I didn't read this series. This is not my area and to be honest even the API doesn't fit my head. I leave this to other reviewers. Just a couple of nits after very quick glance. On 09/04, Adalbert Lazăr wrote: > > +static vm_fault_t mirror_vm_fault(struct vm_fault *vmf) > +{ ... > + up_read(¤t->mm->mmap_sem); ... > + down_read(¤t->mm->mmap_sem); > + > + /* expedite retry */ > + if (mmu_interval_check_retry(&view->mmin, seq)) { > + put_page(req_page); > + > + srcu_read_unlock(&fctx->fault_srcu, idx); > + > + goto fault_retry; > + } > + > + /* make sure the VMA hasn't gone away */ > + vma = find_vma(current->mm, vmf->address); > + if (vma == vmf->vma) { vmf->vma can go away, its memory can be freed and re-allocated as another vma returned by find_vma() above. > +int task_remote_map(struct task_struct *task, int fds[]) > +{ ... > + fds[1] = anon_inode_getfd("[pidfd_mem.map]", &pidfd_mem_map_fops, fctx, > + O_RDWR | O_CLOEXEC | O_LARGEFILE); > + if (fds[1] < 0) { > + ret = fds[1]; > + goto out; > + } > + remote_file_context_get(fctx); > + > + map = fget(fds[1]); Another thread can close this file right after fd_install(). fget() can return NULL or another unrelated file. Oleg.