Hi Brian,
could you check the full duplex handling? I'm testing my new nforce-250
Gb (epox 8KDA3J) with a normal 100 MBit autonegotiation link partner and
I'm getting bad performance (30 kB/sec) when sending. Half duplex works.
The nic reports late collisions:
nv_tx_done: looking at packet 233, Flags 0x28200.
-> packet not yet completed
nv_tx_done: looking at packet 233, Flags 0x1a810.
bits 16, 15, 13, 11, 4
error, lastpacket, late collision, deferred, retryerror.
nv_tx_done: looking at packet 234, Flags 0x8000.
-> successful
nv_tx_done: looking at packet 235, Flags 0x28200.
-> not yet completed.
Autonegotiation seems to work correctly: np->linkspeed is set to 65636,
np->duplex to 1.
lspci reports:
00:05.0 Bridge: nVidia Corporation: Unknown device 00df (rev a2)
Subsystem: Unknown device 1695:100c
Flags: bus master, 66Mhz, fast devsel, latency 0, IRQ 11
Memory at ec000000 (32-bit, non-prefetchable)
I/O ports at b400 [size=8]
Capabilities: [44] Power Management version 2
00: de 10 df 00 07 00 b0 00 a2 00 80 06 00 00 00 00
10: 00 00 00 ec 01 b4 00 00 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 95 16 0c 10
30: 00 00 00 00 44 00 00 00 00 00 00 00 0b 01 01 14
Could you give me a hint what's wrong? If you need further register
values, just ask.
Thanks,
--
Manfred
Will take a look at it.
> -----Original Message-----
> From: Manfred Spraul [mailto:[email protected]]
> Sent: Sunday, June 27, 2004 1:33 PM
> To: Brian Lazara
> Cc: [email protected]; Carl-Daniel Hailfinger
> Subject: [PATCH 2.6.7 and 2.4.27-pre6] new device support for
> forcedeth.c
>
>
> Hi Brian,
>
> could you check the full duplex handling? I'm testing my new
> nforce-250 Gb (epox 8KDA3J) with a normal 100 MBit
> autonegotiation link partner and I'm getting bad performance
> (30 kB/sec) when sending. Half duplex works. The nic reports
> late collisions:
>
> nv_tx_done: looking at packet 233, Flags 0x28200.
> -> packet not yet completed
> nv_tx_done: looking at packet 233, Flags 0x1a810.
> bits 16, 15, 13, 11, 4
> error, lastpacket, late collision, deferred, retryerror.
> nv_tx_done: looking at packet 234, Flags 0x8000.
> -> successful
> nv_tx_done: looking at packet 235, Flags 0x28200.
> -> not yet completed.
>
> Autonegotiation seems to work correctly: np->linkspeed is set
> to 65636,
> np->duplex to 1.
> lspci reports:
>
> 00:05.0 Bridge: nVidia Corporation: Unknown device 00df (rev a2)
> Subsystem: Unknown device 1695:100c
> Flags: bus master, 66Mhz, fast devsel, latency 0, IRQ 11
> Memory at ec000000 (32-bit, non-prefetchable)
> I/O ports at b400 [size=8]
> Capabilities: [44] Power Management version 2
> 00: de 10 df 00 07 00 b0 00 a2 00 80 06 00 00 00 00
> 10: 00 00 00 ec 01 b4 00 00 00 00 00 00 00 00 00 00
> 20: 00 00 00 00 00 00 00 00 00 00 00 00 95 16 0c 10
> 30: 00 00 00 00 44 00 00 00 00 00 00 00 0b 01 01 14
>
> Could you give me a hint what's wrong? If you need further register
> values, just ask.
>
> Thanks,
> --
> Manfred
>
The following is the fix for full duplex handling.
In the function nv_update_linkspeed(), there is an "if" statement that
was missing the "else" section:
if (np->duplex == 0)
phyreg |= PHY_HALF;
else
phyreg &= ~PHY_HALF;
Ayaz
> -----Original Message-----
> From: Manfred Spraul [mailto:manfred@xxxxxxxxxxxxxxxx]
> Sent: Sunday, June 27, 2004 1:33 PM
> To: Brian Lazara
> Cc: linux-kernel@xxxxxxxxxxxxxxx; Carl-Daniel Hailfinger
> Subject: [PATCH 2.6.7 and 2.4.27-pre6] new device support for
> forcedeth.c
>
>
> Hi Brian,
>
> could you check the full duplex handling? I'm testing my new
> nforce-250 Gb (epox 8KDA3J) with a normal 100 MBit
> autonegotiation link partner and I'm getting bad performance
> (30 kB/sec) when sending. Half duplex works. The nic reports
> late collisions:
>
> nv_tx_done: looking at packet 233, Flags 0x28200.
> -> packet not yet completed
> nv_tx_done: looking at packet 233, Flags 0x1a810.
> bits 16, 15, 13, 11, 4
> error, lastpacket, late collision, deferred, retryerror.
> nv_tx_done: looking at packet 234, Flags 0x8000.
> -> successful
> nv_tx_done: looking at packet 235, Flags 0x28200.
> -> not yet completed.
>
> Autonegotiation seems to work correctly: np->linkspeed is set
> to 65636,
> np->duplex to 1.
> lspci reports:
>
> 00:05.0 Bridge: nVidia Corporation: Unknown device 00df (rev a2)
> Subsystem: Unknown device 1695:100c
> Flags: bus master, 66Mhz, fast devsel, latency 0, IRQ 11
> Memory at ec000000 (32-bit, non-prefetchable)
> I/O ports at b400 [size=8]
> Capabilities: [44] Power Management version 2
> 00: de 10 df 00 07 00 b0 00 a2 00 80 06 00 00 00 00
> 10: 00 00 00 ec 01 b4 00 00 00 00 00 00 00 00 00 00
> 20: 00 00 00 00 00 00 00 00 00 00 00 00 95 16 0c 10
> 30: 00 00 00 00 44 00 00 00 00 00 00 00 0b 01 01 14
>
> Could you give me a hint what's wrong? If you need further register
> values, just ask.
>
> Thanks,
> --
> Manfred
>