diff mbox series

[v3] net/packet: support mergeable feature of virtio

Message ID 1678168911-337042-1-git-send-email-amy.saq@antgroup.com (mailing list archive)
State Superseded
Delegated to: Netdev Maintainers
Headers show
Series [v3] net/packet: support mergeable feature of virtio | expand

Checks

Context Check Description
netdev/series_format warning Single patches do not need cover letters; Target tree name not specified in the subject
netdev/tree_selection success Guessed tree name to be net-next
netdev/fixes_present success Fixes tag not required for -next series
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 5332 this patch: 5332
netdev/cc_maintainers warning 3 maintainers not CCed: kuba@kernel.org edumazet@google.com pabeni@redhat.com
netdev/build_clang success Errors and warnings before: 1074 this patch: 1074
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/deprecated_api success None detected
netdev/check_selftest success No net selftest shell script
netdev/verify_fixes success No Fixes tag
netdev/build_allmodconfig_warn success Errors and warnings before: 5540 this patch: 5540
netdev/checkpatch warning WARNING: line length of 82 exceeds 80 columns WARNING: line length of 94 exceeds 80 columns WARNING: line length of 96 exceeds 80 columns
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline success Was 0 now: 0

Commit Message

沈安琪(凛玥) March 7, 2023, 6:01 a.m. UTC
From: Jianfeng Tan <henry.tjf@antgroup.com>

Packet sockets, like tap, can be used as the backend for kernel vhost.
In packet sockets, virtio net header size is currently hardcoded to be
the size of struct virtio_net_hdr, which is 10 bytes; however, it is not
always the case: some virtio features, such as mrg_rxbuf, need virtio
net header to be 12-byte long.

Mergeable buffers, as a virtio feature, is worthy of supporting: packets
that are larger than one-mbuf size will be dropped in vhost worker's
handle_rx if mrg_rxbuf feature is not used, but large packets
cannot be avoided and increasing mbuf's size is not economical.

With this mergeable feature enabled by virtio-user, packet sockets with
hardcoded 10-byte virtio net header will parse mac head incorrectly in
packet_snd by taking the last two bytes of virtio net header as part of
mac header.
This incorrect mac header parsing will cause packet to be dropped due to
invalid ether head checking in later under-layer device packet receiving.

By adding extra field vnet_hdr_sz with utilizing holes in struct
packet_sock to record currently used virtio net header size and supporting
extra sockopt PACKET_VNET_HDR_SZ to set specified vnet_hdr_sz, packet
sockets can know the exact length of virtio net header that virtio user
gives.
In packet_snd, tpacket_snd and packet_recvmsg, instead of using
hardcoded virtio net header size, it can get the exact vnet_hdr_sz from
corresponding packet_sock, and parse mac header correctly based on this
information to avoid the packets being mistakenly dropped.

Besides, has_vnet_hdr field in struct packet_sock is removed since all 
the information it provides is covered by vnet_hdr_sz field: a packet
socket has a vnet header if and only if its vnet_hdr_sz is not zero.

Signed-off-by: Jianfeng Tan <henry.tjf@antgroup.com>
Co-developed-by: Anqi Shen <amy.saq@antgroup.com>
Signed-off-by: Anqi Shen <amy.saq@antgroup.com>
---

V2 -> V3:
* remove has_vnet_hdr field and use vnet_hdr_sz to indicate whether
there is a vnet header;
* refactor PACKET_VNET_HDR and PACKET_VNET_HDR_SZ sockopt to remove
redundant code.

 include/uapi/linux/if_packet.h |  1 +
 net/packet/af_packet.c         | 82 ++++++++++++++++++++++++++++--------------
 net/packet/diag.c              |  2 +-
 net/packet/internal.h          |  4 +--
 4 files changed, 60 insertions(+), 29 deletions(-)

Comments

Willem de Bruijn March 7, 2023, 3:49 p.m. UTC | #1
沈安琪(凛玥) wrote:
> From: Jianfeng Tan <henry.tjf@antgroup.com>
> 
> Packet sockets, like tap, can be used as the backend for kernel vhost.
> In packet sockets, virtio net header size is currently hardcoded to be
> the size of struct virtio_net_hdr, which is 10 bytes; however, it is not
> always the case: some virtio features, such as mrg_rxbuf, need virtio
> net header to be 12-byte long.
> 
> Mergeable buffers, as a virtio feature, is worthy of supporting: packets
> that are larger than one-mbuf size will be dropped in vhost worker's
> handle_rx if mrg_rxbuf feature is not used, but large packets
> cannot be avoided and increasing mbuf's size is not economical.
> 
> With this mergeable feature enabled by virtio-user, packet sockets with
> hardcoded 10-byte virtio net header will parse mac head incorrectly in
> packet_snd by taking the last two bytes of virtio net header as part of
> mac header.
> This incorrect mac header parsing will cause packet to be dropped due to
> invalid ether head checking in later under-layer device packet receiving.
> 
> By adding extra field vnet_hdr_sz with utilizing holes in struct
> packet_sock to record currently used virtio net header size and supporting
> extra sockopt PACKET_VNET_HDR_SZ to set specified vnet_hdr_sz, packet
> sockets can know the exact length of virtio net header that virtio user
> gives.
> In packet_snd, tpacket_snd and packet_recvmsg, instead of using
> hardcoded virtio net header size, it can get the exact vnet_hdr_sz from
> corresponding packet_sock, and parse mac header correctly based on this
> information to avoid the packets being mistakenly dropped.
> 
> Besides, has_vnet_hdr field in struct packet_sock is removed since all 
> the information it provides is covered by vnet_hdr_sz field: a packet
> socket has a vnet header if and only if its vnet_hdr_sz is not zero.
> 
> Signed-off-by: Jianfeng Tan <henry.tjf@antgroup.com>
> Co-developed-by: Anqi Shen <amy.saq@antgroup.com>
> Signed-off-by: Anqi Shen <amy.saq@antgroup.com>
> ---
> 
>  /*
> @@ -2310,8 +2310,8 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev,
>  		netoff = TPACKET_ALIGN(po->tp_hdrlen +
>  				       (maclen < 16 ? 16 : maclen)) +
>  				       po->tp_reserve;
> -		if (po->has_vnet_hdr) {
> -			netoff += sizeof(struct virtio_net_hdr);
> +		if (po->vnet_hdr_sz != 0) {
> +			netoff += po->vnet_hdr_sz;

no need to test != 0 here and elsewhere, just if (po->vnet_hdr_sz)

>  			do_vnet = true;
>  		}
>  		macoff = netoff - maclen;
> @@ -2552,16 +2552,27 @@ static int __packet_snd_vnet_parse(struct virtio_net_hdr *vnet_hdr, size_t len)
>  }
>  
>  static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb,
> @@ -2730,6 +2741,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>  	int status = TP_STATUS_AVAILABLE;
>  	int hlen, tlen, copylen = 0;
>  	long timeo = 0;
> +	int vnet_hdr_sz;

since po->vnet_hdr_sz is touched in the hot path anyway, initialize
it here and use the local var everywhere. Else this might race with
updates too header length.
>  
>  	mutex_lock(&po->pg_vec_lock);
>  
> @@ -2780,7 +2792,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>  	size_max = po->tx_ring.frame_size
>  		- (po->tp_hdrlen - sizeof(struct sockaddr_ll));
>  
> -	if ((size_max > dev->mtu + reserve + VLAN_HLEN) && !po->has_vnet_hdr)
> +	if ((size_max > dev->mtu + reserve + VLAN_HLEN) && po->vnet_hdr_sz == 0)
>  		size_max = dev->mtu + reserve + VLAN_HLEN;
>  
>  	reinit_completion(&po->skb_completion);
> @@ -2809,10 +2821,11 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>  		status = TP_STATUS_SEND_REQUEST;
>  		hlen = LL_RESERVED_SPACE(dev);
>  		tlen = dev->needed_tailroom;
> -		if (po->has_vnet_hdr) {
> +		if (po->vnet_hdr_sz != 0) {
>  			vnet_hdr = data;
> -			data += sizeof(*vnet_hdr);
> -			tp_len -= sizeof(*vnet_hdr);
> +			vnet_hdr_sz = po->vnet_hdr_sz;

.. in particularly here, where vnet_hdr_sz is first checked and then copied
to a local var in a separate step.

