[PATCH bpf-next v2] bpf: use VM_MAP instead of VM_ALLOC for ringbuf

[PATCH bpf-next v2] bpf: use VM_MAP instead of VM_ALLOC for ringbuf

[Hou Tao]

After commit 2fd3fb0be1d1 ("kasan, vmalloc: unpoison VM_ALLOC pages
after mapping"), non-VM_ALLOC mappings will be marked as accessible
in __get_vm_area_node() when KASAN is enabled. But now the flag for
ringbuf area is VM_ALLOC, so KASAN will complain out-of-bound access
after vmap() returns. Because the ringbuf area is created by mapping
allocated pages, so use VM_MAP instead.

After the change, info in /proc/vmallocinfo also changes from
  [start]-[end]   24576 ringbuf_map_alloc+0x171/0x290 vmalloc user
to
  [start]-[end]   24576 ringbuf_map_alloc+0x171/0x290 vmap user

Reported-by: syzbot+5ad567a418794b9b5983@syzkaller.appspotmail.com
Signed-off-by: Hou Tao <houtao1@huawei.com>
---
v2:
  * explain why VM_ALLOC will lead to vmalloc-oob access
  * add Reported-by tag
v1: https://lore.kernel.org/bpf/CANUnq3a+sT_qtO1wNQ3GnLGN7FLvSSgvit2UVgqQKRpUvs85VQ@mail.gmail.com/T/#t
---
 kernel/bpf/ringbuf.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/bpf/ringbuf.c b/kernel/bpf/ringbuf.c
index 638d7fd7b375..710ba9de12ce 100644
--- a/kernel/bpf/ringbuf.c
+++ b/kernel/bpf/ringbuf.c
@@ -104,7 +104,7 @@ static struct bpf_ringbuf *bpf_ringbuf_area_alloc(size_t data_sz, int numa_node)
 	}
 
 	rb = vmap(pages, nr_meta_pages + 2 * nr_data_pages,
-		  VM_ALLOC | VM_USERMAP, PAGE_KERNEL);
+		  VM_MAP | VM_USERMAP, PAGE_KERNEL);
 	if (rb) {
 		kmemleak_not_leak(pages);
 		rb->pages = pages;
-- 
2.35.1

Re: [PATCH bpf-next v2] bpf: use VM_MAP instead of VM_ALLOC for ringbuf

[Daniel Borkmann]

[ +Andrey ]

On 2/2/22 7:01 AM, Hou Tao wrote:
> After commit 2fd3fb0be1d1 ("kasan, vmalloc: unpoison VM_ALLOC pages
> after mapping"), non-VM_ALLOC mappings will be marked as accessible
> in __get_vm_area_node() when KASAN is enabled. But now the flag for
> ringbuf area is VM_ALLOC, so KASAN will complain out-of-bound access
> after vmap() returns. Because the ringbuf area is created by mapping
> allocated pages, so use VM_MAP instead.
> 
> After the change, info in /proc/vmallocinfo also changes from
>    [start]-[end]   24576 ringbuf_map_alloc+0x171/0x290 vmalloc user
> to
>    [start]-[end]   24576 ringbuf_map_alloc+0x171/0x290 vmap user
> 
> Reported-by: syzbot+5ad567a418794b9b5983@syzkaller.appspotmail.com
> Signed-off-by: Hou Tao <houtao1@huawei.com>
> ---
> v2:
>    * explain why VM_ALLOC will lead to vmalloc-oob access

Do you know which tree commit 2fd3fb0be1d1 is, looks like it's neither
in bpf nor in bpf-next tree at the moment.

Either way, I presume this fix should be routed via bpf tree rather
than bpf-next? (I can add Fixes tag while applying.)

