Skip to contentSkip to conversation

NC-SI PHY Support

PATCH Superseded Samuel Mendoza-Jonas 2 participants 12 messages Jun 6, 2019 – Jun 12, 2019

Subsystems (3)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover NC-SI PHY Support

This series introduces support for the NC-SI protocol to u-boot, functionality which so far has only been available in vendor trees.

NC-SI (Network Controller Sideband Interface) is a protocol to manage a sideband connection to a proper network interface, for example a BMC (Baseboard Management Controller) sharing the NIC of the host system. Probing and configuration occurs by communicating with the "remote" NIC via NC-SI control frames (Ethernet header 0x88f8).

Since the RFC[0] the changes have been split out into a few different patches to make the overall changes more obvious. Additionally a few small improvements have been added including adding a check to see if NC-SI is already configured, and marking the PHY bus "reset" as invalid since it does not define a bus.

An example of NC-SI setup before a normal network command:

=> dhcp
net_loop: configuring NCSI first
ethernet at 1e660000: link up, 100 Mbps full-duplex mac:0c:c4:7a:d5:48:43
NCSI: probing channels
NCSI: configuring channel 0
NCSI: configuration done!
BOOTP broadcast 1
BOOTP broadcast 2
DHCP client bound to address 10.61.161.188 (255 ms)

[0]: https://patchwork.ozlabs.org/patch/1107087/

Samuel Mendoza-Jonas (4):
phy: Add support for the NC-SI protocol
phy: Include NC-SI in phy setup
net/ftgmac100: Add NC-SI mode support
net: NC-SI setup and handling

 cmd/Kconfig              |   6 +
 cmd/net.c                |  18 +
 drivers/net/ftgmac100.c  |  39 +-
 drivers/net/phy/Kconfig  |   4 +
 drivers/net/phy/Makefile |   1 +
 drivers/net/phy/ncsi.c   | 892 +++++++++++++++++++++++++++++++++++++++
 drivers/net/phy/phy.c    |   8 +
 include/net.h            |   3 +-
 include/net/ncsi-pkt.h   | 415 ++++++++++++++++++
 include/net/ncsi.h       |  14 +
 include/phy.h            |  12 +
 include/phy_interface.h  |   2 +
 net/net.c                |  27 +-
 13 files changed, 1429 insertions(+), 12 deletions(-)
 create mode 100644 drivers/net/phy/ncsi.c
 create mode 100644 include/net/ncsi-pkt.h
 create mode 100644 include/net/ncsi.h
Signature · 1 line
2.21.0
#1 lore raw

PATCH 1/4 phy: Add support for the NC-SI protocol

This introduces support for the NC-SI protocol, modelled as a phy driver for other ethernet drivers to consume.

NC-SI (Network Controller Sideband Interface) is a protocol to manage a sideband connection to a proper network interface, for example a BMC (Baseboard Management Controller) sharing the NIC of the host system. Probing and configuration occurs by communicating with the "remote" NIC via NC-SI control frames (Ethernet header 0x88f8).

This implementation is roughly based on the upstream Linux implementation[0], with a reduced feature set and an emphasis on getting a link up as fast as possible rather than probing the full possible topology of the bus.
The current phy model relies on the network being "up", sending NC-SI command frames via net_send_packet() and receiving them from the net_loop() loop (added in a following patch).

The ncsi-pkt.h header[1] is copied from the Linux kernel for consistent field definitions.

[0]: https://github.com/torvalds/linux/tree/master/net/ncsi
[1]: https://github.com/torvalds/linux/blob/master/net/ncsi/ncsi-pkt.h

---
 drivers/net/phy/Kconfig  |   4 +
 drivers/net/phy/Makefile |   1 +
 drivers/net/phy/ncsi.c   | 892 +++++++++++++++++++++++++++++++++++++++
 include/net.h            |   1 +
 include/net/ncsi-pkt.h   | 415 ++++++++++++++++++
 include/net/ncsi.h       |  14 +
 include/phy.h            |   1 +
 7 files changed, 1328 insertions(+)
 create mode 100644 drivers/net/phy/ncsi.c
 create mode 100644 include/net/ncsi-pkt.h
 create mode 100644 include/net/ncsi.h
drivers/net/phy/Kconfig +4 -0
index 3dc0822d9c..f6dde8af1c 100644
@@ -227,4 +227,8 @@ config PHY_FIXED
 	  on, the link is always up with fixed speed and fixed duplex-setting.
 	  More information: doc/device-tree-bindings/net/fixed-link.txt
 
+config PHY_NCSI
+	bool "NC-SI based PHY"
+	depends on DM_ETH
+
 endif #PHYLIB
drivers/net/phy/Makefile +1 -0
index 555da83630..32a8409a85 100644
@@ -30,3 +30,4 @@ obj-$(CONFIG_PHY_XILINX) += xilinx_phy.o
 obj-$(CONFIG_PHY_VITESSE) += vitesse.o
 obj-$(CONFIG_PHY_MSCC) += mscc.o
 obj-$(CONFIG_PHY_FIXED) += fixed.o
+obj-$(CONFIG_PHY_NCSI) += ncsi.o
drivers/net/phy/ncsi.c new file +892 -0

