From: Neil Horman <nhorman@tuxdriver.com>
To: "Iremonger, Bernard" <bernard.iremonger@intel.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>
Subject: Re: [dpdk-dev] [RFC PATCH 0/6] remove pci driver from vdevs
Date: Mon, 31 Aug 2015 08:41:52 -0400 [thread overview]
Message-ID: <20150831124152.GB32518@hmsreliant.think-freely.org> (raw)
In-Reply-To: <8CEF83825BEC744B83065625E567D7C219F49405@IRSMSX108.ger.corp.intel.com>
On Mon, Aug 31, 2015 at 11:21:46AM +0000, Iremonger, Bernard wrote:
> Hi Keith, Neil,
>
> <snip>
>
> > >> > On Thu, Aug 27, 2015 at 04:40:35PM +0100, Bernard Iremonger wrote:
> > >> > > There is a dummy pci driver in the vdev PMD's at present.
> > >> > > This RFC proposes to remove the pci driver from the vdev PMD's.
> > >> > > Changes have been made to librte_ether to handle vdevs which do
> > >> > > not
> > >> > have a pci driver.
> > >> > >
> > >> > > The pdev PMD's should work as before with the changes to
> > >>librte_ether
> > >> > > The vdev PMD's which still have a pci driver should work as
> > >> > > before
> > >>with the
> > >> > librte_ether changes.
> > >> > >
> > >> > > The following vdev PMD's have had the pci driver removed
> > >> > >
> > >> > > bonding PMD
> > >> > > null PMD
> > >> > > pcap PMD
> > >> > > ring PMD
> > >> >
> > >> > Any reason there is no patch for the af_packet driver?
> > >> >
> > >> > John
> > >>
> > >> I have just modified the Intel vdev PMD's.
> > >> It would be best if the owners of the non Intel vdev's submitted
> > >>patches for their drivers.
> > >>
> > >I disagree. Its ok given that this is an RFC patch I suppose, but if
> > >you intend to actually propose this change for review, you need to
> > >modify all affected drivers in a single commit. Asking individual
> > >driver maintainers to submit patches to not access a struct element
> > >that is removed in a separate patch will by definition cause FTBFS
> > >errors. All references to the structure member being removed must also
> > >be eliminated in the same or a prior commit, preferably the former.
> >
> > +1, if you introduce a chance that effects other places in the
> > code/drivers then you must also make the changes to those parts as well.
> > It really should not be an option IMO.
> > >
> > >Neil
>
> < snip >
>
> > Regards,
> > ++Keith
> > Intel Corporation
> >
> >
> Firstly no struct element has been removed from struct rte_eth_dev {} in the eth_dev patch.
> A dev_flags element has been added to struct rte_eth_dev{}
> A numa_node element has been added to struct rte_eth_dev_data{}.
> Unmodified VDEV's and PDEV's will not be aware of these new elements and will be unaffected by the eth_dev changes.
>
Then NAK to the entire series, as it doesn't address this:
http://dpdk.org/ml/archives/dev/2015-July/022107.html
If that was your intent. The desire that Thomas had was to refactor the bus
code so that a drivers bus type was independent of the devices it registered.
This really just makes filling out the pci driver struct optional (which likely
becomes very messy down the road). I think what's being sought is something
like this:
1) remove the pci driver field from the ethdev struct, perhaps replacing it with
a union pointer to all the bus type structures supported (though I don't think
thats required)
2) modify the driver init routine so that each driver module can register with
whatever bus type it supports.
3) add a per-bus-instance init routine (akin to the probe routine in the linux
and bsd driver models) called once for each instance of that device type found
on a given bus. This routine allocates and registers an ethernet device
(optionally associating the bus type with the device using the pointer suggested
in (1)
There are other ways to do this of course, this is just a suggestion. The point
being, the goal is to remove the explicit relationship of a ethdev to a pci
device, as that relationship is clearly not guaranteed. Doing so allows us to
remove the need to determine bus type at init time, which would be a big
benefit.
Regards
Neil
> Regards,
>
> Bernard.
