Re: [dpdk-dev] [PATCH v3 22/28] eal/pci: Add pci_close_all_drivers

["Qiu, Michael" <michael.qiu@intel.com>]

On 12/9/2014 2:33 PM, Tetsuya Mukawa wrote:
> The function tries to find a driver for the specified device, and then
> close the driver.
>
> Signed-off-by: Tetsuya Mukawa <mukawa@igel.co.jp>
> ---
>  lib/librte_eal/common/eal_common_pci.c | 19 +++++++++++++++++++
>  1 file changed, 19 insertions(+)
>
> diff --git a/lib/librte_eal/common/eal_common_pci.c b/lib/librte_eal/common/eal_common_pci.c
> index 1e3efea..b404ee0 100644
> --- a/lib/librte_eal/common/eal_common_pci.c
> +++ b/lib/librte_eal/common/eal_common_pci.c
> @@ -100,6 +100,7 @@ static struct rte_devargs *pci_devargs_lookup(struct rte_pci_device *dev)
>  }
>  
>  #define INVOKE_PROBE	(0)
> +#define INVOKE_CLOSE	(1)
>  
>  static int
>  pci_invoke_all_drivers(struct rte_pci_device *dev, int type)
> @@ -112,6 +113,11 @@ pci_invoke_all_drivers(struct rte_pci_device *dev, int type)
>  		case INVOKE_PROBE:
>  			rc = rte_eal_pci_probe_one_driver(dr, dev);
>  			break;
> +#if defined(RTE_LIBRTE_EAL_HOTPLUG) && defined(RTE_LIBRTE_EAL_LINUXAPP)
> +		case INVOKE_CLOSE:
> +			rc = rte_eal_pci_close_one_driver(dr, dev);
> +			break;
> +#endif

Here comments as below :)
>  		}
>  		if (rc < 0)
>  			/* negative value is an error */
> @@ -135,6 +141,19 @@ pci_probe_all_drivers(struct rte_pci_device *dev)
>  	return pci_invoke_all_drivers(dev, INVOKE_PROBE);
>  }
>  
> +#if defined(RTE_LIBRTE_EAL_HOTPLUG) && defined(RTE_LIBRTE_EAL_LINUXAPP)
> +/*
> + * If vendor/device ID match, call the devclose() function of all
> + * registered driver for the given device. Return -1 if initialization
> + * failed, return 1 if no driver is found for this device.
> + */
> +static int
> +pci_close_all_drivers(struct rte_pci_device *dev)
> +{
> +	return pci_invoke_all_drivers(dev, INVOKE_CLOSE);
> +}
> +#endif /* RTE_LIBRTE_EAL_HOTPLUG & RTE_LIBRTE_EAL_LINUXAPP */

If we do not just use #endif here, instead we could use:
+#else
+static inline int
+pci_close_all_drivers(struct rte_pci_device *dev) { return -1; }
+#endif

Then we do not need to do macros check when calling this function.
This could be apply to other patch in your patch set.

But just as one advice.

*__*

Thomas,

I don't know why *dpdk* has lots of "#if defined ---> #endif" within a
function. It is so ugly, we can avoid this, and do it in header files.

I just format on patch of vfio to avoid this. Since every place will do
VFIO_PRESENT checking before calling pci_vfio_is_enabled().
I see one of Tetsuya's patch also do this check.

Thanks,
Michael
> +
>  /*
>   * Scan the content of the PCI bus, and call the devinit() function for
>   * all registered drivers that have a matching entry in its id_table