From: Xueming Li <xuemingl@mellanox.com> During vdap device setup, if some error happens, event channel release stuck at polling event channel. Event channel fd is set to nonblocking in cqe setup, so if any error happens before this function and after event channel created, the pooling before releasing resources will stuck. This patch moves event channel to non-blocking mode right after creation. Fixes: 8395927cdf ("vdpa/mlx5: prepare HW queues") Cc: matan@nvidia.com Signed-off-by: Xueming Li <xuemingl@nvidia.com> --- drivers/vdpa/mlx5/mlx5_vdpa_event.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/drivers/vdpa/mlx5/mlx5_vdpa_event.c b/drivers/vdpa/mlx5/mlx5_vdpa_event.c index 5a2d4fb1ec..bda547ffe0 100644 --- a/drivers/vdpa/mlx5/mlx5_vdpa_event.c +++ b/drivers/vdpa/mlx5/mlx5_vdpa_event.c @@ -51,6 +51,8 @@ mlx5_vdpa_event_qp_global_release(struct mlx5_vdpa_priv *priv) static int mlx5_vdpa_event_qp_global_prepare(struct mlx5_vdpa_priv *priv) { + int flags, ret; + if (priv->eventc) return 0; if (mlx5_glue->devx_query_eqn(priv->ctx, 0, &priv->eqn)) { @@ -66,6 +68,12 @@ mlx5_vdpa_event_qp_global_prepare(struct mlx5_vdpa_priv *priv) rte_errno); goto error; } + flags = fcntl(priv->eventc->fd, F_GETFL); + ret = fcntl(priv->eventc->fd, F_SETFL, flags | O_NONBLOCK); + if (ret) { + DRV_LOG(ERR, "Failed to change event channel FD."); + goto error; + } priv->uar = mlx5_glue->devx_alloc_uar(priv->ctx, 0); if (!priv->uar) { rte_errno = errno; @@ -376,7 +384,6 @@ mlx5_vdpa_interrupt_handler(void *cb_arg) int mlx5_vdpa_cqe_event_setup(struct mlx5_vdpa_priv *priv) { - int flags; int ret; if (!priv->eventc) @@ -393,12 +400,6 @@ mlx5_vdpa_cqe_event_setup(struct mlx5_vdpa_priv *priv) return -1; } } - flags = fcntl(priv->eventc->fd, F_GETFL); - ret = fcntl(priv->eventc->fd, F_SETFL, flags | O_NONBLOCK); - if (ret) { - DRV_LOG(ERR, "Failed to change event channel FD."); - goto error; - } priv->intr_handle.fd = priv->eventc->fd; priv->intr_handle.type = RTE_INTR_HANDLE_EXT; if (rte_intr_callback_register(&priv->intr_handle, -- 2.17.1
During vdap device setup, if some error happens, event channel release stuck at polling event channel. Event channel fd is set to nonblocking in cqe setup, so if any error happens before this function and after event channel created, the pooling before releasing resources will stuck. This patch moves event channel to non-blocking mode right after creation. Fixes: 8395927cdf ("vdpa/mlx5: prepare HW queues") Cc: matan@nvidia.com Signed-off-by: Xueming Li <xuemingl@nvidia.com> --- drivers/vdpa/mlx5/mlx5_vdpa_event.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/drivers/vdpa/mlx5/mlx5_vdpa_event.c b/drivers/vdpa/mlx5/mlx5_vdpa_event.c index 5a2d4fb1ec..bda547ffe0 100644 --- a/drivers/vdpa/mlx5/mlx5_vdpa_event.c +++ b/drivers/vdpa/mlx5/mlx5_vdpa_event.c @@ -51,6 +51,8 @@ mlx5_vdpa_event_qp_global_release(struct mlx5_vdpa_priv *priv) static int mlx5_vdpa_event_qp_global_prepare(struct mlx5_vdpa_priv *priv) { + int flags, ret; + if (priv->eventc) return 0; if (mlx5_glue->devx_query_eqn(priv->ctx, 0, &priv->eqn)) { @@ -66,6 +68,12 @@ mlx5_vdpa_event_qp_global_prepare(struct mlx5_vdpa_priv *priv) rte_errno); goto error; } + flags = fcntl(priv->eventc->fd, F_GETFL); + ret = fcntl(priv->eventc->fd, F_SETFL, flags | O_NONBLOCK); + if (ret) { + DRV_LOG(ERR, "Failed to change event channel FD."); + goto error; + } priv->uar = mlx5_glue->devx_alloc_uar(priv->ctx, 0); if (!priv->uar) { rte_errno = errno; @@ -376,7 +384,6 @@ mlx5_vdpa_interrupt_handler(void *cb_arg) int mlx5_vdpa_cqe_event_setup(struct mlx5_vdpa_priv *priv) { - int flags; int ret; if (!priv->eventc) @@ -393,12 +400,6 @@ mlx5_vdpa_cqe_event_setup(struct mlx5_vdpa_priv *priv) return -1; } } - flags = fcntl(priv->eventc->fd, F_GETFL); - ret = fcntl(priv->eventc->fd, F_SETFL, flags | O_NONBLOCK); - if (ret) { - DRV_LOG(ERR, "Failed to change event channel FD."); - goto error; - } priv->intr_handle.fd = priv->eventc->fd; priv->intr_handle.type = RTE_INTR_HANDLE_EXT; if (rte_intr_callback_register(&priv->intr_handle, -- 2.17.1
From: Xueming Li <xuemingl@nvidia.com>
> During vdap device setup, if some error happens, event channel release
> stuck at polling event channel.
