Skip to content

Commit 040e2cd

Browse files
authored
Merge pull request #296 from NVIDIA/stop-nvlsm-pid
2 parents 852dbed + 1d5769c commit 040e2cd

File tree

2 files changed

+30
-0
lines changed

2 files changed

+30
-0
lines changed

ubuntu22.04/precompiled/nvidia-driver

Lines changed: 15 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -239,6 +239,21 @@ _unload_driver() {
239239
fi
240240
fi
241241

242+
if [ -f /var/run/nvidia-fabricmanager/nvlsm.pid ]; then
243+
echo "Stopping NVLink Subnet Manager daemon..."
244+
local pid=$(< /var/run/nvidia-fabricmanager/nvlsm.pid)
245+
246+
kill -SIGTERM "${pid}"
247+
for i in $(seq 1 50); do
248+
kill -0 "${pid}" 2> /dev/null || break
249+
sleep 0.1
250+
done
251+
if [ $i -eq 50 ]; then
252+
echo "Could not stop NVLink Subnet Manager daemon" >&2
253+
return 1
254+
fi
255+
fi
256+
242257
echo "Unloading NVIDIA driver kernel modules..."
243258
if [ -f /sys/module/nvidia_drm/refcnt ]; then
244259
nvidia_drm_refs=$(< /sys/module/nvidia_drm/refcnt)

ubuntu24.04/precompiled/nvidia-driver

Lines changed: 15 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -239,6 +239,21 @@ _unload_driver() {
239239
fi
240240
fi
241241

242+
if [ -f /var/run/nvidia-fabricmanager/nvlsm.pid ]; then
243+
echo "Stopping NVLink Subnet Manager daemon..."
244+
local pid=$(< /var/run/nvidia-fabricmanager/nvlsm.pid)
245+
246+
kill -SIGTERM "${pid}"
247+
for i in $(seq 1 50); do
248+
kill -0 "${pid}" 2> /dev/null || break
249+
sleep 0.1
250+
done
251+
if [ $i -eq 50 ]; then
252+
echo "Could not stop NVLink Subnet Manager daemon" >&2
253+
return 1
254+
fi
255+
fi
256+
242257
echo "Unloading NVIDIA driver kernel modules..."
243258
if [ -f /sys/module/nvidia_drm/refcnt ]; then
244259
nvidia_drm_refs=$(< /sys/module/nvidia_drm/refcnt)

0 commit comments

Comments
 (0)