> +			data += vnet_hdr_sz;
> +			tp_len -= vnet_hdr_sz;
>  			if (tp_len < 0 ||
>  			    __packet_snd_vnet_parse(vnet_hdr, tp_len)) {
>  				tp_len = -EINVAL;
> @@ -2837,7 +2850,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>  					  addr, hlen, copylen, &sockc);
>  		if (likely(tp_len >= 0) &&
>  		    tp_len > dev->mtu + reserve &&
> -		    !po->has_vnet_hdr &&
> +		    (po->vnet_hdr_sz == 0) &&
>  		    !packet_extra_vlan_len_allowed(dev, skb))
>  			tp_len = -EMSGSIZE;
>  
> @@ -2856,7 +2869,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>  			}
>  		}
>  
> -		if (po->has_vnet_hdr) {
> +		if (po->vnet_hdr_sz != 0) {
>  			if (virtio_net_hdr_to_skb(skb, vnet_hdr, vio_le())) {
>  				tp_len = -EINVAL;
>  				goto tpacket_error;
> @@ -2947,6 +2960,7 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
>  	int offset = 0;
>  	struct packet_sock *po = pkt_sk(sk);
>  	bool has_vnet_hdr = false;

remove has_vnet_hdr, which is now superfluous.

> +	int vnet_hdr_sz;
>  	int hlen, tlen, linear;
>  	int extra_len = 0;
>  
> @@ -2990,8 +3004,9 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
>  
>  	if (sock->type == SOCK_RAW)
>  		reserve = dev->hard_header_len;
> -	if (po->has_vnet_hdr) {
> -		err = packet_snd_vnet_parse(msg, &len, &vnet_hdr);
> +	if (po->vnet_hdr_sz != 0) {
> +		vnet_hdr_sz = po->vnet_hdr_sz;

same here

> +		err = packet_snd_vnet_parse(msg, &len, &vnet_hdr, vnet_hdr_sz);
>  		if (err)
>  			goto out_unlock;
>  		has_vnet_hdr = true;
> @@ -3068,7 +3083,7 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
>  		err = virtio_net_hdr_to_skb(skb, &vnet_hdr, vio_le());
>  		if (err)
>  			goto out_free;
> -		len += sizeof(vnet_hdr);
> +		len += vnet_hdr_sz;
>  		virtio_net_hdr_set_proto(skb, &vnet_hdr);
>  	}
>  
> @@ -3451,11 +3466,11 @@ static int packet_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
>  
>  	packet_rcv_try_clear_pressure(pkt_sk(sk));
>  
> -	if (pkt_sk(sk)->has_vnet_hdr) {
> -		err = packet_rcv_vnet(msg, skb, &len);
> +	if (pkt_sk(sk)->vnet_hdr_sz != 0) {
> +		vnet_hdr_len = pkt_sk(sk)->vnet_hdr_sz;

and here

> +		err = packet_rcv_vnet(msg, skb, &len, vnet_hdr_len);
>  		if (err)
>  			goto out_free;
> -		vnet_hdr_len = sizeof(struct virtio_net_hdr);
>  	}
>  
>  	/* You lose any data beyond the buffer you gave. If it worries
> @@ -3921,8 +3936,10 @@ static void packet_flush_mclist(struct sock *sk)
>  		return 0;
>  	}
>  	case PACKET_VNET_HDR:
> +	case PACKET_VNET_HDR_SZ:
>  	{
>  		int val;
> +		int hdr_len = 0;
>  
>  		if (sock->type != SOCK_RAW)
>  			return -EINVAL;
> @@ -3931,11 +3948,21 @@ static void packet_flush_mclist(struct sock *sk)
>  		if (copy_from_sockptr(&val, optval, sizeof(val)))
>  			return -EFAULT;
>  
> +		if (optname == PACKET_VNET_HDR_SZ) {
> +			if (val != sizeof(struct virtio_net_hdr) &&
> +			    val != sizeof(struct virtio_net_hdr_mrg_rxbuf))
> +				return -EINVAL;
> +			hdr_len = val;

val == 0 is also a a correct value

> +		} else {
> +			if (!!val)
> +				hdr_len = sizeof(struct virtio_net_hdr);

no need for !! when testing non-zero.

and instead of initializing to zero on initial variable definition:

    hdr_len = val ? sizeof(struct virtio_net_hdr) : 0;

> +		}
> +
>  		lock_sock(sk);
>  		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) {
>  			ret = -EBUSY;
>  		} else {
> -			po->has_vnet_hdr = !!val;
> +			po->vnet_hdr_sz = hdr_len;
>  			ret = 0;
>  		}
>  		release_sock(sk);
> @@ -4068,7 +4095,10 @@ static int packet_getsockopt(struct socket *sock, int level, int optname,
>  		val = po->origdev;
>  		break;
>  	case PACKET_VNET_HDR:
> -		val = po->has_vnet_hdr;
> +		val = !!po->vnet_hdr_sz;
> +		break;
> +	case PACKET_VNET_HDR_SZ:
> +		val = po->vnet_hdr_sz;
>  		break;
>  	case PACKET_VERSION:
>  		val = po->tp_version;
> diff --git a/net/packet/diag.c b/net/packet/diag.c
> index 07812ae..4e544da 100644
> --- a/net/packet/diag.c
> +++ b/net/packet/diag.c
> @@ -27,7 +27,7 @@ static int pdiag_put_info(const struct packet_sock *po, struct sk_buff *nlskb)
>  		pinfo.pdi_flags |= PDI_AUXDATA;
>  	if (po->origdev)
>  		pinfo.pdi_flags |= PDI_ORIGDEV;
> -	if (po->has_vnet_hdr)
> +	if (po->vnet_hdr_sz != 0)
>  		pinfo.pdi_flags |= PDI_VNETHDR;
>  	if (po->tp_loss)
>  		pinfo.pdi_flags |= PDI_LOSS;
> diff --git a/net/packet/internal.h b/net/packet/internal.h
> index 48af35b..9b52d93 100644
> --- a/net/packet/internal.h
> +++ b/net/packet/internal.h
> @@ -119,9 +119,9 @@ struct packet_sock {
>  	unsigned int		running;	/* bind_lock must be held */
>  	unsigned int		auxdata:1,	/* writer must hold sock lock */
>  				origdev:1,
> -				has_vnet_hdr:1,
>  				tp_loss:1,
> -				tp_tx_has_off:1;
> +				tp_tx_has_off:1,
> +				vnet_hdr_sz:8;

just a separate u8 variable , rather than 8 bits in a u32.

