[PATCH v8 net-next 00/11] net: ethernet: am65-cpsw: Add mqprio, frame pre-emption & coalescing

[PATCH v8 net-next 00/11] net: ethernet: am65-cpsw: Add mqprio, frame pre-emption & coalescing

[Roger Quadros]

Hi,

This series adds mqprio qdisc offload in channel mode,
Frame Pre-emption MAC merge support and RX/TX coalesing
for AM65 CPSW driver.

In v8 following changes were made
- added a new selftest patch to use aggregate stats if
   pMAC stats are not supported.
- added a patch to rename TI_AM65_CPSW_TAS to TI_AM65_CPSW_QOS
   is added.
- added a patch to fix mac stats reporting. we only support
   aggregate stats.
- build issues if TI_AM65_CPSW_TAS is disabled is resolved.
- selftest patches are moved to the beginning of the series.

Changelog information in each patch file.

cheers,
-roger

Grygorii Strashko (2):
  net: ethernet: ti: am65-cpsw: add mqprio qdisc offload in channel mode
  net: ethernet: ti: am65-cpsw: add sw tx/rx irq coalescing based on
    hrtimers

Roger Quadros (7):
  net: ethernet: am65-cpsw: Build am65-cpsw-qos only if required
  net: ethernet: am65-cpsw: Rename TI_AM65_CPSW_TAS to TI_AM65_CPSW_QOS
  net: ethernet: am65-cpsw: cleanup TAPRIO handling
  net: ethernet: ti: am65-cpsw: Move code to avoid forward declaration
  net: ethernet: am65-cpsw: Move register definitions to header file
  net: ethernet: ti: am65-cpsw-qos: Add Frame Preemption MAC Merge
    support
  net: ethernet: ti: am65-cpsw: Fix get_eth_mac_stats

Vladimir Oltean (2):
  selftests: forwarding: ethtool_mm: support devices with higher
    rx-min-frag-size
  selftests: forwarding: ethtool_mm: fall back to aggregate if device
    does not report pMAC stats

 drivers/net/ethernet/ti/Kconfig               |  14 +-
 drivers/net/ethernet/ti/Makefile              |   3 +-
 drivers/net/ethernet/ti/am65-cpsw-ethtool.c   | 249 ++++++
 drivers/net/ethernet/ti/am65-cpsw-nuss.c      |  64 +-
 drivers/net/ethernet/ti/am65-cpsw-nuss.h      |   9 +
 drivers/net/ethernet/ti/am65-cpsw-qos.c       | 708 +++++++++++++-----
 drivers/net/ethernet/ti/am65-cpsw-qos.h       | 186 +++++
 .../selftests/net/forwarding/ethtool_mm.sh    |  48 +-
 tools/testing/selftests/net/forwarding/lib.sh |   9 +
 9 files changed, 1103 insertions(+), 187 deletions(-)


base-commit: 70028b2e51c61d8dda0a31985978f4745da6a11b
-- 
2.34.1

[PATCH v8 net-next 01/11] selftests: forwarding: ethtool_mm: support devices with higher rx-min-frag-size

[Roger Quadros]

From: Vladimir Oltean <vladimir.oltean@nxp.com>

Some devices have errata due to which they cannot report ETH_ZLEN (60)
in the rx-min-frag-size. This was foreseen of course, and lldpad has
logic that when we request it to advertise addFragSize 0, it will round
it up to the lowest value that is _actually_ supported by the hardware.

The problem is that the selftest expects lldpad to report back to us the
same value as we requested.

Make the selftest smarter by figuring out on its own what is a
reasonable value to expect.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Tested-by: Roger Quadros <rogerq@kernel.org>
Signed-off-by: Roger Quadros <rogerq@kernel.org>
---
 .../selftests/net/forwarding/ethtool_mm.sh    | 37 ++++++++++++++++++-
 1 file changed, 35 insertions(+), 2 deletions(-)

Changelog:

v8: no change. Moved to the beginning of series.
v7: initial commit

diff --git a/tools/testing/selftests/net/forwarding/ethtool_mm.sh b/tools/testing/selftests/net/forwarding/ethtool_mm.sh
index 39e736f30322..6212913f4ad1 100755
--- a/tools/testing/selftests/net/forwarding/ethtool_mm.sh
+++ b/tools/testing/selftests/net/forwarding/ethtool_mm.sh
@@ -155,15 +155,48 @@ manual_failed_verification_h2_to_h1()
 	manual_failed_verification $h2 $h1
 }
 
+smallest_supported_add_frag_size()
+{
+	local iface=$1
+	local rx_min_frag_size=
+
+	rx_min_frag_size=$(ethtool --json --show-mm $iface | \
+		jq '.[]."rx-min-frag-size"')
+
+	if [ $rx_min_frag_size -le 60 ]; then
+		echo 0
+	elif [ $rx_min_frag_size -le 124 ]; then
+		echo 1
+	elif [ $rx_min_frag_size -le 188 ]; then
+		echo 2
+	elif [ $rx_min_frag_size -le 252 ]; then
+		echo 3
+	else
+		echo "$iface: RX min frag size $rx_min_frag_size cannot be advertised over LLDP"
+		exit 1
+	fi
+}
+
+expected_add_frag_size()
+{
+	local iface=$1
+	local requested=$2
+	local min=$(smallest_supported_add_frag_size $iface)
+
+	[ $requested -le $min ] && echo $min || echo $requested
+}
+
 lldp_change_add_frag_size()
 {
 	local add_frag_size=$1
+	local pattern=
 
 	lldptool -T -i $h1 -V addEthCaps addFragSize=$add_frag_size >/dev/null
 	# Wait for TLVs to be received
 	sleep 2
-	lldptool -i $h2 -t -n -V addEthCaps | \
-		grep -q "Additional fragment size: $add_frag_size"
+	pattern=$(printf "Additional fragment size: %d" \
+			 $(expected_add_frag_size $h1 $add_frag_size))
+	lldptool -i $h2 -t -n -V addEthCaps | grep -q "$pattern"
 }
 
 lldp()
-- 
2.34.1

[PATCH v8 net-next 02/11] selftests: forwarding: ethtool_mm: fall back to aggregate if device does not report pMAC stats

[Roger Quadros]

From: Vladimir Oltean <vladimir.oltean@nxp.com>

Some devices do not support individual 'pmac' and 'emac' stats.
For such devices, resort to 'aggregate' stats.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Tested-by: Roger Quadros <rogerq@kernel.org>
Signed-off-by: Roger Quadros <rogerq@kernel.org>
---
 tools/testing/selftests/net/forwarding/ethtool_mm.sh | 11 +++++++++++
 tools/testing/selftests/net/forwarding/lib.sh        |  9 +++++++++
 2 files changed, 20 insertions(+)

Changelog:

v8: initial commit

diff --git a/tools/testing/selftests/net/forwarding/ethtool_mm.sh b/tools/testing/selftests/net/forwarding/ethtool_mm.sh
index 6212913f4ad1..50d5bfb17ef1 100755
--- a/tools/testing/selftests/net/forwarding/ethtool_mm.sh
+++ b/tools/testing/selftests/net/forwarding/ethtool_mm.sh
@@ -25,6 +25,10 @@ traffic_test()
 	local after=
 	local delta=
 
+	if [ ${has_pmac_stats[$if]} = false ]; then
+		src="aggregate"
+	fi
+
 	before=$(ethtool_std_stats_get $if "eth-mac" "FramesTransmittedOK" $src)
 
 	$MZ $if -q -c $num_pkts -p 64 -b bcast -t ip -R $PREEMPTIBLE_PRIO
@@ -317,6 +321,13 @@ for netif in ${NETIFS[@]}; do
 		echo "SKIP: $netif does not support MAC Merge"
 		exit $ksft_skip
 	fi
+
+	if check_ethtool_pmac_std_stats_support $netif eth-mac; then
+		has_pmac_stats[$netif]=true
+	else
+		has_pmac_stats[$netif]=false
+		echo "$netif does not report pMAC statistics, falling back to aggregate"
+	fi
 done
 
 trap cleanup EXIT
diff --git a/tools/testing/selftests/net/forwarding/lib.sh b/tools/testing/selftests/net/forwarding/lib.sh
index 8f6ca458af9a..763c262a3453 100755
--- a/tools/testing/selftests/net/forwarding/lib.sh
+++ b/tools/testing/selftests/net/forwarding/lib.sh
@@ -146,6 +146,15 @@ check_ethtool_mm_support()
 	fi
 }
 
+check_ethtool_pmac_std_stats_support()
+{
+	local dev=$1; shift
+	local grp=$1; shift
+
+	[ 0 -ne $(ethtool --json -S $dev --all-groups --src pmac 2>/dev/null \
+		| jq '.[]."$grp" | length') ]
+}
+
 check_locked_port_support()
 {
 	if ! bridge -d link show | grep -q " locked"; then
-- 
2.34.1

Re: [PATCH v8 net-next 02/11] selftests: forwarding: ethtool_mm: fall back to aggregate if device does not report pMAC stats

[Vladimir Oltean]

On Wed, Dec 13, 2023 at 01:07:12PM +0200, Roger Quadros wrote:
> diff --git a/tools/testing/selftests/net/forwarding/lib.sh b/tools/testing/selftests/net/forwarding/lib.sh
> index 8f6ca458af9a..763c262a3453 100755
> --- a/tools/testing/selftests/net/forwarding/lib.sh
> +++ b/tools/testing/selftests/net/forwarding/lib.sh
> @@ -146,6 +146,15 @@ check_ethtool_mm_support()
>  	fi
>  }
>  
> +check_ethtool_pmac_std_stats_support()
> +{
> +	local dev=$1; shift
> +	local grp=$1; shift
> +
> +	[ 0 -ne $(ethtool --json -S $dev --all-groups --src pmac 2>/dev/null \
> +		| jq '.[]."$grp" | length') ]
> +}
> +
>  check_locked_port_support()
>  {
>  	if ! bridge -d link show | grep -q " locked"; then
> -- 
> 2.34.1
>

FYI, there's another submitted patch that touches the exact same spot,
and it looks like it has a good chance of getting merged.
https://patchwork.kernel.org/project/netdevbpf/patch/20231214135029.383595-9-tobias@waldekranz.com/

You need to pay attention to merge conflicts, so you don't waste a patch
iteration just because of that one thing.

I guess you might be able to wing it, because the other patch does this:

diff --git a/tools/testing/selftests/net/forwarding/lib.sh b/tools/testing/selftests/net/forwarding/lib.sh
index 8f6ca458af9a..e3740163c384 100755
--- a/tools/testing/selftests/net/forwarding/lib.sh
+++ b/tools/testing/selftests/net/forwarding/lib.sh
@@ -146,6 +146,15 @@  check_ethtool_mm_support()
 	fi
 }
 
+check_ethtool_counter_group_support()
+{
+	ethtool --help 2>&1| grep -- '--all-groups' &> /dev/null
+	if [[ $? -ne 0 ]]; then
+		echo "SKIP: ethtool too old; it is missing standard counter group support"
+		exit $ksft_skip
+	fi
+}
+
 check_locked_port_support()
 {
 	if ! bridge -d link show | grep -q " locked"; then

which quite coincidentally does not change what your patch sees in its
upper context, aka 3 lines like this:

----
 	fi
 }
 
----

You can check if your patch set applies on top of Tobias', by formatting
it as patch files on top of net-next/main, resetting HEAD to net-next,
applying Tobias' series and then your patches.

Re: [PATCH v8 net-next 02/11] selftests: forwarding: ethtool_mm: fall back to aggregate if device does not report pMAC stats

[Roger Quadros]

On 14/12/2023 16:16, Vladimir Oltean wrote:
> On Wed, Dec 13, 2023 at 01:07:12PM +0200, Roger Quadros wrote:
>> diff --git a/tools/testing/selftests/net/forwarding/lib.sh b/tools/testing/selftests/net/forwarding/lib.sh
>> index 8f6ca458af9a..763c262a3453 100755
>> --- a/tools/testing/selftests/net/forwarding/lib.sh
>> +++ b/tools/testing/selftests/net/forwarding/lib.sh
>> @@ -146,6 +146,15 @@ check_ethtool_mm_support()
>>  	fi
>>  }
>>  
>> +check_ethtool_pmac_std_stats_support()
>> +{
>> +	local dev=$1; shift
>> +	local grp=$1; shift
>> +
>> +	[ 0 -ne $(ethtool --json -S $dev --all-groups --src pmac 2>/dev/null \
>> +		| jq '.[]."$grp" | length') ]
>> +}
>> +
>>  check_locked_port_support()
>>  {
>>  	if ! bridge -d link show | grep -q " locked"; then
>> -- 
>> 2.34.1
>>
> 
> FYI, there's another submitted patch that touches the exact same spot,
> and it looks like it has a good chance of getting merged.
> https://patchwork.kernel.org/project/netdevbpf/patch/20231214135029.383595-9-tobias@waldekranz.com/
> 
> You need to pay attention to merge conflicts, so you don't waste a patch
> iteration just because of that one thing.
> 
> I guess you might be able to wing it, because the other patch does this:
> 
> diff --git a/tools/testing/selftests/net/forwarding/lib.sh b/tools/testing/selftests/net/forwarding/lib.sh
> index 8f6ca458af9a..e3740163c384 100755
> --- a/tools/testing/selftests/net/forwarding/lib.sh
> +++ b/tools/testing/selftests/net/forwarding/lib.sh
> @@ -146,6 +146,15 @@  check_ethtool_mm_support()
>  	fi
>  }
>  
> +check_ethtool_counter_group_support()
> +{
> +	ethtool --help 2>&1| grep -- '--all-groups' &> /dev/null
> +	if [[ $? -ne 0 ]]; then
> +		echo "SKIP: ethtool too old; it is missing standard counter group support"
> +		exit $ksft_skip
> +	fi
> +}
> +
>  check_locked_port_support()
>  {
>  	if ! bridge -d link show | grep -q " locked"; then
> 
> which quite coincidentally does not change what your patch sees in its
> upper context, aka 3 lines like this:
> 
> ----
>  	fi
>  }
>  
> ----
> 
> You can check if your patch set applies on top of Tobias', by formatting
> it as patch files on top of net-next/main, resetting HEAD to net-next,
> applying Tobias' series and then your patches.

Thanks for the heads up :)

-- 
cheers,
-roger

[PATCH v8 net-next 03/11] net: ethernet: am65-cpsw: Build am65-cpsw-qos only if required

[Roger Quadros]

Build am65-cpsw-qos only if CONFIG_TI_AM65_CPSW_TAS is enabled.

Signed-off-by: Roger Quadros <rogerq@kernel.org>
---
 drivers/net/ethernet/ti/Makefile        |  3 ++-
 drivers/net/ethernet/ti/am65-cpsw-qos.c | 12 ------------
 drivers/net/ethernet/ti/am65-cpsw-qos.h | 26 +++++++++++++++++++++++++
 3 files changed, 28 insertions(+), 13 deletions(-)

Changelog:

v8: no change
v7: no change
v6: initial commit

diff --git a/drivers/net/ethernet/ti/Makefile b/drivers/net/ethernet/ti/Makefile
index 27de1d697134..9d7cd84d1e2d 100644
--- a/drivers/net/ethernet/ti/Makefile
+++ b/drivers/net/ethernet/ti/Makefile
@@ -26,7 +26,8 @@ keystone_netcp_ethss-y := netcp_ethss.o netcp_sgmii.o netcp_xgbepcsr.o cpsw_ale.
 obj-$(CONFIG_TI_K3_CPPI_DESC_POOL) += k3-cppi-desc-pool.o
 
 obj-$(CONFIG_TI_K3_AM65_CPSW_NUSS) += ti-am65-cpsw-nuss.o
-ti-am65-cpsw-nuss-y := am65-cpsw-nuss.o cpsw_sl.o am65-cpsw-ethtool.o cpsw_ale.o am65-cpsw-qos.o
+ti-am65-cpsw-nuss-y := am65-cpsw-nuss.o cpsw_sl.o am65-cpsw-ethtool.o cpsw_ale.o
+ti-am65-cpsw-nuss-$(CONFIG_TI_AM65_CPSW_TAS) += am65-cpsw-qos.o
 ti-am65-cpsw-nuss-$(CONFIG_TI_K3_AM65_CPSW_SWITCHDEV) += am65-cpsw-switchdev.o
 obj-$(CONFIG_TI_K3_AM65_CPTS) += am65-cpts.o
 
diff --git a/drivers/net/ethernet/ti/am65-cpsw-qos.c b/drivers/net/ethernet/ti/am65-cpsw-qos.c
index 9ac2ff05d501..4bc611cc4aad 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-qos.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-qos.c
@@ -571,9 +571,6 @@ static int am65_cpsw_setup_taprio(struct net_device *ndev, void *type_data)
 	    taprio->cmd != TAPRIO_CMD_DESTROY)
 		return -EOPNOTSUPP;
 
-	if (!IS_ENABLED(CONFIG_TI_AM65_CPSW_TAS))
-		return -ENODEV;
-
 	if (!netif_running(ndev)) {
 		dev_err(&ndev->dev, "interface is down, link speed unknown\n");
 		return -ENETDOWN;
@@ -599,9 +596,6 @@ static int am65_cpsw_tc_query_caps(struct net_device *ndev, void *type_data)
 	case TC_SETUP_QDISC_TAPRIO: {
 		struct tc_taprio_caps *caps = base->caps;
 
-		if (!IS_ENABLED(CONFIG_TI_AM65_CPSW_TAS))
-			return -EOPNOTSUPP;
-
 		caps->gate_mask_per_txq = true;
 
 		return 0;
@@ -806,9 +800,6 @@ void am65_cpsw_qos_link_up(struct net_device *ndev, int link_speed)
 {
 	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
 
-	if (!IS_ENABLED(CONFIG_TI_AM65_CPSW_TAS))
-		return;
-
 	am65_cpsw_est_link_up(ndev, link_speed);
 	port->qos.link_down_time = 0;
 }
@@ -817,9 +808,6 @@ void am65_cpsw_qos_link_down(struct net_device *ndev)
 {
 	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
 
