arch/riscv/kernel/entry.S: Assembler messages:
arch/riscv/kernel/entry.S:106: Error: illegal operands `andi a0,s1,0x00001800'
This building error is because of the SR_MPP value is too large to be used
as an immediate value for andi. To fix this issue I use li to set the
immediate value to t0, then it can use t0 and s1 to do and operation.
Reported-by: kernel test robot <[email protected]>
Signed-off-by: Greentime Hu <[email protected]>
---
arch/riscv/kernel/entry.S | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
index 6ed579fc1073..000984695cd6 100644
--- a/arch/riscv/kernel/entry.S
+++ b/arch/riscv/kernel/entry.S
@@ -99,7 +99,8 @@ _save_context:
#ifdef CONFIG_CONTEXT_TRACKING
/* If previous state is in user mode, call context_tracking_user_exit. */
- andi a0, s1, SR_SPP
+ li t0, SR_PP
+ and a0, s1, t0
bnez a0, skip_context_tracking
call context_tracking_user_exit
--
2.27.0
This patch simplifies the checking for SR_MPP and SR_SPP. It uses SR_PP in the
code flow for both m-mode and s-mode then we can remove the ifdef here.
Signed-off-by: Greentime Hu <[email protected]>
---
arch/riscv/kernel/entry.S | 7 +------
1 file changed, 1 insertion(+), 6 deletions(-)
diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
index 000984695cd6..597beae0d238 100644
--- a/arch/riscv/kernel/entry.S
+++ b/arch/riscv/kernel/entry.S
@@ -210,13 +210,8 @@ ret_from_syscall_rejected:
ret_from_exception:
REG_L s0, PT_STATUS(sp)
csrc CSR_STATUS, SR_IE
-#ifdef CONFIG_RISCV_M_MODE
- /* the MPP value is too large to be used as an immediate arg for addi */
- li t0, SR_MPP
+ li t0, SR_PP
and s0, s0, t0
-#else
- andi s0, s0, SR_SPP
-#endif
bnez s0, resume_kernel
resume_userspace:
--
2.27.0
On Mon, 13 Jul 2020 01:32:15 PDT (-0700), [email protected] wrote:
> arch/riscv/kernel/entry.S: Assembler messages:
> arch/riscv/kernel/entry.S:106: Error: illegal operands `andi a0,s1,0x00001800'
>
> This building error is because of the SR_MPP value is too large to be used
> as an immediate value for andi. To fix this issue I use li to set the
> immediate value to t0, then it can use t0 and s1 to do and operation.
Thanks. I guess I must have something wrong with my build tests, as it's
supposed to be building the NOMMU stuff. I was just about to fix this up in
the patch, looks like I also lost my own Signed-off-by so I have to ammend it
anyway.
>
> Reported-by: kernel test robot <[email protected]>
> Signed-off-by: Greentime Hu <[email protected]>
> ---
> arch/riscv/kernel/entry.S | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
> index 6ed579fc1073..000984695cd6 100644
> --- a/arch/riscv/kernel/entry.S
> +++ b/arch/riscv/kernel/entry.S
> @@ -99,7 +99,8 @@ _save_context:
>
> #ifdef CONFIG_CONTEXT_TRACKING
> /* If previous state is in user mode, call context_tracking_user_exit. */
> - andi a0, s1, SR_SPP
> + li t0, SR_PP
> + and a0, s1, t0
> bnez a0, skip_context_tracking
> call context_tracking_user_exit
On Mon, 13 Jul 2020 01:32:16 PDT (-0700), [email protected] wrote:
> This patch simplifies the checking for SR_MPP and SR_SPP. It uses SR_PP in the
> code flow for both m-mode and s-mode then we can remove the ifdef here.
>
> Signed-off-by: Greentime Hu <[email protected]>
> ---
> arch/riscv/kernel/entry.S | 7 +------
> 1 file changed, 1 insertion(+), 6 deletions(-)
>
> diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
> index 000984695cd6..597beae0d238 100644
> --- a/arch/riscv/kernel/entry.S
> +++ b/arch/riscv/kernel/entry.S
> @@ -210,13 +210,8 @@ ret_from_syscall_rejected:
> ret_from_exception:
> REG_L s0, PT_STATUS(sp)
> csrc CSR_STATUS, SR_IE
> -#ifdef CONFIG_RISCV_M_MODE
> - /* the MPP value is too large to be used as an immediate arg for addi */
> - li t0, SR_MPP
> + li t0, SR_PP
> and s0, s0, t0
> -#else
> - andi s0, s0, SR_SPP
> -#endif
> bnez s0, resume_kernel
>
> resume_userspace:
This one is actually on a fairly fast path, so I can buy it's worth saving the
cycle.
On Mon, 13 Jul 2020 01:32:15 PDT (-0700), [email protected] wrote:
> arch/riscv/kernel/entry.S: Assembler messages:
> arch/riscv/kernel/entry.S:106: Error: illegal operands `andi a0,s1,0x00001800'
>
> This building error is because of the SR_MPP value is too large to be used
> as an immediate value for andi. To fix this issue I use li to set the
> immediate value to t0, then it can use t0 and s1 to do and operation.
>
> Reported-by: kernel test robot <[email protected]>
> Signed-off-by: Greentime Hu <[email protected]>
> ---
> arch/riscv/kernel/entry.S | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
> index 6ed579fc1073..000984695cd6 100644
> --- a/arch/riscv/kernel/entry.S
> +++ b/arch/riscv/kernel/entry.S
> @@ -99,7 +99,8 @@ _save_context:
>
> #ifdef CONFIG_CONTEXT_TRACKING
> /* If previous state is in user mode, call context_tracking_user_exit. */
> - andi a0, s1, SR_SPP
> + li t0, SR_PP
> + and a0, s1, t0
> bnez a0, skip_context_tracking
> call context_tracking_user_exit
Looks like this one already got fixed, I guess I saw the build report go by and
fixed it? I don't remember if I actually pulled this in, but I ended up with a
3-register andi so I guess I didn't do it that well.
I'm not sure why my build test aren't catching the M-mode stuff, as the
defconfigs are in the list. I'll go take a look...