Re: [PATCH V4 8/8] riscv: Add config of thread stack size
From: Guo Ren
Date: Wed Sep 21 2022 - 02:11:50 EST
On Tue, Sep 20, 2022 at 3:15 PM Arnd Bergmann <arnd@xxxxxxxx> wrote:
>
> On Tue, Sep 20, 2022, at 2:46 AM, Guo Ren wrote:
>
> >
> > How about this one: (only THREAD_SIZE, no THREAD_ORDER&SHIFT.)
> >
> > -
> > /* thread information allocation */
> > -#ifdef CONFIG_64BIT
> > -#define THREAD_SIZE_ORDER (2 + KASAN_STACK_ORDER)
> > -#else
> > -#define THREAD_SIZE_ORDER (1 + KASAN_STACK_ORDER)
> > -#endif
> > -#define THREAD_SIZE (PAGE_SIZE << THREAD_SIZE_ORDER)
> > +#define THREAD_SIZE CONFIG_THREAD_SIZE
>
>
> So far looks fine.
>
> >
> > /*
> > * By aligning VMAP'd stacks to 2 * THREAD_SIZE, we can detect overflow by
> > - * checking sp & (1 << THREAD_SHIFT), which we can do cheaply in the entry
> > - * assembly.
> > + * checking sp & THREAD_SIZE, which we can do cheaply in the entry assembly.
> > */
> > #ifdef CONFIG_VMAP_STACK
> > #define THREAD_ALIGN (2 * THREAD_SIZE)
> > @@ -36,7 +24,6 @@
> > #define THREAD_ALIGN THREAD_SIZE
> > #endif
>
> The THREAD_ALIGN does not, this only works for power-of-two numbers of
> THREAD_SIZE,
We double THREAD_SIZE to simplify the detection. See the commit log:
The overflow detect is performed before any attempt is made to access
the stack and the principle of stack overflow detection: kernel stacks
are aligned to double their size, enabling overflow to be detected with
a single bit test. For example, a 16K stack is aligned to 32K, ensuring
that bit 14 of the SP must be zero. On an overflow (or underflow), this
bit is flipped. Thus, overflow (of less than the size of the stack) can
be detected by testing whether this bit is set.
I would try to optimize the size of VMAP_STACK in another patch.
>
> > diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
> > index 426529b84db0..1e35fb3bdae5 100644
> > --- a/arch/riscv/kernel/entry.S
> > +++ b/arch/riscv/kernel/entry.S
> > @@ -29,8 +29,8 @@ _restore_kernel_tpsp:
> >
> > #ifdef CONFIG_VMAP_STACK
> > addi sp, sp, -(PT_SIZE_ON_STACK)
> > - srli sp, sp, THREAD_SHIFT
> > - andi sp, sp, 0x1
> > + srli sp, sp, PAGE_SHIFT
> > + andi sp, sp, (THREAD_SIZE >> PAGE_SHIFT)
>
> I think this needs to use THREAD_ALIGN, not THREAD_SIZE.
No, it's BIT[14], when THREAD_SIZE = 16K.
>
> Arnd
--
Best Regards
Guo Ren