Message-ID: <78cdb144-bc87-0c1a-cb70-cec69cb34c3d@sberdevices.ru>
Date: Tue, 25 Jul 2023 16:30:09 +0300
From: Arseniy Krasnov <avkrasnov@...rdevices.ru>
To: "Michael S. Tsirkin" <mst@...hat.com>
CC: Stefan Hajnoczi <stefanha@...hat.com>, Stefano Garzarella
	<sgarzare@...hat.com>, "David S. Miller" <davem@...emloft.net>, Eric Dumazet
	<edumazet@...gle.com>, Jakub Kicinski <kuba@...nel.org>, Paolo Abeni
	<pabeni@...hat.com>, Jason Wang <jasowang@...hat.com>, Bobby Eshleman
	<bobby.eshleman@...edance.com>, <kvm@...r.kernel.org>,
	<virtualization@...ts.linux-foundation.org>, <netdev@...r.kernel.org>,
	<linux-kernel@...r.kernel.org>, <kernel@...rdevices.ru>, <oxffffaa@...il.com>
Subject: Re: [PATCH net-next v3 4/4] vsock/virtio: MSG_ZEROCOPY flag support



On 25.07.2023 16:23, Michael S. Tsirkin wrote:
> On Tue, Jul 25, 2023 at 04:10:40PM +0300, Arseniy Krasnov wrote:
>>
>>
>> On 25.07.2023 14:59, Michael S. Tsirkin wrote:
>>> On Tue, Jul 25, 2023 at 11:39:22AM +0300, Arseniy Krasnov wrote:
>>>>
>>>>
>>>> On 25.07.2023 11:25, Michael S. Tsirkin wrote:
>>>>> On Fri, Jul 21, 2023 at 12:42:45AM +0300, Arseniy Krasnov wrote:
>>>>>> This adds handling of MSG_ZEROCOPY flag on transmission path: if this
>>>>>> flag is set and zerocopy transmission is possible (enabled in socket
>>>>>> options and transport allows zerocopy), then non-linear skb will be
>>>>>> created and filled with the pages of user's buffer. Pages of user's
>>>>>> buffer are locked in memory by 'get_user_pages()'. Second thing that
>>>>>> this patch does is replace type of skb owning: instead of calling
>>>>>> 'skb_set_owner_sk_safe()' it calls 'skb_set_owner_w()'. Reason of this
>>>>>> change is that '__zerocopy_sg_from_iter()' increments 'sk_wmem_alloc'
>>>>>> of socket, so to decrease this field correctly proper skb destructor is
>>>>>> needed: 'sock_wfree()'. This destructor is set by 'skb_set_owner_w()'.
>>>>>>
>>>>>> Signed-off-by: Arseniy Krasnov <AVKrasnov@...rdevices.ru>
>>>>>> ---
>>>>>>  Changelog:
>>>>>>  v5(big patchset) -> v1:
>>>>>>   * Refactorings of 'if' conditions.
>>>>>>   * Remove extra blank line.
>>>>>>   * Remove 'frag_off' field unneeded init.
>>>>>>   * Add function 'virtio_transport_fill_skb()' which fills both linear
>>>>>>     and non-linear skb with provided data.
>>>>>>  v1 -> v2:
>>>>>>   * Use original order of last four arguments in 'virtio_transport_alloc_skb()'.
>>>>>>  v2 -> v3:
>>>>>>   * Add new transport callback: 'msgzerocopy_check_iov'. It checks that
>>>>>>     provided 'iov_iter' with data could be sent in a zerocopy mode.
>>>>>>     If this callback is not set in transport - transport allows to send
>>>>>>     any 'iov_iter' in zerocopy mode. Otherwise - if callback returns 'true'
>>>>>>     then zerocopy is allowed. Reason of this callback is that in case of
>>>>>>     G2H transmission we insert whole skb to the tx virtio queue and such
>>>>>>     skb must fit to the size of the virtio queue to be sent in a single
>>>>>>     iteration (may be tx logic in 'virtio_transport.c' could be reworked
>>>>>>     as in vhost to support partial send of current skb). This callback
>>>>>>     will be enabled only for G2H path. For details pls see comment 
>>>>>>     'Check that tx queue...' below.
