r8169: RTL8125B restores default LEDSEL values on Link Up, preventing persistent LED configuration

r8169: RTL8125B restores default LEDSEL values on Link Up, preventing persistent LED configuration

[Huan Zhao]

Hardware: Beelink SER8 mini PC
NIC: Realtek RTL8125B (XID 0x641, RTL_GIGA_MAC_VER_63, rev 05)
Kernel: 6.19.6-arch1-1
Driver: r8169 (in-tree)
CONFIG_R8169_LEDS=y
CONFIG_LEDS_TRIGGER_NETDEV=m

Problem:
The RTL8125B chip restores its LEDSEL registers to hardware default values
on every Link Up event. The r8169 driver does not reapply the LED
configuration after Link Up, making it impossible to persistently configure
LEDs (e.g. disable them) through the standard kernel LED subsystem.

Details:

The Link Up path in the driver is:

  r8169_phylink_handler()       [r8169_main.c:4757]
    -> rtl_link_chg_patch()     [r8169_main.c:1597]
    -> rtl_enable_tx_lpi()
    -> pm_request_resume()

rtl_link_chg_patch() has no handling for RTL_GIGA_MAC_VER_63 and does not
touch the LEDSEL registers. Similarly, rtl_hw_start_8125b() and
rtl_hw_start_8125_common(), which are called on hardware initialization,
do not write to any LEDSEL register (LEDSEL0=0x18, LEDSEL1=0x86,
LEDSEL2=0x84, LEDSEL3=0x96).

The only driver code that writes LEDSEL registers is rtl8125_set_led_mode()
in r8169_leds.c, called exclusively from the LED classdev hw_control
callbacks. Therefore, any LED configuration applied via sysfs is silently
overwritten by the hardware chip's own firmware on every Link Up event,
before the driver has a chance to reapply it.

This was verified by observing that:
1. Writing 0x0000 to all LEDSEL registers (via BAR0 MMIO) before Link Up
   successfully turns off all LEDs.
2. After Link Up completes, the LEDSEL registers revert to their hardware
   defaults, turning the LEDs back on.
3. Writing 0x0000 to LEDSEL registers immediately after Link Up (via
   NetworkManager dispatcher triggered on dhcp4-change) successfully keeps
   the LEDs off persistently.

Original LEDSEL register values (hardware defaults) on this machine:
  LEDSEL0 (0x18) = 0x0002  (LINK_100)
  LEDSEL1 (0x86) = 0x0028  (LINK_2500 + LINK_1000)
  LEDSEL2 (0x84) = 0x022b  (ACT + LINK_2500 + LINK_1000 + LINK_100 + LINK_10)
  LEDSEL3 (0x96) = 0x0020  (LINK_2500)

Secondary issue: CONFIG_LEDS_TRIGGER_NETDEV=m causes silent failure

On Arch Linux, CONFIG_LEDS_TRIGGER_NETDEV=m (module, not built-in).
The r8169 driver uses /* ignore errors */ when calling
led_classdev_register() in rtl8125_setup_led_ldev(). If ledtrig_netdev
is not loaded when r8169 initializes, the registration fails silently.
The result is that /sys/class/leds/enp2s0-* entries appear to exist but
are dummy stubs with no hardware control capability. Writing to their
trigger or brightness attributes has no effect on the physical LEDs.

This was verified by manually loading ledtrig_netdev before r8169, which
resulted in functional LED classdev entries. The fix was to add
ledtrig_netdev to /etc/modules-load.d/ to ensure it loads before r8169.

However, even with ledtrig_netdev loaded correctly and offloaded=1
confirmed, the LED configuration is still lost on every Link Up due to
the hardware reset described above.

Suggested fix:

In r8169_phylink_handler(), after Link Up is confirmed, reapply the
current LED configuration by notifying the LED classdevs. For example:

  if (netif_carrier_ok(ndev)) {
      rtl_link_chg_patch(tp);
      rtl_enable_tx_lpi(tp, tp->phydev->enable_tx_lpi);
      /* Reapply LED configuration after hardware resets LEDSEL on Link Up */
      if (tp->leds)
          led_classdev_notify_clients(tp->leds);  /* or equivalent */
      pm_request_resume(d);
  }

The exact mechanism for reapplying the LED configuration (whether through
the LED classdev framework or by directly re-calling rtl8125_set_led_mode()
with saved values) is left to your discretion.

Thanks,
Huan Zhao

Re: r8169: RTL8125B restores default LEDSEL values on Link Up, preventing persistent LED configuration

[Heiner Kallweit]

