Rayman Pirate-Community

Seems like you just need to change your kernel configuration and turn off stack protection. Can you do "make menuconfig" or edit the .config file?

Where’s that option located? I can’t seem to find it in that maze-like structure of options.

Go to "General Setup" and set "Stack Protector buffer overflow detection" to "None". Can't guarantee that you won't have other problems, mind…

I hope it’ll go ok. Will try when I get back on my Mac, as I’m not at home.

EDIT: I’m still getting errors.
EDIT 2: Stupid me, I’ve ticked “64-bit kernel” and I’m using a 32-bit VM! It’s working fine now.
EDIT 3: I’m getting “low disk space” errors so I’ll have to restart from the beginning

(Whatever.)

Any luck, Fifo?



I keep forgetting to post a screenshot of my current laptop use of OpenSuse 42.2, dang it.. Oh well, currently I've been busy with my server, which is going quite well. Just me not using linux enough to remember commands is what kills me.

Hoodcom wrote:Any luck, Fifo?

No luck, just errors

Guess I’ll have to experiment more.

I’m trying again from scratch.

EDIT: I got the kernel to compile successfully!
EDIT2: It’s finished, the whole system is installed! I’ve also installed GRUB2 and set up networking.

Well done Fifo!

There's something I've been scratching my head about lately, in relation to VMWare. I'm not really expecting to get an answer here on PC, but I just want to say all this in public, because it's quite odd.
Linus seems to love breaking APIs, because every few kernel versions you get some third-party driver that won't compile anymore and needs to be patched. Happens quite often with VMWare's kernel modules. And pretty soon someone gets onto a blog and puts up a quick-and-dirty patch, which gets adopted in some form or another by VMWare themselves shortly after.
So right now, everything's dandy. The latest version of VMWare Player ships with modules that will compile with up to Linux 4.9 (and indeed 4.10 because they didn't break any APIs this cycle). Except it doesn't. The modules "vmnet" and "vmmon" are patched to work with 4.6 or 4.9, as detailed by blog posts such as this one. But there are two other modules, "vmci" and "vsock". They have not been patched this way. They still call the outdated version of "get_user_pages", and also use things like "vmalloc" and "vfree" that have been moved into different headers, and other more headache-inducing API changes.
I can't understand why noone other than me seems to be affected by this. Any blog posts I've found addressed only the "vmnet" and "vmmon" issues. I mean, what's the story? Everyone else is working with the same modules and kernel sources as I am, right? As it stands, VMWare won't start unless I compile a patched (by me) version of "vmci". Failing "vsock" is fine, but it still causes annoying messages. When I tried to patch "vsock" myself before, it compiled (thus getting rid of the annoying messages), but caused vmware-vmx to get stuck in uninterruptible sleep when my Windows Vista VM was shutting down, thus causing me to have to hard-reboot the entire PC.
Do most distros just not need these latter two modules anymore? Am I doing something crazy?

Again, I'm really just thinking out loud here, but I think these things need to be said.

PluMGMK wrote: Linus seems to love breaking APIs,

Yes, Linus breaks hardware too. I've seen him try making a custom router and ruined like... three motherboards in the process.


Anyway. Will Plum catch the joke? We shall never know!

I finally am going to post the screenshot of OpenSuse here on my laptop. Especially since I still have a bit of road trip ahead, on my way back home.

[offtopic]
Oh man, seeing Discord running in Hoodcom’s screenshot is just making me want to tell Hoodcom to join my Rayman server but I know he’s gonna refuse

Well atleast add me as a Discord friend? I’ll PM you my Discord Tag.
[/offtopic]

PluMGMK wrote:Well done Fifo!

Thanks!

Wow, that’s a cool openSUSE setup.

PluMGMK wrote:but I decided it was time to consider the other option: "replace the software". In the back of my mind, based on a recommendation Inco made some time back, I had an idea to just do a clean install of CentOS and abandon Ubuntu.

So from what I read the hint was confirmed to be pretty good, I haven't tried it by myself since I do everything with Slackware and Raspbian, but the dude that gave me tip is an engineer working with servers so I guess he knows his topic.
Damn, I only come to read this today !

PluMGMK wrote:Attempting to mitigate this by uninstalling the proprietary NVIDIA driver only caused the resolution to drop to an unacceptable 640×480 (again, this was over HDMI ).

