swsusp improvements
Subsystems (2)
cover swsusp improvements
Hi,
The following patches improve swsusp and/or prepare it for further cleanups and improvements.
The patches have been tested quite thoroughly and hopefully they can be considered as 2.6.15 material. Please apply.
All three patches have already been ACKed by Pavel.
Greetings,
Rafael
PATCH 1/3 swsusp: rework image freeing
The following patch makes swsusp use PG_nosave and PG_nosave_free flags to mark pages that should be freed after the state of the system has been restored from the image (or in case of an error during suspend).
This allows us to avoid storing metadata in swap twice and to reduce the amount of memory needed by swsusp. Additionally, it allows us to simplify the code by removing a couple of functions that are no longer necessary.
kernel/power/swsusp.c +75 -84
@@ -701,24 +701,28 @@ } } -static void copy_data_pages (void ) {struct zone *zone;unsigned long zone_pfn; -struct pbe * pbe = pagedir_nosave; +struct pbe *pbe = pagedir_nosave, *p;pr_debug ("copy_data_pages(): pages to copy: %d\n" , nr_copy_pages);for_each_zone (zone) {if (is_highmem (zone))continue ;mark_free_pages (zone); +/* This is necessary for swsusp_free() */ + for_each_pb_page (p, pagedir_nosave) +SetPageNosaveFree (virt_to_page (p)); +for_each_pbe (p, pagedir_nosave) +SetPageNosaveFree (virt_to_page (p->address));for (zone_pfn =0 ; zone_pfn < zone->spanned_pages; ++zone_pfn) {if (saveable (zone, &zone_pfn)) {struct page * page; page =pfn_to_page (zone_pfn + zone->zone_start_pfn);BUG_ON (!pbe); - pbe->orig_address = (long )page_address (page); + pbe->orig_address = (unsigned long )page_address (page);/* copy_page is not usable for copying task structs. */ memcpy ((void *)pbe->address, (void *)pbe->orig_address,PAGE_SIZE ); pbe = pbe->next;
@@ -730,15 +734,6 @@/** -* calc_nr - Determine the number of pages needed for a pbe list. -*/ - -static int calc_nr (int nr_copy) -{ -return nr_copy + (nr_copy+PBES_PER_PAGE -2 )/(PBES_PER_PAGE -1 ); -} - -/** * free_pagedir - free pages allocated with alloc_pagedir() */
@@ -748,6 +743,8 @@while (pblist) { pbe = (pblist +PB_PAGE_SKIP )->next; +ClearPageNosave (virt_to_page (pblist)); +ClearPageNosaveFree (virt_to_page (pblist));free_page ((unsigned long )pblist); pblist = pbe; }
@@ -793,6 +790,16 @@pr_debug ("create_pbe_list(): initialized %d PBEs\n" , num); } +static void *alloc_image_page (void ) +{ +void *res = (void *)get_zeroed_page (GFP_ATOMIC | __GFP_COLD); +if (res) { +SetPageNosave (virt_to_page (res)); +SetPageNosaveFree (virt_to_page (res)); + } +return res; +} +/** * alloc_pagedir - Allocate the page directory. *
@@ -815,11 +822,11 @@return NULL ;pr_debug ("alloc_pagedir(): nr_pages = %d\n" , nr_pages); - pblist = (struct pbe *)get_zeroed_page (GFP_ATOMIC | __GFP_COLD); + pblist = (struct pbe *)alloc_image_page ();for (pbe = pblist, num =PBES_PER_PAGE ; pbe && num < nr_pages; pbe = pbe->next, num +=PBES_PER_PAGE ) { pbe +=PB_PAGE_SKIP ; - pbe->next = (struct pbe *)get_zeroed_page (GFP_ATOMIC | __GFP_COLD); + pbe->next = (struct pbe *)alloc_image_page (); }if (!pbe) {/* get_zeroed_page() failed */ free_pagedir (pblist);
