On 12/5/2017 12:03 PM, Theodore Ts'o wrote:
There are a ton of false lockdep splats that were introduced with commit 2dcd5adfb740: "locking/lockdep: Remove the BROKEN flag from CONFIG_LOCKDEP_CROSSRELEASE and CONFIG_LOCKDEP_COMPLETIONS". I'm seeing this triggered by a number of xfstests tests, including shared/298. I see there was some discussion about this here: https://www.spinics.net/lists/linux-xfs/msg10832.html ... but I don't see any solution yet. Is anyone working on this?
Hello, I believe that the commit e319e1fbd9d42 "block, locking/lockdep: Assign a lock_class per gendisk used for wait_for_completion()" solved the false positive. Could you tell me if it doesn't handle it, with the report? Then, I will follow up and try to solve it.
Failing that can we please make the completion lockdep something which is optional since the false positives are incredibly !@#@!?! annoying? That way I can still use most of lockdep, instead of being forced to completely disable lockdep? Thanks, - Ted
-- Thanks, Byungchul
