In commit 121e6f3258fe ("mm/vmalloc: hugepage vmalloc mappings"),
__vmalloc_node_range was changed such that __get_vm_area_node was no
longer called with the requested/real size of the vmalloc allocation, but
rather with a rounded-up size.
This means that __get_vm_area_node called kasan_unpoision_vmalloc() with
a rounded up size rather than the real size. This led to it allowing
access to too much memory and so missing vmalloc OOBs and failing the
kasan kunit tests.
Pass the real size and the desired shift into __get_vm_area_node. This
allows it to round up the size for the underlying allocators while
still unpoisioning the correct quantity of shadow memory.
Adjust the other call-sites to pass in PAGE_SHIFT for the shift value.
Cc: Nicholas Piggin <[email protected]>
Cc: David Gow <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: Andrey Konovalov <[email protected]>
Cc: Uladzislau Rezki (Sony) <[email protected]>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=213335
Fixes: 121e6f3258fe ("mm/vmalloc: hugepage vmalloc mappings")
Signed-off-by: Daniel Axtens <[email protected]>
---
mm/vmalloc.c | 24 ++++++++++++++----------
1 file changed, 14 insertions(+), 10 deletions(-)
diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index aaad569e8963..3471cbeb083c 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -2362,15 +2362,16 @@ static void clear_vm_uninitialized_flag(struct vm_struct *vm)
}
static struct vm_struct *__get_vm_area_node(unsigned long size,
- unsigned long align, unsigned long flags, unsigned long start,
- unsigned long end, int node, gfp_t gfp_mask, const void *caller)
+ unsigned long align, unsigned long shift, unsigned long flags,
+ unsigned long start, unsigned long end, int node,
+ gfp_t gfp_mask, const void *caller)
{
struct vmap_area *va;
struct vm_struct *area;
unsigned long requested_size = size;
BUG_ON(in_interrupt());
- size = PAGE_ALIGN(size);
+ size = ALIGN(size, 1ul << shift);
if (unlikely(!size))
return NULL;
@@ -2402,8 +2403,8 @@ struct vm_struct *__get_vm_area_caller(unsigned long size, unsigned long flags,
unsigned long start, unsigned long end,
const void *caller)
{
- return __get_vm_area_node(size, 1, flags, start, end, NUMA_NO_NODE,
- GFP_KERNEL, caller);
+ return __get_vm_area_node(size, 1, PAGE_SHIFT, flags, start, end,
+ NUMA_NO_NODE, GFP_KERNEL, caller);
}
/**
@@ -2419,7 +2420,8 @@ struct vm_struct *__get_vm_area_caller(unsigned long size, unsigned long flags,
*/
struct vm_struct *get_vm_area(unsigned long size, unsigned long flags)
{
- return __get_vm_area_node(size, 1, flags, VMALLOC_START, VMALLOC_END,
+ return __get_vm_area_node(size, 1, PAGE_SHIFT, flags,
+ VMALLOC_START, VMALLOC_END,
NUMA_NO_NODE, GFP_KERNEL,
__builtin_return_address(0));
}
@@ -2427,7 +2429,8 @@ struct vm_struct *get_vm_area(unsigned long size, unsigned long flags)
struct vm_struct *get_vm_area_caller(unsigned long size, unsigned long flags,
const void *caller)
{
- return __get_vm_area_node(size, 1, flags, VMALLOC_START, VMALLOC_END,
+ return __get_vm_area_node(size, 1, PAGE_SHIFT, flags,
+ VMALLOC_START, VMALLOC_END,
NUMA_NO_NODE, GFP_KERNEL, caller);
}
@@ -2949,9 +2952,9 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,
}
again:
- size = PAGE_ALIGN(size);
- area = __get_vm_area_node(size, align, VM_ALLOC | VM_UNINITIALIZED |
- vm_flags, start, end, node, gfp_mask, caller);
+ area = __get_vm_area_node(real_size, align, shift, VM_ALLOC |
+ VM_UNINITIALIZED | vm_flags, start, end, node,
+ gfp_mask, caller);
if (!area) {
warn_alloc(gfp_mask, NULL,
"vmalloc error: size %lu, vm_struct allocation failed",
@@ -2970,6 +2973,7 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,
*/
clear_vm_uninitialized_flag(area);
+ size = PAGE_ALIGN(size);
kmemleak_vmalloc(area, size, gfp_mask);
return addr;
--
2.30.2
On Thu, Jun 17, 2021 at 4:13 PM Daniel Axtens <[email protected]> wrote:
>
> In commit 121e6f3258fe ("mm/vmalloc: hugepage vmalloc mappings"),
> __vmalloc_node_range was changed such that __get_vm_area_node was no
> longer called with the requested/real size of the vmalloc allocation, but
> rather with a rounded-up size.
