Re: [PATCH] can: ti_hecc: fix close when napi poll is active

[Måns Rullgård]

Jeroen Hofstee  writes:

> When closing this CAN interface while napi poll is active, for example with:
> `ip link set can0 down` several interfaces freeze. This seemed to be caused
> by napi_disable called from ti_hecc_close expecting the scheduled probe to
> either return quota or call napi_complete. Since the poll functions has a
> check for netif_running it returns 0 and doesn't call napi_complete and hence
> violates the napi its expectation.
>
> So remove this check, so either napi_complete is called or quota is returned.
>
> Signed-off-by: Jeroen Hofstee 
> ---
>  drivers/net/can/ti_hecc.c | 3 ---
>  1 file changed, 3 deletions(-)
>
> diff --git a/drivers/net/can/ti_hecc.c b/drivers/net/can/ti_hecc.c
> index db6ea93..42813d3 100644
> --- a/drivers/net/can/ti_hecc.c
> +++ b/drivers/net/can/ti_hecc.c
> @@ -603,9 +603,6 @@ static int ti_hecc_rx_poll(struct napi_struct *napi, int 
> quota)
>   u32 mbx_mask;
>   unsigned long pending_pkts, flags;
>
> - if (!netif_running(ndev))
> - return 0;
> -
>   while ((pending_pkts = hecc_read(priv, HECC_CANRMP)) &&
>   num_pkts < quota) {
>   mbx_mask = BIT(priv->rx_next); /* next rx mailbox to process */
> -- 
> 2.7.4

This seems to have been lost or forgotten.

-- 
Måns Rullgård

[PATCH] can: ti_hecc: fix close when napi poll is active

[Jeroen Hofstee]

When closing this CAN interface while napi poll is active, for example with:
`ip link set can0 down` several interfaces freeze. This seemed to be caused
by napi_disable called from ti_hecc_close expecting the scheduled probe to
either return quota or call napi_complete. Since the poll functions has a
check for netif_running it returns 0 and doesn't call napi_complete and hence
violates the napi its expectation.

So remove this check, so either napi_complete is called or quota is returned.

Signed-off-by: Jeroen Hofstee 
---
 drivers/net/can/ti_hecc.c | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/drivers/net/can/ti_hecc.c b/drivers/net/can/ti_hecc.c
index db6ea93..42813d3 100644
--- a/drivers/net/can/ti_hecc.c
+++ b/drivers/net/can/ti_hecc.c
@@ -603,9 +603,6 @@ static int ti_hecc_rx_poll(struct napi_struct *napi, int 
quota)
u32 mbx_mask;
unsigned long pending_pkts, flags;
 
-   if (!netif_running(ndev))
-   return 0;
-
while ((pending_pkts = hecc_read(priv, HECC_CANRMP)) &&
num_pkts < quota) {
mbx_mask = BIT(priv->rx_next); /* next rx mailbox to process */
-- 
2.7.4

[PATCH] can: ti_hecc: fix close when napi poll is active

[Jeroen Hofstee]

When closing this CAN interface while napi poll is active, for example with:
`ip link set can0 down` several interfaces freeze. This seemed to be caused
by napi_disable called from ti_hecc_close expecting the scheduled probe to
either return quota or call napi_complete. Since the poll functions has a
check for netif_running it returns 0 and doesn't call napi_complete and hence
violates the napi its expectation.

So remove this check, so either napi_complete is called or quota is returned.

Signed-off-by: Jeroen Hofstee 
---
 drivers/net/can/ti_hecc.c | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/drivers/net/can/ti_hecc.c b/drivers/net/can/ti_hecc.c
index db6ea93..42813d3 100644
--- a/drivers/net/can/ti_hecc.c
+++ b/drivers/net/can/ti_hecc.c
@@ -603,9 +603,6 @@ static int ti_hecc_rx_poll(struct napi_struct *napi, int 
quota)
u32 mbx_mask;
unsigned long pending_pkts, flags;
 
-   if (!netif_running(ndev))
-   return 0;
-
while ((pending_pkts = hecc_read(priv, HECC_CANRMP)) &&
num_pkts < quota) {
mbx_mask = BIT(priv->rx_next); /* next rx mailbox to process */
-- 
2.7.4