Re: [Qemu-devel] [PATCH v3 3/5] qcow2: Introduce an option for sufficient L2 cache for the entire image

[Kevin Wolf <kwolf@redhat.com>]

Am 26.07.2018 um 14:24 hat Leonid Bloch geschrieben:
> > > You mean with QDict? I'll look into that now. But already sent v5 before
> > > reading this email.
> > 
> > Yes, with reading it from the QDict. (Or whatever the simplest way is
> > that results in the right external interface, but I suppose this is the
> > one.)
> 
> Well, there is a problem with that: I can easily isolate
> l2-cache-size from QDict, check if it is "full", and if it is - do whatever
> is needed, and delete this option before parsing. But what if it is "foo"?
> It will not get deleted, and the regular QEMU_OPT_SIZE parsing error will
> appear, stating that l2-cache-size "expects a non-negative number..." - no
> word about that it can expect "full" as well. Now, one can try to modify
> local_err->msg for this particular option, but this will require substantial
> additional logic. I think considering this, it would be easier to stick with
> a dedicated option, l2-cache-full.
> 
> Do you think there is a smarter way to parse the l2-cache-size option, so it
> would accept both size and "full", while handling errors correctly? It seems
> more elegant to have a single option, but the internal handling will be more
> elegant and simpler with two mutually exclusive options.

I think we can live with the suboptimal error message for a while. Once
qcow2 is QAPIfied, it should become easy to improve it. Let's not choose
a worse design (that stays forever) for a temporarily better error
message.

> By the way, the L2 cache resizes now on image resize. Will send the changes
> in v6. Thanks for the suggestion!

Sounds good!

Kevin