On Fri, Jan 8, 2021 at 4:59 AM Saeed Mahameed [off-list ref] wrote:
Eric, about two years ago you were totally against sleeping in
ndo_get_stats, what happened ? :)
https://lore.kernel.org/netdev/4cc44e85-cb5e-502c-30f3-c6ea564fe9ac@gmail.com/ (local)
My approach to solve this was much simpler and didn't require a new
mutex nor RTNL lock, all i did is to reduce the rcu critical section to
not include the call to the driver by simply holding the netdev via
dev_hold()
Yeah, and how have you dealt with bonding at that time ?
Look, it seems to me Vladimir's work is more polished.
If you disagree, repost a rebased patch series so that we can
test/compare and choose the best solution.
And make sure to test it with LOCKDEP enabled ;)
Thanks.