[Intel-wired-lan] [PATCH v3 09/11] igc: Add code for PHY support

Neftin, Sasha sasha.neftin at intel.com
Sun Aug 5 08:59:03 UTC 2018


On 7/5/2018 14:24, Neftin, Sasha wrote:
> On 6/29/2018 17:37, Shannon Nelson wrote:
>> On 6/24/2018 1:45 AM, Sasha Neftin wrote:
>>> Sasha Neftin (v2):
>>> minor cosmetic changes
>>>
>>> Alexander Duyck (v3.5):
>>> Made non-gpy versions of phy functions static
>>> Moved xmdio functions to avoid need for prototype function
>>> Moved reset, read_reg, and write_reg ops into e1000_phy_ops_base
>>> Dropped code implying more than one function
>>> Fixed merge conflict with v3.5 of NVM code.
>>>
>>> Sasha Neftin (v3):
>>> Introduced temporary workaround for PHY power down.
>>>
>>> Signed-off-by: Sasha Neftin <sasha.neftin at intel.com>
>>> Signed-off-by: Alexander Duyck <alexander.h.duyck at intel.com>
>>> ---
>>>   drivers/net/ethernet/intel/igc/Makefile        |   3 +-
>>>   drivers/net/ethernet/intel/igc/e1000_base.c    | 126 ++++++-
>>>   drivers/net/ethernet/intel/igc/e1000_base.h    |   1 +
>>>   drivers/net/ethernet/intel/igc/e1000_defines.h |  81 ++++-
>>>   drivers/net/ethernet/intel/igc/e1000_hw.h      |  54 +++
>>>   drivers/net/ethernet/intel/igc/e1000_mac.c     |  48 ++-
>>>   drivers/net/ethernet/intel/igc/e1000_mac.h     |  11 +
>>>   drivers/net/ethernet/intel/igc/e1000_phy.c     | 457 
>>> +++++++++++++++++++++++++
>>>   drivers/net/ethernet/intel/igc/e1000_phy.h     |  21 ++
>>>   drivers/net/ethernet/intel/igc/e1000_regs.h    |   3 +
>>>   drivers/net/ethernet/intel/igc/igc.h           |  16 +
>>>   drivers/net/ethernet/intel/igc/igc_main.c      |  11 +
>>>   12 files changed, 827 insertions(+), 5 deletions(-)
>>>   create mode 100644 drivers/net/ethernet/intel/igc/e1000_phy.c
>>>   create mode 100644 drivers/net/ethernet/intel/igc/e1000_phy.h
>>>
>>> diff --git a/drivers/net/ethernet/intel/igc/Makefile 
>>> b/drivers/net/ethernet/intel/igc/Makefile
>>> index 5147bca1a900..cd5cb1630b1d 100644
>>> --- a/drivers/net/ethernet/intel/igc/Makefile
>>> +++ b/drivers/net/ethernet/intel/igc/Makefile
>>> @@ -7,4 +7,5 @@
>>>   obj-$(CONFIG_IGC) += igc.o
>>> -igc-objs := igc_main.o e1000_mac.o e1000_i225.o e1000_base.o 
>>> e1000_nvm.o
>>> +igc-objs := igc_main.o e1000_mac.o e1000_i225.o e1000_base.o 
>>> e1000_nvm.o \
>>> +e1000_phy.o
>>> diff --git a/drivers/net/ethernet/intel/igc/e1000_base.c 
>>> b/drivers/net/ethernet/intel/igc/e1000_base.c
>>> index 6db31349daa4..8b35e1d6c32e 100644
>>> --- a/drivers/net/ethernet/intel/igc/e1000_base.c
>>> +++ b/drivers/net/ethernet/intel/igc/e1000_base.c
>>> @@ -12,12 +12,31 @@
>>>   /* forward declaration */
>>>   static s32 igc_get_invariants_base(struct e1000_hw *);
>>>   static s32 igc_check_for_link_base(struct e1000_hw *);
>>> +static s32 igc_acquire_phy_base(struct e1000_hw *);
>>> +static void igc_release_phy_base(struct e1000_hw *);
>>> +static s32 igc_get_phy_id_base(struct e1000_hw *);
>>>   static s32 igc_init_hw_base(struct e1000_hw *);
>>>   static s32 igc_reset_hw_base(struct e1000_hw *);
>>>   static s32 igc_set_pcie_completion_timeout(struct e1000_hw *hw);
>>>   static s32 igc_read_mac_addr_base(struct e1000_hw *hw);
>>>   /**
>>> + *  igc_get_phy_id_base - Retrieve PHY addr and id
>>> + *  @hw: pointer to the HW structure
>>> + *
>>> + *  Retrieves the PHY address and ID for both PHY's which do and do 
>>> not use
>>> + *  sgmi interface.
>>> + **/
>>> +static s32 igc_get_phy_id_base(struct e1000_hw *hw)
>>> +{
>>> +    s32  ret_val = 0;
>>> +
>>> +    ret_val = igc_get_phy_id(hw);
>>> +
>>> +    return ret_val;
>>> +}
>>> +
>>> +/**
>>>    *  igc_init_nvm_params_base - Init NVM func ptrs.
>>>    *  @hw: pointer to the HW structure
>>>    **/
>>> @@ -81,6 +100,59 @@ static s32 igc_init_mac_params_base(struct 
>>> e1000_hw *hw)
>>>       return 0;
>>>   }
>>> +/**
>>> + *  igc_init_phy_params_base - Init PHY func ptrs.
>>> + *  @hw: pointer to the HW structure
>>> + **/
>>> +static s32 igc_init_phy_params_base(struct e1000_hw *hw)
>>> +{
>>> +    struct e1000_phy_info *phy = &hw->phy;
>>> +    s32 ret_val = 0;
>>> +    u32 ctrl_ext;
>>> +
>>> +    if (hw->phy.media_type != e1000_media_type_copper) {
>>> +        phy->type = e1000_phy_none;
>>> +        goto out;
>>> +    }
>>> +
>>> +    phy->autoneg_mask       = AUTONEG_ADVERTISE_SPEED_DEFAULT_2500;
>>> +    phy->reset_delay_us     = 100;
>>> +
>>> +    ctrl_ext = rd32(E1000_CTRL_EXT);
>>> +
>>> +    /* set lan id */
>>> +    hw->bus.func = (rd32(E1000_STATUS) & E1000_STATUS_FUNC_MASK) >>
>>> +            E1000_STATUS_FUNC_SHIFT;
>>> +
>>> +    /* Make sure the PHY is in a good state. Several people have 
>>> reported
>>> +     * firmware leaving the PHY's page select register set to something
>>> +     * other than the default of zero, which causes the PHY ID read to
>>> +     * access something other than the intended register.