>>>>>>
>>>>>>  include/net/af_vsock.h                  |   3 +
>>>>>>  net/vmw_vsock/virtio_transport.c        |  39 ++++
>>>>>>  net/vmw_vsock/virtio_transport_common.c | 257 ++++++++++++++++++------
>>>>>>  3 files changed, 241 insertions(+), 58 deletions(-)
>>>>>>
>>>>>> diff --git a/include/net/af_vsock.h b/include/net/af_vsock.h
>>>>>> index 0e7504a42925..a6b346eeeb8e 100644
>>>>>> --- a/include/net/af_vsock.h
>>>>>> +++ b/include/net/af_vsock.h
>>>>>> @@ -177,6 +177,9 @@ struct vsock_transport {
>>>>>>  
>>>>>>  	/* Read a single skb */
>>>>>>  	int (*read_skb)(struct vsock_sock *, skb_read_actor_t);
>>>>>> +
>>>>>> +	/* Zero-copy. */
>>>>>> +	bool (*msgzerocopy_check_iov)(const struct iov_iter *);
>>>>>>  };
>>>>>>  
>>>>>>  /**** CORE ****/
>>>>>> diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
>>>>>> index 7bbcc8093e51..23cb8ed638c4 100644
>>>>>> --- a/net/vmw_vsock/virtio_transport.c
>>>>>> +++ b/net/vmw_vsock/virtio_transport.c
>>>>>> @@ -442,6 +442,43 @@ static void virtio_vsock_rx_done(struct virtqueue *vq)
>>>>>>  	queue_work(virtio_vsock_workqueue, &vsock->rx_work);
>>>>>>  }
>>>>>>  
>>>>>> +static bool virtio_transport_msgzerocopy_check_iov(const struct iov_iter *iov)
>>>>>> +{
>>>>>> +	struct virtio_vsock *vsock;
>>>>>> +	bool res = false;
>>>>>> +
>>>>>> +	rcu_read_lock();
>>>>>> +
>>>>>> +	vsock = rcu_dereference(the_virtio_vsock);
>>>>>> +	if (vsock) {
>>>>>> +		struct virtqueue *vq;
>>>>>> +		int iov_pages;
>>>>>> +
>>>>>> +		vq = vsock->vqs[VSOCK_VQ_TX];
>>>>>> +
>>>>>> +		iov_pages = round_up(iov->count, PAGE_SIZE) / PAGE_SIZE;
>>>>>> +
>>>>>> +		/* Check that tx queue is large enough to keep whole
>>>>>> +		 * data to send. This is needed, because when there is
>>>>>> +		 * not enough free space in the queue, current skb to
>>>>>> +		 * send will be reinserted to the head of tx list of
>>>>>> +		 * the socket to retry transmission later, so if skb
>>>>>> +		 * is bigger than whole queue, it will be reinserted
>>>>>> +		 * again and again, thus blocking other skbs to be sent.
>>>>>> +		 * Each page of the user provided buffer will be added
>>>>>> +		 * as a single buffer to the tx virtqueue, so compare
>>>>>> +		 * number of pages against maximum capacity of the queue.
>>>>>> +		 * +1 means buffer for the packet header.
>>>>>> +		 */
>>>>>> +		if (iov_pages + 1 <= vq->num_max)
>>>>>> +			res = true;
>>>>>
>>>>>
>>>>> Yes but can't there already be buffers in the queue?
>>>>> Then you can't stick num_max there.
>>>>
>>>> I think, that it is not critical, because vhost part always tries to process all
>>>> incoming buffers (yes, 'vhost_exceeds_weight()' breaks at some moment, but it will
>>>> reschedule tx kick ('vhost_vsock_handle_tx_kick()') work again), so current "too
>>>> big" skb will wait until there will be enough space in queue and as it is requeued
>>>> to the head of tx list it will be inserted to tx queue first.
>>>>
>>>> But anyway, I agree that comparing to 'num_free' may be more effective to the whole
>>>> system performance...
>>>>
>>>> Thanks, Arseniy
>>>
>>> Oh I see. It makes sense then - instead of copying just so we can
>>> stick it in the queue, wait a bit and send later.
>>> Also - for stream transports can't the message be split
>>> and sent chunk by chunk? Better than copying ...
>>
>> Technically yes, also we can split message for non-stream sockets (as vhost
>> does when it copies data to rx buffers of the guest),
> 
> Won't breaking up messages break applications though?

No, for seqpacket we have special marker + port in each message, so we
can restore original message at the receiver and pass it to the socket.

Thanks, Arseniy

> 
> 
>> but it requires to rework
>> current implementation by adding buffers one by one to the tx queue. I think
>> it was not implemented here because until MSG_ZEROCOPY all skbs requires one
>> (if it is control message) or two (with payload) buffers, so there is no big
>> sense in processing max two buffers in "one-by-one" mode - we can just wait
>> for space.
>>
>> May be, I can add this logic for non-linear skb's here: 
>>
>> if (skb->len > vq->max_num)
>>     add buffers "one-by-one", incrementing internal offset in skb,
>>    if (new skb insertion fails)
>>        requeue skb, wait for space.
>>
>> In TX done callback I'll call consume skb only when above mentioned internal
>> offset == skb->len. I think this approach allows to get rid of new 'check_iov'
>> callback from this patch.
>>
>>
>> Stefano, what do You think?
>>
>> Thanks, Arseniy
>>
>>>
>>>
>>>>>
>>>>>
>>>>>> +	}
>>>>>> +
>>>>>> +	rcu_read_unlock();
>>>>>> +
>>>>>> +	return res;
>>>>>> +}
>>>>>> +
>>>>>>  static bool virtio_transport_seqpacket_allow(u32 remote_cid);
>>>>>>  
>>>>>>  static struct virtio_transport virtio_transport = {
>>>>>> @@ -475,6 +512,8 @@ static struct virtio_transport virtio_transport = {
>>>>>>  		.seqpacket_allow          = virtio_transport_seqpacket_allow,
>>>>>>  		.seqpacket_has_data       = virtio_transport_seqpacket_has_data,
>>>>>>  
>>>>>> +		.msgzerocopy_check_iov	  = virtio_transport_msgzerocopy_check_iov,
>>>>>> +
>>>>>>  		.notify_poll_in           = virtio_transport_notify_poll_in,
>>>>>>  		.notify_poll_out          = virtio_transport_notify_poll_out,
>>>>>>  		.notify_recv_init         = virtio_transport_notify_recv_init,
>>>>>> diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c
>>>>>> index 26a4d10da205..e4e3d541aff4 100644
>>>>>> --- a/net/vmw_vsock/virtio_transport_common.c
>>>>>> +++ b/net/vmw_vsock/virtio_transport_common.c
>>>>>> @@ -37,73 +37,122 @@ virtio_transport_get_ops(struct vsock_sock *vsk)
>>>>>>  	return container_of(t, struct virtio_transport, transport);
>>>>>>  }
>>>>>>  
>>>>>> -/* Returns a new packet on success, otherwise returns NULL.
>>>>>> - *
>>>>>> - * If NULL is returned, errp is set to a negative errno.
