[0x08000800]: data fabric sync flood event

ninjachicken1 · Yesterday 00:14:10

I've been experiencing the following error on my desktop:

Sep 20 22:17:47 archlinux kernel: x86/amd: Previous system reset reason [0x08000800]: an uncorrected error caused a data fabric sync flood event
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: System Fatal error.
Sep 20 22:17:47 archlinux kernel: fbcon: Taking over console
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: CPU:12 (19:21:2) MC5_STATUS[-|UE|MiscV|AddrV|PCC|TCC|SyndV|-|-|-]: 0xbea0000000000108
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: Error Addr: 0x01ffffffc2188462
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: IPID: 0x000500b000000000, Syndrome: 0x000000004d000000
Sep 20 22:17:47 archlinux kernel: 
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: Execution Unit Ext. Error Code: 0
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: cache level: RESV, tx: GEN, mem-tx: GEN

It can happen randomly, sometimes within 10 minutes of turning my computer on in the morning, sometimes after a few hours. It also happens consistently after launching Palworld (fairly graphically intense game) for a few minutes.

After this happens, regardless of whether it's random or because I attempted to play Palworld, it takes a few seconds to load kernel modules on boot and 13 seconds into that the GPU is initialized and my desktop becomes extremely laggy until I turn it off and leave it a few minutes.

Apparently "[0x08000800]: an uncorrected error caused a data fabric sync flood event" is a common bug that nobody's figured out the exact cause of, with no specific fix that works for everybody. Supposedly it's connected to a linux kernel update.

Comparing the above error to a previous example, I noticed that aside from the CPU core and the Error Addr, the errors are identical:

Sep 20 18:29:22 archlinux kernel: x86/amd: Previous system reset reason [0x08000800]: an uncorrected error caused a data fabric sync flood event
Sep 20 18:29:22 archlinux kernel: [Hardware Error]: System Fatal error.
Sep 20 18:29:22 archlinux kernel: fbcon: Taking over console
Sep 20 18:29:22 archlinux kernel: [Hardware Error]: CPU:3 (19:21:2) MC5_STATUS[-|UE|MiscV|AddrV|PCC|TCC|SyndV|-|-|-]: 0xbea0000001000108
Sep 20 18:29:22 archlinux kernel: [Hardware Error]: Error Addr: 0x00006ffffbc7eaf2
Sep 20 18:29:22 archlinux kernel: [Hardware Error]: IPID: 0x000500b000000000, Syndrome: 0x000000004d000000
Sep 20 18:29:22 archlinux kernel: 
Sep 20 18:29:22 archlinux kernel: [Hardware Error]: Execution Unit Ext. Error Code: 0
Sep 20 18:29:22 archlinux kernel: [Hardware Error]: cache level: RESV, tx: GEN, mem-tx: GEN

Here's my system specs:

Kernel: 7.2.6-arch2-1
DE: Plasma 6.7.5
CPU: AMD Ryzen 7 5800X3D (16) @ 4.552GHz
GPU: NVIDIA GeForce RTX 4090

GPU Driver: nvidia-open-dkms 615.71.09-1

Here's a couple of relevant journals using systemd.log_level=debug:

https://paste.c-net.org/SolvedFiles - in this session I launched Palworld to reproduce the crash. (also, that URL is taunting me)
https://paste.c-net.org/BuckleAttracts - this is the session immediately afterwards, with the lag issue.

Relevant configuration files:
Both of these services launch after a 5 second timer.

[Unit]
Description=Set 5800X3D power limit

[Service]
Type=oneshot
ExecStart=python3 /usr/local/bin/ruv.py -c 8 -o -30

[Unit]
Description=Set NVIDIA power limit

[Service]
Type=oneshot
# ExecStart=/usr/bin/nvidia-smi -pl 405
ExecStart=/usr/bin/nvidia-smi -pl 450

I've tried everything here: https://wiki.archlinux.org/title/Ryzen# … k_freezing
Aside from the boot parameters, which you can see in the logs, I've done some other things:
1. I updated my BIOS. I was on F15, and it turns out there's been a bunch of improvements since then specific to my CPU.
2. I set Power Supply Idle Control to Typical Current Idle as described here. This option wasn't there before the BIOS update.
3. Doing this made some error messages appear that didn't appear before: "GPU has fallen off the bus". A bit of searching suggested this was because my GPU wasn't getting enough power, suggesting the GPU undervolt was responsible. I set the script back to the default 450.

After doing this I played Palworld for about 5 hours without any issues. I went to bed satisfied that I'd fixed it, only for everything to stop working again. That was Friday night and I've been at this all weekend.

