zfs code - Nov 2008 - zfs znode changes getting lost

In place of padding in zfs znode i added a new field. stored an integer value
and am able to see saved information.

but after reboot it is not there.  If i was able to access before reboot so it
must be in memory. I think i need to  save it to disk.
how does one force zfs znode to disk.
right now i dont do anything special for it. Just made an ioctl, accessed znode
and made changes.

example in zfs_ioctl

case add_new:
                zp = VTOZ(vp);
		zp->z_phys->new_field = 2;
               return(0);
-- 
This message posted from opensolaris.org

shelly wrote:
> In place of padding in zfs znode i added a new field. stored an integer
value and am able to see saved information.
> 
> but after reboot it is not there.  If i was able to access before reboot so
it must be in memory. I think i need to  save it to disk.
> how does one force zfs znode to disk.
> right now i dont do anything special for it. Just made an ioctl, accessed
znode and made changes.
> 
> example in zfs_ioctl
> 
> case add_new:
>                 zp = VTOZ(vp);
> 		zp->z_phys->new_field = 2;
>                return(0);
Did you actually add a new field to znode_phys_t or did you take some 
space from the existing zp_pad[3] ?  You can''t just change the size of
a
znode_phys_t every easily without having major on disk compat issues - 
that is why there is a pad area.

There is a knock on impact to changing the size of a znode_phys_t in 
that it impacts the size of a dnode_phys_t as well and will interfere 
with the bonusbuf size.

Why are you attempting to add a new field to the znode_phys_t anyway ? 
What are you attempting to store there ?

-- 
Darren J Moffat

i took space from pad space. made zp_pad[2]  and added another field in place of
it.

So there is no issue of znode_phys_t size changing.
-- 
This message posted from opensolaris.org