[PATCH net-next 00/12] netlink: specs: rtnetlink: adjust specs for C codegen

[PATCH net-next 00/12] netlink: specs: rtnetlink: adjust specs for C codegen

[Jakub Kicinski]

The first patch brings a schema extension allowing specifying
"header" (as in .h file) properties in attribute sets.
This is used for rare cases where we carry attributes from
another family in a nest - we need to include the extra
headers. If we were to generate kernel code we'd also
need to skip it in the uAPI output.

The remaining 11 patches are pretty boring schema adjustments.

Jakub Kicinski (12):
  netlink: specs: allow header properties for attribute sets
  netlink: specs: rt-link: remove the fixed members from attrs
  netlink: specs: rt-link: remove if-netnsid from attr list
  netlink: specs: rt-link: remove duplicated group in attr list
  netlink: specs: rt-link: add C naming info
  netlink: specs: rt-link: adjust AF_ nest for C codegen
  netlink: specs: rt-link: make bond's ipv6 address attribute fixed size
  netlink: specs: rt-link: add notification for newlink
  netlink: specs: rt-neigh: add C naming info
  netlink: specs: rt-neigh: make sure getneigh is consistent
  netlink: specs: rtnetlink: correct notify properties
  netlink: specs: rt-rule: add C naming info

 Documentation/netlink/genetlink-c.yaml      |  3 +
 Documentation/netlink/genetlink-legacy.yaml |  3 +
 Documentation/netlink/netlink-raw.yaml      |  3 +
 Documentation/netlink/specs/rt-link.yaml    | 61 ++++++++++++++-------
 Documentation/netlink/specs/rt-neigh.yaml   | 12 +++-
 Documentation/netlink/specs/rt-rule.yaml    |  8 ++-
 tools/net/ynl/pyynl/ynl_gen_c.py            |  2 +-
 7 files changed, 69 insertions(+), 23 deletions(-)

-- 
2.49.0

[PATCH net-next 01/12] netlink: specs: allow header properties for attribute sets

[Jakub Kicinski]

rt-link has a number of disjoint headers, plus it uses attributes
of other families (e.g. DPLL). Allow declaring a attribute set
as "foreign" by specifying which header its definition is coming
from.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/genetlink-c.yaml      | 3 +++
 Documentation/netlink/genetlink-legacy.yaml | 3 +++
 Documentation/netlink/netlink-raw.yaml      | 3 +++
 tools/net/ynl/pyynl/ynl_gen_c.py            | 2 +-
 4 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/Documentation/netlink/genetlink-c.yaml b/Documentation/netlink/genetlink-c.yaml
index 96fa1f1522ed..5a234e9b5fa2 100644
--- a/Documentation/netlink/genetlink-c.yaml
+++ b/Documentation/netlink/genetlink-c.yaml
@@ -148,6 +148,9 @@ additionalProperties: False
         attr-max-name:
           description: The explicit name for last member of attribute enum.
           type: string
+        header:
+          description: For C-compatible languages, header which already defines this attribute set.
+          type: string
         # End genetlink-c
         attributes:
           description: List of attributes in the space.
diff --git a/Documentation/netlink/genetlink-legacy.yaml b/Documentation/netlink/genetlink-legacy.yaml
index a8c5b521937d..4cbfe666e6f5 100644
--- a/Documentation/netlink/genetlink-legacy.yaml
+++ b/Documentation/netlink/genetlink-legacy.yaml
@@ -193,6 +193,9 @@ additionalProperties: False
         attr-max-name:
           description: The explicit name for last member of attribute enum.
           type: string
+        header:
+          description: For C-compatible languages, header which already defines this attribute set.
+          type: string
         # End genetlink-c
         attributes:
           description: List of attributes in the space.
diff --git a/Documentation/netlink/netlink-raw.yaml b/Documentation/netlink/netlink-raw.yaml
index 1b0772c8e333..e34bf23897fa 100644
--- a/Documentation/netlink/netlink-raw.yaml
+++ b/Documentation/netlink/netlink-raw.yaml
@@ -207,6 +207,9 @@ additionalProperties: False
         attr-max-name:
           description: The explicit name for last member of attribute enum.
           type: string
+        header:
+          description: For C-compatible languages, header which already defines this attribute set.
+          type: string
         # End genetlink-c
         attributes:
           description: List of attributes in the space.
diff --git a/tools/net/ynl/pyynl/ynl_gen_c.py b/tools/net/ynl/pyynl/ynl_gen_c.py
index 0d930c17f963..9613a6135003 100755
--- a/tools/net/ynl/pyynl/ynl_gen_c.py
+++ b/tools/net/ynl/pyynl/ynl_gen_c.py
@@ -2909,7 +2909,7 @@ _C_KW = {
             cw.p(f'#include "{hdr_file}"')
             cw.p('#include "ynl.h"')
         headers = []
-    for definition in parsed['definitions']:
+    for definition in parsed['definitions'] + parsed['attribute-sets']:
         if 'header' in definition:
             headers.append(definition['header'])
     if args.mode == 'user':
-- 
2.49.0

