2021-06-30 14:04:04

by Alexander Potapenko

Subject: [PATCH v3 1/2] kfence: move the size check to the beginning of __kfence_alloc()

Check the allocation size before toggling kfence_allocation_gate.
This way allocations that can't be served by KFENCE will not result in
waiting for another CONFIG_KFENCE_SAMPLE_INTERVAL without allocating
anything.

Suggested-by: Marco Elver <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: Marco Elver <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: [email protected] # 5.12+
Signed-off-by: Alexander Potapenko <[email protected]>
---
mm/kfence/core.c | 10 +++++++---
1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/mm/kfence/core.c b/mm/kfence/core.c
index 4d21ac44d5d35..33bb20d91bf6a 100644
--- a/mm/kfence/core.c
+++ b/mm/kfence/core.c
@@ -733,6 +733,13 @@ void kfence_shutdown_cache(struct kmem_cache *s)

void *__kfence_alloc(struct kmem_cache *s, size_t size, gfp_t flags)
{
+ /*
+ * Perform size check before switching kfence_allocation_gate, so that
+ * we don't disable KFENCE without making an allocation.
+ */
+ if (size > PAGE_SIZE)
+ return NULL;
+
/*
* allocation_gate only needs to become non-zero, so it doesn't make
* sense to continue writing to it and pay the associated contention
@@ -757,9 +764,6 @@ void *__kfence_alloc(struct kmem_cache *s, size_t size, gfp_t flags)
if (!READ_ONCE(kfence_enabled))
return NULL;

- if (size > PAGE_SIZE)
- return NULL;
-
return kfence_guarded_alloc(s, size, flags);
}

--
2.32.0.93.g670b81a890-goog

2021-06-30 14:04:05

by Alexander Potapenko

Subject: [PATCH v3 2/2] kfence: skip all GFP_ZONEMASK allocations

Allocation requests outside ZONE_NORMAL (MOVABLE, HIGHMEM or DNA) cannot
be fulfilled by KFENCE, because KFENCE memory pool is located in a
zone different from the requested one.

Because callers of kmem_cache_alloc() may actually rely on the
allocation to reside in the requested zone (e.g. memory allocations done
with __GFP_DMA must be DMAable), skip all allocations done with
GFP_ZONEMASK and/or respective SLAB flags (SLAB_CACHE_DMA and
SLAB_CACHE_DMA32).

Fixes: 0ce20dd84089 ("mm: add Kernel Electric-Fence infrastructure")
Cc: Andrew Morton <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: Marco Elver <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Souptick Joarder <[email protected]>
Cc: [email protected] # 5.12+
Signed-off-by: Alexander Potapenko <[email protected]>

---

v2:
- added parentheses around the GFP clause, as requested by Marco
v3:
- ignore GFP_ZONEMASK, which also covers __GFP_HIGHMEM and __GFP_MOVABLE
- move the flag check at the beginning of the function, as requested by
Souptick Joarder
---
mm/kfence/core.c | 8 ++++++++
1 file changed, 8 insertions(+)

diff --git a/mm/kfence/core.c b/mm/kfence/core.c
index 33bb20d91bf6a..d51f77329fd3c 100644
--- a/mm/kfence/core.c
+++ b/mm/kfence/core.c
@@ -740,6 +740,14 @@ void *__kfence_alloc(struct kmem_cache *s, size_t size, gfp_t flags)
if (size > PAGE_SIZE)
return NULL;

+ /*
+ * Skip allocations from non-default zones, including DMA. We cannot guarantee that pages
+ * in the KFENCE pool will have the requested properties (e.g. reside in DMAable memory).
+ */
+ if ((flags & GFP_ZONEMASK) ||
+ (s->flags & (SLAB_CACHE_DMA | SLAB_CACHE_DMA32)))
+ return NULL;
+
/*
* allocation_gate only needs to become non-zero, so it doesn't make
* sense to continue writing to it and pay the associated contention
--
2.32.0.93.g670b81a890-goog

2021-06-30 14:06:10

by Marco Elver

Subject: Re: [PATCH v3 2/2] kfence: skip all GFP_ZONEMASK allocations

On Wed, 30 Jun 2021 at 15:53, Alexander Potapenko <[email protected]> wrote:
> Allocation requests outside ZONE_NORMAL (MOVABLE, HIGHMEM or DNA) cannot

s/DNA/DMA/
... but probably no need to do v4 just for this (everyone knows we're
not yet in the business of allocating DNA ;-)).

