Skip to contentSkip to conversation

net: phy: Add link between phy dev and mac dev

PATCH Superseded Xiaolei Wang 5 participants 10 messages Nov 25, 2022 – Nov 29, 2022

Subsystems (3)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover net: phy: Add link between phy dev and mac dev

Compared with v1, put the link between phy dev and mac dev in phy_attach_direct, if the external phy used by current mac interface is managed by another mac interface, so we should create a device link between phy dev and mac dev.

If the external phy used by current mac interface is managed by another mac interface, it means that this network port cannot work independently, especially when the system suspend and resume, the following trace may appear, so we should create a device link between phy dev and mac dev.

  WARNING: CPU: 0 PID: 24 at drivers/net/phy/phy.c:983 phy_error+0x20/0x68
  Modules linked in:
  CPU: 0 PID: 24 Comm: kworker/0:2 Not tainted 6.1.0-rc3-00011-g5aaef24b5c6d-dirty #34
  Hardware name: Freescale i.MX6 SoloX (Device Tree)
  Workqueue: events_power_efficient phy_state_machine
  unwind_backtrace from show_stack+0x10/0x14
  show_stack from dump_stack_lvl+0x68/0x90
  dump_stack_lvl from __warn+0xb4/0x24c
  __warn from warn_slowpath_fmt+0x5c/0xd8
  warn_slowpath_fmt from phy_error+0x20/0x68
  phy_error from phy_state_machine+0x22c/0x23c
  phy_state_machine from process_one_work+0x288/0x744
  process_one_work from worker_thread+0x3c/0x500
  worker_thread from kthread+0xf0/0x114
  kthread from ret_from_fork+0x14/0x28
  Exception stack(0xf0951fb0 to 0xf0951ff8)

Xiaolei Wang (1):
net: phy: Add link between phy dev and mac dev

 drivers/net/phy/phy_device.c | 12 ++++++++++++
 include/linux/phy.h          |  2 ++
 2 files changed, 14 insertions(+)
Signature · 1 line
2.25.1
#1 lore raw

PATCH v2 1/1 net: phy: Add link between phy dev and mac dev

If the external phy used by current mac interface is managed by another mac interface, it means that this network port cannot work independently, especially when the system suspend and resume, the following trace may appear, so we should create a device link between phy dev and mac dev.

  WARNING: CPU: 0 PID: 24 at drivers/net/phy/phy.c:983 phy_error+0x20/0x68
  Modules linked in:
  CPU: 0 PID: 24 Comm: kworker/0:2 Not tainted 6.1.0-rc3-00011-g5aaef24b5c6d-dirty #34
  Hardware name: Freescale i.MX6 SoloX (Device Tree)
  Workqueue: events_power_efficient phy_state_machine
  unwind_backtrace from show_stack+0x10/0x14
  show_stack from dump_stack_lvl+0x68/0x90
  dump_stack_lvl from __warn+0xb4/0x24c
  __warn from warn_slowpath_fmt+0x5c/0xd8
  warn_slowpath_fmt from phy_error+0x20/0x68
  phy_error from phy_state_machine+0x22c/0x23c
  phy_state_machine from process_one_work+0x288/0x744
  process_one_work from worker_thread+0x3c/0x500
  worker_thread from kthread+0xf0/0x114
  kthread from ret_from_fork+0x14/0x28
  Exception stack(0xf0951fb0 to 0xf0951ff8)
---
 drivers/net/phy/phy_device.c | 12 ++++++++++++
 include/linux/phy.h          |  2 ++
 2 files changed, 14 insertions(+)
drivers/net/phy/phy_device.c +12 -0
index 57849ac0384e..ca6d12f37066 100644
@@ -1511,6 +1511,15 @@ int phy_attach_direct(struct net_device *dev, struct phy_device *phydev,
 	phy_resume(phydev);
 	phy_led_triggers_register(phydev);
 
