[PATCH v3 0/2] fstests: add patient module remover

[PATCH v3 0/2] fstests: add patient module remover

[Luis Chamberlain]

This v3 addresses the issues noted by Eryu, it also drops
the patches already merged.

Luis Chamberlain (2):
  common/module: add patient module rmmod support
  common/scsi_debug: use the patient module remover

 README            |   3 ++
 common/config     |  31 ++++++++++++++
 common/module     | 105 ++++++++++++++++++++++++++++++++++++++++++++++
 common/scsi_debug |   6 ++-
 4 files changed, 143 insertions(+), 2 deletions(-)

-- 
2.30.2

[PATCH v3 1/2] common/module: add patient module rmmod support

[Luis Chamberlain]

When we call rmmod it will fail if the refcnt is greater than 0.
This is expected, however, if using test modules such as scsi_debug,
userspace tests may expect that once userspace is done issuing out
commands it can safely remove the module, and the module will be
removed.

This is not true for few reasons. First, a module might take a while
to quiesce after its used. This varies module by module. For example,
at least for scsi_debug there is one patch to help with this but
that is not sufficient to address all the removal issues, it just helps
quiesce the module faster. If something like LVM pvremove is used, as in
the case of generic/108, it may take time before the module's refcnt goes
to 0 even if DM_DEFERRED_REMOVE is *not* used and even if udevadm settle
is used. Even *after* all this... the module refcnt is still very
fickle. For example, any blkdev_open() against a block device will bump
a module refcnt up and we have little control over stopping these
sporadic userspace calls after a test. A failure on module removal then
just becomes an inconvenience on false positives.

This was first observed on scsi_debug [0]. Doug worked on a patch to
help the driver quiesce [1]. Later the issue has been determined to be
generic [2]. The only way to properly resolve these issues is with a
patient module remover. The kernel used to support a wait for the
delete_module() system call, however this was later deprecated into
kmod with a 10 second userspace sleep. That 10 second sleep is long gone
from kmod now though. I've posted patches now for a kmod patient module
remover then [3], in light of the fact that this issue is generic and
the only way to then properly deal with this is implementing a userspace
patient module remover.

Use the kmod patient module remover when supported, otherwise we open
code our own solution inside fstests. We default to a timeout of 100
seconds. Each test can override the timeout by setting the variable
MODPROBE_PATIENT_RM_TIMEOUT_SECONDS or setting it to "forever" if they
wish for the patience to be infinite.

This uses kmod's patient module remover if you have that feature,
otherwise we open code a solution in fstests which is a simplified
version of what has been proposed for kmod.

[0] https://bugzilla.kernel.org/show_bug.cgi?id=212337
[1] https://lore.kernel.org/linux-scsi/20210508230745.27923-1-dgilbert@interlog.com/
[2] https://bugzilla.kernel.org/show_bug.cgi?id=214015
[3] https://lkml.kernel.org/r/20210810051602.3067384-1-mcgrof@kernel.org
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
---
 README        |   3 ++
 common/config |  31 +++++++++++++++
 common/module | 105 ++++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 139 insertions(+)

diff --git a/README b/README
index b9877b7d..ad4f3a18 100644
--- a/README
+++ b/README
@@ -121,6 +121,9 @@ Preparing system for tests:
                filesystem we can create. Setting this parameter will
                skip the tests creating a filesystem less than
                MIN_FSSIZE.
+	     - Set MODPROBE_PATIENT_RM_TIMEOUT_SECONDS to specify the amount of
+	       time we should try a patient module remove. The default is 50
+	       seconds.
 
         - or add a case to the switch in common/config assigning
           these variables based on the hostname of your test
diff --git a/common/config b/common/config
index 005fd50a..164381b7 100644
--- a/common/config
+++ b/common/config
@@ -252,6 +252,37 @@ if [[ "$UDEV_SETTLE_PROG" == "" || ! -d /proc/net ]]; then
 fi
 export UDEV_SETTLE_PROG
 
