Re: [PATCH net v3] usbnet: fix EVENT_RX_PAUSED race in usbnet_skb_return()

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] net: out-of-order packet delivery during RX resume
--

commit 7148b0b2b0b945f31b2120d7f4cbc269acf91635
Author: Ömer Mete Kaya <omermetekaya0@gmail.com>

usbnet: fix EVENT_RX_PAUSED race in usbnet_skb_return()

This commit addresses a race condition in usbnet_skb_return() by protecting
the EVENT_RX_PAUSED flag check and enqueue operations with a double-check
pattern under the rxq_pause lock. It also protects the clear_bit() operation
in usbnet_resume_rx() under the same lock.

> diff --git a/drivers/net/usb/usbnet.c b/drivers/net/usb/usbnet.c
> index a19ecf718f36a..d1ec3ba1e10e6 100644
> --- a/drivers/net/usb/usbnet.c
> +++ b/drivers/net/usb/usbnet.c
[ ... ]
> @@ -700,15 +705,16 @@ void usbnet_resume_rx(struct usbnet *dev)
>  	int num = 0;
>  
>  	local_bh_disable();
> +	spin_lock(&dev->rxq_pause.lock);
>  	clear_bit(EVENT_RX_PAUSED, &dev->flags);
> -
> +	spin_unlock(&dev->rxq_pause.lock);

[Severity: Medium]
If a new packet arrives concurrently on another CPU via usbnet_skb_return(),
could it observe EVENT_RX_PAUSED as false here?

If so, it would bypass the pause queue and immediately deliver the new SKB
to the network stack via netif_rx().

>  	while ((skb = skb_dequeue(&dev->rxq_pause)) != NULL) {
>  		usbnet_skb_return(dev, skb);
>  		num++;
>  	}

This seems like it would cause the newly arrived packet to be delivered before
the older packets in the rxq_pause queue are fully drained by this while loop,
leading to out-of-order delivery which can degrade TCP performance.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261007220311.113827-1-omermetekaya0@gmail.com?part=1