JFFS2 mount time
Gareth Bult (Encryptec)
Gareth at Encryptec.net
Wed Dec 15 20:02:54 EST 2004
Hi,
Sure, the file I've been changing is blkmtd.c.
The device I'm aiming for and testing on is 256Mb with a 128k erase
size.
Currently I'm using the code included in the 2.6.9 kernel source tree
with tools from cvs as of about a week ago.
Is this likely to contain the latest speedups or should I be looking
elsewhere?
tia
Gareth.
On Wed, 2004-12-15 at 18:15 -0600, Josh Boyer wrote:
> Gareth Bult (Encryptec) wrote:
> > Hi,
> >
> > I've been working on the MTD block driver for JFFS2 and now seem to have
> > a relatively stable environment with a JFFS2 root filesystem on a USB
> > flash key device.
> > (mtd block device speed now approximates actual device speed on both
> > read and write)
>
> I assume you mean blkmtd here. The mtdblock driver is used to present
> an IDE type interface to flash, whereas blkmtd is used to present a
> flash interface to an IDE type device.
>
> >
> > It seems to be much quicker (generally) than the previous ext2
> > incarnation.
> >
> > However, I'm looking at 30secs to mount the root filesystem.
>
> How big of a device?
>
> > a. Is it possible (conceptually) to background the mount scan so the
> > system can boot while the scan happens ?
>
> Which version of the code are you using? Mount time has been
> significantly improved in newer versions of the code.
>
> > b. Is there an easy way to bypass the scan and does JFFS3 do this ?
>
> Not that I know of, because the code needs to read the node information
> in order to get the versioning of inodes, etc. correct.
>
> JFFS3 is essentially the exact same code base as JFFS2 right now. It
> was just forked about a week ago.
>
> josh
>
More information about the linux-mtd
mailing list