@@ -829,52 +836,30 @@ }/** -* free_image_pages - Free pages allocated for snapshot +* Free pages we allocated for suspend. Suspend pages are alocated +* before atomic copy, so we need to free them after resume. */ -static void free_image_pages (void ) +void swsusp_free (void ) { -struct pbe * p; +struct zone *zone; +unsigned long zone_pfn; - for_each_pbe (p, pagedir_save) { -if (p->address) { -ClearPageNosave (virt_to_page (p->address)); -free_page (p->address); - p->address =0 ; - } +for_each_zone (zone) { +for (zone_pfn =0 ; zone_pfn < zone->spanned_pages; ++zone_pfn) +if (pfn_valid (zone_pfn + zone->zone_start_pfn)) { +struct page * page; + page =pfn_to_page (zone_pfn + zone->zone_start_pfn); +if (PageNosave (page) &&PageNosaveFree (page)) { +ClearPageNosave (page); +ClearPageNosaveFree (page); +free_page ((long )page_address (page)); + } + } } }/** -* alloc_image_pages - Allocate pages for the snapshot. -*/ - -static int alloc_image_pages (void ) -{ -struct pbe * p; - - for_each_pbe (p, pagedir_save) { - p->address =get_zeroed_page (GFP_ATOMIC | __GFP_COLD); -if (!p->address) -return -ENOMEM ; -SetPageNosave (virt_to_page (p->address)); - } -return 0 ; -} - -/* Free pages we allocated for suspend. Suspend pages are alocated -* before atomic copy, so we need to free them after resume. -*/ -void swsusp_free (void ) -{ -BUG_ON (PageNosave (virt_to_page (pagedir_save))); -BUG_ON (PageNosaveFree (virt_to_page (pagedir_save))); -free_image_pages (); -free_pagedir (pagedir_save); -} - - -/** * enough_free_mem - Make sure we enough free memory to snapshot. * * Returns TRUE or FALSE after checking the number of available
@@ -883,12 +868,9 @@static int enough_free_mem (void ) { -if (nr_free_pages () < (nr_copy_pages +PAGES_FOR_IO )) { -pr_debug ("swsusp: Not enough free pages: Have %d\n" , -nr_free_pages ()); -return 0 ; - } -return 1 ; +pr_debug ("swsusp: available memory: %u pages\n" ,nr_free_pages ()); +return nr_free_pages () > (nr_copy_pages +PAGES_FOR_IO + + nr_copy_pages/PBES_PER_PAGE + !!(nr_copy_pages%PBES_PER_PAGE )); }
@@ -907,33 +889,16 @@struct sysinfo i;si_swapinfo (&i); -if (i.freeswap < (nr_copy_pages +PAGES_FOR_IO )) { -pr_debug ("swsusp: Not enough swap. Need %ld\n" ,i.freeswap); -return 0 ; - } -return 1 ; +pr_debug ("swsusp: available swap: %lu pages\n" , i.freeswap); +return i.freeswap > (nr_copy_pages +PAGES_FOR_IO + + nr_copy_pages/PBES_PER_PAGE + !!(nr_copy_pages%PBES_PER_PAGE )); }static int swsusp_alloc (void ) { -int error; +struct pbe * p; pagedir_nosave =NULL ; - nr_copy_pages =calc_nr (nr_copy_pages); - nr_copy_pages_check = nr_copy_pages; - -pr_debug ("suspend: (pages needed: %d + %d free: %d)\n" , - nr_copy_pages,PAGES_FOR_IO ,nr_free_pages ()); - -if (!enough_free_mem ()) -return -ENOMEM ; - -if (!enough_swap ()) -return -ENOSPC ; - -if (MAX_PBES < nr_copy_pages /PBES_PER_PAGE + - !!(nr_copy_pages %PBES_PER_PAGE )) -return -ENOSPC ;if (!(pagedir_save =alloc_pagedir (nr_copy_pages))) {printk (KERN_ERR "suspend: Allocating pagedir failed.\n" );
@@ -941,10 +906,14 @@ }create_pbe_list (pagedir_save, nr_copy_pages); pagedir_nosave = pagedir_save; -if ((error =alloc_image_pages ())) { -printk (KERN_ERR "suspend: Allocating image pages failed.\n" ); -swsusp_free (); -return error; + + for_each_pbe (p, pagedir_save) { + p->address = (unsigned long )alloc_image_page (); +if (!p->address) { +printk (KERN_ERR "suspend: Allocating image pages failed.\n" ); +swsusp_free (); +return -ENOMEM ; + } }return 0 ;
