lkml.org 
[lkml]   [2021]   [Jun]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH] mm/vmalloc: unbreak kasan vmalloc support
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 <npiggin@gmail.com>
> > Cc: David Gow <davidgow@google.com>
> > Cc: Dmitry Vyukov <dvyukov@google.com>
> > Cc: Andrey Konovalov <andreyknvl@gmail.com>
> > Cc: Uladzislau Rezki (Sony) <urezki@gmail.com>
> > 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 <npiggin@gmail.com>
>
> > Signed-off-by: Daniel Axtens <dja@axtens.net>
> > ---
> > 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) <urezki@gmail.com>

Indeed hugepage mapping was broken in regard to KASAN.

Thanks!

--
Vlad Rezki

\
 
 \ /
  Last update: 2021-06-19 15:03    [W:0.065 / U:0.756 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site