>
> This means that __get_vm_area_node called kasan_unpoision_vmalloc() with
> a rounded up size rather than the real size. This led to it allowing
> access to too much memory and so missing vmalloc OOBs and failing the
> kasan kunit tests.
>
> Pass the real size and the desired shift into __get_vm_area_node. This
> allows it to round up the size for the underlying allocators while
> still unpoisioning the correct quantity of shadow memory.
>
> Adjust the other call-sites to pass in PAGE_SHIFT for the shift value.
>
> Cc: Nicholas Piggin <[email protected]>
> Cc: David Gow <[email protected]>
> Cc: Dmitry Vyukov <[email protected]>
> Cc: Andrey Konovalov <[email protected]>
> Cc: Uladzislau Rezki (Sony) <[email protected]>
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=213335
> Fixes: 121e6f3258fe ("mm/vmalloc: hugepage vmalloc mappings")
> Signed-off-by: Daniel Axtens <[email protected]>
> ---
This fixes the KUnit test failure I was seeing on x86_64, thanks!
Tested-by: David Gow <[email protected]>
Cheers,
-- David
Excerpts from Daniel Axtens's message of June 17, 2021 6:13 pm:
> In commit 121e6f3258fe ("mm/vmalloc: hugepage vmalloc mappings"),
> __vmalloc_node_range was changed such that __get_vm_area_node was no
> longer called with the requested/real size of the vmalloc allocation, but
> rather with a rounded-up size.
>
> This means that __get_vm_area_node called kasan_unpoision_vmalloc() with
> a rounded up size rather than the real size. This led to it allowing
> access to too much memory and so missing vmalloc OOBs and failing the
> kasan kunit tests.
>
> Pass the real size and the desired shift into __get_vm_area_node. This
> allows it to round up the size for the underlying allocators while
> still unpoisioning the correct quantity of shadow memory.
>
> Adjust the other call-sites to pass in PAGE_SHIFT for the shift value.
>
> Cc: Nicholas Piggin <[email protected]>
> Cc: David Gow <[email protected]>
> Cc: Dmitry Vyukov <[email protected]>
> Cc: Andrey Konovalov <[email protected]>
> Cc: Uladzislau Rezki (Sony) <[email protected]>
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=213335
> Fixes: 121e6f3258fe ("mm/vmalloc: hugepage vmalloc mappings")
Thanks Daniel, good debugging.
Reviewed-by: Nicholas Piggin <[email protected]>
> Signed-off-by: Daniel Axtens <[email protected]>
> ---
> mm/vmalloc.c | 24 ++++++++++++++----------
> 1 file changed, 14 insertions(+), 10 deletions(-)
>
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index aaad569e8963..3471cbeb083c 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -2362,15 +2362,16 @@ static void clear_vm_uninitialized_flag(struct vm_struct *vm)
> }
>
> static struct vm_struct *__get_vm_area_node(unsigned long size,
> - unsigned long align, unsigned long flags, unsigned long start,
> - unsigned long end, int node, gfp_t gfp_mask, const void *caller)
> + unsigned long align, unsigned long shift, unsigned long flags,
> + unsigned long start, unsigned long end, int node,
> + gfp_t gfp_mask, const void *caller)
> {
> struct vmap_area *va;
> struct vm_struct *area;
> unsigned long requested_size = size;
>
> BUG_ON(in_interrupt());
> - size = PAGE_ALIGN(size);
> + size = ALIGN(size, 1ul << shift);
> if (unlikely(!size))
