On Nov 24 Huajun Li wrote:
> While unplugging usb disk, scsi_disk(disk)->device may be released
> before sd_revalidate_disk() is called, then there will occur Oops as
> shown below:
[...]
> --- a/drivers/scsi/sd.c
> +++ b/drivers/scsi/sd.c
> @@ -2354,10 +2354,15 @@ static int sd_try_extended_inquiry(struct
> scsi_device *sdp)
> static int sd_revalidate_disk(struct gendisk *disk)
> {
> struct scsi_disk *sdkp = scsi_disk(disk);
> - struct scsi_device *sdp = sdkp->device;
> + struct scsi_device *sdp;
> unsigned char *buffer;
> unsigned flush = 0;
>
> + if (sdkp)
> + sdp = sdkp->device;
> + else
> + goto out;
> +
> SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
> "sd_revalidate_disk\n"));
>
Shouldn't rather the [block -- command-set-driver -- scsi-core -- lld]
stack be structured along the lines that lower-level device instances live
as long as higher levels rely on them?
For about a year now or so, I am seeing patches floating by that add NULL
pointer checks here and there (or patches that clear pointers), and every
time I wonder
- where else such NULL pointer checks might be needed,
- in what way (if at all) it is ensured that a function which is made to
check for a valid pointer at the top does not race with pointer
invalidation further down the road.
--
Stefan Richter
-=====-==-== =-== ==-=-
http://arcgraph.de/sr/
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