>  	int			pressure;
>  	int			ifindex;	/* bound device		*/
>  	__be16			num;
> -- 
> 1.8.3.1
>
沈安琪(凛玥) March 9, 2023, 9:47 a.m. UTC | #2
在 2023/3/7 下午11:49, Willem de Bruijn 写道:
> 沈安琪(凛玥) wrote:
>> From: Jianfeng Tan <henry.tjf@antgroup.com>
>>
>> Packet sockets, like tap, can be used as the backend for kernel vhost.
>> In packet sockets, virtio net header size is currently hardcoded to be
>> the size of struct virtio_net_hdr, which is 10 bytes; however, it is not
>> always the case: some virtio features, such as mrg_rxbuf, need virtio
>> net header to be 12-byte long.
>>
>> Mergeable buffers, as a virtio feature, is worthy of supporting: packets
>> that are larger than one-mbuf size will be dropped in vhost worker's
>> handle_rx if mrg_rxbuf feature is not used, but large packets
>> cannot be avoided and increasing mbuf's size is not economical.
>>
>> With this mergeable feature enabled by virtio-user, packet sockets with
>> hardcoded 10-byte virtio net header will parse mac head incorrectly in
>> packet_snd by taking the last two bytes of virtio net header as part of
>> mac header.
>> This incorrect mac header parsing will cause packet to be dropped due to
>> invalid ether head checking in later under-layer device packet receiving.
>>
>> By adding extra field vnet_hdr_sz with utilizing holes in struct
>> packet_sock to record currently used virtio net header size and supporting
>> extra sockopt PACKET_VNET_HDR_SZ to set specified vnet_hdr_sz, packet
>> sockets can know the exact length of virtio net header that virtio user
>> gives.
>> In packet_snd, tpacket_snd and packet_recvmsg, instead of using
>> hardcoded virtio net header size, it can get the exact vnet_hdr_sz from
>> corresponding packet_sock, and parse mac header correctly based on this
>> information to avoid the packets being mistakenly dropped.
>>
>> Besides, has_vnet_hdr field in struct packet_sock is removed since all
>> the information it provides is covered by vnet_hdr_sz field: a packet
>> socket has a vnet header if and only if its vnet_hdr_sz is not zero.
>>
>> Signed-off-by: Jianfeng Tan <henry.tjf@antgroup.com>
>> Co-developed-by: Anqi Shen <amy.saq@antgroup.com>
>> Signed-off-by: Anqi Shen <amy.saq@antgroup.com>
>> ---
>>
>>   /*
>> @@ -2310,8 +2310,8 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev,
>>   		netoff = TPACKET_ALIGN(po->tp_hdrlen +
>>   				       (maclen < 16 ? 16 : maclen)) +
>>   				       po->tp_reserve;
>> -		if (po->has_vnet_hdr) {
>> -			netoff += sizeof(struct virtio_net_hdr);
>> +		if (po->vnet_hdr_sz != 0) {
>> +			netoff += po->vnet_hdr_sz;
> no need to test != 0 here and elsewhere, just if (po->vnet_hdr_sz)
>
>>   			do_vnet = true;
>>   		}
>>   		macoff = netoff - maclen;
>> @@ -2552,16 +2552,27 @@ static int __packet_snd_vnet_parse(struct virtio_net_hdr *vnet_hdr, size_t len)
>>   }
>>   
>>   static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb,
>> @@ -2730,6 +2741,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>>   	int status = TP_STATUS_AVAILABLE;
>>   	int hlen, tlen, copylen = 0;
>>   	long timeo = 0;
>> +	int vnet_hdr_sz;
> since po->vnet_hdr_sz is touched in the hot path anyway, initialize
> it here and use the local var everywhere. Else this might race with
> updates too header length.
>>   
>>   	mutex_lock(&po->pg_vec_lock);
>>   
>> @@ -2780,7 +2792,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>>   	size_max = po->tx_ring.frame_size
>>   		- (po->tp_hdrlen - sizeof(struct sockaddr_ll));
>>   
>> -	if ((size_max > dev->mtu + reserve + VLAN_HLEN) && !po->has_vnet_hdr)
>> +	if ((size_max > dev->mtu + reserve + VLAN_HLEN) && po->vnet_hdr_sz == 0)
>>   		size_max = dev->mtu + reserve + VLAN_HLEN;
>>   
>>   	reinit_completion(&po->skb_completion);
>> @@ -2809,10 +2821,11 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>>   		status = TP_STATUS_SEND_REQUEST;
>>   		hlen = LL_RESERVED_SPACE(dev);
>>   		tlen = dev->needed_tailroom;
>> -		if (po->has_vnet_hdr) {
>> +		if (po->vnet_hdr_sz != 0) {
>>   			vnet_hdr = data;
>> -			data += sizeof(*vnet_hdr);
>> -			tp_len -= sizeof(*vnet_hdr);
>> +			vnet_hdr_sz = po->vnet_hdr_sz;
> .. in particularly here, where vnet_hdr_sz is first checked and then copied
> to a local var in a separate step.
>
>> +			data += vnet_hdr_sz;
>> +			tp_len -= vnet_hdr_sz;
>>   			if (tp_len < 0 ||
>>   			    __packet_snd_vnet_parse(vnet_hdr, tp_len)) {
>>   				tp_len = -EINVAL;
>> @@ -2837,7 +2850,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>>   					  addr, hlen, copylen, &sockc);
>>   		if (likely(tp_len >= 0) &&
>>   		    tp_len > dev->mtu + reserve &&
>> -		    !po->has_vnet_hdr &&
>> +		    (po->vnet_hdr_sz == 0) &&
>>   		    !packet_extra_vlan_len_allowed(dev, skb))
>>   			tp_len = -EMSGSIZE;
>>   
>> @@ -2856,7 +2869,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>>   			}
>>   		}
>>   
>> -		if (po->has_vnet_hdr) {
>> +		if (po->vnet_hdr_sz != 0) {
>>   			if (virtio_net_hdr_to_skb(skb, vnet_hdr, vio_le())) {
>>   				tp_len = -EINVAL;
>>   				goto tpacket_error;
>> @@ -2947,6 +2960,7 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
>>   	int offset = 0;
>>   	struct packet_sock *po = pkt_sk(sk);
>>   	bool has_vnet_hdr = false;
> remove has_vnet_hdr, which is now superfluous.
>
>> +	int vnet_hdr_sz;
>>   	int hlen, tlen, linear;
>>   	int extra_len = 0;
>>   
>> @@ -2990,8 +3004,9 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
>>   
>>   	if (sock->type == SOCK_RAW)
>>   		reserve = dev->hard_header_len;
>> -	if (po->has_vnet_hdr) {
>> -		err = packet_snd_vnet_parse(msg, &len, &vnet_hdr);
>> +	if (po->vnet_hdr_sz != 0) {
>> +		vnet_hdr_sz = po->vnet_hdr_sz;
> same here
>
>> +		err = packet_snd_vnet_parse(msg, &len, &vnet_hdr, vnet_hdr_sz);
>>   		if (err)
>>   			goto out_unlock;
>>   		has_vnet_hdr = true;
>> @@ -3068,7 +3083,7 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
>>   		err = virtio_net_hdr_to_skb(skb, &vnet_hdr, vio_le());
>>   		if (err)
>>   			goto out_free;
>> -		len += sizeof(vnet_hdr);
>> +		len += vnet_hdr_sz;
>>   		virtio_net_hdr_set_proto(skb, &vnet_hdr);
>>   	}
>>   
>> @@ -3451,11 +3466,11 @@ static int packet_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
>>   
>>   	packet_rcv_try_clear_pressure(pkt_sk(sk));
>>   
>> -	if (pkt_sk(sk)->has_vnet_hdr) {
>> -		err = packet_rcv_vnet(msg, skb, &len);
>> +	if (pkt_sk(sk)->vnet_hdr_sz != 0) {
>> +		vnet_hdr_len = pkt_sk(sk)->vnet_hdr_sz;
> and here
>
>> +		err = packet_rcv_vnet(msg, skb, &len, vnet_hdr_len);
>>   		if (err)
>>   			goto out_free;
>> -		vnet_hdr_len = sizeof(struct virtio_net_hdr);
>>   	}
>>   
>>   	/* You lose any data beyond the buffer you gave. If it worries
>> @@ -3921,8 +3936,10 @@ static void packet_flush_mclist(struct sock *sk)
>>   		return 0;
>>   	}
>>   	case PACKET_VNET_HDR:
>> +	case PACKET_VNET_HDR_SZ:
>>   	{
>>   		int val;
>> +		int hdr_len = 0;
>>   
>>   		if (sock->type != SOCK_RAW)
>>   			return -EINVAL;
>> @@ -3931,11 +3948,21 @@ static void packet_flush_mclist(struct sock *sk)
>>   		if (copy_from_sockptr(&val, optval, sizeof(val)))
>>   			return -EFAULT;
>>   
>> +		if (optname == PACKET_VNET_HDR_SZ) {
>> +			if (val != sizeof(struct virtio_net_hdr) &&
>> +			    val != sizeof(struct virtio_net_hdr_mrg_rxbuf))
>> +				return -EINVAL;
>> +			hdr_len = val;
> val == 0 is also a a correct value
>
>> +		} else {
>> +			if (!!val)
>> +				hdr_len = sizeof(struct virtio_net_hdr);
> no need for !! when testing non-zero.
>
> and instead of initializing to zero on initial variable definition:
>
>      hdr_len = val ? sizeof(struct virtio_net_hdr) : 0;
>
>> +		}
>> +
>>   		lock_sock(sk);
>>   		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) {
>>   			ret = -EBUSY;
>>   		} else {
>> -			po->has_vnet_hdr = !!val;
>> +			po->vnet_hdr_sz = hdr_len;
>>   			ret = 0;
>>   		}
>>   		release_sock(sk);
>> @@ -4068,7 +4095,10 @@ static int packet_getsockopt(struct socket *sock, int level, int optname,
>>   		val = po->origdev;
>>   		break;
>>   	case PACKET_VNET_HDR:
>> -		val = po->has_vnet_hdr;
>> +		val = !!po->vnet_hdr_sz;
>> +		break;
>> +	case PACKET_VNET_HDR_SZ:
>> +		val = po->vnet_hdr_sz;
>>   		break;
>>   	case PACKET_VERSION:
>>   		val = po->tp_version;
>> diff --git a/net/packet/diag.c b/net/packet/diag.c
>> index 07812ae..4e544da 100644
>> --- a/net/packet/diag.c
>> +++ b/net/packet/diag.c
>> @@ -27,7 +27,7 @@ static int pdiag_put_info(const struct packet_sock *po, struct sk_buff *nlskb)
>>   		pinfo.pdi_flags |= PDI_AUXDATA;
>>   	if (po->origdev)
>>   		pinfo.pdi_flags |= PDI_ORIGDEV;
>> -	if (po->has_vnet_hdr)
>> +	if (po->vnet_hdr_sz != 0)
>>   		pinfo.pdi_flags |= PDI_VNETHDR;
>>   	if (po->tp_loss)
>>   		pinfo.pdi_flags |= PDI_LOSS;
>> diff --git a/net/packet/internal.h b/net/packet/internal.h
>> index 48af35b..9b52d93 100644
>> --- a/net/packet/internal.h
>> +++ b/net/packet/internal.h
>> @@ -119,9 +119,9 @@ struct packet_sock {
>>   	unsigned int		running;	/* bind_lock must be held */
>>   	unsigned int		auxdata:1,	/* writer must hold sock lock */
>>   				origdev:1,
>> -				has_vnet_hdr:1,
>>   				tp_loss:1,
>> -				tp_tx_has_off:1;
>> +				tp_tx_has_off:1,
>> +				vnet_hdr_sz:8;
> just a separate u8 variable , rather than 8 bits in a u32.
>
>>   	int			pressure;
>>   	int			ifindex;	/* bound device		*/
>>   	__be16			num;
>> -- 
>> 1.8.3.1
>>

