virtio-comment

 View Only
  • 1.  [PATCH v1 0/2] virtio-net: Clarify virtio_net_hdr size and rx buffer size

    Posted 01-05-2024 08:01
    There are few small issues around size of receive buffer and size of the
    struct virtio_net_hdr as following.

    1. VIRTIO_NET_F_HASH_REPORT is only applicable on the rx side processing.
    However it changes the size of the struct virtio_net_hdr for the tx side
    as well. This was not clarified when VIRTIO_NET_F_HASH_REPORT feature was
    introduced.

    2. Receive buffer size calculation description and requirements did not
    reflect negotiation of VIRTIO_NET_F_HASH_REPORT.

    3. Receive buffer size calculation was incorrect for ipv6 guest gso feature.

    Above issue 2 was reported in [1].
    Since issue 2 and 3 both touches the same set of requirements and description,
    they are updated under a new issue along with tx side fix as well touching the
    clarification needed for VIRTIO_NET_F_HASH_REPORT.

    Hence above three issues are fixed under issue of [2].

    Patch summary:
    patch-1 fixes receive side description and requirement
    patch-2 adds transmit side clarification for VIRTIO_NET_F_HASH_REPORT

    changelog:
    v0->v1:
    - addressed comments from Xuan to rewrite tx normative

    Parav Pandit (2):
    virtio-net: Fix receive buffer size calculation text
    virtio-net: Clarify the size of the struct virtio_net_hdr for tx

    device-types/net/description.tex | 28 +++++++++++++++++++++-------
    1 file changed, 21 insertions(+), 7 deletions(-)

    --
    2.34.1




  • 2.  [PATCH v1 1/2] virtio-net: Fix receive buffer size calculation text

    Posted 01-05-2024 08:01
    Receive buffer size calculation is based on the following
    negotiated features.

    The text has wrong calculation for IPv6 and also it has missed
    VIRTIO_NET_F_HASH_REPORT.

    The problem of igorance of VIRTIO_NET_F_HASH_REPORT is reported
    in [1], however fix for ipv6 payload length must also be
    considered.

    Since for the both the fixes touching same requirements, a
    new issue is created as [2].

    This patch brings following fixes.

    1. Fix annotating struct virtio_net_hdr as field
    2. Fix receive buffer calculation for guest GSO cases to consider
    ipv6 payload length
    3. small grammar corrections for article
    4. reword the requirement to consider the virtio_ndr_hdr which is
    depends on the negotiated feature, hence first clarify the
    struct virtio_net_hdr size

    [1] https://github.com/oasis-tcs/virtio-spec/issues/170
    [2] https://github.com/oasis-tcs/virtio-spec/issues/183

    Fixes: https://github.com/oasis-tcs/virtio-spec/issues/170
    Fixes: https://github.com/oasis-tcs/virtio-spec/issues/183
    Signed-off-by: Parav Pandit <parav@nvidia.com>
    ---
    device-types/net/description.tex | 25 ++++++++++++++++++-------
    1 file changed, 18 insertions(+), 7 deletions(-)

    diff --git a/device-types/net/description.tex b/device-types/net/description.tex
    index aff5e08..a1f40f7 100644
    --- a/device-types/net/description.tex
    +++ b/device-types/net/description.tex
    @@ -657,24 +657,35 @@ \subsubsection{Setting Up Receive Buffers}\label{sec:Device Types / Network Devi
    If the VIRTIO_NET_F_GUEST_TSO4, VIRTIO_NET_F_GUEST_TSO6,
    VIRTIO_NET_F_GUEST_UFO, VIRTIO_NET_F_GUEST_USO4 or VIRTIO_NET_F_GUEST_USO6
    features are used, the maximum incoming packet
    -will be to 65550 bytes long (the maximum size of a
    -TCP or UDP packet, plus the 14 byte ethernet header), otherwise
    -1514 bytes. The 12-byte struct virtio_net_hdr is prepended to this,
    -making for 65562 or 1526 bytes.
    +will of size 65589 bytes long (14 bytes of Ethernet header, plus 40 bytes of
    +the IPv6 header, plus 65535 bytes of maximum IPv6 payload including any
    +extension header) otherwise 1514 bytes.
    +When VIRTIO_NET_F_HASH_REPORT is not negotiated, the required receive buffer
    +size is either 65601 or 1526 bytes.
    +When VIRTIO_NET_F_HASH_REPORT is negotiated, the required receive buffer
    +size is either 65609 or 1534 bytes.

