Am 14.03.2016 um 17:55 hat Max Reitz geschrieben: > On 14.03.2016 12:31, Kevin Wolf wrote: > > Am 08.03.2016 um 17:34 hat Kevin Wolf geschrieben: > >> All users of the block layers are supposed to go through a BlockBackend. > >> The .bdrv_create() implementations are such users, so this series > >> converts them. > >> > >> This series (specifically patch 1) will also help with moving the > >> writethrough cache implementation from BDS to BB, where it really > >> belongs. Once this is moved, it wouldn't be possible to use bdrv_open() > >> to use a standalone BDS writethrough any more. > >> > >> Of course, writethrough doesn't make any sense in .bdrv_create() anyway, > >> so while I'm still converting everything to BB where it would keep > >> working (because that's the Right Thing), the drivers don't actually > >> make use of this fact any more after this series. > > > > Applied to the block branch. > > Now you can't name any BB "image" any more if you want to use e.g. > drive-mirror. Intended?
Not really. :-/ Of course, you're supposed to use blockdev-mirror anyway (does it even exist?), but we should avoid that. Fortunately, your "Further BB work" will tie the name to a monitor reference and BBs without a name become valid then, I think. So this might be a good excuse for getting that series into 2.6. :-) Kevin > Example: > > $ x86_64-softmmu/qemu-system-x86_64 -drive > if=none,id=image,file=null-co://,driver=raw -qmp stdio > {'execute':'qmp_capabilities'} > {"return": {}} > {'execute':'drive-mirror','arguments':{'device':'image','target':'foo.qcow2','sync':'full','format':'qcow2'}} > Formatting 'foo.qcow2', fmt=qcow2 size=1073741824 encryption=off > cluster_size=65536 lazy_refcounts=off refcount_bits=16 > {"error": {"class": "GenericError", "desc": "Device with id 'image' > already exists"}} > > Max >
pgpENkeXDygE5.pgp
Description: PGP signature