MGS4 [BLES00246] hangs on act 4 briefing end

[zorn-v]

Quick summary

RPCS3 Version: 0.0.27-14876-44b3709d Alpha | master

On loading screen it just close game window and open black new one with something like this in log

·E 0:01:17.952747 {PPU[0x1000000] Thread (main_thread) [liblv2: 0x017d29bc]} SIG: Invalid or unsupported instruction (op=0, reg=24, d_size=0, a_size=0x0, i_size=0)
·E 0:01:17.952751 {PPU[0x1000000] Thread (main_thread) [liblv2: 0x017d29bc]} SIG: decode_x64_reg_op(00007f42d9bfa3d2): unsupported opcode: 0xc46270f734034585f64c897518c6853c

·F 0:01:17.952898 {PPU[0x1000000] Thread (main_thread) [liblv2: 0x017d29bc]} SIG: Thread terminated due to fatal error: Segfault reading location 00000003e004322c at 00007f42d9bfa3d2.

Details

No response

Attach a log file

RPCS3.log.gz

Attach capture files for visual issues

No response

System configuration

No response

Other details

No response

[GamingPlutia]

I had that same issue today, I got around it by using a older build

[zorn-v]

What concrete build you are used for pass this issue ?

[GamingPlutia]

I used build 0.0.25-14499. Simply get the save from (dev_hdd0 > home > 0000001 > savedata) of the current build, and put the folder in the older build's savedata folder. Once that's done, simply use the older build to finish the MGS 1 flashback segment and save. Now that the flashback segment is over, you grab the new save file and put in the current build's savedata folder.

And by doing all of this, you're able to finally to play Act 4 of MGS4. It's a lot but pretty simple. Just compiling the PPU modules in the older 0.0.25-14499 build takes a long time.

[zorn-v]

I used build 0.0.25-14499

Thanks, it worked. But you need to completely remove .cache/cache dir to recompile PPU modules (including system).
After compilling you can use latest version of rpcs3.

Just compiling the PPU modules in the older 0.0.25-14499 build takes a long time.

You can disable "PPU LLVM Precompilation" in "Advanced" tab to not precompile all PPU at start game.

[AniLeo]

The error pointed in the OP has been fixed on #13807, it was a bug with Ryzen CPUs, so I'm closing this issue.

Please provide a log with latest build if you see the same error (Invalid or unsupported instruction) with the same repro steps so it can be reopened.

[AniLeo]

Reopening since it seems it's also affected by a different bug according to some users, please provide an updated log with latest RPCS3 build for this crash

If it's a regression, we need to know what build the regression was introduced on, otherwise it's unlikely this will get fixed anytime soon

[zorn-v]

RPCS3 Version: 0.0.27-15030-8db31809 Alpha | master
RPCS3 Version: 0.0.27-15035-73dba6d6 Alpha | master
All works fine with rebuilding cache (i left that save untouched).

[lithiumfox]

Can confirm that 0.0.27-15030 worked after downgrading from v0.0.28-15124-c8577592 master

[Nekotekina]

Can someone post error message from the recent master?

[lithiumfox]

Can someone post error message from the recent master?

Not a useful example, see below.

[AniLeo]

Giga is not supposed to be used to begin with. If the issue cannot be reproduced anymore on Safe or Mega, this issue can be closed.

[lithiumfox]

Hence why I was double checking to make sure it was consistent: It is not.

It still happens on both Safe and Mega. Just for the record: I was only reporting what I experienced with it. Giga did work on 0.0.27 and allowed me to get past this section, but I kept a save to load back to that goes immediately to the MGS1 segment. I'll keep trying to see if its more like a "random chance" type of thing, but it did work fine on 0.0.27 consistently compared to 0.0.28.

Here's the Safe logs:
RPCS3.log
RPCS3.log.gz

Mind again, it doesn't "crash" at least, nothing that closes any windows. It just hangs, FPS doesn't budge, nothing further happens, then it throws a pad failure.

Update: 0.0.27-15030 does appear to be working correctly compared to 0.0.28 on Safe SPU Size. I can still reproduce hang on 0.0.28 on Safe.

[lithiumfox]

If needed, here's the working log as a comparison example from 0.0.27-15030, config should be similar if not exactly the same as above. Takes about 5 seconds to load after the "restart" to load the PS1 emulation. Left the log in the previous comment open for about 2-4 minutes and nothing happened.

Not sure how your logs read, but figure it might help to see where maybe 0.0.28 stops compared to where 0.0.27 ends up.

RPCS3.log.gz

I will state this does seem like a regression.

[lithiumfox]

This was resolved for me by using the imported patch file provided by !explain bot on the discord and not the canary patch on the wiki

Removing the flickering texture patch seems to resolve this for me. Seems like a change recently made that patch no longer function with the Flashback sequence.

