Re: [PATCH for-rc 2/6] RDMA/vmw_pvrdma: Call ib_umem_release on destroy QP path

[Yuval Shaia <yuval.shaia-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>]

On Wed, Dec 13, 2017 at 10:55:18AM -0800, Bryan Tan wrote:
> On Wed, Dec 13, 2017 at 08:40:13PM +0200, Yuval Shaia wrote:
> > On Wed, Dec 13, 2017 at 09:17:55AM -0700, Jason Gunthorpe wrote:
> > > On Wed, Dec 13, 2017 at 10:53:18AM +0200, Yuval Shaia wrote:
> > > > On Fri, Dec 08, 2017 at 11:01:18AM -0800, Bryan Tan wrote:
> > > > > The QP cleanup did not previously call ib_umem_release. Fix this.
> > > > > 
> > > > > Testing Done: ibv_rc_pingpong, rping, perftests.
> > > > > 
> > > > > Fixes: 29c8d9eba550 ("IB: Add vmw_pvrdma driver")
> > > > > Reviewed-by: Adit Ranadive <aditr-pghWNbHTmq7QT0dZR+AlfA@public.gmane.org>
> > > > > Reviewed-by: Aditya Sarwade <asarwade-pghWNbHTmq7QT0dZR+AlfA@public.gmane.org>
> > > > > Reviewed-by: Jorgen Hansen <jhansen-pghWNbHTmq7QT0dZR+AlfA@public.gmane.org>
> > > > > Signed-off-by: Bryan Tan <bryantan-pghWNbHTmq7QT0dZR+AlfA@public.gmane.org>
> > > > >  drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c | 7 +++++++
> > > > >  1 file changed, 7 insertions(+)
> > > > > 
> > > > > diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c
> > > > > index b932b7e..77e7e57 100644
> > > > > +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c
> > > > > @@ -430,6 +430,13 @@ static void pvrdma_free_qp(struct pvrdma_qp *qp)
> > > > >  	atomic_dec(&qp->refcnt);
> > > > >  	wait_event(qp->wait, !atomic_read(&qp->refcnt));
> > > > >  
> > > > > +	if (!qp->is_kernel) {
> > > > > +		if (qp->rumem)
> > > > 
> > > > IS_ERR_OR_NULL?
> > > 
> > > How can we get an ERR_PTR in rumem?
> > 
> > qp->rumem is set with ib_umem_get which can either return ib_umem pointer
> > or ERR_PTR.
> > So i looked at other places which calls ib_umem_release. The first one that
> > pops up is bnxt_re which do this check with IS_ERR_OR_NULL?.
> > 
> > So assuming it can be used here.
> 
> We check in pvrdma_qp.c pvrdma_create_qp the result of ib_umem_get.
> If it's ERR_PTR we fail the QP creation:
> 
> > qp->rumem = ib_umem_get(pd->uobject->context,
> > 			ucmd.rbuf_addr,
> > 			ucmd.rbuf_size, 0, 0);
> > if (IS_ERR(qp->rumem)) {
> > 	ret = PTR_ERR(qp->rumem);
> > 	goto err_qp;
> > }
> 
> so checking IS_ERR_OR_NULL isn't necessary here (and same for sumem).

I see.
Missed that.

> 
> Thanks,
> Bryan
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html