[PATCH] Drop all pre-CBFS rom_address entries in Config.lb/devicetree.cb

[Uwe Hermann]

See patch.

Uwe.

[PATCH] Drop all pre-CBFS rom_address entries in Config.lb/devicetree.cb

[Myles Watson]

On Wed, Oct 28, 2009 at 1:58 PM, Uwe Hermann  wrote:
> See patch.

I think it would be nice to figure out what we do with the "onboard" device
at the same time.

from src/drivers/pci/onboard.c:

static void onboard_enable(device_t dev)
{
    struct drivers_pci_onboard_config *conf;
        conf = dev->chip_info;
    dev->rom_address = conf->rom_address;
}

Does it make sense to change ROM handling so that only "onboard" devices can
have their ROMs run?  That's the way it used to be, right?

Otherwise, maybe we don't need onboard anymore at all?

Thanks,
Myles

[PATCH] Drop all pre-CBFS rom_address entries in Config.lb/devicetree.cb

[Peter Stuge]

Myles Watson wrote:
> Otherwise, maybe we don't need onboard anymore at all?

Is it used for anything but the option ROM?

//Peter

[PATCH] Drop all pre-CBFS rom_address entries in Config.lb/devicetree.cb

[Myles Watson]

On Thu, Oct 29, 2009 at 7:55 PM, Peter Stuge  wrote:
> Myles Watson wrote:
> > Otherwise, maybe we don't need onboard anymore at all?
>
> Is it used for anything but the option ROM?

Not that I can tell.  There's not much code there.

Thanks,
Myles

[PATCH] Drop all pre-CBFS rom_address entries in Config.lb/devicetree.cb

[Myles Watson]

On Wed, Oct 28, 2009 at 2:18 PM, Myles Watson  wrote:
> On Wed, Oct 28, 2009 at 1:58 PM, Uwe Hermann  wrote:
> > See patch.
>
> I think it would be nice to figure out what we do with the "onboard" device
> at the same time.

Ping.  I think this patch is an improvement, but I didn't want the
opportunity to remove more dead code pass by.  Any comments?

from src/drivers/pci/onboard.c:
> static void onboard_enable(device_t dev)
> {
>     struct drivers_pci_onboard_config *conf;
>         conf = dev->chip_info;
>     dev->rom_address = conf->rom_address;
> }
>
> Does it make sense to change ROM handling so that only "onboard" devices
> can have their ROMs run?  That's the way it used to be, right?
>
> Otherwise, maybe we don't need onboard anymore at all?

Thanks,
Myles

[PATCH] Drop all pre-CBFS rom_address entries in Config.lb/devicetree.cb

[Stefan Reinauer]

Myles Watson wrote:
> On Wed, Oct 28, 2009 at 2:18 PM, Myles Watson <[e-mail]
> > wrote:
> On Wed, Oct 28, 2009 at 1:58 PM, Uwe Hermann <[e-mail]
> <mailto:[e-mail]>> wrote:
>
>     See patch.
>
> I think it would be nice to figure out what we do with the
> "onboard" device at the same time.
>
> Ping.  I think this patch is an improvement, but I didn't want the
> opportunity to remove more dead code pass by.  Any comments?

Acked-by: Stefan Reinauer  for removing all the
->onboard stuff.
>  from src/drivers/pci/onboard.c:
>
> static void onboard_enable(device_t dev)
> {
>     struct drivers_pci_onboard_config *conf;
>         conf = dev->chip_info;
>     dev->rom_address = conf->rom_address;
> }
>
> Does it make sense to change ROM handling so that only "onboard"
> devices can have their ROMs run?  That's the way it used to be, right?

No. Plugin cards (graphics cards) need to have their ROM run, too.

