RE: [RFC net-next 8/8] net/mlx5: Add subdev driver to bind to subdev devices

[Parav Pandit]

> -----Original Message-----
> From: Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx>
> Sent: Tuesday, March 5, 2019 1:27 PM
> To: Parav Pandit <parav@xxxxxxxxxxxx>
> Cc: netdev@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> michal.lkml@xxxxxxxxxxx; davem@xxxxxxxxxxxxx; Jiri Pirko
> <jiri@xxxxxxxxxxxx>; Jakub Kicinski <jakub.kicinski@xxxxxxxxxxxxx>
> Subject: Re: [RFC net-next 8/8] net/mlx5: Add subdev driver to bind to
> subdev devices
> 
> On Tue, Mar 05, 2019 at 05:57:58PM +0000, Parav Pandit wrote:
> >
> >
> > > -----Original Message-----
> > > From: Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx>
> > > Sent: Tuesday, March 5, 2019 1:14 AM
> > > To: Parav Pandit <parav@xxxxxxxxxxxx>
> > > Cc: netdev@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> > > michal.lkml@xxxxxxxxxxx; davem@xxxxxxxxxxxxx; Jiri Pirko
> > > <jiri@xxxxxxxxxxxx>; Jakub Kicinski <jakub.kicinski@xxxxxxxxxxxxx>
> > > Subject: Re: [RFC net-next 8/8] net/mlx5: Add subdev driver to bind
> > > to subdev devices
> > >
> > > On Fri, Mar 01, 2019 at 05:21:13PM +0000, Parav Pandit wrote:
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx>
> > > > > Sent: Friday, March 1, 2019 1:22 AM
> > > > > To: Parav Pandit <parav@xxxxxxxxxxxx>
> > > > > Cc: netdev@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> > > > > michal.lkml@xxxxxxxxxxx; davem@xxxxxxxxxxxxx; Jiri Pirko
> > > > > <jiri@xxxxxxxxxxxx>
> > > > > Subject: Re: [RFC net-next 8/8] net/mlx5: Add subdev driver to
> > > > > bind to subdev devices
> > > > >
> > > > > On Thu, Feb 28, 2019 at 11:37:52PM -0600, Parav Pandit wrote:
> > > > > > Add a subdev driver to probe the subdev devices and create
> > > > > > fake netdevice for it.
> > > > >
> > > > > So I'm guessing here is the "meat" of the whole goal here?
> > > > >
> > > > > You just want multiple netdevices per PCI device?  Why can't you
> > > > > do that today in your PCI driver?
> > > > >
> > > > Yes, but it just not multiple netdevices.
> > > > Let me please elaborate in detail.
> > > >
> > > > There is a swichdev mode of a PCI function for netdevices.
> > > > In this mode a given netdev has additional control netdev (called
> > > representor netdevice = rep-ndev).
> > > > This rep-ndev is attached to OVS for adding rules, offloads etc
> > > > using
> > > standard tc, netfilter infra.
> > > > Currently this rep-ndev controls switch side of the settings, but
> > > > not the
> > > host side of netdev.
> > > > So there is discussion to create another netdev or devlink port..
> > > >
> > > > Additionally this subdev has optional rdma device too.
> > > >
> > > > And when we are in switchdev mode, this rdma dev has similar rdma
> > > > rep
> > > device for control.
> > > >
> > > > In some cases we actually don't create netdev when it is in
> > > > InfiniBand
> > > mode.
> > > > Here there is PCI device->rdma_device.
> > > >
> > > > In other case, a given sub device for rdma is dual port device,
> > > > having
> > > netdevice for each that can use existing netdev->dev_port.
> > > >
> > > > Creating 4 devices of two different classes using one iproute2/ip
> > > > or
> > > iproute2/rdma command is horrible thing to do.
> > >
> > > Why is that?
> > >
> > When user creates the device, user tool needs to return a device handle
> that got created.
> > Creating multiple devices doesn't make sense. I haven't seen any tool
> doing such crazy thing.
> 
> And what do you mean by "device handle"?  All you get here is a sysfs device
> tree.
> 
Subdev devices are created using devlink tool that works on device handle.
Device handle is defined using bus/device of a 'struct device'.
It is described in [1].
$ devlink dev add DEV creates new devlink device instance and its holding 'struct device'.
This command returns device handle = new devlink instance bus/name.
Patch 6 in the series returns device handle.
Patch 6 is at [2] with example in it where sysfs name and devlink matches with each other.

