Frequent server crashes after Memory Access Violation
Closed, ResolvedPublic


"Exception code: C0000005 ACCESS_VIOLATION at 006C1EA0
Allocator: C:\Steam\steamapps\common\Arma 3\dll\tbb4malloc_bi.dll
graphics: No
resolution: 160x120x32"


Legacy ID
Steps To Reproduce

Load a mission in multiplayer (using -server option) on a known stable machine.
Wait until the battle is raging and enough people have connected.
Server will then crash with the above (or similar RPT).

Additional Information
banshee edited Steps To Reproduce. (Show Details)Mar 25 2013, 2:21 PM
banshee edited Additional Information. (Show Details)
banshee set Category to Multiplayer.
banshee set Reproducibility to Always.
banshee set Severity to Crash.
banshee set Resolution to Fixed.
banshee set Legacy ID to 3003977051.May 7 2016, 1:05 PM
banshee added a subscriber: banshee.May 7 2016, 1:05 PM

Usually this crash occurs within 5-10 minutes, depending on how fast the server fills up.

Katos added a subscriber: Katos.May 7 2016, 1:05 PM
Katos added a comment.Mar 25 2013, 2:44 PM

Can confirm this, but it took 40 minutes for me (unsure how fast server filled up). Thats about the time it took for my dev branch servers to die - so it seems like the old dev is the new stable. This is on Windows 2008 Server.

If devs somehow missed it, almost every admin who was running dev build migrated back to stable because of really low uptimes. The stable could stay up for a day with the same map/server.

teaCup added a subscriber: teaCup.May 7 2016, 1:05 PM

We updated our Stable Alpha (non-DEV) "dedicated" server this morning, and it has been crashing every 10-30 minutes ever since. Also, when one of our players hosts on their own machine, it crashes pretty soon.

The server crashes with all types of missions, and when it does, all connected clients crash as well.

3 Error logs from our server, with crashdumps.

Katos added a comment.Mar 25 2013, 3:36 PM

Just noted the exception code is always the same - 006C1EA0 with the current stable, 00C21FC2 with the previous stable, 006C18B7 with the previous dev

Getting almost reliable crashes every 40 minutes with latest stable. As the upload here doesn't work for me, uploaded the logs to dropbox:

Aske added a subscriber: Aske.May 7 2016, 1:05 PM
Aske added a comment.Mar 25 2013, 3:58 PM

No crash at the DEV version, after half hour gameplay in multiplayer (4 players).

edit: It was a new game, we just started to test a new mission.

dazhbog added a subscriber: dazhbog.May 7 2016, 1:05 PM

Was the MP mission started by loading a previously saved game? If so, was the mission saved in previous game version? If so, what version?

Katos added a comment.Mar 25 2013, 3:52 PM

I'm always starting a new mission (404wasteland, didn't even know you can save/load?), happened after changing every appearance of Galkin into Ifrit in the mission files too.

Know from chat that others have the same problem with all kinds of custom missions aswell, so nothing related to wasteland alone.

We started Deathmatch/Capture-the-flag missions from zero. The server doesn't even have savefiles for them, so it has nothing to load from, i just checked. Still it crashes within 10 minutes.

Profecy added a subscriber: Profecy.May 7 2016, 1:05 PM

Confirmed issue on my machine, Windows Server 2008. Was fine before the update.

Well ok, I will take a look. But Does somebody have a clearly and "less time consuming" repro than: start server, connect some players and wait for crash? It would be very useful.

Katos added a comment.Mar 25 2013, 4:44 PM

Don't think there is a better reproduction. My previous dev server ran fine without/with one player for hours. But with max 20 players on it it reliably crashed in under an hour. It might be load dependent, so if you could run the server in a really low spec VM that might be faster.

Didn't find one yet.
Limiting the server to 20 players (could hold 26 stable before) didn't help.

I think that it might have to do with the numbers of connected players and some memory leaks or CPU related issue. Means: The servers crashes when under heavy load. I'll try to limit it to 10 players now to check that.

New Dumps (not our server but "Alex Grosse"):

I'd love to help narrowing the problem down, but i can't think of a commonality or pattern related to the crashes. The missions work perfectly fine, until it all goes kaputt. I know for sure there were no new objects created in one mission at runtime, no createVehicle commands, no new weapons referenced. And no player came forward with the usual "i killed a snake/threw a smoke grenade - and then the server crashed" type paranoia either.

Looks like

Fault address: 006C1EA0 01:002C0EA0 C:\Steam\steamapps\common\Arma 3\arma3.exe
file: Tier1_CPG_v21 (__cur_mp)
world: Stratis
Prev. code bytes: C0 7F 37 6A 01 56 8B CF E8 A6 80 FA FF 8B 46 34
Fault code bytes: 83 78 40 01 75 24 81 C7 18 0A 00 00 57 8D 4D 0C

EAX:00000000 EBX:00000001
ECX:01C4C76C EDX:00003F00
ESI:22CAB3C0 EDI:22C96040
SS:ESP:002B:01C4C7A0 EBP:01C4C7AC
DS:002B ES:002B FS:0053 GS:002B

Is consistent troughout all of our crashes. Exept for the Register ID's.

