* [PATCH net-next v5 0/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support
@ 2026-11-14 18:29 Suraj Gupta
2026-11-14 18:29 ` [PATCH net-next v5 1/6] net: xilinx: axienet: Introduce axienet_config for MAC-specific ops Suraj Gupta
` (5 more replies)
0 siblings, 6 replies; 12+ messages in thread
From: Suraj Gupta @ 2026-11-14 18:29 UTC (permalink / raw)
To: Radhey Shyam Pandey, Andrew Lunn, David S . Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley
Cc: Michal Simek, Russell King, netdev, devicetree, linux-arm-kernel,
linux-kernel
This patch series extends the AXI Ethernet driver to support the
XXV Ethernet subsystem.
- Refactor through struct axienet_config so per-MAC behaviour is
expressed as capability flags and callbacks instead of runtime
MAC-type checks in the shared code.
- Document the XXV IP in Devicetree.
- Propagate a device reset failure out of ndo_open, which XXV bring-up
needs, and keep a later ndo_stop safe after a failed resume.
- Add support for the XXV Ethernet subsystem.
- Refactor the statistics into struct axienet_config ops, and add
ethtool / ndo_get_stats64 counter reporting for XXV.
---
Notes:
- This series is developed on top of [1]. Only "net: xilinx: axienet:
Derive RX frame length from residue in dmaengine path" from that series
is still outstanding; the rest is already in net-next. This series
builds and applies without it, but that change is required to get the
correct RX completion length on the dmaengine path, which is the only
path XXV supports. It is not included here to avoid duplication.
- The series advertises the core's active 10G or 25G rate only. Runtime
switching between 10GBASE-R and 25GBASE-R is not supported and can be
added later.
- Pause/flow control is not advertised for XXV yet; it is not validated
on this IP and will be enabled in a later change.
[1]: https://lore.kernel.org/all/20260929081316.803997-1-srinivas.neeli@amd.com/
Changes in v5:
- Patch 1/6:
* Move the comment on why axienet_dma_err_handler() ignores the
mac_init() return value into the patch that adds the call.
* Explain the lp->mii_bus NULL guard in the commit message, and fix the
axienet_config kernel-doc for pcs_ops, stats_update and internal_pcs.
* Keep a single axienet_setoptions() and put the option table in
axienet_config instead of adding a setoptions callback, which every
MAC would have implemented identically.
- Patch 2/6:
* Constrain "managed" to in-band-status and disallow phy-handle,
pcs-handle and the deprecated phy and phy-device spellings.
* Add an optional "ref" GT reference clock, clarify that the second
clock is the GT helper (dclk), and require the first Tx/Rx DMA
channel pair. Note that dmas and dma-names pair up by position.
* Add xlnx,has-statistics-counters: the counters are a PG210 build
option and no ability register reports whether they were included.
* Explain in the commit message why xlnx,rxmem is required: probe reads
it unconditionally and it bounds the largest frame accepted.
* Drop Rob Herring's Reviewed-by, as the binding has changed since v3.
- Patch 3/6 (new):
* Split the ndo_open / ndo_stop / resume changes out of the XXV patch,
as they change 1G behaviour and are a prerequisite for it.
* Propagate the axienet_device_reset() error, previously overwritten
by phylink_fwnode_phy_connect() before being checked.
* Clear reset_in_progress on the __axienet_device_reset() error paths,
which the propagated error would otherwise leave set, freezing the 1G
statistics readers.
* Make axienet_stop() a no-op after a failed open, replacing the
per-pointer NULL checks posted earlier, which covered neither the
NAPI and interrupt paths nor a failed dma_request_chan().
- Patch 4/6:
* Derive the line rate from phy-mode instead of STAT_CORE_SPEED, which
PG210 provides only on 64-bit AXI4-Stream builds.
* Program the MAC RX max packet length from max_frm_size so oversized
frames are truncated rather than spilling across descriptors, and
finalize max_mtu from the MAC's jumbo capability.
* Set default_an_inband, request the optional "ref" clock, and skip the
GT registers in ethtool -d on IP versions that lack them.
* Stop advertising pause until flow control is validated, and demote
the RX block-lock timeout to a debug message.
* Name the XXV status registers after PG210; two of them described the
wrong register (0x0400 is STAT_TX_STATUS, 0x0408 is STAT_STATUS).
- Drop the "Make axienet_rmon_ranges non-static for reuse" patch. The XXV
RMON callback reports only aggregate counters and leaves ranges NULL, so
the table was never reused and can stay static.
- Patch 6/6:
* Gate XAE_FEATURE_STATS on xlnx,has-statistics-counters, and gate the
TICK drain, the periodic work and the final snapshot on it, so a
design without the block is never touched.
* Move the draining tick after GT reset-done, as the readable STAT_*
registers hold unknown data until the first TICK.
* Mask the counter MSB to its low 16 bits, so bits outside the 48-bit
counter are not accumulated.
* Report FrameCheckSequenceErrors / rx_crc_errors from
stat_rx_packet_bad_fcs, and drop "TX Bad FCS" and "TX Frame Error"
from ethtool -S, as both are already folded into tx_errors.
* Keep collection on the periodic work; queries copy the software
accumulator and can therefore lag it.
V4: https://lore.kernel.org/all/20261006110550.1990213-1-suraj.gupta2@amd.com/
Changes in v4:
- Drop the "legacy" nomenclature and use dmaengine / non-dmaengine terminology
as it might confuse with MAC instead of driver flow (Andrew Lunn).
- Patch 1/6: Gate the "xlnx,txcsum" DT property behind the MAC's tx_csum
capability, so TX checksum offload is not advertised for a MAC whose
hardware cannot complete checksums.
- Patch 2/6: Reword the binding description to state that the
PCS/PMA is integrated into the IP, and that MDIO bus unavailability and
SFP module (I2C) management are detailed in the IP (Andrew Lunn).
- Patch 3/6: Reword the dma_err_handler comment from "legacy embedded-DMA
(1G) path" to "non-dmaengine flow (1G MAC)" (Andrew Lunn).
- Advertise only the core's active line rate.
V3: https://lore.kernel.org/all/20260831150816.1020883-1-suraj.gupta2@amd.com/
Changes in v3:
- Dropped patch 1/7 ("clk: Add devm_clk_bulk_get_enable()"); it has been
applied to mainline as commit b698927accb1.
- Patch 2/6 (dt-bindings):
* Added Reviewed-by from Rob Herring. Thanks Rob!
V2: https://lore.kernel.org/all/20260806181153.1213996-1-suraj.gupta2@amd.com/
Changes in v2:
- Patch 1/7:
* Add Reviewed-by: Brian Masney <bmasney@redhat.com>. Thanks Brian!
- Patch 3/7:
* Split the XXV binding into its own schema file
(xlnx,xxv-ethernet.yaml) instead of extending xlnx,axi-ethernet.yaml
with if/then branches, per Rob Herring's review.
* Add a multichannel (MCDMA) example.
- Patch 4/7:
* Require both GT reset-done bits in axienet_xxv_poll_link_ready()
so polling does not exit before both the TX and RX resets complete.
V1: https://lore.kernel.org/all/20260723123838.125145-1-suraj.gupta2@amd.com/
Suraj Gupta (6):
net: xilinx: axienet: Introduce axienet_config for MAC-specific ops
dt-bindings: net: xlnx,xxv-ethernet: Add Xilinx XXV 10G/25G Ethernet
net: xilinx: axienet: Propagate device reset failure from ndo_open
net: xilinx: axienet: Add 10G/25G (XXV) ethernet support
net: xilinx: axienet: Dispatch statistics through axienet_config ops
net: xilinx: axienet: Add statistics support for XXV ethernet
.../bindings/net/xlnx,xxv-ethernet.yaml | 146 ++++
MAINTAINERS | 1 +
drivers/net/ethernet/xilinx/Makefile | 3 +-
drivers/net/ethernet/xilinx/xilinx_axienet.h | 96 ++-
.../net/ethernet/xilinx/xilinx_axienet_main.c | 659 +++++++++++++-----
.../net/ethernet/xilinx/xilinx_axienet_xxv.c | 633 +++++++++++++++++
.../net/ethernet/xilinx/xilinx_axienet_xxv.h | 151 ++++
7 files changed, 1506 insertions(+), 183 deletions(-)
create mode 100644 Documentation/devicetree/bindings/net/xlnx,xxv-ethernet.yaml
create mode 100644 drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
create mode 100644 drivers/net/ethernet/xilinx/xilinx_axienet_xxv.h
--
2.25.1
^ permalink raw reply [flat|nested] 12+ messages in thread
* [PATCH net-next v5 1/6] net: xilinx: axienet: Introduce axienet_config for MAC-specific ops
2026-11-14 18:29 [PATCH net-next v5 0/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support Suraj Gupta
@ 2026-11-14 18:29 ` Suraj Gupta
2026-11-16 4:12 ` netdev-bot+sashiko
2026-11-14 18:29 ` [PATCH net-next v5 2/6] dt-bindings: net: xlnx,xxv-ethernet: Add Xilinx XXV 10G/25G Ethernet Suraj Gupta
` (4 subsequent siblings)
5 siblings, 1 reply; 12+ messages in thread
From: Suraj Gupta @ 2026-11-14 18:29 UTC (permalink / raw)
To: Radhey Shyam Pandey, Andrew Lunn, David S . Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley
Cc: Michal Simek, Russell King, netdev, devicetree, linux-arm-kernel,
linux-kernel
Introduce an axienet_config structure to abstract MAC-type-specific
operations, preparing the driver for supporting additional MAC types
beyond the existing 1G Ethernet.
Add axienet_config with capability flags (mdio, mac_irq, non_dmaengine,
uc_filter, mc_filter, vlan, jumbo, tx_csum, sw_padding, internal_pcs),
the register-dump count, the MAC option table and callbacks (clk_init,
probe_init, gt_reset, mac_init, mac_link_up, get_regs, phylink_set_caps,
pcs_ops, stats_update), and populate it from the OF match table. Convert
the common code to consult these capabilities and callbacks instead of
open-coding 1G behavior, so a new MAC type can be added without adding
MAC-specific conditionals to the shared paths.
The MDIO teardown in axienet_remove() is now guarded by a NULL check on
lp->mii_bus. This is required by the refactor, because MAC types that do
not provide an MDIO bus (such as the XXV MAC added later in this series)
leave lp->mii_bus NULL. It also hardens the existing 1G path: a failed
axienet_mdio_setup() already left lp->mii_bus NULL while probe only warns
and continues (e.g. a 1G device using a fixed-link), so unbinding such a
device would dereference NULL in mdiobus_unregister(). Apart from this
guard, no functional change is intended for the existing 1G path.
Signed-off-by: Suraj Gupta <suraj.gupta2@amd.com>
---
drivers/net/ethernet/xilinx/xilinx_axienet.h | 56 +++
.../net/ethernet/xilinx/xilinx_axienet_main.c | 387 ++++++++++++------
2 files changed, 324 insertions(+), 119 deletions(-)
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet.h b/drivers/net/ethernet/xilinx/xilinx_axienet.h
index c63d99686d0b..a87cfa079502 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet.h
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet.h
@@ -547,6 +547,7 @@ struct skbuf_dma_descriptor {
* @rx_ring_head: RX skb ring buffer head index.
* @rx_ring_tail: RX skb ring buffer tail index.
* @rx_submit_lock: Protects RX ring resubmission vs teardown in dmaengine path.
+ * @axienet_config: MAC-type specific configuration and operations.
*/
struct axienet_local {
struct net_device *ndev;
@@ -629,6 +630,61 @@ struct axienet_local {
int rx_ring_head;
int rx_ring_tail;
spinlock_t rx_submit_lock;
+ const struct axienet_config *axienet_config;
+};
+
+/**
+ * struct axienet_config - MAC-type specific configuration and operations
+ * @mdio: MAC has an MDIO bus for external PHY management
+ * @mac_irq: MAC has a dedicated core interrupt
+ * @non_dmaengine: MAC supports the non-dmaengine flow
+ * @uc_filter: MAC has unicast address filter registers
+ * @mc_filter: MAC has multicast address filter registers
+ * @vlan: MAC supports VLAN frames by default
+ * @jumbo: MAC supports jumbo frames
+ * @tx_csum: MAC supports TX checksum offload
+ * @sw_padding: Short frames must be padded in software before transmit
+ * @internal_pcs: MAC has an internal PCS that phylink should poll; when set,
+ * the PCS is registered in poll mode at probe
+ * @regs_n: Number of MAC registers exposed through ethtool -d
+ * @options: Table mapping the XAE_OPTION_* flags to the MAC register bits
+ * that enable them, terminated by a zero @opt entry
+ * @clk_init: Callback to get and enable the MAC-specific clocks
+ * @probe_init: Callback for MAC-specific probe-time init, or NULL if unused
+ * @gt_reset: Callback to reset the serial transceiver (GT), or NULL if unused
+ * @mac_init: Callback for MAC-specific bring-up after the DMA reset
+ * @mac_link_up: Callback to program link speed and pause on link up, or NULL
+ * @get_regs: Callback to dump the MAC registers for ethtool -d
+ * @phylink_set_caps: Callback to set phylink MAC capabilities
+ * @pcs_ops: phylink PCS operations for this MAC; required when @internal_pcs
+ * is set, as the PCS is registered without a further NULL check
+ * @stats_update: Callback to latch/accumulate the periodic MAC counters, or
+ * NULL if the MAC has no periodic statistics counters
+ */
+struct axienet_config {
+ bool mdio;
+ bool mac_irq;
+ bool non_dmaengine;
+ bool uc_filter;
+ bool mc_filter;
+ bool vlan;
+ bool jumbo;
+ bool tx_csum;
+ bool sw_padding;
+ bool internal_pcs;
+ unsigned int regs_n;
+ const struct axienet_option *options;
+ int (*clk_init)(struct axienet_local *lp);
+ void (*probe_init)(struct axienet_local *lp);
+ void (*gt_reset)(struct axienet_local *lp);
+ int (*mac_init)(struct net_device *ndev);
+ void (*mac_link_up)(struct net_device *ndev, int speed,
+ bool tx_pause, bool rx_pause);
+ void (*get_regs)(struct axienet_local *lp, u32 *data);
+ void (*phylink_set_caps)(struct axienet_local *lp,
+ struct phylink_config *config);
+ const struct phylink_pcs_ops *pcs_ops;
+ void (*stats_update)(struct axienet_local *lp);
};
/**
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
index 5b4998e2ca14..83f35effbf37 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
@@ -66,18 +66,8 @@
static void axienet_rx_submit_desc(struct net_device *ndev);
-/* Match table for of_platform binding */
-static const struct of_device_id axienet_of_match[] = {
- { .compatible = "xlnx,axi-ethernet-1.00.a", },
- { .compatible = "xlnx,axi-ethernet-1.01.a", },
- { .compatible = "xlnx,axi-ethernet-2.01.a", },
- {},
-};
-
-MODULE_DEVICE_TABLE(of, axienet_of_match);
-
/* Option table for setting up Axi Ethernet hardware options */
-static struct axienet_option axienet_options[] = {
+static const struct axienet_option axienet_options[] = {
/* Turn on jumbo packet support for both Rx and Tx */
{
.opt = XAE_OPTION_JUMBO,
@@ -470,6 +460,9 @@ static void axienet_set_mac_address(struct net_device *ndev,
if (!is_valid_ether_addr(ndev->dev_addr))
eth_hw_addr_random(ndev);
+ if (!lp->axienet_config->uc_filter)
+ return;
+
/* Set up unicast MAC address filter set its mac address */
axienet_iow(lp, XAE_UAW0_OFFSET,
(ndev->dev_addr[0]) |
@@ -519,6 +512,9 @@ static void axienet_set_multicast_list(struct net_device *ndev)
u32 reg, af0reg, af1reg;
struct axienet_local *lp = netdev_priv(ndev);
+ if (!lp->axienet_config->mc_filter)
+ return;
+
reg = axienet_ior(lp, XAE_FMI_OFFSET);
reg &= ~XAE_FMI_PM_MASK;
if (ndev->flags & IFF_PROMISC)
@@ -581,14 +577,14 @@ static void axienet_set_multicast_list(struct net_device *ndev)
* The Axi Ethernet core has multiple features which can be selectively turned
* on or off. The typical options could be jumbo frame option, basic VLAN
* option, promiscuous mode option etc. This function is used to set or clear
- * these options in the Axi Ethernet hardware. This is done through
- * axienet_option structure .
+ * these options in the Axi Ethernet hardware. This is done through the
+ * MAC-specific axienet_option table in &struct axienet_config.
*/
static void axienet_setoptions(struct net_device *ndev, u32 options)
{
int reg;
struct axienet_local *lp = netdev_priv(ndev);
- struct axienet_option *tp = &axienet_options[0];
+ const struct axienet_option *tp = lp->axienet_config->options;
while (tp->opt) {
reg = ((axienet_ior(lp, tp->reg)) & ~(tp->m_or));
@@ -627,14 +623,25 @@ static void axienet_stats_update(struct axienet_local *lp, bool reset)
write_seqcount_end(&lp->hw_stats_seqcount);
}
+/**
+ * axienet_1g_stats_update - Refresh 1G MAC hardware statistics
+ * @lp: Pointer to the axienet_local structure
+ */
+static void axienet_1g_stats_update(struct axienet_local *lp)
+{
+ axienet_stats_update(lp, false);
+}
+
static void axienet_refresh_stats(struct work_struct *work)
{
struct axienet_local *lp = container_of(work, struct axienet_local,
stats_work.work);
- mutex_lock(&lp->stats_lock);
- axienet_stats_update(lp, false);
- mutex_unlock(&lp->stats_lock);
+ if (lp->axienet_config->stats_update) {
+ mutex_lock(&lp->stats_lock);
+ lp->axienet_config->stats_update(lp);
+ mutex_unlock(&lp->stats_lock);
+ }
/* Just less than 2^32 bytes at 2.5 GBit/s */
schedule_delayed_work(&lp->stats_work, 13 * HZ);
@@ -745,6 +752,32 @@ static void axienet_dma_stop(struct axienet_local *lp)
axienet_unlock_mii(lp);
}
+/**
+ * axienet_1g_mac_init - 1G MAC-specific bring-up after the DMA reset
+ * @ndev: Pointer to the net_device structure
+ *
+ * Return: 0 on success or a negative error number otherwise.
+ */
+static int axienet_1g_mac_init(struct net_device *ndev)
+{
+ struct axienet_local *lp = netdev_priv(ndev);
+ u32 axienet_status;
+
+ axienet_status = axienet_ior(lp, XAE_RCW1_OFFSET);
+ axienet_status &= ~XAE_RCW1_RX_MASK;
+ axienet_iow(lp, XAE_RCW1_OFFSET, axienet_status);
+
+ axienet_status = axienet_ior(lp, XAE_IP_OFFSET);
+ if (axienet_status & XAE_INT_RXRJECT_MASK)
+ axienet_iow(lp, XAE_IS_OFFSET, XAE_INT_RXRJECT_MASK);
+ axienet_iow(lp, XAE_IE_OFFSET, lp->eth_irq > 0 ?
+ XAE_INT_RECV_ERROR_MASK : 0);
+
+ axienet_iow(lp, XAE_FCC_OFFSET, XAE_FCC_FCRX_MASK);
+
+ return 0;
+}
+
/**
* axienet_device_reset - Reset and initialize the Axi Ethernet hardware.
* @ndev: Pointer to the net_device structure
@@ -760,19 +793,23 @@ static void axienet_dma_stop(struct axienet_local *lp)
*/
static int axienet_device_reset(struct net_device *ndev)
{
- u32 axienet_status;
struct axienet_local *lp = netdev_priv(ndev);
int ret;
+ if (lp->axienet_config->gt_reset)
+ lp->axienet_config->gt_reset(lp);
+
lp->max_frm_size = XAE_MAX_VLAN_FRAME_SIZE;
- lp->options |= XAE_OPTION_VLAN;
+
+ if (lp->axienet_config->vlan)
+ lp->options |= XAE_OPTION_VLAN;
lp->options &= (~XAE_OPTION_JUMBO);
if (ndev->mtu > XAE_MTU && ndev->mtu <= XAE_JUMBO_MTU) {
lp->max_frm_size = ndev->mtu + VLAN_ETH_HLEN +
XAE_TRL_SIZE;
- if (lp->max_frm_size <= lp->rxmem)
+ if (lp->max_frm_size <= lp->rxmem && lp->axienet_config->jumbo)
lp->options |= XAE_OPTION_JUMBO;
}
@@ -789,17 +826,9 @@ static int axienet_device_reset(struct net_device *ndev)
}
}
- axienet_status = axienet_ior(lp, XAE_RCW1_OFFSET);
- axienet_status &= ~XAE_RCW1_RX_MASK;
- axienet_iow(lp, XAE_RCW1_OFFSET, axienet_status);
-
- axienet_status = axienet_ior(lp, XAE_IP_OFFSET);
- if (axienet_status & XAE_INT_RXRJECT_MASK)
- axienet_iow(lp, XAE_IS_OFFSET, XAE_INT_RXRJECT_MASK);
- axienet_iow(lp, XAE_IE_OFFSET, lp->eth_irq > 0 ?
- XAE_INT_RECV_ERROR_MASK : 0);
-
- axienet_iow(lp, XAE_FCC_OFFSET, XAE_FCC_FCRX_MASK);
+ ret = lp->axienet_config->mac_init(ndev);
+ if (ret)
+ return ret;
/* Sync default options with HW but leave receiver and
* transmitter disabled.