892 lines not on this page · raw message

include/net.h +1 -0
index 51c099dae2..4f7ba211b6 100644
@@ -350,6 +350,7 @@ struct vlan_ethernet_hdr {
 #define PROT_VLAN	0x8100		/* IEEE 802.1q protocol		*/
 #define PROT_IPV6	0x86dd		/* IPv6 over bluebook		*/
 #define PROT_PPP_SES	0x8864		/* PPPoE session messages	*/
+#define PROT_NCSI	0x88f8		/* NC-SI control packets        */
 
 #define IPPROTO_ICMP	 1	/* Internet Control Message Protocol	*/
 #define IPPROTO_UDP	17	/* User Datagram Protocol		*/
include/net/ncsi-pkt.h new file +415 -0

415 lines not on this page · raw message

include/net/ncsi.h new file +14 -0
new file mode 100644
index 0000000000..9637beab2a
@@ -0,0 +1,14 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * NC-SI PHY
+ *
+ * Copyright (C) 2019, IBM Corporation.
+ */
+
+#include <common.h>
+#include <phy.h>
+
+bool ncsi_active(void);
+void ncsi_receive(struct ethernet_hdr *et, struct ip_udp_hdr *ip,
+		  unsigned int len);
+void ncsi_probe_packages(void);
include/phy.h +1 -0
index b86fdfb2ce..f0c9df46c6 100644
@@ -17,6 +17,7 @@
 #include <phy_interface.h>
 
 #define PHY_FIXED_ID		0xa5a55a5a
+#define PHY_NCSI_ID		0xbeefcafe
 
 #define PHY_MAX_ADDR 32
 
Signature · 1 line
2.21.0
#2 lore raw

Re: [PATCH 1/4] phy: Add support for the NC-SI protocol

Quoted 18 lines · Samuel Mendoza-Jonas

This introduces support for the NC-SI protocol, modelled as a phy driver for other ethernet drivers to consume.

NC-SI (Network Controller Sideband Interface) is a protocol to manage a sideband connection to a proper network interface, for example a BMC (Baseboard Management Controller) sharing the NIC of the host system. Probing and configuration occurs by communicating with the "remote" NIC via NC-SI control frames (Ethernet header 0x88f8).

This implementation is roughly based on the upstream Linux implementation[0], with a reduced feature set and an emphasis on getting a link up as fast as possible rather than probing the full possible topology of the bus.
The current phy model relies on the network being "up", sending NC-SI command frames via net_send_packet() and receiving them from the net_loop() loop (added in a following patch).

Quoted 7 lines

The ncsi-pkt.h header[1] is copied from the Linux kernel for consistent field definitions.

[0]: https://github.com/torvalds/linux/tree/master/net/ncsi
[1]: https://github.com/torvalds/linux/blob/master/net/ncsi/ncsi-pkt.h

Looks good. Some comments below.

drivers/net/phy/ncsi.c line 208

Quoted 13 lines
drivers/net/phy/ncsi.c new file +12 -0
+static int ncsi_validate_rsp(struct ncsi_rsp_pkt *pkt, int payload)
+{
+       struct ncsi_rsp_pkt_hdr *hdr = &pkt->rsp;
+       __be32 pchecksum;
+       u32 checksum;
 
+       if (ntohs(hdr->common.length) != payload) {
+               printf("NCSI: 0x%02x response has incorrect length %d\n",
+                      hdr->common.type, hdr->common.length);
+               return -1;
+       }
+
+       pchecksum = get_unaligned_be32((void *)(hdr + 1) + payload - 4);

Wheee. So the checksum is the last 4-bytes of the payload. I assume it's always longer than 4?

A clarifying comment might help, or try to write it in a different way:

endp = (void *)hdr + sizeof(hdr) + payload; pchecksum = get_unaligned_be32(endp - sizeof(checksum));

or

    checksum_offset = sizeof(hdr) + payload - sizeof(checksum);
    pchecksum = get_unaligned_be32(payload + checksum_offset);

drivers/net/phy/ncsi.c line 211

Quoted 3 lines
drivers/net/phy/ncsi.c new file +3 -0
+       if (pchecksum != 0) {
+               checksum = ncsi_calculate_checksum((unsigned char *)hdr,
+                                                  sizeof(*hdr) + payload - 4);

And then this can be:

    checksum = ((unsigned char *)hdr, checksum_offset);
Quoted 10 lines
drivers/net/phy/ncsi.c new file +9 -0
+               if (pchecksum != checksum) {
+                       printf("NCSI: 0x%02x response has invalid checksum\n",
+                              hdr->common.type);
+                       return -1;
+               }
+       }
 
+static void ncsi_send_sma(unsigned int np, unsigned int nc)
+{
+       struct ncsi_cmd_sma_pkt cmd;

drivers/net/phy/ncsi.c line 775

Quoted 10 lines
drivers/net/phy/ncsi.c new file +10 -0
+       unsigned char *addr;
+
+       addr = eth_get_ethaddr();
+       if (!addr) {
+               printf("NCSI: no MAC address configured\n");
+               return;
+       }
+
+       memset(&cmd, 0, sizeof(cmd));
+       memcpy(cmd.mac, addr, 6);

Are there endianness issues with addr here?

drivers/net/phy/ncsi.c line 837

Quoted 13 lines
drivers/net/phy/ncsi.c new file +12 -0
+       cmd.index = 1;
+       cmd.at_e = 1;
+
+       ncsi_send_command(np, nc, NCSI_PKT_CMD_SMA,
+                         ((unsigned char *)&cmd)
+                         + sizeof(struct ncsi_cmd_pkt_hdr),
+                         cmd_payload(NCSI_PKT_CMD_SMA), true);
+}
 
+
+int ncsi_probe(struct phy_device *phydev)
+{
+       // TODO Associate per device

Is this required before we can support multiple NICs?

