natsemi.c (linux 2.4.9) - Something Wicked happened! 18000

natsemi.c (linux 2.4.9) - Something Wicked happened! 18000

[Manfred Spraul]

 
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

> After
> installing the driver module for this card and connecting with a
> cross-cable to another card (tested with 10Mbps 3com509 and 100Mbps
> IntelPro) after the first ping a get the error:
Which compiler do you use?
gcc-2.95-1 is known to miscompile the current linux driver, and at least
one 2.95-2 prerelease snapshot.

egcs-1.1.2 and gcc-2.96-85 are ok.

I've just stress tested the driver from 2.4.9 - no problems at all. 9
MB/s receive, 7 MB/s send with Cyrix 6x68L 120MHz.

I have a workaround, but probably upgrading gcc is the better solution.

--
	Manfred

Re: natsemi.c (linux 2.4.9) - Something Wicked happened! 18000

[Tim Hockin]

> > Which compiler do you use?
> > gcc-2.95-1 is known to miscompile the current linux driver, and at least
> 
> i'm using gcc 2.95.2 (from debian stable) and i saw a post about
> that but it made me belive it was fixed at the time. My mistake!


yes - check that before we go any further :)

Re: natsemi.c (linux 2.4.9) - Something Wicked happened! 18000

[Nuno Miguel Fernandes Sucena Almeida]

> Which compiler do you use?
> gcc-2.95-1 is known to miscompile the current linux driver, and at least

i'm using gcc 2.95.2 (from debian stable) and i saw a post about
that but it made me belive it was fixed at the time. My mistake!

natsemi.c (linux 2.4.9) - Something Wicked happened! 18000

[Nuno Miguel Fernandes Sucena Almeida]

Hello,
	i'm not subscribed in linux kernel, so please excuse the
following. I bought a PCI NIC with the national semiconductor
DP83815 chip. The system is a dual PIII running linux 2.4.9 and with only
1 processor (PIII are expensive:/). So, here's the problem. After
installing the driver module for this card and connecting with a
cross-cable to another card (tested with 10Mbps 3com509 and 100Mbps
IntelPro) after the first ping a get the error:

/usr/src/kernels/linux_2.4.9/drivers/net$ping 10.2.0.2
PING 10.2.0.2 (10.2.0.2): 56 data bytes
64 bytes from 10.2.0.2: icmp_seq=0 ttl=255 time=0.8 ms

--- 10.2.0.2 ping statistics ---
5 packets transmitted, 1 packets received, 80% packet loss

and at /var/log/kern.log (or dmesg of course)

kernel: eth2: Setting full-duplex based on negotiated link capability.
kernel: eth2: Something Wicked happened! 18000.

With tcpdump at the other computer i got all the echo send icmp packets
but the NS NIC didn't seem to receive the echo reply packets!

I searched the linux kernel mailing list and found a post with this same
error. Tried to override link negotiation with mii-diag but no luck.

They said that after an upgrade from 2.4.5 to 2.4.6 the error
started to appear. So, i grabbed the natsemi.c source code from 2.4.5,
compiled it and the system is now working fine, even with 100MbpsTx-FD,
giving 10MByte/s.

I saw a recent post from Tim Hockin (thockin@sun.com) with a patch for
this driver for 2.4.8 but after looking at it i saw that it also changed other
 files and not only natsemi.c so i decided not to apply it.

The question is: now i have the system working (i hope) but there seems to
be a patch available for a long time and its very frustrating to find out
that the kernel is not updated with it. I had a similar problem with the
IntelPro 100 NIC on a toshiba PC - had to grab the source from intel.com
but only after spending a lot of time searching the web and with NO
information of this at the driver source :/

--

Re: natsemi.c (linux 2.4.9) - Something Wicked happened! 18000

[Tim Hockin]

> kernel: eth2: Setting full-duplex based on negotiated link capability.
> kernel: eth2: Something Wicked happened! 18000.
> 
> With tcpdump at the other computer i got all the echo send icmp packets
> but the NS NIC didn't seem to receive the echo reply packets!
> 
> I searched the linux kernel mailing list and found a post with this same
> error. Tried to override link negotiation with mii-diag but no luck.
> 
> They said that after an upgrade from 2.4.5 to 2.4.6 the error
> started to appear. So, i grabbed the natsemi.c source code from 2.4.5,
> compiled it and the system is now working fine, even with 100MbpsTx-FD,
> giving 10MByte/s.


aha!  So I'll check the diffs since 2.4.5 - thanks for the data point.  One
thing I did to make this error go away was increase the RX ring - I don't
have the code here, or I'd post a patch.  If you wait until tomorrow, I'll
pop into work and check out the diffs and post you an elementary patch, and
you can tell me if it goes away.