DPDK patches and discussions
 help / color / mirror / Atom feed
From: Nikos Dragazis <ndragazis@arrikto.com>
To: Ilya Maximets <i.maximets@samsung.com>
Cc: dev@dpdk.org, Maxime Coquelin <maxime.coquelin@redhat.com>,
	Tiwei Bie <tiwei.bie@intel.com>,
	Zhihong Wang <zhihong.wang@intel.com>,
	Stefan Hajnoczi <stefanha@redhat.com>,
	Wei Wang <wei.w.wang@intel.com>,
	Stojaczyk Dariusz <dariusz.stojaczyk@intel.com>,
	Vangelis Koukis <vkoukis@arrikto.com>
Subject: Re: [dpdk-dev] [PATCH 00/28] vhost: add virtio-vhost-user transport
Date: Fri, 21 Jun 2019 02:44:30 +0300	[thread overview]
Message-ID: <8e969352-27d2-1a61-d545-0acf37c65af4@arrikto.com> (raw)
In-Reply-To: <e84d27e6-d494-547d-2151-5bc9be96626c@samsung.com>

On 20/6/19 2:32 μ.μ., Ilya Maximets wrote:
> On 19.06.2019 18:14, Nikos Dragazis wrote:
>> Hi everyone,
> Hi. I didn't look at the code, just a few comments inline.
>
>> this patch series introduces the concept of the virtio-vhost-user
>> transport. This is actually a revised version of an earlier RFC
>> implementation that has been proposed by Stefan Hajnoczi [1]. Though
>> this is a great feature, it seems to have been stalled, so I’d like to
>> restart the conversation on this and hopefully get it merged with your
>> help. Let me give you an overview.
>>
>> The virtio-vhost-user transport is a vhost-user transport implementation
>> that is based on the virtio-vhost-user device. Its key difference with
>> the existing transport is that it allows deploying vhost-user targets
>> inside dedicated Storage Appliance VMs instead of host user space. In
>> other words, it allows having guests that act as vhost-user backends for
>> other guests.
>>
>> The virtio-vhost-user device implements the vhost-user control plane
>> (master-slave communication) as follows:
>>
>> 1. it parses the vhost-user messages from the vhost-user unix domain
>>    socket and forwards them to the slave guest through virtqueues
>>
>> 2. it maps the vhost memory regions in QEMU’s process address space and
>>    exposes them to the slave guest as a RAM-backed PCI MMIO region
>>
>> 3. it hooks up doorbells to the callfds. The slave guest can use these
>>    doorbells to interrupt the master guest driver
>>
>> The device code has not yet been merged into upstream QEMU, but this is
>> definitely the end goal. The current state is that we are awaiting for
>> the approval of the virtio spec.
>>
>> I have Cced Darek from the SPDK community who has helped me a lot by
>> reviewing this series. Note that any device type could be implemented
>> over this new transport. So, adding the virtio-vhost-user transport in
>> DPDK would allow using it from SPDK as well.
>>
>> Getting into the code internals, this patch series makes the following
>> changes:
>>
>> 1. introduce a generic interface for the transport-specific operations.
>>    Each of the two available transports, the pre-existing AF_UNIX
>>    transport and the virtio-vhost-user transport, is going to implement
>>    this interface. The AF_UNIX-specific code has been extracted from the
>>    core vhost-user code and is now part of the AF_UNIX transport
>>    implementation in trans_af_unix.c.
>>
>> 2. introduce the virtio-vhost-user transport. The virtio-vhost-user
>>    transport requires a driver for the virtio-vhost-user devices. The
>>    driver along with the transport implementation have been packed into
>>    a separate library in `drivers/virtio_vhost_user/`. The necessary
>>    virtio-pci code has been copied from `drivers/net/virtio/`. Some
>>    additional changes have been made so that the driver can utilize the
>>    additional resources of the virtio-vhost-user device.
>>
>> 3. update librte_vhost public API to enable choosing transport for each
>>    new vhost device. Extend the vhost net driver and vhost-scsi example
>>    application to export this new API to the end user.
>>
>> The primary changes I did to Stefan’s RFC implementation are the
>> following:
>>
>> 1. moved postcopy live migration code into trans_af_unix.c. Postcopy
>>    live migration relies on the userfault fd mechanism, which cannot be
>>    supported by virtio-vhost-user.
>>
>> 2. moved setup of the log memory region into trans_af_unix.c. Setting up
>>    the log memory region involves mapping/unmapping guest memory. This
>>    is an AF_UNIX transport-specific operation.
> Logging dirty pages is the main concept of live migration support. Does it
> mean that the live migration is not supported for virtio-vhost-user at all?

