[ale] Error running RH 7.2

[Joe Steele]

Your error message isn't generated by the kernel.  It's coming from bash.  
The shell has its own implementation for malloc and uses sentinel values at
the beginning and ending of blocks that are checked whenever free() or
realloc() is called.  If the sentinel values are overwritten, the cryptic
error message you've seen is printed and bash aborts.

Unfortunately, this doesn't really explain why the errors are occurring.  
Once it's booted, does it work o.k. until shutdown?

By default, bash uses its own malloc implementation if possible.  But it 
can also be configured/built to use an alternative implementation, or to 
use the standard glibc malloc.  Redhat has built it using the default.
Interestingly, though, bash has a file called NOTES which states:

> Platform-Specific Configuration and Operation Notes
> ===================================================
> 
> 1.  configure --without-gnu-malloc on: 
>       .....  
>       linux (optional, but don't do it if you're using Doug Lea's malloc) 
>       .....  
>       If you are using GNU libc, especially on a linux system
> 

If this note is correct and up-to-date, then they are saying that bash's
malloc should not be used with GNU libc on linux.  (Their terminology is a
little confusing at first.  The "--without-gnu-malloc" is the old config
option name which has been kept for backward compatibility, but which is
now called "--without-bash-malloc" to better describe its purpose.)  The
result from using either "--without-bash-malloc" or "--without-gnu-malloc"
is that the standard glibc malloc is used instead.

As a start, you might try rebuilding bash from the source rpm after editing the
rpm spec file so that it says '%configure --without-bash-malloc' instead of
just '%configure'.  Build and install the new rpm and see what happens (keep a
copy of the old binary handy just in case something goes wrong).  You at least
won't get that error message anymore because it won't be in the binary (you
could confirm this using 'strings' on it).  Nonetheless, errors may manifest
themselves in new ways if, for example, the underlying cause is a hardware
problem.  If so, then it is probably time to play with swapping out parts --
RAM, CPU, etc.

If, by chance, you are still getting the same error message, then it's from
some other package which shares the malloc code contained in bash.  You will
need to figure out which package it is (Hopefully you won't have to resort to
grepping your entire filesystem for the message).

I'll be the first to admit that I see no reason why you should need to
rebuild bash.  If Redhat's bash configuration is broken, then other people
should be having the same problem as you.  But at least it's a fairly
simple test to perform, it will get rid of the error message that's
troubling you, and it might prove whether or not bash's NOTES file is meant
to be taken seriously.

--Joe

-----Original Message-----
From:	Jeff Layton [SMTP:laytonjb at bellsouth.net]
Sent:	Tuesday, April 09, 2002 4:14 PM
To:	ale at ale.org
Subject:	[ale] Error running RH 7.2

Hello,

   I think I posted this before, but just to be on the safe
side, I thought would ask the question again.
   I have a newly built AMD box (Epox 8HKA+ MB).
I could never get the X installer to come up (tried a S3
Savage 2000 card and a Diamond Stealth III/S540 card).
So, I did the text install. I think everything installed OK.
Oh, I've got a WD WD200BB HD on a ATA 100 cable.
   When the box boots, I see the following error message
throughout the boot process,

p->minfo.mi_magic2 == 0x5555


I searched the web and found only one real reference to
this problem. They suggested a new (better) IDE cable.
I tried that and I still get the same error message.
   I also get the same error message during shutdown.
   Anybody have any ideas? Should I email the kernel list?
Alan Cox? (this is a RH 7.2 installation after all :)

TIA,

Jeff

---
This message has been sent through the ALE general discussion list.
See http://www.ale.org/mailing-lists.shtml for more info. Problems should be 
sent to listmaster at ale dot org.