On 15.03.2026 02:46, Huan Zhao wrote:
> Hardware: Beelink SER8 mini PC
> NIC: Realtek RTL8125B (XID 0x641, RTL_GIGA_MAC_VER_63, rev 05)
> Kernel: 6.19.6-arch1-1
> Driver: r8169 (in-tree)
> CONFIG_R8169_LEDS=y
> CONFIG_LEDS_TRIGGER_NETDEV=m
> 
> Problem:
> The RTL8125B chip restores its LEDSEL registers to hardware default values
> on every Link Up event. The r8169 driver does not reapply the LED
> configuration after Link Up, making it impossible to persistently configure
> LEDs (e.g. disable them) through the standard kernel LED subsystem.
> 
> Details:
> 
> The Link Up path in the driver is:
> 
>   r8169_phylink_handler()       [r8169_main.c:4757]
>     -> rtl_link_chg_patch()     [r8169_main.c:1597]
>     -> rtl_enable_tx_lpi()
>     -> pm_request_resume()
> 
> rtl_link_chg_patch() has no handling for RTL_GIGA_MAC_VER_63 and does not
> touch the LEDSEL registers. Similarly, rtl_hw_start_8125b() and
> rtl_hw_start_8125_common(), which are called on hardware initialization,
> do not write to any LEDSEL register (LEDSEL0=0x18, LEDSEL1=0x86,
> LEDSEL2=0x84, LEDSEL3=0x96).
> 
> The only driver code that writes LEDSEL registers is rtl8125_set_led_mode()
> in r8169_leds.c, called exclusively from the LED classdev hw_control
> callbacks. Therefore, any LED configuration applied via sysfs is silently
> overwritten by the hardware chip's own firmware on every Link Up event,
> before the driver has a chance to reapply it.
> 
> This was verified by observing that:
> 1. Writing 0x0000 to all LEDSEL registers (via BAR0 MMIO) before Link Up
>    successfully turns off all LEDs.
> 2. After Link Up completes, the LEDSEL registers revert to their hardware
>    defaults, turning the LEDs back on.
> 3. Writing 0x0000 to LEDSEL registers immediately after Link Up (via
>    NetworkManager dispatcher triggered on dhcp4-change) successfully keeps
>    the LEDs off persistently.
> 
> Original LEDSEL register values (hardware defaults) on this machine:
>   LEDSEL0 (0x18) = 0x0002  (LINK_100)
>   LEDSEL1 (0x86) = 0x0028  (LINK_2500 + LINK_1000)
>   LEDSEL2 (0x84) = 0x022b  (ACT + LINK_2500 + LINK_1000 + LINK_100 + LINK_10)
>   LEDSEL3 (0x96) = 0x0020  (LINK_2500)
> 
Thanks for the report. For preparation of a fix it would be helpful to know
whether this behavior is specific to RTL8125B, or whether this applies
to all chip variants supported by the driver.
A reset to defaults after a deep power-saving state I'd understand, but
a reset to power-up defaults on link-up sounds quirky. But maybe it's
not power-up defaults, but link speed specific defaults.

Hau, Javen: As I don't have access to chip documentation, I'd appreciate
a comment from your side. Thanks!

RE: r8169: RTL8125B restores default LEDSEL values on Link Up, preventing persistent LED configuration

[Javen]

>On 15.03.2026 02:46, Huan Zhao wrote:
>> Hardware: Beelink SER8 mini PC
>> NIC: Realtek RTL8125B (XID 0x641, RTL_GIGA_MAC_VER_63, rev 05)
>> Kernel: 6.19.6-arch1-1
>> Driver: r8169 (in-tree)
>> CONFIG_R8169_LEDS=y
>> CONFIG_LEDS_TRIGGER_NETDEV=m
>>
>> Problem:
>> The RTL8125B chip restores its LEDSEL registers to hardware default
>> values on every Link Up event. The r8169 driver does not reapply the
>> LED configuration after Link Up, making it impossible to persistently
>> configure LEDs (e.g. disable them) through the standard kernel LED
>subsystem.
>>
>> Details:
>>
>> The Link Up path in the driver is:
>>
>>   r8169_phylink_handler()       [r8169_main.c:4757]
>>     -> rtl_link_chg_patch()     [r8169_main.c:1597]
>>     -> rtl_enable_tx_lpi()
>>     -> pm_request_resume()
>>
>> rtl_link_chg_patch() has no handling for RTL_GIGA_MAC_VER_63 and does
>> not touch the LEDSEL registers. Similarly, rtl_hw_start_8125b() and
>> rtl_hw_start_8125_common(), which are called on hardware
>> initialization, do not write to any LEDSEL register (LEDSEL0=0x18,
>> LEDSEL1=0x86, LEDSEL2=0x84, LEDSEL3=0x96).
>>
>> The only driver code that writes LEDSEL registers is
>> rtl8125_set_led_mode() in r8169_leds.c, called exclusively from the
>> LED classdev hw_control callbacks. Therefore, any LED configuration
>> applied via sysfs is silently overwritten by the hardware chip's own
>> firmware on every Link Up event, before the driver has a chance to reapply
>it.
>>
>> This was verified by observing that:
>> 1. Writing 0x0000 to all LEDSEL registers (via BAR0 MMIO) before Link Up
>>    successfully turns off all LEDs.
>> 2. After Link Up completes, the LEDSEL registers revert to their hardware
>>    defaults, turning the LEDs back on.
>> 3. Writing 0x0000 to LEDSEL registers immediately after Link Up (via
>>    NetworkManager dispatcher triggered on dhcp4-change) successfully
>keeps
>>    the LEDs off persistently.
>>
>> Original LEDSEL register values (hardware defaults) on this machine:
>>   LEDSEL0 (0x18) = 0x0002  (LINK_100)
>>   LEDSEL1 (0x86) = 0x0028  (LINK_2500 + LINK_1000)
>>   LEDSEL2 (0x84) = 0x022b  (ACT + LINK_2500 + LINK_1000 + LINK_100 +
>LINK_10)
>>   LEDSEL3 (0x96) = 0x0020  (LINK_2500)
>>
>Thanks for the report. For preparation of a fix it would be helpful to know
>whether this behavior is specific to RTL8125B, or whether this applies to all
>chip variants supported by the driver.
>A reset to defaults after a deep power-saving state I'd understand, but a reset
>to power-up defaults on link-up sounds quirky. But maybe it's not power-up
>defaults, but link speed specific defaults.
>
>Hau, Javen: As I don't have access to chip documentation, I'd appreciate a
>comment from your side. Thanks!

My platform:
 Hardware: ASUSTek Z890-PLUS
 NIC: Realtek RTL8125B (XID 0x641, RTL_GIGA_MAC_VER_63, rev 04)
 Kernel: 7.0.0-070000rc1-generic
 Driver: r8169 (in-tree)

I tried the following steps:
1. link up, LEDs on
2. link down, LED off
3. writing 0x0000 to mmio address through devmem2, link up, LEDS are still off

1. link up, LEDs on
2. echo 0 | sudo tee /sys/class/leds/enp130s0-2::lan/link_1000, LEDs off
3. link down, link up, LEDs are still off

Original LEDSEL register values ​​on my pc: 
LEDSEL0 (0x18) = 0x0201 
LEDSEL1 (0x86) = 0x0202 
LEDSEL2 (0x84) = 0x0208 
LEDSEL3 (0x96) = 0x0220

I can't see the problem on my pc. I want to know if there's anything wrong with the steps I took to reproduce it.

