[PATCH net 0/4] net: sparx5: misc fixes for sparx5 and lan969x

[PATCH net 0/4] net: sparx5: misc fixes for sparx5 and lan969x

[Daniel Machon]

This series fixes various issues in the sparx5 driver, which also
serves lan969x.

Details are in the individual commit descriptions.

Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
---
Daniel Machon (4):
      net: sparx5: defer VCAP debugfs creation until after netdev registration
      net: sparx5: fix sleep in atomic context in MAC table access
      net: sparx5: fix wrong chip ids for TSN SKUs
      net: sparx5: configure serdes for 1000BASE-X in sparx5_port_init()

 drivers/net/ethernet/microchip/sparx5/Makefile     |  3 ++-
 .../net/ethernet/microchip/sparx5/sparx5_debugfs.c | 26 ++++++++++++++++++++++
 .../ethernet/microchip/sparx5/sparx5_mactable.c    | 24 ++++++++++----------
 .../net/ethernet/microchip/sparx5/sparx5_main.c    |  4 ++--
 .../net/ethernet/microchip/sparx5/sparx5_main.h    | 19 +++++++++++-----
 .../net/ethernet/microchip/sparx5/sparx5_port.c    |  3 ++-
 .../ethernet/microchip/sparx5/sparx5_vcap_impl.c   |  6 -----
 7 files changed, 57 insertions(+), 28 deletions(-)
---
base-commit: 98878ed91b68a3150126fccef125ee7b1bb86ab2
change-id: 20260428-misc-fixes-sparx5-lan969x-bc2961a570fb

Best regards,
-- 
Daniel Machon <daniel.machon@microchip.com>

[PATCH net 1/4] net: sparx5: defer VCAP debugfs creation until after netdev registration

[Daniel Machon]

Commit 3a95973e7c79 ("net: sparx5: move VCAP initialization to probe")
moved sparx5_vcap_init() ahead of sparx5_register_netdevs() in probe.
The VCAP init path ends by calling vcap_port_debugfs() for every port,
which uses netdev_name(ndev) as the debugfs file name. At that point
the netdevs have only been allocated, not registered, so dev->name
still holds the "eth%d" template and netdev_name() returns
"(unnamed net_device)". Every port tries to create the same file under
vcaps/, producing a flood of warnings at boot:

  debugfs: '(unnamed net_device)' already exists in 'vcaps'
  debugfs: '(unnamed net_device)' already exists in 'vcaps'
  ...

Move the debugfs setup into a new sparx5_debugfs() helper in
sparx5_debugfs.c, invoked after sparx5_register_notifier_blocks()
succeeds so the netdev names are finalized. sparx5_vcap_init() now
only deals with VCAP state. The sparx5/ debugfs root is created in
the new helper as well.

Fixes: 3a95973e7c79 ("net: sparx5: move VCAP initialization to probe")
Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
---
 drivers/net/ethernet/microchip/sparx5/Makefile     |  3 ++-
 .../net/ethernet/microchip/sparx5/sparx5_debugfs.c | 26 ++++++++++++++++++++++
 .../net/ethernet/microchip/sparx5/sparx5_main.c    |  4 ++--
 .../net/ethernet/microchip/sparx5/sparx5_main.h    |  7 ++++++
 .../ethernet/microchip/sparx5/sparx5_vcap_impl.c   |  6 -----
 5 files changed, 37 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/microchip/sparx5/Makefile b/drivers/net/ethernet/microchip/sparx5/Makefile
index d447f9e84d92..eb5c81527f41 100644
--- a/drivers/net/ethernet/microchip/sparx5/Makefile
+++ b/drivers/net/ethernet/microchip/sparx5/Makefile
@@ -14,7 +14,8 @@ sparx5-switch-y  := sparx5_main.o sparx5_packet.o \
  sparx5_psfp.o sparx5_mirror.o sparx5_regs.o
 
 sparx5-switch-$(CONFIG_SPARX5_DCB) += sparx5_dcb.o
-sparx5-switch-$(CONFIG_DEBUG_FS) += sparx5_vcap_debugfs.o
+sparx5-switch-$(CONFIG_DEBUG_FS) += sparx5_vcap_debugfs.o \
+				    sparx5_debugfs.o
 