+	/**
+	 * If the external phy used by current mac interface is managed by
+	 * another mac interface, so we should create a device link between
+	 * phy dev and mac dev.
+	 */
+	if (phydev->mdio.bus->parent && dev->dev.parent != phydev->mdio.bus->parent)
+		phydev->devlink = device_link_add(dev->dev.parent, &phydev->mdio.dev,
+						  DL_FLAG_PM_RUNTIME | DL_FLAG_STATELESS);
+
 	return err;
 
 error:
@@ -1748,6 +1757,9 @@ void phy_detach(struct phy_device *phydev)
 	struct module *ndev_owner = NULL;
 	struct mii_bus *bus;
 
+	if (phydev->devlink)
+		device_link_del(phydev->devlink);
+
 	if (phydev->sysfs_links) {
 		if (dev)
 			sysfs_remove_link(&dev->dev.kobj, "phydev");
include/linux/phy.h +2 -0
index ddf66198f751..f7f8b909fed0 100644
@@ -617,6 +617,8 @@ struct phy_device {
 	/* And management functions */
 	struct phy_driver *drv;
 
+	struct device_link *devlink;
+
 	u32 phy_id;
 
 	struct phy_c45_device_ids c45_ids;
Signature · 1 line
2.25.1
maintainer
#2 lore raw
Quoted 24 lines · Xiaolei Wang

If the external phy used by current mac interface is managed by another mac interface, it means that this network port cannot work independently, especially when the system suspend and resume, the following trace may appear, so we should create a device link between phy dev and mac dev.

  WARNING: CPU: 0 PID: 24 at drivers/net/phy/phy.c:983 phy_error+0x20/0x68
  Modules linked in:
  CPU: 0 PID: 24 Comm: kworker/0:2 Not tainted 6.1.0-rc3-00011-g5aaef24b5c6d-dirty #34
  Hardware name: Freescale i.MX6 SoloX (Device Tree)
  Workqueue: events_power_efficient phy_state_machine
  unwind_backtrace from show_stack+0x10/0x14
  show_stack from dump_stack_lvl+0x68/0x90
  dump_stack_lvl from __warn+0xb4/0x24c
  __warn from warn_slowpath_fmt+0x5c/0xd8
  warn_slowpath_fmt from phy_error+0x20/0x68
  phy_error from phy_state_machine+0x22c/0x23c
  phy_state_machine from process_one_work+0x288/0x744
  process_one_work from worker_thread+0x3c/0x500
  worker_thread from kthread+0xf0/0x114
  kthread from ret_from_fork+0x14/0x28
  Exception stack(0xf0951fb0 to 0xf0951ff8)
Quoted 1 line

This needs Florians review, since for v1 he thinks it will cause regressions.

Andrew

#3 lore raw

add Florian

thanks

xiaolei

Quoted 33 lines · Andrew Lunn

CAUTION: This email comes from a non Wind River email account!
Do not click links or open attachments unless you recognize the sender and know the content is safe.

Quoted 24 lines · Xiaolei Wang

If the external phy used by current mac interface is managed by another mac interface, it means that this network port cannot work independently, especially when the system suspend and resume, the following trace may appear, so we should create a device link between phy dev and mac dev.

   WARNING: CPU: 0 PID: 24 at drivers/net/phy/phy.c:983 phy_error+0x20/0x68
   Modules linked in:
   CPU: 0 PID: 24 Comm: kworker/0:2 Not tainted 6.1.0-rc3-00011-g5aaef24b5c6d-dirty #34
   Hardware name: Freescale i.MX6 SoloX (Device Tree)
   Workqueue: events_power_efficient phy_state_machine
   unwind_backtrace from show_stack+0x10/0x14
   show_stack from dump_stack_lvl+0x68/0x90
   dump_stack_lvl from __warn+0xb4/0x24c
   __warn from warn_slowpath_fmt+0x5c/0xd8
   warn_slowpath_fmt from phy_error+0x20/0x68
   phy_error from phy_state_machine+0x22c/0x23c
   phy_state_machine from process_one_work+0x288/0x744
   process_one_work from worker_thread+0x3c/0x500
   worker_thread from kthread+0xf0/0x114
   kthread from ret_from_fork+0x14/0x28
   Exception stack(0xf0951fb0 to 0xf0951ff8)
Quoted 1 line

This needs Florians review, since for v1 he thinks it will cause regressions.

