Re: [dpdk-dev] [PATCH v2] net/i40e: fix device hotplug remove

[Ye Xiaolong <xiaolong.ye@intel.com>]

On 11/07, Di, ChenxuX wrote:
>Hi,
>
>> -----Original Message-----
>> From: Ye, Xiaolong
>> Sent: Thursday, November 7, 2019 11:11 AM
>> To: Di, ChenxuX <chenxux.di@intel.com>
>> Cc: dev@dpdk.org; Lu, Wenzhuo <wenzhuo.lu@intel.com>; Yang, Qiming
>> <qiming.yang@intel.com>; Yigit, Ferruh <ferruh.yigit@intel.com>
>> Subject: Re: [dpdk-dev] [PATCH v2] net/i40e: fix device hotplug remove
>> 
>> Hi,
>> 
>> On 11/07, Di, ChenxuX wrote:
>> >Hi, Xiaolong
>> >
>> >About pf and vf for fm10k, e1000, igb, ixgbe, ice, i40e; only i40e pf
>> >and ixgbe vf use
>> 
>> For fm10k, e1000, igb, ice, please check whether they have called
>> rte_eth_dev_release_port in the remove ops.
>> 
>
>The function rte_eth_dev_pci_generic_remove will call rte_eth_dev_release_port
>After the uninit function of drivers.

Emm, yes, just double checked, rte_eth_dev_pci_generic_remove will call rte_eth_dev_pci_release
which finally invokes rte_eth_dev_release_port.

>
>> >Rte_eth_dev_destory() in their remove function, others have already use the
>> generic remove function.
>> >
>> >I have send another patch about ixgbe because they fix different commits.
>> 
>> For i40e and ixgbe, please address the comments I raised in previous mail.
>> 
>
>When app detach a port without a close, it will call the remove function,
>What will uninit the port and rte_eth_dev_release_port.

You'll still need to handle i40e_vf_representor_uninit.


Thanks,
Xiaolong
>
>Thx
>Di Chenxu
>
>> Thanks,
>> Xiaolong
>> 
>> >
>> >Thanks
>> >Di Chenxu
>> >
>> >> -----Original Message-----
>> >> From: Ye, Xiaolong
>> >> Sent: Thursday, November 7, 2019 10:50 AM
>> >> To: Di, ChenxuX <chenxux.di@intel.com>
>> >> Cc: dev@dpdk.org; Lu, Wenzhuo <wenzhuo.lu@intel.com>; Yang, Qiming
>> >> <qiming.yang@intel.com>; Yigit, Ferruh <ferruh.yigit@intel.com>
>> >> Subject: Re: [dpdk-dev] [PATCH v2] net/i40e: fix device hotplug
>> >> remove
>> >>
>> >> On 11/06, Di ChenxuX wrote:
>> >> >testpmd will occur infinite loops when device hotplug remove.
>> >> >We can fix the issue by using the pci generic remove function
>> >> >
>> >> >Fixes: ac89d46096d5 ("net/i40e: release port upon close")
>> >> >
>> >> >Signed-off-by: Di ChenxuX <chenxux.di@intel.com>
>> >> >
>> >> >v2:
>> >> >change the code in driver.
>> >> >---
>> >> > drivers/net/i40e/i40e_ethdev.c | 12 +-----------
>> >> > 1 file changed, 1 insertion(+), 11 deletions(-)
>> >> >
>> >> >diff --git a/drivers/net/i40e/i40e_ethdev.c
>> >> >b/drivers/net/i40e/i40e_ethdev.c index 77a46832c..64842867d 100644
>> >> >--- a/drivers/net/i40e/i40e_ethdev.c
>> >> >+++ b/drivers/net/i40e/i40e_ethdev.c
>> >> >@@ -694,17 +694,7 @@ eth_i40e_pci_probe(struct rte_pci_driver
>> >> >*pci_drv __rte_unused,
>> >> >
>> >> > static int eth_i40e_pci_remove(struct rte_pci_device *pci_dev)  {
>> >> >-	struct rte_eth_dev *ethdev;
>> >> >-
>> >> >-	ethdev = rte_eth_dev_allocated(pci_dev->device.name);
>> >> >-	if (!ethdev)
>> >> >-		return -ENODEV;
>> >> >-
>> >> >-
>> >> >-	if (ethdev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
>> >> >-		return rte_eth_dev_destroy(ethdev,
>> >> i40e_vf_representor_uninit);
>> >> >-	else
>> >> >-		return rte_eth_dev_destroy(ethdev, eth_i40e_dev_uninit);
>> >> >+	return rte_eth_dev_pci_generic_remove(pci_dev,
>> >> >+eth_i40e_dev_uninit);
>> >>
>> >> What about i40e_vf_representor_uninit, you also need to take it into
>> >> consideration.
>> >>
>> >> And you should also call rte_eth_dev_release_port in remove ops in
>> >> case user detach a port directly without closing it first.
>> >>
>> >> This comment also applies your ixgbe fix.
>> >>
>> >>
>> >> Thanks,
>> >> Xiaolong
>> >>
>> >>
>> >> > }
>> >> >
>> >> > static struct rte_pci_driver rte_i40e_pmd = {
>> >> >--
>> >> >2.17.1
>> >> >