linux-kernel - Re: Re: [PATCH] media: vsp1: Fix runtime PM imbalance in vsp1

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: <20200608031134.GO22208@pendragon.ideasonboard.com>
Date:   Mon, 8 Jun 2020 06:11:34 +0300
From:   Laurent Pinchart <laurent.pinchart@...asonboard.com>
To:     dinghao.liu@....edu.cn
Cc:     kjlu@....edu,
        Kieran Bingham <kieran.bingham+renesas@...asonboard.com>,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        linux-media@...r.kernel.org, linux-renesas-soc@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: Re: [PATCH] media: vsp1: Fix runtime PM imbalance in vsp1_probe

Hi Dianghao,

On Mon, Jun 08, 2020 at 11:03:26AM +0800, dinghao.liu@....edu.cn wrote:
> Hi Laurent,
> 
> > > I wonder how many bugs we have today, and how many bugs will keep
> > > appearing in the future, due to this historical design mistake :-( 
> 
> Good question. It's hard to say if this is a design mistake (some use
> of this API does not check its return value and expects it always to
> increment the usage counter). But it does make developers misuse it easier.
> 
> > > This change looks good to me, but we also need a similar change in the
> > > vsp1_device_get() function if I'm not mistaken. Could you combine both
> > > in the same patch ?
> 
> Thank you for your advice! I think you are right and I will fix this in the
> next version of patch. 
> 
> > And actually, after fixing vsp1_device_get(), we should replace the
> > pm_runtime_get_sync() call here with vsp1_device_get(), and the
> > pm_runtime_put_sync() below with vsp1_device_put(), so there would be no
> > need to call pm_runtime_put_sync() manually in the error path here.
> 
> The parameter type of vsp1_device_get() and vsp1_device_put() is "struct 
> vsp1_device". If we want to use these two wrappers, we need to adjust their 
> parameter type to "struct platform_device" or "struct device", which may 
> lead to errors in other callers. Maybe we should leave it as it is.

The vsp1_probe() function has a struct vsp1_device whose dev field is
populated by the time it needs to call pm_runtime_get_sync() and
pm_runtime_get_put(), so I think you can use vsp1_device_get() and
vsp1_device_put() as drop-in replacements without changing the
parameters to these two functions.

-- 
Regards,

Laurent Pinchart