 sparx5-switch-$(CONFIG_LAN969X_SWITCH) += lan969x/lan969x_regs.o \
 					  lan969x/lan969x.o \
diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_debugfs.c b/drivers/net/ethernet/microchip/sparx5/sparx5_debugfs.c
new file mode 100644
index 000000000000..f6cb1eeaab80
--- /dev/null
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_debugfs.c
@@ -0,0 +1,26 @@
+// SPDX-License-Identifier: GPL-2.0+
+/* Microchip Sparx5 Switch driver debug filesystem support
+ *
+ * Copyright (c) 2026 Microchip Technology Inc. and its subsidiaries.
+ */
+
+#include <linux/debugfs.h>
+
+#include "sparx5_main.h"
+#include "vcap_api_debugfs.h"
+
+void sparx5_debugfs(struct sparx5 *sparx5)
+{
+	const struct sparx5_consts *consts = sparx5->data->consts;
+	struct vcap_control *ctrl = sparx5->vcap_ctrl;
+	struct dentry *dir;
+	int idx;
+
+	sparx5->debugfs_root = debugfs_create_dir("sparx5", NULL);
+
+	dir = vcap_debugfs(sparx5->dev, sparx5->debugfs_root, ctrl);
+	for (idx = 0; idx < consts->n_ports; ++idx)
+		if (sparx5->ports[idx])
+			vcap_port_debugfs(sparx5->dev, dir, ctrl,
+					  sparx5->ports[idx]->ndev);
+}
diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_main.c b/drivers/net/ethernet/microchip/sparx5/sparx5_main.c
index dad713e9ddd5..bec07560e6fe 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_main.c
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_main.c
@@ -820,8 +820,6 @@ static int mchp_sparx5_probe(struct platform_device *pdev)
 	/* Default values, some from DT */
 	sparx5->coreclock = SPX5_CORE_CLOCK_DEFAULT;
 
-	sparx5->debugfs_root = debugfs_create_dir("sparx5", NULL);
-
 	ports = of_get_child_by_name(np, "ethernet-ports");
 	if (!ports) {
 		dev_err(sparx5->dev, "no ethernet-ports child node found\n");
@@ -1000,6 +998,8 @@ static int mchp_sparx5_probe(struct platform_device *pdev)
 		goto cleanup_netdevs;
 	}
 
+	sparx5_debugfs(sparx5);
+
 	goto cleanup_config;
 
 cleanup_netdevs:
diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_main.h b/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
index 6a745bb71b5c..d5e6644ff124 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
@@ -565,6 +565,13 @@ void sparx5_get_hwtimestamp(struct sparx5 *sparx5,
 int sparx5_vcap_init(struct sparx5 *sparx5);
 void sparx5_vcap_deinit(struct sparx5 *sparx5);
 
+/* sparx5_debugfs.c */
+#if defined(CONFIG_DEBUG_FS)
+void sparx5_debugfs(struct sparx5 *sparx5);
+#else
+static inline void sparx5_debugfs(struct sparx5 *sparx5) {}
+#endif
+
 /* sparx5_pgid.c */
 enum sparx5_pgid_type {
 	SPX5_PGID_FREE,
diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_vcap_impl.c b/drivers/net/ethernet/microchip/sparx5/sparx5_vcap_impl.c
index 95b93e46a41d..dd446b3a9f20 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_vcap_impl.c
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_vcap_impl.c
@@ -2035,7 +2035,6 @@ int sparx5_vcap_init(struct sparx5 *sparx5)
 	const struct sparx5_vcap_inst *cfg;
 	struct vcap_control *ctrl;
 	struct vcap_admin *admin;
-	struct dentry *dir;
 	int err = 0, idx;
 
 	/* Create a VCAP control instance that owns the platform specific VCAP
@@ -2074,11 +2073,6 @@ int sparx5_vcap_init(struct sparx5 *sparx5)
 			sparx5_vcap_port_key_selection(sparx5, admin);
 		list_add_tail(&admin->list, &ctrl->list);
 	}
-	dir = vcap_debugfs(sparx5->dev, sparx5->debugfs_root, ctrl);
-	for (idx = 0; idx < consts->n_ports; ++idx)
-		if (sparx5->ports[idx])
-			vcap_port_debugfs(sparx5->dev, dir, ctrl,
-					  sparx5->ports[idx]->ndev);
 
 	return err;
 }

-- 
2.34.1

[PATCH net 2/4] net: sparx5: fix sleep in atomic context in MAC table access

[Daniel Machon]

sparx5_mact_learn() is called from .ndo_set_rx_mode with
netif_addr_lock_bh held, but takes sparx5->lock which is a mutex.
A mutex may block, which is not allowed from atomic context.

