2024-04-19 04:54:22

by Hariom Panthi

[permalink] [raw]
Subject: [PATCH 1/1] mm: vmalloc: dump page owner info if page is already mapped

In vmap_pte_range, BUG_ON is called when page is already mapped,
It doesn't give enough information to debug further.
Dumping page owner information alongwith BUG_ON will be more useful
in case of multiple page mapping.

Example:
[ 18.258806] page_owner tracks the page as allocated
[ 18.258970] page last allocated via order 0, migratetype Unmovable, gfp_mask 0xcc0(GFP_KERNEL), pid 81, tgid 81 (insmod), ts 18257811008, free_ts 0
[ 18.259327] prep_new_page+0xa8/0x10c
[ 18.259455] get_page_from_freelist+0x7f8/0x1248
[ 18.259591] __alloc_pages+0x164/0x2b4
[ 18.259703] alloc_pages_mpol+0x88/0x230
[ 18.259821] alloc_pages+0x4c/0x7c
[ 18.259923] load_module+0x74/0x1af4
[ 18.260031] __do_sys_init_module+0x190/0x1fc
[ 18.260154] __arm64_sys_init_module+0x1c/0x28
[ 18.260280] invoke_syscall+0x44/0x108
[ 18.260394] el0_svc_common.constprop.0+0x40/0xe0
[ 18.260530] do_el0_svc_compat+0x1c/0x34
[ 18.260647] el0_svc_compat+0x2c/0x80
[ 18.260761] el0t_32_sync_handler+0x90/0x140
[ 18.260887] el0t_32_sync+0x194/0x198
[ 18.260996] page_owner free stack trace missing
[ 18.261143] ------------[ cut here ]------------
[ 18.261272] kernel BUG at mm/vmalloc.c:113!

Signed-off-by: Hariom Panthi <[email protected]>
---
mm/vmalloc.c | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 68fa001648cc..9c91091887de 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -42,6 +42,7 @@
#include <linux/sched/mm.h>
#include <asm/tlbflush.h>
#include <asm/shmparam.h>
+#include <linux/page_owner.h>

#define CREATE_TRACE_POINTS
#include <trace/events/vmalloc.h>
@@ -96,6 +97,7 @@ static int vmap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
{
pte_t *pte;
u64 pfn;
+ struct page *page;
unsigned long size = PAGE_SIZE;

pfn = phys_addr >> PAGE_SHIFT;
@@ -103,7 +105,13 @@ static int vmap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
if (!pte)
return -ENOMEM;
do {
- BUG_ON(!pte_none(ptep_get(pte)));
+ if (!pte_none(ptep_get(pte))) {
+ if (pfn_valid(pfn)) {
+ page = pfn_to_page(pfn);
+ dump_page_owner(page);
+ }
+ BUG();
+ }

#ifdef CONFIG_HUGETLB_PAGE
size = arch_vmap_pte_range_map_size(addr, end, pfn, max_page_shift);
--
2.25.1



2024-04-22 21:28:48

by Andrew Morton

[permalink] [raw]
Subject: Re: [PATCH 1/1] mm: vmalloc: dump page owner info if page is already mapped

On Fri, 19 Apr 2024 10:09:10 +0530 Hariom Panthi <[email protected]> wrote:

> In vmap_pte_range, BUG_ON is called when page is already mapped,
> It doesn't give enough information to debug further.
> Dumping page owner information alongwith BUG_ON will be more useful
> in case of multiple page mapping.
>
> Example:
> [ 18.258806] page_owner tracks the page as allocated
> [ 18.258970] page last allocated via order 0, migratetype Unmovable, gfp_mask 0xcc0(GFP_KERNEL), pid 81, tgid 81 (insmod), ts 18257811008, free_ts 0
> [ 18.259327] prep_new_page+0xa8/0x10c
> [ 18.259455] get_page_from_freelist+0x7f8/0x1248
> [ 18.259591] __alloc_pages+0x164/0x2b4
> [ 18.259703] alloc_pages_mpol+0x88/0x230
> [ 18.259821] alloc_pages+0x4c/0x7c
> [ 18.259923] load_module+0x74/0x1af4
> [ 18.260031] __do_sys_init_module+0x190/0x1fc
> [ 18.260154] __arm64_sys_init_module+0x1c/0x28
> [ 18.260280] invoke_syscall+0x44/0x108
> [ 18.260394] el0_svc_common.constprop.0+0x40/0xe0
> [ 18.260530] do_el0_svc_compat+0x1c/0x34
> [ 18.260647] el0_svc_compat+0x2c/0x80
> [ 18.260761] el0t_32_sync_handler+0x90/0x140
> [ 18.260887] el0t_32_sync+0x194/0x198
> [ 18.260996] page_owner free stack trace missing
> [ 18.261143] ------------[ cut here ]------------
> [ 18.261272] kernel BUG at mm/vmalloc.c:113!
>
> ...
>
> @@ -103,7 +105,13 @@ static int vmap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
> if (!pte)
> return -ENOMEM;
> do {
> - BUG_ON(!pte_none(ptep_get(pte)));
> + if (!pte_none(ptep_get(pte))) {
> + if (pfn_valid(pfn)) {
> + page = pfn_to_page(pfn);
> + dump_page_owner(page);
> + }
> + BUG();
> + }

Diving straight into dump_page_owner() seems inappropriate. The
higher-level dump_page() interface is more typically used.

Or, even more common, VM_BUG_ON_PAGE(), but that doesn't look to be a
good fit here.



2024-04-23 03:50:28

by Hariom Panthi

[permalink] [raw]
Subject: Re: [PATCH 1/1] mm: vmalloc: dump page owner info if page is already mapped

Hi,

> > ...
> >
> > @@ -103,7 +105,13 @@ static int vmap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
> >         if (!pte)
> >                 return -ENOMEM;
> >         do {
> > -                BUG_ON(!pte_none(ptep_get(pte)));
> > +                if (!pte_none(ptep_get(pte))) {
> > +                        if (pfn_valid(pfn)) {
> > +                                page = pfn_to_page(pfn);
> > +                                dump_page_owner(page);
> > +                        }
> > +                        BUG();
> > +                }
>
> Diving straight into dump_page_owner() seems inappropriate.  The
> higher-level dump_page() interface is more typically used.
>
> Or, even more common, VM_BUG_ON_PAGE(), but that doesn't look to be a
> good fit here.

Ok I will send V2 with dump_page API.

Thanks,
Hariom