[dpdk-dev,v4,1/4] ethdev: add apis to support access device info

  add new apis:
- rte_eth_dev_default_mac_addr_set
- rte_eth_dev_reg_leng
- rte_eth_dev_reg_info
- rte_eth_dev_eeprom_leng
- rte_eth_dev_get_eeprom
- rte_eth_dev_set_eeprom
- rte_eth_dev_get_ringparam
- rte_eth_dev_set_ringparam

to enable reading device parameters (mac-addr, register,
eeprom, ring) based upon ethtool alike
data parameter specification.

Signed-off-by: Liang-Min Larry Wang <liang-min.wang@intel.com>
---
 lib/librte_ether/Makefile              |   1 +
 lib/librte_ether/rte_eth_dev_info.h    |  80 +++++++++++++++++
 lib/librte_ether/rte_ethdev.c          | 159 +++++++++++++++++++++++++++++++++
 lib/librte_ether/rte_ethdev.h          | 158 ++++++++++++++++++++++++++++++++
 lib/librte_ether/rte_ether_version.map |   8 ++
 5 files changed, 406 insertions(+)
 create mode 100644 lib/librte_ether/rte_eth_dev_info.h
  

Sorry for spam and providing feedback here again.

> -----Original Message-----
> From: Liang-Min Larry Wang [mailto:liang-min.wang@intel.com]
> Sent: Wednesday, June 10, 2015 11:10 AM
> To: dev@dpdk.org
> Cc: Andrew Harvey (agh); Roger Melton (rmelton); David Harton (dharton);
> Liang-Min Larry Wang
> Subject: [PATCH v4 1/4] ethdev: add apis to support access device info
> 
> add new apis:
> - rte_eth_dev_default_mac_addr_set
> - rte_eth_dev_reg_leng
> - rte_eth_dev_reg_info
> - rte_eth_dev_eeprom_leng

<dharton> I know this is a bit of a nit but please consider changing "leng" to "length".
Some places already use "length" in the diffs below so using length would be consistent.

> - rte_eth_dev_get_eeprom
> - rte_eth_dev_set_eeprom
> - rte_eth_dev_get_ringparam
> - rte_eth_dev_set_ringparam
> 
> to enable reading device parameters (mac-addr, register, eeprom, ring)
> based upon ethtool alike data parameter specification.
> 
> Signed-off-by: Liang-Min Larry Wang <liang-min.wang@intel.com>
> ---
>  lib/librte_ether/Makefile              |   1 +
>  lib/librte_ether/rte_eth_dev_info.h    |  80 +++++++++++++++++
>  lib/librte_ether/rte_ethdev.c          | 159
> +++++++++++++++++++++++++++++++++
>  lib/librte_ether/rte_ethdev.h          | 158
> ++++++++++++++++++++++++++++++++
>  lib/librte_ether/rte_ether_version.map |   8 ++
>  5 files changed, 406 insertions(+)
>  create mode 100644 lib/librte_ether/rte_eth_dev_info.h
> 
> diff --git a/lib/librte_ether/Makefile b/lib/librte_ether/Makefile index
> c0e5768..05209e9 100644
> --- a/lib/librte_ether/Makefile
> +++ b/lib/librte_ether/Makefile
> @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c  SYMLINK-y-include += rte_ether.h
> SYMLINK-y-include += rte_ethdev.h  SYMLINK-y-include += rte_eth_ctrl.h
> +SYMLINK-y-include += rte_eth_dev_info.h
> 
>  # this lib depends upon:
>  DEPDIRS-y += lib/librte_eal lib/librte_mempool lib/librte_ring
> lib/librte_mbuf diff --git a/lib/librte_ether/rte_eth_dev_info.h
> b/lib/librte_ether/rte_eth_dev_info.h
> new file mode 100644
> index 0000000..002c4b5
> --- /dev/null
> +++ b/lib/librte_ether/rte_eth_dev_info.h
> @@ -0,0 +1,80 @@
> +/*-
> + *   BSD LICENSE
> + *
> + *   Copyright(c) 2015 Intel Corporation. All rights reserved.
> + *   All rights reserved.
> + *
> + *   Redistribution and use in source and binary forms, with or without
> + *   modification, are permitted provided that the following conditions
> + *   are met:
> + *
> + *     * Redistributions of source code must retain the above copyright
> + *       notice, this list of conditions and the following disclaimer.
> + *     * Redistributions in binary form must reproduce the above
> copyright
> + *       notice, this list of conditions and the following disclaimer in
> + *       the documentation and/or other materials provided with the
> + *       distribution.
> + *     * Neither the name of Intel Corporation nor the names of its
> + *       contributors may be used to endorse or promote products derived
> + *       from this software without specific prior written permission.
> + *
> + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
> + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
> + *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
> FOR
> + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
> + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
> INCIDENTAL,
> + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
> + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
> USE,
> + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
> ANY
> + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
> + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
> USE
> + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
> + */
> +
> +#ifndef _RTE_ETH_DEV_INFO_H_
> +#define _RTE_ETH_DEV_INFO_H_
> +
> +
> +/*
> + * Placeholder for accessing device registers  */ struct
> +rte_dev_reg_info {
> +	void *buf; /**< Buffer for register */
> +	uint32_t offset; /**< Offset for 1st register to fetch */
> +	uint32_t leng; /**< Number of registers to fetch */
> +	uint32_t version; /**< Device version */ };
> +
> +/*
> + * Placeholder for accessing device eeprom  */ struct
> +rte_dev_eeprom_info {
> +	void *buf; /**< Buffer for eeprom */
> +	uint32_t offset; /**< Offset for 1st eeprom location to access */
> +	uint32_t leng; /**< Length of eeprom region to access */
> +	uint32_t magic; /**< Device ID */
> +};
> +
> +/*
> + * Placeholder for accessing device ring parameters  */ struct
> +rte_dev_ring_info {
> +	uint32_t rx_pending; /**< Number of outstanding Rx ring */
> +	uint32_t tx_pending; /**< Number of outstanding Tx ring */
> +	uint32_t rx_max_pending; /**< Maximum number of outstanding Rx ring
> */
> +	uint32_t tx_max_pending; /**< Maximum number of outstanding Tx ring
> */
> +};
> +
> +/*
> + * A data structure captures information as defined in struct
> +ifla_vf_info
> + * for user-space api
> + */
> +struct rte_dev_vf_info {
> +	uint32_t vf;
> +	uint8_t mac[ETHER_ADDR_LEN];
> +	uint32_t vlan;
> +	uint32_t tx_rate;
> +	uint32_t spoofchk;
> +};
> +
> +#endif /* _RTE_ETH_DEV_INFO_H_ */
> diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
> index 5a94654..186e85c 100644
> --- a/lib/librte_ether/rte_ethdev.c
> +++ b/lib/librte_ether/rte_ethdev.c
> @@ -2751,6 +2751,32 @@ rte_eth_dev_mac_addr_remove(uint8_t port_id, struct
> ether_addr *addr)  }
> 
>  int
> +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct ether_addr
> +*addr) {
> +	struct rte_eth_dev *dev;
> +	const int index = 0;
> +	const uint32_t pool = 0;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	dev = &rte_eth_devices[port_id];
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_remove, -ENOTSUP);
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -ENOTSUP);
> +
> +	/* Update NIC default MAC address*/
> +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> +
> +	/* Update default address in NIC data structure */
> +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> +
> +	return 0;
> +}
> +
> +int
>  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
>  				uint16_t rx_mode, uint8_t on)
>  {
> @@ -3627,3 +3653,136 @@ rte_eth_remove_tx_callback(uint8_t port_id,
> uint16_t queue_id,
>  	/* Callback wasn't found. */
>  	return -EINVAL;
>  }
> +
> +int
> +rte_eth_dev_reg_leng(uint8_t port_id)
> +{
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length, -ENOTSUP);
> +	return (*dev->dev_ops->get_reg_length)(dev);
> +}
> +
> +int
> +rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info *info) {
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -ENOTSUP);
> +	return (*dev->dev_ops->get_reg)(dev, info); }
> +
> +int
> +rte_eth_dev_eeprom_leng(uint8_t port_id) {
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom_length, -ENOTSUP);
> +	return (*dev->dev_ops->get_eeprom_length)(dev);
> +}
> +
> +int
> +rte_eth_dev_get_eeprom(uint8_t port_id, struct rte_dev_eeprom_info
> +*info) {
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -ENOTSUP);
> +	return (*dev->dev_ops->get_eeprom)(dev, info); }
> +
> +int
> +rte_eth_dev_set_eeprom(uint8_t port_id, struct rte_dev_eeprom_info
> +*info) {
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -ENOTSUP);
> +	return (*dev->dev_ops->set_eeprom)(dev, info); }
> +
> +int
> +rte_eth_dev_get_ringparam(uint8_t port_id, struct rte_dev_ring_info
> +*info) {
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -ENOTSUP);
> +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> +
> +int
> +rte_eth_dev_set_ringparam(uint8_t port_id, struct rte_dev_ring_info
> +*info) {
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_ringparam, -ENOTSUP);
> +	return (*dev->dev_ops->set_ringparam)(dev, info); }
> diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
> index 16dbe00..8f99f0d 100644
> --- a/lib/librte_ether/rte_ethdev.h
> +++ b/lib/librte_ether/rte_ethdev.h
> @@ -182,6 +182,7 @@ extern "C" {
>  #include <rte_devargs.h>
>  #include "rte_ether.h"
>  #include "rte_eth_ctrl.h"
> +#include "rte_eth_dev_info.h"
> 
>  struct rte_mbuf;
> 
> @@ -1228,6 +1229,31 @@ typedef int (*eth_udp_tunnel_del_t)(struct
> rte_eth_dev *dev,
>  				    struct rte_eth_udp_tunnel *tunnel_udp);  /**<
> @internal Delete tunneling UDP info */
> 
> +typedef int (*eth_get_reg_length_t)(struct rte_eth_dev *dev); /**<
> +@internal Retrieve device register count  */
> +
> +typedef int (*eth_get_reg_t)(struct rte_eth_dev *dev,
> +				struct rte_dev_reg_info *info);
> +/**< @internal Retrieve registers  */
> +
> +typedef int (*eth_get_eeprom_length_t)(struct rte_eth_dev *dev); /**<
> +@internal Retrieve eeprom size  */
> +
> +typedef int (*eth_get_eeprom_t)(struct rte_eth_dev *dev,
> +				struct rte_dev_eeprom_info *info);
> +/**< @internal Retrieve eeprom data  */
> +
> +typedef int (*eth_set_eeprom_t)(struct rte_eth_dev *dev,
> +				struct rte_dev_eeprom_info *info);
> +/**< @internal Program eeprom data  */
> +
> +typedef int (*eth_get_ringparam_t)(struct rte_eth_dev *dev,
> +				struct rte_dev_ring_info *info);
> +/**< @internal Retrieve device ring descriptor  */
> +
> +typedef int (*eth_set_ringparam_t)(struct rte_eth_dev *dev,
> +				struct rte_dev_ring_info *info);
> +/**< @internal Set device ring descriptor  */
> 
>  #ifdef RTE_NIC_BYPASS
> 
> @@ -1368,6 +1394,21 @@ struct eth_dev_ops {
>  	reta_update_t reta_update;
>  	/** Query redirection table. */
>  	reta_query_t reta_query;
> +
> +	eth_get_reg_length_t get_reg_length;
> +	/**< Get # of registers */
> +	eth_get_reg_t get_reg;
> +	/**< Get registers */
> +	eth_get_eeprom_length_t get_eeprom_length;
> +	/**< Get eeprom length */
> +	eth_get_eeprom_t get_eeprom;
> +	/**< Get eeprom data */
> +	eth_set_eeprom_t set_eeprom;
> +	/**< Set eeprom */
> +	eth_get_ringparam_t get_ringparam;
> +	/**< Get descriptor ring parameters */
> +	eth_set_ringparam_t set_ringparam;
> +	/**< Set descriptor ring parameters */
>    /* bypass control */
>  #ifdef RTE_NIC_BYPASS
>    bypass_init_t bypass_init;
> @@ -2982,6 +3023,20 @@ int rte_eth_dev_mac_addr_add(uint8_t port, struct
> ether_addr *mac_addr,  int rte_eth_dev_mac_addr_remove(uint8_t port,
> struct ether_addr *mac_addr);
> 
>  /**
> + * Set the default MAC address.
> + *
> + * @param port
> + *   The port identifier of the Ethernet device.
> + * @param mac_addr
> + *   New default MAC address.
> + * @return
> + *   - (0) if successful, or *mac_addr* didn't exist.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port* invalid.
> + */
> +int rte_eth_dev_default_mac_addr_set(uint8_t port, struct ether_addr
> +*mac_addr);
> +
> +/**
>   * Update Redirection Table(RETA) of Receive Side Scaling of Ethernet
> device.
>   *
>   * @param port
> @@ -3611,6 +3666,109 @@ int rte_eth_remove_rx_callback(uint8_t port_id,
> uint16_t queue_id,  int rte_eth_remove_tx_callback(uint8_t port_id,
> uint16_t queue_id,
>  		struct rte_eth_rxtx_callback *user_cb);
> 
> +/**
> + * Retrieve number of available registers for access
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @return
> + *   - (>=0) number of registers if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_reg_leng(uint8_t port_id);
> +
> +/**
> + * Retrieve device registers and register attributes
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @param info
> + *   The template includes buffer for register data and attribute to be
> filled.
> + * @return
> + *   - (0) if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info
> +*info);
> +
> +/**
> + * Retrieve size of device EEPROM
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @return
> + *   - (>=0) EEPROM size if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_eeprom_leng(uint8_t port_id);
> +
> +/**
> + * Retrieve EEPROM and EEPROM attribute
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @param info
> + *   The template includes buffer for return EEPROM data and
> + *   EEPROM attributes to be filled.
> + * @return
> + *   - (0) if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_get_eeprom(uint8_t port_id, struct rte_dev_eeprom_info
> +*info);
> +
> +/**
> + * Program EEPROM with provided data
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @param info
> + *   The template includes EEPROM data for programming and
> + *   EEPROM attributes to be filled
> + * @return
> + *   - (0) if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_set_eeprom(uint8_t port_id, struct rte_dev_eeprom_info
> +*info);
> +
> +/**
> + * Retrieve device descriptor ring information
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @param info
> + *   The template includes device ring descriptor usage information.
> + * @return
> + *   - (0) if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_get_ringparam(uint8_t port_id, struct rte_dev_ring_info
> +*info);
> +
> +/**
> + * Set device ring descriptor parameters
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @param info
> + *   The template includes descriptor ring parameters to be set.
> + * @return
> + *   - (0) if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_set_ringparam(uint8_t port_id, struct rte_dev_ring_info
> +*info);
> +
>  #ifdef __cplusplus
>  }
>  #endif
> diff --git a/lib/librte_ether/rte_ether_version.map
> b/lib/librte_ether/rte_ether_version.map
> index a2d25a6..1b1e092 100644
> --- a/lib/librte_ether/rte_ether_version.map
> +++ b/lib/librte_ether/rte_ether_version.map
> @@ -102,6 +102,14 @@ DPDK_2.0 {
>  	rte_eth_tx_queue_setup;
>  	rte_eth_xstats_get;
>  	rte_eth_xstats_reset;
> +	rte_eth_dev_default_mac_addr_set;
> +	rte_eth_dev_reg_leng;
> +	rte_eth_dev_reg_info;
> +	rte_eth_dev_eeprom_leng;
> +	rte_eth_dev_get_eeprom;
> +	rte_eth_dev_set_eeprom;
> +	rte_eth_dev_get_ringparam;
> +	rte_eth_dev_set_ringparam;
> 
>  	local: *;
>  };
> --
> 2.1.4

Thanks for these changes as they will prove useful in the user space.

I have no other comments beyond the one above.

Regards,
Dave
  

Hi Larry,

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Liang-Min Larry Wang
> Sent: Wednesday, June 10, 2015 4:10 PM
> To: dev@dpdk.org
> Cc: Wang, Liang-min
> Subject: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access device info
> 
> add new apis:
> - rte_eth_dev_default_mac_addr_set
> - rte_eth_dev_reg_leng
> - rte_eth_dev_reg_info
> - rte_eth_dev_eeprom_leng
> - rte_eth_dev_get_eeprom
> - rte_eth_dev_set_eeprom
> - rte_eth_dev_get_ringparam
> - rte_eth_dev_set_ringparam
> 
> to enable reading device parameters (mac-addr, register,
> eeprom, ring) based upon ethtool alike
> data parameter specification.
> 
> Signed-off-by: Liang-Min Larry Wang <liang-min.wang@intel.com>
> ---
>  lib/librte_ether/Makefile              |   1 +
>  lib/librte_ether/rte_eth_dev_info.h    |  80 +++++++++++++++++
>  lib/librte_ether/rte_ethdev.c          | 159 +++++++++++++++++++++++++++++++++
>  lib/librte_ether/rte_ethdev.h          | 158 ++++++++++++++++++++++++++++++++
>  lib/librte_ether/rte_ether_version.map |   8 ++
>  5 files changed, 406 insertions(+)
>  create mode 100644 lib/librte_ether/rte_eth_dev_info.h
> 
> diff --git a/lib/librte_ether/Makefile b/lib/librte_ether/Makefile
> index c0e5768..05209e9 100644
> --- a/lib/librte_ether/Makefile
> +++ b/lib/librte_ether/Makefile
> @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c
>  SYMLINK-y-include += rte_ether.h
>  SYMLINK-y-include += rte_ethdev.h
>  SYMLINK-y-include += rte_eth_ctrl.h
> +SYMLINK-y-include += rte_eth_dev_info.h
> 
>  # this lib depends upon:
>  DEPDIRS-y += lib/librte_eal lib/librte_mempool lib/librte_ring lib/librte_mbuf
> diff --git a/lib/librte_ether/rte_eth_dev_info.h b/lib/librte_ether/rte_eth_dev_info.h
> new file mode 100644
> index 0000000..002c4b5
> --- /dev/null
> +++ b/lib/librte_ether/rte_eth_dev_info.h
> @@ -0,0 +1,80 @@
> +/*-
> + *   BSD LICENSE
> + *
> + *   Copyright(c) 2015 Intel Corporation. All rights reserved.
> + *   All rights reserved.
> + *
> + *   Redistribution and use in source and binary forms, with or without
> + *   modification, are permitted provided that the following conditions
> + *   are met:
> + *
> + *     * Redistributions of source code must retain the above copyright
> + *       notice, this list of conditions and the following disclaimer.
> + *     * Redistributions in binary form must reproduce the above copyright
> + *       notice, this list of conditions and the following disclaimer in
> + *       the documentation and/or other materials provided with the
> + *       distribution.
> + *     * Neither the name of Intel Corporation nor the names of its
> + *       contributors may be used to endorse or promote products derived
> + *       from this software without specific prior written permission.
> + *
> + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
> + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
> + *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
> + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
> + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
> + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
> + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
> + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
> + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
> + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
> + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
> + */
> +
> +#ifndef _RTE_ETH_DEV_INFO_H_
> +#define _RTE_ETH_DEV_INFO_H_
> +
> +
> +/*
> + * Placeholder for accessing device registers
> + */
> +struct rte_dev_reg_info {
> +	void *buf; /**< Buffer for register */
> +	uint32_t offset; /**< Offset for 1st register to fetch */
> +	uint32_t leng; /**< Number of registers to fetch */
> +	uint32_t version; /**< Device version */
> +};
> +
> +/*
> + * Placeholder for accessing device eeprom
> + */
> +struct rte_dev_eeprom_info {
> +	void *buf; /**< Buffer for eeprom */
> +	uint32_t offset; /**< Offset for 1st eeprom location to access */
> +	uint32_t leng; /**< Length of eeprom region to access */
> +	uint32_t magic; /**< Device ID */
> +};
> +
> +/*
> + * Placeholder for accessing device ring parameters
> + */
> +struct rte_dev_ring_info {
> +	uint32_t rx_pending; /**< Number of outstanding Rx ring */
> +	uint32_t tx_pending; /**< Number of outstanding Tx ring */
> +	uint32_t rx_max_pending; /**< Maximum number of outstanding Rx ring */
> +	uint32_t tx_max_pending; /**< Maximum number of outstanding Tx ring */
> +};
> +
> +/*
> + * A data structure captures information as defined in struct ifla_vf_info
> + * for user-space api
> + */
> +struct rte_dev_vf_info {
> +	uint32_t vf;
> +	uint8_t mac[ETHER_ADDR_LEN];
> +	uint32_t vlan;
> +	uint32_t tx_rate;
> +	uint32_t spoofchk;
> +};


Wonder what that structure is for?
I can't see it used in any function below?

> +
> +#endif /* _RTE_ETH_DEV_INFO_H_ */
> diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
> index 5a94654..186e85c 100644
> --- a/lib/librte_ether/rte_ethdev.c
> +++ b/lib/librte_ether/rte_ethdev.c
> @@ -2751,6 +2751,32 @@ rte_eth_dev_mac_addr_remove(uint8_t port_id, struct ether_addr *addr)
>  }
> 
>  int
> +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct ether_addr *addr)
> +{
> +	struct rte_eth_dev *dev;
> +	const int index = 0;
> +	const uint32_t pool = 0;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	dev = &rte_eth_devices[port_id];
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_remove, -ENOTSUP);
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -ENOTSUP);
> +
> +	/* Update NIC default MAC address*/
> +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> +
> +	/* Update default address in NIC data structure */
> +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> +
> +	return 0;
> +}
> +
> +int
>  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
>  				uint16_t rx_mode, uint8_t on)
>  {
> @@ -3627,3 +3653,136 @@ rte_eth_remove_tx_callback(uint8_t port_id, uint16_t queue_id,
>  	/* Callback wasn't found. */
>  	return -EINVAL;
>  }
> +
> +int
> +rte_eth_dev_reg_leng(uint8_t port_id)
> +{
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length, -ENOTSUP);
> +	return (*dev->dev_ops->get_reg_length)(dev);
> +}
> +
> +int
> +rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info *info)
> +{
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -ENOTSUP);
> +	return (*dev->dev_ops->get_reg)(dev, info);
> +}

Seems that *get_reg* stuff, will be really good addition for DPDK debugging abilities.
Though, I'd suggest we change it a bit to make more generic and flexible:

Introduce rte_eth_reg_read/rte_eth_reg_write(),
or probably even better rte_pcidev_reg_read /rte_pcidev_reg_write at EAL.
Something similar to what port_pci_reg_read/port_pci_reg_write() are doing now at testpmd.h.

struct rte_pcidev_reg_info {
   const char *name;
   uint32_t endianes, bar, offset, size, count;
};

int rte_pcidev_reg_read(const struct rte_pci_device *, const struct rte_pcidev_reg_info *, uint64_t *reg_val);

Then:
int rte_eth_dev_get_reg_info(port_id, const struct rte_pcidev_reg_info **info);

So each device would store in info a pointer to an array of it's register descriptions
(finished by zero elem?).

Then your ethtool (or any other upper layer) can do the following to read all device regs:

const struct rte_eth_dev_reg_info *reg_info;
struct rte_eth_dev_info dev_info;

rte_eth_dev_info_get(pid, &dev_info);
rte_eth_dev_get_reg_info(port_id, &reg_info);

for (i = 0; reg_info[i].name != NULL; i++) {
   ...
   rte_pcidev_read_reg(dev_info. pci_dev, reg_info[i], &v);
  ..
}

> +
> +int
> +rte_eth_dev_eeprom_leng(uint8_t port_id)
> +{
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom_length, -ENOTSUP);
> +	return (*dev->dev_ops->get_eeprom_length)(dev);
> +}
> +
> +int
> +rte_eth_dev_get_eeprom(uint8_t port_id, struct rte_dev_eeprom_info *info)
> +{
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -ENOTSUP);
> +	return (*dev->dev_ops->get_eeprom)(dev, info);
> +}
> +
> +int
> +rte_eth_dev_set_eeprom(uint8_t port_id, struct rte_dev_eeprom_info *info)
> +{
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -ENOTSUP);
> +	return (*dev->dev_ops->set_eeprom)(dev, info);
> +}
> +
> +int
> +rte_eth_dev_get_ringparam(uint8_t port_id, struct rte_dev_ring_info *info)
> +{
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -ENOTSUP);
> +	return (*dev->dev_ops->get_ringparam)(dev, info);
> +}

I think it will be a useful addition to the ethdev API  to have an ability to retrieve current RX/TX queue parameters.
Though again, it need to be more generic, so it could be useful for non-ethtool upper layer too.
So I suggest to modify it a bit.
Something like that:

struct rte_eth_tx_queue_info {
    struct rte_eth_txconf txconf;
    uint32_t nb_tx_desc;
    uint32_t nb_max_tx_desc; /*max allowable TXDs for that queue */
    uint32_t nb_tx_free;            /* number of free TXDs at the moment of call. */
    /* other tx queue data. */ 	
};

int rte_etdev_get_tx_queue_info(portid, queue_id, struct rte_eth_tx_queue_info *qinfo)

Then, your upper layer ethtool wrapper, can implement yours ethtool_get_ringparam() by:

 ...
 struct rte_eth_tx_queue_info qinfo;
 rte_ethdev_get_tx_queue_info(port, 0, &qinfo);
ring_param->tx_pending = qinfo.nb_tx_desc -  rte_eth_rx_queue_count(port, 0);

Or probably even:
ring_param->tx_pending = qinfo.nb_tx_desc - qinfo.nb_tx_free;

Same for RX.

> +
> +int
> +rte_eth_dev_set_ringparam(uint8_t port_id, struct rte_dev_ring_info *info)
> +{
> +	struct rte_eth_dev *dev;
> +
> +	if (!rte_eth_dev_is_valid_port(port_id)) {
> +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> +		return -ENODEV;
> +	}
> +
> +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> +		PMD_DEBUG_TRACE("Invalid port device\n");
> +		return -ENODEV;
> +	}
> +
> +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_ringparam, -ENOTSUP);
> +	return (*dev->dev_ops->set_ringparam)(dev, info);
> +}


I am a bit confused, what particular value of rte_dev_ring_info will be allowed to change?
You can't change number of RXD/TXD at runtime.
You have to stop the device first, and then reconfigure particular queue(s).
Again, you can't reset RXF/TXD without stopping RX/TX first.
So what that function is supposed to do?
As I can see currently, none of the PMD in your patch support it. 

Konstantin

> diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
> index 16dbe00..8f99f0d 100644
> --- a/lib/librte_ether/rte_ethdev.h
> +++ b/lib/librte_ether/rte_ethdev.h
> @@ -182,6 +182,7 @@ extern "C" {
>  #include <rte_devargs.h>
>  #include "rte_ether.h"
>  #include "rte_eth_ctrl.h"
> +#include "rte_eth_dev_info.h"
> 
>  struct rte_mbuf;
> 
> @@ -1228,6 +1229,31 @@ typedef int (*eth_udp_tunnel_del_t)(struct rte_eth_dev *dev,
>  				    struct rte_eth_udp_tunnel *tunnel_udp);
>  /**< @internal Delete tunneling UDP info */
> 
> +typedef int (*eth_get_reg_length_t)(struct rte_eth_dev *dev);
> +/**< @internal Retrieve device register count  */
> +
> +typedef int (*eth_get_reg_t)(struct rte_eth_dev *dev,
> +				struct rte_dev_reg_info *info);
> +/**< @internal Retrieve registers  */
> +
> +typedef int (*eth_get_eeprom_length_t)(struct rte_eth_dev *dev);
> +/**< @internal Retrieve eeprom size  */
> +
> +typedef int (*eth_get_eeprom_t)(struct rte_eth_dev *dev,
> +				struct rte_dev_eeprom_info *info);
> +/**< @internal Retrieve eeprom data  */
> +
> +typedef int (*eth_set_eeprom_t)(struct rte_eth_dev *dev,
> +				struct rte_dev_eeprom_info *info);
> +/**< @internal Program eeprom data  */
> +
> +typedef int (*eth_get_ringparam_t)(struct rte_eth_dev *dev,
> +				struct rte_dev_ring_info *info);
> +/**< @internal Retrieve device ring descriptor  */
> +
> +typedef int (*eth_set_ringparam_t)(struct rte_eth_dev *dev,
> +				struct rte_dev_ring_info *info);
> +/**< @internal Set device ring descriptor  */
> 
>  #ifdef RTE_NIC_BYPASS
> 
> @@ -1368,6 +1394,21 @@ struct eth_dev_ops {
>  	reta_update_t reta_update;
>  	/** Query redirection table. */
>  	reta_query_t reta_query;
> +
> +	eth_get_reg_length_t get_reg_length;
> +	/**< Get # of registers */
> +	eth_get_reg_t get_reg;
> +	/**< Get registers */
> +	eth_get_eeprom_length_t get_eeprom_length;
> +	/**< Get eeprom length */
> +	eth_get_eeprom_t get_eeprom;
> +	/**< Get eeprom data */
> +	eth_set_eeprom_t set_eeprom;
> +	/**< Set eeprom */
> +	eth_get_ringparam_t get_ringparam;
> +	/**< Get descriptor ring parameters */
> +	eth_set_ringparam_t set_ringparam;
> +	/**< Set descriptor ring parameters */
>    /* bypass control */
>  #ifdef RTE_NIC_BYPASS
>    bypass_init_t bypass_init;
> @@ -2982,6 +3023,20 @@ int rte_eth_dev_mac_addr_add(uint8_t port, struct ether_addr *mac_addr,
>  int rte_eth_dev_mac_addr_remove(uint8_t port, struct ether_addr *mac_addr);
> 
>  /**
> + * Set the default MAC address.
> + *
> + * @param port
> + *   The port identifier of the Ethernet device.
> + * @param mac_addr
> + *   New default MAC address.
> + * @return
> + *   - (0) if successful, or *mac_addr* didn't exist.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port* invalid.
> + */
> +int rte_eth_dev_default_mac_addr_set(uint8_t port, struct ether_addr *mac_addr);
> +
> +/**
>   * Update Redirection Table(RETA) of Receive Side Scaling of Ethernet device.
>   *
>   * @param port
> @@ -3611,6 +3666,109 @@ int rte_eth_remove_rx_callback(uint8_t port_id, uint16_t queue_id,
>  int rte_eth_remove_tx_callback(uint8_t port_id, uint16_t queue_id,
>  		struct rte_eth_rxtx_callback *user_cb);
> 
> +/**
> + * Retrieve number of available registers for access
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @return
> + *   - (>=0) number of registers if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_reg_leng(uint8_t port_id);
> +
> +/**
> + * Retrieve device registers and register attributes
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @param info
> + *   The template includes buffer for register data and attribute to be filled.
> + * @return
> + *   - (0) if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info *info);
> +
> +/**
> + * Retrieve size of device EEPROM
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @return
> + *   - (>=0) EEPROM size if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_eeprom_leng(uint8_t port_id);
> +
> +/**
> + * Retrieve EEPROM and EEPROM attribute
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @param info
> + *   The template includes buffer for return EEPROM data and
> + *   EEPROM attributes to be filled.
> + * @return
> + *   - (0) if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_get_eeprom(uint8_t port_id, struct rte_dev_eeprom_info *info);
> +
> +/**
> + * Program EEPROM with provided data
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @param info
> + *   The template includes EEPROM data for programming and
> + *   EEPROM attributes to be filled
> + * @return
> + *   - (0) if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_set_eeprom(uint8_t port_id, struct rte_dev_eeprom_info *info);
> +
> +/**
> + * Retrieve device descriptor ring information
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @param info
> + *   The template includes device ring descriptor usage information.
> + * @return
> + *   - (0) if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_get_ringparam(uint8_t port_id, struct rte_dev_ring_info *info);
> +
> +/**
> + * Set device ring descriptor parameters
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @param info
> + *   The template includes descriptor ring parameters to be set.
> + * @return
> + *   - (0) if successful.
> + *   - (-ENOTSUP) if hardware doesn't support.
> + *   - (-ENODEV) if *port_id* invalid.
> + *   - others depends on the specific operations implementation.
> + */
> +int rte_eth_dev_set_ringparam(uint8_t port_id, struct rte_dev_ring_info *info);
> +
>  #ifdef __cplusplus
>  }
>  #endif
> diff --git a/lib/librte_ether/rte_ether_version.map b/lib/librte_ether/rte_ether_version.map
> index a2d25a6..1b1e092 100644
> --- a/lib/librte_ether/rte_ether_version.map
> +++ b/lib/librte_ether/rte_ether_version.map
> @@ -102,6 +102,14 @@ DPDK_2.0 {
>  	rte_eth_tx_queue_setup;
>  	rte_eth_xstats_get;
>  	rte_eth_xstats_reset;
> +	rte_eth_dev_default_mac_addr_set;
> +	rte_eth_dev_reg_leng;
> +	rte_eth_dev_reg_info;
> +	rte_eth_dev_eeprom_leng;
> +	rte_eth_dev_get_eeprom;
> +	rte_eth_dev_set_eeprom;
> +	rte_eth_dev_get_ringparam;
> +	rte_eth_dev_set_ringparam;
> 
>  	local: *;
>  };
> --
> 2.1.4
  

Hi Konstantin,

> -----Original Message-----
> From: Ananyev, Konstantin
> Sent: Thursday, June 11, 2015 8:26 AM
> To: Wang, Liang-min; dev@dpdk.org
> Cc: Wang, Liang-min
> Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> device info
> 
> Hi Larry,
> 
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Liang-Min Larry
> > Wang
> > Sent: Wednesday, June 10, 2015 4:10 PM
> > To: dev@dpdk.org
> > Cc: Wang, Liang-min
> > Subject: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> > device info
> >
> > add new apis:
> > - rte_eth_dev_default_mac_addr_set
> > - rte_eth_dev_reg_leng
> > - rte_eth_dev_reg_info
> > - rte_eth_dev_eeprom_leng
> > - rte_eth_dev_get_eeprom
> > - rte_eth_dev_set_eeprom
> > - rte_eth_dev_get_ringparam
> > - rte_eth_dev_set_ringparam
> >
> > to enable reading device parameters (mac-addr, register, eeprom, ring)
> > based upon ethtool alike data parameter specification.
> >
> > Signed-off-by: Liang-Min Larry Wang <liang-min.wang@intel.com>
> > ---
> >  lib/librte_ether/Makefile              |   1 +
> >  lib/librte_ether/rte_eth_dev_info.h    |  80 +++++++++++++++++
> >  lib/librte_ether/rte_ethdev.c          | 159
> +++++++++++++++++++++++++++++++++
> >  lib/librte_ether/rte_ethdev.h          | 158
> ++++++++++++++++++++++++++++++++
> >  lib/librte_ether/rte_ether_version.map |   8 ++
> >  5 files changed, 406 insertions(+)
> >  create mode 100644 lib/librte_ether/rte_eth_dev_info.h
> >
> > diff --git a/lib/librte_ether/Makefile b/lib/librte_ether/Makefile
> > index c0e5768..05209e9 100644
> > --- a/lib/librte_ether/Makefile
> > +++ b/lib/librte_ether/Makefile
> > @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c  SYMLINK-y-include +=
> > rte_ether.h  SYMLINK-y-include += rte_ethdev.h  SYMLINK-y-include +=
> > rte_eth_ctrl.h
> > +SYMLINK-y-include += rte_eth_dev_info.h
> >
> >  # this lib depends upon:
> >  DEPDIRS-y += lib/librte_eal lib/librte_mempool lib/librte_ring
> > lib/librte_mbuf diff --git a/lib/librte_ether/rte_eth_dev_info.h
> > b/lib/librte_ether/rte_eth_dev_info.h
> > new file mode 100644
> > index 0000000..002c4b5
> > --- /dev/null
> > +++ b/lib/librte_ether/rte_eth_dev_info.h
> > @@ -0,0 +1,80 @@
> > +/*-
> > + *   BSD LICENSE
> > + *
> > + *   Copyright(c) 2015 Intel Corporation. All rights reserved.
> > + *   All rights reserved.
> > + *
> > + *   Redistribution and use in source and binary forms, with or without
> > + *   modification, are permitted provided that the following conditions
> > + *   are met:
> > + *
> > + *     * Redistributions of source code must retain the above copyright
> > + *       notice, this list of conditions and the following disclaimer.
> > + *     * Redistributions in binary form must reproduce the above copyright
> > + *       notice, this list of conditions and the following disclaimer in
> > + *       the documentation and/or other materials provided with the
> > + *       distribution.
> > + *     * Neither the name of Intel Corporation nor the names of its
> > + *       contributors may be used to endorse or promote products derived
> > + *       from this software without specific prior written permission.
> > + *
> > + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> CONTRIBUTORS
> > + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT
> NOT
> > + *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
> FITNESS FOR
> > + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
> COPYRIGHT
> > + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
> INCIDENTAL,
> > + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
> BUT NOT
> > + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
> LOSS OF USE,
> > + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
> AND ON ANY
> > + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
> TORT
> > + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
> OF THE USE
> > + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
> DAMAGE.
> > + */
> > +
> > +#ifndef _RTE_ETH_DEV_INFO_H_
> > +#define _RTE_ETH_DEV_INFO_H_
> > +
> > +
> > +/*
> > + * Placeholder for accessing device registers  */ struct
> > +rte_dev_reg_info {
> > +	void *buf; /**< Buffer for register */
> > +	uint32_t offset; /**< Offset for 1st register to fetch */
> > +	uint32_t leng; /**< Number of registers to fetch */
> > +	uint32_t version; /**< Device version */ };
> > +
> > +/*
> > + * Placeholder for accessing device eeprom  */ struct
> > +rte_dev_eeprom_info {
> > +	void *buf; /**< Buffer for eeprom */
> > +	uint32_t offset; /**< Offset for 1st eeprom location to access */
> > +	uint32_t leng; /**< Length of eeprom region to access */
> > +	uint32_t magic; /**< Device ID */
> > +};
> > +
> > +/*
> > + * Placeholder for accessing device ring parameters  */ struct
> > +rte_dev_ring_info {
> > +	uint32_t rx_pending; /**< Number of outstanding Rx ring */
> > +	uint32_t tx_pending; /**< Number of outstanding Tx ring */
> > +	uint32_t rx_max_pending; /**< Maximum number of outstanding Rx
> ring */
> > +	uint32_t tx_max_pending; /**< Maximum number of outstanding Tx
> ring
> > +*/ };
> > +
> > +/*
> > + * A data structure captures information as defined in struct
> > +ifla_vf_info
> > + * for user-space api
> > + */
> > +struct rte_dev_vf_info {
> > +	uint32_t vf;
> > +	uint8_t mac[ETHER_ADDR_LEN];
> > +	uint32_t vlan;
> > +	uint32_t tx_rate;
> > +	uint32_t spoofchk;
> > +};
> 
> 
> Wonder what that structure is for?
> I can't see it used in any function below?
> 

Good catch, this is designed for other ethtool ops that I did not include in this release, I will remove this from next fix.