>>>>>> - */
>>>>>> -static struct sk_buff *
>>>>>> -virtio_transport_alloc_skb(struct virtio_vsock_pkt_info *info,
>>>>>> -			   size_t len,
>>>>>> -			   u32 src_cid,
>>>>>> -			   u32 src_port,
>>>>>> -			   u32 dst_cid,
>>>>>> -			   u32 dst_port)
>>>>>> -{
>>>>>> -	const size_t skb_len = VIRTIO_VSOCK_SKB_HEADROOM + len;
>>>>>> -	struct virtio_vsock_hdr *hdr;
>>>>>> -	struct sk_buff *skb;
>>>>>> -	void *payload;
>>>>>> -	int err;
>>>>>> +static bool virtio_transport_can_zcopy(struct virtio_vsock_pkt_info *info,
>>>>>> +				       size_t max_to_send)
>>>>>> +{
>>>>>> +	const struct vsock_transport *t;
>>>>>> +	struct iov_iter *iov_iter;
>>>>>>  
>>>>>> -	skb = virtio_vsock_alloc_skb(skb_len, GFP_KERNEL);
>>>>>> -	if (!skb)
>>>>>> -		return NULL;
>>>>>> +	if (!info->msg)
>>>>>> +		return false;
>>>>>>  
>>>>>> -	hdr = virtio_vsock_hdr(skb);
>>>>>> -	hdr->type	= cpu_to_le16(info->type);
>>>>>> -	hdr->op		= cpu_to_le16(info->op);
>>>>>> -	hdr->src_cid	= cpu_to_le64(src_cid);
>>>>>> -	hdr->dst_cid	= cpu_to_le64(dst_cid);
>>>>>> -	hdr->src_port	= cpu_to_le32(src_port);
>>>>>> -	hdr->dst_port	= cpu_to_le32(dst_port);
>>>>>> -	hdr->flags	= cpu_to_le32(info->flags);
>>>>>> -	hdr->len	= cpu_to_le32(len);
>>>>>> +	iov_iter = &info->msg->msg_iter;
>>>>>>  
>>>>>> -	if (info->msg && len > 0) {
>>>>>> -		payload = skb_put(skb, len);
>>>>>> -		err = memcpy_from_msg(payload, info->msg, len);
>>>>>> -		if (err)
>>>>>> -			goto out;
>>>>>> +	t = vsock_core_get_transport(info->vsk);
>>>>>>  
>>>>>> -		if (msg_data_left(info->msg) == 0 &&
>>>>>> -		    info->type == VIRTIO_VSOCK_TYPE_SEQPACKET) {
>>>>>> -			hdr->flags |= cpu_to_le32(VIRTIO_VSOCK_SEQ_EOM);
>>>>>> +	if (t->msgzerocopy_check_iov &&
>>>>>> +	    !t->msgzerocopy_check_iov(iov_iter))
>>>>>> +		return false;
>>>>>>  
>>>>>> -			if (info->msg->msg_flags & MSG_EOR)
>>>>>> -				hdr->flags |= cpu_to_le32(VIRTIO_VSOCK_SEQ_EOR);
>>>>>> -		}
>>>>>> +	/* Data is simple buffer. */
>>>>>> +	if (iter_is_ubuf(iov_iter))
>>>>>> +		return true;
>>>>>> +
>>>>>> +	if (!iter_is_iovec(iov_iter))
>>>>>> +		return false;
>>>>>> +
>>>>>> +	if (iov_iter->iov_offset)
>>>>>> +		return false;
>>>>>> +
>>>>>> +	/* We can't send whole iov. */
>>>>>> +	if (iov_iter->count > max_to_send)
>>>>>> +		return false;
>>>>>> +
>>>>>> +	return true;
>>>>>> +}
>>>>>> +
>>>>>> +static int virtio_transport_init_zcopy_skb(struct vsock_sock *vsk,
>>>>>> +					   struct sk_buff *skb,
>>>>>> +					   struct msghdr *msg,
>>>>>> +					   bool zerocopy)
>>>>>> +{
>>>>>> +	struct ubuf_info *uarg;
>>>>>> +
>>>>>> +	if (msg->msg_ubuf) {
>>>>>> +		uarg = msg->msg_ubuf;
>>>>>> +		net_zcopy_get(uarg);
>>>>>> +	} else {
>>>>>> +		struct iov_iter *iter = &msg->msg_iter;
>>>>>> +		struct ubuf_info_msgzc *uarg_zc;
>>>>>> +		int len;
>>>>>> +
>>>>>> +		/* Only ITER_IOVEC or ITER_UBUF are allowed and
>>>>>> +		 * checked before.
