Re: [PATCH 02/16] fpga: add FPGA device framework

From: Wu Hao
Date: Fri Mar 31 2017 - 08:24:26 EST


On Fri, Mar 31, 2017 at 11:03:28AM +0200, Greg KH wrote:
> On Fri, Mar 31, 2017 at 03:48:42PM +0800, Wu Hao wrote:
> > On Fri, Mar 31, 2017 at 08:09:09AM +0200, Greg KH wrote:
> > > On Thu, Mar 30, 2017 at 08:08:02PM +0800, Wu Hao wrote:
> > > > During FPGA device (e.g PCI-based) discovery, platform devices are
> > > > registered for different FPGA function units. But the device node path
> > > > isn't quite friendly to applications.
> > > >
> > > > Consider this case, applications want to access child device's sysfs file
> > > > for some information.
> > > >
> > > > 1) Access using bus-based path (e.g PCI)
> > > >
> > > > /sys/bus/pci/devices/xxxxx/fpga_func_a.0/sysfs_file
> > > >
> > > > From the path, it's clear which PCI device is the parent, but not perfect
> > > > solution for applications. PCI device BDF is not fixed, application may
> > > > need to search all PCI device to find the actual FPGA Device.
> > > >
> > > > 2) Or access using platform device path
> > > >
> > > > /sys/bus/platform/devices/fpga_func_a.0/sysfs_file
> > > >
> > > > Applications find the actual function by name easily, but no information
> > > > about which fpga device it belongs to. It's quite confusing if multiple
> > > > FPGA devices are in one system.
> > > >
> > > > 'FPGA Device' class is introduced to resolve this problem. Each node under
> > > > this class represents a fpga device, which may have one or more child
> > > > devices. Applications only need to search under this FPGA Device class
> > > > folder to find the child device node it needs.
> > > >
> > > > For example, for the platform has 2 fpga devices, each fpga device has
> > > > 3 child devices, the hierarchy looks like this.
> > > >
> > > > Two nodes are under /sys/class/fpga/:
> > > > /sys/class/fpga/fpga.0
> > > > /sys/class/fpga/fpga.1
> > > >
> > > > Each node has 1 function A device and 2 function B devices:
> > > > /sys/class/fpga/fpga.0/func_a.0
> > > > /sys/class/fpga/fpga.0/func_b.0
> > > > /sys/class/fpga/fpga.0/func_b.1
> > > >
> > > > /sys/class/fpga/fpga.1/func_a.1
> > > > /sys/class/fpga/fpga.1/func_b.2
> > > > /sys/class/fpga/fpga.1/func_b.3
> > > >
> > > > This following APIs are provided by FPGA device framework:
> > > > * fpga_dev_create
> > > > Create fpga device under the given parent device.
> > > > * fpga_dev_destroy
> > > > Destroy fpga device
> > > >
> > > > The following sysfs files are created:
> > > > * /sys/class/fpga/<fpga.x>/name
> > > > Name of the fpga device.
> > >
> > > How does this interact with the existing "fpga class" that is in the
> > > kernel already?
> >
> > The fpga-dev introduced by this patch, is only a container device, and
> > drivers could register different functions under it. Per my understanding,
> > the existing "fpga class", including fpga-region, fpga-bridge and
> > fpga-manager, is used to provide reconfiguration function for FPGA. So
> > driver can create child node using this existing "fpga class" to provide
> > FPGA reconfiguration function, and more nodes under this container for
> > different functions for given FPGA device.
> >
> > For Intel FPGA device, partial reconfiguration is only one function of
> > Intel FPGA Management Engine (FME). FME driver creates fpga_manager under
> > below path for partial reconfiguration, and other interfaces for more
> > functions, e.g power management, virtualization support and etc.
> >
> > /sys/class/fpga/<fpga.x>/<intel-fpga-fme.x>/fpga_manager
>
> So there is now two different levels of fpga class interfaces?
>
> I'm not disagreeing with this, just that it seems a bit confusing, don't
> you think?

I am not so sure, but the main purpose of fpga-dev, is trying to provide
enduser a more clear sysfs hierarchy reflecting the real hardware. And
fpga-things can be registered to fpga-dev directly if the hardware arch
is simple.

>From enduser point of view, he could find everything of this FPGA device
under /sys/class/fpga/<fpga.x>/, including all fpga-regions, fpga-bridges
and fpga-managers. I feel it is not a bad choice. :)

Thanks
Hao

>
> greg k-h