Occurrence of DRAM ECC errors in arria10 SX660 - not consistent


#1

Hi,

I’m working on Arria10 SX660 based custom board with 2GB DDR with ECC enabled.
The board is up and and everything works fine.

When tested multiple times, we are observing the ECC errors SBE or DBE.

SBE is handled by correcting the bit, DBE is resetting the board.

Below is the DBE error which occurs randomly, may be one out of ten times boot in many boards.

4065272 bytes read in 186 ms (20.8 MiB/s)
50922 bytes read in 5 ms (9.7 MiB/s)
FPGA BRIDGES: enable
Kernel image @ 0x008000 [ 0x000000 - 0x3e07f8 ]
## Flattened Device Tree blob at 00808000
** Booting using the fdt blob at 0x808000**
** Loading Device Tree to 01ff0000, end 01fff6e9 … OK**
Error: DRAM ECC DBE occurred
erraddr = 01ff9040
### ERROR ### Please RESET the board ###
Info: DRAM ECC SBE @ 0x01ff9040
Info: DRAM ECC AUTO CORRECTION SBE ADDRESS @ 0x01ff9040
decoder0: erraddr = 10000000
decoder1: erraddr = 00000000
decoder2: erraddr = 01ff9050
decoder3: erraddr = 00000000
data abort
pc : [] lr : []
sp : 7fefeb90 ip : 01ff0000 fp : 00000000
r10: 01ff0000 r9 : ffe38ee8 r8 : 00000000
r7 : 01ffc74c r6 : 01ff0028 r5 : 01ffd000 r4 : 0000006f
r3 : 01ff9044 r2 : fe00ffd8 r1 : 01ff9054 r0 : 01ff0038
Flags: nzCv IRQs off FIQs off Mode SVC_32
Resetting CPU …

resetting …

What could make this DBE to happen randomly in our board.
Do we have any patch for this issue? Is it because of software or hardware? Any help or inputs is appreciated.

Thanks You.
Regards,
Ambika