[PATCH v2 bpf] bpf: enforce BPF ringbuf size to be the power of 2

[PATCH v2 bpf] bpf: enforce BPF ringbuf size to be the power of 2

[Andrii Nakryiko]

BPF ringbuf assumes the size to be a multiple of page size and the power of
2 value. The latter is important to avoid division while calculating position
inside the ring buffer and using (N-1) mask instead. This patch fixes omission
to enforce power-of-2 size rule.

Fixes: 457f44363a88 ("bpf: Implement BPF ring buffer and verifier support for it")
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
---
 kernel/bpf/ringbuf.c | 18 ++++++++----------
 1 file changed, 8 insertions(+), 10 deletions(-)

diff --git a/kernel/bpf/ringbuf.c b/kernel/bpf/ringbuf.c
index 180414bb0d3e..0af88bbc1c15 100644
--- a/kernel/bpf/ringbuf.c
+++ b/kernel/bpf/ringbuf.c
@@ -132,15 +132,6 @@ static struct bpf_ringbuf *bpf_ringbuf_alloc(size_t data_sz, int numa_node)
 {
 	struct bpf_ringbuf *rb;
 
-	if (!data_sz || !PAGE_ALIGNED(data_sz))
-		return ERR_PTR(-EINVAL);
-
-#ifdef CONFIG_64BIT
-	/* on 32-bit arch, it's impossible to overflow record's hdr->pgoff */
-	if (data_sz > RINGBUF_MAX_DATA_SZ)
-		return ERR_PTR(-E2BIG);
-#endif
-
 	rb = bpf_ringbuf_area_alloc(data_sz, numa_node);
 	if (!rb)
 		return ERR_PTR(-ENOMEM);
@@ -166,9 +157,16 @@ static struct bpf_map *ringbuf_map_alloc(union bpf_attr *attr)
 		return ERR_PTR(-EINVAL);
 
 	if (attr->key_size || attr->value_size ||
-	    attr->max_entries == 0 || !PAGE_ALIGNED(attr->max_entries))
+	    !is_power_of_2(attr->max_entries) ||
+	    !PAGE_ALIGNED(attr->max_entries))
 		return ERR_PTR(-EINVAL);
 
+#ifdef CONFIG_64BIT
+	/* on 32-bit arch, it's impossible to overflow record's hdr->pgoff */
+	if (attr->max_entries > RINGBUF_MAX_DATA_SZ)
+		return ERR_PTR(-E2BIG);
+#endif
+
 	rb_map = kzalloc(sizeof(*rb_map), GFP_USER);
 	if (!rb_map)
 		return ERR_PTR(-ENOMEM);
-- 
2.24.1

Re: [PATCH v2 bpf] bpf: enforce BPF ringbuf size to be the power of 2

[Daniel Borkmann]

On 6/30/20 8:15 AM, Andrii Nakryiko wrote:
> BPF ringbuf assumes the size to be a multiple of page size and the power of
> 2 value. The latter is important to avoid division while calculating position
> inside the ring buffer and using (N-1) mask instead. This patch fixes omission
> to enforce power-of-2 size rule.
> 
> Fixes: 457f44363a88 ("bpf: Implement BPF ring buffer and verifier support for it")
> Signed-off-by: Andrii Nakryiko <andriin@fb.com>

Lgtm, applied, thanks!

[...]
> @@ -166,9 +157,16 @@ static struct bpf_map *ringbuf_map_alloc(union bpf_attr *attr)
>   		return ERR_PTR(-EINVAL);
>   
>   	if (attr->key_size || attr->value_size ||
> -	    attr->max_entries == 0 || !PAGE_ALIGNED(attr->max_entries))
> +	    !is_power_of_2(attr->max_entries) ||
> +	    !PAGE_ALIGNED(attr->max_entries))

Technically !IS_ALIGNED(attr->max_entries, PAGE_SIZE) might have been a bit cleaner
since PAGE_ALIGNED() is only intended for pointers, though, not wrong here given
max_entries is u32.

Thanks,
Daniel

Re: [PATCH v2 bpf] bpf: enforce BPF ringbuf size to be the power of 2

[Andrii Nakryiko]

On Tue, Jun 30, 2020 at 7:52 AM Daniel Borkmann <daniel@iogearbox.net> wrote:
>
> On 6/30/20 8:15 AM, Andrii Nakryiko wrote:
> > BPF ringbuf assumes the size to be a multiple of page size and the power of
> > 2 value. The latter is important to avoid division while calculating position
> > inside the ring buffer and using (N-1) mask instead. This patch fixes omission
> > to enforce power-of-2 size rule.
> >
> > Fixes: 457f44363a88 ("bpf: Implement BPF ring buffer and verifier support for it")
> > Signed-off-by: Andrii Nakryiko <andriin@fb.com>
>
> Lgtm, applied, thanks!
>

Thanks, Daniel!

> [...]
> > @@ -166,9 +157,16 @@ static struct bpf_map *ringbuf_map_alloc(union bpf_attr *attr)
> >               return ERR_PTR(-EINVAL);
> >
> >       if (attr->key_size || attr->value_size ||
> > -         attr->max_entries == 0 || !PAGE_ALIGNED(attr->max_entries))
> > +         !is_power_of_2(attr->max_entries) ||
> > +         !PAGE_ALIGNED(attr->max_entries))
>
> Technically !IS_ALIGNED(attr->max_entries, PAGE_SIZE) might have been a bit cleaner
> since PAGE_ALIGNED() is only intended for pointers, though, not wrong here given
> max_entries is u32.

I've found a bunch of uses on non-pointers, e.g., `if
(!PAGE_ALIGNED(fs_info->nodesize)) {` in BTRFS code, so assumed it's
intended to be used more generically. But let me know if you want me
to do IS_ALIGNED change.

>
> Thanks,
> Daniel