Linux/HPEnvy/Ryzen5 - formerly AMD vs nVidia... again

You'se guys really know what you're posting about, hat's off to you! I don't know what systemd messes up, but I've been against it out of principle since so many knowledgeable folks don't like what it does.
Win10 is disgusting, fingered that out for myself. Almost.

Thank you all very much for being such knowledgeable Members at GOL!
 
Systemd has its place - if all you're doing is running a home Linux desktop, using something like Gnome as your DM, and all you want is to have the basics (graphics, sound, internet, printing) it's no worse than any other init system. It does have some advantages - writing systemd service files is marginally easier than writing openrc init scripts. It's just that it doesn't do what I want it to do, and apparently my particular use case falls into the "you're holding it wrong" category...
 
Happy Saturday! Right before my next bit of overseas travel, I now have a laptop that boots successfully. Reading various support threads, Windows users are having a terrible time with this model laptop, with lockups under load, black, blue or white screens, screen corruption, colour bleeding and absolutely naff-all support from HP. Users of the slightly more recently-released Acer/Ryzen 5 laptop aren't much better off.

Seems that the curse of AMD drivers hasn't quite lifted yet.

HP ENVY x360 Convertible 15-bq1xx/83C6, BIOS F.13 11/10/2017

What's new:

Kernel - switched to Dave Airlie's drm-next git sources. This is based on a slight downgrade to 4.15-rc8, but with all the upstream stuff that'll appear in 4.16, in a couple of months.

Mesa/LLVM/Clang - 18.0_rc3, LLVM6, Clang v6 as in the earlier screenshot - this is the absolute bare minimum to get a usable "raven" chipset working. Yes, you can get some love with Mesa 17.3, but it's not good. I've just noticed that I now have LLVM7 in the tree, so will rebuild Mesa against that.

I haven't checked what the changed sources have done to performance. Feels like a slight regression, but I didn't keep detailed notes before. I'm just happy the damn thing boots.

What's not working:
  • The touchscreen and pen
  • Accelerometer (for screen orientation)
  • The F5 key (screen backlight) isn't registering an event, so I can't do fancy automatic backlight stuff. Responds to manual activation though.
  • Suspend/hibernate/resume - still down to the graphics drivers on this, they don't work quite the same as older Radeons and aren't going to sleep. Result: computer has no screen, no response to sysrq.
  • Rather bizarrely, my old JMicron USB to SATA/PATA bridge isn't properly registering. I'm pretty sure I've built the right module... will check against another laptop.

Not tested:
  • HDMI and HDMI audio - EDIT: works fine, sound and vision :D
  • SD Card reader
  • Secure boot / TPM - other people have posted logs showing that this will work, but I can't be bothered.

Everything that isn't mentioned above works just fine.

When I have a spare moment (I'm packing for more travel right now) I'll play some GAMES :D :D :D
 
Last edited:
That sounds mostly successful. Well done, but good luck with the touchscreen; that is a common problem.

Re: suspend & etc: does /proc/sys/kernel/sysrq=1?
 
ThunderRd said:
That sounds mostly successful. Well done, but good luck with the touchscreen; that is a common problem.
Click to expand...
Yes, so it seems. The device is an ELAN0732 touchscreen controller - it registers in sysfs, but the modules hid_multitouch and elants_i2c don't appear to do anything. Apparently there are a load of new touchscreen device IDs being included in the kernel for 4.16, so might be that people far cleverer than me will have already figured this one out and got it working:

tau-ceti /usr/src/linux # cat /sys/devices/LNXSYSTM\:00/subsystem/devices/ELAN0732\:00/hid
ELAN0732

modalias: acpi:ELAN0732: PNP0C50: (nb remove space between : and P)

So while waiting for the people cleverer than me to do clever stuff, I guess I've got to read up a little on the relevant system and see if there are any options I need to apply to the modules to get them to do... something.

Meanwhile, I've also found the accelerometer - there's /dev/HID-SENSOR-2000e1.1.auto and a couple of /dev/iio devices ready for use. Unfortunately, as I've never had such a fancy piece of hardware before, I don't even know where to begin polling these devices...

SD Card reader, at least, should work after I next compile the kernel. Guess who forgot to build the right module...?
ThunderRd said:
Re: suspend & etc: does /proc/sys/kernel/sysrq=1?
Click to expand...
Yes it does. With power management debugging turned on, I can see that part of the system issues a wakeup signal before all the others have finished suspending, so parts are trying to wake up while others trying to sleep. Doesn't end well. More to read about...
 
Last edited:
Fascinating! It's great to read posts that address problems I don't have a clue about. So why are you posting, cloasters?
Because it is very interesting reading, thank you!

