Re: [PATCH v3 09/21] quorum: Add QuorumChild.to_be_replaced

[Kevin Wolf]

Am 30.01.2020 um 22:44 hat Max Reitz geschrieben:
> We will need this to verify that Quorum can let one of its children be
> replaced without breaking anything else.
> 
> Signed-off-by: Max Reitz <address@hidden>
> ---
>  block/quorum.c | 25 +++++++++++++++++++++++++
>  1 file changed, 25 insertions(+)
> 
> diff --git a/block/quorum.c b/block/quorum.c
> index 59cd524502..6a7224c9e4 100644
> --- a/block/quorum.c
> +++ b/block/quorum.c
> @@ -67,6 +67,13 @@ typedef struct QuorumVotes {
>  
>  typedef struct QuorumChild {
>      BdrvChild *child;
> +
> +    /*
> +     * If set, check whether this node can be replaced without any
> +     * other parent noticing: Unshare CONSISTENT_READ, and take the
> +     * WRITE permission.
> +     */
> +    bool to_be_replaced;

I don't understand these permission changes. How does (preparing for)
detaching a node from quorum make its content invalid? And why do we
suddenly need WRITE permissions even if the quorum node is only used
read-only?

The comment is a bit unclear, too. "check whether" implies that both
outcomes could be true, but it doesn't say what happens in either case.
Is this really "make sure that"?

Kevin