>>> +     */
>>> +    ret_val = hw->phy.ops.reset(hw);
>>> +    if (ret_val) {
>>> +        hw_dbg("Error resetting the PHY.\n");
>>> +        goto out;
>>> +    }
>>> +
>>> +    ret_val = igc_get_phy_id_base(hw);
>>> +    if (ret_val)
>>> +        return ret_val;
>>> +
>>> +    /* Verify phy id and set remaining function pointers */
>>> +    switch (phy->id) {
>>> +    case I225_I_PHY_ID:
>>> +        phy->type    = e1000_phy_i225;
>>> +        break;
>>> +    default:
>>> +        ret_val = -E1000_ERR_PHY;
>>> +        goto out;
>>> +    }
>>> +
>>> +out:
>>> +    return ret_val;
>>> +}
>>> +
>>>   static s32 igc_get_invariants_base(struct e1000_hw *hw)
>>>   {
>>>       s32 ret_val = 0;
>>> @@ -105,6 +177,8 @@ static s32 igc_get_invariants_base(struct 
>>> e1000_hw *hw)
>>>           break;
>>>       }
>>> +    /* setup PHY parameters */
>>> +    ret_val = igc_init_phy_params_base(hw);
>>>       if (ret_val)
>>>           goto out;
>>> @@ -113,6 +187,34 @@ static s32 igc_get_invariants_base(struct 
>>> e1000_hw *hw)
>>>   }
>>>   /**
>>> + *  igc_acquire_phy_base - Acquire rights to access PHY
>>> + *  @hw: pointer to the HW structure
>>> + *
>>> + *  Acquire access rights to the correct PHY.  This is a
>>> + *  function pointer entry point called by the api module.
>>> + **/
>>> +static s32 igc_acquire_phy_base(struct e1000_hw *hw)
>>> +{
>>> +    u16 mask = E1000_SWFW_PHY0_SM;
>>> +
>>> +    return hw->mac.ops.acquire_swfw_sync(hw, mask);
>>> +}
>>> +
>>> +/**
>>> + *  igc_release_phy_base - Release rights to access PHY
>>> + *  @hw: pointer to the HW structure
>>> + *
>>> + *  A wrapper to release access rights to the correct PHY.  This is a
>>> + *  function pointer entry point called by the api module.
>>> + **/
>>> +static void igc_release_phy_base(struct e1000_hw *hw)
>>> +{
>>> +    u16 mask = E1000_SWFW_PHY0_SM;
>>> +
>>> +    hw->mac.ops.release_swfw_sync(hw, mask);
>>> +}
>>> +
>>> +/**
>>>    *  igc_get_link_up_info_base - Get link speed/duplex info
>>>    *  @hw: pointer to the HW structure
>>>    *  @speed: stores the current speed
>>> @@ -200,6 +302,20 @@ static s32 igc_read_mac_addr_base(struct 
>>> e1000_hw *hw)
>>>   }
>>>   /**
>>> + *  igc_power_down_phy_copper_base - Remove link during PHY power down
>>> + *  @hw: pointer to the HW structure
>>> + *
>>> + *  In the case of a PHY power down to save power, or to turn off 
>>> link during a
>>> + *  driver unload, or wake on lan is not enabled, remove the link.
>>> + **/
>>> +void igc_power_down_phy_copper_base(struct e1000_hw *hw)
>>> +{
>>> +    /* If the management interface is not enabled, then power down */
>>> +    if (!(igc_enable_mng_pass_thru(hw) || igc_check_reset_block(hw)))
>>> +        igc_power_down_phy_copper(hw);
>>> +}
>>> +
>>> +/**
>>>    *  igc_rx_fifo_flush_base - Clean rx fifo after Rx enable
>>>    *  @hw: pointer to the HW structure
>>>    *
>>> @@ -283,10 +399,18 @@ static struct e1000_mac_operations 
>>> e1000_mac_ops_base = {
>>>       .get_speed_and_duplex    = igc_get_link_up_info_base,
>>>   };
>>> +static const struct e1000_phy_operations e1000_phy_ops_base = {
>>> +    .acquire        = igc_acquire_phy_base,
>>> +    .release        = igc_release_phy_base,
>>> +    .reset            = igc_phy_hw_reset,
>>> +    .read_reg        = igc_read_phy_reg_gpy,
>>> +    .write_reg        = igc_write_phy_reg_gpy,
>>> +};
>>> +
>>>   const struct e1000_info e1000_base_info = {
>>>       .get_invariants    = igc_get_invariants_base,
>>>       .mac_ops    = &e1000_mac_ops_base,
>>> -    /* TODO phy_ops */
>>> +    .phy_ops    = &e1000_phy_ops_base,
>>>   };
>>>   /**
>>> diff --git a/drivers/net/ethernet/intel/igc/e1000_base.h 
>>> b/drivers/net/ethernet/intel/igc/e1000_base.h
>>> index 210f1b5f6012..37e61e033966 100644
>>> --- a/drivers/net/ethernet/intel/igc/e1000_base.h
>>> +++ b/drivers/net/ethernet/intel/igc/e1000_base.h
>>> @@ -6,6 +6,7 @@
>>>   /* forward declaration */
>>>   void igc_rx_fifo_flush_base(struct e1000_hw *hw);
>>> +void igc_power_down_phy_copper_base(struct e1000_hw *hw);
>>>   /* Transmit Descriptor - Advanced */
>>>   union e1000_adv_tx_desc {
>>> diff --git a/drivers/net/ethernet/intel/igc/e1000_defines.h 
>>> b/drivers/net/ethernet/intel/igc/e1000_defines.h
>>> index 31bc85cfa149..40276ae76c9f 100644
>>> --- a/drivers/net/ethernet/intel/igc/e1000_defines.h
>>> +++ b/drivers/net/ethernet/intel/igc/e1000_defines.h
>>> @@ -67,11 +67,14 @@
>>>   #define E1000_ERR_MAC_INIT            5
>>>   #define E1000_ERR_RESET                9
>>>   #define E1000_ERR_MASTER_REQUESTS_PENDING    10
>>> +#define E1000_BLK_PHY_RESET            12
>>
>> This should have the _ERR_ in it like the rest
>>
> Good. fix will be applied in v4.
