[U-Boot] [PATCH 4/7] dfu: MMC specific routines for DFU operation

[Lukasz Majewski]

Hi Marek,

> Dear Tom Rini,
> 
> > On Tue, Jul 03, 2012 at 11:29:31PM +0200, Marek Vasut wrote:
> > > Dear Lukasz Majewski,
> > > 
> > > > Support for MMC storage devices to work with DFU framework.
> > > > 
> > > > Signed-off-by: Lukasz Majewski <l.majewski at samsung.com>
> > > > Signed-off-by: Kyungmin Park <kyungmin.park at samsung.com>
> > > > Cc: Marek Vasut <marex at denx.de>
> > > > ---
> > > 
> > > [...]
> > > 
> > > > +
> > > > +#include <common.h>
> > > > +#include <malloc.h>
> > > > +#include <dfu.h>
> > > > +
> > > > +int dfu_write_medium_mmc(struct dfu_entity *dfu, void *buf,
> > > > long *len) +{
> > > > +	ALLOC_CACHE_ALIGN_BUFFER(char, cmd_buf,
> > > > DFU_CMD_BUF_SIZE); +
> > > > +	memset(cmd_buf, '\0', sizeof(cmd_buf));
> > > > +
> > > > +	switch (dfu->layout) {
> > > > +	case RAW_ADDR:
> > > > +		sprintf(cmd_buf, "mmc write 0x%x %x %x",
> > > > (unsigned int) buf,
> > > > +			dfu->data.mmc.lba_start,
> > > > dfu->data.mmc.lba_size);
> > > > +		break;
> > > > +	case FAT:
> > > > +		sprintf(cmd_buf, "fatwrite mmc %d:%d 0x%x %s
> > > > %lx",
> > > > +			dfu->data.mmc.dev, dfu->data.mmc.part,
> > > > +			(unsigned int) buf, dfu->name, *len);
> > > > +		break;
> > > > +	default:
> > > > +		printf("%s: Wrong layout!\n", __func__);
> > > > +	}
> > > > +
> > > > +	debug("%s: %s 0x%p\n", __func__, cmd_buf, cmd_buf);
> > > > +	run_command(cmd_buf, 0);
> > > 
> > > Holy Moly ... can we not make this into simple calls to those
> > > subsystems ? Instead invoking command is crazy ;-)
> > 
> > Are they really simple?  There's a few other places we do this, and
> > so long as it's documented that DFU depends on CONFIG_FAT_WRITE for
> > writing to fat and so forth.
> 
> Well ain't it easier to call fat_write() or similar?
> 
I've decided to use run_command on a purpose.

This call provides clean and reliable API. It is very unlikely that the
mmc write <dev> <addr> <start> <size> command will change (or any
other). 
On the other hand the fields of struct mmc are changed from time to
time.

Moreover, mmc drivers are also a subject to change (like adding dw_mmc
recently).
Using run_command also takes the burden of mmc_init() related calls. 

Of course the run_command's downside is the speed of execution. But is
it so important when one considers, the firmware update? 

Side note: DFU uses only EP0 (for transfer and configuration), so this
is rather slow communication link. 

I'm open for discussion.


-- 
Best regards,

Lukasz Majewski

Samsung Poland R&D Center | Linux Platform Group