>
> Event channel fd is set to nonblocking in cqe setup, so if any error happens
> before this function and after event channel created, the pooling before
> releasing resources will stuck.
>
> This patch moves event channel to non-blocking mode right after creation.
>
> Fixes: 8395927cdf ("vdpa/mlx5: prepare HW queues")
> Cc: matan@nvidia.com
>
> Signed-off-by: Xueming Li <xuemingl@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
On 8/25/20 11:17 AM, Xueming Li wrote: > During vdap device setup, if some error happens, event channel release s/vdap/vDPA/ > stuck at polling event channel. > > Event channel fd is set to nonblocking in cqe setup, so if any error non-blocking > happens before this function and after event channel created, the > pooling before releasing resources will stuck. > > This patch moves event channel to non-blocking mode right after > creation. > > Fixes: 8395927cdf ("vdpa/mlx5: prepare HW queues") > Cc: matan@nvidia.com > > Signed-off-by: Xueming Li <xuemingl@nvidia.com> > --- > drivers/vdpa/mlx5/mlx5_vdpa_event.c | 15 ++++++++------- > 1 file changed, 8 insertions(+), 7 deletions(-) > > diff --git a/drivers/vdpa/mlx5/mlx5_vdpa_event.c b/drivers/vdpa/mlx5/mlx5_vdpa_event.c > index 5a2d4fb1ec..bda547ffe0 100644 > --- a/drivers/vdpa/mlx5/mlx5_vdpa_event.c > +++ b/drivers/vdpa/mlx5/mlx5_vdpa_event.c > @@ -51,6 +51,8 @@ mlx5_vdpa_event_qp_global_release(struct mlx5_vdpa_priv *priv) > static int > mlx5_vdpa_event_qp_global_prepare(struct mlx5_vdpa_priv *priv) > { > + int flags, ret; > + > if (priv->eventc) > return 0; > if (mlx5_glue->devx_query_eqn(priv->ctx, 0, &priv->eqn)) { > @@ -66,6 +68,12 @@ mlx5_vdpa_event_qp_global_prepare(struct mlx5_vdpa_priv *priv) > rte_errno); > goto error; > } > + flags = fcntl(priv->eventc->fd, F_GETFL); > + ret = fcntl(priv->eventc->fd, F_SETFL, flags | O_NONBLOCK); > + if (ret) { > + DRV_LOG(ERR, "Failed to change event channel FD."); > + goto error; > + } > priv->uar = mlx5_glue->devx_alloc_uar(priv->ctx, 0); > if (!priv->uar) { > rte_errno = errno; > @@ -376,7 +384,6 @@ mlx5_vdpa_interrupt_handler(void *cb_arg) > int > mlx5_vdpa_cqe_event_setup(struct mlx5_vdpa_priv *priv) > { > - int flags; > int ret; > > if (!priv->eventc) > @@ -393,12 +400,6 @@ mlx5_vdpa_cqe_event_setup(struct mlx5_vdpa_priv *priv) > return -1; > } > } > - flags = fcntl(priv->eventc->fd, F_GETFL); > - ret = fcntl(priv->eventc->fd, F_SETFL, flags | O_NONBLOCK); > - if (ret) { > - DRV_LOG(ERR, "Failed to change event channel FD."); > - goto error; > - } > priv->intr_handle.fd = priv->eventc->fd; > priv->intr_handle.type = RTE_INTR_HANDLE_EXT; > if (rte_intr_callback_register(&priv->intr_handle, > Looks good to me, I'll fix the typos while applying. Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> Thanks, Maxime
