19:42 fdobridge: <b​utterflies> my tip would be to take care of the LTS nvidia driver branches and not too much about the STS ones
20:14 fdobridge: <a​irlied> The problem is new hw support is often in sts ones
20:14 fdobridge: <a​irlied> You don't want to wait for an LTS one to enable support
20:15 fdobridge: <k​arolherbst🐧🦀> also.. we might need random updated firmware even from minor releases to enable newer hardware
20:17 fdobridge: <k​arolherbst🐧🦀> anyway.. we might have to track LTS branches once hardware support gets dropped
20:30 fdobridge: <k​arolherbst🐧🦀> oh wow.. bisecting that recovery failing thing and now it's a "kernel crashes for real" thing
20:31 fdobridge: <k​arolherbst🐧🦀> of course I bisected it to a " 28 files changed, 166 insertions(+), 132 deletions(-)" commit 🥲
23:45 fdobridge: <k​arolherbst🐧🦀> I'm curious now... because it runs into a `BUG_ON` thing through validating that a lock was taken, but that code changed a bit later on... but that kinda makes me more confident in my theory that the channel recovery might just be super racy somewhere
23:49 fdobridge: <k​arolherbst🐧🦀> anyway.. I'll try to figure out next week where pre turing channel recovery is broken, but using a commit before c358f53871605a1a8d7ed6e544a05ea00e9c80cb makes a huge difference