[dpdk-dev] [PATCH v1] net/ice: fix memzone leak when device init failed

[dpdk-dev] [PATCH v1] net/ice: fix memzone leak when device init failed

[Haiyue Wang]

When flow engine initialization or FXP resource reset failed, it needs
to free the memory zone and unregister the interrupt callback.

Bugzilla ID: 752
Fixes: 84dc7a95a2d3 ("net/ice: enable flow director engine")
Fixes: 7615a6895009 ("net/ice: rework for generic flow enabling")
Fixes: 7edc7158d771 ("net/ice: cleanup RSS/FDIR profile on device init")
Cc: stable@dpdk.org

Reported-by: David Marchand <david.marchand@redhat.com>
Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
---
 drivers/net/ice/ice_ethdev.c      | 10 ++++++++--
 drivers/net/ice/ice_fdir_filter.c |  2 ++
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ice/ice_ethdev.c b/drivers/net/ice/ice_ethdev.c
index 64ee569525..8d62b84805 100644
--- a/drivers/net/ice/ice_ethdev.c
+++ b/drivers/net/ice/ice_ethdev.c
@@ -2139,20 +2139,26 @@ ice_dev_init(struct rte_eth_dev *dev)
 		ret = ice_flow_init(ad);
 		if (ret) {
 			PMD_INIT_LOG(ERR, "Failed to initialize flow");
-			return ret;
+			goto err_flow_init;
 		}
 	}
 
 	ret = ice_reset_fxp_resource(hw);
 	if (ret) {
 		PMD_INIT_LOG(ERR, "Failed to reset fxp resource");
-		return ret;
+		goto err_flow_init;
 	}
 
 	pf->supported_rxdid = ice_get_supported_rxdid(hw);
 
 	return 0;
 
+err_flow_init:
+	ice_flow_uninit(ad);
+	rte_intr_disable(intr_handle);
+	ice_pf_disable_irq0(hw);
+	rte_intr_callback_unregister(intr_handle,
+				     ice_interrupt_handler, dev);
 err_pf_setup:
 	ice_res_pool_destroy(&pf->msix_pool);
 err_msix_pool_init:
diff --git a/drivers/net/ice/ice_fdir_filter.c b/drivers/net/ice/ice_fdir_filter.c
index 82adb1fc8b..7ba65b9b04 100644
--- a/drivers/net/ice/ice_fdir_filter.c
+++ b/drivers/net/ice/ice_fdir_filter.c
@@ -651,8 +651,10 @@ ice_fdir_teardown(struct ice_pf *pf)
 
 	ice_tx_queue_release(pf->fdir.txq);
 	pf->fdir.txq = NULL;
+	rte_eth_dma_zone_free(eth_dev, "fdir_tx_ring", ICE_FDIR_QUEUE_ID);
 	ice_rx_queue_release(pf->fdir.rxq);
 	pf->fdir.rxq = NULL;
+	rte_eth_dma_zone_free(eth_dev, "fdir_rx_ring", ICE_FDIR_QUEUE_ID);
 	ice_fdir_prof_rm_all(pf);
 	ice_fdir_prof_free(hw);
 	ice_release_vsi(vsi);
-- 
2.32.0

Re: [dpdk-dev] [PATCH v1] net/ice: fix memzone leak when device init failed

[David Marchand]

On Fri, Aug 13, 2021 at 8:45 AM Haiyue Wang <haiyue.wang@intel.com> wrote:
>
> When flow engine initialization or FXP resource reset failed, it needs
> to free the memory zone and unregister the interrupt callback.
>
> Bugzilla ID: 752
> Fixes: 84dc7a95a2d3 ("net/ice: enable flow director engine")
> Fixes: 7615a6895009 ("net/ice: rework for generic flow enabling")
> Fixes: 7edc7158d771 ("net/ice: cleanup RSS/FDIR profile on device init")
> Cc: stable@dpdk.org
>
> Reported-by: David Marchand <david.marchand@redhat.com>
> Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
> ---
>  drivers/net/ice/ice_ethdev.c      | 10 ++++++++--
>  drivers/net/ice/ice_fdir_filter.c |  2 ++
>  2 files changed, 10 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/ice/ice_ethdev.c b/drivers/net/ice/ice_ethdev.c
> index 64ee569525..8d62b84805 100644
> --- a/drivers/net/ice/ice_ethdev.c
> +++ b/drivers/net/ice/ice_ethdev.c
> @@ -2139,20 +2139,26 @@ ice_dev_init(struct rte_eth_dev *dev)
>                 ret = ice_flow_init(ad);
>                 if (ret) {
>                         PMD_INIT_LOG(ERR, "Failed to initialize flow");
> -                       return ret;
> +                       goto err_flow_init;

Is it safe to call flow engine uninit callbacks when ice_flow_init() fails?


