[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [virtio-comment] Re: [PATCH] virtio-net: define feature of per-packet RSS hash delivery
On Wed, Dec 11, 2019 at 8:20 PM Michael S. Tsirkin <mst@redhat.com> wrote:
>
> On Wed, Dec 11, 2019 at 07:41:59PM +0200, Yuri Benditovich wrote:
> > On Wed, Dec 11, 2019 at 6:35 PM Michael S. Tsirkin <mst@redhat.com> wrote:
> > >
> > > On Wed, Dec 04, 2019 at 12:20:01AM -0500, Yuri Benditovich wrote:
> > > >
> > > >
> > > > âââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââ
> > > >
> > > > From: "Michael S. Tsirkin" <mst@redhat.com>
> > > > To: "Yuri Benditovich" <yuri.benditovich@daynix.com>
> > > > Cc: virtio-comment@lists.oasis-open.org
> > > > Sent: Tuesday, December 3, 2019 10:25:30 PM
> > > > Subject: [virtio-comment] Re: [PATCH] virtio-net: define feature of
> > > > per-packet RSS hash delivery
> > > >
> > > > On Tue, Dec 03, 2019 at 08:32:12PM +0200, Yuri Benditovich wrote:
> > > > > Fixes: https://github.com/oasis-tcs/virtio-spec/issues/66
> > > > > Conditional extending of virtio header structure to deliver
> > > > > packet's hash and hash type used for calculation.
> > > > >
> > > > > Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
> > > >
> > > > Functionally this looks mostly good. But we need to change the
> > > > way it's documented slightly.
> > > >
> > > > Probably, because the RSS patch is not merged yet.
> > > >
> > > >
> > > > > ---
> > > > > content.tex | 24 ++++++++++++++++++++++++
> > > > > 1 file changed, 24 insertions(+)
> > > > >
> > > > > diff --git a/content.tex b/content.tex
> > > > > index 01be7df..44975e4 100644
> > > > > --- a/content.tex
> > > > > +++ b/content.tex
> > > > > @@ -2811,6 +2811,9 @@ \subsection{Feature bits}\label{sec:Device Types /
> > > > Network Device / Feature bits
> > > > > \item[VIRTIO_NET_F_CTRL_MAC_ADDR(23)] Set MAC address through control
> > > > > channel.
> > > > >
> > > > > +\item[VIRTIO_NET_F_HASH_REPORT(58)] Device can report per-packet hash
> > > > value
> > > > > + and a type of calculated hash
> > > > > +
> > > > > \item[VIRTIO_NET_F_RSS(60)] Device supports RSS (receive-side scaling)
> > > > > with Toeplitz hash calculation and configurable hash parameters for
> > > > receive steering
> > > > >
> > > > > @@ -2844,6 +2847,8 @@ \subsubsection{Feature bit requirements}\label
> > > > {sec:Device Types / Network Device
> > > > > \item[VIRTIO_NET_F_CTRL_MAC_ADDR] Requires VIRTIO_NET_F_CTRL_VQ.
> > > > > \item[VIRTIO_NET_F_RSC_EXT] Requires VIRTIO_NET_F_HOST_TSO4 or
> > > > VIRTIO_NET_F_HOST_TSO6.
> > > > > \item[VIRTIO_NET_F_RSS] Requires VIRTIO_NET_F_CTRL_VQ.
> > > > > +\item[VIRTIO_NET_F_HASH_REPORT] Requires VIRTIO_NET_F_RSS.
> > > > > +
> > > > > \end{description}
> > > > >
> > > > > \subsubsection{Legacy Interface: Feature bits}\label{sec:Device Types /
> > > > Network Device / Feature bits / Legacy Interface: Feature bits}
> > > > > @@ -3071,6 +3076,8 @@ \subsection{Device Operation}\label{sec:Device
> > > > Types / Network Device / Device O
> > > > > le16 csum_start;
> > > > > le16 csum_offset;
> > > > > le16 num_buffers;
> > > > > + le32 hash_value; (Only if VIRTIO_NET_F_HASH_REPORT negotiated)
> > > > > + le16 hash_type; (Only if VIRTIO_NET_F_HASH_REPORT negotiated)
> > > > > };
> > > > > \end{lstlisting}
> > > > >
> > > > > @@ -3330,6 +3337,15 @@ \subsubsection{Processing of Incoming Packets}\
> > > > label{sec:Device Types / Network
> > > > > set: if so, device has validated the packet checksum.
