[PATCH] arm64: extable: fix null deref in load_unaligned_zeropad.

[PATCH] arm64: extable: fix null deref in load_unaligned_zeropad.

[Evgenii Stepanov]

ex_handler_load_unaligned_zeropad extracts the source and data register
numbers from the wrong field of the exception table.

Fixes: 753b3236
Signed-off-by: Evgenii Stepanov <eugenis@google.com>
---
 arch/arm64/mm/extable.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/mm/extable.c b/arch/arm64/mm/extable.c
index c0181e60cc98..489455309695 100644
--- a/arch/arm64/mm/extable.c
+++ b/arch/arm64/mm/extable.c
@@ -40,8 +40,8 @@ static bool
 ex_handler_load_unaligned_zeropad(const struct exception_table_entry *ex,
 				  struct pt_regs *regs)
 {
-	int reg_data = FIELD_GET(EX_DATA_REG_DATA, ex->type);
-	int reg_addr = FIELD_GET(EX_DATA_REG_ADDR, ex->type);
+	int reg_data = FIELD_GET(EX_DATA_REG_DATA, ex->data);
+	int reg_addr = FIELD_GET(EX_DATA_REG_ADDR, ex->data);
 	unsigned long data, addr, offset;
 
 	addr = pt_regs_read_reg(regs, reg_addr);
-- 
2.35.0.rc0.227.g00780c9af4-goog

Re: [PATCH] arm64: extable: fix null deref in load_unaligned_zeropad.

[Catalin Marinas]

On Fri, Jan 21, 2022 at 06:34:47PM -0800, Evgenii Stepanov wrote:
> ex_handler_load_unaligned_zeropad extracts the source and data register
> numbers from the wrong field of the exception table.

Could you please also include the stack dump if you get one from the
null pointer dereference?

> Fixes: 753b3236

This should be expanded:

Fixes: 753b32368705 ("arm64: extable: add load_unaligned_zeropad() handler")
Cc: <stable@vger.kernel.org> # 5.16.x

Thanks.

-- 
Catalin

Re: [PATCH] arm64: extable: fix null deref in load_unaligned_zeropad.

[Mark Rutland]

On Fri, Jan 21, 2022 at 06:34:47PM -0800, Evgenii Stepanov wrote:
> ex_handler_load_unaligned_zeropad extracts the source and data register
> numbers from the wrong field of the exception table.

Ouch. Did you find this by inspection, or did this show up in testing?

Sorry about this.

I think we should be a little more explicit as to exactly what goes wrong. How
about:

| In ex_handler_load_unaligned_zeropad() we erroneously extract the data and
| addr register indices from ex->type rather than ex->data. As ex->type will
| contain EX_TYPE_LOAD_UNALIGNED_ZEROPAD (i.e. 4):
| 
| * We'll always treat X0 as the address register, since EX_DATA_REG_ADDR is
|   extracted from bits [9:5]. Thus, we may attempt to dereference an arbitrary
|   address as X0 may hold an arbitary value.
| 
| * We'll always treat X4 as the data register, since EX_DATA_REG_DATA is
|   extracted from bits [4:0]. Thus we will corrupt X4 and cause arbitrary
|   behaviour within load_unaligned_zeropad() and its caller.
| 
| Fix this by extracting both values from ex->data as originally intended.

> Fixes: 753b3236

That should be expanded, e.g.

  Fixes: 753b32368705c396 ("arm64: extable: add load_unaligned_zeropad() handler")

With those changes:

Reviewed-by: Mark Rutland <mark.rutland@arm.com>

Thanks,
Mark.