Thanks for the comments! We will soon address these comments in the next 
version of this patch.

Before we send out the next version, we just wonder whether there is any 
other advice on this version of patch from Willem or Michael and we may 
address them as well in the next version. We really appreciate all the 
comments you provided and may provide in the future.
沈安琪(凛玥) March 9, 2023, 11:59 a.m. UTC | #3
在 2023/3/7 下午11:49, Willem de Bruijn 写道:
> 沈安琪(凛玥) wrote:
>> From: Jianfeng Tan <henry.tjf@antgroup.com>
>>
>> Packet sockets, like tap, can be used as the backend for kernel vhost.
>> In packet sockets, virtio net header size is currently hardcoded to be
>> the size of struct virtio_net_hdr, which is 10 bytes; however, it is not
>> always the case: some virtio features, such as mrg_rxbuf, need virtio
>> net header to be 12-byte long.
>>
>> Mergeable buffers, as a virtio feature, is worthy of supporting: packets
>> that are larger than one-mbuf size will be dropped in vhost worker's
>> handle_rx if mrg_rxbuf feature is not used, but large packets
>> cannot be avoided and increasing mbuf's size is not economical.
>>
>> With this mergeable feature enabled by virtio-user, packet sockets with
>> hardcoded 10-byte virtio net header will parse mac head incorrectly in
>> packet_snd by taking the last two bytes of virtio net header as part of
>> mac header.
>> This incorrect mac header parsing will cause packet to be dropped due to
>> invalid ether head checking in later under-layer device packet receiving.
>>
>> By adding extra field vnet_hdr_sz with utilizing holes in struct
>> packet_sock to record currently used virtio net header size and supporting
>> extra sockopt PACKET_VNET_HDR_SZ to set specified vnet_hdr_sz, packet
>> sockets can know the exact length of virtio net header that virtio user
>> gives.
>> In packet_snd, tpacket_snd and packet_recvmsg, instead of using
>> hardcoded virtio net header size, it can get the exact vnet_hdr_sz from
>> corresponding packet_sock, and parse mac header correctly based on this
>> information to avoid the packets being mistakenly dropped.
>>
>> Besides, has_vnet_hdr field in struct packet_sock is removed since all
>> the information it provides is covered by vnet_hdr_sz field: a packet
>> socket has a vnet header if and only if its vnet_hdr_sz is not zero.
>>
>> Signed-off-by: Jianfeng Tan <henry.tjf@antgroup.com>
>> Co-developed-by: Anqi Shen <amy.saq@antgroup.com>
>> Signed-off-by: Anqi Shen <amy.saq@antgroup.com>
>> ---
>>
>>   /*
>> @@ -2310,8 +2310,8 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev,
>>   		netoff = TPACKET_ALIGN(po->tp_hdrlen +
>>   				       (maclen < 16 ? 16 : maclen)) +
>>   				       po->tp_reserve;
>> -		if (po->has_vnet_hdr) {
>> -			netoff += sizeof(struct virtio_net_hdr);
>> +		if (po->vnet_hdr_sz != 0) {
>> +			netoff += po->vnet_hdr_sz;
> no need to test != 0 here and elsewhere, just if (po->vnet_hdr_sz)
>
>>   			do_vnet = true;
>>   		}
>>   		macoff = netoff - maclen;
>> @@ -2552,16 +2552,27 @@ static int __packet_snd_vnet_parse(struct virtio_net_hdr *vnet_hdr, size_t len)
>>   }
>>   
>>   static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb,
>> @@ -2730,6 +2741,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>>   	int status = TP_STATUS_AVAILABLE;
>>   	int hlen, tlen, copylen = 0;
>>   	long timeo = 0;
>> +	int vnet_hdr_sz;
> since po->vnet_hdr_sz is touched in the hot path anyway, initialize
> it here and use the local var everywhere. Else this might race with
> updates too header length.
>>   
>>   	mutex_lock(&po->pg_vec_lock);
>>   
>> @@ -2780,7 +2792,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>>   	size_max = po->tx_ring.frame_size
>>   		- (po->tp_hdrlen - sizeof(struct sockaddr_ll));
>>   
>> -	if ((size_max > dev->mtu + reserve + VLAN_HLEN) && !po->has_vnet_hdr)
>> +	if ((size_max > dev->mtu + reserve + VLAN_HLEN) && po->vnet_hdr_sz == 0)
>>   		size_max = dev->mtu + reserve + VLAN_HLEN;
>>   
>>   	reinit_completion(&po->skb_completion);
>> @@ -2809,10 +2821,11 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>>   		status = TP_STATUS_SEND_REQUEST;
>>   		hlen = LL_RESERVED_SPACE(dev);
>>   		tlen = dev->needed_tailroom;
>> -		if (po->has_vnet_hdr) {
>> +		if (po->vnet_hdr_sz != 0) {
>>   			vnet_hdr = data;
>> -			data += sizeof(*vnet_hdr);
>> -			tp_len -= sizeof(*vnet_hdr);
>> +			vnet_hdr_sz = po->vnet_hdr_sz;
> .. in particularly here, where vnet_hdr_sz is first checked and then copied
> to a local var in a separate step.
>
>> +			data += vnet_hdr_sz;
>> +			tp_len -= vnet_hdr_sz;
>>   			if (tp_len < 0 ||
>>   			    __packet_snd_vnet_parse(vnet_hdr, tp_len)) {
>>   				tp_len = -EINVAL;
>> @@ -2837,7 +2850,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>>   					  addr, hlen, copylen, &sockc);
>>   		if (likely(tp_len >= 0) &&
>>   		    tp_len > dev->mtu + reserve &&
>> -		    !po->has_vnet_hdr &&
>> +		    (po->vnet_hdr_sz == 0) &&
>>   		    !packet_extra_vlan_len_allowed(dev, skb))
>>   			tp_len = -EMSGSIZE;
>>   
>> @@ -2856,7 +2869,7 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
>>   			}
>>   		}
>>   
>> -		if (po->has_vnet_hdr) {
>> +		if (po->vnet_hdr_sz != 0) {
>>   			if (virtio_net_hdr_to_skb(skb, vnet_hdr, vio_le())) {
>>   				tp_len = -EINVAL;
>>   				goto tpacket_error;
>> @@ -2947,6 +2960,7 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
>>   	int offset = 0;
>>   	struct packet_sock *po = pkt_sk(sk);
>>   	bool has_vnet_hdr = false;
> remove has_vnet_hdr, which is now superfluous.
>
>> +	int vnet_hdr_sz;
>>   	int hlen, tlen, linear;
>>   	int extra_len = 0;
>>   
>> @@ -2990,8 +3004,9 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
>>   
>>   	if (sock->type == SOCK_RAW)
>>   		reserve = dev->hard_header_len;
>> -	if (po->has_vnet_hdr) {
>> -		err = packet_snd_vnet_parse(msg, &len, &vnet_hdr);
>> +	if (po->vnet_hdr_sz != 0) {
>> +		vnet_hdr_sz = po->vnet_hdr_sz;
> same here
>
>> +		err = packet_snd_vnet_parse(msg, &len, &vnet_hdr, vnet_hdr_sz);
>>   		if (err)
>>   			goto out_unlock;
>>   		has_vnet_hdr = true;
>> @@ -3068,7 +3083,7 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
>>   		err = virtio_net_hdr_to_skb(skb, &vnet_hdr, vio_le());
>>   		if (err)
>>   			goto out_free;
>> -		len += sizeof(vnet_hdr);
>> +		len += vnet_hdr_sz;
>>   		virtio_net_hdr_set_proto(skb, &vnet_hdr);
>>   	}
>>   
>> @@ -3451,11 +3466,11 @@ static int packet_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
>>   
>>   	packet_rcv_try_clear_pressure(pkt_sk(sk));
>>   
>> -	if (pkt_sk(sk)->has_vnet_hdr) {
>> -		err = packet_rcv_vnet(msg, skb, &len);
>> +	if (pkt_sk(sk)->vnet_hdr_sz != 0) {
>> +		vnet_hdr_len = pkt_sk(sk)->vnet_hdr_sz;
> and here
>
>> +		err = packet_rcv_vnet(msg, skb, &len, vnet_hdr_len);
>>   		if (err)
>>   			goto out_free;
>> -		vnet_hdr_len = sizeof(struct virtio_net_hdr);
>>   	}
>>   
>>   	/* You lose any data beyond the buffer you gave. If it worries
>> @@ -3921,8 +3936,10 @@ static void packet_flush_mclist(struct sock *sk)
>>   		return 0;
>>   	}
>>   	case PACKET_VNET_HDR:
>> +	case PACKET_VNET_HDR_SZ:
>>   	{
>>   		int val;
>> +		int hdr_len = 0;
>>   
>>   		if (sock->type != SOCK_RAW)
>>   			return -EINVAL;
>> @@ -3931,11 +3948,21 @@ static void packet_flush_mclist(struct sock *sk)
>>   		if (copy_from_sockptr(&val, optval, sizeof(val)))
>>   			return -EFAULT;
>>   
>> +		if (optname == PACKET_VNET_HDR_SZ) {
>> +			if (val != sizeof(struct virtio_net_hdr) &&
>> +			    val != sizeof(struct virtio_net_hdr_mrg_rxbuf))
>> +				return -EINVAL;
>> +			hdr_len = val;
> val == 0 is also a a correct value
>
>> +		} else {
>> +			if (!!val)
>> +				hdr_len = sizeof(struct virtio_net_hdr);
> no need for !! when testing non-zero.
>
> and instead of initializing to zero on initial variable definition:
>
>      hdr_len = val ? sizeof(struct virtio_net_hdr) : 0;
>
>> +		}
>> +
>>   		lock_sock(sk);
>>   		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) {
>>   			ret = -EBUSY;
>>   		} else {
>> -			po->has_vnet_hdr = !!val;
>> +			po->vnet_hdr_sz = hdr_len;
>>   			ret = 0;
>>   		}
>>   		release_sock(sk);
>> @@ -4068,7 +4095,10 @@ static int packet_getsockopt(struct socket *sock, int level, int optname,
>>   		val = po->origdev;
>>   		break;
>>   	case PACKET_VNET_HDR:
>> -		val = po->has_vnet_hdr;
>> +		val = !!po->vnet_hdr_sz;
>> +		break;
>> +	case PACKET_VNET_HDR_SZ:
>> +		val = po->vnet_hdr_sz;
>>   		break;
>>   	case PACKET_VERSION:
>>   		val = po->tp_version;
>> diff --git a/net/packet/diag.c b/net/packet/diag.c
>> index 07812ae..4e544da 100644
>> --- a/net/packet/diag.c
>> +++ b/net/packet/diag.c
>> @@ -27,7 +27,7 @@ static int pdiag_put_info(const struct packet_sock *po, struct sk_buff *nlskb)
>>   		pinfo.pdi_flags |= PDI_AUXDATA;
>>   	if (po->origdev)
>>   		pinfo.pdi_flags |= PDI_ORIGDEV;
>> -	if (po->has_vnet_hdr)
>> +	if (po->vnet_hdr_sz != 0)
>>   		pinfo.pdi_flags |= PDI_VNETHDR;
>>   	if (po->tp_loss)
>>   		pinfo.pdi_flags |= PDI_LOSS;
>> diff --git a/net/packet/internal.h b/net/packet/internal.h
>> index 48af35b..9b52d93 100644
>> --- a/net/packet/internal.h
>> +++ b/net/packet/internal.h
>> @@ -119,9 +119,9 @@ struct packet_sock {
>>   	unsigned int		running;	/* bind_lock must be held */
>>   	unsigned int		auxdata:1,	/* writer must hold sock lock */
>>   				origdev:1,
>> -				has_vnet_hdr:1,
>>   				tp_loss:1,
>> -				tp_tx_has_off:1;
>> +				tp_tx_has_off:1,
>> +				vnet_hdr_sz:8;
> just a separate u8 variable , rather than 8 bits in a u32.
>
>>   	int			pressure;
>>   	int			ifindex;	/* bound device		*/