> > > > In case if this sub device has to be a passthrough device, ip link
> > > > command
> > > will fail badly that day, because we are creating some sub device
> > > which is not even a netdevice.
> > >
> > > But it is a network device, right?
> > >
> > When there is passthrough subdevice, there won't be netdevice created.
> > We don't want to create passthrough subdevice using iproute2/ip tool
> which primarily works on netdevices.
> 
> I don't know enough networking to claim anything here, so I'll ignore this :)
> 
> > > > So iproute2/devlink which works on bus+device, mainly PCI today,
> > > > seems
> > > right abstraction point to create sub devices.
> > > > This also extends to map ports of the device, health, registers
> > > > debug, etc
> > > rich infrastructure that is already built.
> > > >
> > > > Additionally, we don't want mlx driver and other drivers to go
> > > > through its
> > > child devices (split logic in netdev and rdma) for power management.
> > >
> > > And how is power management going to work with your new devices?
> > > All you have here is a tiny shim around a driver bus,
> > So subdevices power management is done before their parent's.
> > Vendor driver doesn't need to iterate its child devices to suspend/resume
> it.
> 
> True, so we can just autosuspend these "children" device and the "vendor
> driver" is not going to care?  You are going to care as you are talking to the
> same PCI device.  
Oh, vendor driver certainly care.
subdev vendor driver implements driver->pm callbacks to work on just a specific subdev.
Patch-2 in series at [3] implement shim layer by connecting core pm layer to driver pm callbacks.

> This goes to the other question about "how are you
> sharing PCI device resources?"
> 
Currently its equal distribution among all subdevices.
But when actual user arise to ask for specific resource reservation etc, we add those parameters using existing devlink infra [4].

> > > I do not see any new
> > > functionality, and as others have said, no way to actually share, or
> > > split up, the PCI resources.
> > >
> > devlink tool create command will be able to accept more parameters
> during device creation time to share and split PCI resources.
> > This is just the start of the development and RFC is to agree on direction.
> > devlink tool has parameters options that can be queried/set and existing
> infra will be used for granular device config.
> 
> Pointers to this beast?
> 
[1] and [4].

> > > > Kernel core code does that well today, that we like to leverage
> > > > through
> > > subdev bus or mfd pm callbacks.
> > > >
> > > > So it is lot more than just creating netdevices.
> > >
> > > But that's all you are showing here :)
> > >
> > Starting use case is netdev and rdma, but we don't want to create new
> > tools few months/a year later for passthrough mode or for different
> > link layers etc.
> 
> And I don't want to see duplicated driver model code happen either, which
> is why I point out the MFD layer :)
> 
Yes. Sure.

> > > > > What problem are you trying to solve that others also are having
> > > > > that requires all of this?
> > > > >
> > > > > Adding a new bus type and subsystem is fine, but usually we want
> > > > > more than just one user of it, as this does not really show how
> > > > > it is exercised very well.
> > > > This subdev and devlink infrastructure solves this problem of
> > > > creating
> > > smaller sub devices out of one PCI device.
> > > > Someone has to start.. :-)
> > >
> > > That is what a mfd should allow you to do.
> > >
> > I did cursory look at mfd.
> > It lacks removing specific devices, but that is small. It can be
> > enhanced to remove specific mfd device.
> 
> That should be easy enough, work with the MFD developers.  I think
> something like that should work today as you can use USB devices with MFD,
> right?
> 
> > >
> > > No, do not abuse a platform device.
> > Yes. that is my point mfd devices are platform devices.
> > mfd creates platform devices. and to match to it, platfrom_register_driver()
> have to be called to bind to it.
> > I do not know currently if we have the flexibility to say that instead of
> binding X driver, bind Y driver for platform devices.
> 
> try it :)
> 
> > > You should be able to just use a normal PCI device for this just
> > > fine, and if not, we should be able to make the needed changes to
> > > mfd for that.
> > >
> > Ok. so parent pci device and mfd devices.
> > mfd seems to fit this use case.
> > Do you think 'Platform devices' section is stale in [1] for autonomy, host
> bridge, soc platform etc points?
> 
> Nope, they are still horrible things and I hate them :)
> 
> Maybe we should just make MFD create "virtual" devices (bare ones, no
> need for platform stuff), and that would solve the issue of the platform
> device bloat being drug around everywhere.
> 
If you mean virtual MFD devices in /sys/devices/virtual/, than, it becomes difficult to do their life cycle using devlink because, devlink handle = bus+device.
devlink will fail to work. Inventing new tool and make it work with devlink wouldn't work.

virtual device has bus=NULL.
mfd device currently has bus_type=platform.

We still need to link subdevice to parent pci for power_mgmt to work, right?
And also to see right device hierarchy.
Don't you think subdev bus is actually able to link all the pieces together?
devlink, sysfs, core kernel, vendor drivers..

> > Should we update the documentation to indicate that it can be used for
> > non-autonomous, user created devices and it can be used for creating
> > devices on top of PCI parent device etc?
> 
> Nope, leave it alone please.
> 
> thanks,
> 
> greg k-h

[1] http://man7.org/linux/man-pages/man8/devlink-dev.8.html
[2] https://lore.kernel.org/patchwork/patch/1046995/
[3] https://lore.kernel.org/patchwork/patch/1046996/
[4] https://lore.kernel.org/patchwork/patch/959280/