On Tue, May 04, 2021 at 02:00:01PM +0300, Dan Carpenter wrote:
> On Tue, May 04, 2021 at 12:07:44PM +0200, Fabio M. De Francesco wrote:
> > -static void setup_scsitaskmgmt_handles(struct idr *idrtable, spinlock_t *lock,
> > - struct uiscmdrsp *cmdrsp,
> > +static int setup_scsitaskmgmt_handles(struct xarray *xa, struct uiscmdrsp *cmdrsp,
> > wait_queue_head_t *event, int *result)
> > {
> > - /* specify the event that has to be triggered when this */
> > - /* cmd is complete */
> > - cmdrsp->scsitaskmgmt.notify_handle =
> > - simple_idr_get(idrtable, event, lock);
> > - cmdrsp->scsitaskmgmt.notifyresult_handle =
> > - simple_idr_get(idrtable, result, lock);
> > + int ret;
> > + u32 id;
> > +
> > + /* specify the event that has to be triggered when this cmd is complete */
> > + ret = xa_alloc_irq(xa, &id, event, xa_limit_32b, GFP_KERNEL);
> > + if (ret == 0)
> > + cmdrsp->scsitaskmgmt.notify_handle = id;
> > + else
> > + return ret;
>
> Reverse the condition. Always do error handling, not success handling.
> Try keep the success path at one tab indent and the failure path at two
> tab indents.
>
> ret = xa_alloc_irq(xa, &id, event, xa_limit_32b, GFP_KERNEL);
> if (ret)
> return ret;
> cmdrsp->scsitaskmgmt.notify_handle = id;
>
> ret = xa_alloc_irq(xa, &id, result, xa_limit_32b, GFP_KERNEL);
> if (ret) {
> xa_erase(xa, cmdrsp->scsitaskmgmt.notify_handle);
umm. When you put these things right next to each other, it's obvious
that should be xa_erase_irq().
> > -static void cleanup_scsitaskmgmt_handles(struct idr *idrtable,
> > +static void cleanup_scsitaskmgmt_handles(struct xarray *xa,
> > struct uiscmdrsp *cmdrsp)
> > {
> > - if (cmdrsp->scsitaskmgmt.notify_handle)
> > - idr_remove(idrtable, cmdrsp->scsitaskmgmt.notify_handle);
> > - if (cmdrsp->scsitaskmgmt.notifyresult_handle)
> > - idr_remove(idrtable, cmdrsp->scsitaskmgmt.notifyresult_handle);
> > + xa_erase(xa, cmdrsp->scsitaskmgmt.notify_handle);
> > + xa_erase(xa, cmdrsp->scsitaskmgmt.notifyresult_handle);
If this code were even run, assuming the appropriate lockdep options
were enabled, xarray would point out that you were taking a lock with
interrupts enabled that was previously taken with interrupts disabled.
That's a good thing! Because before, this code was deleting entries
from the IDR with _no_ locking.
So these should both be xa_erase_irq().
> > @@ -1096,8 +1085,6 @@ static void visorhba_remove(struct visor_device *dev)
> > scsi_remove_host(scsihost);
> > scsi_host_put(scsihost);
> >
> > - idr_destroy(&devdata->idr);
>
> Do we not have to call xa_destroy()?
Correct. We covered this in earlier iterations of this series and I
don't feel like typing out the explanation again.
