On Wed, Sep 18, 2019 at 3:24 PM Kever Yang kever.yang@rock-chips.com wrote:
On 2019/9/18 下午1:25, Jagan Teki wrote:
On Wed, Sep 18, 2019 at 10:36 AM Kever Yang kever.yang@rock-chips.com wrote:
On 2019/9/17 下午2:10, Jagan Teki wrote:
SPL sets up SDRAM while in its board_init_f() function, it is possible for the stack to move there before board_init_r() is reached.
So it is required to reserve the stack for SDRAM, with a proper location and size otherwise any operations during SPL handoff would leads to failure.
On, this particular context tinker-rk3288 SPL is failing to launch U-Boot proper on SDRAM due to lack of stack.
If no SPL_STACK_R, it should still use CONFIG_SPL_STACK when system init, right?
SPL_STACK vs SPL_STTACK_R would be different in their usage former is used to setup the stack during early while calling board_init_f but the later used to setup the stack (with DRAM base) while calling board_init_r. if CONFIG_SPL_STACK_R is defined, then at this point the stack and global_data are relocated to below CONFIG_SPL_STACK_R_ADDR.
CONFIG_SPL_STACK_R_ADDR - 0x80000 (this would be as part of sdram map) CONFIG_SPL_STACK - 0xff718000 (this address where SDRAM not been used)
Since the issue, is due to launch U-Boot proper from mmc, the stack that SPL during that time was not set it up.
I think the root cause is: the master of emmc controller can no access SRAM, and the emmc init driver needs to use memory based on CONFIG_SPL_STACK, if this is in SRAM, then the emmc driver will init fail, and 'fifo-mode' can work around this issue. So the CONFIG_SPL_STACK_R is needed for emmc/sd controller driver.
Not sure, because tinker-rk3288 doesn't have eMMC.