Re: [PATCH v3] getrandom system call wrapper [BZ #17252]

[Florian Weimer <fweimer at redhat dot com>]

On 09/09/2016 04:41 PM, Zack Weinberg wrote:
> On Fri, Sep 9, 2016 at 10:28 AM, Florian Weimer <fweimer@redhat.com> wrote:
> > On 09/09/2016 04:21 PM, Torvald Riegel wrote:
> > > On Thu, 2016-09-08 at 13:44 +0200, Florian Weimer wrote:
> > > > I have made the system call wrapper a cancellation point.  (If we
> > > > implement the simpler getentropy interface, it would not be a
> > > > cancellation point.)
> > >
> > >
> > > Why did you do that?
> >
> > I have to, because it can block indefinitely.
>
> Is it practical to make it a cancellation point if and only if called
> in blocking mode?  I suspect that is the behavior basically everyone
> wants.  Unlike with file descriptors, blocking/nonblocking is visible
> in the flags passed to the function, so it might not be hard.

The predominant use case is to call getrandom to request blocking (with 
flags == 0), without expecting blocking (because the kernel pool is 
initialized).  This would have to be a cancellation point, only the 
obscure GRND_NONBLOCK would not be.  So I don't think it's worth the 
complexity.

(The current goal is to provide a minimal system call wrapper.)

> (What does Solaris' implementation do?  OpenBSD's?)

I don't have access to a Solaris system with a compiler.  The manual 
page doesn't go into that level of detail.

Does OpenBSD have getrandom?  I'm not sure if their cancellation 
implementation is comparable, either.

Florian