[Devel] Re: [PATCH 1/2] c/r: Add AF_UNIX support (v5)

[Serge E. Hallyn]

Quoting Dan Smith (danms at us.ibm.com):
> SH> That also caused you to skip a bunch of security_* calls (at the
> SH> least here, at the recv equivalent, do_sock_getname, and at your
> SH> bind at restore).
> 
> SH> I don't think simply inserting them here is the right thing to do,
> SH> bc then as the main code changes this code is likely to fall out
> SH> of sync.  So like Oren says, I think you need to do more re-use of
> SH> the common code.  For the bind() case, for instance, write a
> SH> common helper used by both sys_bind() and your restart bind, which
> SH> does the security check and then calls sock->ops->bind().  It
> SH> makes your patchset a bit more intrusive, but easier to maintain.
> 
> Does it make sense to modify kern_bind() (and friends) to make the
> security_*() calls and then make sys_bind() and my restore code use
> kern_bind()?  I don't know enough about the security stuff to know if
> the other uses of kern_bind() in the kernel would trip up if the
> checks are done there...

No, since kernel_bind() is preciely for use by the kernel to create
sockets and no security checks are necessary (or make sense).  So
you just need to create a new helper shared by your function
and sys_bind() which does the security check and calls
sock->ops->bind().

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