On Tue, Feb 13, 2018 at 12:10:35PM -0500, Dennis Dalessandro wrote:
> On 2/13/2018 5:18 AM, Leon Romanovsky wrote:
> >From: Jason Gunthorpe <jgg@xxxxxxxxxxxx>
> >
> >This is really being used as an asssert that the expected usecnt
> >is being held and implicitly that the usecnt is valid. Rename it to
> >assert_uverbs_usecnt and tighten the checks to only accept valid
> >values of usecnt (eg 0 and < -1 are invalid).
> >
> >Fixes: 3832125624b7 ("IB/core: Add support for idr types")
> >Signed-off-by: Jason Gunthorpe <jgg@xxxxxxxxxxxx>
> >Signed-off-by: Leon Romanovsky <leon@xxxxxxxxxx>
> > drivers/infiniband/core/rdma_core.c | 14 +++++++-------
> > 1 file changed, 7 insertions(+), 7 deletions(-)
> >
> >diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c
> >index 3ec5c7926f1c..7aee1c0b9884 100644
> >+++ b/drivers/infiniband/core/rdma_core.c
> >@@ -412,13 +412,13 @@ static int __must_check remove_commit_fd_uobject(struct ib_uobject *uobj,
> > return ret;
> > }
> >-static void lockdep_check(struct ib_uobject *uobj, bool exclusive)
> >+static void assert_uverbs_usecnt(struct ib_uobject *uobj, bool exclusive)
> > {
> > #ifdef CONFIG_LOCKDEP
> > if (exclusive)
> >- WARN_ON(atomic_read(&uobj->usecnt) > 0);
> >+ WARN_ON(atomic_read(&uobj->usecnt) != -1);
> > else
> >- WARN_ON(atomic_read(&uobj->usecnt) == -1);
> >+ WARN_ON(atomic_read(&uobj->usecnt) <= 0);
> > #endif
> > }
> >@@ -457,7 +457,7 @@ int __must_check rdma_remove_commit_uobject(struct ib_uobject *uobj)
> > WARN(true, "ib_uverbs: Cleanup is running while removing an uobject\n");
> > return 0;
> > }
> >- lockdep_check(uobj, true);
> >+ assert_uverbs_usecnt(uobj, true);
> > ret = _rdma_remove_commit_uobject(uobj, RDMA_REMOVE_DESTROY);
> > up_read(&ucontext->cleanup_rwsem);
> >@@ -487,7 +487,7 @@ int rdma_explicit_destroy(struct ib_uobject *uobject)
> > WARN(true, "ib_uverbs: Cleanup is running while removing an uobject\n");
> > return 0;
> > }
> >- lockdep_check(uobject, true);
> >+ assert_uverbs_usecnt(uobject, true);
> > ret = uobject->type->type_class->remove_commit(uobject,
> > RDMA_REMOVE_DESTROY);
> > if (ret)
> >@@ -541,7 +541,7 @@ int rdma_alloc_commit_uobject(struct ib_uobject *uobj)
> > }
> > /* matches atomic_set(-1) in alloc_uobj */
> >- lockdep_check(uobj, true);
> >+ assert_uverbs_usecnt(uobj, true);
> > atomic_set(&uobj->usecnt, 0);
> > uobj->type->type_class->alloc_commit(uobj);
> >@@ -578,7 +578,7 @@ static void lookup_put_fd_uobject(struct ib_uobject *uobj, bool exclusive)
> > void rdma_lookup_put_uobject(struct ib_uobject *uobj, bool exclusive)
> > {
> >- lockdep_check(uobj, exclusive);
> >+ assert_uverbs_usecnt(uobj, exclusive);
> > uobj->type->type_class->lookup_put(uobj, exclusive);
> > /*
> > * In order to unlock an object, either decrease its usecnt for
> >
>
> -rc material?
This one is a bit border line - but as coded the asserts were not
working right, eg:
- WARN_ON(atomic_read(&uobj->usecnt) > 0);
+ WARN_ON(atomic_read(&uobj->usecnt) != -1);
Didn't trigger if usecnt == 0, which is wrong.
Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
