* [PATCH net-next v5 01/12] net: ethtool: Add support for ethnl_info_init_ntf helper function
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-18 16:19 ` [PATCH net-next v5 02/12] net: pse-pd: Add support for reporting events Kory Maincent
` (10 subsequent siblings)
11 siblings, 0 replies; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
Introduce support for the ethnl_info_init_ntf helper function to enable
initialization of ethtool notifications outside of the netlink.c file.
This change allows for more flexible notification handling.
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
Changes in v4:
- Use the new helper in ethnl_default_notify function.
Changes in v2:
- new patch.
---
net/ethtool/netlink.c | 7 ++++++-
net/ethtool/netlink.h | 2 ++
2 files changed, 8 insertions(+), 1 deletion(-)
diff --git a/net/ethtool/netlink.c b/net/ethtool/netlink.c
index b4c45207fa32..bb1a35494935 100644
--- a/net/ethtool/netlink.c
+++ b/net/ethtool/netlink.c
@@ -758,7 +758,7 @@ static void ethnl_default_notify(struct net_device *dev, unsigned int cmd,
int reply_len;
int ret;
- genl_info_init_ntf(&info, ðtool_genl_family, cmd);
+ ethnl_info_init_ntf(&info, cmd);
if (WARN_ONCE(cmd > ETHTOOL_MSG_KERNEL_MAX ||
!ethnl_default_notify_ops[cmd],
@@ -825,6 +825,11 @@ static void ethnl_default_notify(struct net_device *dev, unsigned int cmd,
typedef void (*ethnl_notify_handler_t)(struct net_device *dev, unsigned int cmd,
const void *data);
+void ethnl_info_init_ntf(struct genl_info *info, u8 cmd)
+{
+ genl_info_init_ntf(info, ðtool_genl_family, cmd);
+}
+
static const ethnl_notify_handler_t ethnl_notify_handlers[] = {
[ETHTOOL_MSG_LINKINFO_NTF] = ethnl_default_notify,
[ETHTOOL_MSG_LINKMODES_NTF] = ethnl_default_notify,
diff --git a/net/ethtool/netlink.h b/net/ethtool/netlink.h
index ff69ca0715de..af20a175e111 100644
--- a/net/ethtool/netlink.h
+++ b/net/ethtool/netlink.h
@@ -322,6 +322,8 @@ struct ethnl_sock_priv {
int ethnl_sock_priv_set(struct sk_buff *skb, struct net_device *dev, u32 portid,
enum ethnl_sock_type type);
+void ethnl_info_init_ntf(struct genl_info *info, u8 cmd);
+
/**
* struct ethnl_request_ops - unified handling of GET and SET requests
* @request_cmd: command id for request (GET)
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* [PATCH net-next v5 02/12] net: pse-pd: Add support for reporting events
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
2025-02-18 16:19 ` [PATCH net-next v5 01/12] net: ethtool: Add support for ethnl_info_init_ntf helper function Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-21 0:42 ` Jakub Kicinski
2025-02-21 8:50 ` Oleksij Rempel
2025-02-18 16:19 ` [PATCH net-next v5 03/12] net: pse-pd: tps23881: Add support for PSE events and interrupts Kory Maincent
` (9 subsequent siblings)
11 siblings, 2 replies; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
Add support for devm_pse_irq_helper() to register PSE interrupts. This aims
to report events such as over-current or over-temperature conditions
similarly to how the regulator API handles them but using a specific PSE
ethtool netlink socket.
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
Change in v4:
- Fix netlink notification message issues.
- Use netlink bitset in ethtool_pse_send_ntf.
- Add kdoc.
Change in v3:
- Remove C33 prefix when it is not in the standards.
- Fix pse_to_regulator_notifs which could not report regulator events
together.
- Fix deadlock issue.
- Save interrupt in pcdev structure for later use.
Change in v2:
- Add support for PSE ethtool notification.
- Saved the attached phy_device in the pse_control structure to know which
interface should have the notification.
- Rethink devm_pse_irq_helper() without devm_regulator_irq_helper() call.
---
Documentation/netlink/specs/ethtool.yaml | 26 ++++
Documentation/networking/ethtool-netlink.rst | 19 +++
drivers/net/mdio/fwnode_mdio.c | 26 ++--
drivers/net/pse-pd/pse_core.c | 157 ++++++++++++++++++++++++-
include/linux/ethtool_netlink.h | 9 ++
include/linux/pse-pd/pse.h | 24 +++-
include/uapi/linux/ethtool.h | 17 +++
include/uapi/linux/ethtool_netlink_generated.h | 10 ++
net/ethtool/common.c | 6 +
net/ethtool/common.h | 2 +
net/ethtool/pse-pd.c | 53 +++++++++
net/ethtool/strset.c | 5 +
12 files changed, 337 insertions(+), 17 deletions(-)
diff --git a/Documentation/netlink/specs/ethtool.yaml b/Documentation/netlink/specs/ethtool.yaml
index 655d8d10fe24..da78c5daf537 100644
--- a/Documentation/netlink/specs/ethtool.yaml
+++ b/Documentation/netlink/specs/ethtool.yaml
@@ -1526,6 +1526,22 @@ attribute-sets:
name: hwtstamp-flags
type: nest
nested-attributes: bitset
+ -
+ name: pse-ntf
+ attr-cnt-name: __ethtool-a-pse-ntf-cnt
+ attributes:
+ -
+ name: unspec
+ type: unused
+ value: 0
+ -
+ name: header
+ type: nest
+ nested-attributes: header
+ -
+ name: events
+ type: nest
+ nested-attributes: bitset
operations:
enum-model: directional
@@ -2382,3 +2398,13 @@ operations:
attributes: *tsconfig
reply:
attributes: *tsconfig
+ -
+ name: pse-ntf
+ doc: Notification for pse events.
+
+ attribute-set: pse-ntf
+
+ event:
+ attributes:
+ - header
+ - events
diff --git a/Documentation/networking/ethtool-netlink.rst b/Documentation/networking/ethtool-netlink.rst
index 3770a2294509..9fc5e29b3928 100644
--- a/Documentation/networking/ethtool-netlink.rst
+++ b/Documentation/networking/ethtool-netlink.rst
@@ -290,6 +290,7 @@ Kernel to userspace:
``ETHTOOL_MSG_PHY_NTF`` Ethernet PHY information change
``ETHTOOL_MSG_TSCONFIG_GET_REPLY`` hw timestamping configuration
``ETHTOOL_MSG_TSCONFIG_SET_REPLY`` new hw timestamping configuration
+ ``ETHTOOL_MSG_PSE_NTF`` PSE events notification
======================================== =================================
``GET`` requests are sent by userspace applications to retrieve device
@@ -1896,6 +1897,24 @@ various existing products that document power consumption in watts rather than
classes. If power limit configuration based on classes is needed, the
conversion can be done in user space, for example by ethtool.
+PSE_NTF
+=======
+
+Notify PSE events.
+
+Notification contents:
+
+ =============================== ====== ========================
+ ``ETHTOOL_A_PSE_HEADER`` nested request header
+ ``ETHTOOL_A_PSE_EVENTS`` bitset PSE events
+ =============================== ====== ========================
+
+When set, the optional ``ETHTOOL_A_PSE_EVENTS`` attribute identifies the
+PSE events.
+
+.. kernel-doc:: include/uapi/linux/ethtool.h
+ :identifiers: ethtool_pse_events
+
RSS_GET
=======
diff --git a/drivers/net/mdio/fwnode_mdio.c b/drivers/net/mdio/fwnode_mdio.c
index aea0f0357568..9b41d4697a40 100644
--- a/drivers/net/mdio/fwnode_mdio.c
+++ b/drivers/net/mdio/fwnode_mdio.c
@@ -18,7 +18,8 @@ MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("FWNODE MDIO bus (Ethernet PHY) accessors");
static struct pse_control *
-fwnode_find_pse_control(struct fwnode_handle *fwnode)
+fwnode_find_pse_control(struct fwnode_handle *fwnode,
+ struct phy_device *phydev)
{
struct pse_control *psec;
struct device_node *np;
@@ -30,7 +31,7 @@ fwnode_find_pse_control(struct fwnode_handle *fwnode)
if (!np)
return NULL;
- psec = of_pse_control_get(np);
+ psec = of_pse_control_get(np, phydev);
if (PTR_ERR(psec) == -ENOENT)
return NULL;
@@ -128,15 +129,9 @@ int fwnode_mdiobus_register_phy(struct mii_bus *bus,
u32 phy_id;
int rc;
- psec = fwnode_find_pse_control(child);
- if (IS_ERR(psec))
- return PTR_ERR(psec);
-
mii_ts = fwnode_find_mii_timestamper(child);
- if (IS_ERR(mii_ts)) {
- rc = PTR_ERR(mii_ts);
- goto clean_pse;
- }
+ if (IS_ERR(mii_ts))
+ return PTR_ERR(mii_ts);
is_c45 = fwnode_device_is_compatible(child, "ethernet-phy-ieee802.3-c45");
if (is_c45 || fwnode_get_phy_id(child, &phy_id))
@@ -169,6 +164,12 @@ int fwnode_mdiobus_register_phy(struct mii_bus *bus,
goto clean_phy;
}
+ psec = fwnode_find_pse_control(child, phy);
+ if (IS_ERR(psec)) {
+ rc = PTR_ERR(psec);
+ goto unregister_phy;
+ }
+
phy->psec = psec;
/* phy->mii_ts may already be defined by the PHY driver. A
@@ -180,12 +181,13 @@ int fwnode_mdiobus_register_phy(struct mii_bus *bus,
return 0;
+unregister_phy:
+ if (is_acpi_node(child) || is_of_node(child))
+ phy_device_remove(phy);
clean_phy:
phy_device_free(phy);
clean_mii_ts:
unregister_mii_timestamper(mii_ts);
-clean_pse:
- pse_control_put(psec);
return rc;
}
diff --git a/drivers/net/pse-pd/pse_core.c b/drivers/net/pse-pd/pse_core.c
index 4602e26eb8c8..10a5ab30afdd 100644
--- a/drivers/net/pse-pd/pse_core.c
+++ b/drivers/net/pse-pd/pse_core.c
@@ -7,6 +7,7 @@
#include <linux/device.h>
#include <linux/ethtool.h>
+#include <linux/ethtool_netlink.h>
#include <linux/of.h>
#include <linux/pse-pd/pse.h>
#include <linux/regulator/driver.h>
@@ -23,6 +24,7 @@ static LIST_HEAD(pse_controller_list);
* @list: list entry for the pcdev's PSE controller list
* @id: ID of the PSE line in the PSE controller device
* @refcnt: Number of gets of this pse_control
+ * @attached_phydev: PHY device pointer attached by the PSE control
*/
struct pse_control {
struct pse_controller_dev *pcdev;
@@ -30,6 +32,7 @@ struct pse_control {
struct list_head list;
unsigned int id;
struct kref refcnt;
+ struct phy_device *attached_phydev;
};
static int of_load_single_pse_pi_pairset(struct device_node *node,
@@ -557,6 +560,151 @@ int devm_pse_controller_register(struct device *dev,
}
EXPORT_SYMBOL_GPL(devm_pse_controller_register);
+struct pse_irq {
+ struct pse_controller_dev *pcdev;
+ struct pse_irq_desc desc;
+ unsigned long *notifs;
+};
+
+/**
+ * pse_to_regulator_notifs - Convert PSE notifications to Regulator
+ * notifications
+ * @notifs: PSE notifications
+ *
+ * Return: Regulator notifications
+ */
+static unsigned long pse_to_regulator_notifs(unsigned long notifs)
+{
+ unsigned long rnotifs = 0;
+
+ if (notifs & ETHTOOL_PSE_EVENT_OVER_CURRENT)
+ rnotifs |= REGULATOR_EVENT_OVER_CURRENT;
+ if (notifs & ETHTOOL_PSE_EVENT_OVER_TEMP)
+ rnotifs |= REGULATOR_EVENT_OVER_TEMP;
+
+ return rnotifs;
+}
+
+/**
+ * pse_control_find_phy_by_id - Find PHY attached to the a pse control id
+ * @pcdev: a pointer to the PSE
+ * @id: index of the PSE control
+ *
+ * Return: PHY device pointer or NULL
+ */
+static struct phy_device *
+pse_control_find_phy_by_id(struct pse_controller_dev *pcdev, int id)
+{
+ struct pse_control *psec;
+
+ mutex_lock(&pse_list_mutex);
+ list_for_each_entry(psec, &pcdev->pse_control_head, list) {
+ if (psec->id == id) {
+ mutex_unlock(&pse_list_mutex);
+ return psec->attached_phydev;
+ }
+ }
+ mutex_unlock(&pse_list_mutex);
+
+ return NULL;
+}
+
+/**
+ * pse_isr - IRQ handler for PSE
+ * @irq: irq number
+ * @data: pointer to user interrupt structure
+ *
+ * Return: irqreturn_t - status of IRQ
+ */
+static irqreturn_t pse_isr(int irq, void *data)
+{
+ struct netlink_ext_ack extack = {};
+ struct pse_controller_dev *pcdev;
+ unsigned long notifs_mask = 0;
+ struct pse_irq_desc *desc;
+ struct pse_irq *h = data;
+ int ret, i;
+
+ desc = &h->desc;
+ pcdev = h->pcdev;
+
+ /* Clear notifs mask */
+ memset(h->notifs, 0, pcdev->nr_lines * sizeof(*h->notifs));
+ mutex_lock(&pcdev->lock);
+ ret = desc->map_event(irq, pcdev, h->notifs, ¬ifs_mask);
+ mutex_unlock(&pcdev->lock);
+ if (ret || !notifs_mask)
+ return IRQ_NONE;
+
+ for_each_set_bit(i, ¬ifs_mask, pcdev->nr_lines) {
+ struct phy_device *phydev;
+ unsigned long notifs, rnotifs;
+
+ /* Do nothing PI not described */
+ if (!pcdev->pi[i].rdev)
+ continue;
+
+ notifs = h->notifs[i];
+ dev_dbg(h->pcdev->dev,
+ "Sending PSE notification EVT 0x%lx\n", notifs);
+
+ phydev = pse_control_find_phy_by_id(pcdev, i);
+ if (phydev)
+ ethnl_pse_send_ntf(phydev, notifs, &extack);
+ rnotifs = pse_to_regulator_notifs(notifs);
+ regulator_notifier_call_chain(pcdev->pi[i].rdev, rnotifs,
+ NULL);
+ }
+
+ return IRQ_HANDLED;
+}
+
+/**
+ * devm_pse_irq_helper - Register IRQ based PSE event notifier
+ *
+ * @pcdev: a pointer to the PSE
+ * @irq: the irq value to be passed to request_irq
+ * @irq_flags: the flags to be passed to request_irq
+ * @d: PSE interrupt description
+ *
+ * Return: 0 on success and failure value on error
+ */
+int devm_pse_irq_helper(struct pse_controller_dev *pcdev, int irq,
+ int irq_flags, const struct pse_irq_desc *d)
+{
+ struct device *dev = pcdev->dev;
+ struct pse_irq *h;
+ int ret;
+
+ if (!d || !d->map_event || !d->name)
+ return -EINVAL;
+
+ h = devm_kzalloc(dev, sizeof(*h), GFP_KERNEL);
+ if (!h)
+ return -ENOMEM;
+
+ h->pcdev = pcdev;
+ h->desc = *d;
+ h->desc.name = devm_kstrdup(dev, d->name, GFP_KERNEL);
+ if (!h->desc.name)
+ return -ENOMEM;
+
+ h->notifs = devm_kcalloc(pcdev->dev, pcdev->nr_lines,
+ sizeof(*h->notifs), GFP_KERNEL);
+ if (!h->notifs)
+ return -ENOMEM;
+
+ ret = devm_request_threaded_irq(dev, irq, NULL, pse_isr,
+ IRQF_ONESHOT | irq_flags,
+ h->desc.name, h);
+ if (ret)
+ dev_err(pcdev->dev, "Failed to request IRQ %d\n", irq);
+
+ pcdev->irq = irq;
+ return ret;
+}
+EXPORT_SYMBOL_GPL(devm_pse_irq_helper);
+
/* PSE control section */
static void __pse_control_release(struct kref *kref)
@@ -599,7 +747,8 @@ void pse_control_put(struct pse_control *psec)
EXPORT_SYMBOL_GPL(pse_control_put);
static struct pse_control *
-pse_control_get_internal(struct pse_controller_dev *pcdev, unsigned int index)
+pse_control_get_internal(struct pse_controller_dev *pcdev, unsigned int index,
+ struct phy_device *phydev)
{
struct pse_control *psec;
int ret;
@@ -638,6 +787,7 @@ pse_control_get_internal(struct pse_controller_dev *pcdev, unsigned int index)
psec->pcdev = pcdev;
list_add(&psec->list, &pcdev->pse_control_head);
psec->id = index;
+ psec->attached_phydev = phydev;
kref_init(&psec->refcnt);
return psec;
@@ -693,7 +843,8 @@ static int psec_id_xlate(struct pse_controller_dev *pcdev,
return pse_spec->args[0];
}
-struct pse_control *of_pse_control_get(struct device_node *node)
+struct pse_control *of_pse_control_get(struct device_node *node,
+ struct phy_device *phydev)
{
struct pse_controller_dev *r, *pcdev;
struct of_phandle_args args;
@@ -743,7 +894,7 @@ struct pse_control *of_pse_control_get(struct device_node *node)
}
/* pse_list_mutex also protects the pcdev's pse_control list */
- psec = pse_control_get_internal(pcdev, psec_id);
+ psec = pse_control_get_internal(pcdev, psec_id, phydev);
out:
mutex_unlock(&pse_list_mutex);
diff --git a/include/linux/ethtool_netlink.h b/include/linux/ethtool_netlink.h
index aba91335273a..0fa1d8f59cf2 100644
--- a/include/linux/ethtool_netlink.h
+++ b/include/linux/ethtool_netlink.h
@@ -43,6 +43,9 @@ void ethtool_aggregate_rmon_stats(struct net_device *dev,
struct ethtool_rmon_stats *rmon_stats);
bool ethtool_dev_mm_supported(struct net_device *dev);
+void ethnl_pse_send_ntf(struct phy_device *phydev, unsigned long notif,
+ struct netlink_ext_ack *extack);
+
#else
static inline int ethnl_cable_test_alloc(struct phy_device *phydev, u8 cmd)
{
@@ -120,6 +123,12 @@ static inline bool ethtool_dev_mm_supported(struct net_device *dev)
return false;
}
+static inline void ethnl_pse_send_ntf(struct phy_device *phydev,
+ unsigned long notif,
+ struct netlink_ext_ack *extack)
+{
+}
+
#endif /* IS_ENABLED(CONFIG_ETHTOOL_NETLINK) */
static inline int ethnl_cable_test_result(struct phy_device *phydev, u8 pair,
diff --git a/include/linux/pse-pd/pse.h b/include/linux/pse-pd/pse.h
index c773eeb92d04..5d41a1c984bd 100644
--- a/include/linux/pse-pd/pse.h
+++ b/include/linux/pse-pd/pse.h
@@ -7,6 +7,7 @@
#include <linux/list.h>
#include <uapi/linux/ethtool.h>
+#include <linux/regulator/driver.h>
/* Maximum current in uA according to IEEE 802.3-2022 Table 145-1 */
#define MAX_PI_CURRENT 1920000
@@ -37,6 +38,19 @@ struct ethtool_c33_pse_pw_limit_range {
u32 max;
};
+/**
+ * struct pse_irq_desc - notification sender description for IRQ based events.
+ *
+ * @name: the visible name for the IRQ
+ * @map_event: driver callback to map IRQ status into PSE devices with events.
+ */
+struct pse_irq_desc {
+ const char *name;
+ int (*map_event)(int irq, struct pse_controller_dev *pcdev,
+ unsigned long *notifs,
+ unsigned long *notifs_mask);
+};
+
/**
* struct pse_control_config - PSE control/channel configuration.
*
@@ -228,6 +242,7 @@ struct pse_pi {
* @types: types of the PSE controller
* @pi: table of PSE PIs described in this controller device
* @no_of_pse_pi: flag set if the pse_pis devicetree node is not used
+ * @irq: PSE interrupt
*/
struct pse_controller_dev {
const struct pse_controller_ops *ops;
@@ -241,6 +256,7 @@ struct pse_controller_dev {
enum ethtool_pse_types types;
struct pse_pi *pi;
bool no_of_pse_pi;
+ int irq;
};
#if IS_ENABLED(CONFIG_PSE_CONTROLLER)
@@ -249,8 +265,11 @@ void pse_controller_unregister(struct pse_controller_dev *pcdev);
struct device;
int devm_pse_controller_register(struct device *dev,
struct pse_controller_dev *pcdev);
+int devm_pse_irq_helper(struct pse_controller_dev *pcdev, int irq,
+ int irq_flags, const struct pse_irq_desc *d);
-struct pse_control *of_pse_control_get(struct device_node *node);
+struct pse_control *of_pse_control_get(struct device_node *node,
+ struct phy_device *phydev);
void pse_control_put(struct pse_control *psec);
int pse_ethtool_get_status(struct pse_control *psec,
@@ -268,7 +287,8 @@ bool pse_has_c33(struct pse_control *psec);
#else
-static inline struct pse_control *of_pse_control_get(struct device_node *node)
+static inline struct pse_control *of_pse_control_get(struct device_node *node,
+ struct phy_device *phydev)
{
return ERR_PTR(-ENOENT);
}
diff --git a/include/uapi/linux/ethtool.h b/include/uapi/linux/ethtool.h
index 2feba0929a8a..8793946ff851 100644
--- a/include/uapi/linux/ethtool.h
+++ b/include/uapi/linux/ethtool.h
@@ -683,6 +683,7 @@ enum ethtool_link_ext_substate_module {
* @ETH_SS_STATS_RMON: names of RMON statistics
* @ETH_SS_STATS_PHY: names of PHY(dev) statistics
* @ETH_SS_TS_FLAGS: hardware timestamping flags
+ * @ETH_SS_PSE_EVENTS: names of PSE events
*
* @ETH_SS_COUNT: number of defined string sets
*/
@@ -710,6 +711,7 @@ enum ethtool_stringset {
ETH_SS_STATS_RMON,
ETH_SS_STATS_PHY,
ETH_SS_TS_FLAGS,
+ ETH_SS_PSE_EVENTS,
/* add new constants above here */
ETH_SS_COUNT
@@ -1002,6 +1004,21 @@ enum ethtool_c33_pse_pw_d_status {
ETHTOOL_C33_PSE_PW_D_STATUS_OTHERFAULT,
};
+/**
+ * enum ethtool_pse_events - event list of the PSE controller.
+ * @ETHTOOL_PSE_EVENT_OVER_CURRENT: PSE output current is too high.
+ * @ETHTOOL_PSE_EVENT_OVER_TEMP: PSE in over temperature state.
+ *
+ * @ETHTOOL_PSE_EVENT_LAST: Last PSE event of the enum.