No, it is supported. To be more precise, it can be supported, it is part
of the on-going virtio device specification:

https://lists.oasis-open.org/archives/virtio-dev/201905/msg00022.html

and it is in my TODO list for the device code. Here is how it works:

the log memory region, just like the other vhost memory regions, is a
portion of the master guest memory. In case of the AF_UNIX transport,
the master sends a VHOST_USER_SET_LOG_BASE message and the vhost target
mmaps the log memory region into the process' virtual address space. In
case of the virtio-vhost-user transport, the virtio-vhost-user device
parses the VHOST_USER_SET_LOG_BASE message and maps the log memory
region into QEMU's process address space. It then exposes the log memory
region to the slave guest as a RAM-backed PCI MMIO region.

So, from the vhost target's viewpoint, the only difference is the means
of accessing the log memory region. In case of the AF_UNIX transport,
the vhost target receives a file descriptor and mmaps this file. In case
of the virtio-vhost-user transport, the device exports the log memory
region to the vhost target (running in slave guest user space) via an
MMIO BAR.

To recap, just to make sure it is clear, the virtio-vhost-user transport
does support precopy live migration, but it doesn't support postcopy
live migration.

>
>> 3. introduced a vhost transport operation for
>>    process_slave_message_reply()
>>
>> 4. moved the virtio-vhost-user transport/driver into a separate library
>>    in `drivers/virtio_vhost_user/`. This required making vhost.h and
>>    vhost_user.h part of librte_vhost public API and exporting some
>>    private symbols via the version script. This looks better to me that
>>    just moving the entire librte_vhost into `drivers/`. I am not sure if
>>    this is the most appropriate solution. I am looking forward to your
>>    suggestions on this.
> Moving the virtio-vhost-user code to a separate driver looks strange for me.
> What is the purpose?

The virtio-vhost-user transport is based on the virtio-vhost-user
device. This means that we need a user space driver for this device
type.  Currently, the driver and the transport implementation are packed
together into `trans_virtio_vhost_user.c`. There are 2 reasons why I
moved all the virtio-vhost-user code (driver + transport implementation)
into `drivers/`:

1. my understanding is that all drivers should be located in `drivers/`.

2. |the driver depends on `rte_bus_pci.h` which gets exported when
   `drivers/` is built whereas librte_vhost is built before `drivers/`.
   However, I think this could be overcomed by just adding this line:

   CFLAGS += -I$(RTE_SDK)/drivers/bus/pci

   in librte_vhost Makefile.||


|
>
> Exporting a lot of vhost internal structures will lead to a frequent API/ABI
> breakages and will slow down accepting changes to releases in general.

I understand that there are many parameters that we have to consider. My
implementation is just a proposal.  I am looking forward to hearing from
you and the community any other suggestions on this.

>
> It looks inconsistent to have 'trans_af_unix.c' in 'lib/librte_vhost/' and
> 'trans_virtio_vhost_user.c' in 'drivers/virtio_vhost_user/' because these
> files should be similar in provided functionality, hence, should be located
> in similar places.

I agree. Actually, I was thinking about separating the driver from the
transport implementation and moving just the driver into `drivers/`. But
this would require exporting some low-level virtio PCI functions. And
this didn't look appropriate to me.