-	if (!IS_ENABLED(CONFIG_TI_AM65_CPSW_TAS))
-		return;
-
 	if (!port->qos.link_down_time)
 		port->qos.link_down_time = ktime_get();
 
diff --git a/drivers/net/ethernet/ti/am65-cpsw-qos.h b/drivers/net/ethernet/ti/am65-cpsw-qos.h
index 0cc2a3b3d7f9..898f13a4a112 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-qos.h
+++ b/drivers/net/ethernet/ti/am65-cpsw-qos.h
@@ -31,11 +31,37 @@ struct am65_cpsw_qos {
 	struct am65_cpsw_ale_ratelimit ale_mc_ratelimit;
 };
 
+#if IS_ENABLED(CONFIG_TI_AM65_CPSW_TAS)
 int am65_cpsw_qos_ndo_setup_tc(struct net_device *ndev, enum tc_setup_type type,
 			       void *type_data);
 void am65_cpsw_qos_link_up(struct net_device *ndev, int link_speed);
 void am65_cpsw_qos_link_down(struct net_device *ndev);
 int am65_cpsw_qos_ndo_tx_p0_set_maxrate(struct net_device *ndev, int queue, u32 rate_mbps);
 void am65_cpsw_qos_tx_p0_rate_init(struct am65_cpsw_common *common);
+#else
+static inline int am65_cpsw_qos_ndo_setup_tc(struct net_device *ndev,
+					     enum tc_setup_type type,
+					     void *type_data)
+{
+	return -EOPNOTSUPP;
+}
+
+static inline void am65_cpsw_qos_link_up(struct net_device *ndev,
+					 int link_speed)
+{ }
+
+static inline void am65_cpsw_qos_link_down(struct net_device *ndev)
+{ }
+
+static inline int am65_cpsw_qos_ndo_tx_p0_set_maxrate(struct net_device *ndev,
+						      int queue,
+						      u32 rate_mbps)
+{
+	return 0;
+}
+
+static inline void am65_cpsw_qos_tx_p0_rate_init(struct am65_cpsw_common *common)
+{ }
+#endif
 
 #endif /* AM65_CPSW_QOS_H_ */
-- 
2.34.1

[PATCH v8 net-next 04/11] net: ethernet: am65-cpsw: Rename TI_AM65_CPSW_TAS to TI_AM65_CPSW_QOS

[Roger Quadros]

We will use this Kconfig option to not only enable TAS/EST offload
but also other QoS features like Multiqueue priority descriptors
and MAC-Merge/Frame Preemption. TI_AM65_CPSW_QOS seems a more
appropriate Kconfig option name than TI_AM65_CPSW_TAS.

Signed-off-by: Roger Quadros <rogerq@kernel.org>
---
 drivers/net/ethernet/ti/Kconfig         | 12 ++++++------
 drivers/net/ethernet/ti/Makefile        |  2 +-
 drivers/net/ethernet/ti/am65-cpsw-qos.h |  2 +-
 3 files changed, 8 insertions(+), 8 deletions(-)

Changelog:

v8: initial commit

diff --git a/drivers/net/ethernet/ti/Kconfig b/drivers/net/ethernet/ti/Kconfig
index e60b557d59b9..49cd96c4f532 100644
--- a/drivers/net/ethernet/ti/Kconfig
+++ b/drivers/net/ethernet/ti/Kconfig
@@ -134,14 +134,14 @@ config TI_K3_AM65_CPTS
 	  protocol, Ethernet Enhanced Scheduled Traffic Operations (CPTS_ESTFn)
 	  and PCIe Subsystem Precision Time Measurement (PTM).
 
-config TI_AM65_CPSW_TAS
-	bool "Enable TAS offload in AM65 CPSW"
+config TI_AM65_CPSW_QOS
+	bool "Enable QoS offload features in AM65 CPSW"
 	depends on TI_K3_AM65_CPSW_NUSS && NET_SCH_TAPRIO && TI_K3_AM65_CPTS
 	help
-	  Say y here to support Time Aware Shaper(TAS) offload in AM65 CPSW.
-	  AM65 CPSW hardware supports Enhanced Scheduled Traffic (EST)
-	  defined in IEEE 802.1Q 2018. The EST scheduler runs on CPTS and the
-	  TAS/EST schedule is updated in the Fetch RAM memory of the CPSW.
+	  This option enables QoS offload features in AM65 CPSW like
+	  Time Aware Shaper (TAS) / Enhanced Scheduled Traffic (EST).
+	  The EST scheduler runs on CPTS and the TAS/EST schedule is
+	  updated in the Fetch RAM memory of the CPSW.
 
 config TI_KEYSTONE_NETCP
 	tristate "TI Keystone NETCP Core Support"
diff --git a/drivers/net/ethernet/ti/Makefile b/drivers/net/ethernet/ti/Makefile
index 9d7cd84d1e2d..d8590304f3df 100644
--- a/drivers/net/ethernet/ti/Makefile
+++ b/drivers/net/ethernet/ti/Makefile
@@ -27,7 +27,7 @@ obj-$(CONFIG_TI_K3_CPPI_DESC_POOL) += k3-cppi-desc-pool.o
 
 obj-$(CONFIG_TI_K3_AM65_CPSW_NUSS) += ti-am65-cpsw-nuss.o
 ti-am65-cpsw-nuss-y := am65-cpsw-nuss.o cpsw_sl.o am65-cpsw-ethtool.o cpsw_ale.o
-ti-am65-cpsw-nuss-$(CONFIG_TI_AM65_CPSW_TAS) += am65-cpsw-qos.o
+ti-am65-cpsw-nuss-$(CONFIG_TI_AM65_CPSW_QOS) += am65-cpsw-qos.o
 ti-am65-cpsw-nuss-$(CONFIG_TI_K3_AM65_CPSW_SWITCHDEV) += am65-cpsw-switchdev.o
 obj-$(CONFIG_TI_K3_AM65_CPTS) += am65-cpts.o
 
diff --git a/drivers/net/ethernet/ti/am65-cpsw-qos.h b/drivers/net/ethernet/ti/am65-cpsw-qos.h
index 898f13a4a112..be4987eb8c51 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-qos.h
+++ b/drivers/net/ethernet/ti/am65-cpsw-qos.h
@@ -31,7 +31,7 @@ struct am65_cpsw_qos {
 	struct am65_cpsw_ale_ratelimit ale_mc_ratelimit;
 };
 
-#if IS_ENABLED(CONFIG_TI_AM65_CPSW_TAS)
+#if IS_ENABLED(CONFIG_TI_AM65_CPSW_QOS)
 int am65_cpsw_qos_ndo_setup_tc(struct net_device *ndev, enum tc_setup_type type,
 			       void *type_data);
 void am65_cpsw_qos_link_up(struct net_device *ndev, int link_speed);
-- 
2.34.1

[PATCH v8 net-next 05/11] net: ethernet: am65-cpsw: cleanup TAPRIO handling

[Roger Quadros]

Handle offloading commands using switch-case in
am65_cpsw_setup_taprio().

Move checks to am65_cpsw_taprio_replace().

Use NL_SET_ERR_MSG_MOD for error messages.
Change error message from "Failed to set cycle time extension"
to "cycle time extension not supported"

Signed-off-by: Roger Quadros <rogerq@kernel.org>
---
 drivers/net/ethernet/ti/am65-cpsw-qos.c | 151 +++++++++++-------------
 1 file changed, 71 insertions(+), 80 deletions(-)

Changelog:

v8: don't initialize ret = 0, tact = TACT_PROG
v7: don't use "\n" in NL_SET_ERR_MSG_MOD()
v6: initial commit

diff --git a/drivers/net/ethernet/ti/am65-cpsw-qos.c b/drivers/net/ethernet/ti/am65-cpsw-qos.c
index 4bc611cc4aad..2c97fa05a852 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-qos.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-qos.c
@@ -428,7 +428,7 @@ static void am65_cpsw_stop_est(struct net_device *ndev)
 	am65_cpsw_timer_stop(ndev);
 }
 
-static void am65_cpsw_purge_est(struct net_device *ndev)
+static void am65_cpsw_taprio_destroy(struct net_device *ndev)
 {
 	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
 
@@ -441,29 +441,66 @@ static void am65_cpsw_purge_est(struct net_device *ndev)
 	port->qos.est_admin = NULL;
 }
 
-static int am65_cpsw_configure_taprio(struct net_device *ndev,
-				      struct am65_cpsw_est *est_new)
+static void am65_cpsw_cp_taprio(struct tc_taprio_qopt_offload *from,
+				struct tc_taprio_qopt_offload *to)
+{
+	int i;
+
+	*to = *from;
+	for (i = 0; i < from->num_entries; i++)
+		to->entries[i] = from->entries[i];
+}
+
+static int am65_cpsw_taprio_replace(struct net_device *ndev,
+				    struct tc_taprio_qopt_offload *taprio)
 {
 	struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+	struct netlink_ext_ack *extack = taprio->mqprio.extack;
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
 	struct am65_cpts *cpts = common->cpts;
-	int ret = 0, tact = TACT_PROG;
+	struct am65_cpsw_est *est_new;
+	int ret, tact;
 
-	am65_cpsw_est_update_state(ndev);
+	if (!netif_running(ndev)) {
+		NL_SET_ERR_MSG_MOD(extack, "interface is down, link speed unknown");
+		return -ENETDOWN;
+	}
 
-	if (est_new->taprio.cmd == TAPRIO_CMD_DESTROY) {
-		am65_cpsw_stop_est(ndev);
-		return ret;
+	if (common->pf_p0_rx_ptype_rrobin) {
+		NL_SET_ERR_MSG_MOD(extack,
+				   "p0-rx-ptype-rrobin flag conflicts with taprio qdisc");
+		return -EINVAL;
+	}
+
+	if (port->qos.link_speed == SPEED_UNKNOWN)
+		return -ENOLINK;
+
+	if (taprio->cycle_time_extension) {
+		NL_SET_ERR_MSG_MOD(extack,
+				   "cycle time extension not supported");
+		return -EOPNOTSUPP;
 	}
 
+	est_new = devm_kzalloc(&ndev->dev,
+			       struct_size(est_new, taprio.entries, taprio->num_entries),
+			       GFP_KERNEL);
+	if (!est_new)
+		return -ENOMEM;
+
+	am65_cpsw_cp_taprio(taprio, &est_new->taprio);
+
+	am65_cpsw_est_update_state(ndev);
+
 	ret = am65_cpsw_est_check_scheds(ndev, est_new);
 	if (ret < 0)
-		return ret;
+		goto fail;
 
 	tact = am65_cpsw_timer_act(ndev, est_new);
 	if (tact == TACT_NEED_STOP) {
-		dev_err(&ndev->dev,
-			"Can't toggle estf timer, stop taprio first");
-		return -EINVAL;
+		NL_SET_ERR_MSG_MOD(extack,
+				   "Can't toggle estf timer, stop taprio first");
+		ret = -EINVAL;
+		goto fail;
 	}
 
 	if (tact == TACT_PROG)
@@ -476,62 +513,24 @@ static int am65_cpsw_configure_taprio(struct net_device *ndev,
 	am65_cpsw_est_set_sched_list(ndev, est_new);
 	am65_cpsw_port_est_assign_buf_num(ndev, est_new->buf);
 
-	am65_cpsw_est_set(ndev, est_new->taprio.cmd == TAPRIO_CMD_REPLACE);
+	am65_cpsw_est_set(ndev, 1);
 
 	if (tact == TACT_PROG) {
 		ret = am65_cpsw_timer_set(ndev, est_new);
 		if (ret) {
-			dev_err(&ndev->dev, "Failed to set cycle time");
-			return ret;
+			NL_SET_ERR_MSG_MOD(extack,
+					   "Failed to set cycle time");
+			goto fail;
 		}
 	}
 
-	return ret;
-}
-
-static void am65_cpsw_cp_taprio(struct tc_taprio_qopt_offload *from,
-				struct tc_taprio_qopt_offload *to)
-{
-	int i;
-
-	*to = *from;
-	for (i = 0; i < from->num_entries; i++)
-		to->entries[i] = from->entries[i];
-}
-
-static int am65_cpsw_set_taprio(struct net_device *ndev, void *type_data)
-{
-	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
-	struct tc_taprio_qopt_offload *taprio = type_data;
-	struct am65_cpsw_est *est_new;
-	int ret = 0;
-
-	if (taprio->cycle_time_extension) {
-		dev_err(&ndev->dev, "Failed to set cycle time extension");
-		return -EOPNOTSUPP;
-	}
-
-	est_new = devm_kzalloc(&ndev->dev,
-			       struct_size(est_new, taprio.entries, taprio->num_entries),
-			       GFP_KERNEL);
-	if (!est_new)
-		return -ENOMEM;
-
-	am65_cpsw_cp_taprio(taprio, &est_new->taprio);
-	ret = am65_cpsw_configure_taprio(ndev, est_new);
-	if (!ret) {
-		if (taprio->cmd == TAPRIO_CMD_REPLACE) {
-			devm_kfree(&ndev->dev, port->qos.est_admin);
+	devm_kfree(&ndev->dev, port->qos.est_admin);
+	port->qos.est_admin = est_new;
 
-			port->qos.est_admin = est_new;
-		} else {
-			devm_kfree(&ndev->dev, est_new);
-			am65_cpsw_purge_est(ndev);
-		}
-	} else {
-		devm_kfree(&ndev->dev, est_new);
-	}
+	return 0;
 
+fail:
+	devm_kfree(&ndev->dev, est_new);
 	return ret;
 }
 
@@ -558,34 +557,26 @@ static void am65_cpsw_est_link_up(struct net_device *ndev, int link_speed)
 	return;
 
 purge_est:
-	am65_cpsw_purge_est(ndev);
+	am65_cpsw_taprio_destroy(ndev);
 }
 
 static int am65_cpsw_setup_taprio(struct net_device *ndev, void *type_data)
 {
-	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
 	struct tc_taprio_qopt_offload *taprio = type_data;
-	struct am65_cpsw_common *common = port->common;
-
-	if (taprio->cmd != TAPRIO_CMD_REPLACE &&
-	    taprio->cmd != TAPRIO_CMD_DESTROY)
-		return -EOPNOTSUPP;
-
-	if (!netif_running(ndev)) {
-		dev_err(&ndev->dev, "interface is down, link speed unknown\n");
-		return -ENETDOWN;
-	}
-
-	if (common->pf_p0_rx_ptype_rrobin) {
-		dev_err(&ndev->dev,
-			"p0-rx-ptype-rrobin flag conflicts with taprio qdisc\n");
-		return -EINVAL;
+	int err = 0;
+
+	switch (taprio->cmd) {
+	case TAPRIO_CMD_REPLACE:
+		err = am65_cpsw_taprio_replace(ndev, taprio);
+		break;
+	case TAPRIO_CMD_DESTROY:
+		am65_cpsw_taprio_destroy(ndev);
+		break;
+	default:
+		err = -EOPNOTSUPP;
 	}
 
-	if (port->qos.link_speed == SPEED_UNKNOWN)
-		return -ENOLINK;
-
-	return am65_cpsw_set_taprio(ndev, type_data);
+	return err;
 }
 
 static int am65_cpsw_tc_query_caps(struct net_device *ndev, void *type_data)
-- 
2.34.1

Re: [PATCH v8 net-next 05/11] net: ethernet: am65-cpsw: cleanup TAPRIO handling

[Vladimir Oltean]

On Wed, Dec 13, 2023 at 01:07:15PM +0200, Roger Quadros wrote:
> +static int am65_cpsw_taprio_replace(struct net_device *ndev,
> +				    struct tc_taprio_qopt_offload *taprio)
>  {
>  	struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
> +	struct netlink_ext_ack *extack = taprio->mqprio.extack;
> +	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
>  	struct am65_cpts *cpts = common->cpts;
> -	int ret = 0, tact = TACT_PROG;
> +	struct am65_cpsw_est *est_new;
> +	int ret, tact;
>  
> -	am65_cpsw_est_update_state(ndev);
> +	if (!netif_running(ndev)) {
> +		NL_SET_ERR_MSG_MOD(extack, "interface is down, link speed unknown");
> +		return -ENETDOWN;
> +	}

I haven't used the runtime PM API that this driver uses. I don't know
much about how it works. What are the rules here? By checking for
netif_running(), are you intending to rely on the pm_runtime_resume_and_get()
call from ndo_open(), which is released with pm_runtime_put() at
ndo_stop() time?

I see some inconsistencies I don't quite understand.

am65_cpsw_nuss_ndo_slave_add_vid() checks for netif_running() then calls
pm_runtime_resume_and_get() anyway.

am65_cpsw_setup_mqprio() allows changing the offload even when the link
is down (which is more user-friendly anyway) and performs the pm_runtime_get_sync()
call itself.

> -}

Re: [PATCH v8 net-next 05/11] net: ethernet: am65-cpsw: cleanup TAPRIO handling

[Roger Quadros]

On 14/12/2023 13:23, Vladimir Oltean wrote:
> On Wed, Dec 13, 2023 at 01:07:15PM +0200, Roger Quadros wrote:
>> +static int am65_cpsw_taprio_replace(struct net_device *ndev,
>> +				    struct tc_taprio_qopt_offload *taprio)
>>  {
>>  	struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
>> +	struct netlink_ext_ack *extack = taprio->mqprio.extack;
>> +	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
>>  	struct am65_cpts *cpts = common->cpts;
>> -	int ret = 0, tact = TACT_PROG;
>> +	struct am65_cpsw_est *est_new;
>> +	int ret, tact;
>>  
>> -	am65_cpsw_est_update_state(ndev);
>> +	if (!netif_running(ndev)) {
>> +		NL_SET_ERR_MSG_MOD(extack, "interface is down, link speed unknown");
>> +		return -ENETDOWN;
>> +	}
> 
> I haven't used the runtime PM API that this driver uses. I don't know
> much about how it works. What are the rules here? By checking for

The only rule is that if network interface is down, the device might be
runtime_suspended so we need to get it back to runtime_active before any
device access.

