Closed kaolpr closed 3 years ago
Concerning this issue our prefered solution is having PS and PL pins connected to the same flash. I.e. shorting the PS & PL pins (post the TXB0108 buffers). This prove to work on our boards and allows for flashing over PCIe.
@arthurspi Is this solution fine for you? Should Enable be driven from MMC, to switch between PS and PL access?
@filipswit The idea is to have access to the flash that holds the board's firmware from the PL. We don't actually need a second memory. If we can apply the same circuit to the existing flash that would be perfect.
If there is a switching it should also be accessible from the PL.
@arthurspi If you want to use flash that already exist (IC37 IC40), don't you have access to it through FPGA_SPI_FLASH_UPDATE?
@filipswit The reason is there is a shared infrastructure at cern to flash over PCI, and it has a direct SPI interface as for the wr-core. by FPGA_SPI_FLASH_UPDATE do you mean the axi interface?
There is an SPI connected to MMC and PL.
To my understanding there is one SPI connection between the MMC and the PL but not between PL and the QSPI flash
@arthurspi sorry, you are right, while making #84, just realized that there are two separate SPIs. I added another from PL to Flash.
CERN insists on having direct access to some non-volatile memory from PL to store WRC calibration data.
Please add a separate flash connected to PL exclusively and make it an assembly variant.