> > +
> > +#endif /* _RTE_ETH_DEV_INFO_H_ */
> > diff --git a/lib/librte_ether/rte_ethdev.c
> > b/lib/librte_ether/rte_ethdev.c index 5a94654..186e85c 100644
> > --- a/lib/librte_ether/rte_ethdev.c
> > +++ b/lib/librte_ether/rte_ethdev.c
> > @@ -2751,6 +2751,32 @@ rte_eth_dev_mac_addr_remove(uint8_t
> port_id,
> > struct ether_addr *addr)  }
> >
> >  int
> > +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct ether_addr
> > +*addr) {
> > +	struct rte_eth_dev *dev;
> > +	const int index = 0;
> > +	const uint32_t pool = 0;
> > +
> > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > +		return -ENODEV;
> > +	}
> > +
> > +	dev = &rte_eth_devices[port_id];
> > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_remove, -
> ENOTSUP);
> > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -
> ENOTSUP);
> > +
> > +	/* Update NIC default MAC address*/
> > +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> > +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> > +
> > +	/* Update default address in NIC data structure */
> > +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> > +
> > +	return 0;
> > +}
> > +
> > +int
> >  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
> >  				uint16_t rx_mode, uint8_t on)
> >  {
> > @@ -3627,3 +3653,136 @@ rte_eth_remove_tx_callback(uint8_t port_id,
> uint16_t queue_id,
> >  	/* Callback wasn't found. */
> >  	return -EINVAL;
> >  }
> > +
> > +int
> > +rte_eth_dev_reg_leng(uint8_t port_id) {
> > +	struct rte_eth_dev *dev;
> > +
> > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > +		return -ENODEV;
> > +	}
> > +
> > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > +		return -ENODEV;
> > +	}
> > +
> > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length, -
> ENOTSUP);
> > +	return (*dev->dev_ops->get_reg_length)(dev);
> > +}
> > +
> > +int
> > +rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info *info)
> > +{
> > +	struct rte_eth_dev *dev;
> > +
> > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > +		return -ENODEV;
> > +	}
> > +
> > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > +		return -ENODEV;
> > +	}
> > +
> > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -ENOTSUP);
> > +	return (*dev->dev_ops->get_reg)(dev, info); }
> 
> Seems that *get_reg* stuff, will be really good addition for DPDK debugging
> abilities.
> Though, I'd suggest we change it a bit to make more generic and flexible:
> 
> Introduce rte_eth_reg_read/rte_eth_reg_write(),
> or probably even better rte_pcidev_reg_read /rte_pcidev_reg_write at EAL.
> Something similar to what port_pci_reg_read/port_pci_reg_write() are
> doing now at testpmd.h.
> 
> struct rte_pcidev_reg_info {
>    const char *name;
>    uint32_t endianes, bar, offset, size, count; };
> 
> int rte_pcidev_reg_read(const struct rte_pci_device *, const struct
> rte_pcidev_reg_info *, uint64_t *reg_val);
> 
> Then:
> int rte_eth_dev_get_reg_info(port_id, const struct rte_pcidev_reg_info
> **info);
> 
> So each device would store in info a pointer to an array of it's register
> descriptions (finished by zero elem?).
> 
> Then your ethtool (or any other upper layer) can do the following to read all
> device regs:
> 

The proposed reg info structure allows future improvement to support individual register read/write.
Also, because each NIC device has a very distinguish register definition.
So, the plan is to have more comprehensive interface to support query operation (for example, register name) before introduce individual/group register access.
Points taken, the support will be in future release.

> const struct rte_eth_dev_reg_info *reg_info; struct rte_eth_dev_info
> dev_info;
> 
> rte_eth_dev_info_get(pid, &dev_info);
> rte_eth_dev_get_reg_info(port_id, &reg_info);
> 
> for (i = 0; reg_info[i].name != NULL; i++) {
>    ...
>    rte_pcidev_read_reg(dev_info. pci_dev, reg_info[i], &v);
>   ..
> }
> 
> > +
> > +int
> > +rte_eth_dev_eeprom_leng(uint8_t port_id) {
> > +	struct rte_eth_dev *dev;
> > +
> > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > +		return -ENODEV;
> > +	}
> > +
> > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > +		return -ENODEV;
> > +	}
> > +
> > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom_length, -
> ENOTSUP);
> > +	return (*dev->dev_ops->get_eeprom_length)(dev);
> > +}
> > +
> > +int
> > +rte_eth_dev_get_eeprom(uint8_t port_id, struct rte_dev_eeprom_info
> > +*info) {
> > +	struct rte_eth_dev *dev;
> > +
> > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > +		return -ENODEV;
> > +	}
> > +
> > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > +		return -ENODEV;
> > +	}
> > +
> > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -
> ENOTSUP);
> > +	return (*dev->dev_ops->get_eeprom)(dev, info); }
> > +
> > +int
> > +rte_eth_dev_set_eeprom(uint8_t port_id, struct rte_dev_eeprom_info
> > +*info) {
> > +	struct rte_eth_dev *dev;
> > +
> > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > +		return -ENODEV;
> > +	}
> > +
> > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > +		return -ENODEV;
> > +	}
> > +
> > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -
> ENOTSUP);
> > +	return (*dev->dev_ops->set_eeprom)(dev, info); }
> > +
> > +int
> > +rte_eth_dev_get_ringparam(uint8_t port_id, struct rte_dev_ring_info
> > +*info) {
> > +	struct rte_eth_dev *dev;
> > +
> > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > +		return -ENODEV;
> > +	}
> > +
> > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > +		return -ENODEV;
> > +	}
> > +
> > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> ENOTSUP);
> > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> 
> I think it will be a useful addition to the ethdev API  to have an ability to
> retrieve current RX/TX queue parameters.
> Though again, it need to be more generic, so it could be useful for non-
> ethtool upper layer too.
> So I suggest to modify it a bit.
> Something like that:
> 
> struct rte_eth_tx_queue_info {
>     struct rte_eth_txconf txconf;
>     uint32_t nb_tx_desc;
>     uint32_t nb_max_tx_desc; /*max allowable TXDs for that queue */
>     uint32_t nb_tx_free;            /* number of free TXDs at the moment of call.
> */
>     /* other tx queue data. */
> };
> 
> int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> rte_eth_tx_queue_info *qinfo)
> 
> Then, your upper layer ethtool wrapper, can implement yours
> ethtool_get_ringparam() by:
> 
>  ...
>  struct rte_eth_tx_queue_info qinfo;
>  rte_ethdev_get_tx_queue_info(port, 0, &qinfo); ring_param->tx_pending
> = qinfo.nb_tx_desc -  rte_eth_rx_queue_count(port, 0);
> 
> Or probably even:
> ring_param->tx_pending = qinfo.nb_tx_desc - qinfo.nb_tx_free;
> 
> Same for RX.
> 
For now, this descriptor ring information is used by the ethtool op. 
To make this interface simple, i.e. caller doesn't need to access other queue information.
The plan it so keep this information as it is. 
For the expansion, the ethdev can invoke the same dev op to add this information.
> > +
> > +int
> > +rte_eth_dev_set_ringparam(uint8_t port_id, struct rte_dev_ring_info
> > +*info) {
> > +	struct rte_eth_dev *dev;
> > +
> > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > +		return -ENODEV;
> > +	}
> > +
> > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > +		return -ENODEV;
> > +	}
> > +
> > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_ringparam, -
> ENOTSUP);
> > +	return (*dev->dev_ops->set_ringparam)(dev, info); }
> 
> 
> I am a bit confused, what particular value of rte_dev_ring_info will be
> allowed to change?
> You can't change number of RXD/TXD at runtime.
> You have to stop the device first, and then reconfigure particular queue(s).
> Again, you can't reset RXF/TXD without stopping RX/TX first.
> So what that function is supposed to do?
> As I can see currently, none of the PMD in your patch support it.
> 

Exactly, setting RXD/TXD will require some change on kernel (RX/TX burst) function.
As you already observe that there is no dev_op registered for set_ring.
The API is created as a placeholder for future implementation.
There are a couple of idea thrown around (such as mono- increase/decrease to avoid buffer overrun).
Any suggestion is welcome. 

> Konstantin
> 
> > diff --git a/lib/librte_ether/rte_ethdev.h
> > b/lib/librte_ether/rte_ethdev.h index 16dbe00..8f99f0d 100644
> > --- a/lib/librte_ether/rte_ethdev.h
> > +++ b/lib/librte_ether/rte_ethdev.h
> > @@ -182,6 +182,7 @@ extern "C" {
> >  #include <rte_devargs.h>
> >  #include "rte_ether.h"
> >  #include "rte_eth_ctrl.h"
> > +#include "rte_eth_dev_info.h"
> >
> >  struct rte_mbuf;
> >
> > @@ -1228,6 +1229,31 @@ typedef int (*eth_udp_tunnel_del_t)(struct
> rte_eth_dev *dev,
> >  				    struct rte_eth_udp_tunnel *tunnel_udp);
> /**< @internal
> > Delete tunneling UDP info */
> >
> > +typedef int (*eth_get_reg_length_t)(struct rte_eth_dev *dev); /**<
> > +@internal Retrieve device register count  */
> > +
> > +typedef int (*eth_get_reg_t)(struct rte_eth_dev *dev,
> > +				struct rte_dev_reg_info *info);
> > +/**< @internal Retrieve registers  */
> > +
> > +typedef int (*eth_get_eeprom_length_t)(struct rte_eth_dev *dev);
> /**<
> > +@internal Retrieve eeprom size  */
> > +
> > +typedef int (*eth_get_eeprom_t)(struct rte_eth_dev *dev,
> > +				struct rte_dev_eeprom_info *info); /**<
> @internal Retrieve eeprom
> > +data  */
> > +
> > +typedef int (*eth_set_eeprom_t)(struct rte_eth_dev *dev,
> > +				struct rte_dev_eeprom_info *info); /**<
> @internal Program eeprom
> > +data  */
> > +
> > +typedef int (*eth_get_ringparam_t)(struct rte_eth_dev *dev,
> > +				struct rte_dev_ring_info *info);
> > +/**< @internal Retrieve device ring descriptor  */
> > +
> > +typedef int (*eth_set_ringparam_t)(struct rte_eth_dev *dev,
> > +				struct rte_dev_ring_info *info);
> > +/**< @internal Set device ring descriptor  */
> >
> >  #ifdef RTE_NIC_BYPASS
> >
> > @@ -1368,6 +1394,21 @@ struct eth_dev_ops {
> >  	reta_update_t reta_update;
> >  	/** Query redirection table. */
> >  	reta_query_t reta_query;
> > +
> > +	eth_get_reg_length_t get_reg_length;
> > +	/**< Get # of registers */
> > +	eth_get_reg_t get_reg;
> > +	/**< Get registers */
> > +	eth_get_eeprom_length_t get_eeprom_length;
> > +	/**< Get eeprom length */
> > +	eth_get_eeprom_t get_eeprom;
> > +	/**< Get eeprom data */
> > +	eth_set_eeprom_t set_eeprom;
> > +	/**< Set eeprom */
> > +	eth_get_ringparam_t get_ringparam;
> > +	/**< Get descriptor ring parameters */
> > +	eth_set_ringparam_t set_ringparam;
> > +	/**< Set descriptor ring parameters */
> >    /* bypass control */
> >  #ifdef RTE_NIC_BYPASS
> >    bypass_init_t bypass_init;
> > @@ -2982,6 +3023,20 @@ int rte_eth_dev_mac_addr_add(uint8_t port,
> > struct ether_addr *mac_addr,  int
> rte_eth_dev_mac_addr_remove(uint8_t
> > port, struct ether_addr *mac_addr);
> >
> >  /**
> > + * Set the default MAC address.
> > + *
> > + * @param port
> > + *   The port identifier of the Ethernet device.
> > + * @param mac_addr
> > + *   New default MAC address.
> > + * @return
> > + *   - (0) if successful, or *mac_addr* didn't exist.
> > + *   - (-ENOTSUP) if hardware doesn't support.
> > + *   - (-ENODEV) if *port* invalid.
> > + */
> > +int rte_eth_dev_default_mac_addr_set(uint8_t port, struct ether_addr
> > +*mac_addr);
> > +
> > +/**
> >   * Update Redirection Table(RETA) of Receive Side Scaling of Ethernet
> device.
> >   *
> >   * @param port
> > @@ -3611,6 +3666,109 @@ int rte_eth_remove_rx_callback(uint8_t
> > port_id, uint16_t queue_id,  int rte_eth_remove_tx_callback(uint8_t
> port_id, uint16_t queue_id,
> >  		struct rte_eth_rxtx_callback *user_cb);
> >
> > +/**
> > + * Retrieve number of available registers for access
> > + *
> > + * @param port_id
> > + *   The port identifier of the Ethernet device.
> > + * @return
> > + *   - (>=0) number of registers if successful.
> > + *   - (-ENOTSUP) if hardware doesn't support.
> > + *   - (-ENODEV) if *port_id* invalid.
> > + *   - others depends on the specific operations implementation.
> > + */
> > +int rte_eth_dev_reg_leng(uint8_t port_id);
> > +
> > +/**
> > + * Retrieve device registers and register attributes
> > + *
> > + * @param port_id
> > + *   The port identifier of the Ethernet device.
> > + * @param info
> > + *   The template includes buffer for register data and attribute to be
> filled.
> > + * @return
> > + *   - (0) if successful.
> > + *   - (-ENOTSUP) if hardware doesn't support.
> > + *   - (-ENODEV) if *port_id* invalid.
> > + *   - others depends on the specific operations implementation.
> > + */
> > +int rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info
> > +*info);
> > +
> > +/**
> > + * Retrieve size of device EEPROM
> > + *
> > + * @param port_id
> > + *   The port identifier of the Ethernet device.
> > + * @return
> > + *   - (>=0) EEPROM size if successful.
> > + *   - (-ENOTSUP) if hardware doesn't support.
> > + *   - (-ENODEV) if *port_id* invalid.
> > + *   - others depends on the specific operations implementation.
> > + */
> > +int rte_eth_dev_eeprom_leng(uint8_t port_id);
> > +
> > +/**
> > + * Retrieve EEPROM and EEPROM attribute
> > + *
> > + * @param port_id
> > + *   The port identifier of the Ethernet device.
> > + * @param info
> > + *   The template includes buffer for return EEPROM data and
> > + *   EEPROM attributes to be filled.
> > + * @return
> > + *   - (0) if successful.
> > + *   - (-ENOTSUP) if hardware doesn't support.
> > + *   - (-ENODEV) if *port_id* invalid.
> > + *   - others depends on the specific operations implementation.
> > + */
> > +int rte_eth_dev_get_eeprom(uint8_t port_id, struct
> > +rte_dev_eeprom_info *info);
> > +
> > +/**
> > + * Program EEPROM with provided data
> > + *
> > + * @param port_id
> > + *   The port identifier of the Ethernet device.
> > + * @param info
> > + *   The template includes EEPROM data for programming and
> > + *   EEPROM attributes to be filled
> > + * @return
> > + *   - (0) if successful.
> > + *   - (-ENOTSUP) if hardware doesn't support.
> > + *   - (-ENODEV) if *port_id* invalid.
> > + *   - others depends on the specific operations implementation.
> > + */
> > +int rte_eth_dev_set_eeprom(uint8_t port_id, struct
> > +rte_dev_eeprom_info *info);
> > +
> > +/**
> > + * Retrieve device descriptor ring information
> > + *
> > + * @param port_id
> > + *   The port identifier of the Ethernet device.
> > + * @param info
> > + *   The template includes device ring descriptor usage information.
> > + * @return
> > + *   - (0) if successful.
> > + *   - (-ENOTSUP) if hardware doesn't support.
> > + *   - (-ENODEV) if *port_id* invalid.
> > + *   - others depends on the specific operations implementation.
> > + */
> > +int rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > +rte_dev_ring_info *info);
> > +
> > +/**
> > + * Set device ring descriptor parameters
> > + *
> > + * @param port_id
> > + *   The port identifier of the Ethernet device.
> > + * @param info
> > + *   The template includes descriptor ring parameters to be set.
> > + * @return
> > + *   - (0) if successful.
> > + *   - (-ENOTSUP) if hardware doesn't support.
> > + *   - (-ENODEV) if *port_id* invalid.
> > + *   - others depends on the specific operations implementation.
> > + */
> > +int rte_eth_dev_set_ringparam(uint8_t port_id, struct
> > +rte_dev_ring_info *info);
> > +
> >  #ifdef __cplusplus
> >  }
> >  #endif
> > diff --git a/lib/librte_ether/rte_ether_version.map
> > b/lib/librte_ether/rte_ether_version.map
> > index a2d25a6..1b1e092 100644
> > --- a/lib/librte_ether/rte_ether_version.map
> > +++ b/lib/librte_ether/rte_ether_version.map
> > @@ -102,6 +102,14 @@ DPDK_2.0 {
> >  	rte_eth_tx_queue_setup;
> >  	rte_eth_xstats_get;
> >  	rte_eth_xstats_reset;
> > +	rte_eth_dev_default_mac_addr_set;
> > +	rte_eth_dev_reg_leng;
> > +	rte_eth_dev_reg_info;
> > +	rte_eth_dev_eeprom_leng;
> > +	rte_eth_dev_get_eeprom;
> > +	rte_eth_dev_set_eeprom;
> > +	rte_eth_dev_get_ringparam;
> > +	rte_eth_dev_set_ringparam;
> >
> >  	local: *;
> >  };
> > --
> > 2.1.4
  

> -----Original Message-----
> From: Wang, Liang-min
> Sent: Thursday, June 11, 2015 1:58 PM
> To: Ananyev, Konstantin; dev@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access device info
> 
> Hi Konstantin,
> 
> > -----Original Message-----
> > From: Ananyev, Konstantin
> > Sent: Thursday, June 11, 2015 8:26 AM
> > To: Wang, Liang-min; dev@dpdk.org
> > Cc: Wang, Liang-min
> > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> > device info
> >
> > Hi Larry,
> >
> > > -----Original Message-----
> > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Liang-Min Larry
> > > Wang
> > > Sent: Wednesday, June 10, 2015 4:10 PM
> > > To: dev@dpdk.org
> > > Cc: Wang, Liang-min
> > > Subject: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> > > device info
> > >
> > > add new apis:
> > > - rte_eth_dev_default_mac_addr_set
> > > - rte_eth_dev_reg_leng
> > > - rte_eth_dev_reg_info
> > > - rte_eth_dev_eeprom_leng
> > > - rte_eth_dev_get_eeprom
> > > - rte_eth_dev_set_eeprom
> > > - rte_eth_dev_get_ringparam
> > > - rte_eth_dev_set_ringparam
> > >
> > > to enable reading device parameters (mac-addr, register, eeprom, ring)
> > > based upon ethtool alike data parameter specification.
> > >
> > > Signed-off-by: Liang-Min Larry Wang <liang-min.wang@intel.com>
> > > ---
> > >  lib/librte_ether/Makefile              |   1 +
> > >  lib/librte_ether/rte_eth_dev_info.h    |  80 +++++++++++++++++
> > >  lib/librte_ether/rte_ethdev.c          | 159
> > +++++++++++++++++++++++++++++++++
> > >  lib/librte_ether/rte_ethdev.h          | 158
> > ++++++++++++++++++++++++++++++++
> > >  lib/librte_ether/rte_ether_version.map |   8 ++
> > >  5 files changed, 406 insertions(+)
> > >  create mode 100644 lib/librte_ether/rte_eth_dev_info.h
> > >
> > > diff --git a/lib/librte_ether/Makefile b/lib/librte_ether/Makefile
> > > index c0e5768..05209e9 100644
> > > --- a/lib/librte_ether/Makefile
> > > +++ b/lib/librte_ether/Makefile
> > > @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c  SYMLINK-y-include +=
> > > rte_ether.h  SYMLINK-y-include += rte_ethdev.h  SYMLINK-y-include +=
> > > rte_eth_ctrl.h
> > > +SYMLINK-y-include += rte_eth_dev_info.h
> > >
> > >  # this lib depends upon:
> > >  DEPDIRS-y += lib/librte_eal lib/librte_mempool lib/librte_ring
> > > lib/librte_mbuf diff --git a/lib/librte_ether/rte_eth_dev_info.h
> > > b/lib/librte_ether/rte_eth_dev_info.h
> > > new file mode 100644
> > > index 0000000..002c4b5
> > > --- /dev/null
> > > +++ b/lib/librte_ether/rte_eth_dev_info.h
> > > @@ -0,0 +1,80 @@
> > > +/*-
> > > + *   BSD LICENSE
> > > + *
> > > + *   Copyright(c) 2015 Intel Corporation. All rights reserved.
> > > + *   All rights reserved.
> > > + *
> > > + *   Redistribution and use in source and binary forms, with or without
> > > + *   modification, are permitted provided that the following conditions
> > > + *   are met:
> > > + *
> > > + *     * Redistributions of source code must retain the above copyright
> > > + *       notice, this list of conditions and the following disclaimer.
> > > + *     * Redistributions in binary form must reproduce the above copyright
> > > + *       notice, this list of conditions and the following disclaimer in
> > > + *       the documentation and/or other materials provided with the
> > > + *       distribution.
> > > + *     * Neither the name of Intel Corporation nor the names of its
> > > + *       contributors may be used to endorse or promote products derived
> > > + *       from this software without specific prior written permission.
> > > + *
> > > + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> > CONTRIBUTORS
> > > + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT
> > NOT
> > > + *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
> > FITNESS FOR
> > > + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
> > COPYRIGHT
> > > + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
> > INCIDENTAL,
> > > + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
> > BUT NOT
> > > + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
> > LOSS OF USE,
> > > + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
> > AND ON ANY
> > > + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
> > TORT
> > > + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
> > OF THE USE
> > > + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
> > DAMAGE.
> > > + */
> > > +
> > > +#ifndef _RTE_ETH_DEV_INFO_H_
> > > +#define _RTE_ETH_DEV_INFO_H_
> > > +
> > > +
> > > +/*
> > > + * Placeholder for accessing device registers  */ struct
> > > +rte_dev_reg_info {
> > > +	void *buf; /**< Buffer for register */
> > > +	uint32_t offset; /**< Offset for 1st register to fetch */
> > > +	uint32_t leng; /**< Number of registers to fetch */
> > > +	uint32_t version; /**< Device version */ };
> > > +
> > > +/*
> > > + * Placeholder for accessing device eeprom  */ struct
> > > +rte_dev_eeprom_info {
> > > +	void *buf; /**< Buffer for eeprom */
> > > +	uint32_t offset; /**< Offset for 1st eeprom location to access */
> > > +	uint32_t leng; /**< Length of eeprom region to access */
> > > +	uint32_t magic; /**< Device ID */
> > > +};
> > > +
> > > +/*
> > > + * Placeholder for accessing device ring parameters  */ struct
> > > +rte_dev_ring_info {
> > > +	uint32_t rx_pending; /**< Number of outstanding Rx ring */
> > > +	uint32_t tx_pending; /**< Number of outstanding Tx ring */
> > > +	uint32_t rx_max_pending; /**< Maximum number of outstanding Rx
> > ring */
> > > +	uint32_t tx_max_pending; /**< Maximum number of outstanding Tx
> > ring
> > > +*/ };
> > > +
> > > +/*
> > > + * A data structure captures information as defined in struct
> > > +ifla_vf_info
> > > + * for user-space api
> > > + */
> > > +struct rte_dev_vf_info {
> > > +	uint32_t vf;
> > > +	uint8_t mac[ETHER_ADDR_LEN];
> > > +	uint32_t vlan;
> > > +	uint32_t tx_rate;
> > > +	uint32_t spoofchk;
> > > +};
> >
> >
> > Wonder what that structure is for?
> > I can't see it used in any function below?
> >
> 
> Good catch, this is designed for other ethtool ops that I did not include in this release, I will remove this from next fix.
> 
> > > +
> > > +#endif /* _RTE_ETH_DEV_INFO_H_ */
> > > diff --git a/lib/librte_ether/rte_ethdev.c
> > > b/lib/librte_ether/rte_ethdev.c index 5a94654..186e85c 100644
> > > --- a/lib/librte_ether/rte_ethdev.c
> > > +++ b/lib/librte_ether/rte_ethdev.c
> > > @@ -2751,6 +2751,32 @@ rte_eth_dev_mac_addr_remove(uint8_t
> > port_id,
> > > struct ether_addr *addr)  }
> > >
> > >  int
> > > +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct ether_addr
> > > +*addr) {
> > > +	struct rte_eth_dev *dev;
> > > +	const int index = 0;
> > > +	const uint32_t pool = 0;
> > > +
> > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	dev = &rte_eth_devices[port_id];
> > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_remove, -
> > ENOTSUP);
> > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -
> > ENOTSUP);
> > > +
> > > +	/* Update NIC default MAC address*/
> > > +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> > > +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> > > +
> > > +	/* Update default address in NIC data structure */
> > > +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> > > +
> > > +	return 0;
> > > +}
> > > +
> > > +int
> > >  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
> > >  				uint16_t rx_mode, uint8_t on)
> > >  {
> > > @@ -3627,3 +3653,136 @@ rte_eth_remove_tx_callback(uint8_t port_id,
> > uint16_t queue_id,
> > >  	/* Callback wasn't found. */
> > >  	return -EINVAL;
> > >  }
> > > +
> > > +int
> > > +rte_eth_dev_reg_leng(uint8_t port_id) {
> > > +	struct rte_eth_dev *dev;
> > > +
> > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length, -
> > ENOTSUP);
> > > +	return (*dev->dev_ops->get_reg_length)(dev);
> > > +}
> > > +
> > > +int
> > > +rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info *info)
> > > +{
> > > +	struct rte_eth_dev *dev;
> > > +
> > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -ENOTSUP);
> > > +	return (*dev->dev_ops->get_reg)(dev, info); }
> >
> > Seems that *get_reg* stuff, will be really good addition for DPDK debugging
> > abilities.
> > Though, I'd suggest we change it a bit to make more generic and flexible:
> >
> > Introduce rte_eth_reg_read/rte_eth_reg_write(),
> > or probably even better rte_pcidev_reg_read /rte_pcidev_reg_write at EAL.
> > Something similar to what port_pci_reg_read/port_pci_reg_write() are
> > doing now at testpmd.h.
> >
> > struct rte_pcidev_reg_info {
> >    const char *name;
> >    uint32_t endianes, bar, offset, size, count; };
> >
> > int rte_pcidev_reg_read(const struct rte_pci_device *, const struct
> > rte_pcidev_reg_info *, uint64_t *reg_val);
> >
> > Then:
> > int rte_eth_dev_get_reg_info(port_id, const struct rte_pcidev_reg_info
> > **info);
> >
> > So each device would store in info a pointer to an array of it's register
> > descriptions (finished by zero elem?).
> >
> > Then your ethtool (or any other upper layer) can do the following to read all
> > device regs:
> >
> 
> The proposed reg info structure allows future improvement to support individual register read/write.
> Also, because each NIC device has a very distinguish register definition.
> So, the plan is to have more comprehensive interface to support query operation (for example, register name) before introduce
> individual/group register access.
> Points taken, the support will be in future release.

Sorry, didn't get you.
So you are ok to make these changes in next patch version?

> 
> > const struct rte_eth_dev_reg_info *reg_info; struct rte_eth_dev_info
> > dev_info;
> >
> > rte_eth_dev_info_get(pid, &dev_info);
> > rte_eth_dev_get_reg_info(port_id, &reg_info);
> >
> > for (i = 0; reg_info[i].name != NULL; i++) {
> >    ...
> >    rte_pcidev_read_reg(dev_info. pci_dev, reg_info[i], &v);
> >   ..
> > }
> >
> > > +
> > > +int
> > > +rte_eth_dev_eeprom_leng(uint8_t port_id) {
> > > +	struct rte_eth_dev *dev;
> > > +
> > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom_length, -
> > ENOTSUP);
> > > +	return (*dev->dev_ops->get_eeprom_length)(dev);
> > > +}
> > > +
> > > +int
> > > +rte_eth_dev_get_eeprom(uint8_t port_id, struct rte_dev_eeprom_info
> > > +*info) {
> > > +	struct rte_eth_dev *dev;
> > > +
> > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -
> > ENOTSUP);
> > > +	return (*dev->dev_ops->get_eeprom)(dev, info); }
> > > +
> > > +int
> > > +rte_eth_dev_set_eeprom(uint8_t port_id, struct rte_dev_eeprom_info
> > > +*info) {
> > > +	struct rte_eth_dev *dev;
> > > +
> > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -
> > ENOTSUP);
> > > +	return (*dev->dev_ops->set_eeprom)(dev, info); }
> > > +
> > > +int
> > > +rte_eth_dev_get_ringparam(uint8_t port_id, struct rte_dev_ring_info
> > > +*info) {
> > > +	struct rte_eth_dev *dev;
> > > +
> > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> > ENOTSUP);
> > > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> >
> > I think it will be a useful addition to the ethdev API  to have an ability to
> > retrieve current RX/TX queue parameters.
> > Though again, it need to be more generic, so it could be useful for non-
> > ethtool upper layer too.
> > So I suggest to modify it a bit.
> > Something like that:
> >
> > struct rte_eth_tx_queue_info {
> >     struct rte_eth_txconf txconf;
> >     uint32_t nb_tx_desc;
> >     uint32_t nb_max_tx_desc; /*max allowable TXDs for that queue */
> >     uint32_t nb_tx_free;            /* number of free TXDs at the moment of call.
> > */
> >     /* other tx queue data. */
> > };
> >
> > int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> > rte_eth_tx_queue_info *qinfo)
> >
> > Then, your upper layer ethtool wrapper, can implement yours
> > ethtool_get_ringparam() by:
> >
> >  ...
> >  struct rte_eth_tx_queue_info qinfo;
> >  rte_ethdev_get_tx_queue_info(port, 0, &qinfo); ring_param->tx_pending
> > = qinfo.nb_tx_desc -  rte_eth_rx_queue_count(port, 0);
> >
> > Or probably even:
> > ring_param->tx_pending = qinfo.nb_tx_desc - qinfo.nb_tx_free;
> >
> > Same for RX.
> >
> For now, this descriptor ring information is used by the ethtool op.
> To make this interface simple, i.e. caller doesn't need to access other queue information.

