(I'm sorry about my previous message with a legal notice, this time I resent it with another mail server that doesn't include the legal notice for me)
Hi all, I finally found why in my mpc5200 based system (and in the lite5200) I had leaks of skbs when receiving ethernet traffic at 100 Mbit/s so that we generated an rxfifo error (to be able to get it I even reduced the speed of the CPU using the PLL jumpers). The problem was: The reception packets are handled in the interrupt handler: mpc52xx_fec_rx_interrupt. So this interrupt is being handled all the time because we are receiving packets continuously. The reception fifo errors are handled in the interrupt handler: mpc52xx_fec_interrupt. This interruption is being handled very often because we are receiving more packets that what we can handle, and that generates a fifo reception error. So it is very usual that the mpc52xx_fec_interrupt handler is executed at the middle of the mpc52xx_fec_rx_interrupt handler. And I think that it is not designed for that purpose. The mpc52xx_fec_rx_interrupt uses bcom_retrieve_buffer and bcom_submit_next_buffer. The mpc52xx_fec_interrupt calls to mpc52xx_fec_reset which calls to mpc52xx_fec_free_rx_buffers and mpc52xx_fec_alloc_rx_buffers, which at the end call bcom_retrieve_buffer and bcom_submit_next_buffer. Then we have two problems because of the same origin: - bcom_retrieve_buffer and bcom_submit_next_buffer doesn't seem to be atomic, so they cannot be called from one interrupt nested from another one being executing those functions. - Even if they were atomic, the mpc52xx_fec_rx_interrupt and the mpc52xx_fec_reset functions are not intended to be called at the same time. Patch I managed to do: I've managed to perform a patch that avoids the leak disabling the irqs, but I have a remaining problem: The last packet to be transmited is waiting for a packet to be received in order to egress (I don't know why but I detected it experimentaly). Anyone with more experience can tell me what's wrong in this patch? Thank you in advance, Asier Signed-off-by: Asier Llano <[email protected]> --- drivers/net/fec_mpc52xx.c | 23 +++++++++++++++++++++-- 1 file changed, 21 insertions(+), 2 deletions(-) diff -urpN linux-2.6.31.2/drivers/net/fec_mpc52xx.c linux-2.6.31.2-fec_mpc52xx_skb_leak/drivers/net/fec_mpc52xx.c ================================================================ --- linux-2.6.31.2/drivers/net/fec_mpc52xx.c +++ linux-2.6.31.2-fec_mpc52xx_skb_leak/drivers/net/fec_mpc52xx.c @@ -85,13 +85,20 @@ MODULE_PARM_DESC(debug, "debugging messa static void mpc52xx_fec_tx_timeout(struct net_device *dev) { + struct mpc52xx_fec_priv *priv = netdev_priv(dev); + unsigned long flags; + dev_warn(&dev->dev, "transmit timed out\n"); + spin_lock_irqsave(&priv->lock, flags); + mpc52xx_fec_reset(dev); dev->stats.tx_errors++; netif_wake_queue(dev); + + spin_unlock_irqrestore(&priv->lock, flags); } static void mpc52xx_fec_set_paddr(struct net_device *dev, u8 *mac) @@ -359,8 +366,9 @@ static irqreturn_t mpc52xx_fec_tx_interr { struct net_device *dev = dev_id; struct mpc52xx_fec_priv *priv = netdev_priv(dev); + unsigned long flags; - spin_lock(&priv->lock); + spin_lock_irqsave(&priv->lock, flags); while (bcom_buffer_done(priv->tx_dmatsk)) { struct sk_buff *skb; @@ -375,7 +383,7 @@ static irqreturn_t mpc52xx_fec_tx_interr netif_wake_queue(dev); - spin_unlock(&priv->lock); + spin_unlock_irqrestore(&priv->lock, flags); return IRQ_HANDLED; } @@ -384,6 +392,9 @@ static irqreturn_t mpc52xx_fec_rx_interr { struct net_device *dev = dev_id; struct mpc52xx_fec_priv *priv = netdev_priv(dev); + unsigned long flags; + + spin_lock_irqsave(&priv->lock, flags); while (bcom_buffer_done(priv->rx_dmatsk)) { struct sk_buff *skb; @@ -445,6 +456,8 @@ static irqreturn_t mpc52xx_fec_rx_interr bcom_submit_next_buffer(priv->rx_dmatsk, skb); } + spin_unlock_irqrestore(&priv->lock, flags); + return IRQ_HANDLED; } @@ -454,6 +467,7 @@ static irqreturn_t mpc52xx_fec_interrupt struct mpc52xx_fec_priv *priv = netdev_priv(dev); struct mpc52xx_fec __iomem *fec = priv->fec; u32 ievent; + unsigned long flags; ievent = in_be32(&fec->ievent); @@ -471,9 +485,14 @@ static irqreturn_t mpc52xx_fec_interrupt if (net_ratelimit() && (ievent & FEC_IEVENT_XFIFO_ERROR)) dev_warn(&dev->dev, "FEC_IEVENT_XFIFO_ERROR\n"); + spin_lock_irqsave(&priv->lock, flags); + mpc52xx_fec_reset(dev); netif_wake_queue(dev); + + spin_unlock_irqrestore(&priv->lock, flags); + return IRQ_HANDLED; } _______________________________________________ Linuxppc-dev mailing list [email protected] https://lists.ozlabs.org/listinfo/linuxppc-dev