>                 }
>         }
>
>         ret = ice_reset_fxp_resource(hw);
>         if (ret) {
>                 PMD_INIT_LOG(ERR, "Failed to reset fxp resource");
> -               return ret;
> +               goto err_flow_init;
>         }
>
>         pf->supported_rxdid = ice_get_supported_rxdid(hw);
>
>         return 0;
>
> +err_flow_init:
> +       ice_flow_uninit(ad);
> +       rte_intr_disable(intr_handle);
> +       ice_pf_disable_irq0(hw);
> +       rte_intr_callback_unregister(intr_handle,
> +                                    ice_interrupt_handler, dev);
>  err_pf_setup:
>         ice_res_pool_destroy(&pf->msix_pool);
>  err_msix_pool_init:


-- 
David Marchand

Re: [dpdk-dev] [PATCH v1] net/ice: fix memzone leak when device init failed

[Wang, Haiyue]

> -----Original Message-----
> From: David Marchand <david.marchand@redhat.com>
> Sent: Tuesday, August 17, 2021 17:19
> To: Wang, Haiyue <haiyue.wang@intel.com>
> Cc: dev <dev@dpdk.org>; Zhang, Qi Z <qi.z.zhang@intel.com>; dpdk stable <stable@dpdk.org>; Yang,
> Qiming <qiming.yang@intel.com>; Xiaolong Ye <xiaolong.ye@intel.com>; Xing, Beilei
> <beilei.xing@intel.com>; Wang, Ying A <ying.a.wang@intel.com>
> Subject: Re: [PATCH v1] net/ice: fix memzone leak when device init failed
> 
> On Fri, Aug 13, 2021 at 8:45 AM Haiyue Wang <haiyue.wang@intel.com> wrote:
> >
> > When flow engine initialization or FXP resource reset failed, it needs
> > to free the memory zone and unregister the interrupt callback.
> >
> > Bugzilla ID: 752
> > Fixes: 84dc7a95a2d3 ("net/ice: enable flow director engine")
> > Fixes: 7615a6895009 ("net/ice: rework for generic flow enabling")
> > Fixes: 7edc7158d771 ("net/ice: cleanup RSS/FDIR profile on device init")
> > Cc: stable@dpdk.org
> >
> > Reported-by: David Marchand <david.marchand@redhat.com>
> > Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
> > ---
> >  drivers/net/ice/ice_ethdev.c      | 10 ++++++++--
> >  drivers/net/ice/ice_fdir_filter.c |  2 ++
> >  2 files changed, 10 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/net/ice/ice_ethdev.c b/drivers/net/ice/ice_ethdev.c
> > index 64ee569525..8d62b84805 100644
> > --- a/drivers/net/ice/ice_ethdev.c
> > +++ b/drivers/net/ice/ice_ethdev.c
> > @@ -2139,20 +2139,26 @@ ice_dev_init(struct rte_eth_dev *dev)
> >                 ret = ice_flow_init(ad);
> >                 if (ret) {
> >                         PMD_INIT_LOG(ERR, "Failed to initialize flow");
> > -                       return ret;
> > +                       goto err_flow_init;
> 
> Is it safe to call flow engine uninit callbacks when ice_flow_init() fails?

If each engine->init/uninit handles its internal setting correctly, yes,
it's safe, if not, this single engine has BUG, let's fix it. ;-)

int
ice_flow_init(struct ice_adapter *ad)
{

		TAILQ_FOREACH_SAFE(engine, &engine_list, node, temp) {
		if (engine->init == NULL) {
			PMD_INIT_LOG(ERR, "Invalid engine type (%d)",
					engine->type);
			return -ENOTSUP;
		}

		ret = engine->init(ad);
		if (ret) {
			PMD_INIT_LOG(ERR, "Failed to initialize engine %d",
					engine->type);
			return ret;
		}
	}

}

void
ice_flow_uninit(struct ice_adapter *ad)
{
	struct ice_pf *pf = &ad->pf;
	struct ice_flow_engine *engine;
	struct rte_flow *p_flow;
	struct ice_flow_parser_node *p_parser;
	void *temp;

	TAILQ_FOREACH_SAFE(engine, &engine_list, node, temp) {
		if (engine->uninit)
			engine->uninit(ad);
	}

}

> 
> 
> >                 }
> >         }
> >
> >         ret = ice_reset_fxp_resource(hw);
> >         if (ret) {
> >                 PMD_INIT_LOG(ERR, "Failed to reset fxp resource");
> > -               return ret;
> > +               goto err_flow_init;
> >         }
> >
> >         pf->supported_rxdid = ice_get_supported_rxdid(hw);
> >
> >         return 0;
> >
> > +err_flow_init:
> > +       ice_flow_uninit(ad);
> > +       rte_intr_disable(intr_handle);
> > +       ice_pf_disable_irq0(hw);
> > +       rte_intr_callback_unregister(intr_handle,
> > +                                    ice_interrupt_handler, dev);
> >  err_pf_setup:
> >         ice_res_pool_destroy(&pf->msix_pool);
> >  err_msix_pool_init:
> 
> 
> --
> David Marchand

