Marshall Dawson wrote:
- You can skip this step in your instructions. Instead, when running
make menuconfig, look in Chipset / ChromeOS and select "Build for ChromeOS". This should cause Depthcharge to appear as a payload option.
I think this is a design issue in coreboot that ought to be fixed.
Depthcharge can be useful on its own, it should not be artificially tied down.
//Peter
On Thu, Mar 1, 2018 at 6:38 AM, Peter Stuge peter@stuge.se wrote:
Marshall Dawson wrote:
- You can skip this step in your instructions. Instead, when running
make menuconfig, look in Chipset / ChromeOS and select "Build for ChromeOS". This should cause Depthcharge to appear as a payload
option.
I think this is a design issue in coreboot that ought to be fixed.
Depthcharge can be useful on its own, it should not be artificially tied down.
Depthcharge is an external (to coreboot) project with the only goal to support booting Chrome OS. You shouldn't really expect it to do anything else unless you want to maintain a fork for that on your own. Any other use that may work now can break upstream (on chromium.googlesource.com) at any moment, and I think we just regularly sync that repository back into the coreboot.org mirror.
That said, CONFIG_CHROMEOS in coreboot doesn't really mean "this has to boot Chrome OS and nothing else"... it rather means "select code and features that are required for Chrome OS devices". Much of that "required for Chrome OS devices" is specifically "required to run depthcharge" (e.g. the lb_gpios in the coreboot table, or the vboot-capable FMAP), so disabling it just means depthcharge won't work right. If you want to build/run depthcharge for whatever reason (even to boot some other Linux distro as long as you package its kernel in a vblock), just leaving CONFIG_CHROMEOS enabled is probably the right thing. In many ways CONFIG_CHROMEOS in coreboot really just means CONFIG_SUPPORT_FOR_DEPTHCHARGE.
The Speedy issue seems to be a simple bug / missing feature in the payloads/external/depthcharge/Makefile. The configs in payload/libpayload/configs/ were originally just added as an FYI that people could manually copy in for make oldconfig. Then someone seems to have later written that depthcharge Makefile with the expectation that there'd always be a config for every board. That wasn't the intention and I don't think we want to go there, because often a config is shared across all boards in a family (e.g. configs/veyron for veyron_jerry and veyron_speedy, or configs/gru for kevin and bob). If we want to have that automatic selection from a Makefile, we should set up a simple translation table somewhere so it can find out which config belongs to which board. (Symlinks would also be an option, but I think some people don't like them in Git because of Windows. I don't think we should just duplicate the files because that always becomes a mess sooner or later.)
While we're at it, I think the depthcharge Makefile should also inject CONFIG_CHROMEOS=y in the libpayload config. That's how they're built for Chrome OS and that's the only way they're guaranteed to be compatible with depthcharge. (Again, it's not so much a "can only boot Chrome OS" option, more of a "select the right defaults for Chromebooks/depthcharge" option.)
-
Julius Werner -
Peter Stuge