An update on this...
There were two sets of problems: One inside the yaffs_fs.c and another in
yaffs_guts.c
The yaffs_fs.c problem seems to be fixed now and fsx runs fine for long
periods (a day) on partitions up to 64MB.
The yaffs_guts.c problem only manifests on devices > 64MB. The good news is
that I can make this happen in the yaffs_direct test harness (without using
the VFS). This should make the problem far easier to isolate.
Thank you very much to Jisheng for helping to highlight this issue.
It's getting late on Christmas Eve for me...
-- CHarles
On Thursday 24 December 2009 08:48:14 Charles Manning wrote:
> Hmmm....
>
> Four CPUs ran the whole night with
>
> modprobe mtd
> modprobe mtdblock
> modprobe nandsim first_id_byte=0x20 second_id_byte=0xa2 third_id_byte=00
> fourth_id_byte=0x15
>
>
> but fails within seconds on
> modprobe nandsim first_id_byte=0x20 second_id_byte=0xac third_id_byte=0x00
> fourth_id_byte=0x25
>
> That would point to something within yaffs, rather than the vfs interaction
> that has been fixed.
>
> Still, what is in cvs now is a big step up from what was there previously.
>
> -- Charles
>
> On Wednesday 23 December 2009 19:15:48 JiSheng Zhang wrote:
> > Hi Charles,
> >
> >
> > Thanks for your time. I have test the newest code, it still failed :(.
> >
> > 2009/12/23 Charles Manning <manningc2@actrix.gen.nz>:
> > > Hi
> > >
> > > The problem is more to do with the interactions between the VFS and
> > > yaffs.
> > >
> > > I just updated cvs with a fix that seems to work pretty well. I have 4
> > > CPUs running fsx on the nandsim. One has been running for 3 hours and
> > > the other 3 for over an hour.
> > >
> > > Previous to this fsx would typically fail within a few minutes.
> >
> > some nandsim parameter and fsx's seed can trigger failure easily. For
> > example,
> >
> > modprobe nandsim first_id_byte=0x20 second_id_byte=0xac
> > third_id_byte=0x00 fourth_id_byte=0x25
> > modprobe mtdblock
> > mount -t yaffs2 /dev/mtdblock0 /mnt/yaffs2
> > ./fsx -N 100000 -S 1596 /mnt/yaffs2/testfsx
> >
> > it will fail at 7xxxx operations.
> >
> > Regards,
> > Jisheng
> >
> > _______________________________________________
> > yaffs mailing list
> > yaffs@lists.aleph1.co.uk
> > http://lists.aleph1.co.uk/cgi-bin/mailman/listinfo/yaffs
>
> _______________________________________________
> yaffs mailing list
> yaffs@lists.aleph1.co.uk
> http://lists.aleph1.co.uk/cgi-bin/mailman/listinfo/yaffs
(text/plain)