Fixed: Kernel 2.0.1+ a.out binaries broken

Fixed: Kernel 2.0.1+ a.out binaries broken

Post by Jorj Bau » Fri, 23 Aug 1996 04:00:00



I finally found it: my iBCS module was somehow screwing up the load chain
for a.out binaries. Strange, but fixed.

Jorj

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Information Systems Specialist              |         200 S. 33rd St.
School of Engineering and Applied Science   |   Moore Building, Room 166-A
University of Pennsylvania                  |     Philadelphia, PA 19104
http://binky.seas.upenn.edu/~jorj           |         (215) 898-0575
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

 
 
 

1. Kernel 2.0.1+: a.out binaries broken!

As much as I like poking in the kernel and figuring out what's broken,
I'm hoping that someone can give me a quick and easy answer to this one...

I just upgraded to kernel 2.0.13 (from 2.0.0), and all of my a.out
binaries broke! This is with a.out and elf support both in the kernel
(not modules). For any a.out binary run, I get something like:

general protection: 0000
CPU:    0
EIP:    0010:[<018247ba>]
EFLAGS: 00010246
eax: 00edff84   ebx: f000ef6f   ecx: 00000000   edx: f000ef6f
esi: 00000018   edi: 00000000   ebp: 00000002   esp: 00edff64
ds: 0018   es: 0018   fs: 0018   gs: 002b   ss: 0018
Process xdm (pid: 118, process nr: 25, stackpage=00edf000)
Stack: f000ef6f 00000000 00edff84 00000020 0182de88 fffffff8 00000003 00000002
       00000003 00a86810 08053b00 0804ed90 00000002 00976000 0012ae21 0000fb5e
       0012ae6e 00000003 00a86810 08053b00 0804ed90 0010a652 0000fb5e 08053b00
Call Trace: [<0182de88>] [<0012ae21>] [<0012ae6e>] [<0010a652>]
Code: 8b 42 04 ff d0 89 c1 83 c4 10 be 2b 00 00 00 66 8e e6 83 f9

... the only difference being the process name.

The last discernable jump made in the kernel is in sys_uselib (from
fs/exec.c presumably to fs/binfmt_aout.c->load_aout_library), to
fn->use_shlib. More than this I don't know.

I assumed that my shared libraries were out of date, but upgrading to the
latest (libc.so.4.7.6, libc.so.5.2.18 or somesuch, ld.so.1.7.14), I have
the same problem. This persists in all kernels from 2.0.1 through (and
including) 2.0.13. I've been playing with removing the updates between
2.0.0 and 2.0.1 individually to figure out what caused the problem with no
success. There's no mention of a problem like this in any of the Changes
documentation that I can find.

ELF binaries run without a hitch. Scripts continue to function normally.
It's just the half-dozen a.out binaries that I have hanging around that
are dead. Some of them I wrote a long time ago, and would hate to have to
re-write (some of the source code was lost in a crash, but that's another
story...).

Has anyone experienced this problem? Better yet, does anyone know what I
can do to fix it? If I can provide more information for anyone, let me
know...

Thanks,
Jorj

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Information Systems Specialist              |         200 S. 33rd St.
School of Engineering and Applied Science   |   Moore Building, Room 166-A
University of Pennsylvania                  |     Philadelphia, PA 19104
http://binky.seas.upenn.edu/~jorj           |         (215) 898-0575
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

2. How does one view quicktime files on LinuxPPC?

3. Q: Compiling Elf binaries and a.out binaries....

4. mod_auth_pam kills child-processes

5. Make broken, top broken, flock errors, kernel modules broken

6. USB Pencam

7. fix VIA Rhine time outs (some)

8. Does Linux have this feature

9. Failed to execute binary (UML); binary compatibility broken?

10. 2.2.13 breaks aout libraries/binaries -- kernel aout guru request

11. new kernel broke old xdvi -- Fixed!

12. Compile ELF kernel under a.out kernel?

13. Can "make" work in ELF-default environment to produce a.out binaries?