Re: [GIT PULL] FPGA Manager additional changes for 5.10

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



> > > 	0006-fpga-dfl-add-support-for-N3000-Nios-private-feature.patch
> > > 		- module parameters are for drivers written in the
> > > 		  1990's.  Please just "do the right thing" and make the
> > > 		  code work properly without having to have custom
> > > 		  options.  Note this option does not really work if you
> > > 		  have multiple devices in the system at once, which is
> > > 		  one reason why we don't use module parameters anymore.
> > 
> > I'm wondering which is the better way to handle the case, but let me
> > first describe it.
> > 
> > The n3000-nios driver will first require the embedded firmware to
> > finish some netdev configuration (via the SPI controller) according to
> > user's option, then the driver takes over the control of the SPI
> > controller and continue to enumerate various devices behind the SPI bus.
> > The firmware will never be active again after the first configuration on
> > power cycle. So I tend to let the driver finish the configuration
> > requirement at probe, then the following drivers for SPI devices could be
> > autoloaded. This is why I choose the module parameter.
> 
> Remember, module parameters affect ALL devices the driver controls, you
> just broke the ability for this driver to handle multiple devices per a
> single driver, with no way of ever fixing that (module parameters are
> userspace apis...)
> 
> > I was considering the sysfs interface option to trigger this one-time
> > configuration. But this makes the SPI devices unavailable until user
> > inputs something to this sysfs node. We may need to add some specific
> > udev rules to make the whole functionalities of the FPGA card enabled
> > automatically.
> 
> Why can't you figure out the state of the device without having to worry
> about a module option?  No one will know how to set that thing, and if
> they do, it will get baked into a bootloader somewhere and no one can
> ever turn it off.

It's the discrete hot-plugable PCIe card which is not managed in host
bootloader. And because of the HW design, the onboard firmware doesn't
get a writable non-volatile memory for the options, so it needs the input
from user.

Anyway, I need to reconsider this solution. But I got learned that "don't
make the user have to configure something out of their scope".

Thanks very much,
Yilun

> 
> Be dynamic and make things "just work" so that the user does not have to
> configure anything at all please.
> 
> > And our customers don't expect different configurations between
> > multiple cards on one system, which also influences my decision.
> 
> They might not say that today, but you just forced them to never be able
> to do that ever, in the next 20+ years :(
> 
> thanks,
> 
> greg k-h



[Index of Archives]     [LM Sensors]     [Linux Sound]     [ALSA Users]     [ALSA Devel]     [Linux Audio Users]     [Linux Media]     [Kernel]     [Gimp]     [Yosemite News]     [Linux Media]

  Powered by Linux