[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250122163154.1225353-1-eperezma@redhat.com>
Date: Wed, 22 Jan 2025 17:31:49 +0100
From: Eugenio Pérez <eperezma@...hat.com>
To: linux-kernel@...r.kernel.org,
mst@...hat.com,
Hanna Reitz <hreitz@...hat.com>,
Stefano Garzarella <sgarzare@...hat.com>,
Xuan Zhuo <xuanzhuo@...ux.alibaba.com>,
Jason Wang <jasowang@...hat.com>,
Vivek Goyal <vgoyal@...hat.com>,
Miklos Szeredi <miklos@...redi.hu>,
German Maglione <gmaglione@...hat.com>,
Stefan Hajnoczi <stefanha@...hat.com>,
virtualization@...ts.linux.dev
Subject: [RFC 0/5] virtiofs: map buffer out of virtqueue lock
This is useful for some setups like swiotlb or VDUSE where the DMA
operations are expensive and/or need to be performed with a write lock.
After applying this patch, fio read test goes from 1201MiB/s to
1211MiB/s.
---
Sending this series to obtain feedback if this is the right way to go.
Q: Do we need virtqueue_map_sgs? XDP directly call dma_map. Need to
profile.
TODO: Profile multiqueue.
TODO: Handling errors.
TODO: Do the same for hiprio queue.
TODO: Can we batch maps? virtiofs always sends many buffers.
Eugenio Pérez (4):
vduse: add virtio_fs to allowed dev id
virtiofs: Move stack sg to fuse_req
virtio_ring: add virtqueue premapped out/in buffers support
virtiofs: perform DMA operations out of the spinlock
Jason Wang (1):
virtio_ring: introduce virtqueue_map/unmap_sgs()
drivers/vdpa/vdpa_user/vduse_dev.c | 1 +
drivers/virtio/virtio_ring.c | 165 ++++++++++++++++++++++++++---
fs/fuse/fuse_i.h | 7 ++
fs/fuse/virtio_fs.c | 68 +++++++-----
include/linux/virtio.h | 17 +++
5 files changed, 220 insertions(+), 38 deletions(-)
--
2.48.1
Powered by blists - more mailing lists