add results

1 files changed, 29 insertions, 0 deletions

diff --git a/results/classifier/gemma3:27b/runtime/1830 b/results/classifier/gemma3:27b/runtime/1830
new file mode 100644
index 00000000..27322faa
--- /dev/null
+++ b/results/classifier/gemma3:27b/runtime/1830
@@ -0,0 +1,29 @@
+
+
+
+command hangs in CentOS 7 arm64 container with Ubuntu 22 amd64 host
+Description of problem:
+The command hangs in the container, taking over the CPU:
+
+```
+$ docker run -it centos:7
+[root@42e655bf3d60 /]# LD_DEBUG=all /lib64/ld-2.17.so --list /usr/bin/true &
+[1] 74
+[root@42e655bf3d60 /]#         74:      file=/usr/bin/true [0];  generating link map
+
+[root@42e655bf3d60 /]# ps -e -o pid,ppid,etime,time,state,args
+    PID    PPID     ELAPSED     TIME S COMMAND
+      1       0       34:59 00:00:00 S /usr/libexec/qemu-binfmt/aarch64-binfmt-P /bin/bash /bin/bash
+     74       1       03:16 00:03:13 R /usr/libexec/qemu-binfmt/aarch64-binfmt-P /lib64/ld-2.17.so /lib64/ld-2.17.so
+     80       1  4-19:34:01 00:00:00 R ps -e -o pid,ppid,etime,time,state,args
+[root@42e655bf3d60 /]#
+```
+Steps to reproduce:
+1. Start container
+2. Run `/lib64/ld-2.17.so --list /usr/bin/true`
+Additional information:
+1. The problem is not observed in an Ubuntu 20.04 host system performing the same scenario.
+2. My team build environment has amd64 native architecture hardware.  I ran a similar scenario on an AWS arm64 native machine (QEMU is not needed) and the command works fine in the container.
+3. My team builds several Linux images daily - about a dozen amd64 and eight arm64.  This is the only image that's causing us this problem.
+4. I built trace-cmd but when I tried to start a trace it told me `No events enabled with kvm`.
+5. I built qemu-8.1.0-rc3 and saw the same behavior but I don't think `/usr/libexec/qemu-binfmt/aarch64-binfmt-P` was replaced with a new version so I don't think the old version was used for my container.