[bug report] virtio_net: xsk: bind/unbind xsk for rx

[bug report] virtio_net: xsk: bind/unbind xsk for rx

[Dan Carpenter]

Hello Xuan Zhuo,

Commit 09d2b3182c8e ("virtio_net: xsk: bind/unbind xsk for rx") from
Jul 8, 2024 (linux-next), leads to the following Smatch static
checker warning:

	drivers/net/virtio_net.c:5893 virtnet_xsk_pool_enable()
	warn: potential bounds check after use 'qid'

drivers/net/virtio_net.c
    5873 static int virtnet_xsk_pool_enable(struct net_device *dev,
    5874                                    struct xsk_buff_pool *pool,
    5875                                    u16 qid)
    5876 {
    5877         struct virtnet_info *vi = netdev_priv(dev);
    5878         struct receive_queue *rq;
    5879         struct device *dma_dev;
    5880         struct send_queue *sq;
    5881         dma_addr_t hdr_dma;
    5882         int err, size;
    5883 
    5884         if (vi->hdr_len > xsk_pool_get_headroom(pool))
    5885                 return -EINVAL;
    5886 
    5887         /* In big_packets mode, xdp cannot work, so there is no need to
    5888          * initialize xsk of rq.
    5889          */
    5890         if (!vi->rq[qid].page_pool)
                             ^^^
flip these two checks around

    5891                 return -ENOENT;
    5892 
--> 5893         if (qid >= vi->curr_queue_pairs)
                     ^^^
So that we check that qid is valid before using it as an index into the
->rq[] array?

    5894                 return -EINVAL;
    5895 
    5896         sq = &vi->sq[qid];
    5897         rq = &vi->rq[qid];
    5898 
    5899         /* xsk assumes that tx and rx must have the same dma device. The af-xdp

This email is a free service from the Smatch-CI project [smatch.sf.net].

regards,
dan carpenter