Andrew

#4 lore raw
Quoted 38 lines · Xiaolei

add Florian

thanks

xiaolei

Quoted 31 lines · Andrew Lunn

CAUTION: This email comes from a non Wind River email account!
Do not click links or open attachments unless you recognize the sender and know the content is safe.

Quoted 24 lines · Xiaolei Wang

If the external phy used by current mac interface is managed by another mac interface, it means that this network port cannot work independently, especially when the system suspend and resume, the following trace may appear, so we should create a device link between phy dev and mac dev.

   WARNING: CPU: 0 PID: 24 at drivers/net/phy/phy.c:983 phy_error+0x20/0x68
   Modules linked in:
   CPU: 0 PID: 24 Comm: kworker/0:2 Not tainted 6.1.0-rc3-00011-g5aaef24b5c6d-dirty #34
   Hardware name: Freescale i.MX6 SoloX (Device Tree)
   Workqueue: events_power_efficient phy_state_machine
   unwind_backtrace from show_stack+0x10/0x14
   show_stack from dump_stack_lvl+0x68/0x90
   dump_stack_lvl from __warn+0xb4/0x24c
   __warn from warn_slowpath_fmt+0x5c/0xd8
   warn_slowpath_fmt from phy_error+0x20/0x68
   phy_error from phy_state_machine+0x22c/0x23c
   phy_state_machine from process_one_work+0x288/0x744
   process_one_work from worker_thread+0x3c/0x500
   worker_thread from kthread+0xf0/0x114
   kthread from ret_from_fork+0x14/0x28
   Exception stack(0xf0951fb0 to 0xf0951ff8)
Quoted 1 line

This needs Florians review, since for v1 he thinks it will cause regressions.

Please give me until Tuesday to give this patch some proper testing, thanks!

Signature · 1 line
Florian
#5 lore raw
Quoted 42 lines · Florian Fainelli

CAUTION: This email comes from a non Wind River email account!
Do not click links or open attachments unless you recognize the sender and know the content is safe.

Quoted 38 lines · Xiaolei

add Florian

thanks

xiaolei

Quoted 31 lines · Andrew Lunn

CAUTION: This email comes from a non Wind River email account! Do not click links or open attachments unless you recognize the sender and know the content is safe.

Quoted 26 lines · Xiaolei Wang

If the external phy used by current mac interface is managed by another mac interface, it means that this network port cannot work independently, especially when the system suspend and resume, the following trace may appear, so we should create a device link between phy dev and mac dev.

   WARNING: CPU: 0 PID: 24 at drivers/net/phy/phy.c:983
phy_error+0x20/0x68
   Modules linked in:
   CPU: 0 PID: 24 Comm: kworker/0:2 Not tainted
6.1.0-rc3-00011-g5aaef24b5c6d-dirty #34
   Hardware name: Freescale i.MX6 SoloX (Device Tree)
   Workqueue: events_power_efficient phy_state_machine
   unwind_backtrace from show_stack+0x10/0x14
   show_stack from dump_stack_lvl+0x68/0x90
   dump_stack_lvl from __warn+0xb4/0x24c
   __warn from warn_slowpath_fmt+0x5c/0xd8
   warn_slowpath_fmt from phy_error+0x20/0x68
   phy_error from phy_state_machine+0x22c/0x23c
   phy_state_machine from process_one_work+0x288/0x744
   process_one_work from worker_thread+0x3c/0x500
   worker_thread from kthread+0xf0/0x114
   kthread from ret_from_fork+0x14/0x28
   Exception stack(0xf0951fb0 to 0xf0951ff8)
Quoted 5 lines
Quoted 3 lines
Quoted 3 lines
Quoted 1 line

This needs Florians review, since for v1 he thinks it will cause regressions.

