Re: getter PTRACE_GETSIGINFO should not modify anything [Re:[PATCH 11/11] ptrace: implement group stop notification for ptracer]

[Tejun Heo]

Hello, Jan.

On Sun, May 15, 2011 at 04:02:32PM +0200, Jan Kratochvil wrote:
> Or there should be some other way how to query the siginfo_t state
> fully transparently (*).
>
> (*) But if there exists such syscall it probably does not make sense to modify
>     anything by PTRACE_GETSIGINFO.

If there's a syscall which doesn't affect notification state, then we
need something which does clear it.  Either way, I suppose you're
saying we need both something which clears the notification and
something which doesn't.

> Imagine various LD_PRELOAD tools which try to wrap system/library calls and
> operate with ptrace while keeping it transparent for the original debugger.
> (I have a bunch of such libraries for testing gdb/strace/etc. written there.)
> 
> Also complicated debuggers with internal OO hierarchy would need to just wrap
> PTRACE_GETSIGINFO into an internal function to make it transparent for calls
> not intending to modify the debuggee state.

We can add a flag or new request for that but I don't know.  Those are
pretty fringe use cases and they don't even strictly require such
feature - Even for LD_PRELOAD, it can simply keep scheduling INTERRUPT
until the application calls the wrapped GETSIGINFO when it detects the
stopped state has changed.  It can be easily done from userland.

Thank you.

-- 
tejun
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/