+ */
+
+enum ethtool_pse_events {
+ ETHTOOL_PSE_EVENT_OVER_CURRENT = 1 << 0,
+ ETHTOOL_PSE_EVENT_OVER_TEMP = 1 << 1,
+
+ ETHTOOL_PSE_EVENT_LAST = ETHTOOL_PSE_EVENT_OVER_TEMP,
+};
+
/**
* enum ethtool_podl_pse_admin_state - operational state of the PoDL PSE
* functions. IEEE 802.3-2018 30.15.1.1.2 aPoDLPSEAdminState
diff --git a/include/uapi/linux/ethtool_netlink_generated.h b/include/uapi/linux/ethtool_netlink_generated.h
index fe24c3459ac0..f03b51766311 100644
--- a/include/uapi/linux/ethtool_netlink_generated.h
+++ b/include/uapi/linux/ethtool_netlink_generated.h
@@ -709,6 +709,15 @@ enum {
ETHTOOL_A_TSCONFIG_MAX = (__ETHTOOL_A_TSCONFIG_CNT - 1)
};
+enum {
+ ETHTOOL_A_PSE_NTF_UNSPEC,
+ ETHTOOL_A_PSE_NTF_HEADER,
+ ETHTOOL_A_PSE_NTF_EVENTS,
+
+ __ETHTOOL_A_PSE_NTF_CNT,
+ ETHTOOL_A_PSE_NTF_MAX = (__ETHTOOL_A_PSE_NTF_CNT - 1)
+};
+
enum {
ETHTOOL_MSG_USER_NONE = 0,
ETHTOOL_MSG_STRSET_GET = 1,
@@ -813,6 +822,7 @@ enum {
ETHTOOL_MSG_PHY_NTF,
ETHTOOL_MSG_TSCONFIG_GET_REPLY,
ETHTOOL_MSG_TSCONFIG_SET_REPLY,
+ ETHTOOL_MSG_PSE_NTF,
__ETHTOOL_MSG_KERNEL_CNT,
ETHTOOL_MSG_KERNEL_MAX = (__ETHTOOL_MSG_KERNEL_CNT - 1)
diff --git a/net/ethtool/common.c b/net/ethtool/common.c
index 7149d07e90c6..8d207ec6456e 100644
--- a/net/ethtool/common.c
+++ b/net/ethtool/common.c
@@ -517,6 +517,12 @@ const char udp_tunnel_type_names[][ETH_GSTRING_LEN] = {
static_assert(ARRAY_SIZE(udp_tunnel_type_names) ==
__ETHTOOL_UDP_TUNNEL_TYPE_CNT);
+const char pse_event_names[][ETH_GSTRING_LEN] = {
+ [const_ilog2(ETHTOOL_PSE_EVENT_OVER_CURRENT)] = "over-current",
+ [const_ilog2(ETHTOOL_PSE_EVENT_OVER_TEMP)] = "over-temperature",
+};
+static_assert(ARRAY_SIZE(pse_event_names) == __PSE_EVENT_CNT);
+
/* return false if legacy contained non-0 deprecated fields
* maxtxpkt/maxrxpkt. rest of ksettings always updated
*/
diff --git a/net/ethtool/common.h b/net/ethtool/common.h
index 58e9e7db06f9..edef4c230cf1 100644
--- a/net/ethtool/common.h
+++ b/net/ethtool/common.h
@@ -14,6 +14,7 @@
#define __SOF_TIMESTAMPING_CNT (const_ilog2(SOF_TIMESTAMPING_LAST) + 1)
#define __HWTSTAMP_FLAG_CNT (const_ilog2(HWTSTAMP_FLAG_LAST) + 1)
+#define __PSE_EVENT_CNT (const_ilog2(ETHTOOL_PSE_EVENT_LAST) + 1)
struct link_mode_info {
int speed;
@@ -41,6 +42,7 @@ extern const char ts_tx_type_names[][ETH_GSTRING_LEN];
extern const char ts_rx_filter_names[][ETH_GSTRING_LEN];
extern const char ts_flags_names[][ETH_GSTRING_LEN];
extern const char udp_tunnel_type_names[][ETH_GSTRING_LEN];
+extern const char pse_event_names[][ETH_GSTRING_LEN];
int __ethtool_get_link(struct net_device *dev);
diff --git a/net/ethtool/pse-pd.c b/net/ethtool/pse-pd.c
index 2819e2ba6be2..e471e577d4b6 100644
--- a/net/ethtool/pse-pd.c
+++ b/net/ethtool/pse-pd.c
@@ -12,6 +12,7 @@
#include <linux/ethtool_netlink.h>
#include <linux/ethtool.h>
#include <linux/phy.h>
+#include "bitset.h"
struct pse_req_info {
struct ethnl_req_info base;
@@ -315,3 +316,55 @@ const struct ethnl_request_ops ethnl_pse_request_ops = {
.set = ethnl_set_pse,
/* PSE has no notification */
};
+
+void ethnl_pse_send_ntf(struct phy_device *phydev, unsigned long notifs,
+ struct netlink_ext_ack *extack)
+{
+ struct net_device *netdev = phydev->attached_dev;
+ struct genl_info info;
+ void *reply_payload;
+ struct sk_buff *skb;
+ int reply_len;
+ int ret;
+
+ if (!netdev || !notifs)
+ return;
+
+ ethnl_info_init_ntf(&info, ETHTOOL_MSG_PSE_NTF);
+ info.extack = extack;
+
+ reply_len = ethnl_reply_header_size();
+ /* _C33_PSE_NTF_EVENTS */
+ ret = ethnl_bitset_size(¬ifs, NULL, __PSE_EVENT_CNT,
+ pse_event_names, 0);
+ if (ret < 0)
+ return;
+
+ reply_len += ret;
+ skb = genlmsg_new(reply_len, GFP_KERNEL);
+ reply_payload = ethnl_bcastmsg_put(skb, ETHTOOL_MSG_PSE_NTF);
+ if (!reply_payload)
+ goto err_skb;
+
+ ret = ethnl_fill_reply_header(skb, netdev,
+ ETHTOOL_A_PSE_NTF_HEADER);
+ if (ret < 0)
+ goto err_skb;
+
+ ret = ethnl_put_bitset(skb, ETHTOOL_A_PSE_NTF_EVENTS, ¬ifs,
+ NULL, __PSE_EVENT_CNT, pse_event_names, 0);
+ if (ret) {
+ WARN_ONCE(ret == -EMSGSIZE,
+ "calculated message payload length (%d) not sufficient\n",
+ reply_len);
+ goto err_skb;
+ }
+
+ genlmsg_end(skb, reply_payload);
+ ethnl_multicast(skb, netdev);
+ return;
+
+err_skb:
+ nlmsg_free(skb);
+}
+EXPORT_SYMBOL_GPL(ethnl_pse_send_ntf);
diff --git a/net/ethtool/strset.c b/net/ethtool/strset.c
index 6b76c05caba4..b71392fa9129 100644
--- a/net/ethtool/strset.c
+++ b/net/ethtool/strset.c
@@ -115,6 +115,11 @@ static const struct strset_info info_template[] = {
.count = __ETHTOOL_A_STATS_PHY_CNT,
.strings = stats_phy_names,
},
+ [ETH_SS_PSE_EVENTS] = {
+ .per_dev = false,
+ .count = __PSE_EVENT_CNT,
+ .strings = pse_event_names,
+ },
};
struct strset_req_info {
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 02/12] net: pse-pd: Add support for reporting events
2025-02-18 16:19 ` [PATCH net-next v5 02/12] net: pse-pd: Add support for reporting events Kory Maincent
@ 2025-02-21 0:42 ` Jakub Kicinski
2025-02-24 12:33 ` Kory Maincent
2025-02-21 8:50 ` Oleksij Rempel
1 sibling, 1 reply; 42+ messages in thread
From: Jakub Kicinski @ 2025-02-21 0:42 UTC (permalink / raw)
To: Kory Maincent
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Tue, 18 Feb 2025 17:19:06 +0100 Kory Maincent wrote:
> From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
>
> Add support for devm_pse_irq_helper() to register PSE interrupts. This aims
> to report events such as over-current or over-temperature conditions
> similarly to how the regulator API handles them but using a specific PSE
> ethtool netlink socket.
I think you should CC HWMON ML on this.
Avoid any surprises.
> diff --git a/Documentation/netlink/specs/ethtool.yaml b/Documentation/netlink/specs/ethtool.yaml
> index 655d8d10fe24..da78c5daf537 100644
> --- a/Documentation/netlink/specs/ethtool.yaml
> +++ b/Documentation/netlink/specs/ethtool.yaml
> @@ -1526,6 +1526,22 @@ attribute-sets:
> name: hwtstamp-flags
> type: nest
> nested-attributes: bitset
> + -
> + name: pse-ntf
> + attr-cnt-name: __ethtool-a-pse-ntf-cnt
> + attributes:
> + -
> + name: unspec
> + type: unused
> + value: 0
Please don't add the unused entries unless your code actually needs
them. YNL will id real ones from 1 anyway.
> + -
> + name: header
> + type: nest
> + nested-attributes: header
> + -
> + name: events
> + type: nest
> + nested-attributes: bitset
Do we really need a bitset here? Much more manual work to make a bitset
than just a uint + enum with the bits. enum is much easier to use with
YNL based user space, and it's more self-documenting than a list of bits
buried in the source of the kernel.
> operations:
> enum-model: directional
> @@ -2382,3 +2398,13 @@ operations:
> attributes: *tsconfig
> reply:
> attributes: *tsconfig
> + -
> + name: pse-ntf
> + doc: Notification for pse events.
s/pse/PSE/
> +
> + attribute-set: pse-ntf
> +
> + event:
> + attributes:
> + - header
> + - events
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 02/12] net: pse-pd: Add support for reporting events
2025-02-21 0:42 ` Jakub Kicinski
@ 2025-02-24 12:33 ` Kory Maincent
2025-02-24 21:47 ` Jakub Kicinski
0 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-24 12:33 UTC (permalink / raw)
To: Jakub Kicinski
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
Hello Jakub,
On Thu, 20 Feb 2025 16:42:01 -0800
Jakub Kicinski <kuba@kernel.org> wrote:
> On Tue, 18 Feb 2025 17:19:06 +0100 Kory Maincent wrote:
> > From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
> >
> > Add support for devm_pse_irq_helper() to register PSE interrupts. This aims
> > to report events such as over-current or over-temperature conditions
> > similarly to how the regulator API handles them but using a specific PSE
> > ethtool netlink socket.
>
> I think you should CC HWMON ML on this.
> Avoid any surprises.
You mean regulator maintainers right?
> > diff --git a/Documentation/netlink/specs/ethtool.yaml
> > b/Documentation/netlink/specs/ethtool.yaml index 655d8d10fe24..da78c5daf537
> > 100644 --- a/Documentation/netlink/specs/ethtool.yaml
> > +++ b/Documentation/netlink/specs/ethtool.yaml
> > @@ -1526,6 +1526,22 @@ attribute-sets:
> > name: hwtstamp-flags
> > type: nest
> > nested-attributes: bitset
> > + -
> > + name: pse-ntf
> > + attr-cnt-name: __ethtool-a-pse-ntf-cnt
> > + attributes:
> > + -
> > + name: unspec
> > + type: unused
> > + value: 0
>
> Please don't add the unused entries unless your code actually needs
> them. YNL will id real ones from 1 anyway.
ok.
> > + -
> > + name: header
> > + type: nest
> > + nested-attributes: header
> > + -
> > + name: events
> > + type: nest
> > + nested-attributes: bitset
>
> Do we really need a bitset here? Much more manual work to make a bitset
> than just a uint + enum with the bits. enum is much easier to use with
> YNL based user space, and it's more self-documenting than a list of bits
> buried in the source of the kernel.
Ok will change it in next version.
> > operations:
> > enum-model: directional
> > @@ -2382,3 +2398,13 @@ operations:
> > attributes: *tsconfig
> > reply:
> > attributes: *tsconfig
> > + -
> > + name: pse-ntf
> > + doc: Notification for pse events.
>
> s/pse/PSE/
Oh thanks!
Regards,
--
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 02/12] net: pse-pd: Add support for reporting events
2025-02-24 12:33 ` Kory Maincent
@ 2025-02-24 21:47 ` Jakub Kicinski
0 siblings, 0 replies; 42+ messages in thread
From: Jakub Kicinski @ 2025-02-24 21:47 UTC (permalink / raw)
To: Kory Maincent
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Mon, 24 Feb 2025 13:33:12 +0100 Kory Maincent wrote:
> > On Tue, 18 Feb 2025 17:19:06 +0100 Kory Maincent wrote:
> > > From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
> > >
> > > Add support for devm_pse_irq_helper() to register PSE interrupts. This aims
> > > to report events such as over-current or over-temperature conditions
> > > similarly to how the regulator API handles them but using a specific PSE
> > > ethtool netlink socket.
> >
> > I think you should CC HWMON ML on this.
> > Avoid any surprises.
>
> You mean regulator maintainers right?
Fair point, I'm not sure who's responsible for reporting over-current
on a regulator. My intuition would be HWMON, but no idea if it was ever
discussed. So maybe CC both lists?
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 02/12] net: pse-pd: Add support for reporting events
2025-02-18 16:19 ` [PATCH net-next v5 02/12] net: pse-pd: Add support for reporting events Kory Maincent
2025-02-21 0:42 ` Jakub Kicinski
@ 2025-02-21 8:50 ` Oleksij Rempel
2025-02-24 11:02 ` Kory Maincent
1 sibling, 1 reply; 42+ messages in thread
From: Oleksij Rempel @ 2025-02-21 8:50 UTC (permalink / raw)
To: Kory Maincent
Cc: Andrew Lunn, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
Hi Kory,
On Tue, Feb 18, 2025 at 05:19:06PM +0100, Kory Maincent wrote:
> From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
>
> Add support for devm_pse_irq_helper() to register PSE interrupts. This aims
> to report events such as over-current or over-temperature conditions
> similarly to how the regulator API handles them but using a specific PSE
> ethtool netlink socket.
Thank you for your work. Here some comments.
...
> --- a/drivers/net/mdio/fwnode_mdio.c
> +++ b/drivers/net/mdio/fwnode_mdio.c
> @@ -18,7 +18,8 @@ MODULE_LICENSE("GPL");
> MODULE_DESCRIPTION("FWNODE MDIO bus (Ethernet PHY) accessors");
>
> static struct pse_control *
> -fwnode_find_pse_control(struct fwnode_handle *fwnode)
> +fwnode_find_pse_control(struct fwnode_handle *fwnode,
> + struct phy_device *phydev)
> {
This change seems to be not directly related to the commit message.
Is it the preparation for the multi-phy support?
> struct pse_control *psec;
> struct device_node *np;
> @@ -30,7 +31,7 @@ fwnode_find_pse_control(struct fwnode_handle *fwnode)
> if (!np)
> return NULL;
>
> - psec = of_pse_control_get(np);
> + psec = of_pse_control_get(np, phydev);
> if (PTR_ERR(psec) == -ENOENT)
> return NULL;
>
> @@ -128,15 +129,9 @@ int fwnode_mdiobus_register_phy(struct mii_bus *bus,
> u32 phy_id;
> int rc;
>
> - psec = fwnode_find_pse_control(child);
> - if (IS_ERR(psec))
> - return PTR_ERR(psec);
> -
> mii_ts = fwnode_find_mii_timestamper(child);
> - if (IS_ERR(mii_ts)) {
> - rc = PTR_ERR(mii_ts);
> - goto clean_pse;
> - }
> + if (IS_ERR(mii_ts))
> + return PTR_ERR(mii_ts);
>
> is_c45 = fwnode_device_is_compatible(child, "ethernet-phy-ieee802.3-c45");
> if (is_c45 || fwnode_get_phy_id(child, &phy_id))
> @@ -169,6 +164,12 @@ int fwnode_mdiobus_register_phy(struct mii_bus *bus,
> goto clean_phy;
> }
>
> + psec = fwnode_find_pse_control(child, phy);
> + if (IS_ERR(psec)) {
> + rc = PTR_ERR(psec);
> + goto unregister_phy;
> + }
> +
> phy->psec = psec;
>
> /* phy->mii_ts may already be defined by the PHY driver. A
> @@ -180,12 +181,13 @@ int fwnode_mdiobus_register_phy(struct mii_bus *bus,
>
> return 0;
>
> +unregister_phy:
> + if (is_acpi_node(child) || is_of_node(child))
> + phy_device_remove(phy);
> clean_phy:
> phy_device_free(phy);
> clean_mii_ts:
> unregister_mii_timestamper(mii_ts);
> -clean_pse:
> - pse_control_put(psec);
>
> return rc;
> }
> diff --git a/drivers/net/pse-pd/pse_core.c b/drivers/net/pse-pd/pse_core.c
> index 4602e26eb8c8..10a5ab30afdd 100644
> --- a/drivers/net/pse-pd/pse_core.c
> +++ b/drivers/net/pse-pd/pse_core.c
> @@ -7,6 +7,7 @@
...
> +/**
> + * pse_to_regulator_notifs - Convert PSE notifications to Regulator
> + * notifications
> + * @notifs: PSE notifications
> + *
> + * Return: Regulator notifications
> + */
> +static unsigned long pse_to_regulator_notifs(unsigned long notifs)
I prefer converting it the other way around to make it reusable for
plain regulator-based PSEs. For example, the podl-pse-regulator driver
won’t have its own interrupt handler but will instead use
devm_regulator_register_notifier().
Even full-fledged PSE controllers like the PD692x0 are just one part of
a larger chain of regulators. An overcurrent event may originate from a
downstream regulator that is not part of the PD692x0 itself. In this
case, we need to process the event from the downstream regulator,
convert it into an ethtool event, and forward it to the user.
Here is one example how devm_regulator_register_notifier() can be used:
https://lore.kernel.org/all/20250220074429.2906141-1-o.rempel@pengutronix.de/
> +{
> + unsigned long rnotifs = 0;
> +
> + if (notifs & ETHTOOL_PSE_EVENT_OVER_CURRENT)
> + rnotifs |= REGULATOR_EVENT_OVER_CURRENT;
> + if (notifs & ETHTOOL_PSE_EVENT_OVER_TEMP)
> + rnotifs |= REGULATOR_EVENT_OVER_TEMP;
> +
> + return rnotifs;
> +}
> +
Other parts look ok for me.
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 02/12] net: pse-pd: Add support for reporting events
2025-02-21 8:50 ` Oleksij Rempel
@ 2025-02-24 11:02 ` Kory Maincent
2025-02-24 18:19 ` Kory Maincent
0 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-24 11:02 UTC (permalink / raw)
To: Oleksij Rempel
Cc: Andrew Lunn, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
Hello Oleksij,
On Fri, 21 Feb 2025 09:50:33 +0100
Oleksij Rempel <o.rempel@pengutronix.de> wrote:
> Hi Kory,
>
> On Tue, Feb 18, 2025 at 05:19:06PM +0100, Kory Maincent wrote:
> > From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
> >
> > Add support for devm_pse_irq_helper() to register PSE interrupts. This aims
> > to report events such as over-current or over-temperature conditions
> > similarly to how the regulator API handles them but using a specific PSE
> > ethtool netlink socket.
>
> Thank you for your work. Here some comments.
>
> ...
>
> > --- a/drivers/net/mdio/fwnode_mdio.c
> > +++ b/drivers/net/mdio/fwnode_mdio.c
> > @@ -18,7 +18,8 @@ MODULE_LICENSE("GPL");
> > MODULE_DESCRIPTION("FWNODE MDIO bus (Ethernet PHY) accessors");
> >
> > static struct pse_control *
> > -fwnode_find_pse_control(struct fwnode_handle *fwnode)
> > +fwnode_find_pse_control(struct fwnode_handle *fwnode,
> > + struct phy_device *phydev)
> > {
>
> This change seems to be not directly related to the commit message.
> Is it the preparation for the multi-phy support?
I need to save the phy_device related to PSE control to use the right network
interface for the ethtool notification. (ethnl_pse_send_ntf())
Indeed I have not described this in the commit message.
> ...
>
> > +/**
> > + * pse_to_regulator_notifs - Convert PSE notifications to Regulator
> > + * notifications
> > + * @notifs: PSE notifications
> > + *
> > + * Return: Regulator notifications
> > + */
> > +static unsigned long pse_to_regulator_notifs(unsigned long notifs)
>
> I prefer converting it the other way around to make it reusable for
> plain regulator-based PSEs. For example, the podl-pse-regulator driver
> won’t have its own interrupt handler but will instead use
> devm_regulator_register_notifier().
The driver PIs part send PSE notifications which will be converted to regulator
events from the core. It is posting events.
If you use devm_regulator_register_notifier() you will registers a listener for
the regulator events. It is two distinct things.
> Even full-fledged PSE controllers like the PD692x0 are just one part of
> a larger chain of regulators. An overcurrent event may originate from a
> downstream regulator that is not part of the PD692x0 itself. In this
> case, we need to process the event from the downstream regulator,
> convert it into an ethtool event, and forward it to the user.
If you want to do something in case of downstream regulator events you will deal
with regulator events not PSE events. I think you want to disable PIs in case of
event like downstream regulator over current.
What policy should we use? Should we disable all the PIs or only disabled the
low priority like the budget evaluation strategy of this series? As it is over
current event not related to budget we don't know how many PIs we should
disable.
Still as said before it is a distinct development that could be tackled later.
Regards,
--
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com
^ permalink raw reply [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 02/12] net: pse-pd: Add support for reporting events
2025-02-24 11:02 ` Kory Maincent
@ 2025-02-24 18:19 ` Kory Maincent
0 siblings, 0 replies; 42+ messages in thread
From: Kory Maincent @ 2025-02-24 18:19 UTC (permalink / raw)
To: Oleksij Rempel
Cc: Andrew Lunn, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Mon, 24 Feb 2025 12:02:28 +0100
Kory Maincent <kory.maincent@bootlin.com> wrote:
> Hello Oleksij,
>
> On Fri, 21 Feb 2025 09:50:33 +0100
> Oleksij Rempel <o.rempel@pengutronix.de> wrote:
>
> > Hi Kory,
> >
> > On Tue, Feb 18, 2025 at 05:19:06PM +0100, Kory Maincent wrote:
> > > From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
> > >
> > > Add support for devm_pse_irq_helper() to register PSE interrupts. This
> > > aims to report events such as over-current or over-temperature conditions
> > > similarly to how the regulator API handles them but using a specific PSE
> > > ethtool netlink socket.
> >
> > Thank you for your work. Here some comments.
> >
> > ...
> >
> > > --- a/drivers/net/mdio/fwnode_mdio.c
> > > +++ b/drivers/net/mdio/fwnode_mdio.c
> > > @@ -18,7 +18,8 @@ MODULE_LICENSE("GPL");
> > > MODULE_DESCRIPTION("FWNODE MDIO bus (Ethernet PHY) accessors");
> > >
> > > static struct pse_control *
> > > -fwnode_find_pse_control(struct fwnode_handle *fwnode)
> > > +fwnode_find_pse_control(struct fwnode_handle *fwnode,
> > > + struct phy_device *phydev)
> > > {
> >
> > This change seems to be not directly related to the commit message.
> > Is it the preparation for the multi-phy support?
>
> I need to save the phy_device related to PSE control to use the right network
> interface for the ethtool notification. (ethnl_pse_send_ntf())
> Indeed I have not described this in the commit message.
In fact, there is another solution. We can go over all the PHYs and look for
the one that matches the psec pointer.
Mmh it is maybe better, it will avoid saving the phy_device pointer into
the newly attached_phydev.
I will go for it in v6.
Regards,
--
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com
^ permalink raw reply [flat|nested] 42+ messages in thread
* [PATCH net-next v5 03/12] net: pse-pd: tps23881: Add support for PSE events and interrupts
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
2025-02-18 16:19 ` [PATCH net-next v5 01/12] net: ethtool: Add support for ethnl_info_init_ntf helper function Kory Maincent
2025-02-18 16:19 ` [PATCH net-next v5 02/12] net: pse-pd: Add support for reporting events Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-18 16:19 ` [PATCH net-next v5 04/12] net: pse-pd: Add support for PSE power domains Kory Maincent
` (8 subsequent siblings)
11 siblings, 0 replies; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
Add support for PSE event reporting through interrupts. Set up the newly
introduced devm_pse_irq_helper helper to register the interrupt. Events are
reported for over-current and over-temperature conditions.
Reviewed-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
Change in v4:
- Small rename of a function.
Change in v3:
- Loop over interruption register to be sure the interruption pin is
freed before exiting the interrupt handler function.
- Add exist variable to not report event for undescribed PIs.
- Used helpers to convert the chan number to the PI port number.
Change in v2:
- Remove support for OSS pin and TPC23881 specific port priority management
---
drivers/net/pse-pd/tps23881.c | 178 +++++++++++++++++++++++++++++++++++++++++-
1 file changed, 177 insertions(+), 1 deletion(-)
diff --git a/drivers/net/pse-pd/tps23881.c b/drivers/net/pse-pd/tps23881.c
index 5e9dda2c0eac..122666719297 100644
--- a/drivers/net/pse-pd/tps23881.c
+++ b/drivers/net/pse-pd/tps23881.c
@@ -17,6 +17,13 @@
#define TPS23881_MAX_CHANS 8
+#define TPS23881_REG_IT 0x0
+#define TPS23881_REG_IT_MASK 0x1
+#define TPS23881_REG_IT_IFAULT BIT(5)
+#define TPS23881_REG_IT_SUPF BIT(7)
+#define TPS23881_REG_FAULT 0x7
+#define TPS23881_REG_SUPF_EVENT 0xb
+#define TPS23881_REG_TSD BIT(7)
#define TPS23881_REG_PW_STATUS 0x10
#define TPS23881_REG_OP_MODE 0x12
#define TPS23881_OP_MODE_SEMIAUTO 0xaaaa
@@ -24,6 +31,7 @@
#define TPS23881_REG_DET_CLA_EN 0x14
#define TPS23881_REG_GEN_MASK 0x17
#define TPS23881_REG_NBITACC BIT(5)
+#define TPS23881_REG_INTEN BIT(7)
#define TPS23881_REG_PW_EN 0x19
#define TPS23881_REG_2PAIR_POL1 0x1e
#define TPS23881_REG_PORT_MAP 0x26
@@ -51,6 +59,7 @@ struct tps23881_port_desc {
u8 chan[2];
bool is_4p;
int pw_pol;
+ bool exist;
};
struct tps23881_priv {
@@ -782,8 +791,10 @@ tps23881_write_port_matrix(struct tps23881_priv *priv,
hw_chan = port_matrix[i].hw_chan[0] % 4;
/* Set software port matrix for existing ports */
- if (port_matrix[i].exist)
+ if (port_matrix[i].exist) {
priv->port[pi_id].chan[0] = lgcl_chan;
+ priv->port[pi_id].exist = true;
+ }
/* Initialize power policy internal value */
priv->port[pi_id].pw_pol = -1;
@@ -1017,6 +1028,165 @@ static int tps23881_flash_sram_fw(struct i2c_client *client)
return 0;
}
+/* Convert interrupt events to 0xff to be aligned with the chan
+ * number.
+ */
+static u8 tps23881_irq_export_chans_helper(u16 reg_val, u8 field_offset)
+{
+ u8 val;
+
+ val = (reg_val >> (4 + field_offset) & 0xf0) |
+ (reg_val >> field_offset & 0x0f);
+
+ return val;
+}
+
+/* Convert chan number to port number */
+static void tps23881_set_notifs_helper(struct tps23881_priv *priv,
+ u8 chans,
+ unsigned long *notifs,
+ unsigned long *notifs_mask,
+ enum ethtool_pse_events event)
+{
+ u8 chan;
+ int i;
+
+ if (!chans)
+ return;
+
+ for (i = 0; i < TPS23881_MAX_CHANS; i++) {
+ if (!priv->port[i].exist)
+ continue;
+ /* No need to look at the 2nd channel in case of PoE4 as
+ * both registers are set.
+ */
+ chan = priv->port[i].chan[0];
+
+ if (BIT(chan) & chans) {
+ *notifs_mask |= BIT(i);
+ notifs[i] |= event;
+ }
+ }
+}
+
+static void tps23881_irq_event_over_temp(struct tps23881_priv *priv,
+ u16 reg_val,
+ unsigned long *notifs,
+ unsigned long *notifs_mask)
+{
+ int i;
+
+ if (reg_val & TPS23881_REG_TSD) {
+ for (i = 0; i < TPS23881_MAX_CHANS; i++) {
+ if (!priv->port[i].exist)
+ continue;
+
+ *notifs_mask |= BIT(i);
+ notifs[i] |= ETHTOOL_PSE_EVENT_OVER_TEMP;
+ }
+ }
+}
+
+static void tps23881_irq_event_over_current(struct tps23881_priv *priv,
+ u16 reg_val,
+ unsigned long *notifs,
+ unsigned long *notifs_mask)
+{
+ u8 chans;
+
+ chans = tps23881_irq_export_chans_helper(reg_val, 0);
+ if (chans)
+ tps23881_set_notifs_helper(priv, chans, notifs, notifs_mask,
+ ETHTOOL_PSE_EVENT_OVER_CURRENT);
+}
+
+static int tps23881_irq_event_handler(struct tps23881_priv *priv, u16 reg,
+ unsigned long *notifs,
+ unsigned long *notifs_mask)
+{
+ struct i2c_client *client = priv->client;
+ int ret;
+
+ /* The Supply event bit is repeated twice so we only need to read
+ * the one from the first byte.
