2024-02-28 09:06:31

by Dawei Li

[permalink] [raw]
Subject: [PATCH] mk68k: Fix broken THREAD_SIZE_ORDER

Current THREAD_SIZE_ORDER implementation for m68k is incorrect, fix it
by ilog2().

Fixes: cddafa3500fd ("m68k/m68knommu: merge MMU and non-MMU thread_info.h")
Signed-off-by: Dawei Li <[email protected]>
Cc: [email protected]
---
arch/m68k/include/asm/thread_info.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/m68k/include/asm/thread_info.h b/arch/m68k/include/asm/thread_info.h
index 31be2ad999ca..50faecd6fc5f 100644
--- a/arch/m68k/include/asm/thread_info.h
+++ b/arch/m68k/include/asm/thread_info.h
@@ -19,7 +19,8 @@
#else
#define THREAD_SIZE PAGE_SIZE
#endif
-#define THREAD_SIZE_ORDER ((THREAD_SIZE / PAGE_SIZE) - 1)
+
+#define THREAD_SIZE_ORDER ilog2(THREAD_SIZE / PAGE_SIZE)

#ifndef __ASSEMBLY__

--
2.27.0



2024-02-28 09:11:25

by Jiri Slaby

[permalink] [raw]
Subject: Re: [PATCH] mk68k: Fix broken THREAD_SIZE_ORDER

On 28. 02. 24, 9:58, Dawei Li wrote:
> Current THREAD_SIZE_ORDER implementation for m68k is incorrect, fix it
> by ilog2().

This is not a good commit log. Incorrect in what way and why is the
fixed version correct? And what is affected? Note you're referring to a
change which was done 14 years ago. It definitely must not be that
incorrect (for everybody).

> Fixes: cddafa3500fd ("m68k/m68knommu: merge MMU and non-MMU thread_info.h")
> Signed-off-by: Dawei Li <[email protected]>
> Cc: [email protected]
> ---
> arch/m68k/include/asm/thread_info.h | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/arch/m68k/include/asm/thread_info.h b/arch/m68k/include/asm/thread_info.h
> index 31be2ad999ca..50faecd6fc5f 100644
> --- a/arch/m68k/include/asm/thread_info.h
> +++ b/arch/m68k/include/asm/thread_info.h
> @@ -19,7 +19,8 @@
> #else
> #define THREAD_SIZE PAGE_SIZE
> #endif
> -#define THREAD_SIZE_ORDER ((THREAD_SIZE / PAGE_SIZE) - 1)
> +
> +#define THREAD_SIZE_ORDER ilog2(THREAD_SIZE / PAGE_SIZE)
>
> #ifndef __ASSEMBLY__
>

--
js
suse labs


2024-02-28 09:38:06

by Dawei Li

[permalink] [raw]
Subject: Re: [PATCH] mk68k: Fix broken THREAD_SIZE_ORDER

Hi Jiri,

Thanks for quick review.

On Wed, Feb 28, 2024 at 10:11:05AM +0100, Jiri Slaby wrote:
> On 28. 02. 24, 9:58, Dawei Li wrote:
> > Current THREAD_SIZE_ORDER implementation for m68k is incorrect, fix it
> > by ilog2().
>
> This is not a good commit log. Incorrect in what way and why is the fixed

Agreed.

> version correct? And what is affected? Note you're referring to a change
> which was done 14 years ago. It definitely must not be that incorrect (for
> everybody).

It's 'right' just for current PAGE_SIZE & THREAD_SIZE configs:

// arch/m68k/include/asm/thread_info.h
#if PAGE_SHIFT < 13
#ifdef CONFIG_4KSTACKS
#define THREAD_SIZE 4096
#else
#define THREAD_SIZE 8192
#endif
#else
#define THREAD_SIZE PAGE_SIZE
#endif
#define THREAD_SIZE_ORDER ((THREAD_SIZE / PAGE_SIZE) - 1)

// arch/m68k/include/asm/page.h
#if defined(CONFIG_SUN3) || defined(CONFIG_COLDFIRE)
#define PAGE_SHIFT 13
#else
#define PAGE_SHIFT 12
#endif
#define PAGE_SIZE (_AC(1, UL) << PAGE_SHIFT)
#define PAGE_MASK (~(PAGE_SIZE-1))
#define PAGE_OFFSET (PAGE_OFFSET_RAW)

But it's incorrect in generic/mathematical way.

Thanks,

Dawei

>
> > Fixes: cddafa3500fd ("m68k/m68knommu: merge MMU and non-MMU thread_info.h")
> > Signed-off-by: Dawei Li <[email protected]>
> > Cc: [email protected]
> > ---
> > arch/m68k/include/asm/thread_info.h | 3 ++-
> > 1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/arch/m68k/include/asm/thread_info.h b/arch/m68k/include/asm/thread_info.h
> > index 31be2ad999ca..50faecd6fc5f 100644
> > --- a/arch/m68k/include/asm/thread_info.h
> > +++ b/arch/m68k/include/asm/thread_info.h
> > @@ -19,7 +19,8 @@
> > #else
> > #define THREAD_SIZE PAGE_SIZE
> > #endif
> > -#define THREAD_SIZE_ORDER ((THREAD_SIZE / PAGE_SIZE) - 1)
> > +
> > +#define THREAD_SIZE_ORDER ilog2(THREAD_SIZE / PAGE_SIZE)
> > #ifndef __ASSEMBLY__
>
> --
> js
> suse labs
>
>

2024-02-28 10:09:31

by Geert Uytterhoeven

[permalink] [raw]
Subject: Re: [PATCH] mk68k: Fix broken THREAD_SIZE_ORDER

Hi Dawei,

On Wed, Feb 28, 2024 at 10:34 AM Dawei Li <[email protected]> wrote:
> On Wed, Feb 28, 2024 at 10:11:05AM +0100, Jiri Slaby wrote:
> > On 28. 02. 24, 9:58, Dawei Li wrote:
> > > Current THREAD_SIZE_ORDER implementation for m68k is incorrect, fix it
> > > by ilog2().
> >
> > This is not a good commit log. Incorrect in what way and why is the fixed
>
> Agreed.
>
> > version correct? And what is affected? Note you're referring to a change
> > which was done 14 years ago. It definitely must not be that incorrect (for
> > everybody).
>
> It's 'right' just for current PAGE_SIZE & THREAD_SIZE configs:
>
> // arch/m68k/include/asm/thread_info.h
> #if PAGE_SHIFT < 13
> #ifdef CONFIG_4KSTACKS
> #define THREAD_SIZE 4096
> #else
> #define THREAD_SIZE 8192
> #endif
> #else
> #define THREAD_SIZE PAGE_SIZE
> #endif
> #define THREAD_SIZE_ORDER ((THREAD_SIZE / PAGE_SIZE) - 1)

Indeed. The only supported values for THREAD_SIZE_ORDER on m68k are
0 and 1.

> But it's incorrect in generic/mathematical way.

True. But does it matter much?
I.e. do you plan to add support for larger values of THREAD_SIZE?

What about changing the #ifdeffery to set THREAD_SIZE_ORDER to an
explicit value, and calculating THREAD_SIZE from THREAD_SIZE_ORDER
instead?

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68korg

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds