[Yaffs-archive] Re: YAFFS list and archive

[Charles Manning]

The yaffs list does indeed seem iffy at present (maybe for the last few days).

> We are planning to include NAND flash on a Linux/PPC based processor
> board we are developing and are considering using YAFFS.  Nick didn't
> mention a list archive so I will take the liberty of asking you a few
> questions directly.
>
> The the paper on the web site was a bit sketchy on details for the
> mount times.  Do you have an estimate for how long it would take to
> scan/mount a 256MB YAFFS partition?  I assume this task is I/O bound
> so it should be dependent mainly on the read and access times.  Do you
> know of any PCI cards which might provide register access to NAND
> flash?  It would be useful for testing before our target hardware is
> built.

It is indeed difficult to give anything but vague answers. I personally have 
not run YAFFS on real NAND under Linux. I have only used it with WinCE. I 
expect similar performance for the scanning though.

The systems I have been playing with use 512MB of NAND as 4x128 parts, 
running YAFFS1. We get fully booted (ie loading a 12MB image, boot, and mount 
YAFFS) in under a minute. I'd have to pull out a stopwatch to give you a 
better breakdown.

Scanning is really a function of how fast you can read the chunk tags.

YAFFS2 will be faster since the chunks are bigger (hence less chunks for a 
given system).

Samsung MLC will be slower because these parts are slower to "seek". The 2kB 
page Samsung stuff looks pretty fast, especially if you use the 16-bit 
interface.

I have also sketched out a form of "checkpointing" whereby a snapshot of the 
internal structures is dumped to NAND at unmount and can be read back on 
mount without a  full scan. This will allow very fast mounting ( ??3 secs?? 
for 512MB). This has not been coded yet due to it not being a high priority. 
It can be made to happen though. I have been known to reprioritise :-).


>
> Because of the rather large total storage we need (512 MB), at some
> point (perhaps not the first revision), we will most likely be using
> the next generation of flash chips which means MLC and/or large block
> parts.  It looks like you plan to support these in YAFFS2.  Any idea
> when that might make it to the working code stage?
>
> Thanks in advance for your response and thanks for your work on YAFFS!

The good feedback makes it all worthwhile.

--Charles


---------------------------------------------------------------------------------------
This mailing list is hosted by Toby Churchill open software (www.toby-churchill.org).
If mailing list membership is no longer wanted you can remove yourself from the list by 
sending an email to yaffs-request@toby-churchill.org with the text "unsubscribe" 
(without the quotes) as the subject.