[PATCH net v2 0/3] net: stmmac: deliver frames with Rx checksum errors to the stack

[PATCH net v2 0/3] net: stmmac: deliver frames with Rx checksum errors to the stack

[Christian Schmidbauer]

On dwmac1000 the MAC drops every frame that fails Rx checksum validation
before the driver sees it. The enhanced descriptor parser would discard
such a frame even if the MAC delivered it. As a result, a VXLAN tunnel
over IPv6 with zero UDP checksums receives nothing while Rx checksum
offload is on, and neither rx_dropped nor rx_errors shows the loss.
Patches 2 and 3 deliver these frames with CHECKSUM_NONE instead, as
commit fe4042797651 ("net: stmmac: dwmac4: stop hardware from dropping
checksum-error packets") did for dwmac4.

Patch 1 makes the AF_XDP zero-copy Rx path honour the checksum error
status. That path already receives such frames on dwmac4, and patches
2 and 3 make them reach it on dwmac1000 as well. Patch 3 depends on
patches 1 and 2.

Tested on an Aruba AP-325 (ipq806x-gmac-dwmac) on the net tree, with
and without the series.

Not tested:

- Patch 1 was only compile-tested, as an AF_XDP zero-copy setup was not
  available.
- dwmac-loongson, which also uses dwmac1000_dma_ops.
- Cores with normal descriptors, such as Rockchip RK3288, RK3328 and
  RK3399.

Changes in v2:

- Rebase on net, where commit 834ec9530832 ("net: stmmac: drop split
  frames on AF_XDP zero-copy RX") fixed the multi-buffer issue that
  Sashiko raised in the v1 review.
- Drop RDES0_LENGTH_ERROR from the Error Summary mask in patch 2, as
  the bit does not contribute to Error Summary and the length check
  further down discards such a frame anyway.
- Explain in the commit message of patch 2 why RDES0 bit 7 is not
  treated as an L2 error.
- Set DMA_CONTROL_DT in patch 3 only with enhanced descriptors, as
  the normal descriptor parser does not report a checksum-only error.

v1: https://lore.kernel.org/netdev/20261020190319.884375-1-opensource@grische.xyz/

Christian Schmidbauer (3):
  net: stmmac: honour Rx checksum errors in the AF_XDP zero-copy path
  net: stmmac: dwmac1000: don't discard frames with only a checksum
    error
  net: stmmac: dwmac1000: stop hardware from dropping checksum-error
    packets

 drivers/net/ethernet/stmicro/stmmac/dwmac1000.h     | 1 +
 drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c | 8 ++++++++
 drivers/net/ethernet/stmicro/stmmac/enh_desc.c      | 8 ++++----
 drivers/net/ethernet/stmicro/stmmac/stmmac_main.c   | 8 +++++---
 4 files changed, 18 insertions(+), 7 deletions(-)


base-commit: 241dec6f9ab74d8c3ef246764f3611666b96b9e1
-- 
2.54.0

[PATCH net v2 1/3] net: stmmac: honour Rx checksum errors in the AF_XDP zero-copy path

[Christian Schmidbauer]

Commit ee0aace5f844 ("net: stmmac: Correctly handle Rx checksum offload
errors") taught stmmac_rx() to honour the csum_none status, but
stmmac_dispatch_skb_zc() still ignores it and sets CHECKSUM_UNNECESSARY
for every IP frame whenever the checksum offload engine (CoE) is
enabled.

Since commit fe4042797651 ("net: stmmac: dwmac4: stop hardware from
dropping checksum-error packets") dwmac4 delivers such frames instead of
dropping them, so this path receives them.

Mark the skb CHECKSUM_NONE when csum_none is set, as stmmac_rx() does.

Fixes: bba2556efad6 ("net: stmmac: Enable RX via AF_XDP zero-copy")
Cc: <stable@vger.kernel.org>
Assisted-by: Claude:claude-opus-5-5
Signed-off-by: Christian Schmidbauer <opensource@grische.xyz>
---
 drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 7ffed295eb14..3e5803db19b4 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -5547,7 +5547,7 @@ static struct sk_buff *stmmac_construct_skb_zc(struct stmmac_channel *ch,
 
 static void stmmac_dispatch_skb_zc(struct stmmac_priv *priv, u32 queue,
 				   struct dma_desc *p, struct dma_desc *np,
-				   struct xdp_buff *xdp)
+				   struct xdp_buff *xdp, int status)
 {
 	struct stmmac_rxq_stats *rxq_stats = &priv->xstats.rxq_stats[queue];
 	struct stmmac_channel *ch = &priv->channel[queue];
@@ -5575,7 +5575,8 @@ static void stmmac_dispatch_skb_zc(struct stmmac_priv *priv, u32 queue,
 
 	skb->protocol = eth_type_trans(skb, priv->dev);
 