+ */
+ if (reg & TPS23881_REG_IT_SUPF) {
+ ret = i2c_smbus_read_word_data(client, TPS23881_REG_SUPF_EVENT);
+ if (ret < 0)
+ return ret;
+ tps23881_irq_event_over_temp(priv, ret, notifs, notifs_mask);
+ }
+
+ if (reg & (TPS23881_REG_IT_IFAULT | TPS23881_REG_IT_IFAULT << 8)) {
+ ret = i2c_smbus_read_word_data(client, TPS23881_REG_FAULT);
+ if (ret < 0)
+ return ret;
+ tps23881_irq_event_over_current(priv, ret, notifs, notifs_mask);
+ }
+
+ return 0;
+}
+
+static int tps23881_irq_handler(int irq, struct pse_controller_dev *pcdev,
+ unsigned long *notifs,
+ unsigned long *notifs_mask)
+{
+ struct tps23881_priv *priv = to_tps23881_priv(pcdev);
+ struct i2c_client *client = priv->client;
+ int ret, it_mask;
+
+ /* Get interruption mask */
+ ret = i2c_smbus_read_word_data(client, TPS23881_REG_IT_MASK);
+ if (ret < 0)
+ return ret;
+ it_mask = ret;
+
+ /* Read interrupt register until it frees the interruption pin. */
+ while (true) {
+ ret = i2c_smbus_read_word_data(client, TPS23881_REG_IT);
+ if (ret < 0)
+ return ret;
+
+ /* No more relevant interruption */
+ if (!(ret & it_mask))
+ return 0;
+
+ ret = tps23881_irq_event_handler(priv, (u16)ret, notifs,
+ notifs_mask);
+ if (ret)
+ return ret;
+ }
+ return 0;
+}
+
+static int tps23881_setup_irq(struct tps23881_priv *priv, int irq)
+{
+ struct i2c_client *client = priv->client;
+ struct pse_irq_desc irq_desc = {
+ .name = "tps23881-irq",
+ .map_event = tps23881_irq_handler,
+ };
+ int ret;
+ u16 val;
+
+ val = TPS23881_REG_IT_IFAULT | TPS23881_REG_IT_SUPF;
+ val |= val << 8;
+ ret = i2c_smbus_write_word_data(client, TPS23881_REG_IT_MASK, val);
+ if (ret)
+ return ret;
+
+ ret = i2c_smbus_read_word_data(client, TPS23881_REG_GEN_MASK);
+ if (ret < 0)
+ return ret;
+
+ val = (u16)(ret | TPS23881_REG_INTEN | TPS23881_REG_INTEN << 8);
+ ret = i2c_smbus_write_word_data(client, TPS23881_REG_GEN_MASK, val);
+ if (ret < 0)
+ return ret;
+
+ return devm_pse_irq_helper(&priv->pcdev, irq, 0, &irq_desc);
+}
+
static int tps23881_i2c_probe(struct i2c_client *client)
{
struct device *dev = &client->dev;
@@ -1097,6 +1267,12 @@ static int tps23881_i2c_probe(struct i2c_client *client)
"failed to register PSE controller\n");
}
+ if (client->irq) {
+ ret = tps23881_setup_irq(priv, client->irq);
+ if (ret)
+ return ret;
+ }
+
return ret;
}
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* [PATCH net-next v5 04/12] net: pse-pd: Add support for PSE power domains
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
` (2 preceding siblings ...)
2025-02-18 16:19 ` [PATCH net-next v5 03/12] net: pse-pd: tps23881: Add support for PSE events and interrupts Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-21 0:43 ` Jakub Kicinski
2025-02-18 16:19 ` [PATCH net-next v5 05/12] net: ethtool: Add support for new power domains index description Kory Maincent
` (7 subsequent siblings)
11 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
Introduce PSE power domain support as groundwork for upcoming port
priority features. Multiple PSE PIs can now be grouped under a single
PSE power domain, enabling future enhancements like defining available
power budgets, port priority modes, and disconnection policies. This
setup will allow the system to assess whether activating a port would
exceed the available power budget, preventing over-budget states
proactively.
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
Changes in v4:
- Add kdoc.
- Fix null dereference in pse_flush_pw_ds function.
Changes in v3:
- Remove pw_budget variable.
Changes in v2:
- new patch.
---
drivers/net/pse-pd/pse_core.c | 114 ++++++++++++++++++++++++++++++++++++++++++
include/linux/pse-pd/pse.h | 2 +
2 files changed, 116 insertions(+)
diff --git a/drivers/net/pse-pd/pse_core.c b/drivers/net/pse-pd/pse_core.c
index 10a5ab30afdd..a654f6eef6ff 100644
--- a/drivers/net/pse-pd/pse_core.c
+++ b/drivers/net/pse-pd/pse_core.c
@@ -15,6 +15,7 @@
static DEFINE_MUTEX(pse_list_mutex);
static LIST_HEAD(pse_controller_list);
+static DEFINE_XARRAY_ALLOC(pse_pw_d_map);
/**
* struct pse_control - a PSE control
@@ -35,6 +36,16 @@ struct pse_control {
struct phy_device *attached_phydev;
};
+/**
+ * struct pse_power_domain - a PSE power domain
+ * @id: ID of the power domain
+ * @supply: Power supply the Power Domain
+ */
+struct pse_power_domain {
+ int id;
+ struct regulator *supply;
+};
+
static int of_load_single_pse_pi_pairset(struct device_node *node,
struct pse_pi *pi,
int pairset_num)
@@ -440,6 +451,103 @@ devm_pse_pi_regulator_register(struct pse_controller_dev *pcdev,
return 0;
}
+/**
+ * pse_flush_pw_ds - flush all PSE power domains of a PSE
+ * @pcdev: a pointer to the initialized PSE controller device
+ */
+static void pse_flush_pw_ds(struct pse_controller_dev *pcdev)
+{
+ struct pse_power_domain *pw_d;
+ int i;
+
+ for (i = 0; i < pcdev->nr_lines; i++) {
+ if (!pcdev->pi[i].pw_d)
+ continue;
+
+ pw_d = xa_load(&pse_pw_d_map, pcdev->pi[i].pw_d->id);
+ if (pw_d) {
+ regulator_put(pw_d->supply);
+ xa_erase(&pse_pw_d_map, pw_d->id);
+ }
+ }
+}
+
+/**
+ * devm_pse_alloc_pw_d - allocate a new PSE power domain for a device
+ * @dev: device that is registering this PSE power domain
+ *
+ * Return: Pointer to the newly allocated PSE power domain or error pointers
+ */
+static struct pse_power_domain *devm_pse_alloc_pw_d(struct device *dev)
+{
+ struct pse_power_domain *pw_d;
+ int index, ret;
+
+ pw_d = devm_kzalloc(dev, sizeof(*pw_d), GFP_KERNEL);
+ if (!pw_d)
+ return ERR_PTR(-ENOMEM);
+
+ ret = xa_alloc(&pse_pw_d_map, &index, pw_d, XA_LIMIT(1, INT_MAX), GFP_KERNEL);
+ if (ret)
+ return ERR_PTR(ret);
+
+ pw_d->id = index;
+ return pw_d;
+}
+
+/**
+ * pse_register_pw_ds - register the PSE power domains for a PSE
+ * @pcdev: a pointer to the PSE controller device
+ *
+ * Return: 0 on success and failure value on error
+ */
+static int pse_register_pw_ds(struct pse_controller_dev *pcdev)
+{
+ int i;
+
+ for (i = 0; i < pcdev->nr_lines; i++) {
+ struct regulator_dev *rdev = pcdev->pi[i].rdev;
+ struct pse_power_domain *pw_d;
+ struct regulator *supply;
+ bool present = false;
+ unsigned long index;
+
+ /* No regulator or regulator parent supply registered.
+ * We need a regulator parent to register a PSE power domain
+ */
+ if (!rdev || !rdev->supply)
+ continue;
+
+ xa_for_each(&pse_pw_d_map, index, pw_d) {
+ /* Power supply already registered as a PSE power
+ * domain.
+ */
+ if (regulator_is_equal(pw_d->supply, rdev->supply)) {
+ present = true;
+ pcdev->pi[i].pw_d = pw_d;
+ break;
+ }
+ }
+ if (present)
+ continue;
+
+ pw_d = devm_pse_alloc_pw_d(pcdev->dev);
+ if (IS_ERR_OR_NULL(pw_d))
+ return PTR_ERR(pw_d);
+
+ supply = regulator_get(&rdev->dev, rdev->supply_name);
+ if (IS_ERR(supply)) {
+ xa_erase(&pse_pw_d_map, pw_d->id);
+ return PTR_ERR(supply);
+ }
+
+ pw_d->supply = supply;
+ pcdev->pi[i].pw_d = pw_d;
+ }
+
+ return 0;
+}
+
/**
* pse_controller_register - register a PSE controller device
* @pcdev: a pointer to the initialized PSE controller device
@@ -499,6 +607,11 @@ int pse_controller_register(struct pse_controller_dev *pcdev)
return ret;
}
+ ret = pse_register_pw_ds(pcdev);
+
+ if (ret)
+ return ret;
+
mutex_lock(&pse_list_mutex);
list_add(&pcdev->list, &pse_controller_list);
mutex_unlock(&pse_list_mutex);
@@ -513,6 +626,7 @@ EXPORT_SYMBOL_GPL(pse_controller_register);
*/
void pse_controller_unregister(struct pse_controller_dev *pcdev)
{
+ pse_flush_pw_ds(pcdev);
pse_release_pis(pcdev);
mutex_lock(&pse_list_mutex);
list_del(&pcdev->list);
diff --git a/include/linux/pse-pd/pse.h b/include/linux/pse-pd/pse.h
index 5d41a1c984bd..5201a0fb3d74 100644
--- a/include/linux/pse-pd/pse.h
+++ b/include/linux/pse-pd/pse.h
@@ -220,12 +220,14 @@ struct pse_pi_pairset {
* @np: device node pointer of the PSE PI node
* @rdev: regulator represented by the PSE PI
* @admin_state_enabled: PI enabled state
+ * @pw_d: Power domain of the PSE PI
*/
struct pse_pi {
struct pse_pi_pairset pairset[2];
struct device_node *np;
struct regulator_dev *rdev;
bool admin_state_enabled;
+ struct pse_power_domain *pw_d;
};
/**
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 04/12] net: pse-pd: Add support for PSE power domains
2025-02-18 16:19 ` [PATCH net-next v5 04/12] net: pse-pd: Add support for PSE power domains Kory Maincent
@ 2025-02-21 0:43 ` Jakub Kicinski
0 siblings, 0 replies; 42+ messages in thread
From: Jakub Kicinski @ 2025-02-21 0:43 UTC (permalink / raw)
To: Kory Maincent
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Tue, 18 Feb 2025 17:19:08 +0100 Kory Maincent wrote:
> + ret = pse_register_pw_ds(pcdev);
> +
> + if (ret)
> + return ret;
nit: unnecessary empty line
^ permalink raw reply [flat|nested] 42+ messages in thread
* [PATCH net-next v5 05/12] net: ethtool: Add support for new power domains index description
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
` (3 preceding siblings ...)
2025-02-18 16:19 ` [PATCH net-next v5 04/12] net: pse-pd: Add support for PSE power domains Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-18 16:19 ` [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies Kory Maincent
` (6 subsequent siblings)
11 siblings, 0 replies; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
Report the index of the newly introduced PSE power domain to the user,
enabling improved management of the power budget for PSE devices.
Reviewed-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
Changes in v3:
- Do not support power domain id = 0 because we can't differentiate with
no PSE power domain.
Changes in v2:
- new patch.
---
Documentation/netlink/specs/ethtool.yaml | 5 +++++
Documentation/networking/ethtool-netlink.rst | 4 ++++
drivers/net/pse-pd/pse_core.c | 3 +++
include/linux/pse-pd/pse.h | 2 ++
include/uapi/linux/ethtool_netlink_generated.h | 1 +
net/ethtool/pse-pd.c | 7 +++++++
6 files changed, 22 insertions(+)
diff --git a/Documentation/netlink/specs/ethtool.yaml b/Documentation/netlink/specs/ethtool.yaml
index da78c5daf537..9b171c2dd1a4 100644
--- a/Documentation/netlink/specs/ethtool.yaml
+++ b/Documentation/netlink/specs/ethtool.yaml
@@ -1366,6 +1366,10 @@ attribute-sets:
type: nest
multi-attr: true
nested-attributes: c33-pse-pw-limit
+ -
+ name: pse-pw-d-id
+ type: u32
+ name-prefix: ethtool-a-
-
name: rss
attr-cnt-name: __ethtool-a-rss-cnt
@@ -2190,6 +2194,7 @@ operations:
- c33-pse-ext-substate
- c33-pse-avail-pw-limit
- c33-pse-pw-limit-ranges
+ - pse-pw-d-id
dump: *pse-get-op
-
name: pse-set
diff --git a/Documentation/networking/ethtool-netlink.rst b/Documentation/networking/ethtool-netlink.rst
index 9fc5e29b3928..dc3f6afc55a4 100644
--- a/Documentation/networking/ethtool-netlink.rst
+++ b/Documentation/networking/ethtool-netlink.rst
@@ -1789,6 +1789,7 @@ Kernel response contents:
limit of the PoE PSE.
``ETHTOOL_A_C33_PSE_PW_LIMIT_RANGES`` nested Supported power limit
configuration ranges.
+ ``ETHTOOL_A_PSE_PW_D_ID`` u32 Index of the PSE power domain
========================================== ====== =============================
When set, the optional ``ETHTOOL_A_PODL_PSE_ADMIN_STATE`` attribute identifies
@@ -1862,6 +1863,9 @@ identifies the C33 PSE power limit ranges through
If the controller works with fixed classes, the min and max values will be
equal.
+The ``ETHTOOL_A_PSE_PW_D_ID`` attribute identifies the index of PSE power
+domain.
+
PSE_SET
=======
diff --git a/drivers/net/pse-pd/pse_core.c b/drivers/net/pse-pd/pse_core.c
index a654f6eef6ff..d42a8fc7e76a 100644
--- a/drivers/net/pse-pd/pse_core.c
+++ b/drivers/net/pse-pd/pse_core.c
@@ -1039,6 +1039,9 @@ int pse_ethtool_get_status(struct pse_control *psec,
pcdev = psec->pcdev;
ops = pcdev->ops;
mutex_lock(&pcdev->lock);
+ if (pcdev->pi[psec->id].pw_d)
+ status->pw_d_id = pcdev->pi[psec->id].pw_d->id;
+
ret = ops->pi_get_admin_state(pcdev, psec->id, &admin_state);
if (ret)
goto out;
diff --git a/include/linux/pse-pd/pse.h b/include/linux/pse-pd/pse.h
index 5201a0fb3d74..ffa6cf9a0072 100644
--- a/include/linux/pse-pd/pse.h
+++ b/include/linux/pse-pd/pse.h
@@ -112,6 +112,7 @@ struct pse_pw_limit_ranges {
/**
* struct ethtool_pse_control_status - PSE control/channel status.
*
+ * @pw_d_id: PSE power domain index.
* @podl_admin_state: operational state of the PoDL PSE
* functions. IEEE 802.3-2018 30.15.1.1.2 aPoDLPSEAdminState
* @podl_pw_status: power detection status of the PoDL PSE.
@@ -133,6 +134,7 @@ struct pse_pw_limit_ranges {
* ranges
*/
struct ethtool_pse_control_status {
+ u32 pw_d_id;
enum ethtool_podl_pse_admin_state podl_admin_state;
enum ethtool_podl_pse_pw_d_status podl_pw_status;
enum ethtool_c33_pse_admin_state c33_admin_state;
diff --git a/include/uapi/linux/ethtool_netlink_generated.h b/include/uapi/linux/ethtool_netlink_generated.h
index f03b51766311..919435c1a924 100644
--- a/include/uapi/linux/ethtool_netlink_generated.h
+++ b/include/uapi/linux/ethtool_netlink_generated.h
@@ -633,6 +633,7 @@ enum {
ETHTOOL_A_C33_PSE_EXT_SUBSTATE,
ETHTOOL_A_C33_PSE_AVAIL_PW_LIMIT,
ETHTOOL_A_C33_PSE_PW_LIMIT_RANGES,
+ ETHTOOL_A_PSE_PW_D_ID,
__ETHTOOL_A_PSE_CNT,
ETHTOOL_A_PSE_MAX = (__ETHTOOL_A_PSE_CNT - 1)
diff --git a/net/ethtool/pse-pd.c b/net/ethtool/pse-pd.c
index e471e577d4b6..eae5b7894613 100644
--- a/net/ethtool/pse-pd.c
+++ b/net/ethtool/pse-pd.c
@@ -84,6 +84,8 @@ static int pse_reply_size(const struct ethnl_req_info *req_base,
const struct ethtool_pse_control_status *st = &data->status;
int len = 0;
+ if (st->pw_d_id > 0)
+ len += nla_total_size(sizeof(u32)); /* _PSE_PW_D_ID */
if (st->podl_admin_state > 0)
len += nla_total_size(sizeof(u32)); /* _PODL_PSE_ADMIN_STATE */
if (st->podl_pw_status > 0)
@@ -149,6 +151,11 @@ static int pse_fill_reply(struct sk_buff *skb,
const struct pse_reply_data *data = PSE_REPDATA(reply_base);
const struct ethtool_pse_control_status *st = &data->status;
+ if (st->pw_d_id > 0 &&
+ nla_put_u32(skb, ETHTOOL_A_PSE_PW_D_ID,
+ st->pw_d_id))
+ return -EMSGSIZE;
+
if (st->podl_admin_state > 0 &&
nla_put_u32(skb, ETHTOOL_A_PODL_PSE_ADMIN_STATE,
st->podl_admin_state))
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
` (4 preceding siblings ...)
2025-02-18 16:19 ` [PATCH net-next v5 05/12] net: ethtool: Add support for new power domains index description Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-21 0:51 ` Jakub Kicinski
2025-02-18 16:19 ` [PATCH net-next v5 07/12] net: ethtool: Add PSE new budget evaluation strategy support feature Kory Maincent
` (5 subsequent siblings)
11 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
This patch introduces the ability to configure the PSE PI budget evaluation
strategies. Budget evaluation strategies is utilized by PSE controllers to
determine which ports to turn off first in scenarios such as power budget
exceedance.
The pis_prio_max value is used to define the maximum priority level
supported by the controller. Both the current priority and the maximum
priority are exposed to the user through the pse_ethtool_get_status call.
This patch add support for two mode of budget evaluation strategies.
1. Static Method:
This method involves distributing power based on PD classification.
It’s straightforward and stable, the PSE core keeping track of the
budget and subtracting the power requested by each PD’s class.
Advantages: Every PD gets its promised power at any time, which
guarantees reliability.
Disadvantages: PD classification steps are large, meaning devices
request much more power than they actually need. As a result, the power
supply may only operate at, say, 50% capacity, which is inefficient and
wastes money.
Priority max value is matching the number of PSE PIs within the PSE.
2. Dynamic Method:
To address the inefficiencies of the static method, vendors like
Microchip have introduced dynamic power budgeting, as seen in the
PD692x0 firmware. This method monitors the current consumption per port
and subtracts it from the available power budget. When the budget is
exceeded, lower-priority ports are shut down.
Advantages: This method optimizes resource utilization, saving costs.
Disadvantages: Low-priority devices may experience instability.
Priority max value is set by the PSE controller driver.
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
Change in v5:
- Save PI previous power allocated in set current limit to be able to
restore the power allocated in case of error.
Change in v4:
- Remove disconnection policy features.
- Rename port priority to budget evaluation strategy.
- Add kdoc
Change in v3:
- Add disconnection policy.
- Add management of disabled port priority in the interrupt handler.
- Move port prio mode in the power domain instead of the PSE.
Change in v2:
- Rethink the port priority support.
---
drivers/net/pse-pd/pse_core.c | 625 ++++++++++++++++++++++++++++++++++++++----
include/linux/pse-pd/pse.h | 46 ++++
include/uapi/linux/ethtool.h | 39 ++-
net/ethtool/common.c | 6 +
4 files changed, 669 insertions(+), 47 deletions(-)
diff --git a/drivers/net/pse-pd/pse_core.c b/drivers/net/pse-pd/pse_core.c
index d42a8fc7e76a..82a222f54abd 100644
--- a/drivers/net/pse-pd/pse_core.c
+++ b/drivers/net/pse-pd/pse_core.c
@@ -40,10 +40,13 @@ struct pse_control {
* struct pse_power_domain - a PSE power domain
* @id: ID of the power domain
* @supply: Power supply the Power Domain
+ * @budget_eval_strategy: Current power budget evaluation strategy of the
+ * power domain
*/
struct pse_power_domain {
int id;
struct regulator *supply;
+ u32 budget_eval_strategy;
};
static int of_load_single_pse_pi_pairset(struct device_node *node,
@@ -222,6 +225,29 @@ static int of_load_pse_pis(struct pse_controller_dev *pcdev)
return ret;
}
+/**
+ * pse_pw_d_is_sw_pw_control - Is power control software managed
+ * @pcdev: a pointer to the PSE controller device
+ * @pw_d: a pointer to the PSE power domain
+ *
+ * Return: true if the power control of the power domain is managed from
+ * the software in the interrupt handler
+ */
+static bool pse_pw_d_is_sw_pw_control(struct pse_controller_dev *pcdev,
+ struct pse_power_domain *pw_d)
+{
+ if (!pw_d)
+ return false;
+
+ if (pw_d->budget_eval_strategy == ETHTOOL_PSE_BUDGET_EVAL_STRAT_STATIC)
+ return true;
+ if (pw_d->budget_eval_strategy == ETHTOOL_PSE_BUDGET_EVAL_STRAT_DISABLED &&
+ pcdev->ops->pi_enable && pcdev->irq)
+ return true;
+
+ return false;
+}
+
static int pse_pi_is_enabled(struct regulator_dev *rdev)
{
struct pse_controller_dev *pcdev = rdev_get_drvdata(rdev);
@@ -235,6 +261,11 @@ static int pse_pi_is_enabled(struct regulator_dev *rdev)
id = rdev_get_id(rdev);
mutex_lock(&pcdev->lock);
+ if (pse_pw_d_is_sw_pw_control(pcdev, pcdev->pi[id].pw_d)) {
+ ret = pcdev->pi[id].admin_state_enabled;
+ goto out;
+ }
+
ret = ops->pi_get_admin_state(pcdev, id, &admin_state);
if (ret)
goto out;
@@ -249,11 +280,260 @@ static int pse_pi_is_enabled(struct regulator_dev *rdev)
return ret;
}
+/**
+ * pse_pi_deallocate_pw_budget - Deallocate power budget of the PI
+ * @pi: a pointer to the PSE PI
+ */
+static void pse_pi_deallocate_pw_budget(struct pse_pi *pi)
+{
+ if (!pi->pw_d || !pi->pw_allocated_mW)
+ return;
+
+ regulator_free_power_budget(pi->pw_d->supply, pi->pw_allocated_mW);
+ pi->pw_allocated_mW = 0;
+}
+
+/**
+ * _pse_pi_disable - Call disable operation. Assumes the PSE lock has been
+ * acquired.
+ * @pcdev: a pointer to the PSE
+ * @id: index of the PSE control
+ *
+ * Return: 0 on success and failure value on error
+ */
+static int _pse_pi_disable(struct pse_controller_dev *pcdev, int id)
+{
+ const struct pse_controller_ops *ops = pcdev->ops;
+ int ret;
+
+ if (!ops->pi_disable)
+ return -EOPNOTSUPP;
+
+ ret = ops->pi_disable(pcdev, id);
+ if (ret)
+ return ret;
+
+ pse_pi_deallocate_pw_budget(&pcdev->pi[id]);
+
+ return 0;
+}
+
+/**
+ * pse_control_find_phy_by_id - Find PHY attached to the a pse control id
+ * @pcdev: a pointer to the PSE
+ * @id: index of the PSE control
+ *
+ * Return: PHY device pointer or NULL
+ */
+static struct phy_device *
+pse_control_find_phy_by_id(struct pse_controller_dev *pcdev, int id)
+{
+ struct pse_control *psec;
+
+ mutex_lock(&pse_list_mutex);
+ list_for_each_entry(psec, &pcdev->pse_control_head, list) {
+ if (psec->id == id) {
+ mutex_unlock(&pse_list_mutex);
+ return psec->attached_phydev;
+ }
+ }
+ mutex_unlock(&pse_list_mutex);
+ return NULL;
+}
+
+/**
+ * pse_disable_pi_pol - Disable a PI on a power budget policy
+ * @pcdev: a pointer to the PSE
+ * @id: index of the PSE PI
+ *
+ * Return: 0 on success and failure value on error
+ */
+static int pse_disable_pi_pol(struct pse_controller_dev *pcdev, int id)
+{
+ unsigned long notifs = ETHTOOL_PSE_EVENT_OVER_BUDGET;
+ struct netlink_ext_ack extack = {};
+ struct phy_device *phydev;
+ int ret;
+
+ dev_dbg(pcdev->dev, "Disabling PI %d to free power budget\n", id);
+
+ NL_SET_ERR_MSG_FMT(&extack,
+ "Disabling PI %d to free power budget", id);
+
+ ret = _pse_pi_disable(pcdev, id);
+ if (ret) {
+ notifs |= ETHTOOL_PSE_EVENT_SW_PW_CONTROL_ERROR;
+ } else {
+ pcdev->pi[id].admin_state_enabled = 0;
+ pcdev->pi[id]._isr_counter_mismatch = 1;
+ }
+
+ phydev = pse_control_find_phy_by_id(pcdev, id);
+ if (phydev)
+ ethnl_pse_send_ntf(phydev, notifs, &extack);
+
+ return ret;
+}
+
+/**
+ * pse_disable_pi_prio - Disable all PIs of a given priority inside a PSE
+ * power domain
+ * @pcdev: a pointer to the PSE
+ * @pw_d: a pointer to the PSE power domain
+ * @prio: priority
+ *
+ * Return: 0 on success and failure value on error
+ */
+static int pse_disable_pi_prio(struct pse_controller_dev *pcdev,
+ struct pse_power_domain *pw_d,
+ int prio)
+{
+ int i;
+
+ for (i = 0; i < pcdev->nr_lines; i++) {
+ int ret;
+
+ if (pcdev->pi[i].prio != prio ||
+ pcdev->pi[i].pw_d != pw_d ||
+ !pcdev->pi[i].admin_state_enabled)
+ continue;
+
+ ret = pse_disable_pi_pol(pcdev, i);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+/**
+ * pse_pi_allocate_pw_budget_static_prio - Allocate power budget for the PI
+ * when the budget eval strategy is
+ * static
+ * @pcdev: a pointer to the PSE
+ * @id: index of the PSE control
+ * @pw_req: power requested in mW
+ * @extack: extack for error reporting
+ *
+ * Return: 0 on success and failure value on error
+ */
+static int
+pse_pi_allocate_pw_budget_static_prio(struct pse_controller_dev *pcdev, int id,
+ int pw_req, struct netlink_ext_ack *extack)
+{
+ struct pse_pi *pi = &pcdev->pi[id];
+ int ret, _prio;
+
+ _prio = pcdev->nr_lines;
+ while (regulator_request_power_budget(pi->pw_d->supply, pw_req) == -ERANGE) {
+ if (_prio <= pi->prio) {
+ NL_SET_ERR_MSG_FMT(extack,
+ "PI %d: not enough power budget available",
+ id);
+ return -ERANGE;
+ }
+
+ ret = pse_disable_pi_prio(pcdev, pi->pw_d, _prio);
+ if (ret < 0)
+ return ret;
+
+ _prio--;
+ }
+
+ pi->pw_allocated_mW = pw_req;
+ return 0;
+}
+
+/**
+ * pse_pi_allocate_pw_budget - Allocate power budget for the PI
+ * @pcdev: a pointer to the PSE
+ * @id: index of the PSE control
+ * @pw_req: power requested in mW
+ * @extack: extack for error reporting
+ *
+ * Return: 0 on success and failure value on error
+ */
+static int pse_pi_allocate_pw_budget(struct pse_controller_dev *pcdev, int id,
+ int pw_req, struct netlink_ext_ack *extack)
+{
+ struct pse_pi *pi = &pcdev->pi[id];
+
+ if (!pi->pw_d)
+ return 0;
+
+ /* ETHTOOL_PSE_BUDGET_EVAL_STRAT_STATIC */
+ if (pi->pw_d->budget_eval_strategy == ETHTOOL_PSE_BUDGET_EVAL_STRAT_STATIC)
+ return pse_pi_allocate_pw_budget_static_prio(pcdev, id, pw_req,
+ extack);
+
+ return 0;
+}
+
+/**
+ * _pse_pi_enable_sw_pw_ctrl - Enable PSE PI in case of software power control.