WHY DID YOU HAD TO REMIND ME OF THIS

Hoodcom wrote:I finally am going to post the screenshot of OpenSuse here on my laptop. Especially since I still have a bit of road trip ahead, on my way back home.

KDE heh ? I used to like it some time ago (back in 2015 it was my main DE until the end of August ) but I rapidly became frustrated by how slow and performance hungry it was, in addition the KDE suite is really big.
My desktop hasn't changed since I switched to Awesome, a few updates in the Lua code but I didn't even switched my wallpaper
Anyway, you should check out Enlightenment, I think it has even more than KDE when it comes to aesthetic, and in addition it is fast and lightweight.
I really don't know how you managed to make Dragon Player work, that shit never ever functioned on my side and I find Mplayer and VLC a lot more better.
You manage to play vidiogames well on that OS Hood?

Fifo wrote:EDIT1: I got the kernel to compile successfully!
EDIT2: It’s finished, the whole system is installed! I’ve also installed GRUB2 and set up networking.

Boi you managed to make a huge step here, you actually had the patience to cook yourself a gentoo, man... And even to compile a kernel by yourself, gosh I hate so much doing this, feels like damn surgery.

That's a mighty fine KDE desktop! Makes me think that rather than spurning KDE I should sit down and take some time to make it beautiful. But it's easier to just keep using i3/sway. (On a somewhat related note I recently disabled all the Windows 7 eye candy on my work computer since I found it distracting!)

incognito wrote:WHY DID YOU HAD TO REMIND ME OF THIS

"Have" to, not "had" to, my dear!

incognito wrote:Boi you managed to make a huge step here, you actually had the patience to cook yourself a gentoo, man... And even to compile a kernel by yourself, gosh I hate so much doing this, feels like damn surgery.

I've been compiling my own kernels for over two years now, it's become fairly mechanical for me!

À propos of nothing, I've been considering dumping VMWare altogether, and trying QEMU with KVM, and possibly passing my graphics card through to a Windows 10 virtual machine. It's already happened a few times that I've wanted to play something for which neither Wine nor Vista will cut it, and dual-booting is a pain in the proverbial. I know I tried something like that back in 2014, and failed rather miserably, but I've learned a lot more about Linux since then.

Forget what I've said about the wallpaper, I've just changed it

Am I going crazy or can I see a face?

You're going nuts

incognito wrote:

Fifo wrote:EDIT1: I got the kernel to compile successfully!
EDIT2: It’s finished, the whole system is installed! I’ve also installed GRUB2 and set up networking.

Boi you managed to make a huge step here, you actually had the patience to cook yourself a gentoo, man... And even to compile a kernel by yourself, gosh I hate so much doing this, feels like damn surgery.

Yeah it was painful

Well, I used snapshots in VMware.

incognito wrote:You're going nuts

Thank goodness!

Fifo wrote:Well, I used snapshots in VMware.

Cheater!

Hmm, I guess my linux based laptop could use some work, since the graphical user interface did freeze up on me during the end of the trip while watching Samurai Jack on Hulu... audio playback was still going like nothing was wrong, mouse cursor was moving like nothing was wrong, it's just nothing else would work. Couldn't even click another window or anything. Nor the KDE taskbar.

@Incognito...

KDE doesn't seem power hungry to me, in fact I think I get better battery life out of it than I do Gnome.

Video games, yes, older ones under WINE anyway, and a couple on Steam. The laptop's not going to handle anything major, being it's from 2009 and has Intel IGP.

That Samurai Jack experience reminds me that the one thing I do miss about Windows is (usually) being able to override everything by going Ctrl+Alt+Delete. Perhaps I should mention that to our friend Lennart…

Better battery life with KDE? I suppose that makes sense, since with GNOME everything's foisted on you, but KDE allows you to customize the hell out of it. Though as I alluded to above I'm not too familiar with the ways in which KDE can be customized.
I just noticed that Inco mentioned Enlightenment. The extent of my familiarity with it is, really, a flamewar one of its developers had with SirCmpwn, the creator of my WM of choice. But I've been on its website a few times, and I do seem to like what I see!

I do like the freedom given by KDE and the like. Though sometimes things do seem a tad unstable, but then again, I don't know if that's KDE or something else playing up on that laptop. I just know I'm clueless as I had no error to go by.

