Proton: Final Fantasy XV (637650)

[DistantThunder]

Hi,

Thanks for your hard work and to many more SteamOS users!

So FFXV won't run. But it display something I've never encountered.
Have you?

Seems like NVIDIA is bound to bother us one way or another...

Now the output is a bit sparse. Attached is the Wine log.
steam-637650.log

Since your distribution does not include winedbg, I'm afraid this is the extent of my investigations for now.

[themayarose]

So I tried running FFXV directly from the Steam Client and, although it got me a bit further, it seems to imply FFXV is a no go.

I can't attach the log file because it's a 60MB file, and GitHub won't allow files bigger than 10MB, but the important part seems to be this:

69284.019:0008:0009:trace:module:load_dll looking for L"usp10.dll" in L"Z:\\home\\ugopozo\\.local\\share\\Steam\\steamapps\\common\\FINAL FANTASY XV;C:\\Program Files (x86)\\Steam;C:\\windows\\system32;C:\\windows\\system;C:\\windows;.;C:\\windows\\system32;C:\\windows;C:\\windows\\system32\\wbem"
69284.020:0008:0009:trace:module:load_dll Found L"C:\\windows\\system32\\usp10.dll" for L"usp10.dll" at 0x7fece3750000, count=-1
69284.031:0008:0009:trace:seh:NtRaiseException code=c0000005 flags=0 addr=0x7fece339a2c2 ip=7fece339a2c2 tid=0009
69284.031:0008:0009:trace:seh:NtRaiseException  info[0]=0000000000000000
69284.031:0008:0009:trace:seh:NtRaiseException  info[1]=ffffffffffffffff
69284.031:0008:0009:trace:seh:NtRaiseException  rax=0000000000000002 rbx=0052006100350030 rcx=00000000c0000033 rdx=0000000000000000
69284.031:0008:0009:trace:seh:NtRaiseException  rsi=0000000001906590 rdi=0000000000226f50 rbp=00610056004e0032 rsp=0000000000226ba8
69284.031:0008:0009:trace:seh:NtRaiseException   r8=000000007bcc40e0  r9=0000000000003fd1 r10=0000000000000000 r11=000000000010ace8
69284.031:0008:0009:trace:seh:NtRaiseException  r12=0065003500390057 r13=0053006e00520044 r14=005a00730039006d r15=0059006a00740057
69284.031:0008:0009:trace:seh:call_vectored_handlers calling handler at 0x6a404550 code=c0000005 flags=0
69284.031:0008:0009:trace:seh:call_vectored_handlers handler at 0x6a404550 returned 0
69284.031:0008:0009:trace:seh:call_vectored_handlers calling handler at 0x6f275d60 code=c0000005 flags=0
69284.031:0008:0009:trace:seh:call_vectored_handlers handler at 0x6f275d60 returned 0
69284.031:0008:0009:trace:seh:dwarf_virtual_unwind function 7fece339a2c2 base 0x7fece339a040 cie 0x7fece33e49e8 len 14 id 0 version 1 aug 'zR' code_align 1 data_align -8 retaddr %rip
69284.031:0008:0009:trace:seh:execute_cfa_instructions 7fece339a040: DW_CFA_def_cfa %rsp, 8
69284.031:0008:0009:trace:seh:execute_cfa_instructions 7fece339a040: DW_CFA_offset %rip, -8
69284.031:0008:0009:trace:seh:dwarf_virtual_unwind fde 0x7fece3410478 len 11c personality (nil) lsda (nil) code 7fece339a040-7fece339acc5
69284.031:0008:0009:trace:seh:execute_cfa_instructions 7fece339a040: DW_CFA_advance_loc 2
69284.031:0008:0009:trace:seh:execute_cfa_instructions 7fece339a042: DW_CFA_def_cfa_offset 16
69284.031:0008:0009:trace:seh:execute_cfa_instructions 7fece339a042: DW_CFA_offset %r15, -16

According to a quick Google search, a c0000005 exception is an ACCESS_VIOLATION kernel failure. Since Final Fantasy XV incorporates Denuvo DRM technology, and Steam Play explicitly warns about invasive DRM as a impediment to running some games, it probably means FFXV is one of the games that won't be able to run on Steam Play. :(

EDIT: There's a misleading error message claiming "The server is not reachable, check your Internet connection and click 'Retry'." prior to the actual issue, and I wasted quite a few hours looking for that, since clicking the "Next" closed the application and seemed to imply that was the issue. However, after you enable logging, you can see a window briefly flash after you click "Next", and analyzing the log gets you to the actual culprit, the c0000005 exception.

