[LTP] [PATCH v6 1/1] thp04: Add ptrace mode and fix tracee stop synchronization

[Cyril Hrubis]

Hi!
> That's a fair point.
> 
> The main goal was to keep thp04 functional on kernels with
> CONFIG_PROC_MEM_FORCE_PTRACE=y, where direct writes via /proc/self/mem
> are no longer possible.
> 
> Do you have a preference for how support for that configuration should
> look?
> 
> Would you prefer:
> 
>   1. Keeping thp04 focused on the original race test and skipping it
>      when only ptrace-based writes are available?
>
>   2. Splitting the ptrace write validation into a separate testcase and
>      leaving thp04 unchanged?
> 
>   3. Some other approach?

1 + 2 please.

The thp04 is a CVE regression test, if there is no chance to trigger the
CVE it should just report TCONF.


> I'm mostly trying to understand the intended coverage for
> CONFIG_PROC_MEM_FORCE_PTRACE=y before making further changes.

For that we should add tests under testcases/syscalls/ptrace/. I suppose
that we need a test that would check that the process cannot modify
itself when that config option is enabled. And another test that would
validate that parent can access child memory.

-- 
Cyril Hrubis
chrubis@suse.cz