>
>
>
>
>
next prev parent reply other threads:[~2015-08-31 12:42 UTC|newest]
Thread overview: 95+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <RFC PATCH>
2015-04-08 10:02 ` [dpdk-dev] [RFC PATCH] librte_pmd_ixgbe: changes to support PCI Port Hotplug Bernard Iremonger
2015-04-22 3:14 ` Qiu, Michael
2015-04-08 14:11 ` [dpdk-dev] [RFC PATCH] librte_pmd_e1000: igb and em1000 PCI Port Hotplug changes Bernard Iremonger
2015-04-30 13:16 ` [dpdk-dev] [RFC PATCH 1/1] librte_pmd_ring: changes to support PCI Port Hotplug Bernard Iremonger
2015-04-30 15:40 ` [dpdk-dev] [RFC PATCH 1/4] librte_pmd_i40e: " Bernard Iremonger
2015-04-30 15:41 ` [dpdk-dev] [RFC PATCH 2/4] librte_pmd_i40e: release vmdq vsi's in dev_close Bernard Iremonger
2015-04-30 15:42 ` [dpdk-dev] [RFC PATCH 3/4] librte_pmd_i40e: increase ASQ_DELAY_MS to 100 in i40evf_wait_cmd_done() Bernard Iremonger
2015-04-30 15:42 ` [dpdk-dev] [RFC PATCH 4/4] librte_pmd_i40e: call _clear_cmd() when error occurs Bernard Iremonger
2015-05-01 14:36 ` [dpdk-dev] [RFC PATCH] librte_pmd_bond: add support for PCI Port Hotplug Bernard Iremonger
2015-05-01 15:17 ` Neil Horman
2015-05-01 15:28 ` [dpdk-dev] [RFC PATCH] librte_pmd_virtio: " Bernard Iremonger
2015-05-05 14:36 ` [dpdk-dev] [RFC PATCH V2] librte_pmd_ring: changes to support " Bernard Iremonger
2015-05-06 16:15 ` Bruce Richardson
2015-05-06 10:22 ` [dpdk-dev] [RFC PATCH V2] librte_pmd_ixgbe: " Bernard Iremonger
2015-05-06 11:20 ` [dpdk-dev] [RFC PATCH V2] librte_pmd_e1000: igb and em1000 PCI Port Hotplug changes Bernard Iremonger
2015-05-27 16:00 ` [dpdk-dev] [RFC PATCH V2 1/2] drivers/net/virtio: add support for PCI Port Hotplug Bernard Iremonger
2015-05-27 16:00 ` [dpdk-dev] [RFC PATCH V2 2/2] drivers/net/virtio: check vq parameter Bernard Iremonger
2015-06-16 1:08 ` Ouyang, Changchun
2015-05-28 12:28 ` [dpdk-dev] [RFC PATCH V2] drivers/net/bonding: add support for PCI Port Hotplug Bernard Iremonger
2015-05-29 13:46 ` [dpdk-dev] [RFC PATCH V3] drivers/net/ring: changes to support " Bernard Iremonger
2015-06-03 15:21 ` Bruce Richardson
2015-06-02 10:18 ` [dpdk-dev] [RFC PATCH V3] ixgbe: " Bernard Iremonger
2015-06-03 14:03 ` [dpdk-dev] [RFC PATCH V3] e1000: igb and em1000 PCI Port Hotplug changes Bernard Iremonger
2015-06-04 16:24 ` [dpdk-dev] [RFC PATCH V2 1/4] i40e: changes to support PCI Port Hotplug Bernard Iremonger
2015-06-24 8:03 ` Qiu, Michael
2015-06-04 16:25 ` [dpdk-dev] [RFC PATCH V2 2/4] i40e: release vmdq vsi's in dev_close Bernard Iremonger
2015-06-04 16:25 ` [dpdk-dev] [RFC PATCH V2 3/4] i40e: increase ASQ_DELAY_MS to 100 in i40evf_wait_cmd_done() Bernard Iremonger
2015-06-04 16:26 ` [dpdk-dev] [RFC PATCH V2 4/4] i40e: call _clear_cmd() when error occurs Bernard Iremonger
2015-06-08 15:47 ` [dpdk-dev] [RFC PATCH V4] ixgbe: changes to support PCI Port Hotplug Bernard Iremonger
2015-06-15 22:31 ` Ananyev, Konstantin
2015-06-17 6:11 ` Zhang, Helin
2015-06-24 7:45 ` Qiu, Michael
2015-06-24 7:46 ` Qiu, Michael
2015-06-10 8:27 ` [dpdk-dev] [RFC PATCH V3 0/5] PCI Port Hotplug changes Bernard Iremonger
2015-06-10 8:27 ` [dpdk-dev] [RFC PATCH V3 1/5] i40e: changes to support PCI Port Hotplug Bernard Iremonger
2015-06-10 8:27 ` [dpdk-dev] [RFC PATCH V3 2/5] i40e: release vmdq vsi's in dev_close Bernard Iremonger
2015-06-10 8:27 ` [dpdk-dev] [RFC PATCH V3 3/5] i40e: increase ASQ_DELAY_MS to 100 in i40evf_wait_cmd_done() Bernard Iremonger
2015-06-10 8:27 ` [dpdk-dev] [RFC PATCH V3 4/5] i40e: call _clear_cmd() when error occurs Bernard Iremonger
2015-06-10 8:27 ` [dpdk-dev] [RFC PATCH V3 5/5] i40e: clear queues in i40evf_dev_stop Bernard Iremonger
2015-06-10 12:09 ` [dpdk-dev] [RFC PATCH V4] e1000: igb and em1000 PCI Port Hotplug changes Bernard Iremonger
2015-08-27 15:40 ` [dpdk-dev] [RFC PATCH 0/6] remove pci driver from vdevs Bernard Iremonger