@@ -967,6 +996,13 @@ axienet_start_xmit_dmaengine(struct sk_buff *skb, struct net_device *ndev)
int sg_len;
int ret;
+ if (lp->axienet_config->sw_padding && eth_skb_pad(skb)) {
+ /* Pad short frames; the MAC does not append Ethernet padding. */
+ ndev->stats.tx_dropped++;
+ ndev->stats.tx_errors++;
+ return NETDEV_TX_OK;
+ }
+
dma_dev = lp->tx_chan->device;
sg_len = skb_shinfo(skb)->nr_frags + 1;
if (CIRC_SPACE(lp->tx_ring_head, lp->tx_ring_tail, TX_BD_NUM_MAX) <= 1) {
@@ -990,19 +1026,21 @@ axienet_start_xmit_dmaengine(struct sk_buff *skb, struct net_device *ndev)
goto xmit_error_drop_skb;
/* Fill up app fields for checksum */
- if (skb->ip_summed == CHECKSUM_PARTIAL) {
- if (lp->features & XAE_FEATURE_FULL_TX_CSUM) {
- /* Tx Full Checksum Offload Enabled */
- app_metadata[0] |= 2;
- } else if (lp->features & XAE_FEATURE_PARTIAL_TX_CSUM) {
- csum_start_off = skb_transport_offset(skb);
- csum_index_off = csum_start_off + skb->csum_offset;
- /* Tx Partial Checksum Offload Enabled */
- app_metadata[0] |= 1;
- app_metadata[1] = (csum_start_off << 16) | csum_index_off;
+ if (lp->axienet_config->tx_csum) {
+ if (skb->ip_summed == CHECKSUM_PARTIAL) {
+ if (lp->features & XAE_FEATURE_FULL_TX_CSUM) {
+ /* Tx Full Checksum Offload Enabled */
+ app_metadata[0] |= 2;
+ } else if (lp->features & XAE_FEATURE_PARTIAL_TX_CSUM) {
+ csum_start_off = skb_transport_offset(skb);
+ csum_index_off = csum_start_off + skb->csum_offset;
+ /* Tx Partial Checksum Offload Enabled */
+ app_metadata[0] |= 1;
+ app_metadata[1] = (csum_start_off << 16) | csum_index_off;
+ }
+ } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
+ app_metadata[0] |= 2; /* Tx Full Checksum Offload Enabled */
}
- } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
- app_metadata[0] |= 2; /* Tx Full Checksum Offload Enabled */
}
dma_tx_desc = dma_dev->device_prep_slave_sg(lp->tx_chan, skbuf_dma->sgl,
@@ -1855,7 +1893,8 @@ static int axienet_stop(struct net_device *ndev)
}
netdev_reset_queue(ndev);
- axienet_iow(lp, XAE_IE_OFFSET, 0);
+ if (lp->axienet_config->mac_irq)
+ axienet_iow(lp, XAE_IE_OFFSET, 0);
if (lp->eth_irq > 0)
free_irq(lp->eth_irq, ndev);
@@ -2024,30 +2063,18 @@ static void axienet_ethtools_get_drvinfo(struct net_device *ndev,
*/
static int axienet_ethtools_get_regs_len(struct net_device *ndev)
{
- return sizeof(u32) * AXIENET_REGS_N;
+ struct axienet_local *lp = netdev_priv(ndev);
+
+ return sizeof(u32) * lp->axienet_config->regs_n;
}
/**
- * axienet_ethtools_get_regs - Dump the contents of all registers present
- * in AxiEthernet core.
- * @ndev: Pointer to net_device structure
- * @regs: Pointer to ethtool_regs structure
- * @ret: Void pointer used to return the contents of the registers.
- *
- * This implements ethtool command for getting the Axi Ethernet register dump.
- * Issue "ethtool -d ethX" to execute this function.
+ * axienet_1g_get_regs - Dump 1G MAC registers for ethtool
+ * @lp: Pointer to the axienet_local structure
+ * @data: Buffer for register values (zeroed and sized by the caller)
*/
-static void axienet_ethtools_get_regs(struct net_device *ndev,
- struct ethtool_regs *regs, void *ret)
+static void axienet_1g_get_regs(struct axienet_local *lp, u32 *data)
{
- u32 *data = (u32 *)ret;
- size_t len = sizeof(u32) * AXIENET_REGS_N;
- struct axienet_local *lp = netdev_priv(ndev);
-
- regs->version = 0;
- regs->len = len;
-
- memset(data, 0, len);
data[0] = axienet_ior(lp, XAE_RAF_OFFSET);
data[1] = axienet_ior(lp, XAE_TPF_OFFSET);
data[2] = axienet_ior(lp, XAE_IFGP_OFFSET);
@@ -2088,6 +2115,31 @@ static void axienet_ethtools_get_regs(struct net_device *ndev,
}
}
+/**
+ * axienet_ethtools_get_regs - Dump the contents of all registers present
+ * in AxiEthernet core.
+ * @ndev: Pointer to net_device structure
+ * @regs: Pointer to ethtool_regs structure
+ * @ret: Void pointer used to return the contents of the registers.
+ *
+ * This implements ethtool command for getting the Axi Ethernet register dump.
+ * Issue "ethtool -d ethX" to execute this function.
+ */
+static void axienet_ethtools_get_regs(struct net_device *ndev,
+ struct ethtool_regs *regs, void *ret)
+{
+ struct axienet_local *lp = netdev_priv(ndev);
+ u32 *data = (u32 *)ret;
+ size_t len;
+
+ len = sizeof(u32) * lp->axienet_config->regs_n;
+ regs->version = 0;
+ regs->len = len;
+
+ memset(data, 0, len);
+ lp->axienet_config->get_regs(lp, data);
+}
+
static void
axienet_ethtools_get_ringparam(struct net_device *ndev,
struct ethtool_ringparam *ering,
@@ -2724,13 +2776,9 @@ static void axienet_mac_link_down(struct phylink_config *config,
/* nothing meaningful to do */
}
-static void axienet_mac_link_up(struct phylink_config *config,
- struct phy_device *phy,
- unsigned int mode, phy_interface_t interface,
- int speed, int duplex,
- bool tx_pause, bool rx_pause)
+static void axienet_1g_mac_link_up(struct net_device *ndev, int speed,
+ bool tx_pause, bool rx_pause)
{
- struct net_device *ndev = to_net_dev(config->dev);
struct axienet_local *lp = netdev_priv(ndev);
u32 emmc_reg, fcc_reg;
@@ -2767,6 +2815,19 @@ static void axienet_mac_link_up(struct phylink_config *config,
axienet_iow(lp, XAE_FCC_OFFSET, fcc_reg);
}
+static void axienet_mac_link_up(struct phylink_config *config,
+ struct phy_device *phy,
+ unsigned int mode, phy_interface_t interface,
+ int speed, int duplex,
+ bool tx_pause, bool rx_pause)
+{
+ struct net_device *ndev = to_net_dev(config->dev);
+ struct axienet_local *lp = netdev_priv(ndev);
+
+ if (lp->axienet_config->mac_link_up)
+ lp->axienet_config->mac_link_up(ndev, speed, tx_pause, rx_pause);
+}
+
static const struct phylink_mac_ops axienet_phylink_ops = {
.mac_select_pcs = axienet_mac_select_pcs,
.mac_config = axienet_mac_config,
@@ -2784,7 +2845,6 @@ static const struct phylink_mac_ops axienet_phylink_ops = {
static void axienet_dma_err_handler(struct work_struct *work)
{
u32 i;
- u32 axienet_status;
struct axidma_bd *cur_p;
struct axienet_local *lp = container_of(work, struct axienet_local,
dma_err_task);
@@ -2821,16 +2881,10 @@ static void axienet_dma_err_handler(struct work_struct *work)
axienet_dma_start(lp);
- axienet_status = axienet_ior(lp, XAE_RCW1_OFFSET);
- axienet_status &= ~XAE_RCW1_RX_MASK;
- axienet_iow(lp, XAE_RCW1_OFFSET, axienet_status);
-
- axienet_status = axienet_ior(lp, XAE_IP_OFFSET);
- if (axienet_status & XAE_INT_RXRJECT_MASK)
- axienet_iow(lp, XAE_IS_OFFSET, XAE_INT_RXRJECT_MASK);
- axienet_iow(lp, XAE_IE_OFFSET, lp->eth_irq > 0 ?
- XAE_INT_RECV_ERROR_MASK : 0);
- axienet_iow(lp, XAE_FCC_OFFSET, XAE_FCC_FCRX_MASK);
+ /* This error handler runs only for the non-dmaengine flow (1G MAC),
+ * whose mac_init() cannot fail, so its return value is not checked.
+ */
+ lp->axienet_config->mac_init(ndev);
/* Sync default options with HW but leave receiver and
* transmitter disabled.
@@ -2844,6 +2898,100 @@ static void axienet_dma_err_handler(struct work_struct *work)
axienet_setoptions(ndev, lp->options);
}
+/**
+ * axienet_1g_phylink_set_caps - Set 1G phylink capabilities and interfaces
+ * @lp: Pointer to the axienet_local structure
+ * @config: Pointer to the phylink_config structure
+ */
+static void axienet_1g_phylink_set_caps(struct axienet_local *lp,
+ struct phylink_config *config)
+{
+ config->mac_capabilities |= MAC_10FD | MAC_100FD | MAC_1000FD;
+ __set_bit(lp->phy_mode, config->supported_interfaces);
+ if (lp->switch_x_sgmii) {
+ __set_bit(PHY_INTERFACE_MODE_1000BASEX,
+ config->supported_interfaces);
+ __set_bit(PHY_INTERFACE_MODE_SGMII,
+ config->supported_interfaces);
+ }
+}
+
+/**
+ * axienet_1g_clk_init - Get and enable the 1G MAC clocks
+ * @lp: Pointer to the axienet_local structure
+ *
+ * Return: 0 on success or a negative error number otherwise.
+ */
+static int axienet_1g_clk_init(struct axienet_local *lp)
+{
+ struct device *dev = lp->dev;
+ int ret;
+
+ lp->axi_clk = devm_clk_get_optional_enabled(dev, "s_axi_lite_clk");
+ if (!lp->axi_clk) {
+ /* For backward compatibility, if named AXI clock is not
+ * present, treat the first clock specified as the AXI clock.
+ */
+ lp->axi_clk = devm_clk_get_optional_enabled(dev, NULL);
+ }
+ if (IS_ERR(lp->axi_clk))
+ return dev_err_probe(dev, PTR_ERR(lp->axi_clk),
+ "could not get AXI clock\n");
+
+ lp->misc_clks[0].id = "axis_clk";
+ lp->misc_clks[1].id = "ref_clk";
+ lp->misc_clks[2].id = "mgt_clk";
+
+ ret = devm_clk_bulk_get_optional_enable(dev, XAE_NUM_MISC_CLOCKS,
+ lp->misc_clks);
+ if (ret)
+ return dev_err_probe(dev, ret,
+ "could not get/enable misc. clocks\n");
+
+ return 0;
+}
+
+/**
+ * axienet_1g_probe_init - 1G MAC-specific probe-time initialization
+ * @lp: Pointer to the axienet_local structure
+ */
+static void axienet_1g_probe_init(struct axienet_local *lp)
+{
+ if (axienet_ior(lp, XAE_ABILITY_OFFSET) & XAE_ABILITY_STATS)
+ lp->features |= XAE_FEATURE_STATS;
+}
+
+static const struct axienet_config axienet_1g_config = {
+ .mdio = true,
+ .mac_irq = true,
+ .non_dmaengine = true,
+ .uc_filter = true,
+ .mc_filter = true,
+ .vlan = true,
+ .jumbo = true,
+ .tx_csum = true,
+ .regs_n = AXIENET_REGS_N,
+ .options = axienet_options,
+ .clk_init = axienet_1g_clk_init,
+ .probe_init = axienet_1g_probe_init,
+ .mac_init = axienet_1g_mac_init,
+ .mac_link_up = axienet_1g_mac_link_up,
+ .get_regs = axienet_1g_get_regs,
+ .phylink_set_caps = axienet_1g_phylink_set_caps,
+ .pcs_ops = &axienet_pcs_ops,
+ .stats_update = axienet_1g_stats_update,
+};
+
+/* Match table for of_platform binding */
+static const struct of_device_id axienet_of_match[] = {
+ { .compatible = "xlnx,axi-ethernet-1.00.a", .data = &axienet_1g_config },
+ { .compatible = "xlnx,axi-ethernet-1.01.a", .data = &axienet_1g_config },
+ { .compatible = "xlnx,axi-ethernet-2.01.a", .data = &axienet_1g_config },
+ {},
+};
+
+MODULE_DEVICE_TABLE(of, axienet_of_match);
+
/**
* axienet_probe - Axi Ethernet probe function.
* @pdev: Pointer to platform device structure.
@@ -2894,27 +3042,13 @@ static int axienet_probe(struct platform_device *pdev)
seqcount_mutex_init(&lp->hw_stats_seqcount, &lp->stats_lock);
INIT_DEFERRABLE_WORK(&lp->stats_work, axienet_refresh_stats);
- lp->axi_clk = devm_clk_get_optional_enabled(&pdev->dev,
- "s_axi_lite_clk");
- if (!lp->axi_clk) {
- /* For backward compatibility, if named AXI clock is not present,
- * treat the first clock specified as the AXI clock.
- */
- lp->axi_clk = devm_clk_get_optional_enabled(&pdev->dev, NULL);
- }
- if (IS_ERR(lp->axi_clk))
- return dev_err_probe(&pdev->dev, PTR_ERR(lp->axi_clk),
- "could not get AXI clock\n");
-
- lp->misc_clks[0].id = "axis_clk";
- lp->misc_clks[1].id = "ref_clk";
- lp->misc_clks[2].id = "mgt_clk";
+ lp->axienet_config = device_get_match_data(&pdev->dev);
+ if (!lp->axienet_config)
+ return -ENODEV;
- ret = devm_clk_bulk_get_optional_enable(&pdev->dev, XAE_NUM_MISC_CLOCKS,
- lp->misc_clks);
+ ret = lp->axienet_config->clk_init(lp);
if (ret)
- return dev_err_probe(&pdev->dev, ret,
- "could not get/enable misc. clocks\n");
+ return ret;
/* Map device registers */
lp->regs = devm_platform_get_and_ioremap_resource(pdev, 0, ðres);
@@ -2925,11 +3059,12 @@ static int axienet_probe(struct platform_device *pdev)
/* Setup checksum offload, but default to off if not specified */
lp->features = 0;
- if (axienet_ior(lp, XAE_ABILITY_OFFSET) & XAE_ABILITY_STATS)
- lp->features |= XAE_FEATURE_STATS;
-
+ /* Only honour the TX checksum offload property on MACs whose DMA path
+ * can complete checksums. Otherwise the stack would hand down partial
+ * checksums that the hardware never finishes.
+ */
ret = of_property_read_u32(pdev->dev.of_node, "xlnx,txcsum", &value);
- if (!ret) {
+ if (!ret && lp->axienet_config->tx_csum) {
switch (value) {
case 1:
lp->features |= XAE_FEATURE_PARTIAL_TX_CSUM;
@@ -3005,7 +3140,17 @@ static int axienet_probe(struct platform_device *pdev)
return -EINVAL;
}
+ if (lp->axienet_config->probe_init)
+ lp->axienet_config->probe_init(lp);
+
if (!of_property_present(pdev->dev.of_node, "dmas")) {
+ /* Non-dmaengine mode is only supported by some MACs */
+ if (!lp->axienet_config->non_dmaengine) {
+ dev_err(&pdev->dev,
+ "Non-dmaengine mode not supported by this MAC, use dmaengine\n");
+ return -EINVAL;
+ }
+
/* Find the DMA node, map the DMA registers, and decode the DMA IRQs */
np = of_parse_phandle(pdev->dev.of_node, "axistream-connected", 0);
@@ -3091,7 +3236,10 @@ static int axienet_probe(struct platform_device *pdev)
struct xilinx_vdma_config cfg;
struct dma_chan *tx_chan;
- lp->eth_irq = platform_get_irq_optional(pdev, 0);
+ if (lp->axienet_config->mac_irq)
+ lp->eth_irq = platform_get_irq_optional(pdev, 0);
+ else
+ lp->eth_irq = -ENXIO;
if (lp->eth_irq < 0 && lp->eth_irq != -ENXIO) {
return lp->eth_irq;
}
@@ -3145,10 +3293,12 @@ static int axienet_probe(struct platform_device *pdev)
lp->tx_dma_cr = axienet_calc_cr(lp, XAXIDMA_DFT_TX_THRESHOLD,
XAXIDMA_DFT_TX_USEC);
- ret = axienet_mdio_setup(lp);
- if (ret)
- dev_warn(&pdev->dev,
- "error registering MDIO bus: %d\n", ret);
+ if (lp->axienet_config->mdio) {
+ ret = axienet_mdio_setup(lp);
+ if (ret)
+ dev_warn(&pdev->dev,
+ "error registering MDIO bus: %d\n", ret);
+ }
if (lp->phy_mode == PHY_INTERFACE_MODE_SGMII ||
lp->phy_mode == PHY_INTERFACE_MODE_1000BASEX) {
@@ -3172,23 +3322,21 @@ static int axienet_probe(struct platform_device *pdev)
goto cleanup_mdio;
}
of_node_put(np);
- lp->pcs.ops = &axienet_pcs_ops;
+ lp->pcs.ops = lp->axienet_config->pcs_ops;
+ lp->pcs.poll = true;
+ }
+
+ if (lp->axienet_config->internal_pcs) {
+ lp->pcs.ops = lp->axienet_config->pcs_ops;
lp->pcs.poll = true;
}
lp->phylink_config.dev = &ndev->dev;
lp->phylink_config.type = PHYLINK_NETDEV;
lp->phylink_config.mac_managed_pm = true;
- lp->phylink_config.mac_capabilities = MAC_SYM_PAUSE | MAC_ASYM_PAUSE |
- MAC_10FD | MAC_100FD | MAC_1000FD;
+ lp->phylink_config.mac_capabilities = MAC_SYM_PAUSE | MAC_ASYM_PAUSE;
- __set_bit(lp->phy_mode, lp->phylink_config.supported_interfaces);
- if (lp->switch_x_sgmii) {
- __set_bit(PHY_INTERFACE_MODE_1000BASEX,
- lp->phylink_config.supported_interfaces);
- __set_bit(PHY_INTERFACE_MODE_SGMII,
- lp->phylink_config.supported_interfaces);
- }
+ lp->axienet_config->phylink_set_caps(lp, &lp->phylink_config);
lp->phylink = phylink_create(&lp->phylink_config, pdev->dev.fwnode,
lp->phy_mode,
@@ -3231,7 +3379,8 @@ static void axienet_remove(struct platform_device *pdev)
if (lp->pcs_phy)
put_device(&lp->pcs_phy->dev);
- axienet_mdio_teardown(lp);
+ if (lp->mii_bus)
+ axienet_mdio_teardown(lp);
}
static void axienet_shutdown(struct platform_device *pdev)
--
2.25.1
^ permalink raw reply related [flat|nested] 12+ messages in thread
* [PATCH net-next v5 2/6] dt-bindings: net: xlnx,xxv-ethernet: Add Xilinx XXV 10G/25G Ethernet
2026-11-14 18:29 [PATCH net-next v5 0/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support Suraj Gupta
2026-11-14 18:29 ` [PATCH net-next v5 1/6] net: xilinx: axienet: Introduce axienet_config for MAC-specific ops Suraj Gupta
@ 2026-11-14 18:29 ` Suraj Gupta
2026-11-16 4:12 ` netdev-bot+sashiko
2026-11-14 18:29 ` [PATCH net-next v5 3/6] net: xilinx: axienet: Propagate device reset failure from ndo_open Suraj Gupta
` (3 subsequent siblings)
5 siblings, 1 reply; 12+ messages in thread
From: Suraj Gupta @ 2026-11-14 18:29 UTC (permalink / raw)
To: Radhey Shyam Pandey, Andrew Lunn, David S . Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley
Cc: Michal Simek, Russell King, netdev, devicetree, linux-arm-kernel,
linux-kernel
The XXV Ethernet IP integrates a 10G/25G Ethernet MAC with an in-core
PCS/PMA driving a high-speed serial transceiver. Its payload is moved by
an external AXI DMA (or multichannel DMA) controller referenced through
the dmas/dma-names properties. Document the IP in a dedicated schema
file.
Add the "xlnx,xxv-ethernet-1.0" compatible with the 10gbase-r and
25gbase-r phy-modes, the rx_core/drp/s_axi clocks and the AXI DMA
channel properties. The GT transceiver reference clock is described as
an optional fourth "ref" clock, for boards where it is supplied by a
Linux-managed clock provider rather than a fixed oscillator.
Add "xlnx,has-statistics-counters" as well. The statistics counters are
an optional part of the IP, selected by the "Include Statistics Counters"
build option (PG210), and unlike the 1G MAC the subsystem has no ability
register that reports whether they were included, so the devicetree has
to describe it.
Require "xlnx,rxmem". The driver reads it unconditionally at probe time
and fails if it is absent, and the value is what bounds the largest frame
the device accepts, both when validating a new MTU and when deciding
whether jumbo frames can be enabled.
Disallow phy-handle, its deprecated phy and phy-device spellings, and
pcs-handle. All are inherited from ethernet-controller.yaml, so
unevaluatedProperties does not reject them, but the subsystem has no MDIO
bus master and its PCS is in-core, so neither an external PHY nor an
external PCS can be attached. Constrain managed to in-band-status for the
same reason; it stays optional, as the link is always managed in-band.
Signed-off-by: Suraj Gupta <suraj.gupta2@amd.com>
---
.../bindings/net/xlnx,xxv-ethernet.yaml | 146 ++++++++++++++++++
MAINTAINERS | 1 +
2 files changed, 147 insertions(+)
create mode 100644 Documentation/devicetree/bindings/net/xlnx,xxv-ethernet.yaml
diff --git a/Documentation/devicetree/bindings/net/xlnx,xxv-ethernet.yaml b/Documentation/devicetree/bindings/net/xlnx,xxv-ethernet.yaml
new file mode 100644
index 000000000000..caf582c72a8d
--- /dev/null
+++ b/Documentation/devicetree/bindings/net/xlnx,xxv-ethernet.yaml
@@ -0,0 +1,146 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/net/xlnx,xxv-ethernet.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Xilinx XXV 10G/25G Ethernet Subsystem
+
+maintainers:
+ - Radhey Shyam Pandey <radhey.shyam.pandey@amd.com>
+ - Suraj Gupta <suraj.gupta2@amd.com>
+
+description:
+ The Xilinx XXV Ethernet Subsystem implements a 10G/25G Ethernet MAC with an
+ in-core PCS/PMA driving a high-speed serial transceiver. It supports the
+ 10GBASE-R and 25GBASE-R line rates, with 25G operation following the 25G
+ Ethernet Consortium specification. Because the PCS/PMA is integrated into
+ the IP, the subsystem has no MDIO bus master and is not managed through an
+ external MDIO PHY; link state is read from the MAC's own status registers.
+ As there is no PHY to manage, the link is always managed in-band.
+ Any SFP module management (e.g. I2C) is handled by separate IP outside this
+ binding. Its payload is moved by an external AXI DMA (or multichannel DMA)
+ controller referenced through the "dmas" property.
+
+properties:
+ compatible:
+ const: xlnx,xxv-ethernet-1.0
+
+ reg:
+ maxItems: 1
+ description: MAC control and status register space.
+
+ clocks:
+ minItems: 3
+ items:
+ - description: RX AXI4-Stream clock.
+ - description: GT helper (dclk) stable clock.
+ - description: AXI4-Lite register clock.
+ - description: GT transceiver reference clock (refclk).
