Re: [Xen-devel] [PATCH 2/4] iommu: generalize iommu_inclusive_mapping

[Roger Pau Monné <roger.pau@xxxxxxxxxx>]

On Tue, Jul 31, 2018 at 03:14:47AM -0600, Jan Beulich wrote:
> >>> On 31.07.18 at 11:05, <roger.pau@xxxxxxxxxx> wrote:
> > On Tue, Jul 31, 2018 at 02:49:19AM -0600, Jan Beulich wrote:
> >> >>> On 31.07.18 at 10:37, <Paul.Durrant@xxxxxxxxxx> wrote:
> >> >>  -----Original Message-----
> >> >> From: Roger Pau Monne
> >> >> Sent: 31 July 2018 09:34
> >> >> To: Paul Durrant <Paul.Durrant@xxxxxxxxxx>
> >> >> Cc: xen-devel@xxxxxxxxxxxxxxxxxxxx; Kevin Tian <kevin.tian@xxxxxxxxx>;
> >> >> Stefano Stabellini <sstabellini@xxxxxxxxxx>; Wei Liu 
> >> >> <wei.liu2@xxxxxxxxxx>;
> >> >> George Dunlap <George.Dunlap@xxxxxxxxxx>; Andrew Cooper
> >> >> <Andrew.Cooper3@xxxxxxxxxx>; Ian Jackson <Ian.Jackson@xxxxxxxxxx>; Tim
> >> >> (Xen.org) <tim@xxxxxxx>; Julien Grall <julien.grall@xxxxxxx>; Jan 
> >> >> Beulich
> >> >> <jbeulich@xxxxxxxx>
> >> >> Subject: Re: [Xen-devel] [PATCH 2/4] iommu: generalize
> >> >> iommu_inclusive_mapping
> >> >> 
> >> >> On Tue, Jul 31, 2018 at 09:27:03AM +0100, Paul Durrant wrote:
> >> >> > > -----Original Message-----
> >> >> > > From: Roger Pau Monne
> >> >> > > Sent: 31 July 2018 09:16
> >> >> > > To: Paul Durrant <Paul.Durrant@xxxxxxxxxx>
> >> >> > > Cc: xen-devel@xxxxxxxxxxxxxxxxxxxx; Kevin Tian 
> >> >> > > <kevin.tian@xxxxxxxxx>;
> >> >> > > Stefano Stabellini <sstabellini@xxxxxxxxxx>; Wei Liu
> >> >> <wei.liu2@xxxxxxxxxx>;
> >> >> > > George Dunlap <George.Dunlap@xxxxxxxxxx>; Andrew Cooper
> >> >> > > <Andrew.Cooper3@xxxxxxxxxx>; Ian Jackson <Ian.Jackson@xxxxxxxxxx>;
> >> >> Tim
> >> >> > > (Xen.org) <tim@xxxxxxx>; Julien Grall <julien.grall@xxxxxxx>; Jan
> >> >> Beulich
> >> >> > > <jbeulich@xxxxxxxx>
> >> >> > > Subject: Re: [Xen-devel] [PATCH 2/4] iommu: generalize
> >> >> > > iommu_inclusive_mapping
> >> >> > >
> >> >> > > On Tue, Jul 31, 2018 at 08:18:36AM +0100, Paul Durrant wrote:
> >> >> > > > > -----Original Message-----
> >> >> > > > > From: Xen-devel [mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxxx]
> >> >> On
> >> >> > > Behalf
> >> >> > > > > Of Roger Pau Monne
> >> >> > > > > Sent: 27 July 2018 16:32
> >> >> > > > > To: xen-devel@xxxxxxxxxxxxxxxxxxxx 
> >> >> > > > > Cc: Kevin Tian <kevin.tian@xxxxxxxxx>; Stefano Stabellini
> >> >> > > > > <sstabellini@xxxxxxxxxx>; Wei Liu <wei.liu2@xxxxxxxxxx>; George
> >> >> Dunlap
> >> >> > > > > <George.Dunlap@xxxxxxxxxx>; Andrew Cooper
> >> >> > > > > <Andrew.Cooper3@xxxxxxxxxx>; Ian Jackson
> >> >> <Ian.Jackson@xxxxxxxxxx>;
> >> >> > > Tim
> >> >> > > > > (Xen.org) <tim@xxxxxxx>; Julien Grall <julien.grall@xxxxxxx>; 
> >> >> > > > > Jan
> >> >> > > Beulich
> >> >> > > > > <jbeulich@xxxxxxxx>; Roger Pau Monne <roger.pau@xxxxxxxxxx>
> >> >> > > > > Subject: [Xen-devel] [PATCH 2/4] iommu: generalize
> >> >> > > > > iommu_inclusive_mapping
> >> >> > > > >
> >> >> > > > > Introduce a new iommu=inclusive generic option that supersedes
> >> >> > > > > iommu_inclusive_mapping. This should be a non-functional change