2015-08-27 15:40 ` [dpdk-dev] [RFC PATCH 1/6] librte_ether: add fields from rte_pci_driver to rte_eth_dev and rte_eth_dev_data Bernard Iremonger
2015-08-31 14:07 ` Thomas Monjalon
2015-09-01 11:38 ` Iremonger, Bernard
2015-09-01 12:03 ` Bruce Richardson
2015-09-01 12:37 ` Ananyev, Konstantin
2015-09-01 12:43 ` Richardson, Bruce
2015-08-27 15:40 ` [dpdk-dev] [RFC PATCH 2/6] librte_ether: handle RTE_ETH_DEV_INTR_LSC for vdevs Bernard Iremonger
2015-08-27 15:40 ` [dpdk-dev] [RFC PATCH 3/6] null: remove pci device driver Bernard Iremonger
2015-08-31 14:11 ` Thomas Monjalon
2015-08-31 16:05 ` Iremonger, Bernard
2015-08-27 15:40 ` [dpdk-dev] [RFC PATCH 4/6] ring: " Bernard Iremonger
2015-08-27 15:40 ` [dpdk-dev] [RFC PATCH 5/6] bonding: " Bernard Iremonger
2015-08-27 17:48 ` Stephen Hemminger
2015-08-28 8:32 ` Iremonger, Bernard
2015-08-28 15:57 ` Stephen Hemminger
2015-08-27 15:40 ` [dpdk-dev] [RFC PATCH 6/6] pcap: " Bernard Iremonger
2015-08-27 17:43 ` [dpdk-dev] [RFC PATCH 0/6] remove pci driver from vdevs John W. Linville
2015-08-28 8:15 ` Iremonger, Bernard
2015-08-28 10:32 ` Neil Horman
2015-08-28 19:48 ` Wiles, Keith
2015-08-31 11:21 ` Iremonger, Bernard
2015-08-31 12:41 ` Neil Horman [this message]
2015-08-28 17:51 ` John W. Linville
2015-08-31 10:23 ` Iremonger, Bernard
2015-08-31 12:59 ` Neil Horman
2015-08-31 14:22 ` Thomas Monjalon
2015-09-01 13:38 ` Iremonger, Bernard
2015-09-01 19:18 ` Neil Horman
2015-09-03 14:02 ` Iremonger, Bernard
2015-09-07 9:22 ` Zende, Amruta S
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 00/18] refactor eal driver registration code Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 01/18] librte_eal: remove type field from rte_driver structure Bernard Iremonger
2015-09-04 13:08 ` Thomas Monjalon
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 02/18] af_packet: " Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 03/18] bnx2x: remove type field and initialise name field in " Bernard Iremonger
2015-09-04 11:26 ` Harish Patil
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 04/18] bonding: remove type field from " Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 05/18] cxgbe: " Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 06/18] e1000: remove type field and initialise name field in rte_driver structures Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 07/18] enic: remove type field and initialise name field in rte_driver structure Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 08/18] fm10k: " Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 09/18] i40e: remove type field and initialise name field in rte_driver structures Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 10/18] ixgbe: remove type field and initialise name field in rte_driver structure Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 11/18] mlx4: remove type field from " Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 12/18] mpipe: remove type field and update name in " Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 13/18] null: remove type field from " Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 14/18] pcap: " Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 15/18] ring: " Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 16/18] virtio_ethdev: remove type field and initialise name field in " Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 17/18] vmxnet3: " Bernard Iremonger
2015-09-04 11:01 ` [dpdk-dev] [RFC PATCH 18/18] xenvirt: remove type field from " Bernard Iremonger
2015-09-04 11:18 ` [dpdk-dev] [RFC PATCH 00/18] refactor eal driver registration code Bruce Richardson
2015-09-04 12:46 ` Iremonger, Bernard
2015-09-04 12:53 ` Bruce Richardson
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20150831124152.GB32518@hmsreliant.think-freely.org \
--to=nhorman@tuxdriver.com \
--cc=bernard.iremonger@intel.com \
--cc=dev@dpdk.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).