[PATCH v2] serial: max310x: Fix out of bounds access

[PATCH v2] serial: max310x: Fix out of bounds access

[Su Kang Yin]

Max310x driver supports up to 4 UART devices but array size of
"struct max310x_one" is set to 1. That leads to out of bounds
access on UART port registration.

This patch fixed it by increase the array size to 4 which is
maximum supported UART.

Signed-off-by: Su Kang Yin <cantona@cantona.net>
---
v2: add Signed-off-by: line.

 drivers/tty/serial/max310x.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/tty/serial/max310x.c b/drivers/tty/serial/max310x.c
index 182549f..b523552 100644
--- a/drivers/tty/serial/max310x.c
+++ b/drivers/tty/serial/max310x.c
@@ -33,6 +33,9 @@
 #define MAX310X_MAJOR			204
 #define MAX310X_MINOR			209
 
+/* Maximum supported device */
+#define MAX_DEVICE			4
+
 /* MAX310X register definitions */
 #define MAX310X_RHR_REG			(0x00) /* RX FIFO */
 #define MAX310X_THR_REG			(0x00) /* TX FIFO */
@@ -273,7 +276,7 @@ struct max310x_port {
 #ifdef CONFIG_GPIOLIB
 	struct gpio_chip	gpio;
 #endif
-	struct max310x_one	p[0];
+	struct max310x_one	p[MAX_DEVICE];
 };
 
 static u8 max310x_port_read(struct uart_port *port, u8 reg)
-- 
1.9.1

Re: [PATCH v2] serial: max310x: Fix out of bounds access

[Alexander Shiyan]

> Среда, 12 августа 2015, 15:36 +08:00 от cantona <cantona@cantona.net>:
> 
> 
> added Alexander Shiyan < shc_work@mail.ru >.
> 
> On 12 August 2015 at 15:22, Su Kang Yin  < cantona@cantona.net > wrote:
> >Max310x driver supports up to 4 UART devices but array size of
> >"struct max310x_one" is set to 1. That leads to out of bounds
> >access on UART port registration.
> >
> >This patch fixed it by increase the array size to 4 which is
> >maximum supported UART.
> >
> >Signed-off-by: Su Kang Yin < cantona@cantona.net >
> >---
...

This seems incorrect. The number of ports is allocated dynamically by:
...
/* Alloc port structure */
s = devm_kzalloc(dev, sizeof(*s) + sizeof(struct max310x_one) * devtype->nr, GFP_KERNEL);
...

Thanks.

---

Re: [PATCH v2] serial: max310x: Fix out of bounds access

[Kang Yin Su]

On 12 August 2015 at 17:46, Alexander Shiyan <shc_work@mail.ru> wrote:
>> Среда, 12 августа 2015, 15:36 +08:00 от cantona <cantona@cantona.net>:
>>
>>
>> added Alexander Shiyan < shc_work@mail.ru >.
>>
>> On 12 August 2015 at 15:22, Su Kang Yin  < cantona@cantona.net > wrote:
>> >Max310x driver supports up to 4 UART devices but array size of
>> >"struct max310x_one" is set to 1. That leads to out of bounds
>> >access on UART port registration.
>> >
>> >This patch fixed it by increase the array size to 4 which is
>> >maximum supported UART.
>> >
>> >Signed-off-by: Su Kang Yin < cantona@cantona.net >
>> >---
> ...
>
> This seems incorrect. The number of ports is allocated dynamically by:
> ...
> /* Alloc port structure */
> s = devm_kzalloc(dev, sizeof(*s) + sizeof(struct max310x_one) * devtype->nr, GFP_KERNEL);
> ...
>
> Thanks.
>
> ---
>


Oh, my bad.  But it looks confusing. So we must ensure "struct
max310x_one p[0]"  must the last element of "struct max310x_port".

Thanks.

Re: [PATCH v2] serial: max310x: Fix out of bounds access

[Greg Kroah-Hartman]

On Wed, Aug 12, 2015 at 07:50:25PM +0800, Kang Yin Su wrote:
> On 12 August 2015 at 17:46, Alexander Shiyan <shc_work@mail.ru> wrote:
> >> Среда, 12 августа 2015, 15:36 +08:00 от cantona <cantona@cantona.net>:
> >>
> >>
> >> added Alexander Shiyan < shc_work@mail.ru >.
> >>
> >> On 12 August 2015 at 15:22, Su Kang Yin  < cantona@cantona.net > wrote:
> >> >Max310x driver supports up to 4 UART devices but array size of
> >> >"struct max310x_one" is set to 1. That leads to out of bounds
> >> >access on UART port registration.
> >> >
> >> >This patch fixed it by increase the array size to 4 which is
> >> >maximum supported UART.
> >> >
> >> >Signed-off-by: Su Kang Yin < cantona@cantona.net >
> >> >---
> > ...
> >
> > This seems incorrect. The number of ports is allocated dynamically by:
> > ...
> > /* Alloc port structure */
> > s = devm_kzalloc(dev, sizeof(*s) + sizeof(struct max310x_one) * devtype->nr, GFP_KERNEL);
> > ...
> >
> > Thanks.
> >
> > ---
> >
> 
> 
> Oh, my bad.  But it looks confusing. So we must ensure "struct
> max310x_one p[0]"  must the last element of "struct max310x_port".

Yes, that is a very common pattern we use in the kernel.