+ * Assumes the PSE lock has been acquired
+ * @pcdev: a pointer to the PSE
+ * @id: index of the PSE control
+ * @extack: extack for error reporting
+ *
+ * Return: 0 on success and failure value on error
+ */
+static int _pse_pi_enable_sw_pw_ctrl(struct pse_controller_dev *pcdev, int id,
+ struct netlink_ext_ack *extack)
+{
+ const struct pse_controller_ops *ops = pcdev->ops;
+ struct pse_pi *pi = &pcdev->pi[id];
+ int ret, pw_req;
+
+ if (!ops->pi_get_pw_req) {
+ /* No power allocation management */
+ ret = ops->pi_enable(pcdev, id);
+ if (ret)
+ NL_SET_ERR_MSG_FMT(extack,
+ "PI %d: enable error %d",
+ id, ret);
+ return ret;
+ }
+
+ ret = ops->pi_get_pw_req(pcdev, id);
+ if (ret < 0)
+ return ret;
+
+ pw_req = ret;
+
+ /* Compare requested power with port power limit and use the lowest
+ * one.
+ */
+ if (ops->pi_get_pw_limit) {
+ ret = ops->pi_get_pw_limit(pcdev, id);
+ if (ret < 0)
+ return ret;
+
+ if (ret < pw_req)
+ pw_req = ret;
+ }
+
+ ret = pse_pi_allocate_pw_budget(pcdev, id, pw_req, extack);
+ if (ret)
+ return ret;
+
+ ret = ops->pi_enable(pcdev, id);
+ if (ret) {
+ pse_pi_deallocate_pw_budget(pi);
+ NL_SET_ERR_MSG_FMT(extack,
+ "PI %d: enable error %d",
+ id, ret);
+ return ret;
+ }
+
+ return 0;
+}
+
static int pse_pi_enable(struct regulator_dev *rdev)
{
struct pse_controller_dev *pcdev = rdev_get_drvdata(rdev);
const struct pse_controller_ops *ops;
- int id, ret;
+ int id, ret = 0;
ops = pcdev->ops;
if (!ops->pi_enable)
@@ -261,6 +541,23 @@ static int pse_pi_enable(struct regulator_dev *rdev)
id = rdev_get_id(rdev);
mutex_lock(&pcdev->lock);
+ if (pse_pw_d_is_sw_pw_control(pcdev, pcdev->pi[id].pw_d)) {
+ /* Manage enabled status by software.
+ * Real enable process will happen if a port is connected.
+ */
+ if (pcdev->pi[id].isr_pd_detected) {
+ struct netlink_ext_ack extack;
+
+ ret = _pse_pi_enable_sw_pw_ctrl(pcdev, id, &extack);
+ if (!ret)
+ pcdev->pi[id].admin_state_enabled = 1;
+ } else {
+ pcdev->pi[id].admin_state_enabled = 1;
+ }
+ mutex_unlock(&pcdev->lock);
+ return ret;
+ }
+
ret = ops->pi_enable(pcdev, id);
if (!ret)
pcdev->pi[id].admin_state_enabled = 1;
@@ -272,21 +569,20 @@ static int pse_pi_enable(struct regulator_dev *rdev)
static int pse_pi_disable(struct regulator_dev *rdev)
{
struct pse_controller_dev *pcdev = rdev_get_drvdata(rdev);
- const struct pse_controller_ops *ops;
+ struct pse_pi *pi;
int id, ret;
- ops = pcdev->ops;
- if (!ops->pi_disable)
- return -EOPNOTSUPP;
-
id = rdev_get_id(rdev);
+ pi = &pcdev->pi[id];
mutex_lock(&pcdev->lock);
- ret = ops->pi_disable(pcdev, id);
- if (!ret)
- pcdev->pi[id].admin_state_enabled = 0;
- mutex_unlock(&pcdev->lock);
+ ret = _pse_pi_disable(pcdev, id);
+ if (!ret) {
+ pi->admin_state_enabled = 0;
+ pcdev->pi[id]._isr_counter_mismatch = 0;
+ }
- return ret;
+ mutex_unlock(&pcdev->lock);
+ return 0;
}
static int _pse_pi_get_voltage(struct regulator_dev *rdev)
@@ -542,6 +838,10 @@ static int pse_register_pw_ds(struct pse_controller_dev *pcdev)
}
pw_d->supply = supply;
+ if (pcdev->supp_budget_eval_strategies)
+ pw_d->budget_eval_strategy = pcdev->supp_budget_eval_strategies;
+ else
+ pw_d->budget_eval_strategy = ETHTOOL_PSE_BUDGET_EVAL_STRAT_DISABLED;
pcdev->pi[i].pw_d = pw_d;
}
@@ -700,27 +1000,45 @@ static unsigned long pse_to_regulator_notifs(unsigned long notifs)
}
/**
- * pse_control_find_phy_by_id - Find PHY attached to the a pse control id
+ * pse_set_config_isr - Set PSE control config according to the PSE
+ * notifications
* @pcdev: a pointer to the PSE
* @id: index of the PSE control
+ * @notifs: PSE event notifications
+ * @extack: extack for error reporting
*
- * Return: PHY device pointer or NULL
+ * Return: 0 on success and failure value on error
*/
-static struct phy_device *
-pse_control_find_phy_by_id(struct pse_controller_dev *pcdev, int id)
+static int pse_set_config_isr(struct pse_controller_dev *pcdev, int id,
+ unsigned long notifs,
+ struct netlink_ext_ack *extack)
{
- struct pse_control *psec;
+ int ret = 0;
- mutex_lock(&pse_list_mutex);
- list_for_each_entry(psec, &pcdev->pse_control_head, list) {
- if (psec->id == id) {
- mutex_unlock(&pse_list_mutex);
- return psec->attached_phydev;
- }
+ if (notifs & ETHTOOL_PSE_BUDGET_EVAL_STRAT_DYNAMIC)
+ return 0;
+
+ if ((notifs & ETHTOOL_C33_PSE_EVENT_DISCONNECTION) &&
+ ((notifs & ETHTOOL_C33_PSE_EVENT_DETECTION) ||
+ (notifs & ETHTOOL_C33_PSE_EVENT_CLASSIFICATION))) {
+ NL_SET_ERR_MSG_FMT(extack,
+ "PI %d: error, connection and disconnection reported simultaneously",
+ id);
+ return -EINVAL;
}
- mutex_unlock(&pse_list_mutex);
- return NULL;
+ if (notifs & ETHTOOL_C33_PSE_EVENT_CLASSIFICATION) {
+ pcdev->pi[id].isr_pd_detected = true;
+ if (pcdev->pi[id].admin_state_enabled)
+ ret = _pse_pi_enable_sw_pw_ctrl(pcdev, id, extack);
+ } else if (notifs & ETHTOOL_C33_PSE_EVENT_DISCONNECTION) {
+ if (pcdev->pi[id].admin_state_enabled &&
+ pcdev->pi[id].isr_pd_detected)
+ ret = _pse_pi_disable(pcdev, id);
+ pcdev->pi[id].isr_pd_detected = false;
+ }
+
+ return ret;
}
/**
@@ -746,9 +1064,10 @@ static irqreturn_t pse_isr(int irq, void *data)
memset(h->notifs, 0, pcdev->nr_lines * sizeof(*h->notifs));
mutex_lock(&pcdev->lock);
ret = desc->map_event(irq, pcdev, h->notifs, ¬ifs_mask);
- mutex_unlock(&pcdev->lock);
- if (ret || !notifs_mask)
+ if (ret || !notifs_mask) {
+ mutex_unlock(&pcdev->lock);
return IRQ_NONE;
+ }
for_each_set_bit(i, ¬ifs_mask, pcdev->nr_lines) {
struct phy_device *phydev;
@@ -759,6 +1078,12 @@ static irqreturn_t pse_isr(int irq, void *data)
continue;
notifs = h->notifs[i];
+ if (pse_pw_d_is_sw_pw_control(pcdev, pcdev->pi[i].pw_d)) {
+ ret = pse_set_config_isr(pcdev, i, notifs, &extack);
+ if (ret)
+ notifs |= ETHTOOL_PSE_EVENT_SW_PW_CONTROL_ERROR;
+ }
+
dev_dbg(h->pcdev->dev,
"Sending PSE notification EVT 0x%lx\n", notifs);
@@ -770,6 +1095,8 @@ static irqreturn_t pse_isr(int irq, void *data)
NULL);
}
+ mutex_unlock(&pcdev->lock);
+
return IRQ_HANDLED;
}
@@ -864,6 +1191,7 @@ static struct pse_control *
pse_control_get_internal(struct pse_controller_dev *pcdev, unsigned int index,
struct phy_device *phydev)
{
+ struct pse_admin_state admin_state = {0};
struct pse_control *psec;
int ret;
@@ -885,6 +1213,23 @@ pse_control_get_internal(struct pse_controller_dev *pcdev, unsigned int index,
goto free_psec;
}
+ if (!pcdev->ops->pi_get_admin_state) {
+ ret = -EOPNOTSUPP;
+ goto free_psec;
+ }
+
+ /* Initialize admin_state_enabled before the regulator_get. This
+ * aims to have the right value reported in the first is_enabled
+ * call in case of control managed by software.
+ */
+ ret = pcdev->ops->pi_get_admin_state(pcdev, index, &admin_state);
+ if (ret)
+ goto free_psec;
+
+ if (admin_state.podl_admin_state == ETHTOOL_PODL_PSE_ADMIN_STATE_ENABLED ||
+ admin_state.c33_admin_state == ETHTOOL_C33_PSE_ADMIN_STATE_ENABLED)
+ pcdev->pi[index].admin_state_enabled = 1;
+
psec->ps = devm_regulator_get_exclusive(pcdev->dev,
rdev_get_name(pcdev->pi[index].rdev));
if (IS_ERR(psec->ps)) {
@@ -892,12 +1237,6 @@ pse_control_get_internal(struct pse_controller_dev *pcdev, unsigned int index,
goto put_module;
}
- ret = regulator_is_enabled(psec->ps);
- if (ret < 0)
- goto regulator_put;
-
- pcdev->pi[index].admin_state_enabled = ret;
-
psec->pcdev = pcdev;
list_add(&psec->list, &pcdev->pse_control_head);
psec->id = index;
@@ -906,8 +1245,6 @@ pse_control_get_internal(struct pse_controller_dev *pcdev, unsigned int index,
return psec;
-regulator_put:
- devm_regulator_put(psec->ps);
put_module:
module_put(pcdev->owner);
free_psec:
@@ -1018,6 +1355,35 @@ struct pse_control *of_pse_control_get(struct device_node *node,
}
EXPORT_SYMBOL_GPL(of_pse_control_get);
+/**
+ * pse_get_sw_admin_state - Convert the software admin state to c33 or podl
+ * admin state value used in the standard
+ * @psec: PSE control pointer
+ * @admin_state: a pointer to the admin_state structure
+ */
+static void pse_get_sw_admin_state(struct pse_control *psec,
+ struct pse_admin_state *admin_state)
+{
+ struct pse_pi *pi = &psec->pcdev->pi[psec->id];
+
+ if (pse_has_podl(psec)) {
+ if (pi->admin_state_enabled)
+ admin_state->podl_admin_state =
+ ETHTOOL_PODL_PSE_ADMIN_STATE_ENABLED;
+ else
+ admin_state->podl_admin_state =
+ ETHTOOL_PODL_PSE_ADMIN_STATE_DISABLED;
+ }
+ if (pse_has_c33(psec)) {
+ if (pi->admin_state_enabled)
+ admin_state->c33_admin_state =
+ ETHTOOL_C33_PSE_ADMIN_STATE_ENABLED;
+ else
+ admin_state->c33_admin_state =
+ ETHTOOL_C33_PSE_ADMIN_STATE_DISABLED;
+ }
+}
+
/**
* pse_ethtool_get_status - get status of PSE control
* @psec: PSE control pointer
@@ -1034,19 +1400,47 @@ int pse_ethtool_get_status(struct pse_control *psec,
struct pse_pw_status pw_status = {0};
const struct pse_controller_ops *ops;
struct pse_controller_dev *pcdev;
+ struct pse_pi *pi;
int ret;
pcdev = psec->pcdev;
ops = pcdev->ops;
+
+ pi = &pcdev->pi[psec->id];
mutex_lock(&pcdev->lock);
- if (pcdev->pi[psec->id].pw_d)
- status->pw_d_id = pcdev->pi[psec->id].pw_d->id;
+ if (pi->pw_d) {
+ status->pw_d_id = pi->pw_d->id;
+ status->budget_eval_strategy = pi->pw_d->budget_eval_strategy;
+ if (pse_pw_d_is_sw_pw_control(pcdev, pi->pw_d)) {
+ pse_get_sw_admin_state(psec, &admin_state);
+ } else {
+ ret = ops->pi_get_admin_state(pcdev, psec->id,
+ &admin_state);
+ if (ret)
+ goto out;
+ }
+ status->podl_admin_state = admin_state.podl_admin_state;
+ status->c33_admin_state = admin_state.c33_admin_state;
- ret = ops->pi_get_admin_state(pcdev, psec->id, &admin_state);
- if (ret)
- goto out;
- status->podl_admin_state = admin_state.podl_admin_state;
- status->c33_admin_state = admin_state.c33_admin_state;
+ switch (pi->pw_d->budget_eval_strategy) {
+ case ETHTOOL_PSE_BUDGET_EVAL_STRAT_STATIC:
+ status->prio_max = pcdev->nr_lines;
+ status->prio = pi->prio;
+ break;
+ case ETHTOOL_PSE_BUDGET_EVAL_STRAT_DYNAMIC:
+ status->prio_max = pcdev->pis_prio_max;
+ if (ops->pi_get_prio) {
+ ret = ops->pi_get_prio(pcdev, psec->id);
+ if (ret < 0)
+ goto out;
+
+ status->prio = ret;
+ }
+ break;
+ default:
+ break;
+ }
+ }
ret = ops->pi_get_pw_status(pcdev, psec->id, &pw_status);
if (ret)
@@ -1121,11 +1515,15 @@ static int pse_ethtool_c33_set_config(struct pse_control *psec,
case ETHTOOL_C33_PSE_ADMIN_STATE_ENABLED:
/* We could have mismatch between admin_state_enabled and
* state reported by regulator_is_enabled. This can occur when
- * the PI is forcibly turn off by the controller. Call
+ * the PI is forcibly turn off by the controller or in power
+ * off case in the interrupt context. Call
* regulator_disable on that case to fix the counters state.
+ * disable action might be called two times consecutively
+ * but that is not a real issue.
*/
- if (psec->pcdev->pi[psec->id].admin_state_enabled &&
- !regulator_is_enabled(psec->ps)) {
+ if ((psec->pcdev->pi[psec->id].admin_state_enabled &&
+ !regulator_is_enabled(psec->ps)) ||
+ psec->pcdev->pi[psec->id]._isr_counter_mismatch) {
err = regulator_disable(psec->ps);
if (err)
break;
@@ -1195,6 +1593,52 @@ int pse_ethtool_set_config(struct pse_control *psec,
}
EXPORT_SYMBOL_GPL(pse_ethtool_set_config);
+/**
+ * pse_pi_update_pw_budget - Update PSE power budget allocated with new
+ * power in mW
+ * @pcdev: a pointer to the PSE controller device
+ * @id: index of the PSE PI
+ * @pw_req: power requested
+ * @extack: extack for reporting useful error messages
+ *
+ * Return: Previous power allocated on success and failure value on error
+ */
+static int pse_pi_update_pw_budget(struct pse_controller_dev *pcdev, int id,
+ const unsigned int pw_req,
+ struct netlink_ext_ack *extack)
+{
+ struct pse_pi *pi = &pcdev->pi[id];
+ int previous_pw_allocated;
+ int pw_diff, ret = 0;
+
+ /* We don't want pw_allocated_mW value change in the middle of an
+ * power budget update
+ */
+ mutex_lock(&pcdev->lock);
+ previous_pw_allocated = pi->pw_allocated_mW;
+ pw_diff = pw_req - previous_pw_allocated;
+ if (!pw_diff) {
+ goto out;
+ } else if (pw_diff > 0) {
+ ret = regulator_request_power_budget(pi->pw_d->supply, pw_diff);
+ if (ret) {
+ NL_SET_ERR_MSG_FMT(extack,
+ "PI %d: not enough power budget available",
+ id);
+ goto out;
+ }
+
+ } else {
+ regulator_free_power_budget(pi->pw_d->supply, -pw_diff);
+ }
+ pi->pw_allocated_mW = pw_req;
+ ret = previous_pw_allocated;
+
+out:
+ mutex_unlock(&pcdev->lock);
+ return ret;
+}
+
/**
* pse_ethtool_set_pw_limit - set PSE control power limit
* @psec: PSE control pointer
@@ -1207,7 +1651,7 @@ int pse_ethtool_set_pw_limit(struct pse_control *psec,
struct netlink_ext_ack *extack,
const unsigned int pw_limit)
{
- int uV, uA, ret;
+ int uV, uA, ret, previous_pw_allocated = 0;
s64 tmp_64;
if (pw_limit > MAX_PI_PW)
@@ -1231,10 +1675,99 @@ int pse_ethtool_set_pw_limit(struct pse_control *psec,
/* uA = mW * 1000000000 / uV */
uA = DIV_ROUND_CLOSEST_ULL(tmp_64, uV);
- return regulator_set_current_limit(psec->ps, 0, uA);
+ /* Update power budget only in software power control case and
+ * if a Power Device is powered.
+ */
+ if (pse_pw_d_is_sw_pw_control(psec->pcdev,
+ psec->pcdev->pi[psec->id].pw_d) &&
+ psec->pcdev->pi[psec->id].admin_state_enabled &&
+ psec->pcdev->pi[psec->id].isr_pd_detected) {
+ ret = pse_pi_update_pw_budget(psec->pcdev, psec->id,
+ pw_limit, extack);
+ if (ret < 0)
+ return ret;
+ previous_pw_allocated = ret;
+ }
+
+ ret = regulator_set_current_limit(psec->ps, 0, uA);
+ if (ret < 0 && previous_pw_allocated) {
+ pse_pi_update_pw_budget(psec->pcdev, psec->id,
+ previous_pw_allocated, extack);
+ }
+
+ return ret;
}
EXPORT_SYMBOL_GPL(pse_ethtool_set_pw_limit);
+/**
+ * pse_ethtool_set_prio - Set PSE PI priority according to the budget
+ * evaluation strategy
+ * @psec: PSE control pointer
+ * @extack: extack for reporting useful error messages
+ * @prio: priovity value
+ *
+ * Return: 0 on success and failure value on error
+ */
+int pse_ethtool_set_prio(struct pse_control *psec,
+ struct netlink_ext_ack *extack,
+ unsigned int prio)
+{
+ struct pse_controller_dev *pcdev = psec->pcdev;
+ const struct pse_controller_ops *ops;
+ int ret = 0;
+
+ if (!pcdev->pi[psec->id].pw_d) {
+ NL_SET_ERR_MSG(extack, "no power domain attached");
+ return -EOPNOTSUPP;
+ }
+
+ /* We don't want priority change in the middle of an
+ * enable/disable call or a priority mode change
+ */
+ mutex_lock(&pcdev->lock);
+ switch (pcdev->pi[psec->id].pw_d->budget_eval_strategy) {
+ case ETHTOOL_PSE_BUDGET_EVAL_STRAT_STATIC:
+ if (prio > pcdev->nr_lines) {
+ NL_SET_ERR_MSG_FMT(extack,
+ "priority %d exceed priority max %d",
+ prio, pcdev->nr_lines);
+ ret = -ERANGE;
+ goto out;
+ }
+
+ pcdev->pi[psec->id].prio = prio;
+ break;
+
+ case ETHTOOL_PSE_BUDGET_EVAL_STRAT_DYNAMIC:
+ ops = psec->pcdev->ops;
+ if (!ops->pi_set_prio) {
+ NL_SET_ERR_MSG(extack,
+ "pse driver does not support setting port priority");
+ ret = -EOPNOTSUPP;
+ goto out;
+ }
+
+ if (prio > pcdev->pis_prio_max) {
+ NL_SET_ERR_MSG_FMT(extack,
+ "priority %d exceed priority max %d",
+ prio, pcdev->pis_prio_max);
+ ret = -ERANGE;
+ goto out;
+ }
+
+ ret = ops->pi_set_prio(pcdev, psec->id, prio);
+ break;
+
+ default:
+ ret = -EOPNOTSUPP;
+ }
+
+out:
+ mutex_unlock(&pcdev->lock);
+ return ret;
+}
+EXPORT_SYMBOL_GPL(pse_ethtool_set_prio);
+
bool pse_has_podl(struct pse_control *psec)
{
return psec->pcdev->types & ETHTOOL_PSE_PODL;
diff --git a/include/linux/pse-pd/pse.h b/include/linux/pse-pd/pse.h
index ffa6cf9a0072..d535e9709656 100644
--- a/include/linux/pse-pd/pse.h
+++ b/include/linux/pse-pd/pse.h
@@ -132,6 +132,10 @@ struct pse_pw_limit_ranges {
* is in charge of the memory allocation
* @c33_pw_limit_nb_ranges: number of supported power limit configuration
* ranges
+ * @budget_eval_strategy: PSE budget evaluation strategy selected.
+ * @prio_max: max priority allowed for the c33_prio variable value.
+ * @prio: priority of the PSE. Managed by PSE core in case of static budget
+ * evaluation strategy.
*/
struct ethtool_pse_control_status {
u32 pw_d_id;
@@ -145,6 +149,9 @@ struct ethtool_pse_control_status {
u32 c33_avail_pw_limit;
struct ethtool_c33_pse_pw_limit_range *c33_pw_limit_ranges;
u32 c33_pw_limit_nb_ranges;
+ enum ethtool_pse_budget_eval_strategies budget_eval_strategy;
+ u32 prio_max;
+ u32 prio;
};
/**
@@ -168,6 +175,11 @@ struct ethtool_pse_control_status {
* range. The driver is in charge of the memory
* allocation and should return the number of
* ranges.
+ * @pi_get_prio: Get the PSE PI priority.
+ * @pi_set_prio: Configure the PSE PI priority.
+ * @pi_get_pw_req: Get the power requested by a PD before enabling the PSE PI.
+ * This is only relevant when an interrupt is registered using
+ * devm_pse_irq_helper helper.
*/
struct pse_controller_ops {
int (*setup_pi_matrix)(struct pse_controller_dev *pcdev);
@@ -188,6 +200,10 @@ struct pse_controller_ops {
int id, int max_mW);
int (*pi_get_pw_limit_ranges)(struct pse_controller_dev *pcdev, int id,
struct pse_pw_limit_ranges *pw_limit_ranges);
+ int (*pi_get_prio)(struct pse_controller_dev *pcdev, int id);
+ int (*pi_set_prio)(struct pse_controller_dev *pcdev, int id,
+ unsigned int prio);
+ int (*pi_get_pw_req)(struct pse_controller_dev *pcdev, int id);
};
struct module;
@@ -223,6 +239,17 @@ struct pse_pi_pairset {
* @rdev: regulator represented by the PSE PI
* @admin_state_enabled: PI enabled state
* @pw_d: Power domain of the PSE PI
+ * @prio: Priority of the PSE PI. Used in static budget evaluation strategy
+ * @isr_pd_detected: PSE PI detection status managed by the interruption
+ * handler. This variable is relevant when the power enabled
+ * management is managed in software like the static
+ * budget evaluation strategy.
+ * @pw_allocated_mW: Power allocated to a PSE PI to manage power budget in
+ * static budget evaluation strategy.
+ * @_isr_counter_mismatch: Internal flag used in PSE core in case of a
+ * counter mismatch between regulator and PSE API.
+ * This is caused by a disable call in the interrupt
+ * context handler.