>>>   #define E1000_ERR_SWFW_SYNC            13
>>>   /* Device Control */
>>>   #define E1000_CTRL_RST        0x04000000  /* Global reset */
>>> +#define E1000_CTRL_PHY_RST    0x80000000  /* PHY Reset */
>>> +
>>>   /* PBA constants */
>>>   #define E1000_PBA_34K            0x0022
>>> @@ -175,6 +178,22 @@
>>>   #define HALF_DUPLEX        1
>>>   #define FULL_DUPLEX        2
>>> +/* 1Gbps and 2.5Gbps half duplex is not supported, nor 
>>> spec-compliant. */
>>> +#define ADVERTISE_10_HALF        0x0001
>>> +#define ADVERTISE_10_FULL        0x0002
>>> +#define ADVERTISE_100_HALF        0x0004
>>> +#define ADVERTISE_100_FULL        0x0008
>>> +#define ADVERTISE_1000_HALF        0x0010 /* Not used, just FYI */
>>> +#define ADVERTISE_1000_FULL        0x0020
>>> +#define ADVERTISE_2500_HALF        0x0040 /* NOT used, just FYI */
>>> +#define ADVERTISE_2500_FULL        0x0080
>>> +
>>> +#define E1000_ALL_SPEED_DUPLEX_2500 ( \
>>> +    ADVERTISE_10_HALF | ADVERTISE_10_FULL | ADVERTISE_100_HALF | \
>>> +    ADVERTISE_100_FULL | ADVERTISE_1000_FULL | ADVERTISE_2500_FULL)
>>> +
>>> +#define AUTONEG_ADVERTISE_SPEED_DEFAULT_2500 
>>> E1000_ALL_SPEED_DUPLEX_2500
>>> +
>>>   /* Interrupt Cause Read */
>>>   #define E1000_ICR_TXDW        0x00000001 /* Transmit desc written 
>>> back */
>>>   #define E1000_ICR_TXQE        0x00000002 /* Transmit Queue empty */
>>> @@ -256,7 +275,8 @@
>>>   #define E1000_FCRTL_XONE        0x80000000
>>>   /* Management Control */
>>> -#define E1000_MANC_RCV_TCO_EN    0x00020000 /* Receive TCO Packets 
>>> Enabled */
>>> +#define E1000_MANC_RCV_TCO_EN        0x00020000 /* Receive TCO 
>>> Enabled */
>>> +#define E1000_MANC_BLK_PHY_RST_ON_IDE    0x00040000 /* Block phy 
>>> resets */
>>>   /* Receive Control */
>>>   #define E1000_RCTL_RST        0x00000001 /* Software reset */
>>> @@ -322,4 +342,63 @@
>>>   #define E1000_RCTL_PMCF        0x00800000 /* pass MAC control 
>>> frames */
>>>   #define E1000_RCTL_SECRC    0x04000000 /* Strip Ethernet CRC */
>>> +/* GPY211 - I225 defines */
>>> +#define GPY_MMD_MASK            0xFFFF0000
>>> +#define GPY_MMD_SHIFT            16
>>> +#define GPY_REG_MASK            0x0000FFFF
>>> +
>>> +#define E1000_MMDAC_FUNC_DATA    0x4000 /* Data, no post increment */
>>> +
>>> +/* MAC definitions */
>>> +#define E1000_FACTPS_MNGCG        0x20000000
>>> +#define E1000_FWSM_MODE_MASK        0xE
>>> +#define E1000_FWSM_MODE_SHIFT        1
>>> +
>>> +/* Management Control */
>>> +#define E1000_MANC_SMBUS_EN    0x00000001 /* SMBus Enabled - RO */
>>> +#define E1000_MANC_ASF_EN    0x00000002 /* ASF Enabled - RO */
>>> +
>>> +/* PHY */
>>> +#define PHY_REVISION_MASK    0xFFFFFFF0
>>> +#define MAX_PHY_REG_ADDRESS    0x1F  /* 5 bit address bus (0-0x1F) */
>>> +#define E1000_GEN_POLL_TIMEOUT    640
>>> +
>>> +/* PHY Control Register */
>>> +#define MII_CR_FULL_DUPLEX    0x0100  /* FDX =1, half duplex =0 */
>>> +#define MII_CR_RESTART_AUTO_NEG    0x0200  /* Restart auto 
>>> negotiation */
>>> +#define MII_CR_POWER_DOWN    0x0800  /* Power down */
>>> +#define MII_CR_AUTO_NEG_EN    0x1000  /* Auto Neg Enable */
>>> +#define MII_CR_LOOPBACK        0x4000  /* 0 = normal, 1 = loopback */
>>> +#define MII_CR_RESET        0x8000  /* 0 = normal, 1 = PHY reset */
>>> +#define MII_CR_SPEED_1000    0x0040
>>> +#define MII_CR_SPEED_100    0x2000
>>> +#define MII_CR_SPEED_10        0x0000
>>> +
>>> +/* PHY Status Register */
>>> +#define MII_SR_LINK_STATUS    0x0004 /* Link Status 1 = link */
>>> +#define MII_SR_AUTONEG_COMPLETE    0x0020 /* Auto Neg Complete */
>>> +
>>> +/* PHY 1000 MII Register/Bit Definitions */
>>> +/* PHY Registers defined by IEEE */
>>> +#define PHY_CONTROL    0x00 /* Control Register */
>>> +#define PHY_STATUS    0x01 /* Status Register */
>>> +#define PHY_ID1        0x02 /* Phy Id Reg (word 1) */
>>> +#define PHY_ID2        0x03 /* Phy Id Reg (word 2) */
>>> +
>>> +/* Bit definitions for valid PHY IDs. I = Integrated E = External */
>>> +#define I225_I_PHY_ID        0x67C9DC00
>>
>> To be consistent, these PHY_ and MII_ and the rest should have the 
>> same prefix as the rest of the driver specific #defines, which I still 
>> think should be IGC_, not E1000_.
>>
> As I answer previously. In regard e1000_ or igc_ prefix, we will discuss 
> with my colleagues and I will reply.
e1000_ prefix has been changed to igc_ prefix for all methods and files.
>>> +
>>> +/* MDI Control */
>>> +#define E1000_MDIC_DATA_MASK    0x0000FFFF
>>> +#define E1000_MDIC_REG_MASK    0x001F0000
>>> +#define E1000_MDIC_REG_SHIFT    16
>>> +#define E1000_MDIC_PHY_MASK    0x03E00000
>>> +#define E1000_MDIC_PHY_SHIFT    21
>>> +#define E1000_MDIC_OP_WRITE    0x04000000
>>> +#define E1000_MDIC_OP_READ    0x08000000
>>> +#define E1000_MDIC_READY    0x10000000
>>> +#define E1000_MDIC_INT_EN    0x20000000
>>> +#define E1000_MDIC_ERROR    0x40000000
