[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CY1PR02MB1692876485F0557FE798C2D9DC540@CY1PR02MB1692.namprd02.prod.outlook.com>
Date: Wed, 25 Jul 2018 09:12:21 +0000
From: Appana Durga Kedareswara Rao <appanad@...inx.com>
To: Alan Tull <atull@...nel.org>
CC: Moritz Fischer <moritz.fischer@...us.com>,
Nava kishore Manne <navam@...inx.com>,
Michal Simek <michals@...inx.com>,
"linux-fpga@...r.kernel.org" <linux-fpga@...r.kernel.org>,
linux-arm-kernel <linux-arm-kernel@...ts.infradead.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: RE: [PATCH v3 2/2] fpga: zynq-fpga: Add support for readback
Hi Alan,
Thanks for the review...
<Snip>
> >
> > <Snip>
> >> > +static bool readback_type;
> >> > +module_param(readback_type, bool, 0644);
> >> > +MODULE_PARM_DESC(readback_type,
> >> > + "readback_type 0-configuration register read "
> >> > + "1- configuration data read (default: 0)");
> >>
> >> Not sure a module param is the best solution here.
> >
> > Need to provide flexibility to the user to switch between reading of FPGA
> registers and configuration data.
>
> I suggested calling the attribute 'image' because I thought it would always be
> a read back of the FPGA image information. I don't think that a sysfs
> attribute's function should change. :)
In Zynq Case it supports two types of the readback (Configuration registers, Configuration data(fpga image)) which may not be the same case for other vendors.
Since I need to support both the use cases I have differentiated them using module param in the zynq-fpga driver.
As you said we shouldn't change the attribute's function, So in the zynq-fpga driver,
I am planning to add another debugfs attribute for reading back of configuration registers as it is vendor specific readback type.
(Configuration registers ---- Usage: /sys/kernel/debug/fpga/zynq-fpga/cfgreg_summary)
(Configuration data(fpga image) ---- Usage: /sys/kernel/debug/fpga/fpga0/image)
Please let me know your thoughts for the same...
>
> > One other option is sysfs. Do you want me to implement sysfs path???
> > Any other suggestions???
>
> You could implement another sysfs attribute for reading FPGA registers
> with a separate ops callback. Please clarify what it is doing (not
> all FPGAs have this function) and what that will look like when the user reads
> the from the sysfs
AFAIK we shouldn't add another sysfs/debugfs attribute for vendor-specific readback type in the FPGA manager ops.
Please correct me if my understanding is wrong.
Regards,
Kedar.
>
> Alan
Powered by blists - more mailing lists