>
>> 5. made use of the virtio PCI capabilities for the additional device
>>    resources (doorbells, shared memory). This required changes in
>>    virtio_pci.c and trans_virtio_vhost_user.c.
>>
>> 6. [minor] changed some commit headlines to comply with
>>    check-git-log.sh.
>>
>> Please, have a look and let me know about your thoughts. Any
>> reviews/pointers/suggestions are welcome.
>>
>> Best regards,
>> Nikos
>>
>> [1] http://mails.dpdk.org/archives/dev/2018-January/088155.html
>>
>>
>> Nikos Dragazis (23):
>>   vhost: introduce vhost transport operations structure
>>   vhost: move socket management code
>>   vhost: move socket fd and un sockaddr
>>   vhost: move vhost-user connection
>>   vhost: move vhost-user reconnection
>>   vhost: move vhost-user fdset
>>   vhost: propagate vhost transport operations
>>   vhost: use a single structure for the device state
>>   vhost: extract socket I/O into transport
>>   vhost: move slave request fd and lock
>>   vhost: move mmap/munmap
>>   vhost: move setup of the log memory region
>>   vhost: remove main fd parameter from msg handlers
>>   vhost: move postcopy live migration code
>>   vhost: support registering additional vhost-user transports
>>   drivers/virtio_vhost_user: add virtio PCI framework
>>   drivers: add virtio-vhost-user transport
>>   drivers/virtio_vhost_user: use additional device resources
>>   vhost: add flag for choosing vhost-user transport
>>   net/vhost: add virtio-vhost-user support
>>   mk: link apps with virtio-vhost-user driver
>>   config: add option for the virtio-vhost-user transport
>>   usertools: add virtio-vhost-user devices to dpdk-devbind.py
>>
>> Stefan Hajnoczi (5):
>>   vhost: allocate per-socket transport state
>>   vhost: move start server/client calls
>>   vhost: add index field in vhost virtqueues
>>   examples/vhost_scsi: add --socket-file argument
>>   examples/vhost_scsi: add virtio-vhost-user support
>>
>>  config/common_base                                 |    6 +
>>  config/common_linux                                |    1 +
>>  drivers/Makefile                                   |    5 +
>>  drivers/net/vhost/rte_eth_vhost.c                  |   13 +
>>  drivers/virtio_vhost_user/Makefile                 |   27 +
>>  .../rte_virtio_vhost_user_version.map              |    4 +
>>  .../virtio_vhost_user/trans_virtio_vhost_user.c    | 1077 +++++++++++++++++++
>>  drivers/virtio_vhost_user/virtio_pci.c             |  520 ++++++++++
>>  drivers/virtio_vhost_user/virtio_pci.h             |  289 ++++++
>>  drivers/virtio_vhost_user/virtio_vhost_user.h      |   18 +
>>  drivers/virtio_vhost_user/virtqueue.h              |  181 ++++
>>  examples/vhost_scsi/vhost_scsi.c                   |  103 +-
>>  lib/librte_vhost/Makefile                          |    4 +-
>>  lib/librte_vhost/rte_vhost.h                       |    1 +
>>  lib/librte_vhost/rte_vhost_version.map             |   11 +
>>  lib/librte_vhost/socket.c                          |  685 +-----------
>>  lib/librte_vhost/trans_af_unix.c                   | 1094 ++++++++++++++++++++
>>  lib/librte_vhost/vhost.c                           |   22 +-
>>  lib/librte_vhost/vhost.h                           |  298 +++++-
>>  lib/librte_vhost/vhost_user.c                      |  474 ++-------
>>  lib/librte_vhost/vhost_user.h                      |   10 +-
>>  mk/rte.app.mk                                      |    6 +
>>  usertools/dpdk-devbind.py                          |    7 +
>>  23 files changed, 3764 insertions(+), 1092 deletions(-)
>>  create mode 100644 drivers/virtio_vhost_user/Makefile
>>  create mode 100644 drivers/virtio_vhost_user/rte_virtio_vhost_user_version.map
>>  create mode 100644 drivers/virtio_vhost_user/trans_virtio_vhost_user.c
>>  create mode 100644 drivers/virtio_vhost_user/virtio_pci.c
>>  create mode 100644 drivers/virtio_vhost_user/virtio_pci.h
>>  create mode 100644 drivers/virtio_vhost_user/virtio_vhost_user.h
>>  create mode 100644 drivers/virtio_vhost_user/virtqueue.h
>>  create mode 100644 lib/librte_vhost/trans_af_unix.c
>>