Convert sparx5->lock to a spinlock, switch the hardware completion poll
to readx_poll_timeout_atomic(), and use spin_lock_bh() since no caller
runs in hardirq context.

Observed with CONFIG_PROVE_LOCKING, CONFIG_DEBUG_SPINLOCK,
CONFIG_DEBUG_MUTEXES and CONFIG_DEBUG_ATOMIC_SLEEP enabled:

  BUG: sleeping function called from invalid context at kernel/locking/mutex.c:591
  in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 217, name: ip
  preempt_count: 201, expected: 0
  Call trace:
   __might_resched+0x144/0x248
   __might_sleep+0x48/0x7c
   __mutex_lock+0x74/0x850
   mutex_lock_nested+0x24/0x30
   sparx5_mact_learn+0x78/0x100
   sparx5_mc_sync+0x40/0x54
   __hw_addr_sync_dev+0xc4/0x170
   sparx5_set_rx_mode+0x4c/0x58
   __dev_set_rx_mode+0x64/0xa4
   __dev_open+0x1ec/0x26c

Fixes: b37a1bae742f ("net: sparx5: add mactable support")
Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
---
 .../ethernet/microchip/sparx5/sparx5_mactable.c    | 24 +++++++++++-----------
 .../net/ethernet/microchip/sparx5/sparx5_main.h    |  2 +-
 2 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_mactable.c b/drivers/net/ethernet/microchip/sparx5/sparx5_mactable.c
index 2bf9c5f64151..0797cfa32916 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_mactable.c
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_mactable.c
@@ -50,7 +50,7 @@ static int sparx5_mact_wait_for_completion(struct sparx5 *sparx5)
 {
 	u32 val;
 
-	return readx_poll_timeout(sparx5_mact_get_status,
+	return readx_poll_timeout_atomic(sparx5_mact_get_status,
 		sparx5, val,
 		LRN_COMMON_ACCESS_CTRL_MAC_TABLE_ACCESS_SHOT_GET(val) == 0,
 		TABLE_UPDATE_SLEEP_US, TABLE_UPDATE_TIMEOUT_US);
@@ -92,7 +92,7 @@ int sparx5_mact_learn(struct sparx5 *sparx5, int pgid,
 		addr = pgid - consts->n_ports;
 	}
 
-	mutex_lock(&sparx5->lock);
+	spin_lock_bh(&sparx5->lock);
 
 	sparx5_mact_select(sparx5, mac, vid);
 
@@ -111,7 +111,7 @@ int sparx5_mact_learn(struct sparx5 *sparx5, int pgid,
 
 	ret = sparx5_mact_wait_for_completion(sparx5);
 
-	mutex_unlock(&sparx5->lock);
+	spin_unlock_bh(&sparx5->lock);
 
 	return ret;
 }
@@ -164,7 +164,7 @@ bool sparx5_mact_getnext(struct sparx5 *sparx5,
 	u32 cfg2;
 	int ret;
 
-	mutex_lock(&sparx5->lock);
+	spin_lock_bh(&sparx5->lock);
 
 	sparx5_mact_select(sparx5, mac, *vid);
 
@@ -183,7 +183,7 @@ bool sparx5_mact_getnext(struct sparx5 *sparx5,
 			*pcfg2 = cfg2;
 	}
 
-	mutex_unlock(&sparx5->lock);
+	spin_unlock_bh(&sparx5->lock);
 
 	return ret == 0;
 }
@@ -194,7 +194,7 @@ int sparx5_mact_find(struct sparx5 *sparx5,
 	int ret;
 	u32 cfg2;
 
-	mutex_lock(&sparx5->lock);
+	spin_lock_bh(&sparx5->lock);
 
 	sparx5_mact_select(sparx5, mac, vid);
 
@@ -212,7 +212,7 @@ int sparx5_mact_find(struct sparx5 *sparx5,
 			ret = -ENOENT;
 	}
 
-	mutex_unlock(&sparx5->lock);
+	spin_unlock_bh(&sparx5->lock);
 
 	return ret;
 }
@@ -222,7 +222,7 @@ int sparx5_mact_forget(struct sparx5 *sparx5,
 {
 	int ret;
 
-	mutex_lock(&sparx5->lock);
+	spin_lock_bh(&sparx5->lock);
 
 	sparx5_mact_select(sparx5, mac, vid);
 
@@ -233,7 +233,7 @@ int sparx5_mact_forget(struct sparx5 *sparx5,
 
 	ret = sparx5_mact_wait_for_completion(sparx5);
 
