[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <62c553f2-2fb3-4b83-86ae-5b038941cdbb@redhat.com>
Date: Sun, 30 Jun 2024 23:16:36 -0400
From: Waiman Long <longman@...hat.com>
To: Kent Overstreet <kent.overstreet@...ux.dev>
Cc: linux-kernel@...r.kernel.org, linux-bcachefs@...r.kernel.org,
Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>,
Will Deacon <will@...nel.org>, Boqun Feng <boqun.feng@...il.com>
Subject: Re: [PATCH] lockdep: lockdep_set_notrack_class()
On 6/30/24 20:10, Kent Overstreet wrote:
> On Sun, Jun 30, 2024 at 06:08:21PM GMT, Waiman Long wrote:
>> On 6/30/24 01:10, Kent Overstreet wrote:
>>> Add a new helper to disable lockdep tracking entirely for a given class.
>>>
>>> This is needed for bcachefs, which takes too many btree node locks for
>>> lockdep to track. Instead, we have a single lockdep_map for "btree_trans
>>> has any btree nodes locked", which makes more since given that we have
>>> centralized lock management and a cycle detector.
>> Could you explain a bit more what the current novalidate_class is lacking
>> WRT to the bcachefs lock? Is it excessive performance overhead or some bogus
>> lockdep warning?
> novalidate just switches off checking of lock ordering, but the fact
> that the locks are held is still tracked.
>
> bcachefs takes more btree node locks than lockdep can track, so I'm
> switching to a single lockdep map for "this btree_trans has any btree
> nodes locked" instead.
I asked because you are adding a new special class that is similar to
no_validate in some way but also a bit different. So we need to document
what each of these special classes are for to avoid confusion.
Cheers,
Longman
Powered by blists - more mailing lists