> return NULL;
>
> @@ -2402,8 +2403,8 @@ struct vm_struct *__get_vm_area_caller(unsigned long size, unsigned long flags,
> unsigned long start, unsigned long end,
> const void *caller)
> {
> - return __get_vm_area_node(size, 1, flags, start, end, NUMA_NO_NODE,
> - GFP_KERNEL, caller);
> + return __get_vm_area_node(size, 1, PAGE_SHIFT, flags, start, end,
> + NUMA_NO_NODE, GFP_KERNEL, caller);
> }
>
> /**
> @@ -2419,7 +2420,8 @@ struct vm_struct *__get_vm_area_caller(unsigned long size, unsigned long flags,
> */
> struct vm_struct *get_vm_area(unsigned long size, unsigned long flags)
> {
> - return __get_vm_area_node(size, 1, flags, VMALLOC_START, VMALLOC_END,
> + return __get_vm_area_node(size, 1, PAGE_SHIFT, flags,
> + VMALLOC_START, VMALLOC_END,
> NUMA_NO_NODE, GFP_KERNEL,
> __builtin_return_address(0));
> }
> @@ -2427,7 +2429,8 @@ struct vm_struct *get_vm_area(unsigned long size, unsigned long flags)
> struct vm_struct *get_vm_area_caller(unsigned long size, unsigned long flags,
> const void *caller)
> {
> - return __get_vm_area_node(size, 1, flags, VMALLOC_START, VMALLOC_END,
> + return __get_vm_area_node(size, 1, PAGE_SHIFT, flags,
> + VMALLOC_START, VMALLOC_END,
> NUMA_NO_NODE, GFP_KERNEL, caller);
> }
>
> @@ -2949,9 +2952,9 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,
> }
>
> again:
> - size = PAGE_ALIGN(size);
> - area = __get_vm_area_node(size, align, VM_ALLOC | VM_UNINITIALIZED |
> - vm_flags, start, end, node, gfp_mask, caller);
> + area = __get_vm_area_node(real_size, align, shift, VM_ALLOC |
> + VM_UNINITIALIZED | vm_flags, start, end, node,
> + gfp_mask, caller);
> if (!area) {
> warn_alloc(gfp_mask, NULL,
> "vmalloc error: size %lu, vm_struct allocation failed",
> @@ -2970,6 +2973,7 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,
> */
> clear_vm_uninitialized_flag(area);
>
> + size = PAGE_ALIGN(size);
> kmemleak_vmalloc(area, size, gfp_mask);
>
> return addr;
> --
> 2.30.2
>
>
On Thu, Jun 17, 2021 at 07:40:49PM +1000, Nicholas Piggin wrote:
> Excerpts from Daniel Axtens's message of June 17, 2021 6:13 pm:
> > In commit 121e6f3258fe ("mm/vmalloc: hugepage vmalloc mappings"),
> > __vmalloc_node_range was changed such that __get_vm_area_node was no
> > longer called with the requested/real size of the vmalloc allocation, but
> > rather with a rounded-up size.
> >
> > This means that __get_vm_area_node called kasan_unpoision_vmalloc() with
> > a rounded up size rather than the real size. This led to it allowing
> > access to too much memory and so missing vmalloc OOBs and failing the
> > kasan kunit tests.
> >
> > Pass the real size and the desired shift into __get_vm_area_node. This
> > allows it to round up the size for the underlying allocators while
> > still unpoisioning the correct quantity of shadow memory.
> >
> > Adjust the other call-sites to pass in PAGE_SHIFT for the shift value.
> >
> > Cc: Nicholas Piggin <[email protected]>
> > Cc: David Gow <[email protected]>
> > Cc: Dmitry Vyukov <[email protected]>
> > Cc: Andrey Konovalov <[email protected]>
> > Cc: Uladzislau Rezki (Sony) <[email protected]>
> > Link: https://bugzilla.kernel.org/show_bug.cgi?id=213335
> > Fixes: 121e6f3258fe ("mm/vmalloc: hugepage vmalloc mappings")
>
> Thanks Daniel, good debugging.