> netif_running(), are you intending to rely on the pm_runtime_resume_and_get()
> call from ndo_open(), which is released with pm_runtime_put() at
> ndo_stop() time?

Actually, this code is already present upstream. I'm only moving it around
in this patch.

Based on the error message and looking at am65_cpsw_est_check_scheds() and
am65_cpsw_est_set_sched_list() which is called later in am65_cpsw_taprio_replace(),
both of which eventually call am65_est_cmd_ns_to_cnt() which expects valid link_speed,
my understanding is that the author intended to have a valid link_speed before
proceeding further.

Although it seems netif_running() check isn't enough to have valid link_speed
as the link could still be down even if the netif is brought up.

Another gap is that in am65_cpsw_est_link_up(), if link was down for more than 1 second
it just abruptly calls am65_cpsw_taprio_destroy().

So I think we need to do the following to improve taprio support in this driver:
1) accept taprio schedule irrespective of netif/link_speed status
2) call pm_runtime_get()/put() before any device access regardless of netif/link_speed state
3) on link_up when if have valid link_speed and taprio_schedule, apply it.
4) on link_down, destroy the taprio schedule form the controller.

But my concern is, this is a decent amount of work and I don't want to delay this series.
My original subject of this patch series was mpqrio/frame-preemption/coalescing. ;)

Can we please defer taprio enhancement to a separate series? Thanks!

> 
> I see some inconsistencies I don't quite understand.
> 
> am65_cpsw_nuss_ndo_slave_add_vid() checks for netif_running() then calls
> pm_runtime_resume_and_get() anyway.
> 
> am65_cpsw_setup_mqprio() allows changing the offload even when the link
> is down (which is more user-friendly anyway) and performs the pm_runtime_get_sync()
> call itself.
> 
>> -}

-- 
cheers,
-roger

Re: [PATCH v8 net-next 05/11] net: ethernet: am65-cpsw: cleanup TAPRIO handling

[Vladimir Oltean]

On Thu, Dec 14, 2023 at 03:36:57PM +0200, Roger Quadros wrote:
> Actually, this code is already present upstream. I'm only moving it around
> in this patch.
> 
> Based on the error message and looking at am65_cpsw_est_check_scheds() and
> am65_cpsw_est_set_sched_list() which is called later in am65_cpsw_taprio_replace(),
> both of which eventually call am65_est_cmd_ns_to_cnt() which expects valid link_speed,
> my understanding is that the author intended to have a valid link_speed before
> proceeding further.
> 
> Although it seems netif_running() check isn't enough to have valid link_speed
> as the link could still be down even if the netif is brought up.
> 
> Another gap is that in am65_cpsw_est_link_up(), if link was down for more than 1 second
> it just abruptly calls am65_cpsw_taprio_destroy().
> 
> So I think we need to do the following to improve taprio support in this driver:
> 1) accept taprio schedule irrespective of netif/link_speed status
> 2) call pm_runtime_get()/put() before any device access regardless of netif/link_speed state
> 3) on link_up when if have valid link_speed and taprio_schedule, apply it.
> 4) on link_down, destroy the taprio schedule form the controller.
> 
> But my concern is, this is a decent amount of work and I don't want to delay this series.
> My original subject of this patch series was mpqrio/frame-preemption/coalescing. ;)
> 
> Can we please defer taprio enhancement to a separate series? Thanks!

Ok, sounds fair to have some further taprio clean-up scheduled for later.
I would also add taprio_offload_get() to the list of improvements that
could be made.

Re: [PATCH v8 net-next 05/11] net: ethernet: am65-cpsw: cleanup TAPRIO handling

[Roger Quadros]

On 14/12/2023 15:41, Vladimir Oltean wrote:
> On Thu, Dec 14, 2023 at 03:36:57PM +0200, Roger Quadros wrote:
>> Actually, this code is already present upstream. I'm only moving it around
>> in this patch.
>>
>> Based on the error message and looking at am65_cpsw_est_check_scheds() and
>> am65_cpsw_est_set_sched_list() which is called later in am65_cpsw_taprio_replace(),
>> both of which eventually call am65_est_cmd_ns_to_cnt() which expects valid link_speed,
>> my understanding is that the author intended to have a valid link_speed before
>> proceeding further.
>>
>> Although it seems netif_running() check isn't enough to have valid link_speed
>> as the link could still be down even if the netif is brought up.
>>
>> Another gap is that in am65_cpsw_est_link_up(), if link was down for more than 1 second
>> it just abruptly calls am65_cpsw_taprio_destroy().
>>
>> So I think we need to do the following to improve taprio support in this driver:
>> 1) accept taprio schedule irrespective of netif/link_speed status
>> 2) call pm_runtime_get()/put() before any device access regardless of netif/link_speed state
>> 3) on link_up when if have valid link_speed and taprio_schedule, apply it.
>> 4) on link_down, destroy the taprio schedule form the controller.
>>
>> But my concern is, this is a decent amount of work and I don't want to delay this series.
>> My original subject of this patch series was mpqrio/frame-preemption/coalescing. ;)
>>
>> Can we please defer taprio enhancement to a separate series? Thanks!
> 
> Ok, sounds fair to have some further taprio clean-up scheduled for later.
> I would also add taprio_offload_get() to the list of improvements that
> could be made.

Noted. Thanks!

-- 
cheers,
-roger

[PATCH v8 net-next 06/11] net: ethernet: ti: am65-cpsw: Move code to avoid forward declaration

[Roger Quadros]

Move this code to the end to avoid forward declaration.
No functional change.

Signed-off-by: Roger Quadros <rogerq@kernel.org>
---
 drivers/net/ethernet/ti/am65-cpsw-qos.c | 86 ++++++++++++-------------
 1 file changed, 43 insertions(+), 43 deletions(-)

Changelog

v7: move am65_cpsw_qos_tx_rate_calc()
v6: no change
v5: no change
v4: initial commit

diff --git a/drivers/net/ethernet/ti/am65-cpsw-qos.c b/drivers/net/ethernet/ti/am65-cpsw-qos.c
index 2c97fa05a852..862a98211caa 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-qos.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-qos.c
@@ -56,6 +56,16 @@ enum timer_act {
 	TACT_SKIP_PROG,		/* just buffer can be updated */
 };
 
+static u32
+am65_cpsw_qos_tx_rate_calc(u32 rate_mbps, unsigned long bus_freq)
+{
+	u32 ir;
+
+	bus_freq /= 1000000;
+	ir = DIV_ROUND_UP(((u64)rate_mbps * 32768),  bus_freq);
+	return ir;
+}
+
 static int am65_cpsw_port_est_enabled(struct am65_cpsw_port *port)
 {
 	return port->qos.est_oper || port->qos.est_admin;
@@ -772,49 +782,6 @@ static int am65_cpsw_qos_setup_tc_block(struct net_device *ndev, struct flow_blo
 					  port, port, true);
 }
 
-int am65_cpsw_qos_ndo_setup_tc(struct net_device *ndev, enum tc_setup_type type,
-			       void *type_data)
-{
-	switch (type) {
-	case TC_QUERY_CAPS:
-		return am65_cpsw_tc_query_caps(ndev, type_data);
-	case TC_SETUP_QDISC_TAPRIO:
-		return am65_cpsw_setup_taprio(ndev, type_data);
-	case TC_SETUP_BLOCK:
-		return am65_cpsw_qos_setup_tc_block(ndev, type_data);
-	default:
-		return -EOPNOTSUPP;
-	}
-}
-
-void am65_cpsw_qos_link_up(struct net_device *ndev, int link_speed)
-{
-	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
-
-	am65_cpsw_est_link_up(ndev, link_speed);
-	port->qos.link_down_time = 0;
-}
-
-void am65_cpsw_qos_link_down(struct net_device *ndev)
-{
-	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
-
-	if (!port->qos.link_down_time)
-		port->qos.link_down_time = ktime_get();
-
-	port->qos.link_speed = SPEED_UNKNOWN;
-}
-
-static u32
-am65_cpsw_qos_tx_rate_calc(u32 rate_mbps, unsigned long bus_freq)
-{
-	u32 ir;
-
-	bus_freq /= 1000000;
-	ir = DIV_ROUND_UP(((u64)rate_mbps * 32768),  bus_freq);
-	return ir;
-}
-
 static void
 am65_cpsw_qos_tx_p0_rate_apply(struct am65_cpsw_common *common,
 			       int tx_ch, u32 rate_mbps)
@@ -916,3 +883,36 @@ void am65_cpsw_qos_tx_p0_rate_init(struct am65_cpsw_common *common)
 		       host->port_base + AM65_CPSW_PN_REG_PRI_CIR(tx_ch));
 	}
 }
+
+int am65_cpsw_qos_ndo_setup_tc(struct net_device *ndev, enum tc_setup_type type,
+			       void *type_data)
+{
+	switch (type) {
+	case TC_QUERY_CAPS:
+		return am65_cpsw_tc_query_caps(ndev, type_data);
+	case TC_SETUP_QDISC_TAPRIO:
+		return am65_cpsw_setup_taprio(ndev, type_data);
+	case TC_SETUP_BLOCK:
+		return am65_cpsw_qos_setup_tc_block(ndev, type_data);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+void am65_cpsw_qos_link_up(struct net_device *ndev, int link_speed)
+{
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+
+	am65_cpsw_est_link_up(ndev, link_speed);
+	port->qos.link_down_time = 0;
+}
+
+void am65_cpsw_qos_link_down(struct net_device *ndev)
+{
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+
+	if (!port->qos.link_down_time)
+		port->qos.link_down_time = ktime_get();
+
+	port->qos.link_speed = SPEED_UNKNOWN;
+}
-- 
2.34.1

Re: [PATCH v8 net-next 06/11] net: ethernet: ti: am65-cpsw: Move code to avoid forward declaration

[Vladimir Oltean]

On Wed, Dec 13, 2023 at 01:07:16PM +0200, Roger Quadros wrote:
> Move this code to the end to avoid forward declaration.
> No functional change.

It's no longer all being moved to the end - see qox_tx_rate_calc().

> 
> Signed-off-by: Roger Quadros <rogerq@kernel.org>
> ---

Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>

[PATCH v8 net-next 07/11] net: ethernet: am65-cpsw: Move register definitions to header file

[Roger Quadros]

Move register definitions to header file. No functional change.

Signed-off-by: Roger Quadros <rogerq@kernel.org>
---
 drivers/net/ethernet/ti/am65-cpsw-qos.c | 35 -------------------------
 drivers/net/ethernet/ti/am65-cpsw-qos.h | 35 +++++++++++++++++++++++++
 2 files changed, 35 insertions(+), 35 deletions(-)

Changelog:

v7: no change
v6: initial commit

diff --git a/drivers/net/ethernet/ti/am65-cpsw-qos.c b/drivers/net/ethernet/ti/am65-cpsw-qos.c
index 862a98211caa..9f0a05e763d1 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-qos.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-qos.c
@@ -15,41 +15,6 @@
 #include "am65-cpts.h"
 #include "cpsw_ale.h"
 
-#define AM65_CPSW_REG_CTL			0x004
-#define AM65_CPSW_PN_REG_CTL			0x004
-#define AM65_CPSW_PN_REG_FIFO_STATUS		0x050
-#define AM65_CPSW_PN_REG_EST_CTL		0x060
-#define AM65_CPSW_PN_REG_PRI_CIR(pri)		(0x140 + 4 * (pri))
-
-/* AM65_CPSW_REG_CTL register fields */
-#define AM65_CPSW_CTL_EST_EN			BIT(18)
-
-/* AM65_CPSW_PN_REG_CTL register fields */
-#define AM65_CPSW_PN_CTL_EST_PORT_EN		BIT(17)
-
-/* AM65_CPSW_PN_REG_EST_CTL register fields */
-#define AM65_CPSW_PN_EST_ONEBUF			BIT(0)
-#define AM65_CPSW_PN_EST_BUFSEL			BIT(1)
-#define AM65_CPSW_PN_EST_TS_EN			BIT(2)
-#define AM65_CPSW_PN_EST_TS_FIRST		BIT(3)
-#define AM65_CPSW_PN_EST_ONEPRI			BIT(4)
-#define AM65_CPSW_PN_EST_TS_PRI_MSK		GENMASK(7, 5)
-
-/* AM65_CPSW_PN_REG_FIFO_STATUS register fields */
-#define AM65_CPSW_PN_FST_TX_PRI_ACTIVE_MSK	GENMASK(7, 0)
-#define AM65_CPSW_PN_FST_TX_E_MAC_ALLOW_MSK	GENMASK(15, 8)
-#define AM65_CPSW_PN_FST_EST_CNT_ERR		BIT(16)
-#define AM65_CPSW_PN_FST_EST_ADD_ERR		BIT(17)
-#define AM65_CPSW_PN_FST_EST_BUFACT		BIT(18)
-
-/* EST FETCH COMMAND RAM */
-#define AM65_CPSW_FETCH_RAM_CMD_NUM		0x80
-#define AM65_CPSW_FETCH_CNT_MSK			GENMASK(21, 8)
-#define AM65_CPSW_FETCH_CNT_MAX			(AM65_CPSW_FETCH_CNT_MSK >> 8)
-#define AM65_CPSW_FETCH_CNT_OFFSET		8
-#define AM65_CPSW_FETCH_ALLOW_MSK		GENMASK(7, 0)
-#define AM65_CPSW_FETCH_ALLOW_MAX		AM65_CPSW_FETCH_ALLOW_MSK
-
 enum timer_act {
 	TACT_PROG,		/* need program timer */
 	TACT_NEED_STOP,		/* need stop first */
diff --git a/drivers/net/ethernet/ti/am65-cpsw-qos.h b/drivers/net/ethernet/ti/am65-cpsw-qos.h
index be4987eb8c51..e3259f27baad 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-qos.h
+++ b/drivers/net/ethernet/ti/am65-cpsw-qos.h
@@ -31,6 +31,41 @@ struct am65_cpsw_qos {
 	struct am65_cpsw_ale_ratelimit ale_mc_ratelimit;
 };
 
+#define AM65_CPSW_REG_CTL			0x004
+#define AM65_CPSW_PN_REG_CTL			0x004
+#define AM65_CPSW_PN_REG_FIFO_STATUS		0x050
+#define AM65_CPSW_PN_REG_EST_CTL		0x060
+#define AM65_CPSW_PN_REG_PRI_CIR(pri)		(0x140 + 4 * (pri))
+
+/* AM65_CPSW_REG_CTL register fields */
+#define AM65_CPSW_CTL_EST_EN			BIT(18)
+
+/* AM65_CPSW_PN_REG_CTL register fields */
+#define AM65_CPSW_PN_CTL_EST_PORT_EN		BIT(17)
+
+/* AM65_CPSW_PN_REG_EST_CTL register fields */
+#define AM65_CPSW_PN_EST_ONEBUF			BIT(0)
+#define AM65_CPSW_PN_EST_BUFSEL			BIT(1)
+#define AM65_CPSW_PN_EST_TS_EN			BIT(2)
+#define AM65_CPSW_PN_EST_TS_FIRST		BIT(3)
+#define AM65_CPSW_PN_EST_ONEPRI			BIT(4)
+#define AM65_CPSW_PN_EST_TS_PRI_MSK		GENMASK(7, 5)
+
+/* AM65_CPSW_PN_REG_FIFO_STATUS register fields */
+#define AM65_CPSW_PN_FST_TX_PRI_ACTIVE_MSK	GENMASK(7, 0)
+#define AM65_CPSW_PN_FST_TX_E_MAC_ALLOW_MSK	GENMASK(15, 8)
+#define AM65_CPSW_PN_FST_EST_CNT_ERR		BIT(16)
+#define AM65_CPSW_PN_FST_EST_ADD_ERR		BIT(17)
+#define AM65_CPSW_PN_FST_EST_BUFACT		BIT(18)
+
+/* EST FETCH COMMAND RAM */
+#define AM65_CPSW_FETCH_RAM_CMD_NUM		0x80
+#define AM65_CPSW_FETCH_CNT_MSK			GENMASK(21, 8)
+#define AM65_CPSW_FETCH_CNT_MAX			(AM65_CPSW_FETCH_CNT_MSK >> 8)
+#define AM65_CPSW_FETCH_CNT_OFFSET		8
+#define AM65_CPSW_FETCH_ALLOW_MSK		GENMASK(7, 0)
+#define AM65_CPSW_FETCH_ALLOW_MAX		AM65_CPSW_FETCH_ALLOW_MSK
+
 #if IS_ENABLED(CONFIG_TI_AM65_CPSW_QOS)
 int am65_cpsw_qos_ndo_setup_tc(struct net_device *ndev, enum tc_setup_type type,
 			       void *type_data);
-- 
2.34.1

Re: [PATCH v8 net-next 07/11] net: ethernet: am65-cpsw: Move register definitions to header file

[Vladimir Oltean]

On Wed, Dec 13, 2023 at 01:07:17PM +0200, Roger Quadros wrote:
> Move register definitions to header file. No functional change.
> 
> Signed-off-by: Roger Quadros <rogerq@kernel.org>
> ---

Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>

[PATCH v8 net-next 08/11] net: ethernet: ti: am65-cpsw: add mqprio qdisc offload in channel mode

[Roger Quadros]

From: Grygorii Strashko <grygorii.strashko@ti.com>

This patch adds MQPRIO Qdisc offload in full 'channel' mode which allows
not only setting up pri:tc mapping, but also configuring TX shapers
(rate-limiting) on external port FIFOs.

The MQPRIO Qdisc offload is expected to work with or without VLAN/priority
tagged packets.

The CPSW external Port FIFO has 8 Priority queues. The rate-limit can be
set for each of these priority queues. Which Priority queue a packet is
assigned to depends on PN_REG_TX_PRI_MAP register which maps header
priority to switch priority.

