On Wed, 27 Jul 2005, Charles Manning wrote:
That's fine and even convenient to have such a script. Something like MTD
guys have.
But there is a huge BUT - is YAFFS supposed to never make it into the
kernel? You know, it's a great FS that works quite fine and fast on NAND
flashes. Furthermore, there is no other as of today. But living outside the
mainstream makes it marginal, there are very few people that even heard about
it's existance :((
I remember how long it took XFS to make it into the kernel. I remember the
fierce fight for now obsolete devfs that even I took part in. But now we do
have XFS (as well as reiserfs and a couple of others) right out of the box.
Devfs also lives in it's second incarnation. Will we ever see YAFFS in a
stock kernel?
It is not all that important for a regular dumb PC user that can happily
live without such a luxury. Anyway, he doesn't use bare chips and all those
funny memory cards come with their drivers. And there is nothing wrong with
using FAT on them as far as PC user concerned. But such an FS is vital for
the stuff we do, we simply can't live without it. It's the only FS that can
hold our entire root FS and we can't tolerate data loss or similar disaster
that might be catastrophic. An early JFFS3 is not an answer also because
no one would tolerate a "hardware box" taking 10 minutes to boot. Even 30
seconds is too much although bearable due to lack of other options. And what
do we have -- a marginal barely known FS, kernel that holds together
exclusively on a set of patches that are well hidden and known only to a
limited number of enlightened ones who perform a monkey drum dance every time
a new kernel shows up to put all those secret ingredients together without
blowing up the entire pot... Yeah, it's great for a job security and
business, but it slows down the progress to almost grinding halt. One
doesn't have time to do anything else because applying all the patches from
a hundred places and making a system work eats up all of his time...
And I don't even mention lack of FS tools... It looks like nobody noticed
that they do not build at all and that "../nand_ecc.c" from their makefile
last time was at that location when dinosaurs ruled the Earth... I had to
spend several days to hack and bandaid them to get the tools build without
a kernel tree and to make them really work. And that's just YAFFS1, tools
for YAFFS2 are still in pre-conceptional stage, aren't they?
So, are you going to make it into the mainstream kernel?
BTW, I can make a tar.gz of YAFFS1 tools that build and work tomorrow if
anybody cares to have those. Sure, one can use nandwrite/flash_erase/etc.
but mkyaffsimage/mkyaffs are much easier and handier.
OK, sorry for this flow of consciousness, pressure to get a new hardware
board up and running was too high... And those @#@$$% from ST sent us 128
MiB NAND samples that were just like the real ones, with all the bells and
whistles, but with only one die working so they were really 64 MiB... An
entire week wasted in desperate attempts to understand why yaffs makes all
the files dissapear on an FS mount...
Keep up a good job guys. Anyway there is no alternative to YAFFS as of
now...
> On Wednesday 27 July 2005 13:15, Sergey Kubushyn wrote:
> > Hi!
> >
> > This one allows to build yaffs2 in main kernel tree. Just copy
> yaffs2/ to
> > the fs/ tree, remove anything but *.[ch] from yaffs2/ (Makefile and 2
> > directories), apply the patch, make menuconfig and enjoy.
>
> Thanx for this Sergey
>
> As previously discussed, we won't put any "patches against kernels"
> into CVS
> because this leads to dependency hell.
>
> Instead, the CVS will have a patch in script which makes for a clean
> partitioning. This might be used to generate "patch-balls" which can be
> applied to the kernel.
>
> The Marty, Luc and I are working towards a cleaned up patch in script
> and
> simplified stuff to go with it.
>
> A lot of this patch set is already in CVS, but one of us will sift
> through it
> to see what is new here.
>
> -- Charles
>
>
---
******************************************************************
* KSI@home KOI8 Net < > The impossible we do immediately. *
* Las Vegas NV, USA < > Miracles require 24-hour notice. *
******************************************************************
(text/plain)