EDIT2: The whole log, zipped: steam-637650.zip.

[themayarose]

So, I did a few more tests, and while it hasn't worked yet, I recant my statement that this is a no go.

I tried a few winetricks on the wineprefix (corefonts, usp10, dotnet46), and they didn't work. However, I got a picture of the window it flashes before the crash (yeah, I know, but I take pictures faster than I can screenshot on Gnome):

I also downloaded and installed Final Fantasy XII, which has the same Denuvo DRM, but it actually works. This window is supposed to display two text boxes, one with an input code and a blank one for the DRM website return code. It worked on Final Fantasy XII, but for some reason it crashes on FFXV. Which is sad, since you can easily find videos of Wine+DXVK running "alternative" versions of FFXV around, which implies that if this issue is fixed, the game will run fine.

Also, there's the issue of why the DRM can't automatically connect to the server and authenticate itself like it does on Windows. Fixing that probably would fix the issue altogether, since we'd never stumble on the codepath that fails here.

So TL;DR:

• Denuvo + other games: ✓
• FFXV without Denuvo: ✓
• Denuvo + FFXV: 🙁

EDIT: My system specs.

[themayarose]

Hi, a couple of updates on trying to run Final Fantasy XV on Proton (yeah, I really want to play this game on Linux 🙂).

Turns out the connection issue was related to the fact I was running with the Steam runtime disabled (STEAM_RUNTIME=0). Proton seems to hardlink against libgnutls.so.26 (as it's named on the runtime, there's no .so file), and my system has version 3.5. This caused Proton to have no encryption support, hence no connection to auth servers. However, running on the runtime wouldn't work either, since apparently some algorithms used by the authentication framework don't work with GnuTLS version 2.6. I was able to bypass that issue simply symlinking my system library to libgnutls.so.26, and that was the end of the connection issues and gnutls warnings. That part is now working.

However, I still can't launch the game, and I'm having a hard time finding the culprit by perusing the logfile:
steam-637650.log. There's a 406d1388 exception in that file, but this exception seems related to setting thread names, and, from my admittedly poor understanding of Wine log files, it doesn't seem to trigger a fatal failure on the launch sequence. Closer to the shutdown are references to winmm.dll and wintrust.dll, but they also seem to be well supported by Wine, so I'm out of ideas for now.

I will keep trying to run this game, though, and I will report any new successes or failures.

[DistantThunder]

@ugopozo There are several Denuvo versions.

Great debugging you're doing here. Thanks a lot.

[Admiral124]

Hello,
I am quite confident the issue with FFXV is Denuvo because I was able to get the game working without it. I have a backup copy of the game on my NAS with files from before the mod update and the debug executable (without Denuvo). I installed the current Steam version, then went offline and replaced the game folder with the older debug folder and, lo and behold, it worked. It's not perfect, there are some graphical issues such as shadows acting funky and entering the "Display" option in the settings menu locks up the game for a bit. Besides that, however, it is "playable".

[neilsimp1]

@Admiral124 Were you using any wineprefixes or anything else special when you got this running? I'm trying something similar now but it's not starting at all.

[Admiral124]

@neilsimp1 Nope. Just tried it again with the current 3.16-7 beta and it runs much smoother than before. I'm not using any launch options either.

[g00sed]

Is there anything to install with winetricks to run the non-Denuvo version of the game?

[Guy1524]

Here's a hack that allows you play it w/ the DRM:
always_succeed.txt

[hristoast]

Confirming the above patch allows the game to run.

Void Linux 5.3.10_1
RX590
AMD Ryzen 5 1600 Six-Core Processor
Proton built from sha: fd5b6b2a0e80270052d430da36a242b57701a11f

I can run it up to the "High" graphical preset and things run relatively well (aside from what I assume is stutter caused by shader compilation). When I try the "Highest" setting, the game runs around or lower than 10FPS. I didn't yet nail down exactly what the offending setting is.

