init(type);
+ else {
+ BUG_ON(type > MAX_SWAPFILES);
+ set_bit(type, need_init);
+ }
}
EXPORT_SYMBOL(__frontswap_init);
-static inline void __frontswap_clear(struct swap_info_struct *sis, pgoff_t offset)
+bool __frontswap_test(struct swap_info_struct *sis,
+ pgoff_t offset)
+{
+ bool ret = false;
+
+ if (frontswap_ops && sis->frontswap_map)
+ ret = test_bit(offset, sis->frontswap_map);
+ return ret;
+}
+EXPORT_SYMBOL(__frontswap_test);
+
+static inline void __frontswap_clear(struct swap_info_struct *sis,
+ pgoff_t offset)
{
- frontswap_clear(sis, offset);
+ clear_bit(offset, sis->frontswap_map);
atomic_dec(&sis->frontswap_pages);
}
@@ -147,13 +221,20 @@ int __frontswap_store(struct page *page)
struct swap_info_struct *sis = swap_info[type];
pgoff_t offset = swp_offset(entry);
+ /*
+ * Return if no backend registed.
+ * Don't need to inc frontswap_failed_stores here.
+ */
+ if (!frontswap_ops)
+ return ret;
+
BUG_ON(!PageLocked(page));
BUG_ON(sis == NULL);
- if (frontswap_test(sis, offset))
+ if (__frontswap_test(sis, offset))
dup = 1;
- ret = frontswap_ops.store(type, offset, page);
+ ret = frontswap_ops->store(type, offset, page);
if (ret == 0) {
- frontswap_set(sis, offset);
+ set_bit(offset, sis->frontswap_map);
inc_frontswap_succ_stores();
if (!dup)
atomic_inc(&sis->frontswap_pages);
@@ -188,13 +269,16 @@ int __frontswap_load(struct page *page)
BUG_ON(!PageLocked(page));
BUG_ON(sis == NULL);
- if (frontswap_test(sis, offset))
- ret = frontswap_ops.load(type, offset, page);
+ /*
+ * __frontswap_test() will check whether there is backend registered
+ */
+ if (__frontswap_test(sis, offset))
+ ret = frontswap_ops->load(type, offset, page);
if (ret == 0) {
inc_frontswap_loads();
if (frontswap_tmem_exclusive_gets_enabled) {
SetPageDirty(page);
- frontswap_clear(sis, offset);
+ __frontswap_clear(sis, offset);
}
}
return ret;
@@ -210,8 +294,11 @@ void __frontswap_invalidate_page(unsigned type, pgoff_t offset)
struct swap_info_struct *sis = swap_info[type];
BUG_ON(sis == NULL);
- if (frontswap_test(sis, offset)) {
- frontswap_ops.invalidate_page(type, offset);
+ /*
+ * __frontswap_test() will check whether there is backend registered
+ */
+ if (__frontswap_test(sis, offset)) {
+ frontswap_ops->invalidate_page(type, offset);
__frontswap_clear(sis, offset);
inc_frontswap_invalidates();
}
@@ -226,12 +313,15 @@ void __frontswap_invalidate_area(unsigned type)
{
struct swap_info_struct *sis = swap_info[type];
- BUG_ON(sis == NULL);
- if (sis->frontswap_map == NULL)
- return;
- frontswap_ops.invalidate_area(type);
- atomic_set(&sis->frontswap_pages, 0);
- memset(sis->frontswap_map, 0, sis->max / sizeof(long));
+ if (frontswap_ops) {
+ BUG_ON(sis == NULL);
+ if (sis->frontswap_map == NULL)
+ return;
+ frontswap_ops->invalidate_area(type);
+ atomic_set(&sis->frontswap_pages, 0);
+ memset(sis->frontswap_map, 0, sis->max / sizeof(long));
+ }
+ clear_bit(type, need_init);
}
EXPORT_SYMBOL(__frontswap_invalidate_area);