On Wednesday, August 15, 2012, Hans de Goede wrote:Hi,
On 08/15/2012 07:13 AM, Miklos Szeredi wrote:Suspend oopses in generic_ide_suspend() because dev_get_drvdata()
returns NULL (dev->p->driver_data == NULL) and this function is not
prepared for this.
I bisected it to 0998d063 (device-core: Ensure drvdata = NULL when no
driver is bound). Reverting it fixes suspend.
First of all, thanks for reporting and bisecting this. With that said,
I must say that this is very weird. The patch in question:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commitdiff;h=0998d063
Only makes dev-drvdata NULL in 2 cases:
1) The probe method of the driver fails
2) The driver has been detached from the device by calling one of:
device_release_driver() or driver_detach()
Note that in both code paths dev->driver also gets set to NULL, and
other generic ide driver callbacks very much depend on that not being
NULL, ie:
static int generic_ide_remove(struct device *dev)
{
ide_drive_t *drive = to_ide_device(dev);
struct ide_driver *drv = to_ide_driver(dev->driver);
if (drv->remove)
drv->remove(drive);
return 0;
}
Also how can a drivers suspend callback get called if dev->driver is NULL,
since that callback would normally be "reached" through dev->driver, so
something weird is going on here ...
No, it wouldn't, because it is a bus type callback and it is invoked for
all devices whose bus type is ide_bus_type, regardless of whether or not
their driver field is NULL.
It clearly should check if drive is not NULL before using that pointer.