Bug 17009 - Buggy initscripts on alpha
Summary: Buggy initscripts on alpha
Status: CLOSED DUPLICATE of bug 11436
Alias: None
Product: Red Hat Linux
Classification: Retired
Component: initscripts   
(Show other bugs)
Version: 6.2
Hardware: alpha
OS: Linux
Target Milestone: ---
Assignee: Bill Nottingham
QA Contact:
Depends On: 17163
TreeView+ depends on / blocked
Reported: 2000-08-27 19:28 UTC by Simon Hill
Modified: 2014-03-17 02:16 UTC (History)
1 user (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2000-08-30 03:21:46 UTC
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)

Description Simon Hill 2000-08-27 19:28:59 UTC
I just installed 6.2 on an alpha, and found a number of problems in the
initscripts. The machine is not on the internet, so I will have to sumarize
them. Each of the bugs would hang
a shutdown, resulting in fsck needing to be run on rebooting.

1) Hung when shutting down eth0. The solution was to comment out the wait
loop with usleep in ifdown. This would also hang when shutting down the
network manually. I changed loop to use 'sleep 1' and count to 5, which
fixed the ifdown problem, but the system still hung on shutdown. I suspect
that the test is not returning the intended value, but I just commented out
the loop rather than track it down. Shutdown is able to shutdown eth0 now,
though it reports failure. This needs to be looked at.

2) Hung when saving mixer settings. Don't know why because the aumix
command works fine manually. Perhaps has to do with runcmd. I just
commented out the whole section. Didn't this used to be run from an rc.x
directory, and was switchable from linuxconf? I liked the ability to turn
it on and off that way.

3) Hung when deactivating swap. This is due to a missing redirect in the
awk statement (should be </proc/swaps, not /proc/swaps).

4) Would have hung in unmounting the file systems, but I noticed the same
missing redirect in the awk statements in that section and fixed them.

Comment 1 Simon Hill 2000-08-28 16:40:03 UTC
Scatch number 2. I reinstalled and ran though the problems again, and saving the
mixer settings worked fine.

Comment 2 Bill Nottingham 2000-08-29 15:01:39 UTC
We haven't seen these on our alphas here. What sort
of alpha do you have?

Comment 3 Simon Hill 2000-08-29 15:21:19 UTC
It's a UDB (Multia). Actually, I'm setting up a 10-way UDB cluster in a ten-unit

I don't see that it matters, though, except maybe wrt the sleep loop in ifdown.
For the other two problems, awk does need a redirect to read from a file,
doesn't it?

Comment 4 Bill Nottingham 2000-08-29 15:55:53 UTC
No, it shouldn't.

If you symlink awk to mawk, does it spontatenously start

Comment 5 Simon Hill 2000-08-30 03:03:07 UTC
Yes and no. It would work fine if mawk were in /bin.

This statement hangs:

SWAPS=`awk '! /^Filename/ { print $1 }' /proc/swaps`

while this one works:

SWAPS=`mawk '! /^Filename/ { print $1 }' /proc/swaps`

However, with awk -> /usr/bin/mawk, the awk's in rc.sysinit don't work until
/usr is mounted. It also stops working after /usr is unmounted in halt.

Comment 6 Simon Hill 2000-08-30 03:21:44 UTC
Of course, the first example should read gawk not awk. Or rather, when I tested
it awk was simlinked to /bin/gawk.

Comment 7 Bill Nottingham 2000-08-30 20:05:47 UTC
OK, this is a problem that occurs with awk *only* on
EV4 alpha machines (i.e., Multias and Jensens.) As far
as we can tell, it's not an awk/initscripts bug; it's
a problem with FP emulation in the kernel. As such, I'm
not going to change initscripts; symlinking awk
to mawk (and possibly moving it to the root partition)
is a valid workaround for now.

*** This bug has been marked as a duplicate of 11436 ***

Comment 8 Simon Hill 2000-09-01 04:17:36 UTC
Yes, gawk was spending about 90% of its time in the kernel.
Applying the EV4 FPE kernel patch fixed it. See bug #17163.

Note You need to log in before you can comment on or make changes to this bug.