Request for review: NandFS

Request for review: NandFS

[Corentin Chary]

Hi,
Here is NandFS, a NAND Flash Filesystem.
NandFS is similar to YAFFS2 (in some ways), but the codebase is much more
cleaner and simpler. And we'll try to go mainline.

Development started at Adeneo (http://www.adeneo-embedded.com/srt/en/home) in 
2007. After a long pause, I re-started to clean the code in october 2008.

With the new UBIFS, we might think than there is no room for another flash 
filesystem. But another GPLed filesystem can't hurt anyone. Some of us may 
have heard some words about UFFS, but this project is in no way related to 
NandFS.

*IMPORTANT*: NandFS is under active development, and can't be considered safe 
or stable !

NandFS aims to be:
- Simple
- Clean
- Ram-Efficient
- Power-Failure-Resistant
- Write-Optimized (write-back + internal caches)
- Wear-Leveling

NandFS work-in-progress:
- writeback and "budgeting"
- kernel-doc

NandFS short term todo:
- (A lot of) Test
- Better locking (no more "Big NandFS Lock")
- checkpoint/summary support
- ram usage optimizations
- xattr support
- acl support

The code can be found at http://git.iksaif.net/

I'm currently working on kernel-doc, but the more important is done.
There is no whitepaper to explain how it works right now, but nandfs.h and 
scan.c explains a lot and they are well commented.

It's all Licenced under GPLv2.

Please post your question/comments/bugs/patches.

-- 
Corentin Chary
http://xf.iksaif.net

Re: Request for review: NandFS

[Artem Bityutskiy]

Corentin Chary wrote:
> Hi,
> Here is NandFS, a NAND Flash Filesystem.
> NandFS is similar to YAFFS2 (in some ways), but the codebase is much more
> cleaner and simpler. And we'll try to go mainline.
> 
> Development started at Adeneo (http://www.adeneo-embedded.com/srt/en/home) in 
> 2007. After a long pause, I re-started to clean the code in october 2008.
> 
> With the new UBIFS, we might think than there is no room for another flash 
> filesystem. But another GPLed filesystem can't hurt anyone. Some of us may 
> have heard some words about UFFS, but this project is in no way related to 
> NandFS.
> 
> *IMPORTANT*: NandFS is under active development, and can't be considered safe 
> or stable !

Probably you have a very good motivation to do work from
scratch, but what is sad that instead of joining us and
improving UBI/UBIFS stack, you spend your time creating new
file-system.

Almost nobody is interested in new unscalable solutions, and
you really waste your time, because we already have 3 of them,
at least.

There are so many things to improve in UBI/UBIFS, and there
is a possibility to create a scalable UBI2. You may do a lot
of exciting things in the area, instead.

Please, think about joining us instead :-)

-- 
Best Regards,
Artem Bityutskiy (Артём Битюцкий)
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Request for review: NandFS

[Bill Gatliff]

Artem Bityutskiy wrote:
> There are so many things to improve in UBI/UBIFS, and there
> is a possibility to create a scalable UBI2. You may do a lot
> of exciting things in the area, instead.
>   

What's in Corentin's code that UBI/UBIFS could learn from?  Or have you 
reviewed it that closely?


> Please, think about joining us instead :-)
>   

Of course, if NandFS is better than UBI/UBIFS, then maybe you should 
think about joining him instead!  :)


b.g.

-- 
Bill Gatliff
bgat@billgatliff.com

Re: Request for review: NandFS

[Corentin Chary]

>> Please, think about joining us instead :-)
>>
>
> Of course, if NandFS is better than UBI/UBIFS, then maybe you should think
> about joining him instead!  :)

Yep, I think you should do that ! :)

Seriously, NandFS may not be scalable, but with checkpoints and
writeback it'll should get better performances than JFFS/YAFFS.
And as it only use mtd, it should be more easily portable to something
else that linux.

But I know that we now need scalable solutions, and it's why I work on
another project ( http://uffs.org/ ). What we are trying to do with
that is totaly different. We will try to do something portable (linux,
windows, fuse (for MacOSX and *BSD)). And more important we wan't to
make it work on raw NAND, *and* on block devices based on NAND. For
that we've got EBM, which is very similar to UBI (and uses UBI on
Linux) to talk to raw NAND or block devices. You may look at
http://uffs.org/pages/documentation .
For the core filesystem, we may make something new or use a modified
version of UBIFS (in this case, the project may be re-named).
Anyway, by working on UFFS, we will contribute to UBI/UBIFS and UBI2
(ebm on block devices will look like UBI2).


-- 
Corentin Chary
http://xf.iksaif.net