Re: [PATCH] xen: add kconfig for event_fifo

[Stefano Stabellini <sstabellini@xxxxxxxxxx>]

On Thu, 20 Mar 2025, Jan Beulich wrote:
> On 19.03.2025 00:40, Stefano Stabellini wrote:
> > Evtchn fifos are not needed on smaller systems; the older interface is
> > lightweight and sufficient. Make it possible to disable evtchn fifo.
> > 
> > Signed-off-by: Stefano Stabellini <stefano.stabellini@xxxxxxx>
> 
> Let me ask a more general question, before considering whether to ack:
> When kconfig was introduced into Xen, the expectation was (iirc) that
> we wouldn't grow its use to the fine grained granularity that it has
> been having in Linux. The more build-time controls we have, the harder
> it is to know whether a certain piece of code was actually included in
> a build someone, say, reports a problem with. Are we knowingly moving
> away from that earlier position? evtchn_fifo.c isn't actually that
> much code to exclude, after all ...


I think we need to be more flexible with build-time controls. The extent
to which we should be flexible is an interesting discussion to have.

However, this patch is not for code size. This code causes unnecessary
runtime anonymous memory allocations, which are highly undesirable. In
fact, it is the primary source of such allocations. Additionally, it
exposes an extra interface to the guest, which is also undesirable
unless necessary.

In other words, the issue is not the size of the code, but its impact.