Re: [PATCH net 1/1] tipc: revert use of copy_from_iter_full()

On Thu, Dec 22, 2016 at 01:21:01AM +0000, Al Viro wrote:

On Wed, Dec 21, 2016 at 08:01:37PM -0500, Jon Maloy wrote:

quoted

commit cbbd26b8b1a6 ("[iov_iter] new primitives - copy_from_iter_full()
and friends") replaced calls to copy_from_iter() in the function
tipc_msg_build(). This causes a an immediate crash as follows:

Very interesting.

quoted

[ 1209.597076] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[ 1209.607025] IP: copy_from_iter_full+0x43/0x290

quoted

[ 1209.689257]  tipc_msg_build+0xe1/0x590 [tipc]
[ 1209.691479]  ? _raw_spin_unlock_bh+0x1e/0x20
[ 1209.694641]  ? tipc_node_find+0x30/0xa0 [tipc]
[ 1209.696789]  __tipc_sendmsg+0x189/0x480 [tipc]
[ 1209.699017]  ? remove_wait_queue+0x4d/0x60
[ 1209.700354]  tipc_connect+0x15f/0x1b0 [tipc]
[ 1209.701684]  SYSC_connect+0xd9/0x110

I don't believe that it's something tipc-specific; could you post an objdump
of copy_from_iter_full() in your kernel?  That smells like a bug in there
and it really ought to be fixed...

FWIW, looking at the tipc, am I reading the trace correctly?  We seem to
have tipc_connect() taking an msghdr with empty payload and hitting this
        switch (sk->sk_state) {
        case TIPC_OPEN:
                /* Send a 'SYN-' to destination */
                m.msg_name = dest;
                m.msg_namelen = destlen;

                /* If connect is in non-blocking case, set MSG_DONTWAIT to
                 * indicate send_msg() is never blocked.
                 */
                if (!timeout)
                        m.msg_flags = MSG_DONTWAIT;

                res = __tipc_sendmsg(sock, &m, 0);
which eventually calls
        rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, &pktchain);
possibly more than once, but with explicit "restore m->msg_iter to what
it was before the first call" before each subsequent call.

What's putting anything into m.msg_iter on that codepath?  AFAICS, it should
be completely empty...  Wait.  AAARRRGH!

OK, I see what's going on there - unlike iterate_and_advance(), which
explicitly skips any work in case of empty iterator, iterate_all_kind()
does not.  Could you check if the following fixes your problem?

diff --git a/lib/iov_iter.c b/lib/iov_iter.c
index 228892dabba6..6a0396b8d47f 100644
--- a/lib/iov_iter.c
+++ b/lib/iov_iter.c

@@ -73,19 +73,21 @@
 }
 
 #define iterate_all_kinds(i, n, v, I, B, K) {			\
-	size_t skip = i->iov_offset;				\
-	if (unlikely(i->type & ITER_BVEC)) {			\
-		struct bio_vec v;				\
-		struct bvec_iter __bi;				\
-		iterate_bvec(i, n, v, __bi, skip, (B))		\
-	} else if (unlikely(i->type & ITER_KVEC)) {		\
-		const struct kvec *kvec;			\
-		struct kvec v;					\
-		iterate_kvec(i, n, v, kvec, skip, (K))		\
-	} else {						\
-		const struct iovec *iov;			\
-		struct iovec v;					\
-		iterate_iovec(i, n, v, iov, skip, (I))		\
+	if (i->count) {						\
+		size_t skip = i->iov_offset;			\
+		if (unlikely(i->type & ITER_BVEC)) {		\
+			struct bio_vec v;			\
+			struct bvec_iter __bi;			\
+			iterate_bvec(i, n, v, __bi, skip, (B))	\
+		} else if (unlikely(i->type & ITER_KVEC)) {	\
+			const struct kvec *kvec;		\
+			struct kvec v;				\
+			iterate_kvec(i, n, v, kvec, skip, (K))	\
+		} else {					\
+			const struct iovec *iov;		\
+			struct iovec v;				\
+			iterate_iovec(i, n, v, iov, skip, (I))	\
+		}						\
 	}							\
 }