*/
struct pse_pi {
struct pse_pi_pairset pairset[2];
@@ -230,6 +257,10 @@ struct pse_pi {
struct regulator_dev *rdev;
bool admin_state_enabled;
struct pse_power_domain *pw_d;
+ int prio;
+ bool isr_pd_detected;
+ int pw_allocated_mW;
+ bool _isr_counter_mismatch;
};
/**
@@ -247,6 +278,9 @@ struct pse_pi {
* @pi: table of PSE PIs described in this controller device
* @no_of_pse_pi: flag set if the pse_pis devicetree node is not used
* @irq: PSE interrupt
+ * @pis_prio_max: Maximum value allowed for the PSE PIs priority
+ * @supp_budget_eval_strategies: budget evaluation strategies supported
+ * by the PSE
*/
struct pse_controller_dev {
const struct pse_controller_ops *ops;
@@ -261,6 +295,8 @@ struct pse_controller_dev {
struct pse_pi *pi;
bool no_of_pse_pi;
int irq;
+ unsigned int pis_prio_max;
+ u32 supp_budget_eval_strategies;
};
#if IS_ENABLED(CONFIG_PSE_CONTROLLER)
@@ -285,6 +321,9 @@ int pse_ethtool_set_config(struct pse_control *psec,
int pse_ethtool_set_pw_limit(struct pse_control *psec,
struct netlink_ext_ack *extack,
const unsigned int pw_limit);
+int pse_ethtool_set_prio(struct pse_control *psec,
+ struct netlink_ext_ack *extack,
+ unsigned int prio);
bool pse_has_podl(struct pse_control *psec);
bool pse_has_c33(struct pse_control *psec);
@@ -322,6 +361,13 @@ static inline int pse_ethtool_set_pw_limit(struct pse_control *psec,
return -EOPNOTSUPP;
}
+static inline int pse_ethtool_set_prio(struct pse_control *psec,
+ struct netlink_ext_ack *extack,
+ unsigned int prio)
+{
+ return -EOPNOTSUPP;
+}
+
static inline bool pse_has_podl(struct pse_control *psec)
{
return false;
diff --git a/include/uapi/linux/ethtool.h b/include/uapi/linux/ethtool.h
index 8793946ff851..0f618600e1fb 100644
--- a/include/uapi/linux/ethtool.h
+++ b/include/uapi/linux/ethtool.h
@@ -1008,6 +1008,20 @@ enum ethtool_c33_pse_pw_d_status {
* enum ethtool_pse_events - event list of the PSE controller.
* @ETHTOOL_PSE_EVENT_OVER_CURRENT: PSE output current is too high.
* @ETHTOOL_PSE_EVENT_OVER_TEMP: PSE in over temperature state.
+ * @ETHTOOL_C33_PSE_EVENT_DETECTION: detection process occur on the PSE.
+ * IEEE 802.3-2022 33.2.5 and 145.2.6 PSE detection of PDs.
+ * IEEE 802.3-202 30.9.1.1.5 aPSEPowerDetectionStatus.
+ * @ETHTOOL_C33_PSE_EVENT_CLASSIFICATION: classification process occur on
+ * the PSE. IEEE 802.3-2022 33.2.6 and 145.2.8 classification of PDs and
+ * mutual identification.
+ * IEEE 802.3-2022 30.9.1.1.8 aPSEPowerClassification.
+ * @ETHTOOL_C33_PSE_EVENT_DISCONNECTION: PD has been disconnected on the PSE.
+ * IEEE 802.3-2022 33.3.8 and 145.3.9 PD Maintain Power Signature.
+ * IEEE 802.3-2022 33.5.1.2.9 MPS Absent.
+ * IEEE 802.3-2022 30.9.1.1.20 aPSEMPSAbsentCounter.
+ * @ETHTOOL_PSE_EVENT_OVER_BUDGET: PSE turned off due to over budget situation.
+ * @ETHTOOL_PSE_EVENT_SW_PW_CONTROL_ERROR: PSE faced an error managing the
+ * power control from software.
*
* @ETHTOOL_PSE_EVENT_LAST: Last PSE event of the enum.
*/
@@ -1015,8 +1029,31 @@ enum ethtool_c33_pse_pw_d_status {
enum ethtool_pse_events {
ETHTOOL_PSE_EVENT_OVER_CURRENT = 1 << 0,
ETHTOOL_PSE_EVENT_OVER_TEMP = 1 << 1,
+ ETHTOOL_C33_PSE_EVENT_DETECTION = 1 << 2,
+ ETHTOOL_C33_PSE_EVENT_CLASSIFICATION = 1 << 3,
+ ETHTOOL_C33_PSE_EVENT_DISCONNECTION = 1 << 4,
+ ETHTOOL_PSE_EVENT_OVER_BUDGET = 1 << 5,
+ ETHTOOL_PSE_EVENT_SW_PW_CONTROL_ERROR = 1 << 6,
- ETHTOOL_PSE_EVENT_LAST = ETHTOOL_PSE_EVENT_OVER_TEMP,
+ ETHTOOL_PSE_EVENT_LAST = ETHTOOL_PSE_EVENT_SW_PW_CONTROL_ERROR,
+};
+
+/**
+ * enum ethtool_pse_budget_eval_strategies - PSE budget evaluation strategies.
+ * @ETHTOOL_PSE_BUDGET_EVAL_STRAT_DISABLED: Budget evaluation strategy disabled.
+ * @ETHTOOL_PSE_BUDGET_EVAL_STRAT_STATIC: PSE static budget evaluation strategy.
+ * Budget evaluation strategy based on the power requested during PD
+ * classification. This strategy is managed by the PSE core.
+ * @ETHTOOL_PSE_BUDGET_EVAL_STRAT_DYNAMIC: PSE dynamic budget evaluation
+ * strategy. Budget evaluation strategy based on the current consumption
+ * per ports compared to the total power budget. This mode is managed by
+ * the PSE controller.
+ */
+
+enum ethtool_pse_budget_eval_strategies {
+ ETHTOOL_PSE_BUDGET_EVAL_STRAT_DISABLED = 1 << 0,
+ ETHTOOL_PSE_BUDGET_EVAL_STRAT_STATIC = 1 << 1,
+ ETHTOOL_PSE_BUDGET_EVAL_STRAT_DYNAMIC = 1 << 2,
};
/**
diff --git a/net/ethtool/common.c b/net/ethtool/common.c
index 8d207ec6456e..f1fe9a7dd735 100644
--- a/net/ethtool/common.c
+++ b/net/ethtool/common.c
@@ -520,6 +520,12 @@ static_assert(ARRAY_SIZE(udp_tunnel_type_names) ==
const char pse_event_names[][ETH_GSTRING_LEN] = {
[const_ilog2(ETHTOOL_PSE_EVENT_OVER_CURRENT)] = "over-current",
[const_ilog2(ETHTOOL_PSE_EVENT_OVER_TEMP)] = "over-temperature",
+ [const_ilog2(ETHTOOL_C33_PSE_EVENT_DETECTION)] = "detection",
+ [const_ilog2(ETHTOOL_C33_PSE_EVENT_CLASSIFICATION)] = "classification",
+ [const_ilog2(ETHTOOL_C33_PSE_EVENT_DISCONNECTION)] = "disconnection",
+ [const_ilog2(ETHTOOL_PSE_EVENT_OVER_BUDGET)] = "over-budget",
+ [const_ilog2(ETHTOOL_PSE_EVENT_SW_PW_CONTROL_ERROR)] =
+ "software-pw-control-error",
};
static_assert(ARRAY_SIZE(pse_event_names) == __PSE_EVENT_CNT);
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-18 16:19 ` [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies Kory Maincent
@ 2025-02-21 0:51 ` Jakub Kicinski
2025-02-24 13:10 ` Kory Maincent
0 siblings, 1 reply; 42+ messages in thread
From: Jakub Kicinski @ 2025-02-21 0:51 UTC (permalink / raw)
To: Kory Maincent
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Tue, 18 Feb 2025 17:19:10 +0100 Kory Maincent wrote:
> This patch introduces the ability to configure the PSE PI budget evaluation
> strategies. Budget evaluation strategies is utilized by PSE controllers to
> determine which ports to turn off first in scenarios such as power budget
> exceedance.
>
> The pis_prio_max value is used to define the maximum priority level
> supported by the controller. Both the current priority and the maximum
> priority are exposed to the user through the pse_ethtool_get_status call.
>
> This patch add support for two mode of budget evaluation strategies.
> 1. Static Method:
The "methods" can be mixed for ports in a single "domain" ?
On a quick read I don't see this explained
--
pw-bot: cr
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-21 0:51 ` Jakub Kicinski
@ 2025-02-24 13:10 ` Kory Maincent
2025-02-24 21:45 ` Jakub Kicinski
0 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-24 13:10 UTC (permalink / raw)
To: Jakub Kicinski
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Thu, 20 Feb 2025 16:51:29 -0800
Jakub Kicinski <kuba@kernel.org> wrote:
> On Tue, 18 Feb 2025 17:19:10 +0100 Kory Maincent wrote:
> > This patch introduces the ability to configure the PSE PI budget evaluation
> > strategies. Budget evaluation strategies is utilized by PSE controllers to
> > determine which ports to turn off first in scenarios such as power budget
> > exceedance.
> >
> > The pis_prio_max value is used to define the maximum priority level
> > supported by the controller. Both the current priority and the maximum
> > priority are exposed to the user through the pse_ethtool_get_status call.
> >
> > This patch add support for two mode of budget evaluation strategies.
> > 1. Static Method:
>
> The "methods" can be mixed for ports in a single "domain" ?
No they can't for now. Even different PSE power domains within the same PSE
controller. I will make it explicit.
Regards,
--
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-24 13:10 ` Kory Maincent
@ 2025-02-24 21:45 ` Jakub Kicinski
2025-02-25 9:25 ` Kory Maincent
0 siblings, 1 reply; 42+ messages in thread
From: Jakub Kicinski @ 2025-02-24 21:45 UTC (permalink / raw)
To: Kory Maincent
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Mon, 24 Feb 2025 14:10:37 +0100 Kory Maincent wrote:
> > The "methods" can be mixed for ports in a single "domain" ?
>
> No they can't for now. Even different PSE power domains within the same PSE
> controller. I will make it explicit.
Sounds like the property is placed at the wrong level of the hierarchy,
then.
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-24 21:45 ` Jakub Kicinski
@ 2025-02-25 9:25 ` Kory Maincent
2025-02-26 1:47 ` Jakub Kicinski
0 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-25 9:25 UTC (permalink / raw)
To: Jakub Kicinski
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Mon, 24 Feb 2025 13:45:22 -0800
Jakub Kicinski <kuba@kernel.org> wrote:
> On Mon, 24 Feb 2025 14:10:37 +0100 Kory Maincent wrote:
> > > The "methods" can be mixed for ports in a single "domain" ?
> >
> > No they can't for now. Even different PSE power domains within the same PSE
> > controller. I will make it explicit.
>
> Sounds like the property is placed at the wrong level of the hierarchy,
> then.
When a PSE controller appears to be able to support mixed budget strategy and
could switch between them it will be better to have it set at the PSE power
domain level. As the budget is per PSE power domain, its strategy should also
be per PSE power domain.
For now, it is simply not configurable and can't be mixed. It is hard-coded by
the PSE driver.
Regards,
--
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-25 9:25 ` Kory Maincent
@ 2025-02-26 1:47 ` Jakub Kicinski
2025-02-26 5:59 ` Oleksij Rempel
0 siblings, 1 reply; 42+ messages in thread
From: Jakub Kicinski @ 2025-02-26 1:47 UTC (permalink / raw)
To: Kory Maincent
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Tue, 25 Feb 2025 10:25:58 +0100 Kory Maincent wrote:
> On Mon, 24 Feb 2025 13:45:22 -0800
> Jakub Kicinski <kuba@kernel.org> wrote:
>
> > > No they can't for now. Even different PSE power domains within the same PSE
> > > controller. I will make it explicit.
> >
> > Sounds like the property is placed at the wrong level of the hierarchy,
> > then.
>
> When a PSE controller appears to be able to support mixed budget strategy and
> could switch between them it will be better to have it set at the PSE power
> domain level. As the budget is per PSE power domain, its strategy should also
> be per PSE power domain.
> For now, it is simply not configurable and can't be mixed. It is hard-coded by
> the PSE driver.
Yes, but uAPI is forever. We will have to live with those domain
attributes duplicated on each port. Presumably these port attributes
will never support a SET operation, since the set should be towards
the domain? The uAPI does not inspire confidence. If we need more
drivers to define a common API maybe a local sysfs API in the driver
will do?
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-26 1:47 ` Jakub Kicinski
@ 2025-02-26 5:59 ` Oleksij Rempel
2025-02-26 6:06 ` Oleksij Rempel
0 siblings, 1 reply; 42+ messages in thread
From: Oleksij Rempel @ 2025-02-26 5:59 UTC (permalink / raw)
To: Jakub Kicinski
Cc: Kory Maincent, Andrew Lunn, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Tue, Feb 25, 2025 at 05:47:52PM -0800, Jakub Kicinski wrote:
> On Tue, 25 Feb 2025 10:25:58 +0100 Kory Maincent wrote:
> > On Mon, 24 Feb 2025 13:45:22 -0800
> > Jakub Kicinski <kuba@kernel.org> wrote:
> >
> > > > No they can't for now. Even different PSE power domains within the same PSE
> > > > controller. I will make it explicit.
> > >
> > > Sounds like the property is placed at the wrong level of the hierarchy,
> > > then.
> >
> > When a PSE controller appears to be able to support mixed budget strategy and
> > could switch between them it will be better to have it set at the PSE power
> > domain level. As the budget is per PSE power domain, its strategy should also
> > be per PSE power domain.
> > For now, it is simply not configurable and can't be mixed. It is hard-coded by
> > the PSE driver.
>
> Yes, but uAPI is forever. We will have to live with those domain
> attributes duplicated on each port. Presumably these port attributes
> will never support a SET operation, since the set should be towards
> the domain? The uAPI does not inspire confidence. If we need more
> drivers to define a common API maybe a local sysfs API in the driver
> will do?
I tend to disagree here. The evaluation/allocation methods should be
per port.
At this step, we support only "hardware"(firmware)-based methods:
1. Static – Plain hardware classification-based power allocation per
port.
2. Dynamic – Hardware classification with constant measurement for
optimization.
For some devices, the dynamic method may not work reliably enough,
so we will need to switch to a fixed allocation method, which is
currently not implemented but will be set via user space. This
should be configurable per port.
At some point, we will need to introduce LLDP-based allocation from
user space. This will be managed by a daemon.
For testing, here’s an example of how LLDP-based power negotiation can
be analyzed:
https://telecomtest.com.au/wp-content/uploads/2016/12/PDA-LLDP-Powered-Device-LLDP-Analyzer.pdf
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-26 5:59 ` Oleksij Rempel
@ 2025-02-26 6:06 ` Oleksij Rempel
2025-02-27 2:42 ` Jakub Kicinski
0 siblings, 1 reply; 42+ messages in thread
From: Oleksij Rempel @ 2025-02-26 6:06 UTC (permalink / raw)
To: Jakub Kicinski
Cc: Kory Maincent, Andrew Lunn, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Wed, Feb 26, 2025 at 06:59:45AM +0100, Oleksij Rempel wrote:
> On Tue, Feb 25, 2025 at 05:47:52PM -0800, Jakub Kicinski wrote:
> > On Tue, 25 Feb 2025 10:25:58 +0100 Kory Maincent wrote:
> > > On Mon, 24 Feb 2025 13:45:22 -0800
> > > Jakub Kicinski <kuba@kernel.org> wrote:
> > >
> > > > > No they can't for now. Even different PSE power domains within the same PSE
> > > > > controller. I will make it explicit.
> > > >
> > > > Sounds like the property is placed at the wrong level of the hierarchy,
> > > > then.
> > >
> > > When a PSE controller appears to be able to support mixed budget strategy and
> > > could switch between them it will be better to have it set at the PSE power
> > > domain level. As the budget is per PSE power domain, its strategy should also
> > > be per PSE power domain.
> > > For now, it is simply not configurable and can't be mixed. It is hard-coded by
> > > the PSE driver.
> >
> > Yes, but uAPI is forever. We will have to live with those domain
> > attributes duplicated on each port. Presumably these port attributes
> > will never support a SET operation, since the set should be towards
> > the domain? The uAPI does not inspire confidence. If we need more
> > drivers to define a common API maybe a local sysfs API in the driver
> > will do?
>
> I tend to disagree here. The evaluation/allocation methods should be
> per port.
>
> At this step, we support only "hardware"(firmware)-based methods:
> 1. Static – Plain hardware classification-based power allocation per
> port.
> 2. Dynamic – Hardware classification with constant measurement for
> optimization.
>
> For some devices, the dynamic method may not work reliably enough,
> so we will need to switch to a fixed allocation method, which is
> currently not implemented but will be set via user space. This
> should be configurable per port.
>
> At some point, we will need to introduce LLDP-based allocation from
> user space. This will be managed by a daemon.
>
> For testing, here’s an example of how LLDP-based power negotiation can
> be analyzed:
> https://telecomtest.com.au/wp-content/uploads/2016/12/PDA-LLDP-Powered-Device-LLDP-Analyzer.pdf
Here is one example how it is done by HP switches:
https://arubanetworking.hpe.com/techdocs/AOS-CX/10.08/HTML/monitoring_6200/Content/Chp_PoE/PoE_cmds/pow-ove-eth-all-by.htm
switch(config)# interface 1/1/1 <---- per interface
switch(config-if)# power-over-ethernet allocate-by usage
switch(config-if)# power-over-ethernet allocate-by class
Cisco example:
https://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus9000/sw/93x/power-over-ethernet/configuration/configuring-power-over-ethernet/m-configuring-power-over-ethernet.html
switch(config)# interface ethernet1/1 <---- per interface
switch(config-if)# power inline auto
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-26 6:06 ` Oleksij Rempel
@ 2025-02-27 2:42 ` Jakub Kicinski
2025-02-27 7:40 ` Oleksij Rempel
0 siblings, 1 reply; 42+ messages in thread
From: Jakub Kicinski @ 2025-02-27 2:42 UTC (permalink / raw)
To: Oleksij Rempel
Cc: Kory Maincent, Andrew Lunn, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Wed, 26 Feb 2025 07:06:55 +0100 Oleksij Rempel wrote:
> Here is one example how it is done by HP switches:
> https://arubanetworking.hpe.com/techdocs/AOS-CX/10.08/HTML/monitoring_6200/Content/Chp_PoE/PoE_cmds/pow-ove-eth-all-by.htm
>
> switch(config)# interface 1/1/1 <---- per interface
> switch(config-if)# power-over-ethernet allocate-by usage
> switch(config-if)# power-over-ethernet allocate-by class
>
> Cisco example:
> https://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus9000/sw/93x/power-over-ethernet/configuration/configuring-power-over-ethernet/m-configuring-power-over-ethernet.html
>
> switch(config)# interface ethernet1/1 <---- per interface
> switch(config-if)# power inline auto
I don't see any mention of a domain in these docs.
This patchset is creating a concept of "domain" but does
not expose it as an object.
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-27 2:42 ` Jakub Kicinski
@ 2025-02-27 7:40 ` Oleksij Rempel
2025-02-27 14:57 ` Kory Maincent
0 siblings, 1 reply; 42+ messages in thread
From: Oleksij Rempel @ 2025-02-27 7:40 UTC (permalink / raw)
To: Jakub Kicinski
Cc: Kory Maincent, Andrew Lunn, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Wed, Feb 26, 2025 at 06:42:57PM -0800, Jakub Kicinski wrote:
> On Wed, 26 Feb 2025 07:06:55 +0100 Oleksij Rempel wrote:
> > Here is one example how it is done by HP switches:
> > https://arubanetworking.hpe.com/techdocs/AOS-CX/10.08/HTML/monitoring_6200/Content/Chp_PoE/PoE_cmds/pow-ove-eth-all-by.htm
> >
> > switch(config)# interface 1/1/1 <---- per interface
> > switch(config-if)# power-over-ethernet allocate-by usage
> > switch(config-if)# power-over-ethernet allocate-by class
> >
> > Cisco example:
> > https://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus9000/sw/93x/power-over-ethernet/configuration/configuring-power-over-ethernet/m-configuring-power-over-ethernet.html
> >
> > switch(config)# interface ethernet1/1 <---- per interface
> > switch(config-if)# power inline auto
>
> I don't see any mention of a domain in these docs.
> This patchset is creating a concept of "domain" but does
> not expose it as an object.
Ok, I see. @Köry, can you please provide regulator_summary with some
inlined comments to regulators related to the PSE components and PSE
related outputs of ethtool (or what ever tool you are using).
I wont to use this examples to answer.
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-27 7:40 ` Oleksij Rempel
@ 2025-02-27 14:57 ` Kory Maincent
2025-02-27 16:40 ` Oleksij Rempel
0 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-27 14:57 UTC (permalink / raw)
To: Oleksij Rempel
Cc: Jakub Kicinski, Andrew Lunn, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Thu, 27 Feb 2025 08:40:25 +0100
Oleksij Rempel <o.rempel@pengutronix.de> wrote:
> On Wed, Feb 26, 2025 at 06:42:57PM -0800, Jakub Kicinski wrote:
> > On Wed, 26 Feb 2025 07:06:55 +0100 Oleksij Rempel wrote:
> > > Here is one example how it is done by HP switches:
> > > https://arubanetworking.hpe.com/techdocs/AOS-CX/10.08/HTML/monitoring_6200/Content/Chp_PoE/PoE_cmds/pow-ove-eth-all-by.htm
> > >
> > > switch(config)# interface 1/1/1 <---- per interface
> > > switch(config-if)# power-over-ethernet allocate-by usage
> > > switch(config-if)# power-over-ethernet allocate-by class
> > >
> > > Cisco example:
> > > https://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus9000/sw/93x/power-over-ethernet/configuration/configuring-power-over-ethernet/m-configuring-power-over-ethernet.html
> > >
> > > switch(config)# interface ethernet1/1 <---- per interface
> > > switch(config-if)# power inline auto
> >
> > I don't see any mention of a domain in these docs.
> > This patchset is creating a concept of "domain" but does
> > not expose it as an object.
>
> Ok, I see. @Köry, can you please provide regulator_summary with some
> inlined comments to regulators related to the PSE components and PSE
> related outputs of ethtool (or what ever tool you are using).
>
> I wont to use this examples to answer.
On my side, I am not close to using sysfs. As we do all configurations through
ethtool I have assumed we should continue with ethtool.
I think we should set the port priority through ethtool, but indeed the PSE
power domain method get and set could be moved to sysfs as it is not something
relative to the port but to a group of ports. Ethtool should still report the
PSE power domain ID of a port to know which domain the port is.
@Oleksij here it is:
# cat /sys/kernel/debug/regulator/regulator_summary
regulator use open bypass opmode voltage current min max
---------------------------------------------------------------------------------------
regulator-dummy 5 4 0 unknown 0mV 0mA 0mV 0mV
d00e0000.sata-target 1 0mA 0mV 0mV
d00e0000.sata-phy 1 0mA 0mV 0mV
d00e0000.sata-ahci 1 0mA 0mV 0mV
spi0.0-vcc 1 0mA 0mV 0mV
pse-reg 1 4 0 unknown 0mV 0mA 0mV 0mV
pse-0-0020_pi0 0 1 0 unknown 53816mV 2369mA 0mV 0mV
0-0020-pse-0-0020_pi0 0 0mA 0mV 0mV
pse-0-0020_pi2 0 1 0 unknown 53816mV 2369mA 0mV 0mV
0-0020-pse-0-0020_pi2 0 0mA 0mV 0mV
pse-0-0020_pi7 0 1 0 unknown 53816mV 2369mA 0mV 0mV
0-0020-pse-0-0020_pi7 0 0mA 0mV 0mV
pse-reg2 1 2 0 unknown 0mV 0mA 0mV 0mV
pse-0-0020_pi1 0 0 0 unknown 53816mV 4738mA 0mV 0mV
vcc_sd1 2 1 0 unknown 1800mV 0mA 1800mV 3300mV
d00d0000.mmc-vqmmc 1 0mA 1800mV 1950mV
# ./ynl/cli.py --spec netlink/specs/ethtool.yaml --no-schema --do pse-get --json
'{"header":{"dev-name":"wan"}}'
{'c33-pse-admin-state': 2,
'c33-pse-avail-pw-limit': 127500,
'c33-pse-pw-d-status': 2,
'c33-pse-pw-limit-ranges': [{'max': 99900, 'min': 2000}],
'header': {'dev-index': 4, 'dev-name': 'wan'},
'pse-budget-eval-strat': 2,
'pse-prio': 0,
'pse-prio-max': 8,
'pse-pw-d-id': 1}
# ./ynl/cli.py --spec netlink/specs/ethtool.yaml --no-schema --do pse-set --json
'{"header":{"dev-name":"wan"}, "pse-prio":1}'
None
# ./ynl/cli.py --spec netlink/specs/ethtool.yaml --no-schema --do pse-set --json
'{"header":{"dev-name":"wan"}, "c33-pse-avail-pw-limit":15000}'
None
Regards,
--
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com
^ permalink raw reply [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-27 14:57 ` Kory Maincent
@ 2025-02-27 16:40 ` Oleksij Rempel
2025-02-27 18:26 ` Kory Maincent
0 siblings, 1 reply; 42+ messages in thread
From: Oleksij Rempel @ 2025-02-27 16:40 UTC (permalink / raw)
To: Kory Maincent
Cc: Jakub Kicinski, Andrew Lunn, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Thu, Feb 27, 2025 at 03:57:27PM +0100, Kory Maincent wrote:
> On Thu, 27 Feb 2025 08:40:25 +0100
> Oleksij Rempel <o.rempel@pengutronix.de> wrote:
>
> > On Wed, Feb 26, 2025 at 06:42:57PM -0800, Jakub Kicinski wrote:
> > > On Wed, 26 Feb 2025 07:06:55 +0100 Oleksij Rempel wrote:
> > > > Here is one example how it is done by HP switches:
> > > > https://arubanetworking.hpe.com/techdocs/AOS-CX/10.08/HTML/monitoring_6200/Content/Chp_PoE/PoE_cmds/pow-ove-eth-all-by.htm
> > > >
> > > > switch(config)# interface 1/1/1 <---- per interface
> > > > switch(config-if)# power-over-ethernet allocate-by usage
> > > > switch(config-if)# power-over-ethernet allocate-by class
> > > >
> > > > Cisco example:
> > > > https://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus9000/sw/93x/power-over-ethernet/configuration/configuring-power-over-ethernet/m-configuring-power-over-ethernet.html
> > > >
> > > > switch(config)# interface ethernet1/1 <---- per interface
> > > > switch(config-if)# power inline auto
> > >
> > > I don't see any mention of a domain in these docs.