We plan to add

+	   u8	vnet_hdr_sz:8;

here.
Is this a proper place to add this field to make sure the cacheline will not be broken?


>>   	__be16			num;
>> -- 
>> 1.8.3.1
>>
Willem de Bruijn March 9, 2023, 3:18 p.m. UTC | #4
沈安琪(凛玥) wrote:
> 
> 在 2023/3/7 下午11:49, Willem de Bruijn 写道:
> > 沈安琪(凛玥) wrote:
> >> From: Jianfeng Tan <henry.tjf@antgroup.com>
> >>
> >> Packet sockets, like tap, can be used as the backend for kernel vhost.
> >> In packet sockets, virtio net header size is currently hardcoded to be
> >> the size of struct virtio_net_hdr, which is 10 bytes; however, it is not
> >> always the case: some virtio features, such as mrg_rxbuf, need virtio
> >> net header to be 12-byte long.
> >>
> >> Mergeable buffers, as a virtio feature, is worthy of supporting: packets
> >> that are larger than one-mbuf size will be dropped in vhost worker's
> >> handle_rx if mrg_rxbuf feature is not used, but large packets
> >> cannot be avoided and increasing mbuf's size is not economical.
> >>
> >> With this mergeable feature enabled by virtio-user, packet sockets with
> >> hardcoded 10-byte virtio net header will parse mac head incorrectly in
> >> packet_snd by taking the last two bytes of virtio net header as part of
> >> mac header.
> >> This incorrect mac header parsing will cause packet to be dropped due to
> >> invalid ether head checking in later under-layer device packet receiving.
> >>
> >> By adding extra field vnet_hdr_sz with utilizing holes in struct
> >> packet_sock to record currently used virtio net header size and supporting
> >> extra sockopt PACKET_VNET_HDR_SZ to set specified vnet_hdr_sz, packet
> >> sockets can know the exact length of virtio net header that virtio user
> >> gives.
> >> In packet_snd, tpacket_snd and packet_recvmsg, instead of using
> >> hardcoded virtio net header size, it can get the exact vnet_hdr_sz from
> >> corresponding packet_sock, and parse mac header correctly based on this
> >> information to avoid the packets being mistakenly dropped.
> >>
> >> Besides, has_vnet_hdr field in struct packet_sock is removed since all
> >> the information it provides is covered by vnet_hdr_sz field: a packet
> >> socket has a vnet header if and only if its vnet_hdr_sz is not zero.
> >>
> >> Signed-off-by: Jianfeng Tan <henry.tjf@antgroup.com>
> >> Co-developed-by: Anqi Shen <amy.saq@antgroup.com>
> >> Signed-off-by: Anqi Shen <amy.saq@antgroup.com>
> >> ---
> >> diff --git a/net/packet/internal.h b/net/packet/internal.h
> >> index 48af35b..9b52d93 100644
> >> --- a/net/packet/internal.h
> >> +++ b/net/packet/internal.h
> >> @@ -119,9 +119,9 @@ struct packet_sock {
> >>   	unsigned int		running;	/* bind_lock must be held */
> >>   	unsigned int		auxdata:1,	/* writer must hold sock lock */
> >>   				origdev:1,
> >> -				has_vnet_hdr:1,
> >>   				tp_loss:1,
> >> -				tp_tx_has_off:1;
> >> +				tp_tx_has_off:1,
> >> +				vnet_hdr_sz:8;
> > just a separate u8 variable , rather than 8 bits in a u32.
> >
> >>   	int			pressure;
> >>   	int			ifindex;	/* bound device		*/
> 
> 
> We plan to add
> 
> +	   u8	vnet_hdr_sz:8;
> 
> here.
> Is this a proper place to add this field to make sure the cacheline will not be broken?

