[PATCH] Do not trace children created with CLONE_UNTRACED flag

Roland McGrath roland at redhat.com
Tue Sep 14 02:29:16 UTC 2010

> Very sorry for my last reply, actually there are indeed some
> problems here. For syscall who has no arguments such as vfork
> accessing tcp->u_arg here may result an undetermined value.

Oh, that makes sense.  I almost thought this was also a problem
for the existing use of ARG_FLAGS in internal_fork.  But, actually
that code only happens in the exiting case, where setbpt already
ran in the entering case and changed the registers.

> What we want to change only occur in clone syscall, so I think
> we may add one more check here, just like the patch below.
> What do you think?

On ia64, it's clone2.  And there might be others later.
It might be better to test sysent[tcp->scno].sys_func == sys_clone.


More information about the Strace-devel mailing list