Thanks

Re: r8169: RTL8125B restores default LEDSEL values on Link Up, preventing persistent LED configuration

[Heiner Kallweit]

On 19.03.2026 03:35, Javen wrote:
>> On 15.03.2026 02:46, Huan Zhao wrote:
>>> Hardware: Beelink SER8 mini PC
>>> NIC: Realtek RTL8125B (XID 0x641, RTL_GIGA_MAC_VER_63, rev 05)
>>> Kernel: 6.19.6-arch1-1
>>> Driver: r8169 (in-tree)
>>> CONFIG_R8169_LEDS=y
>>> CONFIG_LEDS_TRIGGER_NETDEV=m
>>>
>>> Problem:
>>> The RTL8125B chip restores its LEDSEL registers to hardware default
>>> values on every Link Up event. The r8169 driver does not reapply the
>>> LED configuration after Link Up, making it impossible to persistently
>>> configure LEDs (e.g. disable them) through the standard kernel LED
>> subsystem.
>>>
>>> Details:
>>>
>>> The Link Up path in the driver is:
>>>
>>>   r8169_phylink_handler()       [r8169_main.c:4757]
>>>     -> rtl_link_chg_patch()     [r8169_main.c:1597]
>>>     -> rtl_enable_tx_lpi()
>>>     -> pm_request_resume()
>>>
>>> rtl_link_chg_patch() has no handling for RTL_GIGA_MAC_VER_63 and does
>>> not touch the LEDSEL registers. Similarly, rtl_hw_start_8125b() and
>>> rtl_hw_start_8125_common(), which are called on hardware
>>> initialization, do not write to any LEDSEL register (LEDSEL0=0x18,
>>> LEDSEL1=0x86, LEDSEL2=0x84, LEDSEL3=0x96).
>>>
>>> The only driver code that writes LEDSEL registers is
>>> rtl8125_set_led_mode() in r8169_leds.c, called exclusively from the
>>> LED classdev hw_control callbacks. Therefore, any LED configuration
>>> applied via sysfs is silently overwritten by the hardware chip's own
>>> firmware on every Link Up event, before the driver has a chance to reapply
>> it.
>>>
>>> This was verified by observing that:
>>> 1. Writing 0x0000 to all LEDSEL registers (via BAR0 MMIO) before Link Up
>>>    successfully turns off all LEDs.
>>> 2. After Link Up completes, the LEDSEL registers revert to their hardware
>>>    defaults, turning the LEDs back on.
>>> 3. Writing 0x0000 to LEDSEL registers immediately after Link Up (via
>>>    NetworkManager dispatcher triggered on dhcp4-change) successfully
>> keeps
>>>    the LEDs off persistently.
>>>
>>> Original LEDSEL register values (hardware defaults) on this machine:
>>>   LEDSEL0 (0x18) = 0x0002  (LINK_100)
>>>   LEDSEL1 (0x86) = 0x0028  (LINK_2500 + LINK_1000)
>>>   LEDSEL2 (0x84) = 0x022b  (ACT + LINK_2500 + LINK_1000 + LINK_100 +
>> LINK_10)
>>>   LEDSEL3 (0x96) = 0x0020  (LINK_2500)
>>>
>> Thanks for the report. For preparation of a fix it would be helpful to know
>> whether this behavior is specific to RTL8125B, or whether this applies to all
>> chip variants supported by the driver.
>> A reset to defaults after a deep power-saving state I'd understand, but a reset
>> to power-up defaults on link-up sounds quirky. But maybe it's not power-up
>> defaults, but link speed specific defaults.
>>
>> Hau, Javen: As I don't have access to chip documentation, I'd appreciate a
>> comment from your side. Thanks!
> 
> My platform:
>  Hardware: ASUSTek Z890-PLUS
>  NIC: Realtek RTL8125B (XID 0x641, RTL_GIGA_MAC_VER_63, rev 04)
>  Kernel: 7.0.0-070000rc1-generic
>  Driver: r8169 (in-tree)
> 
> I tried the following steps:
> 1. link up, LEDs on
> 2. link down, LED off
> 3. writing 0x0000 to mmio address through devmem2, link up, LEDS are still off
> 
> 1. link up, LEDs on
> 2. echo 0 | sudo tee /sys/class/leds/enp130s0-2::lan/link_1000, LEDs off
> 3. link down, link up, LEDs are still off
> 
> Original LEDSEL register values ​​on my pc: 
> LEDSEL0 (0x18) = 0x0201 
> LEDSEL1 (0x86) = 0x0202 
> LEDSEL2 (0x84) = 0x0208 
> LEDSEL3 (0x96) = 0x0220
> 
> I can't see the problem on my pc. I want to know if there's anything wrong with the steps I took to reproduce it.
> 
Thanks for having a look at this. In the mean time the original poster admitted
that the report was based on a wrong interpretation of an issue.

