Re: [RFC PATCH 0/1] net: phy: skip autoneg of ethernet(fec) on network boot

Re: [RFC PATCH 0/1] net: phy: skip autoneg of ethernet(fec) on network boot

[Florian Fainelli]

On 07/05/2018 11:07 PM, Masahiko Kimoto wrote:
> Hello,
> 
> This patch introduces auto negotiation skipping for Ethernet.
> It is useful to shorten boot time on network boot like the following
> environment;
> 
>  - target board is NXP i.MX6.
>  - NIC is fec.
>  - using u-boot as boot loader.
>  - boot from kernel and initramfs obtained via TFTP.
>  - mount remote file system and switch root to that.
>  - thus all file system is on network.
> 
> In this case, u-boot and kernel initialize NIC three times,
> once in boot loader, once in device attach and once more in phy attach.
> Each causes link auto negotiation and wait several seconds.
> However link state is stable after reset by boot loader, therefore we
> can skip hardware re-initialization of NIC in kernel.
> 
> The patch skips link down in Ethernet(i.MX's fec) driver and initialization
> in PHY layer if kernel option 'anegskip' is supplied.
> 
> By this patch boot time becomes 3secs shorter.
> 
> I think current patch is dirty hack, because;
>  - modification is split into PHY and Ethernet driver.
>  - in the case of two or more Ethernet I/F exist, currently there is
>    no way to specify whith I/F skips autonego.
> 
> I would like to implement such skpping auto negotiation feature in generic
> framework. How should we implement these requirements?

I completely acknowledge and support the use case, but your
implementation is definitively not the way to go. In my experience the
problem is usually that there may be a disagreement on the
Pause/Asym_Pause advertisement bits and that alone is responsible for
triggering a re-negotiation. Can yo check if that is the case here?

Thank you
-- 
Florian

Re: [RFC PATCH 0/1] net: phy: skip autoneg of ethernet(fec) on network boot

[Masahiko KIMOTO]

Hello Florian,

  > I completely acknowledge and support the use case, but your
  > implementation is definitively not the way to go. In my experience the
  > problem is usually that there may be a disagreement on the
  > Pause/Asym_Pause advertisement bits and that alone is responsible for
  > triggering a re-negotiation. Can yo check if that is the case here?

Thank you for your support.

Concening the problem about PAUSE/Asym_PAUSE advertisement you pointed out,
I could not understand your consideration.

Would you tell me more detail about it?


From: Florian Fainelli <f.fainelli@gmail.com>
Subject: Re: [RFC PATCH 0/1] net: phy: skip autoneg of ethernet(fec) on
 network boot
Date: Fri, 6 Jul 2018 11:41:23 -0700

  > 
  > 
  > On 07/05/2018 11:07 PM, Masahiko Kimoto wrote:
  > > Hello,
  > > 
  > > This patch introduces auto negotiation skipping for Ethernet.
  > > It is useful to shorten boot time on network boot like the following
  > > environment;
  > > 
  > >  - target board is NXP i.MX6.
  > >  - NIC is fec.
  > >  - using u-boot as boot loader.
  > >  - boot from kernel and initramfs obtained via TFTP.
  > >  - mount remote file system and switch root to that.
  > >  - thus all file system is on network.
  > > 
  > > In this case, u-boot and kernel initialize NIC three times,
  > > once in boot loader, once in device attach and once more in phy attach.
  > > Each causes link auto negotiation and wait several seconds.
  > > However link state is stable after reset by boot loader, therefore we
  > > can skip hardware re-initialization of NIC in kernel.
  > > 
  > > The patch skips link down in Ethernet(i.MX's fec) driver and initialization
  > > in PHY layer if kernel option 'anegskip' is supplied.
  > > 
  > > By this patch boot time becomes 3secs shorter.
  > > 
  > > I think current patch is dirty hack, because;
  > >  - modification is split into PHY and Ethernet driver.
  > >  - in the case of two or more Ethernet I/F exist, currently there is
  > >    no way to specify whith I/F skips autonego.
  > > 
  > > I would like to implement such skpping auto negotiation feature in generic
  > > framework. How should we implement these requirements?
  > 
  > I completely acknowledge and support the use case, but your
  > implementation is definitively not the way to go. In my experience the
  > problem is usually that there may be a disagreement on the
  > Pause/Asym_Pause advertisement bits and that alone is responsible for
  > triggering a re-negotiation. Can yo check if that is the case here?
  > 
  > Thank you
  > -- 
  > Florian
  > 