>>> +#define E1000_MDIC_DEST        0x80000000
>>> +
>>>   #endif /* _E1000_DEFINES_H_ */
>>> diff --git a/drivers/net/ethernet/intel/igc/e1000_hw.h 
>>> b/drivers/net/ethernet/intel/igc/e1000_hw.h
>>> index ab630e5b3d97..d8fc6fa9eda2 100644
>>> --- a/drivers/net/ethernet/intel/igc/e1000_hw.h
>>> +++ b/drivers/net/ethernet/intel/igc/e1000_hw.h
>>> @@ -11,6 +11,7 @@
>>>   #include "e1000_regs.h"
>>>   #include "e1000_defines.h"
>>>   #include "e1000_mac.h"
>>> +#include "e1000_phy.h"
>>>   #include "e1000_nvm.h"
>>>   #include "e1000_i225.h"
>>>   #include "e1000_base.h"
>>> @@ -18,6 +19,8 @@
>>>   #define E1000_DEV_ID_I225_LM            0x15F2
>>>   #define E1000_DEV_ID_I225_V            0x15F3
>>> +#define E1000_FUNC_0                0
>>> +
>>>   /* Forward declaration */
>>>   struct e1000_hw;
>>> @@ -47,6 +50,12 @@ enum e1000_phy_type {
>>>       e1000_phy_i225,
>>>   };
>>> +enum e1000_media_type {
>>> +    e1000_media_type_unknown = 0,
>>> +    e1000_media_type_copper = 1,
>>> +    e1000_num_media_types
>>> +};
>>> +
>>>   enum e1000_nvm_type {
>>>       e1000_nvm_unknown = 0,
>>>       e1000_nvm_flash_hw,
>>> @@ -112,6 +121,7 @@ struct e1000_mac_info {
>>>       bool adaptive_ifs;
>>>       bool has_fwsm;
>>> +    bool asf_firmware_present;
>>>       bool arc_subsystem_valid;
>>>       bool autoneg;
>>> @@ -129,6 +139,20 @@ struct e1000_nvm_operations {
>>>       s32 (*valid_led_default)(struct e1000_hw *hw, u16 *data);
>>>   };
>>> +struct e1000_phy_operations {
>>> +    s32 (*acquire)(struct e1000_hw *hw);
>>> +    s32 (*check_polarity)(struct e1000_hw *hw);
>>> +    s32 (*check_reset_block)(struct e1000_hw *hw);
>>> +    s32 (*force_speed_duplex)(struct e1000_hw *hw);
>>> +    s32 (*get_cfg_done)(struct e1000_hw *hw);
>>> +    s32 (*get_cable_length)(struct e1000_hw *hw);
>>> +    s32 (*get_phy_info)(struct e1000_hw *hw);
>>> +    s32 (*read_reg)(struct e1000_hw *hw, u32 address, u16 *data);
>>> +    void (*release)(struct e1000_hw *hw);
>>> +    s32 (*reset)(struct e1000_hw *hw);
>>> +    s32 (*write_reg)(struct e1000_hw *hw, u32 address, u16 data);
>>> +};
>>> +
>>>   struct e1000_nvm_info {
>>>       struct e1000_nvm_operations ops;
>>>       enum e1000_nvm_type type;
>>> @@ -143,6 +167,35 @@ struct e1000_nvm_info {
>>>       u16 page_size;
>>>   };
>>> +struct e1000_phy_info {
>>> +    struct e1000_phy_operations ops;
>>> +
>>> +    enum e1000_phy_type type;
>>> +
>>> +    u32 addr;
>>> +    u32 id;
>>> +    u32 reset_delay_us; /* in usec */
>>> +    u32 revision;
>>> +
>>> +    enum e1000_media_type media_type;
>>> +
>>> +    u16 autoneg_advertised;
>>> +    u16 autoneg_mask;
>>> +    u16 cable_length;
>>> +    u16 max_cable_length;
>>> +    u16 min_cable_length;
>>> +    u16 pair_length[4];
>>> +
>>> +    u8 mdix;
>>> +
>>> +    bool disable_polarity_correction;
>>> +    bool is_mdix;
>>> +    bool polarity_correction;
>>> +    bool reset_disable;
>>> +    bool speed_downgraded;
>>> +    bool autoneg_wait_to_complete;
>>> +};
>>> +
>>>   struct e1000_bus_info {
>>>       enum e1000_bus_type type;
>>>       enum e1000_bus_speed speed;
>>> @@ -188,6 +241,7 @@ struct e1000_hw {
>>>       struct e1000_mac_info  mac;
>>>       struct e1000_fc_info   fc;
>>>       struct e1000_nvm_info  nvm;
>>> +    struct e1000_phy_info  phy;
>>>       struct e1000_bus_info bus;
>>> diff --git a/drivers/net/ethernet/intel/igc/e1000_mac.c 
>>> b/drivers/net/ethernet/intel/igc/e1000_mac.c
>>> index 5681c0e3f0c8..fe483211f66f 100644
>>> --- a/drivers/net/ethernet/intel/igc/e1000_mac.c
>>> +++ b/drivers/net/ethernet/intel/igc/e1000_mac.c
>>> @@ -389,7 +389,8 @@ s32 igc_check_for_copper_link(struct e1000_hw *hw)
>>>        * link.  If so, then we want to get the current speed/duplex
>>>        * of the PHY.
>>>        */
>>> -    /* TODO ret_val = igc_phy_has_link(hw, 1, 0, &link); */
>>> +    ret_val = igc_phy_has_link(hw, 1, 0, &link);
>>> +
>>
>> no blank line needed here
>>
> Thanks. fix will be applied in v4.
>>>       if (ret_val)
>>>           goto out;
>>> @@ -401,7 +402,7 @@ s32 igc_check_for_copper_link(struct e1000_hw *hw)
>>>       /* Check if there was DownShift, must be checked
>>>        * immediately after link-up
>>>        */
>>> -    /* TODO igc_check_downshift(hw); */
>>> +    igc_check_downshift(hw);
>>>       /* If we are forcing speed/duplex, then we simply return since
>>>        * we have already determined whether we have link or not.
>>> @@ -542,3 +543,46 @@ void igc_put_hw_semaphore(struct e1000_hw *hw)
>>>       wr32(E1000_SWSM, swsm);
>>>   }
>>> +
>>> +/**
>>> + *  igc_enable_mng_pass_thru - Enable processing of ARP's
>>> + *  @hw: pointer to the HW structure
>>> + *
>>> + *  Verifies the hardware needs to leave interface enabled so that 
>>> frames can
>>> + *  be directed to and from the management interface.
