rescan bug on 2.5.70 ?

rescan bug on 2.5.70 ?

[Steven Dake]

Folks,

I attempted
echo "1" > rescan within a device in the scsi sysfs tree on 2.5.70.

The kernel repeatadly printed out the SDA device information.  The 
system was still usable, but it appeared the rescan code was stuck in a 
loop which dumped printks continuously.  Only way out was a reset.

Anyone else seen this behavior ?

Thanks
-steve

Re: rescan bug on 2.5.70 ?

[Patrick Mansfield]

On Fri, May 30, 2003 at 02:08:23PM -0700, Steven Dake wrote:
> Folks,
> 
> I attempted
> echo "1" > rescan within a device in the scsi sysfs tree on 2.5.70.
> 
> The kernel repeatadly printed out the SDA device information.  The 
> system was still usable, but it appeared the rescan code was stuck in a 
> loop which dumped printks continuously.  Only way out was a reset.
> 
> Anyone else seen this behavior ?

Yes, I tried it out, and hit the same thing. The user program (not sure
what C code) is retrying the write, since we return 0. You can kill the
program, though I couldn't ^C/intr the echo, I killed it from another
terminal.

Trivial patch against current scsi-misc-2.5:

===== drivers/scsi/scsi_sysfs.c 1.21 vs edited =====
--- 1.21/drivers/scsi/scsi_sysfs.c	Mon May 26 08:45:44 2003
+++ edited/drivers/scsi/scsi_sysfs.c	Sat May 31 00:04:45 2003
@@ -204,7 +204,7 @@
 store_rescan_field (struct device *dev, const char *buf, size_t count) 
 {
 	scsi_rescan_device(dev);
-	return 0;
+	return count;
 }
 
 static DEVICE_ATTR(rescan, S_IRUGO | S_IWUSR, show_rescan_field, store_rescan_field)

-- Patrick Mansfield