Quoted 18 lines
drivers/net/phy/ncsi.c new file +18 -0
+       if (!ncsi_priv) {
+               ncsi_priv = malloc(sizeof(struct ncsi));
+               if (!ncsi_priv)
+                       return -ENOMEM;
+               memset(ncsi_priv, 0, sizeof(struct ncsi));
+       }
+
+       phydev->priv = ncsi_priv;
+
+       return 0;
+}
+
+int ncsi_startup(struct phy_device *phydev)
+{
+       /* Set phydev parameters */
+       phydev->speed = SPEED_100;
+       phydev->duplex = DUPLEX_FULL;
+       /* Normal phy reset is N/A */

drivers/net/phy/ncsi.c line 865

Quoted 10 lines
drivers/net/phy/ncsi.c new file +10 -0
+       phydev->flags |= PHY_FLAG_BROKEN_RESET;
+
+       /* Set initial probe state */
+       ncsi_priv->state = NCSI_PROBE_PACKAGE_SP;
+
+       /* No active package/channel yet */
+       ncsi_priv->current_package = NCSI_PACKAGE_MAX;
+       ncsi_priv->current_channel = NCSI_CHANNEL_MAX;
+
+       /* Pretend link works so ftgmac100 sets final bits up */

s/ftgmac100/mac driver/ ?

Quoted 32 lines
drivers/net/phy/ncsi.c new file +27 -0
+       phydev->link = true;
+
+       return 0;
+}
+
+int ncsi_shutdown(struct phy_device *phydev)
+{
+       printf("NCSI: Disabling package %d\n", ncsi_priv->current_package);
+       ncsi_send_dp(ncsi_priv->current_package);
+       return 0;
+}
+
+static struct phy_driver ncsi_driver = {
+       .uid            = PHY_NCSI_ID,
+       .mask           = 0xffffffff,
+       .name           = "NC-SI",
+       .features       = PHY_100BT_FEATURES | PHY_DEFAULT_FEATURES | SUPPORTED_100baseT_Full | SUPPORTED_MII,
+       .probe          = ncsi_probe,
+       .startup        = ncsi_startup,
+       .shutdown       = ncsi_shutdown,
+};
+
+int phy_ncsi_init(void)
+{
+       phy_register(&ncsi_driver);
+       return 0;
+}

include/net/ncsi-pkt.h line 8

Quoted 8 lines
include/net/ncsi-pkt.h new file +8 -0
@@ -0,0 +1,415 @@
+/*
+ * Copyright Gavin Shan, IBM Corporation 2016.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */

If you grab the version from 5.2-rc3 it has been SPDXified.

include/phy.h line 20

Quoted 7 lines
include/phy.h +1 -0
@@ -17,6 +17,7 @@
 #include <phy_interface.h>
 
 #define PHY_FIXED_ID           0xa5a55a5a
+#define PHY_NCSI_ID            0xbeefcafe

hmmm...

Quoted 6 lines
include/phy.h +0 -1
 #define PHY_MAX_ADDR 32
 
--

2.21.0

#3 lore raw

Re: [PATCH 1/4] phy: Add support for the NC-SI protocol

Quoted 44 lines · Joel Stanley
Quoted 17 lines · Samuel Mendoza-Jonas

This introduces support for the NC-SI protocol, modelled as a phy driver for other ethernet drivers to consume.

NC-SI (Network Controller Sideband Interface) is a protocol to manage a sideband connection to a proper network interface, for example a BMC (Baseboard Management Controller) sharing the NIC of the host system. Probing and configuration occurs by communicating with the "remote" NIC via NC-SI control frames (Ethernet header 0x88f8).

This implementation is roughly based on the upstream Linux implementation[0], with a reduced feature set and an emphasis on getting a link up as fast as possible rather than probing the full possible topology of the bus.
The current phy model relies on the network being "up", sending NC-SI command frames via net_send_packet() and receiving them from the net_loop() loop (added in a following patch).

Quoted 7 lines

The ncsi-pkt.h header[1] is copied from the Linux kernel for consistent field definitions.

[0]: https://github.com/torvalds/linux/tree/master/net/ncsi
[1]: https://github.com/torvalds/linux/blob/master/net/ncsi/ncsi-pkt.h

Looks good. Some comments below.

drivers/net/phy/ncsi.c line 208

Quoted 12 lines
drivers/net/phy/ncsi.c new file +12 -0
+static int ncsi_validate_rsp(struct ncsi_rsp_pkt *pkt, int payload)
+{
+       struct ncsi_rsp_pkt_hdr *hdr = &pkt->rsp;
+       __be32 pchecksum;
+       u32 checksum;
+       if (ntohs(hdr->common.length) != payload) {
+               printf("NCSI: 0x%02x response has incorrect length %d\n",
+                      hdr->common.type, hdr->common.length);
+               return -1;
+       }
+
+       pchecksum = get_unaligned_be32((void *)(hdr + 1) + payload - 4);

Wheee. So the checksum is the last 4-bytes of the payload. I assume it's always longer than 4?

