The Linux Kernel Mailing List
 help / color / mirror / Atom feed
From: Wang Liang <wangliang74@huawei.com>
To: Willem de Bruijn <willemdebruijn.kernel@gmail.com>,
	<mst@redhat.com>, <jasowang@redhat.com>,
	<xuanzhuo@linux.alibaba.com>, <eperezma@redhat.com>,
	<pabeni@redhat.com>, <davem@davemloft.net>, <willemb@google.com>,
	<atenart@kernel.org>
Cc: <yuehaibing@huawei.com>, <zhangchangzhong@huawei.com>,
	<netdev@vger.kernel.org>, <virtualization@lists.linux.dev>,
	<linux-kernel@vger.kernel.org>, <steffen.klassert@secunet.com>
Subject: Re: [PATCH net] net: check the minimum value of gso size in virtio_net_hdr_to_skb()
Date: Fri, 25 Jul 2025 15:55:15 +0800	[thread overview]
Message-ID: <bef878c0-4d7f-4e9a-a05d-30f6fde31e3c@huawei.com> (raw)
In-Reply-To: <688235273230f_39271d29430@willemb.c.googlers.com.notmuch>


在 2025/7/24 21:29, Willem de Bruijn 写道:
> Wang Liang wrote:
>> When sending a packet with virtio_net_hdr to tun device, if the gso_type
>> in virtio_net_hdr is SKB_GSO_UDP and the gso_size is less than udphdr
>> size, below crash may happen.
>>
> gso_size is the size of the segment payload, excluding the transport
> header.
>
> This is probably not the right approach.
>
> Not sure how a GSO skb can be built that is shorter than even the
> transport header. Maybe an skb_dump of the GSO skb can be elucidating.
>>   			return -EINVAL;
>>   
>>   		/* Too small packets are not really GSO ones. */
>> -- 
>> 2.34.1
>>

Thanks for your review!

Here is the skb_dump result:

     skb len=4 headroom=98 headlen=4 tailroom=282
     mac=(64,14) mac_len=14 net=(78,20) trans=98
     shinfo(txflags=0 nr_frags=0 gso(size=0 type=0 segs=0))
     csum(0x8c start=140 offset=0 ip_summed=1 complete_sw=0 valid=1 level=0)
     hash(0x0 sw=0 l4=0) proto=0x0800 pkttype=2 iif=4
     priority=0x0 mark=0x0 alloc_cpu=0 vlan_all=0x0
     encapsulation=0 inner(proto=0x0000, mac=0, net=0, trans=0)
     dev name=tun0 feat=0x00002000000048c1
     skb headroom: 00000000: 20 00 00 00 10 00 05 00 c4 2c 83 68 00 00 00 00
     skb headroom: 00000010: 00 00 00 00 04 00 00 00 01 00 00 00 01 00 00 00
     skb headroom: 00000020: 08 00 1d 00 09 00 00 00 09 00 03 00 74 75 6e 30
     skb headroom: 00000030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb headroom: 00000040: 01 80 c2 00 00 00 ff ff ff ff ff ff 08 00 45 00
     skb headroom: 00000050: 00 18 00 00 20 00 00 11 ba d4 00 00 00 00 e0 00
     skb headroom: 00000060: 00 01
     skb linear:   00000000: 00 9c d0 90
     skb tailroom: 00000000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 00000030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 00000040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 00000050: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 00000060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 00000070: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 00000080: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 00000090: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 000000a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 000000b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 000000c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 000000d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 000000e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 000000f0: 00 00 00 00 00 00 00 00 00 00 00 0b 0e 04 80 88
     skb tailroom: 00000100: ff ff 00 00 00 00 00 00 00 00 00 00 00 00 00 00
     skb tailroom: 00000110: 00 00 00 00 00 00 00 00 00 00

The following C code can reproduce this issue:

     #include <string.h>
     #include <fcntl.h>
     #include <sys/ioctl.h>
     #include <linux/if.h>
     #include <linux/if_tun.h>
     #include <linux/virtio_net.h>
     #include <netinet/ip.h>
     #include <netinet/udp.h>

     int main(void)
     {
         // create udp socket, set option UDP_ENCAP_ESPINUDP
         int udp_fd = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP);
         struct sockaddr_in server_addr = {
             .sin_family = AF_INET,
             .sin_port = htons(20004),
             .sin_addr.s_addr = inet_addr("224.0.0.1"),
         };
         bind(udp_fd, (struct sockaddr *)&server_addr, sizeof(server_addr));
         int val = UDP_ENCAP_ESPINUDP;
         setsockopt(udp_fd, IPPROTO_UDP, UDP_ENCAP, &val, sizeof(val));

         // send udp packet to tun/tap dev
         system("ip tuntap add dev tun0 mode tap");
         system("ip link set dev tun0 up");
         int fd = open("/dev/net/tun", O_RDWR);
         struct ifreq ifr = {
             .ifr_flags = IFF_TAP | IFF_NAPI | IFF_NAPI_FRAGS | 
IFF_ONE_QUEUE | IFF_VNET_HDR,
             .ifr_name  = "tun0",
         };
         ioctl(fd, TUNSETIFF, &ifr);
         struct tun_pi pi = { 0 };
         struct virtio_net_hdr gso = {
             .flags = VIRTIO_NET_HDR_F_NEEDS_CSUM,
             .gso_type = VIRTIO_NET_HDR_GSO_UDP,
             .hdr_len = 64,
             .gso_size = 4,
             .csum_start = 76,
             .csum_offset = 0,
         };
         struct ethhdr eth = {
             .h_dest   = {0x01, 0x80, 0xc2, 0x00, 0x00, 0x00},
             .h_source = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
             .h_proto  = htons(0x0800),
         };
         struct iphdr iph = {
             .version = 4,
             .ihl = 5,
             .tot_len = htons(176),
             .protocol = IPPROTO_UDP,
             .saddr = 0,
             .daddr = inet_addr("224.0.0.1"),
             .check = 0x3cda,
         };
         struct udphdr uh = {
             .source = 0,
             .dest = htons(20004),
             .len = htons(156),
             .check = 0x2363,
         };
         char buf[204] = { 0 };
         memcpy(buf, &pi, 4);
         memcpy(buf + 4, &gso, 10);
         memcpy(buf + 14, &eth, 14);
         memcpy(buf + 28, &iph, 20);
         memcpy(buf + 48, &uh, 8);
         write(fd, buf, sizeof(buf));
         close(fd);
         return 0;
     }

------
Best regards
Wang Liang

>

  reply	other threads:[~2025-07-25  7:55 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-07-24  8:30 [PATCH net] net: check the minimum value of gso size in virtio_net_hdr_to_skb() Wang Liang
2025-07-24 13:29 ` Willem de Bruijn
2025-07-25  7:55   ` Wang Liang [this message]
2025-07-27 16:36     ` Willem de Bruijn
2025-07-27 17:15       ` Willem de Bruijn
2025-07-28  3:21         ` Jason Wang
2025-07-28  3:50           ` Willem de Bruijn
2025-07-29  2:40             ` Jason Wang
2025-07-28  3:36         ` Wang Liang
2025-07-28 13:45           ` Willem de Bruijn
2025-07-25  8:55 ` Xuan Zhuo
2025-07-29 12:22   ` Wang Liang

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=bef878c0-4d7f-4e9a-a05d-30f6fde31e3c@huawei.com \
    --to=wangliang74@huawei.com \
    --cc=atenart@kernel.org \
    --cc=davem@davemloft.net \
    --cc=eperezma@redhat.com \
    --cc=jasowang@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mst@redhat.com \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=steffen.klassert@secunet.com \
    --cc=virtualization@lists.linux.dev \
    --cc=willemb@google.com \
    --cc=willemdebruijn.kernel@gmail.com \
    --cc=xuanzhuo@linux.alibaba.com \
    --cc=yuehaibing@huawei.com \
    --cc=zhangchangzhong@huawei.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox