| Message ID | 49F361E4.4050605@nvidia.com |
|---|---|
| State | Accepted, archived |
| Delegated to: | David Miller |
| Headers | show |
From: Ayaz Abdulla <aabdulla@nvidia.com> Date: Sat, 25 Apr 2009 15:17:56 -0400 > This patch fixes the tx_timeout() to properly handle the clean up of > the tx ring. It also sets the tx put pointer back to the correct > position to be in sync with HW. > > Signed-off-by: Ayaz Abdulla <aabdulla@nvidia.com> You don't need to CC: Jeff Garzik as he is no longer handling network driver patches. -- 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
On Sat, 25 Apr 2009 15:17:56 -0400 Ayaz Abdulla <aabdulla@nvidia.com> wrote: > This patch fixes the tx_timeout() to properly handle the clean up of the > tx ring. It also sets the tx put pointer back to the correct position to > be in sync with HW. It's unclear (to me) what the real-world impact of this bugfix is. There isn't enough information here to decide whether the fix should be backported to 2.6.29 (and earlier?). What is the user-visible behaviour change? What hardware does it affect? What is the risk that the change will inadvertently break other setups? etc. Please always provide such information in the changelogs. -- 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
--- old/drivers/net/forcedeth.c 2009-04-25 15:07:31.000000000 -0400 +++ new/drivers/net/forcedeth.c 2009-04-25 15:10:34.000000000 -0400 @@ -1880,6 +1880,7 @@ np->tx_pkts_in_progress = 0; np->tx_change_owner = NULL; np->tx_end_flip = NULL; + np->tx_stop = 0; for (i = 0; i < np->tx_ring_size; i++) { if (!nv_optimized(np)) { @@ -2530,6 +2531,8 @@ struct fe_priv *np = netdev_priv(dev); u8 __iomem *base = get_hwbase(dev); u32 status; + union ring_type put_tx; + int saved_tx_limit; if (np->msi_flags & NV_MSI_X_ENABLED) status = readl(base + NvRegMSIXIrqStatus) & NVREG_IRQSTAT_MASK; @@ -2589,24 +2592,32 @@ /* 1) stop tx engine */ nv_stop_tx(dev); - /* 2) check that the packets were not sent already: */ + /* 2) complete any outstanding tx and do not give HW any limited tx pkts */ + saved_tx_limit = np->tx_limit; + np->tx_limit = 0; /* prevent giving HW any limited pkts */ + np->tx_stop = 0; /* prevent waking tx queue */ if (!nv_optimized(np)) nv_tx_done(dev, np->tx_ring_size); else nv_tx_done_optimized(dev, np->tx_ring_size); - /* 3) if there are dead entries: clear everything */ - if (np->get_tx_ctx != np->put_tx_ctx) { - printk(KERN_DEBUG "%s: tx_timeout: dead entries!\n", dev->name); - nv_drain_tx(dev); - nv_init_tx(dev); - setup_hw_rings(dev, NV_SETUP_TX_RING); - } + /* save current HW postion */ + if (np->tx_change_owner) + put_tx.ex = np->tx_change_owner->first_tx_desc; + else + put_tx = np->put_tx; - netif_wake_queue(dev); + /* 3) clear all tx state */ + nv_drain_tx(dev); + nv_init_tx(dev); + + /* 4) restore state to current HW position */ + np->get_tx = np->put_tx = put_tx; + np->tx_limit = saved_tx_limit; - /* 4) restart tx engine */ + /* 5) restart tx engine */ nv_start_tx(dev); + netif_wake_queue(dev); spin_unlock_irq(&np->lock); }
This patch fixes the tx_timeout() to properly handle the clean up of the tx ring. It also sets the tx put pointer back to the correct position to be in sync with HW. Signed-off-by: Ayaz Abdulla <aabdulla@nvidia.com>