Quoted 1 line

A clarifying comment might help, or try to write it in a different way:

"Wheee" indeed. I kept this roughly as it is in the kernel so I could more easily verify it was doing the right thing; now that it's more tested I'll do something like below so it's less traumatic.

Quoted 30 lines

endp = (void *)hdr + sizeof(hdr) + payload; pchecksum = get_unaligned_be32(endp - sizeof(checksum));

or

    checksum_offset = sizeof(hdr) + payload - sizeof(checksum);
    pchecksum = get_unaligned_be32(payload + checksum_offset);

drivers/net/phy/ncsi.c line 211

Quoted 3 lines
drivers/net/phy/ncsi.c new file +3 -0
+       if (pchecksum != 0) {
+               checksum = ncsi_calculate_checksum((unsigned char *)hdr,
+                                                  sizeof(*hdr) + payload - 4);

And then this can be:

    checksum = ((unsigned char *)hdr, checksum_offset);
Quoted 13 lines
drivers/net/phy/ncsi.c new file +13 -0
+               if (pchecksum != checksum) {
+                       printf("NCSI: 0x%02x response has invalid checksum\n",
+                              hdr->common.type);
+                       return -1;
+               }
+       }
+static void ncsi_send_sma(unsigned int np, unsigned int nc)
+{
+       struct ncsi_cmd_sma_pkt cmd;
+       unsigned char *addr;
+
+       addr = eth_get_ethaddr();
+       if (!addr) {
Quoted 8 lines

drivers/net/phy/ncsi.c line 775

Quoted 6 lines
drivers/net/phy/ncsi.c new file +6 -0
+               printf("NCSI: no MAC address configured\n");
+               return;
+       }
+
+       memset(&cmd, 0, sizeof(cmd));
+       memcpy(cmd.mac, addr, 6);

Are there endianness issues with addr here?

Aha, will fixup.

Quoted 15 lines

drivers/net/phy/ncsi.c line 837

Quoted 13 lines
drivers/net/phy/ncsi.c new file +12 -0
+       cmd.index = 1;
+       cmd.at_e = 1;
+
+       ncsi_send_command(np, nc, NCSI_PKT_CMD_SMA,
+                         ((unsigned char *)&cmd)
+                         + sizeof(struct ncsi_cmd_pkt_hdr),
+                         cmd_payload(NCSI_PKT_CMD_SMA), true);
+}
+
+int ncsi_probe(struct phy_device *phydev)
+{
+       // TODO Associate per device

Is this required before we can support multiple NICs?

Yes, I'll chase this up.

Quoted 23 lines
Quoted 23 lines
drivers/net/phy/ncsi.c new file +22 -0
+       if (!ncsi_priv) {
+               ncsi_priv = malloc(sizeof(struct ncsi));
+               if (!ncsi_priv)
+                       return -ENOMEM;
+               memset(ncsi_priv, 0, sizeof(struct ncsi));
+       }
+
+       phydev->priv = ncsi_priv;
+
+       return 0;
+}
+
+int ncsi_startup(struct phy_device *phydev)
+{
+       /* Set phydev parameters */
+       phydev->speed = SPEED_100;
+       phydev->duplex = DUPLEX_FULL;
+       /* Normal phy reset is N/A */
+       phydev->flags |= PHY_FLAG_BROKEN_RESET;
+
+       /* Set initial probe state */
+       ncsi_priv->state = NCSI_PROBE_PACKAGE_SP;
Quoted 8 lines

drivers/net/phy/ncsi.c line 865

Quoted 6 lines
drivers/net/phy/ncsi.c new file +6 -0
+
+       /* No active package/channel yet */
+       ncsi_priv->current_package = NCSI_PACKAGE_MAX;
+       ncsi_priv->current_channel = NCSI_CHANNEL_MAX;
+
+       /* Pretend link works so ftgmac100 sets final bits up */

s/ftgmac100/mac driver/ ?

Ack

Quoted 33 lines
Quoted 33 lines
drivers/net/phy/ncsi.c new file +27 -0
+       phydev->link = true;
+
+       return 0;
+}
+
+int ncsi_shutdown(struct phy_device *phydev)
+{
+       printf("NCSI: Disabling package %d\n", ncsi_priv->current_package);
+       ncsi_send_dp(ncsi_priv->current_package);
+       return 0;
+}
+
+static struct phy_driver ncsi_driver = {
+       .uid            = PHY_NCSI_ID,
+       .mask           = 0xffffffff,
+       .name           = "NC-SI",
+       .features       = PHY_100BT_FEATURES | PHY_DEFAULT_FEATURES | SUPPORTED_100baseT_Full | SUPPORTED_MII,
+       .probe          = ncsi_probe,
+       .startup        = ncsi_startup,
+       .shutdown       = ncsi_shutdown,
+};
+
+int phy_ncsi_init(void)
+{
+       phy_register(&ncsi_driver);
+       return 0;
+}
include/net/ncsi-pkt.h new file +2 -0
@@ -0,0 +1,415 @@
+/*
+ * Copyright Gavin Shan, IBM Corporation 2016.
Quoted 8 lines

include/net/ncsi-pkt.h line 8

Quoted 6 lines
include/net/ncsi-pkt.h new file +6 -0
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */

If you grab the version from 5.2-rc3 it has been SPDXified.

