You are not logged in.
Pages: 1
Hi!
A whopping ~8G RAM are gone right after boot and before X11/KDE without anything special running. 6.6.35-2 (and prior), amdgpu.
Kernel options I've tried so far: systemd.unit=rescue.target, systemd.zram=0, amdgpu.gttsize=512, amdgpu.gartsize=128. All pretty much to no avail, making only marginal differences at best.
These command where executed in rescue.target:
# free -h
Mem: 62Gi 7.4Gi 57Gi 10Mi 415Mi 55Gi
Swap: 62Gi 0B 62Gi
MemTotal: 65785924 kB
# smem -tk
PID User Command Swap USS PSS RSS
1293 root /usr/lib/systemd/systemd-su 0 816.0K 1.5M 5.1M
1271 systemd-timesync /usr/lib/systemd/systemd-ti 0 1.6M 2.8M 7.4M
1419 root (udev-worker) 0 624.0K 2.9M 7.0M
1364 root ssh-agent 0 2.8M 2.9M 3.8M
711 root /usr/lib/systemd/systemd-ud 0 696.0K 3.9M 10.8M
1311 root bash 0 4.0M 4.5M 6.2M
635 root /usr/lib/systemd/systemd-jo 0 5.6M 6.8M 11.5M
1 root /sbin/init 0 6.4M 8.1M 13.4M
1420 root python /usr/bin/smem -tk 0 13.1M 13.7M 15.6M
-------------------------------------------------------------------------------
9 2 0 35.7M 47.1M 80.9M
MemTotal: 65785924 kB
MemFree: 59814232 kB
MemAvailable: 57983032 kB
Buffers: 3080 kB
Cached: 381892 kB
SwapCached: 0 kB
Active: 71196 kB
Inactive: 327788 kB
Active(anon): 24736 kB
Inactive(anon): 0 kB
Active(file): 46460 kB
Inactive(file): 327788 kB
Unevictable: 8 kB
Mlocked: 0 kB
SwapTotal: 65785852 kB
SwapFree: 65785852 kB
Zswap: 0 kB
Zswapped: 0 kB
Dirty: 7328 kB
Writeback: 0 kB
AnonPages: 14072 kB
Mapped: 20872 kB
Shmem: 10712 kB
KReclaimable: 47800 kB
Slab: 259660 kB
SReclaimable: 47800 kB
SUnreclaim: 211860 kB
KernelStack: 10640 kB
PageTables: 572 kB
SecPageTables: 0 kB
NFS_Unstable: 0 kB
Bounce: 0 kB
WritebackTmp: 0 kB
CommitLimit: 96581660 kB
Committed_AS: 39452 kB
VmallocTotal: 34359738367 kB
VmallocUsed: 601580 kB
VmallocChunk: 0 kB
Percpu: 112128 kB
HardwareCorrupted: 0 kB
AnonHugePages: 2048 kB
ShmemHugePages: 0 kB
ShmemPmdMapped: 0 kB
FileHugePages: 0 kB
FilePmdMapped: 0 kB
CmaTotal: 0 kB
CmaFree: 0 kB
Unaccepted: 0 kB
HugePages_Total: 2048
HugePages_Free: 2048
HugePages_Rsvd: 0
HugePages_Surp: 0
Hugepagesize: 2048 kB
Hugetlb: 4194304 kB
DirectMap4k: 332492 kB
DirectMap2M: 10024960 kB
DirectMap1G: 56623104 kB
Any advise how to proceed further?
RAM consumption on a Manjaro USB is 2.5G on the same machine.
thx
tcn
Last edited by tcn (2024-06-27 08:06:14)
Offline
Integrated GPU that reserves lots of system ram ?
Disliking systemd intensely, but not satisfied with alternatives so focusing on taming systemd.
clean chroot building not flexible enough ?
Try clean chroot manager by graysky
Offline
please post your hardware specs (board, cpu, (igpu), gpu, ram)
Offline
ASUS Zenith Extreme, Threadripper 1950x, 64GB RAM, Radeon RX 570 4GB (which chromium fails to use for video decoding up to this day...)
