2014-07-01 02:14:32

by Nicholas Krause

Subject: [PATCH v2] hid-appleir: Fix kernel panic due to null pointer

Fixes a null pointer in appleir_input_configured due to reading
into wrong size array. Changed the variable to input_dev->keycodemax.

Signed-off-by: Nicholas Krause <[email protected]>
---
drivers/hid/hid-appleir.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/hid/hid-appleir.c b/drivers/hid/hid-appleir.c
index 0e6a42d..cc02df4 100644
--- a/drivers/hid/hid-appleir.c
+++ b/drivers/hid/hid-appleir.c
@@ -272,7 +272,7 @@ static void appleir_input_configured(struct hid_device *hid,
input_dev->evbit[0] = BIT(EV_KEY) | BIT(EV_REP);

memcpy(appleir->keymap, appleir_key_table, sizeof(appleir->keymap));
- for (i = 0; i < ARRAY_SIZE(appleir_key_table); i++)
+ for (i = 0; i < appleir->keymap; i++)
set_bit(appleir->keymap[i], input_dev->keybit);
clear_bit(KEY_RESERVED, input_dev->keybit);
}
--
1.9.1

2014-07-01 06:39:12

by Jiri Kosina

Subject: Re: [PATCH v2] hid-appleir: Fix kernel panic due to null pointer

On Mon, 30 Jun 2014, Nicholas Krause wrote:

> Fixes a null pointer in appleir_input_configured due to reading
> into wrong size array. Changed the variable to input_dev->keycodemax.

This is a stale changelog from the previous buggy version.

> Signed-off-by: Nicholas Krause <[email protected]>
> ---
> drivers/hid/hid-appleir.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/hid/hid-appleir.c b/drivers/hid/hid-appleir.c
> index 0e6a42d..cc02df4 100644
> --- a/drivers/hid/hid-appleir.c
> +++ b/drivers/hid/hid-appleir.c
> @@ -272,7 +272,7 @@ static void appleir_input_configured(struct hid_device *hid,
> input_dev->evbit[0] = BIT(EV_KEY) | BIT(EV_REP);
>
> memcpy(appleir->keymap, appleir_key_table, sizeof(appleir->keymap));
> - for (i = 0; i < ARRAY_SIZE(appleir_key_table); i++)
> + for (i = 0; i < appleir->keymap; i++)

This is wrong. appleir->keymap is an array, and you want to count its
elements.

Third time is the charm perhaps?

--
Jiri Kosina
SUSE Labs

2014-07-01 06:50:32

by Mateusz Guzik

Subject: Re: [PATCH v2] hid-appleir: Fix kernel panic due to null pointer

On Tue, Jul 01, 2014 at 08:39:06AM +0200, Jiri Kosina wrote:
> On Mon, 30 Jun 2014, Nicholas Krause wrote:
>
> > Fixes a null pointer in appleir_input_configured due to reading
> > into wrong size array. Changed the variable to input_dev->keycodemax.
>
> This is a stale changelog from the previous buggy version.
>
> > Signed-off-by: Nicholas Krause <[email protected]>
> > ---
> > drivers/hid/hid-appleir.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/hid/hid-appleir.c b/drivers/hid/hid-appleir.c
> > index 0e6a42d..cc02df4 100644
> > --- a/drivers/hid/hid-appleir.c
> > +++ b/drivers/hid/hid-appleir.c
> > @@ -272,7 +272,7 @@ static void appleir_input_configured(struct hid_device *hid,
> > input_dev->evbit[0] = BIT(EV_KEY) | BIT(EV_REP);
> >
> > memcpy(appleir->keymap, appleir_key_table, sizeof(appleir->keymap));
> > - for (i = 0; i < ARRAY_SIZE(appleir_key_table); i++)
> > + for (i = 0; i < appleir->keymap; i++)
>
> This is wrong. appleir->keymap is an array, and you want to count its
> elements.
>

But where the claim of 'null pointer' is coming from (or reading/writing
past the array for that matter)?

