mm/page_alloc: fix alloc_pages_bulk/set_page_owner panic on irq disabled
Subsystems (3)
PATCH 1/1 mm/page_alloc: fix alloc_pages_bulk/set_page_owner panic on irq disabled
The problem is caused by set_page_owner alloc memory to save stack with GFP_KERNEL in local_riq disabled.
So, we just can't assume that alloc flags should be same with new page, let's split it. But in most situation, same is ok, in alloc_pages_bulk, input GFP_ATOMIC when prep_new_pages
--- include/linux/page_owner.h | 8 ++++---- mm/compaction.c | 2 +- mm/internal.h | 2 +- mm/page_alloc.c | 21 +++++++++++---------- mm/page_owner.c | 6 +++--- 5 files changed, 20 insertions(+), 19 deletions(-)
include/linux/page_owner.h +4 -4
@@ -10,7 +10,7 @@ extern struct page_ext_operations page_owner_ops;extern void __reset_page_owner (struct page *page,unsigned int order);extern void __set_page_owner (struct page *page, -unsigned int order,gfp_t gfp_mask); +unsigned int order,gfp_t alloc_gfp,gfp_t page_gfp_mask);extern void __split_page_owner (struct page *page,unsigned int nr);extern void __copy_page_owner (struct page *oldpage,struct page *newpage);extern void __set_page_owner_migrate_reason (struct page *page,int reason);
@@ -25,10 +25,10 @@ static inline void reset_page_owner(struct page *page, unsigned int order) }static inline void set_page_owner (struct page *page, -unsigned int order,gfp_t gfp_mask) +unsigned int order,gfp_t alloc_gfp,gfp_t page_gfp_mask) {if (static_branch_unlikely (&page_owner_inited)) -__set_page_owner (page, order, gfp_mask); +__set_page_owner (page, order, alloc_gfp, page_gfp_mask); }static inline void split_page_owner (struct page *page,unsigned int nr)
@@ -56,7 +56,7 @@ static inline void reset_page_owner(struct page *page, unsigned int order) { }static inline void set_page_owner (struct page *page, -unsigned int order,gfp_t gfp_mask) +unsigned int order,gfp_t alloc_gfp,gfp_t page_gfp_mask) { } static inline void split_page_owner(struct page *page,
mm/compaction.c +1 -1
@@ -96,7 +96,7 @@ static void split_map_pages(struct list_head *list) order =page_private (page); nr_pages =1 << order; -post_alloc_hook (page, order, __GFP_MOVABLE); +post_alloc_hook (page, order, __GFP_MOVABLE, __GFP_MOVABLE);if (order)split_page (page, order);
mm/internal.h +1 -1
@@ -195,7 +195,7 @@ extern void memblock_free_pages(struct page *page, unsigned long pfn,extern void __free_pages_core (struct page *page,unsigned int order);extern void prep_compound_page (struct page *page,unsigned int order);extern void post_alloc_hook (struct page *page,unsigned int order, -gfp_t gfp_flags); +gfp_t alloc_gfp,gfp_t page_gfp_mask);extern int user_min_free_kbytes;extern void free_unref_page (struct page *page);
mm/page_alloc.c +11 -10
@@ -2322,7 +2322,7 @@ static bool check_new_pages(struct page *page, unsigned int order) }inline void post_alloc_hook (struct page *page,unsigned int order, -gfp_t gfp_flags) +gfp_t alloc_gfp,gfp_t page_gfp_mask) {bool init;
@@ -2344,20 +2344,21 @@ inline void post_alloc_hook(struct page *page, unsigned int order,* kasan_alloc_pages and kernel_init_free_pages must be * kept together to avoid discrepancies in behavior. */ - init = !want_init_on_free () &&want_init_on_alloc (gfp_flags); + init = !want_init_on_free () &&want_init_on_alloc (page_gfp_mask);kasan_alloc_pages (page, order, init);if (init && !kasan_has_integrated_init ())kernel_init_free_pages (page,1 << order); -set_page_owner (page, order, gfp_flags); +set_page_owner (page, order, alloc_gfp, page_gfp_mask); } -static void prep_new_page (struct page *page,unsigned int order,gfp_t gfp_flags, -unsigned int alloc_flags) +static void prep_new_page (struct page *page,unsigned int order, +gfp_t alloc_gfp,gfp_t page_gfp_mask, +unsigned int alloc_flags) { -post_alloc_hook (page, order, gfp_flags); +post_alloc_hook (page, order, alloc_gfp, page_gfp_mask); -if (order && (gfp_flags & __GFP_COMP)) +if (order && (page_gfp_mask & __GFP_COMP))prep_compound_page (page, order);/*
