linux-kernel - Re: [PATCH] opp: Drop unnecessary check frmo dev_pm_opp_attach

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]

Date:   Thu, 27 Aug 2020 14:14:49 +0200
From:   Stephan Gerhold <stephan@...hold.net>
To:     Viresh Kumar <viresh.kumar@...aro.org>
Cc:     Viresh Kumar <vireshk@...nel.org>, Nishanth Menon <nm@...com>,
        Stephen Boyd <sboyd@...nel.org>, linux-pm@...r.kernel.org,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Rafael Wysocki <rjw@...ysocki.net>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] opp: Drop unnecessary check frmo
 dev_pm_opp_attach_genpd()

On Thu, Aug 27, 2020 at 03:35:15PM +0530, Viresh Kumar wrote:
> Since commit c0ab9e0812da ("opp: Allocate genpd_virt_devs from
> dev_pm_opp_attach_genpd()"), the allocation of the virtual devices is
> moved to dev_pm_opp_attach_genpd() and this check isn't required anymore
> as it will always fail. Drop it.
> 

Only partially related to this patch, but actually I noticed that
dev_pm_opp_attach_genpd() does not work correctly if it is called
multiple times.

For example, qcom-cpufreq-nvmem calls this for every CPU because it is
not aware that the OPP table is shared between the CPUs.
dev_pm_opp_attach_genpd() does not check if opp_table->genpd_virt_devs
is already set, so when it is called again for other CPUs we will:

  - Cause a memory leak (opp_table->genpd_virt_devs is just replaced
    with new memory)
  - Attach the power domains multiple times
  - Never detach the power domains from earlier calls
  - Crash when dev_pm_opp_detach_genpd() is called the second time

Oh well. :)

I think the function should just return and do nothing if the power
domains were already attached, just like dev_pm_opp_set_supported_hw()
etc. But this is a bit complicated to implement with the "virt_devs"
parameter, since callers will probably assume that to be valid if we
return success.

Another advantage of my proposal to remove the virt_devs parameter [1] :)

Stephan