Message-ID: <CAJF2gTS6MntY1sa+VqQ-hXbrGRaXOKqSY86xwe1-G1fVZHn7gQ@mail.gmail.com>
Date: Fri, 22 Dec 2023 15:33:38 +0800
From: Guo Ren <guoren@...nel.org>
To: Leonardo Bras <leobras@...hat.com>
Cc: Charlie Jenkins <charlie@...osinc.com>, linux-kernel@...r.kernel.org, 
	paul.walmsley@...ive.com, palmer@...belt.com, alexghiti@...osinc.com, 
	xiao.w.wang@...el.com, david@...hat.com, panqinglin2020@...as.ac.cn, 
	rick.p.edgecombe@...el.com, willy@...radead.org, bjorn@...osinc.com, 
	conor.dooley@...rochip.com, cleger@...osinc.com, 
	linux-riscv@...ts.infradead.org, Guo Ren <guoren@...ux.alibaba.com>, 
	stable@...r.kernel.org
Subject: Re: [PATCH V2 1/4] riscv: mm: Fixup compat mode boot failure

On Fri, Dec 22, 2023 at 12:33 PM Leonardo Bras <leobras@...hat.com> wrote:
>
> On Thu, Dec 21, 2023 at 07:50:27PM -0800, Charlie Jenkins wrote:
> > On Fri, Dec 22, 2023 at 10:57:16AM +0800, Guo Ren wrote:
> > > On Fri, Dec 22, 2023 at 9:51 AM Leonardo Bras <leobras@...hat.com> wrote:
> > > >
> > > > Hello Guo Ren,
> > > >
> > > > On Thu, Dec 21, 2023 at 10:46:58AM -0500, guoren@...nel.org wrote:
> > > > > From: Guo Ren <guoren@...ux.alibaba.com>
> > > > >
> > > > > In COMPAT mode, the STACK_TOP is 0x80000000, but the TASK_SIZE is
> > > > > 0x7fff000. When the user stack is upon 0x7fff000, it will cause a user
> > > > > segment fault. Sometimes, it would cause boot failure when the whole
> > > > > rootfs is rv32.
> > > >
> > > > Checking if I get the scenario:
> > > >
> > > > In pgtable.h:
> > > > #ifdef CONFIG_64BIT
> > > > #define TASK_SIZE_64    (PGDIR_SIZE * PTRS_PER_PGD / 2)
> > > > #define TASK_SIZE_MIN   (PGDIR_SIZE_L3 * PTRS_PER_PGD / 2)
> > > >
> > > > #ifdef CONFIG_COMPAT
> > > > #define TASK_SIZE_32    (_AC(0x80000000, UL) - PAGE_SIZE)
> > > > #define TASK_SIZE       (test_thread_flag(TIF_32BIT) ? \
> > > >                          TASK_SIZE_32 : TASK_SIZE_64)
> > > > #else
> > > > [...]
> > > >
> > > > Meaning CONFIG_COMPAT is only available in CONFIG_64BIT, and TASK_SIZE in
> > > > compat mode is either TASK_SIZE_32 or TASK_SIZE_64 depending on the thread_flag.
> > > >
> > > > from processor.h:
> > > > #ifdef CONFIG_64BIT
> > > > #define DEFAULT_MAP_WINDOW      (UL(1) << (MMAP_VA_BITS - 1))
> > > > #define STACK_TOP_MAX           TASK_SIZE_64
> > > > [...]
> > > > #define STACK_TOP               DEFAULT_MAP_WINDOW
> > > >
> > > >
> > > > where:
> > > > #define MMAP_VA_BITS (is_compat_task() ? VA_BITS_SV32 : MMAP_VA_BITS_64)
> > > > with MMAP_VA_BITS_64 being either 48 or 37.
> > > >
> > > > In compat mode,
> > > > STACK_TOP = 1 << (32 - 1)       -> 0x80000000
> > > > TASK_SIZE = 0x8000000 - 4k      -> 0x7ffff000
> > > >
> > > > IIUC, your suggestion is to make TASK_SIZE = STACK_TOP in compat mode only.
> > > Yes, it causes the problem, which causes the boot to fail.
> >
> > I think what Leonardo is getting at is that it is odd that it would
> > cause boot issues if TASK_SIZE is not equal STACK_TOP. This seems
> > indicative of a different problem. While this may fix the issue, it
> > should be valid for TASK_SIZE to be less than STACK_TOP.
> >
> > - Charlie
> >
>
> That is also a good point, but I am not that acquainted to this to
> actually propose this.
>
> I was thinking more on these questions:
> Is TASK_SIZE and STACK_TOP related somehow?
> If so, would not be better to describe one in terms of the other, like
> #define TASK_SIZE (STACK_TOP - PAGE_SIZE)
TASK_SIZE means the maximum user address space, so it's the limitation
to any kind of mmap or stack ...
So STACK_TOP <= TASK_SIZE

