On Mon, Mar 04, 2019 at 03:39:34PM -0400, Jason Gunthorpe wrote:
> On Mon, Feb 25, 2019 at 08:56:14AM +0200, Leon Romanovsky wrote:
> > From: Leon Romanovsky <leonro@xxxxxxxxxxxx>
> >
> > Refactor the page fault handler to be more readable and extensible,
> > this cleanup was triggered by error reported below. The code structure
> > made unclear to the automatic tools to identify that such flow is not
> > possible in real life because "requestor != NULL" means that "qp != NULL"
> > too.
> >
> > drivers/infiniband/hw/mlx5/odp.c:1254 mlx5_ib_mr_wqe_pfault_handler()
> > error: we previously assumed 'qp' could be null (see line 1230)
> >
> > Fixes: 08100fad5cac ("IB/mlx5: Add ODP SRQ support")
> > Reported-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> > Reviewed-by: Moni Shoua <monis@xxxxxxxxxxxx>
> > Signed-off-by: Leon Romanovsky <leonro@xxxxxxxxxxxx>
> > ---
> > Changes v0->v1:
> > * Instead of simple fix, rewrote whole function to be more clear.
> > ---
> > drivers/infiniband/hw/mlx5/odp.c | 121 ++++++++++++++-----------------
> > 1 file changed, 55 insertions(+), 66 deletions(-)
> >
> > --
> > 2.19.1
> >
> > diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
> > index c20bfc41ecf1..31caced2e8c2 100644
> > --- a/drivers/infiniband/hw/mlx5/odp.c
> > +++ b/drivers/infiniband/hw/mlx5/odp.c
> > @@ -919,7 +919,7 @@ static int pagefault_data_segments(struct mlx5_ib_dev *dev,
> > struct mlx5_pagefault *pfault,
> > void *wqe,
> > void *wqe_end, u32 *bytes_mapped,
> > - u32 *total_wqe_bytes, int receive_queue)
> > + u32 *total_wqe_bytes, bool receive_queue)
> > {
> > int ret = 0, npages = 0;
> > u64 io_virt;
> > @@ -1199,17 +1199,15 @@ static inline struct mlx5_ib_srq *res_to_srq(struct mlx5_core_rsc_common *res)
> > static void mlx5_ib_mr_wqe_pfault_handler(struct mlx5_ib_dev *dev,
> > struct mlx5_pagefault *pfault)
> > {
> > - int ret;
> > - void *wqe, *wqe_end;
> > + bool sq = pfault->type & MLX5_PFAULT_REQUESTOR;
> > + u16 wqe_index = pfault->wqe.wqe_index;
> > + void *wqe = NULL, *wqe_end = NULL;
> > u32 bytes_mapped, total_wqe_bytes;
> > - char *buffer = NULL;
> > + struct mlx5_core_rsc_common *res;
> > int resume_with_error = 1;
> > - u16 wqe_index = pfault->wqe.wqe_index;
> > - int requestor = pfault->type & MLX5_PFAULT_REQUESTOR;
> > - struct mlx5_core_rsc_common *res = NULL;
> > - struct mlx5_ib_qp *qp = NULL;
> > - struct mlx5_ib_srq *srq = NULL;
> > + struct mlx5_ib_qp *qp;
> > size_t bytes_copied;
> > + int ret;
> >
> > res = odp_get_rsc(dev, pfault->wqe.wq_num, pfault->type);
> > if (!res) {
> > @@ -1217,87 +1215,78 @@ static void mlx5_ib_mr_wqe_pfault_handler(struct mlx5_ib_dev *dev,
> > return;
> > }
> >
> > - switch (res->res) {
> > - case MLX5_RES_QP:
> > - qp = res_to_qp(res);
> > - break;
> > - case MLX5_RES_SRQ:
> > - case MLX5_RES_XSRQ:
> > - srq = res_to_srq(res);
> > - break;
> > - default:
> > - mlx5_ib_err(dev, "wqe page fault for unsupported type %d\n", pfault->type);
> > + if (res->res != MLX5_RES_QP && res->res != MLX5_RES_SRQ &&
> > + res->res != MLX5_RES_XSRQ) {
> > + mlx5_ib_err(dev, "wqe page fault for unsupported type %d\n",
> > + pfault->type);
> > goto resolve_page_fault;
> > }
> >
> > - buffer = (char *)__get_free_page(GFP_KERNEL);
> > - if (!buffer) {
> > + wqe = (void *)__get_free_page(GFP_KERNEL);
> > + if (!wqe) {
> > mlx5_ib_err(dev, "Error allocating memory for IO page fault handling.\n");
> > goto resolve_page_fault;
> > }
> >
> > - if (qp) {
> > - if (requestor) {
> > - ret = mlx5_ib_read_user_wqe_sq(qp, wqe_index,
> > - buffer, PAGE_SIZE,
> > - &bytes_copied);
> > - } else {
> > - ret = mlx5_ib_read_user_wqe_rq(qp, wqe_index,
> > - buffer, PAGE_SIZE,
> > - &bytes_copied);
> > - }
> > - } else {
> > - ret = mlx5_ib_read_user_wqe_srq(srq, wqe_index,
> > - buffer, PAGE_SIZE,
> > - &bytes_copied);
> > + qp = (res->res == MLX5_RES_QP) ? res_to_qp(res) : NULL;
> > + if (qp && sq) {
> > + ret = mlx5_ib_read_user_wqe_sq(qp, wqe_index, wqe, PAGE_SIZE,
> > + &bytes_copied);
> > + if (ret)
> > + goto read_user;
> > + ret = mlx5_ib_mr_initiator_pfault_handler(
> > + dev, pfault, qp, &wqe, &wqe_end, bytes_copied);
> > }
> >
> > - if (ret) {
> > - mlx5_ib_err(dev, "Failed reading a WQE following page fault, error=%d, wqe_index=%x, qpn=%x\n",
> > - ret, wqe_index, pfault->token);
> > - goto resolve_page_fault;
> > + if (qp && !sq) {
>
> Why not
>
> else if (..)
>
> In each of these branches? Only one can run, right?
I don't like "else if ()" construction and see it is as dangerous, due
to the fact that it hides code flow. I intentionally avoid "else if" in
my patches.
Thanks
>
> Jason
Attachment:
signature.asc
Description: PGP signature
