On Fri, Jul 30, 2021 at 08:44:29AM -0600, Simon Glass wrote:
Hi Tom,
On Thu, 29 Jul 2021 at 07:22, Tom Rini trini@konsulko.com wrote:
On Wed, Jul 28, 2021 at 07:27:08PM -0600, Simon Glass wrote:
Hi Tom,
On Wed, 28 Jul 2021 at 17:28, Simon Glass sjg@chromium.org wrote:
Hi again,
On Mon, 26 Jul 2021 at 08:06, Simon Glass sjg@chromium.org wrote:
Hi Tom,
On Sun, 25 Jul 2021 at 15:10, Tom Rini trini@konsulko.com wrote:
So, I'm trying to fix the problem on am335x_evm (and some family configs) with needing SPL_OF_CONTROL enabled. This is mostly fine just enabling the option, except on am335x_evm itself, which is the kitchen-sink config and overflows memory. I've gone with switching to SPL_OF_PLATDATA there as am335x in general has all of the U_BOOT_DRVINFO entries it needs I believe. But, with the following patch:
diff --git a/arch/arm/dts/am335x-evm-u-boot.dtsi b/arch/arm/dts/am335x-evm-u-boot.dtsi index 4cf5f9928d58..514f682cac99 100644 --- a/arch/arm/dts/am335x-evm-u-boot.dtsi +++ b/arch/arm/dts/am335x-evm-u-boot.dtsi @@ -8,6 +8,7 @@ &l4_per {
segment@300000 {
u-boot,dm-pre-reloc; target-module@e000 { u-boot,dm-pre-reloc;diff --git a/configs/am335x_boneblack_vboot_defconfig b/configs/am335x_boneblack_vboot_defconfig index a0baeec79edd..ffeefd1a0087 100644 --- a/configs/am335x_boneblack_vboot_defconfig +++ b/configs/am335x_boneblack_vboot_defconfig @@ -31,6 +31,7 @@ CONFIG_CMD_SPL=y # CONFIG_CMD_SETEXPR is not set CONFIG_BOOTP_DNS2=y CONFIG_OF_CONTROL=y +CONFIG_SPL_OF_CONTROL=y CONFIG_ENV_OVERWRITE=y CONFIG_ENV_IS_IN_MMC=y CONFIG_SYS_REDUNDAND_ENVIRONMENT=y diff --git a/configs/am335x_evm_defconfig b/configs/am335x_evm_defconfig index a33efff42a74..f35b2a02f56b 100644 --- a/configs/am335x_evm_defconfig +++ b/configs/am335x_evm_defconfig @@ -37,13 +37,16 @@ CONFIG_MTDIDS_DEFAULT="nand0=nand.0" CONFIG_MTDPARTS_DEFAULT="mtdparts=nand.0:128k(NAND.SPL),128k(NAND.SPL.backup1),128k(NAND.SPL.backup2),128k(NAND.SPL.backup3),256k(NAND.u-boot-spl-os),1m(NAND.u-boot),128k(NAND.u-boot-env),128k(NAND.u-boot-env.backup1),8m(NAND.kernel),-(NAND.file-system)" # CONFIG_SPL_EFI_PARTITION is not set CONFIG_OF_CONTROL=y +CONFIG_SPL_OF_CONTROL=y CONFIG_OF_LIST="am335x-evm am335x-bone am335x-boneblack am335x-evmsk am335x-bonegreen am335x-icev2 am335x-pocketbeagle" +CONFIG_SPL_OF_PLATDATA=y CONFIG_ENV_OVERWRITE=y CONFIG_SYS_RELOC_GD_ENV_ADDR=y CONFIG_ENV_VARS_UBOOT_RUNTIME_CONFIG=y CONFIG_SPL_ENV_IS_NOWHERE=y CONFIG_VERSION_VARIABLE=y CONFIG_BOOTP_SEND_HOSTNAME=y +# CONFIG_SPL_SIMPLE_BUS is not set CONFIG_BOOTCOUNT_LIMIT=y CONFIG_CLK=y CONFIG_CLK_CDCE9XX=y diff --git a/configs/am335x_evm_spiboot_defconfig b/configs/am335x_evm_spiboot_defconfig index 8f0c330674a9..4a2a56a9af9e 100644 --- a/configs/am335x_evm_spiboot_defconfig +++ b/configs/am335x_evm_spiboot_defconfig @@ -32,6 +32,7 @@ CONFIG_BOOTP_DNS2=y CONFIG_CMD_MTDPARTS=y # CONFIG_SPL_EFI_PARTITION is not set CONFIG_OF_CONTROL=y +CONFIG_SPL_OF_CONTROL=y CONFIG_OF_LIST="am335x-evm am335x-bone am335x-boneblack am335x-evmsk am335x-bonegreen am335x-icev2 am335x-pocketbeagle" CONFIG_ENV_OVERWRITE=y # CONFIG_ENV_IS_IN_FAT is not set
