Re: [PATCH v2] LoongArch: KGDB: Rework arch_kgdb_breakpoint() implementation

[Jinyang He <hejinyang@loongson.cn>]

On 2025-03-24 12:09, Huacai Chen wrote:

> Hi, Jinyang,
>
> On Mon, Mar 24, 2025 at 9:42 AM Jinyang He <hejinyang@loongson.cn> wrote:
>> On 2025-03-22 20:51, Huacai Chen wrote:
>>
>>> Hi, Tiezhu & Jinyang,
>>>
>>> On Fri, Mar 21, 2025 at 10:11 PM WangYuli <wangyuli@uniontech.com> wrote:
>>>> The arch_kgdb_breakpoint() function defines the kgdb_breakinst
>>>> symbol using inline assembly.
>>>>
>>>> There's a potential issue where the compiler might inline
>>>> arch_kgdb_breakpoint(), which would then define the kgdb_breakinst
>>>> symbol multiple times, leading to a linker error.
>>>>
>>>> To prevent this, declare arch_kgdb_breakpoint() as noline.
>>>>
>>>> Fix follow error with LLVM-19 *only* when LTO_CLANG_FULL:
>>>>       LD      vmlinux.o
>>>>     ld.lld-19: error: ld-temp.o <inline asm>:3:1: symbol 'kgdb_breakinst' is already defined
>>>>     kgdb_breakinst: break 2
>>>>     ^
>>>>
>>>> Additionally, remove "nop" here because it's meaningless for LoongArch
>>>> here.
>>>>
>>>> Fixes: e14dd076964e ("LoongArch: Add basic KGDB & KDB support")
>>>> Co-developed-by: Winston Wen <wentao@uniontech.com>
>>>> Signed-off-by: Winston Wen <wentao@uniontech.com>
>>>> Co-developed-by: Wentao Guan <guanwentao@uniontech.com>
>>>> Signed-off-by: Wentao Guan <guanwentao@uniontech.com>
>>>> Co-developed-by: Huacai Chen <chenhuacai@loongson.cn>
>>>> Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
>>>> Tested-by: Yuli Wang <wangyuli@uniontech.com>
>>>> Tested-by: Binbin Zhou <zhoubinbin@loongson.cn>
>>>> Signed-off-by: Yuli Wang <wangyuli@uniontech.com>
>>>> ---
>>>> Changelog:
>>>>    *v1->v2:
>>>>       1. Drop the nop which is no effect for LoongArch here.
>>>>       2. Add "STACK_FRAME_NON_STANDARD" for arch_kgdb_breakpoint() to
>>>> avoid the objtool warning.
>>>> ---
>>>>    arch/loongarch/kernel/kgdb.c | 4 ++--
>>>>    1 file changed, 2 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/arch/loongarch/kernel/kgdb.c b/arch/loongarch/kernel/kgdb.c
>>>> index 445c452d72a7..38bd0561d7d5 100644
>>>> --- a/arch/loongarch/kernel/kgdb.c
>>>> +++ b/arch/loongarch/kernel/kgdb.c
>>>> @@ -224,13 +224,13 @@ void kgdb_arch_set_pc(struct pt_regs *regs, unsigned long pc)
>>>>           regs->csr_era = pc;
>>>>    }
>>>>
>>>> -void arch_kgdb_breakpoint(void)
>>>> +noinline void arch_kgdb_breakpoint(void)
>>>>    {
>>>>           __asm__ __volatile__ (                  \
>>>>                   ".globl kgdb_breakinst\n\t"     \
>>>> -               "nop\n"                         \
>>>>                   "kgdb_breakinst:\tbreak 2\n\t"); /* BRK_KDB = 2 */
>>>>    }
>>>> +STACK_FRAME_NON_STANDARD(arch_kgdb_breakpoint);
>>> Is there a better solution than STACK_FRAME_NON_STANDARD()? In the
>>> past we can use annotate_reachable() in arch_kgdb_breakpoint(), but
>>> annotate_reachable() is no longer exist.
>> Maybe we can parse the imm-code of `break` and set diffrent insn_type in
>> objtool.
>> The BRK_KDB imply the PC will go head, while the BRK_BUG imply PC stop.
>> (arch/loongarch/include/uapi/asm/break.h)
>>
>> Tiezhu, how do you think?
> Touching objtool may be a little complicated, is ANNOTATE_REACHABLE()
> the successor of annotate_reachable()? I tried
> ANNOTATE_REACHABLE(kgdb_breakinst) but there was still a warning.
Should it annotate to the next insn of `break`?
Like,

     kgdb_breakinst:
     break 2
     another_label:

ANNOTATE_REACHABLE(another_label)

Jinyang
>
> Huacai
>
>>
>>> Huacai
>>>
>>>>    /*
>>>>     * Calls linux_debug_hook before the kernel dies. If KGDB is enabled,
>>>> --
>>>> 2.49.0
>>>>