Skip to contentSkip to conversation

extended (syscall) error reporting

Ksummit-discuss Johannes Berg 1 participant 1 message Aug 1, 2016

lore
Share & tools
lore
#0 lore raw

I didn't see this proposed, but I don't think it ever got completely resolved either?

Tim Bird's note on the "more useful types" thread reminded me of this; he said something along the lines of "if we knew where each error is generated, we could more easily identify where a specific error could have come from."

The extended (syscall) error reporting would also help with that, and in particular with the netlink functionality there are always a lot of places that return -EINVAL and you have very little idea of where this came from. For netlink, there are various avenues of how to address this, and something special needs to be done there due to the structure (it's not even returned as an error from a syscall.)

However, it seems worthwhile to have a more general discussion here, perhaps trying to come up with a general "framework" or "structure" that everyone could use, regardless of the actual signalling mechanism towards userspace?

johannes