netdev
[Top] [All Lists]

r8169 in netdev experimental

To: Francois Romieu <romieu@xxxxxxxxxxxxx>
Subject: r8169 in netdev experimental
From: Brad House <brad@xxxxxxxxxxxxxxxxxxxxxxx>
Date: Sun, 04 Jan 2004 12:00:51 -0500
Cc: netdev@xxxxxxxxxxx, Jeff Garzik <jgarzik@xxxxxxxxx>, brad_mssw@xxxxxxxxxx
In-reply-to: <20031130014738.A2589@electric-eye.fr.zoreil.com>
References: <20031122183001.GA16993@gtf.org> <20031124000939.A456@electric-eye.fr.zoreil.com> <20031126004550.A25408@electric-eye.fr.zoreil.com> <20031127235143.A16767@electric-eye.fr.zoreil.com> <20031130014738.A2589@electric-eye.fr.zoreil.com>
Sender: netdev-bounce@xxxxxxxxxxx
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.5) Gecko/20031121
Ok, sorry I dropped out of existance for a while.
I just tried the 2.6.0-bk2 netdev experimental
patches, and the r8169 module locks the system on
loading.  Funny thing is I had to unplug the power cable
from the computer for a few seconds and plug it back in,
because an immediate reset wouldn't let to old driver
work :/

Let me know where to start debugging this, as I should
have some time here.
(Been busy getting AMD64 port into 'official' mode
for Gentoo, so I haven't had time to look into this too
much...)

Thanks

-Brad


Francois Romieu wrote:
Hopefully last round of Brad/Realtek's merging.

Patches apply in this order:
1 - r8169-hw_start.patch
2 - r8169-missing-tx-stats.patch
3 - r8169-intr_mask.patch

on top of:

  2.6.0-test11
+ 2.6.0-test9-bk25-netdrvr-exp1
+ r8169-mac-phy-version
+ r8169-init_one
+ r8169-timer

The unconditional calls to rtl8169_{rx/tx}_interrupt in rtl8169_interrupt()
are not integrated. That should not make a huge difference.

--
Ueimor


------------------------------------------------------------------------


Merge of changes from Realtek: - register voodoo in rtl8169_hw_start().


drivers/net/r8169.c | 6 ++++++ 1 files changed, 6 insertions(+)

diff -puN drivers/net/r8169.c~r8169-hw_start drivers/net/r8169.c
--- linux-2.6.0-test11/drivers/net/r8169.c~r8169-hw_start 2003-11-29 20:36:12.000000000 +0100
+++ linux-2.6.0-test11-fr/drivers/net/r8169.c 2003-11-29 20:44:17.000000000 +0100
@@ -1028,6 +1028,12 @@ rtl8169_hw_start(struct net_device *dev)
RTL_W32(TxConfig,
(TX_DMA_BURST << TxDMAShift) | (InterFrameGap <<
TxInterFrameGapShift));
+ RTL_W16(CPlusCmd, RTL_R16(CPlusCmd));
+
+ if (tp->mac_version == RTL_GIGA_MAC_VER_D) {
+ dprintk(KERN_INFO PFX "Set MAC Reg C+CR Offset 0xE0: bit-3 and bit-14 MUST be 1\n");
+ RTL_W16(CPlusCmd, RTL_R16(CPlusCmd) | (1 << 14) | (1 << 3));
+ }
tp->cur_rx = 0;


_


------------------------------------------------------------------------


Driver forgot to update the transmitted bytes counter. Originally done in rtl8169_start_xmit() by Realtek.


drivers/net/r8169.c | 5 ++++- 1 files changed, 4 insertions(+), 1 deletion(-)

diff -puN drivers/net/r8169.c~r8169-missing-tx-stats drivers/net/r8169.c
--- linux-2.6.0-test11/drivers/net/r8169.c~r8169-missing-tx-stats 2003-11-29 22:34:10.000000000 +0100
+++ linux-2.6.0-test11-fr/drivers/net/r8169.c 2003-11-30 00:26:09.000000000 +0100
@@ -1303,10 +1303,13 @@ rtl8169_tx_interrupt(struct net_device *
int cur = dirty_tx % NUM_TX_DESC;
struct sk_buff *skb = tp->Tx_skbuff[cur];
+ /* FIXME: is it really accurate for TxErr ? */
+ tp->stats.tx_bytes += skb->len >= ETH_ZLEN ?
+ skb->len : ETH_ZLEN;
+ tp->stats.tx_packets++;
rtl8169_unmap_tx_skb(tp->pci_dev, tp->Tx_skbuff + cur,
tp->TxDescArray + cur);
dev_kfree_skb_irq(skb);
- tp->stats.tx_packets++;
dirty_tx++;
tx_left--;
entry++;


_


------------------------------------------------------------------------

 drivers/net/r8169.c |    7 ++-----
 1 files changed, 2 insertions(+), 5 deletions(-)

diff -puN drivers/net/r8169.c~r8169-intr_mask drivers/net/r8169.c
--- linux-2.6.0-test11/drivers/net/r8169.c~r8169-intr_mask 2003-11-30 01:16:48.000000000 +0100
+++ linux-2.6.0-test11-fr/drivers/net/r8169.c 2003-11-30 01:18:22.000000000 +0100
@@ -334,8 +334,7 @@ static void rtl8169_tx_timeout(struct ne
static struct net_device_stats *rtl8169_get_stats(struct net_device *netdev);
static const u16 rtl8169_intr_mask =
- SYSErr | PCSTimeout | RxUnderrun | RxOverflow | RxFIFOOver | TxErr | TxOK |
- RxErr | RxOK;
+ RxUnderrun | RxOverflow | RxFIFOOver | TxErr | TxOK | RxErr | RxOK;
static const unsigned int rtl8169_rx_config =
(RX_FIFO_THRESH << RxCfgFIFOShift) | (RX_DMA_BURST << RxCfgDMAShift);
@@ -1445,9 +1444,7 @@ rtl8169_interrupt(int irq, void *dev_ins
RTL_W16(IntrStatus,
(status & RxFIFOOver) ? (status | RxOverflow) : status);
- if ((status &
- (SYSErr | PCSTimeout | RxUnderrun | RxOverflow | RxFIFOOver
- | TxErr | TxOK | RxErr | RxOK)) == 0)
+ if (!(status & rtl8169_intr_mask))
break;
// Rx interrupt


_



<Prev in Thread] Current Thread [Next in Thread>