2023-09-14 14:48:50

by Arseniy Krasnov

[permalink] [raw]
Subject: Re: [PATCH net-next v8 4/4] vsock/virtio: MSG_ZEROCOPY flag support



On 14.09.2023 16:57, Stefano Garzarella wrote:
> On Mon, Sep 11, 2023 at 11:22:34PM +0300, Arseniy Krasnov wrote:
>> This adds handling of MSG_ZEROCOPY flag on transmission path:
>>
>> 1) 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()'.
>> 2) Replaces way of skb owning: instead of '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()'.
>> 3) Adds new callback to 'struct virtio_transport': 'can_msgzerocopy'.
>>   If this callback is set, then transport needs extra check to be able
>>   to send provided number of buffers in zerocopy mode. Currently, the
>>   only transport that needs this callback set is virtio, because this
>>   transport adds new buffers to the virtio queue and we need to check,
>>   that number of these buffers is less than size of the queue (it is
>>   required by virtio spec). vhost and loopback transports don't need
>>   this check.
>>
>> Signed-off-by: Arseniy Krasnov <[email protected]>
>> ---
>> 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.
>> v3 -> v4:
>>  * 'msgzerocopy_check_iov' moved from 'struct vsock_transport' to
>>    'struct virtio_transport' as it is virtio specific callback and
>>    never needed in other transports.
>> v4 -> v5:
>>  * 'msgzerocopy_check_iov' renamed to 'can_msgzerocopy' and now it
>>    uses number of buffers to send as input argument. I think there is
>>    no need to pass iov to this callback (at least today, it is used only
>>    by guest side of virtio transport), because the only thing that this
>>    callback does is comparison of number of buffers to be inserted to
>>    the tx queue and size of this queue.
>>  * Remove any checks for type of current 'iov_iter' with payload (is it
>>    'iovec' or 'ubuf'). These checks left from the earlier versions where I
>>    didn't use already implemented kernel API which handles every type of
>>    'iov_iter'.
>> v5 -> v6:
>>  * Refactor 'virtio_transport_fill_skb()'.
>>  * Add 'WARN_ON_ONCE()' and comment on invalid combination of destination
>>    socket and payload in 'virtio_transport_alloc_skb()'.
>> v7 -> v8:
>>  * Move '+1' addition from 'can_msgzerocopy' callback body to the caller.
>>    This addition means packet header.
>>  * In 'virtio_transport_can_zcopy()' rename 'max_to_send' argument to
>>    'pkt_len'.
>>  * Update commit message by adding details about new 'can_msgzerocopy'
>>    callback.
>>  * In 'virtio_transport_init_hdr()' move 'len' argument directly after
>>    'info'.
>>  * Add comment about processing last skb in tx loop.
>>  * Update comment for 'can_msgzerocopy' callback for more details.
>>
>> include/linux/virtio_vsock.h            |   9 +
>> net/vmw_vsock/virtio_transport.c        |  32 +++
>> net/vmw_vsock/virtio_transport_common.c | 256 ++++++++++++++++++------
>> 3 files changed, 239 insertions(+), 58 deletions(-)
>>
>> diff --git a/include/linux/virtio_vsock.h b/include/linux/virtio_vsock.h
>> index a91fbdf233e4..ebb3ce63d64d 100644
>> --- a/include/linux/virtio_vsock.h
>> +++ b/include/linux/virtio_vsock.h
>> @@ -160,6 +160,15 @@ struct virtio_transport {
>>
>>     /* Takes ownership of the packet */
>>     int (*send_pkt)(struct sk_buff *skb);
>> +
>> +    /* Used in MSG_ZEROCOPY mode. Checks, that provided data
>> +     * (number of buffers) could be transmitted with zerocopy
>> +     * mode. If this callback is not implemented for the current
>> +     * transport - this means that this transport doesn't need
>> +     * extra checks and can perform zerocopy transmission by
>> +     * default.
>> +     */
>> +    bool (*can_msgzerocopy)(int bufs_num);
>> };
>>
>> ssize_t
>> diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
>> index 73d730156349..09ba3128e759 100644
>> --- a/net/vmw_vsock/virtio_transport.c
>> +++ b/net/vmw_vsock/virtio_transport.c
>> @@ -455,6 +455,37 @@ static void virtio_vsock_rx_done(struct virtqueue *vq)
>>     queue_work(virtio_vsock_workqueue, &vsock->rx_work);
>> }
>>
>> +static bool virtio_transport_can_msgzerocopy(int bufs_num)
>> +{
>> +    struct virtio_vsock *vsock;
>> +    bool res = false;
>> +
>> +    rcu_read_lock();
>> +
>> +    vsock = rcu_dereference(the_virtio_vsock);
>> +    if (vsock) {
>> +        struct virtqueue *vq = vsock->vqs[VSOCK_VQ_TX];
>> +
>> +        /* 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.
>> +         */
>> +        if (bufs_num <= vq->num_max)
>> +            res = true;
>> +    }
>> +
>> +    rcu_read_unlock();
>> +
>> +    return res;
>> +}
>> +
>> static bool virtio_transport_seqpacket_allow(u32 remote_cid);
>>
>> static struct virtio_transport virtio_transport = {
>> @@ -504,6 +535,7 @@ static struct virtio_transport virtio_transport = {
>>     },
>>
>>     .send_pkt = virtio_transport_send_pkt,
>> +    .can_msgzerocopy = virtio_transport_can_msgzerocopy,
>> };
>>
>> static bool virtio_transport_seqpacket_allow(u32 remote_cid)
>> diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c
>> index 3a48e48a99ac..e358f118b07e 100644
>> --- a/net/vmw_vsock/virtio_transport_common.c
>> +++ b/net/vmw_vsock/virtio_transport_common.c
>> @@ -37,73 +37,110 @@ 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.
>> - */
>
> Why we are removing this comment?

I think I can return it like:
/* Returns a new sk_buff on success, otherwise returns NULL. */

Line related to 'errp' is not needed i guess.

>
>> -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 pkt_len)
>> +{
>> +    const struct virtio_transport *t_ops;
>> +    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;
>> +    if (iov_iter->iov_offset)
>> +        return false;
>>
>> -        if (msg_data_left(info->msg) == 0 &&
>> -            info->type == VIRTIO_VSOCK_TYPE_SEQPACKET) {
>> -            hdr->flags |= cpu_to_le32(VIRTIO_VSOCK_SEQ_EOM);
>> +    /* We can't send whole iov. */
>> +    if (iov_iter->count > pkt_len)
>> +        return false;
>>
>> -            if (info->msg->msg_flags & MSG_EOR)
>> -                hdr->flags |= cpu_to_le32(VIRTIO_VSOCK_SEQ_EOR);
>> -        }
>> +    /* Check that transport can send data in zerocopy mode. */
>> +    t_ops = virtio_transport_get_ops(info->vsk);
>
> While reviewing I was wondering why here we don't check if `t_ops` is
> NULL.
>
> Then I realized that the only caller of this function
> (virtio_transport_send_pkt_info()) already get the vsk ops calling
> virtio_transport_get_ops() and also checks if it can be null.
>
> So what about passing the ops as function parameter to avoid to call
> virtio_transport_get_ops() again?

Ack

>
>> +
>> +    if (t_ops->can_msgzerocopy) {
>> +        int pages_in_iov = iov_iter_npages(iov_iter, MAX_SKB_FRAGS);
>> +        int pages_to_send = min(pages_in_iov, MAX_SKB_FRAGS);
>> +
>> +        /* +1 is for packet header. */
>> +        return t_ops->can_msgzerocopy(pages_to_send + 1);
>>     }
>>
>> -    if (info->reply)
>> -        virtio_vsock_skb_set_reply(skb);
>> +    return true;
>> +}
>>
>> -    trace_virtio_transport_alloc_pkt(src_cid, src_port,
>> -                     dst_cid, dst_port,
>> -                     len,
>> -                     info->type,
>> -                     info->op,
>> -                     info->flags);
>> +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 (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;
>> +    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;
>> +
>> +        uarg = msg_zerocopy_realloc(sk_vsock(vsk),
>> +                        iter->count,
>> +                        NULL);
>> +        if (!uarg)
>> +            return -1;
>> +
>> +        uarg_zc = uarg_to_msgzc(uarg);
>> +        uarg_zc->zerocopy = zerocopy ? 1 : 0;
>>     }
>>
>> -    return skb;
>> +    skb_zcopy_init(skb, uarg);
>>
>> -out:
>> -    kfree_skb(skb);
>> -    return NULL;
>> +    return 0;
>> +}
>> +
>> +static int virtio_transport_fill_skb(struct sk_buff *skb,
>> +                     struct virtio_vsock_pkt_info *info,
>> +                     size_t len,
>> +                     bool zcopy)
>> +{
>> +    void *payload;
>> +    int err;
>> +
>> +    if (zcopy)
>> +        return __zerocopy_sg_from_iter(info->msg, NULL, skb,
>> +                           &info->msg->msg_iter,
>> +                           len);
>> +
>> +    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;
>
> We are returning 0 in any case, what is the purpose of this check?

Right, just forget to remove it.

>
>> +
>> +    return 0;
>> +}
>> +
>> +static void virtio_transport_init_hdr(struct sk_buff *skb,
>> +                      struct virtio_vsock_pkt_info *info,
>> +                      size_t payload_len,
>> +                      u32 src_cid,
>> +                      u32 src_port,
>> +                      u32 dst_cid,
>> +                      u32 dst_port)
>> +{
>> +    struct virtio_vsock_hdr *hdr;
>> +
>> +    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(payload_len);
>> }
>>
>> static void virtio_transport_copy_nonlinear_skb(const struct sk_buff *skb,
>> @@ -214,6 +251,77 @@ 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,
>
> Before this patch we used `info->vsk` in virtio_transport_alloc_skb().
> Is it now really necessary to add vsk as a parameter? If so, why?

Agree, I think previous version of virtio_transport_alloc_skb() could be
used.

>
>> +                          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, payload_len, src_cid, src_port,
>> +                  dst_cid, dst_port);
>> +
>> +    /* If 'vsk' != NULL then payload is always present, so we
>> +     * will never call '__zerocopy_sg_from_iter()' below without
>> +     * setting skb owner in 'skb_set_owner_w()'. The only case
>> +     * when 'vsk' == NULL is VIRTIO_VSOCK_OP_RST control message
>> +     * without payload.
>> +     */
>> +    WARN_ON_ONCE(!(vsk && (info->msg && payload_len)) && zcopy);
>> +
>> +    /* 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) {
>
> Before this patch, we did these steps only if
> `msg_data_left(info->msg) == 0`, why now we do it in any case?

Ack, it is bug

>
>> +            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);
>
> Maybe now we should trace also `zcopy`.

Ack

>
>> +
>> +    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 +330,12 @@ 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 +364,30 @@ 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 +395,21 @@ static int virtio_transport_send_pkt_info(struct vsock_sock *vsk,
>>             break;
>>         }
>>
>> +        /* We process buffer part by part, allocating skb on
>> +         * each iteration. If this is last skb for this buffer
>> +         * and MSG_ZEROCOPY mode is in use - we must allocate
>> +         * completion for the current syscall.
>> +         */
>> +        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);
>> @@ -985,7 +1125,7 @@ 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
>>
>