[2450] in Release_Engineering
Fwd: PMAX kernel fixes?
daemon@ATHENA.MIT.EDU (Bill Cattey)
Thu Aug 30 14:42:18 1990
Date: Thu, 30 Aug 90 14:41:44 -0400 (EDT)
From: Bill Cattey <wdc@ATHENA.MIT.EDU>
To: rel-eng@ATHENA.MIT.EDU
Have we run all these fixes into our PMAX kernel?
We will need them to run Andrew.
-wdc
---------- Forwarded message begins here ----------
Received: by E40-PO.MIT.EDU (5.45/4.7) id AA01790; Thu, 30 Aug 90 08:57:28 EDT
Received: from PO2.ANDREW.CMU.EDU by ATHENA.MIT.EDU with SMTP
id AA25672; Thu, 30 Aug 90 08:57:25 EDT
Received: by po2.andrew.cmu.edu (5.54/3.15) id <AA01102> for
wdc@athena.mit.edu; Thu, 30 Aug 90 08:57:13 EDT
Received: via switchmail; Thu, 30 Aug 90 08:57:11 -0400 (EDT)
Received: from violet.fs.andrew.cmu.edu via qmail
ID </afs/andrew.cmu.edu/service/mailqs/q002/QF.4arEo0W00Ud5M0QkNv>;
Thu, 30 Aug 90 08:56:34 -0400 (EDT)
Received: from violet.fs.andrew.cmu.edu via qmail
ID </afs/andrew.cmu.edu/usr4/cat/.Outgoing/QF.sarEnQu00Ud5Q0gJZy>;
Thu, 30 Aug 90 08:55:57 -0400 (EDT)
Received: from
Messages.7.8.N.CUILIB.3.45.SNAP.NOT.LINKED.violet.fs.andrew.cmu.edu.pmax.
3
via MS.5.6.violet.fs.andrew.cmu.edu.pmax_3;
Thu, 30 Aug 90 08:55:56 -0400 (EDT)
Message-Id: <warEnQm00Ud5I0gJQc@andrew.cmu.edu>
Date: Thu, 30 Aug 90 08:55:56 -0400 (EDT)
From: Chris Alan Thyberg <cat+@andrew.cmu.edu>
To: Bill Cattey <wdc@ATHENA.MIT.EDU>
Subject: Re: PMAX kernel fixes?
In-Reply-To: <sar8z8oGG02=9XeT4u@athena.mit.edu>
References: <sar8z8oGG02=9XeT4u@athena.mit.edu>
Bill,
Of course, I remember you. Here's the scoop.
With ATK 14.6 on the pmax we were getting tons of random coredumps (like
console staying up for a minute or two and then quietly dying), seg
faults when you try to run an ATK application for the first few attempts
and then the program runs on the very next try, programs going under
with "pid 42 (foo) was killed on unaligined access", "illegal
instruction" errors during execution, an unacceptably high number of
kernel panics, etc.
We took four patches from DEC's Atlanta Service Center to be applied to
Ultrix 3.1 for RISC systems. I'm more or less quoting from the DEC
blurb that can with the tar file.
1) vm_mem.o
This patch is said to fix a problem encountered by programs that are
self-modifying or which execute data which they have written (ala
CLASS). It is seen if the executable data is paged out and then paged
in for execution. The result is corrupting the shared text segment and
can corrupt the on-disk image of the executable itself.
2) uipc_smem.o
This patch corrects the problem of processes being able to access a
memory segment after they have detached from it.
3) vm_swap.o vm_text.o
This fixes the problem of processes getting swapped out and never
swapped back in.
4) uipc_usrreq.o
This patch fixes the unp_gc panic.
===========
We've installed these in a test kernel and things look much better,
though with the start of semester madness we really haven't had the
man-power to test the new kernel and ATK 14.6 extensively. Beyond 14.6
there's patch level 6 == ATK 15.0, which we haven't even built yet.
However, organizational changes that have a deep and direct impact on
the groups that provide ATK (and AFS and everything else about our
"Andrew/X11/Motif/Unix" environment) make the future of ATK releases
questionable: upper management here are looking for ways to freeze
things, cut effort, eliminate software the "can't be proved to be
required" and so on. We will be going ahead with the kernel fixes as
soon as possible. But I cannot say for sure if we are going to follow
that with newer versions of ATK - we'll see.
As for the kernel patches. I would suggest that you use your DEC
contacts to request these fixes from them. Why? Because this will give
you the chance to beat DEC bloody for not keeping us all informed that
there are kernel fixes we should be getting. This whole notion that we
have to first hit the bug and then poll DEC for fixes is absurd.
Fortunately, the net (and AFS and AMS bboards :-) help keep some folks
in touch so that we don't all have to go through all of the pain all of
the time. But it is still ridiculous that DEC doesn't have a way simply
to TELL US there are bugs fixed and patches to get without our having to
ask them first.
Of course, if time is of the essence and you can't get turn around from
DEC, let me know and we'll pop the tar file over to you.
Take care,
Chris
p.s. As a return favor, please keep me informed (or point me to the
right folks) on what ya'll are doing with Motif and mwm for general
campus use. Is mwm now Athena's default window manager? Any and all
hints, tips, docs about how you "package" and support your X11
environments would be most appreciated. Things like Xserver start up,
default configuration files, standard .logins, etc. Thanks.