Cheers, will do.

Quoted 10 lines

include/phy.h line 20

Quoted 8 lines
include/phy.h +1 -0
@@ -17,6 +17,7 @@
 #include <phy_interface.h>
 
 #define PHY_FIXED_ID           0xa5a55a5a
+#define PHY_NCSI_ID            0xbeefcafe

hmmm...

Haha - suggestions welcome?

Quoted 6 lines
Quoted 6 lines
include/phy.h +0 -1
 #define PHY_MAX_ADDR 32
 
--

2.21.0

#4 lore raw

PATCH 2/4 phy: Include NC-SI in phy setup

In reply to Samuel Mendoza-Jonas

Add NC-SI to the usual phy handling. This makes two notable changes:
- Somewhat similar to a fixed phy, phy_connect() will create an NC-SI phy if CONFIG_PHY_NCSI is defined.
- An early return is added to phy_read() and phy_write() to handle a case like the NC-SI phy which does not define a bus.

---
 drivers/net/phy/phy.c   |  8 ++++++++
 include/phy.h           | 11 +++++++++++
 include/phy_interface.h |  2 ++
 3 files changed, 21 insertions(+)
drivers/net/phy/phy.c +8 -0
index cda4caa803..c2bc57c45d 100644
@@ -527,6 +527,9 @@ int phy_init(void)
 #endif
 #ifdef CONFIG_PHY_FIXED
 	phy_fixed_init();
+#endif
+#ifdef CONFIG_PHY_NCSI
+	phy_ncsi_init();
 #endif
 	return 0;
 }
@@ -902,6 +905,11 @@ struct phy_device *phy_connect(struct mii_dev *bus, int addr,
 		sn = fdt_next_subnode(gd->fdt_blob, sn);
 	}
 #endif
+
+#ifdef CONFIG_PHY_NCSI
+	phydev = phy_device_create(bus, 0, PHY_NCSI_ID, interface);
+#endif
+
 	if (!phydev)
 		phydev = phy_find_by_mask(bus, 1 << addr, interface);
 
include/phy.h +11 -0
index f0c9df46c6..e1d546a81f 100644
@@ -154,6 +154,11 @@ static inline int phy_read(struct phy_device *phydev, int devad, int regnum)
 {
 	struct mii_dev *bus = phydev->bus;
 
+	if (!bus || !bus->read) {
+		debug("%s: No bus configured\n", __func__);
+		return -1;
+	}
+
 	return bus->read(bus, phydev->addr, devad, regnum);
 }
 
@@ -162,6 +167,11 @@ static inline int phy_write(struct phy_device *phydev, int devad, int regnum,
 {
 	struct mii_dev *bus = phydev->bus;
 
+	if (!bus || !bus->read) {
+		debug("%s: No bus configured\n", __func__);
+		return -1;
+	}
+
 	return bus->write(bus, phydev->addr, devad, regnum, val);
 }
 
@@ -241,6 +251,7 @@ int phy_vitesse_init(void);
 int phy_xilinx_init(void);
 int phy_mscc_init(void);
 int phy_fixed_init(void);
+int phy_ncsi_init(void);
 
 int board_phy_config(struct phy_device *phydev);
 int get_phy_id(struct mii_dev *bus, int addr, int devad, u32 *phy_id);
include/phy_interface.h +2 -0
index c6823189f8..9e334bb47d 100644
@@ -31,6 +31,7 @@ typedef enum {
 	PHY_INTERFACE_MODE_XLAUI,
 	PHY_INTERFACE_MODE_CAUI2,
 	PHY_INTERFACE_MODE_CAUI4,
+	PHY_INTERFACE_MODE_NCSI,
 	PHY_INTERFACE_MODE_NONE,	/* Must be last */
 
 	PHY_INTERFACE_MODE_COUNT,
@@ -58,6 +59,7 @@ static const char * const phy_interface_strings[] = {
 	[PHY_INTERFACE_MODE_XLAUI]		= "xlaui4",
 	[PHY_INTERFACE_MODE_CAUI2]		= "caui2",
 	[PHY_INTERFACE_MODE_CAUI4]		= "caui4",
+	[PHY_INTERFACE_MODE_NCSI]		= "NC-SI",
 	[PHY_INTERFACE_MODE_NONE]		= "",
 };
 
Signature · 1 line
2.21.0
Reviewed-by on v1 2/4
#5 lore raw

Re: [PATCH 2/4] phy: Include NC-SI in phy setup

Quoted 8 lines · Samuel Mendoza-Jonas

Add NC-SI to the usual phy handling. This makes two notable changes:
- Somewhat similar to a fixed phy, phy_connect() will create an NC-SI phy if CONFIG_PHY_NCSI is defined.
- An early return is added to phy_read() and phy_write() to handle a case like the NC-SI phy which does not define a bus.

#6 lore raw

PATCH 3/4 net/ftgmac100: Add NC-SI mode support

In reply to Samuel Mendoza-Jonas

Update the ftgmac100 driver to support NC-SI instead of an mdio phy where available. This is a common setup for Aspeed AST2x00 platforms.

