Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <1537887875-29494-1-git-send-email-vbadigan@codeaurora.org>
 <1537887875-29494-3-git-send-email-vbadigan@codeaurora.org>
 <CAD=FV=VxLR6v3vyimjznrqa1Tj5rPziPQX7Pt+wJRR+xm1N-0A@mail.gmail.com> <e6f0c8c6-cd0f-33b9-30da-ab09953449ca@codeaurora.org>
In-Reply-To: <e6f0c8c6-cd0f-33b9-30da-ab09953449ca@codeaurora.org>
From:   Doug Anderson <dianders@google.com>
Date:   Thu, 27 Sep 2018 08:21:03 -0700
Message-ID: <CAD=FV=ULtdGkpRfmAZRbM0VQu6sJ37_bPBRHKFReK4X+jYjT_Q@mail.gmail.com>
Subject: Re: [PATCH 2/2] mmc: sdhci-msm: Re-initialize DLL if MCLK is gated dynamically
To:     vbadigan@codeaurora.org
Cc:     Adrian Hunter <adrian.hunter@intel.com>,
        Ulf Hansson <ulf.hansson@linaro.org>,
        Rob Herring <robh+dt@kernel.org>,
        Asutosh Das <asutoshd@codeaurora.org>, riteshh@codeaurora.org,
        stummala@codeaurora.org, sayalil@codeaurora.org,
        Evan Green <evgreen@chromium.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi,
On Thu, Sep 27, 2018 at 4:00 AM Veerabhadrarao Badiganti
<vbadigan@codeaurora.org> wrote:
>
> Hi Doug,
>
>
> On 9/26/2018 3:58 AM, Doug Anderson wrote:
> > Hi,
> >
> > On Tue, Sep 25, 2018 at 8:05 AM Veerabhadrarao Badiganti
> > <vbadigan@codeaurora.org> wrote:
> >> +       /*
> >> +        * Whenever core-clock is gated dynamically, it's needed to
> >> +        * re-initialize the DLL when the clock is ungated.
> >> +        */
> >> +       if (msm_host->restore_dll_cfg_needed && msm_host->clk_rate) {
> >> +               msm_host->restore_dll_cfg = true;
> >> +               mmc_retune_needed(host->mmc);
> > Using the boolean "restore_dll_cfg" to communicate like this seems
> > really fragile.  I have no basis in fact, but I worry that something
> > will happen in the meantime that really ought to invalidate the
> > "saved_tuning_phase" but the boolean will still be set.
> >
> > Is there a reason you can't just call msm_config_cm_dll_phase()
> > directly from sdhci_msm_runtime_resume()?  Perhaps after the
> > clk_bulk_prepare_enable() call below?
>
> I can do that.
>
> But we don't need to restore the clock phase for all speed modes since
> every mode
> doesn't use DLL. Only few speed modes (only SDR104 mode for SD card)
> would need to restore DLL setting. So I would need to do some extra checks
> for performing this operation for only required speed modes.
> And we should re-initialize the DLL first before programming the phase.
>
> I wanted to reuse the existing logic which does all these things. So did
> this way.

Personally I'd rather see the extra checks or even restore the DLL
settings some cases when it wasn't truly needed.

In theory, though, I'd expect that there's a hardware register
somewhere that says whether the DLL settings matter or not.  Can't you
just key off of that?  ...or if that register is also lost when the
clock turns off then whenever you change that register to turn the DLL
off then clear your saved phase?


-Doug