You are not logged in.
I got a crash after 1h of testing, only had a video player open.
6.13.2.arch1-7
Edit: the 1-6 test was stable for two days.
Last edited by Amunn (2025-02-14 23:31:16)
Offline
This build contains only mandatory patches that impact performance and stability
Build: linux-amdgpu-testing-6.13.2.arch1-8 - stable but worse performance compared to 1-6.
Included patches:
- Optimize mutex protected blocks in amdgpu_vm_flush
- Simplify GFXOFF handling
- Add PG state change for GFX9 Compute rings (in this patch gfx_v9_0_set_powergating_state() is called directly, without extra checks and loops)
- Remove workaround for TLB seq race
Build: linux-amdgpu-testing-6.13.2.arch1-9 - stable, but higher power consumption and worse performance compared to 1-6 and 1-8 builds.
Included patches:
- Optimize mutex protected blocks in amdgpu_vm_flush
- Add PG workaround for Raven Compute rings
- Refactor GFXOFF handler (another approach with latency optimizations)
laikm from pacoandres for reporting issues
Last edited by Mechanicus (2025-02-18 20:23:21)
Offline
Performance is slightly down again compared to 1-6 test. 1-8 is still stable after 5h of uptime.
When I ran the aquarium test, this showed up in the logs, but the system didn't crash:
Feb 15 18:11:47 user systemd-coredump[15032]: Process 15026 (spotify) of user 1000 terminated abnormally with signal 6/ABRT, processing...
Feb 15 18:11:47 user systemd[1]: Started Process Core Dump (PID 15032/UID 0).
Feb 15 18:11:47 user systemd-coredump[15041]: Process 15035 (spotify) of user 1000 terminated abnormally with signal 6/ABRT, processing...
Feb 15 18:11:47 user systemd[1]: Started Process Core Dump (PID 15041/UID 0).
Feb 15 18:11:47 user systemd-coredump[15050]: Process 15042 (spotify) of user 1000 terminated abnormally with signal 6/ABRT, processing...
Feb 15 18:11:47 user systemd[1]: Started Process Core Dump (PID 15050/UID 0).
Feb 15 18:11:47 user systemd-coredump[15060]: Process 15054 (spotify) of user 1000 terminated abnormally with signal 6/ABRT, processing...
Feb 15 18:11:47 user systemd[1]: Started Process Core Dump (PID 15060/UID 0).
Feb 15 18:11:47 user systemd-coredump[15069]: Process 15061 (spotify) of user 1000 terminated abnormally with signal 6/ABRT, processing...
Feb 15 18:11:47 user systemd[1]: Started Process Core Dump (PID 15069/UID 0).
Feb 15 18:11:47 user systemd-coredump[15079]: Process 15073 (spotify) of user 1000 terminated abnormally with signal 6/ABRT, processing...
Feb 15 18:11:47 user systemd[1]: Started Process Core Dump (PID 15079/UID 0).
Feb 15 18:11:47 user systemd-coredump[15088]: Process 15080 (spotify) of user 1000 terminated abnormally with signal 6/ABRT, processing...
Feb 15 18:11:47 user systemd[1]: Started Process Core Dump (PID 15088/UID 0).
Feb 15 18:11:47 user systemd-coredump[15098]: Process 15092 (spotify) of user 1000 terminated abnormally with signal 6/ABRT, processing...
Feb 15 18:11:47 user systemd[1]: Started Process Core Dump (PID 15098/UID 0).
Feb 15 18:11:47 user systemd-coredump[15106]: Process 15099 (spotify) of user 1000 terminated abnormally with signal 6/ABRT, processing...
Feb 15 18:11:47 user systemd[1]: systemd-coredump.socket: Too many incoming connections (9) from source UID 0, dropping connection.
Feb 15 18:11:48 user systemd-coredump[15052]: [?] Process 15042 (spotify) of user 1000 dumped core.
