[pve-devel] [PATCH qemu-server 1/1] fix #6400: pci: allow other pci domains than 0000 for NVIDIA vGPUs
Dominik Csapak
d.csapak at proxmox.com
Thu May 15 15:00:34 CEST 2025
when creating or cleaning up NVIDIA vGPUs, we mistakenly assumed a
PCI domain of 0000, but this might be different.
Use 'normalize_pci_id' from PVE::SysFSTools, which handles this already
correctly.
Signed-off-by: Dominik Csapak <d.csapak at proxmox.com>
---
depends on the pve-common patch
PVE/QemuServer.pm | 2 +-
PVE/QemuServer/PCI.pm | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/PVE/QemuServer.pm b/PVE/QemuServer.pm
index 577959a4..7f937a38 100644
--- a/PVE/QemuServer.pm
+++ b/PVE/QemuServer.pm
@@ -6131,7 +6131,7 @@ sub cleanup_pci_devices {
my $reservations = PVE::QemuServer::PCI::get_reservations($vmid);
# clean up nvidia devices
for my $id ($reservations->@*) {
- $id = '0000:'.$id if $id !~ m/^0000:/;
+ $id = PVE::SysFSTools::normalize_pci_id($id);
my $create_path = "/sys/bus/pci/devices/$id/nvidia/current_vgpu_type";
diff --git a/PVE/QemuServer/PCI.pm b/PVE/QemuServer/PCI.pm
index afb60ad2..5e74bed3 100644
--- a/PVE/QemuServer/PCI.pm
+++ b/PVE/QemuServer/PCI.pm
@@ -528,7 +528,7 @@ sub parse_hostpci_devices {
my sub create_nvidia_device {
my ($id, $model) = @_;
- $id = '0000:'.$id if $id !~ m/^0000:/;
+ $id = PVE::SysFSTools::normalize_pci_id($id);
my $creation = "/sys/bus/pci/devices/$id/nvidia/current_vgpu_type";
--
2.39.5
More information about the pve-devel
mailing list