>
> Reviewed-by: Nicholas Piggin <[email protected]>
>
> > Signed-off-by: Daniel Axtens <[email protected]>
> > ---
> > mm/vmalloc.c | 24 ++++++++++++++----------
> > 1 file changed, 14 insertions(+), 10 deletions(-)
> >
> > diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> > index aaad569e8963..3471cbeb083c 100644
> > --- a/mm/vmalloc.c
> > +++ b/mm/vmalloc.c
> > @@ -2362,15 +2362,16 @@ static void clear_vm_uninitialized_flag(struct vm_struct *vm)
> > }
> >
> > static struct vm_struct *__get_vm_area_node(unsigned long size,
> > - unsigned long align, unsigned long flags, unsigned long start,
> > - unsigned long end, int node, gfp_t gfp_mask, const void *caller)
> > + unsigned long align, unsigned long shift, unsigned long flags,
> > + unsigned long start, unsigned long end, int node,
> > + gfp_t gfp_mask, const void *caller)
> > {
> > struct vmap_area *va;
> > struct vm_struct *area;
> > unsigned long requested_size = size;
> >
> > BUG_ON(in_interrupt());
> > - size = PAGE_ALIGN(size);
> > + size = ALIGN(size, 1ul << shift);
> > if (unlikely(!size))
> > return NULL;
> >
> > @@ -2402,8 +2403,8 @@ struct vm_struct *__get_vm_area_caller(unsigned long size, unsigned long flags,
> > unsigned long start, unsigned long end,
> > const void *caller)
> > {
> > - return __get_vm_area_node(size, 1, flags, start, end, NUMA_NO_NODE,
> > - GFP_KERNEL, caller);
> > + return __get_vm_area_node(size, 1, PAGE_SHIFT, flags, start, end,
> > + NUMA_NO_NODE, GFP_KERNEL, caller);
> > }
> >
> > /**
> > @@ -2419,7 +2420,8 @@ struct vm_struct *__get_vm_area_caller(unsigned long size, unsigned long flags,
> > */
> > struct vm_struct *get_vm_area(unsigned long size, unsigned long flags)
> > {
> > - return __get_vm_area_node(size, 1, flags, VMALLOC_START, VMALLOC_END,
> > + return __get_vm_area_node(size, 1, PAGE_SHIFT, flags,
> > + VMALLOC_START, VMALLOC_END,
> > NUMA_NO_NODE, GFP_KERNEL,
> > __builtin_return_address(0));
> > }
> > @@ -2427,7 +2429,8 @@ struct vm_struct *get_vm_area(unsigned long size, unsigned long flags)
> > struct vm_struct *get_vm_area_caller(unsigned long size, unsigned long flags,
> > const void *caller)
> > {
> > - return __get_vm_area_node(size, 1, flags, VMALLOC_START, VMALLOC_END,
> > + return __get_vm_area_node(size, 1, PAGE_SHIFT, flags,
> > + VMALLOC_START, VMALLOC_END,
> > NUMA_NO_NODE, GFP_KERNEL, caller);
> > }
> >
> > @@ -2949,9 +2952,9 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,
> > }
> >
> > again:
> > - size = PAGE_ALIGN(size);
> > - area = __get_vm_area_node(size, align, VM_ALLOC | VM_UNINITIALIZED |
> > - vm_flags, start, end, node, gfp_mask, caller);
> > + area = __get_vm_area_node(real_size, align, shift, VM_ALLOC |
> > + VM_UNINITIALIZED | vm_flags, start, end, node,
> > + gfp_mask, caller);
> > if (!area) {
> > warn_alloc(gfp_mask, NULL,
> > "vmalloc error: size %lu, vm_struct allocation failed",
> > @@ -2970,6 +2973,7 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,
> > */
> > clear_vm_uninitialized_flag(area);
> >
> > + size = PAGE_ALIGN(size);
> > kmemleak_vmalloc(area, size, gfp_mask);
> >
> > return addr;
> > --
> > 2.30.2
> >
> >
Reviewed-by: Uladzislau Rezki (Sony) <[email protected]>
Indeed hugepage mapping was broken in regard to KASAN.
Thanks!
--
Vlad Rezki
On Thu, Jun 17, 2021 at 11:13 AM Daniel Axtens <[email protected]> wrote:
>
> In commit 121e6f3258fe ("mm/vmalloc: hugepage vmalloc mappings"),
> __vmalloc_node_range was changed such that __get_vm_area_node was no
> longer called with the requested/real size of the vmalloc allocation, but
> rather with a rounded-up size.
>
> This means that __get_vm_area_node called kasan_unpoision_vmalloc() with
> a rounded up size rather than the real size. This led to it allowing
> access to too much memory and so missing vmalloc OOBs and failing the
> kasan kunit tests.
>
> Pass the real size and the desired shift into __get_vm_area_node. This
> allows it to round up the size for the underlying allocators while
> still unpoisioning the correct quantity of shadow memory.
