On Tue, 4 Mar 2008, Stefan Richter wrote:
> Alan Stern wrote:
> > So I need a
> > way to know whether anything is mounted on a device below a SCSI host.
>
> Actually I have been thinking of something related for a while but
> didn't send a patch yet. Consider this:
>
> [include/scsi/scsi_host.h]
>
> struct scsi_host_template {
> ...
> + int (* slave_sdev_get)(struct scsi_device *);
> + void (* slave_sdev_put)(struct scsi_device *);
> ...
> }
>
> [drivers/scsi/scsi.c]
>
> int scsi_device_get(struct scsi_device *sdev)
> {
> + int err = 0;
> +
> if (sdev->sdev_state == SDEV_DEL)
> return -ENXIO;
> if (!get_device(&sdev->sdev_gendev))
> return -ENXIO;
> /* We can fail this if we're doing SCSI operations
> * from module exit (like cache flush) */
> try_module_get(sdev->host->hostt->module);
>
> - return 0;
> + if (sdev->host->hostt->slave_sdev_get)
> + err = sdev->host->hostt->slave_sdev_get(sdev);
> + if (err) {
> + ...some ifdef'd module_put stuff...;
> + put_device(&sdev->sdev_gendev);
> + }
> + return err;
> }
>
> and equivalent additions to scsi_device_put().
That would be okay. It does a little more than I need, but the
difference doesn't matter.
> Example usage:
>
> [drivers/ieee1394/sbp2.c]
>
> +int sbp2_slave_sdev_get(struct scsi_device *sdev)
> +{
> + struct sbp2_lu *lu = (struct sbp2_lu *)sdev->host->hostdata[0];
> +
> + if (!try_module_get(lu->hi->host->driver->owner))
> + return -ENOMEM;
> + return 0;
> +}
>
> and an equivalent sbp2_slave_sdev_put(). These examples would prevent
> unloading ohci1394 while somebody has mounted a filesystem on a device
> driven by sbp2, or otherwise opened or referenced that device. (At the
> moment sbp2 unintelligently gets the host->driver->owner as long as it
> is logged in into a device. This is somewhat necessary because there
> are unrelated drivers in the ieee1394 stack which when unloaded with
> modprobe would cause ohci1394 to be unloaded too.)
>
> You could use slave_sdev_get()/ slave_sdev_put() callbacks in
> usb-storage to do an own bookkeeping of references there.
>
> (Since I haven't looked at your other current work, all of the above may
> be utterly besides the point...)
It's a good idea. I'd be willing to use it, if you submit it.
Alan Stern
--
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
