[PATCH] net/i40e: fix null dereference in raw flow item

[PATCH] net/i40e: fix null dereference in raw flow item

[Ciara Loftus]

When a RTE_FLOW_ITEM_TYPE_RAW item is used with a non-zero length
and a NULL pattern pointer in either the spec or mask, a segfault
occurs as the pattern bytes are accessed unconditionally in a loop.
Add NULL checks for both spec->pattern and mask->pattern before entering
the loop that uses those fields.

Bugzilla ID: 1155
Fixes: 6ced3dd72f5f ("net/i40e: support flexible payload parsing for FDIR")
Cc: stable@dpdk.org

Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
---
 drivers/net/intel/i40e/i40e_flow.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/drivers/net/intel/i40e/i40e_flow.c b/drivers/net/intel/i40e/i40e_flow.c
index 2374b9bbca..6ad3013573 100644
--- a/drivers/net/intel/i40e/i40e_flow.c
+++ b/drivers/net/intel/i40e/i40e_flow.c
@@ -2350,6 +2350,24 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
 				return -rte_errno;
 			}
 
+			if (raw_spec->length != 0) {
+				if (raw_spec->pattern == NULL) {
+					rte_flow_error_set(error, EINVAL,
+							   RTE_FLOW_ERROR_TYPE_ITEM,
+							   item,
+							   "NULL RAW spec pattern");
+					return -rte_errno;
+				}
+
+				if (raw_mask->pattern == NULL) {
+					rte_flow_error_set(error, EINVAL,
+							   RTE_FLOW_ERROR_TYPE_ITEM,
+							   item,
+							   "NULL RAW mask pattern");
+					return -rte_errno;
+				}
+			}
+
 			for (i = 0; i < raw_spec->length; i++) {
 				j = i + next_dst_off;
 				if (j >= RTE_ETH_FDIR_MAX_FLEXLEN ||
-- 
2.43.0

Re: [PATCH] net/i40e: fix null dereference in raw flow item

[Burakov, Anatoly]

On 3/9/2026 2:28 PM, Ciara Loftus wrote:
> When a RTE_FLOW_ITEM_TYPE_RAW item is used with a non-zero length
> and a NULL pattern pointer in either the spec or mask, a segfault
> occurs as the pattern bytes are accessed unconditionally in a loop.
> Add NULL checks for both spec->pattern and mask->pattern before entering
> the loop that uses those fields.
> 
> Bugzilla ID: 1155
> Fixes: 6ced3dd72f5f ("net/i40e: support flexible payload parsing for FDIR")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
> ---
>   drivers/net/intel/i40e/i40e_flow.c | 18 ++++++++++++++++++
>   1 file changed, 18 insertions(+)
> 
> diff --git a/drivers/net/intel/i40e/i40e_flow.c b/drivers/net/intel/i40e/i40e_flow.c
> index 2374b9bbca..6ad3013573 100644
> --- a/drivers/net/intel/i40e/i40e_flow.c
> +++ b/drivers/net/intel/i40e/i40e_flow.c
> @@ -2350,6 +2350,24 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
>   				return -rte_errno;
>   			}
>   
> +			if (raw_spec->length != 0) {
> +				if (raw_spec->pattern == NULL) {
> +					rte_flow_error_set(error, EINVAL,
> +							   RTE_FLOW_ERROR_TYPE_ITEM,
> +							   item,
> +							   "NULL RAW spec pattern");
> +					return -rte_errno;
> +				}
> +
> +				if (raw_mask->pattern == NULL) {
> +					rte_flow_error_set(error, EINVAL,
> +							   RTE_FLOW_ERROR_TYPE_ITEM,
> +							   item,
> +							   "NULL RAW mask pattern");
> +					return -rte_errno;
> +				}
> +			}
> +
>   			for (i = 0; i < raw_spec->length; i++) {
>   				j = i + next_dst_off;
>   				if (j >= RTE_ETH_FDIR_MAX_FLEXLEN ||

Maybe also check if raw_mask length matches raw_spec?

Otherwise,

Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>

-- 
Thanks,
Anatoly

[PATCH v2] net/i40e: validate raw flow items before dereferencing

[Ciara Loftus]

When a RTE_FLOW_ITEM_TYPE_RAW item is used with a non-zero length,
the spec and mask pattern pointers are dereferenced unconditionally
in a loop, causing a segfault if either is NULL. Additionally, no
check is made that the spec and mask have equal lengths before
iterating, which could result in out-of-bounds access.

Add validation before the loop: reject the item if either pattern
pointer is NULL, or if the spec and mask lengths differ.