The header priority of a packet is assigned via the RX_PRI_MAP_REG which
maps packet priority to header priority.

The packet priority is either the VLAN priority (for VLAN tagged packets)
or the thread/channel offset.

For simplicity, we assign the same priority queue to all queues of a
Traffic Class so it can be rate-limited correctly.

Configuration example:
 ethtool -L eth1 tx 5
 ethtool --set-priv-flags eth1 p0-rx-ptype-rrobin off

 tc qdisc add dev eth1 parent root handle 100: mqprio num_tc 3 \
 map 0 0 1 2 0 0 0 0 0 0 0 0 0 0 0 0 \
 queues 1@0 1@1 1@2 hw 1 mode channel \
 shaper bw_rlimit min_rate 0 100mbit 200mbit max_rate 0 101mbit 202mbit

 tc qdisc replace dev eth2 handle 100: parent root mqprio num_tc 1 \
 map 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 queues 1@0 hw 1

 ip link add link eth1 name eth1.100 type vlan id 100
 ip link set eth1.100 type vlan egress 0:0 1:1 2:2 3:3 4:4 5:5 6:6 7:7

In the above example two ports share the same TX CPPI queue 0 for low
priority traffic. 3 traffic classes are defined for eth1 and mapped to:
TC0 - low priority, TX CPPI queue 0 -> ext Port 1 fifo0, no rate limit
TC1 - prio 2, TX CPPI queue 1 -> ext Port 1 fifo1, CIR=100Mbit/s, EIR=1Mbit/s
TC2 - prio 3, TX CPPI queue 2 -> ext Port 1 fifo2, CIR=200Mbit/s, EIR=2Mbit/s

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Roger Quadros <rogerq@kernel.org>
---
 drivers/net/ethernet/ti/Kconfig          |   3 +-
 drivers/net/ethernet/ti/am65-cpsw-nuss.c |   3 +
 drivers/net/ethernet/ti/am65-cpsw-qos.c  | 255 ++++++++++++++++++++++-
 drivers/net/ethernet/ti/am65-cpsw-qos.h  |  20 ++
 4 files changed, 277 insertions(+), 4 deletions(-)

Changelog

v8:
- optimize TO_MBPS()
- mention MQPRIO support in TI_AM65_CPSW_QOS Kconfig help text

v7:
- don't return -EINVAL if num_tc is 0 in setup

v6:
- fixed typo in comment "packet_priorit"->"packet_priority"
- don't call netif_set_real_num_tx_queues() from am65_cpsw_reset_tc_mqprio()
- reset TX queue priorities to zero at init. AM65_CPSW_PN_REG_TX_PRI_MAP = 0
- don't ignore return value in am65_cpsw_setup_mqprio()

v5:
- fix warning at NL_SET_ERR_MSG_FMT_MOD()

v4:
- use macro TO_MBPS()
- use bool for 'shaper_en'
- split code movement into separate patch
- drop rate vs link speed check as mqprio can be set at link down
- call am65_cpsw_reset_tc_mqprio() after pm_runtime_get_sync()

v2:
- clean up commit message
- avoid forward declarations
- use tc_mqprio_qopt_offload :: extack for error message reporting
- avoid unnecessary mqprio->qopt.hw == 0 checks
- set struct tc_mqprio_caps :: validate_queue_counts = true and get rid of
  am65_cpsw_mqprio_verify()
- do not modify AM65_CPSW_P0_REG_RX_PRI_MAP register.


diff --git a/drivers/net/ethernet/ti/Kconfig b/drivers/net/ethernet/ti/Kconfig
index 49cd96c4f532..f76532443298 100644
--- a/drivers/net/ethernet/ti/Kconfig
+++ b/drivers/net/ethernet/ti/Kconfig
@@ -139,7 +139,8 @@ config TI_AM65_CPSW_QOS
 	depends on TI_K3_AM65_CPSW_NUSS && NET_SCH_TAPRIO && TI_K3_AM65_CPTS
 	help
 	  This option enables QoS offload features in AM65 CPSW like
-	  Time Aware Shaper (TAS) / Enhanced Scheduled Traffic (EST).
+	  Time Aware Shaper (TAS) / Enhanced Scheduled Traffic (EST)
+	  and MQPRIO qdisc offload.
 	  The EST scheduler runs on CPTS and the TAS/EST schedule is
 	  updated in the Fetch RAM memory of the CPSW.
 
diff --git a/drivers/net/ethernet/ti/am65-cpsw-nuss.c b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
index 7651f90f51f2..2e374745f239 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-nuss.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
@@ -2127,6 +2127,9 @@ static int am65_cpsw_nuss_init_slave_ports(struct am65_cpsw_common *common)
 				dev_err(dev, "Use random MAC address\n");
 			}
 		}
+
+		/* Reset all Queue priorities to 0 */
+		writel(0, port->port_base + AM65_CPSW_PN_REG_TX_PRI_MAP);
 	}
 	of_node_put(node);
 
diff --git a/drivers/net/ethernet/ti/am65-cpsw-qos.c b/drivers/net/ethernet/ti/am65-cpsw-qos.c
index 9f0a05e763d1..65d69bd890ee 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-qos.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-qos.c
@@ -7,6 +7,7 @@
  */
 
 #include <linux/pm_runtime.h>
+#include <linux/math.h>
 #include <linux/time.h>
 #include <net/pkt_cls.h>
 
@@ -15,6 +16,8 @@
 #include "am65-cpts.h"
 #include "cpsw_ale.h"
 
+#define TO_MBPS(x)	DIV_ROUND_UP((x), BYTES_PER_MBIT)
+
 enum timer_act {
 	TACT_PROG,		/* need program timer */
 	TACT_NEED_STOP,		/* need stop first */
@@ -31,6 +34,232 @@ am65_cpsw_qos_tx_rate_calc(u32 rate_mbps, unsigned long bus_freq)
 	return ir;
 }
 
+static void am65_cpsw_tx_pn_shaper_reset(struct am65_cpsw_port *port)
+{
+	int prio;
+
+	for (prio = 0; prio < AM65_CPSW_PN_FIFO_PRIO_NUM; prio++) {
+		writel(0, port->port_base + AM65_CPSW_PN_REG_PRI_CIR(prio));
+		writel(0, port->port_base + AM65_CPSW_PN_REG_PRI_EIR(prio));
+	}
+}
+
+static void am65_cpsw_tx_pn_shaper_apply(struct am65_cpsw_port *port)
+{
+	struct am65_cpsw_mqprio *p_mqprio = &port->qos.mqprio;
+	struct am65_cpsw_common *common = port->common;
+	struct tc_mqprio_qopt_offload *mqprio;
+	bool enable, shaper_susp = false;
+	u32 rate_mbps;
+	int tc, prio;
+
+	mqprio = &p_mqprio->mqprio_hw;
+	/* takes care of no link case as well */
+	if (p_mqprio->max_rate_total > port->qos.link_speed)
+		shaper_susp = true;
+
+	am65_cpsw_tx_pn_shaper_reset(port);
+
+	enable = p_mqprio->shaper_en && !shaper_susp;
+	if (!enable)
+		return;
+
+	/* Rate limit is specified per Traffic Class but
+	 * for CPSW, rate limit can be applied per priority
+	 * at port FIFO.
+	 *
+	 * We have assigned the same priority (TCn) to all queues
+	 * of a Traffic Class so they share the same shaper
+	 * bandwidth.
+	 */
+	for (tc = 0; tc < mqprio->qopt.num_tc; tc++) {
+		prio = tc;
+
+		rate_mbps = TO_MBPS(mqprio->min_rate[tc]);
+		rate_mbps = am65_cpsw_qos_tx_rate_calc(rate_mbps,
+						       common->bus_freq);
+		writel(rate_mbps,
+		       port->port_base + AM65_CPSW_PN_REG_PRI_CIR(prio));
+
+		rate_mbps = 0;
+
+		if (mqprio->max_rate[tc]) {
+			rate_mbps = mqprio->max_rate[tc] - mqprio->min_rate[tc];
+			rate_mbps = TO_MBPS(rate_mbps);
+			rate_mbps = am65_cpsw_qos_tx_rate_calc(rate_mbps,
+							       common->bus_freq);
+		}
+
+		writel(rate_mbps,
+		       port->port_base + AM65_CPSW_PN_REG_PRI_EIR(prio));
+	}
+}
+
+static int am65_cpsw_mqprio_verify_shaper(struct am65_cpsw_port *port,
+					  struct tc_mqprio_qopt_offload *mqprio)
+{
+	struct am65_cpsw_mqprio *p_mqprio = &port->qos.mqprio;
+	struct netlink_ext_ack *extack = mqprio->extack;
+	u64 min_rate_total = 0, max_rate_total = 0;
+	u32 min_rate_msk = 0, max_rate_msk = 0;
+	bool has_min_rate, has_max_rate;
+	int num_tc, i;
+
+	if (!(mqprio->flags & TC_MQPRIO_F_SHAPER))
+		return 0;
+
+	if (mqprio->shaper != TC_MQPRIO_SHAPER_BW_RATE)
+		return 0;
+
+	has_min_rate = !!(mqprio->flags & TC_MQPRIO_F_MIN_RATE);
+	has_max_rate = !!(mqprio->flags & TC_MQPRIO_F_MAX_RATE);
+
+	if (!has_min_rate && has_max_rate) {
+		NL_SET_ERR_MSG_MOD(extack, "min_rate is required with max_rate");
+		return -EOPNOTSUPP;
+	}
+
+	if (!has_min_rate)
+		return 0;
+
+	num_tc = mqprio->qopt.num_tc;
+
+	for (i = num_tc - 1; i >= 0; i--) {
+		u32 ch_msk;
+
+		if (mqprio->min_rate[i])
+			min_rate_msk |= BIT(i);
+		min_rate_total +=  mqprio->min_rate[i];
+
+		if (has_max_rate) {
+			if (mqprio->max_rate[i])
+				max_rate_msk |= BIT(i);
+			max_rate_total +=  mqprio->max_rate[i];
+
+			if (!mqprio->min_rate[i] && mqprio->max_rate[i]) {
+				NL_SET_ERR_MSG_FMT_MOD(extack,
+						       "TX tc%d rate max>0 but min=0\n",
+						       i);
+				return -EINVAL;
+			}
+
+			if (mqprio->max_rate[i] &&
+			    mqprio->max_rate[i] < mqprio->min_rate[i]) {
+				NL_SET_ERR_MSG_FMT_MOD(extack,
+						       "TX tc%d rate min(%llu)>max(%llu)\n",
+						       i, mqprio->min_rate[i],
+						       mqprio->max_rate[i]);
+				return -EINVAL;
+			}
+		}
+
+		ch_msk = GENMASK(num_tc - 1, i);
+		if ((min_rate_msk & BIT(i)) && (min_rate_msk ^ ch_msk)) {
+			NL_SET_ERR_MSG_FMT_MOD(extack,
+					       "Min rate must be set sequentially hi->lo tx_rate_msk%x\n",
+					       min_rate_msk);
+			return -EINVAL;
+		}
+
+		if ((max_rate_msk & BIT(i)) && (max_rate_msk ^ ch_msk)) {
+			NL_SET_ERR_MSG_FMT_MOD(extack,
+					       "Max rate must be set sequentially hi->lo tx_rate_msk%x\n",
+					       max_rate_msk);
+			return -EINVAL;
+		}
+	}
+
+	min_rate_total = TO_MBPS(min_rate_total);
+	max_rate_total = TO_MBPS(max_rate_total);
+
+	p_mqprio->shaper_en = true;
+	p_mqprio->max_rate_total = max_t(u64, min_rate_total, max_rate_total);
+
+	return 0;
+}
+
+static void am65_cpsw_reset_tc_mqprio(struct net_device *ndev)
+{
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+	struct am65_cpsw_mqprio *p_mqprio = &port->qos.mqprio;
+
+	p_mqprio->shaper_en = false;
+	p_mqprio->max_rate_total = 0;
+
+	am65_cpsw_tx_pn_shaper_reset(port);
+	netdev_reset_tc(ndev);
+
+	/* Reset all Queue priorities to 0 */
+	writel(0, port->port_base + AM65_CPSW_PN_REG_TX_PRI_MAP);
+}
+
+static int am65_cpsw_setup_mqprio(struct net_device *ndev, void *type_data)
+{
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+	struct am65_cpsw_mqprio *p_mqprio = &port->qos.mqprio;
+	struct tc_mqprio_qopt_offload *mqprio = type_data;
+	struct am65_cpsw_common *common = port->common;
+	struct tc_mqprio_qopt *qopt = &mqprio->qopt;
+	int i, tc, offset, count, prio, ret;
+	u8 num_tc = qopt->num_tc;
+	u32 tx_prio_map = 0;
+
+	memcpy(&p_mqprio->mqprio_hw, mqprio, sizeof(*mqprio));
+
+	ret = pm_runtime_get_sync(common->dev);
+	if (ret < 0) {
+		pm_runtime_put_noidle(common->dev);
+		return ret;
+	}
+
+	if (!num_tc) {
+		am65_cpsw_reset_tc_mqprio(ndev);
+		ret = 0;
+		goto exit_put;
+	}
+
+	ret = am65_cpsw_mqprio_verify_shaper(port, mqprio);
+	if (ret)
+		goto exit_put;
+
+	netdev_set_num_tc(ndev, num_tc);
+
+	/* Multiple Linux priorities can map to a Traffic Class
+	 * A Traffic Class can have multiple contiguous Queues,
+	 * Queues get mapped to Channels (thread_id),
+	 *	if not VLAN tagged, thread_id is used as packet_priority
+	 *	if VLAN tagged. VLAN priority is used as packet_priority
+	 * packet_priority gets mapped to header_priority in p0_rx_pri_map,
+	 * header_priority gets mapped to switch_priority in pn_tx_pri_map.
+	 * As p0_rx_pri_map is left at defaults (0x76543210), we can
+	 * assume that Queue_n gets mapped to header_priority_n. We can then
+	 * set the switch priority in pn_tx_pri_map.
+	 */
+
+	for (tc = 0; tc < num_tc; tc++) {
+		prio = tc;
+
+		/* For simplicity we assign the same priority (TCn) to
+		 * all queues of a Traffic Class.
+		 */
+		for (i = qopt->offset[tc]; i < qopt->offset[tc] + qopt->count[tc]; i++)
+			tx_prio_map |= prio << (4 * i);
+
+		count = qopt->count[tc];
+		offset = qopt->offset[tc];
+		netdev_set_tc_queue(ndev, tc, count, offset);
+	}
+
+	writel(tx_prio_map, port->port_base + AM65_CPSW_PN_REG_TX_PRI_MAP);
+
+	am65_cpsw_tx_pn_shaper_apply(port);
+
+exit_put:
+	pm_runtime_put(common->dev);
+
+	return ret;
+}
+
 static int am65_cpsw_port_est_enabled(struct am65_cpsw_port *port)
 {
 	return port->qos.est_oper || port->qos.est_admin;
@@ -414,6 +643,8 @@ static void am65_cpsw_taprio_destroy(struct net_device *ndev)
 
 	port->qos.est_oper = NULL;
 	port->qos.est_admin = NULL;
+
+	am65_cpsw_reset_tc_mqprio(ndev);
 }
 
 static void am65_cpsw_cp_taprio(struct tc_taprio_qopt_offload *from,
@@ -462,6 +693,10 @@ static int am65_cpsw_taprio_replace(struct net_device *ndev,
 	if (!est_new)
 		return -ENOMEM;
 
+	ret = am65_cpsw_setup_mqprio(ndev, &taprio->mqprio);
+	if (ret)
+		return ret;
+
 	am65_cpsw_cp_taprio(taprio, &est_new->taprio);
 
 	am65_cpsw_est_update_state(ndev);
@@ -505,6 +740,7 @@ static int am65_cpsw_taprio_replace(struct net_device *ndev,
 	return 0;
 
 fail:
+	am65_cpsw_reset_tc_mqprio(ndev);
 	devm_kfree(&ndev->dev, est_new);
 	return ret;
 }
@@ -515,7 +751,6 @@ static void am65_cpsw_est_link_up(struct net_device *ndev, int link_speed)
 	ktime_t cur_time;
 	s64 delta;
 
-	port->qos.link_speed = link_speed;
 	if (!am65_cpsw_port_est_enabled(port))
 		return;
 
@@ -559,6 +794,14 @@ static int am65_cpsw_tc_query_caps(struct net_device *ndev, void *type_data)
 	struct tc_query_caps_base *base = type_data;
 
 	switch (base->type) {
+	case TC_SETUP_QDISC_MQPRIO: {
+		struct tc_mqprio_caps *caps = base->caps;
+
+		caps->validate_queue_counts = true;
+
+		return 0;
+	}
+
 	case TC_SETUP_QDISC_TAPRIO: {
 		struct tc_taprio_caps *caps = base->caps;
 
@@ -857,6 +1100,8 @@ int am65_cpsw_qos_ndo_setup_tc(struct net_device *ndev, enum tc_setup_type type,
 		return am65_cpsw_tc_query_caps(ndev, type_data);
 	case TC_SETUP_QDISC_TAPRIO:
 		return am65_cpsw_setup_taprio(ndev, type_data);
+	case TC_SETUP_QDISC_MQPRIO:
+		return am65_cpsw_setup_mqprio(ndev, type_data);
 	case TC_SETUP_BLOCK:
 		return am65_cpsw_qos_setup_tc_block(ndev, type_data);
 	default:
@@ -868,6 +1113,9 @@ void am65_cpsw_qos_link_up(struct net_device *ndev, int link_speed)
 {
 	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
 
+	port->qos.link_speed = link_speed;
+	am65_cpsw_tx_pn_shaper_apply(port);
+
 	am65_cpsw_est_link_up(ndev, link_speed);
 	port->qos.link_down_time = 0;
 }
@@ -876,8 +1124,9 @@ void am65_cpsw_qos_link_down(struct net_device *ndev)
 {
 	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
 