+
+ clock-names:
+ minItems: 3
+ items:
+ - const: rx_core
+ - const: drp
+ - const: s_axi
+ - const: ref
+
+ phy-mode:
+ enum:
+ - 10gbase-r
+ - 25gbase-r
+
+ # The subsystem has no MDIO bus master and its PCS is in-core, so neither
+ # an external PHY nor an external PCS can be attached. phy and phy-device
+ # are the deprecated spellings of phy-handle.
+ phy-handle: false
+ phy: false
+ phy-device: false
+ pcs-handle: false
+
+ managed:
+ const: in-band-status
+
+ xlnx,rxmem:
+ description:
+ Size in bytes of the Rx/Tx packet buffer memory built into the IP. It is
+ fixed by the hardware design and bounds the largest frame (jumbo MTU) the
+ device can accept.
+ $ref: /schemas/types.yaml#/definitions/uint32
+
+ xlnx,has-statistics-counters:
+ description:
+ The statistics counters are an optional part of the IP, selected by the
+ "Include Statistics Counters" build option, and the subsystem does not
+ advertise whether they were included. Present when the design was built
+ with them, in which case the driver reports the MAC counters through
+ ethtool and ndo_get_stats64.
+ type: boolean
+
+ dmas:
+ minItems: 2
+ maxItems: 32
+ description:
+ TX and RX DMA channel phandles, one per entry in "dma-names".
+
+ dma-names:
+ minItems: 2
+ maxItems: 32
+ items:
+ pattern: "^[tr]x_chan([0-9]|1[0-5])$"
+ allOf:
+ - contains:
+ const: tx_chan0
+ - contains:
+ const: rx_chan0
+ description:
+ Should be "tx_chan0", "tx_chan1" ... "tx_chan15" for DMA Tx channel and
+ "rx_chan0", "rx_chan1" ... "rx_chan15" for DMA Rx channel. The first
+ Tx/Rx channel pair ("tx_chan0" and "rx_chan0") is always required. There
+ is one name per entry in "dmas", in the same order.
+
+required:
+ - compatible
+ - reg
+ - clocks
+ - clock-names
+ - phy-mode
+ - xlnx,rxmem
+ - dmas
+ - dma-names
+
+allOf:
+ - $ref: /schemas/net/ethernet-controller.yaml#
+
+unevaluatedProperties: false
+
+examples:
+ # Single AXI DMA channel pair
+ - |
+ ethernet@80000000 {
+ compatible = "xlnx,xxv-ethernet-1.0";
+ reg = <0x80000000 0x10000>;
+ clock-names = "rx_core", "drp", "s_axi";
+ clocks = <&rx_clk>, <&drp_clk>, <&axi_clk>;
+ dma-names = "tx_chan0", "rx_chan0";
+ dmas = <&axi_dma 0>, <&axi_dma 1>;
+ phy-mode = "25gbase-r";
+ managed = "in-band-status";
+ xlnx,rxmem = <0x8000>;
+ xlnx,has-statistics-counters;
+ };
+ # Multichannel DMA (MCDMA) with multiple Tx/Rx channels
+ - |
+ ethernet@80010000 {
+ compatible = "xlnx,xxv-ethernet-1.0";
+ reg = <0x80010000 0x10000>;
+ clock-names = "rx_core", "drp", "s_axi";
+ clocks = <&rx_clk>, <&drp_clk>, <&axi_clk>;
+ dma-names = "tx_chan0", "tx_chan1", "rx_chan0", "rx_chan1";
+ dmas = <&axi_mcdma 0>, <&axi_mcdma 1>,
+ <&axi_mcdma 16>, <&axi_mcdma 17>;
+ phy-mode = "10gbase-r";
+ managed = "in-band-status";
+ xlnx,rxmem = <0x8000>;
+ };
diff --git a/MAINTAINERS b/MAINTAINERS
index 893b0670a7b6..b33fd70d242f 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -30683,6 +30683,7 @@ XILINX AXI ETHERNET DRIVER
M: Radhey Shyam Pandey <radhey.shyam.pandey@amd.com>
S: Maintained
F: Documentation/devicetree/bindings/net/xlnx,axi-ethernet.yaml
+F: Documentation/devicetree/bindings/net/xlnx,xxv-ethernet.yaml
F: drivers/net/ethernet/xilinx/xilinx_axienet*
XILINX CAN DRIVER
--
2.25.1
^ permalink raw reply related [flat|nested] 12+ messages in thread
* [PATCH net-next v5 3/6] net: xilinx: axienet: Propagate device reset failure from ndo_open
2026-11-14 18:29 [PATCH net-next v5 0/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support Suraj Gupta
2026-11-14 18:29 ` [PATCH net-next v5 1/6] net: xilinx: axienet: Introduce axienet_config for MAC-specific ops Suraj Gupta
2026-11-14 18:29 ` [PATCH net-next v5 2/6] dt-bindings: net: xlnx,xxv-ethernet: Add Xilinx XXV 10G/25G Ethernet Suraj Gupta
@ 2026-11-14 18:29 ` Suraj Gupta
2026-11-16 4:12 ` netdev-bot+sashiko
2026-11-14 18:29 ` [PATCH net-next v5 4/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support Suraj Gupta
` (2 subsequent siblings)
5 siblings, 1 reply; 12+ messages in thread
From: Suraj Gupta @ 2026-11-14 18:29 UTC (permalink / raw)
To: Radhey Shyam Pandey, Andrew Lunn, David S . Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley
Cc: Michal Simek, Russell King, netdev, devicetree, linux-arm-kernel,
linux-kernel
axienet_open() overwrites the axienet_device_reset() return value with the
phylink_fwnode_phy_connect() one before it is ever checked, so a failed
reset brings the device up anyway. Check and propagate it.
That makes a failed ndo_open reachable, which matters for axienet_resume():
it calls ndo_open directly, so unlike the netdev core it leaves the device
running when open fails. A later ndo_stop then tears down resources that
open had already unwound and never re-created. For the dmaengine flow that
is a double free of the skb rings and a second dma_release_channel(); for
the non-dmaengine flow it is napi_disable() on an already disabled NAPI,
which never returns, and free_irq() on interrupts that were never
requested.
Track whether ndo_open completed and make axienet_stop() a no-op otherwise.
axienet_resume() now propagates the error, but it still has to undo the
netif_device_detach() that axienet_suspend() did. Leaving the device
detached while it is still IFF_UP wedges the interface: ndo_stop is a
no-op, so bringing it down appears to work, while bringing it up again
fails with -ENODEV because the device is not present. Drop IFF_UP with
netif_close() before reattaching, so the interface can be brought up again
once the cause of the failure is resolved, and so the stack cannot
transmit on the datapath open has already torn down.
Clearing the error path of __axienet_device_reset() goes with this: it
jumps over the block that clears lp->reset_in_progress, and the statistics
work that used to clear it on the next run is no longer scheduled once the
reset error is propagated. Without this, axienet_stat() would keep
returning the frozen base after a reset timeout.
Signed-off-by: Suraj Gupta <suraj.gupta2@amd.com>
---
drivers/net/ethernet/xilinx/xilinx_axienet.h | 2 +
.../net/ethernet/xilinx/xilinx_axienet_main.c | 44 +++++++++++++++++--
2 files changed, 42 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet.h b/drivers/net/ethernet/xilinx/xilinx_axienet.h
index a87cfa079502..140383749996 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet.h
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet.h
@@ -526,6 +526,7 @@ struct skbuf_dma_descriptor {
* @stopping: Set when we are about to stop the device: makes @dma_err_task
* a no-op (legacy DMA path) and fences RX descriptor
* resubmission in axienet_dma_rx_cb() (dmaengine path).
+ * @datapath_up: Set while the resources allocated by ndo_open are live.
* @tx_irq: Axidma TX IRQ number
* @rx_irq: Axidma RX IRQ number
* @eth_irq: Ethernet core IRQ number
@@ -608,6 +609,7 @@ struct axienet_local {
struct work_struct dma_err_task;
bool stopping;
+ bool datapath_up;
int tx_irq;
int rx_irq;
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
index 83f35effbf37..f338029cf870 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
@@ -702,6 +702,17 @@ static int __axienet_device_reset(struct axienet_local *lp)
}
out:
+ /* On failure the block above is skipped, so clear the flag here.
+ * axienet_open() now returns on a reset error without scheduling the
+ * statistics work that used to clear it, which would otherwise leave
+ * axienet_stat() returning the frozen base until a reset succeeds.
+ */
+ if (ret && (lp->features & XAE_FEATURE_STATS)) {
+ write_seqcount_begin(&lp->hw_stats_seqcount);
+ lp->reset_in_progress = false;
+ write_seqcount_end(&lp->hw_stats_seqcount);
+ }
+
mutex_unlock(&lp->stats_lock);
return ret;
}
@@ -1772,6 +1783,10 @@ static int axienet_open(struct net_device *ndev)
axienet_lock_mii(lp);
ret = axienet_device_reset(ndev);
axienet_unlock_mii(lp);
+ if (ret) {
+ dev_err(lp->dev, "axienet_device_reset() failed: %d\n", ret);
+ return ret;
+ }
ret = phylink_fwnode_phy_connect(lp->phylink, dev_fwnode(lp->dev), 0);
if (ret) {
@@ -1802,6 +1817,8 @@ static int axienet_open(struct net_device *ndev)
goto err_phy;
}
+ lp->datapath_up = true;
+
return 0;
err_free_eth_irq:
@@ -1830,6 +1847,16 @@ static int axienet_stop(struct net_device *ndev)
struct axienet_local *lp = netdev_priv(ndev);
int i;
+ /* axienet_resume() calls ndo_open directly, so unlike the netdev core
+ * it can leave the device running after a failed open. Tearing down
+ * below would then stop a phylink that open already stopped and
+ * disconnected, and free interrupts it never requested.
+ */
+ if (!lp->datapath_up)
+ return 0;
+
+ lp->datapath_up = false;
+
if (!lp->use_dmaengine) {
WRITE_ONCE(lp->stopping, true);
flush_work(&lp->dma_err_task);
@@ -3415,17 +3442,26 @@ static int axienet_suspend(struct device *dev)
static int axienet_resume(struct device *dev)
{
struct net_device *ndev = dev_get_drvdata(dev);
+ int ret;
if (!netif_running(ndev))
return 0;
rtnl_lock();
- axienet_open(ndev);
- rtnl_unlock();
-
+ ret = axienet_open(ndev);
+ if (ret) {
+ netdev_err(ndev, "failed to reopen device on resume: %d\n", ret);
+ /* Drop IFF_UP before reattaching, so that the interface can be
+ * brought up again once whatever made open fail is resolved.
+ * Reattaching while still up would also let the stack transmit
+ * on the datapath open has already torn down.
+ */
+ netif_close(ndev);
+ }
netif_device_attach(ndev);
+ rtnl_unlock();
- return 0;
+ return ret;
}
static DEFINE_SIMPLE_DEV_PM_OPS(axienet_pm_ops,
--
2.25.1
^ permalink raw reply related [flat|nested] 12+ messages in thread
* [PATCH net-next v5 4/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support
2026-11-14 18:29 [PATCH net-next v5 0/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support Suraj Gupta
` (2 preceding siblings ...)
2026-11-14 18:29 ` [PATCH net-next v5 3/6] net: xilinx: axienet: Propagate device reset failure from ndo_open Suraj Gupta
@ 2026-11-14 18:29 ` Suraj Gupta
2026-11-16 4:12 ` netdev-bot+sashiko
2026-11-14 18:29 ` [PATCH net-next v5 5/6] net: xilinx: axienet: Dispatch statistics through axienet_config ops Suraj Gupta
2026-11-14 18:29 ` [PATCH net-next v5 6/6] net: xilinx: axienet: Add statistics support for XXV ethernet Suraj Gupta
5 siblings, 1 reply; 12+ messages in thread
From: Suraj Gupta @ 2026-11-14 18:29 UTC (permalink / raw)
To: Radhey Shyam Pandey, Andrew Lunn, David S . Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley
Cc: Michal Simek, Russell King, netdev, devicetree, linux-arm-kernel,
linux-kernel
The AXI Ethernet driver currently supports only the AXI 1G Ethernet IP.
Add support for the Xilinx XXV Ethernet IP[1], which implements a 10G/25G
Ethernet MAC with PCS/PMA driving a high-speed serial transceiver, as
specified by the 25G Ethernet Consortium.
Add the XXV-specific logic in new xilinx_axienet_xxv.c/.h files and
describe the MAC through an axienet_10g25g_config that provides the
capability flags, the MAC option table and callbacks (GT reset,
link-ready polling, phylink caps, PCS ops, register dump and probe-time
init). The shared paths consult those capabilities and callbacks instead
of MAC-type checks; the only interface-specific addition is handling the
10GBASER and 25GBASER modes in axienet_mac_select_pcs().
The register, RX and GT DRP clocks are required. The GT reference clock
is requested as optional, because it is usually a fixed oscillator that
Linux does not model; where a board does describe it as a gated provider
it has to be enabled, as the GT otherwise never reports reset-done.
The advertised rate comes from the required phy-mode property rather than
from the STAT_CORE_SPEED register, which PG210 provides only on the 64-bit
AXI4-Stream builds; a node declaring any other phy-mode is rejected at
probe. Cores built with runtime rate switching are not reconfigured to the
other rate, and flow control is not wired up, so pause support is not
advertised either.
The XXV mac_init() can fail on a GT reset-done timeout, which is why
ndo_open has to propagate a device reset failure as of the preceding
patch.
[1]: https://docs.amd.com/r/en-US/pg210-25g-ethernet
Signed-off-by: Suraj Gupta <suraj.gupta2@amd.com>
---
drivers/net/ethernet/xilinx/Makefile | 3 +-
drivers/net/ethernet/xilinx/xilinx_axienet.h | 9 +
.../net/ethernet/xilinx/xilinx_axienet_main.c | 23 +-
.../net/ethernet/xilinx/xilinx_axienet_xxv.c | 351 ++++++++++++++++++
.../net/ethernet/xilinx/xilinx_axienet_xxv.h | 93 +++++
5 files changed, 470 insertions(+), 9 deletions(-)
create mode 100644 drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
create mode 100644 drivers/net/ethernet/xilinx/xilinx_axienet_xxv.h
diff --git a/drivers/net/ethernet/xilinx/Makefile b/drivers/net/ethernet/xilinx/Makefile
index 7d7dc1771423..db90e691fec5 100644
--- a/drivers/net/ethernet/xilinx/Makefile
+++ b/drivers/net/ethernet/xilinx/Makefile
@@ -6,5 +6,6 @@
ll_temac-objs := ll_temac_main.o ll_temac_mdio.o
obj-$(CONFIG_XILINX_LL_TEMAC) += ll_temac.o
obj-$(CONFIG_XILINX_EMACLITE) += xilinx_emaclite.o
-xilinx_emac-objs := xilinx_axienet_main.o xilinx_axienet_mdio.o
+xilinx_emac-objs := xilinx_axienet_main.o xilinx_axienet_mdio.o \
+ xilinx_axienet_xxv.o
obj-$(CONFIG_XILINX_AXI_EMAC) += xilinx_emac.o
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet.h b/drivers/net/ethernet/xilinx/xilinx_axienet.h
index 140383749996..56570b49574e 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet.h
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet.h
@@ -17,6 +17,8 @@
#include <linux/phylink.h>
#include <linux/skbuff.h>
+#include "xilinx_axienet_xxv.h"
+
/* Packet size info */
#define XAE_HDR_SIZE 14 /* Size of Ethernet header */
#define XAE_TRL_SIZE 4 /* Size of Ethernet trailer (FCS) */
@@ -548,6 +550,7 @@ struct skbuf_dma_descriptor {
* @rx_ring_head: RX skb ring buffer head index.
* @rx_ring_tail: RX skb ring buffer tail index.
* @rx_submit_lock: Protects RX ring resubmission vs teardown in dmaengine path.
+ * @xxv_ip_version: XXV IP version.
* @axienet_config: MAC-type specific configuration and operations.
*/
struct axienet_local {
@@ -632,6 +635,7 @@ struct axienet_local {
int rx_ring_head;
int rx_ring_tail;
spinlock_t rx_submit_lock;
+ u32 xxv_ip_version;
const struct axienet_config *axienet_config;
};
@@ -689,6 +693,11 @@ struct axienet_config {
void (*stats_update)(struct axienet_local *lp);
};
+static inline struct axienet_local *pcs_to_axienet_local(struct phylink_pcs *pcs)
+{
+ return container_of(pcs, struct axienet_local, pcs);
+}
+
/**
* struct axienet_option - Used to set axi ethernet hardware options
* @opt: Option to be set.
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
index f338029cf870..a37aad4432ce 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
@@ -800,6 +800,12 @@ static int axienet_1g_mac_init(struct net_device *ndev)
* Ethernet core. No separate hardware reset is done for the Axi Ethernet
* core.
*
+ * It also recomputes @max_frm_size from the current MTU and runs the
+ * MAC-specific mac_init() callback, so it is the point at which length-related
+ * MAC configuration (such as the XXV RX max-packet-length) is reprogrammed.
+ * Because it is reached from the open path, a new MTU takes effect on the next
+ * interface up.
+ *
* Return: 0 on success or a negative error number otherwise.
*/
static int axienet_device_reset(struct net_device *ndev)
@@ -2720,11 +2726,6 @@ static const struct ethtool_ops axienet_ethtool_dmaengine_ops = {
.get_rmon_stats = axienet_ethtool_get_rmon_stats,
};
-static struct axienet_local *pcs_to_axienet_local(struct phylink_pcs *pcs)
-{
- return container_of(pcs, struct axienet_local, pcs);
-}
-
static void axienet_pcs_get_state(struct phylink_pcs *pcs,
unsigned int neg_mode,
struct phylink_link_state *state)
@@ -2784,7 +2785,9 @@ static struct phylink_pcs *axienet_mac_select_pcs(struct phylink_config *config,
struct axienet_local *lp = netdev_priv(ndev);
if (interface == PHY_INTERFACE_MODE_1000BASEX ||
- interface == PHY_INTERFACE_MODE_SGMII)
+ interface == PHY_INTERFACE_MODE_SGMII ||
+ interface == PHY_INTERFACE_MODE_10GBASER ||
+ interface == PHY_INTERFACE_MODE_25GBASER)
return &lp->pcs;
return NULL;
@@ -3014,6 +3017,7 @@ static const struct of_device_id axienet_of_match[] = {
{ .compatible = "xlnx,axi-ethernet-1.00.a", .data = &axienet_1g_config },
{ .compatible = "xlnx,axi-ethernet-1.01.a", .data = &axienet_1g_config },
{ .compatible = "xlnx,axi-ethernet-2.01.a", .data = &axienet_1g_config },
+ { .compatible = "xlnx,xxv-ethernet-1.0", .data = &axienet_10g25g_config },
{},
};
@@ -3051,9 +3055,10 @@ static int axienet_probe(struct platform_device *pdev)
SET_NETDEV_DEV(ndev, &pdev->dev);
ndev->features = NETIF_F_SG;
- /* MTU range: 64 - 9000 */
+ /* MTU range: min is 64; max is finalized once the MAC config (which
+ * tells us whether jumbo frames are supported) is known, see below.
+ */
ndev->min_mtu = 64;
- ndev->max_mtu = XAE_JUMBO_MTU;
lp = netdev_priv(ndev);
lp->ndev = ndev;
@@ -3073,6 +3078,8 @@ static int axienet_probe(struct platform_device *pdev)
if (!lp->axienet_config)
return -ENODEV;
+ ndev->max_mtu = lp->axienet_config->jumbo ? XAE_JUMBO_MTU : XAE_MTU;
+
ret = lp->axienet_config->clk_init(lp);
if (ret)
return ret;
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
new file mode 100644
index 000000000000..3c204dbbf6ae
--- /dev/null
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
@@ -0,0 +1,351 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * XXV (10G/25G) support
+ *
+ * Copyright (C) 2026 Advanced Micro Devices, Inc.
+ */
+
+#include <linux/bitfield.h>
+#include <linux/clk.h>
+#include <linux/delay.h>
+#include <linux/iopoll.h>
+#include <linux/netdevice.h>
+#include <linux/phylink.h>
+
+#include "xilinx_axienet.h"
+#include "xilinx_axienet_xxv.h"
+
+/* readl_poll_timeout() sleep interval (us) for the GT reset-done and PCS
+ * block-lock polls, which both run to a DELAY_OF_ONE_MILLISEC timeout. The
+ * interval and that bound come from ZynqMP/Versal bring-up; PG210 does not
+ * specify either.
+ */
+#define XXV_LINK_POLL_INTERVAL_US 10
+
+/* Number of XXV (10G/25G) MAC registers exposed via ethtool -d.
+ * Covers configuration and status registers per Xilinx PG210.
