linux-kernel - Re: [PATCH] power: Print wakeup_count instead of event

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAOGSYL0ga-H=0a5EEDmOWSZjnmG1Qc=9AxAbM_-ifHDj_FbREg@mail.gmail.com>
Date:   Sun, 3 Jun 2018 10:14:09 -0700
From:   Ravi Chandra Sadineni <ravisadineni@...gle.com>
To:     "Rafael J. Wysocki" <rafael@...nel.org>
Cc:     Ravi Chandra Sadineni <ravisadineni@...omium.org>,
        "Rafael J. Wysocki" <rjw@...ysocki.net>, chenhong3@...wei.com,
        Pavel Machek <pavel@....cz>, Dmitry Torokhov <dtor@...gle.com>,
        Len Brown <len.brown@...el.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Todd Broch <tbroch@...gle.com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Linux PM <linux-pm@...r.kernel.org>,
        Rajat Jain <rajatja@...gle.com>,
        Benson Leung <bleung@...gle.com>
Subject: Re: [PATCH] power: Print wakeup_count instead of event_count in the
 sysfs attribute.

Hi Rafael,

On Sun, Jun 3, 2018 at 1:05 AM, Rafael J. Wysocki <rafael@...nel.org> wrote:
> On Sat, Jun 2, 2018 at 4:32 AM, Ravi Chandra Sadineni
> <ravisadineni@...omium.org> wrote:
>> Currently we show event_count instead of wakeup_count as part of per
>> device wakeup_count sysfs attribute. Change it to wakeup_count to make
>> it more meaningful.
>
> More information, please.
>
> In particular, why it is more meaningful.
Wakeup_count increments only when events_check_enabled is set. This
bool is set whenever we write current wakeup count to
/sys/power/wakeup_count from the user land. Also events_check_enabled
is cleared on every resume. My understanding is that, userland is
expected to write to this just before suspend. This way
pm_wakeup_event() when called from irqs will increment the
wakeup_count only if we are in system wide suspend resume cycle and
should give a fair approximation of how many times a device might have
caused a wake from S3/S0iX.  event_count on the other hand will
increment every time pm_wakeup_event() is called irrespective of
whether we are in a suspend/resume cycle.  For example when I try
doing something like this (https://lkml.org/lkml/2018/6/1/890), we see
the wakeup_count sysfs attribute for the particular device
incrementing every time there is a irq. If it is important to expose
event_count via sysfs attribute, should we create another attribute ?
Also we do expose each of these counters via
debugfs(/sys/kernel/debug/wake_sources).

Please correct me if I am wrong or missing something. Also if there is
a better way to do this, please let me know.
>
> Thanks,
> Rafael