Stack trace of thread 15042:
#0 0x00007a921065a624 n/a (libc.so.6 + 0x97624)
#1 0x00007a9210600ba0 raise (libc.so.6 + 0x3dba0)
#2 0x00007a92105e8582 abort (libc.so.6 + 0x25582)
#3 0x00007a92105e93bf n/a (libc.so.6 + 0x263bf)
#4 0x00007a92106e8419 __fortify_fail (libc.so.6 + 0x125419)
#5 0x00007a92106e9714 __stack_chk_fail (libc.so.6 + 0x126714)
#6 0x00005ed0b8657d70 n/a (n/a + 0x0)
ELF object binary architecture: AMD x86-64
Feb 15 18:11:48 user systemd[1]: systemd-coredump@3-15050-0.service: Deactivated successfully.
Feb 15 18:11:48 user systemd[1]: systemd-coredump@3-15050-0.service: Consumed 421ms CPU time, 139.1M memory peak.
Feb 15 18:11:48 user systemd-coredump[15044]: [?] Process 15035 (spotify) of user 1000 dumped core.
Stack trace of thread 15035:
#0 0x00007a921065a624 n/a (n/a + 0x0)
#1 0x00007a9210600ba0 n/a (n/a + 0x0)
#2 0x00007a92105e8582 n/a (n/a + 0x0)
#3 0x00007a92105e93bf n/a (n/a + 0x0)
#4 0x00007a92106e8419 n/a (n/a + 0x0)
#5 0x00007a92106e9714 n/a (n/a + 0x0)
#6 0x00005ed0b8657d70 n/a (n/a + 0x0)
ELF object binary architecture: AMD x86-64
Feb 15 18:11:48 user systemd-coredump[15067]: [?] Process 15054 (spotify) of user 1000 dumped core.
Stack trace of thread 15054:
#0 0x00007a921065a624 n/a (n/a + 0x0)
#1 0x00007a9210600ba0 n/a (n/a + 0x0)
#2 0x00007a92105e8582 n/a (n/a + 0x0)
#3 0x00007a92105e93bf n/a (n/a + 0x0)
#4 0x00007a92106e8419 n/a (n/a + 0x0)
#5 0x00007a92106e9714 n/a (n/a + 0x0)
#6 0x00005ed0b8657d70 n/a (n/a + 0x0)
ELF object binary architecture: AMD x86-64
Feb 15 18:11:48 user systemd[1]: systemd-coredump@2-15041-0.service: Deactivated successfully.
Feb 15 18:11:48 user systemd[1]: systemd-coredump@2-15041-0.service: Consumed 415ms CPU time, 132.5M memory peak.
Feb 15 18:11:48 user systemd[1]: systemd-coredump@4-15060-0.service: Deactivated successfully.
Feb 15 18:11:48 user systemd[1]: systemd-coredump@4-15060-0.service: Consumed 385ms CPU time, 130.8M memory peak.
Feb 15 18:11:48 user (sd-parse-elf)[15123]: Could not parse core file, dwfl_core_file_report() failed: (null)
Feb 15 18:11:48 user (sd-parse-elf)[15123]: Failed to inspect core file: Invalid argument
Feb 15 18:11:48 user systemd-coredump[15033]: [?] Process 15026 (spotify) of user 1000 dumped core.
Feb 15 18:11:48 user systemd[1]: systemd-coredump@1-15032-0.service: Deactivated successfully.
Feb 15 18:11:48 user systemd[1]: systemd-coredump@1-15032-0.service: Consumed 415ms CPU time, 130.8M memory peak.
Feb 15 18:11:48 user systemd-coredump[15071]: [?] Process 15061 (spotify) of user 1000 dumped core.
Stack trace of thread 15061:
#0 0x00007a921065a624 n/a (n/a + 0x0)
#1 0x00007a9210600ba0 n/a (n/a + 0x0)
#2 0x00007a92105e8582 n/a (n/a + 0x0)
#3 0x00007a92105e93bf n/a (n/a + 0x0)
#4 0x00007a92106e8419 n/a (n/a + 0x0)
#5 0x00007a92106e9714 n/a (n/a + 0x0)
#6 0x00005ed0b8657d70 n/a (n/a + 0x0)
ELF object binary architecture: AMD x86-64
Feb 15 18:11:48 user systemd[1]: systemd-coredump@5-15069-0.service: Deactivated successfully.
Feb 15 18:11:48 user systemd[1]: systemd-coredump@5-15069-0.service: Consumed 400ms CPU time, 139M memory peak.
Feb 15 18:11:48 user systemd-coredump[15087]: [?] Process 15073 (spotify) of user 1000 dumped core.
Stack trace of thread 15073:
#0 0x00007a921065a624 n/a (n/a + 0x0)
#1 0x00007a9210600ba0 n/a (n/a + 0x0)
#2 0x00007a92105e8582 n/a (n/a + 0x0)
#3 0x00007a92105e93bf n/a (n/a + 0x0)
#4 0x00007a92106e8419 n/a (n/a + 0x0)
#5 0x00007a92106e9714 n/a (n/a + 0x0)
#6 0x00005ed0b8657d70 n/a (n/a + 0x0)
ELF object binary architecture: AMD x86-64
Feb 15 18:11:48 user systemd[1]: systemd-coredump@6-15079-0.service: Deactivated successfully.
Feb 15 18:11:48 user systemd[1]: systemd-coredump@6-15079-0.service: Consumed 385ms CPU time, 130.9M memory peak.
Feb 15 18:11:48 user systemd-coredump[15090]: [?] Process 15080 (spotify) of user 1000 dumped core.
Stack trace of thread 15080:
#0 0x00007a921065a624 n/a (n/a + 0x0)
#1 0x00007a9210600ba0 n/a (n/a + 0x0)
#2 0x00007a92105e8582 n/a (n/a + 0x0)
#3 0x00007a92105e93bf n/a (n/a + 0x0)
#4 0x00007a92106e8419 n/a (n/a + 0x0)
#5 0x00007a92106e9714 n/a (n/a + 0x0)
#6 0x00005ed0b8657d70 n/a (n/a + 0x0)
ELF object binary architecture: AMD x86-64
Feb 15 18:11:48 user systemd[1]: systemd-coredump@7-15088-0.service: Deactivated successfully.
Feb 15 18:11:48 user systemd[1]: systemd-coredump@7-15088-0.service: Consumed 379ms CPU time, 139.1M memory peak.
Feb 15 18:11:48 user systemd-coredump[15105]: [?] Process 15092 (spotify) of user 1000 dumped core.
Stack trace of thread 15092:
#0 0x00007a921065a624 n/a (n/a + 0x0)
#1 0x00007a9210600ba0 n/a (n/a + 0x0)
#2 0x00007a92105e8582 n/a (n/a + 0x0)
#3 0x00007a92105e93bf n/a (n/a + 0x0)
#4 0x00007a92106e8419 n/a (n/a + 0x0)
#5 0x00007a92106e9714 n/a (n/a + 0x0)
#6 0x00005ed0b8657d70 n/a (n/a + 0x0)
ELF object binary architecture: AMD x86-64
Feb 15 18:11:48 user systemd[1]: systemd-coredump@8-15098-0.service: Deactivated successfully.
Feb 15 18:11:48 user systemd[1]: systemd-coredump@8-15098-0.service: Consumed 343ms CPU time, 130.6M memory peak.
Feb 15 18:12:46 user rtkit-daemon[854]: Supervising 11 threads of 6 processes of 1 users.
Feb 15 18:12:46 user rtkit-daemon[854]: Supervising 11 threads of 6 processes of 1 users.
Feb 15 18:13:42 user sudo[15280]: user : TTY=pts/0 ; PWD=/home/user ; USER=root ; COMMAND=/usr/bin/turbostat -s PkgWatt
Feb 15 18:13:42 user sudo[15280]: pam_unix(sudo:session): session opened for user root(uid=0) by user(uid=1000)
Feb 15 18:14:12 user sudo[15280]: pam_unix(sudo:session): session closed for user root
Feb 15 18:18:55 user foot[719]: quirks: applying wl_surface_damage_buffer() workaround for Sway
Offline
@Amunn: Please check back, another release was edited into the post above.
Edit 1: I've been having a stable system, including graphics reset and S3 Suspend, using linux-amdgpu-testing-6.13.1.arch1-18 patches, but applied to linux-zen-6.13.2-zen1-1-zen. Rock solid.
I'm now spinning linux-zen-6.13.2-zen1-1-zen with your linux-amdgpu-testing-6.13.2.arch1-9 applied, and will test those. If that fails, I'll switch back to the previous kernel and wait for further patches.
