batman-adv: Avoid spurious warnings from bat_v neigh_cmp implementation
Commit Message
The neighbor compare API implementation for B.A.T.M.A.N. V checks whether
the neigh_ifinfo for this neighbor on a specific interface exists. A
warning is printed when it isn't found.
But it is not called inside a lock which would prevent that this
information is lost right before batadv_neigh_ifinfo_get. It must therefore
be expected that batadv_v_neigh_(cmp|is_sob) might not be able to get the
requested neigh_ifinfo.
A WARN_ON for such a situation seems not to be appropriate because this
will only flood the kernel logs. The warnings must therefore be removed.
Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com>
---
Cc: Antonio Quartulli <a@unstable.cc>
This patch is for https://www.open-mesh.org/issues/343
---
net/batman-adv/bat_v.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
Comments
On 16/10/17 15:48, Sven Eckelmann wrote:
> The neighbor compare API implementation for B.A.T.M.A.N. V checks whether
> the neigh_ifinfo for this neighbor on a specific interface exists. A
> warning is printed when it isn't found.
>
> But it is not called inside a lock which would prevent that this
> information is lost right before batadv_neigh_ifinfo_get. It must therefore
> be expected that batadv_v_neigh_(cmp|is_sob) might not be able to get the
> requested neigh_ifinfo.
>
> A WARN_ON for such a situation seems not to be appropriate because this
> will only flood the kernel logs. The warnings must therefore be removed.
>
> Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com>
Yeah, this makes sense. As a read only operation, the lock is not held
on purpose, therefore such situation has to be expected.
Acked-by: Antonio Quartulli <a@unstable.cc>
On Montag, 16. Oktober 2017 09:48:03 CEST Sven Eckelmann wrote:
> The neighbor compare API implementation for B.A.T.M.A.N. V checks whether
> the neigh_ifinfo for this neighbor on a specific interface exists. A
> warning is printed when it isn't found.
>
> But it is not called inside a lock which would prevent that this
> information is lost right before batadv_neigh_ifinfo_get. It must therefore
> be expected that batadv_v_neigh_(cmp|is_sob) might not be able to get the
> requested neigh_ifinfo.
>
> A WARN_ON for such a situation seems not to be appropriate because this
> will only flood the kernel logs. The warnings must therefore be removed.
>
> Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com>
> ---
> Cc: Antonio Quartulli <a@unstable.cc>
>
> This patch is for https://www.open-mesh.org/issues/343
Patch was applied in 0dee8aba [1]
Kind regards,
Sven
[1] https://git.open-mesh.org/batman-adv.git/commit/0dee8aba4702f82197ed3428ede6b3884fdff5ca
@@ -623,11 +623,11 @@ static int batadv_v_neigh_cmp(struct batadv_neigh_node *neigh1,
int ret = 0;
ifinfo1 = batadv_neigh_ifinfo_get(neigh1, if_outgoing1);
- if (WARN_ON(!ifinfo1))
+ if (!ifinfo1)
goto err_ifinfo1;
ifinfo2 = batadv_neigh_ifinfo_get(neigh2, if_outgoing2);
- if (WARN_ON(!ifinfo2))
+ if (!ifinfo2)
goto err_ifinfo2;
ret = ifinfo1->bat_v.throughput - ifinfo2->bat_v.throughput;
@@ -649,11 +649,11 @@ static bool batadv_v_neigh_is_sob(struct batadv_neigh_node *neigh1,
bool ret = false;
ifinfo1 = batadv_neigh_ifinfo_get(neigh1, if_outgoing1);
- if (WARN_ON(!ifinfo1))
+ if (!ifinfo1)
goto err_ifinfo1;
ifinfo2 = batadv_neigh_ifinfo_get(neigh2, if_outgoing2);
- if (WARN_ON(!ifinfo2))
+ if (!ifinfo2)
goto err_ifinfo2;
threshold = ifinfo1->bat_v.throughput / 4;