SD power-control necessary?

SD power-control necessary?

[Hein_Tibosch]

Hello,

Maybe a stupid question, but when designing new hardware with a slot
for an sd-card, how essential is it that the driver is able to power off/on
the card?

I would say essential, because I've seen sd-cards in a state in which they
didn't respond to MMC_GO_IDLE_STATE anymore, until they were re-inserted.

Can anyone shed a light on this? How could an sd-card get into such a state?

Will the mmc driver fall back to mmc_rescan() after an initialized card
becomes non-responsive?

Thanks, Hein

Re: SD power-control necessary?

[Edward A. Falk]

Hein_Tibosch wrote:
> Hello,
> 
> Maybe a stupid question, but when designing new hardware with a slot
> for an sd-card, how essential is it that the driver is able to power off/on
> the card?

Just a random thought: allowing the driver to power off the SD card will 
defeat the "power-on memory protect" feature.  This is how the 
supposedly secure G2 was jailbroken.

	-ed falk

Re: SD power-control necessary?

[Chris Ball]

On Fri, Jan 07, 2011 at 07:37:25AM -0800, Edward A. Falk wrote:
>> Maybe a stupid question, but when designing new hardware with a slot
>> for an sd-card, how essential is it that the driver is able to power off/on
>> the card?
>
> Just a random thought: allowing the driver to power off the SD card will  
> defeat the "power-on memory protect" feature.  This is how the  
> supposedly secure G2 was jailbroken.

It wasn't powered off by the SD (it's actually soldered-down eMMC) driver,
though -- the eMMC controller itself was reset by a sideband GPIO, and
then the MMC stack was modified to be able to bring the controller back
up without disrupting userspace.

So, not entirely relevant to the case of controlling card power from
inside the MMC driver.

-- 
Chris Ball   <cjb@laptop.org>   <http://printf.net/>
One Laptop Per Child

Re: SD power-control necessary?

[Edward Falk]

On 01/07/2011 07:37 AM, Edward A. Falk wrote:

> Just a random thought: allowing the driver to power off the SD card will
> defeat the "power-on memory protect" feature. This is how the supposedly
> secure G2 was jailbroken.

Ahh, I see we're talking about a removable SD card.  In that case, 
jailbreaking is not an issue, so yes, the driver should be able to 
remove power from the device.

Re: SD power-control necessary?

[Michał Mirosław]

2011/1/7 Hein_Tibosch <hein_tibosch@yahoo.es>:
> Maybe a stupid question, but when designing new hardware with a slot
> for an sd-card, how essential is it that the driver is able to power off/on
> the card?

For one, this is needed for power-management. In powered suspend case,
when card's slot does not provide card change detection, power-cycle
is the only way sure way to get a card to known state in case it can
be removed in that time.

> I would say essential, because I've seen sd-cards in a state in which they
> didn't respond to MMC_GO_IDLE_STATE anymore, until they were re-inserted.
>
> Can anyone shed a light on this? How could an sd-card get into such a state?

After CMD15 (GO_INACTIVE_STATE) or because of a card bug.

>
> Will the mmc driver fall back to mmc_rescan() after an initialized card
> becomes non-responsive?

When the card goes off-protocol it's in undefined state, so a reset is
the only way to get it back to known state. Since in SD/MMC there's no
reset signal - only power cycle can reset the card.

Best Regards,
Michał Mirosław