+	port->qos.link_speed = SPEED_UNKNOWN;
+	am65_cpsw_tx_pn_shaper_apply(port);
+
 	if (!port->qos.link_down_time)
 		port->qos.link_down_time = ktime_get();
-
-	port->qos.link_speed = SPEED_UNKNOWN;
 }
diff --git a/drivers/net/ethernet/ti/am65-cpsw-qos.h b/drivers/net/ethernet/ti/am65-cpsw-qos.h
index e3259f27baad..b54c34c0aa7f 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-qos.h
+++ b/drivers/net/ethernet/ti/am65-cpsw-qos.h
@@ -9,6 +9,7 @@
 #include <net/pkt_sched.h>
 
 struct am65_cpsw_common;
+struct am65_cpsw_port;
 
 struct am65_cpsw_est {
 	int buf;
@@ -16,6 +17,12 @@ struct am65_cpsw_est {
 	struct tc_taprio_qopt_offload taprio;
 };
 
+struct am65_cpsw_mqprio {
+	struct tc_mqprio_qopt_offload mqprio_hw;
+	u64 max_rate_total;
+	bool shaper_en;
+};
+
 struct am65_cpsw_ale_ratelimit {
 	unsigned long cookie;
 	u64 rate_packet_ps;
@@ -26,6 +33,7 @@ struct am65_cpsw_qos {
 	struct am65_cpsw_est *est_oper;
 	ktime_t link_down_time;
 	int link_speed;
+	struct am65_cpsw_mqprio mqprio;
 
 	struct am65_cpsw_ale_ratelimit ale_bc_ratelimit;
 	struct am65_cpsw_ale_ratelimit ale_mc_ratelimit;
@@ -36,6 +44,15 @@ struct am65_cpsw_qos {
 #define AM65_CPSW_PN_REG_FIFO_STATUS		0x050
 #define AM65_CPSW_PN_REG_EST_CTL		0x060
 #define AM65_CPSW_PN_REG_PRI_CIR(pri)		(0x140 + 4 * (pri))
+#define AM65_CPSW_P0_REG_PRI_EIR(pri)		(0x160 + 4 * (pri))
+
+#define AM65_CPSW_PN_REG_CTL			0x004
+#define AM65_CPSW_PN_REG_TX_PRI_MAP		0x018
+#define AM65_CPSW_PN_REG_RX_PRI_MAP		0x020
+#define AM65_CPSW_PN_REG_FIFO_STATUS		0x050
+#define AM65_CPSW_PN_REG_EST_CTL		0x060
+#define AM65_CPSW_PN_REG_PRI_CIR(pri)		(0x140 + 4 * (pri))
+#define AM65_CPSW_PN_REG_PRI_EIR(pri)		(0x160 + 4 * (pri))
 
 /* AM65_CPSW_REG_CTL register fields */
 #define AM65_CPSW_CTL_EST_EN			BIT(18)
@@ -66,6 +83,9 @@ struct am65_cpsw_qos {
 #define AM65_CPSW_FETCH_ALLOW_MSK		GENMASK(7, 0)
 #define AM65_CPSW_FETCH_ALLOW_MAX		AM65_CPSW_FETCH_ALLOW_MSK
 
+/* number of priority queues per port FIFO */
+#define AM65_CPSW_PN_FIFO_PRIO_NUM		8
+
 #if IS_ENABLED(CONFIG_TI_AM65_CPSW_QOS)
 int am65_cpsw_qos_ndo_setup_tc(struct net_device *ndev, enum tc_setup_type type,
 			       void *type_data);
-- 
2.34.1

Re: [PATCH v8 net-next 08/11] net: ethernet: ti: am65-cpsw: add mqprio qdisc offload in channel mode

[Vladimir Oltean]

On Wed, Dec 13, 2023 at 01:07:18PM +0200, Roger Quadros wrote:
> +	for (i = num_tc - 1; i >= 0; i--) {
> +		u32 ch_msk;
> +
> +		if (mqprio->min_rate[i])
> +			min_rate_msk |= BIT(i);
> +		min_rate_total +=  mqprio->min_rate[i];
> +
> +		if (has_max_rate) {
> +			if (mqprio->max_rate[i])
> +				max_rate_msk |= BIT(i);
> +			max_rate_total +=  mqprio->max_rate[i];
> +
> +			if (!mqprio->min_rate[i] && mqprio->max_rate[i]) {
> +				NL_SET_ERR_MSG_FMT_MOD(extack,
> +						       "TX tc%d rate max>0 but min=0\n",
> +						       i);
> +				return -EINVAL;
> +			}
> +
> +			if (mqprio->max_rate[i] &&
> +			    mqprio->max_rate[i] < mqprio->min_rate[i]) {
> +				NL_SET_ERR_MSG_FMT_MOD(extack,
> +						       "TX tc%d rate min(%llu)>max(%llu)\n",
> +						       i, mqprio->min_rate[i],
> +						       mqprio->max_rate[i]);
> +				return -EINVAL;
> +			}
> +		}
> +
> +		ch_msk = GENMASK(num_tc - 1, i);
> +		if ((min_rate_msk & BIT(i)) && (min_rate_msk ^ ch_msk)) {
> +			NL_SET_ERR_MSG_FMT_MOD(extack,
> +					       "Min rate must be set sequentially hi->lo tx_rate_msk%x\n",
> +					       min_rate_msk);
> +			return -EINVAL;
> +		}
> +
> +		if ((max_rate_msk & BIT(i)) && (max_rate_msk ^ ch_msk)) {
> +			NL_SET_ERR_MSG_FMT_MOD(extack,
> +					       "Max rate must be set sequentially hi->lo tx_rate_msk%x\n",
> +					       max_rate_msk);
> +			return -EINVAL;
> +		}
> +	}

Netlink extack strings don't need \n at the end.

[PATCH v8 net-next 09/11] net: ethernet: ti: am65-cpsw-qos: Add Frame Preemption MAC Merge support

[Roger Quadros]

Add driver support for viewing / changing the MAC Merge sublayer
parameters and seeing the verification state machine's current state
via ethtool.

As hardware does not support interrupt notification for verification
events we resort to polling on link up. On link up we try a couple of
times for verification success and if unsuccessful then give up.

The Frame Preemption feature is described in the Technical Reference
Manual [1] in section:
	12.3.1.4.6.7 Intersperced Express Traffic (IET – P802.3br/D2.0)

Due to Silicon Errata i2208 [2] we set limit min IET fragment size to
124 (excluding 4 bytes mCRC).

[1] AM62x TRM - https://www.ti.com/lit/ug/spruiv7a/spruiv7a.pdf
[2] AM62x Silicon Errata - https://www.ti.com/lit/er/sprz487c/sprz487c.pdf

Signed-off-by: Roger Quadros <rogerq@kernel.org>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
---
 drivers/net/ethernet/ti/Kconfig             |   5 +-
 drivers/net/ethernet/ti/am65-cpsw-ethtool.c | 167 +++++++++++++++++++
 drivers/net/ethernet/ti/am65-cpsw-nuss.c    |   2 +
 drivers/net/ethernet/ti/am65-cpsw-nuss.h    |   5 +
 drivers/net/ethernet/ti/am65-cpsw-qos.c     | 175 ++++++++++++++++++++
 drivers/net/ethernet/ti/am65-cpsw-qos.h     | 105 ++++++++++++
 6 files changed, 457 insertions(+), 2 deletions(-)

Changelog:
v8:
- fix build if CONFIG_TI_AM65_CPSW_TAS not enabled
- change CONFIG_TI_AM65_CPSW_TAS to CONFIG_TI_AM65_CPSW_QOS
- mention about Frame Preemption support in TI_AM65_CPSW_QOS Kconfig help
- mentioned IET fragment size excludes 4 bytes mCRC in commit log

v7:
- use else if
- drop FIXME comment
- fix lldp kselftest failure by limiting max_verify_time to spec limit of 128ms.
- now passes all ethtool_mm.sh kselftests (patch 8 required)

v6:
- get mutex around am65_cpsw_iet_commit_preemptible_tcs() in
  am65_cpsw_iet_change_preemptible_tcs()
- use "preemption" instead of "pre-emption"
- call am65_cpsw_setup_mqprio() from within am65_cpsw_setup_taprio()
- Now works with kselftest except the last test which fails

v5:
- No change

v4:
- Rebase and include in the same series as mqprio support.

v3:
- Rebase on top of v6.6-rc1 and mqprio support [1]
- Support ethtool_ops :: get_mm_stats()
- drop unused variables cmn_ctrl and verify_cnt
- make am65_cpsw_iet_link_state_update() and
  am65_cpsw_iet_change_preemptible_tcs() static

[1] https://lore.kernel.org/all/20230918075358.5878-1-rogerq@kernel.org/

v2:
- Use proper control bits for PMAC enable (AM65_CPSW_PN_CTL_IET_PORT_EN)
  and TX enable (AM65_CPSW_PN_IET_MAC_PENABLE)
- Common IET Enable (AM65_CPSW_CTL_IET_EN) is set if any port has
  AM65_CPSW_PN_CTL_IET_PORT_EN set.
- Fix workaround for erratum i2208. i.e. Limit rx_min_frag_size to 124
- Fix am65_cpsw_iet_get_verify_timeout_ms() to default to timeout for
  1G link if link is inactive.
- resize the RX FIFO based on pmac_enabled, not tx_enabled.

diff --git a/drivers/net/ethernet/ti/Kconfig b/drivers/net/ethernet/ti/Kconfig
index f76532443298..be01450c20dc 100644
--- a/drivers/net/ethernet/ti/Kconfig
+++ b/drivers/net/ethernet/ti/Kconfig
@@ -139,8 +139,9 @@ config TI_AM65_CPSW_QOS
 	depends on TI_K3_AM65_CPSW_NUSS && NET_SCH_TAPRIO && TI_K3_AM65_CPTS
 	help
 	  This option enables QoS offload features in AM65 CPSW like
-	  Time Aware Shaper (TAS) / Enhanced Scheduled Traffic (EST)
-	  and MQPRIO qdisc offload.
+	  Time Aware Shaper (TAS) / Enhanced Scheduled Traffic (EST),
+	  MQPRIO qdisc offload and Frame-Preemption MAC Merge / Interspersing
+	  Express Traffic (IET).
 	  The EST scheduler runs on CPTS and the TAS/EST schedule is
 	  updated in the Fetch RAM memory of the CPSW.
 
diff --git a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
index b9e1d568604b..8c3badec219b 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
@@ -11,6 +11,7 @@
 #include <linux/pm_runtime.h>
 
 #include "am65-cpsw-nuss.h"
+#include "am65-cpsw-qos.h"
 #include "cpsw_ale.h"
 #include "am65-cpts.h"
 
@@ -740,6 +741,169 @@ static int am65_cpsw_set_ethtool_priv_flags(struct net_device *ndev, u32 flags)
 	return 0;
 }
 
+static void am65_cpsw_port_iet_rx_enable(struct am65_cpsw_port *port, bool enable)
+{
+	u32 val;
+
+	val = readl(port->port_base + AM65_CPSW_PN_REG_CTL);
+	if (enable)
+		val |= AM65_CPSW_PN_CTL_IET_PORT_EN;
+	else
+		val &= ~AM65_CPSW_PN_CTL_IET_PORT_EN;
+
+	writel(val, port->port_base + AM65_CPSW_PN_REG_CTL);
+	am65_cpsw_iet_common_enable(port->common);
+}
+
+static void am65_cpsw_port_iet_tx_enable(struct am65_cpsw_port *port, bool enable)
+{
+	u32 val;
+
+	val = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+	if (enable)
+		val |= AM65_CPSW_PN_IET_MAC_PENABLE;
+	else
+		val &= ~AM65_CPSW_PN_IET_MAC_PENABLE;
+
+	writel(val, port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+}
+
+static int am65_cpsw_get_mm(struct net_device *ndev, struct ethtool_mm_state *state)
+{
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+	struct am65_cpsw_ndev_priv *priv = netdev_priv(ndev);
+	u32 port_ctrl, iet_ctrl, iet_status;
+	u32 add_frag_size;
+
+	if (!IS_ENABLED(CONFIG_TI_AM65_CPSW_QOS))
+		return -EOPNOTSUPP;
+
+	mutex_lock(&priv->mm_lock);
+
+	iet_ctrl = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+	port_ctrl = readl(port->port_base + AM65_CPSW_PN_REG_CTL);
+
+	state->tx_enabled = !!(iet_ctrl & AM65_CPSW_PN_IET_MAC_PENABLE);
+	state->pmac_enabled = !!(port_ctrl & AM65_CPSW_PN_CTL_IET_PORT_EN);
+
+	iet_status = readl(port->port_base + AM65_CPSW_PN_REG_IET_STATUS);
+
+	if (iet_ctrl & AM65_CPSW_PN_IET_MAC_DISABLEVERIFY)
+		state->verify_status = ETHTOOL_MM_VERIFY_STATUS_DISABLED;
+	else if (iet_status & AM65_CPSW_PN_MAC_VERIFIED)
+		state->verify_status = ETHTOOL_MM_VERIFY_STATUS_SUCCEEDED;
+	else if (iet_status & AM65_CPSW_PN_MAC_VERIFY_FAIL)
+		state->verify_status = ETHTOOL_MM_VERIFY_STATUS_FAILED;
+	else
+		state->verify_status = ETHTOOL_MM_VERIFY_STATUS_UNKNOWN;
+
+	add_frag_size = AM65_CPSW_PN_IET_MAC_GET_ADDFRAGSIZE(iet_ctrl);
+	state->tx_min_frag_size = ethtool_mm_frag_size_add_to_min(add_frag_size);
+
+	/* Errata i2208: RX min fragment size cannot be less than 124 */
+	state->rx_min_frag_size = 124;
+
+	/* FPE active if common tx_enabled and verification success or disabled (forced) */
+	state->tx_active = state->tx_enabled &&
+			   (state->verify_status == ETHTOOL_MM_VERIFY_STATUS_SUCCEEDED ||
+			    state->verify_status == ETHTOOL_MM_VERIFY_STATUS_DISABLED);
+	state->verify_enabled = !(iet_ctrl & AM65_CPSW_PN_IET_MAC_DISABLEVERIFY);
+
+	state->verify_time = port->qos.iet.verify_time_ms;
+
+	/* 802.3-2018 clause 30.14.1.6, says that the aMACMergeVerifyTime
+	 * variable has a range between 1 and 128 ms inclusive. Limit to that.
+	 */
+	state->max_verify_time = 128;
+
+	mutex_unlock(&priv->mm_lock);
+
+	return 0;
+}
+
+static int am65_cpsw_set_mm(struct net_device *ndev, struct ethtool_mm_cfg *cfg,
+			    struct netlink_ext_ack *extack)
+{
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+	struct am65_cpsw_ndev_priv *priv = netdev_priv(ndev);
+	struct am65_cpsw_iet *iet = &port->qos.iet;
+	u32 val, add_frag_size;
+	int err;
+
+	if (!IS_ENABLED(CONFIG_TI_AM65_CPSW_QOS))
+		return -EOPNOTSUPP;
+
+	err = ethtool_mm_frag_size_min_to_add(cfg->tx_min_frag_size, &add_frag_size, extack);
+	if (err)
+		return err;
+
+	mutex_lock(&priv->mm_lock);
+
+	if (cfg->pmac_enabled) {
+		/* change TX & RX FIFO MAX_BLKS as per TRM recommendation */
+		if (!iet->original_max_blks)
+			iet->original_max_blks = readl(port->port_base + AM65_CPSW_PN_REG_MAX_BLKS);
+
+		writel(AM65_CPSW_PN_TX_RX_MAX_BLKS_IET,
+		       port->port_base + AM65_CPSW_PN_REG_MAX_BLKS);
+	} else if (iet->original_max_blks) {
+		/* restore RX & TX FIFO MAX_BLKS */
+		writel(iet->original_max_blks,
+		       port->port_base + AM65_CPSW_PN_REG_MAX_BLKS);
+	}
+
+	am65_cpsw_port_iet_rx_enable(port, cfg->pmac_enabled);
+	am65_cpsw_port_iet_tx_enable(port, cfg->tx_enabled);
+
+	val = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+	if (cfg->verify_enabled) {
+		val &= ~AM65_CPSW_PN_IET_MAC_DISABLEVERIFY;
+		/* Reset Verify state machine. Verification won't start here.
+		 * Verification will be done once link-up.
+		 */
+		val |= AM65_CPSW_PN_IET_MAC_LINKFAIL;
+	} else {
+		val |= AM65_CPSW_PN_IET_MAC_DISABLEVERIFY;
+		/* Clear LINKFAIL to allow verify/response packets */
+		val &= ~AM65_CPSW_PN_IET_MAC_LINKFAIL;
+	}
+
+	val &= ~AM65_CPSW_PN_IET_MAC_MAC_ADDFRAGSIZE_MASK;
+	val |= AM65_CPSW_PN_IET_MAC_SET_ADDFRAGSIZE(add_frag_size);
+	writel(val, port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+
+	/* verify_timeout_count can only be set at valid link */
+	if (cfg->verify_time > 0)
+		port->qos.iet.verify_time_ms = cfg->verify_time;
+	else
+		port->qos.iet.verify_time_ms = 10;
+
+	/* enable/disable preemption based on link status */
+	am65_cpsw_iet_commit_preemptible_tcs(port);
+
+	mutex_unlock(&priv->mm_lock);
+
+	return 0;
+}
+
+static void am65_cpsw_get_mm_stats(struct net_device *ndev,
+				   struct ethtool_mm_stats *s)
+{
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+	void __iomem *base = port->stat_base;
+
+	s->MACMergeFrameAssOkCount = readl(base + AM65_CPSW_STATN_IET_RX_ASSEMBLY_OK);
+	s->MACMergeFrameAssErrorCount = readl(base + AM65_CPSW_STATN_IET_RX_ASSEMBLY_ERROR);
+	s->MACMergeFrameSmdErrorCount = readl(base + AM65_CPSW_STATN_IET_RX_SMD_ERROR);
+	/* CPSW Functional Spec states:
+	 * "The IET stat aMACMergeFragCountRx is derived by adding the
+	 *  Receive Assembly Error count to this value. i.e. AM65_CPSW_STATN_IET_RX_FRAG"
+	 */
+	s->MACMergeFragCountRx = readl(base + AM65_CPSW_STATN_IET_RX_FRAG) + s->MACMergeFrameAssErrorCount;
+	s->MACMergeFragCountTx = readl(base + AM65_CPSW_STATN_IET_TX_FRAG);
+	s->MACMergeHoldCount = readl(base + AM65_CPSW_STATN_IET_TX_HOLD);
+}
+
 const struct ethtool_ops am65_cpsw_ethtool_ops_slave = {
 	.begin			= am65_cpsw_ethtool_op_begin,
 	.complete		= am65_cpsw_ethtool_op_complete,
@@ -769,4 +933,7 @@ const struct ethtool_ops am65_cpsw_ethtool_ops_slave = {
 	.get_eee		= am65_cpsw_get_eee,
 	.set_eee		= am65_cpsw_set_eee,
 	.nway_reset		= am65_cpsw_nway_reset,
+	.get_mm			= am65_cpsw_get_mm,
+	.set_mm			= am65_cpsw_set_mm,
+	.get_mm_stats		= am65_cpsw_get_mm_stats,
 };
diff --git a/drivers/net/ethernet/ti/am65-cpsw-nuss.c b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
index 2e374745f239..41e0046a52d5 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-nuss.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
@@ -2194,6 +2194,8 @@ am65_cpsw_nuss_init_port_ndev(struct am65_cpsw_common *common, u32 port_idx)
 	ndev_priv = netdev_priv(port->ndev);
 	ndev_priv->port = port;
 	ndev_priv->msg_enable = AM65_CPSW_DEBUG;
+	mutex_init(&ndev_priv->mm_lock);
+	port->qos.link_speed = SPEED_UNKNOWN;
 	SET_NETDEV_DEV(port->ndev, dev);
 