> Thanks

Re: r8169: RTL8125B restores default LEDSEL values on Link Up, preventing persistent LED configuration

[Heiner Kallweit]

On 15.03.2026 02:46, Huan Zhao wrote:
> Hardware: Beelink SER8 mini PC
> NIC: Realtek RTL8125B (XID 0x641, RTL_GIGA_MAC_VER_63, rev 05)
> Kernel: 6.19.6-arch1-1
> Driver: r8169 (in-tree)
> CONFIG_R8169_LEDS=y
> CONFIG_LEDS_TRIGGER_NETDEV=m
> 
> Problem:
> The RTL8125B chip restores its LEDSEL registers to hardware default values
> on every Link Up event. The r8169 driver does not reapply the LED
> configuration after Link Up, making it impossible to persistently configure
> LEDs (e.g. disable them) through the standard kernel LED subsystem.
> 
> Details:
> 
> The Link Up path in the driver is:
> 
>   r8169_phylink_handler()       [r8169_main.c:4757]
>     -> rtl_link_chg_patch()     [r8169_main.c:1597]
>     -> rtl_enable_tx_lpi()
>     -> pm_request_resume()
> 
> rtl_link_chg_patch() has no handling for RTL_GIGA_MAC_VER_63 and does not
> touch the LEDSEL registers. Similarly, rtl_hw_start_8125b() and
> rtl_hw_start_8125_common(), which are called on hardware initialization,
> do not write to any LEDSEL register (LEDSEL0=0x18, LEDSEL1=0x86,
> LEDSEL2=0x84, LEDSEL3=0x96).
> 
> The only driver code that writes LEDSEL registers is rtl8125_set_led_mode()
> in r8169_leds.c, called exclusively from the LED classdev hw_control
> callbacks. Therefore, any LED configuration applied via sysfs is silently
> overwritten by the hardware chip's own firmware on every Link Up event,
> before the driver has a chance to reapply it.
> 
> This was verified by observing that:
> 1. Writing 0x0000 to all LEDSEL registers (via BAR0 MMIO) before Link Up
>    successfully turns off all LEDs.
> 2. After Link Up completes, the LEDSEL registers revert to their hardware
>    defaults, turning the LEDs back on.
> 3. Writing 0x0000 to LEDSEL registers immediately after Link Up (via
>    NetworkManager dispatcher triggered on dhcp4-change) successfully keeps
>    the LEDs off persistently.
> 
> Original LEDSEL register values (hardware defaults) on this machine:
>   LEDSEL0 (0x18) = 0x0002  (LINK_100)
>   LEDSEL1 (0x86) = 0x0028  (LINK_2500 + LINK_1000)
>   LEDSEL2 (0x84) = 0x022b  (ACT + LINK_2500 + LINK_1000 + LINK_100 + LINK_10)
>   LEDSEL3 (0x96) = 0x0020  (LINK_2500)
> 
> Secondary issue: CONFIG_LEDS_TRIGGER_NETDEV=m causes silent failure
> 
> On Arch Linux, CONFIG_LEDS_TRIGGER_NETDEV=m (module, not built-in).
> The r8169 driver uses /* ignore errors */ when calling
> led_classdev_register() in rtl8125_setup_led_ldev(). If ledtrig_netdev
> is not loaded when r8169 initializes, the registration fails silently.

No, it doesn't fail. Just that no trigger is assigned by default.
Once you load ledtrig-netdev module and assign netdev trigger to
the LED, you can control LED hw modes.


> The result is that /sys/class/leds/enp2s0-* entries appear to exist but
> are dummy stubs with no hardware control capability. Writing to their
> trigger or brightness attributes has no effect on the physical LEDs.
> 
Setting brightness manually isn't supported by HW.

> This was verified by manually loading ledtrig_netdev before r8169, which
> resulted in functional LED classdev entries. The fix was to add
> ledtrig_netdev to /etc/modules-load.d/ to ensure it loads before r8169.
> 
> However, even with ledtrig_netdev loaded correctly and offloaded=1
> confirmed, the LED configuration is still lost on every Link Up due to
> the hardware reset described above.
> 
> Suggested fix:
> 
> In r8169_phylink_handler(), after Link Up is confirmed, reapply the
> current LED configuration by notifying the LED classdevs. For example:
> 
>   if (netif_carrier_ok(ndev)) {
>       rtl_link_chg_patch(tp);
>       rtl_enable_tx_lpi(tp, tp->phydev->enable_tx_lpi);
>       /* Reapply LED configuration after hardware resets LEDSEL on Link Up */
>       if (tp->leds)
>           led_classdev_notify_clients(tp->leds);  /* or equivalent */
>       pm_request_resume(d);
>   }
> 
> The exact mechanism for reapplying the LED configuration (whether through
> the LED classdev framework or by directly re-calling rtl8125_set_led_mode()
> with saved values) is left to your discretion.
> 
> Thanks,
> Huan Zhao

Re: r8169: RTL8125B restores default LEDSEL values on Link Up, preventing persistent LED configuration

[Huan Zhao]

You are correct. I apologize for the incorrect analysis in my previous
email. Further testing has clarified the actual behavior.

Summary of findings after additional testing:

1. The LEDSEL registers do NOT reset on Link Up.

    After writing 0x0000 to all LEDSEL registers via BAR0 MMIO and
    triggering multiple Link Up/Down cycles, the registers retained
    their zero values throughout. The hardware does not restore any
    defaults on link state changes.

2. The real problem is CONFIG_LEDS_TRIGGER_NETDEV=m on Arch Linux.

    As you clarified, the LED classdev registers successfully without
    ledtrig_netdev present — just without a trigger assigned. However,
    the link_* sysfs files (link_100, link_2500, etc.) do not appear
    until the netdev trigger is explicitly assigned. Without these
    files, there is no way to configure the LED modes through the
    standard interface.

    The workaround is to ensure ledtrig_netdev loads before r8169
    via /etc/modules-load.d/.

3. Persistent LED disable works correctly via the standard interface.

    With ledtrig_netdev loaded before r8169, the following sequence
    successfully and persistently disables all LEDs:

      echo netdev > /sys/class/leds/enp2s0-N::lan/trigger
      echo 0 > /sys/class/leds/enp2s0-N::lan/link_10
      echo 0 > /sys/class/leds/enp2s0-N::lan/link_100
      echo 0 > /sys/class/leds/enp2s0-N::lan/link_1000
      echo 0 > /sys/class/leds/enp2s0-N::lan/link_2500
      echo 0 > /sys/class/leds/enp2s0-N::lan/tx
      echo 0 > /sys/class/leds/enp2s0-N::lan/rx

    This survives Link Up/Down cycles and network cable
    insertion/removal without any further intervention.

