Hi Guys
We found an issue with last kernel tag v5.18-rc1 on stm32f746-disco and
stm32h743-disco boards (ARMV7-M SoCs).
Kernel hangs when executing SetPageUptodate(ZERO_PAGE(0)); in mm/filemap.c.
By reverting commit 56a8c8eb1eaf ("tmpfs: do not allocate pages on read"),
kernel boots without any issue.
Thanks
Patrice
Hi Geert,
On 21/4/22 00:44, Geert Uytterhoeven wrote:
> On Wed, Apr 20, 2022 at 3:53 PM Greg Ungerer <[email protected]> wrote:
>> On 16/4/22 10:58, Hugh Dickins wrote:
>>> Just to wrap up this thread: the tentative arch/ patches below did not
>>> go into 5.18-rc2, but 5.18-rc3 will contain
>>> 1bdec44b1eee ("tmpfs: fix regressions from wider use of ZERO_PAGE")
>>> which fixes a further issue, and deletes the line which gave you trouble.
>>>
>>> With arch/h8300 removed from linux-next, and arch/arm losing a page by
>>> the patch below, I don't think it's worth my arguing for those changes.
>>> I'd still prefer arch/m68k to expose its empty_zero_page in ZERO_PAGE(),
>>> or else not allocate it; but I won't be pursuing this further.
>>
>> Thanks for pointing this out. It certainly does look wrong to me for
>> the m68k nommu case. I am not aware of any existing issues caused by
>> this - but there is no good reason not to fix it.
>>
>> So I propose this change. Build and run tested on my m68knommu targets.
>>
>> Regards
>> Greg
>>
>>
>> From f809fb8fbca9e5e637b8fda380955bd799bb3926 Mon Sep 17 00:00:00 2001
>> From: Greg Ungerer <[email protected]>
>> Date: Wed, 20 Apr 2022 23:27:47 +1000
>> Subject: [PATCH] m68knommu: set ZERO_PAGE() allocated zeroed page
>>
>> The non-MMU m68k pagetable ZERO_PAGE() macro is being set to the
>> somewhat non-sensical value of "virt_to_page(0)". The zeroth page
>> is not in any way guaranteed to be a page full of "0". So the result
>> is that ZERO_PAGE() will almost certainly contain random values.
>>
>> We already allocate a real "empty_zero_page" in the mm setup code shared
>> between MMU m68k and non-MMU m68k. It is just not hooked up to the
>> ZERO_PAGE() macro for the non-MMU m68k case.
>>
>> Fix ZERO_PAGE() to use the allocated "empty_zero_page" pointer.
>>
>> I am not aware of any specific issues caused by the old code.
>>
>> Link: https://lore.kernel.org/linux-m68k/[email protected]/T/#t
>> Reported-by: Hugh Dickens <[email protected]>
>> Signed-off-by: Greg Ungerer <[email protected]>
>> ---
>> arch/m68k/include/asm/pgtable_no.h | 3 ++-
>> 1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/arch/m68k/include/asm/pgtable_no.h b/arch/m68k/include/asm/pgtable_no.h
>> index 87151d67d91e..bce5ca56c388 100644
>> --- a/arch/m68k/include/asm/pgtable_no.h
>> +++ b/arch/m68k/include/asm/pgtable_no.h
>> @@ -42,7 +42,8 @@ extern void paging_init(void);
>> * ZERO_PAGE is a global shared page that is always zero: used
>> * for zero-mapped memory areas etc..
>> */
>> -#define ZERO_PAGE(vaddr) (virt_to_page(0))
>> +extern void *empty_zero_page;
>> +#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
>>
>> /*
>> * All 32bit addresses are effectively valid for vmalloc...
>
> And after that (or combined with this?), this can be factored
> out from arch/m68k/include/asm/pgtable_{mm,no}.h into
> arch/m68k/include/asm/pgtable.h.
I think a new patch to do that work on top of this one would be best.
I will work on that.
Regards
Greg
> Gr{oetje,eeting}s,
>
> Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
>
> 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
Hi Greg,
On Wed, Apr 20, 2022 at 3:53 PM Greg Ungerer <[email protected]> wrote:
> On 16/4/22 10:58, Hugh Dickins wrote:
> > Just to wrap up this thread: the tentative arch/ patches below did not
> > go into 5.18-rc2, but 5.18-rc3 will contain
> > 1bdec44b1eee ("tmpfs: fix regressions from wider use of ZERO_PAGE")
> > which fixes a further issue, and deletes the line which gave you trouble.
> >
> > With arch/h8300 removed from linux-next, and arch/arm losing a page by
> > the patch below, I don't think it's worth my arguing for those changes.
> > I'd still prefer arch/m68k to expose its empty_zero_page in ZERO_PAGE(),
> > or else not allocate it; but I won't be pursuing this further.
>
> Thanks for pointing this out. It certainly does look wrong to me for
> the m68k nommu case. I am not aware of any existing issues caused by
> this - but there is no good reason not to fix it.
>
> So I propose this change. Build and run tested on my m68knommu targets.
>
> Regards
> Greg
>
>
> From f809fb8fbca9e5e637b8fda380955bd799bb3926 Mon Sep 17 00:00:00 2001
> From: Greg Ungerer <[email protected]>
> Date: Wed, 20 Apr 2022 23:27:47 +1000
> Subject: [PATCH] m68knommu: set ZERO_PAGE() allocated zeroed page
>
> The non-MMU m68k pagetable ZERO_PAGE() macro is being set to the
> somewhat non-sensical value of "virt_to_page(0)". The zeroth page
> is not in any way guaranteed to be a page full of "0". So the result
> is that ZERO_PAGE() will almost certainly contain random values.
>
> We already allocate a real "empty_zero_page" in the mm setup code shared
> between MMU m68k and non-MMU m68k. It is just not hooked up to the
> ZERO_PAGE() macro for the non-MMU m68k case.
>
> Fix ZERO_PAGE() to use the allocated "empty_zero_page" pointer.
>
> I am not aware of any specific issues caused by the old code.
>
> Link: https://lore.kernel.org/linux-m68k/[email protected]/T/#t
> Reported-by: Hugh Dickens <[email protected]>
> Signed-off-by: Greg Ungerer <[email protected]>
> ---
> arch/m68k/include/asm/pgtable_no.h | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/arch/m68k/include/asm/pgtable_no.h b/arch/m68k/include/asm/pgtable_no.h
> index 87151d67d91e..bce5ca56c388 100644
> --- a/arch/m68k/include/asm/pgtable_no.h
> +++ b/arch/m68k/include/asm/pgtable_no.h
> @@ -42,7 +42,8 @@ extern void paging_init(void);
> * ZERO_PAGE is a global shared page that is always zero: used
> * for zero-mapped memory areas etc..
> */
> -#define ZERO_PAGE(vaddr) (virt_to_page(0))
> +extern void *empty_zero_page;
> +#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
>
> /*
> * All 32bit addresses are effectively valid for vmalloc...
And after that (or combined with this?), this can be factored
out from arch/m68k/include/asm/pgtable_{mm,no}.h into
arch/m68k/include/asm/pgtable.h.
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
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