On 17.03.25 10:20, Marc Kleine-Budde wrote:
On 16.03.2025 13:39:16, Oliver Hartkopp wrote:
On 14.03.25 13:31, Marc Kleine-Budde wrote:
On 14.03.2025 12:39:49, Davide Caratti wrote:
The third column in the output of the following command:
# grep CAN /proc/net/protocols
is systematically '0': use sock_prot_inuse_add() to account for the number
of sockets for each protocol on top of AF_CAN family.
Signed-off-by: Davide Caratti <dcaratti@xxxxxxxxxx>
Applied to linux-can-next.
Maybe too fast? E.g. J1939 is not handled.
https://lore.kernel.org/linux-can/78951192-82b1-45bc-9903-d314c94cd182@xxxxxxxxxxxx/T/#m87afc41fef8ec9099344c753e32b06f302cc0e39
Can IMHO be added in a later patch.
Sure?
can_create() creates all CAN sockets, right?
With this patch the sock_prot_inuse_add() increases the counter by one
for every CAN socket.
But only RAW/BCM/ISOTP are decrementing the prot-in-use counter when
removing the socket.
IMO this patch introduces a bug.
Best regards,
Oliver
