Re: [PATCH v8 1/3] Documentation: netlink: add a YAML spec for nfsd_server

[Lorenzo Bianconi <lorenzo.bianconi@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 4844 bytes --]

> On Mon, Sep 11, 2023 at 02:49:44PM +0200, Lorenzo Bianconi wrote:
> > Introduce nfsd_server.yaml specs to generate uAPI and netlink
> > code for nfsd server.
> > Add rpc-status specs to define message reported by the nfsd server
> > dumping the pending RPC requests.
> > 
> > Tested-by: Jeff Layton <jlayton@kernel.org>
> > Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
> > ---
> >  Documentation/netlink/specs/nfsd_server.yaml | 97 ++++++++++++++++++++
> >  1 file changed, 97 insertions(+)
> >  create mode 100644 Documentation/netlink/specs/nfsd_server.yaml
> 
> I've had a look... the series is simple and short. Thanks!
> 
> My only quibbles right now are cosmetic and naming-related, all
> of which can be addressed when I apply these. So I'm going to
> wait for other review comments to see if we need another version
> or whether I can apply v8 with by-hand clean-ups.
> 
> Comments below are what I might change when applying this one.
> This is not (yet) a request for a new version.

Hi Chuck,

thx for the review. Please let me know if I need to post a v9.

> 
> 
> > diff --git a/Documentation/netlink/specs/nfsd_server.yaml b/Documentation/netlink/specs/nfsd_server.yaml
> > new file mode 100644
> > index 000000000000..e681b493847b
> > --- /dev/null
> > +++ b/Documentation/netlink/specs/nfsd_server.yaml
> > @@ -0,0 +1,97 @@
> > +# SPDX-License-Identifier: ((GPL-2.0 WITH Linux-syscall-note) OR BSD-3-Clause)
> > +
> > +name: nfsd_server
> 
> IMHO "nfsd_server" is redundant. "nfsd" should work.

ack, fine

> 
> 
> > +
> > +doc:
> > +  nfsd server configuration over generic netlink.
> > +
> > +attribute-sets:
> > +  -
> > +    name: rpc-status-comp-op-attr
> > +    enum-name: nfsd-rpc-status-comp-attr
> > +    name-prefix: nfsd-attr-rpc-status-comp-
> > +    attributes:
> > +      -
> > +        name: unspec
> > +        type: unused
> > +        value: 0
> 
> I don't recall whether a zero-value definition is explicitly
> necessary. Maybe "value-start: 1" would work rather than
> these three lines? Why is zero a special attribute value?

I do not think it is mandatory, I added them just to be aligned with other
netlink definitions, but we do not use them.

> 
> 
> > +      -
> > +        name: op
> > +        type: u32
> > +  -
> > +    name: rpc-status-attr
> > +    enum-name: nfsd-rpc-status-attr
> > +    name-prefix: nfsd-attr-rpc-status-
> 
> Specifying all three of these name settings seems a bit
> cluttered.

ack, I would say we can get rid of enum-name, agree?

> 
> 
> > +    attributes:
> > +      -
> > +        name: unspec
> > +        type: unused
> > +        value: 0
> > +      -
> > +        name: xid
> > +        type: u32
> > +        byte-order: big-endian
> > +      -
> > +        name: flags
> > +        type: u32
> > +      -
> > +        name: prog
> > +        type: u32
> > +      -
> > +        name: version
> > +        type: u8
> > +      -
> > +        name: proc
> > +        type: u32
> > +      -
> > +        name: service_time
> > +        type: s64
> > +      -
> > +        name: pad
> > +        type: pad
> > +      -
> > +        name: saddr4
> > +        type: u32
> > +        byte-order: big-endian
> > +        display-hint: ipv4
> > +      -
> > +        name: daddr4
> > +        type: u32
> > +        byte-order: big-endian
> > +        display-hint: ipv4
> > +      -
> > +        name: saddr6
> > +        type: binary
> > +        display-hint: ipv6
> > +      -
> > +        name: daddr6
> > +        type: binary
> > +        display-hint: ipv6
> > +      -
> > +        name: sport
> > +        type: u16
> > +        byte-order: big-endian
> > +      -
> > +        name: dport
> > +        type: u16
> > +        byte-order: big-endian
> > +      -
> > +        name: compond-op
> 
> s/compond-op/compound-op

ack

> 
> > +        type: array-nest
> > +        nested-attributes: rpc-status-comp-op-attr
> 
> So, this is supposed to be a counted array of op numbers? Is there
> an existing type that could be used for this instead?

I think the attribute-sets available types are defined here:

https://github.com/torvalds/linux/blob/master/Documentation/netlink/genetlink-c.yaml#L151

Regards,
Lorenzo

> 
> 
> > +
> > +operations:
> > +  enum-name: nfsd-commands
> > +  name-prefix: nfsd-cmd-
> > +  list:
> > +    -
> > +      name: unspec
> > +      doc: unused
> > +      value: 0
> > +    -
> > +      name: rpc-status-get
> > +      doc: dump pending nfsd rpc
> > +      attribute-set: rpc-status-attr
> > +      dump:
> > +        pre: nfsd-server-nl-rpc-status-get-start
> > +        post: nfsd-server-nl-rpc-status-get-done
> 
> -- 
> Chuck Lever
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]