Fixes for clone_bio and __clone_and_map functions in dm.c

AG> I encountered another problem in the final else
condition of
AG> __clone_and_map() function where the bio vector is
split in to two
AG> parts. This is perhaps based on the assumption that
the next
AG> target can completely accomodate the bio vector.
However, I have
AG> encountered situations when this assumption can be
violated and
AG> have introduced a do-while loop to correct it.

I found this bug last week as it was preventing me from
having targets
that have a split_io of 1 sector.  Interestingly, I had
planned to
code up a solution very similar to yours next week!

I'll test it tomorrow to see if it solves my problem...
Thanks! 

-- 
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danmsus.ibm.com
--
dm-devel mailing list
dm-develredhat.com
http
s://www.redhat.com/mailman/listinfo/dm-devel

On Sun, Mar 19, 2006 at 10:24:00AM -0800, Abhishek Gupta
wrote:
> I found a small bug in the clone_bio() function in dm.c
. The number of bio
> vectors in the cloned bio is incorrectly set to the
index of the last bio
> vector in its range. I have corrected it in the
attached patch.

I don't see that yet - the code is not very transparent
though.
(Look at the places that call the function.)
 
> I encountered another problem in the final else
condition of
> __clone_and_map()
> function where the bio vector is split in to two parts.


Yes, that looks wrong, as Dan reported on IRC this week.
The patch needs a little tidying up though.
[only one split_bvec() call not two?
 reconsider loop entry/exit to avoid artificial-looking
exit?]

Alasdair
-- 
agkredhat.com

--
dm-devel mailing list
dm-develredhat.com
http
s://www.redhat.com/mailman/listinfo/dm-devel