-	mutex_unlock(&sparx5->lock);
+	spin_unlock_bh(&sparx5->lock);
 
 	return ret;
 }
@@ -440,7 +440,7 @@ static void sparx5_mact_pull_work(struct work_struct *work)
 	vid = 0;
 	memset(mac, 0, sizeof(mac));
 	do {
-		mutex_lock(&sparx5->lock);
+		spin_lock_bh(&sparx5->lock);
 		sparx5_mact_select(sparx5, mac, vid);
 		spx5_wr(LRN_SCAN_NEXT_CFG_SCAN_NEXT_UNTIL_FOUND_ENA_SET(1),
 			sparx5, LRN_SCAN_NEXT_CFG);
@@ -451,7 +451,7 @@ static void sparx5_mact_pull_work(struct work_struct *work)
 		ret = sparx5_mact_wait_for_completion(sparx5);
 		if (ret == 0)
 			ret = sparx5_mact_get(sparx5, mac, &vid, &cfg2);
-		mutex_unlock(&sparx5->lock);
+		spin_unlock_bh(&sparx5->lock);
 		if (ret == 0)
 			sparx5_mact_handle_entry(sparx5, mac, vid, cfg2);
 	} while (ret == 0);
@@ -493,7 +493,7 @@ int sparx5_mact_init(struct sparx5 *sparx5)
 {
 	char queue_name[32];
 
-	mutex_init(&sparx5->lock);
+	spin_lock_init(&sparx5->lock);
 
 	/*  Flush MAC table */
 	spx5_wr(LRN_COMMON_ACCESS_CTRL_CPU_ACCESS_CMD_SET(MAC_CMD_CLEAR_ALL) |
diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_main.h b/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
index d5e6644ff124..8ef4050ceecc 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
@@ -373,7 +373,7 @@ struct sparx5 {
 	u32 features;
 	void __iomem *regs[NUM_TARGETS];
 	int port_count;
-	struct mutex lock; /* MAC reg lock */
+	spinlock_t lock; /* MAC reg lock */
 	/* port structures are in net device */
 	struct sparx5_port *ports[SPX5_PORTS];
 	enum sparx5_core_clockfreq coreclock;

-- 
2.34.1

[PATCH net 3/4] net: sparx5: fix wrong chip ids for TSN SKUs

[Daniel Machon]

The TSN SKUs in enum spx5_target_chiptype have incorrect IDs:

  SPX5_TARGET_CT_7546TSN    = 0x47546,
  SPX5_TARGET_CT_7549TSN    = 0x47549,
  SPX5_TARGET_CT_7552TSN    = 0x47552,
  SPX5_TARGET_CT_7556TSN    = 0x47556,
  SPX5_TARGET_CT_7558TSN    = 0x47558,

The value read back from the chip is GCB_CHIP_ID_PART_ID, which is a
GENMASK(27, 12) field, i.e. at most 16 bits wide. It can never match
these IDs, so probing a TSN part fails with a "Target not supported"
error.

Fix the enum to use the actual 16-bit part IDs returned by the
hardware: 0x0546, 0x0549, 0x0552, 0x0556 and 0x0558.

Reported-by: Andrew Lunn <andrew@lunn.ch>
Fixes: 3cfa11bac9bb ("net: sparx5: add the basic sparx5 driver")
Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
---
 drivers/net/ethernet/microchip/sparx5/sparx5_main.h | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_main.h b/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
index 8ef4050ceecc..70ca707b5ce4 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_main.h
@@ -31,11 +31,11 @@ enum spx5_target_chiptype {
 	SPX5_TARGET_CT_7552       = 0x7552,  /* SparX-5-128 Enterprise */
 	SPX5_TARGET_CT_7556       = 0x7556,  /* SparX-5-160 Enterprise */
 	SPX5_TARGET_CT_7558       = 0x7558,  /* SparX-5-200 Enterprise */
-	SPX5_TARGET_CT_7546TSN    = 0x47546, /* SparX-5-64i Industrial */
-	SPX5_TARGET_CT_7549TSN    = 0x47549, /* SparX-5-90i Industrial */
-	SPX5_TARGET_CT_7552TSN    = 0x47552, /* SparX-5-128i Industrial */
-	SPX5_TARGET_CT_7556TSN    = 0x47556, /* SparX-5-160i Industrial */
-	SPX5_TARGET_CT_7558TSN    = 0x47558, /* SparX-5-200i Industrial */
+	SPX5_TARGET_CT_7546TSN    = 0x0546,  /* SparX-5-64i Industrial */
+	SPX5_TARGET_CT_7549TSN    = 0x0549,  /* SparX-5-90i Industrial */
+	SPX5_TARGET_CT_7552TSN    = 0x0552,  /* SparX-5-128i Industrial */
+	SPX5_TARGET_CT_7556TSN    = 0x0556,  /* SparX-5-160i Industrial */
+	SPX5_TARGET_CT_7558TSN    = 0x0558,  /* SparX-5-200i Industrial */
 	SPX5_TARGET_CT_LAN9694    = 0x9694,  /* lan969x-40 */
 	SPX5_TARGET_CT_LAN9691VAO = 0x9691,  /* lan969x-40-VAO */
 	SPX5_TARGET_CT_LAN9694TSN = 0x9695,  /* lan969x-40-TSN */

