Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   "Rafael J. Wysocki" <rjw@rjwysocki.net>
To:     Linux PM <linux-pm@vger.kernel.org>
Cc:     LKML <linux-kernel@vger.kernel.org>,
        Frederic Weisbecker <frederic@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        "Zhou Ti (x2019cwm)" <x2019cwm@stfx.ca>
Subject: [PATCH v1 0/5] cpuidle: Take possible negative "sleep length" values into account
Date:   Mon, 29 Mar 2021 20:12:39 +0200
Message-ID: <2764850.e9J7NaK4W3@kreacher>
MIME-Version: 1.0
Content-Transfer-Encoding: 7Bit
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi All,

As follows from the discussion triggered by the patch at

https://lore.kernel.org/lkml/20210311123708.23501-2-frederic@kernel.org/

the cpuidle governors using tick_nohz_get_sleep_length() assume it to always
return positive values which is not correct in general.

To address this issues, first document the fact that negative values can
be returned by tick_nohz_get_sleep_length() (patch [1/5]).  Then, in
preparation for more substantial changes, change the data type of two
fields in struct cpuidle_state to s64 so they can be used in computations
involving negative numbers safely (patch [2/5]).

Next, adjust the teo governor a bit so that negative "sleep length" values
are counted like zero by it (patch [3/5]) and modify it so as to avoid
mishandling negative "sleep length" values (patch [4/5]).

Finally, make the menu governor take negative "sleep length" values into
account properly (patch [5/5]).

Please see the changelogs of the patches for details.

Thanks!