Please give me until Tuesday to give this patch some proper testing, thanks!

Hi

Test on imx6sx
Before adding the patch:

ifconfig eth1 up
echo enabled > /sys/class/tty/ttymxc0/power/wakeup echo mem > /sys/power/state

The following problems arise:

WARNING: CPU: 0 PID: 24 at drivers/net/phy/phy.c:983
phy_error+0x20/0x68
   Modules linked in:
   CPU: 0 PID: 24 Comm: kworker/0:2 Not tainted
6.1.0-rc3-00011-g5aaef24b5c6d-dirty #34
   Hardware name: Freescale i.MX6 SoloX (Device Tree)
   Workqueue: events_power_efficient phy_state_machine
   unwind_backtrace from show_stack+0x10/0x14
   show_stack from dump_stack_lvl+0x68/0x90
   dump_stack_lvl from __warn+0xb4/0x24c
   __warn from warn_slowpath_fmt+0x5c/0xd8
   warn_slowpath_fmt from phy_error+0x20/0x68
   phy_error from phy_state_machine+0x22c/0x23c
   phy_state_machine from process_one_work+0x288/0x744
   process_one_work from worker_thread+0x3c/0x500
   worker_thread from kthread+0xf0/0x114
   kthread from ret_from_fork+0x14/0x28

   Exception stack (0xf0951fb0 to 0xf0951ff8)

After applying the patch:

ifconfig eth1 up

echo enabled > /sys/class/tty/ttymxc0/power/wakeup echo mem > /sys/power/state

eth1 will link normally

Since I don't have more boards here, I haven't tested and analyzed the different situations of other boards. If you need more test records, please wait, I will collect some hardware to verify different situations.

thanks

xiaolei

Quoted 2 lines
Signature · 1 line
Florian
Reviewed-by Tested-by on v2 1/1
#6 lore raw
In reply to Xiaolei Wang
Quoted 24 lines · Xiaolei Wang

If the external phy used by current mac interface is managed by another mac interface, it means that this network port cannot work independently, especially when the system suspend and resume, the following trace may appear, so we should create a device link between phy dev and mac dev.

   WARNING: CPU: 0 PID: 24 at drivers/net/phy/phy.c:983 phy_error+0x20/0x68
   Modules linked in:
   CPU: 0 PID: 24 Comm: kworker/0:2 Not tainted 6.1.0-rc3-00011-g5aaef24b5c6d-dirty #34
   Hardware name: Freescale i.MX6 SoloX (Device Tree)
   Workqueue: events_power_efficient phy_state_machine
   unwind_backtrace from show_stack+0x10/0x14
   show_stack from dump_stack_lvl+0x68/0x90
   dump_stack_lvl from __warn+0xb4/0x24c
   __warn from warn_slowpath_fmt+0x5c/0xd8
   warn_slowpath_fmt from phy_error+0x20/0x68
   phy_error from phy_state_machine+0x22c/0x23c
   phy_state_machine from process_one_work+0x288/0x744
   process_one_work from worker_thread+0x3c/0x500
   worker_thread from kthread+0xf0/0x114
   kthread from ret_from_fork+0x14/0x28
   Exception stack(0xf0951fb0 to 0xf0951ff8)
Quoted 1 line

Tested with bcmgenet and bcmsysport/bcm_sf2:

- suspend/resume using rtcwake w/ Wake-on-LAN disabled
- suspend/resume using rtcwale w/ Wake-on-LAN enabled
- reboot -f (which does exercise the shutdown path which has ties with device_links)
- binding/unbinding PHY driver

There was no change to the ordering for GENET, however there was a change of ordering for the DSA (bcmsysport/bcm_sf2) combination but it seemed to make more sense the way it was, in that we suspended the switch first and later the Ethernet controller attached to the switch.

