Hello all,

I'm attempting to get MMC over xilinx_spi working on a custom Xilinx Virtex-4 board (that has a working SD/SPI interface, tested with a standalone app).

I'm having issues mounting the device, and after doing a bit of debugging, I'm confused about how things are supposed to work. I'm using 2.6.24-rc8-xlnx from git.xilinx.com, with some patches (for the uartlite (Grant Likely), and other minor things). In addition, I'm using some of a patch from Andrei Konovalov (http://patchwork.ozlabs.org/linuxppc-embedded/patch?id=11577) for Xilinx SPI support in virtex_devices.c (BTW, will this patch hit the main git.xilinx.com tree at some point?)

Here's what I do, and what I see:
- Create a block device: "mknod /dev/m0 b 179 0"
- Access that device ("cat /dev/m0")
- I see the calls into the kernel:
    blkdev_open()
       do_open()
          get_gendisk()
             kobj_lookup(bdev_map, ...)
- The device is not found (bdev_map has been initialized, but the only "probe" at major 179 is base_probe().
- kobj_lookup() returns NULL, and the open returns ENXIO (No such device or address).

Here's what I don't understand: In order for kobj_lookup to successfully return non-NULL, a "probe" has to have been added to bdev_map using kobj_map(). But, kobj_map(bdev_map,...) is called only by blk_register_region(), which is never called by the MMC driver.

I'm clearly missing something here. Has anyone successfully used MMC over SPI or xilinx_spi? How is the MMC driver supposed to work in regards to opening a device?

Also, for reference:

I've added the following configs:
CONFIG_SPI=y
CONFIG_SPI_DEBUG=y
CONFIG_SPI_MASTER=y
CONFIG_SPI_BITBANG=y
CONFIG_SPI_XILINX=y

CONFIG_MMC=y
CONFIG_MMC_BLOCK=y
CONFIG_MMC_BLOCK_BOUNCE=y
CONFIG_MMC_SPI=y

And, /proc/devices does report the mmc device:

# cat /proc/devices
Character devices:
1 mem
4 /dev/vc/0
4 tty
5 /dev/tty
5 /dev/console
5 /dev/ptmx
7 vcs
10 misc
13 input
128 ptm
136 pts
204 ttyUL

Block devices:
1 ramdisk
179 mmc

-- 
Brian Silverman
Concept X, LLC