[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20231124232932.7kgjwi6s2h5oo4a2@moria.home.lan>
Date: Fri, 24 Nov 2023 18:29:32 -0500
From: Kent Overstreet <kent.overstreet@...ux.dev>
To: Peter Zijlstra <peterz@...radead.org>
Cc: linux-kernel@...r.kernel.org, linux-bcachefs@...r.kernel.org,
boqun.feng@...il.com, longman@...hat.com, will@...nel.org,
mingo@...hat.com
Subject: Re: [PATCH 2/6] locking/lockdep: lockdep_set_no_check_recursion()
On Fri, Nov 24, 2023 at 10:58:04AM +0100, Peter Zijlstra wrote:
> On Wed, Nov 22, 2023 at 06:51:09PM -0500, Kent Overstreet wrote:
> > This adds a method to tell lockdep not to check lock ordering within a
> > lock class - but to still check lock ordering w.r.t. other lock types.
> >
> > This is for bcachefs, where for btree node locks we have our own
> > deadlock avoidance strategy w.r.t. other btree node locks (cycle
> > detection), but we still want lockdep to check lock ordering w.r.t.
> > other lock types.
>
> So earlier you added custom sort order.
That was never merged? I've been meaning to revisit that though and
convert some other things to it, the original patchset just converted
bcache, not bcachefs.
But it's not applicable here, bcachefs doesn't have lock ordering, it
has cycle detection.
> Additionally there is the wound-wait mutexes that also have semantics
> similar to what you describe.
Yeah, we talked about that as well. Wait/wound is similar in principle
but aborts much more frequently, since it's just comparing transaction
start time and not doing full cycle detection.
Powered by blists - more mailing lists