Re: 2.5.44: problemn when shutting down, drivers/base/power.c and the global_device_list

[Patrick Mansfield <patmans@us.ibm.com>]

On Sun, Oct 20, 2002 at 01:21:29PM -0700, Patrick Mochel wrote:
> 
> On Sat, 19 Oct 2002, Jurriaan wrote:
> 
> > DEV: registering device: ID = 'scsi0', name = sym53c8xx

Above is from scsi_set_pci_device() call.

> > scsi0 : sym53c8xx-1.7.3c-20010512
> > DEV: registering device: ID = 'scsi0', name = sym53c8xx

Above are from scsi_register_host().

> > sym53c860-0-<1,*>: FAST-20 SCSI 20.0 MB/s (50.0 ns, offset 8)
> >   Vendor: TOSHIBA   Model: DVD-ROM SD-M1401  Rev: 1007
> >   Type:   CD-ROM                             ANSI SCSI revision: 02

> 
> It appears the same device is being added twice. After doing some digging
> in the sym53c8xx and the scsi host code, that suspicion grows, though I'm 
> not positively sure how the host is being added twice. 
> 

Looks like we have two device_register(&shost->host_driverfs_dev) calls,
one in scsi_set_pci_device, and one in scsi_register_host(). I think 
Mike meant to delete the one in scsi_register_host, but I can't tell
for certain which one should go away.

And this must might be why I can't shutdown (i.e. no auto reboot).

So in 2.5.44 have:

	scsi_register_host() calls
		attach [ncr_attach] calls
			scsi_register
			scsi_set_pci_device calls
				device_register(&shost->host_driverfs_dev)
		call to device_register() again in scsi_register_host

This removes the scsi_register_host() one, I can shutdown/reboot
with this patch, I didn't dump anything else or try rmmod's:

--- bleed-2.5/drivers/scsi/hosts.c	Sat Oct 19 10:30:01 2002
+++ bleed-2.5/drivers/scsi/hosts.c-mine	Sun Oct 20 14:46:13 2002
@@ -543,7 +543,6 @@
 				       shost->host_no, dm_name);
 
 				/* first register parent with driverfs */
-				device_register(&shost->host_driverfs_dev);
 				scan_scsis(shost, 0, 0, 0, 0);
 			}
 		}

> scsi_register_host() then loops through the hosts, and registers all the 
> ones that this driver added. (Q: why is that function constantly looping 
> through all the scsi hosts just to operate on the one passed in?).

A Scsi_Host_Template is passed in, it loops over Scsi_Hosts to find
everyone that is using the Scsi_Host_Template, the names are confusing
as well as the code.

> The SCSI code is confusing and likely the culprit. 

Yes and yes ...

-- Patrick Mansfield