On Thu, Jun 01, 2017 at 10:11:16AM +0800, Jia-Ju Bai wrote:
> The value "cqp_request->waiting" indicates whether the sleeping operation
> should be performed, and it is not assigned in i40iw_get_cqp_request, so
> the driver may sleep in interrupt handling. The function call path is:
>
> i40iw_dpc (tasklet_init indicates it handles interrupt)
> i40iw_process_aeq
> i40iw_next_iw_state
> i40iw_hw_modify_qp (call i40iw_get_cqp_request)
> i40iw_handle_cqp_op
> i40iw_wait_event --> may sleep
>
> To fix it, "cqp_request->waiting" is assigned in "else" branch in
> i40iw_get_cqp_request.
>
> Signed-off-by: Jia-Ju Bai <baijiaju1990@xxxxxxx>
> ---
> drivers/infiniband/hw/i40iw/i40iw_utils.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/drivers/infiniband/hw/i40iw/i40iw_utils.c b/drivers/infiniband/hw/i40iw/i40iw_utils.c
> index 409a378..0f4e633 100644
> --- a/drivers/infiniband/hw/i40iw/i40iw_utils.c
> +++ b/drivers/infiniband/hw/i40iw/i40iw_utils.c
> @@ -326,6 +326,7 @@ struct i40iw_cqp_request *i40iw_get_cqp_request(struct i40iw_cqp *cqp, bool wait
> cqp_request->waiting = true;
> } else {
> atomic_set(&cqp_request->refcount, 1);
> + cqp_request->waiting = false;
> }
> return cqp_request;
> }
> --
Hi Jia - cqp_request->waiting is always initialized. If cqp_request object is
allocated using kzalloc, it is initialized to 0. For those cqp_request object that are
retrieved from the cqp list 'cqp->cqp_avail_reqs', the waiting flag is set to false
when it is put back on the list via i40iw_free_cqp_request. So there should be no issue
here.
Shiraz
--
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
