[PATCH v4 0/4] virtio-blk: add iothread-vq-mapping parameter

Stefan Hajnoczi posted 4 patches 4 months, 1 week ago
Failed in applying to current master (apply log)
qapi/virtio.json                     |  29 +++++
hw/block/dataplane/virtio-blk.h      |   3 +
include/hw/qdev-properties-system.h  |   5 +
include/hw/virtio/virtio-blk.h       |   2 +
include/qapi/string-output-visitor.h |   6 +-
hw/block/dataplane/virtio-blk.c      | 155 ++++++++++++++++++++-------
hw/block/virtio-blk.c                |  92 +++++++++++++---
hw/core/qdev-properties-system.c     |  46 ++++++++
hw/core/qdev-properties.c            |  18 ++--
qapi/string-output-visitor.c         |  16 +++
10 files changed, 311 insertions(+), 61 deletions(-)
[PATCH v4 0/4] virtio-blk: add iothread-vq-mapping parameter
Posted by Stefan Hajnoczi 4 months, 1 week ago
v4:
- Use DummyVirtioForceArrays naming in QAPI schema [Markus]
v3:
- Rebased onto Kevin's block branch
- Add StringOutputVisitor "<omitted>" patch to fix "info qtree" crash
- Fix QAPI schema formatting [Markus]
- Eliminate unnecessary local variable in get_iothread_vq_mapping_list() [Markus]

virtio-blk and virtio-scsi devices need a way to specify the mapping between
IOThreads and virtqueues. At the moment all virtqueues are assigned to a single
IOThread or the main loop. This single thread can be a CPU bottleneck, so it is
necessary to allow finer-grained assignment to spread the load. With this
series applied, "pidstat -t 1" shows that guests with -smp 2 or higher are able
to exploit multiple IOThreads.

This series introduces command-line syntax for the new iothread-vq-mapping
property is as follows:

  --device '{"driver":"virtio-blk-pci","iothread-vq-mapping":[{"iothread":"iothread0","vqs":[0,1,2]},...]},...'

IOThreads are specified by name and virtqueues are specified by 0-based
index.

It will be common to simply assign virtqueues round-robin across a set
of IOThreads. A convenient syntax that does not require specifying
individual virtqueue indices is available:

  --device '{"driver":"virtio-blk-pci","iothread-vq-mapping":[{"iothread":"iothread0"},{"iothread":"iothread1"},...]},...'

There is no way to reassign virtqueues at runtime and I expect that to be a
very rare requirement.

Note that JSON --device syntax is required for the iothread-vq-mapping
parameter because it's non-scalar.

Based-on: 81e69329d6a4018f4b37d15b6fc845fbe585a93b (https://repo.or.cz/qemu/kevin.git block)

Stefan Hajnoczi (4):
  qdev-properties: alias all object class properties
  string-output-visitor: show structs as "<omitted>"
  qdev: add IOThreadVirtQueueMappingList property type
  virtio-blk: add iothread-vq-mapping parameter

 qapi/virtio.json                     |  29 +++++
 hw/block/dataplane/virtio-blk.h      |   3 +
 include/hw/qdev-properties-system.h  |   5 +
 include/hw/virtio/virtio-blk.h       |   2 +
 include/qapi/string-output-visitor.h |   6 +-
 hw/block/dataplane/virtio-blk.c      | 155 ++++++++++++++++++++-------
 hw/block/virtio-blk.c                |  92 +++++++++++++---
 hw/core/qdev-properties-system.c     |  46 ++++++++
 hw/core/qdev-properties.c            |  18 ++--
 qapi/string-output-visitor.c         |  16 +++
 10 files changed, 311 insertions(+), 61 deletions(-)

-- 
2.43.0
Re: [PATCH v4 0/4] virtio-blk: add iothread-vq-mapping parameter
Posted by Kevin Wolf 4 months, 1 week ago
Am 20.12.2023 um 14:47 hat Stefan Hajnoczi geschrieben:
> v4:
> - Use DummyVirtioForceArrays naming in QAPI schema [Markus]
> v3:
> - Rebased onto Kevin's block branch
> - Add StringOutputVisitor "<omitted>" patch to fix "info qtree" crash
> - Fix QAPI schema formatting [Markus]
> - Eliminate unnecessary local variable in get_iothread_vq_mapping_list() [Markus]
> 
> virtio-blk and virtio-scsi devices need a way to specify the mapping between
> IOThreads and virtqueues. At the moment all virtqueues are assigned to a single
> IOThread or the main loop. This single thread can be a CPU bottleneck, so it is
> necessary to allow finer-grained assignment to spread the load. With this
> series applied, "pidstat -t 1" shows that guests with -smp 2 or higher are able
> to exploit multiple IOThreads.
> 
> This series introduces command-line syntax for the new iothread-vq-mapping
> property is as follows:
> 
>   --device '{"driver":"virtio-blk-pci","iothread-vq-mapping":[{"iothread":"iothread0","vqs":[0,1,2]},...]},...'
> 
> IOThreads are specified by name and virtqueues are specified by 0-based
> index.
> 
> It will be common to simply assign virtqueues round-robin across a set
> of IOThreads. A convenient syntax that does not require specifying
> individual virtqueue indices is available:
> 
>   --device '{"driver":"virtio-blk-pci","iothread-vq-mapping":[{"iothread":"iothread0"},{"iothread":"iothread1"},...]},...'
> 
> There is no way to reassign virtqueues at runtime and I expect that to be a
> very rare requirement.
> 
> Note that JSON --device syntax is required for the iothread-vq-mapping
> parameter because it's non-scalar.
> 
> Based-on: 81e69329d6a4018f4b37d15b6fc845fbe585a93b (https://repo.or.cz/qemu/kevin.git block)

Thanks, applied to the block branch. We agreed off-list that the
remaining problems can be fixed in follow-up patches.

Kevin