NC-SI mode is determined from the device-tree if either phy-mode sets it or the use-ncsi property exists. If set then normal mdio setup is skipped in favour of the NC-SI phy.

---
 drivers/net/ftgmac100.c | 39 +++++++++++++++++++++++++++++----------
 1 file changed, 29 insertions(+), 10 deletions(-)
drivers/net/ftgmac100.c +29 -10
index 92c38a81bd..c0100e91c7 100644
@@ -18,6 +18,7 @@
 #include <wait_bit.h>
 #include <linux/io.h>
 #include <linux/iopoll.h>
+#include <net/ncsi.h>
 
 #include "ftgmac100.h"
 
@@ -81,6 +82,7 @@ struct ftgmac100_data {
 	struct mii_dev *bus;
 	u32 phy_mode;
 	u32 max_speed;
+	bool ncsi_mode;
 
 	struct clk_bulk clks;
 
@@ -181,7 +183,7 @@ static int ftgmac100_phy_adjust_link(struct ftgmac100_data *priv)
 	struct phy_device *phydev = priv->phydev;
 	u32 maccr;
 
-	if (!phydev->link) {
+	if (!phydev->link && !priv->ncsi_mode) {
 		dev_err(phydev->dev, "No link\n");
 		return -EREMOTEIO;
 	}
@@ -217,7 +219,8 @@ static int ftgmac100_phy_init(struct udevice *dev)
 	if (!phydev)
 		return -ENODEV;
 
-	phydev->supported &= PHY_GBIT_FEATURES;
+	if (!priv->ncsi_mode)
+		phydev->supported &= PHY_GBIT_FEATURES;
 	if (priv->max_speed) {
 		ret = phy_set_supported(phydev, priv->max_speed);
 		if (ret)
@@ -275,7 +278,8 @@ static void ftgmac100_stop(struct udevice *dev)
 
 	writel(0, &ftgmac100->maccr);
 
-	phy_shutdown(priv->phydev);
+	if (!priv->ncsi_mode)
+		phy_shutdown(priv->phydev);
 }
 
 static int ftgmac100_start(struct udevice *dev)
@@ -513,6 +517,7 @@ static int ftgmac100_ofdata_to_platdata(struct udevice *dev)
 	pdata->iobase = devfdt_get_addr(dev);
 	pdata->phy_interface = -1;
 	phy_mode = dev_read_string(dev, "phy-mode");
+
 	if (phy_mode)
 		pdata->phy_interface = phy_get_interface_by_name(phy_mode);
 	if (pdata->phy_interface == -1) {
@@ -537,8 +542,13 @@ static int ftgmac100_probe(struct udevice *dev)
 {
 	struct eth_pdata *pdata = dev_get_platdata(dev);
 	struct ftgmac100_data *priv = dev_get_priv(dev);
+	const char *phy_mode;
 	int ret;
 
+	phy_mode = dev_read_string(dev, "phy-mode");
+	priv->ncsi_mode = dev_read_bool(dev, "use-ncsi") ||
+		(phy_mode && strcmp(phy_mode, "NC-SI") == 0);
+
 	priv->iobase = (struct ftgmac100 *)pdata->iobase;
 	priv->phy_mode = pdata->phy_interface;
 	priv->max_speed = pdata->max_speed;
@@ -548,10 +558,15 @@ static int ftgmac100_probe(struct udevice *dev)
 	if (ret)
 		goto out;
 
-	ret = ftgmac100_mdio_init(dev);
-	if (ret) {
-		dev_err(dev, "Failed to initialize mdiobus: %d\n", ret);
-		goto out;
+	if (priv->ncsi_mode) {
+		printf("%s - NCSI detected\n", __func__);
+	} else {
+		ret = ftgmac100_mdio_init(dev);
+		if (ret) {
+			dev_err(dev, "Failed to initialize mdiobus: %d\n", ret);
+			goto out;
+		}
+
 	}
 
 	ret = ftgmac100_phy_init(dev);
@@ -571,9 +586,13 @@ static int ftgmac100_remove(struct udevice *dev)
 {
 	struct ftgmac100_data *priv = dev_get_priv(dev);
 
-	free(priv->phydev);
-	mdio_unregister(priv->bus);
-	mdio_free(priv->bus);
+	if (!priv->ncsi_mode) {
+		free(priv->phydev);
+		mdio_unregister(priv->bus);
+		mdio_free(priv->bus);
+	} else {
+		free(priv->phydev);
+	}
 	clk_release_bulk(&priv->clks);
 
 	return 0;
Signature · 1 line
2.21.0
#7 lore raw

Re: [PATCH 3/4] net/ftgmac100: Add NC-SI mode support

Quoted 22 lines · Samuel Mendoza-Jonas

Update the ftgmac100 driver to support NC-SI instead of an mdio phy where available. This is a common setup for Aspeed AST2x00 platforms.

NC-SI mode is determined from the device-tree if either phy-mode sets it or the use-ncsi property exists. If set then normal mdio setup is skipped in favour of the NC-SI phy.

