Re: [Xen-devel] [PATCH v5 6/9] ioreq-server: add support for multiple servers

[Paul Durrant <Paul.Durrant@xxxxxxxxxx>]

> -----Original Message-----
> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> Sent: 07 May 2014 15:08
> To: Paul Durrant
> Cc: Ian Campbell; Ian Jackson; Stefano Stabellini; xen-devel@xxxxxxxxxxxxx
> Subject: RE: [PATCH v5 6/9] ioreq-server: add support for multiple servers
> 
> >>> On 07.05.14 at 14:37, <Paul.Durrant@xxxxxxxxxx> wrote:
> >> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> >> >>> On 07.05.14 at 14:25, <Paul.Durrant@xxxxxxxxxx> wrote:
> >> >> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> >> >> >>> On 07.05.14 at 14:06, <Paul.Durrant@xxxxxxxxxx> wrote:
> >> >> >> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> >> >> >> >>> On 01.05.14 at 14:08, <paul.durrant@xxxxxxxxxx> wrote:
> >> >> >> > +static int hvm_access_cfc(
> >> >> >> > +    int dir, uint32_t port, uint32_t bytes, uint32_t *val)
> >> >> >> > +{
> >> >> >> > +    struct vcpu *curr = current;
> >> >> >> > +    struct hvm_domain *hd = &curr->domain->arch.hvm_domain;
> >> >> >> > +    int rc;
> >> >> >> > +
> >> >> >> > +    BUG_ON(port < 0xcfc);
> >> >> >> > +    port -= 0xcfc;
> >> >> >> > +
> >> >> >> > +    spin_lock(&hd->pci_lock);
> >> >> >> > +
> >> >> >> > +    if ( hd->pci_cf8 & (1 << 31) ) {
> >> >> >> > +        /* Fall through to an emulator */
> >> >> >> > +        rc = X86EMUL_UNHANDLEABLE;
> >> >> >> > +    } else {
> >> >> >> > +        /* Config access disabled */
> >> >> >>
> >> >> >> Why does this not also get passed through to an emulator?
> >> >> >>
> >> >> >
> >> >> > I was trying to be consistent with QEMU here. It squashes any data
> >> >> accesses
> >> >> > if cf8 has the top bit set.
> >> >>
> >> >> But afaict with that dropped the entire function can go away.
> >> >>
> >> >
> >> > Yes, it can. Do you not think it would be a good idea to be consistent
> with
> >> > QEMU though?
> >>
> >> By removing the function you're going to be consistent with qemu,
> >> because you're going to have qemu supply the data.
> >>
> >
> > Yes, true, in select_ioreq_server() I could just pass cfc accesses to the
> > default emulator if cf8 has the top bit set.
> 
> I continue to fail to see why you make this dependent on the value
> of the top bit - just have any accesses to port 0xCFC be handled by
> the emulator responsible for the respective device.

Because, according to what I read, unless cf8 has the top bit set, a subsequent 
access to cfc should not be treated as a config cycle.

  Paul


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel