Closed yzt000000 closed 3 years ago
I believe it only uses the lowermost 2 GiB of physical memory, but moving the base to 0x100000000 works fine for me (either with or without an apparent bug fix I just made). I configured pk with --with-mem-start=0x100000000
and ran spike with -m0x100000000:0x100000000
to test.
I believe it only uses the lowermost 2 GiB of physical memory, but moving the base to 0x100000000 works fine for me (either with or without an apparent bug fix I just made). I configured pk with
--with-mem-start=0x100000000
and ran spike with-m0x100000000:0x100000000
to test.
I had configure pk --with-mem-start=0x100000000 , i am run the bbl on fpga platform. only the ddr mapping had been changed from 0x80000000 to 0x100000000, then cpu will enter poweroff .
Given that it works on spike, it sounds like it's not a problem with pk.
I found when I modify the custom.dts 's memory define will lead to fail boot linux, so does the bbl can't handle this type address?
#size-cells = <1>;
reg = <0x80000000 0x10000000>;
};
#size-cells = <2>;
reg = <0x0 0x80000000 0x0 0x10000000>;
};
I found when I modify the custom.dts 's memory define will lead to fail boot linux, so does the bbl can't handle this type address?
- work version: memory @80000000{
address-cells = <1>;
size-cells = <1>;
reg = <0x80000000 0x10000000>; };
- bad version: memory @80000000{
address-cells = <2>;
size-cells = <2>;
reg = <0x0 0x80000000 0x0 0x10000000>; };
in dts file, address-cell use parent node’s define
I had a version work well when the DDR address is 0x80000000, when I change the address to be 0x100000000 which is larger than 2^32. BBL will always run into mtrap , then go into "poweroff".
does current BBL support larger 2^32 address?