Thanks for your patience.

Signature · 1 line
Florian
#8 lore raw
Quoted 5 lines · Jakub Kicinski
Quoted 2 lines · On Mon, 28 Nov 2022 13:05:09 -0800 Florian Fainelli wrote:

Thanks! Is this for next or for net?

I would play it safe and schedule it for net-next so we can be made aware of possible regressions, if any. Platforms affected like the one Xiaolei worked on would likely get this back ported into $vendor tree.

Signature · 1 line
Florian
maintainer
#9 lore raw
In reply to Xiaolei Wang
Quoted 65 lines · Xiaolei Wang

If the external phy used by current mac interface is managed by another mac interface, it means that this network port cannot work independently, especially when the system suspend and resume, the following trace may appear, so we should create a device link between phy dev and mac dev.

  WARNING: CPU: 0 PID: 24 at drivers/net/phy/phy.c:983 phy_error+0x20/0x68
  Modules linked in:
  CPU: 0 PID: 24 Comm: kworker/0:2 Not tainted 6.1.0-rc3-00011-g5aaef24b5c6d-dirty #34
  Hardware name: Freescale i.MX6 SoloX (Device Tree)
  Workqueue: events_power_efficient phy_state_machine
  unwind_backtrace from show_stack+0x10/0x14
  show_stack from dump_stack_lvl+0x68/0x90
  dump_stack_lvl from __warn+0xb4/0x24c
  __warn from warn_slowpath_fmt+0x5c/0xd8
  warn_slowpath_fmt from phy_error+0x20/0x68
  phy_error from phy_state_machine+0x22c/0x23c
  phy_state_machine from process_one_work+0x288/0x744
  process_one_work from worker_thread+0x3c/0x500
  worker_thread from kthread+0xf0/0x114
  kthread from ret_from_fork+0x14/0x28
  Exception stack(0xf0951fb0 to 0xf0951ff8)
---
 drivers/net/phy/phy_device.c | 12 ++++++++++++
 include/linux/phy.h          |  2 ++
 2 files changed, 14 insertions(+)
drivers/net/phy/phy_device.c +12 -0
index 57849ac0384e..ca6d12f37066 100644
@@ -1511,6 +1511,15 @@ int phy_attach_direct(struct net_device *dev, struct phy_device *phydev,
 	phy_resume(phydev);
 	phy_led_triggers_register(phydev);
 
+	/**
+	 * If the external phy used by current mac interface is managed by
+	 * another mac interface, so we should create a device link between
+	 * phy dev and mac dev.
+	 */
+	if (phydev->mdio.bus->parent && dev->dev.parent != phydev->mdio.bus->parent)
+		phydev->devlink = device_link_add(dev->dev.parent, &phydev->mdio.dev,
+						  DL_FLAG_PM_RUNTIME | DL_FLAG_STATELESS);
+
 	return err;
 
 error:
@@ -1748,6 +1757,9 @@ void phy_detach(struct phy_device *phydev)
 	struct module *ndev_owner = NULL;
 	struct mii_bus *bus;
 
+	if (phydev->devlink)
+		device_link_del(phydev->devlink);
+
 	if (phydev->sysfs_links) {
 		if (dev)
 			sysfs_remove_link(&dev->dev.kobj, "phydev");
include/linux/phy.h +0 -0
index ddf66198f751..f7f8b909fed0 100644

include/linux/phy.h struct phy_device { · line 620

Quoted 4 lines
include/linux/phy.h +1 -0
@@ -617,6 +617,8 @@ struct phy_device {
 	/* And management functions */
 	struct phy_driver *drv;
 
+	struct device_link *devlink;

Sorry for the late nit picking, but could you please add the kdoc documentation for this new field?

Also, please specify explicitly the net-next target tree on repost, as per Florian's request.

Thanks,

Paolo