[Qemu-devel] [RfC PATCH 0/2] seabios: going to 256k size

[Qemu-devel] [RfC PATCH 0/2] seabios: going to 256k size

[Gerd Hoffmann]

  Hi,

With the seabios update to the next master branch relase (NOT the stable
branch pull sent out earlier today) seabios rom size will cross the 128k
boundary and we have to deal with that.

So here is a RfC patch series with one aproach to handle this: We'll go
build two seabios binaries, one 128k (will less features), one 256k
(full featured).  Machine types 1.7+ will use the 256k bios while older
machine types will continue to use the 128k sized one.

Series applies on top of the "roms: various build improvements" pull
request sent out earlier today.

Comments?

cheers,
  Gerd

Gerd Hoffmann (2):
  roms: build two seabios binaries
  machine type dependant bios

 hw/i386/pc_piix.c        | 1 +
 hw/i386/pc_q35.c         | 1 +
 hw/i386/pc_sysfw.c       | 4 ++--
 include/hw/i386/pc.h     | 2 ++
 roms/Makefile            | 7 ++++---
 roms/config.seabios      | 1 -
 roms/config.seabios-128k | 5 +++++
 roms/config.seabios-256k | 3 +++
 8 files changed, 18 insertions(+), 6 deletions(-)
 delete mode 100644 roms/config.seabios
 create mode 100644 roms/config.seabios-128k
 create mode 100644 roms/config.seabios-256k

-- 
1.8.3.1

[Qemu-devel] [PATCH 1/2] roms: build two seabios binaries

[Gerd Hoffmann]

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
 roms/Makefile            | 7 ++++---
 roms/config.seabios      | 1 -
 roms/config.seabios-128k | 5 +++++
 roms/config.seabios-256k | 3 +++
 4 files changed, 12 insertions(+), 4 deletions(-)
 delete mode 100644 roms/config.seabios
 create mode 100644 roms/config.seabios-128k
 create mode 100644 roms/config.seabios-256k

diff --git a/roms/Makefile b/roms/Makefile
index 10d5a65..1d84230 100644
--- a/roms/Makefile
+++ b/roms/Makefile
@@ -56,9 +56,10 @@ default:
 	@echo "                    the EfiRom utility from edk2 / tianocore)"
 	@echo "  slof           -- update slof.bin"
 
-bios: build-seabios-config-seabios
-	cp seabios/builds/seabios/bios.bin ../pc-bios/bios.bin
-	cp seabios/builds/seabios/*dsdt.aml ../pc-bios/
+bios: build-seabios-config-seabios-128k build-seabios-config-seabios-256k
+	cp seabios/builds/seabios-128k/bios.bin ../pc-bios/bios-128k.bin
+	cp seabios/builds/seabios-256k/bios.bin ../pc-bios/bios-256k.bin
+	cp seabios/builds/seabios-256k/src/fw/*dsdt.aml ../pc-bios/
 
 seavgabios: $(patsubst %,seavgabios-%,$(vgabios_variants))
 
diff --git a/roms/config.seabios b/roms/config.seabios
deleted file mode 100644
index c373b87..0000000
--- a/roms/config.seabios
+++ /dev/null
@@ -1 +0,0 @@
-# empty, default config works for us
diff --git a/roms/config.seabios-128k b/roms/config.seabios-128k
new file mode 100644
index 0000000..23ca812
--- /dev/null
+++ b/roms/config.seabios-128k
@@ -0,0 +1,5 @@
+# for qemu machine types 1.6 + older
+# need to turn off features (xhci) to make it fit into 128k
+CONFIG_QEMU=y
+CONFIG_ROM_SIZE=128
+CONFIG_USB_XHCI=n
diff --git a/roms/config.seabios-256k b/roms/config.seabios-256k
new file mode 100644
index 0000000..cc37a78
--- /dev/null
+++ b/roms/config.seabios-256k
@@ -0,0 +1,3 @@
+# for qemu machine types 1.7 + newer
+CONFIG_QEMU=y
+CONFIG_ROM_SIZE=256
-- 
1.8.3.1

[Qemu-devel] [PATCH 2/2] machine type dependant bios

[Gerd Hoffmann]