[PATCH net-next 02/12] netlink: specs: rt-link: remove the fixed members from attrs

[Jakub Kicinski]

The purpose of the attribute list is to list the attributes
which will be included in a given message to shrink the objects
for families with huge attr spaces. Fixed headers are always
present in their entirety (between netlink header and the attrs)
so there's no point in listing their members. Current C codegen
doesn't expect them and tries to look them up in the attribute space.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/specs/rt-link.yaml | 13 -------------
 1 file changed, 13 deletions(-)

diff --git a/Documentation/netlink/specs/rt-link.yaml b/Documentation/netlink/specs/rt-link.yaml
index 726dfa083d14..cb7bacbd3d95 100644
--- a/Documentation/netlink/specs/rt-link.yaml
+++ b/Documentation/netlink/specs/rt-link.yaml
@@ -2367,7 +2367,6 @@ protonum: 0
         request:
           value: 16
           attributes: &link-new-attrs
-            - ifi-index
             - ifname
             - net-ns-pid
             - net-ns-fd
@@ -2399,7 +2398,6 @@ protonum: 0
         request:
           value: 17
           attributes:
-            - ifi-index
             - ifname
     -
       name: getlink
@@ -2410,7 +2408,6 @@ protonum: 0
         request:
           value: 18
           attributes:
-            - ifi-index
             - ifname
             - alt-ifname
             - ext-mask
@@ -2418,11 +2415,6 @@ protonum: 0
         reply:
           value: 16
           attributes: &link-all-attrs
-            - ifi-family
-            - ifi-type
-            - ifi-index
-            - ifi-flags
-            - ifi-change
             - address
             - broadcast
             - ifname
@@ -2515,14 +2507,9 @@ protonum: 0
       do:
         request:
           value: 94
-          attributes:
-            - ifindex
         reply:
           value: 92
           attributes: &link-stats-attrs
-            - family
-            - ifindex
-            - filter-mask
             - link-64
             - link-xstats
             - link-xstats-slave
-- 
2.49.0

[PATCH net-next 03/12] netlink: specs: rt-link: remove if-netnsid from attr list

[Jakub Kicinski]

if-netnsid an alias to target-netnsid:

  IFLA_TARGET_NETNSID = IFLA_IF_NETNSID, /* new alias */

We don't have a definition for this attr.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/specs/rt-link.yaml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/Documentation/netlink/specs/rt-link.yaml b/Documentation/netlink/specs/rt-link.yaml
index cb7bacbd3d95..7f411e8cd755 100644
--- a/Documentation/netlink/specs/rt-link.yaml
+++ b/Documentation/netlink/specs/rt-link.yaml
@@ -2460,7 +2460,6 @@ protonum: 0
             - xdp
             - event
             - new-netnsid
-            - if-netnsid
             - target-netnsid
             - carrier-up-count
             - carrier-down-count
-- 
2.49.0

[PATCH net-next 04/12] netlink: specs: rt-link: remove duplicated group in attr list

[Jakub Kicinski]

group is listed twice for newlink.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/specs/rt-link.yaml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/Documentation/netlink/specs/rt-link.yaml b/Documentation/netlink/specs/rt-link.yaml
index 7f411e8cd755..38f439911f94 100644
--- a/Documentation/netlink/specs/rt-link.yaml
+++ b/Documentation/netlink/specs/rt-link.yaml
@@ -2382,7 +2382,6 @@ protonum: 0
             - txqlen
             - operstate
             - linkmode
-            - group
             - gso-max-size
             - gso-max-segs
             - gro-max-size
-- 
2.49.0

[PATCH net-next 05/12] netlink: specs: rt-link: add C naming info

[Jakub Kicinski]

