On 2018-12-18 15:07:45 [+0000], Catalin Marinas wrote: … > It may be worth running some performance/latency tests during kmemleak > scanning (echo scan > /sys/kernel/debug/kmemleak) but at a quick look, > I don't think we'd see any difference with a raw_spin_lock_t. > > With a bit more thinking (though I'll be off until the new year), we > could probably get rid of the kmemleak_lock entirely in scan_block() and > make lookup_object() and the related rbtree code in kmemleak RCU-safe. Okay. So let me apply that patch into my RT tree with your ack (from the other email). And then I hope that it either shows up upstream or gets replaced with RCU in the ende :) Thanks. Sebastian
