Re: [PATCH v1] LoongArch: Remove unnecessary checks in bt_address()

[Tiezhu Yang]

On 2025/12/10 上午10:25, Bibo Mao wrote:
> On 2025/12/10 上午9:28, Tiezhu Yang wrote:
> > On 2025/12/9 下午4:30, Huacai Chen wrote:
> > > Hi, Tiezhu,
> > >
> > > On Tue, Dec 9, 2025 at 2:18 PM Tiezhu Yang <yangtiezhu@xxxxxxxxxxx> 
> > > wrote:
> > > >
> > ...
> > > >          extern unsigned long eentry;
> > > >
> > > > -       if (__kernel_text_address(ra))
> > > > -               return ra;
> > > > -
> > > > -       if (__module_text_address(ra))
> > > > -               return ra;
> > > I think the correct way is to remove the __module_text_address()
> > > condition but keep the __kernel_text_address() condition. Then return
> > > 0 at the end of this function, and remove the __kernel_text_address()
> > > condition out of this function.
> >
> > It can not remove the check of __kernel_text_address() after calling
> > bt_address() because it needs to validate the calculated address for
> > exception, then no need to keep the __kernel_text_address() condition
> > in bt_address() because it will check the PC outside bt_address().
>          state->pc = bt_address(pc);
>          if (!state->pc) {
>                  pr_err("cannot find unwind pc at %p\n", (void *)pc);
>                  goto err;
>          }
>
>          if (!__kernel_text_address(state->pc))
>                  goto err;
> I guess that you both comes from different views :) one treats these 
> piece of code into one, one only views function bt_address().
>
> Especially with if (!state->pc), how could this happen?

IMO, state->pc will be not 0 forever in practice, this check is just an
error path and can be removed too if possible.

Thanks,
Tiezhu