Re: ide-scsi + kobject Badness in 2.5.52

Patrick Mochel (mochel@osdl.org)
Fri, 20 Dec 2002 11:46:21 -0600 (CST)


[linux-kernel cc'ed]

> Dec 19 15:13:19 gargoyle kernel: Badness in kobject_register at lib/kobject.c:113
> Dec 19 15:13:19 gargoyle kernel: Call Trace:
> Dec 19 15:13:19 gargoyle kernel: [<f887fc7c>] idescsi_driver+0x7c/0xd0 [ide_scsi]
> Dec 19 15:13:19 gargoyle kernel: [kobject_register+58/80] kobject_register+0x3a/0x50
> Dec 19 15:13:19 gargoyle kernel: [bus_add_driver+108/192] bus_add_driver+0x6c/0xc0
> Dec 19 15:13:19 gargoyle kernel: [<f887fc6c>] idescsi_driver+0x6c/0xd0 [ide_scsi]
> Dec 19 15:13:19 gargoyle kernel: [<f887fc6c>] idescsi_driver+0x6c/0xd0 [ide_scsi]
> Dec 19 15:13:19 gargoyle kernel: [ide_drive_remove+0/48] ide_drive_remove+0x0/0x30
> Dec 19 15:13:19 gargoyle kernel: [<f887fc00>] idescsi_driver+0x0/0xd0 [ide_scsi]
> Dec 19 15:13:19 gargoyle kernel: [ide_register_driver+218/240] ide_register_driver+0xda/0xf0
> Dec 19 15:13:19 gargoyle kernel: [<f887fc4c>] idescsi_driver+0x4c/0xd0 [ide_scsi]
> Dec 19 15:13:19 gargoyle kernel: [<f888200d>] +0xd/0x20 [ide_scsi]
> Dec 19 15:13:19 gargoyle kernel: [<f887fc00>] idescsi_driver+0x0/0xd0 [ide_scsi]
> Dec 19 15:13:19 gargoyle kernel: [sys_init_module+272/416] sys_init_module+0x110/0x1a0
> Dec 19 15:13:19 gargoyle kernel: [syscall_call+7/11] syscall_call+0x7/0xb

This is a multi-part problem.

For one, ide_unregister_driver() was not calling driver_unregister(), so
the driver's sysfs directory was never getting deleted when the module was
removed the first time. Everything still should have worked, but it was
still incorrect.

Once that was resolved, I started hitting a BUG() in
ide_unregister_driver() because the drive was busy. I also noticed that I
now had _two_ pseudo-scsi devices, instead of just one. It turns out that
idescsi_release(), the tear-down method for SCSI devices was not
unregistering the pseudo host that idescsi_detect() was creating.

The two one-liners below should fix those problems.

However, the question still remains if ide-scsi is relevant any more?

-pat

===== drivers/ide/ide.c 1.44 vs edited =====
--- 1.44/drivers/ide/ide.c Tue Dec 3 12:01:26 2002
+++ edited/drivers/ide/ide.c Fri Dec 20 11:36:40 2002
@@ -2317,6 +2317,8 @@
spin_lock(&drivers_lock);
list_del(&driver->drivers);
spin_unlock(&drivers_lock);
+
+ driver_unregister(&driver->gen_driver);

while(!list_empty(&driver->drives)) {
drive = list_entry(driver->drives.next, ide_drive_t, list);
===== drivers/scsi/ide-scsi.c 1.18 vs edited =====
--- 1.18/drivers/scsi/ide-scsi.c Sat Nov 30 07:44:26 2002
+++ edited/drivers/scsi/ide-scsi.c Fri Dec 20 11:37:49 2002
@@ -658,6 +658,7 @@
if (drive)
DRIVER(drive)->busy--;
}
+ scsi_unregister(host);
return 0;
}

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/