Re: [Qemu-devel] [SeaBIOS] [PATCH v2 4/5] acpi: load and link tables from /etc/acpi/

["Michael S. Tsirkin" <mst@redhat.com>]

On Mon, Jul 15, 2013 at 11:11:59AM +0300, Michael S. Tsirkin wrote:
> On Sun, Jul 14, 2013 at 02:27:14PM -0400, Kevin O'Connor wrote:
> > On Sun, Jul 07, 2013 at 06:42:43PM +0300, Michael S. Tsirkin wrote:
> > > Load files in /etc/acpi/, link them using
> > > a linker script and use for acpi tables, including the RSDP.
> > > Presense of RSDP in this directory completely disables
> > > generating and loading legacy acpi tables.
> > > 
> > > Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> > > ---
> > >  src/acpi.c | 39 +++++++++++++++++++++++++++++++++++++++
> > >  1 file changed, 39 insertions(+)
> > > 
> > > diff --git a/src/acpi.c b/src/acpi.c
> > > index a81f0cb..07d311a 100644
> > > --- a/src/acpi.c
> > > +++ b/src/acpi.c
> > > @@ -27,6 +27,7 @@
> > >  #include "config.h" // CONFIG_*
> > >  #include "paravirt.h" // RamSize
> > >  #include "dev-q35.h"
> > > +#include "linker.h"
> > >  
> > >  #include "acpi-dsdt.hex"
> > >  
> > > @@ -599,6 +600,34 @@ static const struct pci_device_id acpi_find_tbl[] = {
> > >  
> > >  struct rsdp_descriptor *RsdpAddr;
> > >  
> > > +/* Look for RSDP signature in files loaded in FSEG */
> > > +struct rsdp_descriptor *
> > > +acpi_find_rsdp_rom(void)
> > > +{
> > > +    struct romfile_s *file = NULL;
> > > +    struct rsdp_descriptor *rsdp = NULL;
> > > +    for (;;) {
> > > +        file = romfile_findprefix("", file);
> > > +        if (!file)
> > > +            break;
> > > +
> > > +        if (!file->data || !pmm_test_fseg(file->data) ||
> > > +            file->size < sizeof(rsdp->signature))
> > > +            continue;
> > > +
> > > +        void *data;
> > > +
> > > +        for (data = file->data;
> > > +             data + sizeof(*rsdp) <= file->data + file->size;
> > > +             data++) {
> > > +            rsdp = data;
> > > +            if (rsdp->signature == cpu_to_le64(RSDP_SIGNATURE))
> > > +                return rsdp;
> > > +        }
> > > +    }
> > > +    return NULL;
> > > +}
> > > +
> > >  #define MAX_ACPI_TABLES 20
> > >  void
> > >  acpi_setup(void)
> > > @@ -608,6 +637,16 @@ acpi_setup(void)
> > >  
> > >      dprintf(3, "init ACPI tables\n");
> > >  
> > > +    linker_loader_execute("/etc/linker-script");
> > > +
> > > +    RsdpAddr = acpi_find_rsdp_rom();
> > > +
> > > +    if (RsdpAddr) {
> > > +        return;
> > > +    }
> > 
> > I don't understand this.  Why not use the presence of
> > "/etc/linker-script" to determine if ACPI should be produced.
> 
> We can do this but what do we do if linker script is there
> but no rsdp?
> Maybe just add
> 	test -e /etc/linker-script != RsdpAddr
> 	        dprintf(1, "linker script found but no rsdp. Falling back on builtin tables\n");
> 
> Would this address your comment?
> 

And just to clarify, in my eyes it's preferable to use
acpi_find_rsdp_rom which is based on ACPI standard than
look for a specific PV interface.

For example, one can imagine a hypervisor that maps
ACPI tables directly into guest memory. Looking for
rsdp in fseg memory will work, relying on the
linker won't.

> > Also, /etc/linker-script is not very informative - how about something
> > like "/etc/biostables".
> > 
> > -Kevin
> 
> Hmm bios seems redundant, and tables makes one think it's the tables
> themselves.
> 
> "/etc/acpi-linker"
> "/etc/acpi-loader"?
> 
> 
> -- 
> MST