>>>>>> +		 */
>>>>>> +		if (iter_is_iovec(iter))
>>>>>> +			len = iov_length(iter->__iov, iter->nr_segs);
>>>>>> +		else
>>>>>> +			len = iter->count;
>>>>>> +
>>>>>> +		uarg = msg_zerocopy_realloc(sk_vsock(vsk),
>>>>>> +					    len,
>>>>>> +					    NULL);
>>>>>> +		if (!uarg)
>>>>>> +			return -1;
>>>>>> +
>>>>>> +		uarg_zc = uarg_to_msgzc(uarg);
>>>>>> +		uarg_zc->zerocopy = zerocopy ? 1 : 0;
>>>>>>  	}
>>>>>>  
>>>>>> -	if (info->reply)
>>>>>> -		virtio_vsock_skb_set_reply(skb);
>>>>>> +	skb_zcopy_init(skb, uarg);
>>>>>>  
>>>>>> -	trace_virtio_transport_alloc_pkt(src_cid, src_port,
>>>>>> -					 dst_cid, dst_port,
>>>>>> -					 len,
>>>>>> -					 info->type,
>>>>>> -					 info->op,
>>>>>> -					 info->flags);
>>>>>> +	return 0;
>>>>>> +}
>>>>>>  
>>>>>> -	if (info->vsk && !skb_set_owner_sk_safe(skb, sk_vsock(info->vsk))) {
>>>>>> -		WARN_ONCE(1, "failed to allocate skb on vsock socket with sk_refcnt == 0\n");
>>>>>> -		goto out;
>>>>>> +static int virtio_transport_fill_skb(struct sk_buff *skb,
>>>>>> +				     struct virtio_vsock_pkt_info *info,
>>>>>> +				     size_t len,
>>>>>> +				     bool zcopy)
>>>>>> +{
>>>>>> +	if (zcopy) {
>>>>>> +		return __zerocopy_sg_from_iter(info->msg, NULL, skb,
>>>>>> +					      &info->msg->msg_iter,
>>>>>> +					      len);
>>>>>> +	} else {
>>>>>> +		void *payload;
>>>>>> +		int err;
>>>>>> +
>>>>>> +		payload = skb_put(skb, len);
>>>>>> +		err = memcpy_from_msg(payload, info->msg, len);
>>>>>> +		if (err)
>>>>>> +			return -1;
>>>>>> +
>>>>>> +		if (msg_data_left(info->msg))
>>>>>> +			return 0;
>>>>>> +
>>>>>> +		return 0;
>>>>>>  	}
>>>>>> +}
>>>>>>  
>>>>>> -	return skb;
>>>>>> +static void virtio_transport_init_hdr(struct sk_buff *skb,
>>>>>> +				      struct virtio_vsock_pkt_info *info,
>>>>>> +				      u32 src_cid,
>>>>>> +				      u32 src_port,
>>>>>> +				      u32 dst_cid,
>>>>>> +				      u32 dst_port,
>>>>>> +				      size_t len)
>>>>>> +{
>>>>>> +	struct virtio_vsock_hdr *hdr;
>>>>>>  
>>>>>> -out:
>>>>>> -	kfree_skb(skb);
>>>>>> -	return NULL;
>>>>>> +	hdr = virtio_vsock_hdr(skb);
>>>>>> +	hdr->type	= cpu_to_le16(info->type);
>>>>>> +	hdr->op		= cpu_to_le16(info->op);
>>>>>> +	hdr->src_cid	= cpu_to_le64(src_cid);
>>>>>> +	hdr->dst_cid	= cpu_to_le64(dst_cid);
>>>>>> +	hdr->src_port	= cpu_to_le32(src_port);
>>>>>> +	hdr->dst_port	= cpu_to_le32(dst_port);
