Received-SPF: pass (google.com: domain of linux-kernel+bounces-108903-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Precedence: bulk
MIME-Version: 1.0
From: Maxim Kiselev <bigunclemax@gmail.com>
Date: Wed, 20 Mar 2024 15:28:19 +0300
Message-ID: <CALHCpMhc1F5Ue7U_gsDXREHUZRVQJNYRCJxYxoNqbN=-39jf7A@mail.gmail.com>
Subject: Re: [PATCH v8 4/4] riscv: dts: thead: Enable LicheePi 4A eMMC and microSD
To: xry111@xry111.site
Cc: Albert Ou <aou@eecs.berkeley.edu>, Conor Dooley <conor@kernel.org>, devicetree@vger.kernel.org, 
	dfustini@baylibre.com, guoren@kernel.org, jkridner@beagleboard.org, 
	jszhang@kernel.org, Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>, 
	open list <linux-kernel@vger.kernel.org>, linux-riscv@lists.infradead.org, 
	Palmer Dabbelt <palmer@dabbelt.com>, Paul Walmsley <paul.walmsley@sifive.com>, 
	robertcnelson@beagleboard.org, Rob Herring <robh+dt@kernel.org>, wefu@redhat.com, 
	Maksim Kiselev <bigunclemax@gmail.com>
Content-Type: text/plain; charset="UTF-8"

Hi Xi, Drew

I have the same problem with SD on my LicheePi 4A.

After some investigations I found how to fix this tuning error.
Here is the patch that increases tuning loop count from
40(MAX_TUNING_LOOP at sdhci.c) to 128.

diff --git a/drivers/mmc/host/sdhci-of-dwcmshc.c
b/drivers/mmc/host/sdhci-of-dwcmshc.c
index 8d6cfb648096..da8f5820fb69 100644
--- a/drivers/mmc/host/sdhci-of-dwcmshc.c
+++ b/drivers/mmc/host/sdhci-of-dwcmshc.c
@@ -706,6 +706,7 @@ static int th1520_execute_tuning(struct sdhci_host
*host, u32 opcode)

        /* perform tuning */
        sdhci_start_tuning(host);
+       host->tuning_loop_count = 128:
        host->tuning_err = __sdhci_execute_tuning(host, opcode);
        if (host->tuning_err) {
                /* disable auto-tuning upon tuning error */

After that change tuning works fine. The same value of loop count is
used in RevyOS BSP
https://github.com/revyos/thead-kernel/blob/c6d4e5df18a17903d012ffd89e67d0ee5ce6cf2d/drivers/mmc/host/sdhci-of-dwcmshc.c#L185

Honestly, it looks a little bit strange for me.

It seems that the tuning algorithm requires to move through
all the taps of delay line(128 taps?) even if we use THRESHOLD_MODE
instend LARGEST_WIN_MODE (I mean bit 2 in AT_CTRL_R(0x540) register).

Xi, could you also test my fix on your board?

Best regards,
Maksim