Bug 9735 - Sis 190 don't work at 1000 Mbps (sis191)
Summary: Sis 190 don't work at 1000 Mbps (sis191)
Status: RESOLVED OBSOLETE
Alias: None
Product: Drivers
Classification: Unclassified
Component: Network (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: Francois Romieu
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2008-01-12 08:15 UTC by A. Bianchi
Modified: 2015-02-19 19:24 UTC (History)
5 users (show)

See Also:
Kernel Version: 3.2.1
Subsystem:
Regression: No
Bisected commit-id:


Attachments
patched sis190 driver used on my test (9.87 KB, patch)
2008-01-12 08:20 UTC, A. Bianchi
Details | Diff
Output of ethtool for 2.6.39.3 (806 bytes, application/octet-stream)
2012-01-19 16:04 UTC, Sr. Cv.
Details
Output of ethtool for 3.2.1 (806 bytes, application/octet-stream)
2012-01-19 16:05 UTC, Sr. Cv.
Details
Output of lspci -v for 2.6.39.3 (6.38 KB, application/octet-stream)
2012-01-19 16:06 UTC, Sr. Cv.
Details
Output of lspci -v for 3.2.1 (6.37 KB, application/octet-stream)
2012-01-19 16:06 UTC, Sr. Cv.
Details

Description A. Bianchi 2008-01-12 08:15:06 UTC
Latest working kernel version:
none

Hardware Environment:
Acer Extensa E261 (SIS191 on ISA bridge SIS968 with RLT8211BL transceiver) linked to a LinkSys SRW2008P managed switch

Software Environment:
sis190 module with some patches applied (see bug 9386, 9467, 9547, 9548)


Problem Description:
the link don't work if on switch I setup switch port to "Max Capability" or when is flagged "1000 Full"; If I set only "1000 Full" then I can see this on kernel log:

eth0: mii ext = 0000.
eth0: mii lpa = c401 adv = 01e1.
eth0: link on unknown mode.

...the device work when I disable 1000 Mbps at other end!
Comment 1 A. Bianchi 2008-01-12 08:20:25 UTC
Created attachment 14426 [details]
patched sis190 driver used on my test

dmesg:

ACPI: PCI interrupt for device 0000:00:04.0 disabled
sis190 Gigabit Ethernet driver 1.2 loaded.
ACPI: PCI Interrupt 0000:00:04.0[A] -> GSI 19 (level, low) -> IRQ 21
PCI: Setting latency timer of device 0000:00:04.0 to 64
0000:00:04.0: Read MAC address from APC.
0000:00:04.0: Realtek PHY RTL8211BL transceiver at address 1 (001c:c912)
0000:00:04.0: Using transceiver at address 1 as default.
0000:00:04.0: SiS 191 PCI Gigabit Ethernet adapter at f8830000 (IRQ: 21), 00:1c:25:4e:1f:75
eth0: RGMII mode.
eth0: Enabling Auto-negotiation.
eth0: auto-negotiating...
Comment 2 Natalie Protasevich 2008-06-03 23:11:19 UTC
Any updates on this problem please,
Thanks.
Comment 3 Juan Jose Pablos 2008-06-03 23:17:23 UTC
Check Bug 10694 as it is on the same hardware. you could give a bit more light on the issue.
Comment 4 Natalie Protasevich 2008-06-03 23:29:25 UTC
It looks like problem still exists, but is being worked on. Thanks for the pointer Juan.
Comment 5 R.Ghetta 2010-02-18 20:15:08 UTC
Should be fixed in kernel 2.6.31
Comment 6 Sr. Cv. 2012-01-19 16:02:41 UTC
I have a problem with the sis190 driver when connected to a 1Gbps switch. It behaves really strange, I even suspected a hardware failure, but it appears a driver issue.

When my card is connected to a gigabit switch small packets are handled properly. However the big ones are mostly dropped. Especially the outgoing ones. I suspect the incomings are OK, since it does not break even large downloads, but uploads, e.g. during network printing are almost completely stopped.

I tested it using ping -s 1400, its about 50%-60% packet lost. ifconfig shows increasing error and overruns counters. 

I tried to modify the MTU value, but it does not change anything.

However when while pinging I turn on tcpdump in another console to check what happens in the network everything starts to work well -- no lost packets! But turning on promiscuous mode for the interface using ifconfig, without turning the tcpdump on does not make any change -- the packets are still lost. 

The other strange thing is that everything starts to work when I unplug my power supply. When I plug the supply again the packets start to get lost again. I tested it on two different power supplies, with one almost brand new. Its fully repeatable for both of them!

When I boot the system with acpi=off kernel boot flag the problem disappears. However turning off runtime power management for all devices, by writing 'on' to their 'power/control' files does not fix the problem.

I'm using Asus F80S. Normally under gentoo's 2.6.38-tuxonice-r1 But I tested it also on vanilla 2.6.39.3 and 3.2.1 kernels and the behaviour is exactly the same. I tried to connect to at least three different items of 1Gbps switches with the same result.

Shall I open a new bug for that? This one's title seems appropriate.

I attach lspci and ethtool outputs for the two vanilla kernels. I can provide more details and test results if required.
Comment 7 Sr. Cv. 2012-01-19 16:04:43 UTC
Created attachment 72127 [details]
Output of ethtool for 2.6.39.3
Comment 8 Sr. Cv. 2012-01-19 16:05:09 UTC
Created attachment 72128 [details]
Output of ethtool for 3.2.1
Comment 9 Sr. Cv. 2012-01-19 16:06:05 UTC
Created attachment 72129 [details]
Output of lspci -v for 2.6.39.3
Comment 10 Sr. Cv. 2012-01-19 16:06:28 UTC
Created attachment 72130 [details]
Output of lspci -v for 3.2.1
Comment 11 Alan 2015-02-19 15:11:12 UTC
This bug relates to a very old kernel. Closing as obsolete.

[Sr. Cv.: I didn't see your addition to an ancient bug - yours looks to be different entirely]
Comment 12 Juan Jose Pablos 2015-02-19 15:49:28 UTC
Alan,
This bug and bug 9386 Had been closed just because they had been tested on old kernels. But Possibly they still been an issue. As soon as I get this hardware around. I will try to update both.
Comment 13 Alan 2015-02-19 19:24:33 UTC
thanks

Note You need to log in before you can comment on or make changes to this bug.