Re: [dpdk-dev] [PATCH V19 1/4] bus/pci: introduce device hot unplug handle

["Ananyev, Konstantin" <konstantin.ananyev@intel.com>]

> -----Original Message-----
> From: Guo, Jia
> Sent: Friday, April 6, 2018 11:57 AM
> To: stephen@networkplumber.org; Richardson, Bruce <bruce.richardson@intel.com>; Yigit, Ferruh <ferruh.yigit@intel.com>; Ananyev,
> Konstantin <konstantin.ananyev@intel.com>; gaetan.rivet@6wind.com; Wu, Jingjing <jingjing.wu@intel.com>; thomas@monjalon.net;
> motih@mellanox.com; Van Haaren, Harry <harry.van.haaren@intel.com>; Tan, Jianfeng <jianfeng.tan@intel.com>
> Cc: jblunck@infradead.org; shreyansh.jain@nxp.com; dev@dpdk.org; Guo, Jia <jia.guo@intel.com>; Zhang, Helin <helin.zhang@intel.com>
> Subject: [PATCH V19 1/4] bus/pci: introduce device hot unplug handle
> 
> As of device hot unplug, we need some preparatory measures so that we will
> not encounter memory fault after device be plug out of the system,
> and also let we could recover the running data path but not been break.
> This allows the buses to handle device hot unplug event.
> The patch only enable the ops in pci bus, when handle device hot unplug
> event, remap a dummy memory to avoid bus read/write error.
> Other buses could accordingly implement this ops specific by themselves.
> 
> Signed-off-by: Jeff Guo <jia.guo@intel.com>
> ---
> v19->v18:
> fix some typo and squeeze patch
> ---
>  drivers/bus/pci/pci_common.c            | 42 +++++++++++++++++++++++++++++++++
>  drivers/bus/pci/pci_common_uio.c        | 32 +++++++++++++++++++++++++
>  drivers/bus/pci/private.h               | 12 ++++++++++
>  lib/librte_eal/common/include/rte_bus.h | 15 ++++++++++++
>  4 files changed, 101 insertions(+)
> 
> diff --git a/drivers/bus/pci/pci_common.c b/drivers/bus/pci/pci_common.c
> index 2a00f36..09192ed 100644
> --- a/drivers/bus/pci/pci_common.c
> +++ b/drivers/bus/pci/pci_common.c
> @@ -474,6 +474,47 @@ pci_find_device(const struct rte_device *start, rte_dev_cmp_t cmp,
>  }
> 
>  static int
> +pci_handle_hot_unplug(struct rte_device *dev)
> +{
> +	struct rte_pci_device *pdev;
> +	int ret;
> +
> +	if (dev == NULL)
> +		return -EINVAL;
> +
> +	pdev = RTE_DEV_TO_PCI(dev);
> +
> +	/* remap resources for devices */
> +	switch (pdev->kdrv) {
> +	case RTE_KDRV_VFIO:
> +#ifdef VFIO_PRESENT
> +		/* TODO */
> +#endif
> +		break;
> +	case RTE_KDRV_IGB_UIO:
> +	case RTE_KDRV_UIO_GENERIC:
> +		if (rte_eal_using_phys_addrs()) {
> +			/* map resources for devices that use uio */
> +			ret = pci_uio_remap_resource(pdev);
> +		}
> +		break;
> +	case RTE_KDRV_NIC_UIO:
> +		ret = pci_uio_remap_resource(pdev);
> +		break;
> +	default:
> +		RTE_LOG(DEBUG, EAL,
> +			"  Not managed by a supported kernel driver, skipped\n");
> +		ret = -1;
> +		break;
> +	}
> +
> +	if (ret != 0)
> +		RTE_LOG(ERR, EAL, "failed to handle hot unplug of %s",
> +			pdev->name);
> +	return ret;
> +}
> +
> +static int
>  pci_plug(struct rte_device *dev)
>  {
>  	return pci_probe_all_drivers(RTE_DEV_TO_PCI(dev));
> @@ -503,6 +544,7 @@ struct rte_pci_bus rte_pci_bus = {
>  		.unplug = pci_unplug,
>  		.parse = pci_parse,
>  		.get_iommu_class = rte_pci_get_iommu_class,
> +		.handle_hot_unplug = pci_handle_hot_unplug,
>  	},
>  	.device_list = TAILQ_HEAD_INITIALIZER(rte_pci_bus.device_list),
>  	.driver_list = TAILQ_HEAD_INITIALIZER(rte_pci_bus.driver_list),
> diff --git a/drivers/bus/pci/pci_common_uio.c b/drivers/bus/pci/pci_common_uio.c
> index 54bc20b..31a4094 100644
> --- a/drivers/bus/pci/pci_common_uio.c
> +++ b/drivers/bus/pci/pci_common_uio.c
> @@ -146,6 +146,38 @@ pci_uio_unmap(struct mapped_pci_resource *uio_res)
>  	}
>  }
> 
> +/* remap the PCI resource of a PCI device in private virtual memory */
> +int
> +pci_uio_remap_resource(struct rte_pci_device *dev)
> +{
> +	int i;
> +	void *map_address;
> +
> +	if (dev == NULL)
> +		return -1;
> +
> +	/* Remap all BARs */
> +	for (i = 0; i != PCI_MAX_RESOURCE; i++) {
> +		/* skip empty BAR */
> +		if (dev->mem_resource[i].phys_addr == 0)
> +			continue;
> +		pci_unmap_resource(dev->mem_resource[i].addr,
> +				(size_t)dev->mem_resource[i].len);
> +		map_address = pci_map_resource(
> +				dev->mem_resource[i].addr, -1, 0,
> +				(size_t)dev->mem_resource[i].len,
> +				MAP_ANONYMOUS | MAP_FIXED);

Wouldn't it be possible to use  mremap() here?
To do munmap/mmap in one go?


> +		if (map_address == MAP_FAILED) {
> +			RTE_LOG(ERR, EAL,
> +				"Cannot remap resource for device %s\n",
> +				dev->name);
> +			return -1;
> +		}
> +	}
> +
> +	return 0;
> +}
> +