[PATCH] net: fec: account for VLAN header in MAX_FL calculation

[PATCH] net: fec: account for VLAN header in MAX_FL calculation

[Clemens Gruber]

Commit 62b5bb7be7bc ("net: fec: update MAX_FL based on the current MTU")
changed the MAX_FL (maximum frame length) calculation without accounting
for VLAN-tagged frames, leading to RX errors / dropped frames.

Add VLAN_HLEN to the MAX_FL calculation.

Fixes: 62b5bb7be7bc ("net: fec: update MAX_FL based on the current MTU")

Signed-off-by: Clemens Gruber <mail@clemensgruber.at>
Cc: stable@kernel.org
---
 drivers/net/ethernet/freescale/fec_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index e2b75d1970ae..f8f88c592323 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -1151,7 +1151,7 @@ fec_restart(struct net_device *ndev)
 	u32 rcntl = FEC_RCR_MII;
 
 	if (OPT_ARCH_HAS_MAX_FL)
-		rcntl |= (fep->netdev->mtu + ETH_HLEN + ETH_FCS_LEN) << 16;
+		rcntl |= (fep->netdev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN) << 16;
 
 	if (fep->bufdesc_ex)
 		fec_ptp_save_state(fep);
-- 
2.52.0

RE: [PATCH] net: fec: account for VLAN header in MAX_FL calculation

[Wei Fang]

> Commit 62b5bb7be7bc ("net: fec: update MAX_FL based on the current MTU")
> changed the MAX_FL (maximum frame length) calculation without accounting
> for VLAN-tagged frames, leading to RX errors / dropped frames.
> 
> Add VLAN_HLEN to the MAX_FL calculation.
> 
> Fixes: 62b5bb7be7bc ("net: fec: update MAX_FL based on the current MTU")
> 
> Signed-off-by: Clemens Gruber <mail@clemensgruber.at>
> Cc: stable@kernel.org
> ---
>  drivers/net/ethernet/freescale/fec_main.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/freescale/fec_main.c
> b/drivers/net/ethernet/freescale/fec_main.c
> index e2b75d1970ae..f8f88c592323 100644
> --- a/drivers/net/ethernet/freescale/fec_main.c
> +++ b/drivers/net/ethernet/freescale/fec_main.c
> @@ -1151,7 +1151,7 @@ fec_restart(struct net_device *ndev)
>         u32 rcntl = FEC_RCR_MII;
> 
>         if (OPT_ARCH_HAS_MAX_FL)
> -               rcntl |= (fep->netdev->mtu + ETH_HLEN + ETH_FCS_LEN) <<
> 16;
> +               rcntl |= (fep->netdev->mtu + ETH_HLEN + ETH_FCS_LEN +
> VLAN_HLEN) << 16;

The line is more than 80 characters after the change, so you can use
VLAN_ETH_HLEN instead of (ETH_HLEN + VLAN_HLEN).

Additionally, I believe the code in other places also needs to be modified
accordingly. See below.

fec_restart():
if ((fep->quirks & FEC_QUIRK_JUMBO_FRAME) &&
  (ndev->mtu > (PKT_MAXBUF_SIZE - ETH_HLEN - ETH_FCS_LEN)))
		writel(0xF, fep->hwp + FEC_X_WMRK);

fec_change_mtu():
order = get_order(new_mtu + ETH_HLEN + ETH_FCS_LEN
			  + FEC_DRV_RESERVE_SPACE);

fec_probe():
ndev->max_mtu = fep->max_buf_size - ETH_HLEN - ETH_FCS_LEN;

[PATCH v2] net: fec: account for VLAN header in frame length calculations

[Clemens Gruber]

The MAX_FL (maximum frame length) and related calculations used ETH_HLEN,
which does not account for the 4-byte VLAN tag in tagged frames. This
caused the hardware to reject valid VLAN frames as oversized, resulting
in RX errors and dropped packets.

Use VLAN_ETH_HLEN instead of ETH_HLEN in the MAX_FL register setup,
cut-through mode threshold, buffer allocation, and max_mtu calculation.

Cc: stable@kernel.org # v6.18+
Fixes: 62b5bb7be7bc ("net: fec: update MAX_FL based on the current MTU")
Fixes: d466c16026e9 ("net: fec: enable the Jumbo frame support for i.MX8QM")
Fixes: 59e9bf037d75 ("net: fec: add change_mtu to support dynamic buffer allocation")
Fixes: ec2a1681ed4f ("net: fec: use a member variable for maximum buffer size")
Signed-off-by: Clemens Gruber <mail@clemensgruber.at>
---
Changes in v2:
- Fix the other frame length calculations too
- Improved title and commit message

One line in fec_restart (after the FEC_QUIRK_JUMBO_FRAME check) has now
82 characters but I think folding it would make it less readable.
Let me know if this should be changed.
---
 drivers/net/ethernet/freescale/fec_main.c | 13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index a753265961af..797ef6899657 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -1150,7 +1150,7 @@ fec_restart(struct net_device *ndev)
 	u32 rcntl = FEC_RCR_MII;
 
 	if (OPT_ARCH_HAS_MAX_FL)
-		rcntl |= (fep->netdev->mtu + ETH_HLEN + ETH_FCS_LEN) << 16;
+		rcntl |= (fep->netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN) << 16;
 
 	if (fep->bufdesc_ex)
 		fec_ptp_save_state(fep);
@@ -1285,12 +1285,13 @@ fec_restart(struct net_device *ndev)
 
