>OBTW: same problem with modutils 2.1.121: while boot scripts try
>to fsck the IDE disk (which is modularized), the module gets
>inserted, but it looks like fsck gets back control /before/ the
>module has had time to initialize, and so gets something like
>a -ENODEV.
Further investigation revealed that the culprit is the vfork() change
introduced in pre9. In both cases (modprobe and loadkeys), the C
library tried to sys_vfork() upon a call to system() or popen().
In pre8 and earlier kernels, this resulted in -ENOSYS and libc performed
a fork() instead (__libc_system calls __vfork which is defined as
pid_t
__vfork __P ((void))
{
pid_t retval = INLINE_SYSCALL (vfork, 0);
if (retval == (pid_t) -1 && errno == ENOSYS)
retval = INLINE_SYSCALL (fork, 0);
return retval;
}
). (This is GLibc 2.0.110. The problem will also be reported to the
libc-alpha mailing list).
In pre9, the new vfork system call is performed, and this results
(for some reasons that I have yet to investigate) in a seg fault
in the calling process. Backing out the change to entry.S so that
syscall(__NR_vfork) returns ENOSYS resolves the symptom.
Hope this helps,
Thomas.
--
Thomas.Quinot@Cuivre.FR.EU.ORG <URL:http://web.fdn.fr/~tquinot/>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/