Re: [PATCH v1 1/3] drivers/net/i40e: Add ethdev functions

-----Original Message-----
From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Remy Horton
Sent: Wednesday, February 24, 2016 10:32 AM
To: Zhang, Helin; Xie, Huawei
Cc: dev@dpdk.org
Subject: Re: [dpdk-dev] [PATCH v1 1/3] drivers/net/i40e: Add ethdev functions

Comments inline.

..Remy

On 23/02/2016 02:06, Zhang, Helin wrote:
 >
 >> +static inline int
 >> +i40e_read_regs(struct i40e_hw *hw, const struct reg_info *reg,
 >> +	       uint32_t *reg_buf)
 >> +{
 >> +	unsigned int i;
 >> +
 >> +	for (i = 0; i < reg->count; i++)
 >> +		reg_buf[i] = I40E_READ_REG(hw,
 >> +			reg->base_addr + i * reg->stride);
 >> +	return reg->count;
 >> +}
 >  From FVL5, some registers should be read by AQ commands, otherwise
it may fail to
 > read without any warning.
 > Please see my patches of which registers should be read by AQ commands.
 > Please check i40e_osdep.h from below link. Thanks!
 > http://www.dpdk.org/dev/patchwork/patch/10654/

Ok - will change for v2.

I noticed that other patches in the same patchset expose extra registers
- are these new or were they simply not exposed previously?


 >> +	/* Only support doing full dump */
 >> +	if (regs->offset != 0 && 0)
 > '&& 0' means it will never be false, right?
 > Anything wrong here?

Oops - some dead code that slipped through.. :)


 >> +		return -ENOTSUP;
 > A message before this return to tell the uers what happened would be
better.

Will add these into v2.


 >> +static int i40e_get_eeprom_length(__rte_unused struct rte_eth_dev *dev)
 > Why needs __rte_unused?

Good point - surprised the compiler did not complain about them, as they
are not supposed to be there..


 >> +static void i40e_set_default_mac_addr(struct rte_eth_dev *dev,
 >> +				      struct ether_addr *mac_addr)
 >> +{
 >> +	struct i40e_hw *hw = I40E_DEV_PRIVATE_TO_HW(dev->data-
 >>> dev_private);
 >> +
 >> +	/* Flags: 0x3 updates port address */
 >> +	i40e_aq_mac_address_write(hw, 0x3, mac_addr->addr_bytes,
 >> NULL); }
 > Checks are needed before writing the MAC address.

Will look into this.


 >> +struct reg_info {
 >> +	uint32_t base_addr;
 >> +	uint32_t count;
 >> +	uint32_t stride;
 >> +	const char *name;
 >> +} reg_info;
 > I think array definition shouldn't be added into a header file,
otherwise any .c source
 > file which includes that header file will define that.

Since it is quite a large table I think this approach, which is also
used in ixgbe, is the lesser of evils. i40e_ethdev.c itself is already
pretty big, and would prefer to avoid giving a driver-specific table
non-static visibility until it actually has to be used from other
compilation units.

Why not to have a separate .h file, specially for registers table definition?