AMD Staging Another Fix To Try Correcting Some Raven Ridge Systems On Linux

[Chewi]

Then run openSUSE 15.1 + Kernel:stable as with your former.
I do this for ~26 Jahres, now. Appart the times I'm running upstream/devel stuff ;-)

Yeah, done that. So far, so good.

[nuetzel]

Yeah, done that. So far, so good.

Thank's for feedback. I've to migrate one or another system to 15.1, soon.

[r1348]

Yep, that is an excellent reason. Carry on



Put simply, we haven't found a way to get OEMs interested in Linux for anything but servers and embedded systems.

I initially thought it would be as simple as getting server departments to understand that they needed to beat on laptop departments to get the right support, but the missing link in most companies seems to be the realization that developer perception of a HW vendor factors into purchase decisions for servers.

Out here in the real world it's kinda obvious that developers are going to be using laptops and consumer desktop systems for most of their work and so that hardware needs to run the same OSes as will be used for deployment on servers, but most large companies seem to have enough separation between "laptop department", "consumer desktop department" and "server department" to make the connection difficult to see.

It's unfortunate that one of the few executives who *did* seem to understand this connection (Steve Ballmer) wasn't a big Linux fan, was focused on SW rather than HW, and didn't talk about the connection in a sufficiently restrained fashion for other execs to listen.

I see, I happen to work in one of those big companies. Big enough that we have our own servers built on order (btw we just added Epyc to the range of CPUs we offer to customers, so there's a plus).
We also internally use linux as a developer environment (or to whoever wishes to use it, really) and of course we run our corp-ified internal distro, and so had the previous two companies I worked for. I think that purchase would even consider buying AMD solutions if there was internal demand, but to my best knowledge very few developers care about it at all, so maybe you're targeting your evangelizing towards the wrong targets. Maybe those execs would listen more if there was internal demand for it to start with. And maybe linux is just good enough as it is, and OEMs don't see the need to spend resources improving what already (mostly) works.

I will abstain from any comment on Ballmer as I'm pretty sure he has the resources to outlast me in a legal battle.

But back on the main topic, I don't see why this specific issue is OS-related. I mean, SBIOS failing to load a firmware to a GPU component is something that happens indipendetly of the OS that system is running, and it's certainly detectable by a validation tool for any platform.

[bridgman]

But back on the main topic, I don't see why this specific issue is OS-related. I mean, SBIOS failing to load a firmware to a GPU component is something that happens indipendetly of the OS that system is running, and it's certainly detectable by a validation tool for any platform.

I haven't dug into this specific issue but from the context sounds like loading a too-old version of firmware rather than not loading it at all. Not sure.

Validation tools aren't the easy solution you might thing - problems like this generally happen when an OEM locks down their hardware/firmware a bit earlier than everyone else, when the firmware version would have been correct and a validation tool would have said "everything is fine".

[finalzone]

Validation tools aren't the easy solution you might thing - problems like this generally happen when an OEM locks down their hardware/firmware a bit earlier than everyone else, when the firmware version would have been correct and a validation tool would have said "everything is fine".

Now that Linux ecosystem has Linux Vendors Firmware Services designed for hardware vendors themselves, will AMD take advantage of such service?

[bridgman]

Now that Linux ecosystem has Linux Vendors Firmware Services designed for hardware vendors themselves, will AMD take advantage of such service?

I expect so, but not sure what the transition plan is between the linux firmware repo and this.

That said, it wouldn't help for cases like what I believe this one to be, where the firmware is embedded in an OEM SBIOS image. I believe that would still require and OEM SBIOS update but not 100% sure.