Re: [Qemu-devel] [PATCH] acpi: switch to a dummy SSDT

[Michael S. Tsirkin]

On Wed, Jan 10, 2018 at 12:41:52PM +0100, Igor Mammedov wrote:
> On Wed, 10 Jan 2018 13:29:09 +0200
> "Michael S. Tsirkin" <address@hidden> wrote:
> 
> > On Wed, Jan 10, 2018 at 12:25:34PM +0100, Igor Mammedov wrote:
> > > On Tue, 9 Jan 2018 15:58:11 +0200
> > > "Michael S. Tsirkin" <address@hidden> wrote:
> > > 
> > > > We prefer not changing table sizes depending on parameters,
> > > > that's why we create a dummy table rather than just drop
> > > > the MCFG table.
> > > > 
> > > > However, a table named "QEMU" could be put to a better
> > > > use than just a stub, e.g. we could use it to pass
> > > > some QEMU specific info to guests.
> > > > 
> > > > Replace with an SSDT and pad with the Nop opcode
> > > > to preserve the length.
> > > Not sure it's useful, do you have patches on top of that, which will use 
> > > this?
> > 
> > Yes but not me - Stefan (Cc'd) wants to add a custom table.
> Re-purposing dummy table that might become not dummy doesn't look like good 
> approach.
> What if later the guest enables MCFG and reboots, DUMMY SSDT would be 
> replaced by MCFG? 

So what's the issue?  Why is this worse than a dummy QEMU table?
There's actually no re-purposing here at all. Either we create
an MCFG or an SSDT. We never re-purpose.

> It'd be cleaner if Stefan would just use his own dedicated SSDT.

SSDT doesn't always work. A data table is needed so e.g. seabios
can process it easily.

> > > 
> > > > Signed-off-by: Michael S. Tsirkin <address@hidden>
> > > > ---
> > > >  hw/i386/acpi-build.c | 17 ++++++++++++-----
> > > >  1 file changed, 12 insertions(+), 5 deletions(-)
> > > > 
> > > > diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
> > > > index 73519ab..d0afb31e 100644
> > > > --- a/hw/i386/acpi-build.c
> > > > +++ b/hw/i386/acpi-build.c
> > > > @@ -2419,6 +2419,8 @@ build_mcfg_q35(GArray *table_data, BIOSLinker 
> > > > *linker, AcpiMcfgInfo *info)
> > > >  {
> > > >      AcpiTableMcfg *mcfg;
> > > >      const char *sig;
> > > > +    const char *oem;
> > > > +    const char *oem_id;
> > > >      int len = sizeof(*mcfg) + 1 * sizeof(mcfg->allocation[0]);
> > > >  
> > > >      mcfg = acpi_data_push(table_data, len);
> > > > @@ -2431,16 +2433,21 @@ build_mcfg_q35(GArray *table_data, BIOSLinker 
> > > > *linker, AcpiMcfgInfo *info)
> > > >      /* MCFG is used for ECAM which can be enabled or disabled by guest.
> > > >       * To avoid table size changes (which create migration issues),
> > > >       * always create the table even if there are no allocations,
> > > > -     * but set the signature to a reserved value in this case.
> > > > -     * ACPI spec requires OSPMs to ignore such tables.
> > > > +     * but fill it with Noop values.
> > > > +     * OSPMs ignore such tables.
> > > >       */
> > > >      if (info->mcfg_base == PCIE_BASE_ADDR_UNMAPPED) {
> > > > -        /* Reserved signature: ignored by OSPM */
> > > > -        sig = "QEMU";
> > > > +        sig = "SSDT";
> > > > +        oem = "QEMU ";
> > > > +        oem_id = "NOOP";
> > > > +        /* 0xa3 - NoopOp */
> > > > +        memset(&mcfg->reserved, 0xa3, len - offsetof(AcpiTableMcfg, 
> > > > reserved));
> > > >      } else {
> > > >          sig = "MCFG";
> > > > +        oem = NULL;
> > > > +        oem_id = NULL;
> > > >      }
> > > > -    build_header(linker, table_data, (void *)mcfg, sig, len, 1, NULL, 
> > > > NULL);
> > > > +    build_header(linker, table_data, (void *)mcfg, sig, len, 1, oem, 
> > > > oem_id);
> > > >  }
> > > >  
> > > >  /*