Re: [PATCH] ArmPkg/ArmMmuLib ARM: fix Mva to use idx instead of table base

[Ard Biesheuvel <ard.biesheuvel@linaro.org>]

On 20 June 2018 at 21:09, Ard Biesheuvel <ard.biesheuvel@linaro.org> wrote:
> On 20 June 2018 at 18:39, Ard Biesheuvel <ard.biesheuvel@linaro.org> wrote:
>> On 19 June 2018 at 22:52, Chris Co <Christopher.Co@microsoft.com> wrote:
>>> Hi,
>>>
>>> Just checking if there is anything needed on my end to get this patch merged in.
>>>
>>
>> Well, the patch looks obviously correct, but I just tested it and it
>> breaks ArmVirtQemu running in 32-bit mode.
>>
>> I will investigate
>>
>
> OK, I found the issue, it is not a hang but a very long stall. Patch incoming.
>

Now pushed as 8e586296c114f630188cfe4c76df91a1e2b7a5b2

>>>> -----Original Message-----
>>>> From: Ard Biesheuvel <ard.biesheuvel@linaro.org>
>>>> Sent: Thursday, April 19, 2018 5:30 AM
>>>> To: Chris Co <Christopher.Co@microsoft.com>
>>>> Cc: Leif Lindholm <leif.lindholm@linaro.org>; edk2-devel@lists.01.org
>>>> Subject: Re: [PATCH] ArmPkg/ArmMmuLib ARM: fix Mva to use idx instead
>>>> of table base
>>>>
>>>> On 16 April 2018 at 21:45, Chris Co <Christopher.Co@microsoft.com> wrote:
>>>> > Hi Leif,
>>>> >
>>>> >> -----Original Message-----
>>>> >> From: Leif Lindholm <leif.lindholm@linaro.org>
>>>> >> Sent: Monday, April 16, 2018 3:44 AM
>>>> >> To: Chris Co <Christopher.Co@microsoft.com>
>>>> >> Cc: edk2-devel@lists.01.org; Ard Biesheuvel
>>>> >> <ard.biesheuvel@linaro.org>
>>>> >> Subject: Re: [PATCH] ArmPkg/ArmMmuLib ARM: fix Mva to use idx
>>>> instead
>>>> >> of table base
>>>> >>
>>>> >> On Fri, Apr 13, 2018 at 11:43:27PM +0000, Chris Co wrote:
>>>> >> > Mva address calculation should use the left-shifted current section
>>>> >> > index instead of the left-shifted table base address.
>>>> >> >
>>>> >> > Using the table base address here has the side-effect of
>>>> >> > potentially causing an access violation depending on the base address
>>>> value.
>>>> >> >
>>>> >> > Cc: Leif Lindholm <leif.lindholm@linaro.org>
>>>> >> > Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
>>>> >> > Contributed-under: TianoCore Contribution Agreement 1.1
>>>> >> > Signed-off-by: Christopher Co <christopher.co@microsoft.com>
>>>> >> > ---
>>>> >> >  ArmPkg/Library/ArmMmuLib/Arm/ArmMmuLibCore.c | 2 +-
>>>> >> >  1 file changed, 1 insertion(+), 1 deletion(-)
>>>> >> >
>>>> >> > diff --git a/ArmPkg/Library/ArmMmuLib/Arm/ArmMmuLibCore.c
>>>> >> > b/ArmPkg/Library/ArmMmuLib/Arm/ArmMmuLibCore.c
>>>> >> > index 774a7ccf59..9bf4ba03fd 100644
>>>> >> > --- a/ArmPkg/Library/ArmMmuLib/Arm/ArmMmuLibCore.c
>>>> >> > +++ b/ArmPkg/Library/ArmMmuLib/Arm/ArmMmuLibCore.c
>>>> >> > @@ -716,7 +716,7 @@ UpdateSectionEntries (
>>>> >> >        Descriptor |= EntryValue;
>>>> >> >
>>>> >> >        if (CurrentDescriptor  != Descriptor) {
>>>> >> > -        Mva = (VOID *)(UINTN)(((UINTN)FirstLevelTable) <<
>>>> >> TT_DESCRIPTOR_SECTION_BASE_SHIFT);
>>>> >> > +        Mva = (VOID *)(UINTN)(((UINTN)FirstLevelIdx + i) <<
>>>> >> > + TT_DESCRIPTOR_SECTION_BASE_SHIFT);
>>>> >>
>>>> >> So, this clearly looks like you've found a bug - thanks!
>>>> >>
>>>> >> But I am a little bit confused about the patch - should this not need
>>>> >> to incorporate the descriptor size in some way?
>>>> >> I.e. something like
>>>> >>   Mva = (VOID *)(UINTN)(((UINTN)FirstLevelIdx + (i * sizeof(UINTN)))
>>>> >> << TT_DESCRIPTOR_SECTION_BASE_SHIFT);
>>>> >> or
>>>> >>   ...                           &FirstLevelTable[FirstLevelIndex + i] ...
>>>> >>
>>>> >> ?
>>>> >>
>>>> >> Regards,
>>>> >>
>>>> >> Leif
>>>> >>
>>>> > I don't think descriptor size is needed here.
>>>> >
>>>> > My understanding is that Mva is the base address of the current section.
>>>> >
>>>> > FirstLevelidx is derived by the first section's BaseAddress >> 20.
>>>> > The current section index is then (FirstLevelIdx + i), which makes the
>>>> > base address of the current section (FirstLeveLidx + i) << 20.
>>>> >
>>>>
>>>> Indeed. 'Index' is a bit misleading here, given that it is the top level index into
>>>> the entire VA space, and so it is congruent with the virtual base address
>>>> itself. The use of 'FirstLevelTable' in this context is obviously incorrect, given
>>>> that it refers to the [physical] address of the page tables itself, not to the
>>>> virtual region they describe.
>>>>
>>>> Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>