Mirrors/ProxMenux
1
0
Fork 0
mirror of https://github.com/MacRimi/ProxMenux.git synced 2026-11-02 10:52:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(gpu): changing the mode of a GPU leaves the other devices of its slot in their VM

Browse Source
This commit is contained in:
MacRimi
2026-10-12 12:12:59 +02:00
parent de7fc4bd49
commit bff791581c
5 changed files with 207 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files
scripts/gpu_tpu/add_gpu_vm.sh
+17 -11
View File
@@ -458,12 +458,14 @@ _is_pci_slot_assigned_to_vm() {
local pci_full="$1"
local vmid="$2"
local slot_base
slot_base="${pci_full#0000:}"
slot_base="${slot_base%.*}" # 01:00
local bdf slot_base
bdf="${pci_full#0000:}" # 01:00.0
slot_base="${bdf%.*}" # 01:00
# The device itself or its whole slot; another function at the same slot
# is a different device.
qm config "$vmid" 2>/dev/null \
| grep -qE "^hostpci[0-9]+:.*(0000:)?${slot_base}(\\.[0-7])?([,[:space:]]|$)"
| grep -qE "^hostpci[0-9]+:.*(0000:)?(${bdf}|${slot_base})([,;[:space:]]|$)"
}
# Match a specific PCI function when possible.
@@ -1453,7 +1455,9 @@ check_switch_mode() {
local vmid
vmid=$(basename "$conf" .conf)
[[ "$vmid" == "$SELECTED_VMID" ]] && continue # same target VM, no conflict
if grep -qE "hostpci[0-9]+:.*${pci_slot}" "$conf"; then
# Another VM has the GPU when it has the function of the GPU itself;
# a neighbour at the same slot, as a USB controller, is not the GPU.
if [[ -n "$(_vm_gpu_hostpci_indexes "$conf" "$SELECTED_GPU_PCI" "${IOMMU_DEVICES[@]}")" ]]; then
vm_src_id="$vmid"
vm_src_name=$(grep "^name:" "$conf" 2>/dev/null | awk '{print $2}')
break
@@ -2008,12 +2012,14 @@ cleanup_vm_config() {
local src_conf="/etc/pve/qemu-server/${SWITCH_VM_SRC}.conf"
if [[ -f "$src_conf" ]]; then
msg_info "$(translate 'Removing GPU from VM') ${SWITCH_VM_SRC}..."
# Precise regex: slot must be followed by ".<function>" and a
# delimiter. Kept in sync with switch_gpu_mode.sh. A looser
# ".*${pci_slot}" would match the slot as a substring and wipe
# unrelated hostpci entries (e.g. slot "00:02" matching inside
# a dGPU BDF 0000:02:00.0).
sed -E -i "/^hostpci[0-9]+:[[:space:]]*(0000:)?${pci_slot}\.[0-7]([,[:space:]]|$)/d" "$src_conf"
# Only the entries that carry the GPU are removed: its own function,
# the audio at its slot and the devices of its IOMMU group. Other
# functions at the same slot stay in the source VM.
local _gpu_idx
while IFS= read -r _gpu_idx; do
[[ -n "$_gpu_idx" ]] || continue
_vm_remove_hostpci_index "$SWITCH_VM_SRC" "$_gpu_idx" "$LOG_FILE"
done < <(_vm_gpu_hostpci_indexes "$src_conf" "$SELECTED_GPU_PCI" "${IOMMU_DEVICES[@]}")
msg_ok "$(translate 'GPU removed from VM') ${SWITCH_VM_SRC}" | tee -a "$screen_capture"
# Cascade cleanup: detect audio companions orphaned in the