Re: [Qemu-devel] [PATCH] riscv: sifive_test: Add reset functionality

[Palmer Dabbelt]

On Fri, 14 Jun 2019 08:15:51 PDT (-0700), address@hidden wrote:
> This adds a reset opcode for sifive_test device to trigger a system
> reset for testing purpose.
>
> Signed-off-by: Bin Meng <address@hidden>
> ---
>
>  hw/riscv/sifive_test.c         | 4 ++++
>  include/hw/riscv/sifive_test.h | 3 ++-
>  2 files changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/hw/riscv/sifive_test.c b/hw/riscv/sifive_test.c
> index 24a04d7..cd86831 100644
> --- a/hw/riscv/sifive_test.c
> +++ b/hw/riscv/sifive_test.c
> @@ -21,6 +21,7 @@
>  #include "qemu/osdep.h"
>  #include "hw/sysbus.h"
>  #include "qemu/module.h"
> +#include "sysemu/sysemu.h"
>  #include "target/riscv/cpu.h"
>  #include "hw/riscv/sifive_test.h"
>
> @@ -40,6 +41,9 @@ static void sifive_test_write(void *opaque, hwaddr addr,
>              exit(code);
>          case FINISHER_PASS:
>              exit(0);
> +        case FINISHER_RESET:
> +            qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
> +            return;
>          default:
>              break;
>          }
> diff --git a/include/hw/riscv/sifive_test.h b/include/hw/riscv/sifive_test.h
> index 71d4c9f..c186a31 100644
> --- a/include/hw/riscv/sifive_test.h
> +++ b/include/hw/riscv/sifive_test.h
> @@ -34,7 +34,8 @@ typedef struct SiFiveTestState {
>
>  enum {
>      FINISHER_FAIL = 0x3333,
> -    FINISHER_PASS = 0x5555
> +    FINISHER_PASS = 0x5555,
> +    FINISHER_RESET = 0x7777
>  };
>
>  DeviceState *sifive_test_create(hwaddr addr);

Poking around the hardware side, it looks like we're silently ignoring all
codes aside from 0x3333 and 0x5555.  As a result there's really no way to test
if this added reset is going to work, so this should be a "sifive,test1"
instead of a "sifive,test0".  It's backwards compatible, so the compat string
in the device trees should look something like

   compatible = "sifive,test1", "sifive,test0";

With that it should be fine, pending the change to our hardware to reserve this
as an error code within "sifive,test0" devices.

Do you mind also submitting a device tree binding to Linux that describes this
device?  That way we can all stay on the same page about what the pair of
interfaces do.

Thanks!