I will also note that, if I don't set power_dpm_force_performance_level to high as seen on this AMDGPU issue, the game would freeze and I'd have to restart my desktop. I'm pretty sure this isn't specific to FFXV, though.

[doitsujin]

I didn't yet nail down exactly what the offending setting is.

Geomapping. Not sure why but it absolutely murders performance to the point where I'm having a hard time even recording a renderdoc capture of it.

Does anyone experience this on Nvidia by the way?

[doitsujin]

Turns out the performance issue with geomapping is a game bug; it consumes undefined input variables in one of its tessellation shaders. On Windows they are treated as zeroes, but there's no easy way to emulate this behaviour in DXVK.

[hristoast]

@doitsujin - I've rebuilt proton with DXVK 1.4.5 and the results are fantastic; I can play on Highest and my average FPS is in the 40s (or higher even).

I am seeing one issue, though: In some areas it seems like terrain LOD takes a minute to adjust, and while it does my FPS hovers below 20 or so (see the below image).

Another thing I've noticed while playing: when reviewing photos after camping, none of them render. In the photo reel all the images appear blank, and when I open one to view it is just a black screen. I don't know if this is specific to my system or not, the app log didn't seem to have anything helpful in it but I also had a hard time parsing it.

[doitsujin]

@hristoast ACO is buggy with geomapping right now; you might get better results with LLVM or AMDVLK. I haven't investigated that problem in detail yet (and it's bloody hard because my system constantly runs out of memory when trying to capture the damn thing).

Another thing I've noticed while playing: when reviewing photos after camping, none of them render.

They probably just don't get saved to disk properly.

[kust2708]

Hi,

I tried to run this game with the custom proton (4.21-GE-1) with the Guy1524 patch.
I got many time this :

And for an unknown reason, I managed to launch the game, some hours later at my first tentative.
I had to disable FSYNC and ESYNC to launch it.

↑ Parameters : 4k, very high, vsync off

As said @hristoast all photos are black and shaders take a lot of time (around 8 sec) to be applied in game.

After that session, I tried to come back to an official Proton build (4.11-10)... and since then, I can't launch the game anymore with the custom proton (same message than above).

Spec :
Archlinux 5.4.1-arch1-1-fsync
Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz
NVIDIA GeForce GTX 1080 Ti
RAM : 16 Go
Proton used : Proton-4.21-GE-1

logs :
steam-637650.log

[jatin-cbs]

Use Proton GE (custom proton glorious eggroll always works alot better than the default proton, it has that patch already ;)

[coffee-lord]

I've noticed a few fixmes in the logs which show up roughly during an attempt to take a photo:

fixme:wincodecs:copypixels_to_24bppBGR Unimplemented conversion path!

Wine - dlls/windowscodecs/converter.c#L1169

Could this be related to the photo bug?

Also, I've encountered "Sorry, something went wrong..." message as described in https://github.com/ValveSoftware/Proton/issues/74#issuecomment-565704579 after switching from Proton-4.21-GE-1 to Proton-4.21-GE-2. My guess is this happens whenever the wine configuration changes.

[GloriousEggroll]

The "sorry something went wrong" is DRM protection that triggers after launching the game too many times. Many games have this including MHW, SOTTR, Metro Exodus, BL3, and others. When this happens you have to wait about 24 hrs for it to turn off.

[nsivov]

I've noticed a few fixmes in the logs which show up roughly during an attempt to take a photo:

fixme:wincodecs:copypixels_to_24bppBGR Unimplemented conversion path!

Wine - dlls/windowscodecs/converter.c#L1169

Could this be related to the photo bug?

Could you attach a log when this happens, with WINEDEBUG=+wincodecs set, in user_settings.py for example.

[coffee-lord]

Update regarding the photo bug:

I've traced the copypixels_to_24bppBGR parameters and as it turns out it is asked to copy format_32bppRGBA images. Just to confirm my hunch I added format_32bppRGBA to the format_32bppBGR* block and the result is working photo mode and Prompto now takes pictures too.

Manual photo:

Prompto's photo:

Weird hue is due to incorrect pixel byte order, as expected. Shouldn't be too hard to hook up the proper RGBA to BGR conversion.