Add properties needed for C codegen to match names with uAPI headers.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/specs/rt-link.yaml | 30 +++++++++++++++++++++++-
 1 file changed, 29 insertions(+), 1 deletion(-)

diff --git a/Documentation/netlink/specs/rt-link.yaml b/Documentation/netlink/specs/rt-link.yaml
index 38f439911f94..a331eb5eecb2 100644
--- a/Documentation/netlink/specs/rt-link.yaml
+++ b/Documentation/netlink/specs/rt-link.yaml
@@ -2,6 +2,7 @@
 
 name: rt-link
 protocol: netlink-raw
+uapi-header: linux/rtnetlink.h
 protonum: 0
 
 doc:
@@ -11,6 +12,9 @@ protonum: 0
   -
     name: ifinfo-flags
     type: flags
+    header: linux/if.h
+    enum-name: net-device-flags
+    name-prefix: iff-
     entries:
       -
         name: up
@@ -53,6 +57,7 @@ protonum: 0
   -
     name: vlan-protocols
     type: enum
+    enum-name:
     entries:
       -
         name: 8021q
@@ -754,6 +759,7 @@ protonum: 0
   -
     name: vlan-flags
     type: flags
+    enum-name:
     entries:
       - reorder-hdr
       - gvrp
@@ -840,6 +846,7 @@ protonum: 0
   -
     name: ifla-vf-link-state-enum
     type: enum
+    enum-name:
     entries:
       - auto
       - enable
@@ -906,6 +913,7 @@ protonum: 0
   -
     name: rtext-filter
     type: flags
+    enum-name:
     entries:
       - vf
       - brvlan
@@ -918,6 +926,7 @@ protonum: 0
   -
     name: netkit-policy
     type: enum
+    enum-name:
     entries:
       -
         name: forward
@@ -928,6 +937,7 @@ protonum: 0
   -
     name: netkit-mode
     type: enum
+    enum-name: netkit-mode
     entries:
       - name: l2
       - name: l3
@@ -935,6 +945,7 @@ protonum: 0
   -
     name: netkit-scrub
     type: enum
+    enum-name:
     entries:
       - name: none
       - name: default
@@ -1195,6 +1206,7 @@ protonum: 0
         nested-attributes: mctp-attrs
   -
     name: vfinfo-list-attrs
+    name-prefix: ifla-vf-
     attributes:
       -
         name: info
@@ -1203,6 +1215,7 @@ protonum: 0
         multi-attr: true
   -
     name: vfinfo-attrs
+    name-prefix: ifla-vf-
     attributes:
       -
         name: mac
@@ -1257,6 +1270,7 @@ protonum: 0
         type: binary
   -
     name: vf-stats-attrs
+    name-prefix: ifla-vf-stats-
     attributes:
       -
         name: rx-packets
@@ -1288,6 +1302,8 @@ protonum: 0
         type: u64
   -
     name: vf-vlan-attrs
+    name-prefix: ifla-vf-vlan-
+    attr-max-name: ifla-vf-vlan-info-max
     attributes:
       -
         name: info
@@ -1296,12 +1312,15 @@ protonum: 0
         multi-attr: true
   -
     name: vf-ports-attrs
+    name-prefix: ifla-
     attributes: []
   -
     name: port-self-attrs
+    name-prefix: ifla-
     attributes: []
   -
     name: linkinfo-attrs
+    name-prefix: ifla-info-
     attributes:
       -
         name: kind
@@ -1855,6 +1874,7 @@ protonum: 0
   -
     name: linkinfo-vti-attrs
     name-prefix: ifla-vti-
+    header: linux/if_tunnel.h
     attributes:
       -
         name: link
@@ -2107,7 +2127,7 @@ protonum: 0
         byte-order: big-endian
   -
     name: ifla-vlan-qos
-    name-prefix: ifla-vlan-qos
+    name-prefix: ifla-vlan-qos-
     attributes:
       -
         name: mapping
@@ -2123,6 +2143,7 @@ protonum: 0
         type: u32
   -
     name: xdp-attrs
+    name-prefix: ifla-xdp-
     attributes:
       -
         name: fd
@@ -2150,6 +2171,7 @@ protonum: 0
         type: s32
   -
     name: ifla-attrs
+    name-prefix: ifla-inet-
     attributes:
       -
         name: conf
@@ -2157,6 +2179,7 @@ protonum: 0
         struct: ipv4-devconf
   -
     name: ifla6-attrs
+    name-prefix: ifla-inet6-
     attributes:
       -
         name: flags
@@ -2222,6 +2245,7 @@ protonum: 0
         type: binary
   -
     name: link-offload-xstats
+    name-prefix: ifla-offload-xstats-
     attributes:
       -
         name: cpu-hit
@@ -2236,6 +2260,7 @@ protonum: 0
         type: binary
   -
     name: hw-s-info-one
+    name-prefix: ifla-offload-xstats-hw-s-info-
     attributes:
       -
         name: request
@@ -2245,6 +2270,8 @@ protonum: 0
         type: u8
   -
     name: link-dpll-pin-attrs
+    name-prefix: dpll-a-
+    header: linux/dpll.h
     attributes:
       -
         name: id
@@ -2357,6 +2384,7 @@ protonum: 0
 