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
 hw/i386/pc_piix.c    | 1 +
 hw/i386/pc_q35.c     | 1 +
 hw/i386/pc_sysfw.c   | 4 ++--
 include/hw/i386/pc.h | 2 ++
 4 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/hw/i386/pc_piix.c b/hw/i386/pc_piix.c
index 907792b..35188a5 100644
--- a/hw/i386/pc_piix.c
+++ b/hw/i386/pc_piix.c
@@ -238,6 +238,7 @@ static void pc_init_pci(QEMUMachineInitArgs *args)
 
 static void pc_compat_1_6(QEMUMachineInitArgs *args)
 {
+    pc_default_bios_name = "bios-128k.bin";
     has_pci_info = false;
     rom_file_in_ram = false;
 }
diff --git a/hw/i386/pc_q35.c b/hw/i386/pc_q35.c
index ca84e1c..e8b0168 100644
--- a/hw/i386/pc_q35.c
+++ b/hw/i386/pc_q35.c
@@ -222,6 +222,7 @@ static void pc_q35_init(QEMUMachineInitArgs *args)
 
 static void pc_compat_1_6(QEMUMachineInitArgs *args)
 {
+    pc_default_bios_name = "bios-128k.bin";
     has_pci_info = false;
     rom_file_in_ram = false;
 }
diff --git a/hw/i386/pc_sysfw.c b/hw/i386/pc_sysfw.c
index e917c83..deb8869 100644
--- a/hw/i386/pc_sysfw.c
+++ b/hw/i386/pc_sysfw.c
@@ -34,7 +34,7 @@
 #include "hw/block/flash.h"
 #include "sysemu/kvm.h"
 
-#define BIOS_FILENAME "bios.bin"
+const char *pc_default_bios_name = "bios-256k.bin";
 
 typedef struct PcSysFwDevice {
     SysBusDevice busdev;
@@ -112,7 +112,7 @@ static void old_pc_system_rom_init(MemoryRegion *rom_memory, bool isapc_ram_fw)
 
     /* BIOS load */
     if (bios_name == NULL) {
-        bios_name = BIOS_FILENAME;
+        bios_name = pc_default_bios_name;
     }
     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
     if (filename) {
diff --git a/include/hw/i386/pc.h b/include/hw/i386/pc.h
index 9b2ddc4..cf9e47e 100644
--- a/include/hw/i386/pc.h
+++ b/include/hw/i386/pc.h
@@ -12,6 +12,8 @@
 
 /* PC-style peripherals (also used by other machines).  */
 
+extern const char *pc_default_bios_name;
+
 typedef struct PcPciInfo {
     Range w32;
     Range w64;
-- 
1.8.3.1

Re: [Qemu-devel] [PATCH 1/2] roms: build two seabios binaries

[Paolo Bonzini]

Il 30/09/2013 12:46, Gerd Hoffmann ha scritto:
> Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
> ---
>  roms/Makefile            | 7 ++++---
>  roms/config.seabios      | 1 -
>  roms/config.seabios-128k | 5 +++++
>  roms/config.seabios-256k | 3 +++
>  4 files changed, 12 insertions(+), 4 deletions(-)
>  delete mode 100644 roms/config.seabios
>  create mode 100644 roms/config.seabios-128k
>  create mode 100644 roms/config.seabios-256k
> 
> diff --git a/roms/Makefile b/roms/Makefile
> index 10d5a65..1d84230 100644
> --- a/roms/Makefile
> +++ b/roms/Makefile
> @@ -56,9 +56,10 @@ default:
>  	@echo "                    the EfiRom utility from edk2 / tianocore)"
>  	@echo "  slof           -- update slof.bin"
>  
> -bios: build-seabios-config-seabios
> -	cp seabios/builds/seabios/bios.bin ../pc-bios/bios.bin
> -	cp seabios/builds/seabios/*dsdt.aml ../pc-bios/
> +bios: build-seabios-config-seabios-128k build-seabios-config-seabios-256k
> +	cp seabios/builds/seabios-128k/bios.bin ../pc-bios/bios-128k.bin
> +	cp seabios/builds/seabios-256k/bios.bin ../pc-bios/bios-256k.bin
> +	cp seabios/builds/seabios-256k/src/fw/*dsdt.aml ../pc-bios/
>  
>  seavgabios: $(patsubst %,seavgabios-%,$(vgabios_variants))
>  
> diff --git a/roms/config.seabios b/roms/config.seabios
> deleted file mode 100644
> index c373b87..0000000
> --- a/roms/config.seabios
> +++ /dev/null
> @@ -1 +0,0 @@
> -# empty, default config works for us
> diff --git a/roms/config.seabios-128k b/roms/config.seabios-128k
> new file mode 100644
> index 0000000..23ca812
> --- /dev/null
> +++ b/roms/config.seabios-128k
> @@ -0,0 +1,5 @@
> +# for qemu machine types 1.6 + older
> +# need to turn off features (xhci) to make it fit into 128k
> +CONFIG_QEMU=y
> +CONFIG_ROM_SIZE=128
> +CONFIG_USB_XHCI=n

Is it enough to disable Xen or perhaps OHCI?

> diff --git a/roms/config.seabios-256k b/roms/config.seabios-256k
> new file mode 100644
> index 0000000..cc37a78
> --- /dev/null
> +++ b/roms/config.seabios-256k
> @@ -0,0 +1,3 @@
> +# for qemu machine types 1.7 + newer
> +CONFIG_QEMU=y
> +CONFIG_ROM_SIZE=256

Is there already an option to remove ACPI tables generation?  Should it
be enabled for 1.7, irrespective of the 128k->256k switch?

