Donald Becker wrote:
> On Wed, 24 May 2000, Anders K. Pedersen wrote:
> > May 23 21:05:17 socks kernel: 3c59x.c:v0.99H 19May00 Donald Becker
>http://cesdis.gsfc.nasa.gov/linux/drivers/vortex.html
> > May 23 21:05:17 socks kernel: eth1: 3Com 3c590 Vortex 10Mbps at 0x6200,
>00:a0:24:cb:d2:e4, IRQ 10
> > May 23 21:05:17 socks kernel: 8K byte-wide RAM 1:1 Rx:Tx split,
>autoselect/10baseT interface.
>
> Note: this is an old 3c590 card, not a 3c900 series card.
> The transmit (and Tx error recovery) path is different.
Yes, I know.
> > May 23 23:19:56 socks kernel: eth1: Transmit error, Tx status register 90.
>
> Hmmm, it's very curious that this would be happening with a 10Mbps-only
> card. The Tx FIFO is 4KB, and the chip shouldn't start transmitting until
> the whole packet is in the FIFO. That should rule out a FIFO underrun as an
> error source.
The machine is a Pentium 133 (with F0 OF bug), that is running a socks5 server
with 300-600 concurrent connections with load values between 10 and 30 most of
the time. You can find the output from "vortex-diag -aa" in my reply to Andrew
Morton.
> This is a bad patch -- the tbusy should only be cleared for the Vortex
> chips. Move it up a few lines:
<cut patch away>
> But this still should never be happening...
OK - I have been running with this patch since this morning - it also fixes the
problem, but the timeouts still happen from time to time as you can see in the
log dump below.
May 24 07:39:41 socks kernel: epic100.c:v1.07h 8/18/99 Donald Becker
http://cesdis.gsfc.nasa.gov/linux/drivers/epic100.html
May 24 07:39:41 socks kernel: eth0: SMSC EPIC/100 83c170 at 0x6100, IRQ 9,
00:e0:29:22:f2:b7.
May 24 07:39:41 socks kernel: eth0: MII transceiver #3 control 3000 status 7809.
May 24 07:39:41 socks kernel: eth0: Autonegotiation advertising 01e1 link partner
0001.
May 24 07:39:58 socks kernel: 3c59x.c:v0.99H 19May00 Donald Becker
http://cesdis.gsfc.nasa.gov/linux/drivers/vortex.html
May 24 07:39:58 socks kernel: eth1: 3Com 3c590 Vortex 10Mbps at 0x6200,
00:a0:24:cb:d2:e4, IRQ 10
May 24 07:39:58 socks kernel: 8K byte-wide RAM 1:1 Rx:Tx split, autoselect/10baseT
interface.
May 24 08:04:29 socks kernel: eth1: Transmit error, Tx status register 90.
May 24 08:04:35 socks kernel: eth1: transmit timed out, tx_status 00 status e000.
May 24 08:34:38 socks kernel: eth1: Transmit error, Tx status register 90.
May 24 08:34:38 socks kernel: eth1: Transmit error, Tx status register c0.
May 24 08:34:45 socks kernel: eth1: transmit timed out, tx_status 00 status e000.
May 24 08:49:41 socks kernel: eth1: Transmit error, Tx status register 90.
May 24 10:03:36 socks last message repeated 2 times
May 24 10:03:44 socks kernel: eth1: transmit timed out, tx_status 00 status e000.
May 24 10:04:15 socks kernel: eth1: Transmit error, Tx status register 90.
May 24 10:04:24 socks kernel: eth1: transmit timed out, tx_status 00 status e000.
May 24 10:15:56 socks kernel: eth1: Transmit error, Tx status register 90.
May 24 10:16:04 socks kernel: eth1: transmit timed out, tx_status 00 status e000.
May 24 10:32:15 socks kernel: eth1: Transmit error, Tx status register 90.
May 24 10:32:24 socks kernel: eth1: transmit timed out, tx_status 00 status e000.
May 24 11:44:30 socks kernel: eth1: Transmit error, Tx status register 90.
May 24 11:55:11 socks last message repeated 2 times
May 24 11:55:19 socks kernel: eth1: transmit timed out, tx_status 00 status e000.
May 24 12:00:03 socks kernel: eth1: Transmit error, Tx status register 90.
May 24 12:01:02 socks last message repeated 2 times
May 24 12:33:46 socks last message repeated 2 times
May 24 12:33:54 socks kernel: eth1: transmit timed out, tx_status 00 status e000.
Regards,
Anders K. Pedersen
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