>>>>>> +	hdr->flags	= cpu_to_le32(info->flags);
>>>>>> +	hdr->len	= cpu_to_le32(len);
>>>>>>  }
>>>>>>  
>>>>>>  static void virtio_transport_copy_nonlinear_skb(const struct sk_buff *skb,
>>>>>> @@ -214,6 +263,70 @@ static u16 virtio_transport_get_type(struct sock *sk)
>>>>>>  		return VIRTIO_VSOCK_TYPE_SEQPACKET;
>>>>>>  }
>>>>>>  
>>>>>> +static struct sk_buff *virtio_transport_alloc_skb(struct vsock_sock *vsk,
>>>>>> +						  struct virtio_vsock_pkt_info *info,
>>>>>> +						  size_t payload_len,
>>>>>> +						  bool zcopy,
>>>>>> +						  u32 src_cid,
>>>>>> +						  u32 src_port,
>>>>>> +						  u32 dst_cid,
>>>>>> +						  u32 dst_port)
>>>>>> +{
>>>>>> +	struct sk_buff *skb;
>>>>>> +	size_t skb_len;
>>>>>> +
>>>>>> +	skb_len = VIRTIO_VSOCK_SKB_HEADROOM;
>>>>>> +
>>>>>> +	if (!zcopy)
>>>>>> +		skb_len += payload_len;
>>>>>> +
>>>>>> +	skb = virtio_vsock_alloc_skb(skb_len, GFP_KERNEL);
>>>>>> +	if (!skb)
>>>>>> +		return NULL;
>>>>>> +
>>>>>> +	virtio_transport_init_hdr(skb, info, src_cid, src_port,
>>>>>> +				  dst_cid, dst_port,
>>>>>> +				  payload_len);
>>>>>> +
>>>>>> +	/* Set owner here, because '__zerocopy_sg_from_iter()' uses
>>>>>> +	 * owner of skb without check to update 'sk_wmem_alloc'.
>>>>>> +	 */
>>>>>> +	if (vsk)
>>>>>> +		skb_set_owner_w(skb, sk_vsock(vsk));
>>>>>> +
>>>>>> +	if (info->msg && payload_len > 0) {
>>>>>> +		int err;
>>>>>> +
>>>>>> +		err = virtio_transport_fill_skb(skb, info, payload_len, zcopy);
>>>>>> +		if (err)
>>>>>> +			goto out;
>>>>>> +
>>>>>> +		if (info->type == VIRTIO_VSOCK_TYPE_SEQPACKET) {
>>>>>> +			struct virtio_vsock_hdr *hdr = virtio_vsock_hdr(skb);
>>>>>> +
>>>>>> +			hdr->flags |= cpu_to_le32(VIRTIO_VSOCK_SEQ_EOM);
>>>>>> +
>>>>>> +			if (info->msg->msg_flags & MSG_EOR)
>>>>>> +				hdr->flags |= cpu_to_le32(VIRTIO_VSOCK_SEQ_EOR);
>>>>>> +		}
>>>>>> +	}
>>>>>> +
>>>>>> +	if (info->reply)
>>>>>> +		virtio_vsock_skb_set_reply(skb);
>>>>>> +
>>>>>> +	trace_virtio_transport_alloc_pkt(src_cid, src_port,
>>>>>> +					 dst_cid, dst_port,
>>>>>> +					 payload_len,
>>>>>> +					 info->type,
>>>>>> +					 info->op,
>>>>>> +					 info->flags);
>>>>>> +
>>>>>> +	return skb;
>>>>>> +out:
>>>>>> +	kfree_skb(skb);
>>>>>> +	return NULL;
>>>>>> +}
>>>>>> +
>>>>>>  /* This function can only be used on connecting/connected sockets,
>>>>>>   * since a socket assigned to a transport is required.
