Re: linux-mtd digest, Vol 1 #323 - 3 msgs

Re: linux-mtd digest, Vol 1 #323 - 3 msgs

[Geoffrey Espin]

> From: "Sanjay Kumar" <sankumar96@lycos.com>
> Subject: Problem in changing  the starting location of JFFS2 for AMD flash
> I am using AMD flash of 4M size and keeping starting 1/2 MB for RAW data. Then for the remaining I want to use JFFS2. That is JFF2 will start from 0.5 MB to 4MB. For that I am making changes in super.c of JFFS2 for function jffs2_read_super. The changes are -    
> c->free_size = c->mtd->size - jffs2_start_addr;
> c->flash_size = c->mtd->size - jffs2_start_addr;
> c->nr_blocks = c->flash_size / c->sector_size;
> c->blocks[i].offset = i * c->sector_size + jffs2_start_addr;
> where am keeping jffs2_start_addr = 0x80000.
> I am using linux 2.4.2.
> Now I am able to mount the jffs2 file system and able to create a empty file.
> But when I go for writing into the created empty file, it is crashing.

As someone already has replied, don't mess with sources like super.c.

In linux/.config, have these options on:

    CONFIG_MTD=y
    CONFIG_MTD_PARTITIONS=y
    CONFIG_MTD_CHAR=y
    CONFIG_MTD_BLOCK=y
    CONFIG_MTD_<myboard>=y

Under linux/drivers/mtd/maps/ you must create your myboard-flash.c
(and add to Makefile), which has the partition table in it, e.g.:

    #define WINDOW_ADDR 0xbfc00000  /* default 4M part */
    #define WINDOW_SIZE 0x00400000
    ...
    static struct mtd_partition myboard_partitions[] = {
	    {
		    name: "bootstrap and kernel",
		    size: WINDOW_SIZE / 4,  /* 0.5M */
		    offset: 0,
	    },{
		    name: "filesystem",
		    size: MTDPART_SIZ_FULL, /* 2M */
		    offset: MTDPART_OFS_APPEND,
	    }
    };

Then from shell you can:

    # mtd_debug erase /dev/mtd/0 ${OFFSET} ${SIZE}   # 0 is entire device
    # cp empty.jffs2 /dev/mtd/2                      # 2 is second partition
    # mount -t jffs2 /dev/mtdblock/2 /mnt

Where 'empty.jffs2' is a stub file created by mkfs.jffs2,
and OFFSET & SIZE are whatever boundaries you've setup.

You might need to drop in later mtd files into 2.4.2.

Geoff
-- 
Geoffrey Espin
espin@idiom.com

Re: linux-mtd digest, Vol 1 #323 - 3 msgs

[David Woodhouse]

espin@idiom.com said:
>  Under linux/drivers/mtd/maps/ you must create your myboard-flash.c
> (and add to Makefile), which has the partition table in it, e.g.: 

Useful explanation, thanks - except that you omitted the step where if it's 
a commonly available board, they should send me the appropriate 
myboard-flash.c too :)

--
dwmw2

Atmel Flash

[Charles Steinkuehler]

Are the small-sector Atmel flash (ie 29C010/29C040) parts supported by MTD?

These are the parts with 128/256 Byte sector size (ie 1024 128 byte sectors
in the '010, and 2048 256 byte sectors in the '040), with no explicit erase
command required before writing new data.

Charles Steinkuehler
charles@steinkuehler.net

Re: Atmel Flash

[Eric W. Biederman]

"Charles Steinkuehler" <charles@steinkuehler.net> writes:

> Are the small-sector Atmel flash (ie 29C010/29C040) parts supported by MTD?
> 
> These are the parts with 128/256 Byte sector size (ie 1024 128 byte sectors
> in the '010, and 2048 256 byte sectors in the '040), with no explicit erase
> command required before writing new data.

Other atmel parts are.  If they aren't cfi then there is the issue
that they aren't in the table jedec_probe.c.   But it should be trivial
to add them.

Eric

Re: Atmel Flash

[David Woodhouse]

ebiederman@lnxi.com said:
> "Charles Steinkuehler" <charles@steinkuehler.net> writes:
> > Are the small-sector Atmel flash (ie 29C010/29C040) parts supported by
> > MTD?
> 
> > These are the parts with 128/256 Byte sector size (ie 1024 128 byte
> > sectors in the '010, and 2048 256 byte sectors in the '040), with no
> > explicit erase command required before writing new data.

> Other atmel parts are.  If they aren't cfi then there is the issue
> that they aren't in the table jedec_probe.c.   But it should be
> trivial to add them.

I think the other supported Atmel parts are the ones with the conventional 
command set. These ones are different. It's not too hard to write support 
for them though.

--
dwmw2