When in doubt, use pahole (`pahole -C packet_sock net/packet/af_packet.o`).

There currently is a 27-bit hole before pressure. That would be a good spot. 
> 
> >>   	__be16			num;
> >> -- 
> >> 1.8.3.1
> >>
沈安琪(凛玥) March 10, 2023, 2:14 a.m. UTC | #5
在 2023/3/9 下午11:18, Willem de Bruijn 写道:
> 沈安琪(凛玥) wrote:
>> 在 2023/3/7 下午11:49, Willem de Bruijn 写道:
>>> 沈安琪(凛玥) wrote:
>>>> From: Jianfeng Tan <henry.tjf@antgroup.com>
>>>>
>>>> Packet sockets, like tap, can be used as the backend for kernel vhost.
>>>> In packet sockets, virtio net header size is currently hardcoded to be
>>>> the size of struct virtio_net_hdr, which is 10 bytes; however, it is not
>>>> always the case: some virtio features, such as mrg_rxbuf, need virtio
>>>> net header to be 12-byte long.
>>>>
>>>> Mergeable buffers, as a virtio feature, is worthy of supporting: packets
>>>> that are larger than one-mbuf size will be dropped in vhost worker's
>>>> handle_rx if mrg_rxbuf feature is not used, but large packets
>>>> cannot be avoided and increasing mbuf's size is not economical.
>>>>
>>>> With this mergeable feature enabled by virtio-user, packet sockets with
>>>> hardcoded 10-byte virtio net header will parse mac head incorrectly in
>>>> packet_snd by taking the last two bytes of virtio net header as part of
>>>> mac header.
>>>> This incorrect mac header parsing will cause packet to be dropped due to
>>>> invalid ether head checking in later under-layer device packet receiving.
>>>>
>>>> By adding extra field vnet_hdr_sz with utilizing holes in struct
>>>> packet_sock to record currently used virtio net header size and supporting
>>>> extra sockopt PACKET_VNET_HDR_SZ to set specified vnet_hdr_sz, packet
>>>> sockets can know the exact length of virtio net header that virtio user
>>>> gives.
>>>> In packet_snd, tpacket_snd and packet_recvmsg, instead of using
>>>> hardcoded virtio net header size, it can get the exact vnet_hdr_sz from
>>>> corresponding packet_sock, and parse mac header correctly based on this
>>>> information to avoid the packets being mistakenly dropped.
>>>>
>>>> Besides, has_vnet_hdr field in struct packet_sock is removed since all
>>>> the information it provides is covered by vnet_hdr_sz field: a packet
>>>> socket has a vnet header if and only if its vnet_hdr_sz is not zero.
>>>>
>>>> Signed-off-by: Jianfeng Tan <henry.tjf@antgroup.com>
>>>> Co-developed-by: Anqi Shen <amy.saq@antgroup.com>
>>>> Signed-off-by: Anqi Shen <amy.saq@antgroup.com>
>>>> ---
>>>> diff --git a/net/packet/internal.h b/net/packet/internal.h
>>>> index 48af35b..9b52d93 100644
>>>> --- a/net/packet/internal.h
>>>> +++ b/net/packet/internal.h
>>>> @@ -119,9 +119,9 @@ struct packet_sock {
>>>>    	unsigned int		running;	/* bind_lock must be held */
>>>>    	unsigned int		auxdata:1,	/* writer must hold sock lock */
>>>>    				origdev:1,
>>>> -				has_vnet_hdr:1,
>>>>    				tp_loss:1,
>>>> -				tp_tx_has_off:1;
>>>> +				tp_tx_has_off:1,
>>>> +				vnet_hdr_sz:8;
>>> just a separate u8 variable , rather than 8 bits in a u32.
>>>
>>>>    	int			pressure;
>>>>    	int			ifindex;	/* bound device		*/
>>
>> We plan to add
>>
>> +	   u8	vnet_hdr_sz:8;
>>
>> here.
>> Is this a proper place to add this field to make sure the cacheline will not be broken?
> When in doubt, use pahole (`pahole -C packet_sock net/packet/af_packet.o`).
>
> There currently is a 27-bit hole before pressure. That would be a good spot.


Thanks for the advice! We will try the tool.

Besides, we wonder whether it will be better to use unsigned char or u8 
here to be more consistent with other fields.


>>>>    	__be16			num;
>>>> -- 
>>>> 1.8.3.1
>>>>
Willem de Bruijn March 10, 2023, 3:01 p.m. UTC | #6
沈安琪(凛玥) wrote:
> 
> 在 2023/3/9 下午11:18, Willem de Bruijn 写道:
> > 沈安琪(凛玥) wrote:
> >> 在 2023/3/7 下午11:49, Willem de Bruijn 写道:
> >>> 沈安琪(凛玥) wrote:
> >>>> From: Jianfeng Tan <henry.tjf@antgroup.com>
> >>>>
> >>>> Packet sockets, like tap, can be used as the backend for kernel vhost.
> >>>> In packet sockets, virtio net header size is currently hardcoded to be
> >>>> the size of struct virtio_net_hdr, which is 10 bytes; however, it is not
> >>>> always the case: some virtio features, such as mrg_rxbuf, need virtio
> >>>> net header to be 12-byte long.
> >>>>
> >>>> Mergeable buffers, as a virtio feature, is worthy of supporting: packets
> >>>> that are larger than one-mbuf size will be dropped in vhost worker's
> >>>> handle_rx if mrg_rxbuf feature is not used, but large packets
> >>>> cannot be avoided and increasing mbuf's size is not economical.
> >>>>
> >>>> With this mergeable feature enabled by virtio-user, packet sockets with
> >>>> hardcoded 10-byte virtio net header will parse mac head incorrectly in
> >>>> packet_snd by taking the last two bytes of virtio net header as part of
> >>>> mac header.
> >>>> This incorrect mac header parsing will cause packet to be dropped due to
> >>>> invalid ether head checking in later under-layer device packet receiving.
> >>>>
> >>>> By adding extra field vnet_hdr_sz with utilizing holes in struct
> >>>> packet_sock to record currently used virtio net header size and supporting
> >>>> extra sockopt PACKET_VNET_HDR_SZ to set specified vnet_hdr_sz, packet
> >>>> sockets can know the exact length of virtio net header that virtio user
> >>>> gives.
> >>>> In packet_snd, tpacket_snd and packet_recvmsg, instead of using
> >>>> hardcoded virtio net header size, it can get the exact vnet_hdr_sz from
> >>>> corresponding packet_sock, and parse mac header correctly based on this
> >>>> information to avoid the packets being mistakenly dropped.
> >>>>
> >>>> Besides, has_vnet_hdr field in struct packet_sock is removed since all
> >>>> the information it provides is covered by vnet_hdr_sz field: a packet
> >>>> socket has a vnet header if and only if its vnet_hdr_sz is not zero.
> >>>>
> >>>> Signed-off-by: Jianfeng Tan <henry.tjf@antgroup.com>
> >>>> Co-developed-by: Anqi Shen <amy.saq@antgroup.com>
> >>>> Signed-off-by: Anqi Shen <amy.saq@antgroup.com>
> >>>> ---
> >>>> diff --git a/net/packet/internal.h b/net/packet/internal.h
> >>>> index 48af35b..9b52d93 100644
> >>>> --- a/net/packet/internal.h
> >>>> +++ b/net/packet/internal.h
> >>>> @@ -119,9 +119,9 @@ struct packet_sock {
> >>>>    	unsigned int		running;	/* bind_lock must be held */
> >>>>    	unsigned int		auxdata:1,	/* writer must hold sock lock */
> >>>>    				origdev:1,
> >>>> -				has_vnet_hdr:1,
> >>>>    				tp_loss:1,
> >>>> -				tp_tx_has_off:1;
> >>>> +				tp_tx_has_off:1,
> >>>> +				vnet_hdr_sz:8;
> >>> just a separate u8 variable , rather than 8 bits in a u32.
> >>>
> >>>>    	int			pressure;
> >>>>    	int			ifindex;	/* bound device		*/
> >>
> >> We plan to add
> >>
> >> +	   u8	vnet_hdr_sz:8;
> >>
> >> here.
> >> Is this a proper place to add this field to make sure the cacheline will not be broken?
> > When in doubt, use pahole (`pahole -C packet_sock net/packet/af_packet.o`).
> >
> > There currently is a 27-bit hole before pressure. That would be a good spot.
> 
> 
> Thanks for the advice! We will try the tool.
> 
> Besides, we wonder whether it will be better to use unsigned char or u8 
> here to be more consistent with other fields.