>    * add Reported-by tag
> v1: https://lore.kernel.org/bpf/CANUnq3a+sT_qtO1wNQ3GnLGN7FLvSSgvit2UVgqQKRpUvs85VQ@mail.gmail.com/T/#t
> ---
>   kernel/bpf/ringbuf.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/bpf/ringbuf.c b/kernel/bpf/ringbuf.c
> index 638d7fd7b375..710ba9de12ce 100644
> --- a/kernel/bpf/ringbuf.c
> +++ b/kernel/bpf/ringbuf.c
> @@ -104,7 +104,7 @@ static struct bpf_ringbuf *bpf_ringbuf_area_alloc(size_t data_sz, int numa_node)
>   	}
>   
>   	rb = vmap(pages, nr_meta_pages + 2 * nr_data_pages,
> -		  VM_ALLOC | VM_USERMAP, PAGE_KERNEL);
> +		  VM_MAP | VM_USERMAP, PAGE_KERNEL);
>   	if (rb) {
>   		kmemleak_not_leak(pages);
>   		rb->pages = pages;
> 

Re: [PATCH bpf-next v2] bpf: use VM_MAP instead of VM_ALLOC for ringbuf

[Hou Tao]

Hi,

> On 2/2/22 7:01 AM, Hou Tao wrote:
> > After commit 2fd3fb0be1d1 ("kasan, vmalloc: unpoison VM_ALLOC pages
> > after mapping"), non-VM_ALLOC mappings will be marked as accessible
> > in __get_vm_area_node() when KASAN is enabled. But now the flag for
> > ringbuf area is VM_ALLOC, so KASAN will complain out-of-bound access
> > after vmap() returns. Because the ringbuf area is created by mapping
> > allocated pages, so use VM_MAP instead.
> > 
> > After the change, info in /proc/vmallocinfo also changes from
> >    [start]-[end]   24576 ringbuf_map_alloc+0x171/0x290 vmalloc user
> > to
> >    [start]-[end]   24576 ringbuf_map_alloc+0x171/0x290 vmap user
> > 
> > Reported-by: syzbot+5ad567a418794b9b5983@syzkaller.appspotmail.com
> > Signed-off-by: Hou Tao <houtao1@huawei.com>
> > ---
> > v2:
> >    * explain why VM_ALLOC will lead to vmalloc-oob access
> 
> Do you know which tree commit 2fd3fb0be1d1 is, looks like it's neither
> in bpf nor in bpf-next tree at the moment.
> 
It is on linux-next tree:
 
 $ git name-rev 2fd3fb0be1d1
 2fd3fb0be1d1 tags/next-20220201~2^2~96
 
> Either way, I presume this fix should be routed via bpf tree rather
> than bpf-next? (I can add Fixes tag while applying.)
>
Make sense and thanks for that.

Regards,
Tao

> >    * add Reported-by tag
> > v1: https://lore.kernel.org/bpf/CANUnq3a+sT_qtO1wNQ3GnLGN7FLvSSgvit2UVgqQKRpUvs85VQ@mail.gmail.com/T/#t
> > ---
> >   kernel/bpf/ringbuf.c | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/kernel/bpf/ringbuf.c b/kernel/bpf/ringbuf.c
> > index 638d7fd7b375..710ba9de12ce 100644
> > --- a/kernel/bpf/ringbuf.c
> > +++ b/kernel/bpf/ringbuf.c
> > @@ -104,7 +104,7 @@ static struct bpf_ringbuf *bpf_ringbuf_area_alloc(size_t data_sz, int numa_node)
> >   	}
> >   
> >   	rb = vmap(pages, nr_meta_pages + 2 * nr_data_pages,
> > -		  VM_ALLOC | VM_USERMAP, PAGE_KERNEL);
> > +		  VM_MAP | VM_USERMAP, PAGE_KERNEL);
> >   	if (rb) {
> >   		kmemleak_not_leak(pages);
> >   		rb->pages = pages;
> > 
> 
> 

Re: [PATCH bpf-next v2] bpf: use VM_MAP instead of VM_ALLOC for ringbuf

[Andrii Nakryiko]

