linux-kernel - [PATCH 4.19 125/131] UAS: fix deadlock in error handling and PM flushing work

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <20200428182240.988236185@linuxfoundation.org>
Date:   Tue, 28 Apr 2020 20:25:37 +0200
From:   Greg Kroah-Hartman <gregkh@...uxfoundation.org>
To:     linux-kernel@...r.kernel.org
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        stable@...r.kernel.org, Oliver Neukum <oneukum@...e.com>
Subject: [PATCH 4.19 125/131] UAS: fix deadlock in error handling and PM flushing work

From: Oliver Neukum <oneukum@...e.com>

commit f6cc6093a729ede1ff5658b493237c42b82ba107 upstream.

A SCSI error handler and block runtime PM must not allocate
memory with GFP_KERNEL. Furthermore they must not wait for
tasks allocating memory with GFP_KERNEL.
That means that they cannot share a workqueue with arbitrary tasks.

Fix this for UAS using a private workqueue.

Signed-off-by: Oliver Neukum <oneukum@...e.com>
Fixes: f9dc024a2da1f ("uas: pre_reset and suspend: Fix a few races")
Cc: stable <stable@...r.kernel.org>
Link: https://lore.kernel.org/r/20200415141750.811-2-oneukum@suse.com
Signed-off-by: Greg Kroah-Hartman <gregkh@...uxfoundation.org>

---
 drivers/usb/storage/uas.c |   43 ++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 40 insertions(+), 3 deletions(-)

--- a/drivers/usb/storage/uas.c
+++ b/drivers/usb/storage/uas.c
@@ -81,6 +81,19 @@ static void uas_free_streams(struct uas_
 static void uas_log_cmd_state(struct scsi_cmnd *cmnd, const char *prefix,
 				int status);
 
+/*
+ * This driver needs its own workqueue, as we need to control memory allocation.
+ *
+ * In the course of error handling and power management uas_wait_for_pending_cmnds()
+ * needs to flush pending work items. In these contexts we cannot allocate memory
+ * by doing block IO as we would deadlock. For the same reason we cannot wait
+ * for anything allocating memory not heeding these constraints.
+ *
+ * So we have to control all work items that can be on the workqueue we flush.
+ * Hence we cannot share a queue and need our own.
+ */
+static struct workqueue_struct *workqueue;
+
 static void uas_do_work(struct work_struct *work)
 {
 	struct uas_dev_info *devinfo =
@@ -109,7 +122,7 @@ static void uas_do_work(struct work_stru
 		if (!err)
 			cmdinfo->state &= ~IS_IN_WORK_LIST;
 		else
-			schedule_work(&devinfo->work);
+			queue_work(workqueue, &devinfo->work);
 	}
 out:
 	spin_unlock_irqrestore(&devinfo->lock, flags);
@@ -134,7 +147,7 @@ static void uas_add_work(struct uas_cmd_
 
 	lockdep_assert_held(&devinfo->lock);
 	cmdinfo->state |= IS_IN_WORK_LIST;
-	schedule_work(&devinfo->work);
+	queue_work(workqueue, &devinfo->work);
 }
 
 static void uas_zap_pending(struct uas_dev_info *devinfo, int result)
@@ -1236,7 +1249,31 @@ static struct usb_driver uas_driver = {
 	.id_table = uas_usb_ids,
 };
 
-module_usb_driver(uas_driver);
+static int __init uas_init(void)
+{
+	int rv;
+
+	workqueue = alloc_workqueue("uas", WQ_MEM_RECLAIM, 0);
+	if (!workqueue)
+		return -ENOMEM;
+
+	rv = usb_register(&uas_driver);
+	if (rv) {
+		destroy_workqueue(workqueue);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static void __exit uas_exit(void)
+{
+	usb_deregister(&uas_driver);
+	destroy_workqueue(workqueue);
+}
+
+module_init(uas_init);
+module_exit(uas_exit);
 
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR(