On Wed, Jul 25, 2018 at 06:23:45PM +0300, Leonid Bloch wrote: > On 07/25/2018 04:32 PM, Kevin Wolf wrote: > > > Another interesting question is whether 'full' shouldn't keep > > > meaning > > > full throughout the lifetime of the BlockDriverState, i.e. > > > should it > > > keep adapting to the new size when the image size changes? > > > > > > > > > Do we even resize the cache now for image size changes? If we use an > > > enum, > > > we could have two different values depending on whether the chosen > > > cache > > > size remains fixed or also tries to resize when the image grows. > > > > We don't because we only support absolute cache sizes today. 'full' > > would be the first one that is relative to the image size. > > > > > Is it even possible to change the virtual disk image size online? > > > > Yes, this is what qcow2_co_truncate() does (can be invoked, amongst > > others, with the QMP command 'block_resize'). > > Cool! This does look like a good idea to resize the L2 cache accordingly, > but maybe this is out of scope for now as well? The purpose of the current > series is just to provide an option to automatically calculate the needed L2 > cache size for covering the entire image, instead of using an external > script to do that and feed the output to l2-cache-size.
Personally if I saw the description of the option and then found it didn't "do the right thing" when resizing I'd consider the option broken. So I think we should make it deal with resizes right away rather than implementing a known bug. Regards, Daniel -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|