Re: [PATCH v4 3/4] arch, mm: restore dependency of __kernel_map_pages() of DEBUG_PAGEALLOC

From: Mike Rapoport
Date: Thu Nov 05 2020 - 06:42:26 EST


On Wed, Nov 04, 2020 at 07:02:20PM +0100, Vlastimil Babka wrote:
> On 11/3/20 5:20 PM, Mike Rapoport wrote:
> > From: Mike Rapoport <rppt@xxxxxxxxxxxxx>
>
> Subject should have "on DEBUG_PAGEALLOC" ?
>
> > The design of DEBUG_PAGEALLOC presumes that __kernel_map_pages() must never
> > fail. With this assumption is wouldn't be safe to allow general usage of
> > this function.
> >
> > Moreover, some architectures that implement __kernel_map_pages() have this
> > function guarded by #ifdef DEBUG_PAGEALLOC and some refuse to map/unmap
> > pages when page allocation debugging is disabled at runtime.
> >
> > As all the users of __kernel_map_pages() were converted to use
> > debug_pagealloc_map_pages() it is safe to make it available only when
> > DEBUG_PAGEALLOC is set.
> >
> > Signed-off-by: Mike Rapoport <rppt@xxxxxxxxxxxxx>
> > Acked-by: David Hildenbrand <david@xxxxxxxxxx>
> > Acked-by: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx>
> > ---
> > arch/Kconfig | 3 +++
> > arch/arm64/Kconfig | 4 +---
> > arch/arm64/mm/pageattr.c | 8 ++++++--
> > arch/powerpc/Kconfig | 5 +----
> > arch/riscv/Kconfig | 4 +---
> > arch/riscv/include/asm/pgtable.h | 2 --
> > arch/riscv/mm/pageattr.c | 2 ++
> > arch/s390/Kconfig | 4 +---
> > arch/sparc/Kconfig | 4 +---
> > arch/x86/Kconfig | 4 +---
> > arch/x86/mm/pat/set_memory.c | 2 ++
> > include/linux/mm.h | 10 +++++++---
> > 12 files changed, 26 insertions(+), 26 deletions(-)
> >
> > diff --git a/arch/Kconfig b/arch/Kconfig
> > index 56b6ccc0e32d..56d4752b6db6 100644
> > --- a/arch/Kconfig
> > +++ b/arch/Kconfig
> > @@ -1028,6 +1028,9 @@ config HAVE_STATIC_CALL_INLINE
> > bool
> > depends on HAVE_STATIC_CALL
> > +config ARCH_SUPPORTS_DEBUG_PAGEALLOC
> > + bool
> > +
> > source "kernel/gcov/Kconfig"
> > source "scripts/gcc-plugins/Kconfig"
> > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> > index 1d466addb078..a932810cfd90 100644
> > --- a/arch/arm64/Kconfig
> > +++ b/arch/arm64/Kconfig
> > @@ -71,6 +71,7 @@ config ARM64
> > select ARCH_USE_QUEUED_RWLOCKS
> > select ARCH_USE_QUEUED_SPINLOCKS
> > select ARCH_USE_SYM_ANNOTATIONS
> > + select ARCH_SUPPORTS_DEBUG_PAGEALLOC
> > select ARCH_SUPPORTS_MEMORY_FAILURE
> > select ARCH_SUPPORTS_SHADOW_CALL_STACK if CC_HAVE_SHADOW_CALL_STACK
> > select ARCH_SUPPORTS_ATOMIC_RMW
> > @@ -1025,9 +1026,6 @@ config HOLES_IN_ZONE
> > source "kernel/Kconfig.hz"
> > -config ARCH_SUPPORTS_DEBUG_PAGEALLOC
> > - def_bool y
> > -
> > config ARCH_SPARSEMEM_ENABLE
> > def_bool y
> > select SPARSEMEM_VMEMMAP_ENABLE
> > diff --git a/arch/arm64/mm/pageattr.c b/arch/arm64/mm/pageattr.c
> > index 1b94f5b82654..439325532be1 100644
> > --- a/arch/arm64/mm/pageattr.c
> > +++ b/arch/arm64/mm/pageattr.c
> > @@ -155,7 +155,7 @@ int set_direct_map_invalid_noflush(struct page *page)
> > .clear_mask = __pgprot(PTE_VALID),
> > };
> > - if (!rodata_full)
> > + if (!debug_pagealloc_enabled() && !rodata_full)
> > return 0;
> > return apply_to_page_range(&init_mm,
> > @@ -170,7 +170,7 @@ int set_direct_map_default_noflush(struct page *page)
> > .clear_mask = __pgprot(PTE_RDONLY),
> > };
> > - if (!rodata_full)
> > + if (!debug_pagealloc_enabled() && !rodata_full)
> > return 0;
> > return apply_to_page_range(&init_mm,
>
> I don't understand these two hunks. Previous patch calls this for
> hibernation when CONFIG_ARCH_HAS_SET_DIRECT_MAP, which is true for arm64.
> Why suddenly this starts to depend on debug_pagealloc_enabled()?

I was confused about this for quite a long :)

On arm64 the changes to direct^w linear map are allowed when

debug_page_alloc() || rodata_full

In hibernation we essentially have now

if (1)
set_direct_map(something)
else
debug_page_alloc_map()

With debug_pagealloc enabled but with rodata_full disabled arm64
versions of set_direct_map_*() will become a nop, so a page that was
unmapped by debug_pagealloc() will not be mapped back.

I'm still puzzled how hibernation might ever need to save a free page,
but that's another story.

--
Sincerely yours,
Mike.