[Devel] Re: [PATCH 1/4] signals: SEND_SIG_NOINFO should be considered as SI_FROMUSER()

Oleg Nesterov oleg at redhat.com
Wed Oct 7 04:30:49 PDT 2009


On 10/06, Roland McGrath wrote:
>
> > Perhaps, we should add the comment to explain that both SI_FROMUSER()
> > and si_fromuser() are only valid in the sending pathes.
>
> Yes.  Also now that you put them both in a sentence together, it is clear
> that it is insane to have two different things with those two names that
> differ only in capitalization.

I think this doesn't matter because we need more cleanups. As for naming
I agree, si_fromuser() sucks and I'd be happy to send the patch which
renames it (or re-send these 2 patches).

The problem is, both SI_FROMUSER() and SI_FROMKERNEL() must die imho.
In fact I think they should never exist.

	SI_FROMUSER(siptr)      ((siptr)->si_code <= 0)

note "<=", this means this helper is unuseable. What we need is another
macro/inline which checks "si_code < 0" (or >= 0 depending on naming),
this helper should be used by sys_sigqueueinfo/etc which can not not use
SI_FROMXXX() because SI_USER is rightly forbidden. __send_signal() can
use the new helper too.

Other cleanups which imho makes sense:

	- rename SEND_SIG_XXX

	- redefine them to make sure SEND_SIG_NOINFO != NULL

Oleg.

_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers




More information about the Devel mailing list