Re: Another SCSI Sync

Re: Another SCSI Sync

[Linus Torvalds]

Hmm.. Some part of the last syncs seem to break for a lot of people. See
the linux kernel mailing list. One report (Rob Radez <rob@osinvestor.com>)  
chased it down some:

	On Tue, Mar 04, 2003 at 11:42:41PM -0600, David van Hoose wrote:
	> Kernels 2.5.63 and 2.5.64 lock up when displaying the information for my 
	> ZipDrive during SCSI probe. Attached is my .config for 2.5.64.

	I suspect I'm seeing something similar.  On SPARC, I'm getting a hang
	after displaying the information for my SCSI CD-ROM drive.

	scsi0 : Sparc ESP100A-FAST
	Vendor: SEAGATE   Model: ST31200W SUN1.05  Rev: 8724
	Type:   Direct-Access                      ANSI SCSI revision: 02
	Vendor: SEAGATE   Model: ST32155W SUN2.1G  Rev: 8456
	Type:   Direct-Access                      ANSI SCSI revision: 02
	Vendor: TOSHIBA   Model: XM-4101TASUNSLCD  Rev: 1755
	Type:   CD-ROM                             ANSI SCSI revision: 02
	<hang>

	Worked fine in 2.5.62, hangs in 2.5.63. Reverting the scsi changes
	listed at
		http://osinvestor.com/sparc/patch/2.5.63-revertscsi.diff
	allows 2.5.63 to boot.

Please check it out.. The common thread seems to be "non-disk".

		Linus

Re: Another SCSI Sync

[Mike Anderson]

I am looking at this now. I switched my config from new AIC to old AIC
and see a failure similar to what Andrew Morton was seeing when he switched.

It looks to be a bug in error handling while probing.

Linus Torvalds [torvalds@transmeta.com] wrote:
> 
> Hmm.. Some part of the last syncs seem to break for a lot of people. See
> the linux kernel mailing list. One report (Rob Radez <rob@osinvestor.com>)  
> chased it down some:
> 
> 	On Tue, Mar 04, 2003 at 11:42:41PM -0600, David van Hoose wrote:
> 	> Kernels 2.5.63 and 2.5.64 lock up when displaying the information for my 
> 	> ZipDrive during SCSI probe. Attached is my .config for 2.5.64.
> 
> 	I suspect I'm seeing something similar.  On SPARC, I'm getting a hang
> 	after displaying the information for my SCSI CD-ROM drive.
> 
> 	scsi0 : Sparc ESP100A-FAST
> 	Vendor: SEAGATE   Model: ST31200W SUN1.05  Rev: 8724
> 	Type:   Direct-Access                      ANSI SCSI revision: 02
> 	Vendor: SEAGATE   Model: ST32155W SUN2.1G  Rev: 8456
> 	Type:   Direct-Access                      ANSI SCSI revision: 02
> 	Vendor: TOSHIBA   Model: XM-4101TASUNSLCD  Rev: 1755
> 	Type:   CD-ROM                             ANSI SCSI revision: 02
> 	<hang>
> 
> 	Worked fine in 2.5.62, hangs in 2.5.63. Reverting the scsi changes
> 	listed at
> 		http://osinvestor.com/sparc/patch/2.5.63-revertscsi.diff
> 	allows 2.5.63 to boot.
> 
> Please check it out.. The common thread seems to be "non-disk".
> 
> 		Linus
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
-andmike
--
Michael Anderson
andmike@us.ibm.com

Re: Another SCSI Sync

[Mike Anderson]

The patch below fixed the problem on my system. I had my list empty
checks reversed if aborting and bus device reset failed. The condition
that causes the error handler to run is still unknown. I will look at it
when I get a chance.

