Re: [PATCH v4 1/3] lib/librte_ether: add support for port reset

From: "Zhao1, Wei" <wei.zhao1@intel.com>
To: Thomas Monjalon <thomas.monjalon@6wind.com>
Cc: "Ananyev, Konstantin" <konstantin.ananyev@intel.com>,
	"Mcnamara, John" <john.mcnamara@intel.com>,
	"dev@dpdk.org" <dev@dpdk.org>,
	"Lu, Wenzhuo" <wenzhuo.lu@intel.com>
Subject: Re: [PATCH v4 1/3] lib/librte_ether: add support for port reset
Date: Fri, 14 Apr 2017 08:03:04 +0000	[thread overview]
Message-ID: <A2573D2ACFCADC41BB3BE09C6DE313CA0207ED75@PGSMSX103.gar.corp.intel.com> (raw)
In-Reply-To: <3836931.6Eq3YNmgkB@xps13>

Hi, Thomas

> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Friday, April 14, 2017 2:31 PM
> To: Zhao1, Wei <wei.zhao1@intel.com>
> Cc: Ananyev, Konstantin <konstantin.ananyev@intel.com>; Mcnamara, John
> <john.mcnamara@intel.com>; dev@dpdk.org; Lu, Wenzhuo
> <wenzhuo.lu@intel.com>
> Subject: Re: [dpdk-dev] [PATCH v4 1/3] lib/librte_ether: add support for port
> reset
> 
> 2017-04-14 01:29, Zhao1, Wei:
> > From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> > > 2017-04-13 08:55, Zhao1, Wei:
> > > > From: Ananyev, Konstantin
> > > > > From: Zhao1, Wei
> > > > > > From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> > > > > > > 2017-04-06 02:57, Zhao1, Wei:
> > > > > > > > >   /**
> > > > > > > > > > + * Reset an ethernet device when it's not working.
> > > > > > > > > > + One scenario is, after PF
> > > > > > > > > > + * port is down and up, the related VF port should be reset.
> > > > > > > > > > + * The API will stop the port, clear the rx/tx
> > > > > > > > > > + queues, re-setup the rx/tx
> > > > > > > > > > + * queues, restart the port.
> > > > > > > > >
> > > > > > > > > s/The API/This function/
> > > > > > > > >
> > > > > > > > > Please explain exactly the responsibility of this
> > > > > > > > > function, and how it is different from calling
> stop/configure/start.
> > > > > > > >
> > > > > > > > In this reset feature, reset function can do the calling
> > > > > > > > stop/configure/start process, but also It can also do some
> > > > > > > > restore work for the port, for example, it can restore the
> > > > > > > > added parameters  of
> > > > > > > vlan,  mac_addrs, promisc_unicast_enabled falg and
> > > > > > > promisc_multicast_enabled flag.
> > > > >
> > > > > Ok, but why start/stop can't do these things?
> > > > > Konstantin
> > > >
> > > > This is because in i40e PMD code, start and stop process do not
> > > > have the process of store and restore the added key parameters.
> > > > Not  only i40e but also other PMD code. So, in the function
> > > > pointed to by dev_reset,
> > > we add specific function do store and restore of some of the
> > > important parameters  listed above.
> > >
> > > Why store and restore cannot be implemented in start/stop functions?
> >
> > Because reset and  start/stop are used for two purposes,  for example:
> > Some user maybe just start/stop the port  and he do not care what key
> > parameters has been configuration last time, and even worse when he
> > want to clear all the configuration last time , if we add specific
> > function do store and restore in  that two function, it is useless for them,
> and may cause a result that user do not expect.
> 
> Is it said somewhere in the doc that the configuration is lost when stopping a
> device?
> Can we say which configuration parameter is kept and which one is lost?
> 
> rte_eth_dev_config_restore() is called in rte_eth_dev_start().

Port reset process not only involve the rte_eth_dev_start() and rte_eth_dev_stop().
It also involve eth_dev_uninit() and eth_dev_init() process，
in which PMD device uninit and init. In this case, for example, data->mac_addrs buffer is freed so it need to add store and restore function.
BUT, if you only call stop/configure/start process, that is not strictly what named "device reset".