> Signed-off-by: Evgenii Stepanov <eugenis@google.com>
> ---
>  arch/arm64/mm/extable.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/arm64/mm/extable.c b/arch/arm64/mm/extable.c
> index c0181e60cc98..489455309695 100644
> --- a/arch/arm64/mm/extable.c
> +++ b/arch/arm64/mm/extable.c
> @@ -40,8 +40,8 @@ static bool
>  ex_handler_load_unaligned_zeropad(const struct exception_table_entry *ex,
>  				  struct pt_regs *regs)
>  {
> -	int reg_data = FIELD_GET(EX_DATA_REG_DATA, ex->type);
> -	int reg_addr = FIELD_GET(EX_DATA_REG_ADDR, ex->type);
> +	int reg_data = FIELD_GET(EX_DATA_REG_DATA, ex->data);
> +	int reg_addr = FIELD_GET(EX_DATA_REG_ADDR, ex->data);
>  	unsigned long data, addr, offset;
>  
>  	addr = pt_regs_read_reg(regs, reg_addr);
> -- 
> 2.35.0.rc0.227.g00780c9af4-goog
> 

Re: [PATCH] arm64: extable: fix null deref in load_unaligned_zeropad.

[Mark Rutland]

On Tue, Jan 25, 2022 at 03:23:20PM +0000, Mark Rutland wrote:
> On Fri, Jan 21, 2022 at 06:34:47PM -0800, Evgenii Stepanov wrote:
> > ex_handler_load_unaligned_zeropad extracts the source and data register
> > numbers from the wrong field of the exception table.
> 
> Ouch. Did you find this by inspection, or did this show up in testing?
> 
> Sorry about this.
> 
> I think we should be a little more explicit as to exactly what goes wrong. How
> about:
> 
> | In ex_handler_load_unaligned_zeropad() we erroneously extract the data and
> | addr register indices from ex->type rather than ex->data. As ex->type will
> | contain EX_TYPE_LOAD_UNALIGNED_ZEROPAD (i.e. 4):
> | 
> | * We'll always treat X0 as the address register, since EX_DATA_REG_ADDR is
> |   extracted from bits [9:5]. Thus, we may attempt to dereference an arbitrary
> |   address as X0 may hold an arbitary value.
> | 
> | * We'll always treat X4 as the data register, since EX_DATA_REG_DATA is
> |   extracted from bits [4:0]. Thus we will corrupt X4 and cause arbitrary
> |   behaviour within load_unaligned_zeropad() and its caller.
> | 
> | Fix this by extracting both values from ex->data as originally intended.
> 
> > Fixes: 753b3236
> 
> That should be expanded, e.g.
> 
>   Fixes: 753b32368705c396 ("arm64: extable: add load_unaligned_zeropad() handler")
> 
> With those changes:
> 
> Reviewed-by: Mark Rutland <mark.rutland@arm.com>

Looking again, sicne this isn't jsut a null-deref, can we also rework the
title, something like:

| arm64: extable: fix load_unaligned_zeropad() reg indices

Thanks,
Mark.

Re: [PATCH] arm64: extable: fix null deref in load_unaligned_zeropad.

[Evgenii Stepanov]

On Tue, Jan 25, 2022 at 7:50 AM Mark Rutland <mark.rutland@arm.com> wrote:
>
> On Tue, Jan 25, 2022 at 03:23:20PM +0000, Mark Rutland wrote:
> > On Fri, Jan 21, 2022 at 06:34:47PM -0800, Evgenii Stepanov wrote:
> > > ex_handler_load_unaligned_zeropad extracts the source and data register
> > > numbers from the wrong field of the exception table.
> >
> > Ouch. Did you find this by inspection, or did this show up in testing?
> >
> > Sorry about this.
> >
> > I think we should be a little more explicit as to exactly what goes wrong. How
> > about:
> >
> > | In ex_handler_load_unaligned_zeropad() we erroneously extract the data and
> > | addr register indices from ex->type rather than ex->data. As ex->type will
> > | contain EX_TYPE_LOAD_UNALIGNED_ZEROPAD (i.e. 4):
> > |
> > | * We'll always treat X0 as the address register, since EX_DATA_REG_ADDR is
> > |   extracted from bits [9:5]. Thus, we may attempt to dereference an arbitrary
> > |   address as X0 may hold an arbitary value.
> > |
> > | * We'll always treat X4 as the data register, since EX_DATA_REG_DATA is
> > |   extracted from bits [4:0]. Thus we will corrupt X4 and cause arbitrary
> > |   behaviour within load_unaligned_zeropad() and its caller.
> > |
> > | Fix this by extracting both values from ex->data as originally intended.
> >
> > > Fixes: 753b3236
> >
> > That should be expanded, e.g.
> >
> >   Fixes: 753b32368705c396 ("arm64: extable: add load_unaligned_zeropad() handler")
> >
> > With those changes:
> >
> > Reviewed-by: Mark Rutland <mark.rutland@arm.com>
>
> Looking again, sicne this isn't jsut a null-deref, can we also rework the
> title, something like:
>
> | arm64: extable: fix load_unaligned_zeropad() reg indices

