Hi all, I come back to this thread, since I managed to make the driver work on OMAP processor.
To be more precise, the driver is working as far as the MMC driver uses polling method. But with the last kernel I can use ( 2.6.29 ), the MMC HCD uses DMA to write/read the SDIO card. Consequently, data transfers have to be blocks of size modulo 4. Transfer of block-size non-modulo 4 is not implemeted. Adding some traces could let me see that the AR6000 driver sometimes writes blocks of size 1 and 3, through the HIFReadWrite function. I am a bit lost about how to modify the driver in order to make the block transfers fit my MMC HCD. Do I only have to modify the HIFReadWrite function, or can there be any deeper modification to make ? Any idea would help. Thanks a lot. Michael Trimarchi wrote: > > Hi, > > benson.dong wrote: >> On Wed, 08 Apr 2009 16:44:16 +0200, Michael Trimarchi >> <[email protected]> wrote: >> >>> Sorry again, >>> I have a bit differnt code, do you pass the GetTargetInfo call, >>> >>> if not can you print the entire credit. Credit is a four byte, maybe >>> depend >>> in how the spi is configured MSB etc and the bus width. So it arrives >>> but it is invalid >>> >>> >> >> Sorry about my careless. >> I think i have posted a wrong link. >> My issue is here: >> http://lists.openmoko.org/nabble.html#nabble-td2548680 >> However, it's a bit similar with OMAP3's issue. >> (If possible could you post in above thread?) >> >> I think i'm using the same code as you which is from Openmoko's lastest >> kernel 2.6.29. >> The issue is that ar6000 driver never passed the GetTargetInfo call. >> I'm sure that the Credit value read in host driver in low level is 0. >> So it will not cause by the "& 0xff". >> > Ok, because I can't try... I will take a look again to the code. > Michael > > > > -- View this message in context: http://n2.nabble.com/-AR6000--Openmoko-driver-for-OMAP3-%28-TI-%29-tp2591890p3011915.html Sent from the Openmoko Kernel mailing list archive at Nabble.com.