I just repeat what I said to you in off-line conversation:
ethdev API is not equal ethtool API.
It is ok to add  a new function/structure to ethdev if it really needed, but we should do mechanical one to one copy.
It is much better to add  a function/structure that would be more generic, and suit other users, not only ethtool.
There is no point to have dozen functions in rte_ethdev API providing similar information.
BTW, I don't see how API I proposed is much more  complex, then yours one.

Konstantin


> The plan it so keep this information as it is.
> For the expansion, the ethdev can invoke the same dev op to add this information.
> > > +
> > > +int
> > > +rte_eth_dev_set_ringparam(uint8_t port_id, struct rte_dev_ring_info
> > > +*info) {
> > > +	struct rte_eth_dev *dev;
> > > +
> > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_ringparam, -
> > ENOTSUP);
> > > +	return (*dev->dev_ops->set_ringparam)(dev, info); }
> >
> >
> > I am a bit confused, what particular value of rte_dev_ring_info will be
> > allowed to change?
> > You can't change number of RXD/TXD at runtime.
> > You have to stop the device first, and then reconfigure particular queue(s).
> > Again, you can't reset RXF/TXD without stopping RX/TX first.
> > So what that function is supposed to do?
> > As I can see currently, none of the PMD in your patch support it.
> >
> 
> Exactly, setting RXD/TXD will require some change on kernel (RX/TX burst) function.
> As you already observe that there is no dev_op registered for set_ring.
> The API is created as a placeholder for future implementation.
> There are a couple of idea thrown around (such as mono- increase/decrease to avoid buffer overrun).
> Any suggestion is welcome.
> 
> > Konstantin
> >
> > > diff --git a/lib/librte_ether/rte_ethdev.h
> > > b/lib/librte_ether/rte_ethdev.h index 16dbe00..8f99f0d 100644
> > > --- a/lib/librte_ether/rte_ethdev.h
> > > +++ b/lib/librte_ether/rte_ethdev.h
> > > @@ -182,6 +182,7 @@ extern "C" {
> > >  #include <rte_devargs.h>
> > >  #include "rte_ether.h"
> > >  #include "rte_eth_ctrl.h"
> > > +#include "rte_eth_dev_info.h"
> > >
> > >  struct rte_mbuf;
> > >
> > > @@ -1228,6 +1229,31 @@ typedef int (*eth_udp_tunnel_del_t)(struct
> > rte_eth_dev *dev,
> > >  				    struct rte_eth_udp_tunnel *tunnel_udp);
> > /**< @internal
> > > Delete tunneling UDP info */
> > >
> > > +typedef int (*eth_get_reg_length_t)(struct rte_eth_dev *dev); /**<
> > > +@internal Retrieve device register count  */
> > > +
> > > +typedef int (*eth_get_reg_t)(struct rte_eth_dev *dev,
> > > +				struct rte_dev_reg_info *info);
> > > +/**< @internal Retrieve registers  */
> > > +
> > > +typedef int (*eth_get_eeprom_length_t)(struct rte_eth_dev *dev);
> > /**<
> > > +@internal Retrieve eeprom size  */
> > > +
> > > +typedef int (*eth_get_eeprom_t)(struct rte_eth_dev *dev,
> > > +				struct rte_dev_eeprom_info *info); /**<
> > @internal Retrieve eeprom
> > > +data  */
> > > +
> > > +typedef int (*eth_set_eeprom_t)(struct rte_eth_dev *dev,
> > > +				struct rte_dev_eeprom_info *info); /**<
> > @internal Program eeprom
> > > +data  */
> > > +
> > > +typedef int (*eth_get_ringparam_t)(struct rte_eth_dev *dev,
> > > +				struct rte_dev_ring_info *info);
> > > +/**< @internal Retrieve device ring descriptor  */
> > > +
> > > +typedef int (*eth_set_ringparam_t)(struct rte_eth_dev *dev,
> > > +				struct rte_dev_ring_info *info);
> > > +/**< @internal Set device ring descriptor  */
> > >
> > >  #ifdef RTE_NIC_BYPASS
> > >
> > > @@ -1368,6 +1394,21 @@ struct eth_dev_ops {
> > >  	reta_update_t reta_update;
> > >  	/** Query redirection table. */
> > >  	reta_query_t reta_query;
> > > +
> > > +	eth_get_reg_length_t get_reg_length;
> > > +	/**< Get # of registers */
> > > +	eth_get_reg_t get_reg;
> > > +	/**< Get registers */
> > > +	eth_get_eeprom_length_t get_eeprom_length;
> > > +	/**< Get eeprom length */
> > > +	eth_get_eeprom_t get_eeprom;
> > > +	/**< Get eeprom data */
> > > +	eth_set_eeprom_t set_eeprom;
> > > +	/**< Set eeprom */
> > > +	eth_get_ringparam_t get_ringparam;
> > > +	/**< Get descriptor ring parameters */
> > > +	eth_set_ringparam_t set_ringparam;
> > > +	/**< Set descriptor ring parameters */
> > >    /* bypass control */
> > >  #ifdef RTE_NIC_BYPASS
> > >    bypass_init_t bypass_init;
> > > @@ -2982,6 +3023,20 @@ int rte_eth_dev_mac_addr_add(uint8_t port,
> > > struct ether_addr *mac_addr,  int
> > rte_eth_dev_mac_addr_remove(uint8_t
> > > port, struct ether_addr *mac_addr);
> > >
> > >  /**
> > > + * Set the default MAC address.
> > > + *
> > > + * @param port
> > > + *   The port identifier of the Ethernet device.
> > > + * @param mac_addr
> > > + *   New default MAC address.
> > > + * @return
> > > + *   - (0) if successful, or *mac_addr* didn't exist.
> > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > + *   - (-ENODEV) if *port* invalid.
> > > + */
> > > +int rte_eth_dev_default_mac_addr_set(uint8_t port, struct ether_addr
> > > +*mac_addr);
> > > +
> > > +/**
> > >   * Update Redirection Table(RETA) of Receive Side Scaling of Ethernet
> > device.
> > >   *
> > >   * @param port
> > > @@ -3611,6 +3666,109 @@ int rte_eth_remove_rx_callback(uint8_t
> > > port_id, uint16_t queue_id,  int rte_eth_remove_tx_callback(uint8_t
> > port_id, uint16_t queue_id,
> > >  		struct rte_eth_rxtx_callback *user_cb);
> > >
> > > +/**
> > > + * Retrieve number of available registers for access
> > > + *
> > > + * @param port_id
> > > + *   The port identifier of the Ethernet device.
> > > + * @return
> > > + *   - (>=0) number of registers if successful.
> > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > + *   - (-ENODEV) if *port_id* invalid.
> > > + *   - others depends on the specific operations implementation.
> > > + */
> > > +int rte_eth_dev_reg_leng(uint8_t port_id);
> > > +
> > > +/**
> > > + * Retrieve device registers and register attributes
> > > + *
> > > + * @param port_id
> > > + *   The port identifier of the Ethernet device.
> > > + * @param info
> > > + *   The template includes buffer for register data and attribute to be
> > filled.
> > > + * @return
> > > + *   - (0) if successful.
> > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > + *   - (-ENODEV) if *port_id* invalid.
> > > + *   - others depends on the specific operations implementation.
> > > + */
> > > +int rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info
> > > +*info);
> > > +
> > > +/**
> > > + * Retrieve size of device EEPROM
> > > + *
> > > + * @param port_id
> > > + *   The port identifier of the Ethernet device.
> > > + * @return
> > > + *   - (>=0) EEPROM size if successful.
> > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > + *   - (-ENODEV) if *port_id* invalid.
> > > + *   - others depends on the specific operations implementation.
> > > + */
> > > +int rte_eth_dev_eeprom_leng(uint8_t port_id);
> > > +
> > > +/**
> > > + * Retrieve EEPROM and EEPROM attribute
> > > + *
> > > + * @param port_id
> > > + *   The port identifier of the Ethernet device.
> > > + * @param info
> > > + *   The template includes buffer for return EEPROM data and
> > > + *   EEPROM attributes to be filled.
> > > + * @return
> > > + *   - (0) if successful.
> > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > + *   - (-ENODEV) if *port_id* invalid.
> > > + *   - others depends on the specific operations implementation.
> > > + */
> > > +int rte_eth_dev_get_eeprom(uint8_t port_id, struct
> > > +rte_dev_eeprom_info *info);
> > > +
> > > +/**
> > > + * Program EEPROM with provided data
> > > + *
> > > + * @param port_id
> > > + *   The port identifier of the Ethernet device.
> > > + * @param info
> > > + *   The template includes EEPROM data for programming and
> > > + *   EEPROM attributes to be filled
> > > + * @return
> > > + *   - (0) if successful.
> > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > + *   - (-ENODEV) if *port_id* invalid.
> > > + *   - others depends on the specific operations implementation.
> > > + */
> > > +int rte_eth_dev_set_eeprom(uint8_t port_id, struct
> > > +rte_dev_eeprom_info *info);
> > > +
> > > +/**
> > > + * Retrieve device descriptor ring information
> > > + *
> > > + * @param port_id
> > > + *   The port identifier of the Ethernet device.
> > > + * @param info
> > > + *   The template includes device ring descriptor usage information.
> > > + * @return
> > > + *   - (0) if successful.
> > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > + *   - (-ENODEV) if *port_id* invalid.
> > > + *   - others depends on the specific operations implementation.
> > > + */
> > > +int rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > +rte_dev_ring_info *info);
> > > +
> > > +/**
> > > + * Set device ring descriptor parameters
> > > + *
> > > + * @param port_id
> > > + *   The port identifier of the Ethernet device.
> > > + * @param info
> > > + *   The template includes descriptor ring parameters to be set.
> > > + * @return
> > > + *   - (0) if successful.
> > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > + *   - (-ENODEV) if *port_id* invalid.
> > > + *   - others depends on the specific operations implementation.
> > > + */
> > > +int rte_eth_dev_set_ringparam(uint8_t port_id, struct
> > > +rte_dev_ring_info *info);
> > > +
> > >  #ifdef __cplusplus
> > >  }
> > >  #endif
> > > diff --git a/lib/librte_ether/rte_ether_version.map
> > > b/lib/librte_ether/rte_ether_version.map
> > > index a2d25a6..1b1e092 100644
> > > --- a/lib/librte_ether/rte_ether_version.map
> > > +++ b/lib/librte_ether/rte_ether_version.map
> > > @@ -102,6 +102,14 @@ DPDK_2.0 {
> > >  	rte_eth_tx_queue_setup;
> > >  	rte_eth_xstats_get;
> > >  	rte_eth_xstats_reset;
> > > +	rte_eth_dev_default_mac_addr_set;
> > > +	rte_eth_dev_reg_leng;
> > > +	rte_eth_dev_reg_info;
> > > +	rte_eth_dev_eeprom_leng;
> > > +	rte_eth_dev_get_eeprom;
> > > +	rte_eth_dev_set_eeprom;
> > > +	rte_eth_dev_get_ringparam;
> > > +	rte_eth_dev_set_ringparam;
> > >
> > >  	local: *;
> > >  };
> > > --
> > > 2.1.4
  

> > > +
> > > +int
> > > +rte_eth_dev_set_ringparam(uint8_t port_id, struct rte_dev_ring_info
> > > +*info) {
> > > +	struct rte_eth_dev *dev;
> > > +
> > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_ringparam, -
> > ENOTSUP);
> > > +	return (*dev->dev_ops->set_ringparam)(dev, info); }
> >
> >
> > I am a bit confused, what particular value of rte_dev_ring_info will be
> > allowed to change?
> > You can't change number of RXD/TXD at runtime.
> > You have to stop the device first, and then reconfigure particular queue(s).
> > Again, you can't reset RXF/TXD without stopping RX/TX first.
> > So what that function is supposed to do?
> > As I can see currently, none of the PMD in your patch support it.
> >
> 
> Exactly, setting RXD/TXD will require some change on kernel (RX/TX burst) function.
> As you already observe that there is no dev_op registered for set_ring.
> The API is created as a placeholder for future implementation.
> There are a couple of idea thrown around (such as mono- increase/decrease to avoid buffer overrun).
> Any suggestion is welcome.

If you don't plan to implement it in current patch-set, then it is better to remove it for now.
About suggestions - as I said, I don't think it is doable without stopping (and reconfiguring) the queue.
Again not sure, why do you want to do that run-tme.
Konstantin
  

> -----Original Message-----
> From: Ananyev, Konstantin
> Sent: Thursday, June 11, 2015 9:14 AM
> To: Wang, Liang-min; dev@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> device info
> 
> 
> > > > +
> > > > +int
> > > > +rte_eth_dev_set_ringparam(uint8_t port_id, struct
> > > > +rte_dev_ring_info
> > > > +*info) {
> > > > +	struct rte_eth_dev *dev;
> > > > +
> > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_ringparam, -
> > > ENOTSUP);
> > > > +	return (*dev->dev_ops->set_ringparam)(dev, info); }
> > >
> > >
> > > I am a bit confused, what particular value of rte_dev_ring_info will
> > > be allowed to change?
> > > You can't change number of RXD/TXD at runtime.
> > > You have to stop the device first, and then reconfigure particular
> queue(s).
> > > Again, you can't reset RXF/TXD without stopping RX/TX first.
> > > So what that function is supposed to do?
> > > As I can see currently, none of the PMD in your patch support it.
> > >
> >
> > Exactly, setting RXD/TXD will require some change on kernel (RX/TX burst)
> function.
> > As you already observe that there is no dev_op registered for set_ring.
> > The API is created as a placeholder for future implementation.
> > There are a couple of idea thrown around (such as mono-
> increase/decrease to avoid buffer overrun).
> > Any suggestion is welcome.
> 
> If you don't plan to implement it in current patch-set, then it is better to
> remove it for now.
> About suggestions - as I said, I don't think it is doable without stopping (and
> reconfiguring) the queue.
> Again not sure, why do you want to do that run-tme.
> Konstantin

That's a very valid point. To support run-time ring descriptor size adjustment, we need to stop traffic in some way.
  

> -----Original Message-----
> From: Ananyev, Konstantin
> Sent: Thursday, June 11, 2015 9:07 AM
> To: Wang, Liang-min; dev@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> device info
> 
> 
> 
> > -----Original Message-----
> > From: Wang, Liang-min
> > Sent: Thursday, June 11, 2015 1:58 PM
> > To: Ananyev, Konstantin; dev@dpdk.org
> > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > access device info
> >
> > Hi Konstantin,
> >
> > > -----Original Message-----
> > > From: Ananyev, Konstantin
> > > Sent: Thursday, June 11, 2015 8:26 AM
> > > To: Wang, Liang-min; dev@dpdk.org
> > > Cc: Wang, Liang-min
> > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > access device info
> > >
> > > Hi Larry,
> > >
> > > > -----Original Message-----
> > > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Liang-Min
> > > > Larry Wang
> > > > Sent: Wednesday, June 10, 2015 4:10 PM
> > > > To: dev@dpdk.org
> > > > Cc: Wang, Liang-min
> > > > Subject: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > access device info
> > > >
> > > > add new apis:
> > > > - rte_eth_dev_default_mac_addr_set
> > > > - rte_eth_dev_reg_leng
> > > > - rte_eth_dev_reg_info
> > > > - rte_eth_dev_eeprom_leng
> > > > - rte_eth_dev_get_eeprom
> > > > - rte_eth_dev_set_eeprom
> > > > - rte_eth_dev_get_ringparam
> > > > - rte_eth_dev_set_ringparam
> > > >
> > > > to enable reading device parameters (mac-addr, register, eeprom,
> > > > ring) based upon ethtool alike data parameter specification.
> > > >
> > > > Signed-off-by: Liang-Min Larry Wang <liang-min.wang@intel.com>
> > > > ---
> > > >  lib/librte_ether/Makefile              |   1 +
> > > >  lib/librte_ether/rte_eth_dev_info.h    |  80 +++++++++++++++++
> > > >  lib/librte_ether/rte_ethdev.c          | 159
> > > +++++++++++++++++++++++++++++++++
> > > >  lib/librte_ether/rte_ethdev.h          | 158
> > > ++++++++++++++++++++++++++++++++
> > > >  lib/librte_ether/rte_ether_version.map |   8 ++
> > > >  5 files changed, 406 insertions(+)  create mode 100644
> > > > lib/librte_ether/rte_eth_dev_info.h
> > > >
> > > > diff --git a/lib/librte_ether/Makefile b/lib/librte_ether/Makefile
> > > > index c0e5768..05209e9 100644
> > > > --- a/lib/librte_ether/Makefile
> > > > +++ b/lib/librte_ether/Makefile
> > > > @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c  SYMLINK-y-include +=
> > > > rte_ether.h  SYMLINK-y-include += rte_ethdev.h  SYMLINK-y-include
> > > > += rte_eth_ctrl.h
> > > > +SYMLINK-y-include += rte_eth_dev_info.h
> > > >
> > > >  # this lib depends upon:
> > > >  DEPDIRS-y += lib/librte_eal lib/librte_mempool lib/librte_ring
> > > > lib/librte_mbuf diff --git a/lib/librte_ether/rte_eth_dev_info.h
> > > > b/lib/librte_ether/rte_eth_dev_info.h
> > > > new file mode 100644
> > > > index 0000000..002c4b5
> > > > --- /dev/null
> > > > +++ b/lib/librte_ether/rte_eth_dev_info.h
> > > > @@ -0,0 +1,80 @@
> > > > +/*-
> > > > + *   BSD LICENSE
> > > > + *
> > > > + *   Copyright(c) 2015 Intel Corporation. All rights reserved.
> > > > + *   All rights reserved.
> > > > + *
> > > > + *   Redistribution and use in source and binary forms, with or without
> > > > + *   modification, are permitted provided that the following conditions
> > > > + *   are met:
> > > > + *
> > > > + *     * Redistributions of source code must retain the above copyright
> > > > + *       notice, this list of conditions and the following disclaimer.
> > > > + *     * Redistributions in binary form must reproduce the above
> copyright
> > > > + *       notice, this list of conditions and the following disclaimer in
> > > > + *       the documentation and/or other materials provided with the
> > > > + *       distribution.
> > > > + *     * Neither the name of Intel Corporation nor the names of its
> > > > + *       contributors may be used to endorse or promote products
> derived
> > > > + *       from this software without specific prior written permission.
> > > > + *
> > > > + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> > > CONTRIBUTORS
> > > > + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING,
> BUT
> > > NOT
> > > > + *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
> > > FITNESS FOR
> > > > + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
> > > COPYRIGHT
> > > > + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
> > > INCIDENTAL,
> > > > + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
> (INCLUDING,
> > > BUT NOT
> > > > + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
> > > LOSS OF USE,
> > > > + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
> CAUSED
> > > AND ON ANY
> > > > + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
> OR
> > > TORT
> > > > + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
> OUT
> > > OF THE USE
> > > > + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
> > > DAMAGE.
> > > > + */
> > > > +
> > > > +#ifndef _RTE_ETH_DEV_INFO_H_
> > > > +#define _RTE_ETH_DEV_INFO_H_
> > > > +
> > > > +
> > > > +/*
> > > > + * Placeholder for accessing device registers  */ struct
> > > > +rte_dev_reg_info {
> > > > +	void *buf; /**< Buffer for register */
> > > > +	uint32_t offset; /**< Offset for 1st register to fetch */
> > > > +	uint32_t leng; /**< Number of registers to fetch */
> > > > +	uint32_t version; /**< Device version */ };
> > > > +
> > > > +/*
> > > > + * Placeholder for accessing device eeprom  */ struct
> > > > +rte_dev_eeprom_info {
> > > > +	void *buf; /**< Buffer for eeprom */
> > > > +	uint32_t offset; /**< Offset for 1st eeprom location to access */
> > > > +	uint32_t leng; /**< Length of eeprom region to access */
> > > > +	uint32_t magic; /**< Device ID */ };
> > > > +
> > > > +/*
> > > > + * Placeholder for accessing device ring parameters  */ struct
> > > > +rte_dev_ring_info {
> > > > +	uint32_t rx_pending; /**< Number of outstanding Rx ring */
> > > > +	uint32_t tx_pending; /**< Number of outstanding Tx ring */
> > > > +	uint32_t rx_max_pending; /**< Maximum number of outstanding Rx
> > > ring */
> > > > +	uint32_t tx_max_pending; /**< Maximum number of outstanding Tx
> > > ring
> > > > +*/ };
> > > > +
> > > > +/*
> > > > + * A data structure captures information as defined in struct
> > > > +ifla_vf_info
> > > > + * for user-space api
> > > > + */
> > > > +struct rte_dev_vf_info {
> > > > +	uint32_t vf;
> > > > +	uint8_t mac[ETHER_ADDR_LEN];
> > > > +	uint32_t vlan;
> > > > +	uint32_t tx_rate;
> > > > +	uint32_t spoofchk;
> > > > +};
> > >
> > >
> > > Wonder what that structure is for?
> > > I can't see it used in any function below?
> > >
> >
> > Good catch, this is designed for other ethtool ops that I did not include in
> this release, I will remove this from next fix.
> >
> > > > +
> > > > +#endif /* _RTE_ETH_DEV_INFO_H_ */
> > > > diff --git a/lib/librte_ether/rte_ethdev.c
> > > > b/lib/librte_ether/rte_ethdev.c index 5a94654..186e85c 100644
> > > > --- a/lib/librte_ether/rte_ethdev.c
> > > > +++ b/lib/librte_ether/rte_ethdev.c
> > > > @@ -2751,6 +2751,32 @@ rte_eth_dev_mac_addr_remove(uint8_t
> > > port_id,
> > > > struct ether_addr *addr)  }
> > > >
> > > >  int
> > > > +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct
> > > > +ether_addr
> > > > +*addr) {
> > > > +	struct rte_eth_dev *dev;
> > > > +	const int index = 0;
> > > > +	const uint32_t pool = 0;
> > > > +
> > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	dev = &rte_eth_devices[port_id];
> > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_remove, -
> > > ENOTSUP);
> > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -
> > > ENOTSUP);
> > > > +
> > > > +	/* Update NIC default MAC address*/
> > > > +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> > > > +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> > > > +
> > > > +	/* Update default address in NIC data structure */
> > > > +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> > > > +
> > > > +	return 0;
> > > > +}
> > > > +
> > > > +int
> > > >  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
> > > >  				uint16_t rx_mode, uint8_t on)  { @@ -3627,3
> +3653,136 @@
> > > > rte_eth_remove_tx_callback(uint8_t port_id,
> > > uint16_t queue_id,
> > > >  	/* Callback wasn't found. */
> > > >  	return -EINVAL;
> > > >  }
> > > > +
> > > > +int
> > > > +rte_eth_dev_reg_leng(uint8_t port_id) {
> > > > +	struct rte_eth_dev *dev;
> > > > +
> > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length, -
> > > ENOTSUP);
> > > > +	return (*dev->dev_ops->get_reg_length)(dev);
> > > > +}
> > > > +
> > > > +int
> > > > +rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info
> > > > +*info) {
> > > > +	struct rte_eth_dev *dev;
> > > > +
> > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -ENOTSUP);
> > > > +	return (*dev->dev_ops->get_reg)(dev, info); }
> > >
> > > Seems that *get_reg* stuff, will be really good addition for DPDK
> > > debugging abilities.
> > > Though, I'd suggest we change it a bit to make more generic and flexible:
> > >
> > > Introduce rte_eth_reg_read/rte_eth_reg_write(),
> > > or probably even better rte_pcidev_reg_read /rte_pcidev_reg_write at
> EAL.
> > > Something similar to what port_pci_reg_read/port_pci_reg_write() are
> > > doing now at testpmd.h.
> > >
> > > struct rte_pcidev_reg_info {
> > >    const char *name;
> > >    uint32_t endianes, bar, offset, size, count; };
> > >
> > > int rte_pcidev_reg_read(const struct rte_pci_device *, const struct
> > > rte_pcidev_reg_info *, uint64_t *reg_val);
> > >
> > > Then:
> > > int rte_eth_dev_get_reg_info(port_id, const struct
> > > rte_pcidev_reg_info **info);
> > >
> > > So each device would store in info a pointer to an array of it's
> > > register descriptions (finished by zero elem?).
> > >
> > > Then your ethtool (or any other upper layer) can do the following to
> > > read all device regs:
> > >
> >
> > The proposed reg info structure allows future improvement to support
> individual register read/write.
> > Also, because each NIC device has a very distinguish register definition.
> > So, the plan is to have more comprehensive interface to support query
> > operation (for example, register name) before introduce individual/group
> register access.
> > Points taken, the support will be in future release.
> 
> Sorry, didn't get you.
> So you are ok to make these changes in next patch version?
> 
I would like to get a consensus from dpdk community on how to provide register information.
Currently, it's designed for debug dumping. The register information is very hardware dependent.
Need to consider current supported NIC device and future devices for DPDK, so we won't make it a bulky interface.
> >
> > > const struct rte_eth_dev_reg_info *reg_info; struct rte_eth_dev_info
> > > dev_info;
> > >
> > > rte_eth_dev_info_get(pid, &dev_info);
> > > rte_eth_dev_get_reg_info(port_id, &reg_info);
> > >
> > > for (i = 0; reg_info[i].name != NULL; i++) {
> > >    ...
> > >    rte_pcidev_read_reg(dev_info. pci_dev, reg_info[i], &v);
> > >   ..
> > > }
> > >
> > > > +
> > > > +int
> > > > +rte_eth_dev_eeprom_leng(uint8_t port_id) {
> > > > +	struct rte_eth_dev *dev;
> > > > +
> > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom_length, -
> > > ENOTSUP);
> > > > +	return (*dev->dev_ops->get_eeprom_length)(dev);
> > > > +}
> > > > +
> > > > +int
> > > > +rte_eth_dev_get_eeprom(uint8_t port_id, struct
> > > > +rte_dev_eeprom_info
> > > > +*info) {
> > > > +	struct rte_eth_dev *dev;
> > > > +
> > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -
> > > ENOTSUP);
> > > > +	return (*dev->dev_ops->get_eeprom)(dev, info); }
> > > > +
> > > > +int
> > > > +rte_eth_dev_set_eeprom(uint8_t port_id, struct
> > > > +rte_dev_eeprom_info
> > > > +*info) {
> > > > +	struct rte_eth_dev *dev;
> > > > +
> > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -
> > > ENOTSUP);
> > > > +	return (*dev->dev_ops->set_eeprom)(dev, info); }
> > > > +
> > > > +int
> > > > +rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > > +rte_dev_ring_info
> > > > +*info) {
> > > > +	struct rte_eth_dev *dev;
> > > > +
> > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> > > ENOTSUP);
> > > > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> > >
> > > I think it will be a useful addition to the ethdev API  to have an
> > > ability to retrieve current RX/TX queue parameters.
> > > Though again, it need to be more generic, so it could be useful for
> > > non- ethtool upper layer too.
> > > So I suggest to modify it a bit.
> > > Something like that:
> > >
> > > struct rte_eth_tx_queue_info {
> > >     struct rte_eth_txconf txconf;
> > >     uint32_t nb_tx_desc;
> > >     uint32_t nb_max_tx_desc; /*max allowable TXDs for that queue */
> > >     uint32_t nb_tx_free;            /* number of free TXDs at the moment of
> call.
> > > */
> > >     /* other tx queue data. */
> > > };
> > >
> > > int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> > > rte_eth_tx_queue_info *qinfo)
> > >
> > > Then, your upper layer ethtool wrapper, can implement yours
> > > ethtool_get_ringparam() by:
> > >
> > >  ...
> > >  struct rte_eth_tx_queue_info qinfo;
> > > rte_ethdev_get_tx_queue_info(port, 0, &qinfo);
> > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > rte_eth_rx_queue_count(port, 0);
> > >
> > > Or probably even:
> > > ring_param->tx_pending = qinfo.nb_tx_desc - qinfo.nb_tx_free;
> > >
> > > Same for RX.
> > >
> > For now, this descriptor ring information is used by the ethtool op.
> > To make this interface simple, i.e. caller doesn't need to access other
> queue information.
> 
> I just repeat what I said to you in off-line conversation:
> ethdev API is not equal ethtool API.
> It is ok to add  a new function/structure to ethdev if it really needed, but we
> should do mechanical one to one copy.
> It is much better to add  a function/structure that would be more generic,
> and suit other users, not only ethtool.
> There is no point to have dozen functions in rte_ethdev API providing similar
> information.
> BTW, I don't see how API I proposed is much more  complex, then yours one.
The ring parameter is a run-time information which is different than data structure described in this discussion.
It's the desire of this patch to separate each data structure to avoid cross dependency.

