From: "Michael S. Tsirkin" <mst@redhat.com>
To: Jason Wang <jasowang@redhat.com>
Cc: "Liu, Jing2" <jing2.liu@linux.intel.com>,
Zha Bin <zhabin@linux.alibaba.com>,
linux-kernel@vger.kernel.org, virtio-dev@lists.oasis-open.org,
slp@redhat.com, qemu-devel@nongnu.org,
chao.p.peng@linux.intel.com, gerry@linux.alibaba.com
Subject: Re: [virtio-dev] Re: [PATCH v2 4/5] virtio-mmio: add MSI interrupt feature support
Date: Wed, 12 Feb 2020 04:16:39 -0500 [thread overview]
Message-ID: <20200212041554-mutt-send-email-mst@kernel.org> (raw)
In-Reply-To: <44209f3c-613c-3766-ca83-321b77b0f0dd@redhat.com>
On Wed, Feb 12, 2020 at 05:06:52PM +0800, Jason Wang wrote:
>
> On 2020/2/12 上午11:54, Liu, Jing2 wrote:
> >
> >
> > On 2/11/2020 3:40 PM, Jason Wang wrote:
> > >
> > > On 2020/2/11 下午2:02, Liu, Jing2 wrote:
> > > >
> > > >
> > > > On 2/11/2020 12:02 PM, Jason Wang wrote:
> > > > >
> > > > > On 2020/2/11 上午11:35, Liu, Jing2 wrote:
> > > > > >
> > > > > > On 2/11/2020 11:17 AM, Jason Wang wrote:
> > > > > > >
> > > > > > > On 2020/2/10 下午5:05, Zha Bin wrote:
> > > > > > > > From: Liu Jiang<gerry@linux.alibaba.com>
> > > > > > > >
> > > > > > > > Userspace VMMs (e.g. Qemu microvm, Firecracker)
> > > > > > > > take advantage of using
> > > > > > > > virtio over mmio devices as a lightweight machine model for modern
> > > > > > > > cloud. The standard virtio over MMIO transport
> > > > > > > > layer only supports one
> > > > > > > > legacy interrupt, which is much heavier than
> > > > > > > > virtio over PCI transport
> > > > > > > > layer using MSI. Legacy interrupt has long work
> > > > > > > > path and causes specific
> > > > > > > > VMExits in following cases, which would considerably slow down the
> > > > > > > > performance:
> > > > > > > >
> > > > > > > > 1) read interrupt status register
> > > > > > > > 2) update interrupt status register
> > > > > > > > 3) write IOAPIC EOI register
> > > > > > > >
> > > > > > > > We proposed to add MSI support for virtio over MMIO via new feature
> > > > > > > > bit VIRTIO_F_MMIO_MSI[1] which increases the interrupt performance.
> > > > > > > >
> > > > > > > > With the VIRTIO_F_MMIO_MSI feature bit
> > > > > > > > supported, the virtio-mmio MSI
> > > > > > > > uses msi_sharing[1] to indicate the event and vector mapping.
> > > > > > > > Bit 1 is 0: device uses non-sharing and fixed
> > > > > > > > vector per event mapping.
> > > > > > > > Bit 1 is 1: device uses sharing mode and dynamic mapping.
> > > > > > >
> > > > > > >
> > > > > > > I believe dynamic mapping should cover the case of fixed vector?
> > > > > > >
> > > > > > Actually this bit *aims* for msi sharing or msi non-sharing.
> > > > > >
> > > > > > It means, when msi sharing bit is 1, device doesn't want
> > > > > > vector per queue
> > > > > >
> > > > > > (it wants msi vector sharing as name) and doesn't want a
> > > > > > high interrupt rate.
> > > > > >
> > > > > > So driver turns to !per_vq_vectors and has to do dynamical mapping.
> > > > > >
> > > > > > So they are opposite not superset.
> > > > > >
> > > > > > Thanks!
> > > > > >
> > > > > > Jing
> > > > >
> > > > >
> > > > > I think you need add more comments on the command.
