Linux 2.1.27

shendrix@escape.widomaker.com
Sun, 02 Mar 1997 02:33:13 -0500


My experiences with 2.1.27 have been fairly good, but I do have questions.

* heavy SCSI activity locked up my machine completely. I'd give out the
error messages and warnings, but there are none. In kernel 2.0.29 I
usually get longer uptimes (when running a system-killer IO script)

* where do these errors come from when booting:

Mar 1 18:34:39 escape kernel: Error seeking in /dev/kmem
Mar 1 18:34:39 escape kernel: Error adding kernel module table entry.

I could not find them in the kernel sources.

* in.h and un.h do not have sa_family_t defined... I fixed this by
including linux/socket.h in each of them

* I get lots of these:

Mar 1 23:00:02 escape kernel: TCPv4 bad checksum from 42027bcc:0050 to 229aa1ce
:0467, len=1480/1480/1500

This occurs in the middle of a bunch of auth/tcp server failing messages.
I don't remember them in the recent past.

* I get a message when disconnecting from my ISP about pppd not
properly specifying its route, and this is another error message
that I cannot find.

* traceroute runs for me... but all legs marked unreachable... I have
the net-tools packages recommended in the changelog

* when using ftp now, traffic is much higher than before at all times
except file transfer, which seems normal (if a tiny bit slower)
I'm running PPP, not ethernet.

* nfs modules on my system no longer work... failed dependencies, even
when it looks to me like everything should be fine.

* has anyone noticed if X operations seem slower? I'll have time for
real testing next week... but for right now, I at least ``feel''
like X slowed down.

I should mention I've had SCSI trouble since the first day I upgraded
to a DPT 2044W PCI controller. As kernel revisions marched forward
from 1.3.9x, the problems have gradually gone away, but I still have
occasional lockups with no warnings errors and/or scsi bus lockups after
warnings about media sense and/or timeouts.

It is not my hardware, cables, drives, etc. The system works flawlessly
with a slower (1540B) controller. Something wicked yet lurks in the
SCSI code, IMHO.

I did the full upgrade of all compiler/library/loader/etc packages and
it runs fairly well but for the problems above.