[v1] fsl/fman: Validate cell-index value obtained from Device Tree

[PATCH] fsl/fman: Validate cell-index value obtained from Device Tree

Posted by Aleksandr Mishin 1 year, 5 months ago

Cell-index value is obtained from Device Tree and then used to calculate
the index for accessing arrays port_mfl[], mac_mfl[] and intr_mng[].
In case of broken DT due to any error cell-index can contain any value
and it is possible to go beyond the array boundaries which can lead
at least to memory corruption.
Validate cell-index value obtained from Device Tree.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes: 414fd46e7762 ("fsl/fman: Add FMan support")
Signed-off-by: Aleksandr Mishin <amishin@t-argos.ru>
---
 drivers/net/ethernet/freescale/fman/fman.c | 7 +++++++
 drivers/net/ethernet/freescale/fman/fman.h | 2 ++
 drivers/net/ethernet/freescale/fman/mac.c  | 5 +++++
 3 files changed, 14 insertions(+)

diff --git a/drivers/net/ethernet/freescale/fman/fman.c b/drivers/net/ethernet/freescale/fman/fman.c
index d96028f01770..6929bca3f768 100644
--- a/drivers/net/ethernet/freescale/fman/fman.c
+++ b/drivers/net/ethernet/freescale/fman/fman.c
@@ -2933,3 +2933,10 @@ module_exit(fman_unload);
 
 MODULE_LICENSE("Dual BSD/GPL");
 MODULE_DESCRIPTION("Freescale DPAA Frame Manager driver");
+
+int check_mac_id(u32 mac_id)
+{
+	if (mac_id >= MAX_NUM_OF_MACS)
+		return -EINVAL;
+	return 0;
+}
diff --git a/drivers/net/ethernet/freescale/fman/fman.h b/drivers/net/ethernet/freescale/fman/fman.h
index 2ea575a46675..3cedde4851e1 100644
--- a/drivers/net/ethernet/freescale/fman/fman.h
+++ b/drivers/net/ethernet/freescale/fman/fman.h
@@ -372,6 +372,8 @@ u16 fman_get_max_frm(void);
 
 int fman_get_rx_extra_headroom(void);
 
+int check_mac_id(u32 mac_id);
+
 #ifdef CONFIG_DPAA_ERRATUM_A050385
 bool fman_has_errata_a050385(void);
 #endif
diff --git a/drivers/net/ethernet/freescale/fman/mac.c b/drivers/net/ethernet/freescale/fman/mac.c
index 9767586b4eb3..7a67b4c887e2 100644
--- a/drivers/net/ethernet/freescale/fman/mac.c
+++ b/drivers/net/ethernet/freescale/fman/mac.c
@@ -247,6 +247,11 @@ static int mac_probe(struct platform_device *_of_dev)
 		dev_err(dev, "failed to read cell-index for %pOF\n", mac_node);
 		return -EINVAL;
 	}
+	err = check_mac_id(val);
+	if (err) {
+		dev_err(dev, "cell-index value is out of range for %pOF\n", mac_node);
+		return err;
+	}
 	priv->cell_index = (u8)val;
 
