Add keepalive feature.#592
Conversation
This allows a debugger to request that a given hart stays available for debugging (e.g. is not powered down). Fixes #519.
Ernie Edgar (ernie-sifive)
left a comment
There was a problem hiding this comment.
I think this simpler approach is the right thing to do. Just one minor nit...
Co-authored-by: Ernie Edgar <43148441+ernie-sifive@users.noreply.github.com>
|
Paul Donahue (@pdonahue-ventana) does this look OK? |
Feedback from Robert Chyla.
|
Robert Chyla (@RobertChyla): You made the original request so what are your thoughts? This allows a debugger to request that harts should stay available but there's no feedback about whether this feature is implemented or if it will be respected. As such, I think that debuggers are not simplified since they'll still have to deal with things going unavailable unexpectedly. At least now with stickyunavail they know that this happened. But it's a nice feature to at least allow the debugger to make the request. I think that some sort of discovery mechanism would be an improvement to at least know if the DM supports making the request (and whether the power controller will honor it may be beyond the scope of the DM spec). |
|
I contacted Robert by e-mail, and he suggested the slight rewording, and said "it is OK as it is now." Even if the feature is supported, some hardware will not be able to guarantee that the hart will remain available, so debuggers will still need to deal with harts disappearing. If we had a config structure I'd just say stick some bits in there that describe possible behavior, although that still doesn't handle the case where the DM itself doesn't even know whether the system it is in can honor the request. Seeing how we're freezing on Monday, I'd like to merge this as-is, and you can open an issue for 1.1 to address discoverability. Thoughts? |
|
I was surprised that my innocent email paragraph end-up as spec change. |
OK |
Addresses #519. Went with the simple version without reading the keepalive values.