[CRIU] [PATCH] ptrace: add ability to retrieve signals without removing them from a queue

[Andrey Wagin]

2013/2/16 Pedro Alves <palves at redhat.com>:
> Forgot to reply to this bit:
>
> On 02/15/2013 07:43 PM, Oleg Nesterov wrote:
>>> We'd miss the poke
>>> > variant, but that looks like something that could be always be added
>>> > later.
>> Yes. _POKE_ or _QUEUE_ or _DEQUEUE_, we can add more features if user-
>> space wants them.
>
> In general, IMO, I agree with Roland at https://lkml.org/lkml/2002/12/20/84
> in that it's good to have setters for completeness, so that you can
> change all the state via ptrace that you can read via ptrace.
>
> But I'm not doing any of this work, hence my "could always be
> added later" comment instead of actually requesting it.  But if
> we had it, we could make e.g., gdb inspect the signal queues,
> and then be able to tweak a realtime signal before it is
> delivered.

PTRACE_POKESIGINFO is more complicated than PTRACE_PEEKSIGINFO.
Looks like PTRACE_POKESIGINFO should replace a siginfo with the
specified sequence number. Should it be able to change signo? If it is
able, what to do with signalfd, which already got a notification about
the previous signal?..

My opinion is that it "could always be added later", when we will
understand what exactly we want to have.

>
> --
> Pedro Alves