Re: [virtio-dev] [PATCH v2 2/2] content: net: steering mode: Add RSS

[Sameeh Jubran <sameeh@daynix.com>]

On Tue, Apr 17, 2018 at 5:14 PM, Michael S. Tsirkin <mst@redhat.com> wrote:

On Tue, Apr 17, 2018 at 02:50:15PM +0300, Sameeh Jubran wrote:
>
>
> On Tue, Apr 17, 2018 at 4:50 AM, Michael S. Tsirkin <mst@redhat.com> wrote:
>
>     On Mon, Apr 16, 2018 at 04:05:26PM +0300, Sameeh Jubran wrote:
>     > From: Sameeh Jubran <sjubran@redhat.com>
>     >
>     > This commit introduces the RSS feature into virtio-net. It is introduced
>     > as a sub mode for a general command which configures the steering mode.
>     >
>     > Most modern high end network devices today support configurable hash
>     functions,
>     > this commit introduces RSS - Receive Side Scaling - [1] to virtio net
>     device.
>     >
>     > The RSS is a technology from Microsoft that boosts network device
>     performance
>     > by efficiently distributing the traffic among the CPUs in a
>     multiprocessor
>     > system.
>     >
>     > This feature is supported in most of the modern network cards as well as
>     most
>     > modern OSes including Linux and Windows. It is worth mentioning that both
>     DPDK
>     > and Hyper-v support RSS too.
>     >
>     > [1] https://docs.microsoft.com/en-us/windows-hardware/drivers/network/
>     ndis-receive-side-scaling2
>     >
>     > Signed-off-by: Sameeh Jubran <sjubran@redhat.com>
>     > ---
>     >  content.tex | 114 ++++++++++++++++++++++++++++++
>     ++++++++++++++++++++++++++++++
>     >  1 file changed, 114 insertions(+)
>     >
>     > diff --git a/content.tex b/content.tex
>     > index 3d538e8..8076147 100644
>     > --- a/content.tex
>     > +++ b/content.tex
>     > @@ -4017,6 +4017,7 @@ The device MUST NOT queue packets on receive queues
>     greater than
>     >  \begin{lstlisting}
>     >  // steering mode flags
>     >  #define STEERING_MODE_AUTO          1
>     > +#define STEERING_MODE_RSS           2
>     > 
>     >  struct virtio_net_steering_modes {
>     >  le32 steering_modes;
>     > @@ -4027,6 +4028,7 @@ le32 steering_mode;
>     >  le32 command;
>     > 
>     >      union {
>     > +    struct virtio_net_rss rss_conf;
>     >      }
>     >  };
>     > 
>     > @@ -4066,6 +4068,118 @@ If this feature has been negotiated, the virtio
>     header has an additional
>     > 
>     >  This is the default steering mode, please refer to the "Automatic
>     receive steering in multiqueue" section.
>     > 
>     > +\subparagraph{Receive Side Scaling}{Device Types / Network Device /
>     Device Operation / Control Virtqueue / Steering mode / Receive Side
>     Scaling}
>     > +
>     > +\begin{lstlisting}
>     > +#define RSS_HASH_FUNCTION_NONE      1
>     > +#define RSS_HASH_FUNCTION_TOEPLITZ  2
>     > +#define RSS_HASH_FUNCTION_SYMMETRIC 3
>     > +
>     > +// Hash function fields
>     > +#define RSS_HASH_FIELDS_IPV4          0x00000100
>     > +#define RSS_HASH_FIELDS_TCP_IPV4      0x00000200
>     > +#define RSS_HASH_FIELDS_IPV6          0x00000400
>     > +#define RSS_HASH_FIELDS_IPV6_EX       0x00000800
>     > +#define RSS_HASH_FIELDS_TCP_IPV6      0x00001000
>     > +#define RSS_HASH_FIELDS_TCP_IPV6_EX   0x00002000
>     > +
>     > +struct virtio_net_rss_supported_hash{
>     > +le32 hash_function;
>     > +}
>     > +
>     > +struct virtio_net_rss {
>     > +le32 hash_function;
>     > +le32 hash_function_flags;
>     > +le32 hash_key_length;
>     > +le32 indirection_table_length;
>     > +     struct {
>     > +             le32 hash_key[hash_key_length];
>     > +             le32 indirection_table[indirection_table_length];
>     > +     }
>     > +};
>     > +
>     > +#define VIRTIO_NET_SM_CTRL_RSS_GET_SUPPORTED_FUNCTIONS   0
>     > +#define VIRTIO_NET_SM_CTRL_RSS_SET                       1
>
>
>     Please add comments in the code as well.
>
> Do you mean that I should add more comments to the code above which explains
> the structures and defines?