@@ -956,7 +925,7 @@pr_debug ("swsusp: critical section: \n" );if (save_highmem ()) { -printk (KERN_CRIT "Suspend machine: Not enough free pages for highmem\n" ); +printk (KERN_CRIT "swsusp: Not enough free pages for highmem\n" );restore_highmem ();return -ENOMEM ; }
@@ -964,6 +933,28 @@drain_local_pages ();count_data_pages ();printk ("swsusp: Need to copy %u pages\n" , nr_copy_pages); + nr_copy_pages_check = nr_copy_pages; + +pr_debug ("swsusp: pages needed: %u + %lu + %u, free: %u\n" , + nr_copy_pages, + nr_copy_pages/PBES_PER_PAGE + !!(nr_copy_pages%PBES_PER_PAGE ), +PAGES_FOR_IO ,nr_free_pages ()); + +if (!enough_free_mem ()) { +printk (KERN_ERR "swsusp: Not enough free memory\n" ); +return -ENOMEM ; + } + +if (MAX_PBES < nr_copy_pages /PBES_PER_PAGE + + !!(nr_copy_pages %PBES_PER_PAGE )) { +printk (KERN_ERR "swsusp: Too many image pages\n" ); +return -ENOSPC ; + } + +if (!enough_swap ()) { +printk (KERN_ERR "swsusp: Not enough free swap\n" ); +return -ENOSPC ; + } error =swsusp_alloc ();if (error)
PATCH 2/3 swsusp: move snapshot functionality to separate file
In reply to Rafael J. WysockiThe following patch moves the functionality of swsusp related to creating and handling the snapshot of memory to a separate file, snapshot.c
This should enable us to untangle the code in the future and eventually to implement some parts of swsusp.c in the user space.
The patch does not change the code.
include/linux/suspend.h +6 -0
@@ -74,4 +74,10 @@extern unsigned long get_usable_page (unsigned gfp_mask);extern void free_eaten_memory (void ); +/* +* XXX: We try to keep some more pages free so that I/O operations succeed +* without paging. Might this be more? +*/ +#define PAGES_FOR_IO 512 +#endif /* _LINUX_SWSUSP_H */
kernel/power/Makefile +1 -1
@@ -4,7 +4,7 @@ endif obj-y := main.o process.o console.o pm.o -obj-$(CONFIG_SOFTWARE_SUSPEND) += swsusp.o disk.o +obj-$(CONFIG_SOFTWARE_SUSPEND) += swsusp.o disk.o snapshot.o obj-$(CONFIG_SUSPEND_SMP) += smp.o
kernel/power/power.h +17 -0
@@ -53,3 +53,20 @@extern int pm_prepare_console (void );extern void pm_restore_console (void ); + + +/* References to section boundaries */ +extern const void __nosave_begin, __nosave_end; + +extern unsigned int nr_copy_pages; +extern suspend_pagedir_t *pagedir_nosave; +extern suspend_pagedir_t *pagedir_save; + +extern asmlinkage int swsusp_arch_suspend(void); +extern asmlinkage int swsusp_arch_resume(void); + +extern int restore_highmem (void ); +extern void free_pagedir (struct pbe *pblist); +extern struct pbe *alloc_pagedir (unsigned nr_pages); +extern void create_pbe_list (struct pbe *pblist,unsigned nr_pages); +extern int enough_swap (void );
kernel/power/snapshot.c new file +464 -0
464 lines not on this page · raw message
kernel/power/swsusp.c +4 -428
489 lines not on this page · raw message
PATCH 3/3 swsusp: rework memory freeing on resume
In reply to Rafael J. WysockiThe following patch makes swsusp use the PG_nosave and PG_nosave_free flags to mark pages that should be freed in case of an error during resume.