-	if (unlikely(!coe) || !stmmac_has_ip_ethertype(skb))
+	if (unlikely(!coe) || !stmmac_has_ip_ethertype(skb) ||
+	    (status & csum_none))
 		skb_checksum_none_assert(skb);
 	else
 		skb->ip_summed = CHECKSUM_UNNECESSARY;
@@ -5789,7 +5790,8 @@ static int stmmac_rx_zc(struct stmmac_priv *priv, int limit, u32 queue)
 
 		switch (res) {
 		case STMMAC_XDP_PASS:
-			stmmac_dispatch_skb_zc(priv, queue, p, np, buf->xdp);
+			stmmac_dispatch_skb_zc(priv, queue, p, np, buf->xdp,
+					       status);
 			xsk_buff_free(buf->xdp);
 			break;
 		case STMMAC_XDP_CONSUMED:
-- 
2.54.0

[PATCH net v2 2/3] net: stmmac: dwmac1000: don't discard frames with only a checksum error

[Christian Schmidbauer]

On dwmac1000 with enhanced descriptors, the hardware sets the Error
Summary (ES) bit of the Rx descriptor not only for L2 errors but also
for an L3/L4 checksum error. enh_desc_get_rx_status() returns
discard_frame whenever ES is set, and since commit 1b746ce8b397 ("net:
stmmac: don't overwrite discard_frame status") the checksum status is
only evaluated for good frames.

Return discard_frame only when ES comes with one of the L2 error bits.
A frame with only a checksum error is then reported as csum_none and
reaches the stack as CHECKSUM_NONE, so the stack verifies the checksum
itself. This matches dwmac4, whose Error Summary bit does not cover
checksum errors.

RDES0 bit 7 is not in the mask, because depending on the core and on
Rx checksum offload it reports a timestamp, an IP header checksum error
or a giant frame, and enh_desc_get_rx_status() cannot tell which. In
threshold Rx mode, a giant frame that fits one buffer therefore reaches
the stack with a verified FCS.

Also remove the pr_err that enh_desc_get_rx_status() prints when bit 7
is set. Once the MAC stops dropping checksum-error frames, any host on
the link can flood the kernel log by sending frames with a bad IPv4
header checksum.

This prepares for setting DMA_CONTROL_DT, which makes the MAC deliver
checksum-error frames instead of dropping them in its Rx FIFO.

Fixes: 1b746ce8b397 ("net: stmmac: don't overwrite discard_frame status")
Cc: <stable@vger.kernel.org> # 6.6.x: ee0aace5f844: net: stmmac: Correctly handle Rx checksum offload errors
Cc: <stable@vger.kernel.org> # 6.6.x
Assisted-by: Claude:claude-opus-4-8
Signed-off-by: Christian Schmidbauer <opensource@grische.xyz>
---
v2:
- Drop RDES0_LENGTH_ERROR from the mask of Error Summary bits. The
  length check further down already discards such a frame.
- Explain in the commit message why RDES0 bit 7 is not treated as an
  L2 error.
v1: https://lore.kernel.org/netdev/20261020190319.884375-1-opensource@grische.xyz/
 drivers/net/ethernet/stmicro/stmmac/enh_desc.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
index 7ece8d4157a8..d0622f2605aa 100644
--- a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
@@ -216,9 +216,6 @@ static int enh_desc_get_rx_status(struct stmmac_extra_stats *x,
 		if (unlikely(rdes0 & RDES0_OVERFLOW_ERROR))
 			x->rx_gmac_overflow++;
 
-		if (unlikely(rdes0 & RDES0_IPC_CSUM_ERROR))
-			pr_err("\tIPC Csum Error/Giant frame\n");
-
 		if (unlikely(rdes0 & RDES0_COLLISION))
 			x->rx_collision++;
 		if (unlikely(rdes0 & RDES0_RECEIVE_WATCHDOG))
@@ -230,7 +227,10 @@ static int enh_desc_get_rx_status(struct stmmac_extra_stats *x,
 		if (unlikely(rdes0 & RDES0_CRC_ERROR)) {
 			x->rx_crc_errors++;
 		}
-		ret = discard_frame;
+		if (rdes0 & (RDES0_DESCRIPTOR_ERROR | RDES0_OVERFLOW_ERROR |
+			     RDES0_COLLISION | RDES0_RECEIVE_WATCHDOG |
+			     RDES0_MII_ERROR | RDES0_CRC_ERROR))
+			ret = discard_frame;
 	}
 
 	/* After a payload csum error, the ES bit is set.
-- 
2.54.0

[PATCH net v2 3/3] net: stmmac: dwmac1000: stop hardware from dropping checksum-error packets

[Christian Schmidbauer]

When the checksum offload engine (CoE) is enabled, the dwmac1000 MAC
silently drops any frame for which the engine detects a checksum error,
before the driver sees it.

This breaks VXLAN over IPv6 with zero UDP checksums, which RFC 6935 and
RFC 6936 allow for tunnels.

Set DMA_CONTROL_DT so that such frames are delivered instead of dropped,
as commit fe4042797651 ("net: stmmac: dwmac4: stop hardware from
dropping checksum-error packets") did for dwmac4. CoE remains enabled,
so the driver still sees the error status and marks the skb
CHECKSUM_NONE for the stack to verify.

Set the bit only with enhanced descriptors. The normal descriptor parser
does not report a checksum-only error to the stack, so cores that use
it keep dropping such frames in hardware anyway.