> > > This patchset is creating a concept of "domain" but does
> > > not expose it as an object.
> >
> > Ok, I see. @Köry, can you please provide regulator_summary with some
> > inlined comments to regulators related to the PSE components and PSE
> > related outputs of ethtool (or what ever tool you are using).
> >
> > I wont to use this examples to answer.
>
> On my side, I am not close to using sysfs. As we do all configurations
> through ethtool I have assumed we should continue with ethtool.
Yes, I agree. But it won't be possible to do it for all components.
> I think we should set the port priority through ethtool.
ack
> but indeed the PSE power domain method get and set could be moved to
> sysfs as it is not something relative to the port but to a group of
> ports.
I would prefer to have it in the for of devlink or use regulator netlink
interface. But, we do not need to do this discussion right now.
> Ethtool should still report the PSE power domain ID of a port to know
> which domain the port is.
Exactly.
@Jakub, at current implementation stage, user need to know the domain
id, because ports (and priorities) are grouped by the top level
regulators (pse-regX in the regulator_summary), they are our top-level
bottlenecks.
HP and Cisco switch either use different PSE controllers, or just didn't
exposed this nuance to the user. Let's assume, they have only one
global power domain.
So, in current patch set I would expect (not force :) ) implementation for
following fields:
- per port:
- priority (valid within the power domain)
- power reservation/allocation methods. First of all, because all
already supported controllers have different implemented/default
methods: microchip - dynamic, TI - static, regulator-pse - fixed (no
classification is supported).
At same time, in the future, we will need be able switch between
(static or dynamic) and fixed for LLPD or manual configuration.
Yes, at this point all ports show the same information and it seems
to be duplicated.
- power domain ID.
@Jakub, did I answered you question, or missed the point? :)
> @Oleksij here it is:
Thank you!
I do not expect it to be the primer user interface, but it can provide
additional diagnostic information. I wonted to see how it is aligns
with current ethtool UAPI implementation and if it possible to combine
it for diagnostics.
> # cat /sys/kernel/debug/regulator/regulator_summary
> regulator use open bypass opmode voltage current min max
> ---------------------------------------------------------------------------------------
> regulator-dummy 5 4 0 unknown 0mV 0mA 0mV 0mV
> d00e0000.sata-target 1 0mA 0mV 0mV
> d00e0000.sata-phy 1 0mA 0mV 0mV
> d00e0000.sata-ahci 1 0mA 0mV 0mV
> spi0.0-vcc 1 0mA 0mV 0mV
> pse-reg 1 4 0 unknown 0mV 0mA 0mV 0mV
pse-regX should be attached to the main supply regulator for better full
picture. And use different name to be better identified as PSE power domains with ID?
> pse-0-0020_pi0 0 1 0 unknown 53816mV 2369mA 0mV 0mV
> 0-0020-pse-0-0020_pi0 0 0mA 0mV 0mV
> pse-0-0020_pi2 0 1 0 unknown 53816mV 2369mA 0mV 0mV
> 0-0020-pse-0-0020_pi2 0 0mA 0mV 0mV
> pse-0-0020_pi7 0 1 0 unknown 53816mV 2369mA 0mV 0mV
> 0-0020-pse-0-0020_pi7 0 0mA 0mV 0mV
> pse-reg2 1 2 0 unknown 0mV 0mA 0mV 0mV
> pse-0-0020_pi1 0 0 0 unknown 53816mV 4738mA 0mV 0mV
> vcc_sd1 2 1 0 unknown 1800mV 0mA 1800mV 3300mV
> d00d0000.mmc-vqmmc 1 0mA 1800mV 1950mV
>
> # ./ynl/cli.py --spec netlink/specs/ethtool.yaml --no-schema --do pse-get --json
> '{"header":{"dev-name":"wan"}}'
> {'c33-pse-admin-state': 2,
> 'c33-pse-avail-pw-limit': 127500,
> 'c33-pse-pw-d-status': 2,
> 'c33-pse-pw-limit-ranges': [{'max': 99900, 'min': 2000}],
> 'header': {'dev-index': 4, 'dev-name': 'wan'},
> 'pse-budget-eval-strat': 2,
> 'pse-prio': 0,
> 'pse-prio-max': 8,
> 'pse-pw-d-id': 1}
>
> # ./ynl/cli.py --spec netlink/specs/ethtool.yaml --no-schema --do pse-set --json
> '{"header":{"dev-name":"wan"}, "pse-prio":1}'
> None
> # ./ynl/cli.py --spec netlink/specs/ethtool.yaml --no-schema --do pse-set --json
> '{"header":{"dev-name":"wan"}, "c33-pse-avail-pw-limit":15000}'
Best Regards,
Oleksij
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-27 16:40 ` Oleksij Rempel
@ 2025-02-27 18:26 ` Kory Maincent
2025-03-01 13:00 ` Oleksij Rempel
0 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-27 18:26 UTC (permalink / raw)
To: Oleksij Rempel
Cc: Jakub Kicinski, Andrew Lunn, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Thu, 27 Feb 2025 17:40:42 +0100
Oleksij Rempel <o.rempel@pengutronix.de> wrote:
> On Thu, Feb 27, 2025 at 03:57:27PM +0100, Kory Maincent wrote:
> > On Thu, 27 Feb 2025 08:40:25 +0100
> > Oleksij Rempel <o.rempel@pengutronix.de> wrote:
> >
> > > On Wed, Feb 26, 2025 at 06:42:57PM -0800, Jakub Kicinski wrote:
> [...]
> [...]
> [...]
> > >
> > > Ok, I see. @Köry, can you please provide regulator_summary with some
> > > inlined comments to regulators related to the PSE components and PSE
> > > related outputs of ethtool (or what ever tool you are using).
> > >
> > > I wont to use this examples to answer.
> >
> > On my side, I am not close to using sysfs. As we do all configurations
> > through ethtool I have assumed we should continue with ethtool.
>
> Yes, I agree. But it won't be possible to do it for all components.
>
> > I think we should set the port priority through ethtool.
>
> ack
>
> > but indeed the PSE power domain method get and set could be moved to
> > sysfs as it is not something relative to the port but to a group of
> > ports.
>
> I would prefer to have it in the for of devlink or use regulator netlink
> interface. But, we do not need to do this discussion right now.
If we want to report the method we should discuss it now. We shouldn't add
BUDGET_EVAL_STRAT uAPI to ethtool if we use another way to get and set the
method later.
We could also not report the method for now and assume the user knows it for
the two controllers currently supported.
> > Ethtool should still report the PSE power domain ID of a port to know
> > which domain the port is.
>
> Exactly.
>
> @Jakub, at current implementation stage, user need to know the domain
> id, because ports (and priorities) are grouped by the top level
> regulators (pse-regX in the regulator_summary), they are our top-level
> bottlenecks.
>
> HP and Cisco switch either use different PSE controllers, or just didn't
> exposed this nuance to the user. Let's assume, they have only one
> global power domain.
>
> So, in current patch set I would expect (not force :) ) implementation for
> following fields:
> - per port:
> - priority (valid within the power domain)
> - power reservation/allocation methods. First of all, because all
> already supported controllers have different implemented/default
> methods: microchip - dynamic, TI - static, regulator-pse - fixed (no
> classification is supported).
> At same time, in the future, we will need be able switch between
> (static or dynamic) and fixed for LLPD or manual configuration.
> Yes, at this point all ports show the same information and it seems
> to be duplicated.
> - power domain ID.
>
> @Jakub, did I answered you question, or missed the point? :)
>
> > @Oleksij here it is:
>
> Thank you!
>
> I do not expect it to be the primer user interface, but it can provide
> additional diagnostic information. I wonted to see how it is aligns
> with current ethtool UAPI implementation and if it possible to combine
> it for diagnostics.
>
> > # cat /sys/kernel/debug/regulator/regulator_summary
> > regulator use open bypass opmode voltage current
> > min max
> > ---------------------------------------------------------------------------------------
> > regulator-dummy 5 4 0 unknown 0mV 0mA
> > 0mV 0mV d00e0000.sata-target 1
> > 0mA 0mV 0mV d00e0000.sata-phy 1
> > 0mA 0mV 0mV d00e0000.sata-ahci 1
> > 0mA 0mV 0mV spi0.0-vcc 1
> > 0mA 0mV 0mV pse-reg
> > 1 4 0 unknown 0mV 0mA 0mV 0mV
>
> pse-regX should be attached to the main supply regulator for better full
> picture. And use different name to be better identified as PSE power domains
> with ID?
This is the regulator name set in the devicetree description so we can set
whatever we want.
Regards,
--
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-02-27 18:26 ` Kory Maincent
@ 2025-03-01 13:00 ` Oleksij Rempel
2025-03-03 13:40 ` Kory Maincent
0 siblings, 1 reply; 42+ messages in thread
From: Oleksij Rempel @ 2025-03-01 13:00 UTC (permalink / raw)
To: Kory Maincent
Cc: Jakub Kicinski, Andrew Lunn, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Thu, Feb 27, 2025 at 07:26:40PM +0100, Kory Maincent wrote:
> On Thu, 27 Feb 2025 17:40:42 +0100
> Oleksij Rempel <o.rempel@pengutronix.de> wrote:
>
> > On Thu, Feb 27, 2025 at 03:57:27PM +0100, Kory Maincent wrote:
> > > On Thu, 27 Feb 2025 08:40:25 +0100
> > > Oleksij Rempel <o.rempel@pengutronix.de> wrote:
> > >
> > > > On Wed, Feb 26, 2025 at 06:42:57PM -0800, Jakub Kicinski wrote:
> > [...]
> > [...]
> > [...]
> > > >
> > > > Ok, I see. @Köry, can you please provide regulator_summary with some
> > > > inlined comments to regulators related to the PSE components and PSE
> > > > related outputs of ethtool (or what ever tool you are using).
> > > >
> > > > I wont to use this examples to answer.
> > >
> > > On my side, I am not close to using sysfs. As we do all configurations
> > > through ethtool I have assumed we should continue with ethtool.
> >
> > Yes, I agree. But it won't be possible to do it for all components.
> >
> > > I think we should set the port priority through ethtool.
> >
> > ack
> >
> > > but indeed the PSE power domain method get and set could be moved to
> > > sysfs as it is not something relative to the port but to a group of
> > > ports.
> >
> > I would prefer to have it in the for of devlink or use regulator netlink
> > interface. But, we do not need to do this discussion right now.
>
> If we want to report the method we should discuss it now. We shouldn't add
> BUDGET_EVAL_STRAT uAPI to ethtool if we use another way to get and set the
> method later.
Ok, I assume we are talking about different things. I mean - not port
specific configurations and diagnostic, will have different interface.
BUDGET_EVAL_STRAT is port specific. HP and Cisco implement it as port
specific. PD692x0 Protocol manual describe it as port specific too:
3.3.6 Set BT Port Parameters
Bits [3..0]—BT port PM mode
0x0: The port power that is used for power management purposes is
dynamic (Iport x Vmain).
0x1: The port power that is used for power management purposes is port
TPPL_BT.
0x2: The port power that is used for power management purposes is
dynamic for non LLDP/CDP/Autoclass ports and TPPL_BT for LLDP/CDP/Autoclass ports.
0xF: Do not change settings.
> We could also not report the method for now and assume the user knows it for
> the two controllers currently supported.
On one side: it is not just status, but also active configuration. By
implementing the interface we may break default configuration and user
expectations.
On other side: PD692x0 seems to need more then just setting prios to
manage them correctly. For example power bank limits should be set,
otherwise internal firmware won't be able to perform budget calculations.
So, I assume, critical components are missing anyway.
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-03-01 13:00 ` Oleksij Rempel
@ 2025-03-03 13:40 ` Kory Maincent
2025-03-04 1:12 ` Jakub Kicinski
0 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-03-03 13:40 UTC (permalink / raw)
To: Oleksij Rempel
Cc: Jakub Kicinski, Andrew Lunn, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Sat, 1 Mar 2025 14:00:43 +0100
Oleksij Rempel <o.rempel@pengutronix.de> wrote:
> On Thu, Feb 27, 2025 at 07:26:40PM +0100, Kory Maincent wrote:
> > On Thu, 27 Feb 2025 17:40:42 +0100
> > Oleksij Rempel <o.rempel@pengutronix.de> wrote:
> > > I would prefer to have it in the for of devlink or use regulator netlink
> > > interface. But, we do not need to do this discussion right now.
> >
> > If we want to report the method we should discuss it now. We shouldn't add
> > BUDGET_EVAL_STRAT uAPI to ethtool if we use another way to get and set the
> > method later.
>
> Ok, I assume we are talking about different things. I mean - not port
> specific configurations and diagnostic, will have different interface.
>
> BUDGET_EVAL_STRAT is port specific. HP and Cisco implement it as port
> specific. PD692x0 Protocol manual describe it as port specific too:
> 3.3.6 Set BT Port Parameters
> Bits [3..0]—BT port PM mode
> 0x0: The port power that is used for power management purposes is
> dynamic (Iport x Vmain).
> 0x1: The port power that is used for power management purposes is port
> TPPL_BT.
> 0x2: The port power that is used for power management purposes is
> dynamic for non LLDP/CDP/Autoclass ports and TPPL_BT for
> LLDP/CDP/Autoclass ports. 0xF: Do not change settings.
I don't really understand how that can be port specific when the power budget is
per PD69208 manager. Maybe I am missing information here.
> > We could also not report the method for now and assume the user knows it for
> > the two controllers currently supported.
>
> On one side: it is not just status, but also active configuration. By
> implementing the interface we may break default configuration and user
> expectations.
Yes we should not implement the budget method get/set interface in this series.
> On other side: PD692x0 seems to need more then just setting prios to
> manage them correctly. For example power bank limits should be set,
> otherwise internal firmware won't be able to perform budget calculations.
Patch 8 is already configuring the power PD692x0 bank limit according to PSE
power domain budget.
> So, I assume, critical components are missing anyway.
As we are not supporting the budget method configured by the user in this
series, I agreed we should not add any uAPI related to it that could be broken
or confusing later.
I will remove it and send v6.
Regards,
--
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies
2025-03-03 13:40 ` Kory Maincent
@ 2025-03-04 1:12 ` Jakub Kicinski
0 siblings, 0 replies; 42+ messages in thread
From: Jakub Kicinski @ 2025-03-04 1:12 UTC (permalink / raw)
To: Kory Maincent
Cc: Oleksij Rempel, Andrew Lunn, David S. Miller, Eric Dumazet,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Mon, 3 Mar 2025 14:40:51 +0100 Kory Maincent wrote:
> > Ok, I assume we are talking about different things. I mean - not port
> > specific configurations and diagnostic, will have different interface.
> >
> > BUDGET_EVAL_STRAT is port specific. HP and Cisco implement it as port
> > specific. PD692x0 Protocol manual describe it as port specific too:
> > 3.3.6 Set BT Port Parameters
> > Bits [3..0]—BT port PM mode
> > 0x0: The port power that is used for power management purposes is
> > dynamic (Iport x Vmain).
> > 0x1: The port power that is used for power management purposes is port
> > TPPL_BT.
> > 0x2: The port power that is used for power management purposes is
> > dynamic for non LLDP/CDP/Autoclass ports and TPPL_BT for
> > LLDP/CDP/Autoclass ports. 0xF: Do not change settings.
>
> I don't really understand how that can be port specific when the power budget is
> per PD69208 manager. Maybe I am missing information here.
+1
> > So, I assume, critical components are missing anyway.
>
> As we are not supporting the budget method configured by the user in this
> series, I agreed we should not add any uAPI related to it that could be broken
> or confusing later.
>
> I will remove it and send v6.
v6 sounds like a good idea.
^ permalink raw reply [flat|nested] 42+ messages in thread
* [PATCH net-next v5 07/12] net: ethtool: Add PSE new budget evaluation strategy support feature
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
` (5 preceding siblings ...)
2025-02-18 16:19 ` [PATCH net-next v5 06/12] net: pse-pd: Add support for budget evaluation strategies Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-21 13:49 ` Oleksij Rempel
2025-02-18 16:19 ` [PATCH net-next v5 08/12] net: pse-pd: pd692x0: Add support for PSE PI priority feature Kory Maincent
` (4 subsequent siblings)
11 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
This patch expands the status information provided by ethtool for PSE c33
with current port priority and max port priority. It also adds a call to
pse_ethtool_set_prio() to configure the PSE port priority.
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
Change in v4:
- Remove disconnection policy features.
- Rename port priority to budget evaluation strategy.
Change in v3:
- Add disconnection policy.
Change in v2:
- Improve port priority documentation.
- Add port priority modes.
---
Documentation/netlink/specs/ethtool.yaml | 16 ++++++
Documentation/networking/ethtool-netlink.rst | 67 ++++++++++++++++++++++++++
include/uapi/linux/ethtool_netlink_generated.h | 3 ++
net/ethtool/pse-pd.c | 26 ++++++++++
4 files changed, 112 insertions(+)
diff --git a/Documentation/netlink/specs/ethtool.yaml b/Documentation/netlink/specs/ethtool.yaml
index 9b171c2dd1a4..387eff675d7f 100644
--- a/Documentation/netlink/specs/ethtool.yaml
+++ b/Documentation/netlink/specs/ethtool.yaml
@@ -1370,6 +1370,18 @@ attribute-sets:
name: pse-pw-d-id
type: u32
name-prefix: ethtool-a-
+ -
+ name: pse-budget-eval-strat
+ type: u32
+ name-prefix: ethtool-a-
+ -
+ name: pse-prio-max
+ type: u32
+ name-prefix: ethtool-a-
+ -
+ name: pse-prio
+ type: u32
+ name-prefix: ethtool-a-
-
name: rss
attr-cnt-name: __ethtool-a-rss-cnt
@@ -2195,6 +2207,9 @@ operations:
- c33-pse-avail-pw-limit
- c33-pse-pw-limit-ranges
- pse-pw-d-id
+ - pse-budget-eval-strat
+ - pse-prio-max
+ - pse-prio
dump: *pse-get-op
-
name: pse-set
@@ -2209,6 +2224,7 @@ operations:
- podl-pse-admin-control
- c33-pse-admin-control
- c33-pse-avail-pw-limit
+ - pse-prio
-
name: rss-get
doc: Get RSS params.
diff --git a/Documentation/networking/ethtool-netlink.rst b/Documentation/networking/ethtool-netlink.rst
index dc3f6afc55a4..148323417fc9 100644
--- a/Documentation/networking/ethtool-netlink.rst
+++ b/Documentation/networking/ethtool-netlink.rst
@@ -1790,6 +1790,12 @@ Kernel response contents:
``ETHTOOL_A_C33_PSE_PW_LIMIT_RANGES`` nested Supported power limit
configuration ranges.
``ETHTOOL_A_PSE_PW_D_ID`` u32 Index of the PSE power domain
+ ``ETHTOOL_A_C33_PSE_BUDGET_EVAL_STRAT`` u32 Budget evaluation strategy
+ of the PSE
+ ``ETHTOOL_A_C33_PSE_PRIO_MAX`` u32 Priority maximum configurable
+ on the PoE PSE
+ ``ETHTOOL_A_C33_PSE_PRIO`` u32 Priority of the PoE PSE
+ currently configured
========================================== ====== =============================
When set, the optional ``ETHTOOL_A_PODL_PSE_ADMIN_STATE`` attribute identifies
@@ -1866,6 +1872,51 @@ equal.
The ``ETHTOOL_A_PSE_PW_D_ID`` attribute identifies the index of PSE power
domain.
+When set, the optional ``ETHTOOL_A_C33_PSE_PRIO_SUPP_MODES`` attribute
+identifies the priority mode supported by the C33 PSE.
+When set, the optional ``ETHTOOL_A_C33_PSE_BUDGET_EVAL_STRAT`` attributes is used to
+identifies the currently configured C33 PSE budget evaluation strategy.
+The available strategies are:
+
+1. Disabled:
+
+ In this mode, the port is excluded from active budget evaluation. It
+ allows the port to violate the budget and is intended primarily for testing
+ purposes.
+
+2. Static Method:
+
+ This method involves distributing power based on PD classification. It’s
+ straightforward and stable, with the PSE core keeping track of the budget
+ and subtracting the power requested by each PD’s class. This is the
+ safest option and should be used by default.
+
+ Advantages: Every PD gets its promised power at any time, which guarantees
+ reliability.
+
+ Disadvantages: PD classification steps are large, meaning devices request
+ much more power than they actually need. As a result, the power supply may
+ only operate at, say, 50% capacity, which is inefficient and wastes money.
+
+3. Dynamic Method:
+
+ This method monitors the current consumption per port and subtracts it from
+ the available power budget. When the budget is exceeded, lower-priority
+ ports are shut down. This method is managed by the PSE controller itself.
+
+ Advantages: This method optimizes resource utilization, saving costs.
+
+ Disadvantages: Low-priority devices may experience instability.
+
+.. kernel-doc:: include/uapi/linux/ethtool.h
+ :identifiers: ethtool_pse_budget_eval_strategies
+
+When set, the optional ``ETHTOOL_A_C33_PSE_PRIO_MAX`` attribute identifies
+the C33 PSE maximum priority value.
+When set, the optional ``ETHTOOL_A_C33_PSE_PRIO`` attributes is used to
+identifies the currently configured C33 PSE priority.
+For a description of PSE priority attributes, see ``PSE_SET``.
+
PSE_SET
=======
@@ -1879,6 +1930,8 @@ Request contents:
``ETHTOOL_A_C33_PSE_ADMIN_CONTROL`` u32 Control PSE Admin state
``ETHTOOL_A_C33_PSE_AVAIL_PWR_LIMIT`` u32 Control PoE PSE available
power limit
+ ``ETHTOOL_A_C33_PSE_PRIO`` u32 Control priority of the
+ PoE PSE
====================================== ====== =============================
When set, the optional ``ETHTOOL_A_PODL_PSE_ADMIN_CONTROL`` attribute is used
@@ -1901,6 +1954,20 @@ various existing products that document power consumption in watts rather than
classes. If power limit configuration based on classes is needed, the
conversion can be done in user space, for example by ethtool.
+When set, the optional ``ETHTOOL_A_C33_PSE_PRIO`` attributes is used to
+control the C33 PSE priority. Allowed priority value are between zero
+and the value of ``ETHTOOL_A_C33_PSE_PRIO_MAX`` attribute.
+
+A lower value indicates a higher priority, meaning that a priority value
+of 0 corresponds to the highest port priority.
+Port priority serves two functions:
+
+ - Power-up Order: After a reset, ports are powered up in order of their
+ priority from highest to lowest. Ports with higher priority
+ (lower values) power up first.
+ - Shutdown Order: When the power budget is exceeded, ports with lower
+ priority (higher values) are turned off first.
+
PSE_NTF
=======
diff --git a/include/uapi/linux/ethtool_netlink_generated.h b/include/uapi/linux/ethtool_netlink_generated.h
index 919435c1a924..5e65b73043c6 100644
--- a/include/uapi/linux/ethtool_netlink_generated.h
+++ b/include/uapi/linux/ethtool_netlink_generated.h
@@ -634,6 +634,9 @@ enum {
ETHTOOL_A_C33_PSE_AVAIL_PW_LIMIT,
ETHTOOL_A_C33_PSE_PW_LIMIT_RANGES,
ETHTOOL_A_PSE_PW_D_ID,
+ ETHTOOL_A_PSE_BUDGET_EVAL_STRAT,
+ ETHTOOL_A_PSE_PRIO_MAX,
+ ETHTOOL_A_PSE_PRIO,
__ETHTOOL_A_PSE_CNT,
ETHTOOL_A_PSE_MAX = (__ETHTOOL_A_PSE_CNT - 1)
diff --git a/net/ethtool/pse-pd.c b/net/ethtool/pse-pd.c
index eae5b7894613..70ac11810c2a 100644
--- a/net/ethtool/pse-pd.c
+++ b/net/ethtool/pse-pd.c
@@ -112,6 +112,12 @@ static int pse_reply_size(const struct ethnl_req_info *req_base,
len += st->c33_pw_limit_nb_ranges *
(nla_total_size(0) +
nla_total_size(sizeof(u32)) * 2);
+ if (st->budget_eval_strategy)
+ /* _PSE_BUDGET_EVAL_STRAT */
+ len += nla_total_size(sizeof(u32));
+ if (st->prio_max)
+ /* _PSE_PRIO_MAX + _PSE_PRIO */
+ len += nla_total_size(sizeof(u32)) * 2;
return len;
}
@@ -206,6 +212,16 @@ static int pse_fill_reply(struct sk_buff *skb,
pse_put_pw_limit_ranges(skb, st))
return -EMSGSIZE;
+ if (st->budget_eval_strategy > 0 &&
+ nla_put_u32(skb, ETHTOOL_A_PSE_BUDGET_EVAL_STRAT,
+ st->budget_eval_strategy))
+ return -EMSGSIZE;
+
+ if (st->prio_max > 0 &&
+ (nla_put_u32(skb, ETHTOOL_A_PSE_PRIO_MAX, st->prio_max) ||
+ nla_put_u32(skb, ETHTOOL_A_PSE_PRIO, st->prio)))
+ return -EMSGSIZE;
+
return 0;
}
@@ -227,6 +243,7 @@ const struct nla_policy ethnl_pse_set_policy[ETHTOOL_A_PSE_MAX + 1] = {
NLA_POLICY_RANGE(NLA_U32, ETHTOOL_C33_PSE_ADMIN_STATE_DISABLED,
ETHTOOL_C33_PSE_ADMIN_STATE_ENABLED),
[ETHTOOL_A_C33_PSE_AVAIL_PW_LIMIT] = { .type = NLA_U32 },
+ [ETHTOOL_A_PSE_PRIO] = { .type = NLA_U32 },
};
static int
@@ -275,6 +292,15 @@ ethnl_set_pse(struct ethnl_req_info *req_info, struct genl_info *info)
if (ret)
return ret;
+ if (tb[ETHTOOL_A_PSE_PRIO]) {
+ unsigned int prio;
+
+ prio = nla_get_u32(tb[ETHTOOL_A_PSE_PRIO]);
+ ret = pse_ethtool_set_prio(phydev->psec, info->extack, prio);
+ if (ret)
+ return ret;
+ }
+
if (tb[ETHTOOL_A_C33_PSE_AVAIL_PW_LIMIT]) {
unsigned int pw_limit;
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 07/12] net: ethtool: Add PSE new budget evaluation strategy support feature
2025-02-18 16:19 ` [PATCH net-next v5 07/12] net: ethtool: Add PSE new budget evaluation strategy support feature Kory Maincent
@ 2025-02-21 13:49 ` Oleksij Rempel
2025-02-24 13:13 ` Kory Maincent
0 siblings, 1 reply; 42+ messages in thread
From: Oleksij Rempel @ 2025-02-21 13:49 UTC (permalink / raw)
To: Kory Maincent
Cc: Andrew Lunn, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
Hi Kory,
On Tue, Feb 18, 2025 at 05:19:11PM +0100, Kory Maincent wrote:
> From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
>
> This patch expands the status information provided by ethtool for PSE c33
> with current port priority and max port priority. It also adds a call to
> pse_ethtool_set_prio() to configure the PSE port priority.