>>> + **/
>>> +bool igc_enable_mng_pass_thru(struct e1000_hw *hw)
>>> +{
>>> +    u32 manc;
>>> +    u32 fwsm, factps;
>>> +    bool ret_val = false;
>>> +
>>> +    if (!hw->mac.asf_firmware_present)
>>> +        goto out;
>>> +
>>> +    manc = rd32(E1000_MANC);
>>> +
>>> +    if (!(manc & E1000_MANC_RCV_TCO_EN))
>>> +        goto out;
>>> +
>>> +    if (hw->mac.arc_subsystem_valid) {
>>> +        fwsm = rd32(E1000_FWSM);
>>> +        factps = rd32(E1000_FACTPS);
>>> +
>>> +        if (!(factps & E1000_FACTPS_MNGCG) &&
>>> +            ((fwsm & E1000_FWSM_MODE_MASK) ==
>>> +            (e1000_mng_mode_pt << E1000_FWSM_MODE_SHIFT))) {
>>> +            ret_val = true;
>>> +            goto out;
>>> +        }
>>> +    } else {
>>> +        if ((manc & E1000_MANC_SMBUS_EN) &&
>>> +            !(manc & E1000_MANC_ASF_EN)) {
>>> +            ret_val = true;
>>> +            goto out;
>>> +        }
>>> +    }
>>> +
>>> +out:
>>> +    return ret_val;
>>> +}
>>> diff --git a/drivers/net/ethernet/intel/igc/e1000_mac.h 
>>> b/drivers/net/ethernet/intel/igc/e1000_mac.h
>>> index f18f5221199f..0f17a8443125 100644
>>> --- a/drivers/net/ethernet/intel/igc/e1000_mac.h
>>> +++ b/drivers/net/ethernet/intel/igc/e1000_mac.h
>>> @@ -5,6 +5,7 @@
>>>   #define _E1000_MAC_H_
>>>   #include "e1000_hw.h"
>>> +#include "e1000_phy.h"
>>>   #include "e1000_defines.h"
>>>   #ifndef E1000_REMOVED
>>> @@ -28,4 +29,14 @@ s32 igc_get_bus_info_pcie(struct e1000_hw *hw);
>>>   s32 igc_get_speed_and_duplex_copper(struct e1000_hw *hw, u16 *speed,
>>>                       u16 *duplex);
>>> +bool igc_enable_mng_pass_thru(struct e1000_hw *hw);
>>> +
>>> +enum e1000_mng_mode {
>>> +    e1000_mng_mode_none = 0,
>>> +    e1000_mng_mode_asf,
>>> +    e1000_mng_mode_pt,
>>> +    e1000_mng_mode_ipmi,
>>> +    e1000_mng_mode_host_if_only
>>> +};
>>> +
>>>   #endif
>>> diff --git a/drivers/net/ethernet/intel/igc/e1000_phy.c 
>>> b/drivers/net/ethernet/intel/igc/e1000_phy.c
>>> new file mode 100644
>>> index 000000000000..c95efc4145e8
>>> --- /dev/null
>>> +++ b/drivers/net/ethernet/intel/igc/e1000_phy.c
>>> @@ -0,0 +1,457 @@
>>> +// SPDX-License-Identifier: GPL-2.0
>>> +/* Copyright (c)  2018 Intel Corporation */
>>> +
>>> +#include "e1000_phy.h"
>>> +
>>> +/**
>>> + *  igc_check_reset_block - Check if PHY reset is blocked
>>> + *  @hw: pointer to the HW structure
>>> + *
>>> + *  Read the PHY management control register and check whether a PHY 
>>> reset
>>> + *  is blocked.  If a reset is not blocked return 0, otherwise
>>> + *  return E1000_BLK_PHY_RESET (12).
>>> + **/
>>> +s32 igc_check_reset_block(struct e1000_hw *hw)
>>> +{
>>> +    u32 manc;
>>> +
>>> +    manc = rd32(E1000_MANC);
>>> +
>>> +    return (manc & E1000_MANC_BLK_PHY_RST_ON_IDE) ? 
>>> E1000_BLK_PHY_RESET : 0;
>>> +}
>>> +
>>> +/**
>>> + *  igc_get_phy_id - Retrieve the PHY ID and revision
>>> + *  @hw: pointer to the HW structure
>>> + *
>>> + *  Reads the PHY registers and stores the PHY ID and possibly the PHY
>>> + *  revision in the hardware structure.
>>> + **/
>>> +s32 igc_get_phy_id(struct e1000_hw *hw)
>>> +{
>>> +    struct e1000_phy_info *phy = &hw->phy;
>>> +    s32 ret_val = 0;
>>> +    u16 phy_id;
>>> +
>>> +    ret_val = phy->ops.read_reg(hw, PHY_ID1, &phy_id);
>>> +    if (ret_val)
>>> +        goto out;
>>> +
>>> +    phy->id = (u32)(phy_id << 16);
>>> +    usleep_range(200, 500);
>>> +    ret_val = phy->ops.read_reg(hw, PHY_ID2, &phy_id);
>>> +    if (ret_val)
>>> +        goto out;
>>> +
>>> +    phy->id |= (u32)(phy_id & PHY_REVISION_MASK);
>>> +    phy->revision = (u32)(phy_id & ~PHY_REVISION_MASK);
>>> +
>>> +out:
>>> +    return ret_val;
>>> +}
>>> +
>>> +/**
>>> + *  igc_phy_has_link - Polls PHY for link
>>> + *  @hw: pointer to the HW structure
>>> + *  @iterations: number of times to poll for link
>>> + *  @usec_interval: delay between polling attempts
>>> + *  @success: pointer to whether polling was successful or not
>>> + *
>>> + *  Polls the PHY status register for link, 'iterations' number of 
>>> times.
>>> + **/
>>> +s32 igc_phy_has_link(struct e1000_hw *hw, u32 iterations,
>>> +             u32 usec_interval, bool *success)
>>> +{
>>> +    s32 ret_val = 0;
>>> +    u16 i, phy_status;
>>> +
>>> +    for (i = 0; i < iterations; i++) {
>>> +        /* Some PHYs require the PHY_STATUS register to be read
>>> +         * twice due to the link bit being sticky.  No harm doing
>>> +         * it across the board.
>>> +         */
>>> +        ret_val = hw->phy.ops.read_reg(hw, PHY_STATUS, &phy_status);
>>> +        if (ret_val && usec_interval > 0) {
>>> +            /* If the first read fails, another entity may have
>>> +             * ownership of the resources, wait and try again to
>>> +             * see if they have relinquished the resources yet.
>>> +             */
>>> +            if (usec_interval >= 1000)
>>> +                mdelay(usec_interval / 1000);
>>> +            else
>>> +                udelay(usec_interval);
>>> +        }
>>> +        ret_val = hw->phy.ops.read_reg(hw, PHY_STATUS, &phy_status);
>>> +        if (ret_val)
>>> +            break;
>>> +        if (phy_status & MII_SR_LINK_STATUS)
>>> +            break;
>>> +        if (usec_interval >= 1000)
>>> +            mdelay(usec_interval / 1000);
>>> +        else
>>> +            udelay(usec_interval);
>>> +    }
>>> +
>>> +    *success = (i < iterations) ? true : false;
>>> +
>>> +    return ret_val;
>>> +}
>>> +
>>> +/**
>>> + *  igc_power_up_phy_copper - Restore copper link in case of PHY 
>>> power down
>>> + *  @hw: pointer to the HW structure
>>> + *
>>> + *  In the case of a PHY power down to save power, or to turn off 
>>> link during a
>>> + *  driver unload, restore the link to previous settings.
>>> + **/
>>> +void igc_power_up_phy_copper(struct e1000_hw *hw)
>>> +{
>>> +    u16 mii_reg = 0;
>>> +
>>> +    /* The PHY will retain its settings across a power down/up cycle */
>>> +    hw->phy.ops.read_reg(hw, PHY_CONTROL, &mii_reg);
>>> +    mii_reg &= ~MII_CR_POWER_DOWN;
>>> +    hw->phy.ops.write_reg(hw, PHY_CONTROL, mii_reg);
>>> +}
>>> +
>>> +/**
>>> + *  igc_power_down_phy_copper - Power down copper PHY
>>> + *  @hw: pointer to the HW structure
>>> + *
>>> + *  Power down PHY to save power when interface is down and wake on lan
>>> + *  is not enabled.
>>> + **/
>>> +void igc_power_down_phy_copper(struct e1000_hw *hw)
>>> +{
>>> +    u16 mii_reg = 0;
>>> +
>>> +    /* The PHY will retain its settings across a power down/up cycle */
>>> +    hw->phy.ops.read_reg(hw, PHY_CONTROL, &mii_reg);
>>> +    mii_reg |= MII_CR_POWER_DOWN;
>>> +
>>> +    /* Temporary workaround - should be removed when PHY will implement
>>> +     * IEEE registers as properly
>>> +     */
>>
>> Maybe this should be marked with a TODO: as that is a more standard 
>> way of tagging future work.