Offline
Integrated GPU that reserves lots of system ram ?
Please post your complete system journal for the boot:
sudo journalctl -b | curl -F 'file=@-' 0x0.st
Offline
Offline
amdgpu grants itself half:
Jun 27 07:23:05 box kernel: [drm] amdgpu: 4096M of VRAM memory ready
Jun 27 07:23:05 box kernel: [drm] amdgpu: 32122M of GTT memory ready.
And zram is in full force:
Jun 27 07:23:06 box kernel: Adding 65785852k swap on /dev/zram0. Priority:100 extents:1 across:65785852k SSDsc
And there's oc a lot of stuff running, 8GB isn't implausible in this situation at all.
This here is whacko:
MemTotal: 65785924 kB
MemFree: 59814232 kB
MemAvailable: 57983032 kB
you've more free than available ram…
What does /proc/meminfo here actually look like?
systemd.unit=rescue.target, systemd.zram=0, amdgpu.gttsize=512, amdgpu.gartsize=128
Offline
And there's oc a lot of stuff running
There is? Where? I thought I had disabled it all a long time ago
What does /proc/meminfo here actually look like?
systemd.unit=rescue.target, systemd.zram=0, amdgpu.gttsize=512, amdgpu.gartsize=128
There you go:
Jun 27 17:24:06 box kernel: Linux version 6.6.35-2-lts (linux-lts@archlinux) (gcc (GCC) 14.1.1 20240522, GNU ld (GNU Binutils) 2.42.0) #1 SMP PREEMPT_DYNAMIC Fri, 21 Jun 2024 21:05:23 +0000
Jun 27 17:24:06 box kernel: Command line: initrd=\amd-ucode.img initrd=\initramfs-linux-lts.img root=PARTUUID=f100f562-2098-4949-b9d1-4ab250db7653 rootfstype=btrfs rw add_efi_memmap audit=0 radeon.modeset=1 pcie_aspm=off systemd.zram=0 systemd.unit=rescue.target amdgpu.gttsize=512 amdgpu.gartsize=128
MemTotal: 65785932 kB
MemFree: 60253320 kB
MemAvailable: 58385056 kB
Buffers: 3080 kB
Cached: 309512 kB
SwapCached: 0 kB
Active: 69956 kB
Inactive: 256724 kB
Active(anon): 24600 kB
Inactive(anon): 0 kB
Active(file): 45356 kB
Inactive(file): 256724 kB
Unevictable: 0 kB
Mlocked: 0 kB
SwapTotal: 0 kB
SwapFree: 0 kB
Zswap: 0 kB
Zswapped: 0 kB
Dirty: 2016 kB
Writeback: 0 kB
AnonPages: 13580 kB
Mapped: 20648 kB
Shmem: 10512 kB
KReclaimable: 46208 kB
Slab: 251720 kB
SReclaimable: 46208 kB
SUnreclaim: 205512 kB
KernelStack: 10640 kB
PageTables: 516 kB
SecPageTables: 0 kB
NFS_Unstable: 0 kB
Bounce: 0 kB
WritebackTmp: 0 kB
CommitLimit: 30795812 kB
Committed_AS: 39084 kB
VmallocTotal: 34359738367 kB
VmallocUsed: 284076 kB
VmallocChunk: 0 kB
Percpu: 110080 kB
HardwareCorrupted: 0 kB
AnonHugePages: 0 kB
ShmemHugePages: 0 kB
ShmemPmdMapped: 0 kB
FileHugePages: 0 kB
FilePmdMapped: 0 kB
CmaTotal: 0 kB
CmaFree: 0 kB
Unaccepted: 0 kB
HugePages_Total: 2048
HugePages_Free: 2048
HugePages_Rsvd: 0
HugePages_Surp: 0
Hugepagesize: 2048 kB
Hugetlb: 4194304 kB
DirectMap4k: 262860 kB
DirectMap2M: 9046016 kB
DirectMap1G: 57671680 kB
It's still over 7GB. While booting an out-of-the-box Manjaro live USB on the same system with the same open-source amdgpu drivers gets me down to 2.5. And I think that's even after KDE...
