net: dsa: mv88e6xxx: Allow hwstamping on the master port
Subsystems (4)
PATCH 1/1 net: dsa: mv88e6xxx: Allow hwstamping on the master port
From: Steffen Bätz <[email protected]>
Currently, it is not possible to run SIOCGHWTSTAMP or SIOCSHWTSTAMP ioctls on the dsa master interface if the port_hwtstamp_set/get() hooks are present, as implemented in net/dsa/master.c:
static int dsa_master_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) {
...
case SIOCGHWTSTAMP: case SIOCSHWTSTAMP: /* Deny PTP operations on master if there is at least one * switch in the tree that is PTP capable. */ list_for_each_entry(dp, &dst->ports, list) if (dp->ds->ops->port_hwtstamp_get || dp->ds->ops->port_hwtstamp_set) return -EBUSY; break; }
Even if the hwtstamping functionality is disabled in the mv88e6xxx driver by not setting CONFIG_NET_DSA_MV88E6XXX_PTP, the functions port_hwtstamp_set() port_hwtstamp_get() are still present due to their stub declarations.
Fix this problem, by removing the stub declarations and guarding these functions wih CONFIG_NET_DSA_MV88E6XXX_PTP.
Without this change:
# hwstamp_ctl -i eth0 SIOCGHWTSTAMP failed: Device or resource busy
With the change applied, it is possible to set and get the timestamping options:
# hwstamp_ctl -i eth0 current settings: tx_type 0 rx_filter 0
# hwstamp_ctl -i eth0 -r 1 -t 1 current settings: tx_type 0 rx_filter 0 new settings: tx_type 1 rx_filter 1
Tested on a custom i.MX8MN board with a 88E6320 switch.
--- drivers/net/dsa/mv88e6xxx/chip.c | 2 ++ drivers/net/dsa/mv88e6xxx/hwtstamp.h | 12 ------------ 2 files changed, 2 insertions(+), 12 deletions(-)
drivers/net/dsa/mv88e6xxx/chip.c +2 -0
@@ -6925,8 +6925,10 @@ static const struct dsa_switch_ops mv88e6xxx_switch_ops = { .port_mirror_del = mv88e6xxx_port_mirror_del, .crosschip_bridge_join = mv88e6xxx_crosschip_bridge_join, .crosschip_bridge_leave = mv88e6xxx_crosschip_bridge_leave, +#ifdef CONFIG_NET_DSA_MV88E6XXX_PTP .port_hwtstamp_set = mv88e6xxx_port_hwtstamp_set, .port_hwtstamp_get = mv88e6xxx_port_hwtstamp_get, +#endif .port_txtstamp = mv88e6xxx_port_txtstamp, .port_rxtstamp = mv88e6xxx_port_rxtstamp, .get_ts_info = mv88e6xxx_get_ts_info,
drivers/net/dsa/mv88e6xxx/hwtstamp.h +0 -12
@@ -132,18 +132,6 @@ int mv88e6165_global_disable(struct mv88e6xxx_chip *chip);#else /* !CONFIG_NET_DSA_MV88E6XXX_PTP */ -static inline int mv88e6xxx_port_hwtstamp_set (struct dsa_switch *ds, -int port,struct ifreq *ifr) -{ -return -EOPNOTSUPP ; -} - -static inline int mv88e6xxx_port_hwtstamp_get (struct dsa_switch *ds, -int port,struct ifreq *ifr) -{ -return -EOPNOTSUPP ; -} -static inline bool mv88e6xxx_port_rxtstamp (struct dsa_switch *ds,int port,struct sk_buff *clone,unsigned int type)
Signature · 1 line
2.25.1
Quoted 49 lines · Fabio Estevam
From: Steffen Bätz <[email protected]>
Currently, it is not possible to run SIOCGHWTSTAMP or SIOCSHWTSTAMP ioctls on the dsa master interface if the port_hwtstamp_set/get() hooks are present, as implemented in
net/dsa/master.c:static int
dsa_master_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd){
...case SIOCGHWTSTAMP: case SIOCSHWTSTAMP: /* Deny PTP operations on master if there is at least one * switch in the tree that is PTP capable. */ list_for_each_entry(dp, &dst->ports, list) if (dp->ds->ops->port_hwtstamp_get || dp->ds->ops->port_hwtstamp_set) return -EBUSY; break; }Even if the hwtstamping functionality is disabled in the mv88e6xxx driver by not setting
CONFIG_NET_DSA_MV88E6XXX_PTP, the functionsport_hwtstamp_set()port_hwtstamp_get()are still present due to their stub declarations.Fix this problem, by removing the stub declarations and guarding these functions wih
CONFIG_NET_DSA_MV88E6XXX_PTP.Without this change:
# hwstamp_ctl -i eth0 SIOCGHWTSTAMP failed: Device or resource busyWith the change applied, it is possible to set and get the timestamping options:
# hwstamp_ctl -i eth0 current settings: tx_type 0 rx_filter 0# hwstamp_ctl -i eth0 -r 1 -t 1 current settings: tx_type 0 rx_filter 0 new settings: tx_type 1 rx_filter 1
Quoted 5 lines
Tested on a custom i.MX8MN board with a 88E6320 switch.
Signed-off-by Steffen Bätz <[email protected]>Signed-off-by Fabio Estevam <[email protected]>---
NACK.
Please extend dsa_master_ioctl() to "see through" the dp->ds->ops->port_hwtstamp_get() pointer, similar to what dsa_port_can_configure_learning() does. Create a fake struct ifreq, call port_hwtstamp_get(), see if it returned -EOPNOTSUPP, and wrap that into a dsa_port_supports_hwtstamp() function, and call that instead of the current simplistic checks for the function pointers.
