[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [PATCH 00/30] virtiofs daemon (base)
From: |
Dr. David Alan Gilbert |
Subject: |
Re: [PATCH 00/30] virtiofs daemon (base) |
Date: |
Thu, 24 Oct 2019 17:19:21 +0100 |
User-agent: |
Mutt/1.12.1 (2019-06-15) |
* Michael S. Tsirkin (address@hidden) wrote:
> On Mon, Oct 21, 2019 at 11:58:02AM +0100, Dr. David Alan Gilbert (git) wrote:
> > From: "Dr. David Alan Gilbert" <address@hidden>
> >
> > Hi,
> > This is the 1st set for the virtiofsd - a daemon
> > that implements the user space side of virtiofs.
> >
> > The kernel and qemu device parts recently went in,
> > so the daemon is the only thing missing to have a working
> > set.
> >
> > This set is the absolute minimal base set of patches;
> > it's not yet safe to use (from security or correctness);
> >
> > I'll follow up with ~3 more series in the next few days
> > with:
> >
> > a) Security patches that add sandboxing and checking
> > compared with normal fuse - that makes it safe.
> > b) Performance improvements including threading
> > c) Other fixes, including correctness.
> >
> > but, this is a good start and gets things rolling.
> >
> > The set pulls in a big chunk of the upstream libfuse library
> > (unmodified so that it's easy to check it really is upstream),
> > chops all the stuff out we don't need and then adds the
> > new transport we need.
> >
> > For new files I've formatted the code according to qemu
> > standards; for files that are from upstream libfuse
> > I've kept with their standards for ease of future updating.
>
> Thinking of future updates, have you given any thought to
> automating them? Something along the lines of
> update-linux-headers maybe?
The problem is we're changing the code in some of them;
libfuse changes slowly enough that rebasing on a new libfuse isn't too
bad; but once committed, just copying new versions in like update-linux-headers
does isn't going to work because we have to merge in our changes.
I expect to have to cherry-pick changes. Libfuse has had 75 patches in
the last year; so it doesn't worry me too much.
Dave
>
> > We can't just link with libfuse, since we have to make ABI incompatible
> > changes for the new transport.
> >
> > Running this daemon is typically done with:
> >
> > ./virtiofsd -o vhost_user_socket=/path/socket -o source=/path/to/fs
> >
> > connected to a qemu that's then started with:
> > -chardev socket,id=char0,path=/path/socket -device
> > vhost-user-fs-pci,queue-size=1024,chardev=char0,tag=myfs
> >
> > and then in the guest mount with:
> > mount -t virtiofs myfs /mnt
> >
> > Our development branch is:
> > https://gitlab.com/virtio-fs/qemu/tree/virtio-fs-dev
> >
> > Dave
> >
> >
> > Dr. David Alan Gilbert (22):
> > virtiofsd: Pull in upstream headers
> > virtiofsd: Pull in kernel's fuse.h
> > virtiofsd: Add auxiliary .c's
> > virtiofsd: Add fuse_lowlevel.c
> > virtiofsd: Add passthrough_ll
> > virtiofsd: Trim down imported files
> > virtiofsd: Fix fuse_daemonize ignored return values
> > virtiofsd: Fix common header and define for QEMU builds
> > virtiofsd: fuse: Make iov_length usable outside fuse_lowlevel.c
> > virtiofsd: Add options for virtio
> > virtiofsd: Open vhost connection instead of mounting
> > virtiofsd: Start wiring up vhost-user
> > virtiofsd: Add main virtio loop
> > virtiofsd: get/set features callbacks
> > virtiofsd: Start queue threads
> > virtiofsd: Poll kick_fd for queue
> > virtiofsd: Start reading commands from queue
> > virtiofsd: Send replies to messages
> > virtiofsd: Keep track of replies
> > virtiofsd: Add Makefile wiring for virtiofsd contrib
> > virtiofsd: Fast path for virtio read
> > virtiofs: Add maintainers entry
> >
> > Stefan Hajnoczi (7):
> > virtiofsd: remove mountpoint dummy argument
> > virtiofsd: remove unused notify reply support