> -----Original Message----- > From: Maxime Coquelin <maxime.coquelin@redhat.com> > Sent: Friday, September 18, 2020 5:42 PM > To: Xueming(Steven) Li <xuemingl@nvidia.com>; Matan Azrad > <matan@nvidia.com>; Slava Ovsiienko <viacheslavo@nvidia.com> > Cc: dev@dpdk.org; Asaf Penso <asafp@nvidia.com> > Subject: Re: [dpdk-dev] [v1] vdpa/mlx5: fix event channel setup > > > > On 8/25/20 11:17 AM, Xueming Li wrote: > > During vdap device setup, if some error happens, event channel release > > s/vdap/vDPA/ > > > stuck at polling event channel. > > > > Event channel fd is set to nonblocking in cqe setup, so if any error > > non-blocking > > > happens before this function and after event channel created, the > > pooling before releasing resources will stuck. > > > > This patch moves event channel to non-blocking mode right after > > creation. > > > > Fixes: 8395927cdf ("vdpa/mlx5: prepare HW queues") > > Cc: matan@nvidia.com > > > > Signed-off-by: Xueming Li <xuemingl@nvidia.com> > > --- > > drivers/vdpa/mlx5/mlx5_vdpa_event.c | 15 ++++++++------- > > 1 file changed, 8 insertions(+), 7 deletions(-) > > > > diff --git a/drivers/vdpa/mlx5/mlx5_vdpa_event.c > > b/drivers/vdpa/mlx5/mlx5_vdpa_event.c > > index 5a2d4fb1ec..bda547ffe0 100644 > > --- a/drivers/vdpa/mlx5/mlx5_vdpa_event.c > > +++ b/drivers/vdpa/mlx5/mlx5_vdpa_event.c > > @@ -51,6 +51,8 @@ mlx5_vdpa_event_qp_global_release(struct > > mlx5_vdpa_priv *priv) static int > > mlx5_vdpa_event_qp_global_prepare(struct mlx5_vdpa_priv *priv) { > > + int flags, ret; > > + > > if (priv->eventc) > > return 0; > > if (mlx5_glue->devx_query_eqn(priv->ctx, 0, &priv->eqn)) { @@ -66,6 > > +68,12 @@ mlx5_vdpa_event_qp_global_prepare(struct mlx5_vdpa_priv > *priv) > > rte_errno); > > goto error; > > } > > + flags = fcntl(priv->eventc->fd, F_GETFL); > > + ret = fcntl(priv->eventc->fd, F_SETFL, flags | O_NONBLOCK); > > + if (ret) { > > + DRV_LOG(ERR, "Failed to change event channel FD."); > > + goto error; > > + } > > priv->uar = mlx5_glue->devx_alloc_uar(priv->ctx, 0); > > if (!priv->uar) { > > rte_errno = errno; > > @@ -376,7 +384,6 @@ mlx5_vdpa_interrupt_handler(void *cb_arg) int > > mlx5_vdpa_cqe_event_setup(struct mlx5_vdpa_priv *priv) { > > - int flags; > > int ret; > > > > if (!priv->eventc) > > @@ -393,12 +400,6 @@ mlx5_vdpa_cqe_event_setup(struct > mlx5_vdpa_priv *priv) > > return -1; > > } > > } > > - flags = fcntl(priv->eventc->fd, F_GETFL); > > - ret = fcntl(priv->eventc->fd, F_SETFL, flags | O_NONBLOCK); > > - if (ret) { > > - DRV_LOG(ERR, "Failed to change event channel FD."); > > - goto error; > > - } > > priv->intr_handle.fd = priv->eventc->fd; > > priv->intr_handle.type = RTE_INTR_HANDLE_EXT; > > if (rte_intr_callback_register(&priv->intr_handle, > > > > Looks good to me, I'll fix the typos while applying. > Appreciate! > Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> > > Thanks, > Maxime
On 8/25/20 11:17 AM, Xueming Li wrote:
> During vdap device setup, if some error happens, event channel release
> stuck at polling event channel.
>
> Event channel fd is set to nonblocking in cqe setup, so if any error
> happens before this function and after event channel created, the
> pooling before releasing resources will stuck.
>
> This patch moves event channel to non-blocking mode right after
> creation.
>
> Fixes: 8395927cdf ("vdpa/mlx5: prepare HW queues")
> Cc: matan@nvidia.com
>
> Signed-off-by: Xueming Li <xuemingl@nvidia.com>
> ---
> drivers/vdpa/mlx5/mlx5_vdpa_event.c | 15 ++++++++-------
> 1 file changed, 8 insertions(+), 7 deletions(-)
Applied to dpdk-next-virtio/master.
Thanks,
Maxime