[EXT] Re: [PATCH] mmc: Add power stabilization delay in mmc_power_on()

[Luke Wang]

> -----Original Message-----
> From: Marek Vasut <marek.vasut at mailbox.org>
> Sent: Tuesday, December 30, 2025 9:37 PM
> To: Luke Wang <ziniu.wang_1 at nxp.com>; Peng Fan <peng.fan at nxp.com>;
> jh80.chung at samsung.com; trini at konsulko.com
> Cc: marek.vasut+renesas at mailbox.org; bhimeswararao.matsa at gmail.com;
> xypron.glpk at gmx.de; kauschluss at disroot.org; andrew.goodbody at linaro.org;
> tanmay.kathpalia at altera.com; Ye Li <ye.li at nxp.com>; Bough Chen
> <haibo.chen at nxp.com>; u-boot at lists.denx.de
> Subject: [EXT] Re: [PATCH] mmc: Add power stabilization delay in
> mmc_power_on()
> 
> Caution: This is an external email. Please take care when clicking links or
> opening attachments. When in doubt, report the message using the 'Report this
> email' button
> 
> 
> On 12/30/25 11:39 AM, ziniu.wang_1 at nxp.com wrote:
> > From: Luke Wang <ziniu.wang_1 at nxp.com>
> >
> > After enabling vmmc supply, add 10ms delay to allow voltage to
> > stabilize before SD initialization starts. Without this delay,
> > some SD cards fail to initialize because the initial clock cycles
> > are sent while power is still ramping up.
> Details please. Which exact cards ? With which controller/SoC ? Does
> this problem happen on another SoC/controller ?

MMC initialization failed on IMX7ULP-EVK with the SanDisk Ultra
64GB card (CID: 035344534436344785569a5ab5018400). However,
SanDisk Ultra 32GB and 16GB cards did not encounter this issue
(both have CID: 035344534331364780c86726e1012700).

The root cause of this issue is that this SD card requires at
least 74 clock cycles after power-up, otherwise CMD8 will not
receive a response. For other SD cards that initialize normally,
the initial clock cycles are not mandatory - they can still
initialize successfully even without them.

On the IMX7ULP-EVK, the SD power-up time takes approximately 2ms
to reach VDD MIN voltage. However, the current code does not wait
for the voltage to stabilize before sending the initial clock
cycles. This results in the VDD voltage being only around 1V when
the initial clock cycles are sent, causing the SD card to
completely miss these initial clock cycles.

We could add regulator-ramp-delay or startup-delay-us values in
the dts vmmc regulator, but I have checked some dts files and
found that many of them do not configure these two values - not
just for the IMX series, but across different platforms. If we
were to add them in dts, it would require modifying a very large
number of dts files.

Alternatively, we could follow the approach in the Linux kernel's
mmc_power_up() function and wait 10ms after power on.

This card does not encounter issues on the IMX8MM platform because
the IMX8MM's SD power only needs about 0.5ms to stabilize. Sending
1ms of initial clock cycles ensures that the SD card can still
receive at least 74 clock cycles after the voltage has stabilized.
Even though no issue was encountered, sending initial clocks during
the power ramp-up phase still does not comply with the description
in Section 6.4.1 of the SD Physical Layer Specification Version
3.00.

Additionally, I believe the 1000us delay at the beginning of the
mmc_go_idle() function exists for a similar reason. Without this
1000us delay, CMD0 would be sent even before the voltage reaches
VDD_MIN.

This should be a common issue, not limited to the IMX platform.
We can either add a delay in the code or add regulator-ramp-delay/
startup-delay-us in numerous dts files.

Regards,
Luke

> 
> Without useful details, this is a NAK.