Re: [PATCH rdma-next 2/5] RDMA/mlx5: Avoid memory leak in case of XRCD dealloc failure

[Jason Gunthorpe <jgg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>]

On Sun, Jan 28, 2018 at 03:37:55PM +0200, Leon Romanovsky wrote:
> On Sun, Jan 28, 2018 at 03:25:58PM +0200, Yuval Shaia wrote:
> > On Sun, Jan 28, 2018 at 02:52:08PM +0200, Leon Romanovsky wrote:
> > > On Sun, Jan 28, 2018 at 01:07:18PM +0200, Yuval Shaia wrote:
> > > > On Sun, Jan 28, 2018 at 11:25:30AM +0200, Leon Romanovsky wrote:
> > > > > From: Leon Romanovsky <leonro-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> > > > >
> > > > > Failure in XRCD FW deallocation command leaves memory leaked and
> > > > > returns error to the user which he can't do anything about it.
> > > > >
> > > > > This patch changes behavior to always free memory and always return
> > > > > success to the user.
> > > > >
> > > > > Fixes: e126ba97dba9 ("mlx5: Add driver for Mellanox Connect-IB adapters")
> > > > > Reviewed-by: Majd Dibbiny <majd-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> > > > > Signed-off-by: Leon Romanovsky <leonro-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> > > > >  drivers/infiniband/hw/mlx5/qp.c | 5 +----
> > > > >  1 file changed, 1 insertion(+), 4 deletions(-)
> > > > >
> > > > > diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
> > > > > index b06a2f9a8df4..39d24bf694a8 100644
> > > > > +++ b/drivers/infiniband/hw/mlx5/qp.c
> > > > > @@ -5066,13 +5066,10 @@ int mlx5_ib_dealloc_xrcd(struct ib_xrcd *xrcd)
> > > > >  	int err;
> > > > >
> > > > >  	err = mlx5_core_xrcd_dealloc(dev->mdev, xrcdn);
> > > > > -	if (err) {
> > > > > +	if (err)
> > > > >  		mlx5_ib_warn(dev, "failed to dealloc xrcdn 0x%x\n", xrcdn);
> > > > > -		return err;
> > > > > -	}
> > > > >
> > > > >  	kfree(xrcd);
> > > > > -
> > > > >  	return 0;
> > > >
> > > > In this case suggesting to make the function return void.
> > >
> > > It is part of global flow:
> > > ib_uverbs_dealloc_xrcd -> ib_dealloc_xrcd -> xrcd->device->dealloc_xrcd
> >
> > I see, missed that code path.
> >
> > But for my understanding, as of now - all implementations of this hook
> > (two) returns 0 always, right?
> 
> Right, both mlx4 and mlx5 always return 0, and it is related to mistake
> in verbs API - various free and destroy functions return errors without
> actual need for it.

It would be nice to send another patch to remove the return code.

Long term I want to have all driver dealloc's unfailable..

Jason
--
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