Hello,
I'm running 2.6.0-test1-ac3 on my sis-based laptop, and I've some
problems with the sis 900 networking card: in a low-bandwith usage
(like a ssh to another computer, or a copy from another computer with
a 10Mbps card), I get some network freezes, and after a few seconds it
starts again (with the following messages in the syslog):
Aug 3 11:55:13 kesto kernel: eth0: Media Link Off
Aug 3 11:55:22 kesto kernel: NETDEV WATCHDOG: eth0: transmit timed out
Aug 3 11:55:22 kesto kernel: eth0: Transmit timeout, status 00000004 00000000
Aug 3 11:55:23 kesto kernel: eth0: Media Link On 100mbps full-duplex
When I copy files to another 100 Mbps laptop (with a cross-cable),
after a short while the network stops completly to work, and I cannot
even rmmod sis900 (device or ressource busy).
If you want the /proc/pci entry for the network card:
Bus 0, device 1, function 1:
Ethernet controller: Silicon Integrated S SiS900 10/100 Ethern (rev 130).
IRQ 10.
Master Capable. Latency=128. Min Gnt=52.Max Lat=11.
I/O at 0x3200 [0x32ff].
Non-prefetchable 32 bit memory at 0x44000000 [0x44000fff].
I've enabled preempt and local APIC in my kernel configuration, but I
used to patch my 2.4 kernels with preempt before switching to 2.6 and
I never had any network-related problem.
As a side-node, I can tell you that when we (a friend and I) ported
the sis900 driver to GNU Mach, we had to add a delay loop (I know,
it's ugly, but it's just a quick-and-dirty patch while waiting for the
OSKit-based GNU Mach) in the sis900_start_xmit function. With no delay
at all the driver stopped to work after one packet has been
transmitted, with a small delay it worked for low-bandwidth program
(say, a "ping 192.168.1.1") but not for a file transfert, I had to add
a bigger one to make it work fine.
I'm available if you need additional informations.
--
Gael Le Mignot "Kilobug" - [email protected] - http://kilobug.free.fr
GSM : 06.71.47.18.22 (in France) ICQ UIN : 7299959
Fingerprint : 1F2C 9804 7505 79DF 95E6 7323 B66B F67B 7103 C5DA
Member of HurdFr: http://hurdfr.org - The GNU Hurd: http://hurd.gnu.org
Can you check if the IRQ allocated to the SiS900 is the same on kernels
where it does work, and try without ACPI support, and/or any IO-APIC support
disabled?
I've seen this before with the network card in one of my systems - the
IO-APIC setup code, or ACPI table parsing was misprogramming the IRQ routing
tables, and it was being allocated the wrong level-triggered IRQ line.
Dan
--
Daniel J Blueman
COMPUTERBILD 15/03: Premium-e-mail-Dienste im Test
--------------------------------------------------
1. GMX TopMail - Platz 1 und Testsieger!
2. GMX ProMail - Platz 2 und Preis-Qualit?tssieger!
3. Arcor - 4. web.de - 5. T-Online - 6. freenet.de - 7. daybyday - 8. e-Post
Please see the attached file for the SiS IO-APIC workaround.
-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of Daniel Blueman
Sent: Monday, August 04, 2003 4:11 PM
To: [email protected]
Subject: Re: [BUG] 2.6.0-t1 sis900 timeout
Can you check if the IRQ allocated to the SiS900 is the same on kernels
where it does work, and try without ACPI support, and/or any IO-APIC support
disabled?
I've seen this before with the network card in one of my systems - the
IO-APIC setup code, or ACPI table parsing was misprogramming the IRQ routing
tables, and it was being allocated the wrong level-triggered IRQ line.
Dan
--
Daniel J Blueman
COMPUTERBILD 15/03: Premium-e-mail-Dienste im Test
--------------------------------------------------
1. GMX TopMail - Platz 1 und Testsieger!
2. GMX ProMail - Platz 2 und Preis-Qualit?tssieger!
3. Arcor - 4. web.de - 5. T-Online - 6. freenet.de - 7. daybyday - 8. e-Post
> Can you check if the IRQ allocated to the SiS900 is the same on kernels
> where it does work, and try without ACPI support, and/or any IO-APIC support
> disabled?
This IRQ is the same (10) with 2.4.20 (which works fine).
--
Gael Le Mignot "Kilobug" - [email protected] - http://kilobug.free.fr
GSM : 06.71.47.18.22 (in France) ICQ UIN : 7299959
Fingerprint : 1F2C 9804 7505 79DF 95E6 7323 B66B F67B 7103 C5DA
Member of HurdFr: http://hurdfr.org - The GNU Hurd: http://hurd.gnu.org