feat: configurable link section attribute for irqs

[sethp]

This change introduces a new config field that allows svd2rust to target which linker sections get assigned to the __INTERRUPTS static, with reasonable defaults.

Previously on RISC-V, the choice was always left up to the compiler, and it seemed to always pick .rodata. Unfortunately, in my context, that meant placing the LUT in a memory range that had a lot of highly variable latency, which cost not just time but predictability in servicing interrupts.

With this change in place, I'm able to target a particular section (e.g. .data, or .trap.rodata) for the placement of the static, which grants more granular control over the ultimate loaded memory address.

For the full details about the problem, please see: https://github.com/esp-rs/esp-hal/pull/534/commits/e29f3d547dc210e1b73313be6053a2122239a467

[burrbull]

I hope this is backward compatible?

[sethp]

Oh, great point: the code generator changes are in the sense that passing None for that parameter (or not setting a value, which should default to None) doesn't change the generated code.

Using the new parameter may not be backwards compatible, depending on what link section is being set: as long as the new section gets output in some part of the resulting binary it sure could be, but it's somewhat easy to rely on a particular layout, either accidentally or on purpose.

[burrbull]

bors r+

[bors[bot]]

Build succeeded!

The publicly hosted instance of bors-ng is deprecated and will go away soon.

If you want to self-host your own instance, instructions are here. For more help, visit the forum.

If you want to switch to GitHub's built-in merge queue, visit their help page.

• CI
• Rustfmt