On Wed, Feb 02, 2000 at 12:17:08AM +0000, Thorsten Kranzkowski wrote:
> > > DU-netscape (stuck right after exec, with 100% cpu-time).
> >
> > Can you strace it?
>
> of course :-) :
>
[ lots of info deleted ]
I've found that any ecoff executable with .text address =
(any_address | 0x200000000) behaves exactly like Thorsten described
(100% system time, not even jumping into the executable).
But it's killable.
Situation is much worse with elf binaries whose start address has
this evil bit set, they cause hard lockup.
The bug appeared somewhere between 2.3.23 and 2.3.30.
[ink@jurassic tst]$ cat a.c
int __start()
{
return 0;
}
[ink@jurassic tst]$ gcc -c -o a.o a.c
[ink@jurassic tst]$ ld -oformat ecoff-littlealpha -o a a.o -Ttext 0x124000001c0
[ink@jurassic tst]$ ./a
Segmentation fault (core dumped)
^ thats ok, return address is bogus
[ink@jurassic tst]$ ld -oformat ecoff-littlealpha -o a a.o -Ttext 0x126000001c0
[ink@jurassic tst]$ ./a
^ stuck here...
^C
[ink@jurassic tst]$
Any ideas?
Ivan.
-- To unsubscribe: send e-mail to axp-list-request@redhat.com with 'unsubscribe' as the subject. Do not send it to axp-list@redhat.com
This archive was generated by hypermail version 2a22 on Wed Mar 1 06:26:22 2000 PST
Send any problems or questions about this archive to webmaster@alphalinux.org.