    This is implemented as a systemd oneshot service running after
    network.target, at which point all link_* files are immediately
    writable with no waiting required.

4. Remaining question: udev timing issue.

    When the same script runs from a udev RUN rule (triggered on
    ACTION=="move" for the network interface rename), the link_*
    files are not yet writable at that point, even after
    trigger=netdev is set. The files only become writable after
    a delay of up to several seconds. Is there a reliable udev
    event or sysfs condition that indicates hw_control is fully
    ready for writing?

Thanks,
Huan Zhao

On Sun, Mar 15, 2026 at 02:55:17PM +0100, Heiner Kallweit wrote:
>On 15.03.2026 02:46, Huan Zhao wrote:
>> Hardware: Beelink SER8 mini PC
>> NIC: Realtek RTL8125B (XID 0x641, RTL_GIGA_MAC_VER_63, rev 05)
>> Kernel: 6.19.6-arch1-1
>> Driver: r8169 (in-tree)
>> CONFIG_R8169_LEDS=y
>> CONFIG_LEDS_TRIGGER_NETDEV=m
>>
>> Problem:
>> The RTL8125B chip restores its LEDSEL registers to hardware default values
>> on every Link Up event. The r8169 driver does not reapply the LED
>> configuration after Link Up, making it impossible to persistently configure
>> LEDs (e.g. disable them) through the standard kernel LED subsystem.
>>
>> Details:
>>
>> The Link Up path in the driver is:
>>
>>   r8169_phylink_handler()       [r8169_main.c:4757]
>>     -> rtl_link_chg_patch()     [r8169_main.c:1597]
>>     -> rtl_enable_tx_lpi()
>>     -> pm_request_resume()
>>
>> rtl_link_chg_patch() has no handling for RTL_GIGA_MAC_VER_63 and does not
>> touch the LEDSEL registers. Similarly, rtl_hw_start_8125b() and
>> rtl_hw_start_8125_common(), which are called on hardware initialization,
>> do not write to any LEDSEL register (LEDSEL0=0x18, LEDSEL1=0x86,
>> LEDSEL2=0x84, LEDSEL3=0x96).
>>
>> The only driver code that writes LEDSEL registers is rtl8125_set_led_mode()
>> in r8169_leds.c, called exclusively from the LED classdev hw_control
>> callbacks. Therefore, any LED configuration applied via sysfs is silently
>> overwritten by the hardware chip's own firmware on every Link Up event,
>> before the driver has a chance to reapply it.
>>
>> This was verified by observing that:
>> 1. Writing 0x0000 to all LEDSEL registers (via BAR0 MMIO) before Link Up
>>    successfully turns off all LEDs.
>> 2. After Link Up completes, the LEDSEL registers revert to their hardware
>>    defaults, turning the LEDs back on.
>> 3. Writing 0x0000 to LEDSEL registers immediately after Link Up (via
>>    NetworkManager dispatcher triggered on dhcp4-change) successfully keeps
>>    the LEDs off persistently.
>>
>> Original LEDSEL register values (hardware defaults) on this machine:
>>   LEDSEL0 (0x18) = 0x0002  (LINK_100)
>>   LEDSEL1 (0x86) = 0x0028  (LINK_2500 + LINK_1000)
>>   LEDSEL2 (0x84) = 0x022b  (ACT + LINK_2500 + LINK_1000 + LINK_100 + LINK_10)
>>   LEDSEL3 (0x96) = 0x0020  (LINK_2500)
>>
>> Secondary issue: CONFIG_LEDS_TRIGGER_NETDEV=m causes silent failure
>>
>> On Arch Linux, CONFIG_LEDS_TRIGGER_NETDEV=m (module, not built-in).
>> The r8169 driver uses /* ignore errors */ when calling
>> led_classdev_register() in rtl8125_setup_led_ldev(). If ledtrig_netdev
>> is not loaded when r8169 initializes, the registration fails silently.
>
>No, it doesn't fail. Just that no trigger is assigned by default.
>Once you load ledtrig-netdev module and assign netdev trigger to
>the LED, you can control LED hw modes.
>
>
>> The result is that /sys/class/leds/enp2s0-* entries appear to exist but
>> are dummy stubs with no hardware control capability. Writing to their
>> trigger or brightness attributes has no effect on the physical LEDs.
>>
>Setting brightness manually isn't supported by HW.
>
>> This was verified by manually loading ledtrig_netdev before r8169, which
>> resulted in functional LED classdev entries. The fix was to add
>> ledtrig_netdev to /etc/modules-load.d/ to ensure it loads before r8169.
>>
>> However, even with ledtrig_netdev loaded correctly and offloaded=1
>> confirmed, the LED configuration is still lost on every Link Up due to
>> the hardware reset described above.
>>
>> Suggested fix:
>>
>> In r8169_phylink_handler(), after Link Up is confirmed, reapply the
>> current LED configuration by notifying the LED classdevs. For example:
>>
>>   if (netif_carrier_ok(ndev)) {
>>       rtl_link_chg_patch(tp);
>>       rtl_enable_tx_lpi(tp, tp->phydev->enable_tx_lpi);
>>       /* Reapply LED configuration after hardware resets LEDSEL on Link Up */
>>       if (tp->leds)
>>           led_classdev_notify_clients(tp->leds);  /* or equivalent */
>>       pm_request_resume(d);
>>   }
>>
>> The exact mechanism for reapplying the LED configuration (whether through
>> the LED classdev framework or by directly re-calling rtl8125_set_led_mode()
>> with saved values) is left to your discretion.
>>
>> Thanks,
>> Huan Zhao
>

Re: r8169: RTL8125B restores default LEDSEL values on Link Up, preventing persistent LED configuration

