On Mon, Aug 19, 2019 at 09:38:41AM -0300, Jason Gunthorpe wrote: > On Mon, Aug 19, 2019 at 07:24:09PM +1000, Dave Chinner wrote: > > > So that leaves just the normal close() syscall exit case, where the > > application has full control of the order in which resources are > > released. We've already established that we can block in this > > context. Blocking in an interruptible state will allow fatal signal > > delivery to wake us, and then we fall into the > > fatal_signal_pending() case if we get a SIGKILL while blocking. > > The major problem with RDMA is that it doesn't always wait on close() for the > MR holding the page pins to be destoyed. This is done to avoid a > deadlock of the form: > > uverbs_destroy_ufile_hw() > mutex_lock() > [..] > mmput() > exit_mmap() > remove_vma() > fput(); > file_operations->release() > ib_uverbs_close() > uverbs_destroy_ufile_hw() > mutex_lock() <-- Deadlock > > But, as I said to Ira earlier, I wonder if this is now impossible on > modern kernels and we can switch to making the whole thing > synchronous. That would resolve RDMA's main problem with this. I'm still looking into this... but my bigger concern is that the RDMA FD can be passed to other processes via SCM_RIGHTS. Which means the process holding the pin may _not_ be the one with the open file and layout lease... Ira