[ItsLaguna]

Can reproduce, but what I'll post is subjective ig since other people have had this issue for a while on builds older than the one I'm reporting. Not using any patches.

Latest working (for me): 15120.log
Broken: 15124-broken.log [PR #13840 ]

In 0.0.28-15120, after the briefing, you normally get a loading screen and the game window will close and open again to load the MGS1 flashback, and after you're done with the section, same thing will happen and you'll go back to MGS4.

In 0.0.28-15124, you'll get a blackscreen when the flashback is supposed to happen.

[solarmystic]

Can reproduce, but what I'll post is subjective ig since other people have had this issue for a while on builds older than the one I'm reporting. Not using any patches.

I can confirm this new finding. It straight up freezes with a blackscreen now (after the game window closes and opens again to load the flashback) instead of a long pause before the MGS1 flashback starts up.

Extra logs and screenshots from my end -

Working in 15120 RPCS3.log.gz

Not working in 15124 RPCS3.log.gz

EDIT:- Final error message before the black screen freeze is
E {PPU[0x1000000] Thread (main_thread) [HLE:0x015d1184, LR:0x0001c808]} SYS: 'cellPadSetActDirect' failed with 0x80121107 : CELL_PAD_ERROR_NO_DEVICE [1]

[AniLeo]

@brian218 can you check? Above report flags your PR as the broken one

[solarmystic]

Attached savedata to easily reproduce this issue for devs working on it.
BLUS30109G64680987.zip

[brian218]

Can reproduce, but what I'll post is subjective ig since other people have had this issue for a while on builds older than the one I'm reporting. Not using any patches.

I can confirm this new finding. It straight up freezes with a blackscreen now (after the game window closes and opens again to load the flashback) instead of a long pause before the MGS1 flashback starts up.

Extra logs and screenshots from my end -

Working in 15120 RPCS3.log.gz

Not working in 15124 RPCS3.log.gz

EDIT:- Final error message before the black screen freeze is E {PPU[0x1000000] Thread (main_thread) [HLE:0x015d1184, LR:0x0001c808]} SYS: 'cellPadSetActDirect' failed with 0x80121107 : CELL_PAD_ERROR_NO_DEVICE [1]

Please retest with master (a hotfix related to #13840 has just been merged two days ago).

[ItsLaguna]

Working in latest (15169).

EDIT: Nevermind, upon booting it again the same section, it freezes. Clearing caches allowed me to play through it again.

Also out of curiosity, I tried with master 15145 (the one with the hotfix) and it still hanged in that one, even with cleared caches. 15145.log

[brian218]

Working in latest (15169).

EDIT: Nevermind, upon booting it again the same section, it freezes. Clearing caches allowed me to play through it again.

Also out of curiosity, I tried with master 15145 (the one with the hotfix) and it still hanged in that one, even with cleared caches. 15145.log

Please confirm if sspu_spu.elf actually exists in your game dump.
I saw it in your log but I don't have that game so I cannot confirm.

[ItsLaguna]

Working in latest (15169).
EDIT: Nevermind, upon booting it again the same section, it freezes. Clearing caches allowed me to play through it again.
Also out of curiosity, I tried with master 15145 (the one with the hotfix) and it still hanged in that one, even with cleared caches. 15145.log

Please confirm if sspu_spu.elf actually exists in your game dump. I saw it in your log but I don't have that game so I cannot confirm.

No, it doesn't exist.

Also, I could get into the flashback again without clearing caches in latest, but what seems to happen is that it'll hang if you reload the section after you go through it and compile the next part of the game.

[brian218]

Working in latest (15169).
EDIT: Nevermind, upon booting it again the same section, it freezes. Clearing caches allowed me to play through it again.
Also out of curiosity, I tried with master 15145 (the one with the hotfix) and it still hanged in that one, even with cleared caches. 15145.log

Please confirm if sspu_spu.elf actually exists in your game dump. I saw it in your log but I don't have that game so I cannot confirm.

No, it doesn't exist.

Also, I could get into the flashback again without clearing caches in latest, but what seems to happen is that it'll hang if you reload the section after you go through it and compile the next part of the game.

Or try looking for it in caches.

[ItsLaguna]

Or try looking for it in caches.

Nothing, made a search in both the emulator folder and game folder.
I'll try clearing caches and compiling again, then leaving the game when I boot in the section and see if that file pops somehow.

EDIT: Nothing, also this time it'll freeze even without having gone past the section 😕

[brian218]

Please test #13995.

[ItsLaguna]

Please test #13995.

It's working, ran it a second time just to make sure and didn't face any hanging.

[brian218]

Please test #13995.

It's working, ran it a second time just to make sure and didn't face any hanging.

Nice! Glad to help.😄

[solarmystic]

Please test #13995.

It's working, ran it a second time just to make sure and didn't face any hanging.

Confirmed @ItsALH's findings, the MGS1 flashback now works perfectly with PR #13995. Thanks for the fix @brian218