Offline
You're actually just using shy above 300MB (the brunt of that being the file cache) - the RAM is most likely going to the GPU - "somehow".
Try to add "transparent_hugepage=never" to the kernel parameters.
Offline
Makes no difference.
# cat /sys/kernel/mm/transparent_hugepage/enabled
always madvise [never]
MemTotal: 65785932 kB
MemFree: 59191596 kB
MemAvailable: 57469056 kB
Buffers: 3080 kB
Cached: 586652 kB
SwapCached: 0 kB
Active: 326836 kB
Inactive: 334800 kB
Active(anon): 89648 kB
Inactive(anon): 0 kB
Active(file): 237188 kB
Inactive(file): 334800 kB
Unevictable: 1872 kB
Mlocked: 1872 kB
SwapTotal: 65785852 kB
SwapFree: 65785852 kB
Zswap: 0 kB
Zswapped: 0 kB
Dirty: 88 kB
Writeback: 0 kB
AnonPages: 74036 kB
Mapped: 92348 kB
Shmem: 15860 kB
KReclaimable: 67744 kB
Slab: 308272 kB
SReclaimable: 67744 kB
SUnreclaim: 240528 kB
KernelStack: 12016 kB
PageTables: 3108 kB
SecPageTables: 0 kB
NFS_Unstable: 0 kB
Bounce: 0 kB
WritebackTmp: 0 kB
CommitLimit: 96581664 kB
Committed_AS: 689832 kB
VmallocTotal: 34359738367 kB
VmallocUsed: 610724 kB
VmallocChunk: 0 kB
Percpu: 131584 kB
HardwareCorrupted: 0 kB
AnonHugePages: 0 kB
ShmemHugePages: 0 kB
ShmemPmdMapped: 0 kB
FileHugePages: 0 kB
FilePmdMapped: 0 kB
CmaTotal: 0 kB
CmaFree: 0 kB
Unaccepted: 0 kB
HugePages_Total: 2048
HugePages_Free: 2048
HugePages_Rsvd: 0
HugePages_Surp: 0
Hugepagesize: 2048 kB
Hugetlb: 4194304 kB
DirectMap4k: 379596 kB
DirectMap2M: 11026432 kB
DirectMap1G: 55574528 kB
Offline
Just to isolate the cause, same when booting "nomodeset" or outright blacklisting amdgpu?
Offline
Again, no effect (nomodeset).
# lsmod | grep -c amdgpu
0
MemTotal: 65785932 kB
MemFree: 59076428 kB
MemAvailable: 57399576 kB
Buffers: 2996 kB
Cached: 675740 kB
SwapCached: 0 kB
Active: 330680 kB
Inactive: 431732 kB
Active(anon): 99232 kB
Inactive(anon): 0 kB
Active(file): 231448 kB
Inactive(file): 431732 kB
Unevictable: 4864 kB
Mlocked: 1792 kB
SwapTotal: 65785852 kB
SwapFree: 65785852 kB
Zswap: 0 kB
Zswapped: 0 kB
Dirty: 384 kB
Writeback: 0 kB
AnonPages: 85916 kB
Mapped: 87368 kB
Shmem: 13872 kB
KReclaimable: 67924 kB
Slab: 296644 kB
SReclaimable: 67924 kB
SUnreclaim: 228720 kB
KernelStack: 11216 kB
PageTables: 3044 kB
SecPageTables: 0 kB
NFS_Unstable: 0 kB
Bounce: 0 kB
WritebackTmp: 0 kB
CommitLimit: 96581664 kB
Committed_AS: 679604 kB
VmallocTotal: 34359738367 kB
VmallocUsed: 556156 kB
VmallocChunk: 0 kB
Percpu: 131584 kB
HardwareCorrupted: 0 kB
AnonHugePages: 10240 kB
ShmemHugePages: 0 kB
ShmemPmdMapped: 0 kB
FileHugePages: 12288 kB
FilePmdMapped: 2048 kB
CmaTotal: 0 kB
CmaFree: 0 kB
Unaccepted: 0 kB
HugePages_Total: 2048
HugePages_Free: 2048
HugePages_Rsvd: 0
HugePages_Surp: 0
Hugepagesize: 2048 kB
Hugetlb: 4194304 kB
DirectMap4k: 377548 kB
DirectMap2M: 11028480 kB
DirectMap1G: 55574528 kB
Offline
Ok, wIld guess: spurious tmpfs usage?