Exactly.

>      
>
>   
>     > +\end{lstlisting}
>     > +
>     > +If the VIRTIO_NET_F_CTRL_STEERING_MODE is negotiated the driver can send
>     control
>     > +commands for the RSS configuration.
>
>     Confused.  Does VIRTIO_NET_F_CTRL_STEERING_MODE imply steering mode
>     generally or RSS specifically?
>     How does a device with streering mode ctrl but without RSS look?
>
> The steering mode provides an infrastructure to multiple modes and it is not
> specific to RSS.


Then maybe this sentence should not imply that VIRTIO_NET_F_CTRL_STEERING_MODE
allows RSS.

> The default mode of the steering mode is the Automatic mode
> which is the default mode when MQ is enabled. I though this would be the best
> option to provide backward compatibility and makes sense as we already have
> "Automatic steering mode" in the spec.
> VIRTIO_NET_F_CTRL_STEERING_MODE  gives each mode it's own subset of commands as
> is the case in RSS.
>
>
>
>     > For configuring RSS the virtio_net_steering_mode
>     > +should be filled. The \field{steering_mode} field should be filled with
>     the STEERING_MODE_RSS
>     > +flag along with one of the VIRTIO_NET_SM_CTRL_RSS commands in the \field
>     {command} field. The
>     > +\field{rss_conf} field should be used.
>     > +
>     > +The class VIRTIO_NET_CTRL_RSS has two commands:
>     > +
>     > +\begin{enumerate}
>     > +\item VIRTIO_NET_SM_CTRL_RSS_GET_SUPPORTED_FUNCTIONS returns the hash
>     functions
>     > +     supported by the device to the driver.
>     > +\item VIRTIO_NET_SM_CTRL_RSS_SET applies the new RSS configuration. The
>     command is
>     > +     used by the driver for setting RSS hash function, hash key and
>     > +     indirection table in the device.
>     > +\end{enumerate}
>     > +
>     > +\devicenormative{\subparagraph}{Receive Side Scaling}{Device Types /
>     Network Device / Device Operation / Control Virtqueue / Steering mode /
>     Receive Side Scaling}
>     > +
>     > +The device MUST fill the virtio_net_rss_supported_hash structure with
>     the hash
>     > +functions it supports and return the structure to the driver. Zero or
>     more
>     > +flags of the RSS_HASH_FUNCTION flags MUST be used to fill the \field
>     {hash_function}
>     > +field.
>     > +
>     > +The device MUST drop all previous RSS configuration upon receiving
>     > +VIRTIO_NET_SM_CTRL_RSS_SET command.
>     > +
>     > +The device MUST set the RSS configuration according to the settings
>     provided as
>     > +follows, once the configuration process is completed the device SHOULD
>     apply
>     > +the hash function to each of the incoming packets and distribute the
>     packets
>     > +through the virqueues using the calculated hash and the indirection
>     table
>     > +that were earlier provided by the driver.
>     > +
>     > +Setting RSS configuration
>     > +\begin{enumerate}
>     > +\item The driver fills all of the fields and passes them through the
>     control
>     > +     queue to the device.
>     > +
>     > +\item The device sets the RSS configuration as provided by the driver.
>     > +
>     > +\item If the device successfully applied the configuration, on each
>     packet
>     > +     received the device MUST calculate the hashing for the packet and
>     > +     store it in the virtio-net header in the \field{hash} field and the
>     > +     hash fields used in the calculation in rss_hash_type.
>     > +\end{enumerate}
>     > +
>     > +In case of any unexpected values/ unsupported hash function the driver
>     > +MUST return VIRTIO_NET_ERR in the \field{ack} field.
>
>     driver or the device? 
>
>
>     All MUST statements belon in correct normative sections.
>
> True, this is obviously a mistake, it should be the device instead of the
> driver. 
>
>
>     Generally what is the text trying to say here?
>
> That the device should handle errors and inform the drivers if there are any
> errors during the execution of the commands. 
>
>   
>     > +
>     > +\drivernormative{\subparagraph}{Receive Side Scaling}{Device Types /
>     Network Device / Device Operation / Control Virtqueue / Steering mode /
>     Receive Side Scaling}
>     > +
>     > +If the driver wants to set RSS hash it should fill the RSS structure
>     fields
>     > +as follows:
>     > +
>     > +\begin{itemize}
>     > +\item The driver SHOULD choose the hash function that SHOULD be used and
>     fill
>     > +     it in the \field{hash_function} field along with the appropriate
>     flags
>     > +     in the \field{hash_function_flags} field. These flags indicate to
>     the
>     > +     device which packet fields MUST be used in the calculation process
>     of
>     > +     the hash.
>     > +\item Once the hash function has been chosen a suitable hash key should
>     be set
>     > +     in the \field{hash_key} field,
>
>     which key is suitable?
>
> depends on the configuration, in Windows case the OS provide us with the key,
> however, in general it could be anything. 
>

We need to somehow describe it in the spec :)

>     > the length of the key should be stored
>     > +     in the \field{hash_key_length} field.
>
>     in 4 byte units?
>     are there limits on the length?
>
> I think one byte can be used as well as now it is 40 bytes in Windows. What is
> the best size that should be chosen in order to keep it future proof?

As a 1st step, could you add spec text explaining what it is?


>     > +\item Lastly the driver should fill the indirection table array
>
>     is an indirection table required? why not have a function produce
>     the queue number?
>
> Yes it is required as in Windows it is given by the OS. the indirection table
> is basically a map between the hash and cpu ids which correspond to queue ids. 
>
>   
>     > in the
>     > +     \field{indirection_table} field while setting the array length in
>     > +     \field{indirection_table_length}.
>
>     any limits on the length here?
>
> currently the indirection table is 128 bytes in Windows. We can change the
> sizes to fit that case, but what about future compatibility? What do you think
> should be the optimal size of the fields in order to keep this future proof
> while keeping the overhead minimal.

As host must maintain this in memory, you can have host specify the
maximum. guest will bail out if host does not support what it needs.


>   
>     > This structure is used by the device
>     > +     for determining in which RX virt queue the packet should be placed.
>
>     How?
>
> By applying the hash function on the packet while using the provided key in
> order to determine the hash for the packet and then use the indirection table
> to place the packet inside the suitable queue. This is hash function dependent
> implementation, but more elaboration can be added indeed. 

I guess indirection table is independent?

>   
>     > +\end{itemize}
>     > +Once the configuration phase is over successfully, the packets SHOULD
>     have the
>     > +\field{hash} field with the hash value that was calculated by the
>     device.
>
>     Why not make this optional? I suspect it's not really useful e.g. for
>     dpdk or xdp which do not generally use hardware offloads.
>
> Good point, will make it optional. 
>
>   
>     > +
>     > +Whenever the driver wants to discard the current RSS settings, it can
>     send an
>     > +VIRTIO_NET_SM_CTRL_RSS_SET command along with rss structure that has
>     > +RSS_HASH_FUNCTION_NONE the \field{hash_function} field.
>
>     And then what happens?
>
> and then the device is configured with RSS settings


so NONE will enable RSS?

> and MUST start distributing
> the packets into the queues accordingly while attaching the calculated hash to
> the header.
>
>   
>     > +
>     > +The driver MUST check the \field{ack} field value provided by the
>     device, in
>     > +case the value is not VIRTIO_NET_OK then the driver MUST handle failure
>     and
>     > +retry another hash function or else give up.
>
>     Is there anything special here?
>
> Not really just describing the error handling in detail. 

If it applies to all commands maybe add it to generic code.

>   
>     > +
>     >  \subparagraph{Legacy Interface: Automatic receive steering in multiqueue
>     mode}\label{sec:Device Types / Network Device / Device Operation / Control
>     Virtqueue / Automatic receive steering in multiqueue mode / Legacy
>     Interface: Automatic receive steering in multiqueue mode}
>     >  When using the legacy interface, transitional devices and drivers
>     >  MUST format \field{virtqueue_pairs}
>
>     conformance statements will need to be updated.
>
> which statements for example?

I mean links are needed in conformance.tex

Do you think that the Steering Mode should be in conformance?  

>   
>     > --
>     > 2.13.6
>     >
>     >
>     > ---------------------------------------------------------------------
>     > To unsubscribe, e-mail: virtio-dev-unsubscribe@lists.oasis-open.org
>     > For additional commands, e-mail: virtio-dev-help@lists.oasis-open.org
>
>
>
>
> --
> Respectfully,
> Sameeh Jubran
> Linkedin
> Software Engineer @ Daynix.

--

Respectfully,

Sameeh Jubran

Linkedin

Software Engineer @ Daynix.