This allows us to simplify the code and to use swsusp_free() in all of the swsusp's resume error paths, which makes them actually work.
kernel/power/swsusp.c +36 -74
@@ -630,6 +630,11 @@* execution continues at place where swsusp_arch_suspend was called */ BUG_ON (!error); +/* The only reason why swsusp_arch_resume() can fail is memory being +* very tight, so we have to free it as soon as we can to avoid +* subsequent failures +*/ +swsusp_free ();restore_processor_state ();restore_highmem ();touch_softlockup_watchdog ();
@@ -645,54 +650,28 @@* * We don't know which pages are usable until we allocate them. * -* Allocated but unusable (ie eaten) memory pages are linked together -* to create a list, so that we can free them easily -* -* We could have used a type other than (void *) -* for this purpose, but ... +* Allocated but unusable (ie eaten) memory pages are marked so that +* swsusp_free() can release them */ -static void **eaten_memory =NULL ; - -static inline void eat_page (void *page) -{ -void **c; - - c = eaten_memory; - eaten_memory = page; - *eaten_memory = c; -} -unsigned long get_usable_page (unsigned gfp_mask) +unsigned long get_safe_page (unsigned gfp_mask) {unsigned long m; - m =get_zeroed_page (gfp_mask); -while (!PageNosaveFree (virt_to_page (m))) { -eat_page ((void *)m); +do { m =get_zeroed_page (gfp_mask); -if (!m) -break ; +if (m &&PageNosaveFree (virt_to_page (m))) +/* This is for swsusp_free() */ +SetPageNosave (virt_to_page (m)); + }while (m &&PageNosaveFree (virt_to_page (m))); +if (m) { +/* This is for swsusp_free() */ +SetPageNosave (virt_to_page (m)); +SetPageNosaveFree (virt_to_page (m)); }return m; } -void free_eaten_memory (void ) -{ -unsigned long m; -void **c; -int i =0 ; - - c = eaten_memory; -while (c) { - m = (unsigned long )c; - c = *c; -free_page (m); - i++; - } - eaten_memory =NULL ; -pr_debug ("swsusp: %d unused pages freed\n" , i); -} -/** * check_pagedir - We ensure here that pages that the PBEs point to * won't collide with pages where we're going to restore from the loaded
@@ -710,7 +689,7 @@ p->address =0UL ; for_each_pbe (p, pblist) { - p->address =get_usable_page (GFP_ATOMIC ); + p->address =get_safe_page (GFP_ATOMIC );if (!p->address)return -ENOMEM ; }
@@ -729,7 +708,7 @@unsigned long zone_pfn;struct pbe *pbpage, *tail, *p;void *m; -int rel =0 , error =0 ; +int rel =0 ;if (!pblist)/* a sanity check */ return NULL ;
@@ -737,41 +716,37 @@pr_debug ("swsusp: Relocating pagedir (%lu pages to check)\n" , swsusp_info.pagedir_pages); -/* Set page flags */ +/* Clear page flags */ for_each_zone (zone) {for (zone_pfn =0 ; zone_pfn < zone->spanned_pages; ++zone_pfn) -SetPageNosaveFree (pfn_to_page (zone_pfn + +if (pfn_valid (zone_pfn + zone->zone_start_pfn)) +ClearPageNosaveFree (pfn_to_page (zone_pfn + zone->zone_start_pfn)); } -/* Clear orig addresses */ +/* Mark orig addresses */ for_each_pbe (p, pblist) -ClearPageNosaveFree (virt_to_page (p->orig_address)); +SetPageNosaveFree (virt_to_page (p->orig_address)); tail = pblist +PB_PAGE_SKIP ;/* Relocate colliding pages */ for_each_pb_page (pbpage, pblist) { -if (!PageNosaveFree (virt_to_page ((unsigned long )pbpage))) { - m = (void *)get_usable_page (GFP_ATOMIC | __GFP_COLD); -if (!m) { - error = -ENOMEM ; -break ; - } +if (PageNosaveFree (virt_to_page ((unsigned long )pbpage))) { + m = (void *)get_safe_page (GFP_ATOMIC | __GFP_COLD); +if (!m) +return NULL ;memcpy (m, (void *)pbpage,PAGE_SIZE );if (pbpage == pblist) pblist = (struct pbe *)m;else tail->next = (struct pbe *)m; - -eat_page ((void *)pbpage); pbpage = (struct pbe *)m;/* We have to link the PBEs again */ -for (p = pbpage; p < pbpage +PB_PAGE_SKIP ; p++)if (p->next)/* needed to save the end */ p->next = p +1 ;
