James Bottomley wrote:
> On Tue, 2004-10-26 at 14:08, Mike Christie wrote:
> 
>>>The null state and and oops are becuase of this
>>>http://marc.theaimsgroup.com/?l=linux-scsi&m=109733573729283&w=2
>>
>>Oh yeah. that patch is not correct, but if you correctly modify it to
>>use device_for_each_child per Christoph's suggestion, I seem to be
>>getting some refcounting errors. For some reason the sdev will be
>>released, but the sd.c still thinks it is there.
> 
> 
> Actually, he suggested using shost_for_each_device.  The reason being
> that you can't have nested device_for_each_child (because it takes the
> bus semaphore).
> 
> The attached should do this, if someone would care to try it out.
> 
> James
> 
> ===== hosts.c 1.102 vs edited =====
> --- 1.102/drivers/scsi/hosts.c	2004-10-11 10:03:45 -05:00
> +++ edited/hosts.c	2004-10-26 20:49:51 -05:00
> @@ -50,11 +50,6 @@
>  	.release	= scsi_host_cls_release,
>  };
>  
> -static int scsi_device_cancel_cb(struct device *dev, void *data)
> -{
> -	return scsi_device_cancel(to_scsi_device(dev), *(int *)data);
> -}
> -
>  /**
>   * scsi_host_cancel - cancel outstanding IO to this host
>   * @shost:	pointer to struct Scsi_Host
> @@ -62,9 +57,12 @@
>   **/
>  void scsi_host_cancel(struct Scsi_Host *shost, int recovery)
>  {
> +	struct scsi_device *sdev;
> +
>  	set_bit(SHOST_CANCEL, &shost->shost_state);
> -	device_for_each_child(&shost->shost_gendev, &recovery,
> -			      scsi_device_cancel_cb);
> +	shost_for_each_device(sdev, shost) {
> +		scsi_device_cancel(sdev, recovery);
> +	}
>  	wait_event(shost->host_wait, (!test_bit(SHOST_RECOVERY,
>  						&shost->shost_state)));
>  }
> 

James,
I have been seeing an oops (or badness) since about the lk 2.5
series if I left my USB Sandisk CF (ImageMate) plug in with a
mounted file system open when I halted the machine. As the logs
had already been closed down there was nothing captured.

So I thought your patch was worth a try.
This test was run on lk 2.6.10-rc1 with the above patch applied.
The result is attached.
To capture this I did the following:
   - comnected the ImageMate [became 4:0:0:0]
   - mounted a fs on it
   - echo 1 > /sys/class/scsi_device/4\:0\:0\:0/device/delete
      [lsscsi showed the device was gone after this]
   - unplugged the ImageMate (and "badness" occurred)

Doug Gilbert