>>
> Ah... we just decided kee without /* TODO... */ so I marked this as 
> comment. In next step of silicon this issue should be resolved and I 
> will send another patch.
>>> +    /* hw->phy.ops.write_reg(hw, PHY_CONTROL, mii_reg);*/
>>> +    usleep_range(1000, 2000);
>>> +}
>>> +
>>> +/**
>>> + *  igc_check_downshift - Checks whether a downshift in speed occurred
>>> + *  @hw: pointer to the HW structure
>>> + *
>>> + *  Success returns 0, Failure returns 1
>>> + *
>>> + *  A downshift is detected by querying the PHY link health.
>>> + **/
>>> +s32 igc_check_downshift(struct e1000_hw *hw)
>>> +{
>>> +    struct e1000_phy_info *phy = &hw->phy;
>>> +    s32 ret_val;
>>> +    u16 phy_data, offset, mask;
>>> +
>>> +    switch (phy->type) {
>>> +    case e1000_phy_i225:
>>> +    default:
>>> +        /* speed downshift not supported */
>>> +        phy->speed_downgraded = false;
>>> +        ret_val = 0;
>>> +        goto out;
>>> +    }
>>
>> This filter looks wrong, I don't think both i225 and default should 
>> end up with "not supported".  Either you're filtering specific devices 
>> for not supported and the default should go on to check for downgrade, 
>> or only specific devices should go on to check for downgrade and the 
>> rest should default to not supported.
>>
> Ok. I need to check more depth with design and run HW experiment.
>>> +
>>> +    ret_val = phy->ops.read_reg(hw, offset, &phy_data);
>>> +
>>> +    if (!ret_val)
>>> +        phy->speed_downgraded = (phy_data & mask) ? true : false;
>>> +
>>> +out:
>>> +    return ret_val;
>>> +}
>>> +
>>> +/**
>>> + *  igc_phy_hw_reset - PHY hardware reset
>>> + *  @hw: pointer to the HW structure
>>> + *
>>> + *  Verify the reset block is not blocking us from resetting.  Acquire
>>> + *  semaphore (if necessary) and read/set/write the device control 
>>> reset
>>> + *  bit in the PHY.  Wait the appropriate delay time for the device to
>>> + *  reset and release the semaphore (if necessary).
>>> + **/
>>> +s32 igc_phy_hw_reset(struct e1000_hw *hw)
>>> +{
>>> +    struct e1000_phy_info *phy = &hw->phy;
>>> +    s32  ret_val;
>>> +    u32 ctrl;
>>> +
>>> +    ret_val = igc_check_reset_block(hw);
>>> +    if (ret_val) {
>>> +        ret_val = 0;
>>> +        goto out;
>>> +    }
>>> +
>>> +    ret_val = phy->ops.acquire(hw);
>>> +    if (ret_val)
>>> +        goto out;
>>> +
>>> +    ctrl = rd32(E1000_CTRL);
>>> +    wr32(E1000_CTRL, ctrl | E1000_CTRL_PHY_RST);
>>> +    wrfl();
>>> +
>>> +    udelay(phy->reset_delay_us); > +
>>> +    wr32(E1000_CTRL, ctrl);
>>> +    wrfl();
>>
>> You have to specifically clear the PHY_RST bit?  Doesn't the reset 
>> clear the bit?
>>
> I keep align with other drivers. Our PHY's FW not completely done yet - 
> so, please wait when I will check this.
>>> +
>>> +    usleep_range(1500, 2000);
>>> +
>>> +    phy->ops.release(hw);
>>> +
>>> +    ret_val = phy->ops.get_cfg_done(hw);
>>> +
>>> +out:
>>> +    return ret_val;
>>> +}
>>> +
>>> +/**
>>> + *  igc_read_phy_reg_mdic - Read MDI control register
>>> + *  @hw: pointer to the HW structure
>>> + *  @offset: register offset to be read
>>> + *  @data: pointer to the read data
>>> + *
>>> + *  Reads the MDI control register in the PHY at offset and stores the
>>> + *  information read to data.
>>> + **/
>>> +static s32 igc_read_phy_reg_mdic(struct e1000_hw *hw, u32 offset, 
>>> u16 *data)
>>> +{
>>> +    struct e1000_phy_info *phy = &hw->phy;
>>> +    u32 i, mdic = 0;
>>> +    s32 ret_val = 0;
>>> +
>>> +    if (offset > MAX_PHY_REG_ADDRESS) {
>>> +        hw_dbg("PHY Address %d is out of range\n", offset);
>>> +        ret_val = -E1000_ERR_PARAM;
>>> +        goto out;
>>> +    }
>>> +
>>> +    /* Set up Op-code, Phy Address, and register offset in the MDI
>>> +     * Control register.  The MAC will take care of interfacing with 
>>> the
>>> +     * PHY to retrieve the desired data.
>>> +     */
>>> +    mdic = ((offset << E1000_MDIC_REG_SHIFT) |
>>> +        (phy->addr << E1000_MDIC_PHY_SHIFT) |
>>> +        (E1000_MDIC_OP_READ));
>>> +
>>> +    wr32(E1000_MDIC, mdic);
>>> +
>>> +    /* Poll the ready bit to see if the MDI read completed
>>> +     * Increasing the time out as testing showed failures with
>>> +     * the lower time out
>>
>> Why not just make the #define a larger number?
>>
> Possible, but why do not keep same approach as another drivers.
fix will be applied in v5.
>>> +     */
>>> +    for (i = 0; i < (E1000_GEN_POLL_TIMEOUT * 3); i++) {
>>> +        usleep_range(500, 1000);
>>> +        mdic = rd32(E1000_MDIC);
>>> +        if (mdic & E1000_MDIC_READY)
>>> +            break;
>>> +    }
>>> +    if (!(mdic & E1000_MDIC_READY)) {
>>> +        hw_dbg("MDI Read did not complete\n");
>>> +        ret_val = -E1000_ERR_PHY;
>>> +        goto out;
>>> +    }
>>> +    if (mdic & E1000_MDIC_ERROR) {
>>> +        hw_dbg("MDI Error\n");
>>> +        ret_val = -E1000_ERR_PHY;
>>> +        goto out;
>>> +    }
>>> +    *data = (u16)mdic;
>>> +
>>> +out:
>>> +    return ret_val;
>>> +}
>>> +
>>> +/**
>>> + *  igc_write_phy_reg_mdic - Write MDI control register
>>> + *  @hw: pointer to the HW structure
>>> + *  @offset: register offset to write to
>>> + *  @data: data to write to register at offset
>>> + *
>>> + *  Writes data to MDI control register in the PHY at offset.
