Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Altera_Forum
Honored Contributor I
1,490 Views

FPGA-to-HPS SDRAM Interface Problem

Hello! 

 

I have system where I want to transfer data to HPS memory using custom IP in FPGA during the boot process (before U-Boot stage). Does the Preloader (U-Boot SPL) initialize the FPGA-to-HPS SDRAM Interface? In Signaltap the f2h_sdram0/avm_waitrequest signal is always high. Do I have to modify the SPL-code to enable the FPGA-to-HPS SDRAM Interface? 

 

regards, 

Joonas
0 Kudos
1 Reply
Altera_Forum
Honored Contributor I
178 Views

If i understand correctly: If you are using SoCEDS 13.1 or later the bridges are indeed initialized during boot. If your u-boot.scr contains the line "run bridge_enable_handoff;" the bridges should be initialised. (all bridges except for the fpga2sdram bridge, read on) 

 

from: https://github.com/mathworks/altera-uboot/blob/master/doc/readme.socfpga 

 

--- Quote Start ---  

 

Usage of registers SYSMGR.ISWGRP.HANDOFF 

---------------------------------------- 

Below are "Preloader to bootloader" handoff register and handoff value. 

These handoff value are provided by the tools (to Preloader) based on user Qsys 

design. As of now, these handoff registers mainly contain the info for bridges 

between the HPS and FPGA as below 

 

 

SYSMGR.ISWGRP.HANDOFF0 - hps axi bridges (RSTMGR.BRGMODRST) 

SYSMGR.ISWGRP.HANDOFF1 - hps l3 remap (L3REGS.REMAP) 

SYSMGR.ISWGRP.HANDOFF2 - hps peripheral controller to fpga bridge 

(SYSMGR.FPGAINTF.MODULE) 

SYSMGR.ISWGRP.HANDOFF3 - fpga to hps sdram bridge (SDR.CTRLGRP.FPGAPORTRST) 

SYSMGR.ISWGRP.HANDOFF4 - row bits handed off by Altera tools 

SYSMGR.ISWGRP.HANDOFF5 - unused 

SYSMGR.ISWGRP.HANDOFF6 - unused 

SYSMGR.ISWGRP.HANDOFF7 - unused 

 

 

If user reprogram the FPGA at U-Boot, all the HPS and FPGA bridges will be 

disabled. This is to prevent the noise generated during FPGA programming from 

going into HPS domain. But U-Boot will not enable back the bridge later as 

U-Boot won't have the knowledge about the bridges used by SOF file. User need 

to release the HPS and FPGA bridges based on their design. 

 

But if the SOF programmed by U-Boot matched with the design to generate 

Preloader handoff files, user can program the bridges at U-Boot based on these 

handoff value. It provide a faster mean for user to re-enable the bridges. 

These additional steps can be saved as series of action for FPGA programming 

(which as part of U-Boot environment too). 

 

--- 

 

As example, below are steps if U-Boot programming the FPGA with the same design 

that used to generate Preloader handoff. 

 

> fatload mmc 0:1 $fpgadata my_design.rbf 

> fpga load 0 $fpgadata $filesize 

> run bridge_enable_handoff 

 

 

where bridge_enable_handoff will execute below steps 

> mw $fpgaintf $fpgaintf_handoff 

> go $fpga2sdram_apply 

> mw $fpga2sdram $fpga2sdram_handoff 

> mw $axibridge $axibridge_handoff 

> mw $l3remap $l3remap_handoff 

> md $fpgaintf 

> md $fpga2sdram 

> md $axibridge 

 

* For your info, below U-Boot environment are referring to particular address 

- "fpgaintf" - SYSMGR.FPGAINTF.MODULE 

- "fpga2sdram" - SDR.CTRLGRP.FPGAPORTRST 

- "axibridge" - RSTMGR.BRGMODRST 

- "l3remap" - L3REGS.REMAP 

--- Quote End ---  

 

 

also usefull: https://www.altera.com/support/support-resources/knowledge-base/embedded/2016/how-and-when-can-i-ena... 

 

However, I just realise I also have a design where the waitrequest signal is always high. If somebody has a fix for this, that would be helpful 

 

EDIT: you, my friend are a lifesaver. thanks to your waitrequestremark i was able to debug something that has been troubling me for weeks! 

you will most likely find this topic very handy 

http://www.alteraforum.com/forum/showthread.php?t=51200 

 

a small remark on what is mentioned in this topic; I ran : 

setenv fpga2sdram_handoff 0x311; saveenv run bridge_enable_handoff; 

After this i programmed the FPGA and booted linux. 

 

after this i was able to use the F2S port to write to SDRAM
Reply