Replacing appleir_key_table with appleir->keymap is a noop anyway
because:
unsigned short keymap[ARRAY_SIZE(appleir_key_table)];

although one may argue is slightly nicer since it is used in memcpy line
earlier.
--
Mateusz Guzik

2014-07-01 06:55:18

by Jiri Kosina

Subject: Re: [PATCH v2] hid-appleir: Fix kernel panic due to null pointer

On Tue, 1 Jul 2014, Mateusz Guzik wrote:

> > > Fixes a null pointer in appleir_input_configured due to reading
> > > into wrong size array. Changed the variable to input_dev->keycodemax.
> >
> > This is a stale changelog from the previous buggy version.
> >
> > > Signed-off-by: Nicholas Krause <[email protected]>
> > > ---
> > > drivers/hid/hid-appleir.c | 2 +-
> > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/hid/hid-appleir.c b/drivers/hid/hid-appleir.c
> > > index 0e6a42d..cc02df4 100644
> > > --- a/drivers/hid/hid-appleir.c
> > > +++ b/drivers/hid/hid-appleir.c
> > > @@ -272,7 +272,7 @@ static void appleir_input_configured(struct hid_device *hid,
> > > input_dev->evbit[0] = BIT(EV_KEY) | BIT(EV_REP);
> > >
> > > memcpy(appleir->keymap, appleir_key_table, sizeof(appleir->keymap));
> > > - for (i = 0; i < ARRAY_SIZE(appleir_key_table); i++)
> > > + for (i = 0; i < appleir->keymap; i++)
> >
> > This is wrong. appleir->keymap is an array, and you want to count its
> > elements.
> >
>
> But where the claim of 'null pointer' is coming from (or reading/writing
> past the array for that matter)?
>
> Replacing appleir_key_table with appleir->keymap is a noop anyway
> because:
> unsigned short keymap[ARRAY_SIZE(appleir_key_table)];

It's not really equivalent. ARRAY_SIZE counts the number of short-sized
elements, sizeof() counts the number of bytes that fit into the data
structure (which is twice as much, as sizeof(unsigned short) == 2).

--
Jiri Kosina
SUSE Labs

2014-07-01 22:21:26

by Nicholas Krause

Subject: Re: [PATCH v2] hid-appleir: Fix kernel panic due to null pointer

So I am wondering them should I can it to sizeof(appleir_keymap)?
Nick

On Tue, Jul 1, 2014 at 2:55 AM, Jiri Kosina <[email protected]> wrote:
> On Tue, 1 Jul 2014, Mateusz Guzik wrote:
>
>> > > Fixes a null pointer in appleir_input_configured due to reading
>> > > into wrong size array. Changed the variable to input_dev->keycodemax.
>> >
>> > This is a stale changelog from the previous buggy version.
>> >
>> > > Signed-off-by: Nicholas Krause <[email protected]>
>> > > ---
>> > > drivers/hid/hid-appleir.c | 2 +-
>> > > 1 file changed, 1 insertion(+), 1 deletion(-)
>> > >
>> > > diff --git a/drivers/hid/hid-appleir.c b/drivers/hid/hid-appleir.c
>> > > index 0e6a42d..cc02df4 100644
>> > > --- a/drivers/hid/hid-appleir.c
>> > > +++ b/drivers/hid/hid-appleir.c
>> > > @@ -272,7 +272,7 @@ static void appleir_input_configured(struct hid_device *hid,
>> > > input_dev->evbit[0] = BIT(EV_KEY) | BIT(EV_REP);
>> > >
>> > > memcpy(appleir->keymap, appleir_key_table, sizeof(appleir->keymap));
>> > > - for (i = 0; i < ARRAY_SIZE(appleir_key_table); i++)
>> > > + for (i = 0; i < appleir->keymap; i++)
>> >
>> > This is wrong. appleir->keymap is an array, and you want to count its
>> > elements.
>> >
>>
>> But where the claim of 'null pointer' is coming from (or reading/writing
>> past the array for that matter)?
>>
>> Replacing appleir_key_table with appleir->keymap is a noop anyway
>> because:
>> unsigned short keymap[ARRAY_SIZE(appleir_key_table)];
>
> It's not really equivalent. ARRAY_SIZE counts the number of short-sized
> elements, sizeof() counts the number of bytes that fit into the data
> structure (which is twice as much, as sizeof(unsigned short) == 2).
>
> --
> Jiri Kosina
> SUSE Labs