[PATCH] RDMA/mlx4: Remove unnessesary check in mlx4_ib_modify_wq()

[PATCH] RDMA/mlx4: Remove unnessesary check in mlx4_ib_modify_wq()

[Jiapeng Chong]

cur_state and new_state are enums and when GCC considers
them as unsigned, the conditions are never met.

Clean up the following smatch warning:

drivers/infiniband/hw/mlx4/qp.c:4258 mlx4_ib_modify_wq() warn: unsigned
'cur_state' is never less than zero.

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
---
 drivers/infiniband/hw/mlx4/qp.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 92ddbcc..162aa59 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -4255,8 +4255,7 @@ int mlx4_ib_modify_wq(struct ib_wq *ibwq, struct ib_wq_attr *wq_attr,
 						     ibwq->state;
 	new_state = wq_attr_mask & IB_WQ_STATE ? wq_attr->wq_state : cur_state;
 
-	if (cur_state  < IB_WQS_RESET || cur_state  > IB_WQS_ERR ||
-	    new_state < IB_WQS_RESET || new_state > IB_WQS_ERR)
+	if (cur_state > IB_WQS_ERR || new_state > IB_WQS_ERR)
 		return -EINVAL;
 
 	if ((new_state == IB_WQS_RDY) && (cur_state == IB_WQS_ERR))
-- 
1.8.3.1

Re: [PATCH] RDMA/mlx4: Remove unnessesary check in mlx4_ib_modify_wq()

[Leon Romanovsky]

On Fri, May 07, 2021 at 06:22:41PM +0800, Jiapeng Chong wrote:
> cur_state and new_state are enums and when GCC considers
> them as unsigned, the conditions are never met.
> 
> Clean up the following smatch warning:
> 
> drivers/infiniband/hw/mlx4/qp.c:4258 mlx4_ib_modify_wq() warn: unsigned
> 'cur_state' is never less than zero.
> 
> Reported-by: Abaci Robot <abaci@linux.alibaba.com>
> Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
> ---
>  drivers/infiniband/hw/mlx4/qp.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)

Thanks,
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>

Re: [PATCH] RDMA/mlx4: Remove unnessesary check in mlx4_ib_modify_wq()

[Leon Romanovsky]

On Fri, May 07, 2021 at 06:22:41PM +0800, Jiapeng Chong wrote:
> cur_state and new_state are enums and when GCC considers
> them as unsigned, the conditions are never met.
> 
> Clean up the following smatch warning:
> 
> drivers/infiniband/hw/mlx4/qp.c:4258 mlx4_ib_modify_wq() warn: unsigned
> 'cur_state' is never less than zero.
> 
> Reported-by: Abaci Robot <abaci@linux.alibaba.com>
> Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
> ---
>  drivers/infiniband/hw/mlx4/qp.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
> index 92ddbcc..162aa59 100644
> --- a/drivers/infiniband/hw/mlx4/qp.c
> +++ b/drivers/infiniband/hw/mlx4/qp.c
> @@ -4255,8 +4255,7 @@ int mlx4_ib_modify_wq(struct ib_wq *ibwq, struct ib_wq_attr *wq_attr,
>  						     ibwq->state;
>  	new_state = wq_attr_mask & IB_WQ_STATE ? wq_attr->wq_state : cur_state;
>  
> -	if (cur_state  < IB_WQS_RESET || cur_state  > IB_WQS_ERR ||
> -	    new_state < IB_WQS_RESET || new_state > IB_WQS_ERR)
> +	if (cur_state > IB_WQS_ERR || new_state > IB_WQS_ERR)
>  		return -EINVAL;

Actually the more robust change will be to move this change to the ib_uverbs_ex_modify_wq().

Thanks

>  
>  	if ((new_state == IB_WQS_RDY) && (cur_state == IB_WQS_ERR))
> -- 
> 1.8.3.1
> 

Re: [PATCH] RDMA/mlx4: Remove unnessesary check in mlx4_ib_modify_wq()

[Jason Gunthorpe]

On Fri, May 07, 2021 at 06:22:41PM +0800, Jiapeng Chong wrote:
> cur_state and new_state are enums and when GCC considers
> them as unsigned, the conditions are never met.

But doesn't gcc consider enums to be 'int' as the standard requires?

This change looks really sketchy to me, cur_state and new_state are
both userspace controlled data. We should not make assumptions about
the underlying signedness of an enum when validating user data.

Jason

Re: [PATCH] RDMA/mlx4: Remove unnessesary check in mlx4_ib_modify_wq()

[Leon Romanovsky]

On Tue, May 11, 2021 at 02:43:02PM -0300, Jason Gunthorpe wrote:
> On Fri, May 07, 2021 at 06:22:41PM +0800, Jiapeng Chong wrote:
> > cur_state and new_state are enums and when GCC considers
> > them as unsigned, the conditions are never met.
> 
> But doesn't gcc consider enums to be 'int' as the standard requires?

Ohh, I missed that.

> 
> This change looks really sketchy to me, cur_state and new_state are
> both userspace controlled data. We should not make assumptions about
> the underlying signedness of an enum when validating user data.

I still think that the right change should be in
ib_uverbs_ex_modify_wq(), so both mlx4 and mlx5 will be protected.

Thanks

> 
> Jason