> 
> Konstantin
> 
> 
> > The plan it so keep this information as it is.
> > For the expansion, the ethdev can invoke the same dev op to add this
> information.
> > > > +
> > > > +int
> > > > +rte_eth_dev_set_ringparam(uint8_t port_id, struct
> > > > +rte_dev_ring_info
> > > > +*info) {
> > > > +	struct rte_eth_dev *dev;
> > > > +
> > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > +		return -ENODEV;
> > > > +	}
> > > > +
> > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_ringparam, -
> > > ENOTSUP);
> > > > +	return (*dev->dev_ops->set_ringparam)(dev, info); }
> > >
> > >
> > > I am a bit confused, what particular value of rte_dev_ring_info will
> > > be allowed to change?
> > > You can't change number of RXD/TXD at runtime.
> > > You have to stop the device first, and then reconfigure particular
> queue(s).
> > > Again, you can't reset RXF/TXD without stopping RX/TX first.
> > > So what that function is supposed to do?
> > > As I can see currently, none of the PMD in your patch support it.
> > >
> >
> > Exactly, setting RXD/TXD will require some change on kernel (RX/TX burst)
> function.
> > As you already observe that there is no dev_op registered for set_ring.
> > The API is created as a placeholder for future implementation.
> > There are a couple of idea thrown around (such as mono-
> increase/decrease to avoid buffer overrun).
> > Any suggestion is welcome.
> >
> > > Konstantin
> > >
> > > > diff --git a/lib/librte_ether/rte_ethdev.h
> > > > b/lib/librte_ether/rte_ethdev.h index 16dbe00..8f99f0d 100644
> > > > --- a/lib/librte_ether/rte_ethdev.h
> > > > +++ b/lib/librte_ether/rte_ethdev.h
> > > > @@ -182,6 +182,7 @@ extern "C" {
> > > >  #include <rte_devargs.h>
> > > >  #include "rte_ether.h"
> > > >  #include "rte_eth_ctrl.h"
> > > > +#include "rte_eth_dev_info.h"
> > > >
> > > >  struct rte_mbuf;
> > > >
> > > > @@ -1228,6 +1229,31 @@ typedef int (*eth_udp_tunnel_del_t)(struct
> > > rte_eth_dev *dev,
> > > >  				    struct rte_eth_udp_tunnel *tunnel_udp);
> > > /**< @internal
> > > > Delete tunneling UDP info */
> > > >
> > > > +typedef int (*eth_get_reg_length_t)(struct rte_eth_dev *dev);
> > > > +/**< @internal Retrieve device register count  */
> > > > +
> > > > +typedef int (*eth_get_reg_t)(struct rte_eth_dev *dev,
> > > > +				struct rte_dev_reg_info *info); /**<
> @internal Retrieve
> > > > +registers  */
> > > > +
> > > > +typedef int (*eth_get_eeprom_length_t)(struct rte_eth_dev *dev);
> > > /**<
> > > > +@internal Retrieve eeprom size  */
> > > > +
> > > > +typedef int (*eth_get_eeprom_t)(struct rte_eth_dev *dev,
> > > > +				struct rte_dev_eeprom_info *info); /**<
> > > @internal Retrieve eeprom
> > > > +data  */
> > > > +
> > > > +typedef int (*eth_set_eeprom_t)(struct rte_eth_dev *dev,
> > > > +				struct rte_dev_eeprom_info *info); /**<
> > > @internal Program eeprom
> > > > +data  */
> > > > +
> > > > +typedef int (*eth_get_ringparam_t)(struct rte_eth_dev *dev,
> > > > +				struct rte_dev_ring_info *info); /**<
> @internal Retrieve
> > > > +device ring descriptor  */
> > > > +
> > > > +typedef int (*eth_set_ringparam_t)(struct rte_eth_dev *dev,
> > > > +				struct rte_dev_ring_info *info); /**<
> @internal Set device
> > > > +ring descriptor  */
> > > >
> > > >  #ifdef RTE_NIC_BYPASS
> > > >
> > > > @@ -1368,6 +1394,21 @@ struct eth_dev_ops {
> > > >  	reta_update_t reta_update;
> > > >  	/** Query redirection table. */
> > > >  	reta_query_t reta_query;
> > > > +
> > > > +	eth_get_reg_length_t get_reg_length;
> > > > +	/**< Get # of registers */
> > > > +	eth_get_reg_t get_reg;
> > > > +	/**< Get registers */
> > > > +	eth_get_eeprom_length_t get_eeprom_length;
> > > > +	/**< Get eeprom length */
> > > > +	eth_get_eeprom_t get_eeprom;
> > > > +	/**< Get eeprom data */
> > > > +	eth_set_eeprom_t set_eeprom;
> > > > +	/**< Set eeprom */
> > > > +	eth_get_ringparam_t get_ringparam;
> > > > +	/**< Get descriptor ring parameters */
> > > > +	eth_set_ringparam_t set_ringparam;
> > > > +	/**< Set descriptor ring parameters */
> > > >    /* bypass control */
> > > >  #ifdef RTE_NIC_BYPASS
> > > >    bypass_init_t bypass_init;
> > > > @@ -2982,6 +3023,20 @@ int rte_eth_dev_mac_addr_add(uint8_t
> port,
> > > > struct ether_addr *mac_addr,  int
> > > rte_eth_dev_mac_addr_remove(uint8_t
> > > > port, struct ether_addr *mac_addr);
> > > >
> > > >  /**
> > > > + * Set the default MAC address.
> > > > + *
> > > > + * @param port
> > > > + *   The port identifier of the Ethernet device.
> > > > + * @param mac_addr
> > > > + *   New default MAC address.
> > > > + * @return
> > > > + *   - (0) if successful, or *mac_addr* didn't exist.
> > > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > > + *   - (-ENODEV) if *port* invalid.
> > > > + */
> > > > +int rte_eth_dev_default_mac_addr_set(uint8_t port, struct
> > > > +ether_addr *mac_addr);
> > > > +
> > > > +/**
> > > >   * Update Redirection Table(RETA) of Receive Side Scaling of
> > > > Ethernet
> > > device.
> > > >   *
> > > >   * @param port
> > > > @@ -3611,6 +3666,109 @@ int rte_eth_remove_rx_callback(uint8_t
> > > > port_id, uint16_t queue_id,  int
> > > > rte_eth_remove_tx_callback(uint8_t
> > > port_id, uint16_t queue_id,
> > > >  		struct rte_eth_rxtx_callback *user_cb);
> > > >
> > > > +/**
> > > > + * Retrieve number of available registers for access
> > > > + *
> > > > + * @param port_id
> > > > + *   The port identifier of the Ethernet device.
> > > > + * @return
> > > > + *   - (>=0) number of registers if successful.
> > > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > > + *   - (-ENODEV) if *port_id* invalid.
> > > > + *   - others depends on the specific operations implementation.
> > > > + */
> > > > +int rte_eth_dev_reg_leng(uint8_t port_id);
> > > > +
> > > > +/**
> > > > + * Retrieve device registers and register attributes
> > > > + *
> > > > + * @param port_id
> > > > + *   The port identifier of the Ethernet device.
> > > > + * @param info
> > > > + *   The template includes buffer for register data and attribute to be
> > > filled.
> > > > + * @return
> > > > + *   - (0) if successful.
> > > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > > + *   - (-ENODEV) if *port_id* invalid.
> > > > + *   - others depends on the specific operations implementation.
> > > > + */
> > > > +int rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info
> > > > +*info);
> > > > +
> > > > +/**
> > > > + * Retrieve size of device EEPROM
> > > > + *
> > > > + * @param port_id
> > > > + *   The port identifier of the Ethernet device.
> > > > + * @return
> > > > + *   - (>=0) EEPROM size if successful.
> > > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > > + *   - (-ENODEV) if *port_id* invalid.
> > > > + *   - others depends on the specific operations implementation.
> > > > + */
> > > > +int rte_eth_dev_eeprom_leng(uint8_t port_id);
> > > > +
> > > > +/**
> > > > + * Retrieve EEPROM and EEPROM attribute
> > > > + *
> > > > + * @param port_id
> > > > + *   The port identifier of the Ethernet device.
> > > > + * @param info
> > > > + *   The template includes buffer for return EEPROM data and
> > > > + *   EEPROM attributes to be filled.
> > > > + * @return
> > > > + *   - (0) if successful.
> > > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > > + *   - (-ENODEV) if *port_id* invalid.
> > > > + *   - others depends on the specific operations implementation.
> > > > + */
> > > > +int rte_eth_dev_get_eeprom(uint8_t port_id, struct
> > > > +rte_dev_eeprom_info *info);
> > > > +
> > > > +/**
> > > > + * Program EEPROM with provided data
> > > > + *
> > > > + * @param port_id
> > > > + *   The port identifier of the Ethernet device.
> > > > + * @param info
> > > > + *   The template includes EEPROM data for programming and
> > > > + *   EEPROM attributes to be filled
> > > > + * @return
> > > > + *   - (0) if successful.
> > > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > > + *   - (-ENODEV) if *port_id* invalid.
> > > > + *   - others depends on the specific operations implementation.
> > > > + */
> > > > +int rte_eth_dev_set_eeprom(uint8_t port_id, struct
> > > > +rte_dev_eeprom_info *info);
> > > > +
> > > > +/**
> > > > + * Retrieve device descriptor ring information
> > > > + *
> > > > + * @param port_id
> > > > + *   The port identifier of the Ethernet device.
> > > > + * @param info
> > > > + *   The template includes device ring descriptor usage information.
> > > > + * @return
> > > > + *   - (0) if successful.
> > > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > > + *   - (-ENODEV) if *port_id* invalid.
> > > > + *   - others depends on the specific operations implementation.
> > > > + */
> > > > +int rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > > +rte_dev_ring_info *info);
> > > > +
> > > > +/**
> > > > + * Set device ring descriptor parameters
> > > > + *
> > > > + * @param port_id
> > > > + *   The port identifier of the Ethernet device.
> > > > + * @param info
> > > > + *   The template includes descriptor ring parameters to be set.
> > > > + * @return
> > > > + *   - (0) if successful.
> > > > + *   - (-ENOTSUP) if hardware doesn't support.
> > > > + *   - (-ENODEV) if *port_id* invalid.
> > > > + *   - others depends on the specific operations implementation.
> > > > + */
> > > > +int rte_eth_dev_set_ringparam(uint8_t port_id, struct
> > > > +rte_dev_ring_info *info);
> > > > +
> > > >  #ifdef __cplusplus
> > > >  }
> > > >  #endif
> > > > diff --git a/lib/librte_ether/rte_ether_version.map
> > > > b/lib/librte_ether/rte_ether_version.map
> > > > index a2d25a6..1b1e092 100644
> > > > --- a/lib/librte_ether/rte_ether_version.map
> > > > +++ b/lib/librte_ether/rte_ether_version.map
> > > > @@ -102,6 +102,14 @@ DPDK_2.0 {
> > > >  	rte_eth_tx_queue_setup;
> > > >  	rte_eth_xstats_get;
> > > >  	rte_eth_xstats_reset;
> > > > +	rte_eth_dev_default_mac_addr_set;
> > > > +	rte_eth_dev_reg_leng;
> > > > +	rte_eth_dev_reg_info;
> > > > +	rte_eth_dev_eeprom_leng;
> > > > +	rte_eth_dev_get_eeprom;
> > > > +	rte_eth_dev_set_eeprom;
> > > > +	rte_eth_dev_get_ringparam;
> > > > +	rte_eth_dev_set_ringparam;
> > > >
> > > >  	local: *;
> > > >  };
> > > > --
> > > > 2.1.4
  

> -----Original Message-----
> From: Wang, Liang-min
> Sent: Thursday, June 11, 2015 10:51 PM
> To: Ananyev, Konstantin; dev@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access device info
> 
> 
> 
> > -----Original Message-----
> > From: Ananyev, Konstantin
> > Sent: Thursday, June 11, 2015 9:07 AM
> > To: Wang, Liang-min; dev@dpdk.org
> > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> > device info
> >
> >
> >
> > > -----Original Message-----
> > > From: Wang, Liang-min
> > > Sent: Thursday, June 11, 2015 1:58 PM
> > > To: Ananyev, Konstantin; dev@dpdk.org
> > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > access device info
> > >
> > > Hi Konstantin,
> > >
> > > > -----Original Message-----
> > > > From: Ananyev, Konstantin
> > > > Sent: Thursday, June 11, 2015 8:26 AM
> > > > To: Wang, Liang-min; dev@dpdk.org
> > > > Cc: Wang, Liang-min
> > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > access device info
> > > >
> > > > Hi Larry,
> > > >
> > > > > -----Original Message-----
> > > > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Liang-Min
> > > > > Larry Wang
> > > > > Sent: Wednesday, June 10, 2015 4:10 PM
> > > > > To: dev@dpdk.org
> > > > > Cc: Wang, Liang-min
> > > > > Subject: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > > access device info
> > > > >
> > > > > add new apis:
> > > > > - rte_eth_dev_default_mac_addr_set
> > > > > - rte_eth_dev_reg_leng
> > > > > - rte_eth_dev_reg_info
> > > > > - rte_eth_dev_eeprom_leng
> > > > > - rte_eth_dev_get_eeprom
> > > > > - rte_eth_dev_set_eeprom
> > > > > - rte_eth_dev_get_ringparam
> > > > > - rte_eth_dev_set_ringparam
> > > > >
> > > > > to enable reading device parameters (mac-addr, register, eeprom,
> > > > > ring) based upon ethtool alike data parameter specification.
> > > > >
> > > > > Signed-off-by: Liang-Min Larry Wang <liang-min.wang@intel.com>
> > > > > ---
> > > > >  lib/librte_ether/Makefile              |   1 +
> > > > >  lib/librte_ether/rte_eth_dev_info.h    |  80 +++++++++++++++++
> > > > >  lib/librte_ether/rte_ethdev.c          | 159
> > > > +++++++++++++++++++++++++++++++++
> > > > >  lib/librte_ether/rte_ethdev.h          | 158
> > > > ++++++++++++++++++++++++++++++++
> > > > >  lib/librte_ether/rte_ether_version.map |   8 ++
> > > > >  5 files changed, 406 insertions(+)  create mode 100644
> > > > > lib/librte_ether/rte_eth_dev_info.h
> > > > >
> > > > > diff --git a/lib/librte_ether/Makefile b/lib/librte_ether/Makefile
> > > > > index c0e5768..05209e9 100644
> > > > > --- a/lib/librte_ether/Makefile
> > > > > +++ b/lib/librte_ether/Makefile
> > > > > @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c  SYMLINK-y-include +=
> > > > > rte_ether.h  SYMLINK-y-include += rte_ethdev.h  SYMLINK-y-include
> > > > > += rte_eth_ctrl.h
> > > > > +SYMLINK-y-include += rte_eth_dev_info.h
> > > > >
> > > > >  # this lib depends upon:
> > > > >  DEPDIRS-y += lib/librte_eal lib/librte_mempool lib/librte_ring
> > > > > lib/librte_mbuf diff --git a/lib/librte_ether/rte_eth_dev_info.h
> > > > > b/lib/librte_ether/rte_eth_dev_info.h
> > > > > new file mode 100644
> > > > > index 0000000..002c4b5
> > > > > --- /dev/null
> > > > > +++ b/lib/librte_ether/rte_eth_dev_info.h
> > > > > @@ -0,0 +1,80 @@
> > > > > +/*-
> > > > > + *   BSD LICENSE
> > > > > + *
> > > > > + *   Copyright(c) 2015 Intel Corporation. All rights reserved.
> > > > > + *   All rights reserved.
> > > > > + *
> > > > > + *   Redistribution and use in source and binary forms, with or without
> > > > > + *   modification, are permitted provided that the following conditions
> > > > > + *   are met:
> > > > > + *
> > > > > + *     * Redistributions of source code must retain the above copyright
> > > > > + *       notice, this list of conditions and the following disclaimer.
> > > > > + *     * Redistributions in binary form must reproduce the above
> > copyright
> > > > > + *       notice, this list of conditions and the following disclaimer in
> > > > > + *       the documentation and/or other materials provided with the
> > > > > + *       distribution.
> > > > > + *     * Neither the name of Intel Corporation nor the names of its
> > > > > + *       contributors may be used to endorse or promote products
> > derived
> > > > > + *       from this software without specific prior written permission.
> > > > > + *
> > > > > + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> > > > CONTRIBUTORS
> > > > > + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING,
> > BUT
> > > > NOT
> > > > > + *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
> > > > FITNESS FOR
> > > > > + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
> > > > COPYRIGHT
> > > > > + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
> > > > INCIDENTAL,
> > > > > + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
> > (INCLUDING,
> > > > BUT NOT
> > > > > + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
> > > > LOSS OF USE,
> > > > > + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
> > CAUSED
> > > > AND ON ANY
> > > > > + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
> > OR
> > > > TORT
> > > > > + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
> > OUT
> > > > OF THE USE
> > > > > + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
> > > > DAMAGE.
> > > > > + */
> > > > > +
> > > > > +#ifndef _RTE_ETH_DEV_INFO_H_
> > > > > +#define _RTE_ETH_DEV_INFO_H_
> > > > > +
> > > > > +
> > > > > +/*
> > > > > + * Placeholder for accessing device registers  */ struct
> > > > > +rte_dev_reg_info {
> > > > > +	void *buf; /**< Buffer for register */
> > > > > +	uint32_t offset; /**< Offset for 1st register to fetch */
> > > > > +	uint32_t leng; /**< Number of registers to fetch */
> > > > > +	uint32_t version; /**< Device version */ };
> > > > > +
> > > > > +/*
> > > > > + * Placeholder for accessing device eeprom  */ struct
> > > > > +rte_dev_eeprom_info {
> > > > > +	void *buf; /**< Buffer for eeprom */
> > > > > +	uint32_t offset; /**< Offset for 1st eeprom location to access */
> > > > > +	uint32_t leng; /**< Length of eeprom region to access */
> > > > > +	uint32_t magic; /**< Device ID */ };
> > > > > +
> > > > > +/*
> > > > > + * Placeholder for accessing device ring parameters  */ struct
> > > > > +rte_dev_ring_info {
> > > > > +	uint32_t rx_pending; /**< Number of outstanding Rx ring */
> > > > > +	uint32_t tx_pending; /**< Number of outstanding Tx ring */
> > > > > +	uint32_t rx_max_pending; /**< Maximum number of outstanding Rx
> > > > ring */
> > > > > +	uint32_t tx_max_pending; /**< Maximum number of outstanding Tx
> > > > ring
> > > > > +*/ };
> > > > > +
> > > > > +/*
> > > > > + * A data structure captures information as defined in struct
> > > > > +ifla_vf_info
> > > > > + * for user-space api
> > > > > + */
> > > > > +struct rte_dev_vf_info {
> > > > > +	uint32_t vf;
> > > > > +	uint8_t mac[ETHER_ADDR_LEN];
> > > > > +	uint32_t vlan;
> > > > > +	uint32_t tx_rate;
> > > > > +	uint32_t spoofchk;
> > > > > +};
> > > >
> > > >
> > > > Wonder what that structure is for?
> > > > I can't see it used in any function below?
> > > >
> > >
> > > Good catch, this is designed for other ethtool ops that I did not include in
> > this release, I will remove this from next fix.
> > >
> > > > > +
> > > > > +#endif /* _RTE_ETH_DEV_INFO_H_ */
> > > > > diff --git a/lib/librte_ether/rte_ethdev.c
> > > > > b/lib/librte_ether/rte_ethdev.c index 5a94654..186e85c 100644
> > > > > --- a/lib/librte_ether/rte_ethdev.c
> > > > > +++ b/lib/librte_ether/rte_ethdev.c
> > > > > @@ -2751,6 +2751,32 @@ rte_eth_dev_mac_addr_remove(uint8_t
> > > > port_id,
> > > > > struct ether_addr *addr)  }
> > > > >
> > > > >  int
> > > > > +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct
> > > > > +ether_addr
> > > > > +*addr) {
> > > > > +	struct rte_eth_dev *dev;
> > > > > +	const int index = 0;
> > > > > +	const uint32_t pool = 0;
> > > > > +
> > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	dev = &rte_eth_devices[port_id];
> > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_remove, -
> > > > ENOTSUP);
> > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -
> > > > ENOTSUP);
> > > > > +
> > > > > +	/* Update NIC default MAC address*/
> > > > > +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> > > > > +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> > > > > +
> > > > > +	/* Update default address in NIC data structure */
> > > > > +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> > > > > +
> > > > > +	return 0;
> > > > > +}
> > > > > +
> > > > > +int
> > > > >  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
> > > > >  				uint16_t rx_mode, uint8_t on)  { @@ -3627,3
> > +3653,136 @@
> > > > > rte_eth_remove_tx_callback(uint8_t port_id,
> > > > uint16_t queue_id,
> > > > >  	/* Callback wasn't found. */
> > > > >  	return -EINVAL;
> > > > >  }
> > > > > +
> > > > > +int
> > > > > +rte_eth_dev_reg_leng(uint8_t port_id) {
> > > > > +	struct rte_eth_dev *dev;
> > > > > +
> > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length, -
> > > > ENOTSUP);
> > > > > +	return (*dev->dev_ops->get_reg_length)(dev);
> > > > > +}
> > > > > +
> > > > > +int
> > > > > +rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info
> > > > > +*info) {
> > > > > +	struct rte_eth_dev *dev;
> > > > > +
> > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -ENOTSUP);
> > > > > +	return (*dev->dev_ops->get_reg)(dev, info); }
> > > >
> > > > Seems that *get_reg* stuff, will be really good addition for DPDK
> > > > debugging abilities.
> > > > Though, I'd suggest we change it a bit to make more generic and flexible:
> > > >
> > > > Introduce rte_eth_reg_read/rte_eth_reg_write(),
> > > > or probably even better rte_pcidev_reg_read /rte_pcidev_reg_write at
> > EAL.
> > > > Something similar to what port_pci_reg_read/port_pci_reg_write() are
> > > > doing now at testpmd.h.
> > > >
> > > > struct rte_pcidev_reg_info {
> > > >    const char *name;
> > > >    uint32_t endianes, bar, offset, size, count; };
> > > >
> > > > int rte_pcidev_reg_read(const struct rte_pci_device *, const struct
> > > > rte_pcidev_reg_info *, uint64_t *reg_val);
> > > >
> > > > Then:
> > > > int rte_eth_dev_get_reg_info(port_id, const struct
> > > > rte_pcidev_reg_info **info);
> > > >
> > > > So each device would store in info a pointer to an array of it's
> > > > register descriptions (finished by zero elem?).
> > > >
> > > > Then your ethtool (or any other upper layer) can do the following to
> > > > read all device regs:
> > > >
> > >
> > > The proposed reg info structure allows future improvement to support
> > individual register read/write.
> > > Also, because each NIC device has a very distinguish register definition.
> > > So, the plan is to have more comprehensive interface to support query
> > > operation (for example, register name) before introduce individual/group
> > register access.
> > > Points taken, the support will be in future release.
> >
> > Sorry, didn't get you.
> > So you are ok to make these changes in next patch version?
> >
> I would like to get a consensus from dpdk community on how to provide register information.

Well, that' ok, but if it is just a trial patch that is not intended to be applied, then you should mark it as RFC.

> Currently, it's designed for debug dumping. The register information is very hardware dependent.
> Need to consider current supported NIC device and future devices for DPDK, so we won't make it a bulky interface.

Ok, could you explain what exactly  concerns you in the approach described above?
What part you feel is bulky?

> > >
> > > > const struct rte_eth_dev_reg_info *reg_info; struct rte_eth_dev_info
> > > > dev_info;
> > > >
> > > > rte_eth_dev_info_get(pid, &dev_info);
> > > > rte_eth_dev_get_reg_info(port_id, &reg_info);
> > > >
> > > > for (i = 0; reg_info[i].name != NULL; i++) {
> > > >    ...
> > > >    rte_pcidev_read_reg(dev_info. pci_dev, reg_info[i], &v);
> > > >   ..
> > > > }
> > > >
> > > > > +
> > > > > +int
> > > > > +rte_eth_dev_eeprom_leng(uint8_t port_id) {
> > > > > +	struct rte_eth_dev *dev;
> > > > > +
> > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom_length, -
> > > > ENOTSUP);
> > > > > +	return (*dev->dev_ops->get_eeprom_length)(dev);
> > > > > +}
> > > > > +
> > > > > +int
> > > > > +rte_eth_dev_get_eeprom(uint8_t port_id, struct
> > > > > +rte_dev_eeprom_info
> > > > > +*info) {
> > > > > +	struct rte_eth_dev *dev;
> > > > > +
> > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -
> > > > ENOTSUP);
> > > > > +	return (*dev->dev_ops->get_eeprom)(dev, info); }
> > > > > +
> > > > > +int
> > > > > +rte_eth_dev_set_eeprom(uint8_t port_id, struct
> > > > > +rte_dev_eeprom_info
> > > > > +*info) {
> > > > > +	struct rte_eth_dev *dev;
> > > > > +
> > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -
> > > > ENOTSUP);
> > > > > +	return (*dev->dev_ops->set_eeprom)(dev, info); }
> > > > > +
> > > > > +int
> > > > > +rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > > > +rte_dev_ring_info
> > > > > +*info) {
> > > > > +	struct rte_eth_dev *dev;
> > > > > +
> > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > +		return -ENODEV;
> > > > > +	}
> > > > > +
> > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> > > > ENOTSUP);
> > > > > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> > > >
> > > > I think it will be a useful addition to the ethdev API  to have an
> > > > ability to retrieve current RX/TX queue parameters.
> > > > Though again, it need to be more generic, so it could be useful for
> > > > non- ethtool upper layer too.
> > > > So I suggest to modify it a bit.
> > > > Something like that:
> > > >
> > > > struct rte_eth_tx_queue_info {
> > > >     struct rte_eth_txconf txconf;
> > > >     uint32_t nb_tx_desc;
> > > >     uint32_t nb_max_tx_desc; /*max allowable TXDs for that queue */
> > > >     uint32_t nb_tx_free;            /* number of free TXDs at the moment of
> > call.
> > > > */
> > > >     /* other tx queue data. */
> > > > };
> > > >
> > > > int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> > > > rte_eth_tx_queue_info *qinfo)
> > > >
> > > > Then, your upper layer ethtool wrapper, can implement yours
> > > > ethtool_get_ringparam() by:
> > > >
> > > >  ...
> > > >  struct rte_eth_tx_queue_info qinfo;
> > > > rte_ethdev_get_tx_queue_info(port, 0, &qinfo);
> > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > rte_eth_rx_queue_count(port, 0);
> > > >
> > > > Or probably even:
> > > > ring_param->tx_pending = qinfo.nb_tx_desc - qinfo.nb_tx_free;
> > > >
> > > > Same for RX.
> > > >
> > > For now, this descriptor ring information is used by the ethtool op.
> > > To make this interface simple, i.e. caller doesn't need to access other
> > queue information.
> >
> > I just repeat what I said to you in off-line conversation:
> > ethdev API is not equal ethtool API.
> > It is ok to add  a new function/structure to ethdev if it really needed, but we
> > should do mechanical one to one copy.
> > It is much better to add  a function/structure that would be more generic,
> > and suit other users, not only ethtool.
> > There is no point to have dozen functions in rte_ethdev API providing similar
> > information.
> > BTW, I don't see how API I proposed is much more  complex, then yours one.
> The ring parameter is a run-time information which is different than data structure described in this discussion.

I don't see how they are different.
Looking at ixgbe_get_ringparam(), it returns:
rx_max_pending - that's a static IXGBE PMD value (max possible number of RXDs per one queue).
rx_pending - number of RXD currently in use by the HW for queue 0 (that information can be changed at each call).

With the approach I suggesting - you can get same information for each RX queue by
calling rte_ethdev_get_rx_queue_info() and rte_eth_rx_queue_count(). 
Plus you are getting other RX queue data.

Another thing - what is practical usage of the information you retrieving now by get_ringparam()?
Let say it returned to you: rx_max_pending=4096; rx_pending=128;
How that information would help to understand what is going on with the device?
Without knowing  value of nb_tx_desc for the queue, you can't say is you queue full or not.
Again, it could be that all your traffic going through some other queue (not 0).
So from my point rte_eth_dev_get_ringparam()  usage is very limited, and doesn't
provide enough information about current queue state.

Same thing applies for TX. 

> It's the desire of this patch to separate each data structure to avoid cross dependency.

That's too cryptic to me.
Could you explain what cross dependency you are talking about?

Konstantin
  

> -----Original Message-----
> From: Ananyev, Konstantin
> Sent: Friday, June 12, 2015 8:31 AM
> To: Wang, Liang-min; dev@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> device info
> 
> 
> 
> > -----Original Message-----
> > From: Wang, Liang-min
> > Sent: Thursday, June 11, 2015 10:51 PM
> > To: Ananyev, Konstantin; dev@dpdk.org
> > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > access device info
> >
> >
> >
> > > -----Original Message-----
> > > From: Ananyev, Konstantin
> > > Sent: Thursday, June 11, 2015 9:07 AM
> > > To: Wang, Liang-min; dev@dpdk.org
> > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > access device info
> > >
> > >
> > >
> > > > -----Original Message-----
> > > > From: Wang, Liang-min
> > > > Sent: Thursday, June 11, 2015 1:58 PM
> > > > To: Ananyev, Konstantin; dev@dpdk.org
> > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > access device info
> > > >
> > > > Hi Konstantin,
> > > >
> > > > > -----Original Message-----
> > > > > From: Ananyev, Konstantin
> > > > > Sent: Thursday, June 11, 2015 8:26 AM
> > > > > To: Wang, Liang-min; dev@dpdk.org
> > > > > Cc: Wang, Liang-min
> > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > support access device info
> > > > >
> > > > > Hi Larry,
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Liang-Min
> > > > > > Larry Wang
> > > > > > Sent: Wednesday, June 10, 2015 4:10 PM
> > > > > > To: dev@dpdk.org
> > > > > > Cc: Wang, Liang-min
> > > > > > Subject: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > > > access device info
> > > > > >
> > > > > > add new apis:
> > > > > > - rte_eth_dev_default_mac_addr_set
> > > > > > - rte_eth_dev_reg_leng
> > > > > > - rte_eth_dev_reg_info
> > > > > > - rte_eth_dev_eeprom_leng
> > > > > > - rte_eth_dev_get_eeprom
> > > > > > - rte_eth_dev_set_eeprom
> > > > > > - rte_eth_dev_get_ringparam
> > > > > > - rte_eth_dev_set_ringparam
> > > > > >
> > > > > > to enable reading device parameters (mac-addr, register,
> > > > > > eeprom,
> > > > > > ring) based upon ethtool alike data parameter specification.
> > > > > >
> > > > > > Signed-off-by: Liang-Min Larry Wang <liang-min.wang@intel.com>
> > > > > > ---
> > > > > >  lib/librte_ether/Makefile              |   1 +
> > > > > >  lib/librte_ether/rte_eth_dev_info.h    |  80 +++++++++++++++++
> > > > > >  lib/librte_ether/rte_ethdev.c          | 159
> > > > > +++++++++++++++++++++++++++++++++
> > > > > >  lib/librte_ether/rte_ethdev.h          | 158
> > > > > ++++++++++++++++++++++++++++++++
> > > > > >  lib/librte_ether/rte_ether_version.map |   8 ++
> > > > > >  5 files changed, 406 insertions(+)  create mode 100644
> > > > > > lib/librte_ether/rte_eth_dev_info.h
> > > > > >
> > > > > > diff --git a/lib/librte_ether/Makefile
> > > > > > b/lib/librte_ether/Makefile index c0e5768..05209e9 100644
> > > > > > --- a/lib/librte_ether/Makefile
> > > > > > +++ b/lib/librte_ether/Makefile
> > > > > > @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c  SYMLINK-y-include +=
> > > > > > rte_ether.h  SYMLINK-y-include += rte_ethdev.h
> > > > > > SYMLINK-y-include
> > > > > > += rte_eth_ctrl.h
> > > > > > +SYMLINK-y-include += rte_eth_dev_info.h
> > > > > >
> > > > > >  # this lib depends upon:
> > > > > >  DEPDIRS-y += lib/librte_eal lib/librte_mempool
> > > > > > lib/librte_ring lib/librte_mbuf diff --git
> > > > > > a/lib/librte_ether/rte_eth_dev_info.h
> > > > > > b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > new file mode 100644
> > > > > > index 0000000..002c4b5
> > > > > > --- /dev/null
> > > > > > +++ b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > @@ -0,0 +1,80 @@
> > > > > > +/*-
> > > > > > + *   BSD LICENSE
> > > > > > + *
> > > > > > + *   Copyright(c) 2015 Intel Corporation. All rights reserved.
> > > > > > + *   All rights reserved.
> > > > > > + *
> > > > > > + *   Redistribution and use in source and binary forms, with or
> without
> > > > > > + *   modification, are permitted provided that the following
> conditions
> > > > > > + *   are met:
> > > > > > + *
> > > > > > + *     * Redistributions of source code must retain the above
> copyright
> > > > > > + *       notice, this list of conditions and the following disclaimer.
> > > > > > + *     * Redistributions in binary form must reproduce the above
> > > copyright
> > > > > > + *       notice, this list of conditions and the following disclaimer in
> > > > > > + *       the documentation and/or other materials provided with the
> > > > > > + *       distribution.
> > > > > > + *     * Neither the name of Intel Corporation nor the names of its
> > > > > > + *       contributors may be used to endorse or promote products
> > > derived
> > > > > > + *       from this software without specific prior written permission.
> > > > > > + *
> > > > > > + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> > > > > CONTRIBUTORS
> > > > > > + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
> INCLUDING,
> > > BUT
> > > > > NOT
> > > > > > + *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
> AND
> > > > > FITNESS FOR
> > > > > > + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL
> THE
> > > > > COPYRIGHT
> > > > > > + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
> INDIRECT,
> > > > > INCIDENTAL,
> > > > > > + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
> > > (INCLUDING,
> > > > > BUT NOT
> > > > > > + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
> SERVICES;
> > > > > LOSS OF USE,
> > > > > > + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
> > > CAUSED
> > > > > AND ON ANY
> > > > > > + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
> LIABILITY,
> > > OR
> > > > > TORT
> > > > > > + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY
> WAY
> > > OUT
> > > > > OF THE USE
> > > > > > + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
> SUCH
> > > > > DAMAGE.
> > > > > > + */
> > > > > > +
> > > > > > +#ifndef _RTE_ETH_DEV_INFO_H_
> > > > > > +#define _RTE_ETH_DEV_INFO_H_
> > > > > > +
> > > > > > +
> > > > > > +/*
> > > > > > + * Placeholder for accessing device registers  */ struct
> > > > > > +rte_dev_reg_info {
> > > > > > +	void *buf; /**< Buffer for register */
> > > > > > +	uint32_t offset; /**< Offset for 1st register to fetch */
> > > > > > +	uint32_t leng; /**< Number of registers to fetch */
> > > > > > +	uint32_t version; /**< Device version */ };
> > > > > > +
> > > > > > +/*
> > > > > > + * Placeholder for accessing device eeprom  */ struct
> > > > > > +rte_dev_eeprom_info {
> > > > > > +	void *buf; /**< Buffer for eeprom */
> > > > > > +	uint32_t offset; /**< Offset for 1st eeprom location to access
> */
> > > > > > +	uint32_t leng; /**< Length of eeprom region to access */
> > > > > > +	uint32_t magic; /**< Device ID */ };
> > > > > > +
> > > > > > +/*
> > > > > > + * Placeholder for accessing device ring parameters  */
> > > > > > +struct rte_dev_ring_info {
> > > > > > +	uint32_t rx_pending; /**< Number of outstanding Rx ring */
> > > > > > +	uint32_t tx_pending; /**< Number of outstanding Tx ring */
> > > > > > +	uint32_t rx_max_pending; /**< Maximum number of
> outstanding
> > > > > > +Rx
> > > > > ring */
> > > > > > +	uint32_t tx_max_pending; /**< Maximum number of
> outstanding
> > > > > > +Tx
> > > > > ring
> > > > > > +*/ };
> > > > > > +
> > > > > > +/*
> > > > > > + * A data structure captures information as defined in struct
> > > > > > +ifla_vf_info
> > > > > > + * for user-space api
> > > > > > + */
> > > > > > +struct rte_dev_vf_info {
> > > > > > +	uint32_t vf;
> > > > > > +	uint8_t mac[ETHER_ADDR_LEN];
> > > > > > +	uint32_t vlan;
> > > > > > +	uint32_t tx_rate;
> > > > > > +	uint32_t spoofchk;
> > > > > > +};
> > > > >
> > > > >
> > > > > Wonder what that structure is for?
> > > > > I can't see it used in any function below?
> > > > >
> > > >
> > > > Good catch, this is designed for other ethtool ops that I did not
> > > > include in
> > > this release, I will remove this from next fix.
> > > >
> > > > > > +
> > > > > > +#endif /* _RTE_ETH_DEV_INFO_H_ */
> > > > > > diff --git a/lib/librte_ether/rte_ethdev.c
> > > > > > b/lib/librte_ether/rte_ethdev.c index 5a94654..186e85c 100644
> > > > > > --- a/lib/librte_ether/rte_ethdev.c
> > > > > > +++ b/lib/librte_ether/rte_ethdev.c
> > > > > > @@ -2751,6 +2751,32 @@ rte_eth_dev_mac_addr_remove(uint8_t
> > > > > port_id,
> > > > > > struct ether_addr *addr)  }
> > > > > >
> > > > > >  int
> > > > > > +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct
> > > > > > +ether_addr
> > > > > > +*addr) {
> > > > > > +	struct rte_eth_dev *dev;
> > > > > > +	const int index = 0;
> > > > > > +	const uint32_t pool = 0;
> > > > > > +
> > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> port_id);
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	dev = &rte_eth_devices[port_id];
> > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> >mac_addr_remove, -
> > > > > ENOTSUP);
> > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -
> > > > > ENOTSUP);
> > > > > > +
> > > > > > +	/* Update NIC default MAC address*/
> > > > > > +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> > > > > > +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> > > > > > +
> > > > > > +	/* Update default address in NIC data structure */
> > > > > > +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> > > > > > +
> > > > > > +	return 0;
> > > > > > +}
> > > > > > +
> > > > > > +int
> > > > > >  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
> > > > > >  				uint16_t rx_mode, uint8_t on)  { @@
> -3627,3
> > > +3653,136 @@
> > > > > > rte_eth_remove_tx_callback(uint8_t port_id,
> > > > > uint16_t queue_id,
> > > > > >  	/* Callback wasn't found. */
> > > > > >  	return -EINVAL;
> > > > > >  }
> > > > > > +
> > > > > > +int
> > > > > > +rte_eth_dev_reg_leng(uint8_t port_id) {
> > > > > > +	struct rte_eth_dev *dev;
> > > > > > +
> > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> port_id);
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length,
> -
> > > > > ENOTSUP);
> > > > > > +	return (*dev->dev_ops->get_reg_length)(dev);
> > > > > > +}
> > > > > > +
> > > > > > +int
> > > > > > +rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info
> > > > > > +*info) {
> > > > > > +	struct rte_eth_dev *dev;
> > > > > > +
> > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> port_id);
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -
> ENOTSUP);
> > > > > > +	return (*dev->dev_ops->get_reg)(dev, info); }
> > > > >
> > > > > Seems that *get_reg* stuff, will be really good addition for
> > > > > DPDK debugging abilities.
> > > > > Though, I'd suggest we change it a bit to make more generic and
> flexible:
> > > > >
> > > > > Introduce rte_eth_reg_read/rte_eth_reg_write(),
> > > > > or probably even better rte_pcidev_reg_read
> > > > > /rte_pcidev_reg_write at
> > > EAL.
> > > > > Something similar to what port_pci_reg_read/port_pci_reg_write()
> > > > > are doing now at testpmd.h.
> > > > >
> > > > > struct rte_pcidev_reg_info {
> > > > >    const char *name;
> > > > >    uint32_t endianes, bar, offset, size, count; };
> > > > >
> > > > > int rte_pcidev_reg_read(const struct rte_pci_device *, const
> > > > > struct rte_pcidev_reg_info *, uint64_t *reg_val);
> > > > >
> > > > > Then:
> > > > > int rte_eth_dev_get_reg_info(port_id, const struct
> > > > > rte_pcidev_reg_info **info);
> > > > >
> > > > > So each device would store in info a pointer to an array of it's
> > > > > register descriptions (finished by zero elem?).
> > > > >
> > > > > Then your ethtool (or any other upper layer) can do the
> > > > > following to read all device regs:
> > > > >
> > > >
> > > > The proposed reg info structure allows future improvement to
> > > > support
> > > individual register read/write.
> > > > Also, because each NIC device has a very distinguish register definition.
> > > > So, the plan is to have more comprehensive interface to support
> > > > query operation (for example, register name) before introduce
> > > > individual/group
> > > register access.
> > > > Points taken, the support will be in future release.
> > >
> > > Sorry, didn't get you.
> > > So you are ok to make these changes in next patch version?
> > >
> > I would like to get a consensus from dpdk community on how to provide
> register information.
> 
> Well, that' ok, but if it is just a trial patch that is not intended to be applied,
> then you should mark it as RFC.
> 
> > Currently, it's designed for debug dumping. The register information is very
> hardware dependent.
> > Need to consider current supported NIC device and future devices for
> DPDK, so we won't make it a bulky interface.
> 
> Ok, could you explain what exactly  concerns you in the approach described
> above?
> What part you feel is bulky?
> 
> > > >
> > > > > const struct rte_eth_dev_reg_info *reg_info; struct
> > > > > rte_eth_dev_info dev_info;
> > > > >
> > > > > rte_eth_dev_info_get(pid, &dev_info);
> > > > > rte_eth_dev_get_reg_info(port_id, &reg_info);
> > > > >
> > > > > for (i = 0; reg_info[i].name != NULL; i++) {
> > > > >    ...
> > > > >    rte_pcidev_read_reg(dev_info. pci_dev, reg_info[i], &v);
> > > > >   ..
> > > > > }
> > > > >
> > > > > > +
> > > > > > +int
> > > > > > +rte_eth_dev_eeprom_leng(uint8_t port_id) {
> > > > > > +	struct rte_eth_dev *dev;
> > > > > > +
> > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> port_id);
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> >get_eeprom_length, -
> > > > > ENOTSUP);
> > > > > > +	return (*dev->dev_ops->get_eeprom_length)(dev);
> > > > > > +}
> > > > > > +
> > > > > > +int
> > > > > > +rte_eth_dev_get_eeprom(uint8_t port_id, struct
> > > > > > +rte_dev_eeprom_info
> > > > > > +*info) {
> > > > > > +	struct rte_eth_dev *dev;
> > > > > > +
> > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> port_id);
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -
> > > > > ENOTSUP);
> > > > > > +	return (*dev->dev_ops->get_eeprom)(dev, info); }
> > > > > > +
> > > > > > +int
> > > > > > +rte_eth_dev_set_eeprom(uint8_t port_id, struct
> > > > > > +rte_dev_eeprom_info
> > > > > > +*info) {
> > > > > > +	struct rte_eth_dev *dev;
> > > > > > +
> > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> port_id);
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -
> > > > > ENOTSUP);
> > > > > > +	return (*dev->dev_ops->set_eeprom)(dev, info); }
> > > > > > +
> > > > > > +int
> > > > > > +rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > > > > +rte_dev_ring_info
> > > > > > +*info) {
> > > > > > +	struct rte_eth_dev *dev;
> > > > > > +
> > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> port_id);
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > +		return -ENODEV;
> > > > > > +	}
> > > > > > +
> > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> > > > > ENOTSUP);
> > > > > > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> > > > >
> > > > > I think it will be a useful addition to the ethdev API  to have
> > > > > an ability to retrieve current RX/TX queue parameters.
> > > > > Though again, it need to be more generic, so it could be useful
> > > > > for
> > > > > non- ethtool upper layer too.
> > > > > So I suggest to modify it a bit.
> > > > > Something like that:
> > > > >
> > > > > struct rte_eth_tx_queue_info {
> > > > >     struct rte_eth_txconf txconf;
> > > > >     uint32_t nb_tx_desc;
> > > > >     uint32_t nb_max_tx_desc; /*max allowable TXDs for that queue */
> > > > >     uint32_t nb_tx_free;            /* number of free TXDs at the moment
> of
> > > call.
> > > > > */
> > > > >     /* other tx queue data. */
> > > > > };
> > > > >
> > > > > int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> > > > > rte_eth_tx_queue_info *qinfo)
> > > > >
> > > > > Then, your upper layer ethtool wrapper, can implement yours
> > > > > ethtool_get_ringparam() by:
> > > > >
> > > > >  ...
> > > > >  struct rte_eth_tx_queue_info qinfo;
> > > > > rte_ethdev_get_tx_queue_info(port, 0, &qinfo);
> > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > rte_eth_rx_queue_count(port, 0);
> > > > >
> > > > > Or probably even:
> > > > > ring_param->tx_pending = qinfo.nb_tx_desc - qinfo.nb_tx_free;
> > > > >
> > > > > Same for RX.
> > > > >
> > > > For now, this descriptor ring information is used by the ethtool op.
> > > > To make this interface simple, i.e. caller doesn't need to access
> > > > other
> > > queue information.
> > >
> > > I just repeat what I said to you in off-line conversation:
> > > ethdev API is not equal ethtool API.
> > > It is ok to add  a new function/structure to ethdev if it really
> > > needed, but we should do mechanical one to one copy.
> > > It is much better to add  a function/structure that would be more
> > > generic, and suit other users, not only ethtool.
> > > There is no point to have dozen functions in rte_ethdev API
> > > providing similar information.
> > > BTW, I don't see how API I proposed is much more  complex, then yours
> one.
> > The ring parameter is a run-time information which is different than data
> structure described in this discussion.
> 
> I don't see how they are different.
> Looking at ixgbe_get_ringparam(), it returns:
> rx_max_pending - that's a static IXGBE PMD value (max possible number of
> RXDs per one queue).
> rx_pending - number of RXD currently in use by the HW for queue 0 (that
> information can be changed at each call).
> 
> With the approach I suggesting - you can get same information for each RX
> queue by calling rte_ethdev_get_rx_queue_info() and
> rte_eth_rx_queue_count().
> Plus you are getting other RX queue data.
> 
> Another thing - what is practical usage of the information you retrieving now
> by get_ringparam()?
> Let say it returned to you: rx_max_pending=4096; rx_pending=128; How that
> information would help to understand what is going on with the device?
> Without knowing  value of nb_tx_desc for the queue, you can't say is you
> queue full or not.
> Again, it could be that all your traffic going through some other queue (not
> 0).
> So from my point rte_eth_dev_get_ringparam()  usage is very limited, and
> doesn't provide enough information about current queue state.
> 
> Same thing applies for TX.
> 

