Skip to contentSkip to conversation

SiS 191 failure with MTU > 1496

DISCUSSION Pascal Terjan 2 participants 6 messages May 29, 2009 – Jun 2, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

On a laptop with such hardware, with default 1500 MTU I can't connect to local machines, error bit LIMIT is set when receiving packets.

By decreasing the MTU to 1496 everything works fine.

Jun 28 20:03:13 localhost kernel: sis190 Gigabit Ethernet driver 1.2
loaded.
Jun 28 20:03:13 localhost kernel: sis190 0000:00:04.0: PCI INT A -> GSI
19 (level, low) -> IRQ 19
Jun 28 20:03:13 localhost kernel: 0000:00:04.0: Read MAC address from
EEPROM
Jun 28 20:03:13 localhost kernel: 0000:00:04.0: Error EEPROM read 0.
Jun 28 20:03:13 localhost kernel: 0000:00:04.0: Read MAC address from
APC.
Jun 28 20:03:13 localhost kernel: 0000:00:04.0: Realtek PHY RTL8201
transceiver at address 1.
Jun 28 20:03:13 localhost kernel: 0000:00:04.0: Using transceiver at
address 1 as default.
Jun 28 20:03:13 localhost kernel: 0000:00:04.0: SiS 191 PCI Gigabit
Ethernet adapter at f08c0c00 (IRQ: 19), 00:21:85:da:e1:51
Jun 28 20:03:13 localhost kernel: eth0: GMII mode.
Jun 28 20:03:13 localhost kernel: eth0: Enabling Auto-negotiation.

00:04.0 0200: 1039:0191 (rev 02)
Subsystem: 1462:63f8
Flags: bus master, medium devsel, latency 0, IRQ 19 Memory at fddfcc00 (32-bit, non-prefetchable) [size=128] I/O ports at cc00 [size=128]
Capabilities: [40] Power Management version 2 Kernel driver in use: sis190
Kernel modules: sis190

#2 lore raw
In reply to Pascal Terjan
Quoted 11 lines · Pascal Terjan
Quoted 5 lines · Pascal Terjan

On a laptop with such hardware, with default 1500 MTU I can't connect to local machines, error bit LIMIT is set when receiving packets.

By decreasing the MTU to 1496 everything works fine.

This looks like a known issue

http://bugzilla.kernel.org/show_bug.cgi?id=10895
https://bugs.launchpad.net/ubuntu/+bug/369513

Oh and even
http://bugzilla.kernel.org/show_bug.cgi?id=10694

Should I add some info there ?

#5 lore raw
Quoted 8 lines · Pascal Terjan
Quoted 7 lines · Francois Romieu

Pascal Terjan <[email protected]> :
[...]

Quoted 2 lines

Oh and even
http://bugzilla.kernel.org/show_bug.cgi?id=10694

Yes, it sucks.

Quoted 7 lines
Quoted 5 lines
Quoted 1 line

Should I add some info there ?

Not yet but could you retry with RX_BUF_SIZE set to 1544, 1528 and 1520 ?

1544, 1528 and 1520 fail the same way

By the way there is some strange code:

tp->rx_buf_sz = (mtu > RX_BUF_SIZE) ? mtu + ETH_HLEN + 8 : RX_BUF_SIZE;

Isn't there a (different from the one I'm interested in :) ) bug when 1500 < mtu <= RX_BUF_SIZE

Shouldn't the test be (mtu > ETH_DATA_LEN) ?