Why not set stack-pointer to end of stack?

[cobratbq]

I noticed that the stack-pointer is initialized to just below the end of the stack. Is there a specific reason for this? My understanding is that the stack-pointer is decremented upon each use of the stack, so setting it to TK1_RAM_BASE+TK1_RAM_SIZE (or ORIGIN(RAM) + LENGTH(RAM) in terms of the linker-script), should be optimal.

libcrt0/crt0.S:

/* init stack below 0x40020000 (TK1_RAM_BASE+TK1_RAM_SIZE) */
li sp, 0x4001fff0

Am I missing something?

[mchack-work]

I marked this as a potential bug a while ago. It's not serious and doesn't affect much, possibly just wasting a few bytes of RAM.

We have different behaviour in firmware's crt0 and tkey-libs, at least.

This needs some investigation. Maybe it's just a misreading of the RISC-V spec of stack alignment, or maybe it's something else, like I replied in e-mail earlier.

Since we plan on using tkey-libs in firmware, too, this needs to be looked at before that. See tillitis/tillitis-key1#129

[cobratbq]

I understand. FWIW, everything I've been doing with the TKey had the starting value set to the boundary. For me it works out. I also documented my deviation from your recommendations at https://crates.io/crates/rustkey. So far, I have had no complaints and no issues.