Edit:

Patch with correct BGR format:

diff --git a/dlls/windowscodecs/converter.c b/dlls/windowscodecs/converter.c
index dcc2f05..c6ace9c 100644
--- a/dlls/windowscodecs/converter.c
+++ b/dlls/windowscodecs/converter.c
@@ -1035,6 +1035,7 @@ static HRESULT copypixels_to_24bppBGR(struct FormatConverter *This, const WICRec
             return hr;
         }
         return S_OK;
+    case format_32bppRGBA:
     case format_32bppBGR:
     case format_32bppBGRA:
     case format_32bppPBGRA:
@@ -1073,6 +1074,8 @@ static HRESULT copypixels_to_24bppBGR(struct FormatConverter *This, const WICRec
                     srcrow += srcstride;
                     dstrow += cbStride;
                 }
+                if (source_format == format_32bppRGBA)
+                    reverse_bgr8(3, pbBuffer, prc->Width, prc->Height, cbStride);
             }

             HeapFree(GetProcessHeap(), 0, srcdata);

There could be a better way to do this but it works.

[nsivov]

I'd rather have it unrolled, going twice through seems unnecessary. Here's untested diff, thanks for tracking this issue down.

converter.txt

[RodrigoLeiteF]

Game instantly crashes after the main menu for me on Proton-GE 5.9. I assume it's not a DRM issue, since it would crash on launch if that were the case.

[Darklink999999]

I have the exact same problem @RodrigoLeiteF . I've tried multiple versions of Proton, from 5.8-GE-1 to 5.9-GE-4,5.9-GE-5, 5.11-GE-3 and 5.14-TKG. The game will just crash when loading the world.

[awsdert]

Hi,

Thanks for your hard work and to many more SteamOS users!

So FFXV won't run. But it display something I've never encountered.
Have you?

Seems like NVIDIA is bound to bother us one way or another...

Now the output is a bit sparse. Attached is the Wine log.
steam-637650.log

Since your distribution does not include winedbg, I'm afraid this is the extent of my investigations for now.

I think I have a similar problem, I recently bought a new computer, did the usual destroy windows and install linux (cus' linux rocks), eventually thought to boot steam via terminal and launch the game at the same time and I got the following output (with home directory removed for ~ because I feel uncomfortable putting that on the net)
[steam_FFXV.log](https://github.com/ValveSoftware/Proton/files/5152789/steam_FFXV.log)
, btw my new computer has a ryzen 7 8 core CPU with a Radeon RX 480 if that is of any help, also (because I deleted ALL old data) the crash occurs before it even reaches the title screen, it's either at language selection or just after displaying that "A Final Fantasy for returners and first timers" text that the crash occurs

[awsdert]

I just found this small log while I was hunting for potential reasons why FFXV won't get to the title screen for me:

2020-09-08T23:24:35.259[I]: Core.cpp:29     #0 Creating GfeSDK Version 1.0.132.d2f5f552
2020-09-08T23:24:35.265[I]: TaskRunner.cpp:42   #1 Start TaskRunner thread
2020-09-08T23:24:35.276[I]: FactoryImpl.cpp:17  #0 Creating default MessageBus wrapper
2020-09-08T23:24:35.282[E]: MessagebusWrapper.cpp:34    #0 Error creating message bus interface
2020-09-08T23:24:35.287[E]: Core.cpp:122    #0 Unexpected exception during NVGSDK_Create
2020-09-08T23:24:35.293[E]: CoreCInterface.cpp:90   #0 Error: failed initializing SDK -1020
2020-09-08T23:24:35.297[I]: Core.cpp:34     #0 Destroyed GfeSDK

I'm using a Radeon RX 480, vulkan works just fine (checked via vulkaninfo), any ideas?

[awsdert]

Was under this path
~/steamapps/compatdata/637650/pfx/drive_c/users/steamuser/Local Settings/Application Data/NVIDIA Corporation/GfeSDK/ffxv_s_208.log

[ShadeTechnik]

Same issue as above, game just crashes trying to load, tried GE 5.9-1 and previous, just crashes. The game worked some months back, not sure what happened.

