On Wed, Apr 17, 2019 at 11:02:38AM +0300, Gal Pressman wrote:
> On 31-Mar-19 19:10, Shamir Rabinovitch wrote:
> > the uverbs_attr_bundle with the ucontext is sent down to the drivers
> > ib_x destroy path as ib_udata. next patch will use the ib_udata to
> > free the drivers destroy path from the dependency in 'uobject->context'
> > as we already did for the create path.
> >
> > Signed-off-by: Shamir Rabinovitch <shamir.rabinovitch@xxxxxxxxxx>
>
> Hi Shamir,
> When running over for-next branch (with this patch), it looks like the driver
> udata passed to dereg_mr callback is corrupted (would expect it to
> be cleared),
Cleared as in equal to NULL or cleared as in the values of the
structure don't make sense?
Hum. There is this weird thing where methods that don't have a udata
defined don't always get a filled in udata pointer.
Does this fix it?
diff --git a/drivers/infiniband/core/uverbs_ioctl.c b/drivers/infiniband/core/uverbs_ioctl.c
index cfbef25b3a73af..829b0c6944d842 100644
--- a/drivers/infiniband/core/uverbs_ioctl.c
+++ b/drivers/infiniband/core/uverbs_ioctl.c
@@ -453,6 +453,8 @@ static int ib_uverbs_run_method(struct bundle_priv *pbundle,
uverbs_fill_udata(&pbundle->bundle,
&pbundle->bundle.driver_udata,
UVERBS_ATTR_UHW_IN, UVERBS_ATTR_UHW_OUT);
+ else
+ pbundle->bundle.driver_udata = (struct ib_udata){};
if (destroy_bkey != UVERBS_API_ATTR_BKEY_LEN) {
struct uverbs_obj_attr *destroy_attr =
