All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH v2] x86/vlapic: accept a logical destination of all ones as a broadcast
@ 2026-11-17 23:36 Stefano Stabellini
  2026-11-18  8:46 ` Jan Beulich
                   ` (2 more replies)
  0 siblings, 3 replies; 5+ messages in thread
From: Stefano Stabellini @ 2026-11-17 23:36 UTC (permalink / raw)
  To: xen-devel
  Cc: stefano.stabellini, Jan Beulich, Andrew Cooper,
	Roger Pau Monné, Teddy Astie

In logical destination mode, a destination of all ones is a broadcast
accepted by every local APIC: 0xff in xAPIC mode, in both the flat and
the cluster model (AMD APM vol. 2, "Receiving System and IPI
Interrupts"), and 0xffffffff in x2APIC mode (AMD APM vol. 2, "x2APIC
Interrupt Command Register (ICR) Operations").

vlapic_match_logical_addr() treated these as ordinary logical
destinations, so a broadcast reached only the vLAPICs whose logical ID
happened to match it. In particular, a vLAPIC whose logical ID is 0
received no logical broadcast at all.

Check for the broadcast in vlapic_match_dest() before telling the two
destination modes apart, as was already done for physical destinations.
A vLAPIC whose destination format register names neither model now
also accepts the broadcast.

Fixes: f9e0cccf7b35 ("x86/HVM: fix ID handling of x2APIC emulation")
Signed-off-by: Stefano Stabellini <stefano.stabellini@amd.com>
---
CC: Jan Beulich <jbeulich@suse.com>,
CC: Andrew Cooper <andrew.cooper3@citrix.com>,
CC: Roger Pau Monné <roger@xenproject.org>,
CC: Teddy Astie <teddy.astie@vates.tech>

v2:
 - Check for the broadcast once in vlapic_match_dest(), before the two
   destination modes are told apart, rather than in each model of
   vlapic_match_logical_addr() (Jan). As a result, a logical broadcast
   now also matches a vLAPIC whose DFR names neither model.
 - Cite the AMD APM by section title (Jan).
 - Add Fixes: tags.

 xen/arch/x86/hvm/vlapic.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/xen/arch/x86/hvm/vlapic.c b/xen/arch/x86/hvm/vlapic.c
index 6698779e22..50b515d254 100644
--- a/xen/arch/x86/hvm/vlapic.c
+++ b/xen/arch/x86/hvm/vlapic.c
@@ -247,10 +247,12 @@ bool vlapic_match_dest(
     switch ( short_hand )
     {
     case APIC_DEST_NOSHORT:
+        /* All ones is a broadcast in either destination mode. */
+        if ( dest == _VLAPIC_ID(target, 0xffffffffU) )
+            return true;
         if ( dest_mode )
             return vlapic_match_logical_addr(target, dest);
-        return (dest == _VLAPIC_ID(target, 0xffffffffU)) ||
-               (dest == VLAPIC_ID(target));
+        return dest == VLAPIC_ID(target);
 
     case APIC_DEST_SELF:
         return (target == source);
-- 
2.47.3



^ permalink raw reply related	[flat|nested] 5+ messages in thread

* Re: [PATCH v2] x86/vlapic: accept a logical destination of all ones as a broadcast
  2026-11-17 23:36 [PATCH v2] x86/vlapic: accept a logical destination of all ones as a broadcast Stefano Stabellini
@ 2026-11-18  8:46 ` Jan Beulich
  2026-11-18  9:31 ` Teddy Astie
  2026-11-19  9:52 ` Jan Beulich
  2 siblings, 0 replies; 5+ messages in thread
From: Jan Beulich @ 2026-11-18  8:46 UTC (permalink / raw)
  To: Stefano Stabellini
  Cc: Andrew Cooper, Roger Pau Monné, Teddy Astie, xen-devel

On 18.11.2026 00:36, Stefano Stabellini wrote:
> In logical destination mode, a destination of all ones is a broadcast
> accepted by every local APIC: 0xff in xAPIC mode, in both the flat and
> the cluster model (AMD APM vol. 2, "Receiving System and IPI
> Interrupts"), and 0xffffffff in x2APIC mode (AMD APM vol. 2, "x2APIC
> Interrupt Command Register (ICR) Operations").
> 
> vlapic_match_logical_addr() treated these as ordinary logical
> destinations, so a broadcast reached only the vLAPICs whose logical ID
> happened to match it. In particular, a vLAPIC whose logical ID is 0
> received no logical broadcast at all.
> 
> Check for the broadcast in vlapic_match_dest() before telling the two
> destination modes apart, as was already done for physical destinations.
> A vLAPIC whose destination format register names neither model now
> also accepts the broadcast.

