[PATCH net-next v4 0/3] gtp: ynl spec for the GTP netlink family, a reply-command fix, and a selftest

[Anil Kaushik <anilkaushikwireless@gmail.com>]

This adds a ynl spec for the gtp generic netlink family, a driver fix the
spec exposed, and a selftest that uses it.

While writing a ynl-based selftest for the existing gtp family I found
that the GET and dump replies put the genl family id in the command
field instead of a GTP_CMD_* value. libgtpnl ignores the command so it
never mattered, but ynl rejects the reply as an unknown command and
drops it. Patch 1 fixes the driver; patches 2 and 3 then describe and
test the family.

Patch 1 sets a valid command (GTP_CMD_GETPDP) in the get and dump
replies.

Patch 2 adds a genetlink-legacy ynl spec for the gtp family
(NEWPDP/DELPDP/GETPDP/ECHOREQ). Description only, no uapi change.

Patch 3 adds a selftest built on the spec: it creates a gtp device, adds
PDP contexts with newpdp, checks getpdp dumps them and delpdp removes
them, by TEI and by MS address, and that deleting an unknown context
fails with ENOENT. It skips when the kernel or iproute2 lack GTP
support.

Changes since v3 (thanks to the Sashiko review):
 - spec: echoreq request now matches gtp_genl_send_echo_req(): add
   ms-address (required by the handler) and net-ns-fd (read by
   gtp_find_dev()), drop peer-addr6 and family (never read).
 - spec: delpdp request now lists the attributes the lookup in
   gtp_find_pdp() reads, the same as the getpdp request, instead of
   all the newpdp ones.
 - spec: add link to the getpdp reply, gtp_genl_fill_info() always
   emits it.
 - spec: commit message reworded, the per-command attributes now
   follow the handlers.
 - selftest: create the gtp device, which loads gtp.ko, before the
   family handle: ynl resolves family ids once per process with a dump,
   which does not autoload modules, so with CONFIG_GTP=m the test
   aborted with a traceback. Without GTP support it now skips.
 - selftest: only consider contexts on the test's own gtp device.
 - selftest: delete by TEI only, so the GTPv1 TEI lookup is used, and
   delete by MS address in the two-contexts case.
 - selftest: check that deleting an unknown context fails with ENOENT,
   both by TEI and by MS address.
 - Patch 1 is unchanged from v3.

v3: https://lore.kernel.org/netdev/20261009060615.4177402-1-anilkaushikwireless@gmail.com/
v2: https://lore.kernel.org/netdev/20261008181031.4129029-1-anilkaushikwireless@gmail.com/

Anil Kaushik (3):
  gtp: set a valid genl command in PDP context get and dump replies
  netlink: specs: add genetlink-legacy spec for GTP
  selftests: net: add a test for the gtp netlink family

 Documentation/netlink/specs/gtp.yaml          | 171 ++++++++++++++++++
 MAINTAINERS                                   |   2 +
 drivers/net/gtp.c                             |   4 +-
 tools/testing/selftests/net/Makefile          |   1 +
 tools/testing/selftests/net/config            |   1 +
 tools/testing/selftests/net/gtp.py            | 115 ++++++++++++
 .../testing/selftests/net/lib/py/__init__.py  |   5 +-
 tools/testing/selftests/net/lib/py/ynl.py     |   8 +-
 8 files changed, 302 insertions(+), 5 deletions(-)
 create mode 100644 Documentation/netlink/specs/gtp.yaml
 create mode 100755 tools/testing/selftests/net/gtp.py

-- 
2.25.1