Re: [PATCH] net: dsa: tag_ocelot_8021q: fix broken reception

[Vladimir Oltean]

On Wed, Dec 11, 2024 at 01:46:56PM +0100, Robert Hodaszi wrote:
> Commit dcfe7673787b4bfea2c213df443d312aa754757b ("net: dsa: tag_sja1105:
> absorb logic for not overwriting precise info into dsa_8021q_rcv()")
> added support to let the DSA switch driver set source_port and
> switch_id. tag_8021q's logic overrides the previously set source_port
> and switch_id only if they are marked as "invalid" (-1). sja1105 and
> vsc73xx drivers are doing that properly, but ocelot_8021q driver doesn't
> initialize those variables. That causes dsa_8021q_rcv() doesn't set
> them, and they remain unassigned.
> 
> Initialize them as invalid to so dsa_8021q_rcv() can return with the
> proper values.
> 
> Signed-off-by: Robert Hodaszi <robert.hodaszi@xxxxxxxx>
> Cc: stable@xxxxxxxxxxxxxxx
> ---

Thank you for the patch, and sorry for the breakage. I suggest the
following alternative commit message:

  The blamed commit changed the dsa_8021q_rcv() calling convention to
  accept pre-populated source_port and switch_id arguments. If those are
  not available, as in the case of tag_ocelot_8021q, the arguments must
  be preinitialized with -1.

  Due to the bug of passing uninitialized arguments, dsa_8021q_rcv()
  does not detect that it needs to populate the source_port and
  switch_id, and this makes dsa_conduit_find_user() fail, and leads to
  packet loss on reception.

  Fixes: dcfe7673787b ("net: dsa: tag_sja1105: absorb logic for not overwriting precise info into dsa_8021q_rcv()")
  Signed-off-by: Robert Hodaszi <robert.hodaszi@xxxxxxxx>
  Cc: stable@xxxxxxxxxxxxxxx