linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: "Michael S. Tsirkin" <mst@redhat.com>
To: Joerg Roedel <joro@8bytes.org>
Cc: Avi Kivity <avi@redhat.com>, Tom Lyon <pugs@cisco.com>,
	linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
	chrisw@sous-sol.org, hjk@linutronix.de, gregkh@suse.de,
	aafabbri@cisco.com, scofeldm@cisco.com
Subject: Re: [PATCH] VFIO driver: Non-privileged user level PCI drivers
Date: Wed, 2 Jun 2010 13:21:44 +0300	[thread overview]
Message-ID: <20100602102144.GD29023@redhat.com> (raw)
In-Reply-To: <20100602101940.GG964@8bytes.org>

On Wed, Jun 02, 2010 at 12:19:40PM +0200, Joerg Roedel wrote:
> On Wed, Jun 02, 2010 at 12:53:12PM +0300, Michael S. Tsirkin wrote:
> > On Wed, Jun 02, 2010 at 11:42:01AM +0200, Joerg Roedel wrote:
> 
> > > IMO a seperate iommu-userspace driver is a nightmare for a userspace
> > > interface. It is just too complicated to use.
> > 
> > One advantage would be that we can reuse the uio framework
> > for the devices themselves. So an existing app can just program
> > an iommu for DMA and keep using uio for interrupts and access.
> 
> The driver is called UIO and not U-INTR-MMIO ;-) So I think handling
> IOMMU mappings belongs there.
> 
> > > We can solve the problem
> > > of multiple devices-per-domain with an ioctl which allows binding one
> > > uio-device to the address-space on another.
> > 
> > This would imply switching an iommu domain for a device while
> > it could potentially be doing DMA. No idea whether this can be done
> > in a safe manner.
> 
> It can. The worst thing that can happen is an io-page-fault.

devices might not be able to recover from this.

> > Forcing iommu assignment to be done as a first step seems much saner.
> 
> If we force it, there is no reason why not doing it implicitly.

What you describe below does 3 ioctls for what can be done with 1.

> We can do something like this then:
> 
> dev1 = open();
> ioctl(dev1, IOMMU_MAP, ...); /* creates IOMMU domain and assigns dev1 to
> 			        it*/
> 
> dev2 = open();
> ioctl(dev2, IOMMU_MAP, ...);
> 
> /* Now dev1 and dev2 are in seperate domains */
> 
> ioctl(dev2, IOMMU_SHARE, dev1); /* destroys all mapping for dev2 and
> 				   assigns it to the same domain as
> 				   dev1. Domain has a refcount of two
> 				   now */

Or maybe it destroys mapping for dev1?
How do you remember?

> close(dev1); /* domain refcount goes down to one */
> close(dev2); /* domain refcount is zero and domain gets destroyed */
> 
> 
> 	Joerg

Also, no way to unshare? That seems limiting.

-- 
MST

  reply	other threads:[~2010-06-02 10:26 UTC|newest]

Thread overview: 66+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-05-28 23:07 [PATCH] VFIO driver: Non-privileged user level PCI drivers Tom Lyon
2010-05-28 23:36 ` Randy Dunlap
2010-05-28 23:56 ` Randy Dunlap
2010-05-29 11:55 ` Arnd Bergmann
2010-05-29 12:16   ` Avi Kivity
2010-05-30 12:19 ` Michael S. Tsirkin
2010-05-30 12:27   ` Avi Kivity
2010-05-30 12:49     ` Michael S. Tsirkin
2010-05-30 13:01       ` Avi Kivity
2010-05-30 13:03         ` Michael S. Tsirkin
2010-05-30 13:13           ` Avi Kivity
2010-05-30 14:53             ` Michael S. Tsirkin
2010-05-31 11:50               ` Avi Kivity
2010-05-31 17:10                 ` Michael S. Tsirkin
2010-06-01  8:10                   ` Avi Kivity
2010-06-01  9:55                     ` Michael S. Tsirkin
2010-06-01 10:28                       ` Avi Kivity
2010-06-01 10:46                         ` Michael S. Tsirkin
2010-06-01 12:41                           ` Avi Kivity
2010-06-02  9:45                             ` Joerg Roedel
2010-06-02  9:49                               ` Avi Kivity
2010-06-02 10:04                                 ` Joerg Roedel
2010-06-02 10:09                                   ` Michael S. Tsirkin
2010-06-02 11:21                                   ` Avi Kivity
2010-06-02 16:53                                     ` Chris Wright
2010-06-06 13:44                                       ` Avi Kivity
2010-06-02 10:15                               ` Michael S. Tsirkin
2010-06-02 10:26                                 ` Joerg Roedel
2010-06-01 21:26                           ` Tom Lyon
2010-06-02  2:59                             ` Avi Kivity
2010-06-02  5:29                               ` Chris Wright
2010-06-02  5:40                                 ` Avi Kivity
2010-06-02  4:29                         ` Alex Williamson
2010-06-02  4:59                           ` Tom Lyon
2010-06-02  5:08                             ` Avi Kivity
2010-06-02  9:53                             ` Joerg Roedel
2010-06-02  9:42                       ` Joerg Roedel
2010-06-02  9:50                         ` Avi Kivity
2010-06-02  9:53                         ` Michael S. Tsirkin
2010-06-02 10:19                           ` Joerg Roedel
2010-06-02 10:21                             ` Michael S. Tsirkin [this message]
2010-06-02 10:35                               ` Joerg Roedel
2010-06-02 10:38                                 ` Michael S. Tsirkin
2010-06-02 11:12                                   ` Joerg Roedel
2010-06-02 11:21                                     ` Michael S. Tsirkin
2010-06-02 12:19                                       ` Joerg Roedel
2010-06-02 12:25                                         ` Avi Kivity
2010-06-02 12:50                                           ` Joerg Roedel
2010-06-02 13:06                                             ` Avi Kivity
2010-06-02 13:53                                               ` Joerg Roedel
2010-06-02 13:17                                             ` Michael S. Tsirkin
2010-06-02 14:01                                               ` Joerg Roedel
2010-06-02 12:34                                         ` Michael S. Tsirkin
2010-06-02 13:02                                           ` Joerg Roedel
2010-06-02 17:46                                         ` Chris Wright
2010-06-02 18:09                                           ` Tom Lyon
2010-06-02 19:46                                             ` Joerg Roedel
2010-06-03  6:23                                           ` Avi Kivity
2010-06-03 21:41                                             ` Tom Lyon
2010-06-06  9:54                                               ` Michael S. Tsirkin
2010-06-07 19:01                                                 ` Tom Lyon
2010-06-08 21:22                                                   ` Michael S. Tsirkin
2010-06-02 10:44                             ` Michael S. Tsirkin
2010-05-30 12:59 ` Avi Kivity
2010-05-31 17:17 ` Alan Cox
2010-06-01 21:29   ` Tom Lyon

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=20100602102144.GD29023@redhat.com \
    --to=mst@redhat.com \
    --cc=aafabbri@cisco.com \
    --cc=avi@redhat.com \
    --cc=chrisw@sous-sol.org \
    --cc=gregkh@suse.de \
    --cc=hjk@linutronix.de \
    --cc=joro@8bytes.org \
    --cc=kvm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=pugs@cisco.com \
    --cc=scofeldm@cisco.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).