---------------------------------------------------------------------------
Masahiko KIMOTO,Ph.D. / Cybertrust Japan Co., Ltd. 
mail: masahiko.kimoto@cybertrust.co.jp
web: https://www.cybertrust.co.jp/ https://www.miraclelinux.com/

Re: [RFC PATCH 0/1] net: phy: skip autoneg of ethernet(fec) on network boot

[Andrew Lunn]

On Mon, Jul 09, 2018 at 06:29:02PM +0900, Masahiko KIMOTO wrote:
> Hello Florian,
> 
>   > I completely acknowledge and support the use case, but your
>   > implementation is definitively not the way to go. In my experience the
>   > problem is usually that there may be a disagreement on the
>   > Pause/Asym_Pause advertisement bits and that alone is responsible for
>   > triggering a re-negotiation. Can yo check if that is the case here?
> 
> Thank you for your support.
> 
> Concening the problem about PAUSE/Asym_PAUSE advertisement you pointed out,
> I could not understand your consideration.
> 
> Would you tell me more detail about it?

Hi Masahiko

You said that the PHY is being configured 3 times, and might be doing
auto-neg three times. Often a phy will not actually perform auto-neg,
or do it much faster, if the advertised bits do not change with each
re-configuration.

So check how u-boot is configuring what is advertised, and look at
what Linux is configuring. In particularly, look at the pause
configuration. If pause is configured different in u-boot and Linux,
try to make it the same. Things might go faster than.

    Andrew

Re: [RFC PATCH 0/1] net: phy: skip autoneg of ethernet(fec) on network boot

[Masahiko KIMOTO]

  > >   > I completely acknowledge and support the use case, but your
  > >   > implementation is definitively not the way to go. In my experience the
  > >   > problem is usually that there may be a disagreement on the
  > >   > Pause/Asym_Pause advertisement bits and that alone is responsible for
  > >   > triggering a re-negotiation. Can yo check if that is the case here?
  > > 
  > > Thank you for your support.
  > > 
  > > Concening the problem about PAUSE/Asym_PAUSE advertisement you pointed out,
  > > I could not understand your consideration.
  > > 
  > > Would you tell me more detail about it?
  > 
  > Hi Masahiko
  > 
  > You said that the PHY is being configured 3 times, and might be doing
  > auto-neg three times. Often a phy will not actually perform auto-neg,
  > or do it much faster, if the advertised bits do not change with each
  > re-configuration.
  > 
  > So check how u-boot is configuring what is advertised, and look at
  > what Linux is configuring. In particularly, look at the pause
  > configuration. If pause is configured different in u-boot and Linux,
  > try to make it the same. Things might go faster than.

Thank you.

I'll try to check out u-boot side by mii command and Linux(server)
side by ethtool.

---------------------------------------------------------------------------
Masahiko KIMOTO,Ph.D. / Cybertrust Japan Co., Ltd. 
mail: masahiko.kimoto@cybertrust.co.jp
web: https://www.cybertrust.co.jp/ https://www.miraclelinux.com/

Re: [RFC PATCH 0/1] net: phy: skip autoneg of ethernet(fec) on network boot

[Masahiko KIMOTO]

  >   > You said that the PHY is being configured 3 times, and might be doing
  >   > auto-neg three times. Often a phy will not actually perform auto-neg,
  >   > or do it much faster, if the advertised bits do not change with each
  >   > re-configuration.
  >   > 
  >   > So check how u-boot is configuring what is advertised, and look at
  >   > what Linux is configuring. In particularly, look at the pause
  >   > configuration. If pause is configured different in u-boot and Linux,
  >   > try to make it the same. Things might go faster than.
  > 
  > Thank you.
  > 
  > I'll try to check out u-boot side by mii command and Linux(server)
  > side by ethtool.