After careful review the suggestion in this comment, and review the existing dpdk source code. 
I came to realize that neither rte_ethdev_get_rx_queue_info, rte_ethdev_get_tx_queue_info, struct rte_eth_rx_queue_info and struct rte_eth_tx_queue_info are available in the existing dpdk source code. I could not make a patch based upon a set of non-existent API and data structure.

> > It's the desire of this patch to separate each data structure to avoid cross
> dependency.
> 
> That's too cryptic to me.
> Could you explain what cross dependency you are talking about?
> 
> Konstantin
  

> -----Original Message-----
> From: Wang, Liang-min
> Sent: Monday, June 15, 2015 2:26 PM
> To: Ananyev, Konstantin; dev@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access device info
> 
> 
> 
> > -----Original Message-----
> > From: Ananyev, Konstantin
> > Sent: Friday, June 12, 2015 8:31 AM
> > To: Wang, Liang-min; dev@dpdk.org
> > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> > device info
> >
> >
> >
> > > -----Original Message-----
> > > From: Wang, Liang-min
> > > Sent: Thursday, June 11, 2015 10:51 PM
> > > To: Ananyev, Konstantin; dev@dpdk.org
> > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > access device info
> > >
> > >
> > >
> > > > -----Original Message-----
> > > > From: Ananyev, Konstantin
> > > > Sent: Thursday, June 11, 2015 9:07 AM
> > > > To: Wang, Liang-min; dev@dpdk.org
> > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > access device info
> > > >
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Wang, Liang-min
> > > > > Sent: Thursday, June 11, 2015 1:58 PM
> > > > > To: Ananyev, Konstantin; dev@dpdk.org
> > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > > access device info
> > > > >
> > > > > Hi Konstantin,
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: Ananyev, Konstantin
> > > > > > Sent: Thursday, June 11, 2015 8:26 AM
> > > > > > To: Wang, Liang-min; dev@dpdk.org
> > > > > > Cc: Wang, Liang-min
> > > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > support access device info
> > > > > >
> > > > > > Hi Larry,
> > > > > >
> > > > > > > -----Original Message-----
> > > > > > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Liang-Min
> > > > > > > Larry Wang
> > > > > > > Sent: Wednesday, June 10, 2015 4:10 PM
> > > > > > > To: dev@dpdk.org
> > > > > > > Cc: Wang, Liang-min
> > > > > > > Subject: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > > > > access device info
> > > > > > >
> > > > > > > add new apis:
> > > > > > > - rte_eth_dev_default_mac_addr_set
> > > > > > > - rte_eth_dev_reg_leng
> > > > > > > - rte_eth_dev_reg_info
> > > > > > > - rte_eth_dev_eeprom_leng
> > > > > > > - rte_eth_dev_get_eeprom
> > > > > > > - rte_eth_dev_set_eeprom
> > > > > > > - rte_eth_dev_get_ringparam
> > > > > > > - rte_eth_dev_set_ringparam
> > > > > > >
> > > > > > > to enable reading device parameters (mac-addr, register,
> > > > > > > eeprom,
> > > > > > > ring) based upon ethtool alike data parameter specification.
> > > > > > >
> > > > > > > Signed-off-by: Liang-Min Larry Wang <liang-min.wang@intel.com>
> > > > > > > ---
> > > > > > >  lib/librte_ether/Makefile              |   1 +
> > > > > > >  lib/librte_ether/rte_eth_dev_info.h    |  80 +++++++++++++++++
> > > > > > >  lib/librte_ether/rte_ethdev.c          | 159
> > > > > > +++++++++++++++++++++++++++++++++
> > > > > > >  lib/librte_ether/rte_ethdev.h          | 158
> > > > > > ++++++++++++++++++++++++++++++++
> > > > > > >  lib/librte_ether/rte_ether_version.map |   8 ++
> > > > > > >  5 files changed, 406 insertions(+)  create mode 100644
> > > > > > > lib/librte_ether/rte_eth_dev_info.h
> > > > > > >
> > > > > > > diff --git a/lib/librte_ether/Makefile
> > > > > > > b/lib/librte_ether/Makefile index c0e5768..05209e9 100644
> > > > > > > --- a/lib/librte_ether/Makefile
> > > > > > > +++ b/lib/librte_ether/Makefile
> > > > > > > @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c  SYMLINK-y-include +=
> > > > > > > rte_ether.h  SYMLINK-y-include += rte_ethdev.h
> > > > > > > SYMLINK-y-include
> > > > > > > += rte_eth_ctrl.h
> > > > > > > +SYMLINK-y-include += rte_eth_dev_info.h
> > > > > > >
> > > > > > >  # this lib depends upon:
> > > > > > >  DEPDIRS-y += lib/librte_eal lib/librte_mempool
> > > > > > > lib/librte_ring lib/librte_mbuf diff --git
> > > > > > > a/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > new file mode 100644
> > > > > > > index 0000000..002c4b5
> > > > > > > --- /dev/null
> > > > > > > +++ b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > @@ -0,0 +1,80 @@
> > > > > > > +/*-
> > > > > > > + *   BSD LICENSE
> > > > > > > + *
> > > > > > > + *   Copyright(c) 2015 Intel Corporation. All rights reserved.
> > > > > > > + *   All rights reserved.
> > > > > > > + *
> > > > > > > + *   Redistribution and use in source and binary forms, with or
> > without
> > > > > > > + *   modification, are permitted provided that the following
> > conditions
> > > > > > > + *   are met:
> > > > > > > + *
> > > > > > > + *     * Redistributions of source code must retain the above
> > copyright
> > > > > > > + *       notice, this list of conditions and the following disclaimer.
> > > > > > > + *     * Redistributions in binary form must reproduce the above
> > > > copyright
> > > > > > > + *       notice, this list of conditions and the following disclaimer in
> > > > > > > + *       the documentation and/or other materials provided with the
> > > > > > > + *       distribution.
> > > > > > > + *     * Neither the name of Intel Corporation nor the names of its
> > > > > > > + *       contributors may be used to endorse or promote products
> > > > derived
> > > > > > > + *       from this software without specific prior written permission.
> > > > > > > + *
> > > > > > > + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> > > > > > CONTRIBUTORS
> > > > > > > + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
> > INCLUDING,
> > > > BUT
> > > > > > NOT
> > > > > > > + *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
> > AND
> > > > > > FITNESS FOR
> > > > > > > + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL
> > THE
> > > > > > COPYRIGHT
> > > > > > > + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
> > INDIRECT,
> > > > > > INCIDENTAL,
> > > > > > > + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
> > > > (INCLUDING,
> > > > > > BUT NOT
> > > > > > > + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
> > SERVICES;
> > > > > > LOSS OF USE,
> > > > > > > + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
> > > > CAUSED
> > > > > > AND ON ANY
> > > > > > > + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
> > LIABILITY,
> > > > OR
> > > > > > TORT
> > > > > > > + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY
> > WAY
> > > > OUT
> > > > > > OF THE USE
> > > > > > > + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
> > SUCH
> > > > > > DAMAGE.
> > > > > > > + */
> > > > > > > +
> > > > > > > +#ifndef _RTE_ETH_DEV_INFO_H_
> > > > > > > +#define _RTE_ETH_DEV_INFO_H_
> > > > > > > +
> > > > > > > +
> > > > > > > +/*
> > > > > > > + * Placeholder for accessing device registers  */ struct
> > > > > > > +rte_dev_reg_info {
> > > > > > > +	void *buf; /**< Buffer for register */
> > > > > > > +	uint32_t offset; /**< Offset for 1st register to fetch */
> > > > > > > +	uint32_t leng; /**< Number of registers to fetch */
> > > > > > > +	uint32_t version; /**< Device version */ };
> > > > > > > +
> > > > > > > +/*
> > > > > > > + * Placeholder for accessing device eeprom  */ struct
> > > > > > > +rte_dev_eeprom_info {
> > > > > > > +	void *buf; /**< Buffer for eeprom */
> > > > > > > +	uint32_t offset; /**< Offset for 1st eeprom location to access
> > */
> > > > > > > +	uint32_t leng; /**< Length of eeprom region to access */
> > > > > > > +	uint32_t magic; /**< Device ID */ };
> > > > > > > +
> > > > > > > +/*
> > > > > > > + * Placeholder for accessing device ring parameters  */
> > > > > > > +struct rte_dev_ring_info {
> > > > > > > +	uint32_t rx_pending; /**< Number of outstanding Rx ring */
> > > > > > > +	uint32_t tx_pending; /**< Number of outstanding Tx ring */
> > > > > > > +	uint32_t rx_max_pending; /**< Maximum number of
> > outstanding
> > > > > > > +Rx
> > > > > > ring */
> > > > > > > +	uint32_t tx_max_pending; /**< Maximum number of
> > outstanding
> > > > > > > +Tx
> > > > > > ring
> > > > > > > +*/ };
> > > > > > > +
> > > > > > > +/*
> > > > > > > + * A data structure captures information as defined in struct
> > > > > > > +ifla_vf_info
> > > > > > > + * for user-space api
> > > > > > > + */
> > > > > > > +struct rte_dev_vf_info {
> > > > > > > +	uint32_t vf;
> > > > > > > +	uint8_t mac[ETHER_ADDR_LEN];
> > > > > > > +	uint32_t vlan;
> > > > > > > +	uint32_t tx_rate;
> > > > > > > +	uint32_t spoofchk;
> > > > > > > +};
> > > > > >
> > > > > >
> > > > > > Wonder what that structure is for?
> > > > > > I can't see it used in any function below?
> > > > > >
> > > > >
> > > > > Good catch, this is designed for other ethtool ops that I did not
> > > > > include in
> > > > this release, I will remove this from next fix.
> > > > >
> > > > > > > +
> > > > > > > +#endif /* _RTE_ETH_DEV_INFO_H_ */
> > > > > > > diff --git a/lib/librte_ether/rte_ethdev.c
> > > > > > > b/lib/librte_ether/rte_ethdev.c index 5a94654..186e85c 100644
> > > > > > > --- a/lib/librte_ether/rte_ethdev.c
> > > > > > > +++ b/lib/librte_ether/rte_ethdev.c
> > > > > > > @@ -2751,6 +2751,32 @@ rte_eth_dev_mac_addr_remove(uint8_t
> > > > > > port_id,
> > > > > > > struct ether_addr *addr)  }
> > > > > > >
> > > > > > >  int
> > > > > > > +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct
> > > > > > > +ether_addr
> > > > > > > +*addr) {
> > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > +	const int index = 0;
> > > > > > > +	const uint32_t pool = 0;
> > > > > > > +
> > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > port_id);
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	dev = &rte_eth_devices[port_id];
> > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> > >mac_addr_remove, -
> > > > > > ENOTSUP);
> > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -
> > > > > > ENOTSUP);
> > > > > > > +
> > > > > > > +	/* Update NIC default MAC address*/
> > > > > > > +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> > > > > > > +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> > > > > > > +
> > > > > > > +	/* Update default address in NIC data structure */
> > > > > > > +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> > > > > > > +
> > > > > > > +	return 0;
> > > > > > > +}
> > > > > > > +
> > > > > > > +int
> > > > > > >  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
> > > > > > >  				uint16_t rx_mode, uint8_t on)  { @@
> > -3627,3
> > > > +3653,136 @@
> > > > > > > rte_eth_remove_tx_callback(uint8_t port_id,
> > > > > > uint16_t queue_id,
> > > > > > >  	/* Callback wasn't found. */
> > > > > > >  	return -EINVAL;
> > > > > > >  }
> > > > > > > +
> > > > > > > +int
> > > > > > > +rte_eth_dev_reg_leng(uint8_t port_id) {
> > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > +
> > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > port_id);
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length,
> > -
> > > > > > ENOTSUP);
> > > > > > > +	return (*dev->dev_ops->get_reg_length)(dev);
> > > > > > > +}
> > > > > > > +
> > > > > > > +int
> > > > > > > +rte_eth_dev_reg_info(uint8_t port_id, struct rte_dev_reg_info
> > > > > > > +*info) {
> > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > +
> > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > port_id);
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -
> > ENOTSUP);
> > > > > > > +	return (*dev->dev_ops->get_reg)(dev, info); }
> > > > > >
> > > > > > Seems that *get_reg* stuff, will be really good addition for
> > > > > > DPDK debugging abilities.
> > > > > > Though, I'd suggest we change it a bit to make more generic and
> > flexible:
> > > > > >
> > > > > > Introduce rte_eth_reg_read/rte_eth_reg_write(),
> > > > > > or probably even better rte_pcidev_reg_read
> > > > > > /rte_pcidev_reg_write at
> > > > EAL.
> > > > > > Something similar to what port_pci_reg_read/port_pci_reg_write()
> > > > > > are doing now at testpmd.h.
> > > > > >
> > > > > > struct rte_pcidev_reg_info {
> > > > > >    const char *name;
> > > > > >    uint32_t endianes, bar, offset, size, count; };
> > > > > >
> > > > > > int rte_pcidev_reg_read(const struct rte_pci_device *, const
> > > > > > struct rte_pcidev_reg_info *, uint64_t *reg_val);
> > > > > >
> > > > > > Then:
> > > > > > int rte_eth_dev_get_reg_info(port_id, const struct
> > > > > > rte_pcidev_reg_info **info);
> > > > > >
> > > > > > So each device would store in info a pointer to an array of it's
> > > > > > register descriptions (finished by zero elem?).
> > > > > >
> > > > > > Then your ethtool (or any other upper layer) can do the
> > > > > > following to read all device regs:
> > > > > >
> > > > >
> > > > > The proposed reg info structure allows future improvement to
> > > > > support
> > > > individual register read/write.
> > > > > Also, because each NIC device has a very distinguish register definition.
> > > > > So, the plan is to have more comprehensive interface to support
> > > > > query operation (for example, register name) before introduce
> > > > > individual/group
> > > > register access.
> > > > > Points taken, the support will be in future release.
> > > >
> > > > Sorry, didn't get you.
> > > > So you are ok to make these changes in next patch version?
> > > >
> > > I would like to get a consensus from dpdk community on how to provide
> > register information.
> >
> > Well, that' ok, but if it is just a trial patch that is not intended to be applied,
> > then you should mark it as RFC.
> >
> > > Currently, it's designed for debug dumping. The register information is very
> > hardware dependent.
> > > Need to consider current supported NIC device and future devices for
> > DPDK, so we won't make it a bulky interface.
> >
> > Ok, could you explain what exactly  concerns you in the approach described
> > above?
> > What part you feel is bulky?
> >
> > > > >
> > > > > > const struct rte_eth_dev_reg_info *reg_info; struct
> > > > > > rte_eth_dev_info dev_info;
> > > > > >
> > > > > > rte_eth_dev_info_get(pid, &dev_info);
> > > > > > rte_eth_dev_get_reg_info(port_id, &reg_info);
> > > > > >
> > > > > > for (i = 0; reg_info[i].name != NULL; i++) {
> > > > > >    ...
> > > > > >    rte_pcidev_read_reg(dev_info. pci_dev, reg_info[i], &v);
> > > > > >   ..
> > > > > > }
> > > > > >
> > > > > > > +
> > > > > > > +int
> > > > > > > +rte_eth_dev_eeprom_leng(uint8_t port_id) {
> > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > +
> > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > port_id);
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> > >get_eeprom_length, -
> > > > > > ENOTSUP);
> > > > > > > +	return (*dev->dev_ops->get_eeprom_length)(dev);
> > > > > > > +}
> > > > > > > +
> > > > > > > +int
> > > > > > > +rte_eth_dev_get_eeprom(uint8_t port_id, struct
> > > > > > > +rte_dev_eeprom_info
> > > > > > > +*info) {
> > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > +
> > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > port_id);
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -
> > > > > > ENOTSUP);
> > > > > > > +	return (*dev->dev_ops->get_eeprom)(dev, info); }
> > > > > > > +
> > > > > > > +int
> > > > > > > +rte_eth_dev_set_eeprom(uint8_t port_id, struct
> > > > > > > +rte_dev_eeprom_info
> > > > > > > +*info) {
> > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > +
> > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > port_id);
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -
> > > > > > ENOTSUP);
> > > > > > > +	return (*dev->dev_ops->set_eeprom)(dev, info); }
> > > > > > > +
> > > > > > > +int
> > > > > > > +rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > > > > > +rte_dev_ring_info
> > > > > > > +*info) {
> > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > +
> > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > port_id);
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > +		return -ENODEV;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> > > > > > ENOTSUP);
> > > > > > > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> > > > > >
> > > > > > I think it will be a useful addition to the ethdev API  to have
> > > > > > an ability to retrieve current RX/TX queue parameters.
> > > > > > Though again, it need to be more generic, so it could be useful
> > > > > > for
> > > > > > non- ethtool upper layer too.
> > > > > > So I suggest to modify it a bit.
> > > > > > Something like that:
> > > > > >
> > > > > > struct rte_eth_tx_queue_info {
> > > > > >     struct rte_eth_txconf txconf;
> > > > > >     uint32_t nb_tx_desc;
> > > > > >     uint32_t nb_max_tx_desc; /*max allowable TXDs for that queue */
> > > > > >     uint32_t nb_tx_free;            /* number of free TXDs at the moment
> > of
> > > > call.
> > > > > > */
> > > > > >     /* other tx queue data. */
> > > > > > };
> > > > > >
> > > > > > int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> > > > > > rte_eth_tx_queue_info *qinfo)
> > > > > >
> > > > > > Then, your upper layer ethtool wrapper, can implement yours
> > > > > > ethtool_get_ringparam() by:
> > > > > >
> > > > > >  ...
> > > > > >  struct rte_eth_tx_queue_info qinfo;
> > > > > > rte_ethdev_get_tx_queue_info(port, 0, &qinfo);
> > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > rte_eth_rx_queue_count(port, 0);
> > > > > >
> > > > > > Or probably even:
> > > > > > ring_param->tx_pending = qinfo.nb_tx_desc - qinfo.nb_tx_free;
> > > > > >
> > > > > > Same for RX.
> > > > > >
> > > > > For now, this descriptor ring information is used by the ethtool op.
> > > > > To make this interface simple, i.e. caller doesn't need to access
> > > > > other
> > > > queue information.
> > > >
> > > > I just repeat what I said to you in off-line conversation:
> > > > ethdev API is not equal ethtool API.
> > > > It is ok to add  a new function/structure to ethdev if it really
> > > > needed, but we should do mechanical one to one copy.
> > > > It is much better to add  a function/structure that would be more
> > > > generic, and suit other users, not only ethtool.
> > > > There is no point to have dozen functions in rte_ethdev API
> > > > providing similar information.
> > > > BTW, I don't see how API I proposed is much more  complex, then yours
> > one.
> > > The ring parameter is a run-time information which is different than data
> > structure described in this discussion.
> >
> > I don't see how they are different.
> > Looking at ixgbe_get_ringparam(), it returns:
> > rx_max_pending - that's a static IXGBE PMD value (max possible number of
> > RXDs per one queue).
> > rx_pending - number of RXD currently in use by the HW for queue 0 (that
> > information can be changed at each call).
> >
> > With the approach I suggesting - you can get same information for each RX
> > queue by calling rte_ethdev_get_rx_queue_info() and
> > rte_eth_rx_queue_count().
> > Plus you are getting other RX queue data.
> >
> > Another thing - what is practical usage of the information you retrieving now
> > by get_ringparam()?
> > Let say it returned to you: rx_max_pending=4096; rx_pending=128; How that
> > information would help to understand what is going on with the device?
> > Without knowing  value of nb_tx_desc for the queue, you can't say is you
> > queue full or not.
> > Again, it could be that all your traffic going through some other queue (not
> > 0).
> > So from my point rte_eth_dev_get_ringparam()  usage is very limited, and
> > doesn't provide enough information about current queue state.
> >
> > Same thing applies for TX.
> >
> 
> After careful review the suggestion in this comment, and review the existing dpdk source code.
> I came to realize that neither rte_ethdev_get_rx_queue_info, rte_ethdev_get_tx_queue_info, struct rte_eth_rx_queue_info and
> struct rte_eth_tx_queue_info are available in the existing dpdk source code. I could not make a patch based upon a set of non-
> existent API and data structure.

Right now, in dpdk.org source code, struct  rte_eth_dev_ring_info, struct  rte_dev_eeprom_info and struct  rte_dev_reg_info
don't exist also.
Same as  all these functions:

rte_eth_dev_default_mac_addr_set
rte_eth_dev_reg_length
rte_eth_dev_reg_info
rte_eth_dev_eeprom_length
rte_eth_dev_get_eeprom
rte_eth_dev_set_eeprom
rte_eth_dev_get_ringparam

All this is a new API that's you are trying to add.
But, by some reason you consider it is ok to add 'struct  rte_eth_dev_ring_info', but couldn't add  struct 'rte_ethdev_get_tx_queue_info'
That just doesn't make any sense to me.
In fact, I think our conversation is going in cycles.
If you are not happy with the suggested approach, please provide some meaningful reason why.
Konstantin

> 
> > > It's the desire of this patch to separate each data structure to avoid cross
> > dependency.
> >
> > That's too cryptic to me.
> > Could you explain what cross dependency you are talking about?
> >
> > Konstantin
  