I get the following failure and I don't see how to debug this: DTOC spl/dts/dt-plat.c Traceback (most recent call last): File "./tools/dtoc/dtoc", line 115, in <module> args.phase, instantiate=args.instantiate) File "/home/trini/work/u-boot/u-boot/tools/dtoc/../dtoc/dtb_platdata.py", line 1223, in run_steps outfile.method(plat) File "/home/trini/work/u-boot/u-boot/tools/dtoc/../dtoc/dtb_platdata.py", line 1081, in generate_plat self.output_node_plat(node) File "/home/trini/work/u-boot/u-boot/tools/dtoc/../dtoc/dtb_platdata.py", line 1023, in output_node_plat self._output_values(node) File "/home/trini/work/u-boot/u-boot/tools/dtoc/../dtoc/dtb_platdata.py", line 812, in _output_values self._output_prop(node, node.props[pname]) File "/home/trini/work/u-boot/u-boot/tools/dtoc/../dtoc/dtb_platdata.py", line 798, in _output_prop self._output_list(node, prop) File "/home/trini/work/u-boot/u-boot/tools/dtoc/../dtoc/dtb_platdata.py", line 628, in _output_list vals.append(get_value(prop.type, val)) File "/home/trini/work/u-boot/u-boot/tools/dtoc/../dtoc/dtb_platdata.py", line 126, in get_value val = '%#x' % fdt_util.fdt32_to_cpu(value) File "/home/trini/work/u-boot/u-boot/tools/dtoc/../dtoc/fdt_util.py", line 28, in fdt32_to_cpu return struct.unpack('>I', val)[0] TypeError: a bytes-like object is required, not 'bool' scripts/Makefile.spl:352: recipe for target 'spl/dts/dt-plat.c' failed make[1]: *** [spl/dts/dt-plat.c] Error 1 make[1]: *** Deleting file 'spl/dts/dt-plat.c' Makefile:1999: recipe for target 'spl/u-boot-spl' failed make: *** [spl/u-boot-spl] Error 2
That seems like a bug, where perhaps it is seeing a property with no value so it thinks it is a bool, but then somehow tries to gets its int value.
I added a print to _output_list()
else: print('node', node.path, prop.name) for val in prop.value: vals.append(get_value(prop.type, val))and see that it is the 'ranges' property. Definitely seems like a bug but I'll have to dig into it I think. The Prop() class selects a bool type when the value is empty, but somehow that isn't happening.
This is actually a little tricky, but I'm looking at it.
OK I think I have figured this out. I sent a series.
That gets everything building, thanks! It also shows that some parts of the kitchen-sink config there depend on OF_CONTROL features, really. What I'm going to prove out next (but will require a little bit of time due to having to set up lab stuff and dig out notes) is that for this case, we can just go with however ROM configured the PHY. We were also still over SRAM space, with PLATDATA enabled too.
OK good. I think more users would help reduce the number of rough edges .
Can you please point me to the tree with before/after? I would like to look at SRAM space just in case I can see something.
With SPL_OF_CONTROL, we're 4900 bytes over (and it's just adding SPL_OF_CONTROL to the current config). PLATDATA isn't an option since both omap_hsmmc.c and drivers/net/ti/cpsw.c need gpio by name/etc functionality. The mmc case could be kludged (assume card is writable if needed) but cpsw is harder at first glance.