May be worth expanding this a little to indicate that other DFR values
are undefined as to behavior.

> Fixes: f9e0cccf7b35 ("x86/HVM: fix ID handling of x2APIC emulation")
> Signed-off-by: Stefano Stabellini <stefano.stabellini@amd.com>

Reviewed-by: Jan Beulich <jbeulich@suse.com>



^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH v2] x86/vlapic: accept a logical destination of all ones as a broadcast
  2026-11-17 23:36 [PATCH v2] x86/vlapic: accept a logical destination of all ones as a broadcast Stefano Stabellini
  2026-11-18  8:46 ` Jan Beulich
@ 2026-11-18  9:31 ` Teddy Astie
  2026-11-19  9:52 ` Jan Beulich
  2 siblings, 0 replies; 5+ messages in thread
From: Teddy Astie @ 2026-11-18  9:31 UTC (permalink / raw)
  To: Stefano Stabellini, xen-devel
  Cc: Jan Beulich, Andrew Cooper, Roger Pau Monné

[-- Attachment #1: Type: text/plain, Size: 2963 bytes --]

Le 18/11/2026 à 00:39, Stefano Stabellini a écrit :
> In logical destination mode, a destination of all ones is a broadcast
> accepted by every local APIC: 0xff in xAPIC mode, in both the flat and
> the cluster model (AMD APM vol. 2, "Receiving System and IPI
> Interrupts"), and 0xffffffff in x2APIC mode (AMD APM vol. 2, "x2APIC
> Interrupt Command Register (ICR) Operations").
> 

Intel SDM also appears to say the same.
For flat model
 > A broadcast to all APICs is achieved by setting the MDA to 1s.
(MDA is "message destination address")

There are many occurences in SDM that all 1s destination is broadcast.

With a additional note that Intel follows the same behavior :
Reviewed-by: Teddy Astie <teddy.astie@vates.tech>

> vlapic_match_logical_addr() treated these as ordinary logical
> destinations, so a broadcast reached only the vLAPICs whose logical ID
> happened to match it. In particular, a vLAPIC whose logical ID is 0
> received no logical broadcast at all.
> 
> Check for the broadcast in vlapic_match_dest() before telling the two
> destination modes apart, as was already done for physical destinations.
> A vLAPIC whose destination format register names neither model now
> also accepts the broadcast.
> 
> Fixes: f9e0cccf7b35 ("x86/HVM: fix ID handling of x2APIC emulation")
> Signed-off-by: Stefano Stabellini <stefano.stabellini@amd.com>
> ---
> CC: Jan Beulich <jbeulich@suse.com>,
> CC: Andrew Cooper <andrew.cooper3@citrix.com>,
> CC: Roger Pau Monné <roger@xenproject.org>,
> CC: Teddy Astie <teddy.astie@vates.tech>
> 
> v2:
>   - Check for the broadcast once in vlapic_match_dest(), before the two
>     destination modes are told apart, rather than in each model of
>     vlapic_match_logical_addr() (Jan). As a result, a logical broadcast
>     now also matches a vLAPIC whose DFR names neither model.
>   - Cite the AMD APM by section title (Jan).
>   - Add Fixes: tags.
> 
>   xen/arch/x86/hvm/vlapic.c | 6 ++++--
>   1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/xen/arch/x86/hvm/vlapic.c b/xen/arch/x86/hvm/vlapic.c
> index 6698779e22..50b515d254 100644
> --- a/xen/arch/x86/hvm/vlapic.c
> +++ b/xen/arch/x86/hvm/vlapic.c
> @@ -247,10 +247,12 @@ bool vlapic_match_dest(
>       switch ( short_hand )
>       {
>       case APIC_DEST_NOSHORT:
> +        /* All ones is a broadcast in either destination mode. */
> +        if ( dest == _VLAPIC_ID(target, 0xffffffffU) )
> +            return true;
>           if ( dest_mode )
>               return vlapic_match_logical_addr(target, dest);
> -        return (dest == _VLAPIC_ID(target, 0xffffffffU)) ||
> -               (dest == VLAPIC_ID(target));
> +        return dest == VLAPIC_ID(target);
>   
>       case APIC_DEST_SELF:
>           return (target == source);