Tested on an Aruba AP-325 (ipq806x-gmac-dwmac) with Rx checksum offload
on.

Link: https://lore.kernel.org/all/20250625132117.1b3264e8@kernel.org/
Link: https://github.com/freifunk-gluon/gluon/issues/3783
Fixes: 47dd7a540b8a ("net: add support for STMicroelectronics Ethernet controllers.")
Cc: <stable@vger.kernel.org> # 6.6.x: ee0aace5f844: net: stmmac: Correctly handle Rx checksum offload errors
Cc: <stable@vger.kernel.org> # 6.6.x
Assisted-by: Claude:claude-opus-4-8
Signed-off-by: Christian Schmidbauer <opensource@grische.xyz>
---
Depends on patches 1 and 2 of this series.

v2:
- Set DMA_CONTROL_DT only with enhanced descriptors. The normal
  descriptor parser does not report a checksum-only error, so cores
  with normal descriptors and Type 2 CoE keep the hardware drop.
v1: https://lore.kernel.org/netdev/20261020190319.884375-1-opensource@grische.xyz/
 drivers/net/ethernet/stmicro/stmmac/dwmac1000.h     | 1 +
 drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c | 8 ++++++++
 2 files changed, 9 insertions(+)

diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h b/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h
index 1de1f929d61c..004e6e30a68b 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h
@@ -163,6 +163,7 @@ enum rx_tx_priority_ratio {
 
 /* DMA operation mode defines (start/stop tx/rx are placed in common header)*/
 /* Disable Drop TCP/IP csum error */
+#define DMA_CONTROL_DT		0x04000000
 #define DMA_CONTROL_RSF		0x02000000	/* Receive Store and Forward */
 #define DMA_CONTROL_DFF		0x01000000	/* Disaable flushing */
 /* Threshold for Activating the FC */
diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c b/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c
index 3ac7a7949529..31d5f79ef560 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c
@@ -15,6 +15,7 @@
 #include <linux/io.h>
 #include "dwmac1000.h"
 #include "dwmac_dma.h"
+#include "stmmac.h"
 
 static void dwmac1000_dma_axi(void __iomem *ioaddr, struct stmmac_axi *axi)
 {
@@ -129,6 +130,13 @@ static void dwmac1000_dma_operation_mode_rx(struct stmmac_priv *priv,
 {
 	u32 csr6 = readl(ioaddr + DMA_CHAN_CONTROL(channel));
 
+	/* Deliver checksum-error frames only with enhanced
+	 * descriptors, because the normal descriptor parser does not
+	 * report a checksum-only error.
+	 */
+	if (priv->plat->enh_desc)
+		csr6 |= DMA_CONTROL_DT;
+
 	if (mode == SF_DMA_MODE) {
 		pr_debug("GMAC: enable RX store and forward mode\n");
 		csr6 |= DMA_CONTROL_RSF | DMA_CONTROL_DFF;
-- 
2.54.0

Re: [PATCH net v2 3/3] net: stmmac: dwmac1000: stop hardware from dropping checksum-error packets

[Jakub Kicinski]

On Wed,  4 Nov 2026 16:19:13 +0100 Christian Schmidbauer wrote:
>  /* DMA operation mode defines (start/stop tx/rx are placed in common header)*/
>  /* Disable Drop TCP/IP csum error */
> +#define DMA_CONTROL_DT		0x04000000

Maybe just me, since I don't know the HW, but what does _DT stand for
in this context? Other bits seem to be documented with a comment,
let's document the new one as well.

Not sure what the comment above the new define is deferring to, either.
Should it be removed?

>  #define DMA_CONTROL_RSF		0x02000000	/* Receive Store and Forward */
>  #define DMA_CONTROL_DFF		0x01000000	/* Disaable flushing */
-- 
pw-bot: cr

Re: [PATCH net v2 3/3] net: stmmac: dwmac1000: stop hardware from dropping checksum-error packets

[Christian Schmidbauer]

On Tue, Nov 10, 2026 at 9:25 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Wed,  4 Nov 2026 16:19:13 +0100 Christian Schmidbauer wrote:
> >  /* DMA operation mode defines (start/stop tx/rx are placed in common header)*/
> >  /* Disable Drop TCP/IP csum error */
> > +#define DMA_CONTROL_DT               0x04000000
>
> Maybe just me, since I don't know the HW, but what does _DT stand for
> in this context? Other bits seem to be documented with a comment,
> let's document the new one as well.
>
> Not sure what the comment above the new define is deferring to, either.
> Should it be removed?
>

DT is the Synopsys GMAC field name for bit 26 of the DMA Operation
Mode register, see [1]. The line above the define is actually its
description. The define was originally added in 47dd7a540b8a ("net:
add support for STMicroelectronics Ethernet controllers.") with the
comment on the same line, and when the define got removed in
5a78fd3debad ("net: stmmac: remove unused definitions") earlier this
year, the comment stayed behind.

I will move the comment back onto the line of the define in v3.

[1] https://docs.altera.com/v/u/resources/r642178/cyclone-v-hps-register-map