> > > > > In case of multiple encapsulated protocols, one level of checksums
> > > > > has been validated.
> > > > > +\item If VIRTIO_NET_F_HASH_REPORT was negotiated and a device has
> > > > calculated
> > > > > + a hash for the packet, \field{hash_value} contains calculated hash
> > > > value and
> > > > > + \field{hash_type} contains exact hash type.
> > > > > +
> > > > > + If the hash was not calculated, \field{hash_type} contains zero.
> > > > > +
> > > > > + For defined hash types and their meaning, see \ref{sec:Device Types /
> > > > Network Device / Device configuration layout / RSS}.
> > > > > +
> > > > > + For the procedure of hash calculation, see \ref{sec:Device Types /
> > > > Network Device / Device Operation / Control Virtqueue / Receive-side
> > > > scaling (RSS) / RSS hash types}.
> > > > > \end{enumerate}
> > > > >
> > > > > Additionally, VIRTIO_NET_F_GUEST_CSUM, TSO4, TSO6, UDP and ECN
> > > > > @@ -3909,6 +3925,14 @@ \subsubsection{Control Virtqueue}\label{sec:Device
> > > > Types / Network Device / Devi
> > > > > \item Else skip IPv6 extension headers and calculate the hash as defined
> > > > above for IPv6 packet without extension headers
> > > > > \end{itemize}
> > > > >
> > > > > +If VIRTIO_NET_F_HASH_REPORT was negotiated, the device reports
> > > > calculated hash information in fields of virtio_net_hdr as follows:
> > > > > +
> > > > > +Exact hash type is populated in \field{hash_type}
> > > > > +
> > > > > +Hash value is populated in \field{hash_value}
> > > > > +
> > > > > +If, due to any reason, the device did not calculate the hash, it sets \
> > > > field{hash_type} to zero.
> > > > > +
> > > > > \drivernormative{\subparagraph}{Setting RSS parameters}{Device Types /
> > > > Network Device / Device Operation / Control Virtqueue / Receive-side
> > > > scaling (RSS) }
> > > > >
> > > > > A driver MUST NOT send the VIRTIO_NET_CTRL_MQ_RSS_CONFIG command if the
> > > > feature VIRTIO_NET_F_RSS has not been negotiated.
> > > >
> > > >
> > > > This is understandably a minimal change, but imho messes things up:
> > > > hash calculation is no longer part of RSS so really should be moved
> > > > from out of there.
> > > >
> > > > Please advise why "hash calculation is no longer part of RSS".
> > >
> > > I might be misunderstanding things. It looks like with this
> > > one can enable hash even with a single RX queue.
> > > No?
> >
> > Not exactly.
> > Hash calculation (even with single queue, which is possible corner
> > case) requires a key to be configured with
> > VIRTIO_NET_CTRL_MQ_RSS_CONFIG command.
> > So, if the device with one queue is capable to deliver the hash it
> > indicates VIRTIO_NET_F_RSS and VIRTIO_NET_F_HASH_REPORT.
> > I will add this.
> > And probably the device should deny VIRTIO_NET_F_HASH_REPORT without
> > VIRTIO_NET_F_RSS.
>
> Oh I see. Interesting. Do we want to tie the two features like this though?
> I wonder why?
> If yes we can certainly make VIRTIO_NET_F_HASH_REPORT depend on RSS ...
> If not there could be a variant of VIRTIO_NET_CTRL_MQ_RSS_CONFIG
> without RSS tables that's allowed without RSS.
>
IMO:
VIRTIO_NET_CTRL_MQ_RSS_CONFIG defines 2 things mandatory for hash calculation:
1) a key
2) set of allowed hashes - this may change the calculation even on the
same packet
So IMO, VIRTIO_NET_F_HASH_REPORT depends on VIRTIO_NET_F_RSS.
If you have a use case that make sense of having only
VIRTIO_NET_F_HASH_REPORT, please describe it.
> > >
> > > >
> > > >
> > > > So we'd have a chapter for hash calculation, defining hash types
> > > > etc.
> > > >
> > > > A chapter for hash calculation is in previous patch
> > > > https://lists.oasis-open.org/archives/virtio-comment/201911/msg00014.html
> > > >
> > >
> > > Can't see it there. Do you mean
> > > +\subparagraph{RSS hash types}\label{sec:Device Types / Network Device / Device Operation / Control Virtqueue / Receive-side scaling (RSS) / RSS hash types}
> > > ?
> > > Sure, what I am saying it needs to be moved from out of RSS then.
> > >
> >
> > I do not see how hash calculation can be moved out of RSS, it requires
> > the configuration that comes with VIRTIO_NET_CTRL_MQ_RSS_CONFIG.
>
> Allow VIRTIO_NET_CTRL_MQ_RSS_CONFIG with no RSS parameters
> when RSS is off?
>
When in VIRTIO_NET_CTRL_MQ_RSS_CONFIG the max_tx_vq=1:
* if VIRTIO_NET_F_HASH_REPORT is not negotiated the rest of parameters
can be ignored as there is nothing to do with them.
* if VIRTIO_NET_F_HASH_REPORT is negotiated, a key and hash_types are used.
So, for consistency, I would not invent new format of
VIRTIO_NET_CTRL_MQ_RSS_CONFIG
> > > >
> > > > Then both RSS and the new chapter refer to that.
> > > >
> > > > Also, I am guessing supported hash types should be valid
> > > > in config space with the new feature - isn't that right?
> > > >
> > > > Yes, of course.
> > >
> > > So we need to say this in the spec I think.
> >
> > OK, I will add this.
> > To avoid misunderstandings I think it is better to wait until the RSS
> > patch (mentioned above) is merged.
> > When approx. we can expect that?
>
> Could of days, my laptop was in repair, so there's a backlog of
> things to merge.
>
> > >
> > > > But this is not related to hash delivery.
> > > > Packet hash is calculated according to allowed hash types.
> > > > If hash delivery is negotiated, the device reports the calculated hash type
> > > > which is, naturally, one of valid ones.
> > > >
> > > >
> > > >
> > > > > --
> > > > > 2.17.2
> > > >
> > > >
> > > > This publicly archived list offers a means to provide input to the
> > > > OASIS Virtual I/O Device (VIRTIO) TC.
> > > >
> > > > In order to verify user consent to the Feedback License terms and
> > > > to minimize spam in the list archive, subscription is required
> > > > before posting.
> > > >
> > > > Subscribe: virtio-comment-subscribe@lists.oasis-open.org
> > > > Unsubscribe: virtio-comment-unsubscribe@lists.oasis-open.org
> > > > List help: virtio-comment-help@lists.oasis-open.org
> > > > List archive: https://lists.oasis-open.org/archives/virtio-comment/
> > > > Feedback License: https://www.oasis-open.org/who/ipr/feedback_license.pdf
> > > > List Guidelines: https://www.oasis-open.org/policies-guidelines/
> > > > mailing-lists
> > > > Committee: https://www.oasis-open.org/committees/virtio/
> > > > Join OASIS: https://www.oasis-open.org/join/
> > > >
> > > >
> > > >
> > >
>
>
> This publicly archived list offers a means to provide input to the
> OASIS Virtual I/O Device (VIRTIO) TC.
>
> In order to verify user consent to the Feedback License terms and
> to minimize spam in the list archive, subscription is required
> before posting.
>
> Subscribe: virtio-comment-subscribe@lists.oasis-open.org
> Unsubscribe: virtio-comment-unsubscribe@lists.oasis-open.org
> List help: virtio-comment-help@lists.oasis-open.org
> List archive: https://lists.oasis-open.org/archives/virtio-comment/
> Feedback License: https://www.oasis-open.org/who/ipr/feedback_license.pdf
> List Guidelines: https://www.oasis-open.org/policies-guidelines/mailing-lists
> Committee: https://www.oasis-open.org/committees/virtio/
> Join OASIS: https://www.oasis-open.org/join/
>
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]