> be fulfilled by KFENCE, because KFENCE memory pool is located in a
> zone different from the requested one.
>
> Because callers of kmem_cache_alloc() may actually rely on the
> allocation to reside in the requested zone (e.g. memory allocations done
> with __GFP_DMA must be DMAable), skip all allocations done with
> GFP_ZONEMASK and/or respective SLAB flags (SLAB_CACHE_DMA and
> SLAB_CACHE_DMA32).
>
> Fixes: 0ce20dd84089 ("mm: add Kernel Electric-Fence infrastructure")
> Cc: Andrew Morton <[email protected]>
> Cc: Dmitry Vyukov <[email protected]>
> Cc: Marco Elver <[email protected]>
> Cc: Greg Kroah-Hartman <[email protected]>
> Cc: Souptick Joarder <[email protected]>
> Cc: [email protected] # 5.12+
> Signed-off-by: Alexander Potapenko <[email protected]>

With the change below, you can add:

Reviewed-by: Marco Elver <[email protected]>


> ---
>
> v2:
> - added parentheses around the GFP clause, as requested by Marco
> v3:
> - ignore GFP_ZONEMASK, which also covers __GFP_HIGHMEM and __GFP_MOVABLE
> - move the flag check at the beginning of the function, as requested by
> Souptick Joarder
> ---
> mm/kfence/core.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/mm/kfence/core.c b/mm/kfence/core.c
> index 33bb20d91bf6a..d51f77329fd3c 100644
> --- a/mm/kfence/core.c
> +++ b/mm/kfence/core.c
> @@ -740,6 +740,14 @@ void *__kfence_alloc(struct kmem_cache *s, size_t size, gfp_t flags)
> if (size > PAGE_SIZE)
> return NULL;
>
> + /*
> + * Skip allocations from non-default zones, including DMA. We cannot guarantee that pages
> + * in the KFENCE pool will have the requested properties (e.g. reside in DMAable memory).

Comments should still be 80 cols, like the rest of the file. :-/

> + */
> + if ((flags & GFP_ZONEMASK) ||
> + (s->flags & (SLAB_CACHE_DMA | SLAB_CACHE_DMA32)))
> + return NULL;
> +
> /*
> * allocation_gate only needs to become non-zero, so it doesn't make
> * sense to continue writing to it and pay the associated contention
> --
> 2.32.0.93.g670b81a890-goog
>

2021-06-30 14:07:42

by Marco Elver

Subject: Re: [PATCH v3 1/2] kfence: move the size check to the beginning of __kfence_alloc()

On Wed, 30 Jun 2021 at 15:53, Alexander Potapenko <[email protected]> wrote:
>
> Check the allocation size before toggling kfence_allocation_gate.
> This way allocations that can't be served by KFENCE will not result in
> waiting for another CONFIG_KFENCE_SAMPLE_INTERVAL without allocating
> anything.
>
> Suggested-by: Marco Elver <[email protected]>
> Cc: Andrew Morton <[email protected]>
> Cc: Dmitry Vyukov <[email protected]>
> Cc: Marco Elver <[email protected]>
> Cc: Greg Kroah-Hartman <[email protected]>
> Cc: [email protected] # 5.12+
> Signed-off-by: Alexander Potapenko <[email protected]>

Reviewed-by: Marco Elver <[email protected]>


> ---
> mm/kfence/core.c | 10 +++++++---
> 1 file changed, 7 insertions(+), 3 deletions(-)
>
> diff --git a/mm/kfence/core.c b/mm/kfence/core.c
> index 4d21ac44d5d35..33bb20d91bf6a 100644
> --- a/mm/kfence/core.c
> +++ b/mm/kfence/core.c
> @@ -733,6 +733,13 @@ void kfence_shutdown_cache(struct kmem_cache *s)
>
> void *__kfence_alloc(struct kmem_cache *s, size_t size, gfp_t flags)
> {
> + /*
> + * Perform size check before switching kfence_allocation_gate, so that
> + * we don't disable KFENCE without making an allocation.
> + */
> + if (size > PAGE_SIZE)
> + return NULL;
> +
> /*
> * allocation_gate only needs to become non-zero, so it doesn't make
> * sense to continue writing to it and pay the associated contention
> @@ -757,9 +764,6 @@ void *__kfence_alloc(struct kmem_cache *s, size_t size, gfp_t flags)
> if (!READ_ONCE(kfence_enabled))
> return NULL;
>
> - if (size > PAGE_SIZE)
> - return NULL;
> -
> return kfence_guarded_alloc(s, size, flags);
> }
>
> --
> 2.32.0.93.g670b81a890-goog
>