> > > > >
> > > > > E.g if I want to map vector 0 to queue 1, how do I need to do?
> > > > >
> > > > > write(1, queue_sel);
> > > > > write(0, vector_sel);
> > > >
> > > > That's true. Besides, two commands are used for msi sharing mode,
> > > >
> > > > VIRTIO_MMIO_MSI_CMD_MAP_CONFIG and VIRTIO_MMIO_MSI_CMD_MAP_QUEUE.
> > > >
> > > > "To set up the event and vector mapping for MSI sharing mode,
> > > > driver SHOULD write a valid MsiVecSel followed by
> > > > VIRTIO_MMIO_MSI_CMD_MAP_CONFIG/VIRTIO_MMIO_MSI_CMD_MAP_QUEUE
> > > > command to map the configuration change/selected queue events
> > > > respectively. " (See spec patch 5/5)
> > > >
> > > > So if driver detects the msi sharing mode, when it does setup
> > > > vq, writes the queue_sel (this already exists in setup vq),
> > > > vector sel and then MAP_QUEUE command to do the queue event
> > > > mapping.
> > > >
> > >
> > > So actually the per vq msix could be done through this.
> >
> > Right, per vq msix can also be mapped by the 2 commands if we want.
> >
> > The current design benefits for those devices requesting per vq msi that
> > driver
> >
> > doesn't have to map during setup each queue,
> >
> > since we define the relationship by default.
> >
>
> Well since you've defined the dynamic mapping, having some "default" mapping
> won't help to reduce the complexity but increase it.
>
>
> >
> > > I don't get why you need to introduce MSI_SHARING_MASK which is the
> > > charge of driver instead of device.
> >
> > MSI_SHARING_MASK is just for identifying the msi_sharing bit in
> > readl(MsiState) (0x0c4). The device tells whether it wants msi_sharing.
> >
> > MsiState register: R
> >
> > le32 {
> > msi_enabled : 1;
> > msi_sharing: 1;
> > reserved : 30;
> > };
> >
>
> The question is why device want such information.
>
>
> >
> > > The interrupt rate should have no direct relationship with whether
> > > it has been shared or not.
> >
> > >
> > > Btw, you introduce mask/unmask without pending, how to deal with the
> > > lost interrupt during the masking then?
> > >
> > >
> > > > For msi non-sharing mode, no special action is needed because we
> > > > make the rule of per_vq_vector and fixed relationship.
> > > >
> > > > Correct me if this is not that clear for spec/code comments.
> > > >
> > >
> > > The ABI is not as straightforward as PCI did. Why not just reuse the
> > > PCI layout?
> > >
> > > E.g having
> > >
> > > queue_sel
> > > queue_msix_vector
> > > msix_config
> > >
> > > for configuring map between msi vector and queues/config
> >
> > Thanks for the advice. :)
> >
> > Actually when looking into pci, the queue_msix_vector/msix_config is the
> > msi vector index, which is the same as the mmio register MsiVecSel
> > (0x0d0).
> >
> > So we don't introduce two extra registers for mapping even in sharing
> > mode.
> >
> > What do you think?
> >
>
> I'm not sure I get the point, but I still prefer the separate vector_sel
> from queue_msix_vector.
>
> Btw, Michael propose per vq registers which could also work.
>
> Thanks
>
Right and I'd even ask a question: do we need shared MSI at all?
Is it somehow better than legacy interrupt? And why?
Performance numbers please.
> >
> > >
> > > Then
> > >
> > > vector_sel
> > > address
> > > data
> > > pending
> > > mask
> > > unmask
> > >
> > > for configuring msi table?
> >
> > PCI-like msix table is not introduced to device and instead simply use
> > commands to tell the mask/configure/enable.
> >
> > Thanks!
> >
> > Jing
> >
> > >
> > > Thanks
> > >
> > >
> > > > Thanks!
> > > >
> > > > Jing
> > > >
> > > >
> > > > >
> > > > > ?
