[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200205094838.GI14879@hirez.programming.kicks-ass.net>
Date: Wed, 5 Feb 2020 10:48:38 +0100
From: Peter Zijlstra <peterz@...radead.org>
To: Waiman Long <longman@...hat.com>
Cc: Ingo Molnar <mingo@...hat.com>, Will Deacon <will.deacon@....com>,
linux-kernel@...r.kernel.org, Bart Van Assche <bvanassche@....org>
Subject: Re: [PATCH v5 6/7] locking/lockdep: Reuse freed chain_hlocks entries
On Tue, Feb 04, 2020 at 11:57:09AM -0500, Waiman Long wrote:
> Wait, it is possible that we can have deadlock like this:
>
> cpu 0 cpu 1
> ----- -----
> lock A lock B
> <irq> <irq>
> lock B lock A
>
> If we eliminate 1-entry chain, will that impact our ability to detect this
> kind of deadlock?
I'm thinking that should trigger irq-inversion (irq-on vs in-irq) on
either A or B (depending on timing).
AFAICT the irq-state tracking is outside of validate_chain().
This is also why I think your separate_irq_context() change is not
needed.
validate_chain() really only checks the per-context lock nesting, and
there, a single lock doesn't do very much. Hence my proposed patch.
Powered by blists - more mailing lists