---
 drivers/net/ftgmac100.c | 39 +++++++++++++++++++++++++++++----------
drivers/net/ftgmac100.c +1 -1
@@ -181,7 +183,7 @@ static int ftgmac100_phy_adjust_link(struct ftgmac100_data *priv)
        struct phy_device *phydev = priv->phydev;
        u32 maccr;
 
-       if (!phydev->link) {
+       if (!phydev->link && !priv->ncsi_mode) {
                dev_err(phydev->dev, "No link\n");
                return -EREMOTEIO;
        }

drivers/net/ftgmac100.c static int ftgmac100_phy_init(struct udevice *dev) · line 223

Quoted 6 lines
drivers/net/ftgmac100.c +2 -1
@@ -217,7 +219,8 @@ static int ftgmac100_phy_init(struct udevice *dev)
        if (!phydev)
                return -ENODEV;
 
-       phydev->supported &= PHY_GBIT_FEATURES;
+       if (!priv->ncsi_mode)
+               phydev->supported &= PHY_GBIT_FEATURES;

Instead of hiding this behind ncsi_mode I was assuming the ncsi "phy" driver would answer these questions through the phy_read callbacks.

If you don't think that is going to work then this solution isn't too invasive, and we've been using this in Linux for a while, so we know it works.

#8 lore raw

Re: [PATCH 3/4] net/ftgmac100: Add NC-SI mode support

Quoted 31 lines · Joel Stanley
Quoted 20 lines · Samuel Mendoza-Jonas

Update the ftgmac100 driver to support NC-SI instead of an mdio phy where available. This is a common setup for Aspeed AST2x00 platforms.

NC-SI mode is determined from the device-tree if either phy-mode sets it or the use-ncsi property exists. If set then normal mdio setup is skipped in favour of the NC-SI phy.

---
 drivers/net/ftgmac100.c | 39 +++++++++++++++++++++++++++++----------
drivers/net/ftgmac100.c +1 -1
@@ -181,7 +183,7 @@ static int ftgmac100_phy_adjust_link(struct ftgmac100_data *priv)
        struct phy_device *phydev = priv->phydev;
        u32 maccr;
 
-       if (!phydev->link) {
+       if (!phydev->link && !priv->ncsi_mode) {
                dev_err(phydev->dev, "No link\n");
                return -EREMOTEIO;
        }

drivers/net/ftgmac100.c static int ftgmac100_phy_init(struct udevice *dev) · line 223

Quoted 6 lines
drivers/net/ftgmac100.c +2 -1
@@ -217,7 +219,8 @@ static int ftgmac100_phy_init(struct udevice *dev)
        if (!phydev)
                return -ENODEV;
 
-       phydev->supported &= PHY_GBIT_FEATURES;
+       if (!priv->ncsi_mode)
+               phydev->supported &= PHY_GBIT_FEATURES;

Instead of hiding this behind ncsi_mode I was assuming the ncsi "phy" driver would answer these questions through the phy_read callbacks.

Quoted 3 lines

If you don't think that is going to work then this solution isn't too invasive, and we've been using this in Linux for a while, so we know it works.

Having this purely transparent would indeed be nice, but it's been a little awkward fitting NCSI into the u-boot model. phy_read, etc don't really make sense for a NCSI "PHY" like in the Linux model since there aren't any registers as such to query - thus checking the device-tree and skipping the normal mdio bus setup.
I'll have a bit more of a think about the driver model, and probably exactly how Aspeed (for example) hooks this up; maybe there's a way we can fit in to the more usual model.

#9 lore raw

PATCH 4/4 net: NC-SI setup and handling

In reply to Samuel Mendoza-Jonas

Add the handling of NC-SI ethernet frames, and add a check at the start of net_loop() to configure NC-SI before starting other network commands. This also adds an "ncsi" command to manually start NC-SI configuration.

---
 cmd/Kconfig   |  6 ++++++
 cmd/net.c     | 18 ++++++++++++++++++
 include/net.h |  2 +-
 net/net.c     | 27 ++++++++++++++++++++++++++-
 4 files changed, 51 insertions(+), 2 deletions(-)
cmd/Kconfig +6 -0
index ea1a325eb3..0b5df0db90 100644
@@ -1296,6 +1296,12 @@ config CMD_LINK_LOCAL
 	help
 	  Acquire a network IP address using the link-local protocol
 
+config CMD_NCSI
+	bool "ncsi"
+	depends on PHY_NCSI
+	help
+	  Configure the attached NIC via NC-SI
+
 endif
 
 config CMD_ETHSW
cmd/net.c +18 -0
index 89721b8f8b..7d2c21ba4d 100644
@@ -457,3 +457,21 @@ U_BOOT_CMD(
 );
 
 #endif  /* CONFIG_CMD_LINK_LOCAL */
+
+#if defined(CONFIG_CMD_NCSI)
+static int do_ncsi(cmd_tbl_t *cmdtp, int flag, int argc,
+		   char * const argv[])
+{
+	if (net_loop(NCSI) < 0)
+		return CMD_RET_FAILURE;
+
+	return CMD_RET_SUCCESS;
+}
+
+U_BOOT_CMD(
+	ncsi,	1,	1,	do_ncsi,
+	"Configure attached NIC via NC-SI",
+	""
+);
+
+#endif  /* CONFIG_CMD_NCSI */
include/net.h +1 -1
index 4f7ba211b6..82a71c5219 100644
@@ -538,7 +538,7 @@ extern int		net_restart_wrap;	/* Tried all network devices */
 
