Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Sat, 24 Apr 2021 20:16:52 +1200
From:   Daniel Beer <dlbeer@gmail.com>
To:     ulf.hansson@linaro.org
Cc:     linux-mmc@vger.kernel.org, linux-kernel@vger.kernel.org,
        Adrian Hunter <adrian.hunter@intel.com>,
        Ben Chuang <ben.chuang@genesyslogic.com.tw>,
        Daniel Beer <dlbeer@gmail.com>
Subject: [PATCH] mmc: sdhci-pci-gli: increase 1.8V regulator wait
Message-ID: <20210424081652.GA16047@nyquist.nev>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Precedence: bulk

Inserting an SD-card on an Intel NUC10i3FNK4 (which contains a GL9755)
results in the message:

    mmc0: 1.8V regulator output did not become stable

Following this message, some cards work (sometimes), but most cards fail
with EILSEQ. This behaviour is observed on Debian 10 running kernel
4.19.188, but also with 5.8.18 and 5.11.15.

The driver currently waits 5ms after switching on the 1.8V regulator for
it to become stable. Increasing this to 10ms gets rid of the warning
about stability, but most cards still fail. Increasing it to 20ms gets
some cards working (a 32GB Samsung micro SD works, a 128GB ADATA
doesn't). At 50ms, the ADATA works most of the time, and at 100ms both
cards work reliably.

Signed-off-by: Daniel Beer <dlbeer@gmail.com>
---
 drivers/mmc/host/sdhci-pci-gli.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/mmc/host/sdhci-pci-gli.c b/drivers/mmc/host/sdhci-pci-gli.c
index 592d79082f58..061618aa247f 100644
--- a/drivers/mmc/host/sdhci-pci-gli.c
+++ b/drivers/mmc/host/sdhci-pci-gli.c
@@ -627,8 +627,13 @@ static void sdhci_gli_voltage_switch(struct sdhci_host *host)
 	 *
 	 * Wait 5ms after set 1.8V signal enable in Host Control 2 register
 	 * to ensure 1.8V signal enable bit is set by GL9750/GL9755.
+	 *
+	 * ...however, the controller in the NUC10i3FNK4 (a 9755) requires
+	 * slightly longer than 5ms before the control register reports that
+	 * 1.8V is ready, and far longer still before the card will actually
+	 * work reliably.
 	 */
-	usleep_range(5000, 5500);
+	usleep_range(100000, 110000);
 }
 
 static void sdhci_gl9750_reset(struct sdhci_host *host, u8 mask)
-- 
2.20.1