u8 is good
沈安琪(凛玥) March 13, 2023, 5:51 a.m. UTC | #7
在 2023/3/10 下午11:01, Willem de Bruijn 写道:
> 沈安琪(凛玥) wrote:
>> 在 2023/3/9 下午11:18, Willem de Bruijn 写道:
>>> 沈安琪(凛玥) wrote:
>>>> 在 2023/3/7 下午11:49, Willem de Bruijn 写道:
>>>>> 沈安琪(凛玥) wrote:
>>>>>> From: Jianfeng Tan <henry.tjf@antgroup.com>
>>>>>>
>>>>>> Packet sockets, like tap, can be used as the backend for kernel vhost.
>>>>>> In packet sockets, virtio net header size is currently hardcoded to be
>>>>>> the size of struct virtio_net_hdr, which is 10 bytes; however, it is not
>>>>>> always the case: some virtio features, such as mrg_rxbuf, need virtio
>>>>>> net header to be 12-byte long.
>>>>>>
>>>>>> Mergeable buffers, as a virtio feature, is worthy of supporting: packets
>>>>>> that are larger than one-mbuf size will be dropped in vhost worker's
>>>>>> handle_rx if mrg_rxbuf feature is not used, but large packets
>>>>>> cannot be avoided and increasing mbuf's size is not economical.
>>>>>>
>>>>>> With this mergeable feature enabled by virtio-user, packet sockets with
>>>>>> hardcoded 10-byte virtio net header will parse mac head incorrectly in
>>>>>> packet_snd by taking the last two bytes of virtio net header as part of
>>>>>> mac header.
>>>>>> This incorrect mac header parsing will cause packet to be dropped due to
>>>>>> invalid ether head checking in later under-layer device packet receiving.
>>>>>>
>>>>>> By adding extra field vnet_hdr_sz with utilizing holes in struct
>>>>>> packet_sock to record currently used virtio net header size and supporting
>>>>>> extra sockopt PACKET_VNET_HDR_SZ to set specified vnet_hdr_sz, packet
>>>>>> sockets can know the exact length of virtio net header that virtio user
>>>>>> gives.
>>>>>> In packet_snd, tpacket_snd and packet_recvmsg, instead of using
>>>>>> hardcoded virtio net header size, it can get the exact vnet_hdr_sz from
>>>>>> corresponding packet_sock, and parse mac header correctly based on this
>>>>>> information to avoid the packets being mistakenly dropped.
>>>>>>
>>>>>> Besides, has_vnet_hdr field in struct packet_sock is removed since all
>>>>>> the information it provides is covered by vnet_hdr_sz field: a packet
>>>>>> socket has a vnet header if and only if its vnet_hdr_sz is not zero.
>>>>>>
>>>>>> Signed-off-by: Jianfeng Tan <henry.tjf@antgroup.com>
>>>>>> Co-developed-by: Anqi Shen <amy.saq@antgroup.com>
>>>>>> Signed-off-by: Anqi Shen <amy.saq@antgroup.com>
>>>>>> ---
>>>>>> diff --git a/net/packet/internal.h b/net/packet/internal.h
>>>>>> index 48af35b..9b52d93 100644
>>>>>> --- a/net/packet/internal.h
>>>>>> +++ b/net/packet/internal.h
>>>>>> @@ -119,9 +119,9 @@ struct packet_sock {
>>>>>>     	unsigned int		running;	/* bind_lock must be held */
>>>>>>     	unsigned int		auxdata:1,	/* writer must hold sock lock */
>>>>>>     				origdev:1,
>>>>>> -				has_vnet_hdr:1,
>>>>>>     				tp_loss:1,
>>>>>> -				tp_tx_has_off:1;
>>>>>> +				tp_tx_has_off:1,
>>>>>> +				vnet_hdr_sz:8;
>>>>> just a separate u8 variable , rather than 8 bits in a u32.
>>>>>
>>>>>>     	int			pressure;
>>>>>>     	int			ifindex;	/* bound device		*/
>>>> We plan to add
>>>>
>>>> +	   u8	vnet_hdr_sz:8;
>>>>
>>>> here.
>>>> Is this a proper place to add this field to make sure the cacheline will not be broken?
>>> When in doubt, use pahole (`pahole -C packet_sock net/packet/af_packet.o`).
>>>
>>> There currently is a 27-bit hole before pressure. That would be a good spot.
>>
>> Thanks for the advice! We will try the tool.
>>
>> Besides, we wonder whether it will be better to use unsigned char or u8
>> here to be more consistent with other fields.
> u8 is good


Got it. Thanks. We will address the comments and send out v4 soon.
diff mbox series

Patch

diff --git a/include/uapi/linux/if_packet.h b/include/uapi/linux/if_packet.h
index 78c981d..9efc423 100644
--- a/include/uapi/linux/if_packet.h
+++ b/include/uapi/linux/if_packet.h
@@ -59,6 +59,7 @@  struct sockaddr_ll {
 #define PACKET_ROLLOVER_STATS		21
 #define PACKET_FANOUT_DATA		22
 #define PACKET_IGNORE_OUTGOING		23
+#define PACKET_VNET_HDR_SZ		24
 
 #define PACKET_FANOUT_HASH		0
 #define PACKET_FANOUT_LB		1
diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c
index 8ffb19c..ce2feff 100644
--- a/net/packet/af_packet.c
+++ b/net/packet/af_packet.c
@@ -2092,18 +2092,18 @@  static unsigned int run_filter(struct sk_buff *skb,
 }
 
 static int packet_rcv_vnet(struct msghdr *msg, const struct sk_buff *skb,
-			   size_t *len)
+			   size_t *len, int vnet_hdr_sz)
 {
-	struct virtio_net_hdr vnet_hdr;
+	struct virtio_net_hdr_mrg_rxbuf vnet_hdr = { .num_buffers = 0 };
 
-	if (*len < sizeof(vnet_hdr))
+	if (*len < vnet_hdr_sz)
 		return -EINVAL;
-	*len -= sizeof(vnet_hdr);
+	*len -= vnet_hdr_sz;
 
-	if (virtio_net_hdr_from_skb(skb, &vnet_hdr, vio_le(), true, 0))
+	if (virtio_net_hdr_from_skb(skb, (struct virtio_net_hdr *)&vnet_hdr, vio_le(), true, 0))
 		return -EINVAL;
 
-	return memcpy_to_msg(msg, (void *)&vnet_hdr, sizeof(vnet_hdr));
+	return memcpy_to_msg(msg, (void *)&vnet_hdr, vnet_hdr_sz);
 }
 
 /*
@@ -2310,8 +2310,8 @@  static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev,
 		netoff = TPACKET_ALIGN(po->tp_hdrlen +
 				       (maclen < 16 ? 16 : maclen)) +
 				       po->tp_reserve;
-		if (po->has_vnet_hdr) {
-			netoff += sizeof(struct virtio_net_hdr);
+		if (po->vnet_hdr_sz != 0) {
+			netoff += po->vnet_hdr_sz;
 			do_vnet = true;
 		}
 		macoff = netoff - maclen;
@@ -2552,16 +2552,27 @@  static int __packet_snd_vnet_parse(struct virtio_net_hdr *vnet_hdr, size_t len)
 }
 
 static int packet_snd_vnet_parse(struct msghdr *msg, size_t *len,
-				 struct virtio_net_hdr *vnet_hdr)
+				 struct virtio_net_hdr *vnet_hdr, int vnet_hdr_sz)
 {
-	if (*len < sizeof(*vnet_hdr))
+	int ret;
+
+	if (*len < vnet_hdr_sz)
 		return -EINVAL;
-	*len -= sizeof(*vnet_hdr);
+	*len -= vnet_hdr_sz;
 
 	if (!copy_from_iter_full(vnet_hdr, sizeof(*vnet_hdr), &msg->msg_iter))
 		return -EFAULT;
 
