[PATCH] net/ice: fix assumption about tag placement order

[PATCH] net/ice: fix assumption about tag placement order

[Bruce Richardson]

The specific placement of outer/inner VLAN tags in E810 and related NICs
is configurable. Therefore, remove the assumption that if the L2Tag2
field is filled in, that the L2Tag1 must also be. Instead, check the
existing mbuf VLAN flags, and move tags and set flags as appropriate.
This fixes an issue where, with QinQ packets with different Tag ethtypes
(0x88a8 vs 0x8100), we get an mbuf reporting two valid tags, but only
having had one tag stripped.

Fixes: e0dcf94a0d7f ("net/ice: support VLAN ops")
Cc: stable@dpdk.org

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 drivers/net/intel/ice/ice_rxtx.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/drivers/net/intel/ice/ice_rxtx.c b/drivers/net/intel/ice/ice_rxtx.c
index da508592aa..f965aab6ee 100644
--- a/drivers/net/intel/ice/ice_rxtx.c
+++ b/drivers/net/intel/ice/ice_rxtx.c
@@ -1835,9 +1835,13 @@ ice_rxd_to_vlan_tci(struct rte_mbuf *mb, volatile union ci_rx_flex_desc *rxdp)
 #ifndef RTE_NET_INTEL_USE_16BYTE_DESC
 	if (rte_le_to_cpu_16(rxdp->wb.status_error1) &
 	    (1 << ICE_RX_FLEX_DESC_STATUS1_L2TAG2P_S)) {
-		mb->ol_flags |= RTE_MBUF_F_RX_QINQ_STRIPPED | RTE_MBUF_F_RX_QINQ |
-				RTE_MBUF_F_RX_VLAN_STRIPPED | RTE_MBUF_F_RX_VLAN;
-		mb->vlan_tci_outer = mb->vlan_tci;
+		if ((mb->ol_flags & RTE_MBUF_F_RX_VLAN_STRIPPED) == 0) {
+			mb->ol_flags |= RTE_MBUF_F_RX_VLAN | RTE_MBUF_F_RX_VLAN_STRIPPED;
+		} else {
+			/* if two tags, move Tag1 to outer tag field */
+			mb->ol_flags |= RTE_MBUF_F_RX_QINQ_STRIPPED | RTE_MBUF_F_RX_QINQ;
+			mb->vlan_tci_outer = mb->vlan_tci;
+		}
 		mb->vlan_tci = rte_le_to_cpu_16(rxdp->wb.l2tag2_2nd);
 		PMD_RX_LOG(DEBUG, "Descriptor l2tag2_1: %u, l2tag2_2: %u",
 			   rte_le_to_cpu_16(rxdp->wb.l2tag2_1st),
-- 
2.48.1

Re: [PATCH] net/ice: fix assumption about tag placement order

[David Marchand]

On Wed, Jul 16, 2025 at 7:39 PM Bruce Richardson
<bruce.richardson@intel.com> wrote:
>
> The specific placement of outer/inner VLAN tags in E810 and related NICs
> is configurable. Therefore, remove the assumption that if the L2Tag2
> field is filled in, that the L2Tag1 must also be. Instead, check the
> existing mbuf VLAN flags, and move tags and set flags as appropriate.
> This fixes an issue where, with QinQ packets with different Tag ethtypes
> (0x88a8 vs 0x8100), we get an mbuf reporting two valid tags, but only
> having had one tag stripped.
>
> Fixes: e0dcf94a0d7f ("net/ice: support VLAN ops")
> Cc: stable@dpdk.org
>
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>

The ice code seems copy/pasted from i40e.
So I guess i40e is affected as well.


-- 
David Marchand

Re: [PATCH] net/ice: fix assumption about tag placement order

[Bruce Richardson]

On Thu, Jul 17, 2025 at 10:19:24AM +0200, David Marchand wrote:
> On Wed, Jul 16, 2025 at 7:39 PM Bruce Richardson
> <bruce.richardson@intel.com> wrote:
> >
> > The specific placement of outer/inner VLAN tags in E810 and related NICs
> > is configurable. Therefore, remove the assumption that if the L2Tag2
> > field is filled in, that the L2Tag1 must also be. Instead, check the
> > existing mbuf VLAN flags, and move tags and set flags as appropriate.
> > This fixes an issue where, with QinQ packets with different Tag ethtypes
> > (0x88a8 vs 0x8100), we get an mbuf reporting two valid tags, but only
> > having had one tag stripped.
> >
> > Fixes: e0dcf94a0d7f ("net/ice: support VLAN ops")
> > Cc: stable@dpdk.org
> >
> > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> 
> The ice code seems copy/pasted from i40e.
> So I guess i40e is affected as well.
> 

I'll investigate more. These fixes I intend for 25.11 anyway, since I'd
rather not risk them late in the 25.07 release cycle.

/Bruce

Re: [PATCH] net/ice: fix assumption about tag placement order

[David Marchand]

On Thu, Jul 17, 2025 at 10:33 AM Bruce Richardson
<bruce.richardson@intel.com> wrote:
>
> On Thu, Jul 17, 2025 at 10:19:24AM +0200, David Marchand wrote:
> > On Wed, Jul 16, 2025 at 7:39 PM Bruce Richardson
> > <bruce.richardson@intel.com> wrote:
> > >
> > > The specific placement of outer/inner VLAN tags in E810 and related NICs
> > > is configurable. Therefore, remove the assumption that if the L2Tag2
> > > field is filled in, that the L2Tag1 must also be. Instead, check the
> > > existing mbuf VLAN flags, and move tags and set flags as appropriate.
> > > This fixes an issue where, with QinQ packets with different Tag ethtypes
> > > (0x88a8 vs 0x8100), we get an mbuf reporting two valid tags, but only
> > > having had one tag stripped.
> > >
> > > Fixes: e0dcf94a0d7f ("net/ice: support VLAN ops")
> > > Cc: stable@dpdk.org
> > >
> > > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> >
> > The ice code seems copy/pasted from i40e.
> > So I guess i40e is affected as well.
> >
>
> I'll investigate more. These fixes I intend for 25.11 anyway, since I'd

iavf has a similar helper.

> rather not risk them late in the 25.07 release cycle.

Yes, too late for the current release.


-- 
David Marchand