df -h
Offline
Filesystem Size Used Avail Use% Mounted on
dev 32G 0 32G 0% /dev
run 32G 11M 32G 1% /run
efivarfs 128K 61K 63K 50% /sys/firmware/efi/efivars
/dev/nvme1n1p2 477G 231G 207G 53% /
tmpfs 32G 3,7G 28G 12% /dev/shm
/dev/nvme1n1p2 477G 231G 207G 53% /home
/dev/nvme1n1p2 477G 231G 207G 53% /var/cache
/dev/nvme1n1p2 477G 231G 207G 53% /var/lib/postgres
/dev/nvme1n1p2 477G 231G 207G 53% /var/log
/dev/nvme1n1p1 511M 373M 139M 73% /boot
/dev/nvme1n1p2 477G 231G 207G 53% /var/tmp
/dev/nvme0n1p1 477G 427G 50G 90% /mnt/970pro
tmpfs 2,0G 368M 1,7G 18% /tmp
tmpfs 100M 0 100M 0% /var/cache/tcn/chromium
tmpfs 100M 0 100M 0% /var/cache/tcn/vivaldi
tmpfs 100M 0 100M 0% /var/cache/tcn/mozilla
tmpfs 6,3G 172K 6,3G 1% /run/user/1024
qp:/homes/tcn 3,0T 1,9T 1,2T 63% /home/qnap/tcn
qp:/Multimedia/photo 3,0T 1,9T 1,2T 63% /home/qnap/photo
Offline
tmpfs 32G 3,7G 28G 12% /dev/shm
that's usually zero if the system is not doing much.
/dev/nvme1n1p2 477G 231G 207G 53% /
/dev/nvme1n1p2 477G 231G 207G 53% /home
/dev/nvme1n1p2 477G 231G 207G 53% /var/cache
/dev/nvme1n1p2 477G 231G 207G 53% /var/lib/postgres
/dev/nvme1n1p2 477G 231G 207G 53% /var/log
/dev/nvme1n1p2 477G 231G 207G 53% /var/tmp
Don't think i've seen that before, how and why did you configure that ?
Disliking systemd intensely, but not satisfied with alternatives so focusing on taming systemd.
clean chroot building not flexible enough ?
Try clean chroot manager by graysky
Offline
tmpfs 32G 3,7G 28G 12% /dev/shm
that's usually zero if the system is not doing much.
You're right, mea culpa, took it from a running env. It's empty in rescue.target.
Don't think i've seen that before, how and why did you configure that ?
These are just btrfs subvolumes.
Offline
So status quo:
You've 62.7G total, 57.5G free, 55.7G available w/ 0.55G actually used by userspace and kernel.
The available calculation is obviously nonsense, so against the *free* RAM, you're losing ~4.7G (one DVD…)
* It's not the kernel version (you tested LTS? Can you try to just downgrade the kernel to the manjaro version?)
* It's not amdgpu, GART/GTT
* it's not tmpfs usage
* [edit] it's not THP
Compare lsmod on the manjaro and arch system - something's clearly leaking lots of RAM.
If you're using any OOT modules, start there.
Last edited by seth (2024-06-28 15:48:15)
Offline
* It's not amdgpu, GART/GTT
I just noticed that amdgpu.gttsize < 32 is ignored (at least as of now). Setting it to 32 renders the machine unusable (at least on wayland), dmesg busy-loops:
[amdgpu] *ERROR* Not enough memory for command submission!
Offline
Pages: 1