On Thu, 2012-05-17 at 16:51 +0200, Tomas Henzl wrote:
> >>> diff --git a/drivers/base/driver.c b/drivers/base/driver.c
> >>> index 3ec3896..246e652 100644
> >>> --- a/drivers/base/driver.c
> >>> +++ b/drivers/base/driver.c
> >>> @@ -10,6 +10,7 @@
> >>> *
> >>> */
> >>>
> >>> +#include <linux/async.h>
> >>> #include <linux/device.h>
> >>> #include <linux/module.h>
> >>> #include <linux/errno.h>
> >>> @@ -199,6 +200,13 @@ EXPORT_SYMBOL_GPL(driver_register);
> >>> */
> >>> void driver_unregister(struct device_driver *drv)
> >>> {
> >>> + /*
> >>> + * make sure all async bits that may be running have completed the way
> >>> + * this works is that try_module_get() now fails, so new async work
> >>> + * should take a reference to the module which now fails, so it should
> >>> + * be safe to remove the driver.
> >>> + */
> >>> + async_synchronize_full();
> >>> if (!drv || !drv->p) {
> >>> WARN(1, "Unexpected driver unregister!\n");
> >>> return;
> Is this safe? I mean, is it possible that a new scan starts after the
> async_synchronize_full() ends?
Not really, because the module is now removing, so the try module get
guard fails
> I wish this were fixed by fixing the real root of this problem which lies
> in scssi_device_get:
> int scsi_device_get(struct scsi_device *sdev)
> {
> if (sdev->sdev_state == SDEV_DEL)
> return -ENXIO;
> if (!get_device(&sdev->sdev_gendev))
> return -ENXIO;
> /* We can fail this if we're doing SCSI operations
> * from module exit (like cache flush) */
> try_module_get(sdev->host->hostt->module);
> here ^ by ignoring the return value. This is why rmmod --wait causes the problem
> and likely every other process which calls scsi_device_get would show the same bug.
I'm not sure how ... this has to be usable in the teardown path by which
point try_module_get will fail (as the comment says). What do you
propose?
James
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