 operations:
   enum-model: directional
+  name-prefix: rtm-
   list:
     -
       name: newlink
-- 
2.49.0

[PATCH net-next 06/12] netlink: specs: rt-link: adjust AF_ nest for C codegen

[Jakub Kicinski]

The AF nest is indexed by AF ID, so it's a bit strange,
but with minor adjustments C codegen deals with it just fine.
Entirely unclear why the names have been in quotes here.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/specs/rt-link.yaml | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/Documentation/netlink/specs/rt-link.yaml b/Documentation/netlink/specs/rt-link.yaml
index a331eb5eecb2..6ab9b876a464 100644
--- a/Documentation/netlink/specs/rt-link.yaml
+++ b/Documentation/netlink/specs/rt-link.yaml
@@ -1188,19 +1188,21 @@ protonum: 0
         multi-attr: true
   -
     name: af-spec-attrs
+    name-prefix: af-
+    attr-max-name: af-max
     attributes:
       -
-        name: "inet"
+        name: inet
         type: nest
         value: 2
         nested-attributes: ifla-attrs
       -
-        name: "inet6"
+        name: inet6
         type: nest
         value: 10
         nested-attributes: ifla6-attrs
       -
-        name: "mctp"
+        name: mctp
         type: nest
         value: 45
         nested-attributes: mctp-attrs
-- 
2.49.0

[PATCH net-next 07/12] netlink: specs: rt-link: make bond's ipv6 address attribute fixed size

[Jakub Kicinski]

ns-ip6-target is an indexed-array. Codegen for variable size binary
array would be a bit tedious, tell C that we know the size of these
attributes, since they are IPv6 addrs.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/specs/rt-link.yaml | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/Documentation/netlink/specs/rt-link.yaml b/Documentation/netlink/specs/rt-link.yaml
index 6ab9b876a464..4b51c5b60d95 100644
--- a/Documentation/netlink/specs/rt-link.yaml
+++ b/Documentation/netlink/specs/rt-link.yaml
@@ -1447,6 +1447,8 @@ protonum: 0
         type: indexed-array
         sub-type: binary
         display-hint: ipv6
+        checks:
+          exact-len: 16
       -
         name: coupled-control
         type: u8
-- 
2.49.0

[PATCH net-next 08/12] netlink: specs: rt-link: add notification for newlink

[Jakub Kicinski]

Add a notification entry for netlink so that we can test ntf handling
in classic netlink and C.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/specs/rt-link.yaml | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/Documentation/netlink/specs/rt-link.yaml b/Documentation/netlink/specs/rt-link.yaml
index 4b51c5b60d95..25f0c3c6a886 100644
--- a/Documentation/netlink/specs/rt-link.yaml
+++ b/Documentation/netlink/specs/rt-link.yaml
@@ -2420,6 +2420,12 @@ protonum: 0
             - gso-ipv4-max-size
             - gro-ipv4-max-size
             - af-spec
+    -
+      name: newlink-ntf
+      doc: Notify that a link has been created
+      value: 16
+      notify: getlink
+      fixed-header: ifinfomsg
     -
       name: dellink
       doc: Delete an existing link.
-- 
2.49.0

[PATCH net-next 09/12] netlink: specs: rt-neigh: add C naming info

[Jakub Kicinski]

