`comedi_device_cleanup()` is called just before freeing a comedi device.
It is possible for the device to still be open in which case the module
reference counts for the core comedi module and possibly the low-level
driver module will remain out of whack because `comedi_close()` will not
find the comedi device and so will not decrement the module counts.
This really needs to be handled better, but for now decrement the module
counts in `comedi_device_cleanup()` according to the number of
outstanding opens.
Signed-off-by: Ian Abbott <abbo...@mev.co.uk>
---
drivers/staging/comedi/comedi_fops.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/drivers/staging/comedi/comedi_fops.c
b/drivers/staging/comedi/comedi_fops.c
index f2bfa0e..70b2034 100644
--- a/drivers/staging/comedi/comedi_fops.c
+++ b/drivers/staging/comedi/comedi_fops.c
@@ -2256,10 +2256,20 @@ static void comedi_device_init(struct comedi_device
*dev)
static void comedi_device_cleanup(struct comedi_device *dev)
{
+ struct module *driver_module = NULL;
+
if (dev == NULL)
return;
mutex_lock(&dev->mutex);
+ if (dev->attached)
+ driver_module = dev->driver->module;
comedi_device_detach(dev);
+ while (dev->use_count > 0) {
+ if (driver_module)
+ module_put(driver_module);
+ module_put(THIS_MODULE);
+ dev->use_count--;
+ }
mutex_unlock(&dev->mutex);
mutex_destroy(&dev->mutex);
}
--
1.8.1.5
_______________________________________________
devel mailing list
devel@linuxdriverproject.org
http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