Edit 2: I tested linux-amdgpu-testing-6.13.2.arch1-9. My machine was booted with the iGPU enabled, also AMD Radeon. It was booted completely headless, and no displays came up when hotplugging DisplayPort. This is contrary to the linux-amdgpu-testing 6.13.1.arch1-18 patch series.
Last edited by kode54 (2025-02-16 07:21:19)
Offline
@kode54 Ok, on 1-9 now. Thanks for writing a new post in this thread, I would have missed it.
Last edited by Amunn (2025-02-16 09:12:57)
Offline
novice here on Manjaro, KDE/Plasma
tested Lone_Wolf's mesa 25.0.0 rc3 driver and it wasn't but a few minutes until the system logged me off (this was a first - usually the desktop freezes, though earlier today it did auto-reboot which was another first)
to install the new driver, i just downloaded the tar.zst file > right clicked > "open with software installed" > reboot
confirmed the driver installed with ...
$ vulkaninfo --summary
GPU0:
[...]
driverInfo = Mesa 25.0.0-rc3 (git-3a8abfa39b) (LLVM 19.1.7)
Operating System: Manjaro Linux
KDE Plasma Version: 6.2.5
KDE Frameworks Version: 6.10.0
Qt Version: 6.8.2
Kernel Version: 6.12.12-2-MANJARO (64-bit)
Graphics Platform: X11
Processors: 8 × AMD Ryzen 5 3400G with Radeon Vega Graphics
Memory: 13.6 GiB of RAM
Graphics Processor: AMD Radeon Vega 11 Graphics
Manufacturer: Micro-Star International Co., Ltd
Product Name: MS-7C02
System Version: 1.0
Offline
The issue with Manjaro is that Manjaro has different repos and Mesa needs to be built towards the right llvm version, i don't know if Manjaro is using the same llvm version of Arch.
Offline
the 'llvm' package is not installed on my system, but Manjaro llvm-libs version is 19.1.7-1 and it appears the Arch version in the extra repo is the same (https://geo.mirror.pkgbuild.com/extra/os/x86_64/)
Offline
@fixitharder
Manjaro afaik has released an update today, which apparently fixes the last huge upgrade before in some places.
Have you already installed this update?
As far as I know the Manjaro users use mesa (25.x) from the AUR to fix the problem.
I still have Manjaro on another disc, but I no longer use it. Tonight I could update the system and install mesa test-git 25.0.0 rc3 there and see if I can reproduce your result.
But first you should install the Manjaro upgrade from today.
Offline
@orbit-oc - yes, i did install the 16/feb stable update and am pretty sure the system logged me off shortly thereafter - also had a very weird issue while responding to your post where the text editor became unresponsive, but the rest of the system *seemed* to be responsive - still wasn't able to do a graceful restart however
looking at the update log for Manjaro 16/feb stable (https://gitlab.manjaro.org/-/snippets/1088/raw), mesa isn't in there
i do see a mesa-git 25.0.0_devel.200442.bfa6b9b6551.d41d8cd in the AUR, but it's a bit older than the rc3 version Lone_Wolf posted - if anyone has any feedback on this, i'm all ears
Last edited by fixitharder (2025-02-16 15:19:32)
Offline
First impressions about linux-amdgpu-testing-6.13.2.arch1-9 here
Offline
i do see a mesa-git 25.0.0_devel.200442.bfa6b9b6551.d41d8cd in the AUR, but it's a bit older than the rc3 version Lone_Wolf posted - if anyone has any feedback on this, i'm all ears
@fixitharder
see @odie @mufflon etc...
https://forum.manjaro.org/t/stable-upda … /173772/91
Offline
This build supposed to be stable. The goal is to make it high efficient as well.
Build: linux-amdgpu-testing-6.13.2.arch1-10, linux-amdgpu-testing-headers-6.13.2.arch1-10 - stable
Included patches:
- Optimize mutex protected blocks in amdgpu_vm_flush
- Add PG workaround for Raven Compute rings
- Refactor GFXOFF handler v2 (low latency approach using atomics)
- Use gfx_off_ctrl_immediate for GFX9/10/11/12
- Remove workaround for Raven in amdgpu_dpm
- Remove workaround for TLB seq race
Kernel option to keep during testing period: fsck.mode=force
laikm from pacoandres for reporting issues
Last edited by Mechanicus (2025-02-20 19:54:30)
Offline
see #556 - #560, #562
I still have Manjaro on another disc, but I no longer use it. Tonight I could update the system and install mesa test-git 25.0.0 rc3 there and see if I can reproduce your result.