PHY setting in u-boot and after booting Linux is as follows;

We may avoid second and third auto-neg if both settings in u-boot and
in linux make same value, right?

---------------------------------------------------------------------------
=> mii dump 1 4
4.     (11e1)                 -- Autonegotiation advertisement register --
  (8000:0000) 4.15    =     0    next page able
  (4000:0000) 4.14    =     0    (reserved)
  (2000:0000) 4.13    =     0    remote fault
  (1000:1000) 4.12    =     1    (reserved)
  (0800:0000) 4.11    =     0    asymmetric pause
  (0400:0000) 4.10    =     0    pause enable
  (0200:0000) 4. 9    =     0    100BASE-T4 able
  (0100:0100) 4. 8    =     1    100BASE-TX full duplex able
  (0080:0080) 4. 7    =     1    100BASE-TX able
  (0040:0040) 4. 6    =     1    10BASE-T   full duplex able
  (0020:0020) 4. 5    =     1    10BASE-T   able
  (001f:0001) 4. 4- 0 =     1    selector = IEEE 802.3


=> mii dump 1 5
5.     (dd01)                 -- Autonegotiation partner abilities register --
  (8000:8000) 5.15    =     1    next page able
  (4000:4000) 5.14    =     1    acknowledge
  (2000:0000) 5.13    =     0    remote fault
  (1000:1000) 5.12    =     1    (reserved)
  (0800:0800) 5.11    =     1    asymmetric pause able
  (0400:0400) 5.10    =     1    pause able
  (0200:0000) 5. 9    =     0    100BASE-T4 able
  (0100:0100) 5. 8    =     1    100BASE-X full duplex able
  (0080:0000) 5. 7    =     0    100BASE-TX able
  (0040:0000) 5. 6    =     0    10BASE-T full duplex able
  (0020:0000) 5. 5    =     0    10BASE-T able
  (001f:0001) 5. 4- 0 =     1    selector = IEEE 802.3

---------------------------------------------------------------------------
root@imx6qdlsabresd:~# ethtool eth0
Settings for eth0:
        Supported ports: [ TP MII ]
        Supported link modes:   10baseT/Half 10baseT/Full 
                                100baseT/Half 100baseT/Full 
                                1000baseT/Full 
        Supported pause frame use: Symmetric
        Supports auto-negotiation: Yes
        Advertised link modes:  10baseT/Half 10baseT/Full 
                                100baseT/Half 100baseT/Full 
                                1000baseT/Full 
        Advertised pause frame use: Symmetric
        Advertised auto-negotiation: Yes
        Link partner advertised link modes:  100baseT/Full 
                                             1000baseT/Full 
        Link partner advertised pause frame use: Symmetric Receive-only
        Link partner advertised auto-negotiation: Yes
        Speed: 1000Mb/s
        Duplex: Full
        Port: MII
        PHYAD: 1
        Transceiver: internal
        Auto-negotiation: on
        Supports Wake-on: d
        Wake-on: d
        Link detected: yes

root@imx6qdlsabresd:~# ethtool -a eth0
Pause parameters for eth0:
Autonegotiate:  on
RX:             off
TX:             off
RX negotiated:  on
TX negotiated:  on

---------------------------------------------------------------------------

---------------------------------------------------------------------------
Masahiko KIMOTO,Ph.D. / Cybertrust Japan Co., Ltd. 
mail: masahiko.kimoto@cybertrust.co.jp
web: https://www.cybertrust.co.jp/ https://www.miraclelinux.com/