@@ -781,15 +756,13 @@ tail = pbpage +PB_PAGE_SKIP ; } -if (error) { -printk ("\nswsusp: Out of memory\n\n" ); -free_pagedir (pblist); -free_eaten_memory (); - pblist =NULL ; -/* Is this even worth handling? It should never ever happen, and we -have just lost user's state, anyway... */ - }else -printk ("swsusp: Relocated %d pages\n" , rel); +/* This is for swsusp_free() */ + for_each_pb_page (pbpage, pblist) { +SetPageNosave (virt_to_page (pbpage)); +SetPageNosaveFree (virt_to_page (pbpage)); + } + +printk ("swsusp: Relocated %d pages\n" , rel);return pblist; }
@@ -1007,9 +980,7 @@break ; } -if (error) -free_pagedir (pblist); -else +if (!error)BUG_ON (i != swsusp_info.pagedir_pages);return error;
@@ -1052,15 +1023,6 @@if (!error) error =data_read (pagedir_nosave); -if (error) {/* We fail cleanly */ -free_eaten_memory (); - for_each_pbe (p, pagedir_nosave) -if (p->address) { -free_page (p->address); - p->address =0UL ; - } -free_pagedir (pagedir_nosave); - }return error; }
kernel/power/disk.c +7 -7
@@ -30,7 +30,6 @@extern int swsusp_read (void );extern void swsusp_close (void );extern int swsusp_resume (void ); -extern int swsusp_free (void );static int noresume =0 ;
@@ -252,14 +251,17 @@pr_debug ("PM: Reading swsusp image.\n" ); -if ((error =swsusp_read ())) - goto Cleanup; +if ((error =swsusp_read ())) { +swsusp_free (); + goto Thaw; + }pr_debug ("PM: Preparing devices for restore.\n" );if ((error =device_suspend (PMSG_FREEZE ))) {printk ("Some devices failed to suspend\n" ); - goto Free; +swsusp_free (); + goto Thaw; }mb ();
@@ -268,9 +270,7 @@swsusp_resume ();pr_debug ("PM: Restore failed, recovering.n" );device_resume (); - Free: -swsusp_free (); - Cleanup: + Thaw:unprepare_processes (); Done:/* For success case, the suspend path will release the lock */
kernel/power/power.h +1 -1
@@ -66,7 +66,7 @@ extern asmlinkage int swsusp_arch_resume(void);extern int restore_highmem (void ); -extern void free_pagedir (struct pbe *pblist);extern struct pbe *alloc_pagedir (unsigned nr_pages);extern void create_pbe_list (struct pbe *pblist,unsigned nr_pages); +extern void swsusp_free (void );extern int enough_swap (void );
kernel/power/snapshot.c +1 -1
@@ -239,7 +239,7 @@* free_pagedir - free pages allocated with alloc_pagedir() */ -void free_pagedir (struct pbe *pblist) +static void free_pagedir (struct pbe *pblist) {struct pbe *pbe;
arch/x86_64/kernel/suspend.c +19 -65
125 lines not on this page · raw message
include/linux/suspend.h +1 -2
@@ -71,8 +71,7 @@struct saved_context ;void __save_processor_state (struct saved_context *ctxt);void __restore_processor_state (struct saved_context *ctxt); -extern unsigned long get_usable_page (unsigned gfp_mask); -extern void free_eaten_memory (void ); +extern unsigned long get_safe_page (unsigned gfp_mask);/* * XXX: We try to keep some more pages free so that I/O operations succeed