+ */
+static const u32 axienet_xxv_reg_dump_offsets[XXV_REGS_N] = {
+ XXV_GT_RESET_OFFSET,
+ XXV_RESET_OFFSET,
+ XXV_MODE_OFFSET,
+ XXV_TC_OFFSET,
+ XXV_RCW1_OFFSET,
+ XXV_JUM_OFFSET,
+ XXV_VL_LENGTH_OFFSET,
+ XXV_TICKREG_OFFSET,
+ XXV_CONFIG_REVISION_OFFSET,
+ XXV_CONFIG_1588_OFFSET,
+ XXV_CONFIG_TX_FLOW_CTRL1_OFFSET,
+ XXV_CONFIG_RX_FLOW_CTRL1_OFFSET,
+ XXV_CONFIG_RX_FLOW_CTRL2_OFFSET,
+ XXV_CONFIG_RSFEC_OFFSET,
+ XXV_CONFIG_FEC_OFFSET,
+ XXV_AN_CTL1_OFFSET,
+ XXV_AN_CTL2_OFFSET,
+ XXV_AN_ABILITY_OFFSET,
+ XXV_LT_CTL1_OFFSET,
+ XXV_SWITCH_CORE_SPEED_OFFSET,
+ XXV_CONFIG_1588_32BIT_OFFSET,
+ XXV_TX_CONFIG_1588_OFFSET,
+ XXV_RX_CONFIG_1588_OFFSET,
+ XXV_GT_CTRL_OFFSET,
+ XXV_STAT_TX_STATUS_OFFSET,
+ XXV_STAT_RX_STATUS_OFFSET,
+ XXV_STAT_STATUS_OFFSET,
+ XXV_STAT_RX_BLOCK_LOCK_OFFSET,
+ XXV_STAT_RX_RSFEC_STATUS_OFFSET,
+ XXV_STAT_RX_FEC_STATUS_OFFSET,
+ XXV_STAT_TX_RSFEC_STATUS_OFFSET,
+ XXV_STAT_TX_FLOW_CTRL1_OFFSET,
+ XXV_STAT_RX_FLOW_CTRL1_OFFSET,
+ XXV_STAT_AN_STS_OFFSET,
+ XXV_STAT_AN_ABILITY_OFFSET,
+ XXV_STAT_AN_LINK_CTL_OFFSET,
+ XXV_STAT_LT_STATUS1_OFFSET,
+ XXV_STAT_LT_STATUS2_OFFSET,
+ XXV_STAT_LT_STATUS3_OFFSET,
+ XXV_STAT_LT_STATUS4_OFFSET,
+ XXV_STAT_LT_COEFF0_OFFSET,
+ XXV_STAT_RX_VALID_CTRL_CODE_OFFSET,
+ XXV_STAT_CORE_SPEED_OFFSET,
+ XXV_STAT_TSN_OFFSET,
+ XXV_STAT_GT_OFFSET,
+ XXV_STAT_AN_LINK_CTL2_OFFSET,
+};
+
+/* Option table for setting up XXV Ethernet hardware options */
+static const struct axienet_option xxvenet_options[] = {
+ { /* Turn on FCS stripping on receive packets */
+ .opt = XAE_OPTION_FCS_STRIP,
+ .reg = XXV_RCW1_OFFSET,
+ .m_or = XXV_RCW1_FCS_MASK,
+ }, { /* Turn on FCS insertion on transmit packets */
+ .opt = XAE_OPTION_FCS_INSERT,
+ .reg = XXV_TC_OFFSET,
+ .m_or = XXV_TC_FCS_MASK,
+ }, { /* Enable transmitter */
+ .opt = XAE_OPTION_TXEN,
+ .reg = XXV_TC_OFFSET,
+ .m_or = XXV_TC_TX_MASK,
+ }, { /* Enable receiver */
+ .opt = XAE_OPTION_RXEN,
+ .reg = XXV_RCW1_OFFSET,
+ .m_or = XXV_RCW1_RX_MASK,
+ },
+ {}
+};
+
+static bool axienet_xxv_ip_has_gt_status(u32 ip_version)
+{
+ u8 minor = FIELD_GET(XXV_MIN_MASK, ip_version);
+ u8 maj = FIELD_GET(XXV_MAJ_MASK, ip_version);
+
+ if (maj == XXV_IP_VER_GT_MAJ_MIN)
+ return minor >= XXV_IP_VER_GT_MIN_MIN;
+ return maj > XXV_IP_VER_GT_MAJ_MIN;
+}
+
+/**
+ * axienet_xxv_probe_init - Initialize XXV-specific state at probe time
+ * @lp: Pointer to the axienet_local structure
+ */
+static void axienet_xxv_probe_init(struct axienet_local *lp)
+{
+ lp->xxv_ip_version = axienet_ior(lp, XXV_CONFIG_REVISION_OFFSET);
+}
+
+/**
+ * axienet_xxv_gt_reset - Pulse the XXV GT reset line
+ * @lp: Pointer to the axienet_local structure
+ */
+static void axienet_xxv_gt_reset(struct axienet_local *lp)
+{
+ u32 val;
+
+ val = axienet_ior(lp, XXV_GT_RESET_OFFSET);
+ val |= XXV_GT_RESET_MASK;
+ axienet_iow(lp, XXV_GT_RESET_OFFSET, val);
+ /* Give the GT time to enter reset so that axienet_xxv_poll_link_ready()
+ * cannot sample a reset-done left over from before the pulse. The delay
+ * is not a documented hold time (see timeout note above).
+ */
+ usleep_range(1000, 2000);
+ val = axienet_ior(lp, XXV_GT_RESET_OFFSET);
+ val &= ~XXV_GT_RESET_MASK;
+ axienet_iow(lp, XXV_GT_RESET_OFFSET, val);
+}
+
+/**
+ * axienet_xxv_poll_link_ready - Wait for XXV GT and PCS block lock
+ * @ndev: Pointer to the net_device structure
+ *
+ * Poll GT reset-done on IP v3.2+, then RX PCS block lock. GT reset-done
+ * depends only on the GT reference clock, so its timeout is treated as fatal.
+ * Block-lock additionally needs a live link partner, so its failure is logged
+ * but not propagated: an unplugged cable must not fail device bring-up.
+ *
+ * Return: 0 on success, or a negative error if GT reset-done times out.
+ */
+static int axienet_xxv_poll_link_ready(struct net_device *ndev)
+{
+ struct axienet_local *lp = netdev_priv(ndev);
+ u32 val;
+ int ret;
+
+ /* Confirm XXV Ethernet is up: the GT status register exists only
+ * on IP v3.2+, so wait there for GT reset-done, then poll until
+ * RX PCS block lock is asserted.
+ */
+ if (axienet_xxv_ip_has_gt_status(lp->xxv_ip_version)) {
+ ret = readl_poll_timeout(lp->regs + XXV_STAT_GT_OFFSET,
+ val,
+ (val & XXV_GT_RESET_DONE) == XXV_GT_RESET_DONE,
+ XXV_LINK_POLL_INTERVAL_US,
+ DELAY_OF_ONE_MILLISEC);
+ if (ret) {
+ netdev_err(ndev, "XXV MAC GT reset not complete! Cross-check the MAC ref clock configuration\n");
+ return ret;
+ }
+ }
+
+ ret = readl_poll_timeout(lp->regs + XXV_STAT_RX_BLOCK_LOCK_OFFSET,
+ val, (val & XXV_STAT_RX_BLOCK_LOCK_MASK),
+ XXV_LINK_POLL_INTERVAL_US,
+ DELAY_OF_ONE_MILLISEC);
+ if (ret)
+ netdev_dbg(ndev, "XXV RX block lock not yet asserted (no link partner?)\n");
+
+ return 0;
+}
+
+/**
+ * axienet_xxv_mac_init - XXV MAC-specific bring-up after the DMA reset
+ * @ndev: Pointer to the net_device structure
+ *
+ * Return: 0 on success or a negative error number otherwise.
+ */
+static int axienet_xxv_mac_init(struct net_device *ndev)
+{
+ struct axienet_local *lp = netdev_priv(ndev);
+ int ret;
+
+ ret = axienet_xxv_poll_link_ready(ndev);
+ if (ret)
+ return ret;
+
+ /* Bound the MAC's RX max packet length to the posted buffer size so an
+ * oversized frame is truncated to a single buffer instead of spilling
+ * across several DMA descriptors (PG210 CONFIGURATION_RX_MTU).
+ */
+ axienet_iow(lp, XXV_JUM_OFFSET,
+ FIELD_PREP(XXV_RX_MAX_LEN_MASK, lp->max_frm_size) |
+ XXV_RX_MIN_LEN_DEFAULT);
+
+ lp->options |= XAE_OPTION_FCS_STRIP;
+ lp->options |= XAE_OPTION_FCS_INSERT;
+
+ return 0;
+}
+
+static void axienet_xxv_phylink_set_capabilities(struct axienet_local *lp,
+ struct phylink_config *cfg)
+{
+ switch (lp->phy_mode) {
+ case PHY_INTERFACE_MODE_10GBASER:
+ cfg->mac_capabilities |= MAC_10000FD;
+ break;
+ case PHY_INTERFACE_MODE_25GBASER:
+ cfg->mac_capabilities |= MAC_25000FD;
+ break;
+ default:
+ /* The binding allows only the two rates above. Leave
+ * supported_interfaces empty rather than guessing a rate;
+ * phylink_create() then rejects the devicetree and probe
+ * fails.
+ */
+ dev_err(lp->dev, "unsupported phy-mode %s for XXV ethernet\n",
+ phy_modes(lp->phy_mode));
+ return;
+ }
+
+ __set_bit(lp->phy_mode, cfg->supported_interfaces);
+
+ /* The integrated PCS/PMA manages the link in-band and there is no
+ * phy-handle, so resolve to MLO_AN_INBAND.
+ */
+ cfg->default_an_inband = true;
+
+ /* Flow control is not wired up for XXV yet; do not advertise pause. */
+ cfg->mac_capabilities &= ~(MAC_SYM_PAUSE | MAC_ASYM_PAUSE);
+}
+
+static void axienet_xxv_pcs_get_state(struct phylink_pcs *pcs,
+ unsigned int neg_mode,
+ struct phylink_link_state *state)
+{
+ struct axienet_local *lp = pcs_to_axienet_local(pcs);
+
+ state->duplex = DUPLEX_FULL;
+ state->an_complete = !!(axienet_ior(lp, XXV_STAT_AN_STS_OFFSET) &
+ XXV_AN_COMPLETE_MASK);
+ state->link = axienet_ior(lp, XXV_STAT_RX_STATUS_OFFSET) &
+ XXV_STAT_RX_STATUS_MASK;
+
+ /* The core runs at the single rate phylink was configured for, so
+ * derive the speed from the interface rather than from STAT_CORE_SPEED,
+ * which PG210 provides only on the 64-bit AXI4-Stream builds.
+ */
+ if (state->interface == PHY_INTERFACE_MODE_10GBASER)
+ state->speed = SPEED_10000;
+ else
+ state->speed = SPEED_25000;
+}
+
+static int axienet_xxv_pcs_config(struct phylink_pcs *pcs, unsigned int neg_mode,
+ phy_interface_t interface,
+ const unsigned long *advertising,
+ bool permit_pause_to_mac)
+{
+ return 0;
+}
+
+static const struct phylink_pcs_ops axienet_xxv_pcs_ops = {
+ .pcs_get_state = axienet_xxv_pcs_get_state,
+ .pcs_config = axienet_xxv_pcs_config,
+};
+
+/**
+ * axienet_xxv_get_regs - Dump XXV MAC registers for ethtool
+ * @lp: Pointer to the axienet_local structure
+ * @data: Buffer for register values (zeroed and sized by the caller)
+ */
+static void axienet_xxv_get_regs(struct axienet_local *lp, u32 *data)
+{
+ bool has_gt_status = axienet_xxv_ip_has_gt_status(lp->xxv_ip_version);
+ unsigned int i;
+
+ for (i = 0; i < ARRAY_SIZE(axienet_xxv_reg_dump_offsets); i++) {
+ u32 off = axienet_xxv_reg_dump_offsets[i];
+
+ /* The GT registers only exist on the IP versions that report
+ * reset-done. Leave their slots zero on the others so the
+ * dump keeps a fixed layout.
+ */
+ if (!has_gt_status && (off == XXV_GT_CTRL_OFFSET ||
+ off == XXV_STAT_GT_OFFSET))
+ continue;
+
+ data[i] = axienet_ior(lp, off);
+ }
+}
+
+static int axienet_10g25g_clk_init(struct axienet_local *lp)
+{
+ struct device *dev = lp->dev;
+ struct clk *ref_clk;
+ int ret;
+
+ /* The register, RX and GT DRP clocks are mandatory for MAC+PCS
+ * operation, so fetch them as required.
+ */
+ lp->axi_clk = devm_clk_get_enabled(dev, "s_axi");
+ if (IS_ERR(lp->axi_clk))
+ return dev_err_probe(dev, PTR_ERR(lp->axi_clk),
+ "could not get AXI clock\n");
+
+ lp->misc_clks[0].id = "rx_core";
+ lp->misc_clks[1].id = "drp";
+
+ ret = devm_clk_bulk_get_enable(dev, XXV_NUM_MISC_CLOCKS,
+ lp->misc_clks);
+ if (ret)
+ return dev_err_probe(dev, ret,
+ "could not get/enable XXV clocks\n");
+
+ /* The GT reference clock is usually a fixed oscillator that Linux does
+ * not model, but enable it for boards that describe it as a gated
+ * provider. Without it the GT never reports reset-done.
+ */
+ ref_clk = devm_clk_get_optional_enabled(dev, "ref");
+ if (IS_ERR(ref_clk))
+ return dev_err_probe(dev, PTR_ERR(ref_clk),
+ "could not get/enable GT reference clock\n");
+
+ return 0;
+}
+
+const struct axienet_config axienet_10g25g_config = {
+ .jumbo = true,
+ .sw_padding = true,
+ .internal_pcs = true,
+ .regs_n = XXV_REGS_N,
+ .options = xxvenet_options,
+ .clk_init = axienet_10g25g_clk_init,
+ .probe_init = axienet_xxv_probe_init,
+ .gt_reset = axienet_xxv_gt_reset,
+ .mac_init = axienet_xxv_mac_init,
+ .get_regs = axienet_xxv_get_regs,
+ .phylink_set_caps = axienet_xxv_phylink_set_capabilities,
+ .pcs_ops = &axienet_xxv_pcs_ops,
+};
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.h b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.h
new file mode 100644
index 000000000000..6530344e40c7
--- /dev/null
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.h
@@ -0,0 +1,93 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * XXV (10G/25G) support
+ *
+ * Copyright (C) 2026 Advanced Micro Devices, Inc.
+ */
+
+#ifndef XILINX_AXIENET_XXV_H
+#define XILINX_AXIENET_XXV_H
+
+#include <linux/bits.h>
+
+struct axienet_local;
+struct axienet_config;
+
+#define XXV_REGS_N 46
+
+/* RX core and GT DRP clocks, fetched through lp->misc_clks */
+#define XXV_NUM_MISC_CLOCKS 2
+
+/* XXV MAC Register Definitions */
+#define XXV_GT_RESET_OFFSET 0x00000000
+#define XXV_RESET_OFFSET 0x00000004
+#define XXV_MODE_OFFSET 0x00000008
+#define XXV_TC_OFFSET 0x0000000C
+#define XXV_RCW1_OFFSET 0x00000014
+#define XXV_JUM_OFFSET 0x00000018
+#define XXV_VL_LENGTH_OFFSET 0x0000001C
+#define XXV_TICKREG_OFFSET 0x00000020
+#define XXV_CONFIG_REVISION_OFFSET 0x00000024
+#define XXV_CONFIG_1588_OFFSET 0x00000038
+#define XXV_CONFIG_TX_FLOW_CTRL1_OFFSET 0x00000040
+#define XXV_CONFIG_RX_FLOW_CTRL1_OFFSET 0x00000094
+#define XXV_CONFIG_RX_FLOW_CTRL2_OFFSET 0x00000098
+#define XXV_CONFIG_RSFEC_OFFSET 0x000000D0
+#define XXV_CONFIG_FEC_OFFSET 0x000000D4
+#define XXV_AN_CTL1_OFFSET 0x000000E0
+#define XXV_AN_CTL2_OFFSET 0x000000E4
+#define XXV_AN_ABILITY_OFFSET 0x000000F8
+#define XXV_LT_CTL1_OFFSET 0x00000100
+#define XXV_SWITCH_CORE_SPEED_OFFSET 0x00000138
+#define XXV_CONFIG_1588_32BIT_OFFSET 0x0000013C
+#define XXV_TX_CONFIG_1588_OFFSET 0x00000140
+#define XXV_RX_CONFIG_1588_OFFSET 0x00000144
+#define XXV_GT_CTRL_OFFSET 0x00000154
+#define XXV_STAT_TX_STATUS_OFFSET 0x00000400
+#define XXV_STAT_RX_STATUS_OFFSET 0x00000404
+#define XXV_STAT_STATUS_OFFSET 0x00000408
+#define XXV_STAT_RX_BLOCK_LOCK_OFFSET 0x0000040C
+#define XXV_STAT_RX_RSFEC_STATUS_OFFSET 0x0000043C
+#define XXV_STAT_RX_FEC_STATUS_OFFSET 0x00000448
+#define XXV_STAT_TX_RSFEC_STATUS_OFFSET 0x0000044C
+#define XXV_STAT_TX_FLOW_CTRL1_OFFSET 0x00000450
+#define XXV_STAT_RX_FLOW_CTRL1_OFFSET 0x00000454
+#define XXV_STAT_AN_STS_OFFSET 0x00000458
+#define XXV_STAT_AN_ABILITY_OFFSET 0x0000045C
+#define XXV_STAT_AN_LINK_CTL_OFFSET 0x00000460
+#define XXV_STAT_LT_STATUS1_OFFSET 0x00000464
+#define XXV_STAT_LT_STATUS2_OFFSET 0x00000468
+#define XXV_STAT_LT_STATUS3_OFFSET 0x0000046C
+#define XXV_STAT_LT_STATUS4_OFFSET 0x00000470
+#define XXV_STAT_LT_COEFF0_OFFSET 0x00000474
+#define XXV_STAT_RX_VALID_CTRL_CODE_OFFSET 0x00000494
+#define XXV_STAT_CORE_SPEED_OFFSET 0x00000498
+#define XXV_STAT_TSN_OFFSET 0x0000049C
+#define XXV_STAT_GT_OFFSET 0x000004A0
+#define XXV_STAT_AN_LINK_CTL2_OFFSET 0x000009F0
+
+/* XXV MAC Register Mask Definitions */
+#define XXV_GT_RESET_MASK BIT(0)
+#define XXV_TC_TX_MASK BIT(0)
+#define XXV_RCW1_RX_MASK BIT(0)
+#define XXV_RCW1_FCS_MASK BIT(1)
+#define XXV_TC_FCS_MASK BIT(1)
+#define XXV_STAT_RX_BLOCK_LOCK_MASK BIT(0)
+#define XXV_GT_RESET_DONE (BIT(0) | BIT(1))
+#define XXV_MAJ_MASK GENMASK(7, 0)
+#define XXV_MIN_MASK GENMASK(15, 8)
+/* The GT control and status registers exist from IP v3.2 on (PG210). */
+#define XXV_IP_VER_GT_MAJ_MIN 3
+#define XXV_IP_VER_GT_MIN_MIN 2
+#define XXV_AN_COMPLETE_MASK BIT(2)
+#define XXV_STAT_RX_STATUS_MASK BIT(0)
+
+/* CONFIGURATION_RX_MTU (XXV_JUM_OFFSET) field layout (PG210): bits [30:16]
+ * hold ctl_rx_max_packet_len, bits [7:0] hold ctl_rx_min_packet_len.
+ */
+#define XXV_RX_MAX_LEN_MASK GENMASK(30, 16)
+#define XXV_RX_MIN_LEN_DEFAULT 64
+
+extern const struct axienet_config axienet_10g25g_config;
+
+#endif /* XILINX_AXIENET_XXV_H */
--
2.25.1
^ permalink raw reply related [flat|nested] 12+ messages in thread
* [PATCH net-next v5 5/6] net: xilinx: axienet: Dispatch statistics through axienet_config ops
2026-11-14 18:29 [PATCH net-next v5 0/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support Suraj Gupta
` (3 preceding siblings ...)
2026-11-14 18:29 ` [PATCH net-next v5 4/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support Suraj Gupta
@ 2026-11-14 18:29 ` Suraj Gupta
2026-11-14 18:29 ` [PATCH net-next v5 6/6] net: xilinx: axienet: Add statistics support for XXV ethernet Suraj Gupta
5 siblings, 0 replies; 12+ messages in thread
From: Suraj Gupta @ 2026-11-14 18:29 UTC (permalink / raw)
To: Radhey Shyam Pandey, Andrew Lunn, David S . Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley
Cc: Michal Simek, Russell King, netdev, devicetree, linux-arm-kernel,
linux-kernel
The ndo_get_stats64() and ethtool statistics callbacks read the 1G MAC
hardware counters directly. Supporting a second MAC type this way would
require an "if (mactype == ...)" dispatch branch in each of the eight
statistics callbacks.
Add statistics operation function pointers to struct axienet_config and
route get_stats64, get_ethtool_stats, get_strings, get_sset_count,
get_pause_stats, get_eth_mac_stats, get_eth_ctrl_stats and
get_rmon_stats through them. Factor the existing 1G counter reads into
axienet_1g_*() helpers and populate axienet_1g_config with them.
A NULL op means the MAC does not expose that statistics group, so a MAC
without control-frame counters simply leaves get_eth_ctrl_stats unset
instead of open-coding the special case in the callback.
No functional change intended.
