[dpdk-dev] Cannot allocate mempool with --no-huge since DPDK 16.07

[dpdk-dev] Cannot allocate mempool with --no-huge since DPDK 16.07

[tom.barbette]

Hi all,

I get a EINVAL since DPDK 16.07 when trying to allocate a mempool with rte_pktmbuf_pool_create. This is when I use --no-huge --vdev=eth_ring0. But this happens before accessing the virtual device in any ways.

Which, btw shows the documentation is wrong as it indicates : 
EINVAL - cache size provided is too large, or priv_size is not aligned.
Which is not the problem here (tried cache_size of 0, 64,256 and priv_size is 0). EINVAL can obviously come from elswhere.

Any idea? Something changed in involved subsystems since 16.07?

Thanks,
Tom

Re: [dpdk-dev] Cannot allocate mempool with --no-huge since DPDK 16.07

[Olivier MATZ]

Hi Tom,

On Thu, Sep 21, 2017 at 05:21:08PM +0200, tom.barbette@ulg.ac.be wrote:
> Hi all,
> 
> I get a EINVAL since DPDK 16.07 when trying to allocate a mempool with rte_pktmbuf_pool_create. This is when I use --no-huge --vdev=eth_ring0. But this happens before accessing the virtual device in any ways.
> 
> Which, btw shows the documentation is wrong as it indicates : 
> EINVAL - cache size provided is too large, or priv_size is not aligned.
> Which is not the problem here (tried cache_size of 0, 64,256 and priv_size is 0). EINVAL can obviously come from elswhere.
> 
> Any idea? Something changed in involved subsystems since 16.07?

I first though your issue was related to this one:
http://dpdk.org/ml/archives/dev/2017-June/067678.html

But maybe it's just a memory issue. Can you try with less mbufs?
  testpmd --no-huge [...] -- --total-num-mbufs=4096 [...]

If that fixes the issue, I'll check if we can change the return
value to something clearer (ENOMEM?).

Thanks,
Olivier