>>> + **/
>>> +static s32 igc_write_phy_reg_mdic(struct e1000_hw *hw, u32 offset, 
>>> u16 data)
>>> +{
>>> +    struct e1000_phy_info *phy = &hw->phy;
>>> +    u32 i, mdic = 0;
>>> +    s32 ret_val = 0;
>>> +
>>> +    if (offset > MAX_PHY_REG_ADDRESS) {
>>> +        hw_dbg("PHY Address %d is out of range\n", offset);
>>> +        ret_val = -E1000_ERR_PARAM;
>>> +        goto out;
>>> +    }
>>> +
>>> +    /* Set up Op-code, Phy Address, and register offset in the MDI
>>> +     * Control register.  The MAC will take care of interfacing with 
>>> the
>>> +     * PHY to retrieve the desired data.
>>
>> s/retrieve/write/
>>
> fix will be applied in v4.
>>> +     */
>>> +    mdic = (((u32)data) |
>>> +        (offset << E1000_MDIC_REG_SHIFT) |
>>> +        (phy->addr << E1000_MDIC_PHY_SHIFT) |
>>> +        (E1000_MDIC_OP_WRITE));
>>> +
>>> +    wr32(E1000_MDIC, mdic);
>>> +
>>> +    /* Poll the ready bit to see if the MDI read completed
>>> +     * Increasing the time out as testing showed failures with
>>> +     * the lower time out
>>
>> Yeah, just fix the #define
>>
> Same as above.
>>> +     */
>>> +    for (i = 0; i < (E1000_GEN_POLL_TIMEOUT * 3); i++) {
>>> +        usleep_range(500, 1000);
>>> +        mdic = rd32(E1000_MDIC);
>>> +        if (mdic & E1000_MDIC_READY)
>>> +            break;
>>> +    }
>>> +    if (!(mdic & E1000_MDIC_READY)) {
>>> +        hw_dbg("MDI Write did not complete\n");
>>> +        ret_val = -E1000_ERR_PHY;
>>> +        goto out;
>>> +    }
>>> +    if (mdic & E1000_MDIC_ERROR) {
>>> +        hw_dbg("MDI Error\n");
>>> +        ret_val = -E1000_ERR_PHY;
>>> +        goto out;
>>> +    }
>>> +
>>> +out:
>>> +    return ret_val;
>>> +}
>>> +
>>> +/**
>>> + *  __igc_access_xmdio_reg - Read/write XMDIO register
>>> + *  @hw: pointer to the HW structure
>>> + *  @address: XMDIO address to program
>>> + *  @dev_addr: device address to program
>>> + *  @data: pointer to value to read/write from/to the XMDIO address
>>> + *  @read: boolean flag to indicate read or write
>>> + **/
>>> +static s32 __igc_access_xmdio_reg(struct e1000_hw *hw, u16 address,
>>> +                  u8 dev_addr, u16 *data, bool read)
>>> +{
>>> +    s32 ret_val;
>>> +
>>> +    ret_val = hw->phy.ops.write_reg(hw, E1000_MMDAC, dev_addr);
>>> +    if (ret_val)
>>> +        return ret_val;
>>> +
>>> +    ret_val = hw->phy.ops.write_reg(hw, E1000_MMDAAD, address);
>>> +    if (ret_val)
>>> +        return ret_val;
>>> +
>>> +    ret_val = hw->phy.ops.write_reg(hw, E1000_MMDAC, 
>>> E1000_MMDAC_FUNC_DATA |
>>> +                    dev_addr);
>>> +    if (ret_val)
>>> +        return ret_val;
>>> +
>>> +    if (read)
>>> +        ret_val = hw->phy.ops.read_reg(hw, E1000_MMDAAD, data);
>>
>> Is there no wait needed before the read?
>>
> ok. I can say that this piece of code work as properly on our HW without 
> delay. I don't saw any problem.
>>> +    else
>>> +        ret_val = hw->phy.ops.write_reg(hw, E1000_MMDAAD, *data);
>>> +    if (ret_val)
>>> +        return ret_val;
>>> +
>>> +    /* Recalibrate the device back to 0 */
>>> +    ret_val = hw->phy.ops.write_reg(hw, E1000_MMDAC, 0);
>>> +    if (ret_val)
>>> +        return ret_val;
>>> +
>>> +    return ret_val;
>>> +}
>>> +
>>> +/**
>>> + *  igc_read_xmdio_reg - Read XMDIO register
>>> + *  @hw: pointer to the HW structure
>>> + *  @addr: XMDIO address to program
>>> + *  @dev_addr: device address to program
>>> + *  @data: value to be read from the EMI address
>>> + **/
>>> +static s32 igc_read_xmdio_reg(struct e1000_hw *hw, u16 addr,
>>> +                  u8 dev_addr, u16 *data)
>>> +{
>>> +    return __igc_access_xmdio_reg(hw, addr, dev_addr, data, true);
>>> +}
>>> +
>>> +/**
>>> + *  igc_write_xmdio_reg - Write XMDIO register
>>> + *  @hw: pointer to the HW structure
>>> + *  @addr: XMDIO address to program
>>> + *  @dev_addr: device address to program
>>> + *  @data: value to be written to the XMDIO address
>>> + **/
>>> +static s32 igc_write_xmdio_reg(struct e1000_hw *hw, u16 addr,
>>> +                   u8 dev_addr, u16 data)
>>> +{
>>> +    return __igc_access_xmdio_reg(hw, addr, dev_addr, &data, false);
>>> +}
>>> +
>>> +/**
>>> + *  igc_write_phy_reg_gpy - Write GPY PHY register
>>> + *  @hw: pointer to the HW structure
>>> + *  @offset: register offset to write to
>>> + *  @data: data to write at register offset
>>> + *
>>> + *  Acquires semaphore, if necessary, then writes the data to PHY 
>>> register
>>> + *  at the offset.  Release any acquired semaphores before exiting.
>>> + **/
>>> +s32 igc_write_phy_reg_gpy(struct e1000_hw *hw, u32 offset, u16 data)
>>> +{
>>> +    s32 ret_val;
>>> +    u8 dev_addr = (offset & GPY_MMD_MASK) >> GPY_MMD_SHIFT;
>>> +
>>> +    offset = offset & GPY_REG_MASK;
>>> +
>>> +    if (!dev_addr) {
>>> +        ret_val = hw->phy.ops.acquire(hw);
>>> +        if (ret_val)
>>> +            return ret_val;
>>> +        ret_val = igc_write_phy_reg_mdic(hw, offset, data);
>>> +        if (ret_val)
>>> +            return ret_val;
>>> +        hw->phy.ops.release(hw);
>>> +    } else {
>>> +        ret_val = igc_write_xmdio_reg(hw, (u16)offset, dev_addr,
>>> +                          data);
>>> +    }
>>> +    return ret_val;
>>> +}
>>> +
>>> +/**
>>> + *  igc_read_phy_reg_gpy - Read GPY PHY register
>>> + *  @hw: pointer to the HW structure
>>> + *  @offset: lower half is register offset to read to
>>> + *     upper half is MMD to use.
>>> + *  @data: data to read at register offset
>>> + *
>>> + *  Acquires semaphore, if necessary, then reads the data in the PHY 
>>> register
>>> + *  at the offset.  Release any acquired semaphores before exiting.