 	eth_hw_addr_set(port->ndev, port->slave.mac_addr);
diff --git a/drivers/net/ethernet/ti/am65-cpsw-nuss.h b/drivers/net/ethernet/ti/am65-cpsw-nuss.h
index f3dad2ab9828..1e4a045057fc 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-nuss.h
+++ b/drivers/net/ethernet/ti/am65-cpsw-nuss.h
@@ -145,6 +145,7 @@ struct am65_cpsw_common {
 	bool			pf_p0_rx_ptype_rrobin;
 	struct am65_cpts	*cpts;
 	int			est_enabled;
+	bool			iet_enabled;
 
 	bool		is_emac_mode;
 	u16			br_members;
@@ -170,6 +171,10 @@ struct am65_cpsw_ndev_priv {
 	struct am65_cpsw_port	*port;
 	struct am65_cpsw_ndev_stats __percpu *stats;
 	bool offload_fwd_mark;
+	/* Serialize access to MAC Merge state between ethtool requests
+	 * and link state updates
+	 */
+	struct mutex		mm_lock;
 };
 
 #define am65_ndev_to_priv(ndev) \
diff --git a/drivers/net/ethernet/ti/am65-cpsw-qos.c b/drivers/net/ethernet/ti/am65-cpsw-qos.c
index 65d69bd890ee..3570e22eed6d 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-qos.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-qos.c
@@ -4,11 +4,13 @@
  *
  * quality of service module includes:
  * Enhanced Scheduler Traffic (EST - P802.1Qbv/D2.2)
+ * Interspersed Express Traffic (IET - P802.3br/D2.0)
  */
 
 #include <linux/pm_runtime.h>
 #include <linux/math.h>
 #include <linux/time.h>
+#include <linux/units.h>
 #include <net/pkt_cls.h>
 
 #include "am65-cpsw-nuss.h"
@@ -24,6 +26,8 @@ enum timer_act {
 	TACT_SKIP_PROG,		/* just buffer can be updated */
 };
 
+static void am65_cpsw_iet_change_preemptible_tcs(struct am65_cpsw_port *port, u8 preemptible_tcs);
+
 static u32
 am65_cpsw_qos_tx_rate_calc(u32 rate_mbps, unsigned long bus_freq)
 {
@@ -191,6 +195,8 @@ static void am65_cpsw_reset_tc_mqprio(struct net_device *ndev)
 
 	/* Reset all Queue priorities to 0 */
 	writel(0, port->port_base + AM65_CPSW_PN_REG_TX_PRI_MAP);
+
+	am65_cpsw_iet_change_preemptible_tcs(port, 0);
 }
 
 static int am65_cpsw_setup_mqprio(struct net_device *ndev, void *type_data)
@@ -253,6 +259,7 @@ static int am65_cpsw_setup_mqprio(struct net_device *ndev, void *type_data)
 	writel(tx_prio_map, port->port_base + AM65_CPSW_PN_REG_TX_PRI_MAP);
 
 	am65_cpsw_tx_pn_shaper_apply(port);
+	am65_cpsw_iet_change_preemptible_tcs(port, mqprio->preemptible_tcs);
 
 exit_put:
 	pm_runtime_put(common->dev);
@@ -260,6 +267,171 @@ static int am65_cpsw_setup_mqprio(struct net_device *ndev, void *type_data)
 	return ret;
 }
 
+static int am65_cpsw_iet_set_verify_timeout_count(struct am65_cpsw_port *port)
+{
+	int verify_time_ms = port->qos.iet.verify_time_ms;
+	u32 val;
+
+	/* The number of wireside clocks contained in the verify
+	 * timeout counter. The default is 0x1312d0
+	 * (10ms at 125Mhz in 1G mode).
+	 */
+	val = 125 * HZ_PER_MHZ;	/* assuming 125MHz wireside clock */
+
+	val /= MILLIHZ_PER_HZ;		/* count per ms timeout */
+	val *= verify_time_ms;		/* count for timeout ms */
+
+	if (val > AM65_CPSW_PN_MAC_VERIFY_CNT_MASK)
+		return -EINVAL;
+
+	writel(val, port->port_base + AM65_CPSW_PN_REG_IET_VERIFY);
+
+	return 0;
+}
+
+static int am65_cpsw_iet_verify_wait(struct am65_cpsw_port *port)
+{
+	u32 ctrl, status;
+	int try;
+
+	try = 20;
+	do {
+		/* Reset the verify state machine by writing 1
+		 * to LINKFAIL
+		 */
+		ctrl = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+		ctrl |= AM65_CPSW_PN_IET_MAC_LINKFAIL;
+		writel(ctrl, port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+
+		/* Clear MAC_LINKFAIL bit to start Verify. */
+		ctrl = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+		ctrl &= ~AM65_CPSW_PN_IET_MAC_LINKFAIL;
+		writel(ctrl, port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+
+		msleep(port->qos.iet.verify_time_ms);
+
+		status = readl(port->port_base + AM65_CPSW_PN_REG_IET_STATUS);
+		if (status & AM65_CPSW_PN_MAC_VERIFIED)
+			return 0;
+
+		if (status & AM65_CPSW_PN_MAC_VERIFY_FAIL) {
+			netdev_dbg(port->ndev,
+				   "MAC Merge verify failed, trying again\n");
+			continue;
+		}
+
+		if (status & AM65_CPSW_PN_MAC_RESPOND_ERR) {
+			netdev_dbg(port->ndev, "MAC Merge respond error\n");
+			return -ENODEV;
+		}
+
+		if (status & AM65_CPSW_PN_MAC_VERIFY_ERR) {
+			netdev_dbg(port->ndev, "MAC Merge verify error\n");
+			return -ENODEV;
+		}
+	} while (try-- > 0);
+
+	netdev_dbg(port->ndev, "MAC Merge verify timeout\n");
+	return -ETIMEDOUT;
+}
+
+static void am65_cpsw_iet_set_preempt_mask(struct am65_cpsw_port *port, u8 preemptible_tcs)
+{
+	u32 val;
+
+	val = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+	val &= ~AM65_CPSW_PN_IET_MAC_PREMPT_MASK;
+	val |= AM65_CPSW_PN_IET_MAC_SET_PREEMPT(preemptible_tcs);
+	writel(val, port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+}
+
+/* enable common IET_ENABLE only if at least 1 port has rx IET enabled.
+ * UAPI doesn't allow tx enable without rx enable.
+ */
+void am65_cpsw_iet_common_enable(struct am65_cpsw_common *common)
+{
+	struct am65_cpsw_port *port;
+	bool rx_enable = false;
+	u32 val;
+	int i;
+
+	for (i = 0; i < common->port_num; i++) {
+		port = &common->ports[i];
+		val = readl(port->port_base + AM65_CPSW_PN_REG_CTL);
+		rx_enable = !!(val & AM65_CPSW_PN_CTL_IET_PORT_EN);
+		if (rx_enable)
+			break;
+	}
+
+	val = readl(common->cpsw_base + AM65_CPSW_REG_CTL);
+
+	if (rx_enable)
+		val |= AM65_CPSW_CTL_IET_EN;
+	else
+		val &= ~AM65_CPSW_CTL_IET_EN;
+
+	writel(val, common->cpsw_base + AM65_CPSW_REG_CTL);
+	common->iet_enabled = rx_enable;
+}
+
+/* CPSW does not have an IRQ to notify changes to the MAC Merge TX status
+ * (active/inactive), but the preemptible traffic classes should only be
+ * committed to hardware once TX is active. Resort to polling.
+ */
+void am65_cpsw_iet_commit_preemptible_tcs(struct am65_cpsw_port *port)
+{
+	u8 preemptible_tcs;
+	int err;
+	u32 val;
+
+	if (port->qos.link_speed == SPEED_UNKNOWN)
+		return;
+
+	val = readl(port->port_base + AM65_CPSW_PN_REG_CTL);
+	if (!(val & AM65_CPSW_PN_CTL_IET_PORT_EN))
+		return;
+
+	/* update common IET enable */
+	am65_cpsw_iet_common_enable(port->common);
+
+	/* update verify count */
+	err = am65_cpsw_iet_set_verify_timeout_count(port);
+	if (err) {
+		netdev_err(port->ndev, "couldn't set verify count: %d\n", err);
+		return;
+	}
+
+	val = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+	if (!(val & AM65_CPSW_PN_IET_MAC_DISABLEVERIFY)) {
+		err = am65_cpsw_iet_verify_wait(port);
+		if (err)
+			return;
+	}
+
+	preemptible_tcs = port->qos.iet.preemptible_tcs;
+	am65_cpsw_iet_set_preempt_mask(port, preemptible_tcs);
+}
+
+static void am65_cpsw_iet_change_preemptible_tcs(struct am65_cpsw_port *port, u8 preemptible_tcs)
+{
+	struct am65_cpsw_ndev_priv *priv = am65_ndev_to_priv(port->ndev);
+
+	port->qos.iet.preemptible_tcs = preemptible_tcs;
+	mutex_lock(&priv->mm_lock);
+	am65_cpsw_iet_commit_preemptible_tcs(port);
+	mutex_unlock(&priv->mm_lock);
+}
+
+static void am65_cpsw_iet_link_state_update(struct net_device *ndev)
+{
+	struct am65_cpsw_ndev_priv *priv = am65_ndev_to_priv(ndev);
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+
+	mutex_lock(&priv->mm_lock);
+	am65_cpsw_iet_commit_preemptible_tcs(port);
+	mutex_unlock(&priv->mm_lock);
+}
+
 static int am65_cpsw_port_est_enabled(struct am65_cpsw_port *port)
 {
 	return port->qos.est_oper || port->qos.est_admin;
@@ -736,6 +908,7 @@ static int am65_cpsw_taprio_replace(struct net_device *ndev,
 
 	devm_kfree(&ndev->dev, port->qos.est_admin);
 	port->qos.est_admin = est_new;
+	am65_cpsw_iet_change_preemptible_tcs(port, taprio->mqprio.preemptible_tcs);
 
 	return 0;
 
@@ -1115,6 +1288,7 @@ void am65_cpsw_qos_link_up(struct net_device *ndev, int link_speed)
 
 	port->qos.link_speed = link_speed;
 	am65_cpsw_tx_pn_shaper_apply(port);
+	am65_cpsw_iet_link_state_update(ndev);
 
 	am65_cpsw_est_link_up(ndev, link_speed);
 	port->qos.link_down_time = 0;
@@ -1126,6 +1300,7 @@ void am65_cpsw_qos_link_down(struct net_device *ndev)
 
 	port->qos.link_speed = SPEED_UNKNOWN;
 	am65_cpsw_tx_pn_shaper_apply(port);
+	am65_cpsw_iet_link_state_update(ndev);
 
 	if (!port->qos.link_down_time)
 		port->qos.link_down_time = ktime_get();
diff --git a/drivers/net/ethernet/ti/am65-cpsw-qos.h b/drivers/net/ethernet/ti/am65-cpsw-qos.h
index b54c34c0aa7f..b328e56c5b2b 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-qos.h
+++ b/drivers/net/ethernet/ti/am65-cpsw-qos.h
@@ -23,6 +23,12 @@ struct am65_cpsw_mqprio {
 	bool shaper_en;
 };
 
+struct am65_cpsw_iet {
+	u8 preemptible_tcs;
+	u32 original_max_blks;
+	int verify_time_ms;
+};
+
 struct am65_cpsw_ale_ratelimit {
 	unsigned long cookie;
 	u64 rate_packet_ps;
@@ -34,6 +40,7 @@ struct am65_cpsw_qos {
 	ktime_t link_down_time;
 	int link_speed;
 	struct am65_cpsw_mqprio mqprio;
+	struct am65_cpsw_iet iet;
 
 	struct am65_cpsw_ale_ratelimit ale_bc_ratelimit;
 	struct am65_cpsw_ale_ratelimit ale_mc_ratelimit;
@@ -93,6 +100,8 @@ void am65_cpsw_qos_link_up(struct net_device *ndev, int link_speed);
 void am65_cpsw_qos_link_down(struct net_device *ndev);
 int am65_cpsw_qos_ndo_tx_p0_set_maxrate(struct net_device *ndev, int queue, u32 rate_mbps);
 void am65_cpsw_qos_tx_p0_rate_init(struct am65_cpsw_common *common);
+void am65_cpsw_iet_commit_preemptible_tcs(struct am65_cpsw_port *port);
+void am65_cpsw_iet_common_enable(struct am65_cpsw_common *common);
 #else
 static inline int am65_cpsw_qos_ndo_setup_tc(struct net_device *ndev,
 					     enum tc_setup_type type,
@@ -117,6 +126,102 @@ static inline int am65_cpsw_qos_ndo_tx_p0_set_maxrate(struct net_device *ndev,
 
 static inline void am65_cpsw_qos_tx_p0_rate_init(struct am65_cpsw_common *common)
 { }
+static inline void am65_cpsw_iet_commit_preemptible_tcs(struct am65_cpsw_port *port)
+{ }
+static inline void am65_cpsw_iet_common_enable(struct am65_cpsw_common *common)
+{ }
 #endif
 
