summary refs log tree commit diff stats
path: root/gitlab/issues/target_missing/host_missing/accel_missing/1705.toml
diff options
context:
space:
mode:
Diffstat (limited to 'gitlab/issues/target_missing/host_missing/accel_missing/1705.toml')
-rw-r--r--gitlab/issues/target_missing/host_missing/accel_missing/1705.toml74
1 files changed, 74 insertions, 0 deletions
diff --git a/gitlab/issues/target_missing/host_missing/accel_missing/1705.toml b/gitlab/issues/target_missing/host_missing/accel_missing/1705.toml
new file mode 100644
index 000000000..bfebefadd
--- /dev/null
+++ b/gitlab/issues/target_missing/host_missing/accel_missing/1705.toml
@@ -0,0 +1,74 @@
+id = 1705
+title = "Illegal instruction when I want to numactl --cpubind=0 --membind=1 to CXL Memory"
+state = "closed"
+created_at = "2023-06-13T02:50:15.126Z"
+closed_at = "2023-09-22T05:30:36.680Z"
+labels = []
+url = "https://gitlab.com/qemu-project/qemu/-/issues/1705"
+host-os = "Ubuntu 20.04"
+host-arch = "x86"
+qemu-version = "8.0.50"
+guest-os = "Ubuntu 22.04"
+guest-arch = "x86"
+description = """I ran QEMU for simulating CXL DRAM and when I tried to run `numactl --cpubind=0 --membind=1  ls` , I got `Illegal instruction`
+The numa node 1 was the CXL DRAM simulated by QEMU.
+
+> root@8003:~# numactl -H
+> available: 2 nodes (0-1)
+> node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
+> node 0 size: 32090 MB
+> node 0 free: 31325 MB
+> node 1 cpus:
+> node 1 size: 32768 MB
+> node 1 free: 32768 MB
+> node distances:
+> node 0 1
+> 0: 10 20
+> 1: 20 10
+
+When I ran on numa node 1, no failed
+
+> root@8003:~# numactl --membind=0 ls
+> ndctl
+
+When I ran on numa node 1(CXL DRAM),it failed.
+
+> root@8003:~# numactl --membind=1 ls
+> [ 913.975032] traps: ls[667] trap invalid opcode ip:7fdec255d180 sp:7ffd3c507288 error:0 in ld-linux-x86-64.so.2[7fdec2546000+2a000]
+> **Illegal instruction**"""
+reproduce = """1. start the guest
+2. cxl list  (we could see the simulated CXL DRAM)
+> root@8003:~# cxl list
+> [
+>   {
+>     "memdev":"mem0",
+>     "ram_size":34359738368,
+>     "serial":0,
+>     "host":"0000:0d:00.0"
+>   }
+> ]
+3. cxl create-region -t ram -d decoder0.0 -m mem0
+4. daxctl reconfigure-device dax0.0 --mode=system-ram
+5. numactl -H
+> root@8003:~# numactl -H
+> available: 2 nodes (0-1)
+> node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
+> node 0 size: 32090 MB
+> node 0 fr
+> ee: 31254 MB
+> node 1 cpus:
+> node 1 size: 32768 MB
+> node 1 free: 32768 MB
+> node distances:
+> node   0   1 
+>   0:  10  20 
+>   1:  20  10 
+6. numactl --membind=1 ls
+> root@8003:~# numactl --membind=1 ls
+> [38441.892140] **traps: ls[861] trap invalid opcode ip:7f15db6ac180 sp:7ffc648755c8 **error:0 in ld-linux-x86-64.so.2[7f15db695000+2a000]
+> **Illegal instruction**"""
+additional = """When I run dmesg, I found an error.
+> root@8003:~# dmesg|grep error
+> [    2.321130] platform regulatory.0: Direct firmware load for regulatory.db failed with error -2
+
+Since my CPU is a Xeon III, not a Xeon IV with CXL support, **I'm wondering if it's because the CPU doesn't support CXL instructions, or if the Xeon III can emulate it, just because my settings don't make sense**. If this is my settings problem, could you help me to deal this? Or it just caused by my Xeon III,I will update it to  Xeon IV."""