> >> >> on
> >> >> > > > > Intel hardware, while AMD hardware will gain the same 
> >> >> > > > > functionality
> >> >> of
> >> >> > > > > mapping almost everything below the 4GB boundary.
> >> >> > > > >
> >> >> > > > > Note that is a noop for ARM hardware.
> >> >> > > > >
> >> >> > > > > Signed-off-by: Roger Pau Monné <roger.pau@xxxxxxxxxx>
> >> >> > > > > ---
> >> >> > > > > Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
> >> >> > > > > Cc: George Dunlap <George.Dunlap@xxxxxxxxxxxxx>
> >> >> > > > > Cc: Ian Jackson <ian.jackson@xxxxxxxxxxxxx>
> >> >> > > > > Cc: Jan Beulich <jbeulich@xxxxxxxx>
> >> >> > > > > Cc: Julien Grall <julien.grall@xxxxxxx>
> >> >> > > > > Cc: Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>
> >> >> > > > > Cc: Stefano Stabellini <sstabellini@xxxxxxxxxx>
> >> >> > > > > Cc: Tim Deegan <tim@xxxxxxx>
> >> >> > > > > Cc: Wei Liu <wei.liu2@xxxxxxxxxx>
> >> >> > > > > Cc: Kevin Tian <kevin.tian@xxxxxxxxx>
> >> >> > > > > ---
> >> >> > > > >  docs/misc/xen-command-line.markdown   | 14 ++++++
> >> >> > > > >  xen/drivers/passthrough/arm/iommu.c   |  4 ++
> >> >> > > > >  xen/drivers/passthrough/iommu.c       |  6 +++
> >> >> > > > >  xen/drivers/passthrough/vtd/extern.h  |  2 -
> >> >> > > > >  xen/drivers/passthrough/vtd/iommu.c   |  6 ---
> >> >> > > > >  xen/drivers/passthrough/vtd/x86/vtd.c | 66 
> >> >> > > > > +------------------------
> >> >> > > > >  xen/drivers/passthrough/x86/iommu.c   | 70
> >> >> > > > > +++++++++++++++++++++++++++
> >> >> > > > >  xen/include/xen/iommu.h               |  2 +
> >> >> > > > >  8 files changed, 97 insertions(+), 73 deletions(-)
> >> >> > > > >
> >> >> > > > > diff --git a/docs/misc/xen-command-line.markdown
> >> >> b/docs/misc/xen-
> >> >> > > > > command-line.markdown
> >> >> > > > > index 65b4754418..91a8bfc9a6 100644
> >> >> > > > > --- a/docs/misc/xen-command-line.markdown
> >> >> > > > > +++ b/docs/misc/xen-command-line.markdown
> >> >> > > > > @@ -1198,6 +1198,17 @@ detection of systems known to misbehave
> >> >> > > upon
> >> >> > > > > accesses to that port.
> >> >> > > > >
> >> >> > > > >  >> Enable IOMMU debugging code (implies `verbose`).
> >> >> > > > >
> >> >> > > > > +> `inclusive`
> >> >> > > >
> >> >> > > > This is a dom0 (or hwdom) specific setting so perhaps 
> >> >> > > > dom0-inclusive?
> >> >> > > >
> >> >> > > > Actually the dom0 iommu options are starting to get unwieldy as 
> >> >> > > > they
> >> >> are
> >> >> > > conflated with the general host iommu options so I think it may be
> >> >> > > worthwhile splitting things out into a separate 'dom0-iommu=' top 
> >> >> > > level
> >> >> > > parameter at this stage. (My reasons are slightly selfish as I 
> >> >> > > intend to 
> >> > add
> >> >> > > another dom0 iommu option to give it just reserved regions, to avoid
> >> >> > > unnecessary set-up if we know it will be using PV-IOMMU).
> >> >> > >
> >> >> > > Mapping just the reserved regions is what I actually do for PVH with