Signed-off-by: Suraj Gupta <suraj.gupta2@amd.com>
---
drivers/net/ethernet/xilinx/xilinx_axienet.h | 23 +++
.../net/ethernet/xilinx/xilinx_axienet_main.c | 171 +++++++++++++-----
2 files changed, 146 insertions(+), 48 deletions(-)
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet.h b/drivers/net/ethernet/xilinx/xilinx_axienet.h
index 56570b49574e..7468fed1416e 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet.h
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet.h
@@ -666,6 +666,15 @@ struct axienet_local {
* is set, as the PCS is registered without a further NULL check
* @stats_update: Callback to latch/accumulate the periodic MAC counters, or
* NULL if the MAC has no periodic statistics counters
+ * @get_stats64: Callback to read MAC counters into rtnl_link_stats64
+ * @get_ethtool_stats: Callback to read MAC counters for ethtool -S
+ * @get_strings: Callback to fill the ethtool -S statistic name strings
+ * @get_sset_count: Callback returning the number of ethtool -S statistics
+ * @get_pause_stats: Callback to read MAC pause-frame counters
+ * @get_eth_mac_stats: Callback to read IEEE 802.3 MAC counters
+ * @get_eth_ctrl_stats: Callback to read MAC-control frame counters, or NULL
+ * if the MAC exposes no control-frame counters
+ * @get_rmon_stats: Callback to read RMON counters and histogram ranges
*/
struct axienet_config {
bool mdio;
@@ -691,6 +700,20 @@ struct axienet_config {
struct phylink_config *config);
const struct phylink_pcs_ops *pcs_ops;
void (*stats_update)(struct axienet_local *lp);
+ void (*get_stats64)(struct axienet_local *lp,
+ struct rtnl_link_stats64 *stats);
+ void (*get_ethtool_stats)(struct axienet_local *lp, u64 *data);
+ void (*get_strings)(u8 *data);
+ int (*get_sset_count)(void);
+ void (*get_pause_stats)(struct axienet_local *lp,
+ struct ethtool_pause_stats *pause_stats);
+ void (*get_eth_mac_stats)(struct axienet_local *lp,
+ struct ethtool_eth_mac_stats *mac_stats);
+ void (*get_eth_ctrl_stats)(struct axienet_local *lp,
+ struct ethtool_eth_ctrl_stats *ctrl_stats);
+ void (*get_rmon_stats)(struct axienet_local *lp,
+ struct ethtool_rmon_stats *rmon_stats,
+ const struct ethtool_rmon_hist_range **ranges);
};
static inline struct axienet_local *pcs_to_axienet_local(struct phylink_pcs *pcs)
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
index a37aad4432ce..329b6adf2475 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
@@ -1992,29 +1992,11 @@ static int axienet_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
return phylink_mii_ioctl(lp->phylink, rq, cmd);
}
-static void
-axienet_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
+static void axienet_1g_get_stats64(struct axienet_local *lp,
+ struct rtnl_link_stats64 *stats)
{
- struct axienet_local *lp = netdev_priv(dev);
unsigned int start;
- netdev_stats_to_stats64(stats, &dev->stats);
-
- do {
- start = u64_stats_fetch_begin(&lp->rx_stat_sync);
- stats->rx_packets = u64_stats_read(&lp->rx_packets);
- stats->rx_bytes = u64_stats_read(&lp->rx_bytes);
- } while (u64_stats_fetch_retry(&lp->rx_stat_sync, start));
-
- do {
- start = u64_stats_fetch_begin(&lp->tx_stat_sync);
- stats->tx_packets = u64_stats_read(&lp->tx_packets);
- stats->tx_bytes = u64_stats_read(&lp->tx_bytes);
- } while (u64_stats_fetch_retry(&lp->tx_stat_sync, start));
-
- if (!(lp->features & XAE_FEATURE_STATS))
- return;
-
do {
start = read_seqcount_begin(&lp->hw_stats_seqcount);
stats->rx_length_errors =
@@ -2042,6 +2024,33 @@ axienet_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
} while (read_seqcount_retry(&lp->hw_stats_seqcount, start));
}
+static void
+axienet_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
+{
+ struct axienet_local *lp = netdev_priv(dev);
+ unsigned int start;
+
+ netdev_stats_to_stats64(stats, &dev->stats);
+
+ do {
+ start = u64_stats_fetch_begin(&lp->rx_stat_sync);
+ stats->rx_packets = u64_stats_read(&lp->rx_packets);
+ stats->rx_bytes = u64_stats_read(&lp->rx_bytes);
+ } while (u64_stats_fetch_retry(&lp->rx_stat_sync, start));
+
+ do {
+ start = u64_stats_fetch_begin(&lp->tx_stat_sync);
+ stats->tx_packets = u64_stats_read(&lp->tx_packets);
+ stats->tx_bytes = u64_stats_read(&lp->tx_bytes);
+ } while (u64_stats_fetch_retry(&lp->tx_stat_sync, start));
+
+ if (!(lp->features & XAE_FEATURE_STATS))
+ return;
+
+ if (lp->axienet_config->get_stats64)
+ lp->axienet_config->get_stats64(lp, stats);
+}
+
static const struct net_device_ops axienet_netdev_ops = {
.ndo_open = axienet_open,
.ndo_stop = axienet_stop,
@@ -2474,11 +2483,8 @@ static int axienet_ethtools_nway_reset(struct net_device *dev)
return phylink_ethtool_nway_reset(lp->phylink);
}
-static void axienet_ethtools_get_ethtool_stats(struct net_device *dev,
- struct ethtool_stats *stats,
- u64 *data)
+static void axienet_1g_get_ethtool_stats(struct axienet_local *lp, u64 *data)
{
- struct axienet_local *lp = netdev_priv(dev);
unsigned int start;
do {
@@ -2495,6 +2501,17 @@ static void axienet_ethtools_get_ethtool_stats(struct net_device *dev,
} while (read_seqcount_retry(&lp->hw_stats_seqcount, start));
}
+static void axienet_ethtools_get_ethtool_stats(struct net_device *dev,
+ struct ethtool_stats *stats,
+ u64 *data)
+{
+ struct axienet_local *lp = netdev_priv(dev);
+
+ if (lp->features & XAE_FEATURE_STATS &&
+ lp->axienet_config->get_ethtool_stats)
+ lp->axienet_config->get_ethtool_stats(lp, data);
+}
+
static const char axienet_ethtool_stats_strings[][ETH_GSTRING_LEN] = {
"Received bytes",
"Transmitted bytes",
@@ -2507,12 +2524,26 @@ static const char axienet_ethtool_stats_strings[][ETH_GSTRING_LEN] = {
"User Defined Counter 2",
};
+static void axienet_1g_get_strings(u8 *data)
+{
+ memcpy(data, axienet_ethtool_stats_strings,
+ sizeof(axienet_ethtool_stats_strings));
+}
+
+static int axienet_1g_get_sset_count(void)
+{
+ return ARRAY_SIZE(axienet_ethtool_stats_strings);
+}
+
static void axienet_ethtools_get_strings(struct net_device *dev, u32 stringset, u8 *data)
{
+ struct axienet_local *lp = netdev_priv(dev);
+
switch (stringset) {
case ETH_SS_STATS:
- memcpy(data, axienet_ethtool_stats_strings,
- sizeof(axienet_ethtool_stats_strings));
+ if (lp->features & XAE_FEATURE_STATS &&
+ lp->axienet_config->get_strings)
+ lp->axienet_config->get_strings(data);
break;
}
}
@@ -2523,24 +2554,20 @@ static int axienet_ethtools_get_sset_count(struct net_device *dev, int sset)
switch (sset) {
case ETH_SS_STATS:
- if (lp->features & XAE_FEATURE_STATS)
- return ARRAY_SIZE(axienet_ethtool_stats_strings);
+ if (lp->features & XAE_FEATURE_STATS &&
+ lp->axienet_config->get_sset_count)
+ return lp->axienet_config->get_sset_count();
fallthrough;
default:
return -EOPNOTSUPP;
}
}
-static void
-axienet_ethtools_get_pause_stats(struct net_device *dev,
- struct ethtool_pause_stats *pause_stats)
+static void axienet_1g_get_pause_stats(struct axienet_local *lp,
+ struct ethtool_pause_stats *pause_stats)
{
- struct axienet_local *lp = netdev_priv(dev);
unsigned int start;
- if (!(lp->features & XAE_FEATURE_STATS))
- return;
-
do {
start = read_seqcount_begin(&lp->hw_stats_seqcount);
pause_stats->tx_pause_frames =
@@ -2551,15 +2578,23 @@ axienet_ethtools_get_pause_stats(struct net_device *dev,
}
static void
-axienet_ethtool_get_eth_mac_stats(struct net_device *dev,
- struct ethtool_eth_mac_stats *mac_stats)
+axienet_ethtools_get_pause_stats(struct net_device *dev,
+ struct ethtool_pause_stats *pause_stats)
{
struct axienet_local *lp = netdev_priv(dev);
- unsigned int start;
if (!(lp->features & XAE_FEATURE_STATS))
return;
+ if (lp->axienet_config->get_pause_stats)
+ lp->axienet_config->get_pause_stats(lp, pause_stats);
+}
+
+static void axienet_1g_get_eth_mac_stats(struct axienet_local *lp,
+ struct ethtool_eth_mac_stats *mac_stats)
+{
+ unsigned int start;
+
do {
start = read_seqcount_begin(&lp->hw_stats_seqcount);
mac_stats->FramesTransmittedOK =
@@ -2596,15 +2631,24 @@ axienet_ethtool_get_eth_mac_stats(struct net_device *dev,
}
static void
-axienet_ethtool_get_eth_ctrl_stats(struct net_device *dev,
- struct ethtool_eth_ctrl_stats *ctrl_stats)
+axienet_ethtool_get_eth_mac_stats(struct net_device *dev,
+ struct ethtool_eth_mac_stats *mac_stats)
{
struct axienet_local *lp = netdev_priv(dev);
- unsigned int start;
if (!(lp->features & XAE_FEATURE_STATS))
return;
+ if (lp->axienet_config->get_eth_mac_stats)
+ lp->axienet_config->get_eth_mac_stats(lp, mac_stats);
+}
+
+static void
+axienet_1g_get_eth_ctrl_stats(struct axienet_local *lp,
+ struct ethtool_eth_ctrl_stats *ctrl_stats)
+{
+ unsigned int start;
+
do {
start = read_seqcount_begin(&lp->hw_stats_seqcount);
ctrl_stats->MACControlFramesTransmitted =
@@ -2616,6 +2660,19 @@ axienet_ethtool_get_eth_ctrl_stats(struct net_device *dev,
} while (read_seqcount_retry(&lp->hw_stats_seqcount, start));
}
+static void
+axienet_ethtool_get_eth_ctrl_stats(struct net_device *dev,
+ struct ethtool_eth_ctrl_stats *ctrl_stats)
+{
+ struct axienet_local *lp = netdev_priv(dev);
+
+ if (!(lp->features & XAE_FEATURE_STATS))
+ return;
+
+ if (lp->axienet_config->get_eth_ctrl_stats)
+ lp->axienet_config->get_eth_ctrl_stats(lp, ctrl_stats);
+}
+
static const struct ethtool_rmon_hist_range axienet_rmon_ranges[] = {
{ 64, 64 },
{ 65, 127 },
@@ -2628,16 +2685,12 @@ static const struct ethtool_rmon_hist_range axienet_rmon_ranges[] = {
};
static void
-axienet_ethtool_get_rmon_stats(struct net_device *dev,
- struct ethtool_rmon_stats *rmon_stats,
- const struct ethtool_rmon_hist_range **ranges)
+axienet_1g_get_rmon_stats(struct axienet_local *lp,
+ struct ethtool_rmon_stats *rmon_stats,
+ const struct ethtool_rmon_hist_range **ranges)
{
- struct axienet_local *lp = netdev_priv(dev);
unsigned int start;
- if (!(lp->features & XAE_FEATURE_STATS))
- return;
-
do {
start = read_seqcount_begin(&lp->hw_stats_seqcount);
rmon_stats->undersize_pkts =
@@ -2681,6 +2734,20 @@ axienet_ethtool_get_rmon_stats(struct net_device *dev,
*ranges = axienet_rmon_ranges;
}
+static void
+axienet_ethtool_get_rmon_stats(struct net_device *dev,
+ struct ethtool_rmon_stats *rmon_stats,
+ const struct ethtool_rmon_hist_range **ranges)
+{
+ struct axienet_local *lp = netdev_priv(dev);
+
+ if (!(lp->features & XAE_FEATURE_STATS))
+ return;
+
+ if (lp->axienet_config->get_rmon_stats)
+ lp->axienet_config->get_rmon_stats(lp, rmon_stats, ranges);
+}
+
static const struct ethtool_ops axienet_ethtool_ops = {
.supported_coalesce_params = ETHTOOL_COALESCE_MAX_FRAMES |
ETHTOOL_COALESCE_USECS |
@@ -3010,6 +3077,14 @@ static const struct axienet_config axienet_1g_config = {
.phylink_set_caps = axienet_1g_phylink_set_caps,
.pcs_ops = &axienet_pcs_ops,
.stats_update = axienet_1g_stats_update,
+ .get_stats64 = axienet_1g_get_stats64,
+ .get_ethtool_stats = axienet_1g_get_ethtool_stats,
+ .get_strings = axienet_1g_get_strings,
+ .get_sset_count = axienet_1g_get_sset_count,
+ .get_pause_stats = axienet_1g_get_pause_stats,
+ .get_eth_mac_stats = axienet_1g_get_eth_mac_stats,
+ .get_eth_ctrl_stats = axienet_1g_get_eth_ctrl_stats,
+ .get_rmon_stats = axienet_1g_get_rmon_stats,
};
/* Match table for of_platform binding */
--
2.25.1
^ permalink raw reply related [flat|nested] 12+ messages in thread
* [PATCH net-next v5 6/6] net: xilinx: axienet: Add statistics support for XXV ethernet
2026-11-14 18:29 [PATCH net-next v5 0/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support Suraj Gupta
` (4 preceding siblings ...)
2026-11-14 18:29 ` [PATCH net-next v5 5/6] net: xilinx: axienet: Dispatch statistics through axienet_config ops Suraj Gupta
@ 2026-11-14 18:29 ` Suraj Gupta
2026-11-16 4:12 ` netdev-bot+sashiko
5 siblings, 1 reply; 12+ messages in thread
From: Suraj Gupta @ 2026-11-14 18:29 UTC (permalink / raw)
To: Radhey Shyam Pandey, Andrew Lunn, David S . Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley
Cc: Michal Simek, Russell King, netdev, devicetree, linux-arm-kernel,
linux-kernel
The XXV ethernet exposes a dedicated statistics block whose counters are
clear-on-tick (PG210): writing TICK_REG latches the internal accumulators
into the readable registers and clears them, so each post-TICK read yields
the count for the interval since the previous TICK. Software accumulates
those intervals, so no last-counter delta (as used by the 1G MAC) is
needed. The readable registers are not reset along with the accumulators
and hold unknown data until the first TICK, so axienet_xxv_mac_init()
drains them once the GT is ready and before RX/TX are enabled.
Feed both ndo_get_stats64() and the ethtool statistics (-S, pause, MAC and
RMON) from this accumulator path through the axienet_config statistics ops.
MAC control statistics stay unsupported as the hardware exposes no matching
counters.
The counters are an optional part of the IP, selected by the "Include
Statistics Counters" build option. Unlike the 1G MAC, which advertises
XAE_ABILITY_STATS, the XXV register map does not report whether they were
included, so gate XAE_FEATURE_STATS on the "xlnx,has-statistics-counters"
devicetree property instead. Gate the TICK_REG drain and the periodic
statistics work on that same flag, so that a design built without the
block is never touched in the statistics register range. The work
previously keyed off the presence of the stats_update hook alone, which
a config installs unconditionally.
Readers copy the software accumulator and do not tick. The periodic
statistics work latches it, and axienet_stop() takes a final snapshot.
ethtool -S and the pause, MAC and RMON callbacks use that same copy, so
they can lag the work interval; the work is deferrable, so on an idle CPU
the lag can exceed it. ndo_get_stats64() may run atomically and cannot tick
these clear-on-tick counters itself. PG210 defines the counters as 48 bits
(LSB [31:0], MSB [15:0] holds [47:32]), saturating at 0xFFFFFFFFFFFF, so
the existing work interval cannot wrap them at 25 Gbit/s. Mask the MSB
register to those 16 bits when a counter is read, so that bits outside the
counter cannot be accumulated into the software base.
The final statistics snapshot added to axienet_stop() is gated on
XAE_FEATURE_STATS, so it also runs on the 1G MAC when that MAC implements
the statistics block; it is therefore not strictly XXV-only.
Signed-off-by: Suraj Gupta <suraj.gupta2@amd.com>
---
drivers/net/ethernet/xilinx/xilinx_axienet.h | 6 +-
.../net/ethernet/xilinx/xilinx_axienet_main.c | 36 ++-
.../net/ethernet/xilinx/xilinx_axienet_xxv.c | 282 ++++++++++++++++++
.../net/ethernet/xilinx/xilinx_axienet_xxv.h | 58 ++++
4 files changed, 378 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet.h b/drivers/net/ethernet/xilinx/xilinx_axienet.h
index 7468fed1416e..10cb73bc460d 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet.h
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet.h
@@ -520,7 +520,7 @@ struct skbuf_dma_descriptor {
* @reset_in_progress: Set while we are performing a reset and statistics
* counters may be invalid
* @hw_stats_seqcount: Sequence counter for @hw_stat_base, @hw_last_counter,
- * and @reset_in_progress.
+ * @reset_in_progress and, on XXV, @xxv_stat_base.
* @stats_lock: Lock for @hw_stats_seqcount
* @stats_work: Work for reading the hardware statistics counters often enough
* to catch overflows.
@@ -551,6 +551,9 @@ struct skbuf_dma_descriptor {
* @rx_ring_tail: RX skb ring buffer tail index.
* @rx_submit_lock: Protects RX ring resubmission vs teardown in dmaengine path.
* @xxv_ip_version: XXV IP version.
+ * @xxv_stat_base: Software accumulators for the XXV clear-on-tick counters.
+ * Serialized by @stats_lock and published under
+ * @hw_stats_seqcount.
* @axienet_config: MAC-type specific configuration and operations.
*/
struct axienet_local {
@@ -605,6 +608,7 @@ struct axienet_local {
u64 hw_stat_base[STAT_COUNT];
u32 hw_last_counter[STAT_COUNT];
+ u64 xxv_stat_base[XXV_STAT_COUNT];
seqcount_mutex_t hw_stats_seqcount;
struct mutex stats_lock;
struct delayed_work stats_work;
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
index 329b6adf2475..c2ee2cc17e4e 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
@@ -637,13 +637,23 @@ static void axienet_refresh_stats(struct work_struct *work)
struct axienet_local *lp = container_of(work, struct axienet_local,
stats_work.work);
- if (lp->axienet_config->stats_update) {
+ /* Gate on XAE_FEATURE_STATS as axienet_stop() does: a config installs
+ * the hook unconditionally, but the counters themselves are optional,
+ * and on XXV the registers are absent when they were not built in.
+ */
+ if (lp->features & XAE_FEATURE_STATS && lp->axienet_config->stats_update) {
mutex_lock(&lp->stats_lock);
lp->axienet_config->stats_update(lp);
mutex_unlock(&lp->stats_lock);
}
- /* Just less than 2^32 bytes at 2.5 GBit/s */
+ /* Re-arm the poll before any counter can overflow between updates: the
+ * 1G byte counter reaches just under 2^32 in this interval at 2.5 Gbit/s.
+ * The XXV counters are 48 bits (PG210): the LSB register holds [31:0] and
+ * MSB[15:0] holds [47:32]. They saturate at 0xFFFFFFFFFFFF and clear on
+ * tick, so this interval cannot wrap them at 25 Gbit/s. This work is
+ * what latches them between queries.
+ */
schedule_delayed_work(&lp->stats_work, 13 * HZ);
}
@@ -652,7 +662,16 @@ static int __axienet_device_reset(struct axienet_local *lp)
u32 value;
int ret;
- /* Save statistics counters in case they will be reset */
+ /* Save statistics counters in case they will be reset.
+ *
+ * The XAE_FEATURE_STATS blocks in this function read the TEMAC
+ * statistics registers at XAE_STATS_OFFSET, i.e. they assume the 1G
+ * counter layout. This is safe because __axienet_device_reset() is
+ * only reached on the non-dmaengine path, which the
+ * non_dmaengine capability restricts to the 1G MAC. MACs that use their
+ * own statistics mechanism (e.g. XXV, which is dmaengine-only) never
+ * execute this path.
+ */
mutex_lock(&lp->stats_lock);
if (lp->features & XAE_FEATURE_STATS)
axienet_stats_update(lp, true);
@@ -1880,6 +1899,17 @@ static int axienet_stop(struct net_device *ndev)
axienet_setoptions(ndev, lp->options &
~(XAE_OPTION_TXEN | XAE_OPTION_RXEN));
+ /* Final snapshot after RX/TX is disabled, so no frames are missed.
+ * Gate on XAE_FEATURE_STATS rather than the stats_update hook: a config
+ * always installs the hook, but the counters are optional in both MACs,
+ * so only probe_init() knows whether they are there.
+ */
+ if (lp->features & XAE_FEATURE_STATS && lp->axienet_config->stats_update) {
+ mutex_lock(&lp->stats_lock);
+ lp->axienet_config->stats_update(lp);
+ mutex_unlock(&lp->stats_lock);
+ }
+
if (!lp->use_dmaengine) {
axienet_dma_stop(lp);
cancel_work_sync(&lp->dma_err_task);
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
index 3c204dbbf6ae..83782bfbce4c 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
@@ -8,9 +8,12 @@
#include <linux/bitfield.h>
#include <linux/clk.h>
#include <linux/delay.h>
+#include <linux/ethtool.h>
#include <linux/iopoll.h>
#include <linux/netdevice.h>
#include <linux/phylink.h>
+#include <linux/property.h>
+#include <linux/string.h>
#include "xilinx_axienet.h"
#include "xilinx_axienet_xxv.h"
@@ -96,6 +99,66 @@ static const struct axienet_option xxvenet_options[] = {
{}
};
+/* PG210 statistics-block offsets indexed by enum xxv_stat. Each entry is
+ * the 32-bit LSB register address; the corresponding MSB is at +4.
+ */
+static const u32 axienet_xxv_stat_offsets[XXV_STAT_COUNT] = {
+ [XXV_STAT_RX_TOTAL_PACKETS] = XXV_STAT_RX_TOTAL_PACKETS_OFFSET,
+ [XXV_STAT_RX_TOTAL_GOOD_PACKETS] = XXV_STAT_RX_TOTAL_GOOD_PACKETS_OFFSET,
+ [XXV_STAT_RX_TOTAL_BYTES] = XXV_STAT_RX_TOTAL_BYTES_OFFSET,
+ [XXV_STAT_RX_TOTAL_GOOD_BYTES] = XXV_STAT_RX_TOTAL_GOOD_BYTES_OFFSET,
+ [XXV_STAT_RX_PACKET_BAD_FCS] = XXV_STAT_RX_PACKET_BAD_FCS_OFFSET,
+ [XXV_STAT_RX_INRANGEERR] = XXV_STAT_RX_INRANGEERR_OFFSET,
+ [XXV_STAT_RX_MULTICAST] = XXV_STAT_RX_MULTICAST_OFFSET,
+ [XXV_STAT_RX_BROADCAST] = XXV_STAT_RX_BROADCAST_OFFSET,
+ [XXV_STAT_RX_UNDERSIZE] = XXV_STAT_RX_UNDERSIZE_OFFSET,
+ [XXV_STAT_RX_FRAGMENT] = XXV_STAT_RX_FRAGMENT_OFFSET,
+ [XXV_STAT_RX_JABBER] = XXV_STAT_RX_JABBER_OFFSET,
+ [XXV_STAT_RX_OVERSIZE] = XXV_STAT_RX_OVERSIZE_OFFSET,
+ [XXV_STAT_RX_PAUSE] = XXV_STAT_RX_PAUSE_OFFSET,
+ [XXV_STAT_TX_TOTAL_PACKETS] = XXV_STAT_TX_TOTAL_PACKETS_OFFSET,
+ [XXV_STAT_TX_TOTAL_GOOD_PACKETS] = XXV_STAT_TX_TOTAL_GOOD_PACKETS_OFFSET,
+ [XXV_STAT_TX_TOTAL_BYTES] = XXV_STAT_TX_TOTAL_BYTES_OFFSET,
+ [XXV_STAT_TX_TOTAL_GOOD_BYTES] = XXV_STAT_TX_TOTAL_GOOD_BYTES_OFFSET,
+ [XXV_STAT_TX_BAD_FCS] = XXV_STAT_TX_BAD_FCS_OFFSET,
+ [XXV_STAT_TX_FRAME_ERROR] = XXV_STAT_TX_FRAME_ERROR_OFFSET,
+ [XXV_STAT_TX_MULTICAST] = XXV_STAT_TX_MULTICAST_OFFSET,
+ [XXV_STAT_TX_BROADCAST] = XXV_STAT_TX_BROADCAST_OFFSET,
+ [XXV_STAT_TX_PAUSE] = XXV_STAT_TX_PAUSE_OFFSET,
+};
+
+/* XXV (PG210) counters exposed via ethtool -S. Only counters that have no
+ * standard uAPI are listed here; the totals include errored frames, so they
+ * are distinct from the good-frame counts in rtnl_link_stats64 and the
+ * IEEE MAC statistics. Every other XXV counter is reported through the
+ * dedicated stats64, pause, MAC and RMON callbacks instead.