On Wed, Feb 2, 2022 at 9:14 PM Hou Tao <hotforest@gmail.com> wrote:
>
> Hi,
>
> > On 2/2/22 7:01 AM, Hou Tao wrote:
> > > After commit 2fd3fb0be1d1 ("kasan, vmalloc: unpoison VM_ALLOC pages
> > > after mapping"), non-VM_ALLOC mappings will be marked as accessible
> > > in __get_vm_area_node() when KASAN is enabled. But now the flag for
> > > ringbuf area is VM_ALLOC, so KASAN will complain out-of-bound access
> > > after vmap() returns. Because the ringbuf area is created by mapping
> > > allocated pages, so use VM_MAP instead.
> > >
> > > After the change, info in /proc/vmallocinfo also changes from
> > >    [start]-[end]   24576 ringbuf_map_alloc+0x171/0x290 vmalloc user
> > > to
> > >    [start]-[end]   24576 ringbuf_map_alloc+0x171/0x290 vmap user
> > >
> > > Reported-by: syzbot+5ad567a418794b9b5983@syzkaller.appspotmail.com
> > > Signed-off-by: Hou Tao <houtao1@huawei.com>
> > > ---
> > > v2:
> > >    * explain why VM_ALLOC will lead to vmalloc-oob access
> >
> > Do you know which tree commit 2fd3fb0be1d1 is, looks like it's neither
> > in bpf nor in bpf-next tree at the moment.
> >
> It is on linux-next tree:
>
>  $ git name-rev 2fd3fb0be1d1
>  2fd3fb0be1d1 tags/next-20220201~2^2~96
>
> > Either way, I presume this fix should be routed via bpf tree rather
> > than bpf-next? (I can add Fixes tag while applying.)
> >
> Make sense and thanks for that.

Added

Fixes: 457f44363a88 ("bpf: Implement BPF ring buffer and verifier
support for it")

and pushed to bpf tree, thanks.

>
> Regards,
> Tao
>
> > >    * add Reported-by tag
> > > v1: https://lore.kernel.org/bpf/CANUnq3a+sT_qtO1wNQ3GnLGN7FLvSSgvit2UVgqQKRpUvs85VQ@mail.gmail.com/T/#t
> > > ---
> > >   kernel/bpf/ringbuf.c | 2 +-
> > >   1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/kernel/bpf/ringbuf.c b/kernel/bpf/ringbuf.c
> > > index 638d7fd7b375..710ba9de12ce 100644
> > > --- a/kernel/bpf/ringbuf.c
> > > +++ b/kernel/bpf/ringbuf.c
> > > @@ -104,7 +104,7 @@ static struct bpf_ringbuf *bpf_ringbuf_area_alloc(size_t data_sz, int numa_node)
> > >     }
> > >
> > >     rb = vmap(pages, nr_meta_pages + 2 * nr_data_pages,
> > > -             VM_ALLOC | VM_USERMAP, PAGE_KERNEL);
> > > +             VM_MAP | VM_USERMAP, PAGE_KERNEL);
> > >     if (rb) {
> > >             kmemleak_not_leak(pages);
> > >             rb->pages = pages;
> > >
> >
> >

Re: [PATCH bpf-next v2] bpf: use VM_MAP instead of VM_ALLOC for ringbuf

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to bpf/bpf.git (master)
by Andrii Nakryiko <andrii@kernel.org>:

On Wed,  2 Feb 2022 14:01:58 +0800 you wrote:
> After commit 2fd3fb0be1d1 ("kasan, vmalloc: unpoison VM_ALLOC pages
> after mapping"), non-VM_ALLOC mappings will be marked as accessible
> in __get_vm_area_node() when KASAN is enabled. But now the flag for
> ringbuf area is VM_ALLOC, so KASAN will complain out-of-bound access
> after vmap() returns. Because the ringbuf area is created by mapping
> allocated pages, so use VM_MAP instead.
> 
> [...]

Here is the summary with links:
  - [bpf-next,v2] bpf: use VM_MAP instead of VM_ALLOC for ringbuf
    https://git.kernel.org/bpf/bpf/c/b293dcc473d2

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html