Re: [dpdk-dev] [PATCH v1] net/ice: fix memzone leak when device init failed

[David Marchand]

On Wed, Aug 18, 2021 at 2:46 AM Wang, Haiyue <haiyue.wang@intel.com> wrote:
>
> > -----Original Message-----
> > From: David Marchand <david.marchand@redhat.com>
> > Sent: Tuesday, August 17, 2021 17:19
> > To: Wang, Haiyue <haiyue.wang@intel.com>
> > Cc: dev <dev@dpdk.org>; Zhang, Qi Z <qi.z.zhang@intel.com>; dpdk stable <stable@dpdk.org>; Yang,
> > Qiming <qiming.yang@intel.com>; Xiaolong Ye <xiaolong.ye@intel.com>; Xing, Beilei
> > <beilei.xing@intel.com>; Wang, Ying A <ying.a.wang@intel.com>
> > Subject: Re: [PATCH v1] net/ice: fix memzone leak when device init failed
> >
> > On Fri, Aug 13, 2021 at 8:45 AM Haiyue Wang <haiyue.wang@intel.com> wrote:
> > >
> > > When flow engine initialization or FXP resource reset failed, it needs
> > > to free the memory zone and unregister the interrupt callback.
> > >
> > > Bugzilla ID: 752
> > > Fixes: 84dc7a95a2d3 ("net/ice: enable flow director engine")
> > > Fixes: 7615a6895009 ("net/ice: rework for generic flow enabling")
> > > Fixes: 7edc7158d771 ("net/ice: cleanup RSS/FDIR profile on device init")
> > > Cc: stable@dpdk.org
> > >
> > > Reported-by: David Marchand <david.marchand@redhat.com>
> > > Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
> > > ---
> > >  drivers/net/ice/ice_ethdev.c      | 10 ++++++++--
> > >  drivers/net/ice/ice_fdir_filter.c |  2 ++
> > >  2 files changed, 10 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/drivers/net/ice/ice_ethdev.c b/drivers/net/ice/ice_ethdev.c
> > > index 64ee569525..8d62b84805 100644
> > > --- a/drivers/net/ice/ice_ethdev.c
> > > +++ b/drivers/net/ice/ice_ethdev.c
> > > @@ -2139,20 +2139,26 @@ ice_dev_init(struct rte_eth_dev *dev)
> > >                 ret = ice_flow_init(ad);
> > >                 if (ret) {
> > >                         PMD_INIT_LOG(ERR, "Failed to initialize flow");
> > > -                       return ret;
> > > +                       goto err_flow_init;
> >
> > Is it safe to call flow engine uninit callbacks when ice_flow_init() fails?
>
> If each engine->init/uninit handles its internal setting correctly, yes,
> it's safe, if not, this single engine has BUG, let's fix it. ;-)

That was my understanding, but I preferred to ask.
Then, patch lgtm, thanks Haiyue.


-- 
David Marchand

Re: [dpdk-dev] [PATCH v1] net/ice: fix memzone leak when device init failed

[Zhang, Qi Z]

> -----Original Message-----
> From: Wang, Haiyue <haiyue.wang@intel.com>
> Sent: Friday, August 13, 2021 2:22 PM
> To: dev@dpdk.org
> Cc: Zhang, Qi Z <qi.z.zhang@intel.com>; Wang, Haiyue
> <haiyue.wang@intel.com>; stable@dpdk.org; David Marchand
> <david.marchand@redhat.com>; Yang, Qiming <qiming.yang@intel.com>;
> Xiaolong Ye <xiaolong.ye@intel.com>; Xing, Beilei <beilei.xing@intel.com>;
> Wang, Ying A <ying.a.wang@intel.com>
> Subject: [PATCH v1] net/ice: fix memzone leak when device init failed
> 
> When flow engine initialization or FXP resource reset failed, it needs to free the
> memory zone and unregister the interrupt callback.
> 
> Bugzilla ID: 752
> Fixes: 84dc7a95a2d3 ("net/ice: enable flow director engine")
> Fixes: 7615a6895009 ("net/ice: rework for generic flow enabling")
> Fixes: 7edc7158d771 ("net/ice: cleanup RSS/FDIR profile on device init")
> Cc: stable@dpdk.org
> 
> Reported-by: David Marchand <david.marchand@redhat.com>
> Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>

Acked-by: Qi Zhang <qi.z.zhang@intel.com>

Applied dpdk-next-net-intel.

Thanks
Qi