I don't know if it did this when it did work, but I enabled "show framerate" in the options, it reports that I have 24Gb of Ram yet I only have 16. It reports not using any VRAM at all, and it reports that it I am using over 14Gb ram just sitting at the title screen

[gyscos]

I don't know if it did this when it did work, but I enabled "show framerate" in the options, it reports that I have 24Gb of Ram yet I only have 16. It reports not using any VRAM at all, and it reports that it I am using over 14Gb ram just sitting at the title screen

In system memory it includes swap, which may account for the difference. Vram reporting is broken but it shouldn't prevent the game from working.

I finally got it working just using proton GE 5.9-8. The 5.11 branch still doesn't work, neither does the main proton bundled with steam.

Since that I only had 1 crash after loading, but running the game again worked just fine.

[RodrigoLeiteF]

Game instantly crashes after the main menu for me on Proton-GE 5.9. I assume it's not a DRM issue, since it would crash on launch if that were the case.

The game works perfectly now, using Proton-5.9-GE-8-ST. Not sure if that's what fixed it, since I (unfortunately) had to wipe my machine in the meantime, but it's something.

[ShadeTechnik]

I'm still not getting it to load with 5.9-GE-8. Starts the menu but any attempt to load the game crashes. I am on an AMD GPU btw, 5700xt

[adrianlshaw]

Seems to work pretty well with the new Proton 5.13. Is anyone else getting variable framerates? (RTX 2070 @ 1920 x 1080)

[ShadeTechnik]

Tried Proton 5.13, and just about every custom version of proton I can find and it still crashes Running Steam from terminal I notice this error when it crashes:

mesa: for the --simplifycfg-sink-common option: may only occur zero or one times! mesa: for the --global-isel-abort option: may only occur zero or one times! mesa: for the --amdgpu-atomic-optimizations option: may only occur zero or one times! wine: Unhandled page fault on read access to 0000000000000008 at address 00007FE64A3309F3 (thread 012c), starting debugger... ERROR: ld.so: object '/home/justin/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.

[JacoG-RH]

@adrianlshaw Which kernel are you using? I'm experiencing similar issues to others now where the game does not load after the menu. Wondering what other factors there may be.

[ShadeTechnik]

I figured out my problem, well mostly. It has something to do with the
vulkan ICD loader. In order to get doom eternal working early on I had to
switch ICD loaders. I can't remember the details but now I can get the game
running if I use Lutris to launch the Steam runner and selecting an ICD
loader since it allows multiple choices.

If necessary I can try to get more details but it runs perfectly this way.

On Wed, Oct 28, 2020, 8:16 AM JacoG-RH [e-mail] wrote:

@adrianlshaw https://github.com/adrianlshaw Which kernel are you using?
I'm experiencing similar issues to others now where the game does not load
after the menu. Wondering what other factors there may be.

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/74#issuecomment-717894475,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ANIOO6T6P2SLJEGGPGAUCDTSNADS3ANCNFSM4FQ5XXKA
.

[adrianlshaw]

@JacoG-RH 5.4.0-52-generic, which is part of Ubuntu 20.04. Since the game uses a huge amount of disk space it's installed on a HDD. Will see if performance improves much on the SSD. It averages around 30fps but haven't found any crashes or weird bugs yet. Some motion blur effects do cause the framerate to be volatile at times.

[ShadeTechnik]

Not sure how well the reporting works with fps on but the game reports
using alot of RAM, over 16gb. Runs at a solid 60fps for me @1440p on a
5700xt with everything maxed out and the Nvidia stuff turned off. I'm on
kernel 5.8

On Wed, Oct 28, 2020, 10:19 AM Adrian L. Shaw [e-mail]
wrote:

@JacoG-RH https://github.com/JacoG-RH 5.4.0-52-generic, which is part
of Ubuntu 20.04. Since the game uses a huge amount of disk space it's
installed on a HDD. Will see if performance improves much on the SSD. It
averages around 30fps but haven't found any crashes or weird bugs yet. Some
motion blur effects do cause the framerate to be volatile at times.

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
https://github.com/ValveSoftware/Proton/issues/74#issuecomment-717964845,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ANIOO6RP2URNDIXYR3RZMVDSNASAFANCNFSM4FQ5XXKA
.