+# Set MODPROBE_PATIENT_RM_TIMEOUT_SECONDS to "forever" if you want the patient
+# modprobe removal to run forever trying to remove a module.
+MODPROBE_REMOVE_PATIENT=""
+modprobe --help | grep -q -1 "remove-patiently"
+if [[ $? -ne 0 ]]; then
+	if [[ -z "$MODPROBE_PATIENT_RM_TIMEOUT_SECONDS" ]]; then
+		# We will open code our own implementation of patient module
+		# remover in fstests. Use a 50 second default.
+		export MODPROBE_PATIENT_RM_TIMEOUT_SECONDS="50"
+	fi
+else
+	MODPROBE_RM_PATIENT_TIMEOUT_ARGS=""
+	if [[ ! -z "$MODPROBE_PATIENT_RM_TIMEOUT_SECONDS" ]]; then
+		if [[ "$MODPROBE_PATIENT_RM_TIMEOUT_SECONDS" != "forever" ]]; then
+			MODPROBE_PATIENT_RM_TIMEOUT_MS="$((MODPROBE_PATIENT_RM_TIMEOUT_SECONDS * 1000))"
+			MODPROBE_RM_PATIENT_TIMEOUT_ARGS="-t $MODPROBE_PATIENT_RM_TIMEOUT_MS"
+		fi
+	else
+		# We export MODPROBE_PATIENT_RM_TIMEOUT_SECONDS here for parity
+		# with environments without support for modprobe -p, but we
+		# only really need it exported right now for environments which
+		# don't have support for modprobe -p to implement our own
+		# patient module removal support within fstests.
+		export MODPROBE_PATIENT_RM_TIMEOUT_SECONDS="50"
+		MODPROBE_PATIENT_RM_TIMEOUT_MS="$((MODPROBE_PATIENT_RM_TIMEOUT_SECONDS * 1000))"
+		MODPROBE_RM_PATIENT_TIMEOUT_ARGS="-t $MODPROBE_PATIENT_RM_TIMEOUT_MS"
+	fi
+	MODPROBE_REMOVE_PATIENT="modprobe -p $MODPROBE_RM_TIMEOUT_ARGS"
+fi
+export MODPROBE_REMOVE_PATIENT
+
 export MKFS_XFS_PROG=$(type -P mkfs.xfs)
 export MKFS_EXT4_PROG=$(type -P mkfs.ext4)
 export MKFS_UDF_PROG=$(type -P mkudffs)
diff --git a/common/module b/common/module
index 39e4e793..8ccc2fa4 100644
--- a/common/module
+++ b/common/module
@@ -81,3 +81,108 @@ _get_fs_module_param()
 {
 	cat /sys/module/${FSTYP}/parameters/${1} 2>/dev/null
 }
+
+# checks the refcount and returns 0 if we can safely remove the module. rmmod
+# does this check for us, but we can use this to also iterate checking for this
+# refcount before we even try to remove the module. This is useful when using
+# debug test modules which take a while to quiesce.
+_patient_rmmod_check_refcnt()
+{
+	local module=$1
+	local refcnt=0
+
+	if [[ -f /sys/module/$module/refcnt ]]; then
+		refcnt=$(cat /sys/module/$module/refcnt 2>/dev/null)
+		if [[ $? -ne 0 || $refcnt -eq 0 ]]; then
+			return 0
+		fi
+		return 1
+	fi
+	return 0
+}
+
+# Patiently tries to wait to remove a module by ensuring first
+# the refcnt is 0 and then trying to persistently remove the module within
+# the time allowed. The timeout is configurable per test, just set
+# MODPROBE_PATIENT_RM_TIMEOUT_SECONDS prior to including this file.
+# If you want this to try forever just set MODPROBE_PATIENT_RM_TIMEOUT_SECONDS
+# to the special value of "forever". This applies to both cases where kmod
+# supports the patient module remover (modrobe -p) and where it does not.
+#
+# If your version of kmod supports modprobe -p, we instead use that
+# instead. Otherwise we have to implement a patient module remover
+# ourselves.
+_patient_rmmod()
+{
+	local module=$1
+	local max_tries_max=$MODPROBE_PATIENT_RM_TIMEOUT_SECONDS
+	local max_tries=0
+	local mod_ret=0
+	local refcnt_is_zero=0
+
+	if [[ ! -z $MODPROBE_REMOVE_PATIENT ]]; then
+		$MODPROBE_REMOVE_PATIENT $module
+		mod_ret=$?
+		if [[ $mod_ret -ne 0 ]]; then
+			echo "kmod patient module removal for $module timed out waiting for refcnt to become 0 using timeout of $max_tries_max returned $mod_ret"
+		fi
+		return $mod_ret
+	fi
+
+	max_tries=$max_tries_max
+
+	while [[ "$max_tries" != "0" ]]; do
+		_patient_rmmod_check_refcnt $module
+		if [[ $? -eq 0 ]]; then
+			refcnt_is_zero=1
+			break
+		fi
+		sleep 1
+		if [[ "$max_tries" == "forever" ]]; then
+			continue
+		fi
+		let max_tries=$max_tries-1
+	done
+
+	if [[ $refcnt_is_zero -ne 1 ]]; then
+		echo "custom patient module removal for $module timed out waiting for refcnt to become 0 using timeout of $max_tries_max"
+		return -1
+	fi
+
+	# If we ran out of time but our refcnt check confirms we had
+	# a refcnt of 0, just try to remove the module once.
+	if [[ "$max_tries" == "0" ]]; then
+		modprobe -r $module
+		return $?
+	fi
+
+	# If we have extra time left. Use the time left to now try to
+	# persistently remove the module. We do this because although through
+	# the above we found refcnt to be 0, removal can still fail since
+	# userspace can always race to bump the refcnt. An example is any
+	# blkdev_open() calls against a block device. These issues have been
+	# tracked and documented in the following bug reports, which justifies
+	# our need to do this in userspace:
+	# https://bugzilla.kernel.org/show_bug.cgi?id=212337
+	# https://bugzilla.kernel.org/show_bug.cgi?id=214015
+	while [[ $max_tries != 0 ]]; do
+		if [[ -d /sys/module/$module ]]; then
+			modprobe -r $module 2> /dev/null
+			mod_ret=$?
+			if [[ $mod_ret == 0 ]]; then
+				break;
+			fi
+			sleep 1
+			if [[ "$max_tries" == "forever" ]]; then
+				continue
+			fi
+			let max_tries=$max_tries-1
+		fi
+	done
+
+	if [[ $mod_ret -ne 0 ]]; then
+		echo "custom patient module removal for $module timed out trying to remove $module using timeout of $max_tries_max last try returned $mod_ret"
+	fi
+
+	return $mod_ret
+}
-- 
2.30.2

