On 7/12/2019 12:02 PM, Vamsi Krishna Attunuru wrote:
> 
> 
> 
> --------------------------------------------------------------------------------
> *From:* dev <dev-boun...@dpdk.org> on behalf of Ferruh Yigit
> <ferruh.yi...@intel.com>
> *Sent:* Thursday, July 11, 2019 9:52 PM
> *To:* Vamsi Krishna Attunuru; dev@dpdk.org
> *Cc:* olivier.m...@6wind.com; arybche...@solarflare.com
> *Subject:* [EXT] Re: [dpdk-dev] [PATCH v6 2/4] lib/kni: add PCI related 
> information
>  
> External Email
> 
> ----------------------------------------------------------------------
> On 6/25/2019 4:56 AM, vattun...@marvell.com wrote:
>> From: Vamsi Attunuru <vattun...@marvell.com>
>> 
>> PCI related information is needed in KNI kernel module,
>> since it requires iommu domain info for address
>> translations(using iommu_iova_to_phys() call) when
>> KNI runs in IOVA = VA mode.
>> 
>> Signed-off-by: Vamsi Attunuru <vattun...@marvell.com>
>> ---
>>  lib/librte_eal/linux/eal/include/rte_kni_common.h | 7 +++++++
>>  lib/librte_kni/rte_kni.c                          | 5 +++++
>>  2 files changed, 12 insertions(+)
>> 
>> diff --git a/lib/librte_eal/linux/eal/include/rte_kni_common.h 
>> b/lib/librte_eal/linux/eal/include/rte_kni_common.h
>> index 91a1c14..5db5a13 100644
>> --- a/lib/librte_eal/linux/eal/include/rte_kni_common.h
>> +++ b/lib/librte_eal/linux/eal/include/rte_kni_common.h
>> @@ -111,6 +111,13 @@ struct rte_kni_device_info {
>>        void * mbuf_va;
>>        phys_addr_t mbuf_phys;
>>  
>> +     /* PCI info */
>> +     uint16_t vendor_id;           /**< Vendor ID or PCI_ANY_ID. */
>> +     uint16_t device_id;           /**< Device ID or PCI_ANY_ID. */
>> +     uint8_t bus;                  /**< Device bus */
>> +     uint8_t devid;                /**< Device ID */
>> +     uint8_t function;             /**< Device function. */
>> +
>>        uint16_t group_id;            /**< Group ID */
>>        uint32_t core_id;             /**< core ID to bind for kernel thread 
>>*/
>>  
>> diff --git a/lib/librte_kni/rte_kni.c b/lib/librte_kni/rte_kni.c
>> index e29d0cc..99c4bf5 100644
>> --- a/lib/librte_kni/rte_kni.c
>> +++ b/lib/librte_kni/rte_kni.c
>> @@ -242,6 +242,11 @@ rte_kni_alloc(struct rte_mempool *pktmbuf_pool,
>>                kni->ops.port_id = UINT16_MAX;
>>  
>>        memset(&dev_info, 0, sizeof(dev_info));
>> +     dev_info.bus = conf->addr.bus;
>> +     dev_info.devid = conf->addr.devid;
>> +     dev_info.function = conf->addr.function;
>> +     dev_info.vendor_id = conf->id.vendor_id;
>> +     dev_info.device_id = conf->id.device_id;
> 
> Sample application part to set 'conf' values were also removed, need to add 
> them
> back too.
> 
> Ack, adding in next version of patches.
> 
> The device part was removed from KNI but I can see why this patch requires it 
> back.
> 
> Yes, rte_kni_device_info is the only struct carries all required info to 
> kernel
> module,
> all these device info can be protected under the config check though.

But still, lets make the device info requirement only for 'iova_mode', not for
all KNI

Reply via email to