From: Joerg Roedel <joro@8bytes.org>
To: Will Deacon <will.deacon@arm.com>
Cc: "iommu@lists.linux-foundation.org"
<iommu@lists.linux-foundation.org>, Kukjin Kim <kgene@kernel.org>,
David Woodhouse <dwmw2@infradead.org>,
Heiko Stuebner <heiko@sntech.de>, Hiroshi Doyu <hdoyu@nvidia.com>,
Thierry Reding <thierry.reding@gmail.com>,
Alex Williamson <alex.williamson@redhat.com>,
Arnd Bergmann <arnd@arndb.de>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
Robin Murphy <Robin.Murphy@arm.com>,
Laurent Pinchart <laurent.pinchart@ideasonboard.com>,
"jroedel@suse.de" <jroedel@suse.de>
Subject: Re: [PATCH 2/5] iommu: Allocate a default domain for iommu groups
Date: Fri, 30 Jan 2015 13:25:30 +0100 [thread overview]
Message-ID: <20150130122530.GB3702@8bytes.org> (raw)
In-Reply-To: <20150128143006.GQ1569@arm.com>
On Wed, Jan 28, 2015 at 02:30:06PM +0000, Will Deacon wrote:
> On Tue, Jan 27, 2015 at 12:08:56AM +0000, Joerg Roedel wrote:
> > + if (group->default_domain == NULL)
> > + group->default_domain = __iommu_domain_alloc(dev->bus,
> > + IOMMU_DOMAIN_DMA);
>
> Having a per-group domain is wasteful for IOMMUs that only support a modest
> number of concurrent domains, so in reality I think we need to have one
> domain per IOMMU instance. Is that possible?
Well, you could make sure that there are no more groups behind one IOMMU
than the number of concurent domains it can handle. But that would be
too static. But once we have an per-iommu-descriptor in the IOMMU core I
see no reason to allocate a default domain per iommu only, based on a
policy exported by the driver.
Joerg
WARNING: multiple messages have this Message-ID (diff)
From: Joerg Roedel <joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org>
To: Will Deacon <will.deacon-5wv7dgnIgG8@public.gmane.org>
Cc: "jroedel-l3A5Bk7waGM@public.gmane.org"
<jroedel-l3A5Bk7waGM@public.gmane.org>,
Laurent Pinchart
<laurent.pinchart-ryLnwIuWjnjg/C1BVhZhaw@public.gmane.org>,
Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org>,
Arnd Bergmann <arnd-r2nGTMty4D4@public.gmane.org>,
"linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org"
<linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>,
"iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org"
<iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org>,
Thierry Reding
<thierry.reding-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>,
Kukjin Kim <kgene-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>,
David Woodhouse <dwmw2-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>
Subject: Re: [PATCH 2/5] iommu: Allocate a default domain for iommu groups
Date: Fri, 30 Jan 2015 13:25:30 +0100 [thread overview]
Message-ID: <20150130122530.GB3702@8bytes.org> (raw)
In-Reply-To: <20150128143006.GQ1569-5wv7dgnIgG8@public.gmane.org>
On Wed, Jan 28, 2015 at 02:30:06PM +0000, Will Deacon wrote:
> On Tue, Jan 27, 2015 at 12:08:56AM +0000, Joerg Roedel wrote:
> > + if (group->default_domain == NULL)
> > + group->default_domain = __iommu_domain_alloc(dev->bus,
> > + IOMMU_DOMAIN_DMA);
>
> Having a per-group domain is wasteful for IOMMUs that only support a modest
> number of concurrent domains, so in reality I think we need to have one
> domain per IOMMU instance. Is that possible?
Well, you could make sure that there are no more groups behind one IOMMU
than the number of concurent domains it can handle. But that would be
too static. But once we have an per-iommu-descriptor in the IOMMU core I
see no reason to allocate a default domain per iommu only, based on a
policy exported by the driver.
Joerg
next prev parent reply other threads:[~2015-01-30 12:25 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-01-27 0:08 [RFC PATCH 0/5] iommu: Introduce default domains for iommu groups Joerg Roedel
2015-01-27 0:08 ` Joerg Roedel
2015-01-27 0:08 ` [PATCH 1/5] iommu: Add default domain to iommu-groups Joerg Roedel
2015-01-27 0:08 ` Joerg Roedel
2015-01-27 0:08 ` [PATCH 2/5] iommu: Allocate a default domain for iommu groups Joerg Roedel
2015-01-27 0:08 ` Joerg Roedel
2015-01-28 14:30 ` Will Deacon
2015-01-28 14:30 ` Will Deacon
2015-01-28 15:11 ` Robin Murphy
2015-01-28 15:11 ` Robin Murphy
2015-01-30 12:25 ` Joerg Roedel [this message]
2015-01-30 12:25 ` Joerg Roedel
2015-01-27 0:08 ` [PATCH 3/5] iommu: Limit iommu_attach/detach_device to devices with their own group Joerg Roedel
2015-01-27 0:08 ` Joerg Roedel
2015-01-28 14:35 ` Will Deacon
2015-01-28 14:35 ` Will Deacon
2015-01-30 12:28 ` Joerg Roedel
2015-01-30 12:28 ` Joerg Roedel
2015-02-02 16:45 ` Will Deacon
2015-02-02 16:45 ` Will Deacon
2015-02-03 12:25 ` Thierry Reding
2015-02-03 12:25 ` Thierry Reding
2015-02-03 12:59 ` Joerg Roedel
2015-02-03 12:59 ` Joerg Roedel
2015-01-27 0:08 ` [PATCH 4/5] iommu: Make sure a device is always attached to a domain Joerg Roedel
2015-01-27 0:08 ` Joerg Roedel
2015-01-28 14:38 ` Will Deacon
2015-01-28 14:38 ` Will Deacon
2015-01-30 12:29 ` Joerg Roedel
2015-01-30 12:29 ` Joerg Roedel
2015-01-27 0:08 ` [PATCH 5/5] iommu: Add iommu_get_domain_for_dev function Joerg Roedel
2015-01-27 0:08 ` Joerg Roedel
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20150130122530.GB3702@8bytes.org \
--to=joro@8bytes.org \
--cc=Robin.Murphy@arm.com \
--cc=alex.williamson@redhat.com \
--cc=arnd@arndb.de \
--cc=dwmw2@infradead.org \
--cc=hdoyu@nvidia.com \
--cc=heiko@sntech.de \
--cc=iommu@lists.linux-foundation.org \
--cc=jroedel@suse.de \
--cc=kgene@kernel.org \
--cc=laurent.pinchart@ideasonboard.com \
--cc=linux-kernel@vger.kernel.org \
--cc=thierry.reding@gmail.com \
--cc=will.deacon@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.