check: check reference count for modprobe --remove --wait success case

The commit "check,common/*: replace module removal with patient module
removal" introduced the new helper function _patient_rmmod() which calls
modprobe command with --wait option to do patient module removal.
However, the modprobe command can return a zero exit status even when
the module removal fails. In such cases, the failure remains unreported
and hidden. This behavior was observed during the execution of blktests
srp test group using rdma_rxe driver on a kernel affected by the
rdma_rxe module unload failure bug, which was addressed by the recent
patch [1].

To address this problem, check the reference count of the target module
after calling the modprobe command in _patient_rmmod(). If the module's
reference count indicates a removal failure, print an error message to
stderr. While at it, change the print target stream from stdout to
stderr for other error messages in _patient_rmmod() to ensure the
messages are printed on failure.

[1] https://lore.kernel.org/linux-rdma/[email protected]/

Reviewed-by: Bart Van Assche <[email protected]>
Reviewed-by: Daniel Wagner <[email protected]>
Signed-off-by: Shin'ichiro Kawasaki <[email protected]>

Author: kawasaki

check

@@ -530,7 +530,10 @@ _patient_rmmod()
 		modprobe --remove --wait="${timeout_ms}" "$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"
+			echo "kmod patient module removal for $module timed out waiting for refcnt to become 0 using timeout of $max_tries_max returned $mod_ret" >&2
+		elif ! _patient_rmmod_check_refcnt "$module_sys"; then
+			echo "modprobe with --wait option succeeded but still $module has references" >&2
+			mod_ret=1
 		fi
 		return $mod_ret
 	fi
@@ -544,7 +547,7 @@ _patient_rmmod()
 	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"
+		echo "custom patient module removal for $module timed out waiting for refcnt to become 0 using timeout of $max_tries_max" 2>&1
 		return 1
 	fi
 
@@ -575,7 +578,7 @@ _patient_rmmod()
 	done
 
 	if [[ $mod_ret -ne 0 ]]; then
-		echo "custom patient module removal for $module timed out trying to remove using timeout of $max_tries_max last try returned $mod_ret"
+		echo "custom patient module removal for $module timed out trying to remove using timeout of $max_tries_max last try returned $mod_ret" 2>&1
 	fi
 
 	return $mod_ret