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

["Guo, Jia" <jia.guo@intel.com>]

On 4/10/2018 1:47 AM, Ananyev, Konstantin wrote:
>
>> -----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?
>
since i don't thinks is need to add a new function to remap resource, so 
just use the exist map function with specif flag to handler it. what i 
want to say is if my implement have problem or what is
the mremap's  advantage for that, if so please let me know, thank.
>> +		if (map_address == MAP_FAILED) {
>> +			RTE_LOG(ERR, EAL,
>> +				"Cannot remap resource for device %s\n",
>> +				dev->name);
>> +			return -1;
>> +		}
>> +	}
>> +
>> +	return 0;
>> +}
>> +