[PATCH 4.7 138/141] Revert "usbtmc: convert to devm_kzalloc"

From: Greg Kroah-Hartman
Date: Thu Oct 06 2016 - 04:49:09 EST

4.7-stable review patch. If anyone has any objections, please let me know.


From: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>

commit ab21b63e8aedfc73565dd9cdd51eb338341177cb upstream.

This reverts commit e6c7efdcb76f11b04e3d3f71c8d764ab75c9423b.

Turns out it was totally wrong. The memory is supposed to be bound to
the kref, as the original code was doing correctly, not the
device/driver binding as the devm_kzalloc() would cause.

This fixes an oops when read would be called after the device was
unbound from the driver.

Reported-by: Ladislav Michl <ladis@xxxxxxxxxxxxxx>
Cc: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>

drivers/usb/class/usbtmc.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

--- a/drivers/usb/class/usbtmc.c
+++ b/drivers/usb/class/usbtmc.c
@@ -141,6 +141,7 @@ static void usbtmc_delete(struct kref *k
struct usbtmc_device_data *data = to_usbtmc_data(kref);

+ kfree(data);

static int usbtmc_open(struct inode *inode, struct file *filp)
@@ -1379,7 +1380,7 @@ static int usbtmc_probe(struct usb_inter

dev_dbg(&intf->dev, "%s called\n", __func__);

- data = devm_kzalloc(&intf->dev, sizeof(*data), GFP_KERNEL);
+ data = kmalloc(sizeof(*data), GFP_KERNEL);
if (!data)
return -ENOMEM;