 	/* Get the MAC address */
-- 
2.30.2

Re: [PATCH] fsl/fman: Validate cell-index value obtained from Device Tree

Posted by Simon Horman 1 year, 5 months ago

On Tue, Jul 02, 2024 at 12:50:34PM +0300, Aleksandr Mishin wrote:
> Cell-index value is obtained from Device Tree and then used to calculate
> the index for accessing arrays port_mfl[], mac_mfl[] and intr_mng[].
> In case of broken DT due to any error cell-index can contain any value
> and it is possible to go beyond the array boundaries which can lead
> at least to memory corruption.
> Validate cell-index value obtained from Device Tree.
> 
> Found by Linux Verification Center (linuxtesting.org) with SVACE.
> 
> Fixes: 414fd46e7762 ("fsl/fman: Add FMan support")
> Signed-off-by: Aleksandr Mishin <amishin@t-argos.ru>
> ---
>  drivers/net/ethernet/freescale/fman/fman.c | 7 +++++++
>  drivers/net/ethernet/freescale/fman/fman.h | 2 ++
>  drivers/net/ethernet/freescale/fman/mac.c  | 5 +++++
>  3 files changed, 14 insertions(+)
> 
> diff --git a/drivers/net/ethernet/freescale/fman/fman.c b/drivers/net/ethernet/freescale/fman/fman.c
> index d96028f01770..6929bca3f768 100644
> --- a/drivers/net/ethernet/freescale/fman/fman.c
> +++ b/drivers/net/ethernet/freescale/fman/fman.c
> @@ -2933,3 +2933,10 @@ module_exit(fman_unload);
>  
>  MODULE_LICENSE("Dual BSD/GPL");
>  MODULE_DESCRIPTION("Freescale DPAA Frame Manager driver");
> +
> +int check_mac_id(u32 mac_id)
> +{
> +	if (mac_id >= MAX_NUM_OF_MACS)
> +		return -EINVAL;
> +	return 0;
> +}
> diff --git a/drivers/net/ethernet/freescale/fman/fman.h b/drivers/net/ethernet/freescale/fman/fman.h
> index 2ea575a46675..3cedde4851e1 100644
> --- a/drivers/net/ethernet/freescale/fman/fman.h
> +++ b/drivers/net/ethernet/freescale/fman/fman.h
> @@ -372,6 +372,8 @@ u16 fman_get_max_frm(void);
>  
>  int fman_get_rx_extra_headroom(void);
>  
> +int check_mac_id(u32 mac_id);
> +
>  #ifdef CONFIG_DPAA_ERRATUM_A050385
>  bool fman_has_errata_a050385(void);
>  #endif
> diff --git a/drivers/net/ethernet/freescale/fman/mac.c b/drivers/net/ethernet/freescale/fman/mac.c
> index 9767586b4eb3..7a67b4c887e2 100644
> --- a/drivers/net/ethernet/freescale/fman/mac.c
> +++ b/drivers/net/ethernet/freescale/fman/mac.c
> @@ -247,6 +247,11 @@ static int mac_probe(struct platform_device *_of_dev)
>  		dev_err(dev, "failed to read cell-index for %pOF\n", mac_node);
>  		return -EINVAL;
>  	}
> +	err = check_mac_id(val);

Hi Aleksandr, all,

It seems that this breaks linking with allmodconfig builds on x86_64,
perhaps it is better to simply make check_mac_id a static function in mac.c ?

> +	if (err) {
> +		dev_err(dev, "cell-index value is out of range for %pOF\n", mac_node);

Although other instances exist in this function before this patch,
it does look like this leaks a reference to of_dev taken by the
call to of_find_device_by_node() on line 194.

Maybe it is intentional, I'm unsure.
Perhaps this can be investigated separately to the fix proposed by this
patch?

Flagged by Coccinelle (this one is on line 253):

 .../mac.c:238:2-8: ERROR: missing put_device; call of_find_device_by_node on line 194, but without a corresponding object release within this function.
 .../mac.c:242:2-8: ERROR: missing put_device; call of_find_device_by_node on line 194, but without a corresponding object release within this function.
 .../mac.c:248:2-8: ERROR: missing put_device; call of_find_device_by_node on line 194, but without a corresponding object release within this function.
 .../mac.c:253:2-8: ERROR: missing put_device; call of_find_device_by_node on line 194, but without a corresponding object release within this function.
 .../mac.c:267:2-8: ERROR: missing put_device; call of_find_device_by_node on line 194, but without a corresponding object release within this function.
 .../mac.c:273:2-8: ERROR: missing put_device; call of_find_device_by_node on line 194, but without a corresponding object release within this function.
 .../mac.c:282:3-9: ERROR: missing put_device; call of_find_device_by_node on line 194, but without a corresponding object release within this function.
 .../mac.c:320:2-8: ERROR: missing put_device; call of_find_device_by_node on line 194, but without a corresponding object release within this function.
 .../mac.c:333:1-7: ERROR: missing put_device; call of_find_device_by_node on line 194, but without a corresponding object release within this function.
 .../mac.c:337:1-7: ERROR: missing put_device; call of_find_device_by_node on line 194, but without a corresponding object release within this function.
 .../mac.c:282:3-9: ERROR: missing put_device; call of_find_device_by_node on line 285, but without a corresponding object release within this function.
 .../mac.c:320:2-8: ERROR: missing put_device; call of_find_device_by_node on line 285, but without a corresponding object release within this function.
 .../mac.c:333:1-7: ERROR: missing put_device; call of_find_device_by_node on line 285, but without a corresponding object release within this function.
 .../mac.c:337:1-7: ERROR: missing put_device; call of_find_device_by_node on line 285, but without a corresponding object release within this function.