Bugzilla ID: 1155
Fixes: 6ced3dd72f5f ("net/i40e: support flexible payload parsing for FDIR")
Cc: stable@dpdk.org

Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
---
v2:
* Add an additional check for spec and mask length mismatch
---
 drivers/net/intel/i40e/i40e_flow.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/drivers/net/intel/i40e/i40e_flow.c b/drivers/net/intel/i40e/i40e_flow.c
index 2374b9bbca..4624edc758 100644
--- a/drivers/net/intel/i40e/i40e_flow.c
+++ b/drivers/net/intel/i40e/i40e_flow.c
@@ -2350,6 +2350,30 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
 				return -rte_errno;
 			}
 
+			if (raw_spec->length != 0) {
+				if (raw_spec->pattern == NULL) {
+					rte_flow_error_set(error, EINVAL,
+							   RTE_FLOW_ERROR_TYPE_ITEM,
+							   item,
+							   "NULL RAW spec pattern");
+					return -rte_errno;
+				}
+				if (raw_mask->pattern == NULL) {
+					rte_flow_error_set(error, EINVAL,
+							   RTE_FLOW_ERROR_TYPE_ITEM,
+							   item,
+							   "NULL RAW mask pattern");
+					return -rte_errno;
+				}
+				if (raw_spec->length != raw_mask->length) {
+					rte_flow_error_set(error, EINVAL,
+							   RTE_FLOW_ERROR_TYPE_ITEM,
+							   item,
+							   "RAW spec and mask length mismatch");
+					return -rte_errno;
+				}
+			}
+
 			for (i = 0; i < raw_spec->length; i++) {
 				j = i + next_dst_off;
 				if (j >= RTE_ETH_FDIR_MAX_FLEXLEN ||
-- 
2.43.0

Re: [PATCH v2] net/i40e: validate raw flow items before dereferencing

[Bruce Richardson]

On Mon, Mar 09, 2026 at 03:29:19PM +0000, Ciara Loftus wrote:
> When a RTE_FLOW_ITEM_TYPE_RAW item is used with a non-zero length,
> the spec and mask pattern pointers are dereferenced unconditionally
> in a loop, causing a segfault if either is NULL. Additionally, no
> check is made that the spec and mask have equal lengths before
> iterating, which could result in out-of-bounds access.
> 
> Add validation before the loop: reject the item if either pattern
> pointer is NULL, or if the spec and mask lengths differ.
> 
> Bugzilla ID: 1155
> Fixes: 6ced3dd72f5f ("net/i40e: support flexible payload parsing for FDIR")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>

Missed carrying-forward Anatoly's ack from v1:
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
> ---
> v2:
> * Add an additional check for spec and mask length mismatch
> ---
>  drivers/net/intel/i40e/i40e_flow.c | 24 ++++++++++++++++++++++++
>  1 file changed, 24 insertions(+)
> 
> diff --git a/drivers/net/intel/i40e/i40e_flow.c b/drivers/net/intel/i40e/i40e_flow.c
> index 2374b9bbca..4624edc758 100644
> --- a/drivers/net/intel/i40e/i40e_flow.c
> +++ b/drivers/net/intel/i40e/i40e_flow.c
> @@ -2350,6 +2350,30 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
>  				return -rte_errno;
>  			}
>  
> +			if (raw_spec->length != 0) {
> +				if (raw_spec->pattern == NULL) {
> +					rte_flow_error_set(error, EINVAL,
> +							   RTE_FLOW_ERROR_TYPE_ITEM,
> +							   item,
> +							   "NULL RAW spec pattern");
> +					return -rte_errno;
> +				}
> +				if (raw_mask->pattern == NULL) {
> +					rte_flow_error_set(error, EINVAL,
> +							   RTE_FLOW_ERROR_TYPE_ITEM,
> +							   item,
> +							   "NULL RAW mask pattern");
> +					return -rte_errno;
> +				}
> +				if (raw_spec->length != raw_mask->length) {
> +					rte_flow_error_set(error, EINVAL,
> +							   RTE_FLOW_ERROR_TYPE_ITEM,
> +							   item,
> +							   "RAW spec and mask length mismatch");
> +					return -rte_errno;
> +				}
> +			}
> +
>  			for (i = 0; i < raw_spec->length; i++) {
>  				j = i + next_dst_off;
>  				if (j >= RTE_ETH_FDIR_MAX_FLEXLEN ||
> -- 
> 2.43.0
> 

Re: [PATCH v2] net/i40e: validate raw flow items before dereferencing

[Bruce Richardson]

On Wed, Mar 11, 2026 at 03:21:35PM +0000, Bruce Richardson wrote:
> On Mon, Mar 09, 2026 at 03:29:19PM +0000, Ciara Loftus wrote:
> > When a RTE_FLOW_ITEM_TYPE_RAW item is used with a non-zero length,
> > the spec and mask pattern pointers are dereferenced unconditionally
> > in a loop, causing a segfault if either is NULL. Additionally, no
> > check is made that the spec and mask have equal lengths before
> > iterating, which could result in out-of-bounds access.
> > 
> > Add validation before the loop: reject the item if either pattern
> > pointer is NULL, or if the spec and mask lengths differ.
> > 
> > Bugzilla ID: 1155
> > Fixes: 6ced3dd72f5f ("net/i40e: support flexible payload parsing for FDIR")
> > Cc: stable@dpdk.org
> > 
> > Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
> 
> Missed carrying-forward Anatoly's ack from v1:
> Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>

Applied to dpdk-next-net-intel.
Thanks,
/Bruce