[RFC PATCH 02/13] fs/userfaultfd: fix wrong file usage with iouring

[Nadav Amit <nadav.amit@gmail.com>]

From: Nadav Amit <namit@vmware.com>

Using io-uring with userfaultfd for reads can lead upon a fork event to
the installation of the userfaultfd file descriptor on the worker kernel
thread instead of the process that initiated the read. io-uring assumes
that no new file descriptors are installed during read.

As a result the controlling process would not be able to access the
new forked process userfaultfd file descriptor.

To solve this problem, Save the files_struct of the process that
initiated userfaultfd syscall in the context and reload it when needed.

Cc: Jens Axboe <axboe@kernel.dk>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: io-uring@vger.kernel.org
Cc: linux-fsdevel@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: linux-mm@kvack.org
Fixes: 2b188cc1bb85 ("Add io_uring IO interface")
Signed-off-by: Nadav Amit <namit@vmware.com>
---
 fs/userfaultfd.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c
index c8ed4320370e..4fe07c1a44c6 100644
--- a/fs/userfaultfd.c
+++ b/fs/userfaultfd.c
@@ -27,6 +27,7 @@
 #include <linux/ioctl.h>
 #include <linux/security.h>
 #include <linux/hugetlb.h>
+#include <linux/fdtable.h>
 
 int sysctl_unprivileged_userfaultfd __read_mostly = 1;
 
@@ -76,6 +77,8 @@ struct userfaultfd_ctx {
 	bool mmap_changing;
 	/* mm with one ore more vmas attached to this userfaultfd_ctx */
 	struct mm_struct *mm;
+	/* controlling process files as they might be different than current */
+	struct files_struct *files;
 };
 
 struct userfaultfd_fork_ctx {
@@ -173,6 +176,7 @@ static void userfaultfd_ctx_put(struct userfaultfd_ctx *ctx)
 		VM_BUG_ON(spin_is_locked(&ctx->fd_wqh.lock));
 		VM_BUG_ON(waitqueue_active(&ctx->fd_wqh));
 		mmdrop(ctx->mm);
+		put_files_struct(ctx->files);
 		kmem_cache_free(userfaultfd_ctx_cachep, ctx);
 	}
 }
@@ -666,6 +670,8 @@ int dup_userfaultfd(struct vm_area_struct *vma, struct list_head *fcs)
 		ctx->mm = vma->vm_mm;
 		mmgrab(ctx->mm);
 
+		ctx->files = octx->files;
+		atomic_inc(&ctx->files->count);
 		userfaultfd_ctx_get(octx);
 		WRITE_ONCE(octx->mmap_changing, true);
 		fctx->orig = octx;
@@ -976,10 +982,32 @@ static int resolve_userfault_fork(struct userfaultfd_ctx *ctx,
 				  struct userfaultfd_ctx *new,
 				  struct uffd_msg *msg)
 {
+	struct files_struct *files = NULL;
 	int fd;
 
+	BUG_ON(new->files == NULL);
+
+	/*
+	 * This function can be called from another context than the controlling
+	 * process, for instance, for an io-uring submission kernel thread. If
+	 * that is the case we must ensure the correct files are being used.
+	 */
+	if (current->files != new->files) {
+		task_lock(current);
+		files = current->files;
+		current->files = new->files;
+		task_unlock(current);
+	}
+
 	fd = anon_inode_getfd("[userfaultfd]", &userfaultfd_fops, new,
 			      O_RDWR | (new->flags & UFFD_SHARED_FCNTL_FLAGS));
+
+	if (files != NULL) {
+		task_lock(current);
+		current->files = files;
+		task_unlock(current);
+	}
+
 	if (fd < 0)
 		return fd;
 
@@ -1986,6 +2014,8 @@ SYSCALL_DEFINE1(userfaultfd, int, flags)
 	/* prevent the mm struct to be freed */
 	mmgrab(ctx->mm);
 
+	ctx->files = get_files_struct(current);
+
 	fd = anon_inode_getfd("[userfaultfd]", &userfaultfd_fops, ctx,
 			      O_RDWR | (flags & UFFD_SHARED_FCNTL_FLAGS));
 	if (fd < 0) {
-- 
2.25.1