Re: [PATCH] Fixes and enhancements for the MMC SPI driver

[Matt Fleming <matt@console-pimps.org>]

On Thu, Jan 08, 2009 at 09:22:08AM +0100, Wolfgang M?es wrote:
> Matt,
> 
> Am Dienstag, 6. Januar 2009 schrieb Matt Fleming:
> > Which cards are you having issues with?
> 
> This was a noname 4 GByte SDHC card. If you write a continious stream of data 
> at max. speed, the card generates a BUSY signal every 10 seconds for nearly a 
> whole second.
> 

Ouch :-/

> > Modifying the code to no longer abide by the standard doesn't seem like a
> > win to me.
> 
> I strongly disagree with you.
> 
> While each and every SD card should conform to the standard, a driver for a 
> host controller should be as forgiving as possible. If the driver cannot cope 
> with cards sold today, the user has a bad experience.
> 
> I code drivers for people and not for standards.
> 
> Remember, this is only a timeout. Lengthen the timeout will have no 
> consequences for cards which behave.

My concern is that completely broken cards may be able to get away with more
if the timeout is extended. There is clearly a tradeoff here between
supporting non-conforming cards and easily detecting broken cards.

Someone suggested to me off list that having a configuration option to
enable/disable the longer timeout is a good compromise, and I agree.