On 04/21/2017 06:37 PM, Dalon Westergreen wrote:
[...]
My proposal is revert back to its original code, revert socfpga_bridges_reset() return type to void and don't check error from fpgamgr_test_fpga_ready(). So, we can support Gen5 boot without program FPGA.
Isn't the difference between CV/AV and A10 such that the former can boot without programmed FPGA while the later can NOT boot without programmed FPGA ? So this code is A10 specific ?
Yes, at a minimum in A10 the IO ring needs to be configured prior to boot.
We do the IOCSR configuration on C5/A5 too, so what's the difference ?
in c5/a5 the iocsr config is done directly, the iocsr data is compiled into spl, for a10 this is all done by the fpga bitstream, there is no direct iocsr config.
So that's why the FPGA must always be programmed with something to get the A10 working ? I see ... thanks for clarifying.
Since these functions (socfpga_bridges_reset) are different for c5/a5 and a10, i would suggest changing the code to enable the bridges IF the fpga is configured. perhaps just:
If it works ... fine.
... } else { writel(0, &sysmgr_regs-
iswgrp_handoff[0]);
writel(l3mask, &sysmgr_regs-iswgrp_handoff[1]);
/* Check signal from FPGA.*/ if (!fpgamgr_test_fpga_ready()) { /* brdmodrst */ writel(0, &reset_manager_base-
brg_mod_reset);
}...
I think this actually mirrors the behavior of the 2013.01.01 fork where if the fpga is configured, spl enables the bridges.
Super :)
--dalon