Did you check the kernel log with dmesg? Or any other logs?

PluMGMK wrote:À propos of nothing, I've been considering dumping VMWare altogether, and trying QEMU with KVM, and possibly passing my graphics card through to a Windows 10 virtual machine. It's already happened a few times that I've wanted to play something for which neither Wine nor Vista will cut it, and dual-booting is a pain in the proverbial. I know I tried something like that back in 2014, and failed rather miserably, but I've learned a lot more about Linux since then.

I succeeded! I bloody succeeded!
Thanks to this blog I was able to get this set up with QEMU, KVM and LibVirt within the space of about seven hours. It would've been a lot quicker on a distro, since being on Linux from Scratch meant I had to compile QEMU and LibVirt myself, along with myriad dependencies!
Please note that to understand this post fully, you probably will have to consult that blog I just mentioned! I found it quite accessible though.
I figured the first thing I had to do was compile a kernel with the vfio-pci driver, and so I waited for Linux 4.11 to come out since this occasion would mean I would be compiling a new kernel anyway. This was rather silly in hindsight, since I could've at least compiled QEMU, LibVirt and their dependencies on Sunday, and had less work to do yesterday!
But anyway, I built that kernel, with the addition of the vfio-pci and i915 (Intel Graphics) drivers. I then rebooted, went into the BIOS (or UEFI I guess) settings, and changed the main GPU from PCIe graphics to the integrated one. I also connected the VGA port on my motherboard to the secondart input of my monitor. The primary input is, and has always been, connected to the DVI port on my NVIDIA card.
I noticed that I had two "firmware" options for my VM. One was a traditional BIOS, in which case I would need to use the kernel's VGA arbitrator, and therefore apply an extra kernel patch to make that work with the Intel driver. The second option was UEFI, using an OVMF binary. I decided to do the latter, and set about acquiring one of those binaries. The blog says that compiling one from source is a pain, and that it would be better to just get a prebuilt one. I decided I knew better, since the prebuilt ones come in RPMs, which I'm not set up to handle anyway! I set up the toolchain for compiling my OVMF binary, and went about doing it. It's supposed to be done with GCC 5 or earlier, but I was able to hack it to compile with my GCC 6.3!
I went ahead and compiled QEMU and LibVirt and various dependencies, and tried out LibVirt's graphical Virtual Machine Manager. I was able to create a test VM, but then it was unable to start the virtual network, complaining of being unable to initialize a valid firewall backend. Since doing that would require installing more dependencies, which could also screw about with the host network configuration (which I had built on the KISS principle), I decided to leave networking out for the time being.
After that, there were various other problems which led me to need to go back and compile more dependencies. Fine. Eventually I was able to start my test VM and look at it in the Virtual Machine Manager. But all I saw was black. I kept trying various things, but eventually I concluded that my GCC-6-compiled UEFI firmware was, in fact, braindead! I went back and installed a little tool to extract RPMs and downloaded a prebuilt OVMF binary. I changed the LibVirt settings to point to this new firmware, then recreated the VM from scratch. And… it booted! The first thing I tested it with was a CentOS ISO image, but once that worked, I was ready to try to get it to boot Windows!
Even though, as I said, dual-booting is a pain, my PC does have a hard drive with Windows 10 on it. I finally caved in and put it there last summer, when I bought the Solus Project and discovered that it couldn't run under Wine, and that VMWare's emulated GPU just wouldn't cut it! Anyway, it didn't take long to figure out how to attach this physical hard drive to my VM. This couldn't be done using the Virtual Machine Manager GUI, so I had to edit the VM's LibVirt XML file. No problem.
I proceeded to try to boot Windows. It seemed to work, but it kept complaining of errors and going into recovery mode. At first I thought it was because the VM's firmware was only seeing the Recovery partition and not the main one, and I couldn't imagine how I'd crack that nut.* But then I tried the "Reset my PC" option, figuring it couldn't do that much harm. It informed me that the hard drive was locked. So I checked what partitions were mounted in Linux, and sure enough, there was the partition with my Steam library, which is on the same physical disk. I was quite alarmed, since I was sure I'd already unmounted it, knowing that attaching a mounted drive to a VM is very dangerous.
Anyway, I unmounted the partition, shut off the VM and started it again. This time, Windows booted! Great! The firmware and the OS were working!