[Heiner Kallweit]

On 15.03.2026 18:41, Huan Zhao wrote:
> You are correct. I apologize for the incorrect analysis in my previous
> email. Further testing has clarified the actual behavior.
> 
> Summary of findings after additional testing:
> 
> 1. The LEDSEL registers do NOT reset on Link Up.
> 
>    After writing 0x0000 to all LEDSEL registers via BAR0 MMIO and
>    triggering multiple Link Up/Down cycles, the registers retained
>    their zero values throughout. The hardware does not restore any
>    defaults on link state changes.
> 
> 2. The real problem is CONFIG_LEDS_TRIGGER_NETDEV=m on Arch Linux.
> 
>    As you clarified, the LED classdev registers successfully without
>    ledtrig_netdev present — just without a trigger assigned. However,
>    the link_* sysfs files (link_100, link_2500, etc.) do not appear
>    until the netdev trigger is explicitly assigned. Without these
>    files, there is no way to configure the LED modes through the
>    standard interface.
> 
>    The workaround is to ensure ledtrig_netdev loads before r8169
>    via /etc/modules-load.d/.

The netdev trigger module can be loaded also after r8169, and then
be assigned as trigger.

> 
> 3. Persistent LED disable works correctly via the standard interface.
> 
>    With ledtrig_netdev loaded before r8169, the following sequence
>    successfully and persistently disables all LEDs:
> 
>      echo netdev > /sys/class/leds/enp2s0-N::lan/trigger

How is it if you add a small delay here? I'd have check whether
assigning a trigger is fully synchronous.

>      echo 0 > /sys/class/leds/enp2s0-N::lan/link_10
>      echo 0 > /sys/class/leds/enp2s0-N::lan/link_100
>      echo 0 > /sys/class/leds/enp2s0-N::lan/link_1000
>      echo 0 > /sys/class/leds/enp2s0-N::lan/link_2500
>      echo 0 > /sys/class/leds/enp2s0-N::lan/tx
>      echo 0 > /sys/class/leds/enp2s0-N::lan/rx
> 
>    This survives Link Up/Down cycles and network cable
>    insertion/removal without any further intervention.
> 
>    This is implemented as a systemd oneshot service running after
>    network.target, at which point all link_* files are immediately
>    writable with no waiting required.
> 
> 4. Remaining question: udev timing issue.
> 
>    When the same script runs from a udev RUN rule (triggered on
>    ACTION=="move" for the network interface rename), the link_*
>    files are not yet writable at that point, even after
>    trigger=netdev is set. The files only become writable after
>    a delay of up to several seconds. Is there a reliable udev
>    event or sysfs condition that indicates hw_control is fully
>    ready for writing?
> 
> Thanks,
> Huan Zhao
> 
> On Sun, Mar 15, 2026 at 02:55:17PM +0100, Heiner Kallweit wrote:
>> On 15.03.2026 02:46, Huan Zhao wrote:
>>> Hardware: Beelink SER8 mini PC
>>> NIC: Realtek RTL8125B (XID 0x641, RTL_GIGA_MAC_VER_63, rev 05)
>>> Kernel: 6.19.6-arch1-1
>>> Driver: r8169 (in-tree)
>>> CONFIG_R8169_LEDS=y
>>> CONFIG_LEDS_TRIGGER_NETDEV=m
>>>
>>> Problem:
>>> The RTL8125B chip restores its LEDSEL registers to hardware default values
>>> on every Link Up event. The r8169 driver does not reapply the LED
>>> configuration after Link Up, making it impossible to persistently configure
>>> LEDs (e.g. disable them) through the standard kernel LED subsystem.
>>>
>>> Details:
>>>
>>> The Link Up path in the driver is:
>>>
>>>   r8169_phylink_handler()       [r8169_main.c:4757]
>>>     -> rtl_link_chg_patch()     [r8169_main.c:1597]
>>>     -> rtl_enable_tx_lpi()
>>>     -> pm_request_resume()
>>>
>>> rtl_link_chg_patch() has no handling for RTL_GIGA_MAC_VER_63 and does not
>>> touch the LEDSEL registers. Similarly, rtl_hw_start_8125b() and
>>> rtl_hw_start_8125_common(), which are called on hardware initialization,
>>> do not write to any LEDSEL register (LEDSEL0=0x18, LEDSEL1=0x86,
>>> LEDSEL2=0x84, LEDSEL3=0x96).
>>>
>>> The only driver code that writes LEDSEL registers is rtl8125_set_led_mode()
>>> in r8169_leds.c, called exclusively from the LED classdev hw_control
>>> callbacks. Therefore, any LED configuration applied via sysfs is silently
>>> overwritten by the hardware chip's own firmware on every Link Up event,
>>> before the driver has a chance to reapply it.
>>>
>>> This was verified by observing that:
>>> 1. Writing 0x0000 to all LEDSEL registers (via BAR0 MMIO) before Link Up
>>>    successfully turns off all LEDs.
>>> 2. After Link Up completes, the LEDSEL registers revert to their hardware
>>>    defaults, turning the LEDs back on.
>>> 3. Writing 0x0000 to LEDSEL registers immediately after Link Up (via
>>>    NetworkManager dispatcher triggered on dhcp4-change) successfully keeps
>>>    the LEDs off persistently.
>>>
>>> Original LEDSEL register values (hardware defaults) on this machine:
>>>   LEDSEL0 (0x18) = 0x0002  (LINK_100)
>>>   LEDSEL1 (0x86) = 0x0028  (LINK_2500 + LINK_1000)
>>>   LEDSEL2 (0x84) = 0x022b  (ACT + LINK_2500 + LINK_1000 + LINK_100 + LINK_10)
>>>   LEDSEL3 (0x96) = 0x0020  (LINK_2500)
>>>
>>> Secondary issue: CONFIG_LEDS_TRIGGER_NETDEV=m causes silent failure
>>>
>>> On Arch Linux, CONFIG_LEDS_TRIGGER_NETDEV=m (module, not built-in).
>>> The r8169 driver uses /* ignore errors */ when calling
>>> led_classdev_register() in rtl8125_setup_led_ldev(). If ledtrig_netdev
>>> is not loaded when r8169 initializes, the registration fails silently.
>>
>> No, it doesn't fail. Just that no trigger is assigned by default.
>> Once you load ledtrig-netdev module and assign netdev trigger to
>> the LED, you can control LED hw modes.
>>
>>
>>> The result is that /sys/class/leds/enp2s0-* entries appear to exist but
>>> are dummy stubs with no hardware control capability. Writing to their
>>> trigger or brightness attributes has no effect on the physical LEDs.
>>>
>> Setting brightness manually isn't supported by HW.
>>
>>> This was verified by manually loading ledtrig_netdev before r8169, which
>>> resulted in functional LED classdev entries. The fix was to add
>>> ledtrig_netdev to /etc/modules-load.d/ to ensure it loads before r8169.
>>>
>>> However, even with ledtrig_netdev loaded correctly and offloaded=1
>>> confirmed, the LED configuration is still lost on every Link Up due to
>>> the hardware reset described above.
>>>
>>> Suggested fix:
>>>
>>> In r8169_phylink_handler(), after Link Up is confirmed, reapply the
>>> current LED configuration by notifying the LED classdevs. For example:
>>>
>>>   if (netif_carrier_ok(ndev)) {
>>>       rtl_link_chg_patch(tp);
>>>       rtl_enable_tx_lpi(tp, tp->phydev->enable_tx_lpi);
>>>       /* Reapply LED configuration after hardware resets LEDSEL on Link Up */
>>>       if (tp->leds)
>>>           led_classdev_notify_clients(tp->leds);  /* or equivalent */
>>>       pm_request_resume(d);
>>>   }
>>>
>>> The exact mechanism for reapplying the LED configuration (whether through
>>> the LED classdev framework or by directly re-calling rtl8125_set_led_mode()
>>> with saved values) is left to your discretion.
>>>
>>> Thanks,
>>> Huan Zhao
>>