Thank you! Here are some comments...
> --- a/Documentation/networking/ethtool-netlink.rst
> +++ b/Documentation/networking/ethtool-netlink.rst
> @@ -1790,6 +1790,12 @@ Kernel response contents:
> ``ETHTOOL_A_C33_PSE_PW_LIMIT_RANGES`` nested Supported power limit
> configuration ranges.
> ``ETHTOOL_A_PSE_PW_D_ID`` u32 Index of the PSE power domain
> + ``ETHTOOL_A_C33_PSE_BUDGET_EVAL_STRAT`` u32 Budget evaluation strategy
> + of the PSE
> + ``ETHTOOL_A_C33_PSE_PRIO_MAX`` u32 Priority maximum configurable
> + on the PoE PSE
> + ``ETHTOOL_A_C33_PSE_PRIO`` u32 Priority of the PoE PSE
> + currently configured
Please remove _C33_ from these fields, as they are not specific to Clause 33.
> ========================================== ====== =============================
>
> When set, the optional ``ETHTOOL_A_PODL_PSE_ADMIN_STATE`` attribute identifies
> @@ -1866,6 +1872,51 @@ equal.
> The ``ETHTOOL_A_PSE_PW_D_ID`` attribute identifies the index of PSE power
> domain.
>
> +When set, the optional ``ETHTOOL_A_C33_PSE_PRIO_SUPP_MODES`` attribute
> +identifies the priority mode supported by the C33 PSE.
> +When set, the optional ``ETHTOOL_A_C33_PSE_BUDGET_EVAL_STRAT`` attributes is used to
> +identifies the currently configured C33 PSE budget evaluation strategy.
> +The available strategies are:
> +
> +1. Disabled:
> +
> + In this mode, the port is excluded from active budget evaluation. It
> + allows the port to violate the budget and is intended primarily for testing
> + purposes.
> +
> +2. Static Method:
> +
> + This method involves distributing power based on PD classification. It’s
> + straightforward and stable, with the PSE core keeping track of the budget
> + and subtracting the power requested by each PD’s class. This is the
> + safest option and should be used by default.
> +
> + Advantages: Every PD gets its promised power at any time, which guarantees
> + reliability.
> +
> + Disadvantages: PD classification steps are large, meaning devices request
> + much more power than they actually need. As a result, the power supply may
> + only operate at, say, 50% capacity, which is inefficient and wastes money.
> +
> +3. Dynamic Method:
> +
> + This method monitors the current consumption per port and subtracts it from
> + the available power budget. When the budget is exceeded, lower-priority
> + ports are shut down. This method is managed by the PSE controller itself.
> +
> + Advantages: This method optimizes resource utilization, saving costs.
> +
> + Disadvantages: Low-priority devices may experience instability.
> +
> +.. kernel-doc:: include/uapi/linux/ethtool.h
> + :identifiers: ethtool_pse_budget_eval_strategies
> +
> +When set, the optional ``ETHTOOL_A_C33_PSE_PRIO_MAX`` attribute identifies
> +the C33 PSE maximum priority value.
> +When set, the optional ``ETHTOOL_A_C33_PSE_PRIO`` attributes is used to
> +identifies the currently configured C33 PSE priority.
> +For a description of PSE priority attributes, see ``PSE_SET``.
> +
> PSE_SET
> =======
>
> @@ -1879,6 +1930,8 @@ Request contents:
> ``ETHTOOL_A_C33_PSE_ADMIN_CONTROL`` u32 Control PSE Admin state
> ``ETHTOOL_A_C33_PSE_AVAIL_PWR_LIMIT`` u32 Control PoE PSE available
> power limit
> + ``ETHTOOL_A_C33_PSE_PRIO`` u32 Control priority of the
> + PoE PSE
Please remove _C33_ from these field, as they are not specific to
Clause 33.
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 42+ messages in thread
* Re: [PATCH net-next v5 07/12] net: ethtool: Add PSE new budget evaluation strategy support feature
2025-02-21 13:49 ` Oleksij Rempel
@ 2025-02-24 13:13 ` Kory Maincent
0 siblings, 0 replies; 42+ messages in thread
From: Kory Maincent @ 2025-02-24 13:13 UTC (permalink / raw)
To: Oleksij Rempel
Cc: Andrew Lunn, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Jonathan Corbet, Donald Hunter, Rob Herring,
Andrew Lunn, Simon Horman, Heiner Kallweit, Russell King,
Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni, netdev,
linux-doc, Kyle Swenson, Dent Project, kernel, Maxime Chevallier,
devicetree, linux-kernel
On Fri, 21 Feb 2025 14:49:21 +0100
Oleksij Rempel <o.rempel@pengutronix.de> wrote:
> Hi Kory,
>
> On Tue, Feb 18, 2025 at 05:19:11PM +0100, Kory Maincent wrote:
> > From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
> >
> > This patch expands the status information provided by ethtool for PSE c33
> > with current port priority and max port priority. It also adds a call to
> > pse_ethtool_set_prio() to configure the PSE port priority.
>
> Thank you! Here are some comments...
>
> > --- a/Documentation/networking/ethtool-netlink.rst
> > +++ b/Documentation/networking/ethtool-netlink.rst
> > @@ -1790,6 +1790,12 @@ Kernel response contents:
> > ``ETHTOOL_A_C33_PSE_PW_LIMIT_RANGES`` nested Supported power limit
> > configuration ranges.
> > ``ETHTOOL_A_PSE_PW_D_ID`` u32 Index of the PSE
> > power domain
> > + ``ETHTOOL_A_C33_PSE_BUDGET_EVAL_STRAT`` u32 Budget evaluation
> > strategy
> > + of the PSE
> > + ``ETHTOOL_A_C33_PSE_PRIO_MAX`` u32 Priority maximum
> > configurable
> > + on the PoE PSE
> > + ``ETHTOOL_A_C33_PSE_PRIO`` u32 Priority of the PoE
> > PSE
> > + currently configured
> >
>
> Please remove _C33_ from these fields, as they are not specific to Clause 33.
Oops, forgot to update the documentation accordingly. Thanks for spotting it.
--
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com
^ permalink raw reply [flat|nested] 42+ messages in thread
* [PATCH net-next v5 08/12] net: pse-pd: pd692x0: Add support for PSE PI priority feature
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
` (6 preceding siblings ...)
2025-02-18 16:19 ` [PATCH net-next v5 07/12] net: ethtool: Add PSE new budget evaluation strategy support feature Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-18 16:19 ` [PATCH net-next v5 09/12] net: pse-pd: pd692x0: Add support for controller and manager power supplies Kory Maincent
` (3 subsequent siblings)
11 siblings, 0 replies; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
This patch extends the PSE callbacks by adding support for the newly
introduced pi_set_prio() callback, enabling the configuration of PSE PI
priorities. The current port priority is now also included in the status
information returned to users.
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
Changes in v3:
- New patch
---
drivers/net/pse-pd/pd692x0.c | 205 +++++++++++++++++++++++++++++++++++++++++++
1 file changed, 205 insertions(+)
diff --git a/drivers/net/pse-pd/pd692x0.c b/drivers/net/pse-pd/pd692x0.c
index 7d60a714ca53..44ded2aa6fca 100644
--- a/drivers/net/pse-pd/pd692x0.c
+++ b/drivers/net/pse-pd/pd692x0.c
@@ -12,6 +12,8 @@
#include <linux/of.h>
#include <linux/platform_device.h>
#include <linux/pse-pd/pse.h>
+#include <linux/regulator/driver.h>
+#include <linux/regulator/machine.h>
#define PD692X0_PSE_NAME "pd692x0_pse"
@@ -76,6 +78,8 @@ enum {
PD692X0_MSG_GET_PORT_CLASS,
PD692X0_MSG_GET_PORT_MEAS,
PD692X0_MSG_GET_PORT_PARAM,
+ PD692X0_MSG_GET_POWER_BANK,
+ PD692X0_MSG_SET_POWER_BANK,
/* add new message above here */
PD692X0_MSG_CNT
@@ -95,6 +99,8 @@ struct pd692x0_priv {
unsigned long last_cmd_key_time;
enum ethtool_c33_pse_admin_state admin_state[PD692X0_MAX_PIS];
+ struct regulator_dev *manager_reg[PD692X0_MAX_MANAGERS];
+ int manager_pw_budget[PD692X0_MAX_MANAGERS];
};
/* Template list of communication messages. The non-null bytes defined here
@@ -170,6 +176,16 @@ static const struct pd692x0_msg pd692x0_msg_template_list[PD692X0_MSG_CNT] = {
.data = {0x4e, 0x4e, 0x4e, 0x4e,
0x4e, 0x4e, 0x4e, 0x4e},
},
+ [PD692X0_MSG_GET_POWER_BANK] = {
+ .key = PD692X0_KEY_REQ,
+ .sub = {0x07, 0x0b, 0x57},
+ .data = { 0, 0x4e, 0x4e, 0x4e,
+ 0x4e, 0x4e, 0x4e, 0x4e},
+ },
+ [PD692X0_MSG_SET_POWER_BANK] = {
+ .key = PD692X0_KEY_CMD,
+ .sub = {0x07, 0x0b, 0x57},
+ },
};
static u8 pd692x0_build_msg(struct pd692x0_msg *msg, u8 echo)
@@ -739,6 +755,29 @@ pd692x0_pi_get_actual_pw(struct pse_controller_dev *pcdev, int id)
return (buf.data[0] << 4 | buf.data[1]) * 100;
}
+static int
+pd692x0_pi_get_prio(struct pse_controller_dev *pcdev, int id)
+{
+ struct pd692x0_priv *priv = to_pd692x0_priv(pcdev);
+ struct pd692x0_msg msg, buf = {0};
+ int ret;
+
+ ret = pd692x0_fw_unavailable(priv);
+ if (ret)
+ return ret;
+
+ msg = pd692x0_msg_template_list[PD692X0_MSG_GET_PORT_PARAM];
+ msg.sub[2] = id;
+ ret = pd692x0_sendrecv_msg(priv, &msg, &buf);
+ if (ret < 0)
+ return ret;
+ if (buf.data[2] < 1 || 3 < buf.data[2])
+ return -ERANGE;
+
+ /* PSE core priority start at 0 */
+ return buf.data[2] - 1;
+}
+
static struct pd692x0_msg_ver pd692x0_get_sw_version(struct pd692x0_priv *priv)
{
struct device *dev = &priv->client->dev;
@@ -766,6 +805,7 @@ static struct pd692x0_msg_ver pd692x0_get_sw_version(struct pd692x0_priv *priv)
struct pd692x0_manager {
struct device_node *port_node[PD692X0_MAX_MANAGER_PORTS];
+ struct device_node *node;
int nports;
};
@@ -857,6 +897,8 @@ pd692x0_of_get_managers(struct pd692x0_priv *priv,
if (ret)
goto out;
+ of_node_get(node);
+ manager[manager_id].node = node;
nmanagers++;
}
@@ -869,6 +911,8 @@ pd692x0_of_get_managers(struct pd692x0_priv *priv,
of_node_put(manager[i].port_node[j]);
manager[i].port_node[j] = NULL;
}
+ of_node_put(manager[i].node);
+ manager[i].node = NULL;
}
of_node_put(node);
@@ -876,6 +920,130 @@ pd692x0_of_get_managers(struct pd692x0_priv *priv,
return ret;
}
+static const struct regulator_ops dummy_ops;
+
+static struct regulator_dev *
+pd692x0_register_manager_regulator(struct device *dev, char *reg_name,
+ struct device_node *node)
+{
+ struct regulator_init_data *rinit_data;
+ struct regulator_config rconfig = {0};
+ struct regulator_desc *rdesc;
+ struct regulator_dev *rdev;
+
+ rinit_data = devm_kzalloc(dev, sizeof(*rinit_data),
+ GFP_KERNEL);
+ if (!rinit_data)
+ return ERR_PTR(-ENOMEM);
+
+ rdesc = devm_kzalloc(dev, sizeof(*rdesc), GFP_KERNEL);
+ if (!rdesc)
+ return ERR_PTR(-ENOMEM);
+
+ rdesc->name = reg_name;
+ rdesc->type = REGULATOR_VOLTAGE;
+ rdesc->ops = &dummy_ops;
+ rdesc->owner = THIS_MODULE;
+
+ rinit_data->supply_regulator = "vmain";
+
+ rconfig.dev = dev;
+ rconfig.init_data = rinit_data;
+ rconfig.of_node = node;
+
+ rdev = devm_regulator_register(dev, rdesc, &rconfig);
+ if (IS_ERR(rdev)) {
+ dev_err_probe(dev, PTR_ERR(rdev),
+ "Failed to register regulator\n");
+ return rdev;
+ }
+
+ return rdev;
+}
+
+static int
+pd692x0_register_managers_regulator(struct pd692x0_priv *priv,
+ const struct pd692x0_manager *manager,
+ int nmanagers)
+{
+ struct device *dev = &priv->client->dev;
+ size_t reg_name_len;
+ int i;
+
+ /* Each regulator name len is dev name + 12 char +
+ * int max digit number (10) + 1
+ */
+ reg_name_len = strlen(dev_name(dev)) + 23;
+
+ for (i = 0; i < nmanagers; i++) {
+ struct regulator_dev *rdev;
+ char *reg_name;
+
+ reg_name = devm_kzalloc(dev, reg_name_len, GFP_KERNEL);
+ if (!reg_name)
+ return -ENOMEM;
+ snprintf(reg_name, 26, "pse-%s-manager%d", dev_name(dev), i);
+ rdev = pd692x0_register_manager_regulator(dev, reg_name,
+ manager[i].node);
+ if (IS_ERR(rdev))
+ return PTR_ERR(rdev);
+
+ priv->manager_reg[i] = rdev;
+ }
+
+ return 0;
+}
+
+static int
+pd692x0_conf_manager_power_budget(struct pd692x0_priv *priv, int id, int pw)
+{
+ struct pd692x0_msg msg, buf;
+ int ret, pw_mW = pw / 1000;
+
+ msg = pd692x0_msg_template_list[PD692X0_MSG_GET_POWER_BANK];
+ msg.data[0] = id;
+ ret = pd692x0_sendrecv_msg(priv, &msg, &buf);
+ if (ret < 0)
+ return ret;
+
+ msg = pd692x0_msg_template_list[PD692X0_MSG_SET_POWER_BANK];
+ msg.data[0] = id;
+ msg.data[1] = pw_mW >> 8;
+ msg.data[2] = pw_mW & 0xff;
+ msg.data[3] = buf.sub[2];
+ msg.data[4] = buf.data[0];
+ msg.data[5] = buf.data[1];
+ msg.data[6] = buf.data[2];
+ msg.data[7] = buf.data[3];
+ return pd692x0_sendrecv_msg(priv, &msg, &buf);
+}
+
+static int
+pd692x0_configure_managers(struct pd692x0_priv *priv, int nmanagers)
+{
+ int i, ret;
+
+ for (i = 0; i < nmanagers; i++) {
+ struct regulator *supply = priv->manager_reg[i]->supply;
+ int pw_budget;
+
+ pw_budget = regulator_get_unclaimed_power_budget(supply);
+ /* Max power budget per manager */
+ if (pw_budget > 6000000)
+ pw_budget = 6000000;
+ ret = regulator_request_power_budget(supply, pw_budget);
+ if (ret < 0)
+ return ret;
+
+ priv->manager_pw_budget[i] = pw_budget;
+ ret = pd692x0_conf_manager_power_budget(priv, i, pw_budget);
+ if (ret < 0)
+ return ret;
+ }
+
+ return 0;
+}
+
static int
pd692x0_set_port_matrix(const struct pse_pi_pairset *pairset,
const struct pd692x0_manager *manager,
@@ -998,6 +1166,14 @@ static int pd692x0_setup_pi_matrix(struct pse_controller_dev *pcdev)
return ret;
nmanagers = ret;
+ ret = pd692x0_register_managers_regulator(priv, manager, nmanagers);
+ if (ret)
+ goto out;
+
+ ret = pd692x0_configure_managers(priv, nmanagers);
+ if (ret)
+ goto out;
+
ret = pd692x0_set_ports_matrix(priv, manager, nmanagers, port_matrix);
if (ret)
goto out;
@@ -1008,8 +1184,14 @@ static int pd692x0_setup_pi_matrix(struct pse_controller_dev *pcdev)
out:
for (i = 0; i < nmanagers; i++) {
+ struct regulator *supply = priv->manager_reg[i]->supply;
+
+ regulator_free_power_budget(supply,
+ priv->manager_pw_budget[i]);
+
for (j = 0; j < manager[i].nports; j++)
of_node_put(manager[i].port_node[j]);
+ of_node_put(manager[i].node);
}
return ret;
}
@@ -1071,6 +1253,25 @@ static int pd692x0_pi_set_pw_limit(struct pse_controller_dev *pcdev,
return pd692x0_sendrecv_msg(priv, &msg, &buf);
}
+static int pd692x0_pi_set_prio(struct pse_controller_dev *pcdev, int id,
+ unsigned int prio)
+{
+ struct pd692x0_priv *priv = to_pd692x0_priv(pcdev);
+ struct pd692x0_msg msg, buf = {0};
+ int ret;
+
+ ret = pd692x0_fw_unavailable(priv);
+ if (ret)
+ return ret;
+
+ msg = pd692x0_msg_template_list[PD692X0_MSG_SET_PORT_PARAM];
+ msg.sub[2] = id;
+ /* Controller priority from 1 to 3 */
+ msg.data[4] = prio + 1;
+
+ return pd692x0_sendrecv_msg(priv, &msg, &buf);
+}
+
static const struct pse_controller_ops pd692x0_ops = {
.setup_pi_matrix = pd692x0_setup_pi_matrix,
.pi_get_admin_state = pd692x0_pi_get_admin_state,
@@ -1084,6 +1285,8 @@ static const struct pse_controller_ops pd692x0_ops = {
.pi_get_pw_limit = pd692x0_pi_get_pw_limit,
.pi_set_pw_limit = pd692x0_pi_set_pw_limit,
.pi_get_pw_limit_ranges = pd692x0_pi_get_pw_limit_ranges,
+ .pi_get_prio = pd692x0_pi_get_prio,
+ .pi_set_prio = pd692x0_pi_set_prio,
};
#define PD692X0_FW_LINE_MAX_SZ 0xff
@@ -1500,6 +1703,8 @@ static int pd692x0_i2c_probe(struct i2c_client *client)
priv->pcdev.ops = &pd692x0_ops;
priv->pcdev.dev = dev;
priv->pcdev.types = ETHTOOL_PSE_C33;
+ priv->pcdev.supp_budget_eval_strategies = ETHTOOL_PSE_BUDGET_EVAL_STRAT_DYNAMIC;
+ priv->pcdev.pis_prio_max = 2;
ret = devm_pse_controller_register(dev, &priv->pcdev);
if (ret)
return dev_err_probe(dev, ret,
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* [PATCH net-next v5 09/12] net: pse-pd: pd692x0: Add support for controller and manager power supplies
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
` (7 preceding siblings ...)
2025-02-18 16:19 ` [PATCH net-next v5 08/12] net: pse-pd: pd692x0: Add support for PSE PI priority feature Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-24 12:42 ` Maxime Chevallier
2025-02-18 16:19 ` [PATCH net-next v5 10/12] dt-bindings: net: pse-pd: microchip,pd692x0: Add manager regulator supply Kory Maincent
` (2 subsequent siblings)
11 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
Add support for managing the VDD and VDDA power supplies for the PD692x0
PSE controller, as well as the VAUX5 and VAUX3P3 power supplies for the
PD6920x PSE managers.
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
Changes in v5:
- New patch
---
drivers/net/pse-pd/pd692x0.c | 20 ++++++++++++++++++++
1 file changed, 20 insertions(+)
diff --git a/drivers/net/pse-pd/pd692x0.c b/drivers/net/pse-pd/pd692x0.c
index 44ded2aa6fca..c9fa60b314ce 100644
--- a/drivers/net/pse-pd/pd692x0.c
+++ b/drivers/net/pse-pd/pd692x0.c
@@ -976,8 +976,10 @@ pd692x0_register_managers_regulator(struct pd692x0_priv *priv,
reg_name_len = strlen(dev_name(dev)) + 23;
for (i = 0; i < nmanagers; i++) {
+ static const char * const regulators[] = { "vaux5", "vaux3p3" };
struct regulator_dev *rdev;
char *reg_name;
+ int ret;
reg_name = devm_kzalloc(dev, reg_name_len, GFP_KERNEL);
if (!reg_name)
@@ -988,6 +990,17 @@ pd692x0_register_managers_regulator(struct pd692x0_priv *priv,
if (IS_ERR(rdev))
return PTR_ERR(rdev);
+ /* VMAIN is described as main supply for the manager.
+ * Add other VAUX power supplies and link them to the
+ * virtual device rdev->dev.
+ */
+ ret = devm_regulator_bulk_get_enable(&rdev->dev,
+ ARRAY_SIZE(regulators),
+ regulators);
+ if (ret)
+ return dev_err_probe(&rdev->dev, ret,
+ "Failed to enable regulators\n");
+
priv->manager_reg[i] = rdev;
}
@@ -1640,6 +1653,7 @@ static const struct fw_upload_ops pd692x0_fw_ops = {
static int pd692x0_i2c_probe(struct i2c_client *client)
{
+ static const char * const regulators[] = { "vdd", "vdda" };
struct pd692x0_msg msg, buf = {0}, zero = {0};
struct device *dev = &client->dev;
struct pd692x0_msg_ver ver;
@@ -1647,6 +1661,12 @@ static int pd692x0_i2c_probe(struct i2c_client *client)
struct fw_upload *fwl;
int ret;
+ ret = devm_regulator_bulk_get_enable(dev, ARRAY_SIZE(regulators),
+ regulators);
+ if (ret)
+ return dev_err_probe(dev, ret,
+ "Failed to enable regulators\n");
+
if (!i2c_check_functionality(client->adapter, I2C_FUNC_I2C)) {
dev_err(dev, "i2c check functionality failed\n");
return -ENXIO;
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 09/12] net: pse-pd: pd692x0: Add support for controller and manager power supplies
2025-02-18 16:19 ` [PATCH net-next v5 09/12] net: pse-pd: pd692x0: Add support for controller and manager power supplies Kory Maincent
@ 2025-02-24 12:42 ` Maxime Chevallier
2025-02-24 12:49 ` Russell King (Oracle)
0 siblings, 1 reply; 42+ messages in thread
From: Maxime Chevallier @ 2025-02-24 12:42 UTC (permalink / raw)
To: Kory Maincent
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni,
netdev, linux-doc, Kyle Swenson, Dent Project, kernel, devicetree,
linux-kernel
Hi Köry,
On Tue, 18 Feb 2025 17:19:13 +0100
Kory Maincent <kory.maincent@bootlin.com> wrote:
> From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
>
> Add support for managing the VDD and VDDA power supplies for the PD692x0
> PSE controller, as well as the VAUX5 and VAUX3P3 power supplies for the
> PD6920x PSE managers.
>
> Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
> ---
>
> Changes in v5:
> - New patch
> ---
> drivers/net/pse-pd/pd692x0.c | 20 ++++++++++++++++++++
> 1 file changed, 20 insertions(+)
>
> diff --git a/drivers/net/pse-pd/pd692x0.c b/drivers/net/pse-pd/pd692x0.c
> index 44ded2aa6fca..c9fa60b314ce 100644
> --- a/drivers/net/pse-pd/pd692x0.c
> +++ b/drivers/net/pse-pd/pd692x0.c
> @@ -976,8 +976,10 @@ pd692x0_register_managers_regulator(struct pd692x0_priv *priv,
> reg_name_len = strlen(dev_name(dev)) + 23;
>
> for (i = 0; i < nmanagers; i++) {
> + static const char * const regulators[] = { "vaux5", "vaux3p3" };
Looks like the 'static' is not needed here :)
> struct regulator_dev *rdev;
> char *reg_name;
> + int ret;
>
> reg_name = devm_kzalloc(dev, reg_name_len, GFP_KERNEL);
> if (!reg_name)
> @@ -988,6 +990,17 @@ pd692x0_register_managers_regulator(struct pd692x0_priv *priv,
> if (IS_ERR(rdev))
> return PTR_ERR(rdev);
>
> + /* VMAIN is described as main supply for the manager.
> + * Add other VAUX power supplies and link them to the
> + * virtual device rdev->dev.