@fixitharder
I have connected the drive and updated Manjaro to the current version of 25.0.0 “Zetar”. After the reboot I installed 'mesa-test-git 25.0.0_rc3.201182.3a8abfa39b7-1' and removed 'mesa' and 'vulkan-radeon'.
Another reboot shows that this mesa-test-git runs on the current Manjaro. I am now writing this Post with this system and will then switch back to my original system.
Offline
@orbit-oc - thanks! - i downgraded mesa to 1:24.2.7-1 (24.2.8 wasn't in the repo?) and llvm-libs to 18.1.8-4 ... we'll see what happens
just a note for others looking for solutions (i didn't read every post here)...
* my understanding is that this is an AMD Ryzen issue, possibly limited to older APUs (integrated graphics)
* OpenSUSE Tumbleweed may be experiencing the same problem
* EndevourOS is of course having the same problem (personally tested)
* the mesa 25.0.0 rc3 driver linked by Lone_Wolf didn't work for me (https://bbs.archlinux.org/viewtopic.php … 0#p2223890)
* kernel versions 6.1, 6.6 and 6.12 didn't solve the problem
* wayland and x11 are both affected
Offline
@orbit-oc - oh, you didn't have to do that - i already tested mesa-test-git 25.0.0_rc3 (https://bbs.archlinux.org/viewtopic.php … 9#p2226739)
didn't work for me, but i expect the downgrade will (mesa 1:24.2.7-1, llvm-libs 18.1.8-4)
Offline
Tested 1-9 kenel, stable, performance increased for me, same as the 6.13.2.arch1-6 test. Interestingly, power consumption did not increase.
Switching to 6.13.2.arch1-10. (updated firmware and ucode, I forgot to pacman -Syu before test 1-9)
Last edited by Amunn (2025-02-17 06:47:52)
Offline
First results with linux-amdgpu-testing-6.13.2-arch1-10 here: https://github.com/pacoandres/laikm/iss … 2662382633
Offline
Idle is slightly improved at 2w, 30k fish @ 27 fps gives 12.5w. Uptime, 2h.
Offline
On my Athlon 200GE the new lock-free approach for GFXOFF handling showed 26.5% performance boost compared to default and linux-amdgpu-testing-6.13.2-arch1-6 kernels. Tested with new linux-firmware: 20250210.5bc5868b-1, which reduced performance drastically compared to previous one.
Last edited by Mechanicus (2025-02-17 09:39:10)
Offline
still freezing
mesa 24.2.7-1
lvm-libs 18.1.8-4
kernel 6.1.127-2
Offline
@fixitharder
That's exactly why it was important for me to reproduce your problem with mesa-test-git 25.0 rc3, as other Manjaro users may also be reading here. See post #564.
So you probably have another problem besides the problem with mesa 24.3.x ...
Offline
@orbit-oc - sorry, i'm not understanding - in #564 you said you loaded mesa-test-git, but no mention of whether the system froze or not
were you able to reproduce the problem, or do you think mesa-test-git 25.0.0_rc3 is the solution?
and yes, i think i do have other problems (https://forum.manjaro.org/t/constant-sy … /174105/36)
Offline
EDIT: I removed this post to avoid any confusion.
Last edited by NuSkool (2025-02-17 19:38:36)
Scripts I use: https://github.com/Cody-Learner
$ glxinfo | grep Device Device: AMD Radeon Vega 11 Graphics (radeonsi, raven, LLVM 19.1.7, DRM 3.60, 6.13.3-arch1-1.1) (0x15dd)
$ sudo dmesg | awk '/drm/ && /gfx/' [ 6.427009] [drm] add ip block number 6 <gfx_v9_0>
Offline
@NuSkool - thanks for the tip, but where is the mesa-test-git-25.0.0_devel.200908.66775c89fce-1-x86_64.pkg.tar.zst file? Lone-Wolf's post you linked to doesn't link that file
Offline