+#define AM65_CPSW_REG_CTL			0x004
+#define AM65_CPSW_PN_REG_CTL			0x004
+#define AM65_CPSW_PN_REG_MAX_BLKS		0x008
+#define AM65_CPSW_PN_REG_TX_PRI_MAP		0x018
+#define AM65_CPSW_PN_REG_RX_PRI_MAP		0x020
+#define AM65_CPSW_PN_REG_IET_CTRL		0x040
+#define AM65_CPSW_PN_REG_IET_STATUS		0x044
+#define AM65_CPSW_PN_REG_IET_VERIFY		0x048
+#define AM65_CPSW_PN_REG_FIFO_STATUS		0x050
+#define AM65_CPSW_PN_REG_EST_CTL		0x060
+#define AM65_CPSW_PN_REG_PRI_CIR(pri)		(0x140 + 4 * (pri))
+#define AM65_CPSW_PN_REG_PRI_EIR(pri)		(0x160 + 4 * (pri))
+
+/* AM65_CPSW_REG_CTL register fields */
+#define AM65_CPSW_CTL_IET_EN			BIT(17)
+#define AM65_CPSW_CTL_EST_EN			BIT(18)
+
+/* AM65_CPSW_PN_REG_CTL register fields */
+#define AM65_CPSW_PN_CTL_IET_PORT_EN		BIT(16)
+#define AM65_CPSW_PN_CTL_EST_PORT_EN		BIT(17)
+
+/* AM65_CPSW_PN_REG_EST_CTL register fields */
+#define AM65_CPSW_PN_EST_ONEBUF			BIT(0)
+#define AM65_CPSW_PN_EST_BUFSEL			BIT(1)
+#define AM65_CPSW_PN_EST_TS_EN			BIT(2)
+#define AM65_CPSW_PN_EST_TS_FIRST		BIT(3)
+#define AM65_CPSW_PN_EST_ONEPRI			BIT(4)
+#define AM65_CPSW_PN_EST_TS_PRI_MSK		GENMASK(7, 5)
+
+/* AM65_CPSW_PN_REG_IET_CTRL register fields */
+#define AM65_CPSW_PN_IET_MAC_PENABLE		BIT(0)
+#define AM65_CPSW_PN_IET_MAC_DISABLEVERIFY	BIT(2)
+#define AM65_CPSW_PN_IET_MAC_LINKFAIL		BIT(3)
+#define AM65_CPSW_PN_IET_MAC_MAC_ADDFRAGSIZE_MASK	GENMASK(10, 8)
+#define AM65_CPSW_PN_IET_MAC_MAC_ADDFRAGSIZE_OFFSET	8
+#define AM65_CPSW_PN_IET_MAC_PREMPT_MASK		GENMASK(23, 16)
+#define AM65_CPSW_PN_IET_MAC_PREMPT_OFFSET		16
+
+#define AM65_CPSW_PN_IET_MAC_SET_ADDFRAGSIZE(n)	(((n) << AM65_CPSW_PN_IET_MAC_MAC_ADDFRAGSIZE_OFFSET) & \
+						  AM65_CPSW_PN_IET_MAC_MAC_ADDFRAGSIZE_MASK)
+#define AM65_CPSW_PN_IET_MAC_GET_ADDFRAGSIZE(n)	(((n) & AM65_CPSW_PN_IET_MAC_MAC_ADDFRAGSIZE_MASK) >> \
+						  AM65_CPSW_PN_IET_MAC_MAC_ADDFRAGSIZE_OFFSET)
+#define AM65_CPSW_PN_IET_MAC_SET_PREEMPT(n)	(((n) << AM65_CPSW_PN_IET_MAC_PREMPT_OFFSET) & \
+						 AM65_CPSW_PN_IET_MAC_PREMPT_MASK)
+#define AM65_CPSW_PN_IET_MAC_GET_PREEMPT(n)	(((n) & AM65_CPSW_PN_IET_MAC_PREMPT_MASK) >> \
+						 AM65_CPSW_PN_IET_MAC_PREMPT_OFFSET)
+
+/* AM65_CPSW_PN_REG_IET_STATUS register fields */
+#define AM65_CPSW_PN_MAC_STATUS			GENMASK(3, 0)
+#define AM65_CPSW_PN_MAC_VERIFIED		BIT(0)
+#define AM65_CPSW_PN_MAC_VERIFY_FAIL		BIT(1)
+#define AM65_CPSW_PN_MAC_RESPOND_ERR		BIT(2)
+#define AM65_CPSW_PN_MAC_VERIFY_ERR		BIT(3)
+
+/* AM65_CPSW_PN_REG_IET_VERIFY register fields */
+#define AM65_CPSW_PN_MAC_VERIFY_CNT_MASK	GENMASK(23, 0)
+#define AM65_CPSW_PN_MAC_GET_VERIFY_CNT(n)	((n) & AM65_CPSW_PN_MAC_VERIFY_CNT_MASK)
+/* 10 msec converted to NSEC */
+#define AM65_CPSW_IET_VERIFY_CNT_MS		(10)
+#define AM65_CPSW_IET_VERIFY_CNT_NS		(AM65_CPSW_IET_VERIFY_CNT_MS * \
+						 NSEC_PER_MSEC)
+
+/* AM65_CPSW_PN_REG_FIFO_STATUS register fields */
+#define AM65_CPSW_PN_FST_TX_PRI_ACTIVE_MSK	GENMASK(7, 0)
+#define AM65_CPSW_PN_FST_TX_E_MAC_ALLOW_MSK	GENMASK(15, 8)
+#define AM65_CPSW_PN_FST_EST_CNT_ERR		BIT(16)
+#define AM65_CPSW_PN_FST_EST_ADD_ERR		BIT(17)
+#define AM65_CPSW_PN_FST_EST_BUFACT		BIT(18)
+
+/* EST FETCH COMMAND RAM */
+#define AM65_CPSW_FETCH_RAM_CMD_NUM		0x80
+#define AM65_CPSW_FETCH_CNT_MSK			GENMASK(21, 8)
+#define AM65_CPSW_FETCH_CNT_MAX			(AM65_CPSW_FETCH_CNT_MSK >> 8)
+#define AM65_CPSW_FETCH_CNT_OFFSET		8
+#define AM65_CPSW_FETCH_ALLOW_MSK		GENMASK(7, 0)
+#define AM65_CPSW_FETCH_ALLOW_MAX		AM65_CPSW_FETCH_ALLOW_MSK
+
+/* AM65_CPSW_PN_REG_MAX_BLKS fields for IET and No IET cases */
+/* 7 blocks for pn_rx_max_blks, 13 for pn_tx_max_blks*/
+#define AM65_CPSW_PN_TX_RX_MAX_BLKS_IET		0xD07
+
+/* Slave IET Stats. register offsets */
+#define AM65_CPSW_STATN_IET_RX_ASSEMBLY_ERROR	0x140
+#define AM65_CPSW_STATN_IET_RX_ASSEMBLY_OK	0x144
+#define AM65_CPSW_STATN_IET_RX_SMD_ERROR	0x148
+#define AM65_CPSW_STATN_IET_RX_FRAG		0x14c
+#define AM65_CPSW_STATN_IET_TX_HOLD		0x150
+#define AM65_CPSW_STATN_IET_TX_FRAG		0x154
+
+/* number of priority queues per port FIFO */
+#define AM65_CPSW_PN_FIFO_PRIO_NUM		8
+
 #endif /* AM65_CPSW_QOS_H_ */
-- 
2.34.1

Re: [PATCH v8 net-next 09/11] net: ethernet: ti: am65-cpsw-qos: Add Frame Preemption MAC Merge support

[Vladimir Oltean]

On Wed, Dec 13, 2023 at 01:07:19PM +0200, Roger Quadros wrote:
> - Fix am65_cpsw_iet_get_verify_timeout_ms() to default to timeout for
>   1G link if link is inactive.
> +
> +static int am65_cpsw_set_mm(struct net_device *ndev, struct ethtool_mm_cfg *cfg,
> +			    struct netlink_ext_ack *extack)
> +{
> +	val = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
> +	if (cfg->verify_enabled) {
> +		val &= ~AM65_CPSW_PN_IET_MAC_DISABLEVERIFY;
> +		/* Reset Verify state machine. Verification won't start here.
> +		 * Verification will be done once link-up.
> +		 */
> +		val |= AM65_CPSW_PN_IET_MAC_LINKFAIL;
> +	} else {
> +		val |= AM65_CPSW_PN_IET_MAC_DISABLEVERIFY;
> +		/* Clear LINKFAIL to allow verify/response packets */
> +		val &= ~AM65_CPSW_PN_IET_MAC_LINKFAIL;
> +	}
> +
> +	val &= ~AM65_CPSW_PN_IET_MAC_MAC_ADDFRAGSIZE_MASK;
> +	val |= AM65_CPSW_PN_IET_MAC_SET_ADDFRAGSIZE(add_frag_size);
> +	writel(val, port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
> +
> +	/* verify_timeout_count can only be set at valid link */
> +	if (cfg->verify_time > 0)
> +		port->qos.iet.verify_time_ms = cfg->verify_time;
> +	else
> +		port->qos.iet.verify_time_ms = 10;

I don't think I understand what the check is for? The netlink policy for
ETHTOOL_A_MM_VERIFY_TIME limits the range between 1 ms and 128 ms.
How can it be 0?

> +
> +	/* enable/disable preemption based on link status */
> +	am65_cpsw_iet_commit_preemptible_tcs(port);
> +
> +	return 0;
> +}

Re: [PATCH v8 net-next 09/11] net: ethernet: ti: am65-cpsw-qos: Add Frame Preemption MAC Merge support

[Roger Quadros]

On 14/12/2023 13:04, Vladimir Oltean wrote:
> On Wed, Dec 13, 2023 at 01:07:19PM +0200, Roger Quadros wrote:
>> - Fix am65_cpsw_iet_get_verify_timeout_ms() to default to timeout for
>>   1G link if link is inactive.
>> +
>> +static int am65_cpsw_set_mm(struct net_device *ndev, struct ethtool_mm_cfg *cfg,
>> +			    struct netlink_ext_ack *extack)
>> +{
>> +	val = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
>> +	if (cfg->verify_enabled) {
>> +		val &= ~AM65_CPSW_PN_IET_MAC_DISABLEVERIFY;
>> +		/* Reset Verify state machine. Verification won't start here.
>> +		 * Verification will be done once link-up.
>> +		 */
>> +		val |= AM65_CPSW_PN_IET_MAC_LINKFAIL;
>> +	} else {
>> +		val |= AM65_CPSW_PN_IET_MAC_DISABLEVERIFY;
>> +		/* Clear LINKFAIL to allow verify/response packets */
>> +		val &= ~AM65_CPSW_PN_IET_MAC_LINKFAIL;
>> +	}
>> +
>> +	val &= ~AM65_CPSW_PN_IET_MAC_MAC_ADDFRAGSIZE_MASK;
>> +	val |= AM65_CPSW_PN_IET_MAC_SET_ADDFRAGSIZE(add_frag_size);
>> +	writel(val, port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
>> +
>> +	/* verify_timeout_count can only be set at valid link */
>> +	if (cfg->verify_time > 0)
>> +		port->qos.iet.verify_time_ms = cfg->verify_time;
>> +	else
>> +		port->qos.iet.verify_time_ms = 10;
> 
> I don't think I understand what the check is for? The netlink policy for
> ETHTOOL_A_MM_VERIFY_TIME limits the range between 1 ms and 128 ms.
> How can it be 0?

I didn't check ETHTOOL_A_MM_VERIFY_TIME before.
Indeed, this check is pointless.

> 
>> +
>> +	/* enable/disable preemption based on link status */
>> +	am65_cpsw_iet_commit_preemptible_tcs(port);
>> +
>> +	return 0;
>> +}

-- 
cheers,
-roger

Re: [PATCH v8 net-next 09/11] net: ethernet: ti: am65-cpsw-qos: Add Frame Preemption MAC Merge support

[Vladimir Oltean]

On Thu, Dec 14, 2023 at 03:44:03PM +0200, Roger Quadros wrote:
> On 14/12/2023 13:04, Vladimir Oltean wrote:
> >> +	/* verify_timeout_count can only be set at valid link */
> >> +	if (cfg->verify_time > 0)
> >> +		port->qos.iet.verify_time_ms = cfg->verify_time;
> >> +	else
> >> +		port->qos.iet.verify_time_ms = 10;
> > 
> > I don't think I understand what the check is for? The netlink policy for
> > ETHTOOL_A_MM_VERIFY_TIME limits the range between 1 ms and 128 ms.
> > How can it be 0?
> 
> I didn't check ETHTOOL_A_MM_VERIFY_TIME before.

Not even when the exact same policy was rejecting the verify-time set by
LLDP in v6, for being larger than the upper limit of 128 ms? :-/

Re: [PATCH v8 net-next 09/11] net: ethernet: ti: am65-cpsw-qos: Add Frame Preemption MAC Merge support

[Roger Quadros]

On 14/12/2023 15:50, Vladimir Oltean wrote:
> On Thu, Dec 14, 2023 at 03:44:03PM +0200, Roger Quadros wrote:
>> On 14/12/2023 13:04, Vladimir Oltean wrote:
>>>> +	/* verify_timeout_count can only be set at valid link */
>>>> +	if (cfg->verify_time > 0)
>>>> +		port->qos.iet.verify_time_ms = cfg->verify_time;
>>>> +	else
>>>> +		port->qos.iet.verify_time_ms = 10;
>>>
>>> I don't think I understand what the check is for? The netlink policy for
>>> ETHTOOL_A_MM_VERIFY_TIME limits the range between 1 ms and 128 ms.
>>> How can it be 0?
>>
>> I didn't check ETHTOOL_A_MM_VERIFY_TIME before.
> 
> Not even when the exact same policy was rejecting the verify-time set by
> LLDP in v6, for being larger than the upper limit of 128 ms? :-/

Sorry. I did check it before but overlooked it here. My memory seems to be failing me. :(

-- 
cheers,
-roger

[PATCH v8 net-next 10/11] net: ethernet: ti: am65-cpsw: add sw tx/rx irq coalescing based on hrtimers

[Roger Quadros]

From: Grygorii Strashko <grygorii.strashko@ti.com>

Add SW IRQ coalescing based on hrtimers for TX and RX data path which
can be enabled by ethtool commands:

- RX coalescing
  ethtool -C eth1 rx-usecs 50

- TX coalescing can be enabled per TX queue

  - by default enables coalesing for TX0
  ethtool -C eth1 tx-usecs 50
  - configure TX0
  ethtool -Q eth0 queue_mask 1 --coalesce tx-usecs 100
  - configure TX1
  ethtool -Q eth0 queue_mask 2 --coalesce tx-usecs 100
  - configure TX0 and TX1
  ethtool -Q eth0 queue_mask 3 --coalesce tx-usecs 100 --coalesce tx-usecs 100

  show configuration for TX0 and TX1:
  ethtool -Q eth0 queue_mask 3 --show-coalesce

Comparing to gro_flush_timeout and napi_defer_hard_irqs, this patch
allows to enable IRQ coalesing for RX path separately.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Roger Quadros <rogerq@kernel.org>
---
 drivers/net/ethernet/ti/am65-cpsw-ethtool.c | 79 +++++++++++++++++++++
 drivers/net/ethernet/ti/am65-cpsw-nuss.c    | 59 ++++++++++++---
 drivers/net/ethernet/ti/am65-cpsw-nuss.h    |  4 ++
 3 files changed, 134 insertions(+), 8 deletions(-)

Changelog:

v8: no change
v7: no change
v6: no change
v5: initial commit

diff --git a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
index 8c3badec219b..d2baffb05d55 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
@@ -904,6 +904,80 @@ static void am65_cpsw_get_mm_stats(struct net_device *ndev,
 	s->MACMergeHoldCount = readl(base + AM65_CPSW_STATN_IET_TX_HOLD);
 }
 
+static int am65_cpsw_get_coalesce(struct net_device *ndev, struct ethtool_coalesce *coal,
+				  struct kernel_ethtool_coalesce *kernel_coal,
+				  struct netlink_ext_ack *extack)
+{
+	struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+	struct am65_cpsw_tx_chn *tx_chn;
+
+	tx_chn = &common->tx_chns[0];
+
+	coal->rx_coalesce_usecs = common->rx_pace_timeout / 1000;
+	coal->tx_coalesce_usecs = tx_chn->tx_pace_timeout / 1000;
+
+	return 0;
+}
+
+static int am65_cpsw_get_per_queue_coalesce(struct net_device *ndev, u32 queue,
+					    struct ethtool_coalesce *coal)
+{
+	struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+	struct am65_cpsw_tx_chn *tx_chn;
+
+	if (queue >= AM65_CPSW_MAX_TX_QUEUES)
+		return -EINVAL;
+
+	tx_chn = &common->tx_chns[queue];
+
+	coal->tx_coalesce_usecs = tx_chn->tx_pace_timeout / 1000;
+
+	return 0;
+}
+
+static int am65_cpsw_set_coalesce(struct net_device *ndev, struct ethtool_coalesce *coal,
+				  struct kernel_ethtool_coalesce *kernel_coal,
+				  struct netlink_ext_ack *extack)
+{
+	struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+	struct am65_cpsw_tx_chn *tx_chn;
+
+	tx_chn = &common->tx_chns[0];
+
+	if (coal->rx_coalesce_usecs && coal->rx_coalesce_usecs < 20)
+		return -EINVAL;
+
+	if (coal->tx_coalesce_usecs && coal->tx_coalesce_usecs < 20)
+		return -EINVAL;
+
+	common->rx_pace_timeout = coal->rx_coalesce_usecs * 1000;
+	tx_chn->tx_pace_timeout = coal->tx_coalesce_usecs * 1000;
+
+	return 0;
+}
+
+static int am65_cpsw_set_per_queue_coalesce(struct net_device *ndev, u32 queue,
+					    struct ethtool_coalesce *coal)
+{
+	struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+	struct am65_cpsw_tx_chn *tx_chn;
+
+	if (queue >= AM65_CPSW_MAX_TX_QUEUES)
+		return -EINVAL;
+
+	tx_chn = &common->tx_chns[queue];
+
+	if (coal->tx_coalesce_usecs && coal->tx_coalesce_usecs < 20) {
+		dev_info(common->dev, "defaulting to min value of 20us for tx-usecs for tx-%u\n",
+			 queue);
+		coal->tx_coalesce_usecs = 20;
+	}
+
+	tx_chn->tx_pace_timeout = coal->tx_coalesce_usecs * 1000;
+
+	return 0;
+}
+
 const struct ethtool_ops am65_cpsw_ethtool_ops_slave = {
 	.begin			= am65_cpsw_ethtool_op_begin,
 	.complete		= am65_cpsw_ethtool_op_complete,
@@ -922,6 +996,11 @@ const struct ethtool_ops am65_cpsw_ethtool_ops_slave = {
 	.get_ts_info		= am65_cpsw_get_ethtool_ts_info,
 	.get_priv_flags		= am65_cpsw_get_ethtool_priv_flags,
 	.set_priv_flags		= am65_cpsw_set_ethtool_priv_flags,
+	.supported_coalesce_params = ETHTOOL_COALESCE_USECS,
+	.get_coalesce           = am65_cpsw_get_coalesce,
+	.set_coalesce           = am65_cpsw_set_coalesce,
+	.get_per_queue_coalesce = am65_cpsw_get_per_queue_coalesce,
+	.set_per_queue_coalesce = am65_cpsw_set_per_queue_coalesce,
 
 	.get_link		= ethtool_op_get_link,
 	.get_link_ksettings	= am65_cpsw_get_link_ksettings,
diff --git a/drivers/net/ethernet/ti/am65-cpsw-nuss.c b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
index 41e0046a52d5..faa0561e988e 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-nuss.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
@@ -596,8 +596,10 @@ static int am65_cpsw_nuss_common_stop(struct am65_cpsw_common *common)
 					msecs_to_jiffies(1000));
 	if (!i)
 		dev_err(common->dev, "tx timeout\n");
-	for (i = 0; i < common->tx_ch_num; i++)
+	for (i = 0; i < common->tx_ch_num; i++) {
 		napi_disable(&common->tx_chns[i].napi_tx);
+		hrtimer_cancel(&common->tx_chns[i].tx_hrtimer);
+	}
 
 	for (i = 0; i < common->tx_ch_num; i++) {
 		k3_udma_glue_reset_tx_chn(common->tx_chns[i].tx_chn,
@@ -616,6 +618,7 @@ static int am65_cpsw_nuss_common_stop(struct am65_cpsw_common *common)
 	}
 
 	napi_disable(&common->napi_rx);
+	hrtimer_cancel(&common->rx_hrtimer);
 
 	for (i = 0; i < AM65_CPSW_MAX_RX_FLOWS; i++)
 		k3_udma_glue_reset_rx_chn(common->rx_chns.rx_chn, i,
@@ -885,6 +888,15 @@ static int am65_cpsw_nuss_rx_packets(struct am65_cpsw_common *common,
 	return ret;
 }
 
