Strange Linux Problem

[David Ronis <ronis at ronispc dot chem dot mcgill dot ca>]

I've started following the egcs CVS's, and have currently built 

gcc version egcs-2.93.19 19990419 (gcc2 ss-980929 experimental)

Last night I tried the benchmarks.  With gcc version egcs-2.91.66
19990314 (egcs-1.1.2 release) things ran fine; however, when I tried
the same test with the CVS version, I ended up crashing the machine
with the following (Linux-2.2.6 on an DELL i586) message being
endlessly scrolled on the screen:

kmem_alloc: Bad slab magic (corrupt) (name=skbuff_head_cache).

The test that was running had to do with a heavily recursive routine.
I had to do a hardware reboot and ended up with a corrupted disk.

OK, having not learned my lesson, I've just installed gdb-4.18 and
it's testsuite; again, the 1.1.2 versions work as expected.  However,
make check using the egcs-2.93.19 19990419 results in a kernel oops:

Apr 22 14:56:25 ronispc kernel: Unable to handle kernel NULL pointer 
    dereference at virtual address 00000001
Apr 22 14:56:25 ronispc kernel: current->tss.cr3 = 023de000, Lr3 = 023de000
Apr 22 14:56:25 ronispc kernel: *pde = 00000000
Apr 22 14:56:25 ronispc kernel: Oops: 0002
Apr 22 14:56:25 ronispc kernel: CPU:    0
Apr 22 14:56:25 ronispc kernel: EIP:    0010:[<c01a4010>]
Apr 22 14:56:25 ronispc kernel: EFLAGS: 00010202
Apr 22 14:56:25 ronispc kernel: eax: 00000001   ebx: 00000001   
    ecx: c037f000   edx: c037f000
Apr 22 14:56:25 ronispc kernel: esi: 40009058   edi: c1b8e000   
    ebp: 00000000   esp: c224dedc
Apr 22 14:56:25 ronispc kernel: ds: 0018   es: 0018   ss: 0018
Apr 22 14:56:25 ronispc kernel: Process gdb (pid: 31414,
    process nr: 59, stackpage=c224d000)
Apr 22 14:56:25 ronispc kernel: Stack: c01a6000 c224c000
    7fffffff c1b8e968 c1b8e000 00000019 c01a6000 c037f000 
Apr 22 14:56:25 ronispc kernel:        c1b8e000 c1b8e000 c224df8c
    c0179e85 00000000 00000000 c01710f1 c1b8e000 
Apr 22 14:56:25 ronispc kernel:        c0172faa 00000203 c1b8e000
		c01731b0 c1b8e000 40009000 00000001 00000001 
Apr 22 14:56:25 ronispc kernel: Call Trace: [<c0179e85>] [<c01710f1>] [<c0172faa>] [<c01731b0>] [<c016f392>] [<c0173098>] [<c01220a0>] 
Apr 22 14:56:25 ronispc kernel:        [<c016f224>] [<c0107984>] 
Apr 22 14:56:25 ronispc kernel: Code: 08 08 08 08 08 08 08 08 08 28 28 28 28 28 08 08 08 08 08 08 

What I don't get is why am I able to do this at all (I'm not running as root)? 
Moreover is this an egcs problem or a linux problem?

I don't subscribe to this list, so a direct reply would be appreciated.

Thanks


David Ronis