Mirrors/criu
1
0
Fork 0
mirror of https://github.com/checkpoint-restore/criu.git synced 2026-07-18 00:58:31 +00:00
Code Issues Projects Releases Wiki Activity

plugin/amdgpu: Allow dump with victim unable to see all gpus

Browse source 
In container setups, particularly k8s, cgroup may be used to
restrict what GPUs a process has access to. CRIU may then be used
to checkpoint that process while CRIU itself has full access to
all GPUs.

In this case, the kfd CRIU ioctl will return only a subset of the
ioctls visible through sysfs.

Don't write these devices or any links connected to them to the
tology dump files.

Signed-off-by: David Francis <David.Francis@amd.com>
This commit is contained in:
David Francis 2026-02-03 14:22:52 -05:00 • committed by Andrei Vagin
parent cfccc83872
commit fc2cddbdb9
1 changed files with 11 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

13
plugins/amdgpu/amdgpu_plugin.c
View file

@ -181,7 +181,7 @@ static int allocate_bo_entries(CriuKfd *e, int num_bos, struct kfd_criu_bo_bucke
int topology_to_devinfo(struct tp_system *sys, struct device_maps *maps, KfdDeviceEntry **deviceEntries)
{
uint32_t devinfo_index = 0;
struct tp_node *node;
struct tp_node *node, *node2;
list_for_each_entry(node, &sys->nodes, listm_system) {
KfdDeviceEntry *devinfo = deviceEntries[devinfo_index++];
@ -191,7 +191,7 @@ int topology_to_devinfo(struct tp_system *sys, struct device_maps *maps, KfdDevi
if (NODE_IS_GPU(node)) {
devinfo->gpu_id = maps_get_dest_gpu(maps, node->gpu_id);
if (!devinfo->gpu_id)
return -EINVAL;
continue;
devinfo->simd_count = node->simd_count;
devinfo->mem_banks_count = node->mem_banks_count;
@ -233,9 +233,18 @@ int topology_to_devinfo(struct tp_system *sys, struct device_maps *maps, KfdDevi
return -ENOMEM;
list_for_each_entry(iolink, &node->iolinks, listm) {
bool link_to_present_node = false;
if (!iolink->valid)
continue;
list_for_each_entry(node2, &sys->nodes, listm_system)
if (node2->id == iolink->node_to_id && maps_get_dest_gpu(maps, node2->gpu_id) != 0)
link_to_present_node = true;
if (!link_to_present_node)
continue;
devinfo->iolinks[iolink_index] = xmalloc(sizeof(DevIolink));
if (!devinfo->iolinks[iolink_index])
return -ENOMEM;