+static enum hrtimer_restart am65_cpsw_nuss_rx_timer_callback(struct hrtimer *timer)
+{
+	struct am65_cpsw_common *common =
+			container_of(timer, struct am65_cpsw_common, rx_hrtimer);
+
+	enable_irq(common->rx_chns.irq);
+	return HRTIMER_NORESTART;
+}
+
 static int am65_cpsw_nuss_rx_poll(struct napi_struct *napi_rx, int budget)
 {
 	struct am65_cpsw_common *common = am65_cpsw_napi_to_common(napi_rx);
@@ -912,7 +924,13 @@ static int am65_cpsw_nuss_rx_poll(struct napi_struct *napi_rx, int budget)
 	if (num_rx < budget && napi_complete_done(napi_rx, num_rx)) {
 		if (common->rx_irq_disabled) {
 			common->rx_irq_disabled = false;
-			enable_irq(common->rx_chns.irq);
+			if (unlikely(common->rx_pace_timeout)) {
+				hrtimer_start(&common->rx_hrtimer,
+					      ns_to_ktime(common->rx_pace_timeout),
+					      HRTIMER_MODE_REL_PINNED);
+			} else {
+				enable_irq(common->rx_chns.irq);
+			}
 		}
 	}
 
@@ -968,7 +986,7 @@ static void am65_cpsw_nuss_tx_wake(struct am65_cpsw_tx_chn *tx_chn, struct net_d
 }
 
 static int am65_cpsw_nuss_tx_compl_packets(struct am65_cpsw_common *common,
-					   int chn, unsigned int budget)
+					   int chn, unsigned int budget, bool *tdown)
 {
 	struct device *dev = common->dev;
 	struct am65_cpsw_tx_chn *tx_chn;
@@ -991,6 +1009,7 @@ static int am65_cpsw_nuss_tx_compl_packets(struct am65_cpsw_common *common,
 		if (cppi5_desc_is_tdcm(desc_dma)) {
 			if (atomic_dec_and_test(&common->tdown_cnt))
 				complete(&common->tdown_complete);
+			*tdown = true;
 			break;
 		}
 
@@ -1013,7 +1032,7 @@ static int am65_cpsw_nuss_tx_compl_packets(struct am65_cpsw_common *common,
 }
 
 static int am65_cpsw_nuss_tx_compl_packets_2g(struct am65_cpsw_common *common,
-					      int chn, unsigned int budget)
+					      int chn, unsigned int budget, bool *tdown)
 {
 	struct device *dev = common->dev;
 	struct am65_cpsw_tx_chn *tx_chn;
@@ -1034,6 +1053,7 @@ static int am65_cpsw_nuss_tx_compl_packets_2g(struct am65_cpsw_common *common,
 		if (cppi5_desc_is_tdcm(desc_dma)) {
 			if (atomic_dec_and_test(&common->tdown_cnt))
 				complete(&common->tdown_complete);
+			*tdown = true;
 			break;
 		}
 
@@ -1059,21 +1079,40 @@ static int am65_cpsw_nuss_tx_compl_packets_2g(struct am65_cpsw_common *common,
 	return num_tx;
 }
 
+static enum hrtimer_restart am65_cpsw_nuss_tx_timer_callback(struct hrtimer *timer)
+{
+	struct am65_cpsw_tx_chn *tx_chns =
+			container_of(timer, struct am65_cpsw_tx_chn, tx_hrtimer);
+
+	enable_irq(tx_chns->irq);
+	return HRTIMER_NORESTART;
+}
+
 static int am65_cpsw_nuss_tx_poll(struct napi_struct *napi_tx, int budget)
 {
 	struct am65_cpsw_tx_chn *tx_chn = am65_cpsw_napi_to_tx_chn(napi_tx);
+	bool tdown = false;
 	int num_tx;
 
 	if (AM65_CPSW_IS_CPSW2G(tx_chn->common))
-		num_tx = am65_cpsw_nuss_tx_compl_packets_2g(tx_chn->common, tx_chn->id, budget);
+		num_tx = am65_cpsw_nuss_tx_compl_packets_2g(tx_chn->common, tx_chn->id,
+							    budget, &tdown);
 	else
-		num_tx = am65_cpsw_nuss_tx_compl_packets(tx_chn->common, tx_chn->id, budget);
+		num_tx = am65_cpsw_nuss_tx_compl_packets(tx_chn->common,
+							 tx_chn->id, budget, &tdown);
 
 	if (num_tx >= budget)
 		return budget;
 
-	if (napi_complete_done(napi_tx, num_tx))
-		enable_irq(tx_chn->irq);
+	if (napi_complete_done(napi_tx, num_tx)) {
+		if (unlikely(tx_chn->tx_pace_timeout && !tdown)) {
+			hrtimer_start(&tx_chn->tx_hrtimer,
+				      ns_to_ktime(tx_chn->tx_pace_timeout),
+				      HRTIMER_MODE_REL_PINNED);
+		} else {
+			enable_irq(tx_chn->irq);
+		}
+	}
 
 	return 0;
 }
@@ -1705,6 +1744,8 @@ static int am65_cpsw_nuss_ndev_add_tx_napi(struct am65_cpsw_common *common)
 
 		netif_napi_add_tx(common->dma_ndev, &tx_chn->napi_tx,
 				  am65_cpsw_nuss_tx_poll);
+		hrtimer_init(&tx_chn->tx_hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_PINNED);
+		tx_chn->tx_hrtimer.function = &am65_cpsw_nuss_tx_timer_callback;
 
 		ret = devm_request_irq(dev, tx_chn->irq,
 				       am65_cpsw_nuss_tx_irq,
@@ -1930,6 +1971,8 @@ static int am65_cpsw_nuss_init_rx_chns(struct am65_cpsw_common *common)
 
 	netif_napi_add(common->dma_ndev, &common->napi_rx,
 		       am65_cpsw_nuss_rx_poll);
+	hrtimer_init(&common->rx_hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_PINNED);
+	common->rx_hrtimer.function = &am65_cpsw_nuss_rx_timer_callback;
 
 	ret = devm_request_irq(dev, rx_chn->irq,
 			       am65_cpsw_nuss_rx_irq,
diff --git a/drivers/net/ethernet/ti/am65-cpsw-nuss.h b/drivers/net/ethernet/ti/am65-cpsw-nuss.h
index 1e4a045057fc..7da0492dc091 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-nuss.h
+++ b/drivers/net/ethernet/ti/am65-cpsw-nuss.h
@@ -75,6 +75,8 @@ struct am65_cpsw_tx_chn {
 	struct k3_cppi_desc_pool *desc_pool;
 	struct k3_udma_glue_tx_channel *tx_chn;
 	spinlock_t lock; /* protect TX rings in multi-port mode */
+	struct hrtimer tx_hrtimer;
+	unsigned long tx_pace_timeout;
 	int irq;
 	u32 id;
 	u32 descs_num;
@@ -138,6 +140,8 @@ struct am65_cpsw_common {
 	struct napi_struct	napi_rx;
 
 	bool			rx_irq_disabled;
+	struct hrtimer		rx_hrtimer;
+	unsigned long		rx_pace_timeout;
 
 	u32			nuss_ver;
 	u32			cpsw_ver;
-- 
2.34.1

[PATCH v8 net-next 11/11] net: ethernet: ti: am65-cpsw: Fix get_eth_mac_stats

[Roger Quadros]

We do not support individual stats for PMAC and EMAC so
report only aggregate stats.

Fixes: 67372d7a85fc ("net: ethernet: am65-cpsw: Add standard Ethernet MAC stats to ethtool")
Signed-off-by: Roger Quadros <rogerq@kernel.org>
---
 drivers/net/ethernet/ti/am65-cpsw-ethtool.c | 3 +++
 1 file changed, 3 insertions(+)

Changelog:

v8: initial commit

diff --git a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
index d2baffb05d55..35e318458b0c 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
@@ -671,6 +671,9 @@ static void am65_cpsw_get_eth_mac_stats(struct net_device *ndev,
 
 	stats = port->stat_base;
 
+	if (s->src != ETHTOOL_MAC_STATS_SRC_AGGREGATE)
+		return;
+
 	s->FramesTransmittedOK = readl_relaxed(&stats->tx_good_frames);
 	s->SingleCollisionFrames = readl_relaxed(&stats->tx_single_coll_frames);
 	s->MultipleCollisionFrames = readl_relaxed(&stats->tx_mult_coll_frames);
-- 
2.34.1

Re: [PATCH v8 net-next 11/11] net: ethernet: ti: am65-cpsw: Fix get_eth_mac_stats

[Vladimir Oltean]

On Wed, Dec 13, 2023 at 01:07:21PM +0200, Roger Quadros wrote:
> We do not support individual stats for PMAC and EMAC so
> report only aggregate stats.
> 
> Fixes: 67372d7a85fc ("net: ethernet: am65-cpsw: Add standard Ethernet MAC stats to ethtool")
> Signed-off-by: Roger Quadros <rogerq@kernel.org>
> ---
>  drivers/net/ethernet/ti/am65-cpsw-ethtool.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> Changelog:
> 
> v8: initial commit
> 
> diff --git a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
> index d2baffb05d55..35e318458b0c 100644
> --- a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
> +++ b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
> @@ -671,6 +671,9 @@ static void am65_cpsw_get_eth_mac_stats(struct net_device *ndev,
>  
>  	stats = port->stat_base;
>  
> +	if (s->src != ETHTOOL_MAC_STATS_SRC_AGGREGATE)
> +		return;
> +
>  	s->FramesTransmittedOK = readl_relaxed(&stats->tx_good_frames);
>  	s->SingleCollisionFrames = readl_relaxed(&stats->tx_single_coll_frames);
>  	s->MultipleCollisionFrames = readl_relaxed(&stats->tx_mult_coll_frames);
> -- 
> 2.34.1
>

Fixes are only fixes if they address a visible issue. And the blamed
commit is the one that made the issue visible - the same one that
"git bisect" would lead to - not necessarily the commit that introduced
the code being changed.

If you look at net/ethtool/stats.c, it will only accept ETHTOOL_MAC_STATS_SRC_AGGREGATE
for drivers that don't support the MAC Merge layer.

	if ((src == ETHTOOL_MAC_STATS_SRC_EMAC ||
	     src == ETHTOOL_MAC_STATS_SRC_PMAC) &&
	    !__ethtool_dev_mm_supported(dev)) {
		NL_SET_ERR_MSG_MOD(info->extack,
				   "Device does not support MAC merge layer");
		ethnl_ops_complete(dev);
		return -EOPNOTSUPP;
	}

So, there was nothing broken in commit 67372d7a85fc ("net: ethernet:
am65-cpsw: Add standard Ethernet MAC stats to ethtool").

The first broken commit is when you add support for get_mm(), such that
__ethtool_dev_mm_supported() returns true.

And because you don't add bugs in the code just to fix them later in the
series, you need to order the patches such that all the dependencies for
get_mm() are in place before the get_mm() support is added.

Translated to your case, this patch must not be 11/11, and it must have
the Fixes: tag dropped, and it must explain in the commit message that
it is preparatory work.

--
pw-bot: changes-requested

Re: [PATCH v8 net-next 11/11] net: ethernet: ti: am65-cpsw: Fix get_eth_mac_stats

[Roger Quadros]

On 13/12/2023 15:54, Vladimir Oltean wrote:
> On Wed, Dec 13, 2023 at 01:07:21PM +0200, Roger Quadros wrote:
>> We do not support individual stats for PMAC and EMAC so
>> report only aggregate stats.
>>
>> Fixes: 67372d7a85fc ("net: ethernet: am65-cpsw: Add standard Ethernet MAC stats to ethtool")
>> Signed-off-by: Roger Quadros <rogerq@kernel.org>
>> ---
>>  drivers/net/ethernet/ti/am65-cpsw-ethtool.c | 3 +++
>>  1 file changed, 3 insertions(+)
>>
>> Changelog:
>>
>> v8: initial commit
>>
>> diff --git a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
>> index d2baffb05d55..35e318458b0c 100644
>> --- a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
>> +++ b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
>> @@ -671,6 +671,9 @@ static void am65_cpsw_get_eth_mac_stats(struct net_device *ndev,
>>  
>>  	stats = port->stat_base;
>>  
>> +	if (s->src != ETHTOOL_MAC_STATS_SRC_AGGREGATE)
>> +		return;
>> +
>>  	s->FramesTransmittedOK = readl_relaxed(&stats->tx_good_frames);
>>  	s->SingleCollisionFrames = readl_relaxed(&stats->tx_single_coll_frames);
>>  	s->MultipleCollisionFrames = readl_relaxed(&stats->tx_mult_coll_frames);
>> -- 
>> 2.34.1
>>
> 
> Fixes are only fixes if they address a visible issue. And the blamed
> commit is the one that made the issue visible - the same one that
> "git bisect" would lead to - not necessarily the commit that introduced
> the code being changed.
> 
> If you look at net/ethtool/stats.c, it will only accept ETHTOOL_MAC_STATS_SRC_AGGREGATE
> for drivers that don't support the MAC Merge layer.
> 
> 	if ((src == ETHTOOL_MAC_STATS_SRC_EMAC ||
> 	     src == ETHTOOL_MAC_STATS_SRC_PMAC) &&
> 	    !__ethtool_dev_mm_supported(dev)) {
> 		NL_SET_ERR_MSG_MOD(info->extack,
> 				   "Device does not support MAC merge layer");
> 		ethnl_ops_complete(dev);
> 		return -EOPNOTSUPP;
> 	}

Got it.

> 
> So, there was nothing broken in commit 67372d7a85fc ("net: ethernet:
> am65-cpsw: Add standard Ethernet MAC stats to ethtool").
> 
> The first broken commit is when you add support for get_mm(), such that
> __ethtool_dev_mm_supported() returns true.
> 
> And because you don't add bugs in the code just to fix them later in the
> series, you need to order the patches such that all the dependencies for
> get_mm() are in place before the get_mm() support is added.
> 
> Translated to your case, this patch must not be 11/11, and it must have
> the Fixes: tag dropped, and it must explain in the commit message that
> it is preparatory work.

OK. I take care of this in next spin. Thanks.
> 
> --
> pw-bot: changes-requested

-- 
cheers,
-roger

Re: [PATCH v8 net-next 11/11] net: ethernet: ti: am65-cpsw: Fix get_eth_mac_stats

[Roger Quadros]

On 14/12/2023 12:04, Roger Quadros wrote:
> 
> 
> On 13/12/2023 15:54, Vladimir Oltean wrote:
>> On Wed, Dec 13, 2023 at 01:07:21PM +0200, Roger Quadros wrote:
>>> We do not support individual stats for PMAC and EMAC so
>>> report only aggregate stats.
>>>
>>> Fixes: 67372d7a85fc ("net: ethernet: am65-cpsw: Add standard Ethernet MAC stats to ethtool")
>>> Signed-off-by: Roger Quadros <rogerq@kernel.org>
>>> ---
>>>  drivers/net/ethernet/ti/am65-cpsw-ethtool.c | 3 +++
>>>  1 file changed, 3 insertions(+)
>>>
>>> Changelog:
>>>
>>> v8: initial commit
>>>
>>> diff --git a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
>>> index d2baffb05d55..35e318458b0c 100644
>>> --- a/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
>>> +++ b/drivers/net/ethernet/ti/am65-cpsw-ethtool.c
>>> @@ -671,6 +671,9 @@ static void am65_cpsw_get_eth_mac_stats(struct net_device *ndev,
>>>  
>>>  	stats = port->stat_base;
>>>  
>>> +	if (s->src != ETHTOOL_MAC_STATS_SRC_AGGREGATE)
>>> +		return;
>>> +
>>>  	s->FramesTransmittedOK = readl_relaxed(&stats->tx_good_frames);
>>>  	s->SingleCollisionFrames = readl_relaxed(&stats->tx_single_coll_frames);
>>>  	s->MultipleCollisionFrames = readl_relaxed(&stats->tx_mult_coll_frames);
>>> -- 
>>> 2.34.1
>>>
>>
>> Fixes are only fixes if they address a visible issue. And the blamed
>> commit is the one that made the issue visible - the same one that
>> "git bisect" would lead to - not necessarily the commit that introduced
>> the code being changed.
>>
>> If you look at net/ethtool/stats.c, it will only accept ETHTOOL_MAC_STATS_SRC_AGGREGATE
>> for drivers that don't support the MAC Merge layer.
>>
>> 	if ((src == ETHTOOL_MAC_STATS_SRC_EMAC ||
>> 	     src == ETHTOOL_MAC_STATS_SRC_PMAC) &&
>> 	    !__ethtool_dev_mm_supported(dev)) {
>> 		NL_SET_ERR_MSG_MOD(info->extack,
>> 				   "Device does not support MAC merge layer");
>> 		ethnl_ops_complete(dev);
>> 		return -EOPNOTSUPP;
>> 	}
> 
> Got it.
> 
>>
>> So, there was nothing broken in commit 67372d7a85fc ("net: ethernet:
>> am65-cpsw: Add standard Ethernet MAC stats to ethtool").
>>
>> The first broken commit is when you add support for get_mm(), such that
>> __ethtool_dev_mm_supported() returns true.
>>
>> And because you don't add bugs in the code just to fix them later in the
>> series, you need to order the patches such that all the dependencies for
>> get_mm() are in place before the get_mm() support is added.
>>
>> Translated to your case, this patch must not be 11/11, and it must have
>> the Fixes: tag dropped, and it must explain in the commit message that
>> it is preparatory work.
> 
> OK. I take care of this in next spin. Thanks.

Just a heads up. I have decided to merge this patch with mac-merge patch.
One less patch to deal with ;).

-- 
cheers,
-roger