Enabling stack smashing protection boot loops esp32

[okhsunrog]

Is the protection supposed to work on all chips? If not, maybe we shouldn't allow enabling stack smashing protection for non-supported chips?

[okhsunrog]

Steps to reproduce - generating project for esp32 with esp-generate with stack smashing protection and running it

[bugadani]

Working theory is that this is a duplicate of #194

[okhsunrog]

@bugadani yeah, maybe, I had this thought too. I'll test after your PR is merged

[bugadani]

No actually this bug still occurs, and the sad part is that handling the panic caused by SSP triggers SSP an exception recursively so it doesn't end up printing anything useful.

[bjoernQ]

Interesting - pretty sure it worked on all targets back when it was initially added (long time ago): esp-rs/esp-hal#1135

[bjoernQ]

Given it's mostly an LLVM feature it might be related to some change there

[bugadani]

Given it's mostly an LLVM feature it might be related to some change there

Did you miss that this got fixed in esp-rs/esp-hal#3635 ?

[bjoernQ]

Did you miss that this got fixed in esp-rs/esp-hal#3635 ?

I mean stack-smash-protection is an LLVM feature - i.e. LLVM will instrument functions - from a brief look it seems Xtensa does something different than RISC-V now (while it did the same before)

[bugadani]

The feature does the same, the issue was that the linker behaves differently a bit and placed the stack guard symbol to 0x7FFsomething because the linker script misunderstood what we wanted.

[bjoernQ]

ok maybe I got confused by the last message here

No actually this bug still occurs, and the sad part is that handling the panic caused by SSP triggers SSP an exception recursively so it doesn't end up printing anything useful.

and I think I tested this on main before that got merged .... anyways would be good if it's fixed