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;
Cc:     shawn.lin@rock-chips.com, Jaehoon Chung <jh80.chung@samsung.com>,
        Ulf Hansson <ulf.hansson@linaro.org>,
        linux-mmc@vger.kernel.org, linux-kernel@vger.kernel.org,
        Zhangfei Gao <zhangfei.gao@linaro.org>
Subject: Re: [PATCH] mmc: dw_mmc-k3: Fix DDR52 mode by setting required clock
 divisor
To:     Ryan Grachek <ryan@edited.us>
References: <20180329182423.21201-1-ryan@edited.us>
 <e3ec3a77-f2f3-2164-4dbb-80e1ba3ebc63@rock-chips.com>
 <CAGoBSJq+tSpvFZsJLPffTaFBg5o2Nn2eeXXvHqt4Us_dd7==1w@mail.gmail.com>
From:   Shawn Lin <shawn.lin@rock-chips.com>
Message-ID: <bed16723-6bf0-4bcf-f44c-ebcf7a1e7cc5@rock-chips.com>
Date:   Thu, 5 Apr 2018 08:51:14 +0800
User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:52.0) Gecko/20100101
 Thunderbird/52.6.0
MIME-Version: 1.0
In-Reply-To: <CAGoBSJq+tSpvFZsJLPffTaFBg5o2Nn2eeXXvHqt4Us_dd7==1w@mail.gmail.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

[+ Zhangfei Gao who added support for hi6220]

On 2018/4/4 23:31, Ryan Grachek wrote:
> On Tue, Apr 3, 2018 at 6:31 AM, Shawn Lin <shawn.lin@rock-chips.com 
> <mailto:shawn.lin@rock-chips.com>> wrote:
> 
>     On 2018/3/30 2:24, oscardagrach wrote:
> 
>     Need at least one line commit body.
> 
>         Signed-off-by: oscardagrach <ryan@edited.us <mailto:ryan@edited.us>>
>         ---
>            drivers/mmc/host/dw_mmc-k3.c | 10 ++++++++--
>            1 file changed, 8 insertions(+), 2 deletions(-)
> 
>         diff --git a/drivers/mmc/host/dw_mmc-k3.c
>         b/drivers/mmc/host/dw_mmc-k3.c
>         index 89cdb3d533bb..efc546cb4db8 100644
>         --- a/drivers/mmc/host/dw_mmc-k3.c
>         +++ b/drivers/mmc/host/dw_mmc-k3.c
>         @@ -194,8 +194,14 @@ static void dw_mci_hi6220_set_ios(struct
>         dw_mci *host, struct mmc_ios *ios)
>                  int ret;
>                  unsigned int clock;
>            -     clock = (ios->clock <= 25000000) ? 25000000 : ios->clock;
>         -
>         +       /* CLKDIV must be 1 for DDR52/8-bit mode */
>         +       if (ios->bus_width == MMC_BUS_WIDTH_8 &&
>         +               ios->timing == MMC_TIMING_MMC_DDR52) {
>         +               mci_writel(host, CLKDIV, 0x1);
>         +               clock = ios->clock;
>         +       } else {
>         +               clock = (ios->clock <= 25000000) ? 25000000 :
>         ios->clock;
>         +       }
> 
> 
>     I undertand DDR52/8-bit need CLKDIV fixed 1, but shouldn't the following
>     change is more sensible?
> 
>     if (ios->bus_width == MMC_BUS_WIDTH_8 && ios->timing ==
>     MMC_TIMING_MMC_DDR52)
>              clock = ios->clock * 2;
>     else
>              clock = (ios->clock <= 25000000) ? 25000000 : ios->clock;
> 
> 
>     The reason is ios->clock is 52MHz and you could claim 104MHz from the
>     clock provider and let dw_mmc core take care of the divder to be 1.
>     Otherwise, you just force it to be DDR52/8-bit with a clk rate of 26MHz.
> 
> 
>                  ret = clk_set_rate(host->biu_clk, clock);
>                  if (ret)
>                          dev_warn(host->dev, "failed to set rate
>         %uHz\n", clock);
> 
> 
> 

For future wise, please use plain mode mail, but not HTML format.

> Your feedback is correct. I see the Rockchip dwmmc driver has a similar
> implementation. After applying your suggested changes, however, my board
> reports "dwmmc_k3 f723d000.dwmmc0: failed to set rate 104000000Hz"
> during intialization of eMMC. In addition, I do not see CLKDIV being
> set to 1. clk_set_rate fails and I wonder if this is out-of-scope of
> the driver.
> 
> If I set CLKDIV where I did prior, with your changes, the device fails
> to set the clock and falls back to 52 MHz (26 MHz) and works fine, but
> again, CLKDIV is reported as 0 (even though it is 1.) One thing of
> interest to note is when I manually set the clock by doing:
> (echo 104000000 > /sys/kernel/debug/mmc0/clock) the device reports back
> 'mmc_host mmc0: Bus speed (slot 0) = 198400000Hz (slot req 104000000Hz,
>   actual 99200000HZ div = 1)' which works reliably and clk_set_rate does
> not report any error.
> 

When looking closely into the code, at least dw_mci_hi6220_set_ios
goes wrong with the bus_hz, since it should be ciu_clk but not biu_clk.
"b" stands for bus, and "c" stands for card IMHO, however bus_hz
describs the clock to the card, provided by controller. Does the
following patch help?


diff --git a/drivers/mmc/host/dw_mmc-k3.c b/drivers/mmc/host/dw_mmc-k3.c
index 89cdb3d..9e78cf2 100644
--- a/drivers/mmc/host/dw_mmc-k3.c
+++ b/drivers/mmc/host/dw_mmc-k3.c
@@ -194,13 +194,21 @@ static void dw_mci_hi6220_set_ios(struct dw_mci 
*host, struct mmc_ios *ios)
         int ret;
         unsigned int clock;

-       clock = (ios->clock <= 25000000) ? 25000000 : ios->clock;
+       if (ios->bus_width == MMC_BUS_WIDTH_8 &&
+           ios->timing == MMC_TIMING_MMC_DDR52)
+               clock = ios->clock * 2;
+       else
+               clock = (ios->clock <= 25000000) ? 25000000 : ios->clock;

-       ret = clk_set_rate(host->biu_clk, clock);
+       ret = clk_set_rate(host->ciu_clk, clock);
         if (ret)
                 dev_warn(host->dev, "failed to set rate %uHz\n", clock);

-       host->bus_hz = clk_get_rate(host->biu_clk);
+       clock = clk_get_rate(host->ciu_clk);
+       if (clock != host->bus_hz) {
+               host->bus_hz = clock;
+               host->current_speed = 0;
+       }
  }


> I am not sure where to begin debugging these clock issues and welcome
> any feedback.