Troughout all the logfiles (from diffrent servers, diffrent missions) this is a reoccuring sceme:

Fault address: 006C1EA0 01:002C0EA0
Fault address: 006C1EA0 01:002C0EA0
Fault address: 006C1EA0 01:002C0EA0
Fault address: 006C1EA0 01:002C0EA0
Fault address: 006C1EA0 01:002C0EA0
Fault address: 006C1EA0 01:002C0EA0
Fault address: 006C1EA0 01:002C0EA0

So it's crashing with the same memory issues on the same memory adress.

More crashlogs:

Doesn't seem to be corelated with the number of players. Those crashlogs are after a couple of minutes with only 5-7 players.

Possible steps to reproduce better:

Start mission
Throw HE grenade(s)
Results in server crash regardless of how many ppl. are on there.

Reproducable on our server with diffrent missions (reproduced 4 times in a row now)

Crashes on Smokegrenades as well.

Just adding a note for clarification - this is related to dedicated server (correct me if i'm wrong).

It's related to running with "-server" option. Yes. That's why the category is set to multiplayer. :)

dazhbog, if you need more direct feedback, you can join our TS. Dwarden has the details.

A bit more verbose: If a player throws two grenades shortly after each other, the server will crash.
If two players throw one grenade each at the same time, the server won't crash.

BierAIG added a subscriber: BierAIG.May 7 2016, 1:05 PM

After the first patch in the Arma 3 Alpha when man throws two grenades crashes the game respectively. the server. We have tested this with several missions, and always the same.

Arma 3 Alpha v.0.52

I can provide logs for 3 separate server instances if required. As mentioned above it seems to be the same:

Exception code: C0000005 ACCESS_VIOLATION at 006C1EA0


Anywhere upto an hour it seems before the server resets although it if is a grenade throwing issue I can see why the range in times. Using a grenade-stop script as well that prevents throwing in base.. wonder if that has any effect?

It's affecting only the Stable branch, the one updated today (25th), the Dev branch doesn't crash the servers. Crashes both dedicated servers and so called "hosted" (both client and server on the same PC, same process).

Not sure if grenade trowing causes the crash, but we tested it with 4-5 players, and we managed to crash the server every time. Sometimes all it took was a single person throwing a single grenade and BAM, server and most of the clients crashed instantly as soon as the hand grenade was "fired". The player pulls his arm back to throw it, and the crash occurs, grenade doesn't even get to fly, land, explode, etc..

We could not reproduce the issue without grenades yet. Our guys are playing on a now passworded server for 2+ hrs without any crashes (everyone knows about the grenades).

Good to know thanks banshee... will do the same

edit: someone forgot and threw some it does crash for that as well :)

Has anyone tried using a different memory allocator?

I personally use ned_malloc, and I've not had any issues, but I don't have a server to test.

Update: With a private server and a "no nades" policy, our server runs without any crashes since 20:10 CET (5 hrs now).

Seems to be an infrequent frag or satchel. I've played on some servers where LAWS deployed by others and a grenade deployed by me were fine.

I heard a larger explosion (ie. charge/satchel?) from another player just prior to this "C0000005 ACCESS_VIOLATION at 006C1EA0" error.

Sounds like this should be a quick fix once they trace it down within the code. They likely won't need additional info as it's quickly reproducible.

Shame they don't allow us with Intel i7 (3.5Ghz) CPU's to have an option to run the game within a debuger. (ie. Statically link the program to run within gdb with a flag switch.) We would have had it traced for them already. But does appear to have a bulk of debug info already.

Well, due to this bug, stable branch is unplayable on all servers. Let's hope there's a fix very soon or likely today on this!

So far it seems that the problem is fixed already in the latest dev version, please confirm.

That's true, the problem was only in Stable, since the March 25 update. But Stable itself was ruined by it, becoming anything but. It forces all dedicated servers to switch to Dev, and consider that many players don't even know how to switch between Stable/Dev branches for their ArmA3 client.

Not really wanting to go to Dev version on servers.. this will only confuse players.

Also confirming it happens on any throw-able item. Tried the grenadestop.sqf script to cover the entire island but no difference.

This kind of surprised me, as it crashed pretty reliable without throwing frags as far as I knew. But I'm getting the same mem alloc. addy's so I believe it. I'm really stoked we have so many people rocking and rolling on an alpha and it makes the devs jobs a lot faster trying to find and troubleshoot this stuff. I'm hoping for a stable hotfix, so I dont have to switch server to dev.

Hotfix in progress, we don't want to force you to switch to dev servers as well :).

In fact.. We're working on a new stable release build which could be distributed this during week

Hotfix has been released!!! Thanks for the fast response to this issue BIS.

I can confirm this has been fixed with the Hotfix for the "stable" branch.
Good job!

Yep 5+ hours and running no crashes (dancing banana gif)

Mass close.

Add Comment