[dpdk-dev] [PATCH 0/2] Allow application set mempool handle

[Jerin Jacob]

-----Original Message-----
> Date: Tue, 20 Jun 2017 16:07:17 +0530
> From: Hemant Agrawal <hemant.agrawal at nxp.com>
> To: Jerin Jacob <jerin.jacob at caviumnetworks.com>
> CC: Santosh Shukla <santosh.shukla at caviumnetworks.com>,
>  olivier.matz at 6wind.com, dev at dpdk.org
> Subject: Re: [PATCH 0/2] Allow application set mempool handle
> User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:45.0) Gecko/20100101
>  Thunderbird/45.8.0
> 
> On 6/19/2017 6:31 PM, Jerin Jacob wrote:
> > -----Original Message-----
> > > Date: Mon, 19 Jun 2017 17:22:46 +0530
> > > From: Hemant Agrawal <hemant.agrawal at nxp.com>
> > > To: Santosh Shukla <santosh.shukla at caviumnetworks.com>,
> > >  olivier.matz at 6wind.com, dev at dpdk.org
> > > CC: jerin.jacob at caviumnetworks.com
> > > Subject: Re: [PATCH 0/2] Allow application set mempool handle
> > > User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:45.0) Gecko/20100101
> > >  Thunderbird/45.8.0
> > > 
> > > On 6/1/2017 1:35 PM, Santosh Shukla wrote:
> > > > Some platform can have two different NICs for example external PCI Intel
> > > > 40G card and Integrated NIC like vNIC/octeontx/dpaa2.
> > > > 
> > > > Both NICs like to use their preferred pool e.g. external PCI card/ vNIC's
> > > > preferred pool would be the ring based pool and octeontx/dpaa2 preferred would
> > > > be ext-mempools.
> > > > Right now, Framework doesn't support such case. Only one pool can be
> > > > used across two different NIC's. For that, user has to statically set
> > > > CONFIG_RTE_MEMPOOL_DEFAULT_OPS=<pool-name>.
> > > > 
> > > > So proposing two approaches:
> > > > Patch 1) Introducing eal option --pkt-mempool=<pool-name>
> > > > Patch 2) Introducing ethdev API called _get_preferred_pool(), where PMD driver
> > > > gets a chance to advertise their pool capability to the application. And based
> > > > on that hint- application creates pools for that driver.
> 
> If the system is having more than one heterogeneous ethernet device with
> different mempool, the application has to create different mempool for each
> of the ethernet device.
> 
> However, let's take a case
> As system has a DPAA2 eth device, which only work with dpaa2 mempools.

dpaa2 ethdev will return dpaa2 mempool as preferred handler.

> System also detect a standard PCI NIC, which can work with any software
> mempool (e.g ring_mp_mc) or with dpaa2 mempool. Given the preference, PCI
> NIC will have preferred as software mempool.
> how the application will choose between these, if it want to create only one
> mempool?

We need add some policy in common code to help application to choose
in case if application interested in creating in one pool for
heterogeneous cases. It is more of application problem, ethdev can
return the preferred handler, let application choose interested in one.
ethdev is depended on the specific mempool not any other object.

We will provide option 1(eal argument based one) as one policy.More sophisticated
policies we need add in application.


> Or, how the scheme will work if the application want to create only one
> mempool?

option 1 (eal argument based) or we need to change the application to
choose from available ethdev count and its preferred mempool handler.

> 
> 
> > > > 
> > > The idea is good. it will help the vendors with hw mempool support.
> > > 
> > > On a similar line, I also submitted a patch to check the existence of a
> > > mempool instance.
> > > http://dpdk.org/dev/patchwork/patch/15877/
> > > 
> > > Option 1) requires manual knowledge of underlying NIC and different commands
> > > for different machines.
> > > 
> > > Option 2) this will help more as it allows the application to take decision
> > > autonomously.
> > > 
> > > In addition to it, we can also extend the overall MEMPOOL_OPS support.
> > >  3)  currently we support defining only one "RTE_MBUF_DEFAULT_MEMPOOL_OPS"
> > >     this can be supported to publish a priority list of MEMPOOL_OPS in
> > > config. if one is not available, application can try the next one in
> > > priority list as supported by the platform.
> > > 
> > > 4) we can also try something, where the existing application can also be
> > > supported.
> > >         - default mempool is configured as alias. This is with empty ops.
> > >         - based on the mempool detections on the bus, the bus configure the
> > > mempool ops internally with the actual ones.
> > 
> > What if both HW are on PCIe bus(That the case for us)? Any scheme to fix
> > that?
> 
> Nothing without a hackish approach. In our case, there is only a one mempool
> type on one type of platform specific bus.

Yes. But in  future a bus may have >1 mempool device(which already case for PCI now)
Could you send a patch for you patch, We can review it. May be we can
add it as one policy. I think, this may help to fix the problem avoiding the
application change on specific use case.

I think, it is difficult to integrate external mempool without
application change. We can some policy like option 1 or bus scheme but
it may not working in all the use cases without application change(option
2(new api).So I think, more polices and API(option 2) will fix the
problem. Let application choose what they want.

> 
> > 
> > 
> > > 
> > > 
> > > > Santosh Shukla (2):
> > > >   eal: Introducing option to set mempool handle
> > > >   ether/ethdev: Allow pmd to advertise preferred pool capability
> > > > 
> > > >  lib/librte_eal/bsdapp/eal/eal.c                 |  9 +++++++
> > > >  lib/librte_eal/bsdapp/eal/rte_eal_version.map   |  7 +++++
> > > >  lib/librte_eal/common/eal_common_options.c      |  3 +++
> > > >  lib/librte_eal/common/eal_internal_cfg.h        |  2 ++
> > > >  lib/librte_eal/common/eal_options.h             |  2 ++
> > > >  lib/librte_eal/common/include/rte_eal.h         |  9 +++++++
> > > >  lib/librte_eal/linuxapp/eal/eal.c               | 36 +++++++++++++++++++++++++
> > > >  lib/librte_eal/linuxapp/eal/rte_eal_version.map |  7 +++++
> > > >  lib/librte_ether/rte_ethdev.c                   | 16 +++++++++++
> > > >  lib/librte_ether/rte_ethdev.h                   | 21 +++++++++++++++
> > > >  lib/librte_ether/rte_ether_version.map          |  7 +++++
> > > >  lib/librte_mbuf/rte_mbuf.c                      |  8 ++++--
> > > >  12 files changed, 125 insertions(+), 2 deletions(-)
> > > > 
> > > 
> > > 
> > 
> 
>