Message ID | 1380281068-13269-1-git-send-email-ricardo.ribalda@gmail.com |
---|---|
State | Rejected, archived |
Delegated to: | David Miller |
Headers | show |
From: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com> Date: Fri, 27 Sep 2013 13:24:28 +0200 > The dma descriptors are only initialized on the probe function. > > If a packet is on the buffer when temac_stop is called, the dma > descriptors can be left on a incorrect status where no other package can > be sent. > > So an interface could be left in an usable state after ifdow/ifup. > > This patch makes sure that the descriptors are in a proper status when > the device is started. > > Signed-off-by: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com> This analysis is not correct. In the current driver, the descriptors are allocated and initialized in the open function, not the probe function. I'm not applying this patch. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hello David lp->tx_bd_ci, lp->tx_bd_next... are only initialized to zero on temac_of_probe (inside alloc_etherdev). Those vars are used to index the dma descriptors. The initialization of lp->tx_bd_v[i].app0 = 0; is redundant, because it is already done on dma_zalloc_coherent in temac_dma_bd_init called on open. What if I move lp->tx_bd_ci = 0; lp->tx_bd_next = 0; lp->tx_bd_tail = 0; lp->rx_bd_ci = 0; to temac_dma_bd_init? Will this be more correct? Without this patch a script like these kills the card in 1-10 iterations: ifdown eth0 ifdown eth1 while true do ifconfig eth1 10.100.10.100 udhcpc -i eth0 ping 192.168.2.1 -c 5 || break ifconfig eth0 down ifconfig eth0 10.100.10.100 udhcpc -i eth1 ping 192.168.2.1 -c 5 || break ifconfig eth1 down done Regards On Tue, Oct 1, 2013 at 6:21 AM, David Miller <davem@davemloft.net> wrote: > From: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com> > Date: Fri, 27 Sep 2013 13:24:28 +0200 > >> The dma descriptors are only initialized on the probe function. >> >> If a packet is on the buffer when temac_stop is called, the dma >> descriptors can be left on a incorrect status where no other package can >> be sent. >> >> So an interface could be left in an usable state after ifdow/ifup. >> >> This patch makes sure that the descriptors are in a proper status when >> the device is started. >> >> Signed-off-by: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com> > > This analysis is not correct. > > In the current driver, the descriptors are allocated and initialized > in the open function, not the probe function. > > I'm not applying this patch.
From: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com> Date: Tue, 1 Oct 2013 07:46:31 +0200 > What if I move > lp->tx_bd_ci = 0; > lp->tx_bd_next = 0; > lp->tx_bd_tail = 0; > lp->rx_bd_ci = 0; > > to temac_dma_bd_init? Will this be more correct? Yes, that would be a lot better. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Just send a v2 of the patch http://patchwork.ozlabs.org/patch/279339/ Thanks! On Tue, Oct 1, 2013 at 8:01 AM, David Miller <davem@davemloft.net> wrote: > From: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com> > Date: Tue, 1 Oct 2013 07:46:31 +0200 > >> What if I move >> lp->tx_bd_ci = 0; >> lp->tx_bd_next = 0; >> lp->tx_bd_tail = 0; >> lp->rx_bd_ci = 0; >> >> to temac_dma_bd_init? Will this be more correct? > > Yes, that would be a lot better.
diff --git a/drivers/net/ethernet/xilinx/ll_temac_main.c b/drivers/net/ethernet/xilinx/ll_temac_main.c index b88121f..8bae87f 100644 --- a/drivers/net/ethernet/xilinx/ll_temac_main.c +++ b/drivers/net/ethernet/xilinx/ll_temac_main.c @@ -229,6 +229,25 @@ static void temac_dma_bd_release(struct net_device *ndev) } /** + * temac_dma_bd_reset - Reset buffer descriptor rings + */ +static void temac_dma_bd_reset(struct net_device *ndev) +{ + struct temac_local *lp = netdev_priv(ndev); + int i; + + for (i = 0; i < TX_BD_NUM; i++) + lp->tx_bd_v[i].app0 = 0; + + lp->tx_bd_ci = 0; + lp->tx_bd_next = 0; + lp->tx_bd_tail = 0; + lp->rx_bd_ci = 0; + + return; +} + +/** * temac_dma_bd_init - Setup buffer descriptor rings */ static int temac_dma_bd_init(struct net_device *ndev) @@ -864,6 +883,8 @@ static int temac_open(struct net_device *ndev) if (rc) goto err_rx_irq; + temac_dma_bd_reset(ndev); + return 0; err_rx_irq:
The dma descriptors are only initialized on the probe function. If a packet is on the buffer when temac_stop is called, the dma descriptors can be left on a incorrect status where no other package can be sent. So an interface could be left in an usable state after ifdow/ifup. This patch makes sure that the descriptors are in a proper status when the device is started. Signed-off-by: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com> --- drivers/net/ethernet/xilinx/ll_temac_main.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+)