[BUG] linux-2.6.0-test9: slave_destroy called too early

[BUG] linux-2.6.0-test9: slave_destroy called too early

[Heiko Carstens]

Hi,

in linux-2.6.0-test9 there seems to be a bug in scsi_remove_device: each time
I try to remove a scsi disk the sd driver fails to synchronize the SCSI cache.
Reason for this is that the slave_destroy function of the LLD gets called
before sd_shutdown of the sd driver gets called.
I would suggest the patch below. Please apply or let me know that this is the
wrong thing to do.

Thanks,
Heiko

diff -urN linux-2.6.0-test9/drivers/scsi/scsi_sysfs.c 
linux-2.6.0-test9-sd/drivers/scsi/scsi_sysfs.c
--- linux-2.6.0-test9/drivers/scsi/scsi_sysfs.c	Sat Oct 25 20:43:04 2003
+++ linux-2.6.0-test9-sd/drivers/scsi/scsi_sysfs.c	Tue Nov 11 11:43:27 2003
@@ -410,9 +410,9 @@
 	if (class) {
 		down_write(&class->subsys.rwsem);
 		set_bit(SDEV_DEL, &sdev->sdev_state);
+		device_del(&sdev->sdev_gendev);
 		if (sdev->host->hostt->slave_destroy)
 			sdev->host->hostt->slave_destroy(sdev);
-		device_del(&sdev->sdev_gendev);
 		up_write(&class->subsys.rwsem);
 	}

Re: [BUG] linux-2.6.0-test9: slave_destroy called too early

[Mike Anderson]

Heiko Carstens [heiko.carstens@de.ibm.com] wrote:
> Hi,
> 
> in linux-2.6.0-test9 there seems to be a bug in scsi_remove_device: each time
> I try to remove a scsi disk the sd driver fails to synchronize the SCSI cache.
> Reason for this is that the slave_destroy function of the LLD gets called
> before sd_shutdown of the sd driver gets called.
> I would suggest the patch below. Please apply or let me know that this is the
> wrong thing to do.
> 

The patch I previously posted would give the same effect. You can find
it in the archives.
http://marc.theaimsgroup.com/?l=linux-scsi&m=106810957502435&w=2

I will repost the patch with a update based on the feedback I received
on the previous thread (inline the scsi_free_sdev function).  This
updated patch will look very similiar to the sequence you have in your
patch.

-andmike
--
Michael Anderson
andmike@us.ibm.com