State backend doesn't necessarily point at a device tree node that has a device associated, e.g. when we have a top-level fixed-partitions node with the barebox-specific partuuid binding. As of_find_path_by_node will iterate over all available cdev's we need not handle the error and can just call of_find_path_by_node. Signed-off-by: Ahmad Fatoum <a.fatoum@xxxxxxxxxxxxxx> --- v1 -> v2: - new patch --- common/state/state.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) diff --git a/common/state/state.c b/common/state/state.c index 6b4acbb32bcc..85854f564d48 100644 --- a/common/state/state.c +++ b/common/state/state.c @@ -618,9 +618,15 @@ struct state *state_new_from_node(struct device_node *node, bool readonly) } #ifdef __BAREBOX__ - ret = of_partition_ensure_probed(partition_node); - if (ret) - goto out_release_state; + /* + * On EFI, where devices are not instantiated from device tree, the + * state backend may point at a top-level fixed-partitions partition + * subnode with a partuuid property, which will be looked up globally. + * + * In order to support this binding, we do not early exit when + * of_partition_ensure_probed fails, but instead try the custom binding. + */ + (void)of_partition_ensure_probed(partition_node); ret = of_find_path_by_node(partition_node, &state->backend_path, 0); #else -- 2.39.2
