On Tue, 27 Jun 2017, Marek Vasut wrote:
On 06/27/2017 05:15 PM, matthew.gerl...@linux.intel.com wrote:
On Tue, 27 Jun 2017, Marek Vasut wrote:
On 06/26/2017 06:13 PM, matthew.gerl...@linux.intel.com wrote:
From: Matthew Gerlach <matthew.gerl...@linux.intel.com>
Just wrap it into the Altera QSPI driver , no need for separate platform
driver IMO.
Hi Marek,
I answered this question when you asked why the header file was
necessary, but I think further discussion could be helpful, since this
problem is becoming more prevelent. The Altera Quadspi component is a
soft IP in a
FPGA, and the processor using the component may or may not have device
tree support compiled into the Linux kernel. Since device tree support
may or may not be available, the device tree specific code must be
separated
from the core driver code.
I see, that's fine, although there is no PCIe or other support in this
submission. Is that planned ?
You probably would not see a PCIe driver for a card with a FPGA that would
only have the Altera Quadspi component on it. Usually a FPGA has several
components, each requiring their own driver which are considered
sub-drivers of the PCIe driver. I will be adding Altera Quadspi support
to the Intel-FPGA PCIe driver that is currently under review. We have
also seen people use the Altera Quadspi with a NIOS-II soft processor, and
I expect someone would want to use the component with ARM SOCFPGAs.
Matthew Gerlach
One can certainly make the case, that device tree support could/should
be available everywhere, but the current reality is most x86 Linux
kernel configurations do not include device tree support.
For the record, I believe device trees, and more specifically device
tree overlays, are the best way for Linux to use FPGAs, but I have to
deal with the current realities.
Thanks again for all the great feedback.
Matthew Gerlach
[...]
--
Best regards,
Marek Vasut