Hi Vladimir,
Quoted 8 lines · Vladimir Oltean
NACK.
Please extend
dsa_master_ioctl()to "see through" thedp->ds->ops->port_hwtstamp_get()pointer, similar to whatdsa_port_can_configure_learning()does. Create a fake struct ifreq, callport_hwtstamp_get(), see if it returned-EOPNOTSUPP, and wrap that into adsa_port_supports_hwtstamp()function, and call that instead of the current simplistic checks for the function pointers.
I tried to implement what you suggested in the attached patch, but I am not sure it is correct.
If it doesn't work, please cook a patch so we can try it.
I am not familiar with the DSA code, sorry.
Thanks
From 6f45968dd2e4585818ea6acf8420bfed38246604 Mon Sep 17 00:00:00 2001
From: Fabio Estevam <[email protected]>
Date: Thu, 10 Nov 2022 18:50:59 -0300
Subject: [PATCH net-next] net: dsa: Introduce dsa_port_supports_hwtstamp()
Currently, during SIOCGHWTSTAMP/SIOCSHWTSTAMP ioctl operations, the criteria for deciding whether PTP operations on the master port is allowed or not is simply based on the presence of the port_hwtstamp_get/set() operations.
This is not a robust method because on the mv88e6xxx driver, for example, the port_hwtstamp_get()/set() hooks are always implemented.
Even when CONFIG_NET_DSA_MV88E6XXX_PTP is disabled there are still stub implementations for these functions that return -EOPNOTSUPP.
Instead of relying on the presence of port_hwtstamp_get/set(), introduce the dsa_port_supports_hwtstamp() function, which checks the return value of port_hwtstamp_get/set() to decide.
Without this change:
# hwstamp_ctl -i eth0 SIOCGHWTSTAMP failed: Device or resource busy
With this change applied, it is possible to set and get the timestamping options:
# hwstamp_ctl -i eth0 current settings: tx_type 0 rx_filter 0
# hwstamp_ctl -i eth0 -r 1 -t 1 current settings: tx_type 0 rx_filter 0 new settings: tx_type 1 rx_filter 1
Tested on a custom i.MX8MN board with a 88E6320 switch.
--- net/dsa/master.c | 40 ++++++++++++++++++++++++++++------------ 1 file changed, 28 insertions(+), 12 deletions(-)
net/dsa/master.c +28 -12
@@ -187,29 +187,45 @@ static void dsa_master_get_strings(struct net_device *dev, uint32_t stringset, } } -static int dsa_master_ioctl (struct net_device *dev,struct ifreq *ifr,int cmd) +static bool dsa_port_supports_hwtstamp (struct net_device *dev,struct ifreq *ifr,int cmd) {struct dsa_port *cpu_dp = dev->dsa_ptr;struct dsa_switch *ds = cpu_dp->ds; -struct dsa_switch_tree *dst; +int port = cpu_dp->index;int err = -EOPNOTSUPP ; -struct dsa_port *dp; - - dst = ds->dst;switch (cmd) {case SIOCGHWTSTAMP : +if (ds->ops->port_hwtstamp_get) { + err = ds->ops->port_hwtstamp_get (ds, port, ifr); +if (err == -EOPNOTSUPP ) +return false; + } +break ;case SIOCSHWTSTAMP : -/* Deny PTP operations on master if there is at least one -* switch in the tree that is PTP capable. -*/ -list_for_each_entry (dp, &dst->ports, list) -if (dp->ds->ops->port_hwtstamp_get || - dp->ds->ops->port_hwtstamp_set) -return -EBUSY ; +if (ds->ops->port_hwtstamp_set) { + err = ds->ops->port_hwtstamp_set (ds, port, ifr); +if (err == -EOPNOTSUPP ) +return false; + }break ; } +return !!err; +} + +static int dsa_master_ioctl (struct net_device *dev,struct ifreq *ifr,int cmd) +{ +struct dsa_port *cpu_dp = dev->dsa_ptr; +struct dsa_switch *ds = cpu_dp->ds; +struct dsa_switch_tree *dst; +int err = -EOPNOTSUPP ; + + dst = ds->dst; + +if (dsa_port_supports_hwtstamp (dev, ifr, cmd)) +return -EBUSY ; +if (dev->netdev_ops->ndo_eth_ioctl) err = dev->netdev_ops->ndo_eth_ioctl (dev, ifr, cmd);
Signature · 1 line
2.25.1
Vladimir Oltean
maintainerQuoted 13 lines · Fabio Estevam
Hi Vladimir,
Quoted 8 lines · Vladimir Oltean
NACK.
Please extend
dsa_master_ioctl()to "see through" thedp->ds->ops->port_hwtstamp_get()pointer, similar to whatdsa_port_can_configure_learning()does. Create a fake struct ifreq, callport_hwtstamp_get(), see if it returned-EOPNOTSUPP, and wrap that into adsa_port_supports_hwtstamp()function, and call that instead of the current simplistic checks for the function pointers.
Quoted 4 lines
I tried to implement what you suggested in the attached patch, but I am not sure it is correct.
If it doesn't work, please cook a patch so we can try it.
Yeah, this is not what I meant. I posted a patch illustrating what I meant here:
https://patchwork.kernel.org/project/netdevbpf/patch/[email protected]/
Hi Vladimir,
Quoted 3 lines · Vladimir Oltean
Yeah, this is not what I meant. I posted a patch illustrating what I meant here:
https://patchwork.kernel.org/project/netdevbpf/patch/[email protected]/
Thanks for your patch, appreciated.
I tested your patch and it works fine.
Cheers