On Wed, Aug 16, 2017 at 06:59:25AM -0500, Eric Blake wrote:
On 08/16/2017 04:41 AM, Manos Pitsidianakis wrote:+## +# @block-insert-node: +# +# Insert a filter node between a specific edge in the block driver state graph. +# @parent: the name of the parent node or device +# @node: the name of the node to insert under parent +# @child: the name of the child of both node and parentIs this always going to be between two existing nodes, or can this command also be used to insert at the end of the chain (for example, if parent or child is omitted)?If this is used for filter nodes, I suppose only between would make sense (for now). Is there a use case for the latter?Perhaps. Given a qcow2 image backing chain: base <- active there are four BDS (2 format, 2 protocol). Ideally, I could add filtering to any one of those four nodes (a filter on the base protocol level restricts how much guest data can be used from the backing image, but with no limits on the qcow2 metadata; a filter on the base format level restricts metadata reads as well; similarly for filters on the active protocol and format layers). But adding a filter on 'active' at the format level has no pre-existing parent (I'm adding the filter as the new top-level). Or am I missing something?
The parent in this case is the storage device (disk / cdrom), whose name is specified as the parent. The first example in the qapi/block-core.json is such a case. In code I check blk_by_name(parent) and if that doesn't exist, I try with bdrv_find_node(parent). Perhaps I should reword the documentation or did I misunderstand what you wrote?
signature.asc
Description: PGP signature
