Re: [dpdk-dev] [PATCH v2 8/8] ethdev: Decouple interrupt handling from PCI device

[Shreyansh Jain <shreyansh.jain@nxp.com>]

On Tuesday 20 December 2016 04:21 PM, Jan Blunck wrote:
> On Tue, Nov 22, 2016 at 1:57 PM, Shreyansh Jain <shreyansh.jain@nxp.com> wrote:
>> On Monday 21 November 2016 10:25 PM, Jan Blunck wrote:
>>>
>>> The struct rte_intr_handle is an abstraction layer for different types of
>>> interrupt mechanisms. It is embedded in the low-level device (e.g. PCI).
>>> On allocation of a struct rte_eth_dev a reference to the intr_handle
>>> should be stored for devices supporting interrupts.
>>>
>>> Signed-off-by: Jan Blunck <jblunck@infradead.org>
>>> ---
>>>  lib/librte_ether/rte_ethdev.c | 18 ++++++++++++++++--
>>>  lib/librte_ether/rte_ethdev.h |  1 +
>>>  2 files changed, 17 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
>>> index 4288577..4ecea50 100644
>>> --- a/lib/librte_ether/rte_ethdev.c
>>> +++ b/lib/librte_ether/rte_ethdev.c
>>> @@ -258,6 +258,7 @@ rte_eth_dev_pci_probe(struct rte_pci_driver *pci_drv,
>>>                         rte_panic("Cannot allocate memzone for private
>>> port data\n");
>>>         }
>>>         eth_dev->pci_dev = pci_dev;
>>> +       eth_dev->intr_handle = &pci_dev->intr_handle;

(#) See below.

>>>         eth_dev->driver = eth_drv;
>>>         eth_dev->data->rx_mbuf_alloc_failed = 0;
>>>
>>> @@ -2543,7 +2544,13 @@ rte_eth_dev_rx_intr_ctl(uint8_t port_id, int epfd,
>>> int op, void *data)
>>>         RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
>>>
>>>         dev = &rte_eth_devices[port_id];
>>> -       intr_handle = &dev->pci_dev->intr_handle;
>>> +
>>> +       if (!dev->intr_handle) {
>>> +               RTE_PMD_DEBUG_TRACE("RX Intr handle unset\n");
>>> +               return -ENOTSUP;
>>> +       }
>>> +
>>> +       intr_handle = dev->intr_handle;
>>>         if (!intr_handle->intr_vec) {
>>>                 RTE_PMD_DEBUG_TRACE("RX Intr vector unset\n");
>>>                 return -EPERM;
>>> @@ -2603,7 +2610,12 @@ rte_eth_dev_rx_intr_ctl_q(uint8_t port_id, uint16_t
>>> queue_id,
>>>                 return -EINVAL;
>>>         }
>>>
>>> -       intr_handle = &dev->pci_dev->intr_handle;
>>> +       if (!dev->intr_handle) {
>>> +               RTE_PMD_DEBUG_TRACE("RX Intr handle unset\n");
>>> +               return -ENOTSUP;
>>> +       }
>>> +
>>> +       intr_handle = dev->intr_handle;
>>>         if (!intr_handle->intr_vec) {
>>>                 RTE_PMD_DEBUG_TRACE("RX Intr vector unset\n");
>>>                 return -EPERM;
>>> @@ -3205,6 +3217,8 @@ rte_eth_copy_pci_info(struct rte_eth_dev *eth_dev,
>>> struct rte_pci_device *pci_de
>>>                 return;
>>>         }
>>>
>>> +       eth_dev->intr_handle = &pci_dev->intr_handle;
>>> +
>>>         eth_dev->data->dev_flags = 0;
>>>         if (pci_dev->driver->drv_flags & RTE_PCI_DRV_INTR_LSC)
>>>                 eth_dev->data->dev_flags |= RTE_ETH_DEV_INTR_LSC;
>>> diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
>>> index 3adbb2b..f1f656a 100644
>>> --- a/lib/librte_ether/rte_ethdev.h
>>> +++ b/lib/librte_ether/rte_ethdev.h
>>> @@ -1629,6 +1629,7 @@ struct rte_eth_dev {
>>>         const struct eth_driver *driver;/**< Driver for this device */
>>>         const struct eth_dev_ops *dev_ops; /**< Functions exported by PMD
>>> */
>>>         struct rte_pci_device *pci_dev; /**< PCI info. supplied by probing
>>> */
>>> +       struct rte_intr_handle *intr_handle; /**< Device interrupt handle
>>> */
>>>         /** User application callbacks for NIC interrupts */
>>>         struct rte_eth_dev_cb_list link_intr_cbs;
>>>         /**
>>>
>>
>> Is there another patch which replaces all uses of
>> eth_dev->pci_dev->intr_handle with that of eth_dev->intr_handle?
>>
>> Now that eth_dev has a reference which is initialized as early as probe, we
>> should use that rather than pci_dev->intr_handle for PCI PMDs.
>>
>
> I've added this indirection because it is required for the ethdev
> function. The drivers shouldn't use the indirection through ethdev to
> access the intr_handle because they do have direct access to the
> device.

This is my mistake - I posted my review at a wrong location. It should 
have been at (#) above.

My intention was that we should use ETH_DEV_INTR_HANDLE like macro for 
accessing intr_handle.

Anyways, I see that you have already posted v3. I will review that. You 
can consider this comment as retracted from my side.

>
>
>> OR maybe, ETH_DEV_INTR_HANDLE() like macro which you have introduced in
>> i40e_ethdev.h.
>>
>> -
>> Shreyansh
>

-
Shreyansh