> +		return err;
> +	}
>  	priv->cell_index = (u8)val;
>  
>  	/* Get the MAC address */

-- 
pw-bot: changes-requested

Re: [PATCH] fsl/fman: Validate cell-index value obtained from Device Tree

Posted by Jakub Kicinski 1 year, 5 months ago

On Tue, 2 Jul 2024 14:36:51 +0100 Simon Horman wrote:
> Maybe it is intentional, I'm unsure.
> Perhaps this can be investigated separately to the fix proposed by this
> patch?

+1, please fix this first

[PATCH v2] fsl/fman: Validate cell-index value obtained from Device Tree

Posted by Aleksandr Mishin 1 year, 5 months ago

Cell-index value is obtained from Device Tree and then used to calculate
the index for accessing arrays port_mfl[], mac_mfl[] and intr_mng[].
In case of broken DT due to any error cell-index can contain any value
and it is possible to go beyond the array boundaries which can lead
at least to memory corruption.
Validate cell-index value obtained from Device Tree.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes: 414fd46e7762 ("fsl/fman: Add FMan support")
Signed-off-by: Aleksandr Mishin <amishin@t-argos.ru>
---
v1->v2: Move check to mac.c to avoid allmodconfig build errors and reference leaks

 drivers/net/ethernet/freescale/fman/fman.c | 1 -
 drivers/net/ethernet/freescale/fman/fman.h | 3 +++
 drivers/net/ethernet/freescale/fman/mac.c  | 4 ++++
 3 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/freescale/fman/fman.c b/drivers/net/ethernet/freescale/fman/fman.c
index d96028f01770..fb416d60dcd7 100644
--- a/drivers/net/ethernet/freescale/fman/fman.c
+++ b/drivers/net/ethernet/freescale/fman/fman.c
@@ -24,7 +24,6 @@
 
 /* General defines */
 #define FMAN_LIODN_TBL			64	/* size of LIODN table */
-#define MAX_NUM_OF_MACS			10
 #define FM_NUM_OF_FMAN_CTRL_EVENT_REGS	4
 #define BASE_RX_PORTID			0x08
 #define BASE_TX_PORTID			0x28
diff --git a/drivers/net/ethernet/freescale/fman/fman.h b/drivers/net/ethernet/freescale/fman/fman.h
index 2ea575a46675..74eb62eba0d7 100644
--- a/drivers/net/ethernet/freescale/fman/fman.h
+++ b/drivers/net/ethernet/freescale/fman/fman.h
@@ -74,6 +74,9 @@
 #define BM_MAX_NUM_OF_POOLS		64 /* Buffers pools */
 #define FMAN_PORT_MAX_EXT_POOLS_NUM	8  /* External BM pools per Rx port */
 
+/* General defines */
+#define MAX_NUM_OF_MACS			10
+
 struct fman; /* FMan data */
 
 /* Enum for defining port types */
diff --git a/drivers/net/ethernet/freescale/fman/mac.c b/drivers/net/ethernet/freescale/fman/mac.c
index 9767586b4eb3..ac9ad5e67b44 100644
--- a/drivers/net/ethernet/freescale/fman/mac.c
+++ b/drivers/net/ethernet/freescale/fman/mac.c
@@ -247,6 +247,10 @@ static int mac_probe(struct platform_device *_of_dev)
 		dev_err(dev, "failed to read cell-index for %pOF\n", mac_node);
 		return -EINVAL;
 	}
+	if (val >= MAX_NUM_OF_MACS) {
+		dev_err(dev, "cell-index value is too big for %pOF\n", mac_node);
+		return -EINVAL;
+	}
 	priv->cell_index = (u8)val;
 