>
> Adjust the other call-sites to pass in PAGE_SHIFT for the shift value.
>
> Cc: Nicholas Piggin <[email protected]>
> Cc: David Gow <[email protected]>
> Cc: Dmitry Vyukov <[email protected]>
> Cc: Andrey Konovalov <[email protected]>
> Cc: Uladzislau Rezki (Sony) <[email protected]>
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=213335
> Fixes: 121e6f3258fe ("mm/vmalloc: hugepage vmalloc mappings")
> Signed-off-by: Daniel Axtens <[email protected]>
> ---
> mm/vmalloc.c | 24 ++++++++++++++----------
> 1 file changed, 14 insertions(+), 10 deletions(-)
>
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index aaad569e8963..3471cbeb083c 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -2362,15 +2362,16 @@ static void clear_vm_uninitialized_flag(struct vm_struct *vm)
> }
>
> static struct vm_struct *__get_vm_area_node(unsigned long size,
> - unsigned long align, unsigned long flags, unsigned long start,
> - unsigned long end, int node, gfp_t gfp_mask, const void *caller)
> + unsigned long align, unsigned long shift, unsigned long flags,
> + unsigned long start, unsigned long end, int node,
> + gfp_t gfp_mask, const void *caller)
> {
> struct vmap_area *va;
> struct vm_struct *area;
> unsigned long requested_size = size;
>
> BUG_ON(in_interrupt());
> - size = PAGE_ALIGN(size);
> + size = ALIGN(size, 1ul << shift);
> if (unlikely(!size))
> return NULL;
>
> @@ -2402,8 +2403,8 @@ struct vm_struct *__get_vm_area_caller(unsigned long size, unsigned long flags,
> unsigned long start, unsigned long end,
> const void *caller)
> {
> - return __get_vm_area_node(size, 1, flags, start, end, NUMA_NO_NODE,
> - GFP_KERNEL, caller);
> + return __get_vm_area_node(size, 1, PAGE_SHIFT, flags, start, end,
> + NUMA_NO_NODE, GFP_KERNEL, caller);
> }
>
> /**
> @@ -2419,7 +2420,8 @@ struct vm_struct *__get_vm_area_caller(unsigned long size, unsigned long flags,
> */
> struct vm_struct *get_vm_area(unsigned long size, unsigned long flags)
> {
> - return __get_vm_area_node(size, 1, flags, VMALLOC_START, VMALLOC_END,
> + return __get_vm_area_node(size, 1, PAGE_SHIFT, flags,
> + VMALLOC_START, VMALLOC_END,
> NUMA_NO_NODE, GFP_KERNEL,
> __builtin_return_address(0));
> }
> @@ -2427,7 +2429,8 @@ struct vm_struct *get_vm_area(unsigned long size, unsigned long flags)
> struct vm_struct *get_vm_area_caller(unsigned long size, unsigned long flags,
> const void *caller)
> {
> - return __get_vm_area_node(size, 1, flags, VMALLOC_START, VMALLOC_END,
> + return __get_vm_area_node(size, 1, PAGE_SHIFT, flags,
> + VMALLOC_START, VMALLOC_END,
> NUMA_NO_NODE, GFP_KERNEL, caller);
> }
>
> @@ -2949,9 +2952,9 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,
> }
>
> again:
> - size = PAGE_ALIGN(size);
> - area = __get_vm_area_node(size, align, VM_ALLOC | VM_UNINITIALIZED |
> - vm_flags, start, end, node, gfp_mask, caller);
> + area = __get_vm_area_node(real_size, align, shift, VM_ALLOC |
> + VM_UNINITIALIZED | vm_flags, start, end, node,
> + gfp_mask, caller);
> if (!area) {
> warn_alloc(gfp_mask, NULL,
> "vmalloc error: size %lu, vm_struct allocation failed",
> @@ -2970,6 +2973,7 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,
> */
> clear_vm_uninitialized_flag(area);
>
> + size = PAGE_ALIGN(size);
> kmemleak_vmalloc(area, size, gfp_mask);
>
> return addr;
> --
> 2.30.2
>
This fixes the vmalloc_oob test for me. Thank you, Daniel!
Tested-by: Andrey Konovalov <[email protected]>
Acked-by: Andrey Konovalov <[email protected]>