>>>>>>   *
>>>>>> @@ -222,10 +335,12 @@ static u16 virtio_transport_get_type(struct sock *sk)
>>>>>>  static int virtio_transport_send_pkt_info(struct vsock_sock *vsk,
>>>>>>  					  struct virtio_vsock_pkt_info *info)
>>>>>>  {
>>>>>> +	u32 max_skb_len = VIRTIO_VSOCK_MAX_PKT_BUF_SIZE;
>>>>>>  	u32 src_cid, src_port, dst_cid, dst_port;
>>>>>>  	const struct virtio_transport *t_ops;
>>>>>>  	struct virtio_vsock_sock *vvs;
>>>>>>  	u32 pkt_len = info->pkt_len;
>>>>>> +	bool can_zcopy = false;
>>>>>>  	u32 rest_len;
>>>>>>  	int ret;
>>>>>>  
>>>>>> @@ -254,15 +369,30 @@ static int virtio_transport_send_pkt_info(struct vsock_sock *vsk,
>>>>>>  	if (pkt_len == 0 && info->op == VIRTIO_VSOCK_OP_RW)
>>>>>>  		return pkt_len;
>>>>>>  
>>>>>> +	if (info->msg) {
>>>>>> +		/* If zerocopy is not enabled by 'setsockopt()', we behave as
>>>>>> +		 * there is no MSG_ZEROCOPY flag set.
>>>>>> +		 */
>>>>>> +		if (!sock_flag(sk_vsock(vsk), SOCK_ZEROCOPY))
>>>>>> +			info->msg->msg_flags &= ~MSG_ZEROCOPY;
>>>>>> +
>>>>>> +		if (info->msg->msg_flags & MSG_ZEROCOPY)
>>>>>> +			can_zcopy = virtio_transport_can_zcopy(info, pkt_len);
>>>>>> +
>>>>>> +		if (can_zcopy)
>>>>>> +			max_skb_len = min_t(u32, VIRTIO_VSOCK_MAX_PKT_BUF_SIZE,
>>>>>> +					    (MAX_SKB_FRAGS * PAGE_SIZE));
>>>>>> +	}
>>>>>> +
>>>>>>  	rest_len = pkt_len;
>>>>>>  
>>>>>>  	do {
>>>>>>  		struct sk_buff *skb;
>>>>>>  		size_t skb_len;
>>>>>>  
>>>>>> -		skb_len = min_t(u32, VIRTIO_VSOCK_MAX_PKT_BUF_SIZE, rest_len);
>>>>>> +		skb_len = min(max_skb_len, rest_len);
>>>>>>  
>>>>>> -		skb = virtio_transport_alloc_skb(info, skb_len,
>>>>>> +		skb = virtio_transport_alloc_skb(vsk, info, skb_len, can_zcopy,
>>>>>>  						 src_cid, src_port,
>>>>>>  						 dst_cid, dst_port);
>>>>>>  		if (!skb) {
>>>>>> @@ -270,6 +400,17 @@ static int virtio_transport_send_pkt_info(struct vsock_sock *vsk,
>>>>>>  			break;
>>>>>>  		}
>>>>>>  
>>>>>> +		/* This is last skb to send this portion of data. */
>>>>>> +		if (info->msg && info->msg->msg_flags & MSG_ZEROCOPY &&
>>>>>> +		    skb_len == rest_len && info->op == VIRTIO_VSOCK_OP_RW) {
>>>>>> +			if (virtio_transport_init_zcopy_skb(vsk, skb,
>>>>>> +							    info->msg,
>>>>>> +							    can_zcopy)) {
>>>>>> +				ret = -ENOMEM;
>>>>>> +				break;
>>>>>> +			}
>>>>>> +		}
>>>>>> +
>>>>>>  		virtio_transport_inc_tx_pkt(vvs, skb);
>>>>>>  
>>>>>>  		ret = t_ops->send_pkt(skb);
>>>>>> @@ -934,7 +1075,7 @@ static int virtio_transport_reset_no_sock(const struct virtio_transport *t,
>>>>>>  	if (!t)
>>>>>>  		return -ENOTCONN;
>>>>>>  
>>>>>> -	reply = virtio_transport_alloc_skb(&info, 0,
>>>>>> +	reply = virtio_transport_alloc_skb(NULL, &info, 0, false,
>>>>>>  					   le64_to_cpu(hdr->dst_cid),
>>>>>>  					   le32_to_cpu(hdr->dst_port),
>>>>>>  					   le64_to_cpu(hdr->src_cid),
>>>>>> -- 
>>>>>> 2.25.1
>>>>>
>>>
> 

Powered by blists - more mailing lists