> -----Original Message-----
> From: Ananyev, Konstantin
> Sent: Monday, June 15, 2015 9:46 AM
> To: Wang, Liang-min; dev@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> device info
> 
> 
> 
> > -----Original Message-----
> > From: Wang, Liang-min
> > Sent: Monday, June 15, 2015 2:26 PM
> > To: Ananyev, Konstantin; dev@dpdk.org
> > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > access device info
> >
> >
> >
> > > -----Original Message-----
> > > From: Ananyev, Konstantin
> > > Sent: Friday, June 12, 2015 8:31 AM
> > > To: Wang, Liang-min; dev@dpdk.org
> > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > access device info
> > >
> > >
> > >
> > > > -----Original Message-----
> > > > From: Wang, Liang-min
> > > > Sent: Thursday, June 11, 2015 10:51 PM
> > > > To: Ananyev, Konstantin; dev@dpdk.org
> > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > access device info
> > > >
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Ananyev, Konstantin
> > > > > Sent: Thursday, June 11, 2015 9:07 AM
> > > > > To: Wang, Liang-min; dev@dpdk.org
> > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > support access device info
> > > > >
> > > > >
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: Wang, Liang-min
> > > > > > Sent: Thursday, June 11, 2015 1:58 PM
> > > > > > To: Ananyev, Konstantin; dev@dpdk.org
> > > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > support access device info
> > > > > >
> > > > > > Hi Konstantin,
> > > > > >
> > > > > > > -----Original Message-----
> > > > > > > From: Ananyev, Konstantin
> > > > > > > Sent: Thursday, June 11, 2015 8:26 AM
> > > > > > > To: Wang, Liang-min; dev@dpdk.org
> > > > > > > Cc: Wang, Liang-min
> > > > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > > support access device info
> > > > > > >
> > > > > > > Hi Larry,
> > > > > > >
> > > > > > > > -----Original Message-----
> > > > > > > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of
> > > > > > > > Liang-Min Larry Wang
> > > > > > > > Sent: Wednesday, June 10, 2015 4:10 PM
> > > > > > > > To: dev@dpdk.org
> > > > > > > > Cc: Wang, Liang-min
> > > > > > > > Subject: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > > > support access device info
> > > > > > > >
> > > > > > > > add new apis:
> > > > > > > > - rte_eth_dev_default_mac_addr_set
> > > > > > > > - rte_eth_dev_reg_leng
> > > > > > > > - rte_eth_dev_reg_info
> > > > > > > > - rte_eth_dev_eeprom_leng
> > > > > > > > - rte_eth_dev_get_eeprom
> > > > > > > > - rte_eth_dev_set_eeprom
> > > > > > > > - rte_eth_dev_get_ringparam
> > > > > > > > - rte_eth_dev_set_ringparam
> > > > > > > >
> > > > > > > > to enable reading device parameters (mac-addr, register,
> > > > > > > > eeprom,
> > > > > > > > ring) based upon ethtool alike data parameter specification.
> > > > > > > >
> > > > > > > > Signed-off-by: Liang-Min Larry Wang
> > > > > > > > <liang-min.wang@intel.com>
> > > > > > > > ---
> > > > > > > >  lib/librte_ether/Makefile              |   1 +
> > > > > > > >  lib/librte_ether/rte_eth_dev_info.h    |  80
> +++++++++++++++++
> > > > > > > >  lib/librte_ether/rte_ethdev.c          | 159
> > > > > > > +++++++++++++++++++++++++++++++++
> > > > > > > >  lib/librte_ether/rte_ethdev.h          | 158
> > > > > > > ++++++++++++++++++++++++++++++++
> > > > > > > >  lib/librte_ether/rte_ether_version.map |   8 ++
> > > > > > > >  5 files changed, 406 insertions(+)  create mode 100644
> > > > > > > > lib/librte_ether/rte_eth_dev_info.h
> > > > > > > >
> > > > > > > > diff --git a/lib/librte_ether/Makefile
> > > > > > > > b/lib/librte_ether/Makefile index c0e5768..05209e9 100644
> > > > > > > > --- a/lib/librte_ether/Makefile
> > > > > > > > +++ b/lib/librte_ether/Makefile
> > > > > > > > @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c
> > > > > > > > SYMLINK-y-include += rte_ether.h  SYMLINK-y-include +=
> > > > > > > > rte_ethdev.h SYMLINK-y-include
> > > > > > > > += rte_eth_ctrl.h
> > > > > > > > +SYMLINK-y-include += rte_eth_dev_info.h
> > > > > > > >
> > > > > > > >  # this lib depends upon:
> > > > > > > >  DEPDIRS-y += lib/librte_eal lib/librte_mempool
> > > > > > > > lib/librte_ring lib/librte_mbuf diff --git
> > > > > > > > a/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > new file mode 100644
> > > > > > > > index 0000000..002c4b5
> > > > > > > > --- /dev/null
> > > > > > > > +++ b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > @@ -0,0 +1,80 @@
> > > > > > > > +/*-
> > > > > > > > + *   BSD LICENSE
> > > > > > > > + *
> > > > > > > > + *   Copyright(c) 2015 Intel Corporation. All rights reserved.
> > > > > > > > + *   All rights reserved.
> > > > > > > > + *
> > > > > > > > + *   Redistribution and use in source and binary forms, with or
> > > without
> > > > > > > > + *   modification, are permitted provided that the following
> > > conditions
> > > > > > > > + *   are met:
> > > > > > > > + *
> > > > > > > > + *     * Redistributions of source code must retain the above
> > > copyright
> > > > > > > > + *       notice, this list of conditions and the following disclaimer.
> > > > > > > > + *     * Redistributions in binary form must reproduce the above
> > > > > copyright
> > > > > > > > + *       notice, this list of conditions and the following disclaimer
> in
> > > > > > > > + *       the documentation and/or other materials provided with
> the
> > > > > > > > + *       distribution.
> > > > > > > > + *     * Neither the name of Intel Corporation nor the names of
> its
> > > > > > > > + *       contributors may be used to endorse or promote
> products
> > > > > derived
> > > > > > > > + *       from this software without specific prior written
> permission.
> > > > > > > > + *
> > > > > > > > + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS
> AND
> > > > > > > CONTRIBUTORS
> > > > > > > > + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
> > > INCLUDING,
> > > > > BUT
> > > > > > > NOT
> > > > > > > > + *   LIMITED TO, THE IMPLIED WARRANTIES OF
> MERCHANTABILITY
> > > AND
> > > > > > > FITNESS FOR
> > > > > > > > + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT
> SHALL
> > > THE
> > > > > > > COPYRIGHT
> > > > > > > > + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
> > > INDIRECT,
> > > > > > > INCIDENTAL,
> > > > > > > > + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
> > > > > (INCLUDING,
> > > > > > > BUT NOT
> > > > > > > > + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
> > > SERVICES;
> > > > > > > LOSS OF USE,
> > > > > > > > + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
> HOWEVER
> > > > > CAUSED
> > > > > > > AND ON ANY
> > > > > > > > + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
> > > LIABILITY,
> > > > > OR
> > > > > > > TORT
> > > > > > > > + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY
> > > WAY
> > > > > OUT
> > > > > > > OF THE USE
> > > > > > > > + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY
> OF
> > > SUCH
> > > > > > > DAMAGE.
> > > > > > > > + */
> > > > > > > > +
> > > > > > > > +#ifndef _RTE_ETH_DEV_INFO_H_ #define
> _RTE_ETH_DEV_INFO_H_
> > > > > > > > +
> > > > > > > > +
> > > > > > > > +/*
> > > > > > > > + * Placeholder for accessing device registers  */ struct
> > > > > > > > +rte_dev_reg_info {
> > > > > > > > +	void *buf; /**< Buffer for register */
> > > > > > > > +	uint32_t offset; /**< Offset for 1st register to fetch */
> > > > > > > > +	uint32_t leng; /**< Number of registers to fetch */
> > > > > > > > +	uint32_t version; /**< Device version */ };
> > > > > > > > +
> > > > > > > > +/*
> > > > > > > > + * Placeholder for accessing device eeprom  */ struct
> > > > > > > > +rte_dev_eeprom_info {
> > > > > > > > +	void *buf; /**< Buffer for eeprom */
> > > > > > > > +	uint32_t offset; /**< Offset for 1st eeprom location to
> > > > > > > > +access
> > > */
> > > > > > > > +	uint32_t leng; /**< Length of eeprom region to access */
> > > > > > > > +	uint32_t magic; /**< Device ID */ };
> > > > > > > > +
> > > > > > > > +/*
> > > > > > > > + * Placeholder for accessing device ring parameters  */
> > > > > > > > +struct rte_dev_ring_info {
> > > > > > > > +	uint32_t rx_pending; /**< Number of outstanding Rx ring */
> > > > > > > > +	uint32_t tx_pending; /**< Number of outstanding Tx ring */
> > > > > > > > +	uint32_t rx_max_pending; /**< Maximum number of
> > > outstanding
> > > > > > > > +Rx
> > > > > > > ring */
> > > > > > > > +	uint32_t tx_max_pending; /**< Maximum number of
> > > outstanding
> > > > > > > > +Tx
> > > > > > > ring
> > > > > > > > +*/ };
> > > > > > > > +
> > > > > > > > +/*
> > > > > > > > + * A data structure captures information as defined in
> > > > > > > > +struct ifla_vf_info
> > > > > > > > + * for user-space api
> > > > > > > > + */
> > > > > > > > +struct rte_dev_vf_info {
> > > > > > > > +	uint32_t vf;
> > > > > > > > +	uint8_t mac[ETHER_ADDR_LEN];
> > > > > > > > +	uint32_t vlan;
> > > > > > > > +	uint32_t tx_rate;
> > > > > > > > +	uint32_t spoofchk;
> > > > > > > > +};
> > > > > > >
> > > > > > >
> > > > > > > Wonder what that structure is for?
> > > > > > > I can't see it used in any function below?
> > > > > > >
> > > > > >
> > > > > > Good catch, this is designed for other ethtool ops that I did
> > > > > > not include in
> > > > > this release, I will remove this from next fix.
> > > > > >
> > > > > > > > +
> > > > > > > > +#endif /* _RTE_ETH_DEV_INFO_H_ */
> > > > > > > > diff --git a/lib/librte_ether/rte_ethdev.c
> > > > > > > > b/lib/librte_ether/rte_ethdev.c index 5a94654..186e85c
> > > > > > > > 100644
> > > > > > > > --- a/lib/librte_ether/rte_ethdev.c
> > > > > > > > +++ b/lib/librte_ether/rte_ethdev.c
> > > > > > > > @@ -2751,6 +2751,32 @@
> rte_eth_dev_mac_addr_remove(uint8_t
> > > > > > > port_id,
> > > > > > > > struct ether_addr *addr)  }
> > > > > > > >
> > > > > > > >  int
> > > > > > > > +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct
> > > > > > > > +ether_addr
> > > > > > > > +*addr) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +	const int index = 0;
> > > > > > > > +	const uint32_t pool = 0;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	dev = &rte_eth_devices[port_id];
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> > > >mac_addr_remove, -
> > > > > > > ENOTSUP);
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -
> > > > > > > ENOTSUP);
> > > > > > > > +
> > > > > > > > +	/* Update NIC default MAC address*/
> > > > > > > > +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> > > > > > > > +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> > > > > > > > +
> > > > > > > > +	/* Update default address in NIC data structure */
> > > > > > > > +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> > > > > > > > +
> > > > > > > > +	return 0;
> > > > > > > > +}
> > > > > > > > +
> > > > > > > > +int
> > > > > > > >  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
> > > > > > > >  				uint16_t rx_mode, uint8_t on)  { @@
> > > -3627,3
> > > > > +3653,136 @@
> > > > > > > > rte_eth_remove_tx_callback(uint8_t port_id,
> > > > > > > uint16_t queue_id,
> > > > > > > >  	/* Callback wasn't found. */
> > > > > > > >  	return -EINVAL;
> > > > > > > >  }
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_reg_leng(uint8_t port_id) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length,
> > > -
> > > > > > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->get_reg_length)(dev);
> > > > > > > > +}
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_reg_info(uint8_t port_id, struct
> > > > > > > > +rte_dev_reg_info
> > > > > > > > +*info) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -
> > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->get_reg)(dev, info); }
> > > > > > >
> > > > > > > Seems that *get_reg* stuff, will be really good addition for
> > > > > > > DPDK debugging abilities.
> > > > > > > Though, I'd suggest we change it a bit to make more generic
> > > > > > > and
> > > flexible:
> > > > > > >
> > > > > > > Introduce rte_eth_reg_read/rte_eth_reg_write(),
> > > > > > > or probably even better rte_pcidev_reg_read
> > > > > > > /rte_pcidev_reg_write at
> > > > > EAL.
> > > > > > > Something similar to what
> > > > > > > port_pci_reg_read/port_pci_reg_write()
> > > > > > > are doing now at testpmd.h.
> > > > > > >
> > > > > > > struct rte_pcidev_reg_info {
> > > > > > >    const char *name;
> > > > > > >    uint32_t endianes, bar, offset, size, count; };
> > > > > > >
> > > > > > > int rte_pcidev_reg_read(const struct rte_pci_device *, const
> > > > > > > struct rte_pcidev_reg_info *, uint64_t *reg_val);
> > > > > > >
> > > > > > > Then:
> > > > > > > int rte_eth_dev_get_reg_info(port_id, const struct
> > > > > > > rte_pcidev_reg_info **info);
> > > > > > >
> > > > > > > So each device would store in info a pointer to an array of
> > > > > > > it's register descriptions (finished by zero elem?).
> > > > > > >
> > > > > > > Then your ethtool (or any other upper layer) can do the
> > > > > > > following to read all device regs:
> > > > > > >
> > > > > >
> > > > > > The proposed reg info structure allows future improvement to
> > > > > > support
> > > > > individual register read/write.
> > > > > > Also, because each NIC device has a very distinguish register
> definition.
> > > > > > So, the plan is to have more comprehensive interface to
> > > > > > support query operation (for example, register name) before
> > > > > > introduce individual/group
> > > > > register access.
> > > > > > Points taken, the support will be in future release.
> > > > >
> > > > > Sorry, didn't get you.
> > > > > So you are ok to make these changes in next patch version?
> > > > >
> > > > I would like to get a consensus from dpdk community on how to
> > > > provide
> > > register information.
> > >
> > > Well, that' ok, but if it is just a trial patch that is not intended
> > > to be applied, then you should mark it as RFC.
> > >
> > > > Currently, it's designed for debug dumping. The register
> > > > information is very
> > > hardware dependent.
> > > > Need to consider current supported NIC device and future devices
> > > > for
> > > DPDK, so we won't make it a bulky interface.
> > >
> > > Ok, could you explain what exactly  concerns you in the approach
> > > described above?
> > > What part you feel is bulky?
> > >
> > > > > >
> > > > > > > const struct rte_eth_dev_reg_info *reg_info; struct
> > > > > > > rte_eth_dev_info dev_info;
> > > > > > >
> > > > > > > rte_eth_dev_info_get(pid, &dev_info);
> > > > > > > rte_eth_dev_get_reg_info(port_id, &reg_info);
> > > > > > >
> > > > > > > for (i = 0; reg_info[i].name != NULL; i++) {
> > > > > > >    ...
> > > > > > >    rte_pcidev_read_reg(dev_info. pci_dev, reg_info[i], &v);
> > > > > > >   ..
> > > > > > > }
> > > > > > >
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_eeprom_leng(uint8_t port_id) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> > > >get_eeprom_length, -
> > > > > > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->get_eeprom_length)(dev);
> > > > > > > > +}
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_get_eeprom(uint8_t port_id, struct
> > > > > > > > +rte_dev_eeprom_info
> > > > > > > > +*info) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -
> > > > > > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->get_eeprom)(dev, info); }
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_set_eeprom(uint8_t port_id, struct
> > > > > > > > +rte_dev_eeprom_info
> > > > > > > > +*info) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -
> > > > > > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->set_eeprom)(dev, info); }
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > > > > > > +rte_dev_ring_info
> > > > > > > > +*info) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> > > > > > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> > > > > > >
> > > > > > > I think it will be a useful addition to the ethdev API  to
> > > > > > > have an ability to retrieve current RX/TX queue parameters.
> > > > > > > Though again, it need to be more generic, so it could be
> > > > > > > useful for
> > > > > > > non- ethtool upper layer too.
> > > > > > > So I suggest to modify it a bit.
> > > > > > > Something like that:
> > > > > > >
> > > > > > > struct rte_eth_tx_queue_info {
> > > > > > >     struct rte_eth_txconf txconf;
> > > > > > >     uint32_t nb_tx_desc;
> > > > > > >     uint32_t nb_max_tx_desc; /*max allowable TXDs for that
> queue */
> > > > > > >     uint32_t nb_tx_free;            /* number of free TXDs at the
> moment
> > > of
> > > > > call.
> > > > > > > */
> > > > > > >     /* other tx queue data. */ };
> > > > > > >
> > > > > > > int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> > > > > > > rte_eth_tx_queue_info *qinfo)
> > > > > > >
> > > > > > > Then, your upper layer ethtool wrapper, can implement yours
> > > > > > > ethtool_get_ringparam() by:
> > > > > > >
> > > > > > >  ...
> > > > > > >  struct rte_eth_tx_queue_info qinfo;
> > > > > > > rte_ethdev_get_tx_queue_info(port, 0, &qinfo);
> > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > rte_eth_rx_queue_count(port, 0);
> > > > > > >
> > > > > > > Or probably even:
> > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > qinfo.nb_tx_free;
> > > > > > >
> > > > > > > Same for RX.
> > > > > > >
> > > > > > For now, this descriptor ring information is used by the ethtool op.
> > > > > > To make this interface simple, i.e. caller doesn't need to
> > > > > > access other
> > > > > queue information.
> > > > >
> > > > > I just repeat what I said to you in off-line conversation:
> > > > > ethdev API is not equal ethtool API.
> > > > > It is ok to add  a new function/structure to ethdev if it really
> > > > > needed, but we should do mechanical one to one copy.
> > > > > It is much better to add  a function/structure that would be
> > > > > more generic, and suit other users, not only ethtool.
> > > > > There is no point to have dozen functions in rte_ethdev API
> > > > > providing similar information.
> > > > > BTW, I don't see how API I proposed is much more  complex, then
> > > > > yours
> > > one.
> > > > The ring parameter is a run-time information which is different
> > > > than data
> > > structure described in this discussion.
> > >
> > > I don't see how they are different.
> > > Looking at ixgbe_get_ringparam(), it returns:
> > > rx_max_pending - that's a static IXGBE PMD value (max possible
> > > number of RXDs per one queue).
> > > rx_pending - number of RXD currently in use by the HW for queue 0
> > > (that information can be changed at each call).
> > >
> > > With the approach I suggesting - you can get same information for
> > > each RX queue by calling rte_ethdev_get_rx_queue_info() and
> > > rte_eth_rx_queue_count().
> > > Plus you are getting other RX queue data.
> > >
> > > Another thing - what is practical usage of the information you
> > > retrieving now by get_ringparam()?
> > > Let say it returned to you: rx_max_pending=4096; rx_pending=128; How
> > > that information would help to understand what is going on with the
> device?
> > > Without knowing  value of nb_tx_desc for the queue, you can't say is
> > > you queue full or not.
> > > Again, it could be that all your traffic going through some other
> > > queue (not 0).
> > > So from my point rte_eth_dev_get_ringparam()  usage is very limited,
> > > and doesn't provide enough information about current queue state.
> > >
> > > Same thing applies for TX.
> > >
> >
> > After careful review the suggestion in this comment, and review the
> existing dpdk source code.
> > I came to realize that neither rte_ethdev_get_rx_queue_info,
> > rte_ethdev_get_tx_queue_info, struct rte_eth_rx_queue_info and struct
> > rte_eth_tx_queue_info are available in the existing dpdk source code. I
> could not make a patch based upon a set of non- existent API and data
> structure.
> 
> Right now, in dpdk.org source code, struct  rte_eth_dev_ring_info, struct
> rte_dev_eeprom_info and struct  rte_dev_reg_info don't exist also.
> Same as  all these functions:
> 
> rte_eth_dev_default_mac_addr_set
> rte_eth_dev_reg_length
> rte_eth_dev_reg_info
> rte_eth_dev_eeprom_length
> rte_eth_dev_get_eeprom
> rte_eth_dev_set_eeprom
> rte_eth_dev_get_ringparam
> 
> All this is a new API that's you are trying to add.
> But, by some reason you consider it is ok to add 'struct
> rte_eth_dev_ring_info', but couldn't add  struct
> 'rte_ethdev_get_tx_queue_info'
> That just doesn't make any sense to me.
> In fact, I think our conversation is going in cycles.
> If you are not happy with the suggested approach, please provide some
> meaningful reason why.

All the API's and data structure that you have questions in this comment are available from the submitted patch, you could run to see if it works.
What I learned from your comments are a couple of API name and incomplete data structure description, so I could not make my patch according to your suggestion.
If you strongly feel the API's that you are proposing are useful for get_ringparam(), please create a patch and submit it for evaluation.
As coded in this patch, it's a simple and clean interface and most important, it can be validated and it works.

> Konstantin
> 
> >
> > > > It's the desire of this patch to separate each data structure to
> > > > avoid cross
> > > dependency.
> > >
> > > That's too cryptic to me.
> > > Could you explain what cross dependency you are talking about?
> > >
> > > Konstantin
  

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Ananyev, Konstantin
> Sent: Monday, June 15, 2015 9:46 AM
> To: Wang, Liang-min; dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> device info
> 
> 
> 
> > -----Original Message-----
> > From: Wang, Liang-min
> > Sent: Monday, June 15, 2015 2:26 PM
> > To: Ananyev, Konstantin; dev@dpdk.org
> > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > access device info
> >
> >
> >
> > > -----Original Message-----
> > > From: Ananyev, Konstantin
> > > Sent: Friday, June 12, 2015 8:31 AM
> > > To: Wang, Liang-min; dev@dpdk.org
> > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > access device info
> > >
> > >
> > >
> > > > -----Original Message-----
> > > > From: Wang, Liang-min
> > > > Sent: Thursday, June 11, 2015 10:51 PM
> > > > To: Ananyev, Konstantin; dev@dpdk.org
> > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > access device info
> > > >
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Ananyev, Konstantin
> > > > > Sent: Thursday, June 11, 2015 9:07 AM
> > > > > To: Wang, Liang-min; dev@dpdk.org
> > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > support access device info
> > > > >
> > > > >
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: Wang, Liang-min
> > > > > > Sent: Thursday, June 11, 2015 1:58 PM
> > > > > > To: Ananyev, Konstantin; dev@dpdk.org
> > > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > support access device info
> > > > > >
> > > > > > Hi Konstantin,
> > > > > >
> > > > > > > -----Original Message-----
> > > > > > > From: Ananyev, Konstantin
> > > > > > > Sent: Thursday, June 11, 2015 8:26 AM
> > > > > > > To: Wang, Liang-min; dev@dpdk.org
> > > > > > > Cc: Wang, Liang-min
> > > > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > > support access device info
> > > > > > >
> > > > > > > Hi Larry,
> > > > > > >
> > > > > > > > -----Original Message-----
> > > > > > > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of
> > > > > > > > Liang-Min Larry Wang
> > > > > > > > Sent: Wednesday, June 10, 2015 4:10 PM
> > > > > > > > To: dev@dpdk.org
> > > > > > > > Cc: Wang, Liang-min
> > > > > > > > Subject: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > > > support access device info
> > > > > > > >
> > > > > > > > add new apis:
> > > > > > > > - rte_eth_dev_default_mac_addr_set
> > > > > > > > - rte_eth_dev_reg_leng
> > > > > > > > - rte_eth_dev_reg_info
> > > > > > > > - rte_eth_dev_eeprom_leng
> > > > > > > > - rte_eth_dev_get_eeprom
> > > > > > > > - rte_eth_dev_set_eeprom
> > > > > > > > - rte_eth_dev_get_ringparam
> > > > > > > > - rte_eth_dev_set_ringparam
> > > > > > > >
> > > > > > > > to enable reading device parameters (mac-addr, register,
> > > > > > > > eeprom,
> > > > > > > > ring) based upon ethtool alike data parameter specification.
> > > > > > > >
> > > > > > > > Signed-off-by: Liang-Min Larry Wang
> > > > > > > > <liang-min.wang@intel.com>
> > > > > > > > ---
> > > > > > > >  lib/librte_ether/Makefile              |   1 +
> > > > > > > >  lib/librte_ether/rte_eth_dev_info.h    |  80
> +++++++++++++++++
> > > > > > > >  lib/librte_ether/rte_ethdev.c          | 159
> > > > > > > +++++++++++++++++++++++++++++++++
> > > > > > > >  lib/librte_ether/rte_ethdev.h          | 158
> > > > > > > ++++++++++++++++++++++++++++++++
> > > > > > > >  lib/librte_ether/rte_ether_version.map |   8 ++
> > > > > > > >  5 files changed, 406 insertions(+)  create mode 100644
> > > > > > > > lib/librte_ether/rte_eth_dev_info.h
> > > > > > > >
> > > > > > > > diff --git a/lib/librte_ether/Makefile
> > > > > > > > b/lib/librte_ether/Makefile index c0e5768..05209e9 100644
> > > > > > > > --- a/lib/librte_ether/Makefile
> > > > > > > > +++ b/lib/librte_ether/Makefile
> > > > > > > > @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c
> > > > > > > > SYMLINK-y-include += rte_ether.h  SYMLINK-y-include +=
> > > > > > > > rte_ethdev.h SYMLINK-y-include
> > > > > > > > += rte_eth_ctrl.h
> > > > > > > > +SYMLINK-y-include += rte_eth_dev_info.h
> > > > > > > >
> > > > > > > >  # this lib depends upon:
> > > > > > > >  DEPDIRS-y += lib/librte_eal lib/librte_mempool
> > > > > > > > lib/librte_ring lib/librte_mbuf diff --git
> > > > > > > > a/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > new file mode 100644
> > > > > > > > index 0000000..002c4b5
> > > > > > > > --- /dev/null
> > > > > > > > +++ b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > @@ -0,0 +1,80 @@
> > > > > > > > +/*-
> > > > > > > > + *   BSD LICENSE
> > > > > > > > + *
> > > > > > > > + *   Copyright(c) 2015 Intel Corporation. All rights
> reserved.
> > > > > > > > + *   All rights reserved.
> > > > > > > > + *
> > > > > > > > + *   Redistribution and use in source and binary forms,
> with or
> > > without
> > > > > > > > + *   modification, are permitted provided that the
> following
> > > conditions
> > > > > > > > + *   are met:
> > > > > > > > + *
> > > > > > > > + *     * Redistributions of source code must retain the
> above
> > > copyright
> > > > > > > > + *       notice, this list of conditions and the following
> disclaimer.
> > > > > > > > + *     * Redistributions in binary form must reproduce the
> above
> > > > > copyright
> > > > > > > > + *       notice, this list of conditions and the following
> disclaimer in
> > > > > > > > + *       the documentation and/or other materials provided
> with the
> > > > > > > > + *       distribution.
> > > > > > > > + *     * Neither the name of Intel Corporation nor the
> names of its
> > > > > > > > + *       contributors may be used to endorse or promote
> products
> > > > > derived
> > > > > > > > + *       from this software without specific prior written
> permission.
> > > > > > > > + *
> > > > > > > > + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> > > > > > > CONTRIBUTORS
> > > > > > > > + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
> > > INCLUDING,
> > > > > BUT
> > > > > > > NOT
> > > > > > > > + *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
> > > AND
> > > > > > > FITNESS FOR
> > > > > > > > + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL
> > > THE
> > > > > > > COPYRIGHT
> > > > > > > > + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
> > > INDIRECT,
> > > > > > > INCIDENTAL,
> > > > > > > > + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
> > > > > (INCLUDING,
> > > > > > > BUT NOT
> > > > > > > > + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
> > > SERVICES;
> > > > > > > LOSS OF USE,
> > > > > > > > + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
> > > > > CAUSED
> > > > > > > AND ON ANY
> > > > > > > > + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
> > > LIABILITY,
> > > > > OR
> > > > > > > TORT
> > > > > > > > + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY
> > > WAY
> > > > > OUT
> > > > > > > OF THE USE
> > > > > > > > + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY
> OF
> > > SUCH
> > > > > > > DAMAGE.
> > > > > > > > + */
> > > > > > > > +
> > > > > > > > +#ifndef _RTE_ETH_DEV_INFO_H_ #define _RTE_ETH_DEV_INFO_H_
> > > > > > > > +
> > > > > > > > +
> > > > > > > > +/*
> > > > > > > > + * Placeholder for accessing device registers  */ struct
> > > > > > > > +rte_dev_reg_info {
> > > > > > > > +	void *buf; /**< Buffer for register */
> > > > > > > > +	uint32_t offset; /**< Offset for 1st register to fetch
> */
> > > > > > > > +	uint32_t leng; /**< Number of registers to fetch */
> > > > > > > > +	uint32_t version; /**< Device version */ };
> > > > > > > > +
> > > > > > > > +/*
> > > > > > > > + * Placeholder for accessing device eeprom  */ struct
> > > > > > > > +rte_dev_eeprom_info {
> > > > > > > > +	void *buf; /**< Buffer for eeprom */
> > > > > > > > +	uint32_t offset; /**< Offset for 1st eeprom location to
> > > > > > > > +access
> > > */
> > > > > > > > +	uint32_t leng; /**< Length of eeprom region to access */
> > > > > > > > +	uint32_t magic; /**< Device ID */ };
> > > > > > > > +
> > > > > > > > +/*
> > > > > > > > + * Placeholder for accessing device ring parameters  */
> > > > > > > > +struct rte_dev_ring_info {
> > > > > > > > +	uint32_t rx_pending; /**< Number of outstanding Rx ring
> */
> > > > > > > > +	uint32_t tx_pending; /**< Number of outstanding Tx ring
> */
> > > > > > > > +	uint32_t rx_max_pending; /**< Maximum number of
> > > outstanding
> > > > > > > > +Rx
> > > > > > > ring */
> > > > > > > > +	uint32_t tx_max_pending; /**< Maximum number of
> > > outstanding
> > > > > > > > +Tx
> > > > > > > ring
> > > > > > > > +*/ };
> > > > > > > > +
> > > > > > > > +/*
> > > > > > > > + * A data structure captures information as defined in
> > > > > > > > +struct ifla_vf_info
> > > > > > > > + * for user-space api
> > > > > > > > + */
> > > > > > > > +struct rte_dev_vf_info {
> > > > > > > > +	uint32_t vf;
> > > > > > > > +	uint8_t mac[ETHER_ADDR_LEN];
> > > > > > > > +	uint32_t vlan;
> > > > > > > > +	uint32_t tx_rate;
> > > > > > > > +	uint32_t spoofchk;
> > > > > > > > +};
> > > > > > >
> > > > > > >
> > > > > > > Wonder what that structure is for?
> > > > > > > I can't see it used in any function below?
> > > > > > >
> > > > > >
> > > > > > Good catch, this is designed for other ethtool ops that I did
> > > > > > not include in
> > > > > this release, I will remove this from next fix.
> > > > > >
> > > > > > > > +
> > > > > > > > +#endif /* _RTE_ETH_DEV_INFO_H_ */
> > > > > > > > diff --git a/lib/librte_ether/rte_ethdev.c
> > > > > > > > b/lib/librte_ether/rte_ethdev.c index 5a94654..186e85c
> > > > > > > > 100644
> > > > > > > > --- a/lib/librte_ether/rte_ethdev.c
> > > > > > > > +++ b/lib/librte_ether/rte_ethdev.c
> > > > > > > > @@ -2751,6 +2751,32 @@ rte_eth_dev_mac_addr_remove(uint8_t
> > > > > > > port_id,
> > > > > > > > struct ether_addr *addr)  }
> > > > > > > >
> > > > > > > >  int
> > > > > > > > +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct
> > > > > > > > +ether_addr
> > > > > > > > +*addr) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +	const int index = 0;
> > > > > > > > +	const uint32_t pool = 0;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	dev = &rte_eth_devices[port_id];
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> > > >mac_addr_remove, -
> > > > > > > ENOTSUP);
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -
> > > > > > > ENOTSUP);
> > > > > > > > +
> > > > > > > > +	/* Update NIC default MAC address*/
> > > > > > > > +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> > > > > > > > +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> > > > > > > > +
> > > > > > > > +	/* Update default address in NIC data structure */
> > > > > > > > +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> > > > > > > > +
> > > > > > > > +	return 0;
> > > > > > > > +}
> > > > > > > > +
> > > > > > > > +int
> > > > > > > >  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
> > > > > > > >  				uint16_t rx_mode, uint8_t on)  { @@
> > > -3627,3
> > > > > +3653,136 @@
> > > > > > > > rte_eth_remove_tx_callback(uint8_t port_id,
> > > > > > > uint16_t queue_id,
> > > > > > > >  	/* Callback wasn't found. */
> > > > > > > >  	return -EINVAL;
> > > > > > > >  }
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_reg_leng(uint8_t port_id) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length,
> > > -
> > > > > > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->get_reg_length)(dev);
> > > > > > > > +}
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_reg_info(uint8_t port_id, struct
> > > > > > > > +rte_dev_reg_info
> > > > > > > > +*info) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -
> > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->get_reg)(dev, info); }
> > > > > > >
> > > > > > > Seems that *get_reg* stuff, will be really good addition for
> > > > > > > DPDK debugging abilities.
> > > > > > > Though, I'd suggest we change it a bit to make more generic
> > > > > > > and
> > > flexible:
> > > > > > >
> > > > > > > Introduce rte_eth_reg_read/rte_eth_reg_write(),
> > > > > > > or probably even better rte_pcidev_reg_read
> > > > > > > /rte_pcidev_reg_write at
> > > > > EAL.
> > > > > > > Something similar to what
> > > > > > > port_pci_reg_read/port_pci_reg_write()
> > > > > > > are doing now at testpmd.h.
> > > > > > >
> > > > > > > struct rte_pcidev_reg_info {
> > > > > > >    const char *name;
> > > > > > >    uint32_t endianes, bar, offset, size, count; };
> > > > > > >
> > > > > > > int rte_pcidev_reg_read(const struct rte_pci_device *, const
> > > > > > > struct rte_pcidev_reg_info *, uint64_t *reg_val);
> > > > > > >
> > > > > > > Then:
> > > > > > > int rte_eth_dev_get_reg_info(port_id, const struct
> > > > > > > rte_pcidev_reg_info **info);
> > > > > > >
> > > > > > > So each device would store in info a pointer to an array of
> > > > > > > it's register descriptions (finished by zero elem?).
> > > > > > >
> > > > > > > Then your ethtool (or any other upper layer) can do the
> > > > > > > following to read all device regs:
> > > > > > >
> > > > > >
> > > > > > The proposed reg info structure allows future improvement to
> > > > > > support
> > > > > individual register read/write.
> > > > > > Also, because each NIC device has a very distinguish register
> definition.
> > > > > > So, the plan is to have more comprehensive interface to
> > > > > > support query operation (for example, register name) before
> > > > > > introduce individual/group
> > > > > register access.
> > > > > > Points taken, the support will be in future release.
> > > > >
> > > > > Sorry, didn't get you.
> > > > > So you are ok to make these changes in next patch version?
> > > > >
> > > > I would like to get a consensus from dpdk community on how to
> > > > provide
> > > register information.
> > >
> > > Well, that' ok, but if it is just a trial patch that is not intended
> > > to be applied, then you should mark it as RFC.
> > >
> > > > Currently, it's designed for debug dumping. The register
> > > > information is very
> > > hardware dependent.
> > > > Need to consider current supported NIC device and future devices
> > > > for
> > > DPDK, so we won't make it a bulky interface.
> > >
> > > Ok, could you explain what exactly  concerns you in the approach
> > > described above?
> > > What part you feel is bulky?
> > >
> > > > > >
> > > > > > > const struct rte_eth_dev_reg_info *reg_info; struct
> > > > > > > rte_eth_dev_info dev_info;
> > > > > > >
> > > > > > > rte_eth_dev_info_get(pid, &dev_info);
> > > > > > > rte_eth_dev_get_reg_info(port_id, &reg_info);
> > > > > > >
> > > > > > > for (i = 0; reg_info[i].name != NULL; i++) {
> > > > > > >    ...
> > > > > > >    rte_pcidev_read_reg(dev_info. pci_dev, reg_info[i], &v);
> > > > > > >   ..
> > > > > > > }
> > > > > > >
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_eeprom_leng(uint8_t port_id) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> > > >get_eeprom_length, -
> > > > > > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->get_eeprom_length)(dev);
> > > > > > > > +}
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_get_eeprom(uint8_t port_id, struct
> > > > > > > > +rte_dev_eeprom_info
> > > > > > > > +*info) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -
> > > > > > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->get_eeprom)(dev, info); }
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_set_eeprom(uint8_t port_id, struct
> > > > > > > > +rte_dev_eeprom_info
> > > > > > > > +*info) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -
> > > > > > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->set_eeprom)(dev, info); }
> > > > > > > > +
> > > > > > > > +int
> > > > > > > > +rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > > > > > > +rte_dev_ring_info
> > > > > > > > +*info) {
> > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > +
> > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > port_id);
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > +		return -ENODEV;
> > > > > > > > +	}
> > > > > > > > +
> > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> > > > > > > ENOTSUP);
> > > > > > > > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> > > > > > >
> > > > > > > I think it will be a useful addition to the ethdev API  to
> > > > > > > have an ability to retrieve current RX/TX queue parameters.
> > > > > > > Though again, it need to be more generic, so it could be
> > > > > > > useful for
> > > > > > > non- ethtool upper layer too.
> > > > > > > So I suggest to modify it a bit.
> > > > > > > Something like that:
> > > > > > >
> > > > > > > struct rte_eth_tx_queue_info {
> > > > > > >     struct rte_eth_txconf txconf;
> > > > > > >     uint32_t nb_tx_desc;
> > > > > > >     uint32_t nb_max_tx_desc; /*max allowable TXDs for that
> queue */
> > > > > > >     uint32_t nb_tx_free;            /* number of free TXDs at
> the moment
> > > of
> > > > > call.
> > > > > > > */
> > > > > > >     /* other tx queue data. */ };
> > > > > > >
> > > > > > > int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> > > > > > > rte_eth_tx_queue_info *qinfo)
> > > > > > >
> > > > > > > Then, your upper layer ethtool wrapper, can implement yours
> > > > > > > ethtool_get_ringparam() by:
> > > > > > >
> > > > > > >  ...
> > > > > > >  struct rte_eth_tx_queue_info qinfo;
> > > > > > > rte_ethdev_get_tx_queue_info(port, 0, &qinfo);
> > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > rte_eth_rx_queue_count(port, 0);
> > > > > > >
> > > > > > > Or probably even:
> > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > qinfo.nb_tx_free;
> > > > > > >
> > > > > > > Same for RX.
> > > > > > >
> > > > > > For now, this descriptor ring information is used by the ethtool
> op.
> > > > > > To make this interface simple, i.e. caller doesn't need to
> > > > > > access other
> > > > > queue information.
> > > > >
> > > > > I just repeat what I said to you in off-line conversation:
> > > > > ethdev API is not equal ethtool API.
> > > > > It is ok to add  a new function/structure to ethdev if it really
> > > > > needed, but we should do mechanical one to one copy.
> > > > > It is much better to add  a function/structure that would be
> > > > > more generic, and suit other users, not only ethtool.
> > > > > There is no point to have dozen functions in rte_ethdev API
> > > > > providing similar information.
> > > > > BTW, I don't see how API I proposed is much more  complex, then
> > > > > yours
> > > one.
> > > > The ring parameter is a run-time information which is different
> > > > than data
> > > structure described in this discussion.
> > >
> > > I don't see how they are different.
> > > Looking at ixgbe_get_ringparam(), it returns:
> > > rx_max_pending - that's a static IXGBE PMD value (max possible
> > > number of RXDs per one queue).
> > > rx_pending - number of RXD currently in use by the HW for queue 0
> > > (that information can be changed at each call).
> > >
> > > With the approach I suggesting - you can get same information for
> > > each RX queue by calling rte_ethdev_get_rx_queue_info() and
> > > rte_eth_rx_queue_count().
> > > Plus you are getting other RX queue data.
> > >
> > > Another thing - what is practical usage of the information you
> > > retrieving now by get_ringparam()?
> > > Let say it returned to you: rx_max_pending=4096; rx_pending=128; How
> > > that information would help to understand what is going on with the
> device?
> > > Without knowing  value of nb_tx_desc for the queue, you can't say is
> > > you queue full or not.
> > > Again, it could be that all your traffic going through some other
> > > queue (not 0).
> > > So from my point rte_eth_dev_get_ringparam()  usage is very limited,
> > > and doesn't provide enough information about current queue state.
> > >
> > > Same thing applies for TX.
> > >
> >
> > After careful review the suggestion in this comment, and review the
> existing dpdk source code.
> > I came to realize that neither rte_ethdev_get_rx_queue_info,
> > rte_ethdev_get_tx_queue_info, struct rte_eth_rx_queue_info and struct
> > rte_eth_tx_queue_info are available in the existing dpdk source code. I
> could not make a patch based upon a set of non- existent API and data
> structure.
> 
> Right now, in dpdk.org source code, struct  rte_eth_dev_ring_info, struct
> rte_dev_eeprom_info and struct  rte_dev_reg_info don't exist also.
> Same as  all these functions:
> 
> rte_eth_dev_default_mac_addr_set
> rte_eth_dev_reg_length
> rte_eth_dev_reg_info
> rte_eth_dev_eeprom_length
> rte_eth_dev_get_eeprom
> rte_eth_dev_set_eeprom
> rte_eth_dev_get_ringparam
> 
> All this is a new API that's you are trying to add.
> But, by some reason you consider it is ok to add 'struct
> rte_eth_dev_ring_info', but couldn't add  struct
> 'rte_ethdev_get_tx_queue_info'
> That just doesn't make any sense to me.
> In fact, I think our conversation is going in cycles.
> If you are not happy with the suggested approach, please provide some
> meaningful reason why.
> Konstantin

It seems the new API aims at providing users a mechanism to quickly and 
gracefully migrate from using ethtool/ioctl calls.  The provided get/set 
ring param info is very similar to that of ethtool and facilitates the
ethtool needs.

