ub: oops with preempt ("Sahara Workshop")

ub: oops with preempt ("Sahara Workshop")

[Pete Zaitcev]

I admit that the code should be locked properly instead, but the global plan
is to drop all P3 tagged printks anyway. So let it be guarded for the moment.

Signed-off-by: Pete Zaitcev <zaitcev@yahoo.com>

--- linux-2.6.10-rc2-bk8-ub/drivers/block/ub.c	2004-11-16 17:03:02.000000000 -0800
+++ linux-2.6.10-rc1-ub/drivers/block/ub.c	2004-11-07 19:01:03.000000000 -0800
@@ -1535,8 +1535,11 @@
 
 	ub_revalidate(sc);
 	/* This is pretty much a long term P3 */
-	printk(KERN_INFO "%s: device %u capacity nsec %ld bsize %u\n",
-	    sc->name, sc->dev->devnum, sc->capacity.nsec, sc->capacity.bsize);
+	if (!atomic_read(&sc->poison)) {		/* Cover sc->dev */
+		printk(KERN_INFO "%s: device %u capacity nsec %ld bsize %u\n",
+		    sc->name, sc->dev->devnum,
+		    sc->capacity.nsec, sc->capacity.bsize);
+	}
 
 	/* XXX Support sector size switching like in sr.c */
 	blk_queue_hardsect_size(disk->queue, sc->capacity.bsize);

Re: ub: oops with preempt ("Sahara Workshop")

[Greg KH]

On Tue, Nov 23, 2004 at 10:02:47AM -0800, Pete Zaitcev wrote:
> I admit that the code should be locked properly instead, but the global plan
> is to drop all P3 tagged printks anyway. So let it be guarded for the moment.
> 
> Signed-off-by: Pete Zaitcev <zaitcev@yahoo.com>

Applied, thanks.

greg k-h

Re: ub: oops with preempt ("Sahara Workshop") [u]

[Martin Schlemmer [c]]

[-- Attachment #1: Type: text/plain, Size: 613 bytes --]

On Tue, 2004-11-23 at 10:02 -0800, Pete Zaitcev wrote:

Hi,

> I admit that the code should be locked properly instead, but the global plan
> is to drop all P3 tagged printks anyway. So let it be guarded for the moment.
> 

Sorry for the delay, but I have not had any time to really test this
again.  I did some minor testing, and only after really working it,
I could get an oops, but not nearly the same (think it was deeper into
the scsi layer or maybe kobject stuff).

Will see when I can get some time to try and generate some consistent
traces if any.


Thanks,

-- 
Martin Schlemmer


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: ub: oops with preempt ("Sahara Workshop") [u]

[Pete Zaitcev]

On Sat, 27 Nov 2004 23:38:15 +0200, "Martin Schlemmer [c]" <azarah@nosferatu.za.org> wrote:
> On Tue, 2004-11-23 at 10:02 -0800, Pete Zaitcev wrote:

> > I admit that the code should be locked properly instead, but the global plan
> > is to drop all P3 tagged printks anyway. So let it be guarded for the moment.

> Sorry for the delay, but I have not had any time to really test this
> again.  I did some minor testing, and only after really working it,
> I could get an oops, but not nearly the same (think it was deeper into
> the scsi layer or maybe kobject stuff).

This is strange, because ub hasn't got any connection with SCSI layer.

Without trying to deflect the blame for improper locking in ub, I have to
ask, does it all work without the preempt for you?

-- Pete