2019-07-04 14:46:46

by Ilya Maximets

Subject: [PATCH bpf] xdp: fix possible cq entry leak

Completion queue address reservation could not be undone.
In case of bad 'queue_id' or skb allocation failure, reserved entry
will be leaked reducing the total capacity of completion queue.

Fix that by moving reservation to the point where failure is not
possible. Additionally, 'queue_id' checking moved out from the loop
since there is no point to check it there.

Fixes: 35fcde7f8deb ("xsk: support for Tx")
Signed-off-by: Ilya Maximets <[email protected]>
---
net/xdp/xsk.c | 11 ++++-------
1 file changed, 4 insertions(+), 7 deletions(-)

diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
index f53a6ef7c155..703cf5ea448b 100644
--- a/net/xdp/xsk.c
+++ b/net/xdp/xsk.c
@@ -226,6 +226,9 @@ static int xsk_generic_xmit(struct sock *sk, struct msghdr *m,

mutex_lock(&xs->mutex);

+ if (xs->queue_id >= xs->dev->real_num_tx_queues)
+ goto out;
+
while (xskq_peek_desc(xs->tx, &desc)) {
char *buffer;
u64 addr;
@@ -236,12 +239,6 @@ static int xsk_generic_xmit(struct sock *sk, struct msghdr *m,
goto out;
}

- if (xskq_reserve_addr(xs->umem->cq))
- goto out;
-
- if (xs->queue_id >= xs->dev->real_num_tx_queues)
- goto out;
-
len = desc.len;
skb = sock_alloc_send_skb(sk, len, 1, &err);
if (unlikely(!skb)) {
@@ -253,7 +250,7 @@ static int xsk_generic_xmit(struct sock *sk, struct msghdr *m,
addr = desc.addr;
buffer = xdp_umem_get_data(xs->umem, addr);
err = skb_store_bits(skb, 0, buffer, len);
- if (unlikely(err)) {
+ if (unlikely(err) || xskq_reserve_addr(xs->umem->cq)) {
kfree_skb(skb);
goto out;
}
--
2.17.1

2019-07-05 06:49:35

by Björn Töpel

Subject: Re: [PATCH bpf] xdp: fix possible cq entry leak

On Thu, 4 Jul 2019 at 16:25, Ilya Maximets <[email protected]> wrote:
>
> Completion queue address reservation could not be undone.
> In case of bad 'queue_id' or skb allocation failure, reserved entry
> will be leaked reducing the total capacity of completion queue.
>
> Fix that by moving reservation to the point where failure is not
> possible. Additionally, 'queue_id' checking moved out from the loop
> since there is no point to check it there.
>

Good catch, Ilya! Thanks for the patch!

Acked-by: Björn Töpel <[email protected]>

> Fixes: 35fcde7f8deb ("xsk: support for Tx")
> Signed-off-by: Ilya Maximets <[email protected]>
> ---
> net/xdp/xsk.c | 11 ++++-------
> 1 file changed, 4 insertions(+), 7 deletions(-)
>
> diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
> index f53a6ef7c155..703cf5ea448b 100644
> --- a/net/xdp/xsk.c
> +++ b/net/xdp/xsk.c
> @@ -226,6 +226,9 @@ static int xsk_generic_xmit(struct sock *sk, struct msghdr *m,
>
> mutex_lock(&xs->mutex);
>
> + if (xs->queue_id >= xs->dev->real_num_tx_queues)
> + goto out;
> +
> while (xskq_peek_desc(xs->tx, &desc)) {
> char *buffer;
> u64 addr;
> @@ -236,12 +239,6 @@ static int xsk_generic_xmit(struct sock *sk, struct msghdr *m,
> goto out;
> }
>
> - if (xskq_reserve_addr(xs->umem->cq))
> - goto out;
> -
> - if (xs->queue_id >= xs->dev->real_num_tx_queues)
> - goto out;
> -
> len = desc.len;
> skb = sock_alloc_send_skb(sk, len, 1, &err);
> if (unlikely(!skb)) {
> @@ -253,7 +250,7 @@ static int xsk_generic_xmit(struct sock *sk, struct msghdr *m,
> addr = desc.addr;
> buffer = xdp_umem_get_data(xs->umem, addr);
> err = skb_store_bits(skb, 0, buffer, len);
> - if (unlikely(err)) {
> + if (unlikely(err) || xskq_reserve_addr(xs->umem->cq)) {
> kfree_skb(skb);
> goto out;
> }
> --
> 2.17.1
>

2019-07-08 23:29:38

by William Tu

Subject: Re: [PATCH bpf] xdp: fix possible cq entry leak

On Thu, Jul 4, 2019 at 11:49 PM Björn Töpel <[email protected]> wrote:
>
> On Thu, 4 Jul 2019 at 16:25, Ilya Maximets <[email protected]> wrote:
> >
> > Completion queue address reservation could not be undone.
> > In case of bad 'queue_id' or skb allocation failure, reserved entry
> > will be leaked reducing the total capacity of completion queue.
> >
> > Fix that by moving reservation to the point where failure is not
> > possible. Additionally, 'queue_id' checking moved out from the loop
> > since there is no point to check it there.
> >
>
> Good catch, Ilya! Thanks for the patch!
>
> Acked-by: Björn Töpel <[email protected]>
>
Thanks
Tested-by: William Tu <[email protected]>

2019-07-12 13:08:21

by Daniel Borkmann

Subject: Re: [PATCH bpf] xdp: fix possible cq entry leak

On 07/04/2019 04:25 PM, Ilya Maximets wrote:
> Completion queue address reservation could not be undone.
> In case of bad 'queue_id' or skb allocation failure, reserved entry
> will be leaked reducing the total capacity of completion queue.
>
> Fix that by moving reservation to the point where failure is not
> possible. Additionally, 'queue_id' checking moved out from the loop
> since there is no point to check it there.
>
> Fixes: 35fcde7f8deb ("xsk: support for Tx")
> Signed-off-by: Ilya Maximets <[email protected]>

Applied, thanks!