Re: [ANNOUNCE] libblkio v0.1.0 preview release

["Richard W.M. Jones" <rjones@redhat.com>]

On Thu, Apr 29, 2021 at 03:41:29PM +0100, Stefan Hajnoczi wrote:
> On Thu, Apr 29, 2021 at 03:22:59PM +0100, Richard W.M. Jones wrote:
> > On Thu, Apr 29, 2021 at 03:05:50PM +0100, Stefan Hajnoczi wrote:
> > > The purpose of this preview release is to discuss both the API design
> > > and general direction of the project. API documentation is available
> > > here:
> > > 
> > >   https://gitlab.com/libblkio/libblkio/-/blob/v0.1.0/docs/blkio.rst
> > 
> > libvirt originally, and now libnbd, keep a per-thread error message
> > (stored in thread-local storage).  It's a lot nicer than having to
> > pass &errmsg to every function.  You can just write:
> > 
> >  if (nbd_connect_tcp (nbd, "remote", "nbd") == -1) {
> >    fprintf (stderr,
> >             "failed to connect to remote server: %s (errno = %d)\n",
> >             nbd_get_error (), nbd_get_errno ());
> >    exit (EXIT_FAILURE);
> >  }
> > 
> > (https://libguestfs.org/libnbd.3.html#ERROR-HANDLING)
> > 
> > It means you can extend the range of error information available in
> > future.  Also you can return a 'const char *' and the application
> > doesn't have to worry about lifetimes, at least in the common case.
> 
> Thanks for sharing the idea, I think it would work well for libblkio
> too.
> 
> Do you ignore the dlclose(3) memory leak?

I believe this mechanism in libnbd ensures there is no leak in the
ordinary shared library (not dlopen/dlclose) case:

https://gitlab.com/nbdkit/libnbd/-/blob/f9257a9fdc68706a4079deb4ced61e1d468f28d6/lib/errors.c#L35

However I'm not sure what happens in the dlopen case, so I'm going to
test that out now ...

> > > Examples are available here:
> > > 
> > >   https://gitlab.com/libblkio/libblkio/-/tree/v0.1.0/examples
> > > 
> > > The goal is to eventually include the following drivers:
> > > - Linux io_uring
> > > - NVMe (VFIO and vfio-user)
> > > - virtio-blk (VFIO, vfio-user, vhost-user-blk, and vhost-vdpa-blk)
> > >
> > > There are a few reasons why libblkio is needed:
> > > 
> > > 1. QEMU, Ceph, GlusterFS, MariaDB, and other programs have been adding
> > >    more low-level block I/O code, most of it duplicated. Usually only
> > >    one or two of Linux AIO, io_uring, userspace drivers, vhost-user
> > >    drivers, etc are implemented. This makes it difficult to benefit from
> > >    the latest advances in high-performance block I/O.
> > > 
> > > 2. Coding to a standard API makes it possible to introduce new
> > >    optimizations or hardware interfaces without costly changes to the
> > >    software stack.
> > > 
> > > 3. A client library is needed so applications can take advantage of
> > >    qemu-storage-daemon's vhost-user-blk exports.
> > > 
> > > 4. Implementing block I/O as a library allows QEMU to use Rust for new
> > >    code without messy QEMU internal API bindings. Note that libblkio
> > >    currently does not provide a Rust crate, it only offers a C API.
> > 
> > This is where I get confused about what this library actually does.
> > It's not just a nicer wrapper around io_uring, but what is it actually
> > doing?
> 
> It's a library for what QEMU calls protocol drivers (POSIX files,
> userspace NVMe driver, etc). In particular, anything offering
> multi-queue block I/O fits into libblkio.
> 
> It is not intended to replace libnbd or other network storage libraries.
> libblkio's properties API is synchronous to keep things simple for
> applications. Attaching to network storage needs to be asynchronous,
> although the libblkio API could be extended if people want to support
> network storage.

I think what confuses me is why is NVMe any different from io_uring?
ie would this work?

$ blkio-info --output=json io_uring path=/dev/nvme0

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-df lists disk usage of guests without needing to install any
software inside the virtual machine.  Supports Linux and Windows.
http://people.redhat.com/~rjones/virt-df/