I have written a small device driver [1] which is used to provide a virtual NIC to a user-level protocol stack. In order to pass data between the application process and the kernel a shared pool of page aligned buffers is used. When the application process wishes to send a packet it constructs the packet in one of these shared buffers and then writes the buffer number to a character device. After getting a pointer to the page (in buf) the following code is used to transmit the packet (this works fine). void *buf <-- the buffer. skb = dev_alloc_skb(buf_len); memcpy(skb_put(skb, buf_len), buf, buf_len); skb->dev = dev; skb->protocol = eth_type_trans(skb, dev); skb->ip_summed = CHECKSUM_UNNECESSARY; netif_rx_ni(skb); However, since buf is actually a pointer to a page (allocated with get_free_page) I would like to avoid the extra copy and do something like the following (which doesn't work): skb = dev_alloc_skb(buf_len); // This wastes buf_len bytes. skb_shinfo(skb)->nr_frags = 1; skb_shinfo(skb)->frags[0].page = buf; skb_shinfo(skb)->frags[0].page_offset = ETH_HLEN; skb_shinfo(skb)->frags[0].size = buf_len - ETH_HLEN; skb->len = buf_len - ETH_HLEN; skb->data_len = buf_len - ETH_HLEN; skb->truesize += (buf_len - ETH_HLEN); skb->dev = dev; skb->protocol = htons(ETH_P_IP); skb->ip_summed = CHECKSUM_UNNECESSARY; netif_rx_ni(skb); This always dies with the following stack trace (Complete oops in the form of a PNG screencap of the qemu session is available at http://www.coverfire.com/tmp/oops.png): __pskb_pull_tail ip_rcv netif_receive_skb process_backlog net_rx_action __do_softirq do_softirg Questions: 1) Is it OK if an SKB only contains data in the page frags and no data in the main buffer? 2) Any clue what I am doing wrong? The complete code (without the page stuff I'm trying to get going above) can be found at the below URL. Any pointers would be greatly appreciated. Thanks. [1] - http://projects.coverfire.com/pnet/ -- Dan Siemon <dan@xxxxxxxxxxxxx> OpenPGP: http://www.coverfire.com/files/pubkey.txt Fingerprint: FB0A 2D8A A1E9 11B6 6CA3 0C53 742A 9EA8 891C BD98 -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