Add properties needed for C codegen to match names with uAPI headers.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/specs/rt-neigh.yaml | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/Documentation/netlink/specs/rt-neigh.yaml b/Documentation/netlink/specs/rt-neigh.yaml
index a843caa72259..9b87efaafd15 100644
--- a/Documentation/netlink/specs/rt-neigh.yaml
+++ b/Documentation/netlink/specs/rt-neigh.yaml
@@ -2,6 +2,7 @@
 
 name: rt-neigh
 protocol: netlink-raw
+uapi-header: linux/rtnetlink.h
 protonum: 0
 
 doc:
@@ -48,6 +49,7 @@ protonum: 0
   -
     name: nud-state
     type: flags
+    enum-name:
     entries:
       - incomplete
       - reachable
@@ -60,6 +62,7 @@ protonum: 0
   -
     name: ntf-flags
     type: flags
+    enum-name:
     entries:
       - use
       - self
@@ -72,12 +75,14 @@ protonum: 0
   -
     name: ntf-ext-flags
     type: flags
+    enum-name:
     entries:
       - managed
       - locked
   -
     name: rtm-type
     type: enum
+    enum-name:
     entries:
       - unspec
       - unicast
@@ -179,6 +184,7 @@ protonum: 0
 attribute-sets:
   -
     name: neighbour-attrs
+    name-prefix: nda-
     attributes:
       -
         name: unspec
@@ -241,6 +247,7 @@ protonum: 0
         type: u8
   -
     name: ndt-attrs
+    name-prefix: ndta-
     attributes:
       -
         name: name
@@ -274,6 +281,7 @@ protonum: 0
         type: pad
   -
     name: ndtpa-attrs
+    name-prefix: ndtpa-
     attributes:
       -
         name: ifindex
@@ -335,6 +343,7 @@ protonum: 0
 
 operations:
   enum-model: directional
+  name-prefix: rtm-
   list:
     -
       name: newneigh
-- 
2.49.0

[PATCH net-next 10/12] netlink: specs: rt-neigh: make sure getneigh is consistent

[Jakub Kicinski]

The consistency check complains replies to do and dump don't match
because dump has no value. It doesn't have to by the schema... but
fixing this in code gen would be more code than adjusting the spec.
This is rare.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/specs/rt-neigh.yaml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/Documentation/netlink/specs/rt-neigh.yaml b/Documentation/netlink/specs/rt-neigh.yaml
index 9b87efaafd15..fe34ade6b300 100644
--- a/Documentation/netlink/specs/rt-neigh.yaml
+++ b/Documentation/netlink/specs/rt-neigh.yaml
@@ -402,6 +402,7 @@ protonum: 0
             - ifindex
             - master
         reply:
+          value: 28
           attributes: *neighbour-all
     -
       name: newneigh-ntf
-- 
2.49.0

[PATCH net-next 11/12] netlink: specs: rtnetlink: correct notify properties

[Jakub Kicinski]

The notify property should point at the object the notifications
carry, usually the get object, not the cmd which triggers
the notification:

  notify:
    description: Name of the command sharing the reply type with
                 this notification.

Not treating this as a fix, I think that only C codegen cares.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/specs/rt-neigh.yaml | 2 +-
 Documentation/netlink/specs/rt-rule.yaml  | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/Documentation/netlink/specs/rt-neigh.yaml b/Documentation/netlink/specs/rt-neigh.yaml
index fe34ade6b300..e9cba164e3d1 100644
--- a/Documentation/netlink/specs/rt-neigh.yaml
+++ b/Documentation/netlink/specs/rt-neigh.yaml
@@ -381,7 +381,7 @@ protonum: 0
       name: delneigh-ntf
       doc: Notify a neighbour deletion
       value: 29
-      notify: delneigh
+      notify: getneigh
       fixed-header: ndmsg
     -
       name: getneigh
diff --git a/Documentation/netlink/specs/rt-rule.yaml b/Documentation/netlink/specs/rt-rule.yaml
index de0938d36541..f585654a4d41 100644
--- a/Documentation/netlink/specs/rt-rule.yaml
+++ b/Documentation/netlink/specs/rt-rule.yaml
@@ -234,7 +234,7 @@ protonum: 0
       name: newrule-ntf
       doc: Notify a rule creation
       value: 32
-      notify: newrule
+      notify: getrule
     -
       name: delrule
       doc: Remove an existing FIB rule
@@ -247,7 +247,7 @@ protonum: 0
       name: delrule-ntf
       doc: Notify a rule deletion
       value: 33
-      notify: delrule
+      notify: getrule
     -
       name: getrule
       doc: Dump all FIB rules
-- 
2.49.0