Stefan

[PATCH] Drop all pre-CBFS rom_address entries in Config.lb/devicetree.cb

[Myles Watson]

On Thu, Nov 5, 2009 at 12:12 PM, Stefan Reinauer wrote:
> Myles Watson wrote:
> > On Wed, Oct 28, 2009 at 2:18 PM, Myles Watson <[e-mail]
> > > wrote:
> > On Wed, Oct 28, 2009 at 1:58 PM, Uwe Hermann <[e-mail]
> > <mailto:[e-mail]>> wrote:
> >
> >     See patch.
> >
> > I think it would be nice to figure out what we do with the
> > "onboard" device at the same time.
> >
> > Ping.  I think this patch is an improvement, but I didn't want the
> > opportunity to remove more dead code pass by.  Any comments?
>
> Acked-by: Stefan Reinauer  for removing all the
> ->onboard stuff.
> >  from src/drivers/pci/onboard.c:
> >
> > static void onboard_enable(device_t dev)
> > {
> >     struct drivers_pci_onboard_config *conf;
> >         conf = dev->chip_info;
> >     dev->rom_address = conf->rom_address;
> > }
> >
> > Does it make sense to change ROM handling so that only "onboard"
> > devices can have their ROMs run?  That's the way it used to be,
>
> right?
> >
>
> No. Plugin cards (graphics cards) need to have their ROM run, too.

I meant should "onboard" devices be the only ones allowed to have ROMs in
CBFS, but I can see that that wasn't right either.  It just seems like when
we made the change to CBFS we allowed a lot more devices to have ROMs in
CBFS, and I was wondering if that was intentional, or a side-effect.

Thanks,
Myles

[PATCH] Drop all pre-CBFS rom_address entries in Config.lb/devicetree.cb

[ron minnich]

On Thu, Nov 5, 2009 at 10:22 AM, Myles Watson  wrote:
> I meant should "onboard" devices be the only ones allowed to have ROMs in
> CBFS, but I can see that that wasn't right either.  It just seems like when
> we made the change to CBFS we allowed a lot more devices to have ROMs in
> CBFS, and I was wondering if that was intentional, or a side-effect.

from my point of view, it fixed a problem that we also fixed in v3 --
it was really, really hard to have lots of ROM images in coreboot
before we went to LAR/CBFS. So we fixed a shortcoming.

It was incredibly painful (IMHO) to add rom images before we got CBFS.

What's interesting is we can even add (e.g.) an upgraded ROM image to
CBFS that might over-ride the ROM image on an add-in card. Hence you
can "upgrade" the rom image on a card without having to reflash the
card -- just put it in CBFS. We've never done this but the possibility
is there.

ron

[PATCH] Drop all pre-CBFS rom_address entries in Config.lb/devicetree.cb

[Myles Watson]

On Thu, Nov 5, 2009 at 12:25 PM, ron minnich  wrote:
> On Thu, Nov 5, 2009 at 10:22 AM, Myles Watson  wrote:
> > I meant should "onboard" devices be the only ones allowed to have ROMs in
> > CBFS, but I can see that that wasn't right either.  It just seems like
>
> when
> > we made the change to CBFS we allowed a lot more devices to have ROMs in
> > CBFS, and I was wondering if that was intentional, or a side-effect.
>
> from my point of view, it fixed a problem that we also fixed in v3 --
> it was really, really hard to have lots of ROM images in coreboot
> before we went to LAR/CBFS. So we fixed a shortcoming.
>
> It was incredibly painful (IMHO) to add rom images before we got CBFS.
>
> What's interesting is we can even add (e.g.) an upgraded ROM image to
> CBFS that might over-ride the ROM image on an add-in card. Hence you
> can "upgrade" the rom image on a card without having to reflash the
> card -- just put it in CBFS. We've never done this but the possibility
> is there.

Great.  I just wanted to make sure we didn't need onboard any more.

Uwe:
Acked-by: Myles Watson

Thanks,
Myles

[PATCH] Drop all pre-CBFS rom_address entries in Config.lb/devicetree.cb

[Myles Watson]

> Acked-by: Myles Watson

Rev 4923.

Thanks,
Myles