+ */
+static const enum xxv_stat axienet_xxv_priv_stats[] = {
+ XXV_STAT_RX_TOTAL_PACKETS,
+ XXV_STAT_RX_TOTAL_BYTES,
+ XXV_STAT_TX_TOTAL_PACKETS,
+ XXV_STAT_TX_TOTAL_BYTES,
+};
+
+static const char axienet_xxv_ethtool_stats_strings[][ETH_GSTRING_LEN] = {
+ "RX Total Packets",
+ "RX Total Bytes",
+ "TX Total Packets",
+ "TX Total Bytes",
+};
+
+static u64 axienet_xxv_read_counter(struct axienet_local *lp, enum xxv_stat stat)
+{
+ u32 off = axienet_xxv_stat_offsets[stat];
+ u32 lsb, msb;
+
+ lsb = axienet_ior(lp, off);
+ msb = axienet_ior(lp, off + XXV_STAT_MSB_OFFSET);
+
+ /* MSB[31:16] is not part of the 48-bit counter. */
+ return ((u64)FIELD_GET(XXV_STAT_MSB_MASK, msb) << 32) | lsb;
+}
+
static bool axienet_xxv_ip_has_gt_status(u32 ip_version)
{
u8 minor = FIELD_GET(XXV_MIN_MASK, ip_version);
@@ -113,6 +176,13 @@ static bool axienet_xxv_ip_has_gt_status(u32 ip_version)
static void axienet_xxv_probe_init(struct axienet_local *lp)
{
lp->xxv_ip_version = axienet_ior(lp, XXV_CONFIG_REVISION_OFFSET);
+
+ /* The counters are an optional part of the IP (PG210 "Include
+ * Statistics Counters") and nothing in the register map reports
+ * whether they were built in, so trust the devicetree.
+ */
+ if (device_property_read_bool(lp->dev, "xlnx,has-statistics-counters"))
+ lp->features |= XAE_FEATURE_STATS;
}
/**
@@ -202,6 +272,12 @@ static int axienet_xxv_mac_init(struct net_device *ndev)
FIELD_PREP(XXV_RX_MAX_LEN_MASK, lp->max_frm_size) |
XXV_RX_MIN_LEN_DEFAULT);
+ /* Drain pre-open counts now that GT reset-done is complete and RX/TX
+ * are still disabled, so statistics start from zero.
+ */
+ if (lp->features & XAE_FEATURE_STATS)
+ axienet_iow(lp, XXV_TICKREG_OFFSET, XXV_TICKREG_STATEN_MASK);
+
lp->options |= XAE_OPTION_FCS_STRIP;
lp->options |= XAE_OPTION_FCS_INSERT;
@@ -335,6 +411,204 @@ static int axienet_10g25g_clk_init(struct axienet_local *lp)
return 0;
}
+/**
+ * axienet_xxv_stats_update - Latch and accumulate XXV hardware counters
+ * @lp: Pointer to the axienet_local structure
+ *
+ * The XXV statistics counters are clear-on-tick (PG210): writing TICK_REG
+ * latches the internal accumulators into the readable STAT_*_LSB/MSB registers
+ * and clears the internal accumulators. Each post-TICK read therefore returns
+ * the count for the interval since the previous TICK, so software accumulates
+ * those intervals into lp->xxv_stat_base. A last-counter delta (as used by the
+ * free-running 1G MAC path) is not needed.
+ *
+ * Readers copy @xxv_stat_base and do not tick. ndo_get_stats64() may run in
+ * atomic context, so it cannot take @stats_lock and tick itself. The periodic
+ * statistics work, and the final snapshot in axienet_stop(), are the latches.
+ */
+static void axienet_xxv_stats_update(struct axienet_local *lp)
+{
+ u64 counter[XXV_STAT_COUNT];
+ enum xxv_stat stat;
+
+ /* Latch the clear-on-tick snapshot and read the counters outside the
+ * seqcount write section, so only the accumulator updates run inside it
+ * and the reader retry window stays short.
+ */
+ axienet_iow(lp, XXV_TICKREG_OFFSET, XXV_TICKREG_STATEN_MASK);
+ for (stat = 0; stat < XXV_STAT_COUNT; stat++)
+ counter[stat] = axienet_xxv_read_counter(lp, stat);
+
+ write_seqcount_begin(&lp->hw_stats_seqcount);
+ for (stat = 0; stat < XXV_STAT_COUNT; stat++)
+ lp->xxv_stat_base[stat] += counter[stat];
+ write_seqcount_end(&lp->hw_stats_seqcount);
+}
+
+/**
+ * axienet_xxv_get_stats64 - Fill rtnl_link_stats64 from XXV counters
+ * @lp: Pointer to the axienet_local structure
+ * @stats: Output rtnl_link_stats64 structure to populate
+ */
+static void axienet_xxv_get_stats64(struct axienet_local *lp,
+ struct rtnl_link_stats64 *stats)
+{
+ unsigned int start;
+ u64 tx_hw_errors;
+
+ do {
+ start = read_seqcount_begin(&lp->hw_stats_seqcount);
+ stats->rx_crc_errors =
+ lp->xxv_stat_base[XXV_STAT_RX_PACKET_BAD_FCS];
+ /* Both in-range length errors and frame-too-long (oversize)
+ * frames are IEEE 802.3 length errors, so fold them together
+ * into rx_length_errors; rx_over_errors is reserved for receiver
+ * FIFO/ring overflow, which the XXV MAC does not expose.
+ */
+ stats->rx_length_errors =
+ lp->xxv_stat_base[XXV_STAT_RX_INRANGEERR] +
+ lp->xxv_stat_base[XXV_STAT_RX_OVERSIZE];
+ stats->rx_errors = lp->xxv_stat_base[XXV_STAT_RX_UNDERSIZE] +
+ lp->xxv_stat_base[XXV_STAT_RX_FRAGMENT] +
+ lp->xxv_stat_base[XXV_STAT_RX_JABBER] +
+ stats->rx_crc_errors +
+ stats->rx_length_errors;
+ stats->multicast = lp->xxv_stat_base[XXV_STAT_RX_MULTICAST];
+ tx_hw_errors = lp->xxv_stat_base[XXV_STAT_TX_BAD_FCS] +
+ lp->xxv_stat_base[XXV_STAT_TX_FRAME_ERROR];
+ } while (read_seqcount_retry(&lp->hw_stats_seqcount, start));
+
+ /* Fold the HW TX error count into the software tx_errors seeded from
+ * dev->stats by the caller. The XXV xmit path bumps dev->stats.tx_errors
+ * for frames dropped on pad failure before they reach the MAC, while the
+ * HW counters only cover frames the MAC actually transmitted; the two are
+ * disjoint. This runs once, outside the seqcount retry loop, so a racing
+ * stats update cannot make the accumulation re-add on retry.
+ */
+ stats->tx_errors += tx_hw_errors;
+}
+
+/**
+ * axienet_xxv_get_ethtool_stats - Copy XXV counters for ethtool -S
+ * @lp: Pointer to the axienet_local structure
+ * @data: Output buffer sized for ARRAY_SIZE(axienet_xxv_priv_stats) u64 values
+ */
+static void axienet_xxv_get_ethtool_stats(struct axienet_local *lp, u64 *data)
+{
+ unsigned int i, start;
+
+ do {
+ start = read_seqcount_begin(&lp->hw_stats_seqcount);
+ for (i = 0; i < ARRAY_SIZE(axienet_xxv_priv_stats); i++)
+ data[i] = lp->xxv_stat_base[axienet_xxv_priv_stats[i]];
+ } while (read_seqcount_retry(&lp->hw_stats_seqcount, start));
+}
+
+/**
+ * axienet_xxv_get_strings - Copy XXV ethtool statistics name strings
+ * @data: Output buffer for the statistics name strings
+ */
+static void axienet_xxv_get_strings(u8 *data)
+{
+ memcpy(data, axienet_xxv_ethtool_stats_strings,
+ sizeof(axienet_xxv_ethtool_stats_strings));
+}
+
+/**
+ * axienet_xxv_get_sset_count - Number of XXV ethtool statistics
+ *
+ * Return: Count of XXV hardware statistics reported via ethtool -S.
+ */
+static int axienet_xxv_get_sset_count(void)
+{
+ return ARRAY_SIZE(axienet_xxv_ethtool_stats_strings);
+}
+
+/**
+ * axienet_xxv_get_pause_stats - Fill ethtool pause frame statistics
+ * @lp: Pointer to the axienet_local structure
+ * @pause_stats: Output ethtool_pause_stats structure to populate
+ */
+static void axienet_xxv_get_pause_stats(struct axienet_local *lp,
+ struct ethtool_pause_stats *pause_stats)
+{
+ unsigned int start;
+
+ do {
+ start = read_seqcount_begin(&lp->hw_stats_seqcount);
+ pause_stats->tx_pause_frames =
+ lp->xxv_stat_base[XXV_STAT_TX_PAUSE];
+ pause_stats->rx_pause_frames =
+ lp->xxv_stat_base[XXV_STAT_RX_PAUSE];
+ } while (read_seqcount_retry(&lp->hw_stats_seqcount, start));
+}
+
+/**
+ * axienet_xxv_get_eth_mac_stats - Fill IEEE 802.3 MAC statistics
+ * @lp: Pointer to the axienet_local structure
+ * @mac_stats: Output ethtool_eth_mac_stats structure to populate
+ */
+static void axienet_xxv_get_eth_mac_stats(struct axienet_local *lp,
+ struct ethtool_eth_mac_stats *mac_stats)
+{
+ unsigned int start;
+
+ do {
+ start = read_seqcount_begin(&lp->hw_stats_seqcount);
+ mac_stats->FramesTransmittedOK =
+ lp->xxv_stat_base[XXV_STAT_TX_TOTAL_GOOD_PACKETS];
+ mac_stats->FramesReceivedOK =
+ lp->xxv_stat_base[XXV_STAT_RX_TOTAL_GOOD_PACKETS];
+ mac_stats->OctetsTransmittedOK =
+ lp->xxv_stat_base[XXV_STAT_TX_TOTAL_GOOD_BYTES];
+ mac_stats->OctetsReceivedOK =
+ lp->xxv_stat_base[XXV_STAT_RX_TOTAL_GOOD_BYTES];
+ mac_stats->FrameCheckSequenceErrors =
+ lp->xxv_stat_base[XXV_STAT_RX_PACKET_BAD_FCS];
+ mac_stats->MulticastFramesXmittedOK =
+ lp->xxv_stat_base[XXV_STAT_TX_MULTICAST];
+ mac_stats->BroadcastFramesXmittedOK =
+ lp->xxv_stat_base[XXV_STAT_TX_BROADCAST];
+ mac_stats->MulticastFramesReceivedOK =
+ lp->xxv_stat_base[XXV_STAT_RX_MULTICAST];
+ mac_stats->BroadcastFramesReceivedOK =
+ lp->xxv_stat_base[XXV_STAT_RX_BROADCAST];
+ mac_stats->InRangeLengthErrors =
+ lp->xxv_stat_base[XXV_STAT_RX_INRANGEERR];
+ } while (read_seqcount_retry(&lp->hw_stats_seqcount, start));
+}
+
+/**
+ * axienet_xxv_get_rmon_stats - Fill RMON statistics and histogram ranges
+ * @lp: Pointer to the axienet_local structure
+ * @rmon_stats: Output ethtool_rmon_stats structure to populate
+ * @ranges: Set to NULL; XXV exposes aggregate RMON counters only, not
+ * per-bin histogram ranges.
+ */
+static void axienet_xxv_get_rmon_stats(struct axienet_local *lp,
+ struct ethtool_rmon_stats *rmon_stats,
+ const struct ethtool_rmon_hist_range **ranges)
+{
+ unsigned int start;
+
+ do {
+ start = read_seqcount_begin(&lp->hw_stats_seqcount);
+ rmon_stats->undersize_pkts =
+ lp->xxv_stat_base[XXV_STAT_RX_UNDERSIZE];
+ rmon_stats->oversize_pkts =
+ lp->xxv_stat_base[XXV_STAT_RX_OVERSIZE];
+ rmon_stats->fragments =
+ lp->xxv_stat_base[XXV_STAT_RX_FRAGMENT];
+ rmon_stats->jabbers =
+ lp->xxv_stat_base[XXV_STAT_RX_JABBER];
+ } while (read_seqcount_retry(&lp->hw_stats_seqcount, start));
+
+ /* XXV currently exposes only aggregate RMON counters, not per-bin
+ * histogram buckets. Keep ranges NULL until histogram bins are wired.
+ */
+ *ranges = NULL;
+}
+
const struct axienet_config axienet_10g25g_config = {
.jumbo = true,
.sw_padding = true,
@@ -348,4 +622,12 @@ const struct axienet_config axienet_10g25g_config = {
.get_regs = axienet_xxv_get_regs,
.phylink_set_caps = axienet_xxv_phylink_set_capabilities,
.pcs_ops = &axienet_xxv_pcs_ops,
+ .stats_update = axienet_xxv_stats_update,
+ .get_stats64 = axienet_xxv_get_stats64,
+ .get_ethtool_stats = axienet_xxv_get_ethtool_stats,
+ .get_strings = axienet_xxv_get_strings,
+ .get_sset_count = axienet_xxv_get_sset_count,
+ .get_pause_stats = axienet_xxv_get_pause_stats,
+ .get_eth_mac_stats = axienet_xxv_get_eth_mac_stats,
+ .get_rmon_stats = axienet_xxv_get_rmon_stats,
};
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.h b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.h
index 6530344e40c7..83682d1524ed 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.h
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.h
@@ -66,6 +66,36 @@ struct axienet_config;
#define XXV_STAT_GT_OFFSET 0x000004A0
#define XXV_STAT_AN_LINK_CTL2_OFFSET 0x000009F0
+/* XXV statistics counter registers (PG210). Each counter is a pair of 32-bit
+ * registers; the MSB register is XXV_STAT_MSB_OFFSET bytes above the LSB.
+ * MSB[15:0] holds counter bits [47:32]. The upper half is not part of the
+ * counter and must be masked off before the value is accumulated.
+ */
+#define XXV_STAT_MSB_OFFSET 4
+#define XXV_STAT_MSB_MASK GENMASK(15, 0)
+#define XXV_STAT_TX_FRAME_ERROR_OFFSET 0x000006A0
+#define XXV_STAT_TX_TOTAL_PACKETS_OFFSET 0x00000700
+#define XXV_STAT_TX_TOTAL_GOOD_PACKETS_OFFSET 0x00000708
+#define XXV_STAT_TX_TOTAL_BYTES_OFFSET 0x00000710
+#define XXV_STAT_TX_TOTAL_GOOD_BYTES_OFFSET 0x00000718
+#define XXV_STAT_TX_BAD_FCS_OFFSET 0x000007B8
+#define XXV_STAT_TX_MULTICAST_OFFSET 0x000007D8
+#define XXV_STAT_TX_BROADCAST_OFFSET 0x000007E0
+#define XXV_STAT_TX_PAUSE_OFFSET 0x000007F0
+#define XXV_STAT_RX_TOTAL_PACKETS_OFFSET 0x00000808
+#define XXV_STAT_RX_TOTAL_GOOD_PACKETS_OFFSET 0x00000810
+#define XXV_STAT_RX_TOTAL_BYTES_OFFSET 0x00000818
+#define XXV_STAT_RX_TOTAL_GOOD_BYTES_OFFSET 0x00000820
+#define XXV_STAT_RX_UNDERSIZE_OFFSET 0x00000898
+#define XXV_STAT_RX_FRAGMENT_OFFSET 0x000008A0
+#define XXV_STAT_RX_OVERSIZE_OFFSET 0x000008A8
+#define XXV_STAT_RX_JABBER_OFFSET 0x000008B8
+#define XXV_STAT_RX_PACKET_BAD_FCS_OFFSET 0x000008C8
+#define XXV_STAT_RX_MULTICAST_OFFSET 0x000008E0
+#define XXV_STAT_RX_BROADCAST_OFFSET 0x000008E8
+#define XXV_STAT_RX_PAUSE_OFFSET 0x000008F8
+#define XXV_STAT_RX_INRANGEERR_OFFSET 0x00000908
+
/* XXV MAC Register Mask Definitions */
#define XXV_GT_RESET_MASK BIT(0)
#define XXV_TC_TX_MASK BIT(0)
@@ -73,6 +103,7 @@ struct axienet_config;
#define XXV_RCW1_FCS_MASK BIT(1)
#define XXV_TC_FCS_MASK BIT(1)
#define XXV_STAT_RX_BLOCK_LOCK_MASK BIT(0)
+#define XXV_TICKREG_STATEN_MASK BIT(0)
#define XXV_GT_RESET_DONE (BIT(0) | BIT(1))
#define XXV_MAJ_MASK GENMASK(7, 0)
#define XXV_MIN_MASK GENMASK(15, 8)
@@ -88,6 +119,33 @@ struct axienet_config;
#define XXV_RX_MAX_LEN_MASK GENMASK(30, 16)
#define XXV_RX_MIN_LEN_DEFAULT 64
+/* XXV (10G/25G) MAC statistics counters, used to index xxv_stat_base[]. */
+enum xxv_stat {
+ XXV_STAT_RX_TOTAL_PACKETS = 0,
+ XXV_STAT_RX_TOTAL_GOOD_PACKETS,
+ XXV_STAT_RX_TOTAL_BYTES,
+ XXV_STAT_RX_TOTAL_GOOD_BYTES,
+ XXV_STAT_RX_PACKET_BAD_FCS,
+ XXV_STAT_RX_INRANGEERR,
+ XXV_STAT_RX_MULTICAST,
+ XXV_STAT_RX_BROADCAST,
+ XXV_STAT_RX_UNDERSIZE,
+ XXV_STAT_RX_FRAGMENT,
+ XXV_STAT_RX_JABBER,
+ XXV_STAT_RX_OVERSIZE,
+ XXV_STAT_RX_PAUSE,
+ XXV_STAT_TX_TOTAL_PACKETS,
+ XXV_STAT_TX_TOTAL_GOOD_PACKETS,
+ XXV_STAT_TX_TOTAL_BYTES,
+ XXV_STAT_TX_TOTAL_GOOD_BYTES,
+ XXV_STAT_TX_BAD_FCS,
+ XXV_STAT_TX_FRAME_ERROR,
+ XXV_STAT_TX_MULTICAST,
+ XXV_STAT_TX_BROADCAST,
+ XXV_STAT_TX_PAUSE,
+ XXV_STAT_COUNT,
+};
+
extern const struct axienet_config axienet_10g25g_config;
#endif /* XILINX_AXIENET_XXV_H */
--
2.25.1
^ permalink raw reply related [flat|nested] 12+ messages in thread
* Re: [PATCH net-next v5 1/6] net: xilinx: axienet: Introduce axienet_config for MAC-specific ops
2026-11-14 18:29 ` [PATCH net-next v5 1/6] net: xilinx: axienet: Introduce axienet_config for MAC-specific ops Suraj Gupta
@ 2026-11-16 4:12 ` netdev-bot+sashiko
0 siblings, 0 replies; 12+ messages in thread
From: netdev-bot+sashiko @ 2026-11-16 4:12 UTC (permalink / raw)
To: suraj.gupta2
Cc: radhey.shyam.pandey, andrew+netdev, davem, edumazet, kuba, pabeni,
robh, krzk+dt, conor+dt, michal.simek, linux, netdev, devicetree,
linux-arm-kernel, linux-kernel
Thank you for your contribution! Sashiko AI review found 2 potential
issue(s) to consider.
Critical: 0 · High: 1 · Medium: 1 · Low: 0
If you believe none of the issues are legitimate, please reply to each
issue of Medium severity or higher and explain why the AI is wrong.
If any of them are legitimate and you plan to respin, please let patchwork
know by including "pw-bot: cr" as a separate line at the end of your reply
(one such reply per series is enough).
- [High] net: xilinx: axienet: axienet_probe() fails on software-node
devices
- [Medium] This commit is presented as a no-functional-change refactor,
but it also contains a real fix for a NULL pointer dereference.
--
net: xilinx: axienet: Introduce axienet_config for MAC-specific ops
This commit adds struct axienet_config, which holds MAC capability flags
and MAC-specific callbacks. It is filled in from the OF match table, and
the shared 1G code paths now use it. It also puts a NULL check on
lp->mii_bus around the MDIO teardown in axienet_remove().
> diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> index 7e83c4b5f1931..125ad3dbf6aa9 100644
> --- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> +++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
[ ... ]
> @@ -3149,10 +3297,12 @@ static int axienet_probe(struct platform_device *pdev)
> lp->tx_dma_cr = axienet_calc_cr(lp, XAXIDMA_DFT_TX_THRESHOLD,
> XAXIDMA_DFT_TX_USEC);
>
> - ret = axienet_mdio_setup(lp);
> - if (ret)
> - dev_warn(&pdev->dev,
> - "error registering MDIO bus: %d\n", ret);
> + if (lp->axienet_config->mdio) {
> + ret = axienet_mdio_setup(lp);
> + if (ret)
> + dev_warn(&pdev->dev,
> + "error registering MDIO bus: %d\n", ret);
> + }
>
> if (lp->phy_mode == PHY_INTERFACE_MODE_SGMII ||
> lp->phy_mode == PHY_INTERFACE_MODE_1000BASEX) {
[ ... ]
> @@ -3235,7 +3383,8 @@ static void axienet_remove(struct platform_device *pdev)
> if (lp->pcs_phy)
> put_device(&lp->pcs_phy->dev);
>
> - axienet_mdio_teardown(lp);
> + if (lp->mii_bus)
> + axienet_mdio_teardown(lp);
> }
[Severity: Medium]
Could this NULL check go to net as its own fix, with a Fixes: tag and
Cc: stable, and this refactor then be based on it?
Without the guard, the current 1G driver can oops on unbind. Every
failure path in axienet_mdio_setup() frees the bus and sets lp->mii_bus
to NULL. The failure paths are an MDC divisor overflow, a timeout in
axienet_mdio_wait_until_ready(), and a failed of_mdiobus_register():
drivers/net/ethernet/xilinx/xilinx_axienet_mdio.c:axienet_mdio_setup() {
...
unregister:
of_node_put(mdio_node);
mdiobus_free(bus);
lp->mii_bus = NULL;
return ret;
}
axienet_probe() only warns and continues, so the device binds with
lp->mii_bus == NULL. Then, on unbind or remove:
axienet_remove()
axienet_mdio_teardown(lp)
mdiobus_unregister(lp->mii_bus)
reads bus->state on a NULL bus
The commit message already describes this:
It also hardens the existing 1G path: a failed
axienet_mdio_setup() already left lp->mii_bus NULL while probe only warns
and continues (e.g. a 1G device using a fixed-link), so unbinding such a
device would dereference NULL in mdiobus_unregister().
