The virtio-net code for eBPF RSS is still ignoring errors when
failing to load the eBPF RSS program passed in by the mgmt app
via pre-opened FDs.

This series re-factors the eBPF common code so that it actually
reports using "Error" objects. Then it makes virtio-net treat
a failure to load pre-opened FDs as a fatal problem. When doing
speculative opening of eBPF FDs, QEMU merely prints a warning,
and allows the software fallback to continue.

Trace event coverage is significantly expanded to make this all
much more debuggable too.

Changed in v2:

 - Split 'ebpf_error' probe into multiple probes

Daniel P. Berrangé (7):
  hw/net: fix typo s/epbf/ebpf/ in virtio-net
  ebpf: drop redundant parameter checks in static methods
  ebpf: improve error trace events
  ebpf: add formal error reporting to all APIs
  hw/net: report errors from failing to use eBPF RSS FDs
  ebpf: improve trace event coverage to all key operations
  hw/net: improve tracing of eBPF RSS setup

 ebpf/ebpf_rss.c     | 118 ++++++++++++++++++++++++++++----------------
 ebpf/ebpf_rss.h     |  10 ++--
 ebpf/trace-events   |   8 ++-
 hw/net/trace-events |   8 +--
 hw/net/virtio-net.c |  63 +++++++++++++++--------
 5 files changed, 137 insertions(+), 70 deletions(-)

-- 
2.45.2

On Thu, Sep 05, 2024 at 07:13:23PM +0100, Daniel P. Berrangé wrote:
> The virtio-net code for eBPF RSS is still ignoring errors when
> failing to load the eBPF RSS program passed in by the mgmt app
> via pre-opened FDs.
> 
> This series re-factors the eBPF common code so that it actually
> reports using "Error" objects. Then it makes virtio-net treat
> a failure to load pre-opened FDs as a fatal problem. When doing
> speculative opening of eBPF FDs, QEMU merely prints a warning,
> and allows the software fallback to continue.
> 
> Trace event coverage is significantly expanded to make this all
> much more debuggable too.


looks good
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>

Jason's tree.

> Changed in v2:
> 
>  - Split 'ebpf_error' probe into multiple probes
> 
> Daniel P. Berrangé (7):
>   hw/net: fix typo s/epbf/ebpf/ in virtio-net
>   ebpf: drop redundant parameter checks in static methods
>   ebpf: improve error trace events
>   ebpf: add formal error reporting to all APIs
>   hw/net: report errors from failing to use eBPF RSS FDs
>   ebpf: improve trace event coverage to all key operations
>   hw/net: improve tracing of eBPF RSS setup
> 
>  ebpf/ebpf_rss.c     | 118 ++++++++++++++++++++++++++++----------------
>  ebpf/ebpf_rss.h     |  10 ++--
>  ebpf/trace-events   |   8 ++-
>  hw/net/trace-events |   8 +--
>  hw/net/virtio-net.c |  63 +++++++++++++++--------
>  5 files changed, 137 insertions(+), 70 deletions(-)
> 
> -- 
> 2.45.2

On Fri, Sep 6, 2024 at 5:57 PM Michael S. Tsirkin <mst@redhat.com> wrote:
>
> On Thu, Sep 05, 2024 at 07:13:23PM +0100, Daniel P. Berrangé wrote:
> > The virtio-net code for eBPF RSS is still ignoring errors when
> > failing to load the eBPF RSS program passed in by the mgmt app
> > via pre-opened FDs.
> >
> > This series re-factors the eBPF common code so that it actually
> > reports using "Error" objects. Then it makes virtio-net treat
> > a failure to load pre-opened FDs as a fatal problem. When doing
> > speculative opening of eBPF FDs, QEMU merely prints a warning,
> > and allows the software fallback to continue.
> >
> > Trace event coverage is significantly expanded to make this all
> > much more debuggable too.
>
>
> looks good
> Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
>
> Jason's tree.
>

Queued.

Thanks

On Mon, Sep 09, 2024 at 10:34:32AM +0800, Jason Wang wrote:
> On Fri, Sep 6, 2024 at 5:57 PM Michael S. Tsirkin <mst@redhat.com> wrote:
> >
> > On Thu, Sep 05, 2024 at 07:13:23PM +0100, Daniel P. Berrangé wrote:
> > > The virtio-net code for eBPF RSS is still ignoring errors when
> > > failing to load the eBPF RSS program passed in by the mgmt app
> > > via pre-opened FDs.
> > >
> > > This series re-factors the eBPF common code so that it actually
> > > reports using "Error" objects. Then it makes virtio-net treat
> > > a failure to load pre-opened FDs as a fatal problem. When doing
> > > speculative opening of eBPF FDs, QEMU merely prints a warning,
> > > and allows the software fallback to continue.
> > >
> > > Trace event coverage is significantly expanded to make this all
> > > much more debuggable too.
> >
> >
> > looks good
> > Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
> >
> > Jason's tree.
> >
> 
> Queued.

I'm still not seeing this in master - is any PR planned in near future ?

With regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|