linux-kernel - Re: [PATCH v5 12/22] media: camss: Remove per VFE power domain toggling

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 for Android: free password hash cracker in your pocket

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

Message-ID: <CAG3jFysEkRNXsaXudXd9msLPU3RiFcPpAgS6ecbOjeVfdwpb4A@mail.gmail.com>
Date:   Wed, 24 Feb 2021 15:53:34 +0100
From:   Robert Foss <robert.foss@...aro.org>
To:     Andrey Konovalov <andrey.konovalov@...aro.org>
Cc:     Andy Gross <agross@...nel.org>,
        Bjorn Andersson <bjorn.andersson@...aro.org>,
        Todor Tomov <todor.too@...il.com>,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        Rob Herring <robh+dt@...nel.org>,
        angelogioacchino.delregno@...ainline.org,
        MSM <linux-arm-msm@...r.kernel.org>,
        linux-media <linux-media@...r.kernel.org>,
        "open list:OPEN FIRMWARE AND FLATTENED DEVICE TREE BINDINGS" 
        <devicetree@...r.kernel.org>,
        linux-kernel <linux-kernel@...r.kernel.org>,
        AngeloGioacchino Del Regno <kholk11@...il.com>,
        Sakari Ailus <sakari.ailus@....fi>,
        Nicolas Boichat <drinkcat@...omium.org>,
        Rob Herring <robh@...nel.org>,
        Tomasz Figa <tfiga@...omium.org>,
        Azam Sadiq Pasha Kapatrala Syed <akapatra@...cinc.com>,
        Sarvesh Sridutt <Sarvesh.Sridutt@...rtwirelesscompute.com>,
        Laurent Pinchart <laurent.pinchart@...asonboard.com>,
        Jonathan Marek <jonathan@...ek.ca>
Subject: Re: [PATCH v5 12/22] media: camss: Remove per VFE power domain toggling

Hey Andrey,

>
> On 17.02.2021 14:21, Robert Foss wrote:
> > For Titan ISPs clocks fail to re-enable during vfe_get()
> > after any vfe has been halted and its corresponding power
> > domain power has been detached.
>
> OK.
>
> > Since all of the clocks depend on all of the PDs, per
> > VFE PD detaching is no option for this generation of HW.
>
> But this patch removes camss_pm_domain_on/off calls from
> vfe_get/put() for all the SOCs, not only for sdm845.
> And this looks like a regression (higher power consumption)
> for all the generation1 devices.

Yeah, that is a serious problem with the approach I picked here. The
power difference shouldn't be huge however, since the best case
scenario savings of the previous implementation was being able to
power down 1 VFE when the other one is working. If none of the VFEs is
working, nothing is powered up both in the previous implementation &
using this patch.

>
> Is it possible to handle gen1 and gen2 hardware differently,
> so that gen1 continued to use camss_pm_domain_on/off as
> before?

I hesitated going down this gen1/gen2 split here, due to how deep into
the common code some of this functionality is. Let me have another
look at this though, not having a power regression for gen1 devices
would definitely be preferable.