[PATCH net-next 12/12] netlink: specs: rt-rule: add C naming info

[Jakub Kicinski]

Add properties needed for C codegen to match names with uAPI headers.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 Documentation/netlink/specs/rt-rule.yaml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/Documentation/netlink/specs/rt-rule.yaml b/Documentation/netlink/specs/rt-rule.yaml
index f585654a4d41..003707ca4a3e 100644
--- a/Documentation/netlink/specs/rt-rule.yaml
+++ b/Documentation/netlink/specs/rt-rule.yaml
@@ -2,6 +2,7 @@
 
 name: rt-rule
 protocol: netlink-raw
+uapi-header: linux/fib_rules.h
 protonum: 0
 
 doc:
@@ -56,6 +57,7 @@ protonum: 0
   -
     name: fr-act
     type: enum
+    enum-name:
     entries:
       - unspec
       - to-tbl
@@ -90,6 +92,7 @@ protonum: 0
 attribute-sets:
   -
     name: fib-rule-attrs
+    name-prefix: fra-
     attributes:
       -
         name: dst
@@ -198,6 +201,7 @@ protonum: 0
 operations:
   enum-model: directional
   fixed-header: fib-rule-hdr
+  name-prefix: rtm-
   list:
     -
       name: newrule
-- 
2.49.0

Re: [PATCH net-next 01/12] netlink: specs: allow header properties for attribute sets

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> rt-link has a number of disjoint headers, plus it uses attributes
> of other families (e.g. DPLL). Allow declaring a attribute set
> as "foreign" by specifying which header its definition is coming
> from.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

Re: [PATCH net-next 02/12] netlink: specs: rt-link: remove the fixed members from attrs

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> The purpose of the attribute list is to list the attributes
> which will be included in a given message to shrink the objects
> for families with huge attr spaces. Fixed headers are always
> present in their entirety (between netlink header and the attrs)
> so there's no point in listing their members. Current C codegen
> doesn't expect them and tries to look them up in the attribute space.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

I think my intent when I first wrote this spec was to list the fixed
header fields that are used or required by each op, not realising the
intended codegen purpose of the attributes list.

I guess we could add usage details to the doc string for each op.

> ---
>  Documentation/netlink/specs/rt-link.yaml | 13 -------------
>  1 file changed, 13 deletions(-)
>
> diff --git a/Documentation/netlink/specs/rt-link.yaml b/Documentation/netlink/specs/rt-link.yaml
> index 726dfa083d14..cb7bacbd3d95 100644
> --- a/Documentation/netlink/specs/rt-link.yaml
> +++ b/Documentation/netlink/specs/rt-link.yaml
> @@ -2367,7 +2367,6 @@ protonum: 0
>          request:
>            value: 16
>            attributes: &link-new-attrs
> -            - ifi-index
>              - ifname
>              - net-ns-pid
>              - net-ns-fd
> @@ -2399,7 +2398,6 @@ protonum: 0
>          request:
>            value: 17
>            attributes:
> -            - ifi-index
>              - ifname
>      -
>        name: getlink
> @@ -2410,7 +2408,6 @@ protonum: 0
>          request:
>            value: 18
>            attributes:
> -            - ifi-index
>              - ifname
>              - alt-ifname
>              - ext-mask
> @@ -2418,11 +2415,6 @@ protonum: 0
>          reply:
>            value: 16
>            attributes: &link-all-attrs
> -            - ifi-family
> -            - ifi-type
> -            - ifi-index
> -            - ifi-flags
> -            - ifi-change
>              - address
>              - broadcast
>              - ifname
> @@ -2515,14 +2507,9 @@ protonum: 0
>        do:
>          request:
>            value: 94
> -          attributes:
> -            - ifindex
>          reply:
>            value: 92
>            attributes: &link-stats-attrs
> -            - family
> -            - ifindex
> -            - filter-mask
>              - link-64
>              - link-xstats
>              - link-xstats-slave

Re: [PATCH net-next 03/12] netlink: specs: rt-link: remove if-netnsid from attr list

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> if-netnsid an alias to target-netnsid:
>
>   IFLA_TARGET_NETNSID = IFLA_IF_NETNSID, /* new alias */
>
> We don't have a definition for this attr.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

Re: [PATCH net-next 04/12] netlink: specs: rt-link: remove duplicated group in attr list

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> group is listed twice for newlink.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