Have a safe journey, Kaitain!
 
I'm not much further ahead than you are, in the "having a clue" stakes. Maybe I'll get time and brainpower to dig about in in the next few days.

Greetings from Iraq, by the way.
 
Thank you for the flattery, but I knew you were one smart cookie when we had the good fortune to meet, many years ago.
Please keep your head down, Iraq is still one unhappy place. I surely don't miss being in a conflict zone.
 
I'll post something about where I am and what I'm doing in another thread, probably in a week or two once I'm settled in. Suffice to say, the region I'm in is not a conflict zone, and the security arrangements are very thorough. So no major concerns there.

Ryzen-wise, I'm still getting intermittent lockups in X - sometimes after 5 minutes, sometimes 5 hours. It doesn't seem to matter what the workload is. Reading the mailing lists, this seems to be a common problem for anyone trying to use the "display core next" part of the driver, so Ravens, Vegas, and some SI chips. Unfortunately, for the developers, it's a hard lock so nothing gets reported in the logs :(
 
'Fraid that all of Iraq is still a combat zone. I pray that your security is top notch--and hope you won't need it to be. We did decades worth of damage there, we seem to forget that Iraq was a thriving country, only after our "help" did it become a third world nation. Yes, Sadaam was bad. That was their business to fix, not ours.
Sorry, I'm home after a same day surgery for a hernia. So I'm not all there. Ouch is the proper word.

Intermittent lock ups, carap, problems you surely don't need right now. I seem to remember that supposedly X was no longer needed, maybe it was only a Mint thing. But I betcha the booger is still in there. Thanks for saying howdy, it's always great to hear what's up with you.
 
cloasters said:
Intermittent lock ups, carap, problems you surely don't need right now. I seem to remember that supposedly X was no longer needed, maybe it was only a Mint thing. But I betcha the booger is still in there. Thanks for saying howdy, it's always great to hear what's up with you.
Click to expand...
X is still definitely needed. Wayland is the new black - or black screen, in Gizmo's case - but it's not ready to be a total replacement just yet.

This is partly because a great many applications will need to be updated, and partly because driver support etc. is still some way behind. So... it's possible to run with Wayland and an X shim, but will probably run like treacle on a frosty morning for the time being. Give it a couple of years, though - that project's seeing a lot of effort and is progressing rapidly.
 
Kaitain said:
Wayland is the new black - or black screen, in Gizmo's case - but it's not ready to be a total replacement just yet.
Click to expand...

Yeah, despite the fact that I was ragging on Wayland in the other thread, I fully understand the need to replace X, and it seems like Wayland really is the way to go. It's just not ready yet, and frankly, nVidia's dragging their feet on getting properly functioning Wayland drivers is part of the problem, IMO.

I can't find any numbers, but I would expect AT LEAST 1/2 the Linux desktop user base is running an nVidia card, which means those people are pretty well stuck in X-land, because the performance of nVidia using EGLStreams under Wayland is beyond abysmal. That then means that software developers really have no incentive to build apps that are compatible with Wayland.

Basically, a lot of the video improvements slated for Linux are being held hostage by nVidia, or so it appears to me, and nVidia really don't seem to care.
 
Kaitain said:
Funnily enough https://www.phoronix.com/scan.php?page=news_item&px=XWayland-EGLStreams-Support

I was reading one of the DRM mailing lists - dull, I know, but Iraq's not known for its nightlife - and it seems that nVidia rather likes the idea of Wayland, but aren't about to support any interim stages of development. Kind of, "if you finish it, then we'll come"
Click to expand...

I gave the EGLStreams stuff a whirl, but ended up with a blank screen. TBH, I might not have followed the correct steps, because by the time I got to that point I was very frustrated.
 
Nvidia is infamously SLOW with updates, especially for Linux. For this reason I've avoided their VidCards for years. Which is too bad, really. GeForce is fine but it's not a good feeling to be stuck with them.
 
Gizmo said:
(Have I mentioned recently how much I really, Really, REALLY hate Win 10?)
Click to expand...
Goes without saying big guy. But still it's always nice to hear...":O}
 
Last edited by a moderator:
Gizmo said:
I gave the EGLStreams stuff a whirl, but ended up with a blank screen. TBH, I might not have followed the correct steps, because by the time I got to that point I was very frustrated.
Click to expand...
My God I can relate to this! ":O}
 
... I'm just about at the point of sticking Windows back on this thing for the next 6 months, and trying again in August. But since I really don't want to find alternatives for some of the software I use, I'm considering rolling my / up into a vbox image, building the missing drivers for it to be a guest OS and proceeding from there.

Ho hum...
 
You must log in or register to reply here.
Back
Top