Marek Vasut marex@denx.de schrieb am Do., 13. Juni 2019, 22:40:
On 6/13/19 10:26 PM, Simon Goldschmidt wrote:
On 13.06.19 22:14, Marek Vasut wrote:
On 6/13/19 9:50 PM, Simon Goldschmidt wrote:
This provides an SPL_SIZE_LIMIT that makes the build check that the SPL binary loaded from flash fits into the SRAM (64 KiB) and leaves enough room for global data, heap and stack (512 bytes assumed stack usage).
Signed-off-by: Simon Goldschmidt simon.k.r.goldschmidt@gmail.com
arch/arm/mach-socfpga/Kconfig | 8 ++++++++ 1 file changed, 8 insertions(+)
diff --git a/arch/arm/mach-socfpga/Kconfig b/arch/arm/mach-socfpga/Kconfig index 48f02f08d4..1d914648e3 100644 --- a/arch/arm/mach-socfpga/Kconfig +++ b/arch/arm/mach-socfpga/Kconfig @@ -3,6 +3,12 @@ if ARCH_SOCFPGA config NR_DRAM_BANKS default 1 +config SPL_SIZE_LIMIT
- default 65536 if TARGET_SOCFPGA_GEN5
+config SPL_SIZE_LIMIT_PROVIDE_STACK
- default 0x200 if TARGET_SOCFPGA_GEN5
- config SPL_STACK_R_ADDR default 0x00800000 if TARGET_SOCFPGA_GEN5 @@ -49,6 +55,8 @@ config TARGET_SOCFPGA_GEN5 bool select SPL_ALTERA_SDRAM imply FPGA_SOCFPGA
- imply SPL_SIZE_LIMIT_SUBTRACT_GD
- imply SPL_SIZE_LIMIT_SUBTRACT_MALLOC imply SPL_STACK_R imply SPL_SYS_MALLOC_SIMPLE imply USE_TINY_PRINTF
512 bytes for stack looks like it's too little. I think the SPL started misbehaving when it overgrew 50 kiB, no ?
To 1: Well, I tested the stack usage once, booting from eMMC, and was somewhere below that range. But yes, it's a problem for the future: once we get a more stack-consuming function, we might be lost. Which size would you suggest?
Hmmm, now that I think about it, the stack gets relocated to DRAM quite early too, right ? So maybe we really don't need that much space for stack.
Exactly. The only stack-consuming things before relocation are dts parsing and maybe the ddr driver. I implemented a stack usage check by filling the memory with 0xaa and checking it afterwards and if I remember correctly it resulted in about 400 bytes. It's even more or less independent of the boot type since the ski/mmc drivers are probed only after DDR is up. Same goes for file systems.
Nevertheless, stack usage can increase in the future. That's why I'm not too happy about this constant. Otoh, DM_CLK makes me need every byte and right now I don't see how I can enable secure boot (fit signature check) due to this size limit...
Regards, Simon
To 2: No. The documentation says 60 KiB would be the limit (upper 4 KiB reserved for BootRom usage while loading SPL). I'm running SPL with the DM_CLK driver (still under development) with a binary (code + dtb) of about 58 KiB and it runs fine. Heap size before relocation increases to around 4KiB though, so that ensures the 60 KiB limit isn't breached.
OK, thanks for checking.
-- Best regards, Marek Vasut