    \drivernormative{\paragraph}{Setting Up Receive Buffers}{Device Types / Network Device / Device Operation / Setting Up Receive Buffers}

    +If VIRTIO_NET_F_HASH_REPORT is not negotiated, the size of the field
    +\field{struct virtio_net_hdr} is 12 bytes.
    +
    +If VIRTIO_NET_F_HASH_REPORT is negotiated, the size of the field
    +\field{struct virtio_net_hdr} is 20 bytes.
    +
    \begin{itemize}
    \item If VIRTIO_NET_F_MRG_RXBUF is not negotiated:
    \begin{itemize}
    \item If VIRTIO_NET_F_GUEST_TSO4, VIRTIO_NET_F_GUEST_TSO6, VIRTIO_NET_F_GUEST_UFO,
    VIRTIO_NET_F_GUEST_USO4 or VIRTIO_NET_F_GUEST_USO6 are negotiated, the driver SHOULD populate
    - the receive queue(s) with buffers of at least 65562 bytes.
    + the receive queue(s) with buffers of at least 65589 bytes plus the size
    + of the \field{struct virtio_net_hdr}.
    \item Otherwise, the driver SHOULD populate the receive queue(s)
    - with buffers of at least 1526 bytes.
    + with buffers of at least 1514 bytes plus the size of the
    + \field{struct virtio_net_hdr}.
    \end{itemize}
    \item If VIRTIO_NET_F_MRG_RXBUF is negotiated, each buffer MUST be at
    -least the size of the struct virtio_net_hdr.
    +least the size of the \field{struct virtio_net_hdr}.
    \end{itemize}

    \begin{note}
    --
    2.34.1




  • 3.  Re: [PATCH v1 1/2] virtio-net: Fix receive buffer size calculation text

    Posted 01-12-2024 11:24
    On Fri, Jan 05 2024, Parav Pandit <parav@nvidia.com> wrote:

    > Receive buffer size calculation is based on the following
    > negotiated features.
    >
    > The text has wrong calculation for IPv6 and also it has missed
    > VIRTIO_NET_F_HASH_REPORT.
    >
    > The problem of igorance of VIRTIO_NET_F_HASH_REPORT is reported
    > in [1], however fix for ipv6 payload length must also be
    > considered.
    >
    > Since for the both the fixes touching same requirements, a
    > new issue is created as [2].
    >
    > This patch brings following fixes.
    >
    > 1. Fix annotating struct virtio_net_hdr as field
    > 2. Fix receive buffer calculation for guest GSO cases to consider
    > ipv6 payload length
    > 3. small grammar corrections for article
    > 4. reword the requirement to consider the virtio_ndr_hdr which is
    > depends on the negotiated feature, hence first clarify the
    > struct virtio_net_hdr size
    >
    > [1] https://github.com/oasis-tcs/virtio-spec/issues/170
    > [2] https://github.com/oasis-tcs/virtio-spec/issues/183
    >
    > Fixes: https://github.com/oasis-tcs/virtio-spec/issues/170
    > Fixes: https://github.com/oasis-tcs/virtio-spec/issues/183
    > Signed-off-by: Parav Pandit <parav@nvidia.com>
    > ---
    > device-types/net/description.tex | 25 ++++++++++++++++++-------
    > 1 file changed, 18 insertions(+), 7 deletions(-)
    >
    > diff --git a/device-types/net/description.tex b/device-types/net/description.tex
    > index aff5e08..a1f40f7 100644
    > --- a/device-types/net/description.tex
    > +++ b/device-types/net/description.tex
    > @@ -657,24 +657,35 @@ \subsubsection{Setting Up Receive Buffers}\label{sec:Device Types / Network Devi
    > If the VIRTIO_NET_F_GUEST_TSO4, VIRTIO_NET_F_GUEST_TSO6,
    > VIRTIO_NET_F_GUEST_UFO, VIRTIO_NET_F_GUEST_USO4 or VIRTIO_NET_F_GUEST_USO6
    > features are used, the maximum incoming packet
    > -will be to 65550 bytes long (the maximum size of a
    > -TCP or UDP packet, plus the 14 byte ethernet header), otherwise
    > -1514 bytes. The 12-byte struct virtio_net_hdr is prepended to this,
    > -making for 65562 or 1526 bytes.
    > +will of size 65589 bytes long (14 bytes of Ethernet header, plus 40 bytes of