>>> + **/
>>> +s32 igc_read_phy_reg_gpy(struct e1000_hw *hw, u32 offset, u16 *data)
>>> +{
>>> +    s32 ret_val;
>>> +    u8 dev_addr = (offset & GPY_MMD_MASK) >> GPY_MMD_SHIFT;
>>> +
>>> +    offset = offset & GPY_REG_MASK;
>>> +
>>> +    if (!dev_addr) {
>>> +        ret_val = hw->phy.ops.acquire(hw);
>>> +        if (ret_val)
>>> +            return ret_val;
>>> +        ret_val = igc_read_phy_reg_mdic(hw, offset, data);
>>> +        if (ret_val)
>>> +            return ret_val;
>>> +        hw->phy.ops.release(hw);
>>> +    } else {
>>> +        ret_val = igc_read_xmdio_reg(hw, (u16)offset, dev_addr,
>>> +                         data);
>>> +    }
>>> +    return ret_val;
>>> +}
>>> +
>>> diff --git a/drivers/net/ethernet/intel/igc/e1000_phy.h 
>>> b/drivers/net/ethernet/intel/igc/e1000_phy.h
>>> new file mode 100644
>>> index 000000000000..d04dfd0be7fd
>>> --- /dev/null
>>> +++ b/drivers/net/ethernet/intel/igc/e1000_phy.h
>>> @@ -0,0 +1,21 @@
>>> +/* SPDX-License-Identifier: GPL-2.0 */
>>> +/* Copyright (c)  2018 Intel Corporation */
>>> +
>>> +#ifndef _E1000_PHY_H_
>>> +#define _E1000_PHY_H_
>>> +
>>> +#include "e1000_mac.h"
>>> +
>>> +s32  igc_check_reset_block(struct e1000_hw *hw);
>>> +s32  igc_phy_hw_reset(struct e1000_hw *hw);
>>> +s32  igc_get_phy_id(struct e1000_hw *hw);
>>> +s32  igc_phy_has_link(struct e1000_hw *hw, u32 iterations,
>>> +              u32 usec_interval, bool *success);
>>> +s32 igc_check_downshift(struct e1000_hw *hw);
>>> +void igc_power_up_phy_copper(struct e1000_hw *hw);
>>> +void igc_power_down_phy_copper(struct e1000_hw *hw);
>>> +
>>> +s32  igc_write_phy_reg_gpy(struct e1000_hw *hw, u32 offset, u16 data);
>>> +s32  igc_read_phy_reg_gpy(struct e1000_hw *hw, u32 offset, u16 *data);
>>> +
>>> +#endif
>>> diff --git a/drivers/net/ethernet/intel/igc/e1000_regs.h 
>>> b/drivers/net/ethernet/intel/igc/e1000_regs.h
>>> index 698ce0cac757..3fcaf606ce20 100644
>>> --- a/drivers/net/ethernet/intel/igc/e1000_regs.h
>>> +++ b/drivers/net/ethernet/intel/igc/e1000_regs.h
>>> @@ -55,6 +55,9 @@
>>>   #define E1000_SWSM        0x05B50  /* SW Semaphore */
>>>   #define E1000_FWSM        0x05B54  /* FW Semaphore */
>>> +/* Function Active and Power State to MNG */
>>> +#define E1000_FACTPS        0x05B30
>>> +
>>>   /* Interrupt Register Description */
>>>   #define E1000_PICAUSE        0x05B88  /* PCIe Interrupt Cause - 
>>> RW1/C */
>>>   #define E1000_PIENA        0x05B8C  /* PCIe Interrupt enable - RW */
>>> diff --git a/drivers/net/ethernet/intel/igc/igc.h 
>>> b/drivers/net/ethernet/intel/igc/igc.h
>>> index 735a5e3d0717..91df0a14c3a5 100644
>>> --- a/drivers/net/ethernet/intel/igc/igc.h
>>> +++ b/drivers/net/ethernet/intel/igc/igc.h
>>> @@ -365,6 +365,22 @@ static inline u16 igc_desc_unused(const struct 
>>> igc_ring *ring)
>>>       return ((ntc > ntu) ? 0 : ring->count) + ntc - ntu - 1;
>>>   }
>>> +static inline s32 igc_get_phy_info(struct e1000_hw *hw)
>>> +{
>>> +    if (hw->phy.ops.get_phy_info)
>>> +        return hw->phy.ops.get_phy_info(hw);
>>> +
>>> +    return 0;
>>> +}
>>> +
>>> +static inline s32 igc_reset_phy(struct e1000_hw *hw)
>>> +{
>>> +    if (hw->phy.ops.reset)
>>> +        return hw->phy.ops.reset(hw);
>>> +
>>> +    return 0;
>>> +}
>>> +
>>>   static inline struct netdev_queue *txring_txq(const struct igc_ring 
>>> *tx_ring)
>>>   {
>>>       return netdev_get_tx_queue(tx_ring->netdev, tx_ring->queue_index);
>>> diff --git a/drivers/net/ethernet/intel/igc/igc_main.c 
>>> b/drivers/net/ethernet/intel/igc/igc_main.c
>>> index dcc7e700074f..21aed91454d4 100644
>>> --- a/drivers/net/ethernet/intel/igc/igc_main.c
>>> +++ b/drivers/net/ethernet/intel/igc/igc_main.c
>>> @@ -100,6 +100,8 @@ void igc_reset(struct igc_adapter *adapter)
>>>       if (!netif_running(adapter->netdev))
>>>           igc_power_down_link(adapter);
>>> +
>>> +    igc_get_phy_info(hw);
>>>   }
>>>   /**
>>> @@ -108,6 +110,12 @@ void igc_reset(struct igc_adapter *adapter)
>>>    **/
>>>   void igc_power_up_link(struct igc_adapter *adapter)
>>>   {
>>> +    igc_reset_phy(&adapter->hw);
>>> +
>>> +    if (adapter->hw.phy.media_type == e1000_media_type_copper)
>>> +        igc_power_up_phy_copper(&adapter->hw);
>>> +
>>> +    igc_setup_link(&adapter->hw);
>>>   }
>>>   /**
>>> @@ -116,6 +124,8 @@ void igc_power_up_link(struct igc_adapter *adapter)
>>>    **/
>>>   static void igc_power_down_link(struct igc_adapter *adapter)
>>>   {
>>> +    if (adapter->hw.phy.media_type == e1000_media_type_copper)
>>> +        igc_power_down_phy_copper_base(&adapter->hw);
>>>   }
>>>   /**
>>> @@ -3424,6 +3434,7 @@ static int igc_probe(struct pci_dev *pdev,
>>>       /* Copy the default MAC and PHY function pointers */
>>>       memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
>>> +    memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
>>>       /* Initialize skew-specific constants */
>>>       err = ei->get_invariants(hw);
>>>
> Thanks for your comments.
> 
> _______________________________________________
> Intel-wired-lan mailing list
> Intel-wired-lan at osuosl.org
> https://lists.osuosl.org/mailman/listinfo/intel-wired-lan



More information about the Intel-wired-lan mailing list