Re: [PATCH net-next 05/12] netlink: specs: rt-link: add C naming info

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> Add properties needed for C codegen to match names with uAPI headers.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

Re: [PATCH net-next 06/12] netlink: specs: rt-link: adjust AF_ nest for C codegen

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> The AF nest is indexed by AF ID, so it's a bit strange,
> but with minor adjustments C codegen deals with it just fine.
> Entirely unclear why the names have been in quotes here.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

Re: [PATCH net-next 07/12] netlink: specs: rt-link: make bond's ipv6 address attribute fixed size

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> ns-ip6-target is an indexed-array. Codegen for variable size binary
> array would be a bit tedious, tell C that we know the size of these
> attributes, since they are IPv6 addrs.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

Re: [PATCH net-next 08/12] netlink: specs: rt-link: add notification for newlink

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> Add a notification entry for netlink so that we can test ntf handling

nit: newlink

> in classic netlink and C.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

Re: [PATCH net-next 09/12] netlink: specs: rt-neigh: add C naming info

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> Add properties needed for C codegen to match names with uAPI headers.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

Re: [PATCH net-next 10/12] netlink: specs: rt-neigh: make sure getneigh is consistent

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> The consistency check complains replies to do and dump don't match
> because dump has no value. It doesn't have to by the schema... but
> fixing this in code gen would be more code than adjusting the spec.
> This is rare.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

Re: [PATCH net-next 11/12] netlink: specs: rtnetlink: correct notify properties

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> The notify property should point at the object the notifications
> carry, usually the get object, not the cmd which triggers
> the notification:
>
>   notify:
>     description: Name of the command sharing the reply type with
>                  this notification.
>
> Not treating this as a fix, I think that only C codegen cares.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

Re: [PATCH net-next 12/12] netlink: specs: rt-rule: add C naming info

[Donald Hunter]

Jakub Kicinski <kuba@kernel.org> writes:

> Add properties needed for C codegen to match names with uAPI headers.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Reviewed-by: Donald Hunter <donald.hunter@gmail.com>

Re: [PATCH net-next 00/12] netlink: specs: rtnetlink: adjust specs for C codegen

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Thu, 17 Apr 2025 19:16:54 -0700 you wrote:
> The first patch brings a schema extension allowing specifying
> "header" (as in .h file) properties in attribute sets.
> This is used for rare cases where we carry attributes from
> another family in a nest - we need to include the extra
> headers. If we were to generate kernel code we'd also
> need to skip it in the uAPI output.
> 
> [...]

Here is the summary with links:
  - [net-next,01/12] netlink: specs: allow header properties for attribute sets
    https://git.kernel.org/netdev/net-next/c/7965facefaed
  - [net-next,02/12] netlink: specs: rt-link: remove the fixed members from attrs
    https://git.kernel.org/netdev/net-next/c/43b606d98482
  - [net-next,03/12] netlink: specs: rt-link: remove if-netnsid from attr list
    https://git.kernel.org/netdev/net-next/c/ed43ce6ab222
  - [net-next,04/12] netlink: specs: rt-link: remove duplicated group in attr list
    https://git.kernel.org/netdev/net-next/c/c703d258f626
  - [net-next,05/12] netlink: specs: rt-link: add C naming info
    https://git.kernel.org/netdev/net-next/c/b12b0f41819a
  - [net-next,06/12] netlink: specs: rt-link: adjust AF_ nest for C codegen
    https://git.kernel.org/netdev/net-next/c/e6e1f53f0283
  - [net-next,07/12] netlink: specs: rt-link: make bond's ipv6 address attribute fixed size
    https://git.kernel.org/netdev/net-next/c/1c224f19ff06
  - [net-next,08/12] netlink: specs: rt-link: add notification for newlink
    https://git.kernel.org/netdev/net-next/c/622d7050cfd4
  - [net-next,09/12] netlink: specs: rt-neigh: add C naming info
    https://git.kernel.org/netdev/net-next/c/cd879795c3ee
  - [net-next,10/12] netlink: specs: rt-neigh: make sure getneigh is consistent
    https://git.kernel.org/netdev/net-next/c/eee94a89c55a
  - [net-next,11/12] netlink: specs: rtnetlink: correct notify properties
    https://git.kernel.org/netdev/net-next/c/e3d199d30909
  - [net-next,12/12] netlink: specs: rt-rule: add C naming info
    https://git.kernel.org/netdev/net-next/c/620b38232f43

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html