On Fri, Jan 28, 2022 at 3:53 AM Eric Dumazet [off-list ref] wrote:

On Thu, Jan 27, 2022 at 6:48 PM Jann Horn [off-list ref] wrote:

quoted

When someone is using NET_DEV_REFCNT_TRACKER for slow debugging, they
should also be able to take the performance hit of
CONFIG_PCPU_DEV_REFCNT and rely on the normal increment-from-zero
detection of the generic refcount code, right? (Maybe
NET_DEV_REFCNT_TRACKER should depend on !CONFIG_PCPU_DEV_REFCNT?)

NET_DEV_REFCNT_TRACKER is not slow, I think it has neglectable cost really.
(I could not see any difference in my tests)

Ah, sorry. I misread the Kconfig text.