[dpdk-dev] [PATCH v3] net/i40e: fix allocating hash table on socket 0

Ferruh Yigit ferruh.yigit at intel.com
Mon Feb 20 17:16:32 CET 2017


On 2/20/2017 3:38 PM, Ivan Nardi wrote:
> Hi
> I reported the same issue 2 weeks ago
> http://dpdk.org/ml/archives/dev/2017-February/056930.html
> 
> You can reproduce it forcing testpmd to allocate memory only from node
> 1, for example:
> 
> /tmp/testpmd -d /home/micro/lib/librte_pmd_i40e.so -c 0xFFFC000 -w 0000:81:00.0  -w 0000:81:00.1 -n 4 --socket-mem=0,8192 -- -i --socket-num=1

Thanks Ivan,

I will update commit log with the information provided.

> 
> Ivan
> 
> 
> On 20 February 2017 at 15:45, Ferruh Yigit <ferruh.yigit at intel.com
> <mailto:ferruh.yigit at intel.com>> wrote:
> 
>     On 2/17/2017 8:37 AM, Beilei Xing wrote:
>     > Testpmd failed to start in another hugetlbfs mount point on
>     > i40e,
> 
>     Hi Beilei,
> 
>     Can you please describe the failure more. And how to reproduce it.
> 
>     I prefer initial patch title: "net/i40e: fix fail to start testpmd" but
>     with more details about when it fails to start.
> 
>     the root cause is that hash table is always allocated
>     > on socket 0. Fix the issue by assigning scocket id during
>     > hash parameter defination.
>     >
>     > Fixes: 5c53c82c8174 ("net/i40e: store flow director filter")
>     > Fixes: 425c3325f0b0 ("net/i40e: store tunnel filter")
>     > Fixes: 078259773da9 ("net/i40e: store ethertype filter")
>     >
>     > Signed-off-by: Beilei Xing <beilei.xing at intel.com <mailto:beilei.xing at intel.com>>
>     <...>
> 
> 



More information about the dev mailing list