> + */
> + ret = devm_regulator_bulk_get_enable(&rdev->dev,
> + ARRAY_SIZE(regulators),
> + regulators);
> + if (ret)
> + return dev_err_probe(&rdev->dev, ret,
> + "Failed to enable regulators\n");
> +
> priv->manager_reg[i] = rdev;
> }
>
> @@ -1640,6 +1653,7 @@ static const struct fw_upload_ops pd692x0_fw_ops = {
>
> static int pd692x0_i2c_probe(struct i2c_client *client)
> {
> + static const char * const regulators[] = { "vdd", "vdda" };
And here as well
Thanks,
Maxime
^ permalink raw reply [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 09/12] net: pse-pd: pd692x0: Add support for controller and manager power supplies
2025-02-24 12:42 ` Maxime Chevallier
@ 2025-02-24 12:49 ` Russell King (Oracle)
2025-02-24 13:17 ` Maxime Chevallier
0 siblings, 1 reply; 42+ messages in thread
From: Russell King (Oracle) @ 2025-02-24 12:49 UTC (permalink / raw)
To: Maxime Chevallier
Cc: Kory Maincent, Andrew Lunn, Oleksij Rempel, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Jonathan Corbet,
Donald Hunter, Rob Herring, Andrew Lunn, Simon Horman,
Heiner Kallweit, Krzysztof Kozlowski, Conor Dooley,
Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, devicetree, linux-kernel
On Mon, Feb 24, 2025 at 01:42:22PM +0100, Maxime Chevallier wrote:
> On Tue, 18 Feb 2025 17:19:13 +0100
> Kory Maincent <kory.maincent@bootlin.com> wrote:
> > diff --git a/drivers/net/pse-pd/pd692x0.c b/drivers/net/pse-pd/pd692x0.c
> > index 44ded2aa6fca..c9fa60b314ce 100644
> > --- a/drivers/net/pse-pd/pd692x0.c
> > +++ b/drivers/net/pse-pd/pd692x0.c
> > @@ -976,8 +976,10 @@ pd692x0_register_managers_regulator(struct pd692x0_priv *priv,
> > reg_name_len = strlen(dev_name(dev)) + 23;
> >
> > for (i = 0; i < nmanagers; i++) {
> > + static const char * const regulators[] = { "vaux5", "vaux3p3" };
>
> Looks like the 'static' is not needed here :)
Have you checked the compiler output before saying that?
I've seen plenty of instances where "static" should be there but isn't,
leading to the compiler generating inline code to create the
array/struct on the stack.
--
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
^ permalink raw reply [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 09/12] net: pse-pd: pd692x0: Add support for controller and manager power supplies
2025-02-24 12:49 ` Russell King (Oracle)
@ 2025-02-24 13:17 ` Maxime Chevallier
0 siblings, 0 replies; 42+ messages in thread
From: Maxime Chevallier @ 2025-02-24 13:17 UTC (permalink / raw)
To: Russell King (Oracle)
Cc: Kory Maincent, Andrew Lunn, Oleksij Rempel, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Jonathan Corbet,
Donald Hunter, Rob Herring, Andrew Lunn, Simon Horman,
Heiner Kallweit, Krzysztof Kozlowski, Conor Dooley,
Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, devicetree, linux-kernel
On Mon, 24 Feb 2025 12:49:19 +0000
"Russell King (Oracle)" <linux@armlinux.org.uk> wrote:
> On Mon, Feb 24, 2025 at 01:42:22PM +0100, Maxime Chevallier wrote:
> > On Tue, 18 Feb 2025 17:19:13 +0100
> > Kory Maincent <kory.maincent@bootlin.com> wrote:
> > > diff --git a/drivers/net/pse-pd/pd692x0.c b/drivers/net/pse-pd/pd692x0.c
> > > index 44ded2aa6fca..c9fa60b314ce 100644
> > > --- a/drivers/net/pse-pd/pd692x0.c
> > > +++ b/drivers/net/pse-pd/pd692x0.c
> > > @@ -976,8 +976,10 @@ pd692x0_register_managers_regulator(struct pd692x0_priv *priv,
> > > reg_name_len = strlen(dev_name(dev)) + 23;
> > >
> > > for (i = 0; i < nmanagers; i++) {
> > > + static const char * const regulators[] = { "vaux5", "vaux3p3" };
> >
> > Looks like the 'static' is not needed here :)
>
> Have you checked the compiler output before saying that?
No I have not
> I've seen plenty of instances where "static" should be there but isn't,
> leading to the compiler generating inline code to create the
> array/struct on the stack.
Makes sense then, so it should be good here.
Thanks,
Maxime
^ permalink raw reply [flat|nested] 42+ messages in thread
* [PATCH net-next v5 10/12] dt-bindings: net: pse-pd: microchip,pd692x0: Add manager regulator supply
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
` (8 preceding siblings ...)
2025-02-18 16:19 ` [PATCH net-next v5 09/12] net: pse-pd: pd692x0: Add support for controller and manager power supplies Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-19 7:41 ` Krzysztof Kozlowski
2025-02-18 16:19 ` [PATCH net-next v5 11/12] net: pse-pd: tps23881: Add support for static port priority feature Kory Maincent
2025-02-18 16:19 ` [PATCH net-next v5 12/12] dt-bindings: net: pse-pd: ti,tps23881: Add interrupt description Kory Maincent
11 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
This patch adds the regulator supply parameter of the managers.
It updates also the example as the regulator supply of the PSE PIs
should be the managers itself and not an external regulator.
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
Changes in v5:
- Add description of others power supplies.
Changes in v3:
- New patch
---
.../bindings/net/pse-pd/microchip,pd692x0.yaml | 22 +++++++++++++++++++---
1 file changed, 19 insertions(+), 3 deletions(-)
diff --git a/Documentation/devicetree/bindings/net/pse-pd/microchip,pd692x0.yaml b/Documentation/devicetree/bindings/net/pse-pd/microchip,pd692x0.yaml
index fd4244fceced..ca61cc37a790 100644
--- a/Documentation/devicetree/bindings/net/pse-pd/microchip,pd692x0.yaml
+++ b/Documentation/devicetree/bindings/net/pse-pd/microchip,pd692x0.yaml
@@ -22,6 +22,12 @@ properties:
reg:
maxItems: 1
+ vdd-supply:
+ description: Regulator that provides 3.3V VDD power supply.
+
+ vdda-supply:
+ description: Regulator that provides 3.3V VDDA power supply.
+
managers:
type: object
additionalProperties: false
@@ -68,6 +74,15 @@ properties:
"#size-cells":
const: 0
+ vmain-supply:
+ description: Regulator that provides 44-57V VMAIN power supply.
+
+ vaux5-supply:
+ description: Regulator that provides 5V VAUX5 power supply.
+
+ vaux3p3-supply:
+ description: Regulator that provides 3.3V VAUX3P3 power supply.
+
patternProperties:
'^port@[0-7]$':
type: object
@@ -106,10 +121,11 @@ examples:
#address-cells = <1>;
#size-cells = <0>;
- manager@0 {
+ manager0: manager@0 {
reg = <0>;
#address-cells = <1>;
#size-cells = <0>;
+ vmain-supply = <&pse1_supply>;
phys0: port@0 {
reg = <0>;
@@ -161,7 +177,7 @@ examples:
pairset-names = "alternative-a", "alternative-b";
pairsets = <&phys0>, <&phys1>;
polarity-supported = "MDI", "S";
- vpwr-supply = <&vpwr1>;
+ vpwr-supply = <&manager0>;
};
pse_pi1: pse-pi@1 {
reg = <1>;
@@ -169,7 +185,7 @@ examples:
pairset-names = "alternative-a";
pairsets = <&phys2>;
polarity-supported = "MDI";
- vpwr-supply = <&vpwr2>;
+ vpwr-supply = <&manager0>;
};
};
};
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 10/12] dt-bindings: net: pse-pd: microchip,pd692x0: Add manager regulator supply
2025-02-18 16:19 ` [PATCH net-next v5 10/12] dt-bindings: net: pse-pd: microchip,pd692x0: Add manager regulator supply Kory Maincent
@ 2025-02-19 7:41 ` Krzysztof Kozlowski
0 siblings, 0 replies; 42+ messages in thread
From: Krzysztof Kozlowski @ 2025-02-19 7:41 UTC (permalink / raw)
To: Kory Maincent
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni,
netdev, linux-doc, Kyle Swenson, Dent Project, kernel,
Maxime Chevallier, devicetree, linux-kernel
On Tue, Feb 18, 2025 at 05:19:14PM +0100, Kory Maincent wrote:
> From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
>
> This patch adds the regulator supply parameter of the managers.
In the future:
Please do not use "This commit/patch/change", but imperative mood. See
longer explanation here:
https://elixir.bootlin.com/linux/v5.17.1/source/Documentation/process/submitting-patches.rst#L95
> It updates also the example as the regulator supply of the PSE PIs
> should be the managers itself and not an external regulator.
>
> Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
> ---
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Best regards,
Krzysztof
^ permalink raw reply [flat|nested] 42+ messages in thread
* [PATCH net-next v5 11/12] net: pse-pd: tps23881: Add support for static port priority feature
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
` (9 preceding siblings ...)
2025-02-18 16:19 ` [PATCH net-next v5 10/12] dt-bindings: net: pse-pd: microchip,pd692x0: Add manager regulator supply Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-18 16:19 ` [PATCH net-next v5 12/12] dt-bindings: net: pse-pd: ti,tps23881: Add interrupt description Kory Maincent
11 siblings, 0 replies; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
This patch enhances PSE callbacks by introducing support for the static
port priority feature. It extends interrupt management to handle and report
detection, classification, and disconnection events. Additionally, it
introduces the pi_get_pw_req() callback, which provides information about
the power requested by the Powered Devices.
Interrupt support is essential for the proper functioning of the TPS23881
controller. Without it, after a power-on (PWON), the controller will
no longer perform detection and classification. This could lead to
potential hazards, such as connecting a non-PoE device after a PoE device,
which might result in magic smoke.
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
We may need a fix for the interrupt support in old version of Linux.
Change in v4:
- Fix variable type nit.
Change in v3:
- New patch
---
drivers/net/pse-pd/tps23881.c | 204 +++++++++++++++++++++++++++++++++++++++---
1 file changed, 194 insertions(+), 10 deletions(-)
diff --git a/drivers/net/pse-pd/tps23881.c b/drivers/net/pse-pd/tps23881.c
index 122666719297..6012c58b47e8 100644
--- a/drivers/net/pse-pd/tps23881.c
+++ b/drivers/net/pse-pd/tps23881.c
@@ -19,20 +19,30 @@
#define TPS23881_REG_IT 0x0
#define TPS23881_REG_IT_MASK 0x1
+#define TPS23881_REG_IT_DISF BIT(2)
+#define TPS23881_REG_IT_DETC BIT(3)
+#define TPS23881_REG_IT_CLASC BIT(4)
#define TPS23881_REG_IT_IFAULT BIT(5)
#define TPS23881_REG_IT_SUPF BIT(7)
+#define TPS23881_REG_DET_EVENT 0x5
#define TPS23881_REG_FAULT 0x7
#define TPS23881_REG_SUPF_EVENT 0xb
#define TPS23881_REG_TSD BIT(7)
+#define TPS23881_REG_DISC 0xc
#define TPS23881_REG_PW_STATUS 0x10
#define TPS23881_REG_OP_MODE 0x12
+#define TPS23881_REG_DISC_EN 0x13
#define TPS23881_OP_MODE_SEMIAUTO 0xaaaa
#define TPS23881_REG_DIS_EN 0x13
#define TPS23881_REG_DET_CLA_EN 0x14
#define TPS23881_REG_GEN_MASK 0x17
+#define TPS23881_REG_CLCHE BIT(2)
+#define TPS23881_REG_DECHE BIT(3)
#define TPS23881_REG_NBITACC BIT(5)
#define TPS23881_REG_INTEN BIT(7)
#define TPS23881_REG_PW_EN 0x19
+#define TPS23881_REG_RESET 0x1a
+#define TPS23881_REG_CLRAIN BIT(7)
#define TPS23881_REG_2PAIR_POL1 0x1e
#define TPS23881_REG_PORT_MAP 0x26
#define TPS23881_REG_PORT_POWER 0x29
@@ -177,6 +187,7 @@ static int tps23881_pi_enable(struct pse_controller_dev *pcdev, int id)
struct i2c_client *client = priv->client;
u8 chan;
u16 val;
+ int ret;
if (id >= TPS23881_MAX_CHANS)
return -ERANGE;
@@ -190,7 +201,22 @@ static int tps23881_pi_enable(struct pse_controller_dev *pcdev, int id)
BIT(chan % 4));
}
- return i2c_smbus_write_word_data(client, TPS23881_REG_PW_EN, val);
+ ret = i2c_smbus_write_word_data(client, TPS23881_REG_PW_EN, val);
+ if (ret)
+ return ret;
+
+ /* Enable DC disconnect*/
+ chan = priv->port[id].chan[0];
+ ret = i2c_smbus_read_word_data(client, TPS23881_REG_DISC_EN);
+ if (ret < 0)
+ return ret;
+
+ val = tps23881_set_val(ret, chan, 0, BIT(chan % 4), BIT(chan % 4));
+ ret = i2c_smbus_write_word_data(client, TPS23881_REG_DISC_EN, val);
+ if (ret)
+ return ret;
+
+ return 0;
}
static int tps23881_pi_disable(struct pse_controller_dev *pcdev, int id)
@@ -223,6 +249,17 @@ static int tps23881_pi_disable(struct pse_controller_dev *pcdev, int id)
*/
mdelay(5);
+ /* Disable DC disconnect*/
+ chan = priv->port[id].chan[0];
+ ret = i2c_smbus_read_word_data(client, TPS23881_REG_DISC_EN);
+ if (ret < 0)
+ return ret;
+
+ val = tps23881_set_val(ret, chan, 0, 0, BIT(chan % 4));
+ ret = i2c_smbus_write_word_data(client, TPS23881_REG_DISC_EN, val);
+ if (ret)
+ return ret;
+
/* Enable detection and classification */
ret = i2c_smbus_read_word_data(client, TPS23881_REG_DET_CLA_EN);
if (ret < 0)
@@ -918,6 +955,47 @@ static int tps23881_setup_pi_matrix(struct pse_controller_dev *pcdev)
return ret;
}
+static int tps23881_power_class_table[] = {
+ -ERANGE,
+ 4000,
+ 7000,
+ 15500,
+ 30000,
+ 15500,
+ 15500,
+ -ERANGE,
+ 45000,
+ 60000,
+ 75000,
+ 90000,
+ 15500,
+ 45000,
+ -ERANGE,
+ -ERANGE,
+};
+
+static int tps23881_pi_get_pw_req(struct pse_controller_dev *pcdev, int id)
+{
+ struct tps23881_priv *priv = to_tps23881_priv(pcdev);
+ struct i2c_client *client = priv->client;
+ u8 reg, chan;
+ int ret;
+ u16 val;
+
+ /* For a 4-pair the classification need 5ms to be completed */
+ if (priv->port[id].is_4p)
+ mdelay(5);
+
+ chan = priv->port[id].chan[0];
+ reg = TPS23881_REG_DISC + (chan % 4);
+ ret = i2c_smbus_read_word_data(client, reg);
+ if (ret < 0)
+ return ret;
+
+ val = tps23881_calc_val(ret, chan, 4, 0xf);
+ return tps23881_power_class_table[val];
+}
+
static const struct pse_controller_ops tps23881_ops = {
.setup_pi_matrix = tps23881_setup_pi_matrix,
.pi_enable = tps23881_pi_enable,
@@ -930,6 +1008,7 @@ static const struct pse_controller_ops tps23881_ops = {
.pi_get_pw_limit = tps23881_pi_get_pw_limit,
.pi_set_pw_limit = tps23881_pi_set_pw_limit,
.pi_get_pw_limit_ranges = tps23881_pi_get_pw_limit_ranges,
+ .pi_get_pw_req = tps23881_pi_get_pw_req,
};
static const char fw_parity_name[] = "ti/tps23881/tps23881-parity-14.bin";
@@ -1100,12 +1179,83 @@ static void tps23881_irq_event_over_current(struct tps23881_priv *priv,
ETHTOOL_PSE_EVENT_OVER_CURRENT);
}
+static void tps23881_irq_event_disconnection(struct tps23881_priv *priv,
+ u16 reg_val,
+ unsigned long *notifs,
+ unsigned long *notifs_mask)
+{
+ u8 chans;
+
+ chans = tps23881_irq_export_chans_helper(reg_val, 4);
+ if (chans)
+ tps23881_set_notifs_helper(priv, chans, notifs, notifs_mask,
+ ETHTOOL_C33_PSE_EVENT_DISCONNECTION);
+}
+
+static int tps23881_irq_event_detection(struct tps23881_priv *priv,
+ u16 reg_val,
+ unsigned long *notifs,
+ unsigned long *notifs_mask)
+{
+ enum ethtool_pse_events event;
+ int reg, ret, i, val;
+ unsigned long chans;
+
+ chans = tps23881_irq_export_chans_helper(reg_val, 0);
+ for_each_set_bit(i, &chans, TPS23881_MAX_CHANS) {
+ reg = TPS23881_REG_DISC + (i % 4);
+ ret = i2c_smbus_read_word_data(priv->client, reg);
+ if (ret < 0)
+ return ret;
+
+ val = tps23881_calc_val(ret, i, 0, 0xf);
+ /* If detection valid */
+ if (val == 0x4)
+ event = ETHTOOL_C33_PSE_EVENT_DETECTION;
+ else
+ event = ETHTOOL_C33_PSE_EVENT_DISCONNECTION;
+
+ tps23881_set_notifs_helper(priv, BIT(i), notifs,
+ notifs_mask, event);
+ }
+
+ return 0;
+}
+
+static int tps23881_irq_event_classification(struct tps23881_priv *priv,
+ u16 reg_val,
+ unsigned long *notifs,
+ unsigned long *notifs_mask)
+{
+ int reg, ret, val, i;
+ unsigned long chans;
+
+ chans = tps23881_irq_export_chans_helper(reg_val, 4);
+ for_each_set_bit(i, &chans, TPS23881_MAX_CHANS) {
+ reg = TPS23881_REG_DISC + (i % 4);
+ ret = i2c_smbus_read_word_data(priv->client, reg);
+ if (ret < 0)
+ return ret;
+
+ val = tps23881_calc_val(ret, i, 4, 0xf);
+ /* Do not report classification event for unknown class */
+ if (!val || val == 0x8 || val == 0xf)
+ continue;
+
+ tps23881_set_notifs_helper(priv, BIT(i), notifs,
+ notifs_mask,
+ ETHTOOL_C33_PSE_EVENT_CLASSIFICATION);
+ }
+
+ return 0;
+}
+
static int tps23881_irq_event_handler(struct tps23881_priv *priv, u16 reg,
unsigned long *notifs,
unsigned long *notifs_mask)
{
struct i2c_client *client = priv->client;
- int ret;
+ int ret, val;
/* The Supply event bit is repeated twice so we only need to read
* the one from the first byte.
@@ -1117,13 +1267,33 @@ static int tps23881_irq_event_handler(struct tps23881_priv *priv, u16 reg,
tps23881_irq_event_over_temp(priv, ret, notifs, notifs_mask);
}
- if (reg & (TPS23881_REG_IT_IFAULT | TPS23881_REG_IT_IFAULT << 8)) {
+ if (reg & (TPS23881_REG_IT_IFAULT | TPS23881_REG_IT_IFAULT << 8 |
+ TPS23881_REG_IT_DISF | TPS23881_REG_IT_DISF << 8)) {
ret = i2c_smbus_read_word_data(client, TPS23881_REG_FAULT);
if (ret < 0)
return ret;
tps23881_irq_event_over_current(priv, ret, notifs, notifs_mask);
+
+ tps23881_irq_event_disconnection(priv, ret, notifs, notifs_mask);
}
+ if (reg & (TPS23881_REG_IT_DETC | TPS23881_REG_IT_DETC << 8 |
+ TPS23881_REG_IT_CLASC | TPS23881_REG_IT_CLASC << 8)) {
+ ret = i2c_smbus_read_word_data(client, TPS23881_REG_DET_EVENT);
+ if (ret < 0)
+ return ret;
+
+ val = ret;
+ ret = tps23881_irq_event_detection(priv, val, notifs,
+ notifs_mask);
+ if (ret)
+ return ret;
+
+ ret = tps23881_irq_event_classification(priv, val, notifs,
+ notifs_mask);
+ if (ret)
+ return ret;
+ }
return 0;
}
@@ -1169,7 +1339,14 @@ static int tps23881_setup_irq(struct tps23881_priv *priv, int irq)
int ret;
u16 val;
- val = TPS23881_REG_IT_IFAULT | TPS23881_REG_IT_SUPF;
+ if (!irq) {
+ dev_err(&client->dev, "interrupt is missing");
+ return -EINVAL;
+ }
+
+ val = TPS23881_REG_IT_IFAULT | TPS23881_REG_IT_SUPF |
+ TPS23881_REG_IT_DETC | TPS23881_REG_IT_CLASC |
+ TPS23881_REG_IT_DISF;
val |= val << 8;
ret = i2c_smbus_write_word_data(client, TPS23881_REG_IT_MASK, val);
if (ret)
@@ -1179,11 +1356,19 @@ static int tps23881_setup_irq(struct tps23881_priv *priv, int irq)
if (ret < 0)
return ret;
- val = (u16)(ret | TPS23881_REG_INTEN | TPS23881_REG_INTEN << 8);
+ val = TPS23881_REG_INTEN | TPS23881_REG_CLCHE | TPS23881_REG_DECHE;
+ val |= val << 8;
+ val |= (u16)ret;
ret = i2c_smbus_write_word_data(client, TPS23881_REG_GEN_MASK, val);
if (ret < 0)
return ret;
+ /* Reset interrupts registers */
+ ret = i2c_smbus_write_word_data(client, TPS23881_REG_RESET,
+ TPS23881_REG_CLRAIN);
+ if (ret < 0)
+ return ret;
+
return devm_pse_irq_helper(&priv->pcdev, irq, 0, &irq_desc);
}
@@ -1261,17 +1446,16 @@ static int tps23881_i2c_probe(struct i2c_client *client)
priv->pcdev.dev = dev;
priv->pcdev.types = ETHTOOL_PSE_C33;
priv->pcdev.nr_lines = TPS23881_MAX_CHANS;
+ priv->pcdev.supp_budget_eval_strategies = ETHTOOL_PSE_BUDGET_EVAL_STRAT_STATIC;
ret = devm_pse_controller_register(dev, &priv->pcdev);
if (ret) {
return dev_err_probe(dev, ret,
"failed to register PSE controller\n");
}
- if (client->irq) {
- ret = tps23881_setup_irq(priv, client->irq);
- if (ret)
- return ret;
- }
+ ret = tps23881_setup_irq(priv, client->irq);
+ if (ret)
+ return ret;
return ret;
}
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* [PATCH net-next v5 12/12] dt-bindings: net: pse-pd: ti,tps23881: Add interrupt description
2025-02-18 16:19 [PATCH net-next v5 00/12] Add support for PSE budget evaluation strategy Kory Maincent
` (10 preceding siblings ...)
2025-02-18 16:19 ` [PATCH net-next v5 11/12] net: pse-pd: tps23881: Add support for static port priority feature Kory Maincent
@ 2025-02-18 16:19 ` Kory Maincent
2025-02-19 7:41 ` Krzysztof Kozlowski
11 siblings, 1 reply; 42+ messages in thread
From: Kory Maincent @ 2025-02-18 16:19 UTC (permalink / raw)
To: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley
Cc: Thomas Petazzoni, netdev, linux-doc, Kyle Swenson, Dent Project,
kernel, Maxime Chevallier, devicetree, linux-kernel,
Kory Maincent (Dent Project)
From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
Add an interrupt property to the device tree bindings for the TI TPS23881
PSE controller. The interrupt is primarily used to detect classification
and disconnection events, which are essential for managing the PSE
controller in compliance with the PoE standard.
Interrupt support is essential for the proper functioning of the TPS23881
controller. Without it, after a power-on (PWON), the controller will
no longer perform detection and classification. This could lead to
potential hazards, such as connecting a non-PoE device after a PoE device,
which might result in magic smoke.
Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
---
Change in v5:
- Use standard interrupt flag in the example.
Change in v3:
- New patch
---
Documentation/devicetree/bindings/net/pse-pd/ti,tps23881.yaml | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/Documentation/devicetree/bindings/net/pse-pd/ti,tps23881.yaml b/Documentation/devicetree/bindings/net/pse-pd/ti,tps23881.yaml
index d08abcb01211..3a5f960d8489 100644
--- a/Documentation/devicetree/bindings/net/pse-pd/ti,tps23881.yaml
+++ b/Documentation/devicetree/bindings/net/pse-pd/ti,tps23881.yaml
@@ -20,6 +20,9 @@ properties:
reg:
maxItems: 1
+ interrupts:
+ maxItems: 1
+
'#pse-cells':
const: 1
@@ -62,9 +65,12 @@ unevaluatedProperties: false
required:
- compatible
- reg
+ - interrupts
examples:
- |
+ #include <dt-bindings/interrupt-controller/irq.h>
+
i2c {
#address-cells = <1>;
#size-cells = <0>;
@@ -72,6 +78,8 @@ examples:
ethernet-pse@20 {
compatible = "ti,tps23881";
reg = <0x20>;
+ interrupts = <8 IRQ_TYPE_LEVEL_HIGH>;
+ interrupt-parent = <&gpiog>;
channels {
#address-cells = <1>;
--
2.34.1
^ permalink raw reply related [flat|nested] 42+ messages in thread* Re: [PATCH net-next v5 12/12] dt-bindings: net: pse-pd: ti,tps23881: Add interrupt description
2025-02-18 16:19 ` [PATCH net-next v5 12/12] dt-bindings: net: pse-pd: ti,tps23881: Add interrupt description Kory Maincent
@ 2025-02-19 7:41 ` Krzysztof Kozlowski
0 siblings, 0 replies; 42+ messages in thread
From: Krzysztof Kozlowski @ 2025-02-19 7:41 UTC (permalink / raw)
To: Kory Maincent
Cc: Andrew Lunn, Oleksij Rempel, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Donald Hunter,
Rob Herring, Andrew Lunn, Simon Horman, Heiner Kallweit,
Russell King, Krzysztof Kozlowski, Conor Dooley, Thomas Petazzoni,
netdev, linux-doc, Kyle Swenson, Dent Project, kernel,
Maxime Chevallier, devicetree, linux-kernel
On Tue, Feb 18, 2025 at 05:19:16PM +0100, Kory Maincent wrote:
> From: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
>
> Add an interrupt property to the device tree bindings for the TI TPS23881
> PSE controller. The interrupt is primarily used to detect classification
> and disconnection events, which are essential for managing the PSE
> controller in compliance with the PoE standard.
>
> Interrupt support is essential for the proper functioning of the TPS23881
> controller. Without it, after a power-on (PWON), the controller will
> no longer perform detection and classification. This could lead to
> potential hazards, such as connecting a non-PoE device after a PoE device,
> which might result in magic smoke.
>
> Signed-off-by: Kory Maincent (Dent Project) <kory.maincent@bootlin.com>
> ---
>
> Change in v5:
> - Use standard interrupt flag in the example.
>
> Change in v3:
> - New patch
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Best regards,
Krzysztof
^ permalink raw reply [flat|nested] 42+ messages in thread