Paolo

Re: [Qemu-devel] [PATCH 1/2] roms: build two seabios binaries

[Gerd Hoffmann]

> > diff --git a/roms/config.seabios-128k b/roms/config.seabios-128k
> > new file mode 100644
> > index 0000000..23ca812
> > --- /dev/null
> > +++ b/roms/config.seabios-128k
> > @@ -0,0 +1,5 @@
> > +# for qemu machine types 1.6 + older
> > +# need to turn off features (xhci) to make it fit into 128k
> > +CONFIG_QEMU=y
> > +CONFIG_ROM_SIZE=128
> > +CONFIG_USB_XHCI=n
> 
> Is it enough to disable Xen or perhaps OHCI?

We can try that.  I've picked xhci because it is new in master, so we
don't take something away.

> > diff --git a/roms/config.seabios-256k b/roms/config.seabios-256k
> > new file mode 100644
> > index 0000000..cc37a78
> > --- /dev/null
> > +++ b/roms/config.seabios-256k
> > @@ -0,0 +1,3 @@
> > +# for qemu machine types 1.7 + newer
> > +CONFIG_QEMU=y
> > +CONFIG_ROM_SIZE=256
> 
> Is there already an option to remove ACPI tables generation?  Should it
> be enabled for 1.7, irrespective of the 128k->256k switch?

Not yet, but will be once mst's patches are in.  Yes, we can set that
for the 256k rom.

cheers,
  Gerd

Re: [Qemu-devel] [PATCH 1/2] roms: build two seabios binaries

[Paolo Bonzini]

Il 30/09/2013 15:44, Gerd Hoffmann ha scritto:
>>> > > +# for qemu machine types 1.6 + older
>>> > > +# need to turn off features (xhci) to make it fit into 128k
>>> > > +CONFIG_QEMU=y
>>> > > +CONFIG_ROM_SIZE=128
>>> > > +CONFIG_USB_XHCI=n
>> > 
>> > Is it enough to disable Xen or perhaps OHCI?
> We can try that.  I've picked xhci because it is new in master, so we
> don't take something away.
> 

Ok.  But if we can enable XHCI and disable Xen, that'd be a win because
Xen anyway embeds SeaBIOS within its own hvmloader firmware.

Paolo

Re: [Qemu-devel] [PATCH 2/2] machine type dependant bios

[Anthony Liguori]

[-- Attachment #1: Type: text/plain, Size: 2317 bytes --]

Wouldn't this be logical as a machine option?  They way you are introducing
it makes it impossible to control with -global.

Regards,

Anthony Liguori


On Mon, Sep 30, 2013 at 5:46 AM, Gerd Hoffmann <kraxel@redhat.com> wrote:

> Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
> ---
>  hw/i386/pc_piix.c    | 1 +
>  hw/i386/pc_q35.c     | 1 +
>  hw/i386/pc_sysfw.c   | 4 ++--
>  include/hw/i386/pc.h | 2 ++
>  4 files changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/hw/i386/pc_piix.c b/hw/i386/pc_piix.c
> index 907792b..35188a5 100644
> --- a/hw/i386/pc_piix.c
> +++ b/hw/i386/pc_piix.c
> @@ -238,6 +238,7 @@ static void pc_init_pci(QEMUMachineInitArgs *args)
>
>  static void pc_compat_1_6(QEMUMachineInitArgs *args)
>  {
> +    pc_default_bios_name = "bios-128k.bin";
>      has_pci_info = false;
>      rom_file_in_ram = false;
>  }
> diff --git a/hw/i386/pc_q35.c b/hw/i386/pc_q35.c
> index ca84e1c..e8b0168 100644
> --- a/hw/i386/pc_q35.c
> +++ b/hw/i386/pc_q35.c
> @@ -222,6 +222,7 @@ static void pc_q35_init(QEMUMachineInitArgs *args)
>
>  static void pc_compat_1_6(QEMUMachineInitArgs *args)
>  {
> +    pc_default_bios_name = "bios-128k.bin";
>      has_pci_info = false;
>      rom_file_in_ram = false;
>  }
> diff --git a/hw/i386/pc_sysfw.c b/hw/i386/pc_sysfw.c
> index e917c83..deb8869 100644
> --- a/hw/i386/pc_sysfw.c
> +++ b/hw/i386/pc_sysfw.c
> @@ -34,7 +34,7 @@
>  #include "hw/block/flash.h"
>  #include "sysemu/kvm.h"
>
> -#define BIOS_FILENAME "bios.bin"
> +const char *pc_default_bios_name = "bios-256k.bin";
>
>  typedef struct PcSysFwDevice {
>      SysBusDevice busdev;
> @@ -112,7 +112,7 @@ static void old_pc_system_rom_init(MemoryRegion
> *rom_memory, bool isapc_ram_fw)
>
>      /* BIOS load */
>      if (bios_name == NULL) {
> -        bios_name = BIOS_FILENAME;
> +        bios_name = pc_default_bios_name;
>      }
>      filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
>      if (filename) {
> diff --git a/include/hw/i386/pc.h b/include/hw/i386/pc.h
> index 9b2ddc4..cf9e47e 100644
> --- a/include/hw/i386/pc.h
> +++ b/include/hw/i386/pc.h
> @@ -12,6 +12,8 @@
>
>  /* PC-style peripherals (also used by other machines).  */
>
> +extern const char *pc_default_bios_name;
> +
>  typedef struct PcPciInfo {
>      Range w32;
>      Range w64;
> --
> 1.8.3.1
>
>