    Either "will be 65589 bytes long" or "will be of 65589 bytes size" (I'd
    prefer the former.)

    > +the IPv6 header, plus 65535 bytes of maximum IPv6 payload including any
    > +extension header) otherwise 1514 bytes.

    s/ otherwise/, otherwise/

    > +When VIRTIO_NET_F_HASH_REPORT is not negotiated, the required receive buffer
    > +size is either 65601 or 1526 bytes.
    > +When VIRTIO_NET_F_HASH_REPORT is negotiated, the required receive buffer
    > +size is either 65609 or 1534 bytes.
    >
    > \drivernormative{\paragraph}{Setting Up Receive Buffers}{Device Types / Network Device / Device Operation / Setting Up Receive Buffers}
    >
    > +If VIRTIO_NET_F_HASH_REPORT is not negotiated, the size of the field
    > +\field{struct virtio_net_hdr} is 12 bytes.
    > +
    > +If VIRTIO_NET_F_HASH_REPORT is negotiated, the size of the field
    > +\field{struct virtio_net_hdr} is 20 bytes.

    This reads a bit odd because these two are not normative statements. I'd
    rather include this information into the actual statements.

    > +
    > \begin{itemize}
    > \item If VIRTIO_NET_F_MRG_RXBUF is not negotiated:
    > \begin{itemize}
    > \item If VIRTIO_NET_F_GUEST_TSO4, VIRTIO_NET_F_GUEST_TSO6, VIRTIO_NET_F_GUEST_UFO,
    > VIRTIO_NET_F_GUEST_USO4 or VIRTIO_NET_F_GUEST_USO6 are negotiated, the driver SHOULD populate
    > - the receive queue(s) with buffers of at least 65562 bytes.
    > + the receive queue(s) with buffers of at least 65589 bytes plus the size
    > + of the \field{struct virtio_net_hdr}.

    "with buffers of at least 65609 bytes if VIRTIO_NET_F_HASH_REPORT is
    negotiated, and of at least 65601 bytes if not."

    > \item Otherwise, the driver SHOULD populate the receive queue(s)
    > - with buffers of at least 1526 bytes.
    > + with buffers of at least 1514 bytes plus the size of the
    > + \field{struct virtio_net_hdr}.

    "with buffers of at least 1534 bytes if VIRTIO_NET_F_HASH_REPORT is
    negotiated, and of at least 1526 bytes if not."

    > \end{itemize}
    > \item If VIRTIO_NET_F_MRG_RXBUF is negotiated, each buffer MUST be at
    > -least the size of the struct virtio_net_hdr.
    > +least the size of the \field{struct virtio_net_hdr}.

    If you reformat it like that, I'd drop the "the", and also state the
    lenght explictly:

    "each buffer MUST be at least the size of \field{struct virtio_net_hdr},
    i.e. 20 bytes if VIRTIO_NET_F_HASH_REPORT is negotiated, and 12 bytes if
    not."




  • 4.  [PATCH v1 2/2] virtio-net: Clarify the size of the struct virtio_net_hdr for tx

    Posted 01-05-2024 08:01
    The feature VIRTIO_NET_F_HASH_REPORT only applies to the receive side.
    However, when VIRTIO_NET_F_HASH_REPORT feature was introduced, it was
    not clarified that the size of the struct virtio_net_hdr on the packet
    transmission also uses higher size when VIRTIO_NET_F_HASH_REPORT is
    negotiated.

    Explicitly clarify this.

    Fixes: https://github.com/oasis-tcs/virtio-spec/issues/183
    Signed-off-by: Parav Pandit <parav@nvidia.com>

    ---
    changelog:
    v0->v1:
    - addressed comments from Xuan to rewrite tx normative
    ---
    device-types/net/description.tex | 3 +++
    1 file changed, 3 insertions(+)

    diff --git a/device-types/net/description.tex b/device-types/net/description.tex
    index a1f40f7..d86beee 100644
    --- a/device-types/net/description.tex
    +++ b/device-types/net/description.tex
    @@ -528,6 +528,9 @@ \subsubsection{Packet Transmission}\label{sec:Device Types / Network Device / De

    \drivernormative{\paragraph}{Packet Transmission}{Device Types / Network Device / Device Operation / Packet Transmission}

    +The driver MUST use the same size of the field \field{struct virtio_net_hdr}
    +for transmit and receive packet buffers.
    +
    The driver MUST set \field{num_buffers} to zero.

    If VIRTIO_NET_F_CSUM is not negotiated, the driver MUST set
    --
    2.34.1




  • 5.  RE: [PATCH v1 0/2] virtio-net: Clarify virtio_net_hdr size and rx buffer size

    Posted 01-15-2024 06:00
    Hi Cornelia,

    > From: Parav Pandit <parav@nvidia.com>
    > Sent: Friday, January 5, 2024 1:31 PM
    > To: virtio-comment@lists.oasis-open.org; mst@redhat.com;
    > cohuck@redhat.com
    > Cc: Shahaf Shuler <shahafs@nvidia.com>; xuanzhuo@linux.alibaba.com;
    > yuri.benditovich@daynix.com; Parav Pandit <parav@nvidia.com>
    > Subject: [PATCH v1 0/2] virtio-net: Clarify virtio_net_hdr size and rx buffer size
    >
    > There are few small issues around size of receive buffer and size of the struct
    > virtio_net_hdr as following.
    >
    > 1. VIRTIO_NET_F_HASH_REPORT is only applicable on the rx side processing.
    > However it changes the size of the struct virtio_net_hdr for the tx side
    > as well. This was not clarified when VIRTIO_NET_F_HASH_REPORT feature
    > was
    > introduced.
    >
    > 2. Receive buffer size calculation description and requirements did not
    > reflect negotiation of VIRTIO_NET_F_HASH_REPORT.
    >
    > 3. Receive buffer size calculation was incorrect for ipv6 guest gso feature.
    >
    > Above issue 2 was reported in [1].
    > Since issue 2 and 3 both touches the same set of requirements and
    > description, they are updated under a new issue along with tx side fix as well
    > touching the clarification needed for VIRTIO_NET_F_HASH_REPORT.
    >
    > Hence above three issues are fixed under issue of [2].
    >
    > Patch summary:
    > patch-1 fixes receive side description and requirement
    > patch-2 adds transmit side clarification for VIRTIO_NET_F_HASH_REPORT
    >
    > changelog:
    > v0->v1:
    > - addressed comments from Xuan to rewrite tx normative
    >

    Can you please raise the voting request for these two small description changes?

    > Parav Pandit (2):
    > virtio-net: Fix receive buffer size calculation text
    > virtio-net: Clarify the size of the struct virtio_net_hdr for tx
    >
    > device-types/net/description.tex | 28 +++++++++++++++++++++-------
    > 1 file changed, 21 insertions(+), 7 deletions(-)
    >
    > --
    > 2.34.1