On 18-Sep-98 Simon Kenyon wrote:
>> Somewhere between 1.3.20 and 2.0.1 is where it happened (I'm still tracing
>> down the exact patch that broke it)
>
> could you not try a more recent kernel? 2.0.30 was over two years ago and
> you are using 2.0.1.
You misunderstand. The problems exist in the most recent kernels. At least,
they exist in 2.0.36pre9, and 2.1.121... which are the newest I've tested.
My reference to the older kernels is in an effort to track down just -where-
they broke. I haven't used the Lance driver in Linux until recently, so this
is the first I'm seeing of the problem. Which is why I'm bringing it up now.
I've traced it now to between 1.3.20 and 1.3.50. I suspect it was when
lance.c and lance32.c were merged.
Symptoms are most pronounced when reading/writing to the network at the same
time. They show up for me when untarring a large file from and to an NFS
mount. It was suggested to me to specifically try 2.0.29 and 2.0.34. The
problem shows up in them as well.
Problems include:
- - Tx FIFO errors.
- - Non-atomic function call errors (kmalloc, and alloc_skb in older kernels,
only alloc_skb after Mr. Cox's 2.0.36pre1 patch).
- - Many many NFS timeout (NFS server XXX not responding) messages
In kernels that exhibit the problem (and that have ifconfig which shows it),
there are many Carrier TX errors. There are no TX errors in kernels <1.3.20.
The network cabling has been ruled out, as the same error can be ducplicated
between a small test network of two machines and new cabling. The NFS server
is not the problem, as several different ones have been used. The
motherboards are not the problem, as several have been used. The LANCE cards
are not the problem, as they work perfectly under Windows, and FreeBSD, even
when the ISA bus is loaded with additional (sound + video) traffic.
I'm not going to rest until I get this solved. New hardware is not an
option, and I've already standardized on Linux with other machines, so going
FreeBSD is not an option. If I have to port the Lance driver from 1.3.20 to
2.1.x, then I will. It just might take me ages to figure out how to do this.
=)
- --- __
Kurt Fitzner <kfitzner@nexus.v-wave.com> |_/\
,--,;\)
PGP Fingerprint: 2B 4F 1B DE B9 2C 35 81 ,-"-..._\
(ID 0xD0CBC40D) 94 9E 0B 9A EF 4A 13 86 \_...._( )
|a a )|
"Those who don't write back are subject to /._ / /
my deeseenteegratore." \' ||\'
-----BEGIN PGP SIGNATURE-----
Version: PGP for Personal Privacy 5.0i
Charset: noconv
iQCVAwUBNgL9q73GEEDQy8QNAQH4ugP9HXNAI8/61rb2yc6de332vaVskfCKs74R
iwfjrwq9Oj7Tyo1DI4PCOpkw8dGGbk/B8e1PCiBN93qzhuhTDhQNE9UP300+sgSm
KVF0Wx3rt2zT4HBFPrnupaX23otrE7iWMxkTtnqKFsFrKaawmkBOSwNrbblhK95v
hSLvVNDfcPw=
=QCzN
-----END PGP SIGNATURE-----
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/