 		/* When Jumbo Frame is enabled, the FIFO may not be large enough
 		 * to hold an entire frame. In such cases, if the MTU exceeds
-		 * (PKT_MAXBUF_SIZE - ETH_HLEN - ETH_FCS_LEN), configure the interface
-		 * to operate in cut-through mode, triggered by the FIFO threshold.
+		 * (PKT_MAXBUF_SIZE - VLAN_ETH_HLEN - ETH_FCS_LEN), configure
+		 * the interface to operate in cut-through mode, triggered by
+		 * the FIFO threshold.
 		 * Otherwise, enable the ENET store-and-forward mode.
 		 */
 		if ((fep->quirks & FEC_QUIRK_JUMBO_FRAME) &&
-		    (ndev->mtu > (PKT_MAXBUF_SIZE - ETH_HLEN - ETH_FCS_LEN)))
+		    (ndev->mtu > (PKT_MAXBUF_SIZE - VLAN_ETH_HLEN - ETH_FCS_LEN)))
 			writel(0xF, fep->hwp + FEC_X_WMRK);
 		else
 			writel(FEC_TXWMRK_STRFWD, fep->hwp + FEC_X_WMRK);
@@ -4037,7 +4038,7 @@ static int fec_change_mtu(struct net_device *ndev, int new_mtu)
 	if (netif_running(ndev))
 		return -EBUSY;
 
-	order = get_order(new_mtu + ETH_HLEN + ETH_FCS_LEN
+	order = get_order(new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN
 			  + FEC_DRV_RESERVE_SPACE);
 	fep->rx_frame_size = (PAGE_SIZE << order) - FEC_DRV_RESERVE_SPACE;
 	fep->pagepool_order = order;
@@ -4588,7 +4589,7 @@ fec_probe(struct platform_device *pdev)
 	else
 		fep->max_buf_size = PKT_MAXBUF_SIZE;
 
-	ndev->max_mtu = fep->max_buf_size - ETH_HLEN - ETH_FCS_LEN;
+	ndev->max_mtu = fep->max_buf_size - VLAN_ETH_HLEN - ETH_FCS_LEN;
 
 	ret = register_netdev(ndev);
 	if (ret)
-- 
2.52.0

RE: [PATCH v2] net: fec: account for VLAN header in frame length calculations

[Wei Fang]

> The MAX_FL (maximum frame length) and related calculations used
> ETH_HLEN,
> which does not account for the 4-byte VLAN tag in tagged frames. This
> caused the hardware to reject valid VLAN frames as oversized, resulting
> in RX errors and dropped packets.
> 
> Use VLAN_ETH_HLEN instead of ETH_HLEN in the MAX_FL register setup,
> cut-through mode threshold, buffer allocation, and max_mtu calculation.
> 
> Cc: stable@kernel.org # v6.18+
> Fixes: 62b5bb7be7bc ("net: fec: update MAX_FL based on the current MTU")
> Fixes: d466c16026e9 ("net: fec: enable the Jumbo frame support for
> i.MX8QM")
> Fixes: 59e9bf037d75 ("net: fec: add change_mtu to support dynamic buffer
> allocation")
> Fixes: ec2a1681ed4f ("net: fec: use a member variable for maximum buffer
> size")
> Signed-off-by: Clemens Gruber <mail@clemensgruber.at>
> ---
> Changes in v2:
> - Fix the other frame length calculations too
> - Improved title and commit message
> 
> One line in fec_restart (after the FEC_QUIRK_JUMBO_FRAME check) has now
> 82 characters but I think folding it would make it less readable.
> Let me know if this should be changed.
> ---

Thanks, this patch looks good to me.

Reviewed-by: Wei Fang <wei.fang@nxp.com>

Re: [PATCH v2] net: fec: account for VLAN header in frame length calculations

[Paolo Abeni]

On 1/21/26 9:37 AM, Clemens Gruber wrote:
> The MAX_FL (maximum frame length) and related calculations used ETH_HLEN,
> which does not account for the 4-byte VLAN tag in tagged frames. This
> caused the hardware to reject valid VLAN frames as oversized, resulting
> in RX errors and dropped packets.
> 
> Use VLAN_ETH_HLEN instead of ETH_HLEN in the MAX_FL register setup,
> cut-through mode threshold, buffer allocation, and max_mtu calculation.
> 
> Cc: stable@kernel.org # v6.18+
> Fixes: 62b5bb7be7bc ("net: fec: update MAX_FL based on the current MTU")
> Fixes: d466c16026e9 ("net: fec: enable the Jumbo frame support for i.MX8QM")
> Fixes: 59e9bf037d75 ("net: fec: add change_mtu to support dynamic buffer allocation")
> Fixes: ec2a1681ed4f ("net: fec: use a member variable for maximum buffer size")
> Signed-off-by: Clemens Gruber <mail@clemensgruber.at>

A couple of things to remember for future submissions: you should
include the target tree into the subj prefix ('net' in this case) and
you should avoid sending you revision in reply to older ones as it may
foul patchwork.

No repost needed, the patch looks good, I'm applying it.

Thanks,

Paolo

Re: [PATCH v2] net: fec: account for VLAN header in frame length calculations

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net.git (main)
by Paolo Abeni <pabeni@redhat.com>:

On Wed, 21 Jan 2026 09:37:51 +0100 you wrote:
> The MAX_FL (maximum frame length) and related calculations used ETH_HLEN,
> which does not account for the 4-byte VLAN tag in tagged frames. This
> caused the hardware to reject valid VLAN frames as oversized, resulting
> in RX errors and dropped packets.
> 
> Use VLAN_ETH_HLEN instead of ETH_HLEN in the MAX_FL register setup,
> cut-through mode threshold, buffer allocation, and max_mtu calculation.
> 
> [...]

Here is the summary with links:
  - [v2] net: fec: account for VLAN header in frame length calculations
    https://git.kernel.org/netdev/net/c/ca1bb3fedf26

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html