The patch applies to 2.5.63, 2.5.64 and scsi-misc-2.5

Mike Anderson [andmike@us.ibm.com] wrote:
> I am looking at this now. I switched my config from new AIC to old AIC
> and see a failure similar to what Andrew Morton was seeing when he switched.
> 
> It looks to be a bug in error handling while probing.
> 
> Linus Torvalds [torvalds@transmeta.com] wrote:
> > 
> > Hmm.. Some part of the last syncs seem to break for a lot of people. See
> > the linux kernel mailing list. One report (Rob Radez <rob@osinvestor.com>)  
> > chased it down some:
> > 
> > 	On Tue, Mar 04, 2003 at 11:42:41PM -0600, David van Hoose wrote:
> > 	> Kernels 2.5.63 and 2.5.64 lock up when displaying the information for my 
> > 	> ZipDrive during SCSI probe. Attached is my .config for 2.5.64.
> > 
> > 	I suspect I'm seeing something similar.  On SPARC, I'm getting a hang
> > 	after displaying the information for my SCSI CD-ROM drive.
> > 
> > 	scsi0 : Sparc ESP100A-FAST
> > 	Vendor: SEAGATE   Model: ST31200W SUN1.05  Rev: 8724
> > 	Type:   Direct-Access                      ANSI SCSI revision: 02
> > 	Vendor: SEAGATE   Model: ST32155W SUN2.1G  Rev: 8456
> > 	Type:   Direct-Access                      ANSI SCSI revision: 02
> > 	Vendor: TOSHIBA   Model: XM-4101TASUNSLCD  Rev: 1755
> > 	Type:   CD-ROM                             ANSI SCSI revision: 02
> > 	<hang>
> > 
> > 	Worked fine in 2.5.62, hangs in 2.5.63. Reverting the scsi changes
> > 	listed at
> > 		http://osinvestor.com/sparc/patch/2.5.63-revertscsi.diff
> > 	allows 2.5.63 to boot.
> > 
> > Please check it out.. The common thread seems to be "non-disk".
> > 
> > 		Linus
-andmike
--
Michael Anderson
andmike@us.ibm.com


=====
name:		00_scsi_error_ready_devs-1.diff
version:	2003-03-05.10:39:28-0800
against:	2.5.63

 scsi_error.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

=====
===== drivers/scsi/scsi_error.c 1.38 vs edited =====
--- 1.38/drivers/scsi/scsi_error.c	Sat Feb 22 08:17:01 2003
+++ edited/drivers/scsi/scsi_error.c	Wed Mar  5 10:14:22 2003
@@ -1490,9 +1490,9 @@
 			       struct list_head *work_q,
 			       struct list_head *done_q)
 {
-	if (scsi_eh_bus_device_reset(shost, work_q, done_q))
-		if (scsi_eh_bus_reset(shost, work_q, done_q))
-			if (scsi_eh_host_reset(work_q, done_q))
+	if (!scsi_eh_bus_device_reset(shost, work_q, done_q))
+		if (!scsi_eh_bus_reset(shost, work_q, done_q))
+			if (!scsi_eh_host_reset(work_q, done_q))
 				scsi_eh_offline_sdevs(work_q, done_q);
 }
 

Re: another SCSI sync

[Linus Torvalds]

This fails with 

	drivers/built-in.o(.init.text+0x88df): In function `init_scsi':
	: undefined reference to `scsi_exit_procfs'
	drivers/built-in.o(.init.text+0x88e4): In function `init_scsi':
	: undefined reference to `scsi_exit_queue'
	make: *** [.tmp_vmlinux1] Error 1

for me.

They are marked __exit, which means that they end up being marked unused
_and_ in a segment that doesn't ever get loaded for built-in stuff (since 
clearly the built-in kernel can never unload).

In general, it is just illegal to call an __exit function from something 
that isn't __exit itself. And these functions are called from __init 
function too.

Please fix. In the meantime, my fix is to remove the two bogus "__exit" 
notifiers, and wondering to myself how the hell this ever worked for any 
of you guys.

		Linus

Re: another SCSI sync

[Christoph Hellwig]

On Sat, Feb 22, 2003 at 12:34:46PM -0800, Linus Torvalds wrote:
> They are marked __exit, which means that they end up being marked unused
> _and_ in a segment that doesn't ever get loaded for built-in stuff (since 
> clearly the built-in kernel can never unload).
> 
> In general, it is just illegal to call an __exit function from something 
> that isn't __exit itself. And these functions are called from __init 
> function too.
> 
> Please fix. In the meantime, my fix is to remove the two bogus "__exit" 
> notifiers, and wondering to myself how the hell this ever worked for any 
> of you guys.

Okay, this was my fault.  It seems the old toolchain on SuSE 8.0
(gcc 2.95-based) doesn't seem to notice this yet.  Your fix is correct,
there's no other fixups needed.