riscv: Correct the MODULES_VADDR

  As Documentation/riscv/vm-layout.rst describes, the 2G-sized "modules, BPF"
area should lie right before the "kernel" area. But the current definition
of MODULES_VADDR isn't consistent with that, so correct it.

Before this patch, the size of "modules" from print_vm_layout() is not 2G.

[    0.000000]      modules : 0xffffffff2ff2f000 - 0xffffffffae600000 (2022 MB)
[    0.000000]       lowmem : 0xff60000000000000 - 0xff60000040000000 (1024 MB)
[    0.000000]       kernel : 0xffffffffae600000 - 0xffffffffffffffff (1305 MB)

After this patch, the size is 2G.

[    0.000000]      modules : 0xffffffff3a000000 - 0xffffffffba000000 (2048 MB)
[    0.000000]       lowmem : 0xff60000000000000 - 0xff60000040000000 (1024 MB)
[    0.000000]       kernel : 0xffffffffba000000 - 0xffffffffffffffff (1119 MB)

Signed-off-by: Song Shuai <suagrfillet@gmail.com>
---
 arch/riscv/include/asm/pgtable.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
  

Hi Song,

On Tue, Aug 1, 2023 at 1:10 PM Song Shuai <suagrfillet@gmail.com> wrote:
>
> As Documentation/riscv/vm-layout.rst describes, the 2G-sized "modules, BPF"
> area should lie right before the "kernel" area. But the current definition
> of MODULES_VADDR isn't consistent with that, so correct it.
>
> Before this patch, the size of "modules" from print_vm_layout() is not 2G.
>
> [    0.000000]      modules : 0xffffffff2ff2f000 - 0xffffffffae600000 (2022 MB)
> [    0.000000]       lowmem : 0xff60000000000000 - 0xff60000040000000 (1024 MB)
> [    0.000000]       kernel : 0xffffffffae600000 - 0xffffffffffffffff (1305 MB)
>
> After this patch, the size is 2G.
>
> [    0.000000]      modules : 0xffffffff3a000000 - 0xffffffffba000000 (2048 MB)
> [    0.000000]       lowmem : 0xff60000000000000 - 0xff60000040000000 (1024 MB)
> [    0.000000]       kernel : 0xffffffffba000000 - 0xffffffffffffffff (1119 MB)
>
> Signed-off-by: Song Shuai <suagrfillet@gmail.com>
> ---
>  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 75970ee2bda2..7c57e17fc758 100644
> --- a/arch/riscv/include/asm/pgtable.h
> +++ b/arch/riscv/include/asm/pgtable.h
> @@ -53,8 +53,8 @@
>  #ifdef CONFIG_64BIT
>  /* This is used to define the end of the KASAN shadow region */
>  #define MODULES_LOWEST_VADDR   (KERNEL_LINK_ADDR - SZ_2G)
> -#define MODULES_VADDR          (PFN_ALIGN((unsigned long)&_end) - SZ_2G)
>  #define MODULES_END            (PFN_ALIGN((unsigned long)&_start))
> +#define MODULES_VADDR          (MODULES_END - SZ_2G)
>  #endif
>
>  /*
> --
> 2.20.1
>

The documentation is approximative, the modules must stay within a 2GB
window to *all* the kernel symbols, hence the __end - 2G, not __start.
  