[PATCH v3 2/2] common/scsi_debug: use the patient module remover

[Luis Chamberlain]

If you try to run tests such as generic/108 in a loop
you'll eventually see a failure, but the failure can
be a false positive and the test was just unable to remove
the scsi_debug module.

We need to give some time for the refcnt to become 0. For
instance for the test generic/108 the refcnt lingers between
2 and 1. It should be 0 when we're done but a bit of time
seems to be required. The chance of us trying to run rmmod
when the refcnt is 2 or 1 is low, about 1/30 times if you
run the test in a loop on linux-next today.

Likewise, even when its 0 we just need a tiny breather before
we can remove the module (sleep 10 suffices) but this is
only required on older kernels. Otherwise removing the module
will just fail.

Some of these races are documented on the korg#212337, and
Doug Gilbert has posted at least one patch attempt to try
to help with this [1]. The patch does not resolve all the
issues though, it helps though.

[0] https://bugzilla.kernel.org/show_bug.cgi?id=212337
[1] https://lkml.kernel.org/r/20210508230745.27923-1-dgilbert@interlog.com
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
---
 common/scsi_debug | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/common/scsi_debug b/common/scsi_debug
index e7988469..3c9cd820 100644
--- a/common/scsi_debug
+++ b/common/scsi_debug
@@ -4,11 +4,13 @@
 #
 # Functions useful for tests on unique block devices
 
+. common/module
+
 _require_scsi_debug()
 {
 	# make sure we have the module and it's not already used
 	modinfo scsi_debug 2>&1 > /dev/null || _notrun "scsi_debug module not found"
-	lsmod | grep -wq scsi_debug && (rmmod scsi_debug || _notrun "scsi_debug module in use")
+	lsmod | grep -wq scsi_debug && (_patient_rmmod scsi_debug || _notrun "scsi_debug module in use")
 	# make sure it has the features we need
 	# logical/physical sectors plus unmap support all went in together
 	modinfo scsi_debug | grep -wq sector_size || _notrun "scsi_debug too old"
@@ -53,5 +55,5 @@ _put_scsi_debug_dev()
 		$UDEV_SETTLE_PROG
 		n=$((n-1))
 	done
-	rmmod scsi_debug || _fail "Could not remove scsi_debug module"
+	_patient_rmmod scsi_debug || _fail "Could not remove scsi_debug module"
 }
-- 
2.30.2