 	/* Get the MAC address */
-- 
2.30.2

Re: [PATCH v2] fsl/fman: Validate cell-index value obtained from Device Tree

Posted by Sean Anderson 1 year, 5 months ago

On 7/2/24 10:01, Aleksandr Mishin wrote:
> [You don't often get email from amishin@t-argos.ru. Learn why this is important at https://cas5-0-urlprotect.trendmicro.com:443/wis/clicktime/v1/query?url=https%3a%2f%2faka.ms%2fLearnAboutSenderIdentification&umid=ca782a5b-a3fd-4738-b50c-ffbd8e52e50f&auth=d807158c60b7d2502abde8a2fc01f40662980862-ac0deac714bbe85d389fd2711a17d0113034b3ca ]
>
> Cell-index value is obtained from Device Tree and then used to calculate
> the index for accessing arrays port_mfl[], mac_mfl[] and intr_mng[].
> In case of broken DT due to any error cell-index can contain any value
> and it is possible to go beyond the array boundaries which can lead
> at least to memory corruption.
> Validate cell-index value obtained from Device Tree.
>
> Found by Linux Verification Center (linuxtesting.org) with SVACE.
>
> Fixes: 414fd46e7762 ("fsl/fman: Add FMan support")
> Signed-off-by: Aleksandr Mishin <amishin@t-argos.ru>
> ---
> v1->v2: Move check to mac.c to avoid allmodconfig build errors and reference leaks
>
>  drivers/net/ethernet/freescale/fman/fman.c | 1 -
>  drivers/net/ethernet/freescale/fman/fman.h | 3 +++
>  drivers/net/ethernet/freescale/fman/mac.c  | 4 ++++
>  3 files changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/freescale/fman/fman.c b/drivers/net/ethernet/freescale/fman/fman.c
> index d96028f01770..fb416d60dcd7 100644
> --- a/drivers/net/ethernet/freescale/fman/fman.c
> +++ b/drivers/net/ethernet/freescale/fman/fman.c
> @@ -24,7 +24,6 @@
>
>  /* General defines */
>  #define FMAN_LIODN_TBL                 64      /* size of LIODN table */
> -#define MAX_NUM_OF_MACS                        10
>  #define FM_NUM_OF_FMAN_CTRL_EVENT_REGS 4
>  #define BASE_RX_PORTID                 0x08
>  #define BASE_TX_PORTID                 0x28
> diff --git a/drivers/net/ethernet/freescale/fman/fman.h b/drivers/net/ethernet/freescale/fman/fman.h
> index 2ea575a46675..74eb62eba0d7 100644
> --- a/drivers/net/ethernet/freescale/fman/fman.h
> +++ b/drivers/net/ethernet/freescale/fman/fman.h
> @@ -74,6 +74,9 @@
>  #define BM_MAX_NUM_OF_POOLS            64 /* Buffers pools */
>  #define FMAN_PORT_MAX_EXT_POOLS_NUM    8  /* External BM pools per Rx port */
>
> +/* General defines */
> +#define MAX_NUM_OF_MACS                        10
> +
>  struct fman; /* FMan data */
>
>  /* Enum for defining port types */
> diff --git a/drivers/net/ethernet/freescale/fman/mac.c b/drivers/net/ethernet/freescale/fman/mac.c
> index 9767586b4eb3..ac9ad5e67b44 100644
> --- a/drivers/net/ethernet/freescale/fman/mac.c
> +++ b/drivers/net/ethernet/freescale/fman/mac.c
> @@ -247,6 +247,10 @@ static int mac_probe(struct platform_device *_of_dev)
>                 dev_err(dev, "failed to read cell-index for %pOF\n", mac_node);
>                 return -EINVAL;
>         }
> +       if (val >= MAX_NUM_OF_MACS) {
> +               dev_err(dev, "cell-index value is too big for %pOF\n", mac_node);
> +               return -EINVAL;
> +       }
>         priv->cell_index = (u8)val;
>
>         /* Get the MAC address */
> --
> 2.30.2
>

Reviewed-by: Sean Anderson <sean.anderson@seco.com>

[StudioX, SECO SpA]<https://www.seco.com/applications/studiox>