@@ -3991,7 +3992,7 @@ get_page_from_freelist(gfp_t gfp_mask, unsigned int order, int alloc_flags, page =rmqueue (ac->preferred_zoneref->zone, zone, order, gfp_mask, alloc_flags, ac->migratetype);if (page) { -prep_new_page (page, order, gfp_mask, alloc_flags); +prep_new_page (page, order, gfp_mask, gfp_mask, alloc_flags);/* * If this is a high-order atomic allocation then check
@@ -4211,7 +4212,7 @@ __alloc_pages_direct_compact(gfp_t gfp_mask, unsigned int order,/* Prep a captured page if available */ if (page) -prep_new_page (page, order, gfp_mask, alloc_flags); +prep_new_page (page, order, gfp_mask, gfp_mask, alloc_flags);/* Try get a page from the freelist if available */ if (!page)
@@ -5127,7 +5128,7 @@ unsigned long __alloc_pages_bulk(gfp_t gfp, int preferred_nid,__count_zid_vm_events (PGALLOC ,zone_idx (zone),1 );zone_statistics (ac.preferred_zoneref->zone, zone); -prep_new_page (page,0 , gfp,0 ); +prep_new_page (page,0 ,GFP_ATOMIC , gfp,0 );if (page_list)list_add (&page->lru, page_list);else
mm/page_owner.c +3 -3
@@ -170,7 +170,7 @@ static inline void __set_page_owner_handle(struct page_ext *page_ext, } noinline void __set_page_owner(struct page *page, unsigned int order, -gfp_t gfp_mask) +gfp_t alloc_gfp,gfp_t page_gfp_mask) {struct page_ext *page_ext =lookup_page_ext (page);depot_stack_handle_t handle;
@@ -178,8 +178,8 @@ noinline void __set_page_owner(struct page *page, unsigned int order,if (unlikely (!page_ext))return ; - handle =save_stack (gfp_mask); -__set_page_owner_handle (page_ext, handle, order, gfp_mask); + handle =save_stack (alloc_gfp); +__set_page_owner_handle (page_ext, handle, order, page_gfp_mask); }void __set_page_owner_migrate_reason (struct page *page,int reason)
Signature · 1 line
2.32.0
Andrew Morton
maintainerQuoted 5 lines · On Fri, 9 Jul 2021 17:05:59 +0800 Yang Huan <[email protected]> wrote:
The problem is caused by
set_page_owneralloc memory to save stack withGFP_KERNELinlocal_riqdisabled.
So, we just can't assume that alloc flags should be same with new page, let's split it. But in most situation, same is ok, inalloc_pages_bulk, inputGFP_ATOMICwhenprep_new_pages
Please more fully describe the bug which is being fixed. A link to the sysbot report would be helpful. I assume there was a stack backtrace available? Seeing the will help others to understand the bug.
And if you are able to identify the patch which introduced the bug, a
Fixes: tag would be helpful as well.
Thanks.
Quoted 11 lines
Quoted 7 lines
The problem is caused by
set_page_owneralloc memory to save stack withGFP_KERNELinlocal_riqdisabled.
So, we just can't assume that alloc flags should be same with new page, let's split it. But in most situation, same is ok, inalloc_pages_bulk, inputGFP_ATOMICwhenprep_new_pagesPlease more fully describe the bug which is being fixed. A link to the sysbot report would be helpful. I assume there was a stack backtrace available? Seeing the will help others to understand the bug.
Sorry, here is the back trace
BUG: sleeping function called from invalid context at mm/page_alloc.c:5179 in_atomic(): 0, irqs_disabled(): 1, non_block: 0, pid: 1, name: swapper/0 __dump_stack lib/dump_stack.c:79 [inline] dump_stack_lvl+0xcd/0x134 lib/dump_stack.c:96 ___might_sleep.cold+0x1f1/0x237 kernel/sched/core.c:9153 prepare_alloc_pages+0x3da/0x580 mm/page_alloc.c:5179 __alloc_pages+0x12f/0x500 mm/page_alloc.c:5375 alloc_pages+0x18c/0x2a0 mm/mempolicy.c:2272 stack_depot_save+0x39d/0x4e0 lib/stackdepot.c:303 save_stack+0x15e/0x1e0 mm/page_owner.c:120 __set_page_owner+0x50/0x290 mm/page_owner.c:181 prep_new_page mm/page_alloc.c:2445 [inline] __alloc_pages_bulk+0x8b9/0x1870 mm/page_alloc.c:5313
Quoted 2 lines
And if you are able to identify the patch which introduced the bug, a
Fixes: tag would be helpful as well.
I will send patch again with this descrption.
Thanks
Quoted 3 lines
Thanks.