linux-kernel - Re: [Question] about patch: don't use [delayed_]work

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: <20160901184503.GD12660@htj.duckdns.org>
Date:   Thu, 1 Sep 2016 14:45:03 -0400
From:   Tejun Heo <tj@...nel.org>
To:     qiaozhou <qiaozhou@...micro.com>
Cc:     linux-pm@...r.kernel.org, Wang Wilbur <wilburwang@...micro.com>,
        Wu Gang <gangwu@...micro.com>, linux-kernel@...r.kernel.org
Subject: Re: [Question] about patch: don't use [delayed_]work_pending()

Hello,

On Thu, Sep 01, 2016 at 05:09:36PM +0800, qiaozhou wrote:
> In our system, we do cpu clock init in of_clk_init path, and use pm qos to
> maintain cpu/cci clock. Firstly we init a CCI_CLK_QOS and set a default
> value, then update CCI_CLK_QOS to limit CCI min frequency according to
> current cpu frequency. Before calling pm_qos_update_request, irq is
> disabled, but after the calling, irq is enabled in cancel_delayed_work_sync,
> which causes some inconvenience before Before this patch is applied, it
> checks pending work and won't do cancel_delayed_work_sync in this boot up
> phase.

So, cancel_delayed_work_sync() usually shouldn't be called with irq
disabled as it's a possibly blocking call.

> The simple calling sequence is like this:
> 
> start_kernel -> of_clk_init -> cpu_clk_init -> pm_qos_add_request(xx,
> default_value),
> 
> then pm_qos_update_request.
> 
> I don't know whether it's meaningful to still check pending work here, or
> it's not suggested to use pm_qos_update_request in this early boot up phase.
> Could you help to share some opinions? (I can fix this issue by adding the
> current qos value directly instead of default value, though.)

Hmmm... but I suppose this is super-early in the boot.  Would it make
sense to have a static variable (e.g. bool clk_fully_initailized) to
gate the cancel_delayed_sync() call?

Thanks.

-- 
tejun