Re: [PATCH net-next] ppp: Replace per-CPU recursion counter with lock-owner field

On 6/27/25 12:50 PM, Sebastian Andrzej Siewior wrote:

The per-CPU variable ppp::xmit_recursion is protecting against recursion
due to wrong configuration of the ppp channels. The per-CPU variable
relies on disabled BH for its locking. Without per-CPU locking in
local_bh_disable() on PREEMPT_RT this data structure requires explicit
locking.

The ppp::xmit_recursion is used as a per-CPU boolean. The counter is
checked early in the send routing and the transmit path is only entered
if the counter is zero. Then the counter is incremented to avoid
recursion. It used to detect recursion on channel::downl and
ppp::wlock.

Replace the per-CPU ppp:xmit_recursion counter with an explicit owner
field for both structs.
pch_downl_lock() is helper to check for recursion on channel::downl and
either assign the owner field if there is no recursion.
__ppp_channel_push() is moved into ppp_channel_push() and gets the
recursion check unconditionally because it is based on the lock now.
The recursion check in ppp_xmit_process() is based on ppp::wlock which
is acquired by ppp_xmit_lock(). The locking is moved from
__ppp_xmit_process() into ppp_xmit_lock() to check the owner, lock and
then assign the owner in one spot.
The local_bh_disable() in ppp_xmit_lock() can be removed because
ppp_xmit_lock() disables BH as part of the locking.

Cc: Gao Feng <redacted>
Cc: Guillaume Nault <redacted>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>

Is there any special reason to not use local_lock here? I find this
patch quite hard to read and follow, as opposed to the local_lock usage
pattern. Also the fact that the code change does not affect RT enabled
build only is IMHO a negative thing.

/P