> > > > >
> > > > > Thanks
> > > > >
> > > > >
> > > > > >
> > > > > >
> > > > > > > Thanks
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > ---------------------------------------------------------------------
> > > > > > >
> > > > > > > To unsubscribe, e-mail: virtio-dev-unsubscribe@lists.oasis-open.org
> > > > > > > For additional commands, e-mail:
> > > > > > > virtio-dev-help@lists.oasis-open.org
> > > > > > >
> > > > > >
> > > > >
> > >
next prev parent reply other threads:[~2020-02-12 9:16 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-02-10 9:05 [PATCH v2 0/5] virtio mmio specification enhancement Zha Bin
2020-02-10 9:05 ` [PATCH v2 1/5] virtio-mmio: add notify feature for per-queue Zha Bin
2020-02-11 10:50 ` Michael S. Tsirkin
2020-02-11 11:33 ` Michael S. Tsirkin
2020-02-12 3:39 ` Jason Wang
2020-02-12 8:18 ` Michael S. Tsirkin
2020-02-12 8:53 ` Jason Wang
2020-02-12 9:33 ` Jason Wang
2020-02-12 9:55 ` Michael S. Tsirkin
2020-02-13 3:38 ` Jason Wang
2020-02-10 9:05 ` [PATCH v2 2/5] virtio-mmio: refactor common functionality Zha Bin
2020-02-11 11:19 ` Michael S. Tsirkin
2020-02-12 2:58 ` [virtio-dev] " Liu, Jing2
2020-02-12 7:29 ` Michael S. Tsirkin
2020-02-10 9:05 ` [PATCH v2 3/5] virtio-mmio: create a generic MSI irq domain Zha Bin
2020-02-11 11:16 ` Michael S. Tsirkin
2020-02-12 7:40 ` Michael S. Tsirkin
2020-02-10 9:05 ` [PATCH v2 4/5] virtio-mmio: add MSI interrupt feature support Zha Bin
2020-02-11 3:17 ` Jason Wang
2020-02-11 3:35 ` [virtio-dev] " Liu, Jing2
2020-02-11 4:02 ` Jason Wang
[not found] ` <5522f205-207b-b012-6631-3cc77dde3bfe@linux.intel.com>
2020-02-11 7:40 ` Jason Wang
2020-02-11 11:58 ` Michael S. Tsirkin
2020-02-11 12:04 ` Jason Wang
2020-02-11 12:08 ` Michael S. Tsirkin
2020-02-11 12:18 ` Jason Wang
2020-02-11 14:00 ` Michael S. Tsirkin
2020-02-12 9:03 ` Jason Wang
2020-02-12 9:15 ` Michael S. Tsirkin
[not found] ` <4c19292f-9d25-a859-3dde-6dd5a03fdf0b@linux.intel.com>
2020-02-12 7:33 ` Michael S. Tsirkin
2020-02-12 9:06 ` Jason Wang
2020-02-12 9:16 ` Michael S. Tsirkin [this message]
2020-02-13 3:40 ` Jason Wang
2020-02-11 11:21 ` Michael S. Tsirkin
2020-02-11 11:11 ` Michael S. Tsirkin
2020-02-10 9:05 ` [PATCH v2 5/5] x86: virtio-mmio: support virtio-mmio with MSI for x86 Zha Bin
2020-02-11 11:14 ` Michael S. Tsirkin
2020-02-10 11:44 ` [PATCH v2 0/5] virtio mmio specification enhancement Michael S. Tsirkin
2020-02-11 16:05 ` Chao Peng
2020-02-11 10:57 ` Michael S. Tsirkin
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=20200212041554-mutt-send-email-mst@kernel.org \
--to=mst@redhat.com \
--cc=chao.p.peng@linux.intel.com \
--cc=gerry@linux.alibaba.com \
--cc=jasowang@redhat.com \
--cc=jing2.liu@linux.intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=qemu-devel@nongnu.org \
--cc=slp@redhat.com \
--cc=virtio-dev@lists.oasis-open.org \
--cc=zhabin@linux.alibaba.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).