[PATCH V5 08/11] riscv: Support HAVE_IRQ_EXIT_ON_IRQ_STACK

[Guo Ren]

On Wed, Sep 21, 2022 at 7:56 PM Chen Zhongjin <chenzhongjin at huawei.com> wrote:
>
> Hi,
>
> Sorry to bother again, I just finished the test with your patches on
> mine patch set.
>
> On 2022/9/21 17:53, Guo Ren wrote:
> > On Wed, Sep 21, 2022 at 4:34 PM Chen Zhongjin <chenzhongjin at huawei.com> wrote:
> >> Hi,
> >>
> >> On 2022/9/18 23:52, guoren at kernel.org wrote:
> >>> diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
> >>> index 5f49517cd3a2..426529b84db0 100644
> >>> --- a/arch/riscv/kernel/entry.S
> >>> +++ b/arch/riscv/kernel/entry.S
> >>> @@ -332,6 +332,33 @@ ENTRY(ret_from_kernel_thread)
> >>>        tail syscall_exit_to_user_mode
> >>>    ENDPROC(ret_from_kernel_thread)
> >>>
> >>> +#ifdef CONFIG_IRQ_STACKS
> >>> +ENTRY(call_on_stack)
> >>> +     /* Create a frame record to save our ra and fp */
> >>> +     addi    sp, sp, -RISCV_SZPTR
> >>> +     REG_S   ra, (sp)
> >>> +     addi    sp, sp, -RISCV_SZPTR
> >>> +     REG_S   fp, (sp)
> >>> +
> >>> +     /* Save sp in fp */
> >>> +     move    fp, sp
> >>> +
>
> Considering that s0 points to previous sp normally, I think here we
> should have 'addi fp, sp, 2*RISCV_SZPTR'.
>
> An example below:
>
>      addi    sp, sp, -16
>      sd  ra, 8(sp)
>      sd  s0, 0(sp)
>      addi    s0, sp, 16    <- s0 is set to previous sp
>      ...
>
>      ld  ra,8(sp)
>      ld  s0,0(sp)
>      addi    sp,sp,16
>
> So maybe it's better to save the stack frame as below:
>
>      addi    sp, sp, -2*RISCV_SZPTR
>      REG_S   ra, RISCV_SZPTR(sp)
>      REG_S   s0, (sp)
>
>      /* Save sp in fp */
>      addi    s0, sp, 2*RISCV_SZPTR
>
>      ...
>
>      /*
>       * Restore sp from prev fp, and fp, ra from the frame
>       */
>      addi    sp, s0, -2*RISCV_SZPTR
>      REG_L   ra, RISCV_SZPTR(sp)
>      REG_L   s0, (sp)
>      addi    sp, sp, 2*RISCV_SZPTR
>
>
> Anyway, lets set fp as sp + 2 * RISCV_SZPTR, so that unwinder can
> connect two stacks same as normal function.
>
> I tested this with my patch and the unwinder works properly.
Thx, you got it. My patch broke the fp chain. I would fix it in the
next version.

>
>
> Thanks for your time!
>
> Best,
>
> Chen
>
> >>> +     /* Move to the new stack and call the function there */
> >>> +     li      a3, IRQ_STACK_SIZE
> >>> +     add     sp, a1, a3
> >>> +     jalr    a2
> >>> +
> >>> +     /*
> >>> +      * Restore sp from prev fp, and fp, ra from the frame
> >>> +      */
> >>> +     move    sp, fp
> >>> +     REG_L   fp, (sp)
> >>> +     addi    sp, sp, RISCV_SZPTR
> >>> +     REG_L   ra, (sp)
> >>> +     addi    sp, sp, RISCV_SZPTR
> >>> +     ret
> >>> +ENDPROC(call_on_stack)
> >>> +#endif
> >> Seems my compiler (riscv64-linux-gnu-gcc 8.4.0, cross compiling from
> >> x86) cannot recognize the register `fp`.
> > The whole entry.S uses s0 instead of fp, so I approve of your advice. Thx.
> >
> >> After I changed it to `s0` this can pass compiling.
> >>
> >>
> >> Seems there is nowhere else using `fp`, can this just using `s0` instead?
> >>
> >> Best,
> >>
> >> Chen
> >>
> >>



-- 
Best Regards
 Guo Ren