DRM Updates Land For Linux 4.15; Torvalds: "There's Something Odd About DRM People"

[phoronix]

Phoronix: DRM Updates Land For Linux 4.15; Torvalds: "There's Something Odd About DRM People"

Overnight the Direct Rendering Manager (DRM) display/graphics driver updates were sent out and ultimately pulled into the mainline kernel for Linux 4.15. This doesn't yet include though the separate AMDGPU DC pull request...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-4.15-DRM-Lands

[tildearrow]

GPU reset support for RX Vega GPUs

Awesome. Very little typo:

and a lot of other low-level -improvements.

[Sonadow]

I have this sinking feeling that Linus is going to explode when he pulls in the AMDGPU DC code...and then undo the pull.

[Kendji]

Thumbs up for the DC code

[FireBurn]

It's code that's being used - i think he'll pull it, issue the riot act and say he wants a timeline for the todo list

[coder111]

Well, Christ on a stick indeed.

I agree with Linus on this one. How did that .h file get this far up the maintainer chain without anyone noticing it is unused? Or at least adding a comment with an explanation why it's needed in the kernel tree even though it's unused...

[L_A_G]

It's not the first time David has pissed off Linus by pushing things that have caused him unnecessary headaches, but this soon after the tinyDRM debacle I wouldn't be surprised if Linus' patience is starting to wear thin with David and the other senior DRM maintainers. Remembering that Linus runs Linux development based on a chain of trust model where the proverbial sh*t travels up the chain rather than down (like in most hierarchical organizations) I'm beginning to wonder at what point Linus starts replacing senior DRM maintainers.

I don't want to sound like I'm harping on David, but the guy did dismiss the legitimate criticisms of his last Linus-angering f*ck-up as Linus just being stupid so this obviously won't be the last time he pisses off Linus as what he considers acceptable has quite a bit of overlap with what what Linus considers unacceptable.

[microcode]

Maybe something should be done on the build system side to ditch these auto-generated headers of mysterious logic and origin. Then again, it's possible they're from shared tools, in which case god only knows how complicated it would be to share or replicate the generator.

[mphuZ]

Something tells me that the merger will be postponed to 4.17...

[agd5f]

Linus is just whining to for the sake of whining. It absolutely makes sense to use machine generated register headers. They eliminate human error and enable more people to get involved. If there is a feature someone wants to implement, all of the register information is there and correct. It also makes debugging a lot easier since you have ready access to all of the additional registers you might need to look at or ask someone to dump. Adding it manually on a per feature basis is error prone and makes it harder for someone working with the code to see the full picture and capabilities.