Am 24.10.2014 um 15:57 hat Max Reitz geschrieben: > qemu-img should use QMP commands whenever possible in order to ensure > feature completeness of both online and offline image operations. For > the "commit" command, this is relatively easy, so implement it first > (in the hope that indeed others will follow). > > As qemu-img does not have access to QMP (due to QMP being intertwined > with basically everything in qemu), we cannot directly use QMP, but at > least use the functions the corresponding QMP commands are using (which > would be "block-commit", in this case). > > > v14: > - Patch 3: > - Pull the optimized code out into an own function [Kevin] > - Only run the optimized code if the image header, the reftable, a > single refblock and the L1 table can be described by that single > refblock [Kevin] > - Zero out the L1 table before modifying anything else [Kevin] > - Always set bs->drv to NULL if an error occured which broke the > in-memory refcount information and thus rendered the BDS unusable > (because all of the qcow2 code assumes that the in-memory > information is correct, obviously) [Kevin] > - Do truncation last, because while it is the eventual goal, it is > also the structurally least invasive and therefore least important > operation (we don't want it to fail and leave the BDS unusable) > [Kevin]
Reviewed-by: Kevin Wolf <kw...@redhat.com>