> >> >> > > iommu=inclusive (patch 4/4), so maybe it would make sense to speak
> >> >> about
> >> >> > > the
> >> >> > > naming here in order to use the same naming for PV and PVH.
> >> >> > >
> >> >> > > TBH I don't really like the dom0- prefix, the command line iommu
> >> >> > > options either apply to all domains or Dom0 only, having
> >> >> > > domu-inclusive for example makes no sense IMO.
> >> >> >
> >> >> > No, I think there are some options that you may want to apply to dom0
> >> >> only, but these are more like the dom0_mem or dom0_max_vpus options.
> >> >> Particularly, the inclusive option is probably something that is only 
> > desirable
> >> >> for dom0. Clearly dom0-passthrough and dom0-strict are already defined 
> >> >> to
> >> >> relate to dom0 only, and options such as 'reserved' should only be 
> >> >> specific 
> > on
> >> >> the command line in relation to dom0 IMO. For other domains such an 
> >> >> option
> >> >> should be specified via xl.cfg.
> >> >> 
> >> >> Yes, we already have a bunch of those, so then I think dom0-inclusive
> >> >> and dom0-reserved would be appropriate?
> >> >> 
> >> >> dom0-inclusive-mapping or dom0-reserved-mapping seems too long.
> >> > 
> >> > Yes, those names are ok, but I still think it better in the long run if 
> >> > we 
> >> > have something like:
> >> > 
> >> > dom0_iommu=[inclusive,][reserved,][strict,][none,][relaxed]
> >> > 
> >> > where relaxed is the default and 'none' (I think) is equivalent to the 
> >> > current iommu=dom0-passthrough.
> >> 
> >> Or, along the lines of the other reply just sent, e.g.
> >> 
> >> dom0=pvh,iommu:inclusive;reserved,shadow
> >> 
> >> But perhaps the difference between , and ; gets too confusing then.
> > 
> > So I think we have the following options:
> > 
> > 1. dom0_iommu=[inclusive,][reserved,][strict,][none,][relaxed]
> 
> Nit: dom0-iommu= (no underscores in new options)
> 
> > 2. 
> > dom0=[pvh,][shadow,][iommu=[inclusive;][reserved;][strict;][none;][relaxed]]
> > 3. 
> > dom0=[pvh,][shadow,][iommu-inclusive,][iommu-reserved,][iommu-strict,][iommu-none,][iommu-relaxed]
> > 
> > I don't have a strong preference between 1 and 3, but I would prefer
> > to avoid 2 because I think suboptions inside of options it's too
> > complex IMO.
> 
> While generally I prefer to limit the number of top level options, in
> this case I think I'd prefer 1 after all. Or wait - does any pair of the
> (sub)options actually make sense to be specified?

Yes, for example you can use strict and inclusive at the same time, I
think it's something like:

dom0=[pvh,][shadow,][iommu=[inclusive|reserved;][strict|none|relaxed]]

> Isn't it rather a
> choice of five than an enumeration of up to 5? In which case I'd
> still prefer 2 (as then there's no need for a second separator
> beside comma), the more that we have at least one example with
> such sub-options (cpufreq).

OK, I can do the nested iommu option inside of dom0 if that's the
preference.

Thanks, Roger.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel