On Wed, Sep 13, 2006 at 04:27:07PM +0400, Vitaly Wool wrote: > On 9/13/06, Jonathan McDowell wrote: > >On Tue, Sep 12, 2006 at 04:23:11PM +0400, Vitaly Wool wrote: > >> On 9/12/06, Jonathan McDowell wrote: > >> >Should I expect this to work on filesystem that works fine on 2.6.16.5 > >> >(in particular 2.6.16.5-tcl1 on a Balloon2)? > >> > >> I think you should. If that is not the case, please let me know, let's > >> debug that together :) > > > >Ok. I've got a bit further with is. If I enable CONFIG_YAFFS_DOES_ECC > >in my 2.6.18-rc5 build (it was enabled in the 2.6.16.5) then I can at > >least mount and boot from the yaffs partition. I'm a bit surprised by > >this - YAFFS and MTD handle ECC usage differently? > > So, are you saying that you've got the original filesystem as created > in 2.6.16.5 with CONFIG_YAFFS_DOES_ECC option? Yes; it's the image from: http://balloonboard.org/files/balloon3/distro/test-v0.1/rootfs/balloon3-test.yaffs.gz > If so, then nothing good can happen in 2.6.17+ as yaffs will conflict > in OOB usage with the MTD core. I really think that > CONFIG_YAFFS_DOES_ECC should be marked as obsolete for 2.6.17+ > kernels. Enabling the CONFIG_YAFFS_DOES_ECC option should still work though, shouldn't it? While I agree that using the MTD ECC is probably saner I'm guessing that if yaffs does things slightly differently that method still needs support - eg for other OSes that might use yaffs partitions that need mounted under Linux or bootloaders. J. -- I'm a *daemon*! Demons are evil. Daemons are not!