nopwrites on dmu_sync-ed blocks can result in a panic

After device removal, performing nopwrites on a dmu_sync-ed block
will result in a panic. This panic can show up in two ways:
1. an attempt to issue an IOCTL in vdev_indirect_io_start()
2. a failed comparison of zio->io_bp and zio->io_bp_orig in
   zio_done()
To resolve both of these panics, nopwrites of blocks on indirect
vdevs should be ignored and new allocations should be performed on
concrete vdevs.

Reviewed-by: Igor Kozhukhov <[email protected]>
Reviewed-by: Pavel Zakharov <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Reviewed-by: Don Brady <[email protected]>
Signed-off-by: George Wilson <[email protected]>
Closes #8957 

3 files changed, 89 insertions, 2 deletions

diff --git a/tests/runfiles/linux.run b/tests/runfiles/linux.run
index e5dd044ac..5d33b2058 100644
--- a/tests/runfiles/linux.run
+++ b/tests/runfiles/linux.run
@@ -763,7 +763,7 @@ tags = ['functional', 'refreserv']
 pre =
 tests = ['removal_all_vdev', 'removal_check_space',
     'removal_condense_export', 'removal_multiple_indirection',
-    'removal_remap_deadlists',
+    'removal_nopwrite', 'removal_remap_deadlists',
     'removal_resume_export', 'removal_sanity', 'removal_with_add',
     'removal_with_create_fs', 'removal_with_dedup',
     'removal_with_errors', 'removal_with_export',
diff --git a/tests/zfs-tests/tests/functional/removal/Makefile.am b/tests/zfs-tests/tests/functional/removal/Makefile.am
index dc52bf1f2..52bc42efd 100644
--- a/tests/zfs-tests/tests/functional/removal/Makefile.am
+++ b/tests/zfs-tests/tests/functional/removal/Makefile.am
@@ -18,7 +18,7 @@ pkgdatadir = $(datadir)/@PACKAGE@/zfs-tests/tests/functional/removal
 dist_pkgdata_SCRIPTS = \
 	cleanup.ksh removal_all_vdev.ksh removal_check_space.ksh \
 	removal_condense_export.ksh removal_multiple_indirection.ksh \
-	removal_remap_deadlists.ksh \
+	removal_nopwrite.ksh removal_remap_deadlists.ksh \
 	removal_reservation.ksh removal_resume_export.ksh \
 	removal_sanity.ksh removal_with_add.ksh removal_with_create_fs.ksh \
 	removal_with_dedup.ksh removal_with_errors.ksh \
diff --git a/tests/zfs-tests/tests/functional/removal/removal_nopwrite.ksh b/tests/zfs-tests/tests/functional/removal/removal_nopwrite.ksh
new file mode 100755
index 000000000..cb8bd6b81
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/removal/removal_nopwrite.ksh
@@ -0,0 +1,87 @@
+#! /bin/ksh -p
+#
+# CDDL HEADER START
+#
+# This file and its contents are supplied under the terms of the
+# Common Development and Distribution License ("CDDL"), version 1.0.
+# You may only use this file in accordance with the terms of version
+# 1.0 of the CDDL.
+#
+# A full copy of the text of the CDDL should have accompanied this
+# source.  A copy of the CDDL is also available via the Internet at
+# http://www.illumos.org/license/CDDL.
+#
+# CDDL HEADER END
+#
+
+#
+# Copyright (c) 2019 by Delphix. All rights reserved.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/removal/removal.kshlib
+. $STF_SUITE/tests/functional/nopwrite/nopwrite.shlib
+
+default_setup_noexit "$DISKS"
+log_onexit default_cleanup_noexit
+BLOCKSIZE=8192
+
+origin="$TESTPOOL/$TESTFS"
+
+log_must zfs set compress=on $origin
+log_must zfs set checksum=edonr $origin
+
+log_must zfs set recordsize=8k $origin
+dd if=/dev/urandom of=$TESTDIR/file_8k bs=1024k count=$MEGS oflag=sync \
+    conv=notrunc >/dev/null 2>&1 || log_fail "dd into $TESTDIR/file failed."
+log_must zfs set recordsize=128k $origin
+dd if=/dev/urandom of=$TESTDIR/file_128k bs=1024k count=$MEGS oflag=sync \
+    conv=notrunc >/dev/null 2>&1 || log_fail "dd into $TESTDIR/file failed."
+
+zfs snapshot $origin@a || log_fail "zfs snap failed"
+log_must zfs clone $origin@a $origin/clone
+
+#
+# Verify that nopwrites work prior to removal
+#
+log_must zfs set recordsize=8k $origin/clone
+dd if=/$TESTDIR/file_8k of=/$TESTDIR/clone/file_8k bs=1024k \
+     oflag=sync conv=notrunc >/dev/null 2>&1 || log_fail "dd failed."
+log_must verify_nopwrite $origin $origin@a $origin/clone
+
+log_must zfs set recordsize=128k $origin/clone
+dd if=/$TESTDIR/file_128k of=/$TESTDIR/clone/file_128k bs=1024k \
+     oflag=sync conv=notrunc >/dev/null 2>&1 || log_fail "dd failed."
+log_must verify_nopwrite $origin $origin@a $origin/clone
+
+#
+# Remove a device before testing nopwrites again
+#
+log_must zpool remove $TESTPOOL $REMOVEDISK
+log_must wait_for_removal $TESTPOOL
+log_mustnot vdevs_in_pool $TESTPOOL $REMOVEDISK
+
+#
+# Normally, we expect nopwrites to avoid allocating new blocks, but
+# after a device has been removed the DVAs will get remapped when
+# a L0's indirect bloock is written. This will negate the effects
+# of nopwrite and should result in new allocations.
+#
+
+#
+# Perform a direct zil nopwrite test
+#
+log_must zfs set recordsize=8k $origin/clone
+dd if=/$TESTDIR/file_8k of=/$TESTDIR/clone/file_8k bs=1024k \
+     oflag=sync conv=notrunc >/dev/null 2>&1 || log_fail "dd failed."
+log_mustnot verify_nopwrite $origin $origin@a $origin/clone
+
+#
+# Perform an indirect zil nopwrite test
+#
+log_must zfs set recordsize=128k $origin/clone
+dd if=/$TESTDIR/file_128k of=/$TESTDIR/clone/file_128k bs=1024k \
+     oflag=sync conv=notrunc >/dev/null 2>&1 || log_fail "dd failed."
+log_mustnot verify_nopwrite $origin $origin@a $origin/clone
+
+log_pass "Remove works with nopwrite."