[-- Attachment #2: Type: text/html, Size: 3013 bytes --]

Re: [Qemu-devel] [PATCH 2/2] machine type dependant bios

[Paolo Bonzini]

Il 01/10/2013 00:08, Anthony Liguori ha scritto:
> Wouldn't this be logical as a machine option?  They way you are
> introducing it makes it impossible to control with -global.

You still have -bios.

We should introduce a machine object and convert unknown "-machine"
options to -global on that object.  But until that exists, IMO there's
no reason to do anything different from what has been done so far.

Paolo

Re: [Qemu-devel] [PATCH 2/2] machine type dependant bios

[Anthony Liguori]

[-- Attachment #1: Type: text/plain, Size: 790 bytes --]

On Mon, Sep 30, 2013 at 5:43 PM, Paolo Bonzini <pbonzini@redhat.com> wrote:

> Il 01/10/2013 00:08, Anthony Liguori ha scritto:
> > Wouldn't this be logical as a machine option?  They way you are
> > introducing it makes it impossible to control with -global.
>
> You still have -bios.
>

Which cannot be versioned.


> We should introduce a machine object and convert unknown "-machine"
> options to -global on that object.


There's no need.  -global works on QemuOpts, not QObjects.  By making this
a machine option, you get -global and machine compat for free.

We also have a way to set per-machine default machine options already too.

Regards,

Anthony Liguori



>  But until that exists, IMO there's
> no reason to do anything different from what has been done so far.
>
> Paolo
>

[-- Attachment #2: Type: text/html, Size: 1558 bytes --]

Re: [Qemu-devel] [PATCH 2/2] machine type dependant bios

[Paolo Bonzini]

Il 01/10/2013 00:57, Anthony Liguori ha scritto:
> On Mon, Sep 30, 2013 at 5:43 PM, Paolo Bonzini <pbonzini@redhat.com
> <mailto:pbonzini@redhat.com>> wrote:
> 
>     Il 01/10/2013 00:08, Anthony Liguori ha scritto:
>     > Wouldn't this be logical as a machine option?  They way you are
>     > introducing it makes it impossible to control with -global.
> 
>     You still have -bios.
> 
> Which cannot be versioned.

Yeah, that's right.

>     We should introduce a machine object and convert unknown "-machine"
>     options to -global on that object.
> 
> 
> There's no need.  -global works on QemuOpts, not QObjects.

No, it doesn't. :)

Paolo

Re: [Qemu-devel] [PATCH 2/2] machine type dependant bios

[Anthony Liguori]

[-- Attachment #1: Type: text/plain, Size: 990 bytes --]

On Tue, Oct 1, 2013 at 1:27 AM, Paolo Bonzini <pbonzini@redhat.com> wrote:

> Il 01/10/2013 00:57, Anthony Liguori ha scritto:
> > On Mon, Sep 30, 2013 at 5:43 PM, Paolo Bonzini <pbonzini@redhat.com
> > <mailto:pbonzini@redhat.com>> wrote:
> >
> >     Il 01/10/2013 00:08, Anthony Liguori ha scritto:
> >     > Wouldn't this be logical as a machine option?  They way you are
> >     > introducing it makes it impossible to control with -global.
> >
> >     You still have -bios.
> >
> > Which cannot be versioned.
>
> Yeah, that's right.
>
> >     We should introduce a machine object and convert unknown "-machine"
> >     options to -global on that object.
> >
> >
> > There's no need.  -global works on QemuOpts, not QObjects.
>
> No, it doesn't. :)
>

You are correct.  Too long of a day I guess.

Nonetheless, I still think it belongs as a machine option.  The other
comments about having a way to have default machine arguments still applies.

Regards,

Anthony Liguori


>
> Paolo
>

[-- Attachment #2: Type: text/html, Size: 1820 bytes --]