From d012836fb61654942c9d573c8e0f9def598d4ae2 Mon Sep 17 00:00:00 2001
From: Mateusz Guzik <mjg@FreeBSD.org>
Date: Wed, 5 Apr 2023 20:27:12 +0000
Subject: [PATCH] zfs: fix up EXDEV handling for clone_range

API contract requires VOPs to handle EXDEV internally, worst case by
falling back to the generic copy routine. This broke with the recent
changes.

While here whack custom loop to lock 2 vnodes with vn_lock_pair, which
provides the same functionality internally. write start/finish around
it plays no role so got eliminated.

One difference is that vn_lock_pair always takes an exclusive lock on
both vnodes. I did not patch around it because current code takes an
exclusive lock on the target vnode. zfs supports shared-locking for
writes, so this serializes different calls to the routine as is, despite
range locking inside. At the same time you may notice the source vnode
can get some traffic if only shared-locked, thus once more this goes
the safer route of exclusive-locking. Note this should be patched to
use shared-locking for both once the feature is considered stable.

Technically the switch to vn_lock_pair should be a separate change, but
it would only introduce churn immediately whacked by the rest of the
patch.

[note: technically the review is still in progress, but so is the
active breakage]

Sponsored by:	Rubicon Communications, LLC ("Netgate")
---
 .../module/os/freebsd/zfs/zfs_vnops_os.c      | 56 +++++++++----------
 1 file changed, 27 insertions(+), 29 deletions(-)

diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c
index baa2ee5b3824..67c1a6e3344b 100644
--- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c
+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c
@@ -6250,56 +6250,54 @@ zfs_freebsd_copy_file_range(struct vop_copy_file_range_args *ap)
 	 * need something else than vn_generic_copy_file_range().
 	 */
 
-	/* Lock both vnodes, avoiding risk of deadlock. */
-	do {
-		mp = NULL;
-		error = vn_start_write(outvp, &mp, V_WAIT);
-		if (error == 0) {
-			error = vn_lock(outvp, LK_EXCLUSIVE);
-			if (error == 0) {
-				if (invp == outvp)
-					break;
-				error = vn_lock(invp, LK_SHARED | LK_NOWAIT);
-				if (error == 0)
-					break;
-				VOP_UNLOCK(outvp);
-				if (mp != NULL)
-					vn_finished_write(mp);
-				mp = NULL;
-				error = vn_lock(invp, LK_SHARED);
-				if (error == 0)
-					VOP_UNLOCK(invp);
-			}
+	vn_start_write(outvp, &mp, V_WAIT);
+	if (invp == outvp) {
+		if (vn_lock(outvp, LK_EXCLUSIVE) != 0) {
+			goto bad_write_fallback;
 		}
-		if (mp != NULL)
-			vn_finished_write(mp);
-	} while (error == 0);
-	if (error != 0)
-		return (error);
+	} else {
+		vn_lock_pair(invp, false, outvp, false);
+		if (VN_IS_DOOMED(invp) || VN_IS_DOOMED(outvp)) {
+			goto bad_locked_fallback;
+		}
+	}
+
 #ifdef MAC
 	error = mac_vnode_check_write(curthread->td_ucred, ap->a_outcred,
 	    outvp);
 	if (error != 0)
-		goto unlock;
+		goto out_locked;
 #endif
 
 	io.uio_offset = *ap->a_outoffp;
 	io.uio_resid = *ap->a_lenp;
 	error = vn_rlimit_fsize(outvp, &io, ap->a_fsizetd);
 	if (error != 0)
-		goto unlock;
+		goto out_locked;
 
 	error = zfs_clone_range(VTOZ(invp), ap->a_inoffp, VTOZ(outvp),
 	    ap->a_outoffp, &len, ap->a_outcred);
+	if (error == EXDEV)
+		goto bad_locked_fallback;
 	*ap->a_lenp = (size_t)len;
-
-unlock:
+out_locked:
 	if (invp != outvp)
 		VOP_UNLOCK(invp);
 	VOP_UNLOCK(outvp);
 	if (mp != NULL)
 		vn_finished_write(mp);
+	return (error);
 
+bad_locked_fallback:
+	if (invp != outvp)
+		VOP_UNLOCK(invp);
+	VOP_UNLOCK(outvp);
+bad_write_fallback:
+	if (mp != NULL)
+		vn_finished_write(mp);
+	error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
+	    ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags,
+	    ap->a_incred, ap->a_outcred, ap->a_fsizetd);
 	return (error);
 }