Merge: nvmet: fix ns enable/disable possible hang

MR: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9/-/merge_requests/4444

# Merge Request Required Information

JIRA: https://issues.redhat.com/browse/RHEL-37676

## Summary of Changes


## Approved Development Ticket
All submissions to CentOS Stream must reference an approved ticket in [Red Hat Jira](https://issues.redhat.com/). Please follow the CentOS Stream [contribution documentation](https://docs.centos.org/en-US/stream-contrib/quickstart/) for how to file this ticket and have it approved.

Signed-off-by: Maurizio Lombardi <mlombard@redhat.com>

Approved-by: John Meneghini <jmeneghi@redhat.com>
Approved-by: Chris Leech <cleech@redhat.com>
Approved-by: CKI KWF Bot <cki-ci-bot+kwf-gitlab-com@redhat.com>
Approved-by: Tomas Henzl <thenzl@redhat.com>

Merged-by: Lucas Zampieri <lzampier@redhat.com>

Author: zampierilucas

drivers/nvme/target/configfs.c

@@ -676,10 +676,18 @@ static ssize_t nvmet_ns_enable_store(struct config_item *item,
 	if (kstrtobool(page, &enable))
 		return -EINVAL;
 
+	/*
+	 * take a global nvmet_config_sem because the disable routine has a
+	 * window where it releases the subsys-lock, giving a chance to
+	 * a parallel enable to concurrently execute causing the disable to
+	 * have a misaccounting of the ns percpu_ref.
+	 */
+	down_write(&nvmet_config_sem);
 	if (enable)
 		ret = nvmet_ns_enable(ns);
 	else
 		nvmet_ns_disable(ns);
+	up_write(&nvmet_config_sem);
 
 	return ret ? ret : count;
 }