[U-Boot] [PATCH v7 14/15] doc: Add a readme guide for SiFive FU540

Wed Feb 13 22:39:55 UTC 2019

Atish Patra <Atish.Patra at wdc.com> writes:

>> On Feb 12, 2019, at 4:18 PM, Kevin Hilman <khilman at baylibre.com> wrote:
>> 
>> Anup Patel <Anup.Patel at wdc.com> writes:
>> 
>>> From: Atish Patra <atish.patra at wdc.com>
>>> 
>>> The readme guide describes the procedure to build, flash and boot Linux
>>> using U-Boot on HiFive Unleashed. It also explains the current state of
>>> U-boot support and future action items.
>>> 
>>> Signed-off-by: Atish Patra <atish.patra at wdc.com>
>>> Signed-off-by: Anup Patel <anup.patel at wdc.com>
>>> Reviewed-by: Lukas Auer <lukas.auer at aisec.fraunhofer.de>
>> 
>> I'm testing this with the mainline kernel (v5.0-rc6) and running into
>> some problems getting kernel output on the serial console.
>> 
>
> Unfortunately. 
>

[...]

>> First question: this doc doesn't explain how there is a DT at
>> 0x82200000, and what it contains.
>> 
>
> DT is being passed from the previous stage boot loader to OpenSBI and OpenSBI passed it to U-Boot
> after modifying few entries (hart masks and plic entries).
>
> There is a way to provide custom DT in OpenSBI as well.

And if u-boot were load (or TFTP) a custom DT, this should work too,
right?

>> Trying this with a freshly build u-boot payload with OpenSBI, bootm
>> seems to detect a DT there, but I don't understand how it got there, or
>> what it is in it.
>> 
>> Looking a little closer, it appears that this DT (and the address) is
>> hard-coded in the OpenSBI code.  This should proably be documented here
>> for clarity sake.
>> 
>
> I will update the document to add more clarity.

Thanks.

>>> +## Booting kernel from Legacy Image at 80200000 ...
>>> +   Image Name:   Linux
>>> +   Image Type:   RISC-V Linux Kernel Image (uncompressed)
>>> +   Data Size:    14939068 Bytes = 14.2 MiB
>>> +   Load Address: 80200000
>>> +   Entry Point:  80200000
>>> +   Verifying Checksum ... OK
>>> +## Flattened Device Tree blob at 82200000
>>> +   Booting using the fdt blob at 0x82200000
>>> +   Loading Kernel Image ... OK
>>> +   Using Device Tree in place at 0000000082200000, end 0000000082205c69
>>> +
>>> +Starting kernel ...
>> 
>> Next, I'm able to DHCP and TFTP my uImage just like above, but I don't
>> see any output on the console after the "Starting kernel".
>> 
>> That suggests that whatever DT is present there doesn't have the right
>> settings for the serial console.
>> 
>> I tried setting the u-boot bootargs to "console=ttySIF0 earlyprintk",
>> but I'm still seeing nothing on the console.
>> 
>> Dumping the hard-coded OpenSBI DT from u-boot[1], it seem that the
>> chosen node has the right value (as documented in this patch):
>> 
>> Hmm, maybe I'm missing the obvious... is there even an upstream serial
>> driver for this UART in v5.0-rc6...  (/me goes searching for the
>> compatible)... hmm, doesn't look like it.
>> 
>
> Unfortunately, all the drivers for unleashed are not upstreamed yet. It is a mess and hopefully it will be resolved soon.
>
>>> +[    0.000000] OF: fdt: Ignoring memory range 0x80000000 - 0x80200000
>>> +[    0.000000] Linux version 5.0.0-rc1-00020-g4b51f736 (atish at jedi-01) (gcc version 7.2.0 (GCC)) #262 SMP Mon Jan 21 17:39:27 PST 2019
>> 
>> Looks like you're testing with a handful of out-of-tree kernel patches.
>> Can you give a pointer to where you're building your kernel from?
>> 
>
> Here is my branch based on 5.0-rc5 that should work.

Yes, building a kernel from there, using the config options you
mentioned works.  Thanks.

> https://github.com/atishp04/linux/  v5.0-rc5_unleashed_uboot
>
> 1-8 patches are mostly SMP fixes and currently under review. They should be part of next merge window.
> ------------------------------------------------------------------------------------------
> 1. 4a0edc9b RISC-V: Assign hwcap as per comman capabilities.
> 2. c29e4afa irqchip/irq-sifive-plic: Check and continue in case of an invalid cpuid.
> 3. 6c191098 clocksource/drivers/riscv: Add required checks during clock source init
> 4. d89bfcf6 RISC-V: Compare cpuid with NR_CPUS before mapping.
> 5. 474cb3e3 RISC-V: Allow hartid-to-cpuid function to fail.
> 6. 5e07a2f4 RISC-V: Remove NR_CPUs check during hartid search from DT
> 7. cf5f2ec6 RISC-V: Move cpuid to hartid mapping to SMP.
> 8. 7838fe36 RISC-V: Do not wait indefinitely in __cpu_up
> ------------------------------------------------------------------------------------------
>
> The real mess is the following driver patches for unleashed board. 
>
> 9. d46dc16f spi: sifive: no dma hack
> 10. e8ea1346 spi: add driver for the SiFive SPI controller
> 11. 120d5658 pcie-microsemi: added support for the Vera-board root complex
> 12. afb97b09 RISC-V: Networking fix Hack
> 13. 4ca6e585 pwm-sifive: add a driver for SiFive SoC PWM
> 14. ddb4b49e gpio-sifive: support GPIO on SiFive SoCs
> 15. f8859c7d u54-prci: driver for core U54 clocks
> 16. 27b6fd77 u54-prci: driver for core U54 clocks
> 17. dcd33855 gemgxl-mgmt: implement clock switch for GEM tx_clk
> 18. b25bb020 tty: serial: add driver for the SiFive UART
> 19. fd6f363f dt-bindings: serial: add documentation for the SiFive UART driver
>
> Please make sure following configs are enabled in your config.
>
> CONFIG_SERIAL_SIFIVE=y
> CONFIG_SERIAL_SIFIVE_CONSOLE=y
> CONFIG_SIFIVE_PLIC=y
> CONFIG_SPI=y
> CONFIG_SPI_SIFIVE=y
> CONFIG_GPIOLIB=y
> CONFIG_GPIO_SIFIVE=y
> CONFIG_PWM_SIFIVE=y
> CONFIG_CLK_U54_PRCI=y
> CONFIG_CLK_GEMGXL_MGMT=y
>
> I will update the document to mention about the driver patches as well. Sorry for the inconvenience caused.

Probably should mention the kernel config requirements too, if they are
not the defaults in upstream.

Kevin