 enum proto_t {
 	BOOTP, RARP, ARP, TFTPGET, DHCP, PING, DNS, NFS, CDP, NETCONS, SNTP,
-	TFTPSRV, TFTPPUT, LINKLOCAL, FASTBOOT, WOL
+	TFTPSRV, TFTPPUT, LINKLOCAL, FASTBOOT, WOL, NCSI
 };
 
 extern char	net_boot_file_name[1024];/* Boot File name */
net/net.c +26 -1
index a5a216c3ee..16ca2b0e18 100644
@@ -95,6 +95,7 @@
 #include <net.h>
 #include <net/fastboot.h>
 #include <net/tftp.h>
+#include <net/ncsi.h>
 #if defined(CONFIG_LED_STATUS)
 #include <miiphy.h>
 #include <status_led.h>
@@ -411,6 +412,16 @@ int net_loop(enum proto_t protocol)
 	net_try_count = 1;
 	debug_cond(DEBUG_INT_STATE, "--- net_loop Entry\n");
 
+#ifdef CONFIG_PHY_NCSI
+	if (protocol != NCSI && !ncsi_active()) {
+		printf("%s: configuring NCSI first\n", __func__);
+		if (net_loop(NCSI) < 0)
+			return ret;
+		eth_init_state_only();
+		goto restart;
+	}
+#endif
+
 	bootstage_mark_name(BOOTSTAGE_ID_ETH_START, "eth_start");
 	net_init();
 	if (eth_is_on_demand_init() || protocol != NETCONS) {
@@ -424,6 +435,7 @@ int net_loop(enum proto_t protocol)
 	} else {
 		eth_init_state_only();
 	}
+
 restart:
 #ifdef CONFIG_USB_KEYBOARD
 	net_busy_flag = 0;
@@ -530,6 +542,11 @@ restart:
 		case WOL:
 			wol_start();
 			break;
+#endif
+#if defined(CONFIG_CMD_NCSI)
+		case NCSI:
+			ncsi_probe_packages();
+			break;
 #endif
 		default:
 			break;
@@ -641,7 +658,7 @@ restart:
 				env_set_hex("filesize", net_boot_file_size);
 				env_set_hex("fileaddr", load_addr);
 			}
-			if (protocol != NETCONS)
+			if (protocol != NETCONS && protocol != NCSI)
 				eth_halt();
 			else
 				eth_halt_state_only();
@@ -1324,6 +1341,11 @@ void net_process_received_packet(uchar *in_packet, int len)
 	case PROT_WOL:
 		wol_receive(ip, len);
 		break;
+#endif
+#ifdef CONFIG_PHY_NCSI
+	case PROT_NCSI:
+		ncsi_receive(et, ip, len);
+		break;
 #endif
 	}
 }
@@ -1385,6 +1407,9 @@ common:
 
 #ifdef CONFIG_CMD_RARP
 	case RARP:
+#endif
+#ifdef CONFIG_CMD_NCSI
+	case NCSI:
 #endif
 	case BOOTP:
 	case CDP:
Signature · 1 line
2.21.0
Reviewed-by on v1 4/4
#10 lore raw

Re: [PATCH 4/4] net: NC-SI setup and handling

Quoted 6 lines · Samuel Mendoza-Jonas

Add the handling of NC-SI ethernet frames, and add a check at the start of net_loop() to configure NC-SI before starting other network commands. This also adds an "ncsi" command to manually start NC-SI configuration.

We need the u-boot net specialists to weigh in, but this looks okay to me. Please clarify the help text for the NCSI command, then add:

cmd/Kconfig config CMD_LINK_LOCAL · line 1303

Quoted 11 lines
cmd/Kconfig +5 -0
@@ -1296,6 +1296,12 @@ config CMD_LINK_LOCAL
        help
          Acquire a network IP address using the link-local protocol
 
+config CMD_NCSI
+       bool "ncsi"
+       depends on PHY_NCSI
+       help
+         Configure the attached NIC via NC-SI

Can we clarify in the text if we need the 'ncsi' command to use any NC-SI attached NIC, or if it's more for ncsi and network debugging?

#11 lore raw

Re: [PATCH 4/4] net: NC-SI setup and handling

Quoted 18 lines · Joel Stanley
Quoted 5 lines · Samuel Mendoza-Jonas

Add the handling of NC-SI ethernet frames, and add a check at the start of net_loop() to configure NC-SI before starting other network commands. This also adds an "ncsi" command to manually start NC-SI configuration.

We need the u-boot net specialists to weigh in, but this looks okay to me. Please clarify the help text for the NCSI command, then add:

Quoted 6 lines
cmd/Kconfig +0 -0
@@ -1296,6 +1296,12 @@ config CMD_LINK_LOCAL
        help
          Acquire a network IP address using the link-local protocol
 
Quoted 8 lines

cmd/Kconfig config CMD_LINK_LOCAL · line 1303

Quoted 5 lines
cmd/Kconfig +5 -0
+config CMD_NCSI
+       bool "ncsi"
+       depends on PHY_NCSI
+       help
+         Configure the attached NIC via NC-SI

Can we clarify in the text if we need the 'ncsi' command to use any NC-SI attached NIC, or if it's more for ncsi and network debugging?

Can do - the command manually kicks off configuration of the NC-SI topology, but isn't needed in normal use.