From 29f233cfffe7fbc6672938117ce7e4154a2f515f Mon Sep 17 00:00:00 2001 From: Dan Magenheimer Date: Mon, 9 Apr 2012 17:09:27 -0600 Subject: mm: frontswap: core frontswap functionality This patch, 3of4, provides the core frontswap code that interfaces between the hooks in the swap subsystem and a frontswap backend via frontswap_ops. --- New file added: mm/frontswap.c [v14: add support for writethrough, per suggestion by aarcange@redhat.com] [v11: sjenning@linux.vnet.ibm.com: s/puts/failed_puts/] [v10: sjenning@linux.vnet.ibm.com: fix debugfs calls on 32-bit] [v9: akpm@linux-foundation.org: change "flush" to "invalidate", part 1] [v9: akpm@linux-foundation.org: mark some statics __read_mostly] [v9: akpm@linux-foundation.org: add clarifying comments] [v9: akpm@linux-foundation.org: no need to loop repeating try_to_unuse] [v9: error27@gmail.com: remove superfluous check for NULL] [v8: rebase to 3.0-rc4] [v8: kamezawa.hiroyu@jp.fujitsu.com: add comment to clarify find_next_to_unuse] [v7: rebase to 3.0-rc3] [v7: JBeulich@novell.com: use new static inlines, no-ops if not config'd] [v6: rebase to 3.1-rc1] [v6: lliubbo@gmail.com: use vzalloc] [v6: lliubbo@gmail.com: fix null pointer deref if vzalloc fails] [v6: konrad.wilk@oracl.com: various checks and code clarifications/comments] [v4: rebase to 2.6.39] Signed-off-by: Dan Magenheimer Acked-by: Jan Beulich Acked-by: Seth Jennings Cc: Jeremy Fitzhardinge Cc: Hugh Dickins Cc: Johannes Weiner Cc: Nitin Gupta Cc: Matthew Wilcox Cc: Chris Mason Cc: Rik Riel Cc: Andrew Morton [v12: Squashed s/flush/invalidate/ in] [v15: A bit of cleanup and seperate DEBUGFS] Signed-off-by: Konrad Wilk --- mm/frontswap.c | 314 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 314 insertions(+) create mode 100644 mm/frontswap.c (limited to 'mm/frontswap.c') diff --git a/mm/frontswap.c b/mm/frontswap.c new file mode 100644 index 000000000000..8c0a5f8683f0 --- /dev/null +++ b/mm/frontswap.c @@ -0,0 +1,314 @@ +/* + * Frontswap frontend + * + * This code provides the generic "frontend" layer to call a matching + * "backend" driver implementation of frontswap. See + * Documentation/vm/frontswap.txt for more information. + * + * Copyright (C) 2009-2012 Oracle Corp. All rights reserved. + * Author: Dan Magenheimer + * + * This work is licensed under the terms of the GNU GPL, version 2. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* + * 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); + +/* + * If enabled, frontswap_put will return failure even on success. As + * a result, the swap subsystem will always write the page to swap, in + * effect converting frontswap into a writethrough cache. In this mode, + * there is no direct reduction in swap writes, but a frontswap backend + * can unilaterally "reclaim" any pages in use with no data loss, thus + * providing increases control over maximum memory usage due to frontswap. + */ +static bool frontswap_writethrough_enabled __read_mostly; + +#ifdef CONFIG_DEBUG_FS +/* + * Counters available via /sys/kernel/debug/frontswap (if debugfs is + * properly configured). These are for information only so are not protected + * against increment races. + */ +static u64 frontswap_gets; +static u64 frontswap_succ_puts; +static u64 frontswap_failed_puts; +static u64 frontswap_invalidates; + +static inline void inc_frontswap_gets(void) { + frontswap_gets++; +} +static inline void inc_frontswap_succ_puts(void) { + frontswap_succ_puts++; +} +static inline void inc_frontswap_failed_puts(void) { + frontswap_failed_puts++; +} +static inline void inc_frontswap_invalidates(void) { + frontswap_invalidates++; +} +#else +static inline void inc_frontswap_gets(void) { } +static inline void inc_frontswap_succ_puts(void) { } +static inline void inc_frontswap_failed_puts(void) { } +static inline void inc_frontswap_invalidates(void) { } +#endif +/* + * 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 old = frontswap_ops; + + frontswap_ops = *ops; + frontswap_enabled = true; + return old; +} +EXPORT_SYMBOL(frontswap_register_ops); + +/* + * Enable/disable frontswap writethrough (see above). + */ +void frontswap_writethrough(bool enable) +{ + frontswap_writethrough_enabled = enable; +} +EXPORT_SYMBOL(frontswap_writethrough); + +/* + * Called when a swap device is swapon'd. + */ +void __frontswap_init(unsigned type) +{ + struct swap_info_struct *sis = swap_info[type]; + + BUG_ON(sis == NULL); + if (sis->frontswap_map == NULL) + return; + if (frontswap_enabled) + (*frontswap_ops.init)(type); +} +EXPORT_SYMBOL(__frontswap_init); + +/* + * "Put" data from a page to frontswap and associate it with the page's + * swaptype and offset. Page must be locked and in the swap cache. + * If frontswap already contains a page with matching swaptype and + * offset, the frontswap implmentation may either overwrite the data and + * return success or invalidate the page from frontswap and return failure. + */ +int __frontswap_put_page(struct page *page) +{ + int ret = -1, dup = 0; + swp_entry_t entry = { .val = page_private(page), }; + int type = swp_type(entry); + struct swap_info_struct *sis = swap_info[type]; + pgoff_t offset = swp_offset(entry); + + BUG_ON(!PageLocked(page)); + BUG_ON(sis == NULL); + if (frontswap_test(sis, offset)) + dup = 1; + ret = (*frontswap_ops.put_page)(type, offset, page); + if (ret == 0) { + frontswap_set(sis, offset); + inc_frontswap_succ_puts(); + if (!dup) + atomic_inc(&sis->frontswap_pages); + } else if (dup) { + /* + failed dup always results in automatic invalidate of + the (older) page from frontswap + */ + frontswap_clear(sis, offset); + atomic_dec(&sis->frontswap_pages); + inc_frontswap_failed_puts(); + } else + inc_frontswap_failed_puts(); + if (frontswap_writethrough_enabled) + /* report failure so swap also writes to swap device */ + ret = -1; + return ret; +} +EXPORT_SYMBOL(__frontswap_put_page); + +/* + * "Get" data from frontswap associated with swaptype and offset that were + * specified when the data was put to frontswap and use it to fill the + * specified page with data. Page must be locked and in the swap cache. + */ +int __frontswap_get_page(struct page *page) +{ + int ret = -1; + swp_entry_t entry = { .val = page_private(page), }; + int type = swp_type(entry); + struct swap_info_struct *sis = swap_info[type]; + pgoff_t offset = swp_offset(entry); + + BUG_ON(!PageLocked(page)); + BUG_ON(sis == NULL); + if (frontswap_test(sis, offset)) + ret = (*frontswap_ops.get_page)(type, offset, page); + if (ret == 0) + inc_frontswap_gets(); + return ret; +} +EXPORT_SYMBOL(__frontswap_get_page); + +/* + * Invalidate any data from frontswap associated with the specified swaptype + * and offset so that a subsequent "get" will fail. + */ +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); + atomic_dec(&sis->frontswap_pages); + frontswap_clear(sis, offset); + inc_frontswap_invalidates(); + } +} +EXPORT_SYMBOL(__frontswap_invalidate_page); + +/* + * Invalidate all data from frontswap associated with all offsets for the + * specified swaptype. + */ +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)); +} +EXPORT_SYMBOL(__frontswap_invalidate_area); + +/* + * Frontswap, like a true swap device, may unnecessarily retain pages + * under certain circumstances; "shrink" frontswap is essentially a + * "partial swapoff" and works by calling try_to_unuse to attempt to + * unuse enough frontswap pages to attempt to -- subject to memory + * constraints -- reduce the number of pages in frontswap to the + * number given in the parameter target_pages. + */ +void frontswap_shrink(unsigned long target_pages) +{ + struct swap_info_struct *si = NULL; + int si_frontswap_pages; + unsigned long total_pages = 0, total_pages_to_unuse; + unsigned long pages = 0, pages_to_unuse = 0; + int type; + bool locked = false; + + /* + * we don't want to hold swap_lock while doing a very + * lengthy try_to_unuse, but swap_list may change + * so restart scan from swap_list.head each time + */ + spin_lock(&swap_lock); + locked = true; + total_pages = 0; + for (type = swap_list.head; type >= 0; type = si->next) { + si = swap_info[type]; + total_pages += atomic_read(&si->frontswap_pages); + } + if (total_pages <= target_pages) + goto out; + total_pages_to_unuse = total_pages - target_pages; + for (type = swap_list.head; type >= 0; type = si->next) { + si = swap_info[type]; + si_frontswap_pages = atomic_read(&si->frontswap_pages); + if (total_pages_to_unuse < si_frontswap_pages) + pages = pages_to_unuse = total_pages_to_unuse; + else { + pages = si_frontswap_pages; + pages_to_unuse = 0; /* unuse all */ + } + /* ensure there is enough RAM to fetch pages from frontswap */ + if (security_vm_enough_memory_mm(current->mm, pages)) + continue; + vm_unacct_memory(pages); + break; + } + if (type < 0) + goto out; + locked = false; + spin_unlock(&swap_lock); + try_to_unuse(type, true, pages_to_unuse); +out: + if (locked) + spin_unlock(&swap_lock); + return; +} +EXPORT_SYMBOL(frontswap_shrink); + +/* + * Count and return the number of frontswap pages across all + * swap devices. This is exported so that backend drivers can + * determine current usage without reading debugfs. + */ +unsigned long frontswap_curr_pages(void) +{ + int type; + unsigned long totalpages = 0; + struct swap_info_struct *si = NULL; + + spin_lock(&swap_lock); + for (type = swap_list.head; type >= 0; type = si->next) { + si = swap_info[type]; + totalpages += atomic_read(&si->frontswap_pages); + } + spin_unlock(&swap_lock); + return totalpages; +} +EXPORT_SYMBOL(frontswap_curr_pages); + +static int __init init_frontswap(void) +{ +#ifdef CONFIG_DEBUG_FS + struct dentry *root = debugfs_create_dir("frontswap", NULL); + if (root == NULL) + return -ENXIO; + debugfs_create_u64("gets", S_IRUGO, root, &frontswap_gets); + debugfs_create_u64("succ_puts", S_IRUGO, root, &frontswap_succ_puts); + debugfs_create_u64("failed_puts", S_IRUGO, root, + &frontswap_failed_puts); + debugfs_create_u64("invalidates", S_IRUGO, + root, &frontswap_invalidates); +#endif + return 0; +} + +module_init(init_frontswap); -- cgit v1.2.3