Re: r8169: RTL8125B restores default LEDSEL values on Link Up, preventing persistent LED configuration

[Huan Zhao]

On Sun, Mar 15, 2026 at 07:06:58PM +0100, Heiner Kallweit wrote:
>On 15.03.2026 18:41, Huan Zhao wrote:
>> You are correct. I apologize for the incorrect analysis in my previous
>> email. Further testing has clarified the actual behavior.
>>
>> Summary of findings after additional testing:
>>
>> 1. The LEDSEL registers do NOT reset on Link Up.
>>
>>    After writing 0x0000 to all LEDSEL registers via BAR0 MMIO and
>>    triggering multiple Link Up/Down cycles, the registers retained
>>    their zero values throughout. The hardware does not restore any
>>    defaults on link state changes.
>>
>> 2. The real problem is CONFIG_LEDS_TRIGGER_NETDEV=m on Arch Linux.
>>
>>    As you clarified, the LED classdev registers successfully without
>>    ledtrig_netdev present — just without a trigger assigned. However,
>>    the link_* sysfs files (link_100, link_2500, etc.) do not appear
>>    until the netdev trigger is explicitly assigned. Without these
>>    files, there is no way to configure the LED modes through the
>>    standard interface.
>>
>>    The workaround is to ensure ledtrig_netdev loads before r8169
>>    via /etc/modules-load.d/.
>
>The netdev trigger module can be loaded also after r8169, and then
>be assigned as trigger.
>
Confirmed. After unloading ledtrig_netdev and reloading it after r8169
was already running, assigning trigger=netdev to the LED classdev
immediately activated hw_control (offloaded=1) and all link_* files
appeared. Load order does not matter.

This means /etc/modules-load.d/ledtrig-netdev.conf is not needed.

>>
>> 3. Persistent LED disable works correctly via the standard interface.
>>
>>    With ledtrig_netdev loaded before r8169, the following sequence
>>    successfully and persistently disables all LEDs:
>>
>>      echo netdev > /sys/class/leds/enp2s0-N::lan/trigger
>
>How is it if you add a small delay here? I'd have check whether
>assigning a trigger is fully synchronous.
>
This is exactly what we observed in testing. When the script runs from
a udev RUN rule (ACTION=="move" on network interface rename), writing
trigger=netdev succeeds but the link_* files are not yet writable
immediately after. They become writable after a variable delay of up
to several seconds.

Our current workaround in the script is to poll link_100 until it
becomes writable:

   echo netdev > "$led/trigger"
   while ! echo 0 > "$led/link_100" 2>/dev/null; do
       sleep 0.1
   done

When the script runs as a systemd oneshot service after network.target,
this polling loop completes in zero iterations — link_* files are
immediately writable at that point.

So the udev timing issue is likely caused by trigger assignment being
asynchronous, as you suspected.