While additional enhancements to the API to provide additional details 
such as those you have suggested are certainly possible, I believe Larry 
is stating those ideas are outside the scope he has intended with the 
API introduction and that they should be discussed further and delivered
in a future patch.

Does that seem reasonable?

Thanks,
Dave

> 
> >
> > > > It's the desire of this patch to separate each data structure to
> > > > avoid cross
> > > dependency.
> > >
> > > That's too cryptic to me.
> > > Could you explain what cross dependency you are talking about?
> > >
> > > Konstantin
  

> -----Original Message-----
> From: Wang, Liang-min
> Sent: Monday, June 15, 2015 3:47 PM
> To: Ananyev, Konstantin; dev@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access device info
> 
> 
> 
> > -----Original Message-----
> > From: Ananyev, Konstantin
> > Sent: Monday, June 15, 2015 9:46 AM
> > To: Wang, Liang-min; dev@dpdk.org
> > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> > device info
> >
> >
> >
> > > -----Original Message-----
> > > From: Wang, Liang-min
> > > Sent: Monday, June 15, 2015 2:26 PM
> > > To: Ananyev, Konstantin; dev@dpdk.org
> > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > access device info
> > >
> > >
> > >
> > > > -----Original Message-----
> > > > From: Ananyev, Konstantin
> > > > Sent: Friday, June 12, 2015 8:31 AM
> > > > To: Wang, Liang-min; dev@dpdk.org
> > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > access device info
> > > >
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Wang, Liang-min
> > > > > Sent: Thursday, June 11, 2015 10:51 PM
> > > > > To: Ananyev, Konstantin; dev@dpdk.org
> > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > > access device info
> > > > >
> > > > >
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: Ananyev, Konstantin
> > > > > > Sent: Thursday, June 11, 2015 9:07 AM
> > > > > > To: Wang, Liang-min; dev@dpdk.org
> > > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > support access device info
> > > > > >
> > > > > >
> > > > > >
> > > > > > > -----Original Message-----
> > > > > > > From: Wang, Liang-min
> > > > > > > Sent: Thursday, June 11, 2015 1:58 PM
> > > > > > > To: Ananyev, Konstantin; dev@dpdk.org
> > > > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > > support access device info
> > > > > > >
> > > > > > > Hi Konstantin,
> > > > > > >
> > > > > > > > -----Original Message-----
> > > > > > > > From: Ananyev, Konstantin
> > > > > > > > Sent: Thursday, June 11, 2015 8:26 AM
> > > > > > > > To: Wang, Liang-min; dev@dpdk.org
> > > > > > > > Cc: Wang, Liang-min
> > > > > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > > > support access device info
> > > > > > > >
> > > > > > > > Hi Larry,
> > > > > > > >
> > > > > > > > > -----Original Message-----
> > > > > > > > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of
> > > > > > > > > Liang-Min Larry Wang
> > > > > > > > > Sent: Wednesday, June 10, 2015 4:10 PM
> > > > > > > > > To: dev@dpdk.org
> > > > > > > > > Cc: Wang, Liang-min
> > > > > > > > > Subject: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > > > > support access device info
> > > > > > > > >
> > > > > > > > > add new apis:
> > > > > > > > > - rte_eth_dev_default_mac_addr_set
> > > > > > > > > - rte_eth_dev_reg_leng
> > > > > > > > > - rte_eth_dev_reg_info
> > > > > > > > > - rte_eth_dev_eeprom_leng
> > > > > > > > > - rte_eth_dev_get_eeprom
> > > > > > > > > - rte_eth_dev_set_eeprom
> > > > > > > > > - rte_eth_dev_get_ringparam
> > > > > > > > > - rte_eth_dev_set_ringparam
> > > > > > > > >
> > > > > > > > > to enable reading device parameters (mac-addr, register,
> > > > > > > > > eeprom,
> > > > > > > > > ring) based upon ethtool alike data parameter specification.
> > > > > > > > >
> > > > > > > > > Signed-off-by: Liang-Min Larry Wang
> > > > > > > > > <liang-min.wang@intel.com>
> > > > > > > > > ---
> > > > > > > > >  lib/librte_ether/Makefile              |   1 +
> > > > > > > > >  lib/librte_ether/rte_eth_dev_info.h    |  80
> > +++++++++++++++++
> > > > > > > > >  lib/librte_ether/rte_ethdev.c          | 159
> > > > > > > > +++++++++++++++++++++++++++++++++
> > > > > > > > >  lib/librte_ether/rte_ethdev.h          | 158
> > > > > > > > ++++++++++++++++++++++++++++++++
> > > > > > > > >  lib/librte_ether/rte_ether_version.map |   8 ++
> > > > > > > > >  5 files changed, 406 insertions(+)  create mode 100644
> > > > > > > > > lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > >
> > > > > > > > > diff --git a/lib/librte_ether/Makefile
> > > > > > > > > b/lib/librte_ether/Makefile index c0e5768..05209e9 100644
> > > > > > > > > --- a/lib/librte_ether/Makefile
> > > > > > > > > +++ b/lib/librte_ether/Makefile
> > > > > > > > > @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c
> > > > > > > > > SYMLINK-y-include += rte_ether.h  SYMLINK-y-include +=
> > > > > > > > > rte_ethdev.h SYMLINK-y-include
> > > > > > > > > += rte_eth_ctrl.h
> > > > > > > > > +SYMLINK-y-include += rte_eth_dev_info.h
> > > > > > > > >
> > > > > > > > >  # this lib depends upon:
> > > > > > > > >  DEPDIRS-y += lib/librte_eal lib/librte_mempool
> > > > > > > > > lib/librte_ring lib/librte_mbuf diff --git
> > > > > > > > > a/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > > b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > > new file mode 100644
> > > > > > > > > index 0000000..002c4b5
> > > > > > > > > --- /dev/null
> > > > > > > > > +++ b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > > @@ -0,0 +1,80 @@
> > > > > > > > > +/*-
> > > > > > > > > + *   BSD LICENSE
> > > > > > > > > + *
> > > > > > > > > + *   Copyright(c) 2015 Intel Corporation. All rights reserved.
> > > > > > > > > + *   All rights reserved.
> > > > > > > > > + *
> > > > > > > > > + *   Redistribution and use in source and binary forms, with or
> > > > without
> > > > > > > > > + *   modification, are permitted provided that the following
> > > > conditions
> > > > > > > > > + *   are met:
> > > > > > > > > + *
> > > > > > > > > + *     * Redistributions of source code must retain the above
> > > > copyright
> > > > > > > > > + *       notice, this list of conditions and the following disclaimer.
> > > > > > > > > + *     * Redistributions in binary form must reproduce the above
> > > > > > copyright
> > > > > > > > > + *       notice, this list of conditions and the following disclaimer
> > in
> > > > > > > > > + *       the documentation and/or other materials provided with
> > the
> > > > > > > > > + *       distribution.
> > > > > > > > > + *     * Neither the name of Intel Corporation nor the names of
> > its
> > > > > > > > > + *       contributors may be used to endorse or promote
> > products
> > > > > > derived
> > > > > > > > > + *       from this software without specific prior written
> > permission.
> > > > > > > > > + *
> > > > > > > > > + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS
> > AND
> > > > > > > > CONTRIBUTORS
> > > > > > > > > + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
> > > > INCLUDING,
> > > > > > BUT
> > > > > > > > NOT
> > > > > > > > > + *   LIMITED TO, THE IMPLIED WARRANTIES OF
> > MERCHANTABILITY
> > > > AND
> > > > > > > > FITNESS FOR
> > > > > > > > > + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT
> > SHALL
> > > > THE
> > > > > > > > COPYRIGHT
> > > > > > > > > + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
> > > > INDIRECT,
> > > > > > > > INCIDENTAL,
> > > > > > > > > + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
> > > > > > (INCLUDING,
> > > > > > > > BUT NOT
> > > > > > > > > + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
> > > > SERVICES;
> > > > > > > > LOSS OF USE,
> > > > > > > > > + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
> > HOWEVER
> > > > > > CAUSED
> > > > > > > > AND ON ANY
> > > > > > > > > + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
> > > > LIABILITY,
> > > > > > OR
> > > > > > > > TORT
> > > > > > > > > + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY
> > > > WAY
> > > > > > OUT
> > > > > > > > OF THE USE
> > > > > > > > > + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY
> > OF
> > > > SUCH
> > > > > > > > DAMAGE.
> > > > > > > > > + */
> > > > > > > > > +
> > > > > > > > > +#ifndef _RTE_ETH_DEV_INFO_H_ #define
> > _RTE_ETH_DEV_INFO_H_
> > > > > > > > > +
> > > > > > > > > +
> > > > > > > > > +/*
> > > > > > > > > + * Placeholder for accessing device registers  */ struct
> > > > > > > > > +rte_dev_reg_info {
> > > > > > > > > +	void *buf; /**< Buffer for register */
> > > > > > > > > +	uint32_t offset; /**< Offset for 1st register to fetch */
> > > > > > > > > +	uint32_t leng; /**< Number of registers to fetch */
> > > > > > > > > +	uint32_t version; /**< Device version */ };
> > > > > > > > > +
> > > > > > > > > +/*
> > > > > > > > > + * Placeholder for accessing device eeprom  */ struct
> > > > > > > > > +rte_dev_eeprom_info {
> > > > > > > > > +	void *buf; /**< Buffer for eeprom */
> > > > > > > > > +	uint32_t offset; /**< Offset for 1st eeprom location to
> > > > > > > > > +access
> > > > */
> > > > > > > > > +	uint32_t leng; /**< Length of eeprom region to access */
> > > > > > > > > +	uint32_t magic; /**< Device ID */ };
> > > > > > > > > +
> > > > > > > > > +/*
> > > > > > > > > + * Placeholder for accessing device ring parameters  */
> > > > > > > > > +struct rte_dev_ring_info {
> > > > > > > > > +	uint32_t rx_pending; /**< Number of outstanding Rx ring */
> > > > > > > > > +	uint32_t tx_pending; /**< Number of outstanding Tx ring */
> > > > > > > > > +	uint32_t rx_max_pending; /**< Maximum number of
> > > > outstanding
> > > > > > > > > +Rx
> > > > > > > > ring */
> > > > > > > > > +	uint32_t tx_max_pending; /**< Maximum number of
> > > > outstanding
> > > > > > > > > +Tx
> > > > > > > > ring
> > > > > > > > > +*/ };
> > > > > > > > > +
> > > > > > > > > +/*
> > > > > > > > > + * A data structure captures information as defined in
> > > > > > > > > +struct ifla_vf_info
> > > > > > > > > + * for user-space api
> > > > > > > > > + */
> > > > > > > > > +struct rte_dev_vf_info {
> > > > > > > > > +	uint32_t vf;
> > > > > > > > > +	uint8_t mac[ETHER_ADDR_LEN];
> > > > > > > > > +	uint32_t vlan;
> > > > > > > > > +	uint32_t tx_rate;
> > > > > > > > > +	uint32_t spoofchk;
> > > > > > > > > +};
> > > > > > > >
> > > > > > > >
> > > > > > > > Wonder what that structure is for?
> > > > > > > > I can't see it used in any function below?
> > > > > > > >
> > > > > > >
> > > > > > > Good catch, this is designed for other ethtool ops that I did
> > > > > > > not include in
> > > > > > this release, I will remove this from next fix.
> > > > > > >
> > > > > > > > > +
> > > > > > > > > +#endif /* _RTE_ETH_DEV_INFO_H_ */
> > > > > > > > > diff --git a/lib/librte_ether/rte_ethdev.c
> > > > > > > > > b/lib/librte_ether/rte_ethdev.c index 5a94654..186e85c
> > > > > > > > > 100644
> > > > > > > > > --- a/lib/librte_ether/rte_ethdev.c
> > > > > > > > > +++ b/lib/librte_ether/rte_ethdev.c
> > > > > > > > > @@ -2751,6 +2751,32 @@
> > rte_eth_dev_mac_addr_remove(uint8_t
> > > > > > > > port_id,
> > > > > > > > > struct ether_addr *addr)  }
> > > > > > > > >
> > > > > > > > >  int
> > > > > > > > > +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct
> > > > > > > > > +ether_addr
> > > > > > > > > +*addr) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +	const int index = 0;
> > > > > > > > > +	const uint32_t pool = 0;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	dev = &rte_eth_devices[port_id];
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> > > > >mac_addr_remove, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +
> > > > > > > > > +	/* Update NIC default MAC address*/
> > > > > > > > > +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> > > > > > > > > +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> > > > > > > > > +
> > > > > > > > > +	/* Update default address in NIC data structure */
> > > > > > > > > +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> > > > > > > > > +
> > > > > > > > > +	return 0;
> > > > > > > > > +}
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > >  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
> > > > > > > > >  				uint16_t rx_mode, uint8_t on)  { @@
> > > > -3627,3
> > > > > > +3653,136 @@
> > > > > > > > > rte_eth_remove_tx_callback(uint8_t port_id,
> > > > > > > > uint16_t queue_id,
> > > > > > > > >  	/* Callback wasn't found. */
> > > > > > > > >  	return -EINVAL;
> > > > > > > > >  }
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_reg_leng(uint8_t port_id) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length,
> > > > -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->get_reg_length)(dev);
> > > > > > > > > +}
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_reg_info(uint8_t port_id, struct
> > > > > > > > > +rte_dev_reg_info
> > > > > > > > > +*info) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -
> > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->get_reg)(dev, info); }
> > > > > > > >
> > > > > > > > Seems that *get_reg* stuff, will be really good addition for
> > > > > > > > DPDK debugging abilities.
> > > > > > > > Though, I'd suggest we change it a bit to make more generic
> > > > > > > > and
> > > > flexible:
> > > > > > > >
> > > > > > > > Introduce rte_eth_reg_read/rte_eth_reg_write(),
> > > > > > > > or probably even better rte_pcidev_reg_read
> > > > > > > > /rte_pcidev_reg_write at
> > > > > > EAL.
> > > > > > > > Something similar to what
> > > > > > > > port_pci_reg_read/port_pci_reg_write()
> > > > > > > > are doing now at testpmd.h.
> > > > > > > >
> > > > > > > > struct rte_pcidev_reg_info {
> > > > > > > >    const char *name;
> > > > > > > >    uint32_t endianes, bar, offset, size, count; };
> > > > > > > >
> > > > > > > > int rte_pcidev_reg_read(const struct rte_pci_device *, const
> > > > > > > > struct rte_pcidev_reg_info *, uint64_t *reg_val);
> > > > > > > >
> > > > > > > > Then:
> > > > > > > > int rte_eth_dev_get_reg_info(port_id, const struct
> > > > > > > > rte_pcidev_reg_info **info);
> > > > > > > >
> > > > > > > > So each device would store in info a pointer to an array of
> > > > > > > > it's register descriptions (finished by zero elem?).
> > > > > > > >
> > > > > > > > Then your ethtool (or any other upper layer) can do the
> > > > > > > > following to read all device regs:
> > > > > > > >
> > > > > > >
> > > > > > > The proposed reg info structure allows future improvement to
> > > > > > > support
> > > > > > individual register read/write.
> > > > > > > Also, because each NIC device has a very distinguish register
> > definition.
> > > > > > > So, the plan is to have more comprehensive interface to
> > > > > > > support query operation (for example, register name) before
> > > > > > > introduce individual/group
> > > > > > register access.
> > > > > > > Points taken, the support will be in future release.
> > > > > >
> > > > > > Sorry, didn't get you.
> > > > > > So you are ok to make these changes in next patch version?
> > > > > >
> > > > > I would like to get a consensus from dpdk community on how to
> > > > > provide
> > > > register information.
> > > >
> > > > Well, that' ok, but if it is just a trial patch that is not intended
> > > > to be applied, then you should mark it as RFC.
> > > >
> > > > > Currently, it's designed for debug dumping. The register
> > > > > information is very
> > > > hardware dependent.
> > > > > Need to consider current supported NIC device and future devices
> > > > > for
> > > > DPDK, so we won't make it a bulky interface.
> > > >
> > > > Ok, could you explain what exactly  concerns you in the approach
> > > > described above?
> > > > What part you feel is bulky?
> > > >
> > > > > > >
> > > > > > > > const struct rte_eth_dev_reg_info *reg_info; struct
> > > > > > > > rte_eth_dev_info dev_info;
> > > > > > > >
> > > > > > > > rte_eth_dev_info_get(pid, &dev_info);
> > > > > > > > rte_eth_dev_get_reg_info(port_id, &reg_info);
> > > > > > > >
> > > > > > > > for (i = 0; reg_info[i].name != NULL; i++) {
> > > > > > > >    ...
> > > > > > > >    rte_pcidev_read_reg(dev_info. pci_dev, reg_info[i], &v);
> > > > > > > >   ..
> > > > > > > > }
> > > > > > > >
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_eeprom_leng(uint8_t port_id) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> > > > >get_eeprom_length, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->get_eeprom_length)(dev);
> > > > > > > > > +}
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_get_eeprom(uint8_t port_id, struct
> > > > > > > > > +rte_dev_eeprom_info
> > > > > > > > > +*info) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->get_eeprom)(dev, info); }
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_set_eeprom(uint8_t port_id, struct
> > > > > > > > > +rte_dev_eeprom_info
> > > > > > > > > +*info) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->set_eeprom)(dev, info); }
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > > > > > > > +rte_dev_ring_info
> > > > > > > > > +*info) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> > > > > > > >
> > > > > > > > I think it will be a useful addition to the ethdev API  to
> > > > > > > > have an ability to retrieve current RX/TX queue parameters.
> > > > > > > > Though again, it need to be more generic, so it could be
> > > > > > > > useful for
> > > > > > > > non- ethtool upper layer too.
> > > > > > > > So I suggest to modify it a bit.
> > > > > > > > Something like that:
> > > > > > > >
> > > > > > > > struct rte_eth_tx_queue_info {
> > > > > > > >     struct rte_eth_txconf txconf;
> > > > > > > >     uint32_t nb_tx_desc;
> > > > > > > >     uint32_t nb_max_tx_desc; /*max allowable TXDs for that
> > queue */
> > > > > > > >     uint32_t nb_tx_free;            /* number of free TXDs at the
> > moment
> > > > of
> > > > > > call.
> > > > > > > > */
> > > > > > > >     /* other tx queue data. */ };
> > > > > > > >
> > > > > > > > int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> > > > > > > > rte_eth_tx_queue_info *qinfo)
> > > > > > > >
> > > > > > > > Then, your upper layer ethtool wrapper, can implement yours
> > > > > > > > ethtool_get_ringparam() by:
> > > > > > > >
> > > > > > > >  ...
> > > > > > > >  struct rte_eth_tx_queue_info qinfo;
> > > > > > > > rte_ethdev_get_tx_queue_info(port, 0, &qinfo);
> > > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > > rte_eth_rx_queue_count(port, 0);
> > > > > > > >
> > > > > > > > Or probably even:
> > > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > > qinfo.nb_tx_free;
> > > > > > > >
> > > > > > > > Same for RX.
> > > > > > > >
> > > > > > > For now, this descriptor ring information is used by the ethtool op.
> > > > > > > To make this interface simple, i.e. caller doesn't need to
> > > > > > > access other
> > > > > > queue information.
> > > > > >
> > > > > > I just repeat what I said to you in off-line conversation:
> > > > > > ethdev API is not equal ethtool API.
> > > > > > It is ok to add  a new function/structure to ethdev if it really
> > > > > > needed, but we should do mechanical one to one copy.
> > > > > > It is much better to add  a function/structure that would be
> > > > > > more generic, and suit other users, not only ethtool.
> > > > > > There is no point to have dozen functions in rte_ethdev API
> > > > > > providing similar information.
> > > > > > BTW, I don't see how API I proposed is much more  complex, then
> > > > > > yours
> > > > one.
> > > > > The ring parameter is a run-time information which is different
> > > > > than data
> > > > structure described in this discussion.
> > > >
> > > > I don't see how they are different.
> > > > Looking at ixgbe_get_ringparam(), it returns:
> > > > rx_max_pending - that's a static IXGBE PMD value (max possible
> > > > number of RXDs per one queue).
> > > > rx_pending - number of RXD currently in use by the HW for queue 0
> > > > (that information can be changed at each call).
> > > >
> > > > With the approach I suggesting - you can get same information for
> > > > each RX queue by calling rte_ethdev_get_rx_queue_info() and
> > > > rte_eth_rx_queue_count().
> > > > Plus you are getting other RX queue data.
> > > >
> > > > Another thing - what is practical usage of the information you
> > > > retrieving now by get_ringparam()?
> > > > Let say it returned to you: rx_max_pending=4096; rx_pending=128; How
> > > > that information would help to understand what is going on with the
> > device?
> > > > Without knowing  value of nb_tx_desc for the queue, you can't say is
> > > > you queue full or not.
> > > > Again, it could be that all your traffic going through some other
> > > > queue (not 0).
> > > > So from my point rte_eth_dev_get_ringparam()  usage is very limited,
> > > > and doesn't provide enough information about current queue state.
> > > >
> > > > Same thing applies for TX.
> > > >
> > >
> > > After careful review the suggestion in this comment, and review the
> > existing dpdk source code.
> > > I came to realize that neither rte_ethdev_get_rx_queue_info,
> > > rte_ethdev_get_tx_queue_info, struct rte_eth_rx_queue_info and struct
> > > rte_eth_tx_queue_info are available in the existing dpdk source code. I
> > could not make a patch based upon a set of non- existent API and data
> > structure.
> >
> > Right now, in dpdk.org source code, struct  rte_eth_dev_ring_info, struct
> > rte_dev_eeprom_info and struct  rte_dev_reg_info don't exist also.
> > Same as  all these functions:
> >
> > rte_eth_dev_default_mac_addr_set
> > rte_eth_dev_reg_length
> > rte_eth_dev_reg_info
> > rte_eth_dev_eeprom_length
> > rte_eth_dev_get_eeprom
> > rte_eth_dev_set_eeprom
> > rte_eth_dev_get_ringparam
> >
> > All this is a new API that's you are trying to add.
> > But, by some reason you consider it is ok to add 'struct
> > rte_eth_dev_ring_info', but couldn't add  struct
> > 'rte_ethdev_get_tx_queue_info'
> > That just doesn't make any sense to me.
> > In fact, I think our conversation is going in cycles.
> > If you are not happy with the suggested approach, please provide some
> > meaningful reason why.
> 
> All the API's and data structure that you have questions in this comment are available from the submitted patch, you could run to see
> if it works.
> What I learned from your comments are a couple of API name and incomplete data structure description, so I could not make my
> patch according to your suggestion.
> If you strongly feel the API's that you are proposing are useful for get_ringparam(), please create a patch and submit it for evaluation.

Ok, I'll try to create a patch in next few days.
Hopefully it will make things clearer to you and fit everyone.
Konstantin

> As coded in this patch, it's a simple and clean interface and most important, it can be validated and it works.
> 
> > Konstantin
> >
> > >
> > > > > It's the desire of this patch to separate each data structure to
> > > > > avoid cross
> > > > dependency.
> > > >
> > > > That's too cryptic to me.
> > > > Could you explain what cross dependency you are talking about?
> > > >
> > > > Konstantin
  