> > virtiofsd: add -o source=PATH to help output
> > virtiofsd: add --fd=FDNUM fd passing option
> > virtiofsd: make -f (foreground) the default
> > virtiofsd: add vhost-user.json file
> > virtiofsd: add --print-capabilities option
> >
> > Vivek Goyal (1):
> > virtiofsd: Make fsync work even if only inode is passed in
> >
> > .gitignore | 1 +
> > MAINTAINERS | 8 +
> > Makefile | 9 +
> > Makefile.objs | 1 +
> > contrib/virtiofsd/50-qemu-virtiofsd.json.in | 5 +
> > contrib/virtiofsd/Makefile.objs | 10 +
> > contrib/virtiofsd/buffer.c | 318 +++
> > contrib/virtiofsd/fuse.h | 1268 ++++++++++
> > contrib/virtiofsd/fuse_common.h | 823 +++++++
> > contrib/virtiofsd/fuse_i.h | 131 ++
> > contrib/virtiofsd/fuse_kernel.h | 858 +++++++
> > contrib/virtiofsd/fuse_log.c | 40 +
> > contrib/virtiofsd/fuse_log.h | 82 +
> > contrib/virtiofsd/fuse_loop_mt.c | 54 +
> > contrib/virtiofsd/fuse_lowlevel.c | 2302 +++++++++++++++++++
> > contrib/virtiofsd/fuse_lowlevel.h | 2024 ++++++++++++++++
> > contrib/virtiofsd/fuse_misc.h | 59 +
> > contrib/virtiofsd/fuse_opt.c | 422 ++++
> > contrib/virtiofsd/fuse_opt.h | 271 +++
> > contrib/virtiofsd/fuse_signals.c | 90 +
> > contrib/virtiofsd/fuse_virtio.c | 717 ++++++
> > contrib/virtiofsd/fuse_virtio.h | 33 +
> > contrib/virtiofsd/helper.c | 300 +++
> > contrib/virtiofsd/passthrough_helpers.h | 76 +
> > contrib/virtiofsd/passthrough_ll.c | 1341 +++++++++++
> > docs/interop/vhost-user.json | 4 +-
> > 26 files changed, 11246 insertions(+), 1 deletion(-)
> > create mode 100644 contrib/virtiofsd/50-qemu-virtiofsd.json.in
> > create mode 100644 contrib/virtiofsd/Makefile.objs
> > create mode 100644 contrib/virtiofsd/buffer.c
> > create mode 100644 contrib/virtiofsd/fuse.h
> > create mode 100644 contrib/virtiofsd/fuse_common.h
> > create mode 100644 contrib/virtiofsd/fuse_i.h
> > create mode 100644 contrib/virtiofsd/fuse_kernel.h
> > create mode 100644 contrib/virtiofsd/fuse_log.c
> > create mode 100644 contrib/virtiofsd/fuse_log.h
> > create mode 100644 contrib/virtiofsd/fuse_loop_mt.c
> > create mode 100644 contrib/virtiofsd/fuse_lowlevel.c
> > create mode 100644 contrib/virtiofsd/fuse_lowlevel.h
> > create mode 100644 contrib/virtiofsd/fuse_misc.h
> > create mode 100644 contrib/virtiofsd/fuse_opt.c
> > create mode 100644 contrib/virtiofsd/fuse_opt.h
> > create mode 100644 contrib/virtiofsd/fuse_signals.c
> > create mode 100644 contrib/virtiofsd/fuse_virtio.c
> > create mode 100644 contrib/virtiofsd/fuse_virtio.h
> > create mode 100644 contrib/virtiofsd/helper.c
> > create mode 100644 contrib/virtiofsd/passthrough_helpers.h
> > create mode 100644 contrib/virtiofsd/passthrough_ll.c
> >
> > --
> > 2.23.0
--
Dr. David Alan Gilbert / address@hidden / Manchester, UK
- Re: [PATCH 00/30] virtiofs daemon (base), (continued)
- Re: [PATCH 00/30] virtiofs daemon (base), Dr. David Alan Gilbert, 2019/10/21
- Re: [PATCH 00/30] virtiofs daemon (base), Michael S. Tsirkin, 2019/10/24
- Re: [PATCH 00/30] virtiofs daemon (base), Dr. David Alan Gilbert, 2019/10/24
- Re: [PATCH 00/30] virtiofs daemon (base), Daniel P . Berrangé, 2019/10/24
- Re: [PATCH 00/30] virtiofs daemon (base), Dr. David Alan Gilbert, 2019/10/24
- Re: [PATCH 00/30] virtiofs daemon (base), Michael S. Tsirkin, 2019/10/24
- Re: [PATCH 00/30] virtiofs daemon (base), Dr. David Alan Gilbert, 2019/10/24
- Re: [PATCH 00/30] virtiofs daemon (base), Daniel P . Berrangé, 2019/10/24
- Re: [PATCH 00/30] virtiofs daemon (base), Dr. David Alan Gilbert, 2019/10/24
Re: [PATCH 00/30] virtiofs daemon (base), Michael S. Tsirkin, 2019/10/24
- Re: [PATCH 00/30] virtiofs daemon (base),
Dr. David Alan Gilbert <=
Re: [PATCH 00/30] virtiofs daemon (base), Michael S. Tsirkin, 2019/10/27
Re: [PATCH 00/30] virtiofs daemon (base), Michael S. Tsirkin, 2019/10/29