[U-Boot] [PATCH V4 1/2] ext4fs ls load support

[Graeme Russ]

Hi Manjunatha & Rob

On Wed, Mar 28, 2012 at 12:46 AM, Rob Herring <robherring2 at gmail.com> wrote:
> On 03/27/2012 08:13 AM, manjunatha wrote:
>> Dear Rob Herring/ Wolfgang,
>>
>> As discussed in earlier mail chain, our current ext4 implementation is
>> capable enough to list(ls) and read(load) ext2 partitons as well.
>>
>> Also the current ext4 implementation is optimized to 7-8 times greater
>> read throughput than ext2load.
>
> Cool. IIRC, there was just recently some discussion on the list about
> ext2 performance problems.

I am very keen to see this ext4 support hit maturity as I too am seeing
sub-par performance reading from an ext2 filesystem on an SD card. (I'm a
bit sad that I simply have not had the time to test drive it myself)

>> I propose that we can remove the existing ext2 code and let it be
>> replaced with ext4.
>
> That's exactly what I proposed. Although, I think your patch should be
> done as modifications to cmd_ext2.c rather than deleting it and adding a
> new cmd_ext4.c

That depends on what the delta between the ext2 and ext4 code is. If the
delta is large (i.e. the patches are not intuitive to read) then I would
suggest the following sequence:

 1. Introduce the ext4 code with dedicated ext4ls, ext4load and ext4write
    functions as 100% seperate code (i.e. do not modify existing ext2 code)
 2. Delete the ext2 code and modify ext2ls and ext2load to use the new ext4
    functions (also add ext2write assuming these patches support writing to
    ext2)

>> All users using ext2ls and ext2load command can replace them with ext4ls
>> and ext4load respectively.
>
> I think you need to keep the command names to not break existing users'
> scripts. So either we just stick with ext2ls/ext2load or define
> additional ext4* commands which call the same functions as the ext2
> version. I don't really care, but would lean toward the former.

As above - I don't see using ext2* to access an ext4 filesystem as very
intuitive (or ext4* to access an ext2 filesystem) - The user does not
need to know what happens under the hood, but the command interface
should always make sense.

Regards,

Graeme