> -----Original Message-----
> From: David Harton (dharton) [mailto:dharton@cisco.com]
> Sent: Monday, June 15, 2015 5:05 PM
> To: Ananyev, Konstantin; Wang, Liang-min; dev@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access device info
> 
> 
> 
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Ananyev, Konstantin
> > Sent: Monday, June 15, 2015 9:46 AM
> > To: Wang, Liang-min; dev@dpdk.org
> > Subject: Re: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support access
> > device info
> >
> >
> >
> > > -----Original Message-----
> > > From: Wang, Liang-min
> > > Sent: Monday, June 15, 2015 2:26 PM
> > > To: Ananyev, Konstantin; dev@dpdk.org
> > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > access device info
> > >
> > >
> > >
> > > > -----Original Message-----
> > > > From: Ananyev, Konstantin
> > > > Sent: Friday, June 12, 2015 8:31 AM
> > > > To: Wang, Liang-min; dev@dpdk.org
> > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > access device info
> > > >
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Wang, Liang-min
> > > > > Sent: Thursday, June 11, 2015 10:51 PM
> > > > > To: Ananyev, Konstantin; dev@dpdk.org
> > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to support
> > > > > access device info
> > > > >
> > > > >
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: Ananyev, Konstantin
> > > > > > Sent: Thursday, June 11, 2015 9:07 AM
> > > > > > To: Wang, Liang-min; dev@dpdk.org
> > > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > support access device info
> > > > > >
> > > > > >
> > > > > >
> > > > > > > -----Original Message-----
> > > > > > > From: Wang, Liang-min
> > > > > > > Sent: Thursday, June 11, 2015 1:58 PM
> > > > > > > To: Ananyev, Konstantin; dev@dpdk.org
> > > > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > > support access device info
> > > > > > >
> > > > > > > Hi Konstantin,
> > > > > > >
> > > > > > > > -----Original Message-----
> > > > > > > > From: Ananyev, Konstantin
> > > > > > > > Sent: Thursday, June 11, 2015 8:26 AM
> > > > > > > > To: Wang, Liang-min; dev@dpdk.org
> > > > > > > > Cc: Wang, Liang-min
> > > > > > > > Subject: RE: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > > > support access device info
> > > > > > > >
> > > > > > > > Hi Larry,
> > > > > > > >
> > > > > > > > > -----Original Message-----
> > > > > > > > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of
> > > > > > > > > Liang-Min Larry Wang
> > > > > > > > > Sent: Wednesday, June 10, 2015 4:10 PM
> > > > > > > > > To: dev@dpdk.org
> > > > > > > > > Cc: Wang, Liang-min
> > > > > > > > > Subject: [dpdk-dev] [PATCH v4 1/4] ethdev: add apis to
> > > > > > > > > support access device info
> > > > > > > > >
> > > > > > > > > add new apis:
> > > > > > > > > - rte_eth_dev_default_mac_addr_set
> > > > > > > > > - rte_eth_dev_reg_leng
> > > > > > > > > - rte_eth_dev_reg_info
> > > > > > > > > - rte_eth_dev_eeprom_leng
> > > > > > > > > - rte_eth_dev_get_eeprom
> > > > > > > > > - rte_eth_dev_set_eeprom
> > > > > > > > > - rte_eth_dev_get_ringparam
> > > > > > > > > - rte_eth_dev_set_ringparam
> > > > > > > > >
> > > > > > > > > to enable reading device parameters (mac-addr, register,
> > > > > > > > > eeprom,
> > > > > > > > > ring) based upon ethtool alike data parameter specification.
> > > > > > > > >
> > > > > > > > > Signed-off-by: Liang-Min Larry Wang
> > > > > > > > > <liang-min.wang@intel.com>
> > > > > > > > > ---
> > > > > > > > >  lib/librte_ether/Makefile              |   1 +
> > > > > > > > >  lib/librte_ether/rte_eth_dev_info.h    |  80
> > +++++++++++++++++
> > > > > > > > >  lib/librte_ether/rte_ethdev.c          | 159
> > > > > > > > +++++++++++++++++++++++++++++++++
> > > > > > > > >  lib/librte_ether/rte_ethdev.h          | 158
> > > > > > > > ++++++++++++++++++++++++++++++++
> > > > > > > > >  lib/librte_ether/rte_ether_version.map |   8 ++
> > > > > > > > >  5 files changed, 406 insertions(+)  create mode 100644
> > > > > > > > > lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > >
> > > > > > > > > diff --git a/lib/librte_ether/Makefile
> > > > > > > > > b/lib/librte_ether/Makefile index c0e5768..05209e9 100644
> > > > > > > > > --- a/lib/librte_ether/Makefile
> > > > > > > > > +++ b/lib/librte_ether/Makefile
> > > > > > > > > @@ -51,6 +51,7 @@ SRCS-y += rte_ethdev.c
> > > > > > > > > SYMLINK-y-include += rte_ether.h  SYMLINK-y-include +=
> > > > > > > > > rte_ethdev.h SYMLINK-y-include
> > > > > > > > > += rte_eth_ctrl.h
> > > > > > > > > +SYMLINK-y-include += rte_eth_dev_info.h
> > > > > > > > >
> > > > > > > > >  # this lib depends upon:
> > > > > > > > >  DEPDIRS-y += lib/librte_eal lib/librte_mempool
> > > > > > > > > lib/librte_ring lib/librte_mbuf diff --git
> > > > > > > > > a/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > > b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > > new file mode 100644
> > > > > > > > > index 0000000..002c4b5
> > > > > > > > > --- /dev/null
> > > > > > > > > +++ b/lib/librte_ether/rte_eth_dev_info.h
> > > > > > > > > @@ -0,0 +1,80 @@
> > > > > > > > > +/*-
> > > > > > > > > + *   BSD LICENSE
> > > > > > > > > + *
> > > > > > > > > + *   Copyright(c) 2015 Intel Corporation. All rights
> > reserved.
> > > > > > > > > + *   All rights reserved.
> > > > > > > > > + *
> > > > > > > > > + *   Redistribution and use in source and binary forms,
> > with or
> > > > without
> > > > > > > > > + *   modification, are permitted provided that the
> > following
> > > > conditions
> > > > > > > > > + *   are met:
> > > > > > > > > + *
> > > > > > > > > + *     * Redistributions of source code must retain the
> > above
> > > > copyright
> > > > > > > > > + *       notice, this list of conditions and the following
> > disclaimer.
> > > > > > > > > + *     * Redistributions in binary form must reproduce the
> > above
> > > > > > copyright
> > > > > > > > > + *       notice, this list of conditions and the following
> > disclaimer in
> > > > > > > > > + *       the documentation and/or other materials provided
> > with the
> > > > > > > > > + *       distribution.
> > > > > > > > > + *     * Neither the name of Intel Corporation nor the
> > names of its
> > > > > > > > > + *       contributors may be used to endorse or promote
> > products
> > > > > > derived
> > > > > > > > > + *       from this software without specific prior written
> > permission.
> > > > > > > > > + *
> > > > > > > > > + *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> > > > > > > > CONTRIBUTORS
> > > > > > > > > + *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
> > > > INCLUDING,
> > > > > > BUT
> > > > > > > > NOT
> > > > > > > > > + *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
> > > > AND
> > > > > > > > FITNESS FOR
> > > > > > > > > + *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL
> > > > THE
> > > > > > > > COPYRIGHT
> > > > > > > > > + *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
> > > > INDIRECT,
> > > > > > > > INCIDENTAL,
> > > > > > > > > + *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
> > > > > > (INCLUDING,
> > > > > > > > BUT NOT
> > > > > > > > > + *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
> > > > SERVICES;
> > > > > > > > LOSS OF USE,
> > > > > > > > > + *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
> > > > > > CAUSED
> > > > > > > > AND ON ANY
> > > > > > > > > + *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
> > > > LIABILITY,
> > > > > > OR
> > > > > > > > TORT
> > > > > > > > > + *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY
> > > > WAY
> > > > > > OUT
> > > > > > > > OF THE USE
> > > > > > > > > + *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY
> > OF
> > > > SUCH
> > > > > > > > DAMAGE.
> > > > > > > > > + */
> > > > > > > > > +
> > > > > > > > > +#ifndef _RTE_ETH_DEV_INFO_H_ #define _RTE_ETH_DEV_INFO_H_
> > > > > > > > > +
> > > > > > > > > +
> > > > > > > > > +/*
> > > > > > > > > + * Placeholder for accessing device registers  */ struct
> > > > > > > > > +rte_dev_reg_info {
> > > > > > > > > +	void *buf; /**< Buffer for register */
> > > > > > > > > +	uint32_t offset; /**< Offset for 1st register to fetch
> > */
> > > > > > > > > +	uint32_t leng; /**< Number of registers to fetch */
> > > > > > > > > +	uint32_t version; /**< Device version */ };
> > > > > > > > > +
> > > > > > > > > +/*
> > > > > > > > > + * Placeholder for accessing device eeprom  */ struct
> > > > > > > > > +rte_dev_eeprom_info {
> > > > > > > > > +	void *buf; /**< Buffer for eeprom */
> > > > > > > > > +	uint32_t offset; /**< Offset for 1st eeprom location to
> > > > > > > > > +access
> > > > */
> > > > > > > > > +	uint32_t leng; /**< Length of eeprom region to access */
> > > > > > > > > +	uint32_t magic; /**< Device ID */ };
> > > > > > > > > +
> > > > > > > > > +/*
> > > > > > > > > + * Placeholder for accessing device ring parameters  */
> > > > > > > > > +struct rte_dev_ring_info {
> > > > > > > > > +	uint32_t rx_pending; /**< Number of outstanding Rx ring
> > */
> > > > > > > > > +	uint32_t tx_pending; /**< Number of outstanding Tx ring
> > */
> > > > > > > > > +	uint32_t rx_max_pending; /**< Maximum number of
> > > > outstanding
> > > > > > > > > +Rx
> > > > > > > > ring */
> > > > > > > > > +	uint32_t tx_max_pending; /**< Maximum number of
> > > > outstanding
> > > > > > > > > +Tx
> > > > > > > > ring
> > > > > > > > > +*/ };
> > > > > > > > > +
> > > > > > > > > +/*
> > > > > > > > > + * A data structure captures information as defined in
> > > > > > > > > +struct ifla_vf_info
> > > > > > > > > + * for user-space api
> > > > > > > > > + */
> > > > > > > > > +struct rte_dev_vf_info {
> > > > > > > > > +	uint32_t vf;
> > > > > > > > > +	uint8_t mac[ETHER_ADDR_LEN];
> > > > > > > > > +	uint32_t vlan;
> > > > > > > > > +	uint32_t tx_rate;
> > > > > > > > > +	uint32_t spoofchk;
> > > > > > > > > +};
> > > > > > > >
> > > > > > > >
> > > > > > > > Wonder what that structure is for?
> > > > > > > > I can't see it used in any function below?
> > > > > > > >
> > > > > > >
> > > > > > > Good catch, this is designed for other ethtool ops that I did
> > > > > > > not include in
> > > > > > this release, I will remove this from next fix.
> > > > > > >
> > > > > > > > > +
> > > > > > > > > +#endif /* _RTE_ETH_DEV_INFO_H_ */
> > > > > > > > > diff --git a/lib/librte_ether/rte_ethdev.c
> > > > > > > > > b/lib/librte_ether/rte_ethdev.c index 5a94654..186e85c
> > > > > > > > > 100644
> > > > > > > > > --- a/lib/librte_ether/rte_ethdev.c
> > > > > > > > > +++ b/lib/librte_ether/rte_ethdev.c
> > > > > > > > > @@ -2751,6 +2751,32 @@ rte_eth_dev_mac_addr_remove(uint8_t
> > > > > > > > port_id,
> > > > > > > > > struct ether_addr *addr)  }
> > > > > > > > >
> > > > > > > > >  int
> > > > > > > > > +rte_eth_dev_default_mac_addr_set(uint8_t port_id, struct
> > > > > > > > > +ether_addr
> > > > > > > > > +*addr) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +	const int index = 0;
> > > > > > > > > +	const uint32_t pool = 0;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	dev = &rte_eth_devices[port_id];
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> > > > >mac_addr_remove, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->mac_addr_add, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +
> > > > > > > > > +	/* Update NIC default MAC address*/
> > > > > > > > > +	(*dev->dev_ops->mac_addr_remove)(dev, index);
> > > > > > > > > +	(*dev->dev_ops->mac_addr_add)(dev, addr, index, pool);
> > > > > > > > > +
> > > > > > > > > +	/* Update default address in NIC data structure */
> > > > > > > > > +	ether_addr_copy(addr, &dev->data->mac_addrs[index]);
> > > > > > > > > +
> > > > > > > > > +	return 0;
> > > > > > > > > +}
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > >  rte_eth_dev_set_vf_rxmode(uint8_t port_id,  uint16_t vf,
> > > > > > > > >  				uint16_t rx_mode, uint8_t on)  { @@
> > > > -3627,3
> > > > > > +3653,136 @@
> > > > > > > > > rte_eth_remove_tx_callback(uint8_t port_id,
> > > > > > > > uint16_t queue_id,
> > > > > > > > >  	/* Callback wasn't found. */
> > > > > > > > >  	return -EINVAL;
> > > > > > > > >  }
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_reg_leng(uint8_t port_id) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg_length,
> > > > -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->get_reg_length)(dev);
> > > > > > > > > +}
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_reg_info(uint8_t port_id, struct
> > > > > > > > > +rte_dev_reg_info
> > > > > > > > > +*info) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_reg, -
> > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->get_reg)(dev, info); }
> > > > > > > >
> > > > > > > > Seems that *get_reg* stuff, will be really good addition for
> > > > > > > > DPDK debugging abilities.
> > > > > > > > Though, I'd suggest we change it a bit to make more generic
> > > > > > > > and
> > > > flexible:
> > > > > > > >
> > > > > > > > Introduce rte_eth_reg_read/rte_eth_reg_write(),
> > > > > > > > or probably even better rte_pcidev_reg_read
> > > > > > > > /rte_pcidev_reg_write at
> > > > > > EAL.
> > > > > > > > Something similar to what
> > > > > > > > port_pci_reg_read/port_pci_reg_write()
> > > > > > > > are doing now at testpmd.h.
> > > > > > > >
> > > > > > > > struct rte_pcidev_reg_info {
> > > > > > > >    const char *name;
> > > > > > > >    uint32_t endianes, bar, offset, size, count; };
> > > > > > > >
> > > > > > > > int rte_pcidev_reg_read(const struct rte_pci_device *, const
> > > > > > > > struct rte_pcidev_reg_info *, uint64_t *reg_val);
> > > > > > > >
> > > > > > > > Then:
> > > > > > > > int rte_eth_dev_get_reg_info(port_id, const struct
> > > > > > > > rte_pcidev_reg_info **info);
> > > > > > > >
> > > > > > > > So each device would store in info a pointer to an array of
> > > > > > > > it's register descriptions (finished by zero elem?).
> > > > > > > >
> > > > > > > > Then your ethtool (or any other upper layer) can do the
> > > > > > > > following to read all device regs:
> > > > > > > >
> > > > > > >
> > > > > > > The proposed reg info structure allows future improvement to
> > > > > > > support
> > > > > > individual register read/write.
> > > > > > > Also, because each NIC device has a very distinguish register
> > definition.
> > > > > > > So, the plan is to have more comprehensive interface to
> > > > > > > support query operation (for example, register name) before
> > > > > > > introduce individual/group
> > > > > > register access.
> > > > > > > Points taken, the support will be in future release.
> > > > > >
> > > > > > Sorry, didn't get you.
> > > > > > So you are ok to make these changes in next patch version?
> > > > > >
> > > > > I would like to get a consensus from dpdk community on how to
> > > > > provide
> > > > register information.
> > > >
> > > > Well, that' ok, but if it is just a trial patch that is not intended
> > > > to be applied, then you should mark it as RFC.
> > > >
> > > > > Currently, it's designed for debug dumping. The register
> > > > > information is very
> > > > hardware dependent.
> > > > > Need to consider current supported NIC device and future devices
> > > > > for
> > > > DPDK, so we won't make it a bulky interface.
> > > >
> > > > Ok, could you explain what exactly  concerns you in the approach
> > > > described above?
> > > > What part you feel is bulky?
> > > >
> > > > > > >
> > > > > > > > const struct rte_eth_dev_reg_info *reg_info; struct
> > > > > > > > rte_eth_dev_info dev_info;
> > > > > > > >
> > > > > > > > rte_eth_dev_info_get(pid, &dev_info);
> > > > > > > > rte_eth_dev_get_reg_info(port_id, &reg_info);
> > > > > > > >
> > > > > > > > for (i = 0; reg_info[i].name != NULL; i++) {
> > > > > > > >    ...
> > > > > > > >    rte_pcidev_read_reg(dev_info. pci_dev, reg_info[i], &v);
> > > > > > > >   ..
> > > > > > > > }
> > > > > > > >
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_eeprom_leng(uint8_t port_id) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> > > > >get_eeprom_length, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->get_eeprom_length)(dev);
> > > > > > > > > +}
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_get_eeprom(uint8_t port_id, struct
> > > > > > > > > +rte_dev_eeprom_info
> > > > > > > > > +*info) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_eeprom, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->get_eeprom)(dev, info); }
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_set_eeprom(uint8_t port_id, struct
> > > > > > > > > +rte_dev_eeprom_info
> > > > > > > > > +*info) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->set_eeprom, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->set_eeprom)(dev, info); }
> > > > > > > > > +
> > > > > > > > > +int
> > > > > > > > > +rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > > > > > > > +rte_dev_ring_info
> > > > > > > > > +*info) {
> > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > +
> > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > port_id);
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > +		return -ENODEV;
> > > > > > > > > +	}
> > > > > > > > > +
> > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> > > > > > > > ENOTSUP);
> > > > > > > > > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> > > > > > > >
> > > > > > > > I think it will be a useful addition to the ethdev API  to
> > > > > > > > have an ability to retrieve current RX/TX queue parameters.
> > > > > > > > Though again, it need to be more generic, so it could be
> > > > > > > > useful for
> > > > > > > > non- ethtool upper layer too.
> > > > > > > > So I suggest to modify it a bit.
> > > > > > > > Something like that:
> > > > > > > >
> > > > > > > > struct rte_eth_tx_queue_info {
> > > > > > > >     struct rte_eth_txconf txconf;
> > > > > > > >     uint32_t nb_tx_desc;
> > > > > > > >     uint32_t nb_max_tx_desc; /*max allowable TXDs for that
> > queue */
> > > > > > > >     uint32_t nb_tx_free;            /* number of free TXDs at
> > the moment
> > > > of
> > > > > > call.
> > > > > > > > */
> > > > > > > >     /* other tx queue data. */ };
> > > > > > > >
> > > > > > > > int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> > > > > > > > rte_eth_tx_queue_info *qinfo)
> > > > > > > >
> > > > > > > > Then, your upper layer ethtool wrapper, can implement yours
> > > > > > > > ethtool_get_ringparam() by:
> > > > > > > >
> > > > > > > >  ...
> > > > > > > >  struct rte_eth_tx_queue_info qinfo;
> > > > > > > > rte_ethdev_get_tx_queue_info(port, 0, &qinfo);
> > > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > > rte_eth_rx_queue_count(port, 0);
> > > > > > > >
> > > > > > > > Or probably even:
> > > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > > qinfo.nb_tx_free;
> > > > > > > >
> > > > > > > > Same for RX.
> > > > > > > >
> > > > > > > For now, this descriptor ring information is used by the ethtool
> > op.
> > > > > > > To make this interface simple, i.e. caller doesn't need to
> > > > > > > access other
> > > > > > queue information.
> > > > > >
> > > > > > I just repeat what I said to you in off-line conversation:
> > > > > > ethdev API is not equal ethtool API.
> > > > > > It is ok to add  a new function/structure to ethdev if it really
> > > > > > needed, but we should do mechanical one to one copy.
> > > > > > It is much better to add  a function/structure that would be
> > > > > > more generic, and suit other users, not only ethtool.
> > > > > > There is no point to have dozen functions in rte_ethdev API
> > > > > > providing similar information.
> > > > > > BTW, I don't see how API I proposed is much more  complex, then
> > > > > > yours
> > > > one.
> > > > > The ring parameter is a run-time information which is different
> > > > > than data
> > > > structure described in this discussion.
> > > >
> > > > I don't see how they are different.
> > > > Looking at ixgbe_get_ringparam(), it returns:
> > > > rx_max_pending - that's a static IXGBE PMD value (max possible
> > > > number of RXDs per one queue).
> > > > rx_pending - number of RXD currently in use by the HW for queue 0
> > > > (that information can be changed at each call).
> > > >
> > > > With the approach I suggesting - you can get same information for
> > > > each RX queue by calling rte_ethdev_get_rx_queue_info() and
> > > > rte_eth_rx_queue_count().
> > > > Plus you are getting other RX queue data.
> > > >
> > > > Another thing - what is practical usage of the information you
> > > > retrieving now by get_ringparam()?
> > > > Let say it returned to you: rx_max_pending=4096; rx_pending=128; How
> > > > that information would help to understand what is going on with the
> > device?
> > > > Without knowing  value of nb_tx_desc for the queue, you can't say is
> > > > you queue full or not.
> > > > Again, it could be that all your traffic going through some other
> > > > queue (not 0).
> > > > So from my point rte_eth_dev_get_ringparam()  usage is very limited,
> > > > and doesn't provide enough information about current queue state.
> > > >
> > > > Same thing applies for TX.
> > > >
> > >
> > > After careful review the suggestion in this comment, and review the
> > existing dpdk source code.
> > > I came to realize that neither rte_ethdev_get_rx_queue_info,
> > > rte_ethdev_get_tx_queue_info, struct rte_eth_rx_queue_info and struct
> > > rte_eth_tx_queue_info are available in the existing dpdk source code. I
> > could not make a patch based upon a set of non- existent API and data
> > structure.
> >
> > Right now, in dpdk.org source code, struct  rte_eth_dev_ring_info, struct
> > rte_dev_eeprom_info and struct  rte_dev_reg_info don't exist also.
> > Same as  all these functions:
> >
> > rte_eth_dev_default_mac_addr_set
> > rte_eth_dev_reg_length
> > rte_eth_dev_reg_info
> > rte_eth_dev_eeprom_length
> > rte_eth_dev_get_eeprom
> > rte_eth_dev_set_eeprom
> > rte_eth_dev_get_ringparam
> >
> > All this is a new API that's you are trying to add.
> > But, by some reason you consider it is ok to add 'struct
> > rte_eth_dev_ring_info', but couldn't add  struct
> > 'rte_ethdev_get_tx_queue_info'
> > That just doesn't make any sense to me.
> > In fact, I think our conversation is going in cycles.
> > If you are not happy with the suggested approach, please provide some
> > meaningful reason why.
> > Konstantin
> 
> It seems the new API aims at providing users a mechanism to quickly and
> gracefully migrate from using ethtool/ioctl calls.

I am fine with that goal in general.
But it doesn't mean that all ethool API should be pushed into rte_ethdev layer.
That's why  a shim layer on top of rte_ethdev is created -
it's goal is to provide for the upper layer an ethool-like API and
hide actual implementation based on rte_ethdev API inside.

>  The provided get/set
> ring param info is very similar to that of ethtool and facilitates the
> ethtool needs.

If rte_ethtool shim layer has to provide get/set ring_param API as it is - that's ok with me.
Though I don't see any good reason why rte_ethdev layer API should restrict itself with exactly the same ethtool-like API.
As I said before - current practical usage of rte_eth_dev_get_ringparam() looks quite limited.
Probably it just me, but I don't see how user can conclude what is the device state,
using  information provided by rte_eth_dev_get_ringparam().

My concern is - if we'll introduce rte_eth_dev_get_ringparam() as it suggested by Larry now,
one of two things would happen:
- no one except ethtool shim layer would use it.
- people will complain that it doesn't provide desired information. 
So, in next release we'll have to either introduce a new function and support 2 functions doing similar things
(code duplication), or modify existing API (ABI breakage pain).

It would be much better to introduce a new rte_ethdev API here that would be generic enough
and fit common needs, not only one particular case (ethtool).
After all, I don't think the changes I suggesting  differ that much from current approach.
Konstantin

> While additional enhancements to the API to provide additional details
> such as those you have suggested are certainly possible, I believe Larry
> is stating those ideas are outside the scope he has intended with the
> API introduction and that they should be discussed further and delivered
> in a future patch.

> 
> Does that seem reasonable?
> 
> Thanks,
> Dave
> 
> >
> > >
> > > > > It's the desire of this patch to separate each data structure to
> > > > > avoid cross
> > > > dependency.
> > > >
> > > > That's too cryptic to me.
> > > > Could you explain what cross dependency you are talking about?
> > > >
> > > > Konstantin
  

> > > > > > > > > > +int
> > > > > > > > > > +rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > > > > > > > > +rte_dev_ring_info
> > > > > > > > > > +*info) {
> > > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > > +
> > > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > > port_id);
> > > > > > > > > > +		return -ENODEV;
> > > > > > > > > > +	}
> > > > > > > > > > +
> > > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > > +		return -ENODEV;
> > > > > > > > > > +	}
> > > > > > > > > > +
> > > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> > > > > > > > > ENOTSUP);
> > > > > > > > > > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> > > > > > > > >
> > > > > > > > > I think it will be a useful addition to the ethdev API  to
> > > > > > > > > have an ability to retrieve current RX/TX queue parameters.
> > > > > > > > > Though again, it need to be more generic, so it could be
> > > > > > > > > useful for
> > > > > > > > > non- ethtool upper layer too.
> > > > > > > > > So I suggest to modify it a bit.
> > > > > > > > > Something like that:
> > > > > > > > >
> > > > > > > > > struct rte_eth_tx_queue_info {
> > > > > > > > >     struct rte_eth_txconf txconf;
> > > > > > > > >     uint32_t nb_tx_desc;
> > > > > > > > >     uint32_t nb_max_tx_desc; /*max allowable TXDs for that
> > > queue */
> > > > > > > > >     uint32_t nb_tx_free;            /* number of free TXDs at
> > > the moment
> > > > > of
> > > > > > > call.
> > > > > > > > > */
> > > > > > > > >     /* other tx queue data. */ };
> > > > > > > > >
> > > > > > > > > int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> > > > > > > > > rte_eth_tx_queue_info *qinfo)
> > > > > > > > >
> > > > > > > > > Then, your upper layer ethtool wrapper, can implement yours
> > > > > > > > > ethtool_get_ringparam() by:
> > > > > > > > >
> > > > > > > > >  ...
> > > > > > > > >  struct rte_eth_tx_queue_info qinfo;
> > > > > > > > > rte_ethdev_get_tx_queue_info(port, 0, &qinfo);
> > > > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > > > rte_eth_rx_queue_count(port, 0);
> > > > > > > > >
> > > > > > > > > Or probably even:
> > > > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > > > qinfo.nb_tx_free;
> > > > > > > > >
> > > > > > > > > Same for RX.
> > > > > > > > >
> > > > > > > > For now, this descriptor ring information is used by the ethtool
> > > op.
> > > > > > > > To make this interface simple, i.e. caller doesn't need to
> > > > > > > > access other
> > > > > > > queue information.
> > > > > > >
> > > > > > > I just repeat what I said to you in off-line conversation:
> > > > > > > ethdev API is not equal ethtool API.
> > > > > > > It is ok to add  a new function/structure to ethdev if it really
> > > > > > > needed, but we should do mechanical one to one copy.
> > > > > > > It is much better to add  a function/structure that would be
> > > > > > > more generic, and suit other users, not only ethtool.
> > > > > > > There is no point to have dozen functions in rte_ethdev API
> > > > > > > providing similar information.
> > > > > > > BTW, I don't see how API I proposed is much more  complex, then
> > > > > > > yours
> > > > > one.
> > > > > > The ring parameter is a run-time information which is different
> > > > > > than data
> > > > > structure described in this discussion.
> > > > >
> > > > > I don't see how they are different.
> > > > > Looking at ixgbe_get_ringparam(), it returns:
> > > > > rx_max_pending - that's a static IXGBE PMD value (max possible
> > > > > number of RXDs per one queue).
> > > > > rx_pending - number of RXD currently in use by the HW for queue 0
> > > > > (that information can be changed at each call).
> > > > >
> > > > > With the approach I suggesting - you can get same information for
> > > > > each RX queue by calling rte_ethdev_get_rx_queue_info() and
> > > > > rte_eth_rx_queue_count().
> > > > > Plus you are getting other RX queue data.
> > > > >
> > > > > Another thing - what is practical usage of the information you
> > > > > retrieving now by get_ringparam()?
> > > > > Let say it returned to you: rx_max_pending=4096; rx_pending=128; How
> > > > > that information would help to understand what is going on with the
> > > device?
> > > > > Without knowing  value of nb_tx_desc for the queue, you can't say is
> > > > > you queue full or not.
> > > > > Again, it could be that all your traffic going through some other
> > > > > queue (not 0).
> > > > > So from my point rte_eth_dev_get_ringparam()  usage is very limited,
> > > > > and doesn't provide enough information about current queue state.
> > > > >
> > > > > Same thing applies for TX.
> > > > >
> > > >
> > > > After careful review the suggestion in this comment, and review the
> > > existing dpdk source code.
> > > > I came to realize that neither rte_ethdev_get_rx_queue_info,
> > > > rte_ethdev_get_tx_queue_info, struct rte_eth_rx_queue_info and struct
> > > > rte_eth_tx_queue_info are available in the existing dpdk source code. I
> > > could not make a patch based upon a set of non- existent API and data
> > > structure.
> > >
> > > Right now, in dpdk.org source code, struct  rte_eth_dev_ring_info, struct
> > > rte_dev_eeprom_info and struct  rte_dev_reg_info don't exist also.
> > > Same as  all these functions:
> > >
> > > rte_eth_dev_default_mac_addr_set
> > > rte_eth_dev_reg_length
> > > rte_eth_dev_reg_info
> > > rte_eth_dev_eeprom_length
> > > rte_eth_dev_get_eeprom
> > > rte_eth_dev_set_eeprom
> > > rte_eth_dev_get_ringparam
> > >
> > > All this is a new API that's you are trying to add.
> > > But, by some reason you consider it is ok to add 'struct
> > > rte_eth_dev_ring_info', but couldn't add  struct
> > > 'rte_ethdev_get_tx_queue_info'
> > > That just doesn't make any sense to me.
> > > In fact, I think our conversation is going in cycles.
> > > If you are not happy with the suggested approach, please provide some
> > > meaningful reason why.
> > > Konstantin
> >
> > It seems the new API aims at providing users a mechanism to quickly and
> > gracefully migrate from using ethtool/ioctl calls.
> 
> I am fine with that goal in general.
> But it doesn't mean that all ethool API should be pushed into rte_ethdev layer.
> That's why  a shim layer on top of rte_ethdev is created -
> it's goal is to provide for the upper layer an ethool-like API and
> hide actual implementation based on rte_ethdev API inside.
> 
> >  The provided get/set
> > ring param info is very similar to that of ethtool and facilitates the
> > ethtool needs.
> 

Actually a quick questions to you guys:
Looking at linux struct ethtool_ringparam description, I am seeing:
http://lxr.free-electrons.com/source/include/uapi/linux/ethtool.h:
@rx_pending: Current maximum number of pending entries per RX ring

And all linux driver I looked at (ixgbe,i40e,virtio,vmxbet3) returns number
of configured RX descriptors for queue 0.
In DPDK terms: nb_desc for the queue.

While in Larry's patch,  ixgbe_get_ringparam() for rx_pending returns
number of RX descriptors that are in HW use (for queue 0).
So, did you intentionally change linux ethool get_ringparam() behaviour here,
or is it just a mistake?

Konstantin
  

...
> > > > > > > > > > +
> > > > > > > > > > +int
> > > > > > > > > > +rte_eth_dev_get_ringparam(uint8_t port_id, struct
> > > > > > > > > > +rte_dev_ring_info
> > > > > > > > > > +*info) {
> > > > > > > > > > +	struct rte_eth_dev *dev;
> > > > > > > > > > +
> > > > > > > > > > +	if (!rte_eth_dev_is_valid_port(port_id)) {
> > > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port_id=%d\n",
> > > > > port_id);
> > > > > > > > > > +		return -ENODEV;
> > > > > > > > > > +	}
> > > > > > > > > > +
> > > > > > > > > > +	if ((dev= &rte_eth_devices[port_id]) == NULL) {
> > > > > > > > > > +		PMD_DEBUG_TRACE("Invalid port device\n");
> > > > > > > > > > +		return -ENODEV;
> > > > > > > > > > +	}
> > > > > > > > > > +
> > > > > > > > > > +	FUNC_PTR_OR_ERR_RET(*dev->dev_ops->get_ringparam, -
> > > > > > > > > ENOTSUP);
> > > > > > > > > > +	return (*dev->dev_ops->get_ringparam)(dev, info); }
> > > > > > > > >
> > > > > > > > > I think it will be a useful addition to the ethdev API  to
> > > > > > > > > have an ability to retrieve current RX/TX queue parameters.
> > > > > > > > > Though again, it need to be more generic, so it could be
> > > > > > > > > useful for
> > > > > > > > > non- ethtool upper layer too.
> > > > > > > > > So I suggest to modify it a bit.
> > > > > > > > > Something like that:
> > > > > > > > >
> > > > > > > > > struct rte_eth_tx_queue_info {
> > > > > > > > >     struct rte_eth_txconf txconf;
> > > > > > > > >     uint32_t nb_tx_desc;
> > > > > > > > >     uint32_t nb_max_tx_desc; /*max allowable TXDs for that
> > > queue */
> > > > > > > > >     uint32_t nb_tx_free;            /* number of free TXDs at the
> > > moment
> > > > > of
> > > > > > > call.
> > > > > > > > > */
> > > > > > > > >     /* other tx queue data. */ };
> > > > > > > > >
> > > > > > > > > int rte_etdev_get_tx_queue_info(portid, queue_id, struct
> > > > > > > > > rte_eth_tx_queue_info *qinfo)
> > > > > > > > >
> > > > > > > > > Then, your upper layer ethtool wrapper, can implement yours
> > > > > > > > > ethtool_get_ringparam() by:
> > > > > > > > >
> > > > > > > > >  ...
> > > > > > > > >  struct rte_eth_tx_queue_info qinfo;
> > > > > > > > > rte_ethdev_get_tx_queue_info(port, 0, &qinfo);
> > > > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > > > rte_eth_rx_queue_count(port, 0);
> > > > > > > > >
> > > > > > > > > Or probably even:
> > > > > > > > > ring_param->tx_pending = qinfo.nb_tx_desc -
> > > > > > > > > qinfo.nb_tx_free;
> > > > > > > > >
> > > > > > > > > Same for RX.
> > > > > > > > >
> > > > > > > > For now, this descriptor ring information is used by the ethtool op.
> > > > > > > > To make this interface simple, i.e. caller doesn't need to
> > > > > > > > access other
> > > > > > > queue information.
> > > > > > >
> > > > > > > I just repeat what I said to you in off-line conversation:
> > > > > > > ethdev API is not equal ethtool API.
> > > > > > > It is ok to add  a new function/structure to ethdev if it really
> > > > > > > needed, but we should do mechanical one to one copy.
> > > > > > > It is much better to add  a function/structure that would be
> > > > > > > more generic, and suit other users, not only ethtool.
> > > > > > > There is no point to have dozen functions in rte_ethdev API
> > > > > > > providing similar information.
> > > > > > > BTW, I don't see how API I proposed is much more  complex, then
> > > > > > > yours
> > > > > one.
> > > > > > The ring parameter is a run-time information which is different
> > > > > > than data
> > > > > structure described in this discussion.
> > > > >
> > > > > I don't see how they are different.
> > > > > Looking at ixgbe_get_ringparam(), it returns:
> > > > > rx_max_pending - that's a static IXGBE PMD value (max possible
> > > > > number of RXDs per one queue).
> > > > > rx_pending - number of RXD currently in use by the HW for queue 0
> > > > > (that information can be changed at each call).
> > > > >
> > > > > With the approach I suggesting - you can get same information for
> > > > > each RX queue by calling rte_ethdev_get_rx_queue_info() and
> > > > > rte_eth_rx_queue_count().
> > > > > Plus you are getting other RX queue data.
> > > > >
> > > > > Another thing - what is practical usage of the information you
> > > > > retrieving now by get_ringparam()?
> > > > > Let say it returned to you: rx_max_pending=4096; rx_pending=128; How
> > > > > that information would help to understand what is going on with the
> > > device?
> > > > > Without knowing  value of nb_tx_desc for the queue, you can't say is
> > > > > you queue full or not.
> > > > > Again, it could be that all your traffic going through some other
> > > > > queue (not 0).
> > > > > So from my point rte_eth_dev_get_ringparam()  usage is very limited,
> > > > > and doesn't provide enough information about current queue state.
> > > > >
> > > > > Same thing applies for TX.
> > > > >
> > > >
> > > > After careful review the suggestion in this comment, and review the
> > > existing dpdk source code.
> > > > I came to realize that neither rte_ethdev_get_rx_queue_info,
> > > > rte_ethdev_get_tx_queue_info, struct rte_eth_rx_queue_info and struct
> > > > rte_eth_tx_queue_info are available in the existing dpdk source code. I
> > > could not make a patch based upon a set of non- existent API and data
> > > structure.
> > >
> > > Right now, in dpdk.org source code, struct  rte_eth_dev_ring_info, struct
> > > rte_dev_eeprom_info and struct  rte_dev_reg_info don't exist also.
> > > Same as  all these functions:
> > >
> > > rte_eth_dev_default_mac_addr_set
> > > rte_eth_dev_reg_length
> > > rte_eth_dev_reg_info
> > > rte_eth_dev_eeprom_length
> > > rte_eth_dev_get_eeprom
> > > rte_eth_dev_set_eeprom
> > > rte_eth_dev_get_ringparam
> > >
> > > All this is a new API that's you are trying to add.
> > > But, by some reason you consider it is ok to add 'struct
> > > rte_eth_dev_ring_info', but couldn't add  struct
> > > 'rte_ethdev_get_tx_queue_info'
> > > That just doesn't make any sense to me.
> > > In fact, I think our conversation is going in cycles.
> > > If you are not happy with the suggested approach, please provide some
> > > meaningful reason why.
> >
> > All the API's and data structure that you have questions in this comment are available from the submitted patch, you could run to
> see
> > if it works.
> > What I learned from your comments are a couple of API name and incomplete data structure description, so I could not make my
> > patch according to your suggestion.
> > If you strongly feel the API's that you are proposing are useful for get_ringparam(), please create a patch and submit it for
> evaluation.
> 
> Ok, I'll try to create a patch in next few days.
> Hopefully it will make things clearer to you and fit everyone.
> Konstantin

Here is the patch I submitted for proposed changes at re_ethdev and PMDs, as discussed:
http://dpdk.org/dev/patchwork/patch/5482/
Please have a look.

With that, get_ringraram for ethtool shim layer, would look something like:

int
rte_ethtool_get_ringparam(uint8_t port_id,
	struct ethtool_ringparam *ring_param)
{
	struct rte_eth_rx_qinfo rx_qinfo;
                struct rte_eth_tx_qinfo tx_qinfo;
	int status;

                if ((status = rte_eth_rx_queue_info_get (port_id, 0, &rx_qinfo)) != 0)
		return status;

               if ((status = rte_eth_tx_queue_info_get (port_id, 0, &tx_qinfo)) != 0)
		return status;

              memset(ring_param, 0, sizeof(*ring_param)

              ring_param->rx_pending = rx_qinfo.nb_desc;
              ring_param->tx_pending = tx_qinfo.nb_desc;
              ring_param->rx_max_pending = rx_qinfo.max_desc;
              ring_param->tx_max_pending = tx_qinfo.max_desc;

              return 0;
}


As you can see, the changes at ethtool are minimal, while it provides desired info.
From other side, we have much more generic and easily extendable API at rte_ethdev.

Konstantin

> 
> > As coded in this patch, it's a simple and clean interface and most important, it can be validated and it works.
> >
> > > Konstantin
> > >
> > > >
> > > > > > It's the desire of this patch to separate each data structure to
> > > > > > avoid cross
> > > > > dependency.
> > > > >
> > > > > That's too cryptic to me.
> > > > > Could you explain what cross dependency you are talking about?
> > > > >
> > > > > Konstantin