On Tue, Aug 1, 2023 at 8:05 AM Alexandre Ghiti <alexghiti@rivosinc.com> wrote:
>
> Hi Song,
>
> On Tue, Aug 1, 2023 at 1:10 PM Song Shuai <suagrfillet@gmail.com> wrote:
> >
> > As Documentation/riscv/vm-layout.rst describes, the 2G-sized "modules, BPF"
> > area should lie right before the "kernel" area. But the current definition
> > of MODULES_VADDR isn't consistent with that, so correct it.
> >
> > Before this patch, the size of "modules" from print_vm_layout() is not 2G.
> >
> > [    0.000000]      modules : 0xffffffff2ff2f000 - 0xffffffffae600000 (2022 MB)
> > [    0.000000]       lowmem : 0xff60000000000000 - 0xff60000040000000 (1024 MB)
> > [    0.000000]       kernel : 0xffffffffae600000 - 0xffffffffffffffff (1305 MB)
> >
> > After this patch, the size is 2G.
> >
> > [    0.000000]      modules : 0xffffffff3a000000 - 0xffffffffba000000 (2048 MB)
> > [    0.000000]       lowmem : 0xff60000000000000 - 0xff60000040000000 (1024 MB)
> > [    0.000000]       kernel : 0xffffffffba000000 - 0xffffffffffffffff (1119 MB)
> >
> > Signed-off-by: Song Shuai <suagrfillet@gmail.com>
> > ---
> >  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 75970ee2bda2..7c57e17fc758 100644
> > --- a/arch/riscv/include/asm/pgtable.h
> > +++ b/arch/riscv/include/asm/pgtable.h
> > @@ -53,8 +53,8 @@
> >  #ifdef CONFIG_64BIT
> >  /* This is used to define the end of the KASAN shadow region */
> >  #define MODULES_LOWEST_VADDR   (KERNEL_LINK_ADDR - SZ_2G)
> > -#define MODULES_VADDR          (PFN_ALIGN((unsigned long)&_end) - SZ_2G)
> >  #define MODULES_END            (PFN_ALIGN((unsigned long)&_start))
> > +#define MODULES_VADDR          (MODULES_END - SZ_2G)
> >  #endif
> >
> >  /*
> > --
> > 2.20.1
> >
>
> The documentation is approximative, the modules must stay within a 2GB
> window to *all* the kernel symbols, hence the __end - 2G, not __start.
Thus, the ftrace detour trampoline could cover all kernel symbols.
  

Guo Ren <guoren@kernel.org> 于2023年8月1日周二 23:20写道：
>
> On Tue, Aug 1, 2023 at 8:05 AM Alexandre Ghiti <alexghiti@rivosinc.com> wrote:
> >
> > Hi Song,
> >
> > On Tue, Aug 1, 2023 at 1:10 PM Song Shuai <suagrfillet@gmail.com> wrote:
> > >
> > > As Documentation/riscv/vm-layout.rst describes, the 2G-sized "modules, BPF"
> > > area should lie right before the "kernel" area. But the current definition
> > > of MODULES_VADDR isn't consistent with that, so correct it.
> > >
> > > Before this patch, the size of "modules" from print_vm_layout() is not 2G.
> > >
> > > [    0.000000]      modules : 0xffffffff2ff2f000 - 0xffffffffae600000 (2022 MB)
> > > [    0.000000]       lowmem : 0xff60000000000000 - 0xff60000040000000 (1024 MB)
> > > [    0.000000]       kernel : 0xffffffffae600000 - 0xffffffffffffffff (1305 MB)
> > >
> > > After this patch, the size is 2G.
> > >
> > > [    0.000000]      modules : 0xffffffff3a000000 - 0xffffffffba000000 (2048 MB)
> > > [    0.000000]       lowmem : 0xff60000000000000 - 0xff60000040000000 (1024 MB)
> > > [    0.000000]       kernel : 0xffffffffba000000 - 0xffffffffffffffff (1119 MB)
> > >
> > > Signed-off-by: Song Shuai <suagrfillet@gmail.com>
> > > ---
> > >  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 75970ee2bda2..7c57e17fc758 100644
> > > --- a/arch/riscv/include/asm/pgtable.h
> > > +++ b/arch/riscv/include/asm/pgtable.h
> > > @@ -53,8 +53,8 @@
> > >  #ifdef CONFIG_64BIT
> > >  /* This is used to define the end of the KASAN shadow region */
> > >  #define MODULES_LOWEST_VADDR   (KERNEL_LINK_ADDR - SZ_2G)
> > > -#define MODULES_VADDR          (PFN_ALIGN((unsigned long)&_end) - SZ_2G)
> > >  #define MODULES_END            (PFN_ALIGN((unsigned long)&_start))
> > > +#define MODULES_VADDR          (MODULES_END - SZ_2G)
> > >  #endif
> > >
> > >  /*
> > > --
> > > 2.20.1
> > >
> >
> > The documentation is approximative, the modules must stay within a 2GB
> > window to *all* the kernel symbols, hence the __end - 2G, not __start.
> Thus, the ftrace detour trampoline could cover all kernel symbols.
>
That's another view to help me understand the definition of
MODULES_VADDR,Thanks.
> --
> Best Regards
>  Guo Ren