Re: [PATCH V5 04/22] xen/ioreq: Make x86's IOREQ feature common

[Oleksandr <olekstysh@xxxxxxxxx>]

On 27.01.21 22:46, Stefano Stabellini wrote:

Hi Stefano, all

> On Wed, 27 Jan 2021, Oleksandr wrote:
> > > > > On Mon, 25 Jan 2021 at 19:09, Oleksandr Tyshchenko <olekstysh@xxxxxxxxx>
> > > > > wrote:
> > > > > > ***
> > > > > > Please note, this patch depends on the following which is
> > > > > > on review:
> > > > > > https://patchwork.kernel.org/patch/11816689/
> > > > > > The effort (to get it upstreamed) was paused because of
> > > > > > the security issue around that code (XSA-348).
> > > > > > ***
> > > > > I read this comment as "This series should be applied on top the patch
> > > > > X". However, looking at your branch, I can't see the patch. What did I
> > > > > miss?
> > > > You didn't miss anything. Patch series doesn't contain it. I mentioned
> > > > about this patch in order not to forget about it
> > > > and draw reviewer's attention. Looks like, the activity (to get it
> > > > upstreamed) hasn't been resumed yet and I don't know what we should do
> > > > with that dependency
> > > > in the context of this series...
> > > I will reply what I wrote on IRC privately :). I think this fine to defer it
> > > to after Xen 4.15 because IOREQ will be in tech preview.
> > Thank you. I got a request to make a possibility for user to select IOREQ via
> > the menuconfig on Arm. Saying tech preview do you mean that I also need to put
> > it under CONFIG_EXPERT on Arm?
> Normally I would say that unless it actually takes an expert to enable
> the feature, it is better to make it depend on CONFIG_UNSUPPORTED [1].
>
> However, in this case, it might actually take an expert :-)
> Additional patches are still required to enable the feature at the
> toolstack level, and also the user needs to build a userspace ioreq
> server. So in this case I am fine either way. I'll leave it up to you
> and Julien to pick the best one.
>
> Either way, please add "(EXPERT)" or "(UNSUPPORTED)" in the one-line
> kconfig description for ARM if possible.
>
>
> [1] https://marc.info/?l=xen-devel&m=161168780401884
Thank you for the explanation, personally I would prefer EXPERT, but...

It would be really nice if we could agreed regarding that *common* 
Kconfig option before I submit V6 (which would probably avoid me to send 
V7 I hope).
Now I see it as following (please note, it is a subject for new patch in 
this series, which should come last):
+++ b/xen/common/Kconfig
@@ -137,7 +137,13 @@ config HYPFS_CONFIG
          want to hide the .config contents from dom0.

 config IOREQ_SERVER
-       bool
+       bool "IOREQ support" if EXPERT || X86
+       default X86
+       depends on HVM
+       ---help---
+         Enables generic mechanism for providing emulated devices to 
the guests.
+
+         If unsure, say Y.

 config KEXEC
        bool "kexec support"


Looks like, it works this way for both archs, but for me it looks a bit ugly

I would appreciate your input on that.


--
Regards,

Oleksandr Tyshchenko