WARNING - OLD ARCHIVES

This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
   
 
 
Xen 
 
Home Products Support Community News
 
   
 

xen-devel

[Xen-devel] Re: [PATCH] Enable PCI passthrough with stub domain.

To: "Li, Xin" <xin.li@xxxxxxxxx>, Shohei Fujiwara <fujiwara-sxa@xxxxxxxxxxxxxxx>, "xen-devel@xxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxx>
Subject: [Xen-devel] Re: [PATCH] Enable PCI passthrough with stub domain.
From: Keir Fraser <keir.fraser@xxxxxxxxxxxxx>
Date: Fri, 09 Jan 2009 15:43:08 +0000
Cc:
Delivery-date: Fri, 09 Jan 2009 07:43:29 -0800
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
In-reply-to: <E88DD564E9DC5446A76B2B47C3BCCA15434E9DEC@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>
List-help: <mailto:xen-devel-request@lists.xensource.com?subject=help>
List-id: Xen developer discussion <xen-devel.lists.xensource.com>
List-post: <mailto:xen-devel@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=unsubscribe>
Sender: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
Thread-index: AclyTaE9qFgbXqJVME+VL5aTsIesSQAEWqigAAR7Cko=
Thread-topic: [PATCH] Enable PCI passthrough with stub domain.
User-agent: Microsoft-Entourage/12.14.0.081024
Possibly, but not necessarily. Regardless of that, there's been discussion
about moving responsibility for management of PCI device space forwarding
into pciback (out of xend).

It's not crucial to my criticism of Shohei's patch though: number one issue
is that passthrough-setup operations such as permitting access to iomem and
ioport space should *not* be done in stubdom. I think it'd be better off in
pciback, but xend could perhaps work okay too. Either way the privilege
escalation of the domU is being instigated within dom0 rather than from
stubdom (which we don't want to trust any more than the domU itself).

 -- Keir

On 09/01/2009 13:36, "Li, Xin" <xin.li@xxxxxxxxx> wrote:

> Keir, are you suggesting to have pcifront in stub domain?
> -Xin
> 
>> -----Original Message-----
>> From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
>> [mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Keir Fraser
>> Sent: Friday, January 09, 2009 7:30 PM
>> To: Shohei Fujiwara; xen-devel@xxxxxxxxxxxxxxxxxxx
>> Subject: [Xen-devel] Re: [PATCH] Enable PCI passthrough with stub domain.
>> 
>> One major point of stubdomains is to deprivilege qemu. If we give it access
>> to domctls like {memory,ioport}_mapping it's rather defeating that purpose.
>> Perhaps some thought is needed as to how PCI passthru is architected -- the
>> actual setting up of access permissions (which is implicitly what some of
>> those domctls do) should be done by the toolstack and/or pciback, both of
>> which continue to reside in dom0.
>> 
>> -- Keir
>> 
>> On 09/01/2009 10:08, "Shohei Fujiwara" <fujiwara-sxa@xxxxxxxxxxxxxxx> wrote:
>> 
>>> This patch enables PCI passthrough with stub domain.
>>> 
>>> PCI passthrough with stub domain has failed in the past.
>>> The primary reason is that hypercalls from qemu in stub domain are rejected.
>>> 
>>> This patch allows qemu in stub domain to call the hypercalls which is
>>> needed for PCI passthrough. For security, if target domain of
>>> hypercall is different from that of stub domain, it rejects hypercall.
>>> 
>>> To use PCI passthrough, please configure pciback driver to be
>>> permissive, because qemu has own virtualization code for PCI
>>> passthrough. For example:
>>> 
>>>     # cd /sys/bus/pci/drivers/pciback
>>>     # echo "0000:0d:00.0" > permissive
>>> 
>>> Currently Linux guest only works.
>>> But Windows guest doesn't work, so I'm debugging for now.
>>> I hope this patch will be useful for other developers in community.
>>> 
>>> Thanks,
>>> --
>>> Shohei Fujiwara.
>>> 
>>> 
>>> Signed-off-by: Shohei Fujiwara <fujiwara-sxa@xxxxxxxxxxxxxxx>
>>> 
>>> diff -r b9721b2766c1 xen/arch/x86/irq.c
>>> --- a/xen/arch/x86/irq.c Wed Jan 07 11:25:00 2009 +0000
>>> +++ b/xen/arch/x86/irq.c Thu Jan 08 18:58:36 2009 +0900
>>> @@ -857,7 +857,7 @@ int map_domain_pirq(
>>>      if ( type == MAP_PIRQ_TYPE_MSI )
>>>          return -EINVAL;
>>> 
>>> -    if ( !IS_PRIV(current->domain) )
>>> +    if ( !IS_PRIV_FOR(current->domain, d) )
>>>          return -EPERM;
>>> 
>>>      if ( pirq < 0 || pirq >= NR_IRQS || vector < 0 || vector >= NR_VECTORS
>>> )
>>> @@ -931,7 +931,7 @@ int unmap_domain_pirq(struct domain *d,
>>>      if ( (pirq < 0) || (pirq >= NR_IRQS) )
>>>          return -EINVAL;
>>> 
>>> -    if ( !IS_PRIV(current->domain) )
>>> +    if ( !IS_PRIV_FOR(current->domain, d) )
>>>          return -EINVAL;
>>> 
>>>      ASSERT(spin_is_locked(&pcidevs_lock));
>>> diff -r b9721b2766c1 xen/arch/x86/physdev.c
>>> --- a/xen/arch/x86/physdev.c Wed Jan 07 11:25:00 2009 +0000
>>> +++ b/xen/arch/x86/physdev.c Thu Jan 08 18:58:36 2009 +0900
>>> @@ -34,9 +34,6 @@ static int physdev_map_pirq(struct physd
>>>      struct msi_info _msi;
>>>      void *map_data = NULL;
>>> 
>>> -    if ( !IS_PRIV(current->domain) )
>>> -        return -EPERM;
>>> -
>>>      if ( !map )
>>>          return -EINVAL;
>>> 
>>> @@ -49,6 +46,14 @@ static int physdev_map_pirq(struct physd
>>>      {
>>>          ret = -ESRCH;
>>>          goto free_domain;
>>> +    }
>>> +    else
>>> +    {
>>> +        if ( !IS_PRIV_FOR(current->domain, d) )
>>> +        {
>>> +            ret = -EPERM;
>>> +            goto free_domain;
>>> +        }
>>>      }
>>> 
>>>      /* Verify or get vector. */
>>> @@ -161,8 +166,8 @@ static int physdev_unmap_pirq(struct phy
>>>      struct domain *d;
>>>      int ret;
>>> 
>>> -    if ( !IS_PRIV(current->domain) )
>>> -        return -EPERM;
>>> +    if ( !unmap )
>>> +        return -EINVAL;
>>> 
>>>      if ( unmap->domid == DOMID_SELF )
>>>          d = rcu_lock_domain(current->domain);
>>> @@ -171,6 +176,14 @@ static int physdev_unmap_pirq(struct phy
>>> 
>>>      if ( d == NULL )
>>>          return -ESRCH;
>>> +    else
>>> +    {
>>> +        if ( !IS_PRIV_FOR(current->domain, d) )
>>> +        {
>>> +            rcu_unlock_domain(d);
>>> +            return -EPERM;
>>> +        }
>>> +    }
>>> 
>>>      spin_lock(&pcidevs_lock);
>>>      spin_lock(&d->event_lock);
>>> diff -r b9721b2766c1 xen/common/domctl.c
>>> --- a/xen/common/domctl.c Wed Jan 07 11:25:00 2009 +0000
>>> +++ b/xen/common/domctl.c Thu Jan 08 18:58:36 2009 +0900
>>> @@ -193,14 +193,49 @@ long do_domctl(XEN_GUEST_HANDLE(xen_domc
>>>      long ret = 0;
>>>      struct xen_domctl curop, *op = &curop;
>>> 
>>> -    if ( !IS_PRIV(current->domain) )
>>> -        return -EPERM;
>>> -
>>>      if ( copy_from_guest(op, u_domctl, 1) )
>>>          return -EFAULT;
>>> 
>>>      if ( op->interface_version != XEN_DOMCTL_INTERFACE_VERSION )
>>>          return -EACCES;
>>> +
>>> +    switch ( op->cmd )
>>> +    {
>>> +    case XEN_DOMCTL_pin_mem_cacheattr:
>>> +    case XEN_DOMCTL_memory_mapping:
>>> +    case XEN_DOMCTL_ioport_mapping:
>>> +    case XEN_DOMCTL_assign_device:
>>> +    case XEN_DOMCTL_deassign_device:
>>> +    case XEN_DOMCTL_bind_pt_irq:
>>> +    case XEN_DOMCTL_unbind_pt_irq:
>>> +    {
>>> +        struct domain *d;
>>> +
>>> +        if ( op->domain != DOMID_SELF )
>>> +        {
>>> +            d = rcu_lock_domain_by_id(op->domain);
>>> +
>>> +            if ( d == NULL )
>>> +                return -ESRCH;
>>> +            else
>>> +            {
>>> +                if ( !IS_PRIV_FOR(current->domain, d) )
>>> +                {
>>> +                    rcu_unlock_domain(d);
>>> +                    return -EPERM;
>>> +                }
>>> +                else
>>> +                    rcu_unlock_domain(d);
>>> +            }
>>> +            break;
>>> +        }
>>> +    }
>>> +
>>> +    default:
>>> +        if ( !IS_PRIV(current->domain) )
>>> +            return -EPERM;
>>> +        break;
>>> +    }
>>> 
>>>      spin_lock(&domctl_lock);
>>> 
>> 
>> 
>> 
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@xxxxxxxxxxxxxxxxxxx
>> http://lists.xensource.com/xen-devel



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel