On 11/19/25 3:18 AM, Michael S. Tsirkin wrote:
> On Tue, Nov 18, 2025 at 04:31:09PM -0500, Michael S. Tsirkin wrote:
>>> +static int setup_ip_key_mask(struct virtio_net_ff_selector *selector,
>>> + u8 *key,
>>> + const struct ethtool_rx_flow_spec *fs)
>>> +{
>>> + struct iphdr *v4_m = (struct iphdr *)&selector->mask;
>>> + struct iphdr *v4_k = (struct iphdr *)key;
>>> +
>>> + selector->type = VIRTIO_NET_FF_MASK_TYPE_IPV4;
>>> + selector->length = sizeof(struct iphdr);
>>> +
>>> + if (fs->h_u.usr_ip4_spec.l4_4_bytes ||
>>> + fs->h_u.usr_ip4_spec.tos ||
>>> + fs->h_u.usr_ip4_spec.ip_ver != ETH_RX_NFC_IP4)
>>> + return -EOPNOTSUPP;
>>
>> So include/uapi/linux/ethtool.h says:
>>
>> * struct ethtool_usrip4_spec - general flow specification for IPv4
>> * @ip4src: Source host
>> * @ip4dst: Destination host
>> * @l4_4_bytes: First 4 bytes of transport (layer 4) header
>> * @tos: Type-of-service
>> * @ip_ver: Value must be %ETH_RX_NFC_IP4; mask must be 0
>> * @proto: Transport protocol number; mask must be 0
>>
>> I guess this ETH_RX_NFC_IP4 check validates that userspace follows this
>> documentation? But then shouldn't you check the mask
>> as well? and mask for proto?
>>
>>
>>
>
> in fact, what if e.g. tos is 0 but mask is non-zero? should not
> this be rejected, too?
>
Actually the tos check should be removed, there's no guidance it should
be 0, like the other fields. Our hardware doesn't support it, but this
will be caught in validate_classifier_selectors.