  reply	other threads:[~2019-06-20 23:44 UTC|newest]

Thread overview: 40+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-06-19 15:14 Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 01/28] vhost: introduce vhost transport operations structure Nikos Dragazis
2019-06-19 20:14   ` Aaron Conole
2019-06-20 10:30     ` Bruce Richardson
2019-06-20 18:24       ` Nikos Dragazis
2019-06-20 18:19     ` Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 02/28] vhost: move socket management code Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 03/28] vhost: allocate per-socket transport state Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 04/28] vhost: move socket fd and un sockaddr Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 05/28] vhost: move start server/client calls Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 06/28] vhost: move vhost-user connection Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 07/28] vhost: move vhost-user reconnection Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 08/28] vhost: move vhost-user fdset Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 09/28] vhost: propagate vhost transport operations Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 10/28] vhost: use a single structure for the device state Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 11/28] vhost: extract socket I/O into transport Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 12/28] vhost: move slave request fd and lock Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 13/28] vhost: move mmap/munmap Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 14/28] vhost: move setup of the log memory region Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 15/28] vhost: remove main fd parameter from msg handlers Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 16/28] vhost: move postcopy live migration code Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 17/28] vhost: support registering additional vhost-user transports Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 18/28] drivers/virtio_vhost_user: add virtio PCI framework Nikos Dragazis
2019-09-05 16:34   ` Maxime Coquelin
2019-09-09  8:42     ` Nikos Dragazis
2019-09-09  8:44       ` Maxime Coquelin
2019-06-19 15:14 ` [dpdk-dev] [PATCH 19/28] vhost: add index field in vhost virtqueues Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 20/28] drivers: add virtio-vhost-user transport Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 21/28] drivers/virtio_vhost_user: use additional device resources Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 22/28] vhost: add flag for choosing vhost-user transport Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 23/28] net/vhost: add virtio-vhost-user support Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 24/28] examples/vhost_scsi: add --socket-file argument Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 25/28] examples/vhost_scsi: add virtio-vhost-user support Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 26/28] mk: link apps with virtio-vhost-user driver Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 27/28] config: add option for the virtio-vhost-user transport Nikos Dragazis
2019-06-19 15:14 ` [dpdk-dev] [PATCH 28/28] usertools: add virtio-vhost-user devices to dpdk-devbind.py Nikos Dragazis
     [not found] ` <CGME20190620113240eucas1p22ca4faa64a36bbb7aec38a81298ade56@eucas1p2.samsung.com>
2019-06-20 11:32   ` [dpdk-dev] [PATCH 00/28] vhost: add virtio-vhost-user transport Ilya Maximets
2019-06-20 23:44     ` Nikos Dragazis [this message]
2019-06-20 11:35 ` Maxime Coquelin
2019-06-22 20:26   ` Nikos Dragazis

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=8e969352-27d2-1a61-d545-0acf37c65af4@arrikto.com \
    --to=ndragazis@arrikto.com \
    --cc=dariusz.stojaczyk@intel.com \
    --cc=dev@dpdk.org \
    --cc=i.maximets@samsung.com \
    --cc=maxime.coquelin@redhat.com \
    --cc=stefanha@redhat.com \
    --cc=tiwei.bie@intel.com \
    --cc=vkoukis@arrikto.com \
    --cc=wei.w.wang@intel.com \
    --cc=zhihong.wang@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).