Re: [PATCH v2] x86/HVM: fix forwarding of internally cached requests

From: "Li, Liang Z" <liang.z.li@intel.com>
To: Jan Beulich <JBeulich@suse.com>, jzh Chang <changjzh@gmail.com>,
	"Xu, Quan" <quan.xu@intel.com>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>,
	Paul Durrant <paul.durrant@citrix.com>,
	Keir Fraser <keir@xen.org>,
	xen-devel <xen-devel@lists.xenproject.org>
Subject: Re: [PATCH v2] x86/HVM: fix forwarding of internally cached requests
Date: Mon, 25 Apr 2016 08:40:54 +0000	[thread overview]
Message-ID: <F2CBF3009FA73547804AE4C663CAB28E0418589C@shsmsx102.ccr.corp.intel.com> (raw)
In-Reply-To: <5719EC3602000078000E486B@prv-mh.provo.novell.com>

> >> >>> On 30.03.16 at 09:28, <changjzh@gmail.com> wrote:
> >> > 2016-03-29 18:39 GMT+08:00 Jan Beulich <JBeulich@suse.com>:
> >> >> ---
> >> >> I assume this also addresses the issue which
> >> >>
> >> http://lists.xenproject.org/archives/html/xen-devel/2016-01/msg03189.
> >> html
> >> >> attempted to deal with in a not really acceptable way.
> >> >
> >> > I hope this issue is resolved, but it still exists.
> >>
> >> Mind giving the small below patch a try?
> >>
> >> Jan
> >>
> >> --- unstable.orig/xen/arch/x86/msi.c
> >> +++ unstable/xen/arch/x86/msi.c
> >> @@ -430,8 +430,13 @@ static bool_t msi_set_mask_bit(struct ir
> >>          {
> >>              writel(flag, entry->mask_base +
> >> PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET);
> >>              readl(entry->mask_base +
> >> PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET);
> >> +
> >>              if ( likely(control & PCI_MSIX_FLAGS_ENABLE) )
> >>                  break;
> >> +
> >> +            entry->msi_attrib.host_masked = host;
> >> +            entry->msi_attrib.guest_masked = guest;
> >> +
> >>              flag = 1;
> >>          }
> >>          else if ( flag && !(control & PCI_MSIX_FLAGS_MASKALL) )
> >>
> >>
> >> The issue still exist. But, the host_masked is changed.
> 
> At least something.
> 
> > guest_masked can be changed by guest_mask_msi_irq() function.
> > The function is not called as previous e-mail analysis.
> 
> I have to admit that I had quite a bit of trouble understanding that previous
> patch of yours. The function not being called of course needs to be
> understood, which requires a trace of the writes of the guest to the vector
> control field(s), including the ones before the MSI-X region gets registered.
> Just to double check - was this latest try with the other patch also in place, or
> just the small one I had sent yesterday?
> 

With other patches also in place, still not work. 
Jianzhong  has been left and Quan will take over the task.

Liang

> I do have a debugging patch around for the necessary logging to get added,
> but that's against quite a bit older a hypervisor version, so likely would serve
> only as a reference. Let me know if you would still like me to hand that to you.
> 
> Jan
> 
> > No patch xen log message:
> > (XEN)  MSI-X  114 vec=73 lowest  edge   assert  log lowest dest=00000555
> > mask=1/HG/1
> > (XEN)  MSI-X  115 vec=7b lowest  edge   assert  log lowest dest=00000555
> > mask=1/HG/1
> > (XEN)  MSI-X  116 vec=83 lowest  edge   assert  log lowest dest=00000555
> > mask=1/HG/1
> >
> > Patched-xen log message :
> > (XEN)  MSI-X  114 vec=76 lowest  edge   assert  log lowest dest=00000555
> > mask=1/ G/1
> > (XEN)  MSI-X  115 vec=7e lowest  edge   assert  log lowest dest=00000555
> > mask=1/ G/1
> > (XEN)  MSI-X  116 vec=86 lowest  edge   assert  log lowest dest=00000555
> > mask=1/ G/1
> >
> > --
> > Jianzhong,Chang
> 
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel