Hi Geert, > > Signed-off-by: Wolfram Sang > > Thanks for your patch, which looks useful! Glad you like it! > Can you please put it in a subdir, like all other summaries: Sadly not, because of the paragraph from above: > > Note that mailbox controllers currently can be instantiated at any > > initcall stage. So, per-controller debugfs handling was discarded > > because it is not clear when to create the root "mailbox"-debugfs entry. > > A central file was chosen, similar to the GPIO subsystem, which is > > independent of the initcall stage because it will be accessed only when > > userspace is available anyhow. I would need to use core_initcall for the mailbox-subsys. debugfs itself also uses core_initcall. Then, some drivers (e.g. hi3660-mailbox) use core_initcall. I don't think the subdir is worth all the potential race conditions and fragile solutions which usually come with it. Or do you know a solid solution to this problem? Happy hacking, Wolfram