-	return __packet_snd_vnet_parse(vnet_hdr, *len);
+	ret = __packet_snd_vnet_parse(vnet_hdr, *len);
+
+	if (ret)
+		return ret;
+
+	/* move iter to point to the start of mac header */
+	if (vnet_hdr_sz != sizeof(struct virtio_net_hdr))
+		iov_iter_advance(&msg->msg_iter, vnet_hdr_sz - sizeof(struct virtio_net_hdr));
+
+	return 0;
 }
 
 static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb,
@@ -2730,6 +2741,7 @@  static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
 	int status = TP_STATUS_AVAILABLE;
 	int hlen, tlen, copylen = 0;
 	long timeo = 0;
+	int vnet_hdr_sz;
 
 	mutex_lock(&po->pg_vec_lock);
 
@@ -2780,7 +2792,7 @@  static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
 	size_max = po->tx_ring.frame_size
 		- (po->tp_hdrlen - sizeof(struct sockaddr_ll));
 
-	if ((size_max > dev->mtu + reserve + VLAN_HLEN) && !po->has_vnet_hdr)
+	if ((size_max > dev->mtu + reserve + VLAN_HLEN) && po->vnet_hdr_sz == 0)
 		size_max = dev->mtu + reserve + VLAN_HLEN;
 
 	reinit_completion(&po->skb_completion);
@@ -2809,10 +2821,11 @@  static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
 		status = TP_STATUS_SEND_REQUEST;
 		hlen = LL_RESERVED_SPACE(dev);
 		tlen = dev->needed_tailroom;
-		if (po->has_vnet_hdr) {
+		if (po->vnet_hdr_sz != 0) {
 			vnet_hdr = data;
-			data += sizeof(*vnet_hdr);
-			tp_len -= sizeof(*vnet_hdr);
+			vnet_hdr_sz = po->vnet_hdr_sz;
+			data += vnet_hdr_sz;
+			tp_len -= vnet_hdr_sz;
 			if (tp_len < 0 ||
 			    __packet_snd_vnet_parse(vnet_hdr, tp_len)) {
 				tp_len = -EINVAL;
@@ -2837,7 +2850,7 @@  static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
 					  addr, hlen, copylen, &sockc);
 		if (likely(tp_len >= 0) &&
 		    tp_len > dev->mtu + reserve &&
-		    !po->has_vnet_hdr &&
+		    (po->vnet_hdr_sz == 0) &&
 		    !packet_extra_vlan_len_allowed(dev, skb))
 			tp_len = -EMSGSIZE;
 
@@ -2856,7 +2869,7 @@  static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
 			}
 		}
 
-		if (po->has_vnet_hdr) {
+		if (po->vnet_hdr_sz != 0) {
 			if (virtio_net_hdr_to_skb(skb, vnet_hdr, vio_le())) {
 				tp_len = -EINVAL;
 				goto tpacket_error;
@@ -2947,6 +2960,7 @@  static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
 	int offset = 0;
 	struct packet_sock *po = pkt_sk(sk);
 	bool has_vnet_hdr = false;
+	int vnet_hdr_sz;
 	int hlen, tlen, linear;
 	int extra_len = 0;
 
@@ -2990,8 +3004,9 @@  static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
 
 	if (sock->type == SOCK_RAW)
 		reserve = dev->hard_header_len;
-	if (po->has_vnet_hdr) {
-		err = packet_snd_vnet_parse(msg, &len, &vnet_hdr);
+	if (po->vnet_hdr_sz != 0) {
+		vnet_hdr_sz = po->vnet_hdr_sz;
+		err = packet_snd_vnet_parse(msg, &len, &vnet_hdr, vnet_hdr_sz);
 		if (err)
 			goto out_unlock;
 		has_vnet_hdr = true;
@@ -3068,7 +3083,7 @@  static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)
 		err = virtio_net_hdr_to_skb(skb, &vnet_hdr, vio_le());
 		if (err)
 			goto out_free;
-		len += sizeof(vnet_hdr);
+		len += vnet_hdr_sz;
 		virtio_net_hdr_set_proto(skb, &vnet_hdr);
 	}
 
@@ -3451,11 +3466,11 @@  static int packet_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
 
 	packet_rcv_try_clear_pressure(pkt_sk(sk));
 
-	if (pkt_sk(sk)->has_vnet_hdr) {
-		err = packet_rcv_vnet(msg, skb, &len);
+	if (pkt_sk(sk)->vnet_hdr_sz != 0) {
+		vnet_hdr_len = pkt_sk(sk)->vnet_hdr_sz;
+		err = packet_rcv_vnet(msg, skb, &len, vnet_hdr_len);
 		if (err)
 			goto out_free;
-		vnet_hdr_len = sizeof(struct virtio_net_hdr);
 	}
 
 	/* You lose any data beyond the buffer you gave. If it worries
@@ -3921,8 +3936,10 @@  static void packet_flush_mclist(struct sock *sk)
 		return 0;
 	}
 	case PACKET_VNET_HDR:
+	case PACKET_VNET_HDR_SZ:
 	{
 		int val;
+		int hdr_len = 0;
 
 		if (sock->type != SOCK_RAW)
 			return -EINVAL;
@@ -3931,11 +3948,21 @@  static void packet_flush_mclist(struct sock *sk)
 		if (copy_from_sockptr(&val, optval, sizeof(val)))
 			return -EFAULT;
 
+		if (optname == PACKET_VNET_HDR_SZ) {
+			if (val != sizeof(struct virtio_net_hdr) &&
+			    val != sizeof(struct virtio_net_hdr_mrg_rxbuf))
+				return -EINVAL;
+			hdr_len = val;
+		} else {
+			if (!!val)
+				hdr_len = sizeof(struct virtio_net_hdr);
+		}
+
 		lock_sock(sk);
 		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) {
 			ret = -EBUSY;
 		} else {
-			po->has_vnet_hdr = !!val;
+			po->vnet_hdr_sz = hdr_len;
 			ret = 0;
 		}
 		release_sock(sk);
@@ -4068,7 +4095,10 @@  static int packet_getsockopt(struct socket *sock, int level, int optname,
 		val = po->origdev;
 		break;
 	case PACKET_VNET_HDR:
-		val = po->has_vnet_hdr;
+		val = !!po->vnet_hdr_sz;
+		break;
+	case PACKET_VNET_HDR_SZ:
+		val = po->vnet_hdr_sz;
 		break;
 	case PACKET_VERSION:
 		val = po->tp_version;
diff --git a/net/packet/diag.c b/net/packet/diag.c
index 07812ae..4e544da 100644
--- a/net/packet/diag.c
+++ b/net/packet/diag.c
@@ -27,7 +27,7 @@  static int pdiag_put_info(const struct packet_sock *po, struct sk_buff *nlskb)
 		pinfo.pdi_flags |= PDI_AUXDATA;
 	if (po->origdev)
 		pinfo.pdi_flags |= PDI_ORIGDEV;
-	if (po->has_vnet_hdr)
+	if (po->vnet_hdr_sz != 0)
 		pinfo.pdi_flags |= PDI_VNETHDR;
 	if (po->tp_loss)
 		pinfo.pdi_flags |= PDI_LOSS;
diff --git a/net/packet/internal.h b/net/packet/internal.h
index 48af35b..9b52d93 100644
--- a/net/packet/internal.h
+++ b/net/packet/internal.h
@@ -119,9 +119,9 @@  struct packet_sock {
 	unsigned int		running;	/* bind_lock must be held */
 	unsigned int		auxdata:1,	/* writer must hold sock lock */
 				origdev:1,
-				has_vnet_hdr:1,
 				tp_loss:1,
-				tp_tx_has_off:1;
+				tp_tx_has_off:1,
+				vnet_hdr_sz:8;
 	int			pressure;
 	int			ifindex;	/* bound device		*/
 	__be16			num;