Follow your idea. The question is:
#define TASK_SIZE ((UL(1) << (VA_BITS - 1)) - PAGE_SIZE)

Do we need to reserve one page between userspace & kernel?


>
> Or the other way around.
>
> I mean, if they have any relation it would be much easier to represent them
> that way, and it would avoid having two magical numbers.
>
> Thanks!
> Leo
>
> > >
> > > >
> > > > Then why not:
> > > > #ifdef CONFIG_COMPAT
> > > > #define TASK_SIZE_32    STACK_TOP
> > > Yes, it's the solution that I think at first. But I didn't find any
> > > problem with 0x7ffff000 ~ 0x80000000, and then I removed this gap to
> > > unify it with Sv39 and Sv48.
> > >
> > > >
> > > > With some comments explaining why there is no need to reserve a PAGE_SIZE
> > > > in the TASK_SIZE_32.
> > > At first, I wanted to put a invalid page between the user & kernel
> > > space, but it seems useless.
> > >
> > > >
> > > > Does that make sense?
> > > >
> > > > Thanks!
> > > > Leo
> > > >
> > > > >
> > > > > Freeing unused kernel image (initmem) memory: 2236K
> > > > > Run /sbin/init as init process
> > > > > Starting init: /sbin/init exists but couldn't execute it (error -14)
> > > > > Run /etc/init as init process
> > > > > ...
> > > > >
> > > > > Cc: stable@...r.kernel.org
> > > > > Fixes: add2cc6b6515 ("RISC-V: mm: Restrict address space for sv39,sv48,sv57")
> > > > > Signed-off-by: Guo Ren <guoren@...ux.alibaba.com>
> > > > > Signed-off-by: Guo Ren <guoren@...nel.org>
> > > > > ---
> > > > >  arch/riscv/include/asm/pgtable.h | 2 +-
> > > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > >
> > > > > diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
> > > > > index ab00235b018f..74ffb2178f54 100644
> > > > > --- a/arch/riscv/include/asm/pgtable.h
> > > > > +++ b/arch/riscv/include/asm/pgtable.h
> > > > > @@ -881,7 +881,7 @@ static inline pte_t pte_swp_clear_exclusive(pte_t pte)
> > > > >  #define TASK_SIZE_MIN        (PGDIR_SIZE_L3 * PTRS_PER_PGD / 2)
> > > > >
> > > > >  #ifdef CONFIG_COMPAT
> > > > > -#define TASK_SIZE_32 (_AC(0x80000000, UL) - PAGE_SIZE)
> > > > > +#define TASK_SIZE_32 (_AC(0x80000000, UL))
> > > >
> > > >
> > > >
> > > >
> > > > >  #define TASK_SIZE    (test_thread_flag(TIF_32BIT) ? \
> > > > >                        TASK_SIZE_32 : TASK_SIZE_64)
> > > > >  #else
> > > > > --
> > > > > 2.40.1
> > > > >
> > > >
> > >
> > >
> > > --
> > > Best Regards
> > >  Guo Ren
> >
>


--
Best Regards
 Guo Ren

Powered by blists - more mailing lists