2010-07-30 11:09:52

by Kulikov Vasiliy

Subject: [PATCH 7/9] staging: tidspbridge: fix memory leak

dbll_create() leaks pzl_target if target_obj is NULL. Call kzalloc()
only if memory is used.

Signed-off-by: Kulikov Vasiliy <[email protected]>
---
drivers/staging/tidspbridge/pmgr/dbll.c | 4 ++--
1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/staging/tidspbridge/pmgr/dbll.c b/drivers/staging/tidspbridge/pmgr/dbll.c
index 3636aa3..86840ce 100644
--- a/drivers/staging/tidspbridge/pmgr/dbll.c
+++ b/drivers/staging/tidspbridge/pmgr/dbll.c
@@ -247,9 +247,9 @@ int dbll_create(struct dbll_tar_obj **target_obj,
DBC_REQUIRE(pattrs != NULL);
DBC_REQUIRE(target_obj != NULL);

- /* Allocate DBL target object */
- pzl_target = kzalloc(sizeof(struct dbll_tar_obj), GFP_KERNEL);
if (target_obj != NULL) {
+ /* Allocate DBL target object */
+ pzl_target = kzalloc(sizeof(struct dbll_tar_obj), GFP_KERNEL);
if (pzl_target == NULL) {
*target_obj = NULL;
status = -ENOMEM;
--
1.7.0.4

2010-07-30 12:38:25

by Nishanth Menon

Subject: Re: [PATCH 7/9] staging: tidspbridge: fix memory leak

Kulikov Vasiliy had written, on 07/30/2010 06:09 AM, the following:
> dbll_create() leaks pzl_target if target_obj is NULL. Call kzalloc()
> only if memory is used.
>
> Signed-off-by: Kulikov Vasiliy <[email protected]>
> ---
> drivers/staging/tidspbridge/pmgr/dbll.c | 4 ++--
> 1 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/staging/tidspbridge/pmgr/dbll.c b/drivers/staging/tidspbridge/pmgr/dbll.c
> index 3636aa3..86840ce 100644
> --- a/drivers/staging/tidspbridge/pmgr/dbll.c
> +++ b/drivers/staging/tidspbridge/pmgr/dbll.c
> @@ -247,9 +247,9 @@ int dbll_create(struct dbll_tar_obj **target_obj,
> DBC_REQUIRE(pattrs != NULL);
> DBC_REQUIRE(target_obj != NULL);
>
> - /* Allocate DBL target object */
> - pzl_target = kzalloc(sizeof(struct dbll_tar_obj), GFP_KERNEL);
> if (target_obj != NULL) {
> + /* Allocate DBL target object */
> + pzl_target = kzalloc(sizeof(struct dbll_tar_obj), GFP_KERNEL);
> if (pzl_target == NULL) {
> *target_obj = NULL;
> status = -ENOMEM;
I would suggest to add a else branch for if (target_obj != NULL) adding
a status =-ENOMEM if error.

Overall, I guess we need to kick out the DBC_REQUIRE usage.. it is
misleading as it is active only when DEBUG macro is enabled!

--
Regards,
Nishanth Menon