I figured that, before going for the big prize, namely passing through the graphics card, it would be best to figure out what was going on with the network. A bit of searching online revealed that for the firewall backend, it was sufficient to install iptables, ebtables and dnsmasq. I went ahead and compiled and installed those, but it still wasn't working. I needed to do a bit of debugging, including hacking the LibVirt source code itself, but I figured out that I needed to move the ebtables binary from /usr/sbin to /sbin, and that for it to find the dnsmasq binary, I needed to reconfigure, recompile and reinstall LibVirt. Once I did all that, I was able to set up a nice normal NAT virtual network!
I started up the VM again, confirmed that the network was usable from inside, briefly cursed the fragmentation of settings in Windows ≥ 8 , and installed a TightVNC server. I then installed the TigerVNC client on Linux, and confirmed that I was able to use it to access the VM's user interface. Good. I'd definitely need that once the display was going directly to the NVIDIA card, rather than through the VM manager GUI.
I had been fretting earlier about how I'd go about assigning my NVIDIA card to the vfio-pci driver at boot-time. All the tutorials for doing it seemed to assume the use of a ramdisk, which I don't have since I'm on Linux from Scratch. There was a tutorial with a shell script on a Fedora wiki. I decided, what the hell, I'll just run those commands right now, and see if they work. Sure enough, they did! I was able to unbind the (unused) NVIDIA GPU from the nouveau driver, and all it did was cause the fan speed to change! No lockups, no kernel panics, no nothing! Seemingly the kernel's got more robust since some of these guides were written! So I was then able to just bind the NVIDIA card's GPU and audio device to the vfio-pci driver.
Now, before attaching the NVIDIA card to the VM, I made sure to go and edit the VM's XML file again, to hide KVM from the guest. This is because NVIDIA's Windows driver locks up when it detects KVM. Ostensibly it's an unintentional but WONTFIX bug, but it seems obvious that it's to prevent people from using GeForces in this way when they want VM users to by Quadros instead. But anyway, with the circumvention complete, I was able to attach the card to the VM, boot it, and switch the input on my monitor to DVI. And it worked! Well actually, it booted up in 800×600 at first. Seemingly it took the NVIDIA driver a few minutes to figure out that the card was there, but once it did the VM operated in full HD! I guess the reason was that the PCIe bus position of the card in the VM was different from what it had been on the physical PC.
I soon figured that the best way of interacting with this Windows system was to bring up TigerVNC on the host, put it in full screen, then switch the input on my monitor. Since the VNC window is full screen, and both systems are operating at 1920×1080, the keyboard and mouse are directed to exactly where I'd expect them to be on the "real" screen.
It's still not fully set up. Notably, I need to get the sound working. Currently the only audio device accessible to Windows, as far as I can tell, is the NVIDIA card's HDMI audio, which is obviously unsuitable since my monitor is DVI. PulseAudio for Windows is apparently a thing, but when I see the words "Windows XP" mentioned on the site, I don't feel very encouraged. I think Jack is probably the best way to go. I read about using it for this purpose back in 2014. The other thing I want to do is write a shell-script that automatically unmounts the partition I mentioned above, rebinds the graphics card to the VFIO driver, starts the VM and launches TigerVNC. That's about half-done, and shouldn't be too difficult to complete.
So right now, I've got Windows with access to my high-end graphics card, and Linux running on my integrated Intel GPU. This isn't ideal. I would prefer two dedicated GPUs, one for Windows and one for Linux. I had a separate card before, from when I tried (and failed) to do this in 2014, but it was low-end due to my misunderstanding of AMD's (then) new GPU naming scheme, and I can't seem to figure out where it is now (probably in an anti-static bag in a box somewhere, but which box?). The other problem with this is that my motherboard actually runs both PCIe slots from the CPU root port, which has no ACS isolation. Therefore if I installed two graphics cards they'd be in the same IOMMU group, and so it would be impossible to assign one to the VM without the other, which of course would be pointless! It's possible to cheat and patch the kernel to force them into separate groups, but that's not really good practice. I'll keep looking into it anyway. There's at least one old GPU lying around at home!

*Why have I a Recovery partition, you ask? Well, this Windows 10 installation was actually, er, copied from a hard drive in a laptop that my father has, but doesn't use anymore. So this particular copy of Windows has been moved around quite a lot!