Thanks to the good old rubber duck method I've discovered that "GPU has fallen off the bus" was showing up when I launched Palworld for that journal I posted, just as I'm about to finish typing this and go to bed. Hopefully it will help me figure it out in the morning. If anyone can put a second pair of eyes on this I'd be grateful.

mmy8x · Yesterday 04:40:32

ninjachicken1 wrote:

I've been experiencing the following error on my desktop:
Sep 20 22:17:47 archlinux kernel: x86/amd: Previous system reset reason [0x08000800]: an uncorrected error caused a data fabric sync flood event
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: System Fatal error.
Sep 20 22:17:47 archlinux kernel: fbcon: Taking over console
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: CPU:12 (19:21:2) MC5_STATUS[-|UE|MiscV|AddrV|PCC|TCC|SyndV|-|-|-]: 0xbea0000000000108
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: Error Addr: 0x01ffffffc2188462
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: IPID: 0x000500b000000000, Syndrome: 0x000000004d000000
Sep 20 22:17:47 archlinux kernel: 
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: Execution Unit Ext. Error Code: 0
Sep 20 22:17:47 archlinux kernel: [Hardware Error]: cache level: RESV, tx: GEN, mem-tx: GEN
It can happen randomly, sometimes within 10 minutes of turning my computer on in the morning, sometimes after a few hours. It also happens consistently after launching Palworld (fairly graphically intense game) for a few minutes.
After this happens, regardless of whether it's random or because I attempted to play Palworld, it takes a few seconds to load kernel modules on boot and 13 seconds into that the GPU is initialized and my desktop becomes extremely laggy until I turn it off and leave it a few minutes.

After what happens? My guess is that the machine reboots, then prints this error and becomes sluggish until power cycled?

Random idea: add usbcore.autosuspend=-1 boot parameter and see if it helps. There is a known HW/FW bug in AMD USB controllers which causes machines to freeze or reboot randomly.

seth · Yesterday 06:21:39

Journal of the previous boot ends w/ the nvidia GPU falling off the bus.
The lane before it has an mt7xxx chip which is blamed as cause/trigger by eg. https://pcforum.amd.com/s/question/0D5P … se-further and those chips have a track record for causing bus errors (typically resulting in themselves not showing up)

=> can you live w/o the wifi for a while?

That being said and wrt to #2

Sep 20 22:12:17 MetalGear (udev-worker)[527]: 6:0:0:0: sd-device-monitor(worker): Passed 313 byte to netlink monitor.
Sep 20 22:12:17 MetalGear (udev-worker)[527]: Notify message sent to '@10353203418429332851': "PROCESSED=1"
Sep 20 22:12:17 MetalGear (udev-worker)[536]: 3-3:1.2: sd-device: Created database file '/run/udev/data/+usb:3-3:1.2' for '/devices/pci0000:00/0000:00:08.1/0000:08:00.3/usb3/3-3/3-3:1.2'.
Sep 20 22:12:17 MetalGear (udev-worker)[536]: 3-3:1.2: Running built-in command "kmod load"
Sep 20 22:12:17 MetalGear (udev-worker)[536]: Loading module: usb:v045Ep02EAd0301dcFFdsc47dpD0icFFisc47ipD0in02
Sep 20 22:12:17 MetalGear (udev-worker)[551]: 3-2.3: /usr/lib/udev/rules.d/69-libmtp.rules:38 PROGRAM="/usr/lib/udev/mtp-probe /sys$env{DEVPATH} $attr{busnum} $attr{devnum}": Running command "/usr/lib/udev/mtp-probe /sys/devices/pci0000:00/0000:00:08.1/0000:08:00.3/usb3/3-2/3-2.3 3 7"
Sep 20 22:12:17 MetalGear (udev-worker)[551]: 3-2.3: Found callout binary: "/usr/lib/udev/mtp-probe".
Sep 20 22:12:17 MetalGear (udev-worker)[551]: 3-2.3: Starting '/usr/lib/udev/mtp-probe /sys/devices/pci0000:00/0000:00:08.1/0000:08:00.3/usb3/3-2/3-2.3 3 7'
Sep 20 22:12:17 MetalGear (udev-worker)[566]: hidraw6: /usr/lib/udev/rules.d/60-persistent-hidraw.rules:7 IMPORT{builtin}="usb_id": Importing properties from results of builtin command "usb_id".
Sep 20 22:12:17 MetalGear (udev-worker)[548]: hiddev4: Preserve permissions of /dev/usb/hiddev4, uid=0, gid=0, mode=0600
Sep 20 22:12:17 MetalGear (udev-worker)[548]: hiddev4: Successfully created symlink '/dev/char/180:100' to '/dev/usb/hiddev4'
Sep 20 22:12:17 MetalGear (udev-worker)[543]: hiddev1: Preserve permissions of /dev/usb/hiddev1, uid=0, gid=0, mode=0600
Sep 20 22:12:17 MetalGear (udev-worker)[548]: hiddev4: sd-device: Created database file '/run/udev/data/c180:100' for '/devices/pci0000:00/0000:00:08.1/0000:08:00.3/usb3/3-4/3-4:1.2/usbmisc/hiddev4'.
Sep 20 22:12:17 MetalGear (udev-worker)[543]: hiddev1: Successfully created symlink '/dev/char/180:97' to '/dev/usb/hiddev1'
Sep 20 22:12:17 MetalGear (udev-worker)[548]: hiddev4: sd-device: Created database file '/run/udev/data/c180:100' for '/devices/pci0000:00/0000:00:08.1/0000:08:00.3/usb3/3-4/3-4:1.2/usbmisc/hiddev4'.
Sep 20 22:12:17 MetalGear kernel: ccp 0000:08:00.1: could not enable MSI-X (-22), trying MSI
Sep 20 22:12:17 MetalGear (udev-worker)[548]: hiddev4: Device processed (SEQNUM=3782, ACTION=add, UUID=256a2e88-9a2c-48d7-9d46-453da502eeca)
Sep 20 22:12:17 MetalGear kernel: ccp 0000:08:00.1: could not enable MSI (-22)
Sep 20 22:12:17 MetalGear (udev-worker)[548]: hiddev4: sd-device-monitor(worker): Passed 310 byte to netlink monitor.
Sep 20 22:12:17 MetalGear kernel: ccp 0000:08:00.1: initialization failed
Sep 20 22:12:17 MetalGear (udev-worker)[543]: hiddev1: sd-device: Created database file '/run/udev/data/c180:97' for '/devices/pci0000:00/0000:00:08.1/0000:08:00.3/usb3/3-1/3-1:1.1/usbmisc/hiddev1'.
Sep 20 22:12:17 MetalGear kernel: ccp 0000:08:00.1: probe with driver ccp failed with error -22

