[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20180214083213.GS25181@hirez.programming.kicks-ass.net>
Date: Wed, 14 Feb 2018 09:32:13 +0100
From: Peter Zijlstra <peterz@...radead.org>
To: Sodagudi Prasad <psodagud@...eaurora.org>
Cc: Raghavendra Rao Ananta <rananta@...eaurora.org>, mingo@...hat.com,
acme@...nel.org, alexander.shishkin@...ux.intel.com,
jolsa@...hat.com, namhyung@...nel.org,
linux-kernel@...r.kernel.org, tsoni@...eaurora.org
Subject: Re: [PATCH] perf: Add support for creating offline events
On Tue, Feb 13, 2018 at 02:17:07PM -0800, Sodagudi Prasad wrote:
> > This is horrible.. and you seem to have forgotten to explain why you
> > care about offline CPUs.
>
> Up to 4.9 kernel, drivers can register cpu hotplug notfiters and drivers are
> able to create perf events dynamically based cpu notifies callback events.
> As cpu hot plug is converted to state machine approach from hot plug
> notifiers, every driver need to define a cpuhp state and registers with cpu
> hotplug state machine for creating perf events dynamically.
>
> Qualcomm have use cases to monitor the cpu cycles and other hw events
> continuously on all cpus from kernel and profiling tools.
> So we are thinking that there could be other soc vendors, who are interested
> in perf events preserving across cpu hot plug and perf events creation on
> hot plugged cores.
But _why_ are you hotplugging to begin with? Just don't do that.
Powered by blists - more mailing lists