[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20210520092120.166419561@linuxfoundation.org>
Date: Thu, 20 May 2021 11:18:13 +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, Shuah Khan <skhan@...uxfoundation.org>,
Tom Seewald <tseewald@...il.com>,
syzbot+a93fba6d384346a761e3@...kaller.appspotmail.com
Subject: [PATCH 4.14 001/323] usbip: vudc synchronize sysfs code paths
From: Shuah Khan <skhan@...uxfoundation.org>
commit bd8b82042269a95db48074b8bb400678dbac1815 upstream.
Fuzzing uncovered race condition between sysfs code paths in usbip
drivers. Device connect/disconnect code paths initiated through
sysfs interface are prone to races if disconnect happens during
connect and vice versa.
Use sysfs_lock to protect sysfs paths in vudc.
Cc: stable@...r.kernel.org
Reported-and-tested-by: syzbot+a93fba6d384346a761e3@...kaller.appspotmail.com
Signed-off-by: Shuah Khan <skhan@...uxfoundation.org>
Link: https://lore.kernel.org/r/caabcf3fc87bdae970509b5ff32d05bb7ce2fb15.1616807117.git.skhan@linuxfoundation.org
Signed-off-by: Tom Seewald <tseewald@...il.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
---
drivers/usb/usbip/vudc_dev.c | 1 +
drivers/usb/usbip/vudc_sysfs.c | 5 +++++
2 files changed, 6 insertions(+)
--- a/drivers/usb/usbip/vudc_dev.c
+++ b/drivers/usb/usbip/vudc_dev.c
@@ -586,6 +586,7 @@ static int init_vudc_hw(struct vudc *udc
init_waitqueue_head(&udc->tx_waitq);
spin_lock_init(&ud->lock);
+ mutex_init(&ud->sysfs_lock);
ud->status = SDEV_ST_AVAILABLE;
ud->side = USBIP_VUDC;
--- a/drivers/usb/usbip/vudc_sysfs.c
+++ b/drivers/usb/usbip/vudc_sysfs.c
@@ -125,6 +125,7 @@ static ssize_t store_sockfd(struct devic
dev_err(dev, "no device");
return -ENODEV;
}
+ mutex_lock(&udc->ud.sysfs_lock);
spin_lock_irqsave(&udc->lock, flags);
/* Don't export what we don't have */
if (!udc->driver || !udc->pullup) {
@@ -200,6 +201,8 @@ static ssize_t store_sockfd(struct devic
wake_up_process(udc->ud.tcp_rx);
wake_up_process(udc->ud.tcp_tx);
+
+ mutex_unlock(&udc->ud.sysfs_lock);
return count;
} else {
@@ -220,6 +223,7 @@ static ssize_t store_sockfd(struct devic
}
spin_unlock_irqrestore(&udc->lock, flags);
+ mutex_unlock(&udc->ud.sysfs_lock);
return count;
@@ -229,6 +233,7 @@ unlock_ud:
spin_unlock_irq(&udc->ud.lock);
unlock:
spin_unlock_irqrestore(&udc->lock, flags);
+ mutex_unlock(&udc->ud.sysfs_lock);
return ret;
}
Powered by blists - more mailing lists