>>      echo 0 > /sys/class/leds/enp2s0-N::lan/link_10
>>      echo 0 > /sys/class/leds/enp2s0-N::lan/link_100
>>      echo 0 > /sys/class/leds/enp2s0-N::lan/link_1000
>>      echo 0 > /sys/class/leds/enp2s0-N::lan/link_2500
>>      echo 0 > /sys/class/leds/enp2s0-N::lan/tx
>>      echo 0 > /sys/class/leds/enp2s0-N::lan/rx
>>
>>    This survives Link Up/Down cycles and network cable
>>    insertion/removal without any further intervention.
>>
>>    This is implemented as a systemd oneshot service running after
>>    network.target, at which point all link_* files are immediately
>>    writable with no waiting required.
>>
>> 4. Remaining question: udev timing issue.
>>
>>    When the same script runs from a udev RUN rule (triggered on
>>    ACTION=="move" for the network interface rename), the link_*
>>    files are not yet writable at that point, even after
>>    trigger=netdev is set. The files only become writable after
>>    a delay of up to several seconds. Is there a reliable udev
>>    event or sysfs condition that indicates hw_control is fully
>>    ready for writing?
>>
>> Thanks,
>> Huan Zhao
>>
>> On Sun, Mar 15, 2026 at 02:55:17PM +0100, Heiner Kallweit wrote:
>>> On 15.03.2026 02:46, Huan Zhao wrote:
>>>> Hardware: Beelink SER8 mini PC
>>>> NIC: Realtek RTL8125B (XID 0x641, RTL_GIGA_MAC_VER_63, rev 05)
>>>> Kernel: 6.19.6-arch1-1
>>>> Driver: r8169 (in-tree)
>>>> CONFIG_R8169_LEDS=y
>>>> CONFIG_LEDS_TRIGGER_NETDEV=m
>>>>
>>>> Problem:
>>>> The RTL8125B chip restores its LEDSEL registers to hardware default values
>>>> on every Link Up event. The r8169 driver does not reapply the LED
>>>> configuration after Link Up, making it impossible to persistently configure
>>>> LEDs (e.g. disable them) through the standard kernel LED subsystem.
>>>>
>>>> Details:
>>>>
>>>> The Link Up path in the driver is:
>>>>
>>>>   r8169_phylink_handler()       [r8169_main.c:4757]
>>>>     -> rtl_link_chg_patch()     [r8169_main.c:1597]
>>>>     -> rtl_enable_tx_lpi()
>>>>     -> pm_request_resume()
>>>>
>>>> rtl_link_chg_patch() has no handling for RTL_GIGA_MAC_VER_63 and does not
>>>> touch the LEDSEL registers. Similarly, rtl_hw_start_8125b() and
>>>> rtl_hw_start_8125_common(), which are called on hardware initialization,
>>>> do not write to any LEDSEL register (LEDSEL0=0x18, LEDSEL1=0x86,
>>>> LEDSEL2=0x84, LEDSEL3=0x96).
>>>>
>>>> The only driver code that writes LEDSEL registers is rtl8125_set_led_mode()
>>>> in r8169_leds.c, called exclusively from the LED classdev hw_control
>>>> callbacks. Therefore, any LED configuration applied via sysfs is silently
>>>> overwritten by the hardware chip's own firmware on every Link Up event,
>>>> before the driver has a chance to reapply it.
>>>>
>>>> This was verified by observing that:
>>>> 1. Writing 0x0000 to all LEDSEL registers (via BAR0 MMIO) before Link Up
>>>>    successfully turns off all LEDs.
>>>> 2. After Link Up completes, the LEDSEL registers revert to their hardware
>>>>    defaults, turning the LEDs back on.
>>>> 3. Writing 0x0000 to LEDSEL registers immediately after Link Up (via
>>>>    NetworkManager dispatcher triggered on dhcp4-change) successfully keeps
>>>>    the LEDs off persistently.
>>>>
>>>> Original LEDSEL register values (hardware defaults) on this machine:
>>>>   LEDSEL0 (0x18) = 0x0002  (LINK_100)
>>>>   LEDSEL1 (0x86) = 0x0028  (LINK_2500 + LINK_1000)
>>>>   LEDSEL2 (0x84) = 0x022b  (ACT + LINK_2500 + LINK_1000 + LINK_100 + LINK_10)
>>>>   LEDSEL3 (0x96) = 0x0020  (LINK_2500)
>>>>
>>>> Secondary issue: CONFIG_LEDS_TRIGGER_NETDEV=m causes silent failure
>>>>
>>>> On Arch Linux, CONFIG_LEDS_TRIGGER_NETDEV=m (module, not built-in).
>>>> The r8169 driver uses /* ignore errors */ when calling
>>>> led_classdev_register() in rtl8125_setup_led_ldev(). If ledtrig_netdev
>>>> is not loaded when r8169 initializes, the registration fails silently.
>>>
>>> No, it doesn't fail. Just that no trigger is assigned by default.
>>> Once you load ledtrig-netdev module and assign netdev trigger to
>>> the LED, you can control LED hw modes.
>>>
>>>
>>>> The result is that /sys/class/leds/enp2s0-* entries appear to exist but
>>>> are dummy stubs with no hardware control capability. Writing to their
>>>> trigger or brightness attributes has no effect on the physical LEDs.
>>>>
>>> Setting brightness manually isn't supported by HW.
>>>
>>>> This was verified by manually loading ledtrig_netdev before r8169, which
>>>> resulted in functional LED classdev entries. The fix was to add
>>>> ledtrig_netdev to /etc/modules-load.d/ to ensure it loads before r8169.
>>>>
>>>> However, even with ledtrig_netdev loaded correctly and offloaded=1
>>>> confirmed, the LED configuration is still lost on every Link Up due to
>>>> the hardware reset described above.
>>>>
>>>> Suggested fix:
>>>>
>>>> In r8169_phylink_handler(), after Link Up is confirmed, reapply the
>>>> current LED configuration by notifying the LED classdevs. For example:
>>>>
>>>>   if (netif_carrier_ok(ndev)) {
>>>>       rtl_link_chg_patch(tp);
>>>>       rtl_enable_tx_lpi(tp, tp->phydev->enable_tx_lpi);
>>>>       /* Reapply LED configuration after hardware resets LEDSEL on Link Up */
>>>>       if (tp->leds)
>>>>           led_classdev_notify_clients(tp->leds);  /* or equivalent */
>>>>       pm_request_resume(d);
>>>>   }
>>>>
>>>> The exact mechanism for reapplying the LED configuration (whether through
>>>> the LED classdev framework or by directly re-calling rtl8125_set_led_mode()
>>>> with saved values) is left to your discretion.
>>>>
>>>> Thanks,
>>>> Huan Zhao
>>>
>