08:00 is one of the USBs

Sidebar: please don't use systemd.log_level=debug unless specifically asked to.
It turns the noise to signal ratio into the absurd territory.

ninjachicken1 · Yesterday 14:26:39

mmy8x wrote:

After what happens? My guess is that the machine reboots, then prints this error and becomes sluggish until power cycled?

You're mostly right - usually it's until I let it boot into a sluggish session, then power it off for a few minutes, and also boot into UEFI and then exit without saving for some reason.

Unfortunatley today, I had a crash happen a few minutes in (when it does crash a few minutes in, it's usually first thing in the morning after turning my computer off for the night, maybe that's a clue). That was 4 hours ago and after trying everything I usually do (including leaving my pc off for 30 minutes multiple times), I'm still lagging.

One thing I've noticed today is that my NVME temp is fluctuating wildly, and my GPU has show bursts of high VRAM usage with no apparent temperature changes.

mmy8x wrote:

Random idea: add usbcore.autosuspend=-1 boot parameter and see if it helps. There is a known HW/FW bug in AMD USB controllers which causes machines to freeze or reboot randomly.

I've tried this and my next boot went slightly smoother until I launched my browser. It might be a placebo though.

seth wrote:

can you live w/o the wifi for a while?

Yes. Could you tell me how to disable it? I'm struggling to even use Duckduckgo at the moment.

seth wrote:

Sidebar: please don't use systemd.log_level=debug unless specifically asked to.
It turns the noise to signal ratio into the absurd territory.

Sorry, I've turned it of now. Unfortunately the morning crash happened as soon as I read that sentence and reached to my keyboard to do that, so If I upload that log it was before I disabled it.

seth · Yesterday 17:06:38

Yes. Could you tell me how to disable it?

BIOS/UEFI settings - anything in the OS won't help if it's actually poisoning the bus.
(The alternative would then involve a screwdriver)

ninjachicken1 · Yesterday 20:00:07

seth wrote:

Yes. Could you tell me how to disable it?
BIOS/UEFI settings - anything in the OS won't help if it's actually poisoning the bus.
(The alternative would then involve a screwdriver)

After checking, I don't thing there's an option for that in my BIOS, and the screwdriver alternative isn't an option for me.

Today's the worst this issue's ever been. I only got two sessions that weren't sluggish, and both crashed after a short period of time. I've rolled back to a snapshot from a couple of months ago - I'm now using the linux 6.7.8-arch1-1 kernel and nvidia-dkms 535.113.01-2 driver, and things are stable again. I'm going to start by trying to upgrade my nvidia drivers without updating the kernel and see if it breaks anything.

seth · Yesterday 20:19:37

If it's not the HW directly and we blame the wifi chip you can just "module_blacklist=mt7921e", https://wiki.archlinux.org/title/Kernel … and_line_2

Arch Linux

#1 Yesterday 00:14:10

[0x08000800]: data fabric sync flood event

#2 Yesterday 04:40:32

Re: [0x08000800]: data fabric sync flood event

#3 Yesterday 06:21:39

Re: [0x08000800]: data fabric sync flood event

#4 Yesterday 14:26:39

Re: [0x08000800]: data fabric sync flood event

#5 Yesterday 17:06:38

Re: [0x08000800]: data fabric sync flood event

#6 Yesterday 20:00:07

Re: [0x08000800]: data fabric sync flood event

#7 Yesterday 20:19:37

Re: [0x08000800]: data fabric sync flood event

Board footer