--
Teddy Astie | Vates XCP-ng Developer

XCP-ng & Xen Orchestra - Vates solutions

web: https://vates.tech

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH v2] x86/vlapic: accept a logical destination of all ones as a broadcast
  2026-11-17 23:36 [PATCH v2] x86/vlapic: accept a logical destination of all ones as a broadcast Stefano Stabellini
  2026-11-18  8:46 ` Jan Beulich
  2026-11-18  9:31 ` Teddy Astie
@ 2026-11-19  9:52 ` Jan Beulich
  2026-11-19 21:26   ` Stefano Stabellini
  2 siblings, 1 reply; 5+ messages in thread
From: Jan Beulich @ 2026-11-19  9:52 UTC (permalink / raw)
  To: Stefano Stabellini
  Cc: Andrew Cooper, Roger Pau Monné, Teddy Astie, xen-devel

On 18.11.2026 00:36, Stefano Stabellini wrote:
> In logical destination mode, a destination of all ones is a broadcast
> accepted by every local APIC: 0xff in xAPIC mode, in both the flat and
> the cluster model (AMD APM vol. 2, "Receiving System and IPI
> Interrupts"), and 0xffffffff in x2APIC mode (AMD APM vol. 2, "x2APIC
> Interrupt Command Register (ICR) Operations").
> 
> vlapic_match_logical_addr() treated these as ordinary logical
> destinations, so a broadcast reached only the vLAPICs whose logical ID
> happened to match it. In particular, a vLAPIC whose logical ID is 0
> received no logical broadcast at all.
> 
> Check for the broadcast in vlapic_match_dest() before telling the two
> destination modes apart, as was already done for physical destinations.
> A vLAPIC whose destination format register names neither model now
> also accepts the broadcast.
> 
> Fixes: f9e0cccf7b35 ("x86/HVM: fix ID handling of x2APIC emulation")
> Signed-off-by: Stefano Stabellini <stefano.stabellini@amd.com>
> ---
> CC: Jan Beulich <jbeulich@suse.com>,
> CC: Andrew Cooper <andrew.cooper3@citrix.com>,
> CC: Roger Pau Monné <roger@xenproject.org>,
> CC: Teddy Astie <teddy.astie@vates.tech>
> 
> v2:
>  - Check for the broadcast once in vlapic_match_dest(), before the two
>    destination modes are told apart, rather than in each model of
>    vlapic_match_logical_addr() (Jan). As a result, a logical broadcast
>    now also matches a vLAPIC whose DFR names neither model.
>  - Cite the AMD APM by section title (Jan).
>  - Add Fixes: tags.
> 
>  xen/arch/x86/hvm/vlapic.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/xen/arch/x86/hvm/vlapic.c b/xen/arch/x86/hvm/vlapic.c
> index 6698779e22..50b515d254 100644
> --- a/xen/arch/x86/hvm/vlapic.c
> +++ b/xen/arch/x86/hvm/vlapic.c
> @@ -247,10 +247,12 @@ bool vlapic_match_dest(
>      switch ( short_hand )
>      {
>      case APIC_DEST_NOSHORT:
> +        /* All ones is a broadcast in either destination mode. */
> +        if ( dest == _VLAPIC_ID(target, 0xffffffffU) )

Turns out a very similar patch (referencing the SDM rather than the APM)
was sent before, and Andrew pointed out an issue there [1]: This needs to
use the sender, not the receiver, to determine whether broadcast is meant.