-- 
2.34.1

[PATCH net 4/4] net: sparx5: configure serdes for 1000BASE-X in sparx5_port_init()

[Daniel Machon]

sparx5_port_init() only invokes sparx5_serdes_set() and the associated
shadow-device enable and low-speed device switch for SGMII and QSGMII.
On any port with a high-speed primary device (DEV5G/DEV10G/DEV25G)
configured for 1000BASE-X the serdes is therefore left uninitialized,
the DEV2G5 shadow is never enabled, and the port stays pointed at its
high-speed device rather than the DEV2G5. The PCS1G block looks
healthy in isolation, but no frames reach the link partner.

Add 1000BASE-X to the check so the same three steps run.

Reported-by: Andrew Lunn <andrew@lunn.ch>
Fixes: 946e7fd5053a ("net: sparx5: add port module support")
Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
---
 drivers/net/ethernet/microchip/sparx5/sparx5_port.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_port.c b/drivers/net/ethernet/microchip/sparx5/sparx5_port.c
index 04bc8fffaf96..62c49893de3c 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_port.c
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_port.c
@@ -1128,7 +1128,8 @@ int sparx5_port_init(struct sparx5 *sparx5,
 		DEV2G5_PCS1G_SD_CFG(port->portno));
 
 	if (conf->portmode == PHY_INTERFACE_MODE_QSGMII ||
-	    conf->portmode == PHY_INTERFACE_MODE_SGMII) {
+	    conf->portmode == PHY_INTERFACE_MODE_SGMII ||
+	    conf->portmode == PHY_INTERFACE_MODE_1000BASEX) {
 		err = sparx5_serdes_set(sparx5, port, conf);
 		if (err)
 			return err;

-- 
2.34.1

Re: [PATCH net 2/4] net: sparx5: fix sleep in atomic context in MAC table access

[Sebastian Andrzej Siewior]

On 2026-05-04 16:43:43 [+0200], Daniel Machon wrote:
> index 2bf9c5f64151..0797cfa32916 100644
> --- a/drivers/net/ethernet/microchip/sparx5/sparx5_mactable.c
> +++ b/drivers/net/ethernet/microchip/sparx5/sparx5_mactable.c
> @@ -50,7 +50,7 @@ static int sparx5_mact_wait_for_completion(struct sparx5 *sparx5)
>  {
>  	u32 val;
>  
> -	return readx_poll_timeout(sparx5_mact_get_status,
> +	return readx_poll_timeout_atomic(sparx5_mact_get_status,

If you do _atomic, it becomes atomic. That means it does not sleep as in
TABLE_UPDATE_SLEEP_US for 10 us but spins via udelay(). The
TABLE_UPDATE_TIMEOUT_US is set to 100ms which _might_ be high. 
This is probably just nitpicking (given that there are other drivers
doing the same with a greater timeout (READL_TIMEOUT_US)) so feel free
to ignore it.

>  		sparx5, val,
>  		LRN_COMMON_ACCESS_CTRL_MAC_TABLE_ACCESS_SHOT_GET(val) == 0,
>  		TABLE_UPDATE_SLEEP_US, TABLE_UPDATE_TIMEOUT_US);

Sebastian

Re: [PATCH net 2/4] net: sparx5: fix sleep in atomic context in MAC table access

[Jakub Kicinski]

On Mon, 4 May 2026 16:43:43 +0200 Daniel Machon wrote:
> sparx5_mact_learn() is called from .ndo_set_rx_mode with
> netif_addr_lock_bh held, but takes sparx5->lock which is a mutex.
> A mutex may block, which is not allowed from atomic context.

Could you try to switch to the recently added ndo_set_rx_mode_async?
We added it exactly because most drivers have to sleep / wait for FW
to do something.