
On Wed, Mar 08, 2017 at 11:36:06PM -0500, Trevor Woerner wrote:
On Wed 2017-03-08 @ 06:17:13 PM, Tom Rini wrote:
On Wed, Mar 08, 2017 at 05:32:45PM -0500, Trevor Woerner wrote:
On Wed 2017-03-08 @ 04:33:21 PM, Tom Rini wrote:
Now, you've boot tested this, right? :)
Yes.
Also, what board(s) show this problem exactly?
I'm working with the firefly-rk3288 board.
OK. Can you please turn off CONFIG_ROCKCHIP_SPL_BACK_TO_BROM and see if we can link again without -N? No, this won't boot, but that's not what I'm after just yet, thanks!
Building with binutils-2.28 and CONFIG_ROCKCHIP_SPL_BACK_TO_BROM disabled succeeds.
In case this is where you were heading... ;-)
Building twice (once with binutils-2.27 and once with binutils-2.28) both times without the -N and both times with CONFIG_ROCKCHIP_SPL_BACK_TO_BROM disabled produces a diff in the map files of:
--- 2.27-without-N-without-SPL/u-boot.map 2017-03-08 23:27:38.555324770 -0500 +++ 2.28-without-N-without-SPL/u-boot.map 2017-03-08 23:24:59.383082189 -0500 @@ -7080,6 +7080,7 @@
.data 0x0000000000143688 0x2970 *(.data*)
- .data.rel.ro 0x0000000000143688 0x0 arch/arm/cpu/armv7/start.o .data.rk3288_otg_data 0x0000000000143688 0x2c arch/arm/mach-rockchip/built-in.o .data.rk3288_dmc_ops
@@ -7944,6 +7945,9 @@ *(.rel*) .rel.got 0x0000000000147cd0 0x0 arch/arm/cpu/armv7/start.o .rel.plt 0x0000000000147cd0 0x0 arch/arm/cpu/armv7/start.o
- .rel.bss 0x0000000000147cd0 0x0 arch/arm/cpu/armv7/start.o
- .rel.data.rel.ro
.rel.iplt 0x0000000000147cd0 0x0 arch/arm/cpu/armv7/start.o .rel.text 0x0000000000147cd0 0x20 arch/arm/cpu/armv7/start.o .rel.vectors 0x0000000000147cf0 0x38 arch/arm/cpu/armv7/start.o0x0000000000147cd0 0x0 arch/arm/cpu/armv7/start.o
Adding Khem to this thread as well. The change I asked Trevor to make changed us from linking at -Ttext 0x0 to -Ttext 0x00100000 which is why I'm wondernig if there's not some corner case being tickled now in ld.