The fix is part of a large refactor and has no Fixes: tag, so stable and
LTS kernels can't backport it. No other patch in the series carries it
on its own. Two candidates for the Fixes: tag are the commit that made
MDIO setup failure non-fatal in probe and the commit that clears
lp->mii_bus on failure.
[ ... ]
> @@ -2894,27 +3042,13 @@ static int axienet_probe(struct platform_device *pdev)
> seqcount_mutex_init(&lp->hw_stats_seqcount, &lp->stats_lock);
> INIT_DEFERRABLE_WORK(&lp->stats_work, axienet_refresh_stats);
>
> - lp->axi_clk = devm_clk_get_optional_enabled(&pdev->dev,
> - "s_axi_lite_clk");
> - if (!lp->axi_clk) {
> - /* For backward compatibility, if named AXI clock is not present,
> - * treat the first clock specified as the AXI clock.
> - */
> - lp->axi_clk = devm_clk_get_optional_enabled(&pdev->dev, NULL);
> - }
> - if (IS_ERR(lp->axi_clk))
> - return dev_err_probe(&pdev->dev, PTR_ERR(lp->axi_clk),
> - "could not get AXI clock\n");
> -
> - lp->misc_clks[0].id = "axis_clk";
> - lp->misc_clks[1].id = "ref_clk";
> - lp->misc_clks[2].id = "mgt_clk";
> + lp->axienet_config = device_get_match_data(&pdev->dev);
> + if (!lp->axienet_config)
> + return -ENODEV;
>
> - ret = devm_clk_bulk_get_optional_enable(&pdev->dev, XAE_NUM_MISC_CLOCKS,
> - lp->misc_clks);
> + ret = lp->axienet_config->clk_init(lp);
> if (ret)
> - return dev_err_probe(&pdev->dev, ret,
> - "could not get/enable misc. clocks\n");
> + return ret;
>
> /* Map device registers */
> lp->regs = devm_platform_get_and_ioremap_resource(pdev, 0, ðres);
[Severity: High]
This makes missing match data a hard probe failure. Doesn't that
break devices that are not instantiated from DT?
device_get_match_data() only asks the device's fwnode, through
fwnode_call_ptr_op(dev_fwnode(dev), device_get_match_data, dev). For
an OF node that resolves through axienet_of_match. software_node_ops
does not implement .device_get_match_data, though, and axienet_driver
has no id_table either.
The driver still carries MODULE_ALIAS("platform:" AXIENET_DRIVER_NAME).
Probe also reads its configuration through device_property_read_u32()
and the fwnode helpers, so a parent driver can create the AXI Ethernet
device by name with a software_node attached. The ptp_ocp TimeCard
child device is one example. Such a device has no OF match, so
device_get_match_data() returns NULL. axienet_probe() then returns
-ENODEV before it touches the hardware. Before this patch the same
device probed fine as a 1G MAC.
The commit message says no functional change is intended for the
existing 1G path, and this is one. As far as I can see the later
patches in the series keep this lookup as is, so the regression is
still there at the end of the series.
Since every device bound by name today is a 1G MAC, would it be enough
to fall back to &axienet_1g_config when device_get_match_data() returns
NULL? Or would you rather add a platform_device_id table whose
driver_data points at axienet_1g_config, and have probe use
platform_get_device_id() when there is no fwnode match data?
--
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261114182947.2908628-1-suraj.gupta2%40amd.com
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH net-next v5 2/6] dt-bindings: net: xlnx,xxv-ethernet: Add Xilinx XXV 10G/25G Ethernet
2026-11-14 18:29 ` [PATCH net-next v5 2/6] dt-bindings: net: xlnx,xxv-ethernet: Add Xilinx XXV 10G/25G Ethernet Suraj Gupta
@ 2026-11-16 4:12 ` netdev-bot+sashiko
0 siblings, 0 replies; 12+ messages in thread
From: netdev-bot+sashiko @ 2026-11-16 4:12 UTC (permalink / raw)
To: suraj.gupta2
Cc: radhey.shyam.pandey, andrew+netdev, davem, edumazet, kuba, pabeni,
robh, krzk+dt, conor+dt, michal.simek, linux, netdev, devicetree,
linux-arm-kernel, linux-kernel
Thank you for your contribution! Sashiko AI review found more severe
issue(s) in other patches of this series. This mail is only an FYI, in case
you want to address the nit picks below when you respin.
Low priority / nit picks for your consideration:
- [Low] Documentation/devicetree/bindings/net/xlnx,xxv-ethernet.yaml
blocks the inherited properties that conflict with its stated design:…
--
dt-bindings: net: xlnx,xxv-ethernet: Add Xilinx XXV 10G/25G Ethernet
This commit adds a dedicated schema for the Xilinx XXV 10G/25G Ethernet
subsystem, which has an in-core PCS/PMA and uses an external AXI DMA or
MCDMA controller. It requires xlnx,rxmem and adds the optional
xlnx,has-statistics-counters property. It also disallows phy-handle, phy,
phy-device and pcs-handle, and restricts managed to in-band-status.
> diff --git a/Documentation/devicetree/bindings/net/xlnx,xxv-ethernet.yaml b/Documentation/devicetree/bindings/net/xlnx,xxv-ethernet.yaml
> new file mode 100644
> index 0000000000000..caf582c72a8dc
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/net/xlnx,xxv-ethernet.yaml
[ ... ]
> + phy-mode:
> + enum:
> + - 10gbase-r
> + - 25gbase-r
> +
> + # The subsystem has no MDIO bus master and its PCS is in-core, so neither
> + # an external PHY nor an external PCS can be attached. phy and phy-device
> + # are the deprecated spellings of phy-handle.
> + phy-handle: false
> + phy: false
> + phy-device: false
> + pcs-handle: false
> +
> + managed:
> + const: in-band-status
[Severity: Low]
Should fixed-link also be set to false here?
fixed-link comes in through the ethernet-controller.yaml $ref in the same
way as the other blocked properties, so unevaluatedProperties: false does
not reject it. ethernet-controller.yaml also does not make fixed-link and
managed mutually exclusive. Its only dependencies entry is pcs-handle-names
requiring pcs-handle.
The commit message and the binding description both say "the link is
always managed in-band". Even so, this schema accepts two configurations
that go against that.
The first is a node with fixed-link alone. In drivers/net/phy/phylink.c,
phylink_parse_mode() checks fixed-link after the in-band default:
if (pl->config->default_an_inband)
pl->cfg_link_an_mode = MLO_AN_INBAND;
dn = fwnode_get_named_child_node(fwnode, "fixed-link");
if (dn || fwnode_property_present(fwnode, "fixed-link"))
pl->cfg_link_an_mode = MLO_AN_FIXED;
This overrides the default_an_inband = true that
axienet_xxv_phylink_set_capabilities() sets in the XXV driver patch later
in this series. Carrier would then come from the emulated fixed state, not
from the PCS status that axienet_xxv_pcs_get_state() reads.
The second is a node with fixed-link plus managed = "in-band-status". The
binding allows both properties, but phylink_parse_mode() rejects the
combination:
if (pl->cfg_link_an_mode == MLO_AN_FIXED) {
phylink_err(pl,
"can't use both fixed-link and in-band-status\n");
return -EINVAL;
}
So the device fails to probe from axienet_probe()->phylink_create(), even
though dt_binding_check accepted the node.
Later patches in the series don't change this binding file, so the gap is
still there at the end of the series. Adding "fixed-link: false" next to
the other blocked properties would close it.
--
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261114182947.2908628-1-suraj.gupta2%40amd.com
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH net-next v5 3/6] net: xilinx: axienet: Propagate device reset failure from ndo_open
2026-11-14 18:29 ` [PATCH net-next v5 3/6] net: xilinx: axienet: Propagate device reset failure from ndo_open Suraj Gupta
@ 2026-11-16 4:12 ` netdev-bot+sashiko
0 siblings, 0 replies; 12+ messages in thread
From: netdev-bot+sashiko @ 2026-11-16 4:12 UTC (permalink / raw)
To: suraj.gupta2
Cc: radhey.shyam.pandey, andrew+netdev, davem, edumazet, kuba, pabeni,
robh, krzk+dt, conor+dt, michal.simek, linux, netdev, devicetree,
linux-arm-kernel, linux-kernel
Thank you for your contribution! Sashiko AI review found 5 potential
issue(s) to consider.
Critical: 0 · High: 0 · Medium: 3 · Low: 2
If you believe none of the issues are legitimate, please reply to each
issue of Medium severity or higher and explain why the AI is wrong.
If any of them are legitimate and you plan to respin, please let patchwork
know by including "pw-bot: cr" as a separate line at the end of your reply
(one such reply per series is enough).
- [Medium] The patch fixes user-visible bugs in existing kernels but has
no Fixes: tag, and it is sent as the third patch of an XXV feature…
- [Medium] net: xilinx: axienet: u32 stats underflow on
__axienet_device_reset() failure
- [Medium] net: xilinx: axienet: leaked legacy DMA rings and active DMA
on open failure
- [Low] The commit message presents the resume/stop changes as a side
effect of the new reset check.
- [Low] The comment added to axienet_stop() justifies the datapath_up
guard with "axienet_resume() calls ndo_open directly, so unlike the…
--
net: xilinx: axienet: Propagate device reset failure from ndo_open
With this patch, axienet_open() now returns the error from
axienet_device_reset(). A new datapath_up flag makes axienet_stop() a
no-op when ndo_open did not complete. axienet_resume() now propagates
the open failure and calls netif_close() before it reattaches the device.
The patch also clears reset_in_progress on the error path of
__axienet_device_reset().
> That makes a failed ndo_open reachable, which matters for axienet_resume():
> it calls ndo_open directly, so unlike the netdev core it leaves the device
> running when open fails.
[Severity: Low]
Is this description accurate? Before this patch, axienet_open() could
already fail without the reset being involved:
- phylink_fwnode_phy_connect() failure returns directly
- request_irq(lp->eth_irq) failure on the dmaengine path goes to err_phy
- axienet_init_dmaengine() failure (dma_request_chan() or the ring
allocations) goes to err_free_eth_irq
- request_irq() failures inside axienet_init_legacy_dma() go to
err_phy, after napi_disable()
Since a3de357b087e ("net: axiemac: add PM callbacks to support
suspend/resume"), axienet_resume() has ignored the axienet_open() return
value and always called netif_device_attach(). The device stayed IFF_UP
after any of those failures.
A later ndo_stop then ran the full teardown. On the legacy path that is
napi_disable() on an already disabled NAPI with RTNL held, plus
free_irq() on IRQs that were never requested. On the dmaengine path it is
dma_release_channel() on released channels and a second kfree() of
tx_skb_ring/rx_skb_ring.
Could the commit message describe these as existing bugs, rather than as
a consequence of the new reset check?
> diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> index 125ad3dbf6aa9..a553150faa89b 100644
> --- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> +++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
[ ... ]
> @@ -1834,6 +1851,16 @@ static int axienet_stop(struct net_device *ndev)
> struct axienet_local *lp = netdev_priv(ndev);
> int i;
>
> + /* axienet_resume() calls ndo_open directly, so unlike the netdev core
> + * it can leave the device running after a failed open. Tearing down
> + * below would then stop a phylink that open already stopped and
> + * disconnected, and free interrupts it never requested.
> + */
> + if (!lp->datapath_up)
> + return 0;
[Severity: Low]
This comment says axienet_resume() can leave the device running after a
failed open. The same patch changes axienet_resume() to call
netif_close() on that failure.
Isn't that netif_close() call the path that actually reaches
axienet_stop() with datapath_up == false?
axienet_resume()
axienet_open() <- fails
netif_close()
netif_close_many()
__dev_close_many()
ops->ndo_stop() -> axienet_stop()
The device is still IFF_UP at this point, because axienet_suspend()
calls axienet_stop() directly.
Could the comment describe the netif_close() path instead, so the guard
is not later mistaken for dead code and removed?
> +
> + lp->datapath_up = false;
> +
[ ... ]
> @@ -3419,17 +3446,26 @@ static int axienet_suspend(struct device *dev)
> static int axienet_resume(struct device *dev)
> {
> struct net_device *ndev = dev_get_drvdata(dev);
> + int ret;
>
> if (!netif_running(ndev))
> return 0;
>
> rtnl_lock();
> - axienet_open(ndev);
> - rtnl_unlock();
> -
> + ret = axienet_open(ndev);
> + if (ret) {
> + netdev_err(ndev, "failed to reopen device on resume: %d\n", ret);
> + /* Drop IFF_UP before reattaching, so that the interface can be
> + * brought up again once whatever made open fail is resolved.
> + * Reattaching while still up would also let the stack transmit
> + * on the datapath open has already torn down.
> + */
> + netif_close(ndev);
> + }
> netif_device_attach(ndev);
> + rtnl_unlock();
>
> - return 0;
> + return ret;
> }
[Severity: Medium]
Does this leave the legacy DMA rings allocated and the DMA running
when open fails after the reset?
On the non-dmaengine path, axienet_device_reset() calls
axienet_dma_bd_init(). That allocates lp->tx_bd_v and lp->rx_bd_v
with dma_alloc_coherent(), allocates and maps lp->rx_bd_num RX skbs,
and starts both DMA channels. axienet_device_reset() then runs
mac_init() and enables XAE_OPTION_TXEN | XAE_OPTION_RXEN.
None of the failure exits that follow undo that:
- mac_init() failing inside axienet_device_reset(), which this
patch turns into an open failure
- phylink_fwnode_phy_connect() failing, which returns directly
- axienet_init_legacy_dma() failing, which goes to err_phy
err_phy only cancels the work items and stops and disconnects
phylink. Nothing calls axienet_dma_stop() or axienet_dma_bd_release(),
and TXEN and RXEN stay set.
With the new datapath_up guard, axienet_stop() will not clean this up
either. That includes the case where it is reached through
netif_close() in axienet_resume():
axienet_resume()
axienet_open() <- fails after axienet_dma_bd_init()
netif_close()
axienet_stop() <- datapath_up == false, returns 0
The descriptor rings and the mapped RX skbs are therefore never
freed, and the DMA engine and the MAC keep running while the
interface is down. The next axienet_open() calls
axienet_dma_bd_init() again and overwrites lp->tx_bd_v and
lp->rx_bd_v, which leaks the previous rings.
Should the error paths in axienet_open() unwind axienet_device_reset()
on the legacy path, by stopping the DMA and calling
axienet_dma_bd_release()? Then datapath_up == false would really mean
there is nothing left to tear down.
[Severity: Medium]
This fixes bugs that exist in released kernels. Should it carry Fixes:
tags and go to net as a standalone fix, instead of being the third patch
of the XXV feature series?
The reset return value being overwritten by the
phylink_fwnode_phy_connect() result goes back at least to:
Fixes: 6b1b40f704fc ("net: axienet: Preparatory changes for dmaengine support")
The resume handling that leaves the device IFF_UP after a failed reopen
comes from:
Fixes: a3de357b087e ("net: axiemac: add PM callbacks to support suspend/resume")
Without a Fixes: tag, the stable tooling will not pick up the
datapath_up guard or the netif_close() handling here. In stable kernels,
an ifdown after a failed resume can then hang in napi_disable() with
RTNL held on the legacy DMA path. On the dmaengine path it can release
the DMA channels again and free tx_skb_ring/rx_skb_ring a second time.
[ ... ]
> @@ -702,6 +702,17 @@ static int __axienet_device_reset(struct axienet_local *lp)
> }
>
> out:
> + /* On failure the block above is skipped, so clear the flag here.
> + * axienet_open() now returns on a reset error without scheduling the
> + * statistics work that used to clear it, which would otherwise leave
> + * axienet_stat() returning the frozen base until a reset succeeds.
> + */
> + if (ret && (lp->features & XAE_FEATURE_STATS)) {
> + write_seqcount_begin(&lp->hw_stats_seqcount);
> + lp->reset_in_progress = false;
> + write_seqcount_end(&lp->hw_stats_seqcount);
> + }
> +
> mutex_unlock(&lp->stats_lock);
> return ret;
> }
[Severity: Medium]
Does clearing reset_in_progress here leave hw_last_counter[] stale?
At the top of this function, axienet_stats_update(lp, true) stores
the pre-reset counter values in lp->hw_last_counter[] and sets
lp->reset_in_progress. The XAXIDMA_CR_RESET_MASK write that follows
resets the Ethernet core along with the DMA, which zeroes the
statistics counters. On either timeout, the goto out skips the loop
that folds the reset into hw_stat_base[] and reloads
hw_last_counter[].
With reset_in_progress now cleared, axienet_stat() computes
lp->hw_stat_base[stat] + (counter - lp->hw_last_counter[stat])
where counter is the u32 read back from the core. That value is 0
after the reset. If the DMA reset poll is the one that timed out, it
is whatever the core returns while still held in reset.
For every counter that was non-zero before the reset, the u32
subtraction wraps. ndo_get_stats64 and the ethtool stats then report
values close to 2^32 too high. The next axienet_stats_update(), from
the stats work or the next reset, adds the same wrapped delta into
hw_stat_base[], so the error becomes permanent.
The statistics work this replaces would have hit the same wrap
through axienet_stats_update(lp, false). This patch makes it the
explicit outcome of a failed reset, though.
Should the error path resynchronise hw_last_counter[] from the
hardware when the counters are readable? Or is keeping the frozen
base the safer choice when the core may not have come out of reset?
--
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261114182947.2908628-1-suraj.gupta2%40amd.com
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH net-next v5 4/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support
2026-11-14 18:29 ` [PATCH net-next v5 4/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support Suraj Gupta
@ 2026-11-16 4:12 ` netdev-bot+sashiko
0 siblings, 0 replies; 12+ messages in thread
From: netdev-bot+sashiko @ 2026-11-16 4:12 UTC (permalink / raw)
To: suraj.gupta2
Cc: radhey.shyam.pandey, andrew+netdev, davem, edumazet, kuba, pabeni,
robh, krzk+dt, conor+dt, michal.simek, linux, netdev, devicetree,
linux-arm-kernel, linux-kernel
Thank you for your contribution! Sashiko AI review found 7 potential
issue(s) to consider.
Critical: 0 · High: 1 · Medium: 2 · Low: 4
If you believe none of the issues are legitimate, please reply to each
issue of Medium severity or higher and explain why the AI is wrong.
If any of them are legitimate and you plan to respin, please let patchwork
know by including "pw-bot: cr" as a separate line at the end of your reply
(one such reply per series is enough).
- [High] XXV receive is routed through the shared dmaengine RX callback,
axienet_dma_rx_cb()…
- [Medium] axienet_mac_select_pcs() is shared by all MAC types.
- [Medium] PG210's XXV MAC flags bad-FCS, oversize/length-error and
truncated frames only through rx_axis_tuser at end of packet, and…
- [Low] axienet_xxv_get_regs() reads every STAT_* status register on each
'ethtool -d'.
- [Low] axienet_ethtools_get_regs() always sets regs->version = 0 and
regs->len = sizeof(u32) * regs_n, then dispatches to…
- [Low] The patch builds xilinx_axienet_xxv.o into xilinx_emac under
CONFIG_XILINX_AXI_EMAC, adding 10G/25G XXV support to that option, but…
- [Low] In axienet_open(), axienet_device_reset() runs first.
--
net: xilinx: axienet: Add 10G/25G (XXV) ethernet support
This commit adds support for the Xilinx XXV 10G/25G Ethernet MAC to the
axienet driver. The XXV-specific code goes into new xilinx_axienet_xxv.c/.h
files and is described by an axienet_10g25g_config. The shared paths use
the callbacks from that config, and axienet_mac_select_pcs() now also
handles 10GBASER and 25GBASER.
> diff --git a/drivers/net/ethernet/xilinx/Makefile b/drivers/net/ethernet/xilinx/Makefile
> index 7d7dc1771423e..db90e691fec58 100644
> --- a/drivers/net/ethernet/xilinx/Makefile
> +++ b/drivers/net/ethernet/xilinx/Makefile
> @@ -6,5 +6,6 @@
> ll_temac-objs := ll_temac_main.o ll_temac_mdio.o
> obj-$(CONFIG_XILINX_LL_TEMAC) += ll_temac.o
> obj-$(CONFIG_XILINX_EMACLITE) += xilinx_emaclite.o
> -xilinx_emac-objs := xilinx_axienet_main.o xilinx_axienet_mdio.o
> +xilinx_emac-objs := xilinx_axienet_main.o xilinx_axienet_mdio.o \
> + xilinx_axienet_xxv.o
> obj-$(CONFIG_XILINX_AXI_EMAC) += xilinx_emac.o
[Severity: Low]
This isn't a bug, but should the XILINX_AXI_EMAC Kconfig entry be updated
too? XXV support now builds under this option, but the prompt still reads:
tristate "Xilinx 10/100/1000 AXI Ethernet support"
The help text also still says "This driver supports the 10/100/1000
Ethernet from Xilinx...". It is unchanged at the end of the series, so
nothing points someone configuring an XXV-based board at this option.
> diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> index a553150faa89b..47c84014921c8 100644
> --- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> +++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> @@ -801,6 +801,12 @@ static int axienet_1g_mac_init(struct net_device *ndev)
> * Ethernet core. No separate hardware reset is done for the Axi Ethernet
> * core.
> *
> + * It also recomputes @max_frm_size from the current MTU and runs the
> + * MAC-specific mac_init() callback, so it is the point at which length-related
> + * MAC configuration (such as the XXV RX max-packet-length) is reprogrammed.
> + * Because it is reached from the open path, a new MTU takes effect on the next
> + * interface up.
> + *
> * Return: 0 on success or a negative error number otherwise.
> */
> static int axienet_device_reset(struct net_device *ndev)
[Severity: Low]
Is the ordering in axienet_open() safe for XXV? The XXV receiver is
enabled before any RX DMA buffer is posted:
axienet_open()
axienet_device_reset()
axienet_setoptions(ndev, lp->options) <- sets XXV_RCW1_RX_MASK
phylink_start()
axienet_init_dmaengine() <- RX channel and buffers
PG210 describes the XXV RX AXI4-Stream as having no tready backpressure.
If a frame arrives before the S2MM channel accepts data, could the first
RX buffer capture only the tail of that frame (ending at tlast) and pass
it up as a frame?
On the 1G core, its internal RX FIFO made this ordering safe. With XXV it
depends on the FPGA logic between the MAC and the DMA.
[ ... ]
> @@ -2788,7 +2789,9 @@ static struct phylink_pcs *axienet_mac_select_pcs(struct phylink_config *config,
> struct axienet_local *lp = netdev_priv(ndev);
>
> if (interface == PHY_INTERFACE_MODE_1000BASEX ||
> - interface == PHY_INTERFACE_MODE_SGMII)
> + interface == PHY_INTERFACE_MODE_SGMII ||
> + interface == PHY_INTERFACE_MODE_10GBASER ||
> + interface == PHY_INTERFACE_MODE_25GBASER)
> return &lp->pcs;
>
> return NULL;
[Severity: Medium]
Can this hand phylink a PCS with NULL ops on a 1G MAC?
axienet_mac_select_pcs() is shared by all MAC types. On the 1G config,
axienet_probe() sets lp->pcs.ops only for SGMII/1000BASEX, and
axienet_1g_config has no internal_pcs.
axienet_1g_phylink_set_caps() still does this for any phy_mode:
__set_bit(lp->phy_mode, config->supported_interfaces);
So a 1G compatible node with phy-mode = "10gbase-r" or "25gbase-r" is not
rejected at probe.
With fixed-link, probe still succeeds. phylink_validate() prints
"uninitialised PCS" with a dump_stack() and returns -EINVAL, but both
phylink_create() and phylink_parse_fixedlink() ignore that result. Then
on ifup:
axienet_open()
phylink_start()
phylink_major_config()
pl->pcs = &lp->pcs;
phylink_pcs_pre_config(pl->pcs, ...)
if (pcs && pcs->ops->pcs_pre_config) <- pcs->ops is NULL
Before this patch, these modes returned NULL here. Should &lp->pcs only be
returned when lp->pcs.ops is set?
[ ... ]
> diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
> new file mode 100644
> index 0000000000000..3c204dbbf6aec
> --- /dev/null
> +++ b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
> @@ -0,0 +1,351 @@
[ ... ]
> +static int axienet_xxv_mac_init(struct net_device *ndev)
> +{
> + struct axienet_local *lp = netdev_priv(ndev);
> + int ret;
> +
> + ret = axienet_xxv_poll_link_ready(ndev);
> + if (ret)
> + return ret;
> +
> + /* Bound the MAC's RX max packet length to the posted buffer size so an
> + * oversized frame is truncated to a single buffer instead of spilling
> + * across several DMA descriptors (PG210 CONFIGURATION_RX_MTU).
> + */
> + axienet_iow(lp, XXV_JUM_OFFSET,
> + FIELD_PREP(XXV_RX_MAX_LEN_MASK, lp->max_frm_size) |
> + XXV_RX_MIN_LEN_DEFAULT);
[Severity: High]
Where does XXV get the RX frame length from? axienet_10g25g_config does
not set non_dmaengine, so every XXV RX completion goes through the shared
dmaengine callback:
drivers/net/ethernet/xilinx/xilinx_axienet_main.c:axienet_dma_rx_cb() {
...
if (IS_ERR(app_metadata)) {
if (net_ratelimit())
netdev_err(lp->ndev, "Failed to get RX metadata pointer\n");
dev_kfree_skb_any(skb);
lp->ndev->stats.rx_dropped++;
goto rx_submit;
}
/* TODO: Derive app word index programmatically */
rx_len = (app_metadata[LEN_APP] & 0xFFFF);
skb_put(skb, rx_len);
...
}
app_metadata[LEN_APP] is APP4. Only the 1G AXI Ethernet IP fills that
word, through its RX status stream. The PG210 MAC has no status stream,
and nothing in the XXV config or binding provides one.
In xilinx_dma.c, software never writes the S2MM app words: hw->app is
copied only for DMA_MEM_TO_DEV, and segments are zeroed. For MCDMA, which
the XXV binding example uses, xilinx_dma_get_metadata_ptr() returns
&seg->hw.s2mm_status for DEV_TO_MEM, so index 4 lands on app[2] and not
on a length field.
What happens then depends on the DMA setup:
- Without xlnx,axistream-connected on the DMA, no metadata_ops are
attached. Every frame is dropped with "Failed to get RX metadata
pointer".
- With it, rx_len comes from a zeroed or unrelated word. Each frame
becomes a zero-length skb and rx_bytes stays at 0.
- If that word ever holds a stale non-zero value, rx_len is not checked
against lp->max_frm_size. Anything above the skb tailroom (up to
0xFFFF) reaches skb_put() and hits skb_over_panic().
The DMA does report the real transfer count, via
xilinx_dma_get_residue() into result->residue, but the callback ignores
it.
The comment above says CONFIGURATION_RX_MTU keeps frames within one
buffer. However, the length passed to skb_put() never comes from the MAC.
The last patch in the series ("net: xilinx: axienet: Add statistics
support for XXV ethernet") does not change axienet_dma_rx_cb() either.
Does XXV receive work on any of the documented DMA configurations?
> +
> + lp->options |= XAE_OPTION_FCS_STRIP;
> + lp->options |= XAE_OPTION_FCS_INSERT;
> +
> + return 0;
> +}
[Severity: Medium]
How are errored frames dropped on XXV? Per PG210, the XXV MAC reports
bad-FCS, length-error and truncated frames only through rx_axis_tuser at
end of packet. It still forwards those frames on the stream.
AXI DMA S2MM does not record tuser. axienet_dma_rx_cb() also checks no
per-frame error status before it calls __netif_rx().
With XAE_OPTION_FCS_STRIP set here, the FCS is gone before software
could check it. The oversized frames that ctl_rx_max_packet_len
truncates, per the comment above, would also go up as valid packets.
The 1G core drops errored frames in its own RX FIFO and raises
XAE_INT_RXRJECT, so the shared datapath never needed this check.
Does the FPGA design need a tuser-aware drop stage between the MAC and
the DMA? If so, should that be documented or detected?
[ ... ]
> +static void axienet_xxv_get_regs(struct axienet_local *lp, u32 *data)
> +{
> + bool has_gt_status = axienet_xxv_ip_has_gt_status(lp->xxv_ip_version);
> + unsigned int i;
> +
> + for (i = 0; i < ARRAY_SIZE(axienet_xxv_reg_dump_offsets); i++) {
> + u32 off = axienet_xxv_reg_dump_offsets[i];
> +
> + /* The GT registers only exist on the IP versions that report
> + * reset-done. Leave their slots zero on the others so the
> + * dump keeps a fixed layout.
> + */
> + if (!has_gt_status && (off == XXV_GT_CTRL_OFFSET ||
> + off == XXV_STAT_GT_OFFSET))
> + continue;
> +
> + data[i] = axienet_ior(lp, off);
[Severity: Low]
Can an ethtool -d dump interfere with link reporting? This reads every
STAT_* register, including XXV_STAT_RX_STATUS_OFFSET. PG210 describes
that register's bit 0 as latched and cleared on read.
axienet_xxv_pcs_get_state() uses the same bit for the link:
state->link = axienet_ior(lp, XXV_STAT_RX_STATUS_OFFSET) &
XXV_STAT_RX_STATUS_MASK;
lp->pcs.poll is true for internal_pcs. A dump taken between two phylink
polls could therefore clear a latched link-down event, and phylink would
never see the flap. STAT_RX_BLOCK_LOCK, STAT_TX_STATUS and the AN/LT/FEC
status registers are read the same way.
The skip logic is also uneven. GT_CTRL/STAT_GT are skipped on IP older
than v3.2. But STAT_CORE_SPEED is read unconditionally, and the commit
message says it exists only on 64-bit AXI4-Stream builds. The same goes
for the build-optional RS-FEC/AN/LT/1588 registers.
> + }
> +}
[Severity: Low]
Should the dump version tell the two layouts apart? The caller reports
version 0 for every MAC type:
drivers/net/ethernet/xilinx/xilinx_axienet_main.c:axienet_ethtools_get_regs() {
...
len = sizeof(u32) * lp->axienet_config->regs_n;
regs->version = 0;
regs->len = len;
...
}
After this patch, the driver produces both the 1G layout (AXIENET_REGS_N
words with the DMA registers at fixed indices) and this 46-word PG210
layout, both under version 0. That leaves userspace decoders no way to
tell which format they have.
--
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261114182947.2908628-1-suraj.gupta2%40amd.com
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH net-next v5 6/6] net: xilinx: axienet: Add statistics support for XXV ethernet
2026-11-14 18:29 ` [PATCH net-next v5 6/6] net: xilinx: axienet: Add statistics support for XXV ethernet Suraj Gupta
@ 2026-11-16 4:12 ` netdev-bot+sashiko
0 siblings, 0 replies; 12+ messages in thread
From: netdev-bot+sashiko @ 2026-11-16 4:12 UTC (permalink / raw)
To: suraj.gupta2
Cc: radhey.shyam.pandey, andrew+netdev, davem, edumazet, kuba, pabeni,
robh, krzk+dt, conor+dt, michal.simek, linux, netdev, devicetree,
linux-arm-kernel, linux-kernel
Thank you for your contribution! Sashiko AI review found 6 potential
issue(s) to consider.
Critical: 0 · High: 0 · Medium: 2 · Low: 4
If you believe none of the issues are legitimate, please reply to each
issue of Medium severity or higher and explain why the AI is wrong.
If any of them are legitimate and you plan to respin, please let patchwork
know by including "pw-bot: cr" as a separate line at the end of your reply
(one such reply per series is enough).
- [Medium] None of the XXV statistics readers in
drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c latch the clear-on-…
- [Medium] The new XXV ndo_get_stats64 reader, axienet_xxv_get_stats64(),
runs read_seqcount_begin()/read_seqcount_retry() on…
- [Low] The kernel-doc of axienet_xxv_get_rmon_stats() (xxv.c:585-586)
says "XXV exposes aggregate RMON counters only, not per-bin histogram…
- [Low] axienet_xxv_get_stats64() treats XXV_STAT_RX_OVERSIZE as an IEEE
802.3 frame-too-long error and folds it into rx_length_errors.
- [Low] The patch updates the @hw_stats_seqcount kernel-doc in struct
axienet_local for XXV but leaves @stats_work saying "Work for reading…
- [Low] axienet_stop() writes TICK_REG for the final XXV statistics
snapshot (via axienet_xxv_stats_update) right after…
--
net: xilinx: axienet: Add statistics support for XXV ethernet
This patch adds support for the clear-on-tick XXV (PG210) statistics
block. The counters are accumulated in software by the periodic stats
work and by a final snapshot in axienet_stop(). The result feeds
ndo_get_stats64() and the ethtool -S, pause, MAC and RMON callbacks,
gated on the "xlnx,has-statistics-counters" devicetree property.
> diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet.h b/drivers/net/ethernet/xilinx/xilinx_axienet.h
> index 7468fed1416ef..10cb73bc460d2 100644
> --- a/drivers/net/ethernet/xilinx/xilinx_axienet.h
> +++ b/drivers/net/ethernet/xilinx/xilinx_axienet.h
> @@ -520,7 +520,7 @@ struct skbuf_dma_descriptor {
> * @reset_in_progress: Set while we are performing a reset and statistics
> * counters may be invalid
> * @hw_stats_seqcount: Sequence counter for @hw_stat_base, @hw_last_counter,
> - * and @reset_in_progress.
> + * @reset_in_progress and, on XXV, @xxv_stat_base.
> * @stats_lock: Lock for @hw_stats_seqcount
> * @stats_work: Work for reading the hardware statistics counters often enough
> * to catch overflows.
[Severity: Low]
This isn't a bug, but should the @stats_work description be updated as
well? On XXV this work is now the only latch for the clear-on-tick
counters. The new comment in axienet_refresh_stats() says:
This work is
what latches them between queries.
The @stats_update kernel-doc in struct axienet_config has not changed
either:
* @stats_update: Callback to latch/accumulate the periodic MAC counters, or
* NULL if the MAC has no periodic statistics counters
The new comments in axienet_refresh_stats() and axienet_stop() say
something different: a config installs the hook unconditionally, and
XAE_FEATURE_STATS decides whether the counters exist.
Could these two descriptions be updated to match?
[ ... ]
> diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> index 260755fbe465c..a046290522f4d 100644
> --- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> +++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
[ ... ]
> @@ -1884,6 +1903,17 @@ static int axienet_stop(struct net_device *ndev)
> axienet_setoptions(ndev, lp->options &
> ~(XAE_OPTION_TXEN | XAE_OPTION_RXEN));
>
> + /* Final snapshot after RX/TX is disabled, so no frames are missed.
> + * Gate on XAE_FEATURE_STATS rather than the stats_update hook: a config
> + * always installs the hook, but the counters are optional in both MACs,
> + * so only probe_init() knows whether they are there.
> + */
> + if (lp->features & XAE_FEATURE_STATS && lp->axienet_config->stats_update) {
> + mutex_lock(&lp->stats_lock);
> + lp->axienet_config->stats_update(lp);
> + mutex_unlock(&lp->stats_lock);
> + }
[Severity: Low]
Is "no frames are missed" guaranteed here? axienet_setoptions() only
does a few posted read-modify-write MMIO accesses to clear ctl_tx_enable
and ctl_rx_enable. Nothing waits for the MAC to go idle before
axienet_xxv_stats_update() writes TICK_REG.
Suppose the MAC finishes a frame that is already in flight after the
enable is cleared. A jumbo frame takes about 7.7 us at 10G and about
3 us at 25G. Could that frame be counted into the internal accumulator
after this final tick?
On the next open, axienet_xxv_mac_init() does:
if (lp->features & XAE_FEATURE_STATS)
axienet_iow(lp, XXV_TICKREG_OFFSET, XXV_TICKREG_STATEN_MASK);
This latches the late count into the readable registers. The first
stats_update tick then overwrites those registers without reading them.
Would up to one frame per direction per stop then never reach
xxv_stat_base?
This depends on how PG210 behaves when the enable is deasserted, which
could not be checked from the tree.
> diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
> index 3c204dbbf6aec..83782bfbce4ce 100644
> --- a/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
> +++ b/drivers/net/ethernet/xilinx/xilinx_axienet_xxv.c
[ ... ]
> @@ -335,6 +411,204 @@ static int axienet_10g25g_clk_init(struct axienet_local *lp)
> return 0;
> }
>
> +/**
> + * axienet_xxv_stats_update - Latch and accumulate XXV hardware counters
[ ... ]
> + * Readers copy @xxv_stat_base and do not tick. ndo_get_stats64() may run in
> + * atomic context, so it cannot take @stats_lock and tick itself. The periodic
> + * statistics work, and the final snapshot in axienet_stop(), are the latches.
> + */
> +static void axienet_xxv_stats_update(struct axienet_local *lp)
> +{
[ ... ]
> + write_seqcount_begin(&lp->hw_stats_seqcount);
> + for (stat = 0; stat < XXV_STAT_COUNT; stat++)
> + lp->xxv_stat_base[stat] += counter[stat];
> + write_seqcount_end(&lp->hw_stats_seqcount);
> +}
> +
> +/**
> + * axienet_xxv_get_stats64 - Fill rtnl_link_stats64 from XXV counters
> + * @lp: Pointer to the axienet_local structure
> + * @stats: Output rtnl_link_stats64 structure to populate
> + */
> +static void axienet_xxv_get_stats64(struct axienet_local *lp,
> + struct rtnl_link_stats64 *stats)
> +{
> + unsigned int start;
> + u64 tx_hw_errors;
> +
> + do {
> + start = read_seqcount_begin(&lp->hw_stats_seqcount);
[Severity: Medium]
Can this reader sleep on PREEMPT_RT? hw_stats_seqcount is a
seqcount_mutex_t bound to stats_lock in axienet_probe():
seqcount_mutex_init(&lp->hw_stats_seqcount, &lp->stats_lock);
With CONFIG_PREEMPT_RT, read_seqcount_begin() on a seqcount_mutex_t
calls mutex_lock(s->lock) and then mutex_unlock() when it sees an odd
sequence. On RT the writer also stays preemptible between
write_seqcount_begin() and write_seqcount_end().
The write section in axienet_xxv_stats_update() runs with stats_lock
held, from the stats work or from axienet_stop(). A reader that
overlaps it would block on stats_lock. ndo_get_stats64 can be called
under rcu_read_lock(), for example from /proc/net/dev, where
dev_seq_start() takes rcu_read_lock():
dev_seq_show()
dev_seq_printf_stats()
dev_get_stats()
axienet_get_stats64()
axienet_xxv_get_stats64()
read_seqcount_begin()
mutex_lock(&lp->stats_lock)
Bonding also calls it under its stats spinlock plus RCU.
Would this trigger a "sleeping function called from invalid context"
splat on RT kernels? The kernel-doc above says ndo_get_stats64() cannot
take @stats_lock, but on RT the seqcount takes it anyway.
The 1G reader axienet_1g_get_stats64() already uses the same pattern.
This patch adds a new atomic-context reader whose design assumes that
readers never take stats_lock.
[ ... ]
> +static void axienet_xxv_get_ethtool_stats(struct axienet_local *lp, u64 *data)
> +{
> + unsigned int i, start;
> +
> + do {
> + start = read_seqcount_begin(&lp->hw_stats_seqcount);
> + for (i = 0; i < ARRAY_SIZE(axienet_xxv_priv_stats); i++)
> + data[i] = lp->xxv_stat_base[axienet_xxv_priv_stats[i]];
> + } while (read_seqcount_retry(&lp->hw_stats_seqcount, start));
> +}
[Severity: Medium]
Will ethtool -S on XXV return values that are at least one work
interval old? This reader only copies lp->xxv_stat_base. So do
axienet_xxv_get_pause_stats(), axienet_xxv_get_eth_mac_stats(),
axienet_xxv_get_rmon_stats() and axienet_xxv_get_stats64().
The base only changes in axienet_xxv_stats_update(), which is called
from axienet_refresh_stats() and axienet_stop(). The work is set up
with INIT_DEFERRABLE_WORK() and re-armed every 13 * HZ, so on an idle
CPU the lag can go well past 13 seconds. After a traffic burst, the
ethtool -S totals and the pause, eth-mac and RMON groups would not
change until the work runs.
The commit message explains why ndo_get_stats64() cannot tick. The
ethtool callbacks, though, run in process context under RTNL.
Could they take stats_lock and call stats_update so they return
current values?
For comparison, the 1G readers go through axienet_stat(), which adds
the live register to the software base:
counter = axienet_ior(lp, XAE_STATS_OFFSET + stat * 8);
return lp->hw_stat_base[stat] + (counter - lp->hw_last_counter[stat]);
so the 1G ethtool and stats64 values are current.
In ndo_get_stats64 the software rx/tx packet and byte counters are live,
while the XXV hardware error and multicast counters come from the
cached base. A single snapshot therefore mixes values from different
points in time.
[ ... ]
> + mac_stats->InRangeLengthErrors =
> + lp->xxv_stat_base[XXV_STAT_RX_INRANGEERR];
> + } while (read_seqcount_retry(&lp->hw_stats_seqcount, start));
> +}
[Severity: Low]
Should FrameTooLongErrors be filled in here too? axienet_xxv_get_stats64()
treats XXV_STAT_RX_OVERSIZE as a frame-too-long error and folds it into
rx_length_errors:
/* Both in-range length errors and frame-too-long (oversize)
* frames are IEEE 802.3 length errors, so fold them together
axienet_xxv_get_rmon_stats() also reports it as oversize_pkts.
struct ethtool_eth_mac_stats has a FrameTooLongErrors field, but it is
never set here. As a result, ethtool -S <dev> --groups eth-mac shows it
as unsupported, even though the driver already accumulates the counter.
> +
> +/**
> + * axienet_xxv_get_rmon_stats - Fill RMON statistics and histogram ranges
> + * @lp: Pointer to the axienet_local structure
> + * @rmon_stats: Output ethtool_rmon_stats structure to populate
> + * @ranges: Set to NULL; XXV exposes aggregate RMON counters only, not
> + * per-bin histogram ranges.
> + */
[ ... ]
> + /* XXV currently exposes only aggregate RMON counters, not per-bin
> + * histogram buckets. Keep ranges NULL until histogram bins are wired.
> + */
> + *ranges = NULL;
> +}
[Severity: Low]
Is the kernel-doc accurate about the hardware? The register map this
patch adds in xilinx_axienet_xxv.h has a gap here:
#define XXV_STAT_RX_TOTAL_GOOD_BYTES_OFFSET 0x00000820
#define XXV_STAT_RX_UNDERSIZE_OFFSET 0x00000898
That gap is 14 skipped 8-byte counter slots, 0x828 to 0x890. There is a
similar gap here:
#define XXV_STAT_TX_TOTAL_GOOD_BYTES_OFFSET 0x00000718
#define XXV_STAT_TX_BAD_FCS_OFFSET 0x000007B8
These look like the PG210 per-size-bin packet counters (64, 65-127, ...
8192-9215 bytes, plus large and small). This could not be checked
against the PG210 datasheet from the tree. The inline comment
("currently ... until histogram bins are wired") reads like a driver
limitation rather than missing hardware.
The commit message says the RMON statistics come from this path. It
does not mention that the RX and TX histograms are left out.
Could the kernel-doc and the commit message say that the histogram
counters exist but are not wired up yet?
--
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261114182947.2908628-1-suraj.gupta2%40amd.com
^ permalink raw reply [flat|nested] 12+ messages in thread
end of thread, other threads:[~2026-11-16 4:13 UTC | newest]
Thread overview: 12+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-14 18:29 [PATCH net-next v5 0/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support Suraj Gupta
2026-11-14 18:29 ` [PATCH net-next v5 1/6] net: xilinx: axienet: Introduce axienet_config for MAC-specific ops Suraj Gupta
2026-11-16 4:12 ` netdev-bot+sashiko
2026-11-14 18:29 ` [PATCH net-next v5 2/6] dt-bindings: net: xlnx,xxv-ethernet: Add Xilinx XXV 10G/25G Ethernet Suraj Gupta
2026-11-16 4:12 ` netdev-bot+sashiko
2026-11-14 18:29 ` [PATCH net-next v5 3/6] net: xilinx: axienet: Propagate device reset failure from ndo_open Suraj Gupta
2026-11-16 4:12 ` netdev-bot+sashiko
2026-11-14 18:29 ` [PATCH net-next v5 4/6] net: xilinx: axienet: Add 10G/25G (XXV) ethernet support Suraj Gupta
2026-11-16 4:12 ` netdev-bot+sashiko
2026-11-14 18:29 ` [PATCH net-next v5 5/6] net: xilinx: axienet: Dispatch statistics through axienet_config ops Suraj Gupta
2026-11-14 18:29 ` [PATCH net-next v5 6/6] net: xilinx: axienet: Add statistics support for XXV ethernet Suraj Gupta
2026-11-16 4:12 ` netdev-bot+sashiko
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox