linux-kernel - Re: [PATCH v3 1/5] pwms: pwm-ti*: Remove support for local clock gating

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:	Mon, 29 Feb 2016 15:20:12 -0800
From:	Tony Lindgren <tony@...mide.com>
To:	"Franklin S Cooper Jr." <fcooper@...com>
Cc:	paul@...an.com, t-kristo@...com, vigneshr@...com,
	linux-pwm@...r.kernel.org, devicetree@...r.kernel.org,
	linux-kernel@...r.kernel.org, linux-omap@...r.kernel.org,
	linux-arm-kernel@...ts.infradead.org, linux-clk@...r.kernel.org
Subject: Re: [PATCH v3 1/5] pwms: pwm-ti*: Remove support for local clock
 gating

* Franklin S Cooper Jr. <fcooper@...com> [160229 15:12]:
> 
> 
> On 02/29/2016 04:55 PM, Tony Lindgren wrote:
> > * Franklin S Cooper Jr. <fcooper@...com> [160229 14:31]:
> >> On 02/29/2016 04:04 PM, Tony Lindgren wrote:
> >>> Hmm but why are you also removing the pm_runtime calls? Those
> >>> actually do take care of gating the clocks via the interconnect
> >>> level code that is hwmod in this case.
> >> I removed all PM runtime calls that revolved around
> >> pwmss_submodule_state_change. Originally the driver would do
> >> a pm_runtime_get_sync then call pwmss_submodule_state_change
> >> and then immediately call pm_runtime_put_sync. Without
> >> pwmss_submodule_state_change those calls would be
> >> meaningless.  I also removed pm_runtime calls in error paths
> >> that no longer existed.
> > Typically the interconnect level code can gate the clkctrl bit
> > for the module with PM runtime even with no other driver specific
> > registers. If you remove the pm_runtime calls, that does not
> > happen.
> 
> So the clocks should be unlocked when ever the IP registers are
> being read/written or if the peripheral is being used for
> example
> the pwm signal is being generated. All these cases are already
> being handled.
> 
> Using ecap driver as an example.
> 
> Pm_runtime_get_sync is called within ecap_pwm_enable when
> the pwm signal is to be generated. Pm_runtime_put_sync is called
> when the pwm signal is to be stopped.
> 
> When either the pwm signal polarity is set or pwm
> configuration is made
> then a pm_runtime_get_sync and pm_runtime_put_sync are
> called within
> the same function surrounding calls to the IP's registers.
> 
> Probe is calling pm_runtime_enable while remove is calling
> pm_runtime_disable.

OK good to hear you have considered this. The above answers
my questions then thanks.

> So the correct pm_runtime calls are being made from what I
> can see.
> I'm not sure I understand the concern since removing those
> calls aren't
> creating any kind of imbalance.

OK thanks for checking.

> If I'm not addressing your concern please give me an example
> of where
> you see a possible issue.

No that's fine. I thought you're ripping out all of the the
pm_runtime based on just looking at the patch :)

> > Also, how do you know this change does not affect the other
> > SoC variants using the same driver?
> 
> I've tested these changes on AM335x GP and AM437x GP evms.
> AM335x
> and AM437x were the only other users of this driver. Sorry 
> I should of
> documented this in my cover-letter.

OK good to hear.

Thanks,

Tony