That's a much better commit message, thank you! I'll upload v2 shortly.

This was found by updating to a newer QEMU that correctly delivers MTE
faults from unaligned memory accesses, and triggers this bug reliably
during Android boot. I'll add a stack trace to v2.

>
> Thanks,
> Mark.

Re: [PATCH] arm64: extable: fix null deref in load_unaligned_zeropad.

[Mark Rutland]

On Tue, Jan 25, 2022 at 08:59:07AM -0800, Evgenii Stepanov wrote:
> On Tue, Jan 25, 2022 at 7:50 AM Mark Rutland <mark.rutland@arm.com> wrote:
> >
> > On Tue, Jan 25, 2022 at 03:23:20PM +0000, Mark Rutland wrote:
> > > On Fri, Jan 21, 2022 at 06:34:47PM -0800, Evgenii Stepanov wrote:
> > > > ex_handler_load_unaligned_zeropad extracts the source and data register
> > > > numbers from the wrong field of the exception table.
> > >
> > > Ouch. Did you find this by inspection, or did this show up in testing?
> > >
> > > Sorry about this.
> > >
> > > I think we should be a little more explicit as to exactly what goes wrong. How
> > > about:
> > >
> > > | In ex_handler_load_unaligned_zeropad() we erroneously extract the data and
> > > | addr register indices from ex->type rather than ex->data. As ex->type will
> > > | contain EX_TYPE_LOAD_UNALIGNED_ZEROPAD (i.e. 4):
> > > |
> > > | * We'll always treat X0 as the address register, since EX_DATA_REG_ADDR is
> > > |   extracted from bits [9:5]. Thus, we may attempt to dereference an arbitrary
> > > |   address as X0 may hold an arbitary value.
> > > |
> > > | * We'll always treat X4 as the data register, since EX_DATA_REG_DATA is
> > > |   extracted from bits [4:0]. Thus we will corrupt X4 and cause arbitrary
> > > |   behaviour within load_unaligned_zeropad() and its caller.
> > > |
> > > | Fix this by extracting both values from ex->data as originally intended.
> > >
> > > > Fixes: 753b3236
> > >
> > > That should be expanded, e.g.
> > >
> > >   Fixes: 753b32368705c396 ("arm64: extable: add load_unaligned_zeropad() handler")
> > >
> > > With those changes:
> > >
> > > Reviewed-by: Mark Rutland <mark.rutland@arm.com>
> >
> > Looking again, sicne this isn't jsut a null-deref, can we also rework the
> > title, something like:
> >
> > | arm64: extable: fix load_unaligned_zeropad() reg indices
> 
> That's a much better commit message, thank you! I'll upload v2 shortly.
> 
> This was found by updating to a newer QEMU that correctly delivers MTE
> faults from unaligned memory accesses, and triggers this bug reliably
> during Android boot. I'll add a stack trace to v2.

That'd be great, thanks!

Mark.