Jan

[1] https://lore.kernel.org/xen-devel/95e0e0fb-2978-4c87-b723-fb7c30f36883@citrix.com/

> +            return true;
>          if ( dest_mode )
>              return vlapic_match_logical_addr(target, dest);
> -        return (dest == _VLAPIC_ID(target, 0xffffffffU)) ||
> -               (dest == VLAPIC_ID(target));
> +        return dest == VLAPIC_ID(target);
>  
>      case APIC_DEST_SELF:
>          return (target == source);



^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH v2] x86/vlapic: accept a logical destination of all ones as a broadcast
  2026-11-19  9:52 ` Jan Beulich
@ 2026-11-19 21:26   ` Stefano Stabellini
  0 siblings, 0 replies; 5+ messages in thread
From: Stefano Stabellini @ 2026-11-19 21:26 UTC (permalink / raw)
  To: Jan Beulich
  Cc: Stefano Stabellini, Andrew Cooper, Roger Pau Monné,
	Teddy Astie, xen-devel

[-- Attachment #1: Type: text/plain, Size: 2939 bytes --]

On Thu, 19 Nov 2026, Jan Beulich wrote:
> On 18.11.2026 00:36, Stefano Stabellini wrote:
> > In logical destination mode, a destination of all ones is a broadcast
> > accepted by every local APIC: 0xff in xAPIC mode, in both the flat and
> > the cluster model (AMD APM vol. 2, "Receiving System and IPI
> > Interrupts"), and 0xffffffff in x2APIC mode (AMD APM vol. 2, "x2APIC
> > Interrupt Command Register (ICR) Operations").
> > 
> > vlapic_match_logical_addr() treated these as ordinary logical
> > destinations, so a broadcast reached only the vLAPICs whose logical ID
> > happened to match it. In particular, a vLAPIC whose logical ID is 0
> > received no logical broadcast at all.
> > 
> > Check for the broadcast in vlapic_match_dest() before telling the two
> > destination modes apart, as was already done for physical destinations.
> > A vLAPIC whose destination format register names neither model now
> > also accepts the broadcast.
> > 
> > Fixes: f9e0cccf7b35 ("x86/HVM: fix ID handling of x2APIC emulation")
> > Signed-off-by: Stefano Stabellini <stefano.stabellini@amd.com>
> > ---
> > CC: Jan Beulich <jbeulich@suse.com>,
> > CC: Andrew Cooper <andrew.cooper3@citrix.com>,
> > CC: Roger Pau Monné <roger@xenproject.org>,
> > CC: Teddy Astie <teddy.astie@vates.tech>
> > 
> > v2:
> >  - Check for the broadcast once in vlapic_match_dest(), before the two
> >    destination modes are told apart, rather than in each model of
> >    vlapic_match_logical_addr() (Jan). As a result, a logical broadcast
> >    now also matches a vLAPIC whose DFR names neither model.
> >  - Cite the AMD APM by section title (Jan).
> >  - Add Fixes: tags.
> > 
> >  xen/arch/x86/hvm/vlapic.c | 6 ++++--
> >  1 file changed, 4 insertions(+), 2 deletions(-)
> > 
> > diff --git a/xen/arch/x86/hvm/vlapic.c b/xen/arch/x86/hvm/vlapic.c
> > index 6698779e22..50b515d254 100644
> > --- a/xen/arch/x86/hvm/vlapic.c
> > +++ b/xen/arch/x86/hvm/vlapic.c
> > @@ -247,10 +247,12 @@ bool vlapic_match_dest(
> >      switch ( short_hand )
> >      {
> >      case APIC_DEST_NOSHORT:
> > +        /* All ones is a broadcast in either destination mode. */
> > +        if ( dest == _VLAPIC_ID(target, 0xffffffffU) )
> 
> Turns out a very similar patch (referencing the SDM rather than the APM)
> was sent before, and Andrew pointed out an issue there [1]: This needs to
> use the sender, not the receiver, to determine whether broadcast is meant.
> 
> Jan
> 
> [1] https://lore.kernel.org/xen-devel/95e0e0fb-2978-4c87-b723-fb7c30f36883@citrix.com/

I'll send out a new patch to address this



> > +            return true;
> >          if ( dest_mode )
> >              return vlapic_match_logical_addr(target, dest);
> > -        return (dest == _VLAPIC_ID(target, 0xffffffffU)) ||
> > -               (dest == VLAPIC_ID(target));
> > +        return dest == VLAPIC_ID(target);
> >  
> >      case APIC_DEST_SELF:
> >          return (target == source);
> 

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2026-11-19 21:26 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-17 23:36 [PATCH v2] x86/vlapic: accept a logical destination of all ones as a broadcast Stefano Stabellini
2026-11-18  8:46 ` Jan Beulich
2026-11-18  9:31 ` Teddy Astie
2026-11-19  9:52 ` Jan Beulich
2026-11-19 21:26   ` Stefano Stabellini

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.